Mitoxyperilysis-based molecular subtyping identifies SPP1⁺ macrophages and IL1B as critical mediators of tumor progression and immunotherapy resistance in intrahepatic cholangiocarcinoma

Abstract

Background

Intrahepatic cholangiocarcinoma (ICC) is an aggressive malignancy characterized by poor clinical outcomes. Although mitoxyperilysis has recently emerged as a novel form of regulated cell death, the clinical relevance of mitoxyperilysis-related molecular subtypes in ICC remains largely unexplored. This study aimed to elucidate the clinical significance, immune microenvironment features, and potential therapeutic targets of mitoxyperilysis-related molecular subtypes in ICC.

Methods

Based on the expression profiles of 45 mitoxyperilysis-related genes, ICC patients were classified into two subtypes using consensus clustering, and the robustness of the classification was validated in the independent GSE89749 cohort. Survival analysis, immune microenvironment characterization, single-cell transcriptomic profiling, in vitro/in vivo experiments, and nomogram construction were subsequently performed to investigate the underlying biological mechanisms and therapeutic implications.

Results

Two robust mitoxyperilysis-related molecular subtypes were identified. Patients in the high-mitoxyperilysis subtype exhibited significantly poorer overall survival (OS), and subtype classification remained an independent predictor of adverse prognosis. The high-mitoxyperilysis subtype was closely associated with lymph node metastasis, elevated CA19-9 levels, and KRAS mutation. Furthermore, this subtype demonstrated pronounced inflammatory activation, an immunosuppressive tumor microenvironment (TME), increased infiltration of neutrophils and M2 macrophages, and elevated expression of immune checkpoint molecules. Single-cell transcriptomic analysis revealed a marked enrichment of SPP1⁺ macrophages in the high-mitoxyperilysis subtype, with IL1B and CXCL3 identified as key regulatory molecules. IL1B promoted ICC cell invasion and metastasis. Notably, treatment with diacerein, an IL1B inhibitor, synergized with anti-PD-1 immunotherapy to suppress tumor progression and overcome immunotherapy resistance.

Conclusions

Mitoxyperilysis-related molecular classification provides a clinically relevant framework for prognostic stratification and prediction of immunotherapeutic responsiveness in ICC. SPP1⁺ macrophages and IL1B represent potential therapeutic targets, providing a promising strategy for precise diagnosis and treatment of ICC.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12967-026-08680-3.

Introduction

Intrahepatic cholangiocarcinoma (ICC) is a highly aggressive primary liver malignancy originating from the epithelial cells of the intrahepatic bile ducts, representing the second most common primary hepatic cancer after hepatocellular carcinoma [1]. ICC accounts for approximately 10–20% of primary liver cancers worldwide. Its incidence varies substantially across geographic regions (ranging from 0.3 to 6 per 100,000 inhabitants), with the highest rates reported in Southeast Asia [2, 3]. Over the past two decades, the incidence of ICC has increased steadily, and epidemiological projections suggest that it may rise by as much as 99% in the United States by 2029 [4]. Despite advances in diagnostic techniques, ICC remains a highly lethal malignancy, accounting for approximately 2% of all cancer-related deaths worldwide [5]. Owing to its insidious clinical presentation, most patients are diagnosed at advanced stages, resulting in poor outcomes. The median overall survival (OS) is less than 24 months, with a 5-year survival rate below 20% [6]. Although surgical resection remains the only potentially curative treatment, the majority of patients are ineligible for surgery at diagnosis, and postoperative recurrence rates remain high even among those who undergo resection [7]. Moreover, the efficacy of conventional chemotherapy and targeted therapies is often compromised by substantial molecular heterogeneity and therapeutic resistance. These challenges underscore the urgent need for novel prognostic stratification tools and therapeutic strategies.

Mitochondria are increasingly recognized not only as cellular bioenergetic centers but also as critical command hubs integrating metabolic cues with cell fate decisions. In cancer, mitochondrial dysfunction contributes to metabolic reprogramming, oxidative stress, and the regulation of diverse forms of regulated cell death (RCD) [8]. Beyond the well-characterized pathways of apoptosis and necroptosis, a spectrum of non-apoptotic RCD modalities has been increasingly recognized as critical determinants of tumor progression and treatment outcomes [9]. Among these, mitoxyperilysis is a recently identified form of lytic cell death induced by the synergistic effects of innate immune activation and metabolic disruption. During mitoxyperilysis, mitochondria subjected to BAX/BAK1/BID-dependent oxidative stress maintain prolonged contact with the plasma membrane, resulting in localized oxidative damage and subsequent membrane lysis. Distinct from other inflammatory cell death pathways, including pyroptosis, necroptosis, and ferroptosis, mitoxyperilysis is regulated by mTORC2 signaling [10]. Importantly, mitoxyperilysis lies at the intersection of mitochondrial oxidative stress, inflammatory signaling, and cell death. Given that chronic inflammation and metabolic dysregulation are well-established hallmarks of ICC, mitoxyperilysis may represent a previously unrecognized mechanism linking these pathogenic processes to tumor progression. However, the biological significance of mitoxyperilysis in ICC remains entirely unknown. In particular, whether mitoxyperilysis exerts tumor-suppressive effects or instead promotes disease progression by fostering a pro-inflammatory and immunosuppressive tumor microenvironment (TME) has yet to be elucidated.

