Comprehensive analysis of PANoptosis-related molecular subtypes and prognostic model development of clear cell renal cell carcinoma

Abstract

Background

Clear cell renal cell carcinoma (ccRCC) exhibits strong heterogeneity and variable therapeutic responses. PANoptosis, an integrated form of inflammatory programmed cell death, may influence tumor immunity and prognosis, yet its role in ccRCC remains unclear.

Methods

Multi-omics data from TCGA database were analyzed to characterize PANoptosis-related genes (PRG), define molecular and gene subtypes, and construct a prognostic PRG score using Cox and LASSO regression. Immune infiltration, drug sensitivity, and predicted immunotherapy response were evaluated. Single-cell RNA sequencing analysis was used to map PRG expression across cell populations. In vitro experiments were performed to validate RBCK1 function in ccRCC.

Results

14 PRG showed marked CNV alterations and differential expression. Three PRG molecular subtypes displayed distinct survival outcomes and immune landscapes. A three-gene PRG score (RIPK1, PYCARD, RBCK1) independently stratified prognosis and correlated with immune infiltration, mutation burden, and therapy sensitivity. Lower scores predicted better immunotherapy response and higher drug sensitivity. Single-cell analysis revealed broad PRG expression across macrophages, epithelial cells, endothelial cells, and stem-like cells. RBCK1 was significantly upregulated in ccRCC and promoted proliferation and migration, while its knockdown inhibited tumor cell growth.

Conclusions

We delineated the PANoptosis landscape in ccRCC and developed a robust PRG score with strong prognostic and immunological relevance. RBCK1 functions as a key oncogenic regulator and potential therapeutic target. These findings offer a valuable framework for precision risk assessment and treatment optimization in ccRCC.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-026-04561-9.

Introduction

Clear cell renal cell carcinoma (ccRCC) is the most common and aggressive subtype of renal cell carcinoma, accounting for approximately 70–80% of cases and responsible for the majority of kidney cancer–related deaths [1]. Despite advances in molecular characterization, ccRCC remains a highly heterogeneous malignancy driven by complex genetic and metabolic alterations [2]. Current therapeutic strategies, including immune checkpoint inhibitors, VEGF-targeted therapies, and combination regimens, have significantly improved patient outcomes [3]. However, durable responses occur only in a subset of individuals, and resistance inevitably develops in many cases [4, 5]. Moreover, the lack of reliable biomarkers for prognosis and treatment selection continues to limit the precision management of ccRCC [6]. These challenges underscore the urgent need to explore novel molecular determinants and construct integrative computational models to improve risk stratification and guide individualized therapy.

PANoptosis is an emerging form of regulated cell death characterized by the integrated and coordinated activation of pyroptosis, apoptosis, and necroptosis through a unified molecular machinery [7]. In contrast to traditional views that consider these pathways as independent entities, PANoptosis is executed by a multicomponent signaling platform known as the PANoptosome, which brings together upstream sensors, adaptor proteins, and downstream effectors [8]. Mechanistically, PANoptosis is triggered by diverse infectious, genotoxic, metabolic, or oncogenic stress signals, leading to extensive inflammatory cytokine release, disruption of cellular homeostasis, and amplification of innate immune responses [9, 10]. This integrated cell death modality plays essential roles in pathogen clearance, tissue homeostasis, and the regulation of inflammatory disorders [11–13].

Inducing programmed cell death is an important strategy in cancer treatment [14]. Accumulating evidence indicates that PANoptosis is closely associated with tumorigenesis and prognosis prediction [15]. As an integrated inflammatory cell death pathway, PANoptosis plays a critical role in cancer biology by influencing tumor initiation, immune surveillance, and therapeutic responsiveness [16, 17]. Dysregulation of PANoptosis components can either facilitate tumor progression by promoting immune evasion and shaping a tolerogenic microenvironment or enhance antitumor immunity through the release of proinflammatory cytokines and immunogenic cell death signals [18]. These bidirectional effects underscore the need to further investigate PANoptosis within specific tumor contexts. Although PANoptosis has been implicated in various malignancies, its contribution to ccRCC remains largely undefined [16]. Considering the pronounced immune heterogeneity of ccRCC and its dependence on inflammatory signaling and immune-mediated pathways, delineating the PANoptosis landscape in this tumor type is of particular importance [2, 3]. A deeper understanding of PANoptosis-related molecular regulation may help identify new biomarkers for risk stratification and reveal therapeutic vulnerabilities that could advance precision management of ccRCC.

Although growing evidence has highlighted the importance of PANoptosis-related regulators in shaping tumor cell fate and modulating antitumor immunity, their functional significance and comprehensive regulatory network in ccRCC remain poorly defined. In this study, we performed an integrative multi-omics analysis combining transcriptomic profiling, genomic variation assessment, immune infiltration characterization, and prognostic modeling to delineate the roles of PANoptosis-related genes in ccRCC. We further constructed a PANoptosis-related risk score, validated its prognostic value across training and test cohorts, and explored its association with immunotherapy response, drug sensitivity, and tumor mutational burden. Finally, single-cell RNA sequencing data and in vitro experiments were incorporated to provide cellular-level validation and functional evidence. This comprehensive analysis aims to provide new insights into PANoptosis biology and to identify robust biomarkers that may guide precision stratification and therapeutic decision-making for patients with ccRCC.

Materials and methods

Acquisition and preprocessing of transcriptomic data

In this study, RNA sequencing (RNA-seq) data and corresponding clinical information of normal control (NC) and clear cell renal cell carcinoma (ccRCC) samples were obtained from The Cancer Genome Atlas database (TCGA, https://portal.gdc.cancer.gov/). The raw matrices were initially processed in the Perl environment to extract gene expression values for each sample. Ensembl gene IDs were converted to standard gene symbols based on the human genome annotation file. Subsequently, Count-format data were transformed into Transcripts Per Million (TPM) using the “limma” R package, followed by log2 transformation and normalization. Clinical data, including overall survival time, survival status, age, sex, histological grade (Grade), clinical stage (Stage), and TNM stage, were curated using Perl scripts. Samples with incomplete clinical information were excluded, resulting in a final cohort of 72 NC samples and 530 ccRCC samples for subsequent analyses.

