Deciphering mitotic catastrophe–associated transcriptomic patterns in renal cell carcinoma: prognostic significance and immunotherapy-related associations

Abstract

Background

Renal cell carcinoma (RCC) is a highly heterogeneous malignancy with complex molecular features. Mitotic catastrophe (MC), a key regulator of cell fate during aberrant mitosis, contributes to tumor progression and therapeutic response; however, its prognostic relevance in RCC remains unclear.

Methods

We integrated multi-omics, clinical, and imaging data from TCGA and multiple external cohorts. Prognostic mitotic catastrophe–related genes were identified to construct a mitotic catastrophe–related score (MCRS) using CoxBoost and random survival forest models. Associations with clinicopathologic features, immune contexture, and mutational profiles were analyzed. Single-cell RNA sequencing characterized Cyclin F (CCNF) expression and its association with immunotherapy-treated cohort outcomes, and these findings were further validated in vitro and in murine xenograft models.

Results

The nine-gene MCRG signature stratified patients into high- and low-risk groups with strong prognostic performance (TCGA 1-, 3-, 5-year AUCs > 0.9). High MCRS correlated with advanced stage, adverse CT features, angiogenesis, inflammatory signaling, regulatory T-cell infiltration, and elevated LAG3/TIGIT expression. MTOR mutations were enriched in high-risk tumors. SurvSHAP(t) analysis identified CCNF as the top prognostic contributor. CCNF was upregulated in tumor tissues and enriched in proliferating cells and epithelial cells. High CCNF expression was associated with worse survival and inferior outcomes in immunotherapy-treated cohorts. Overexpression of CCNF promoted RCC cell proliferation, migration, invasion, and in vivo tumor growth, partly through mTOR pathway activation.

Conclusions

We established a robust MC-based prognostic model for RCC and identified CCNF as a key regulator of tumor progression and the immune microenvironment, characterized by increased Treg abundance and checkpoint expression. CCNF may represent a candidate biomarker and therapeutic target in RCC, although its relevance to outcomes in immunotherapy-treated cohorts requires further validation.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13062-026-00863-8.

Highlights

A mitotic catastrophe–based transcriptomic model stratifies prognosis and is associated with immunotherapy-related outcomes in RCC.

High MCRS is associated with aggressive clinicopathologic features, angiogenesis, and immunosuppressive immune infiltration.

CCNF emerges as a central gene associated with mitotic dysregulation and an immune-enriched tumor microenvironment.

CCNF may promote RCC progression partly through mTOR-related signaling.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13062-026-00863-8.

Introduction

According to the global cancer statistics (2022), Renal cell carcinoma (RCC) ranks among the top ten most frequently diagnosed cancers and is a leading cause of cancer-related deaths in the urinary system, with clear cell RCC (ccRCC) comprising nearly 70–80% of all RCC cases [1, 2]. Despite advances in genomics, immuno-oncology, and targeted therapeutics that have substantially improved our understanding of RCC biology and the tumor microenvironment, the prognosis for advanced RCC remains unsatisfactory, with a 5-year survival rate below 15% in metastatic settings [2, 3]. Resistance to therapy and the pronounced heterogeneity of RCC further complicate treatment strategies and patient management [4, 5]. Therefore, there is an urgent need to identify reliable molecular biomarkers and construct robust prognostic models to stratify patient risk and guide individualized therapy. Notably, emerging studies have highlighted the importance of the cell cycle, particularly mitosis-related pathways, in renal cell carcinoma progression and treatment response [6, 7]. However, studies of mitosis, including examination of mitotic gene signatures and their prognostic value in renal cell carcinoma, remain limited.

Uncontrolled proliferation is a fundamental hallmark of cancer, often driven by dysregulated mitotic signaling [8, 9]. Mitosis, as a central component of the eukaryotic cell cycle, involves the precise coordination of DNA replication and nuclear division to ensure faithful cell division. However, in malignant tumors, mitotic regulation is frequently disrupted, resulting in hyperactivation of the cell cycle and abnormal cell division [10]. Mitotic catastrophe (MC), which occurs during dysregulated mitosis, represents a fascinating strategy to specifically eradicate tumor cells [11]. MC ultimately leads to irreversible cell cycle arrest or cell death via apoptosis, necrosis, autophagy, or senescence. Morphologically, MC is characterized by features such as multinucleation, micronuclei formation, polyploidy, and multipolar spindles [12]. Recent studies have highlighted the relevance of MC in cancer therapy, as it offers a distinct mechanism for eliminating tumor cells—especially those resistant to canonical apoptosis [13, 14]. For example, targeting mitotic catastrophe-related targets can effectively ameliorate tumor progression in breast cancer and colorectal cancer.