The TME of ICC is highly heterogeneous and is characterized by extensive stromal remodeling, profound immune suppression, and complex cellular interactions that collectively shape disease progression and therapeutic outcomes [11]. Emerging evidence has implicated myeloid cells, particularly tumor-associated macrophages (TAMs), as central regulators of this immunosuppressive milieu [12]. Among the diverse myeloid subpopulations, SPP1⁺ macrophages have emerged as a distinct TAM subset associated with poor prognosis in ICC [13]. Cholangiocarcinoma-derived SPP1 induces monocyte chemotaxis and M2 polarization through activation of the CD44-PI3K-AKT pathway, thereby promoting tumor progression via the TGF-β/SMAD2/3 axis [14]. Despite these insights, the heterogeneity of myeloid cells in ICC and their mechanistic interplay with specific cell death pathways, including mitoxyperilysis, have not been investigated at single-cell resolution.

This study characterized mitoxyperilysis‑related molecular subtypes in ICC and validated their robustness in an independent external cohort. The prognostic significance, clinicopathological correlations, and immune microenvironment characteristics of these subtypes were comprehensively investigated. Single‑cell RNA sequencing analyses identified SPP1⁺ macrophages as a key cellular executor of the high‑mitoxyperilysis phenotype. In addition, IL1B and CXCL3 were identified as key regulatory molecules, and a prognostic nomogram was developed to facilitate individualized risk assessment. Functional validation further demonstrated that macrophage‑derived IL1B promotes ICC progression and immunotherapy resistance, whereas pharmacological inhibition of IL1B with diacerein enhances therapeutic responsiveness. Collectively, these findings establish mitoxyperilysis-based molecular subtypes as a robust framework for ICC prognostic stratification and immunotherapy guidance and identify SPP1⁺ macrophages and IL1B signaling as promising targets for precision therapy.

Materials and methods

Patient cohorts and clinical data

A total of 244 patients with ICC were enrolled in the study. Their RNA sequencing, somatic mutation profiles, and matched clinicopathological information were obtained from the cohort reported by Dong et al. [15]. An independent validation cohort comprising 83 patients with ICC was retrieved from the GEO database (accession number: GSE89749). Clinical variables collected for analysis included age, gender, tumor size, TNM stage, lymph node metastasis, vascular invasion, perineural invasion, CA19-9, CEA levels, HBV infection status, cholelithiasis, and mutation status of KRAS, IDH1/2, and TP53. OS was defined as the interval from surgical resection to death or the last follow-up. Publicly available single-cell RNA sequencing (scRNA-seq) dataset from 26 ICC patients were downloaded from the Genome Sequence Archive at the National Genomics Data Center(ID: PRJCA007744).

Identification of mitoxyperilysis-related genes

A total of 45 mitoxyperilysis-related genes were retrieved from the literature and publicly available databases (Supplement Table 1). Functional annotation and pathway enrichment analyses were performed to characterize the biological processes and signaling pathways associated with these genes.

Consensus clustering and subtype classification

To stratify ICC patients according to the expression profiles of mitoxyperilysis-related genes, consensus clustering was conducted using the ConsensusClusterPlus package. Two robust molecular subtypes were identified. K-means clustering was subsequently applied to validate the robustness of the classification. To quantify mitoxyperilysis activity at the individual-sample level, a mitoxyperilysis score was calculated using single-sample gene set enrichment analysis (ssGSEA). Patients were stratified into low- and high-mitoxyperilysis subgroups according to the optimal cutoff value.

Survival and prognostic analysis

OS was compared between subgroups using Kaplan-Meier survival analysis, and statistical significance was assessed using the log-rank test. Univariate and multivariate Cox regression analyses were performed to identify independent prognostic factors. In addition, stratified survival analysis was conducted according to clinicopathological features.

Bulk transcriptome analysis

Differentially expressed genes (DEGs) between the low- and high-mitoxyperilysis subgroups were identified using thresholds of |log₂FC| > 1 and adjusted P < 0.05. Functional enrichment analyses, including GO, KEGG, and GSEA, were performed to characterize the biological processes and signaling pathways associated with mitoxyperilysis. Immune cell infiltration was estimated using the xCell algorithm. To comprehensively evaluate the tumor immune microenvironment, the myeloid inflammatory score, immunosuppressive score, M2 macrophage score, and extracellular matrix remodeling score were calculated.

Single-cell RNA sequencing (scRNA-seq) analysis

Publicly available scRNA-seq data from 26 ICC patients were subjected to quality control. Dimensionality reduction and clustering were performed using the Seurat package, and cell types were annotated using canonical marker genes. Myeloid populations were further subclustered to identify macrophage subsets. Mitoxyperilysis activity at single-cell resolution was quantified. The Scissor algorithm was applied to integrate bulk transcriptomic subtypes with scRNA-seq cell populations.

Identification of hub genes and construction of a prognostic nomogram

Candidate hub genes were identified by overlapping DEGs and SPP1⁺ macrophage signature genes. IL1B and CXCL3 were validated as key regulators through logistic regression. A prognostic nomogram incorporating tumor size, IL1B expression, and CXCL3 expression was subsequently constructed. The predictive performance of the nomogram was evaluated using time-dependent ROC curves, calibration curves, and decision curve analysis (DCA).

ssGSEA

To quantify mitoxyperilysis activity in individual patients, ssGSEA was performed using the “GSVA” R package. A predefined gene set comprising 45 mitoxyperilysis-related genes, curated from published literature and public databases, was used to calculate sample-specific enrichment scores. The resulting mitoxyperilysis score was used to represent the relative activity of mitoxyperilysis-associated biological processes in each sample.