Differential expression analysis of PANoptosis-related genes and copy number variation

Based on previous literature, 14 PRG were curated as candidate genes [19–22]. These genes have been repeatedly reported as core regulators of PANoptosis, participating in the coordinated regulation of pyroptosis, apoptosis, and necroptosis, and have been widely adopted in prior PANoptosis-related studies. Copy number variation (CNV) data were retrieved from TCGA database, and analyses were performed using the “maftools” R package to visualize CNV patterns and calculate amplification and deletion frequencies. Genomic regions exhibiting significant CNV were further identified using the GISTIC 2.0 algorithm, and CNV frequency distribution plots and chromosomal localization maps were generated to characterize CNV features and potential driver roles of PRG in ccRCC. Protein–protein interaction (PPI) networks were constructed using the STRING database (https://string-db.org/) to explore potential interactions among PRG. Differential expression between NC samples and ccRCC samples was evaluated using the “limma” R package, with adjusted P values < 0.05 considered statistically significant.

Identification of molecular subtypes and immune microenvironment infiltration analysis

To investigate molecular heterogeneity in ccRCC, consensus clustering was performed based on the expression profiles of 14 PRG using the “ConsensusClusterPlus” R package. The number of clusters (K) was set from 2 to 9, with partitioning around medoids (PAM) as the clustering algorithm and Euclidean distance as the similarity metric. The optimal number of subtypes was determined by evaluating the cumulative distribution function (CDF) and its relative change (ΔCDF). Kaplan–Meier (KM) survival curves were generated using the “survival” R package, and differences between subtypes were assessed with the “log-rank” test. Immune cell infiltration was quantified using single-sample gene set enrichment analysis (ssGSEA) via the “GSVA” R package, based on transcriptomic data and reported immune cell marker genes (Supplementary Table 1). The “estimate” R package was employed to calculate immune scores, stromal scores, ESTIMATE scores, and tumor purity for a comprehensive evaluation of the tumor microenvironment.

Identification of differentially expressed genes among molecular subtypes and pathway enrichment analysis

Differential expression analysis between molecular subtypes was performed using the “limma” R package, with thresholds set at |fold change|≥ 1 and adjusted P values < 0.05. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were conducted using the “clusterProfiler” R package to explore the potential biological functions of differentially expressed genes (DEGs). Based on DEGs expression patterns, gene-based subtypes were further identified using “ConsensusClusterPlus” R package. Principal component analysis (PCA) plots were generated with “ggplot2” R package to visualize inter-subtype heterogeneity, and “pheatmap” R package was used to illustrate DEGs expression patterns and their associations with clinical parameters.

Construction of a PRG scoring index system

To establish a PRG-based prognostic model, survival data were integrated with the expression profiles of differentially expressed genes. Univariate Cox regression analysis was first performed using the “survival” R package to identify PRG significantly associated with overall survival in ccRCC patients, yielding four candidate prognostic genes. These candidates were subsequently subjected to least absolute shrinkage and selection operator (LASSO) regression using the “glmnet” package to reduce redundancy and prevent overfitting, followed by multivariate Cox regression analysis to evaluate their independent prognostic value under mutual adjustment. Three PRG that remained statistically significant in the multivariate model were ultimately selected to construct the PRG score, which was calculated by weighting the expression levels of RIPK1, PYCARD, and RBCK1 according to their corresponding regression coefficients (β) as follows: PRG score = RIPK1 ×  − 0.358 + PYCARD × 0.265 + RBCK1 × 0.456. Based on the optimal cutoff value derived from survival outcomes, patients were stratified into high- and low-PRG score subgroups for subsequent analyses. Kaplan–Meier survival curves were generated using survival to compare prognoses between subgroups. Heatmaps depicting PRG scores were visualized using “pheatmap” R scripts, and alluvial diagrams illustrating relationships among molecular subtypes, PRG scores, and clinical outcomes were generated using “ggalluvial” R scripts.

Internal cohort validation and prognostic model construction

Using the caret R package, samples were randomly divided into training and validation cohorts at a 7:3 ratio for internal validation. The PRG score for each sample was calculated based on the PRG scoring system formula, and samples were stratified into low- and high-PRG score subgroups according to the optimal cutoff value derived from clinical survival outcomes. Time-dependent Receiver Operating Characteristic (ROC) curves were generated using the “timeROC” R package to evaluate the predictive accuracy of the PRG score model for 1-, 3-, and 5-year overall survival. A nomogram prediction model was constructed by integrating clinical-pathological variables with the PRG scoring system using the rms R package, allowing estimation of individual 1-, 3-, and 5-year survival probabilities. The “regplot” R package was used to generate calibration curves to assess the nomogram’s accuracy and predictive reliability. Univariate and multivariate Cox regression analyses were performed using the survival R package to calculate hazard ratios (HRs) and p-values for each variable, thereby evaluating the independent prognostic value of the PRG score relative to other clinical-pathological factors.

Immunotherapy response and drug sensitivity analysis

Based on the The Cancer Immunome Atlas (TCIA; https://tcia.at/home) database, we obtained the Immunophenoscore (IPS) for each sample to reflect the tumor immune microenvironment's immunogenicity and potential response to immunotherapy. Additionally, drug sensitivity was predicted using the Genomics of Drug Sensitivity in Cancer database (GDSC; https://www.cancerrxgene.org/). The “pRRophetic” R package was applied to estimate the half-maximal inhibitory concentration (IC50) of targeted drugs based on the transcriptomic expression profiles of the samples. Somatic mutation data for each sample were extracted and preprocessed using Perl scripts. The “maftools” R package was employed to generate waterfall plots depicting mutation burden and somatic mutation frequency characteristics across different subgroups.