Furthermore, several prognostic models based on MC-related genes have been developed in cancers such as hepatocellular carcinoma, breast cancer, and bladder cancer, revealing strong associations with both clinical outcomes and immune landscape heterogeneity [15–17]. Despite its pivotal biological role and therapeutic potential, the significance of MC in RCC—particularly in the clear cell subtype—remains underexplored. No prognostic stratification model has yet been established based on MC-related gene signatures in RCC.

Materials and methods

Data acquisition

Gene expression profiles, somatic SNV data, and clinical information for ccRCC, pRCC, and chRCC patients were obtained from TCGA via the UCSC Xena platform. A multicenter cohort of 54 fumarate hydratase-deficient RCC (FHRCC) cases (WCH-FHRCC) with RNA-Seq and clinical data was also analyzed. FH deficiency was diagnosed based on IHC evidence of FH loss and/or 2SC positivity, assessed by two experienced genitourinary pathologists (N. Chen and L. Nie), and further confirmed by DNA sequencing. Variables, including age, sex, and TNM stage, were extracted from medical records and summarized in Table 1. We also included a TFE3-rearranged RCC cohort (WCH-TFE3), in which the diagnosis was established by morphologic evaluation, immunohistochemistry, and molecular confirmation of TFE3 rearrangement. The transcriptomic data of TCGA-KIRC, TCGA-KIRP, TCGA-KICH, E-MTAB-1980, CheckMate, and IMmotion151 were obtained from the following repositories:

Table 1.

Baseline clinic characteristics of WCH-FHRCC(n = 54)

Characteristics	MCRS-Low (n = 27)	MCRS-High (n = 27)	P value
pT stage, n (%)			0.171
T < 3	15 (55.6%)	9 (33.3%)
T ≥ 3	12 (44.4%)	18 (66.7%)
pN stage, n (%)			0.271
N0	18 (66.7%)	13 (48.1%)
N1	9 (33.3%)	14 (51.9%)
pM stage, n (%)			0.028
M0	20 (74.1%)	11 (40.7%)
M1	7 (25.9%)	16 (59.3%)
Gender, n (%)			0.773
Female	8 (29.6%)	10 (37.0%)
Male	19 (70.4%)	17 (63.0%)
Age, n (%)			0.119
< 60	27 (100%)	23 (85.2%)
≥ 60	0 (0%)	4 (14.8%)

https://portal.gdc.cancer.gov,

https://www.ebi.ac.uk/biostudies/arrayexpress, https://pubmed.ncbi.nlm.nih.gov/32472114/,

https://ega-archive.org/studies/EGAS00001004353.

Additionally, single-cell RNA sequencing (scRNA-seq) data generated using the 10x Genomics platform were obtained from the GEO database (GSE237425, https://www.ncbi.nlm.nih.gov/). The list of mitosis- and apoptosis-related genes was retrieved from the GeneCards database (https://www.genecards.org/) and is provided in Supplementary Table S1. This study was approved by the Ethics Committee of West China Hospital. Imaging data for RCC patients were obtained from the TCIA database (n = 237; http://www.cancerimagingarchive.net/). The exclusion criteria used to select images for further analysis were specified as follows: (1) non-contrast scans, (2) postoperative scans, and (3) poor-quality scans. Additionally, imaging data without corresponding matched clinical and genomic data in the TCGA-KIRC cohort were also excluded, resulting in a final integrated cohort of 194 ccRCC cases in TCIA-KIRC.

Publicly available spatial transcriptomic data were retrieved from the RCC_Pmid35231421 dataset [18], generated using the 10x Genomics Visium platform. The analyzed specimen (sample ID: c45) corresponded to a primary ccRCC. Spatial gene expression matrices and matched H&E images were integrated for downstream visualization and gene set activity analysis.