To further characterize the TME, two additional functional scores were calculated. The myeloid inflammatory score was defined as the average expression of myeloid inflammation-related signature genes (including S100A8, S100A9, S100A12, VCAN, and LYZ). The immunosuppressive score was calculated based on the expression of immune checkpoint molecules (PD-L1, CTLA4, LAG3, and TIM3) and immunosuppressive cytokines (IL-10 and TGF-β1). These scores were used to evaluate the inflammatory and immunosuppressive characteristics of the TME across different mitoxyperilysis subtypes.

Immune cell infiltration analysis

Immune cell infiltration was estimated from bulk transcriptomic data using the xCell algorithm. xCell is a gene signature-based deconvolution method that infers the relative abundance of immune and stromal cell populations from gene expression profiles. xCell scores were calculated for each sample using the “xCell” R package with default parameters. The resulting enrichment scores were used to compare the immune landscape between the low- and high-mitoxyperilysis subtypes, with particular emphasis on myeloid cell populations (including neutrophils, M2 macrophages, and dendritic cells) and T-cell subsets (such as CD8⁺ T cells and regulatory T cells).

Cell culture and molecular biology experiments

The human monocyte cell line THP-1 and the ICC cell line HCCC-9810 were cultured under standard conditions. THP-1 cells were transfected with lentivirus to generate stable IL1B-overexpression (OV-IL1B) cell lines. Successful overexpression was confirmed by quantitative real-time polymerase chain reaction (qPCR) and Western blot analyses. Transwell migration and invasion assays, together with wound-healing assays, were performed to assess the effects of IL1B-overexpressing macrophages on ICC cell malignant phenotypes. Diacerein was used as a pharmacological inhibitor of IL1B signaling.

In vivo xenograft model and therapeutic intervention

An immunocompetent murine ICC model was established by subcutaneous inoculation of SB-1 cells into C57BL/6J mice. Mice were randomly divided into four groups: Vehicle + IgG, Vehicle + anti-PD-1, Diacerein + IgG, and Diacerein + anti-PD-1. Tumor volume was measured regularly. At the end of treatment, tumors were excised and weighed to evaluate the therapeutic efficacy of diacerein combined with anti-PD-1 immunotherapy.

Statistical analysis

All statistical analyses were performed using R software (version 4.1.0). Continuous variables were compared using either Student’s t-test or Mann-Whitney U test, as appropriate. Categorical variables were analyzed using the chi-square test. P < 0.05 was considered statistically significant.

Results

Identification of mitoxyperilysis-based molecular subtypes in ICC

To investigate the potential role of mitoxyperilysis in ICC, 45 mitoxyperilysis-related genes were collected from published literature and publicly available databases. Enrichment analysis revealed that these genes were primarily associated with inflammatory response, programmed cell death, and the TNF signaling pathway (Fig. 1A), suggesting a potential link between mitoxyperilysis, tumor immunity, and inflammatory regulation. Based on the expression profiles of these genes, consensus clustering and K-means clustering were performed, resulting in the identification of two robust molecular subtypes with distinct transcriptional characteristics (Fig. 1B; Figure S1A-S1B). The high-mitoxyperilysis subgroup exhibited elevated expression of mitoxyperilysis-related genes, whereas the low-mitoxyperilysis subtype displayed comparatively lower expression levels (Fig. 1C). To quantitatively assess mitoxyperilysis activity, a mitoxyperilysis score was calculated for each patient using ssGSEA. The high-mitoxyperilysis subgroup (n = 115) demonstrated significantly higher mitoxyperilysis scores than the low-mitoxyperilysis subgroup (n = 129) (P < 0.001; Fig. 1D).

Fig. 1.

Identification and validation of mitoxyperilysis-related molecular subtypes in intrahepatic cholangiocarcinoma (ICC). (A) Enrichment analysis of 45 mitoxyperilysis-related genes, showing their involvement in inflammatory response, programmed cell death, and TNF signaling pathway. (B) Consensus clustering of ICC patients based on the expression profiles of mitoxyperilysis-related genes. (C) Heatmap of the expression patterns of mitoxyperilysis-related genes across low- and high-mitoxyperilysis subtypes. (D) Comparison of mitoxyperilysis scores between low- and high-mitoxyperilysis subgroups. (E-F) Validation of the mitoxyperilysis-based classification in the independent GSE89749 cohort

To validate the robustness of the mitoxyperilysis-based classification, an independent external cohort comprising 83 ICC patients (GSE89749) was analyzed. Consistent with the findings from the training cohort, patients were reproducibly stratified into high- and low-mitoxyperilysis subtypes, confirming the stability and reproducibility of this classification (Fig. 1E and F; Figure S1C-S1D).

Prognostic value and clinicopathological features of mitoxyperilysis subtypes

Survival analysis demonstrated that patients in the high-mitoxyperilysis subtype experienced significantly worse outcomes than those in the low-mitoxyperilysis subgroup (HR = 1.88, 95%CI: 1.26–2.80, P = 0.0006; Fig. 2A). A similar result was observed in the GSE89749 validation cohort (HR = 2.24, 95%CI: 1.20–4.19, P = 0.0155; Fig. 2B). Comparative clinicopathological analyses revealed that the high-mitoxyperilysis subgroup was significantly associated with adverse clinical characteristics, including regional lymph node metastasis, elevated CA19-9 levels, KRAS mutation, cholelithiasis, and perineural invasion (Fig. 2C; Figure S2A). In contrast, IDH1/2 mutations were more frequently observed in the low-mitoxyperilysis subtype (Figure S2A). These results suggest that the high-mitoxyperilysis subtype is linked to a more aggressive disease phenotype and a distinct mutational profile in ICC, characterized by KRAS mutation enrichment, whereas the low-mitoxyperilysis subtype is enriched for IDH1/2 mutations.