Single-cell RNA sequencing data processing and analysis

The GSE304466 dataset, containing 10 × Genomics scRNA-seq data from three ccRCC samples, was obtained from the Gene Expression Omnibus (GEO) database. Data preprocessing and downstream analyses were performed in R using the “Seurat” package. Raw count matrices were converted into Seurat objects using “CreateSeuratObject” (min.cells = 3, min.features = 200), ensuring that each gene was expressed in at least three cells and each cell contained at least 200 detected genes. Cell quality was assessed by calculating the proportion of mitochondrial (percent.mt) and ribosomal (percent.rb) genes. Low-quality cells and potential doublets were filtered out based on the following criteria: nFeature_RNA between 200 and 10,000, percent.mt ≤ 20%, and percent.rb ≤ 20%. The retained high-quality cells were used for all subsequent analyses. Normalization was performed using the “LogNormalize” method, and the top 2,000 highly variable genes were identified using the “vst” algorithm. Data were centered and scaled via “ScaleData”, followed by dimensionality reduction using PCA plot; the top 20 PCs were selected based on ElbowPlot results. Batch effects among samples were corrected using the Harmony algorithm. For visualization, t-distributed stochastic neighbor embedding (t-SNE) and uniform manifold approximation and projection (UMAP) were applied to the batch-corrected data. Clustering was performed using the Louvain algorithm via “FindNeighbors” and “FindClusters”; clustering stability across resolutions was assessed with the “clustree” package. Differentially expressed genes for each cluster were identified with “FindAllMarkers” (|log₂FC|≥ 0.25, adjusted P value < 0.05). Cell type annotation combined canonical marker genes reported in the literature with automated classification using “SingleR” referencing the Human Primary Cell Atlas, followed by manual curation to finalize assignments. Module scores for specific gene sets across cell types were calculated using “AddModuleScore”, and inter-group differences were visualized using “VlnPlot” with custom scripts.

Cell culture

Human renal proximal tubular epithelial cells HK2, human ccRCC 786-O cells, and human ccRCC 769-P cells were obtained from the American Type Culture Collection (ATCC, USA). HK2 cells were cultured in Dulbecco’ s Modified Eagle Medium/Nutrient Mixture F-12 (DMEM/F12; Gibco, USA), whereas 786-O and 769-P cells were maintained in RPMI-1640 medium (Gibco, USA). All culture media were supplemented with 10% fetal bovine serum (FBS; Gibco, USA) and 1% penicillin–streptomycin (Gibco, USA). Cells were incubated at 37 °C in a humidified atmosphere containing 5% CO₂, and culture media were regularly refreshed to maintain optimal growth conditions. When cells reached approximately 80% confluence, they were detached using 0.25% trypsin–EDTA (Gibco, USA) and passaged at a ratio of 1:3–1:5 to ensure stable cell sources and to avoid microbial contamination during experiments.

Western blot analysis

HK2 cells, 786-O cells, and 769-P cells were cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS; Gibco, USA) at 37 °C in a humidified incubator with 5% CO₂. Cells were harvested at the logarithmic growth phase for protein extraction. Total protein was extracted using RIPA lysis buffer (Beyotime, China, P0013B) and quantified by the BCA assay (Beyotime, China, P0010). Equal amounts of protein (30 μg) were separated by SDS-PAGE and transferred to PVDF membranes (Millipore, USA, IPVH00010). Membranes were blocked with 5% non-fat milk for 1 h at room temperature and incubated overnight at 4 °C with primary antibodies: RBCK1 (Proteintech, 26367-1-AP, 1:1000) and β-Actin (Proteintech, 66009-1-Ig, 1:5000) as the loading control. The following day, membranes were incubated with HRP-conjugated secondary antibody (Proteintech, SA00001-1, 1:5000) at room temperature for 1 h, and protein bands were visualized using ECL chemiluminescence substrate (Thermo Fisher, USA, 34580). Band intensities were quantified using ImageJ software to assess differential RBCK1 expression among HK2, 786-O, and 769-P cells.

siRNA-mediated knockdown of RBCK1

786-O and 769-P cells were cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS; Gibco, USA) at 37 °C in a humidified incubator with 5% CO₂. To establish RBCK1 knockdown models, both 786-O and 769-P cells were transfected with siRNA targeting RBCK1 (GenePharma, Shanghai, China), while a non-targeting siRNA (siNC) served as the negative control. Transfection was performed using Lipofectamine 3000 reagent (Invitrogen, USA) according to the manufacturer’s instructions, with a final siRNA concentration of 50 nM. After 48 h, cells were collected, and total protein was extracted using RIPA lysis buffer (Beyotime, China). Protein concentrations were quantified using the BCA assay, followed by Western blot analysis to evaluate knockdown efficiency. The primary antibodies used included RBCK1 and β-Actin as the internal control. Protein bands were detected with ECL chemiluminescence substrate (Thermo Fisher, USA), and densitometric analysis was performed using ImageJ software to assess RBCK1 silencing efficiency in both 786-O and 769-P cells.

Colony formation assay

786-O and 769-P cells in the logarithmic growth phase were collected 48 h after transfection with siNC or siRBCK1, and the cell density was adjusted to 500–1000 cells per well. Cells were evenly seeded into 6-well plates, with three replicates per group. The cells were cultured in complete medium containing 10% FBS for 10–14 days until visible single-cell colonies formed, and the culture medium was replaced every 2 days. At the end of incubation, the medium was discarded, cells were washed twice with PBS, and then fixed with 4% paraformaldehyde (Beyotime, China, P0099) for 20 min. Colonies were subsequently stained with 0.1% crystal violet (Beyotime, China, C0121) for 15 min, washed, and air-dried. The number of colonies was counted under a light microscope to evaluate the colony-forming ability of both 786-O and 769-P cells.

Transwell assays

786-O and 769-P cells were collected 48 h after transfection with siNC or siRBCK1. Migration assays were performed using 24-well Transwell chambers with 8 μm pore size (Corning, USA, 3422), while invasion assays were conducted by coating the upper surface of the inserts with pre-diluted Matrigel (1:8, Corning, USA, 356234). Prior to use, both the upper and lower chambers were pre-equilibrated with culture medium for 30 min. Cells were resuspended in serum-free medium, and 1 × 10⁵ cells per well were added to the upper chamber, while the lower chamber was filled with complete medium containing 10% FBS as a chemoattractant. Migration assays were incubated for 24 h, and invasion assays for 48 h. After incubation, non-migrated or non-invaded cells in the upper chamber were carefully removed using a cotton swab. Cells that had passed through the membrane to the lower chamber were fixed with 4% paraformaldehyde for 20 min and stained with 0.1% crystal violet for 15 min. Following washing and drying, five random fields per well were observed under a microscope to count the cells, and the numbers were used to compare the migration and invasion capacities between the siRBCK1 and siNC groups in both 786-O and 769-P cells.