Identification of differentially expressed genes (DEGs)

Throughout the study, the “limma” R package was used for differential binning of bulk RNA-seq, with thresholds of an absolute log2FC > 1 and an adjusted P-value < 0.05. The DEGs used for the enrichment analysis were obtained from the differential expression analysis comparing the RCC and normal kidney groups in the TCGA-KIRC and WCH-TFE3 cohorts.

Identification of core modules for DEGs

To refine the DEGs between RCC and normal tissues and highlight those most relevant to tumor phenotypes, we intersected the list of mitotic catastrophe-related genes with prognosis-associated genes derived from the IMmotion151, WCH-TFE3, and TCGA-KIRC cohorts.

Optimal prognostic modeling

Based on the aforementioned core genes, we integrated 10 machine learning algorithms to construct prognostic models [19]. The TCGA-KIRC cohort served as the training set, while the E-MTAB-1980 and WCH-FHRCC cohorts were used as independent external validation sets. The algorithm with the highest average C-index across the three datasets, including the training cohort, was selected as the optimal strategy. Patients in each cohort were stratified into high- and low-risk groups from the training set. The prognostic performance of the model was evaluated using Kaplan–Meier survival analysis and time-dependent ROC curves. Finally, the independence of the risk score as a prognostic factor for RCC patients was confirmed through multivariable Cox regression analysis. The detailed flowchart is shown in Figure S1.

CT imaging assessment and analysis

For the TCGA-KIRC cohort, two abdominal radiologists, blinded to clinical data and survival outcomes, reviewed all CT images. The following imaging features were assessed: (1) tumor size; (2) tumor architecture; (3) tumor margin; (4) spiculated margin; (5) enhancement pattern; (6) intratumoral calcification; (7) intratumoral artery (IA); (8) necrosis; (9) renal vein invasion, and (10) collecting system invasion. Definitions of imaging features are presented in Table S1. All examinations were performed using a workstation and recorded on a picture archiving and communication system.

Single-cell analysis

Single-cell analysis was conducted within the R statistical environment (version 4.2.1). The “Seurat” package facilitated raw data processing and subsequent quality control, including the inclusion of genes detected in at least 3 cells and the exclusion of cells with more than 10% mitochondrial reads. Cells exhibiting fewer than 300 or more than 5000 genes were also removed. Harmony was deployed to mitigate batch effects. Cellular annotations were manually assigned based on known gene expression profiles, with specific marker genes detailed in Table S2.

Differential genes and functional enrichment

Throughout the study, the “limma” R package was used for differential binning of bulk RNA-seq data. The analysis of DEGs of tumor versus normal kidney tissue was based on the WCH-TFE3, TCGA-KIRC, TCGA-KIRP, and TCGA-KICH cohorts, and enrichment analysis was performed using the “ssGSEA” R package.

Immune microenvironment analysis

The relative proportions of 22 immune cell types within the tumor microenvironment were estimated using the CIBERSORT algorithm. The LM22 signature matrix was used to deconvolute bulk RNA-seq expression data from WCH-FHRCC and TCGA-KIRC. Only samples with a CIBERSORT output p-value less than 0.05 were considered robust and included in subsequent analyses.

Human samples and cell lines

The RCC and normal kidney tissues used in this study were obtained from the Department of Urology at West China Hospital of Sichuan University, with informed consent from all subjects. The clinical information (sex, age, CCNF expression) of 15 patients.

The cell lines used in this study, including Caki-1 and OS-RC-2, were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). Cells were maintained in RPMI Medium 1640 (Gibco, USA) supplemented with 10% fetal bovine serum (Pricella, China) and 1% penicillin /streptomycin (Solarbio, China) and mycoplasma removal agent (C0288M, Beyotime, China) in a 5% CO₂ incubator (Thermo Fisher Scientific, USA) at 37 °C.

Multiplex immunofluorescence staining

Multiplex immunofluorescence staining was performed on paraffin-embedded tumor and normal kidney tissue sections to assess the spatial distribution of CCNF. Tissue sections were deparaffinized, rehydrated, and subjected to antigen retrieval using citrate buffer (pH 6.0). After blocking with 5% bovine serum albumin (BSA) in phosphate-buffered saline (PBS) for 1 h at room temperature, sections were incubated overnight at 4 °C with primary antibodies: anti-Cyclin F antibody (1:100, CPA1172, Cohesion). Following primary antibody incubation and thorough washing, sections were probed with species-specific conjugated secondary antibodies (rabbit, 1:100, GB22303; mouse, 1:100, GB21301, Servicebio) for 1 h at room temperature in the dark. Nuclei were counterstained with 4’,6-diamidino-2-phenylindole (DAPI, G1012, Servicebio). Stained sections were visualized and imaged using a digital slide scanner (P250 FLASH, Jinan Tangier Electronics Co., Ltd.). Quantification of fluorescence intensity of CCNF signals was performed using ImageJ software (National Institutes of Health), and the results of the evaluation are in Table S3.