Fig. 2.

Prognostic significance and clinicopathological characteristics of mitoxyperilysis subtypes in ICC. (A) Kaplan-Meier analysis of overall survival (OS) in the training cohort. (B) Validation of survival differences in the GSE89749 cohort. (C) Associations between mitoxyperilysis subtypes and clinicopathological features, including lymph node metastasis, CA19-9 level, KRAS mutation, cholelithiasis, and perineural invasion. (D) Stratified survival analysis according to lymph node metastasis, CA19-9 level, KRAS mutation, CEA level, vascular invasion, and TNM stage (C1: low-mitoxyperilysis; C2: high-mitoxyperilysis). (E) Univariate Cox regression analysis of mitoxyperilysis subtype and clinical features. (F) Multivariate Cox regression analysis identifying the high-mitoxyperilysis subtype as an independent predictor of poor prognosis

To further assess the prognostic significance of mitoxyperilysis across different clinical contexts, stratified survival analyses were performed. Significant survival differences between mitoxyperilysis subtypes were observed among patients stratified by lymph node metastasis, CA19-9 (> 37 U/mL), KRAS mutation, CEA (> 5 µg/L), vascular invasion, and TNM stage (Fig. 2D; Figure S2B). Univariate Cox regression analysis revealed that the mitoxyperilysis subtype was significantly associated with prognosis (HR = 1.91, 95%CI: 1.28–2.87, P = 0.00167; Fig. 2E). After adjustment for gender, age, TNM stage, and other clinicopathological variables, multivariate Cox regression confirmed that the high-mitoxyperilysis subtype remained an independent predictor of poor survival (HR = 1.65, 95%CI: 1.07–2.55, P = 0.0248; Fig. 2F).

Immune microenvironment characteristics of the high-mitoxyperilysis subtype

To elucidate the biological mechanisms underlying the unfavorable prognosis associated with the high-mitoxyperilysis subtype, differential expression analysis was performed between low- and high-mitoxyperilysis tumors. A total of 1,246 upregulated and 319 downregulated genes were identified in the high-mitoxyperilysis subtype (Fig. 3A). GO and KEGG enrichment analyses demonstrated that genes upregulated in the high-mitoxyperilysis subtype were predominantly involved in inflammatory response, chemotaxis, neutrophil degranulation, cytokine-cytokine receptor interaction, IL-10 signaling, and immune regulatory pathways (Fig. 3B). GSEA further confirmed significant activation of the interferon-γ response and IL6_JAK_STAT3 signaling in the high-mitoxyperilysis subtype (Fig. 3C; Figure S3A). Collectively, these findings indicate that the high-mitoxyperilysis subtype is characterized by enhanced inflammation-related immune activation.

Fig. 3.

Immune microenvironment characteristics of the high-mitoxyperilysis subtype. (A) Volcano plot of differentially expressed genes (DEGs) between low- and high-mitoxyperilysis subtypes. (B) GO and KEGG enrichment analyses of genes upregulated in the high-mitoxyperilysis subgroup. (C) GSEA demonstrating activation of IFN-γ response and IL6_JAK_STAT3 signaling in the high-mitoxyperilysis subgroup. (D) Immune cell infiltration profiles estimated using the xCell algorithm. (E) Comparison of myeloid inflammatory and immunosuppressive scores between the two subgroups

To further explore the immune landscape of this subtype, immune cell infiltration was estimated using the xCell algorithm. Compared with the low-mitoxyperilysis subtype, tumors in the high-mitoxyperilysis group exhibited significantly increased infiltration of myeloid cells, including neutrophils, M2 macrophages, and immature dendritic cells (Fig. 3D). In contrast, the low-mitoxyperilysis subtype was enriched in CD8⁺ naive T cells, plasma cells, and Th1 cells. Both the myeloid inflammatory score and the immunosuppressive score were significantly elevated in the high-mitoxyperilysis subtype (Fig. 3E), indicating that the inflammatory milieu of this subtype is predominantly driven by myeloid cell infiltration. In addition, the M2 macrophage score and the extracellular matrix remodeling score were markedly increased in the high-mitoxyperilysis subtype (Figure S3B), suggesting activation of immunosuppressive macrophages and fibroblasts. Notably, immune checkpoint molecules, including PD-L1 and CTLA4, were markedly upregulated in the high-mitoxyperilysis subtype (Figure S3B-S3C). Furthermore, enrichment analyses revealed significant activation of PD-L1 expression and PD-1 checkpoint signaling pathways in the high-mitoxyperilysis subtype, indicating the presence of an immunosuppressive TME that may be more amenable to immune checkpoint blockade therapy.

Single-cell transcriptomic analysis reveals microenvironmental heterogeneity associated with mitoxyperilysis subtypes

To further investigate the cellular basis and microenvironmental heterogeneity underlying mitoxyperilysis subtypes, publicly available scRNA-seq data from 26 ICC patients reported by Zhang et al. were integrated and analyzed. Following rigorous quality control (Figure S4A), dimensionality reduction, clustering, and UMAP visualization were performed (Figure S4B). Based on canonical marker gene expression, nine major cell types were identified, including T cells, epithelial cells, macrophages, neutrophils, mesenchymal cells, B cells, endothelial cells, proliferating cells, and mast cells (Fig. 4A and C).

Fig. 4.