CCK-8 assay

Exponentially growing 786-O and 769-P cells were transfected with siNC or siRBCK1 for 48 h, followed by trypsin digestion and cell counting. The cell density was adjusted to 3 × 10³ cells per well and evenly seeded into 96-well plates, with five replicates per group. After cell attachment, the plates were incubated at 37 °C with 5% CO₂. At predetermined time points (0, 24, 48, 72, and 96 h), 10 μL of CCK-8 reagent (Dojindo, Japan, CK04) was added to each well, and the plates were further incubated for 2 h. Absorbance (OD value) was measured at 450 nm using a microplate reader (Bio-Rad, USA). Care was taken to ensure uniform cell growth in each well and to avoid bubbles that could interfere with signal detection. To ensure data reliability, all experiments were independently repeated three times, with five replicates in each experiment. The mean OD values and standard deviations at each time point were calculated to assess differences in cell proliferation between the siRBCK1 and siNC groups in both 786-O and 769-P cells.

Statistical analysis

In this study, all data preprocessing and analyses were performed using R software (version 4.5.1), GraphPad Prism, and Perl. Pearson correlation analysis was applied to assess the relationships between different variables. Kaplan–Meier survival curves were compared using the log-rank test to evaluate survival differences among subgroups. Student’s t-test or the Wilcoxon rank-sum test was used to compare differences between two groups, while one-way analysis of variance (ANOVA) was employed for comparisons among multiple groups. To minimize false positives due to multiple testing, all P-values were adjusted using the Benjamini–Hochberg (BH) method. All statistical tests were two-sided, with a significance threshold set at P < 0.05. Experimental data are presented as mean ± standard deviation (Mean ± SD), with significance levels indicated as: *P < 0.05, **P < 0.01, ***P < 0.001.

Results

Differential expression and copy number variation of PRG signatures in ccRCC

In this study, a total of 14 PRG were collected to explore their potential roles in ccRCC. CNV analysis revealed notable CNV deletions in CASP8, NLRP3, CASP6, and TNFAIP3, whereas RNF31, RIPK3, and FADD exhibited significant CNV amplifications (Fig. 1A). Chromosomal localization analysis further illustrated the distribution of these 14 PRG across different chromosomes (Fig. 1B). PPI network analysis indicated significant interconnections among the 14 PRG (Fig. 1C). Moreover, differential expression analysis demonstrated that the PRG signature was generally upregulated in ccRCC tissues compared with normal renal tissues, including ZBP1, NLRP3, PYCARD, TNFAIP3, and RBCK1 (Fig. 1D).

Fig. 1.

Differential expression analysis and CNV frequency analysis of PANoptosis-related genes. A CNV frequency analysis of 14 PRG. B Chromosomal localization analysis of PRG. C PPI network of PRG signatures. D Differential expression analysis of PRG signatures between normal and tumor tissues. ns: not significant; *P < 0.05, **P < 0.01, ***P < 0.001

Identification of PRG molecular subtypes characterization in ccRCC

Based on the expression profiles of PRG signatures, we performed consensus clustering analysis to identify PRG-related molecular subtypes in ccRCC. Using the optimal parameters from the unsupervised consensus clustering model, ccRCC samples were classified into three molecular subgroups: subgroup A included 215 samples, subgroup B included 252 samples, and subgroup C included 63 samples (Fig. 2A). PCA plot demonstrated a clear spatial separation among the three PRG molecular subtypes, further highlighting their distinctiveness and independence (Fig. 2B). Clinical survival analysis revealed significant differences in OS rate among the PRG subgroups, with subgroups A and B showing markedly better outcomes than subgroup C (Fig. 2C). Based on the GSVA algorithm, we preliminarily explored the differential regulation of KEGG pathways among the PRG subgroups. Compared with subgroup A, multiple immune-related pathways were significantly downregulated in subgroup B, including the B cell receptor signaling pathway, NOD-like receptor signaling pathway, Toll-like receptor signaling pathway, and T cell receptor signaling pathway. Additionally, compared with subgroup C, subgroup B exhibited significant upregulation of pathways such as proximal tubule bicarbonate reclamation, fatty acid metabolism, and limonene and pinene degradation, whereas tumor-related pathways, including p53 signaling, cell cycle, DNA replication, and mismatch repair, were prominently activated in subgroup C (Fig. 2D, E). Heatmap visualization indicated differential expression of PRG signatures across clinical pathological features, with most PRG significantly upregulated in subgroup A (Fig. 2F). In summary, using consensus clustering analysis, we accurately identified PRG molecular subtypes in ccRCC and demonstrated a significant association between these molecular subgroups and clinical survival outcomes.

Fig. 2.

Identification of PRG molecular subtypes and analysis of potential regulatory mechanisms. A Identification of PRG molecular subtypes of ccRCC samples. B PCA illustrating the heterogeneity and independence among PRG molecular subtypes. C Kaplan–Meier survival analysis of different molecular subgroups. D, E Differential regulation analysis of KEGG pathway enrichment scores. F Expression patterns of PRG signatures across different clinicopathological features and molecular subtypes

Immune microenvironment landscape analysis of PRG subtypes

In recent years, immune checkpoint inhibitors have become a first-line therapy for ccRCC; however, the marked variability in treatment efficacy among patients suggests a complex stratification of the tumor immune microenvironment. To more accurately construct an immune classification system for different molecular subtypes of ccRCC, we systematically characterized the immune infiltration landscape and potential immunotherapy responses using multiple immune infiltration algorithms. The ESTIMATE algorithm results showed that in PRG subgroup B, the ESTIMATE score, stromal score, and immune score were significantly decreased, whereas tumor purity was markedly increased, indicating a potentially immunosuppressive state (Fig. 3A–D). ssGSEA enrichment analysis revealed that most immune cell infiltration levels were significantly reduced in subgroup B, including activated B cells, activated CD4⁺ T cells, activated CD8⁺ T cells, immature B cells, and macrophages (Fig. 3E). Immunotherapy response analysis indicated that, compared with subgroup C, IPS scores were significantly higher in subgroups A and B, indicating a potentially increased responsiveness to PD-1 and CTLA-4 blockade (Fig. 3F–H). Taken together, these results elucidate the immune infiltration characteristics of different PRG molecular subtypes and quantify the infiltration levels of various immune cell populations. Furthermore, we highlight the potential responsiveness of these molecular subtypes to immunotherapy, providing new insights for future clinical management of ccRCC.