Plasmid construction and lentiviral transduction

The lentiviral packaging CCNF shRNAs were purchased from Tsingke (Beijing, China) and then cloned into the pLVX-puro vector (Addgene, #8453) to construct the overexpression plasmids, pLVX-CCNF. The shRNA sequences are listed in Table S4. Caki-1 and OS-RC-2 were infected at a multiplicity of infection of 10 in the presence of 8 µg/mL polybrene, and stably transduced cells were selected with puromycin (2 µg/mL) for 5–7 days.

Western blot analysis

Cellular proteins were extracted with RIPA lysis buffer (Beyotime, P0013B) supplemented with protease and phosphatase inhibitors (Roche, 04693116001, 04906845001). Protein concentrations were measured using the bicinchoninic acid (BCA) assay (ThermoFisher, 23227). SDS-PAGE with a 15% polyacrylamide gel was used to separate total proteins, which were subsequently transferred onto a PVDF membrane (Millipore, IPVH00010). The PVDF membrane was blocked with 5% skim milk for 1 h, then incubated with the primary antibodies: anti-Cyclin F (1:1000, CPA1172, Cohesion) and anti-GAPDH (1:5000, 10494-1-AP, Proteintech) overnight at 4℃. After being washed three times with TBST buffer, the PVDF membrane was incubated with the corresponding secondary rabbit antibody (1:5000, ab97051, Abcam) for 2 h at room temperature. The blots were visualized with an ultra-sensitive ECL chemiluminescence kit (#P0018, Beyotime) using Bio-Rad ChemiDoc Touch (Bio-Rad, USA).

Cell proliferation and colony formation assay

For EdU stain (APExBIO, K1076), Caki-1 and OS-RC-2 cells were seeded into 24-well plates containing coverslips and incubated for 24 h. Following this, the cells were incubated for 2 h with medium containing 50 µM EdU. The coverslips were then removed, and the cells were fixed with 4% paraformaldehyde. Staining was performed according to the manufacturer’s instructions, and imaging was conducted using an inverted microscope.

For the colony formation assay, a total of 1,000 Caki-1 and OS-RC-2 cells in the logarithmic growth phase were seeded into 6-well plates, with the culture medium replaced every 3 days. After 2 weeks of incubation, the cells were fixed with paraformaldehyde and stained with crystal violet. Cell colonies with > 50 cells were counted.

Invasion assay

Caki-1 and OS-RC-2 cells were seeded into Boyden chambers with 8 μm pore membranes in 24-well Transwell plates (Corning, 3526). Matrigel-coated was used for the invasion assay. Tumor cells were seeded into the upper chamber with 200 µL of serum-free medium, while 500 µL of medium containing 10% FBS was added to the lower chamber. After incubation for the appropriate duration, the chambers were fixed with 4% paraformaldehyde for 30 min and stained with crystal violet. Cells in the upper chamber were removed, and imaging was conducted using an inverted microscope.

Wound healing assay

Caki-1 and OS-RC-2 cells were seeded into 6-well plates at a density of 5 × 10⁵ cells/well and cultured in complete medium until reaching 90–100% confluence. A linear scratch was introduced using a sterile 200 µL pipette tip to create a wound across the cell monolayer. Detached cells and debris were gently washed away with phosphate-buffered saline (PBS), and cells were then cultured in serum-free medium to minimize proliferation. Images of the wound area were captured at 0 h and 24 h using an inverted microscope (×100 magnification).

The wound closure rate was quantified using ImageJ software by measuring the area of the wound at each time point. The relative migration rate was calculated using the following formula: Wound Closure (%) = [(Area₀h – Area₂₄h) / Area₀h] × 100%.