Single-cell transcriptomic analysis revealing tumor microenvironment (TME) heterogeneity associated with mitoxyperilysis subtypes. (A-C) UMAP visualization and annotation of nine major cell populations identified in ICC. (D) Comparison of single-cell mitoxyperilysis scores between low- and high-mitoxyperilysis subgroups. (E) Association between mitoxyperilysis status and HBV infection. (F) Differences in cellular composition between subtypes, highlighting increased macrophage and neutrophil infiltration in the high-mitoxyperilysis subgroup

Mitoxyperilysis scores were subsequently calculated for each patient, enabling classification into high-mitoxyperilysis (n = 13) and low-mitoxyperilysis (n = 13) groups. The high-mitoxyperilysis group exhibited significantly elevated mitoxyperilysis scores (P < 0.0001; Fig. 4D), supporting the robustness of the classification framework at single-cell resolution. When cells were annotated according to mitoxyperilysis subtype, distinct distribution patterns were observed in UMAP space, confirming transcriptomic heterogeneity between the two subtypes (Figure S4C).

Clinical correlation analysis revealed a significantly higher prevalence of hepatitis B virus (HBV) infection among patients in the high-mitoxyperilysis group (P = 0.02706; Fig. 4E), suggesting that HBV infection may contribute to the development of the high-mitoxyperilysis subtype. Comparative analysis of cellular composition further demonstrated that high-mitoxyperilysis tumors contained significantly greater proportions of macrophages (P = 0.000559) and neutrophils (P = 0.011), accompanied by a reduced proportion of epithelial cells (P = 0.0324) (Fig. 4F; Figure S4D). No significant differences were observed in the abundance of T cells, B cells, mesenchymal cells, or endothelial cells between the two groups. Taken together, these single-cell findings provide direct evidence that the high-mitoxyperilysis subtype is characterized by extensive infiltration of myeloid immune cells (macrophages and neutrophils). This observation is consistent with the bulk transcriptomic evidence of enhanced inflammation and immunosuppression, providing a cellular basis for the unfavorable prognosis associated with this subtype.

Single-cell dissection of myeloid subpopulations identifies SPP1⁺ macrophages as a key driver of aggressive ICC phenotypes

To further characterize the myeloid cell heterogeneity associated with the high-mitoxyperilysis subtype, myeloid cells from ICC scRNA-seq data were subjected to unsupervised subclustering analysis. Nine distinct myeloid subclusters were identified, including SPP1⁺ macrophages (SPP1⁺Mac), EGR1⁺ macrophages, S100A9⁺ monocytes, ZNF428⁺ macrophages, GPR155⁺ macrophages, SKAP1⁺ macrophages, plasmacytoid dendritic cells (pDC), LAMP3⁺ DCs, and XCR1⁺ DCs (Fig. 5A; Figure S5A). Subsequently, the relative abundance of each myeloid subcluster was compared between high- and low-mitoxyperilysis groups. SPP1⁺Mac exhibited a significantly higher infiltration proportion in the high-mitoxyperilysis group (P = 0.0421; Fig. 5B), whereas no significant differences were observed for the remaining myeloid subpopulations. Functional enrichment analysis revealed that SPP1⁺Mac was most strongly enriched in pathways related to inflammatory response, hypoxia, and M1 macrophage polarization while exhibiting reduced enrichment of anti-tumor immunity and immune surveillance pathways (Fig. 5C). To quantify the SPP1⁺Mac signature at the bulk level, a signature score was calculated for each patient. The SPP1⁺Mac signature score was significantly elevated in the high-mitoxyperilysis subgroup compared with the low-mitoxyperilysis subgroup (P < 0.001; Fig. 5D).

Fig. 5.

Single-cell characterization of myeloid heterogeneity identifies SPP1⁺ macrophages as a key driver of the aggressive ICC phenotype. (A) UMAP plot of nine myeloid cell subclusters. (B) Comparison of SPP1⁺ macrophage abundance between low- and high-mitoxyperilysis subgroups. (C) Functional enrichment of SPP1⁺ macrophages in inflammatory response, hypoxia, and M1 polarization pathways. (D) SPP1⁺ macrophage signature scores in low- and high-mitoxyperilysis subgroups. (E-F) Identification of Scissor⁺ cells and their correlation with SPP1⁺ macrophages. (G-H) Associations between SPP1⁺ macrophages and adverse clinicopathological features, including lymph node metastasis, perineural invasion, and vascular invasion. (I) Prognostic value of the SPP1⁺ macrophage signature

To investigate the relationship between SPP1⁺Mac and the mitoxyperilysis phenotype, the Scissor algorithm was applied to integrate bulk RNA-seq-derived mitoxyperilysis subtypes with myeloid subpopulations identified by scRNA-seq. This analysis identified Scissor⁺ and Scissor⁻ cell populations in ICC (Fig. 5E), and SPP1⁺Mac exhibited the highest proportion of Scissor⁺ cells among all myeloid subpopulations (Fig. 5F). Clinically, patients with a high SPP1⁺Mac signature showed significantly higher incidences of lymph node metastasis (P = 3.729e − 05; Fig. 5G), perineural invasion (P = 6.469e − 05; Fig. 5H), and vascular invasion (P = 0.006489; Fig. 5H). Furthermore, a high SPP1⁺Mac signature was significantly associated with elevated CA19-9, distant metastasis, and larger tumor size (Figure S5B). Collectively, these results indicate that SPP1⁺Mac infiltration is closely associated with the high-mitoxyperilysis subtype and contributes to a more aggressive clinicopathological profile in ICC.