Fig. 3.

Immune microenvironment infiltration landscape and immunotherapy response analysis of PRG molecular subgroups. A–D Quantitative assessment of immune infiltration status across molecular subgroups. E Quantitative analysis of infiltration proportions of 23 immune cell types. F–H Evaluation of IPS scores for different molecular subgroups

Identification of gene subtypes associated with PRG subgroups

To further elucidate the potential molecular mechanisms underlying the PRG molecular subtypes, we performed differential expression analysis among the three PRG subgroups and identified 1,615 DEGs (Fig. 4A). KEGG enrichment analysis indicated that these DEGs were involved in regulating pathways such as IgSF CAM signaling, Salmonella infection, Epstein–Barr virus infection, and protein processing in the endoplasmic reticulum, whereas GO enrichment analysis revealed that DEGs participated in biological processes including regulation of actin cytoskeleton organization, homeostasis of cell numbers, focal adhesion, and GTPase regulator activity (Fig. 4B, C). Based on the expression profiles of the DEGs, we applied a consensus clustering algorithm to identify gene-based subtypes. Using the optimal parameters of the unsupervised consensus clustering model, two gene subtypes were accurately defined: gene subtype A, comprising 171 samples, and gene subtype B, comprising 359 samples (Fig. 4D–F). PCA plot analysis demonstrated a distinct separation in distribution patterns between the two gene subtypes, highlighting their heterogeneity and independence (Fig. 4G). Kaplan–Meier survival analysis revealed that gene subtype A exhibited poorer clinical outcomes compared with gene subtype B (Fig. 4H). Furthermore, heatmap visualization showed the expression characteristics of the DEGs across different clinical features and molecular subgroups, indicating that the expression levels of DEGs were markedly elevated in gene subtype B (Fig. 4I).

Fig. 4.

Gene subtype analysis associated with PRG molecular subgroups. A Identification of DEGs among PRG molecular subgroups. B, C KEGG and GO enrichment analysis of DEGs. D–F Consensus clustering analysis based on DEGs expression profiles to identify gene subtypes. G PCA plot showing the distribution patterns of gene subtypes. H Clinical outcomes analysis of gene subtypes. I Heatmap showing DEG s expression across different subgroups and clinical pathological features

Independent prognostic value of PRG score and nomogram model construction

By integrating overall survival data and the expression profiles of DEGs from ccRCC samples, we applied a univariate Cox–LASSO regression approach to identify prognostic candidate variables. The results indicated that ZBP1 (HR = 1.541 [1.215–1.954], P < 0.001), PYCARD (HR = 1.487 [1.223–1.807], P < 0.001), and RBCK1 (HR = 1.920 [1.456–2.532], P < 0.001) were risk factors associated with unfavorable prognosis in ccRCC, whereas RIPK1 (HR = 0.750 [0.566–0.993], P = 0.045) was associated with improved clinical survival (Fig. 5A, B). Using multivariate Cox regression analysis, we further identified three independent prognostic variables from the four candidates and calculated the PRG score for each sample based on their corresponding regression coefficients. According to the optimal cutoff value determined by survival outcomes, ccRCC patients were stratified into two PRG scoring index subgroups (Fig. 5C). Kaplan–Meier survival analysis demonstrated that the low PRG scoring index subgroup exhibited significantly better survival outcomes compared with the high PRG scoring index subgroup (Fig. 5D, P < 0.001). The Sankey diagram illustrated the potential relationships among PRG subgroups, gene subgroups, PRG scoring index, and clinical outcomes (Fig. 5E). Additionally, we examined the distribution of PRG scores across different molecular and gene-based subtypes. The results showed that PRG scores were markedly elevated in molecular subtype C and gene subtype A—both associated with poorer prognosis—further confirming that higher PRG scoring index values are linked to adverse clinical outcomes in ccRCC (Fig. 5F, G).

Fig. 5.

Construction of the PRG score system based on prognostic signatures. A, B Univariate Cox-LASSO analysis for identifying prognostic signature variables. C Stratification of PRG score subgroups. D Clinical outcomes analysis of PRG score subgroups. E Sankey diagram illustrating the relationships among molecular subgroups, PRG score subgroups, and clinical outcomes. F, G Differential expression analysis of PRG scores across different subgroups

Validation of the PRG scoring system in independent cohorts

In the subsequent analyses, we further evaluated the independent prognostic value of clinical pathological features and the PRG scoring index. Univariate Cox regression revealed that age (HR = 1.023 [1.005–1.041], P = 0.012), grade (HR = 2.242 [1.682–2.988], P < 0.001), stage (HR = 1.862 [1.541–2.251], P < 0.001), T stage (HR = 1.943 [1.538–2.456], P < 0.001), M stage (HR = 4.073 [2.634–6.300], P < 0.001), N stage (HR = 2.932 [1.516–5.668], P < 0.001), and PRG score (HR = 1.583 [1.253–2.000], P < 0.001) were all significantly associated with unfavorable prognosis in ccRCC (Fig. 6A). Multivariate Cox regression further identified grade (HR = 1.404 [1.008–1.955], P = 0.045) and PRG score (HR = 1.354 [1.044–1.758], P = 0.023) as independent prognostic indicators for predicting clinical outcomes in ccRCC (Fig. 6B). Based on these findings, we constructed a nomogram incorporating pathological variables and the PRG scoring index to estimate 1-, 3-, and 5-year survival probabilities for patients with ccRCC (Fig. 6C). Calibration curve analysis demonstrated that the predicted survival probabilities generated by the nomogram were highly consistent with the observed outcomes (Fig. 6D). Taken together, these results indicate that the PRG scoring index, as an independent prognostic factor, provides prognostic information beyond traditional clinicopathological features, and the integrated nomogram model offers accurate estimation of survival probabilities across different time horizons in patients with ccRCC.