Murine xenografts

Five-week-old female BALB/c nude mice (GemPharmatechCo., Ltd, China) were acclimatized for one week under specific pathogen-free conditions prior to experimentation. After acclimation, twelve mice were randomly assigned to two groups (n = 5 per group) and subcutaneously injected in the right flank with 1 × 10⁶ Caki-1 cells stably overexpressing CCNF or control cells, respectively. The cells were suspended in 200 µL of phosphate-buffered saline (PBS) containing 50% Matrigel (C0387, Beyotime). Tumor volumes were measured every three days using a digital caliper and calculated with the formula:

V = (length × width²) / 2.

According to the approved protocol, the maximal permitted tumor burden was 1,500 mm³ (or a maximal diameter of 15 mm). Humane endpoints included ulceration, necrosis, impaired mobility, body weight loss exceeding 20%, or other signs of distress. Mice were euthanized when tumors appeared. Tumor growth was monitored for 5 weeks.

Statistical analysis

All statistical analyses were performed using R software version 4.2.1. A P-value < 0.05 was considered statistically significant. Multiple testing correction was conducted using the Benjamini-Hochberg (BH) method to obtain adjusted P-values. The log-rank test was employed for comparisons between two groups, unless otherwise specified. Correlation analyses were conducted using Spearman’s rank correlation coefficient.

Results

Identification of prognosis-associated MCRGs

Based on gene expression profiles from cancerous and adjacent normal kidney tissues in the WCH-TFE3, TCGA-KIRC, TCGA-KIRP, and TCGA-KICH cohorts, differential expression analysis was conducted using the limma package, yielding 12,348, 14,212, 11,402, and 13,489 differentially expressed genes (DEGs), respectively (Fig. 1A–D).

Fig. 1.

Demonstrating the importance of mitosis in renal carcinogenesis. (A-D) Volcano plot of differentially expressed genes (DEGs) between tumor and normal kidney tissue in WCH-TFE3, TCGA-KIRC, TCGA-KIRP and TCGA-KICH; (E-H) The ridge plot illustrates the results of ssGSEA enrichment analysis for DEGs between tumor and normal renal tissues in WCH-TFE3, TCGA-KIRC, TCGA-KIRP and TCGA-KICH; (I) Spatial transcriptomic visualization of the hallmark mitotic spindle gene set (geometric mean expression) overlaid on an H&E-stained RCC tumor section from a representative case. Warmer colors indicate higher pathway activity. In this illustrative sample, the gene set showed broad activation across the tumor region; (J) The Venn diagram depicts the intersection of mitotic genes with prognosis-associated genes identified in the WCH-TFE3, IMmotion151, and TCGA cohorts; (K) The heatmap illustrates the differential expression of the intersecting genes between tumor and normal renal tissues in TCGA-KIRC

To further investigate the biological mechanisms underlying RCC development, single-sample gene set enrichment analysis (ssGSEA) was performed. This analysis revealed that mitosis-related pathways—including Mitotic DNA Replication, Cell Cycle, and Mitotic Spindle—were significantly upregulated in tumor samples (Fig. 1E–H), suggesting that aberrant mitotic activity is a central feature of tumorigenesis in both TFE3-rearranged renal cell carcinoma (TFE3-tRCC) and ccRCC. Spatial transcriptomic analysis of a representative RCC sample revealed a broad increase in activity of the hallmark mitotic spindle gene set throughout the tumor, with focal enrichment in specific intratumoral niches (Fig. 1I).

Subsequently, prognostic genes were identified through univariate Cox regression analysis in the WCH-TFE3, TCGA-KIRC, and IMmotion151 cohorts. These genes were intersected with mitosis-related candidate genes (MCRGs), resulting in the identification of 30 dysregulated MCRGs (Fig. 1J–K). These 30 genes were selected for further in-depth analysis due to their potential relevance to RCC prognosis.

Construction of machine learning prognostic models

By integrating 10 machine learning algorithms, we constructed 101 prognostic models (Fig. 2A). Based on their concordance index (C-index) values, the top 3 models were selected for further evaluation. Ultimately, the combination of CoxBoost and Random Survival Forest (RSF) was chosen as the optimal model (Fig. 2B). This approach achieved comparable predictive performance while utilizing the fewest genes, resulting in a concise nine-gene signature consisting of BIRC5, CCNF, CDC25C, DDX21, EXO1, FOXM1, PLK1, SPHK1, and TACC3.