Finally, the prognostic significance of the SPP1 + Mac signature in the ICC cohort was evaluated. Patients with a high SPP1⁺Mac signature had significantly poorer OS than those with a low signature (HR = 2.70, 95%CI: 1.81–4.03, Log-rank P < 0.0001; Fig. 5I). These findings identify SPP1⁺Mac as a key cellular mediator of the high-mitoxyperilysis subtype and unfavorable clinical outcomes, highlighting its potential as both a prognostic biomarker and a therapeutic target in ICC.

Identification of key oncogenic drivers and construction of a prognostic nomogram

To identify key molecular regulators underlying the establishment and maintenance of the high-mitoxyperilysis subtype, differential expression and bioinformatic analyses were performed between high- and low-mitoxyperilysis subgroups. Using stringent statistical criteria, 414 DEGs were identified. In addition, 41 highly expressed genes constituting the SPP1⁺Mac signature were obtained. Integration of these datasets yielded nine genes, including IL1B and CXCL3. Correlation analyses with clinicopathological features and the high-mitoxyperilysis subtype were performed, followed by univariate logistic regression. IL1B and CXCL3 were identified as the only genes significantly associated with the high-mitoxyperilysis subtype (Supplement Table 2). Both genes were markedly upregulated in the high-mitoxyperilysis subgroup (P < 0.0001; Fig. 6A and B), suggesting that these cytokines play central roles in mediating mitoxyperilysis-related inflammatory signaling and TME remodeling.

Fig. 6.

Identification of key oncogenic drivers and construction of a prognostic nomogram for ICC. (A-B) Differential expression of IL1B and CXCL3 between low- and high-mitoxyperilysis subgroups. (C) Prognostic significance of IL1B and CXCL3 in ICC. (D) Multivariate logistic regression identifying tumor size, IL1B, and CXCL3 as independent predictors of the high-mitoxyperilysis subtype. (E) Construction of a prognostic nomogram. (F) Time-dependent ROC analyses in the training and validation cohorts. (G-H) Calibration plots of the nomogram. (I) Decision curve analysis (DCA) of the clinical utility of the model

Survival analysis demonstrated that elevated expression of IL1B and CXCL3 was associated with shorter OS in TNM-stratified patient subgroups (Fig. 6C), indicating that these molecules are not only hallmarks of the high-mitoxyperilysis subtype but also potential predictors of poor prognosis in ICC. To account for potential confounding factors, multivariate logistic regression analysis was subsequently performed. Tumor size, IL1B expression, and CXCL3 expression were identified as independent predictors of the high-mitoxyperilysis subtype (Fig. 6D).

Based on these three independent predictors, a quantitative prognostic nomogram was established to facilitate individualized risk stratification and prognosis prediction in patients with ICC (Fig. 6E). Time-dependent ROC analysis demonstrated excellent predictive performance, with an AUC of 0.86 (95%CI: 0.78–0.98) in the training cohort and 0.78 (95%CI: 0.68–0.89) in the validation cohort (Fig. 6F).

Calibration curves showed close agreement between predicted and observed survival probabilities in both cohorts, with curves closely approximating the ideal diagonal line (Fig. 6G and H), indicating good calibration and model robustness. Furthermore, DCA confirmed that the nomogram provided greater clinical net benefits across a broad range of threshold probabilities than any individual predictor (Fig. 6I). Collectively, these results demonstrate that the prognostic nomogram integrating IL1B, CXCL3, and tumor size is a reliable and clinically applicable tool for risk stratification and personalized prognostic assessment in patients with ICC.

IL1B Inhibition combined with anti-PD-1 therapy to enhance anti-tumor efficacy

To prioritize the variables incorporated into the prognostic nomogram, a random forest algorithm was applied to evaluate their relative contributions. IL1B emerged as the most influential variable, exhibiting the highest importance score among all predictors (Fig. 7A). To explore the biological function of IL1B in ICC, a THP-1 macrophage cell line with stable OV-IL1B was generated by lentiviral transfection. qPCR and Western blot analyses confirmed significant upregulation of IL1B at both the mRNA and protein levels in the OV-IL1B group (Fig. 7B and D).

Fig. 7.

Macrophage-derived IL1B promotes ICC progression and immunotherapy resistance, which can be overcome by diacerein. (A) Random forest analysis identified IL1B as the top contributor to the nomogram model. (B-D) THP-1 cells were transduced with an IL1B-overexpression (OV-IL1B) lentivirus. Successful overexpression was confirmed by qPCR (B) and Western blot analyses (C-D). (E) Schematic illustration of the co-culture system. HCCC-9810 cells were co-cultured with control THP-1, OV-IL1B THP-1, or OV-IL1B THP-1 cells pretreated with diacerein for 24 h. (F-G) Transwell migration and invasion assays (F) and wound-healing (G) assays demonstrated that co-culture with OV-IL1B THP-1 cells significantly enhanced the migratory and invasive capacities of HCCC-9810 cells, whereas diacerein treatment markedly attenuated these effects. (H-K) In vivo validation in SB-1 xenograft models. (H) Mice were randomly assigned to four groups: Vehicle + IgG, Vehicle + aPD-1, Diacerein + IgG, and Diacerein + aPD-1. (I) Tumor growth curves, (J) representative images of excised tumors, and (K) final tumor volume measurements demonstrated that combined diacerein and anti-PD-1 treatment exerted a synergistic anti-tumor effect, resulting in significantly greater tumor suppression than either monotherapy alone

To determine the regulatory role of IL1B in ICC progression, a co-culture system consisting of differently treated THP-1 cells and the ICC cell line HCCC-9810 was established (Fig. 7E). Co-culture with OV-IL1B THP-1 cells significantly enhanced the migratory and invasive capacities of HCCC-9810 cells, whereas treatment with diacerein, a specific inhibitor of IL1B, markedly abrogated these tumor-promoting effects (Fig. 7F and G). These results suggest that macrophages promote ICC cell invasion and metastasis by secreting IL1B, and that pharmacological inhibition of IL1B effectively suppresses these effects.