Fig. 6.

Independent prognostic value evaluation and nomogram prediction model construction. A, B Univariate and multivariate Cox analyses based on clinical pathological variables and PRG score. C Construction of a nomogram prediction model integrating clinical pathological features and PRG score. D Calibration curve analysis

Validation of the independence and stability of the PRG score in predicting clinical prognosis

We further validated the stability and consistency of the PRG score system in predicting clinical outcomes of ccRCC in two independent cohorts. Based on a 7:3 random split ratio, all samples were divided into a training set and a validation set. The PRG scoring system was established in the training cohort according to prognostic variables. Using the optimal cutoff values of the PRG score, patients in both cohorts were stratified into low– and high–PRG score subgroups (Fig. 7A, B). Kaplan–Meier survival analysis demonstrated that patients in the low PRG score subgroup exhibited significantly better overall survival than those in the high PRG score subgroup in both the training and validation cohorts (Fig. 7C, D). Time-dependent ROC analyses indicated that the AUCs for predicting 1-, 3-, and 5-year survival were 0.687, 0.662, and 0.662 in the training cohort, and 0.677, 0.617, and 0.640 in the validation cohort, respectively (Fig. 7E, F). Collectively, these results confirm that the PRG score system can serve as a reliable tool for risk stratification and accurate prognostic prediction in patients with ccRCC.

Fig. 7.

Consistency validation of the PRG score system in training and validation cohorts. A, B Construction of PRG score subgroups in the training and validation cohorts. C, D Kaplan–Meier survival analysis of PRG score subgroups in the training and validation cohorts. E, F Time-dependent ROC curve analysis

Immune infiltration landscape and drug sensitivity associated with PRG score

We further investigated the potential association between the PRG scoring index and the immune infiltration landscape of the ccRCC tumor microenvironment. ESTIMATE analysis revealed that, compared with the low PRG score subgroup, the high PRG score subgroup exhibited significantly higher ESTIMATE scores, immune scores, and stromal scores, whereas tumor purity was markedly lower (Fig. 8A–D). Results from ssGSEA indicated that several immune cell populations including activated B cells, activated CD4⁺ T cells, activated CD8⁺ T cells, activated dendritic cells, and CD56^bright natural killer cells were significantly enriched in the high PRG score subgroup. In contrast, eosinophils and neutrophils displayed markedly reduced infiltration levels (Fig. 8E). Using correlation analysis, we further examined the relationship between the PRG scoring index, prognostic signatures, and immune infiltration. The results showed that PYCARD was strongly and positively correlated with most immune cell types, whereas RBCK1 showed strong negative correlations with the majority of immune infiltrating cells (Fig. 8F). Drug sensitivity analysis demonstrated that, compared with the high PRG score subgroup, the low PRG score subgroup exhibited significantly lower IC50 values for Rapamycin, Saracatinib, S-Trityl-L-cysteine, and Z-LLNle-CHO, suggesting that patients with low PRG scores may derive greater therapeutic benefit from these agents (Fig. 8G–J). Collectively, these findings elucidate the relationship between the PRG scoring index and the immune microenvironment in ccRCC and identify several small-molecule compounds with potential therapeutic value, thereby providing novel insights for future personalized treatment strategies.

Fig. 8.

Immune microenvironment infiltration landscape and drug sensitivity analysis of PRG score subgroups. A–D Immune infiltration characteristics of PRG score subgroups. E Quantitative assessment of infiltration proportions of 23 immune cell types. F Correlation analysis between the PRG score, prognostic signatures, and 23 immune-infiltrating cells. G–J Drug sensitivity prediction analysis

Immunotherapy response and tumor mutation burden features

Immunotherapy response analysis based on IPS scores derived from the TCIA database showed that the low PRG score subgroup exhibited significantly higher IPS values, indicating a potentially increased responsiveness to PD-1 and CTLA-4 blockade (Fig. 9A–D). Somatic mutation burden analysis further demonstrated that 75% of samples in the low PRG score subgroup harbored somatic mutations, whereas 83.33% of samples in the high PRG score subgroup exhibited somatic mutations. Notably, VHL (low: 38% vs. high: 46%), TTN (low: 12% vs. high: 20%), SETD2 (low: 6% vs. high: 17%), and BAP1 (low: 3% vs. high: 17%) showed higher mutation frequencies in the high PRG score subgroup, while PBRM1 (low: 38% vs. high: 33%) exhibited a higher mutation frequency in the low PRG score subgroup (Fig. 9E, F).

Fig. 9.

Immune therapy response assessment and mutation burden landscape analysis. A–D IPS score analysis of PRG score subgroups. E, F Somatic mutation burden landscape of PRG score subgroups

Single-cell sequencing reveals cellular subpopulations and prognostic signature expression

Based on single-cell sequencing data, we further evaluated the cellular subpopulation landscape of ccRCC and characterized the expression patterns of prognostic signatures across distinct cell subsets. Using the scRNA-seq dataset GSE304466, we extracted three high-quality ccRCC single-cell profiles for downstream analyses. After performing rigorous quality control and normalization, the top 2,000 highly variable genes were identified for subsequent dimensionality reduction (Fig. 10A, B). According to canonical marker genes for each cell type, a total of 31 cell types were annotated, and violin plot analyses indicated that RBCK1, PYCARD, and RIPK1 were broadly expressed across these 31 cell populations (Fig. 10C). UMAP and t-SNE projections further illustrated the distribution patterns of the 31 identified cell types (Fig. 10D, E). Using the SingleR annotation algorithm, we accurately classified these 31 cell populations into six major cell subgroups, including Macrophages, Endothelial cells, T cells, Tissue stem cells, Monocytes, and Epithelial cells. UMAP and t-SNE analyses were used to visualize the characteristics and spatial distribution of these six subpopulations (Fig. 10F, G). Furthermore, UMAP-based expression profiling revealed that RIPK1 was highly expressed in Endothelial cells, Tissue stem cells, Macrophages, and T cells; PYCARD showed dominant expression in Macrophages; and RBCK1 was markedly expressed in Tissue stem cells, Macrophages, and Epithelial cells (Fig. 10H–J).