Fig. 2.

A consensus MCRS was developed and validated using a combination of bulk-seq data and machine learning methods. (A) A total of 101 prediction models were calculated through a LOOCV framework, and the C-index was further calculated for each model on the training and validation datasets; (B) COXBOOST + RSF is the best model and shows the C-index on the training set and two validation sets; (C-E) Kaplan-Meier OS curves of MCRS models in the TCGA-KIRC training cohort, the EMTAB-1980 validation cohort and WCH-FHRCC validation cohort; (F-H) ROC curves showing AUC values of the model for different survival times in RCC patients in the TCGA-KIRC, EMTAB-1980 and WCH-FHRCC cohort based on MCRS models

Patients in the TCGA-KIRC, E-MTAB-1980, and WCH-FHRCC cohorts were stratified into high-risk and low-risk groups based on the median risk score. Kaplan-Meier survival analysis across all three datasets revealed significantly worse overall survival (OS) in the high-risk group compared to the low-risk group (p < 0.05; Fig. 2C–E), indicating robust risk stratification and broad applicability of the model.

Furthermore, the time-dependent area under the receiver operating characteristic (ROC) curve (AUC) for 1-, 3-, and 5-year OS in the TCGA-KIRC cohort all exceeded 0.9, suggesting near-perfect prognostic performance (Fig. 2F). Despite potential overfitting concerns, the model retained strong predictive performance across both external validation cohorts. In the E-MTAB-1980 cohort, the 1-, 3-, and 5-year AUCs were 0.824, 0.827, and 0.747, respectively, while in the WCH-FHRCC cohort, all AUCs remained above 0.6 (Fig. 2G–H).

Correlation between mitotic catastrophe–related score (MCRS) and clinical baseline

We used pie charts to illustrate the distribution of clinical characteristics in the TCGA-KIRC cohort. As anticipated, the high-risk subgroup exhibited a higher proportion of adverse clinicopathologic features (Fig. 3A). In the WCH-FHRCC cohort, a similar distributional trend was observed, although the associations were more limited and only the pM stage reached statistical significance, likely owing to the relatively small sample size (Fig. 3B; Table 1). Consistently, in the TCGA-KIRC cohort, risk scores were positively associated with higher clinical grade and were also higher in older patients (age ≥ 60) (Fig. 3C). In the WCH-FHRCC cohort, MCRS showed an increasing trend across higher T- and M-stage categories, but no significant differences were observed for several other clinicopathologic variables, including pT stage, pN stage, sex, and age (Fig. 3D; Table 1).

Fig. 3.

Associations of MCRS with clinical and imaging baseline characteristics and the underlying mechanisms. (A-B) Pie charts of the distribution of different clinical characteristics in the MCRS-high and MCRS-low risk groups in TCGA-KIRC and WCH-FHRCC; (C-D) The box plots illustrate the differences in MCRS across various clinicopathologic subgroups in TCGA-KIRC and WCH-FHRCC in MCRS-high and MCRS-low groups; (E) The heatmap illustrates the differences in TCIA-KIRC imaging features between the MCRS-high and MCRS-low groups; (F) The enrichment bar plot depicts the results of GO and KEGG pathway analyses of DEGs between the MCRS-high and MCRS-low groups

In addition, we analyzed 10 contrast-enhanced CT features in the TCGA-KIRC cohort to further characterize the radiologic differences associated with MCRS. Chi-square analysis showed that MCRS was significantly associated with eight imaging features, including tumor size, tumor margin, spiculated margin, enhancement pattern, intratumoral angiogenesis, necrosis, renal vein invasion, and collecting system invasion (all p < 0.05) (Fig. 3E). Descriptively, tumors in the high-MCRS group more frequently exhibited adverse imaging characteristics, including larger tumor size, ill-defined margins, nodular enhancement pattern, and spiculated margins. In addition, intratumoral angiogenesis, necrosis, renal vein invasion, and collecting system invasion were more commonly observed in the high-MCRS group than in the low-MCRS group (Table S1).

MCRS differential gene enrichment analysis

Differentially expressed genes (DEGs) were identified by comparing the high- and low-MCRS groups. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses revealed that these DEGs were enriched in anti-inflammatory processes, vascular-related pathways, and metabolic pathways (Fig.  3F). Collectively, these findings suggest that MCRS gene expression may be involved in regulating the inflammatory response as well as vascular- and metabolism-related biological processes.