The therapeutic potential of IL1B inhibition was subsequently evaluated in vivo using a subcutaneous SB-1 xenograft mouse model (Fig. 7H). Tumor-bearing mice were randomly assigned to four treatment groups: control group (Vehicle + IgG), anti-PD-1 monotherapy group (Vehicle + aPD-1), diacerein monotherapy group (Diacerein + IgG), and combination therapy group (Diacerein + aPD-1). Combination treatment with diacerein and anti-PD-1 produced significantly greater inhibition of tumor growth than either monotherapy alone (Fig. 7I and K). These findings indicate that diacerein can effectively enhance the anti-tumor efficacy of PD-1 blockade.

Given the marked upregulation of IL1B in the high-mitoxyperilysis subtype, these findings further suggest that ICC patients with this molecular subtype may derive greater therapeutic benefit from combined diacerein and anti-PD-1 treatment than from either agent alone.

Discussion

ICC is a highly heterogeneous and aggressive malignancy with limited tools for prognostic stratification and poor responsiveness to immunotherapy. This study characterized, for the first time, the clinical significance and immune regulatory mechanisms of mitoxyperilysis in ICC. Based on 45 mitoxyperilysis‑related genes, two robust molecular subtypes were identified and subsequently validated in an independent cohort. The high‑mitoxyperilysis subtype was significantly associated with unfavorable survival, aggressive clinicopathological features, and an immunosuppressive TME. Single‑cell transcriptomic analysis identified SPP1⁺ macrophages as the predominant myeloid population associated with the high-mitoxyperilysis subtype. In addition, IL1B and CXCL3 were identified as key regulators. A prognostic nomogram integrating these markers with tumor size demonstrated robust predictive performance. Functional experiments demonstrated that macrophage‑derived IL1B promoted ICC progression and resistance to PD-1 blockade, whereas pharmacological inhibition with diacerein effectively reversed these effects. These findings establish a novel molecular subtyping system, identify potential therapeutic targets, and provide a promising strategy for improving immunotherapeutic efficacy in ICC.

The molecular heterogeneity of ICC has been increasingly recognized as a critical determinant of tumor behavior and therapeutic response. Recent multi-omics studies have demonstrated that ICC can be stratified into distinct molecular subtypes with important prognostic and therapeutic implications [15]. Histologically, ICC is classified into large-duct and small-duct subtypes. Large-duct ICC more frequently harbors KRAS and TP53 mutations and exhibits more aggressive behavior, whereas small-duct ICC is associated with IDH1/2 mutations and FGFR2 fusions [16]. Consistent with these observations, the high-mitoxyperilysis subtype identified in this study was characterized by KRAS mutation and aggressive clinicopathological features. By focusing on a distinct cell death pathway closely associated with inflammation and immune regulation, the mitoxyperilysis-based classification complements existing histological and molecular classification systems. Validation in an independent external cohort (GSE89749) further supports the robustness, reproducibility, and potential clinical applicability of this classification system.

Chronic inflammation is a well-established driver of ICC initiation and progression. Multiple etiological factors, including hepatitis B virus (HBV) infection, hepatolithiasis, and parasitic infections, create a pro-inflammatory microenvironment that promotes the malignant transformation of cholangiocytes [17]. This study demonstrated a significant association between the high-mitoxyperilysis subtype and both HBV infection and cholelithiasis, suggesting that chronic inflammatory stimuli may contribute to the activation of mitoxyperilysis-related pathways. As central regulators of cellular metabolism and fate determination, mitochondria integrate metabolic cues and cell fate decisions, and mitochondrial dysfunction has emerged as a critical driver of both inflammatory response and tumor immunity [18]. Mitoxyperilysis, a recently identified lytic cell death pathway induced by innate immune activation and metabolic stress, may represent a key link between chronic inflammation and ICC progression. This hypothesis is supported by the enrichment of myeloid inflammatory signatures and activation of the interferon-γ response and IL6_JAK_STAT3 signaling pathways in the high-mitoxyperilysis subtype, indicating that mitochondrial stress may amplify inflammatory signaling and reshape the TME [19].

The TME of ICC is characterized by extensive myeloid cell infiltration and profound immunosuppression [20]. Consistent findings from both bulk and single-cell transcriptomic analyses demonstrated that the high-mitoxyperilysis subtype exhibited increased infiltration of macrophages and neutrophils, elevated myeloid inflammatory and immunosuppressive scores, and upregulation of immune checkpoint molecules, including PD-L1 and CTLA4. These findings suggest that mitoxyperilysis-associated inflammation paradoxically promotes the establishment of an immunosuppressive TME. This phenomenon is consistent with the well-recognized dual role of chronic inflammation in cancer, whereby persistent inflammatory signaling not only drives tumor progression but also induces immune evasion.