Fig. 10.

Single-cell sequencing analysis revealing cell subpopulation classification and prognostic signature expression profiles. A Quality control and normalization of single-cell sequencing data. B Identification of the top 2,000 highly variable genes. C Expression levels of prognostic signatures across different cell types. D, E UMAP and tSNE plots showing the distribution of 31 cell types. F, G UMAP and tSNE plots showing the distribution of annotated cell subpopulations. H–J UMAP plots displaying the expression profiles of prognostic signatures across different cell subpopulations

RBCK1 knockdown suppresses ccRCC cell proliferation and migration

During the construction of the PRG scoring system, we observed that among the four prognostic signatures, RBCK1 exhibited the highest risk coefficient, suggesting that it may serve as a critical driver of poor prognosis in ccRCC. Therefore, we further investigated its regulatory role through in vitro experiments. Differential expression analysis showed that RBCK1 expression was significantly elevated in tumor tissues compared with normal tissues (Fig. 11A). Furthermore, we assessed the association between RBCK1 expression and clinical outcomes in ccRCC. The results demonstrated that patients with high RBCK1 expression had markedly reduced disease-specific survival (DSS), progression-free interval (PFI), and overall survival (OS) rates compared with those in the low-expression group, indicating that RBCK1 functions as a risk factor associated with unfavorable prognosis in ccRCC (Fig. 11B–D). Analysis of the CPTAC dataset further confirmed that RBCK1 protein expression was significantly higher in ccRCC tissues than in normal tissues (Fig. 11E). In vitro validation using cell lines revealed consistent findings. Western blot analysis demonstrated that RBCK1 protein levels were substantially increased in 786-O and 769-P ccRCC cells compared with normal HK2 cells (Fig. 11F, G). To further elucidate the functional role of RBCK1 in ccRCC, we constructed an siRBCK1 knockdown model in 786-O and 769-P cells using siRNA interference and confirmed efficient silencing (Supplementary Fig. 1A, B). Functional assays showed that RBCK1 knockdown markedly suppressed the proliferation and migration capabilities of 786-O and 769-P cells, as demonstrated by colony formation and Transwell assays (Fig. 11H-K). Additionally, CCK-8 assays indicated that silencing RBCK1 significantly reduced cell viability at 24, 48, 72, and 96 h compared with the siNC group in 786-O and 769-P cells (Fig. 11L, M). Taken together, these findings demonstrate that RBCK1 acts as a high-risk factor associated with poor prognosis in ccRCC. Moreover, RBCK1 inhibition significantly impairs ccRCC cell proliferation and migration, highlighting its potential biological role in the development and progression of ccRCC and supporting its value as a potential therapeutic target.

Fig. 11.

Knockdown of RBCK1 significantly inhibits ccRCC cell proliferation and migration. A Differential expression analysis of RBCK1 between normal and tumor groups. B–D DSS, PFI, and OS survival curves for low- and high-RBCK1 expression subgroups. E Protein expression levels of RBCK1 in normal and tumor groups based on the CPTAC database. F Protein expression levels of RBCK1 in HK2 cells and 786-O cells with quantitative analysis (n = 3). G Protein expression levels of RBCK1 in HK2 cells and 769-P cells with quantitative analysis (n = 3). H, I Colony formation assay and Transwell migration assay in 786-O cells (n = 3). J, K Colony formation assay and Transwell migration assay in 769-P cells (n = 3). L, M CCK-8 assay showing cell viability at different time points in 786-O and 769-P cells (n = 3). Data are presented as mean ± SD; significance: *P < 0.05, **P < 0.01, ***P < 0.001

Discussion

This study highlights PANoptosis-related regulators as important contributors to the molecular heterogeneity and immune landscape of ccRCC. By integrating multi-omics analyses with functional validation, we provide a framework linking PANoptosis activity to prognosis and therapeutic stratification.

Our findings further support the potential involvement of PANoptosis in ccRCC. Previous studies have reported that quantifying PANoptosis-related processes can predict immunotherapy efficacy and prognosis in gastric cancer and colorectal cancer. However, the mechanisms through which PANoptosis influences patient prognosis in ccRCC remain largely undefined [23, 24]. One possible mechanism is through modulation of therapeutic resistance. Our results suggest that PANoptosis-based molecular subtyping may affect drug resistance patterns in ccRCC. Alterations in programmed cell death pathways are recognized drivers of resistance to therapy, enabling cancer cells to evade drug-induced cytotoxicity [25]. Several proteins including cysteine desulfurase NFS1, the m6A methyltransferase WTAP, and Y box binding protein 1 (YBX1) have been reported to influence chemotherapeutic responses through PANoptosis-related mechanisms [26–28]. To date, no studies have examined PANoptosis in ccRCC-associated therapeutic resistance. Notably, YBX1 has been shown to be critical for acquired drug resistance in metastatic ccRCC, implying that PANoptosis-related processes may contribute to resistance development and, consequently, to inter-patient prognostic differences [29].