Immune microenvironment profiling

The tumor immune microenvironment (TIME) plays a pivotal role in tumor initiation, progression, and response to therapy. To characterize the heterogeneity in immune infiltration patterns among patients, we conducted a comprehensive immune landscape analysis using gene expression profiles from the TCGA-KIRC and WCH-FHRCC cohorts.

We first evaluated the relative abundance of 22 immune cell types using bar plots, revealing distinct infiltration patterns across all RCC patients (Fig. 4A-D). Notably, B cells and regulatory T cells (Tregs) were more highly infiltrated in the high-risk subgroup, and their abundance was significantly and positively correlated with the MCRS risk score (Fig. 4E-F).

Fig. 4.

Exploration of the tumor immune microenvironment. (A-B) The box plots illustrate the differences in immune cell infiltration between the MCRS-high and MCRS-low groups in the TCGA-KIRC and WCH-FHRCC cohorts, as estimated by the CIBERSORT algorithm; (C-D) The lollipop plot depicts the correlations between MCRS and various immune cell subsets estimated by the CIBERSORT algorithm in the TCGA-KIRC and WCH-FHRCC cohorts; (E-F) The scatter plots illustrate the correlations between MCRS and several common immune checkpoint molecules in the TCGA-KIRC and WCH-FHRCC cohorts

Given these findings, we further explored the association between MCRS and immune checkpoint expression. In the TCGA-KIRC cohort, MCRS showed significant positive correlations with PDCD1, CTLA4, TIGIT, and LAG3 (all p < 0.05), whereas in the WCH-FHRCC cohort, a strong positive correlation was observed with TIGIT (R = 0.505, p < 0.05).

Malignancy and mutation analysis

We further investigated the genetic alterations associated with MCRS stratification. Comparison of somatic single-nucleotide variant (SNV) profiles between the high- and low-MCRS subgroups revealed higher mutation frequencies of BAP1 (14% vs. 7%), SETD2 (16% vs. 8%), and TP53 in the high-MCRS group (Fig. 5A–B). Among the frequently mutated genes, mTOR mutations were significantly enriched in the high-MCRS subgroup (p < 0.05), whereas the differences observed for BAP1, SETD2 and TP53 also showed a borderline trend toward enrichment (p < 0.05).

Fig. 5.

Mutational landscapes. (A-B) Waterfall graphs of common mutation in TCGA-KIRC in MCRS-high and MCRS-low risk groups; (C) The bar plot shows the frequency of CNVs in model genes, with green and blue dots indicating copy number gain and loss, respectively

We subsequently examined copy-number variation (CNV) in the nine MCRS-related genes. Except for TACC3, all genes exhibited varying degrees of copy number alterations. Notably, CDC25C showed the highest amplification frequency across the cohort, suggesting a potential association between elevated MCRS and genomic instability (Fig. 5C).

Characterization of the core model gene

Based on the results for the model genes, we constructed a survival (t) explanatory model to assess the extent to which each factor contributes to the outcome, ultimately highlighting CCNF as a strong influence (Fig. 6A-C). We then focused on CCNF, which was upregulated in TCGA-KIRC tumor samples at the paired-sample level (Fig. 7A). In addition, we performed direct validation on pathological sections from 15 ccRCC patients using immunofluorescence (IF). As expected, the protein assay results were consistent with the RNA-Seq results (Fig. 7B). Through baseline clinical analysis, we observed that elevated CCNF expression was significantly associated with advanced T, N, and M stages in patients with ccRCC (p < 0.05; Fig. 7C-F). Furthermore, survival analyses in both the TCGA and WCH-FHRCC cohorts demonstrated that high CCNF expression was consistently associated with poorer overall survival (OS) (p < 0.05), supporting its role as a pivotal gene within the mitotic-apoptosis model (Fig. 7G–H). Collectively, these findings suggest that CCNF serves as a potential prognostic biomarker for poor outcomes in both ccRCC and FHRCC populations.

Fig. 6.

Screening for model core gene. (A) SHAP summary plot showing the contribution of each gene to the prediction model; (B) Bar plot displaying the mean absolute SHAP values of representative genes, indicating their overall importance in the model; (C) SHAP force plot illustrating the contribution of individual features to a single prediction, with positive and negative values representing risk-promoting and risk-reducing effects, respectively

Fig. 7.