Among the myeloid cell populations, SPP1⁺ macrophages emerged as the principal cellular mediator of the aggressive ICC phenotype associated with high-mitoxyperilysis subtype. SPP1 (secreted phosphoprotein 1), also known as osteopontin, is a multifunctional cytokine that has been implicated in tumor progression, metastasis, and immune regulation across multiple malignancies [21, 22]. In cholangiocarcinoma, recent studies have shown that tumor-derived SPP1 induces monocyte chemotaxis and M2 polarization through the CD44-PI3K-AKT pathway, whereas polarized macrophages subsequently facilitate tumor proliferation via the TGF-β/SMAD2/3 signaling axis, thereby establishing a reciprocal tumor-macrophage regulatory circuit [8]. These findings extend this model by demonstrating that SPP1⁺ macrophages are significantly enriched in the high-mitoxyperilysis subtype, and that elevated SPP1⁺Mac signatures are associated with lymph node metastasis, perineural invasion, vascular invasion, and poor clinical outcomes. Furthermore, Scissor analysis identified SPP1⁺ macrophages as the myeloid population containing the highest proportion of Scissor⁺ cells, linking bulk mitoxyperilysis subtypes to single-cell populations. These findings provide compelling evidence that SPP1⁺ macrophages constitute a major cellular mediator of the high-mitoxyperilysis subtype. Consistent with these observations, a recent study employing three mouse ICC models and four clinical ICC cohorts identified an M2-like SPP1⁺ macrophage subset that promotes ICC proliferation and invasion via the OSM-OSMR-STAT3-CCND1 and THBS1-CD47-mTOR-DBN1 signaling pathways, further supporting the pivotal role of SPP1⁺ macrophages in ICC progression [23].

The identification of IL1B and CXCL3 as key oncogenic drivers of the high-mitoxyperilysis subtype provides additional mechanistic insights and potential therapeutic opportunities. IL1B, a prototypical pro-inflammatory cytokine, is a well-established mediator of tumor-associated inflammation and cancer progression. In cholangiocarcinoma, hyperglycemia has been reported to increase IL1B expression through activation of the NF-κB signaling pathway, suggesting a link between metabolic stress and inflammatory signaling [24]. CXCL3, a member of the CXC chemokine family, has recently been identified as an oncogenic factor in cholangiocarcinoma, with significantly elevated expression correlating with larger tumor size and reduced infiltration of monocytes and Th17 cells [25]. In this study, both IL1B and CXCL3 were markedly upregulated in the high-mitoxyperilysis subtype and independently predicted this aggressive subtype. Furthermore, a prognostic nomogram integrating tumor size, IL1B, and CXCL3 demonstrated excellent predictive performance, supporting the potential utility of these markers for clinical risk stratification.

Functional experiments further demonstrated that IL1B overexpression in macrophages significantly promoted the migratory and invasive capacities of ICC cells, whereas pharmacological inhibition with diacerein effectively abolished these tumor-promoting effects. In vivo, diacerein synergized with PD-1 blockade to suppress tumor growth, indicating that IL1B inhibition can reverse immunotherapy resistance in the high‑mitoxyperilysis subtype. Given the marked enrichment of IL1B expression and SPP1⁺ macrophage infiltration in the high-mitoxyperilysis subtype, these findings raise the possibility that patients with this molecular subtype may derive greater benefit from combined IL1B inhibition and PD-1 blockade. This provides a promising rationale for biomarker-guided immunotherapy in ICC.

Several limitations should be acknowledged. First, the single-cell cohort comprised only 26 patients. Validation in larger, multicenter prospective cohorts is required to establish the generalizability of the findings. Second, the underlying molecular mechanisms by which mitoxyperilysis regulates SPP1⁺ macrophage polarization and IL1B secretion remain incompletely understood. Future gain- and loss-of-function studies are needed to further elucidate these mechanisms. Third, while diacerein demonstrated synergistic anti-tumor activity with PD-1 blockade in vivo, its therapeutic efficacy in patients with the high-mitoxyperilysis subtype remains to be evaluated in prospective clinical trials. Fourth, the molecular mechanism by which IL1B regulates CXCL3 expression in cholangiocarcinoma cells has not yet been experimentally validated and warrants further investigation. Finally, because this study was retrospective, the prognostic value of the mitoxyperilysis-based classification and its ability to predict immunotherapy response require confirmation in large-scale prospective clinical studies.

In conclusion, this study established a mitoxyperilysis-based molecular classification that enables prognostic stratification and may facilitate patient selection for immunotherapy in ICC. SPP1⁺ macrophages and IL1B were identified as key drivers of the aggressive high-mitoxyperilysis subtype, and pharmacological inhibition of IL1B enhanced the anti-tumor efficacy of PD-1 blockade. These findings provide novel insights into the immunobiology of ICC and offer promising biomarkers and therapeutic targets for precision immunotherapy.

Supplementary Information

Below is the link to the electronic supplementary material.

Authors’ contributions

Yafei Wang, Kai Ding and Jiale Niu performed data analyses and drafted the manuscript. These authors contributed equally to this work. Zhi Li and Xiaobo Zhang designed the study, supervised the experiments, critically revised the manuscript for important intellectual content, and approved the final version of the manuscript. Zhi Li and Xiaobo Zhang contributed equally as corresponding authors. All authors read and approved the final manuscript.

Funding

This study was supported by grants from the Henan Provincial Medical Science and Technology Research Project (Grant No. LHGJ20250537).

Data availability

All data generated or analyzed during this study are included in this published article. Additional data are available from the corresponding author, XZ, upon reasonable request.

Declarations

Ethical approval and consent to participate

Written informed consent was obtained from all participants prior to sample collection. The study protocol was reviewed and approved by the Ethics Committee of The First Affiliated Hospital of Henan University of Science and Technology and was conducted in accordance with the principles of the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.