We also demonstrated that higher RBCK1 expression is associated with poorer PFS and OS in ccRCC. RBCK1 siRNA knockdown experiments in the 769-P cell line yielded results highly consistent with those observed in 786-O cells, thereby further supporting the tumor-promoting role of RBCK1 in ccRCC and strengthening the robustness of our conclusions. Emerging evidence indicates that E3 ubiquitin ligases can influence PANoptosis by regulating the stability of key upstream sensors such as ZBP1, suggesting that ubiquitin-dependent protein turnover may modulate PANoptosis signaling [30, 31]. RBCK1 has been reported to participate in immune regulation and tumor progression in ccRCC [32], including shaping an immunosuppressive tumor microenvironment characterized by reduced infiltration of CD4⁺ T cells and M1 macrophages, which is consistent with our observations [33]. Beyond immune modulation, RBCK1 can promote tumor progression by enhancing p53 polyubiquitination and degradation, thereby facilitating renal cancer cell proliferation in vitro and in vivo [34], and has also been implicated in reduced sensitivity to sunitinib treatment [35]. Notably, RBCK1 has been shown to promote the degradation of MFN2, a key regulator of mitochondrial dynamics [36], while MFN2 has recently been identified as an important mediator of PANoptosis [37]. Taken together, these lines of evidence do not establish a direct mechanistic role for RBCK1 in PANoptosis, but collectively suggest that RBCK1 may be indirectly linked to PANoptosis-related processes through ubiquitin-dependent regulation of mitochondrial dynamics and immune-associated pathways. Such an association provides a biologically plausible hypothesis to explain the observed correlation between elevated RBCK1 expression and unfavorable clinical outcomes in ccRCC.

RIPK1 and PYCARD represent two core components of the PANoptosis machinery with established roles in inflammatory cell death and immune regulation [38–41]. In the context of ccRCC, their inclusion in the prognostic signature suggests that coordinated activation of PANoptosis-related signaling may influence tumor progression through both cell death regulation and modulation of the tumor immune microenvironment. Notably, RIPK1 has been implicated in regulating necroptosis and therapeutic sensitivity in renal cancer [42, 43], whereas PYCARD is closely associated with inflammasome activity and immune cell infiltration [44], [45]. Together, these findings support the notion that RIPK1 and PYCARD may function as key molecular links between inflammatory cell death pathways and immune-related heterogeneity in ccRCC, warranting further mechanistic investigation.

In this study, we observed that the high PRG score subgroup exhibited higher immune and stromal scores; however, it was associated with poorer prognosis and a lower predicted response to immunotherapy. This seemingly paradoxical finding suggests that the quantity of immune cell infiltration within the tumor microenvironment does not necessarily translate into effective antitumor immune activity [46–48]. Tumors with high PRG scores may harbor an immunosuppressive or functionally exhausted inflammatory microenvironment, characterized by the enrichment of regulatory T cells, M2-polarized macrophages, or fibroblast-associated signaling pathways, which collectively impair the cytotoxic function of effector immune cells [49, 50]. Moreover, an abundant stromal component may act as a physical barrier or exert paracrine effects that restrict effective immune cell infiltration and activation, thereby facilitating tumor progression and reducing sensitivity to immunotherapy [51, 52]. Importantly, these inferences are primarily derived from computational analyses and immune infiltration scores obtained from public databases, and thus require further validation in future studies using single-cell sequencing, immunohistochemistry, or functional immune assays to more comprehensively elucidate the relationship between PRG scores and the immune status of the tumor microenvironment.

Pathway enrichment analyses identified immunoglobulin superfamily (IgSF) cell adhesion molecule (CAM) signaling as the most prominently enriched pathway among PRG molecular-related DEGs. IgSF CAMs regulate key biological processes such as cell proliferation, differentiation, and morphogenesis [53]. In cancer cells, IgSF CAMs are frequently overexpressed and can exert protumor effects by activating tyrosine kinase receptors—including FGF, EGF, and NGF receptors—as well as the Wnt/β-catenin signaling pathway [54]. Numerous CAM-related proteins have been implicated in ccRCC cell survival, migration, and invasion [55–58]. Our findings reinforce the potential role of IgSF CAMs in ccRCC. Given their limited distribution in normal tissues, IgSF CAMs represent promising candidates for the development of targeted therapies [59].

Several limitations of this study should be acknowledged. First, the analyses were mainly based on retrospective public datasets, and the lack of an independent external validation cohort may limit the generalizability of our findings, despite the use of rigorous internal validation strategies. Future studies incorporating large, multi-center cohorts are required to further confirm the robustness and clinical applicability of the PANoptosis-related score. Second, although RBCK1 was functionally validated in vitro, RIPK1 and PYCARD were not experimentally explored in the present study, largely due to scope and feasibility constraints. Given their established involvement in necroptosis, inflammasome activation, and immune regulation, further mechanistic investigations are warranted to clarify their precise roles in PANoptosis and ccRCC progression. Third, while single-cell RNA sequencing provided valuable cell-type specific insights, the limited number of ccRCC samples may introduce sampling bias and fail to fully capture inter-patient heterogeneity. Therefore, conclusions derived from the single-cell analyses should be interpreted with caution and validated in larger single-cell or spatial transcriptomic datasets. In addition, since immunotherapy response was inferred using IPS scores obtained from the TCIA database rather than real-world clinical response data, the immunotherapy response results should be interpreted as predictive inference rather than direct clinical validation, and require confirmation in prospective cohorts of ccRCC patients receiving immune checkpoint inhibitors.

Author contributions

W.G. and X.J. conceived and designed the study. P.Z. performed most of the analyses and wrote the manuscript. Z.X. performed some bioinformatic analyses and in vitro analyses. P.Z. and Z.X. contributed equally to this work and share first authorship. All authors participated in the acquisition, analysis, interpretation of data, drafting of the manuscript, and approval of the submitted version.

Funding

No funding was received for conducting this study.

Data availability

All datasets used in this study are publicly available. Bulk RNA-seq data and clinical information for ccRCC were obtained from TCGA-KIRC through the GDC Data Portal (https://portal.gdc.cancer.gov/projects/TCGA-KIRC). Immune-related data were downloaded from TCIA corresponding to the TCGA-KIRC cohort (https://tcia.at/home). Drug sensitivity profiles were retrieved from the Genomics of GDSC database (cancerrxgene.org/downloads/bulk\_download). Single-cell RNA sequencing data for ccRCC were obtained from the Gene Expression Omnibus under accession number GSE304466 (https://www.ncbi.nlm.nih.gov/gds/?term=GSE304466). All external datasets were used in accordance with their respective access and licensing policies.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.