Expression and prognostic significance of CCNF in RCC. (A) CCNF expression was significantly upregulated in tumor compared with adjacent normal tissues in the TCGA-KIRC cohort; (B) Representative IF staining of CCNF in paired tumor and normal tissues from the WCH cohort (n = 15), with quantification shown on the right; (C) Heatmap illustrating the association between CCNF expression and clinical characteristics, including pT stage, pN stage, pM stage, gender, and age; (D–F) Differential expression of CCNF across tumor stages. Bar plots show CCNF expression according to T stage (D), N stage (E), and M stage (F); (G–H) Kaplan–Meier survival analysis of overall survival (OS) stratified by CCNF expression in the TCGA-KIRC and WCH-FHRCC cohorts. Statistical significance was assessed using the log-rank test

Single-cell analysis

Further analysis at single-cell resolution provided us with a deep view of the CCNF expression landscape. We analyzed GSE237425, which contains four ccRCC samples. We classified the single-cell data into nine distinct cell types based on canonical markers, including keratinocytes, epithelial cells, T cells, proliferating T cells, NK cells, neutrophils, myeloid cells, and fibroblasts (Fig. 8A). By mapping CCNF expression across all cell types in RCC, we found that CCNF was predominantly enriched in proliferating cells, epithelial cells and T cells (Fig. 8B). Based on this observation, we further investigated the association between CCNF expression and the tumor immune microenvironment. Using bulk RNA-Seq data from the WCH-FHRCC cohort, we compared CCNF-high and CCNF-low expression groups. Notably, Treg cell abundance was significantly higher in the CCNF-high group, and CCNF expression positively correlated with Treg infiltration (R = 0.18, p < 0.05) (Fig. 8C-D). To validate this finding, we analyzed two independent immunotherapy cohorts—IMmotion151 and CheckMate. As anticipated, Kaplan–Meier survival analysis revealed that higher CCNF expression was associated with worse progression-free survival (PFS) and overall survival (OS) (p < 0.05) (Fig. 8 E-F).

Fig. 8.

Single-cell landscape, immune associations, and immunotherapy relevance of CCNF in RCC. (A) UMAP plot showing the clustering of major cell types in RCC single-cell RNA-seq data; (B) UMAP and dot plots displaying the distribution and expression levels of CCNF across different cell populations; (C) Box plots comparing immune infiltration scores of various immune cell subsets between high and low CCNF expression groups; (D) Correlation analysis between CCNF expression and immune cell subsets estimated by CIBERSORT; (E–F) Kaplan–Meier survival curves of PFS/OS stratified by CCNF expression in the IMmotion151 and CheckMate cohorts, showing that higher CCNF expression was associated with inferior outcomes in these immunotherapy-treated cohorts

CCNF promotes RCC cell proliferation through mTOR

To elucidate the functional role of CCNF, several experiments were conducted. First, we constructed lentiviral vectors for CCNF overexpression (OE-CCNF-1 and OE-CCNF-2) and a negative control (NC) using shRNA, which were subsequently transduced into the Caki-1 and OS-RC-2 renal cancer cell lines. Western blot analysis confirmed successful transduction, showing that CCNF expression was significantly upregulated in both Caki-1 and OS-RC-2 cells following OE-CCNF-1 and OE-CCNF-2 transfection (Fig. 9A, Figure S2). EdU incorporation and colony formation assays demonstrated that CCNF overexpression markedly enhanced the proliferative capacity of Caki-1 and OS-RC-2 cells (Fig. 9B-C). To further assess the impact of CCNF on invasive and migratory behavior, we performed Transwell and wound-healing assays. As shown, CCNF-overexpressing cells exhibited significantly increased invasion and migration abilities compared to the control group (Fig. 9D-E).

Fig. 9.

Functional validation of CCNF in RCC progression. (A) WB confirmed the overexpression efficiency of CCNF in OS-RC-2 and Caki-1 cells; (B) Colony formation assays showing that CCNF overexpression significantly enhanced the clonogenic ability of RCC cells; (C) EdU assays indicating that CCNF overexpression promoted RCC cell proliferation; (D) Transwell assays demonstrating that CCNF overexpression increased the invasion capacity of OS-RC-2 and Caki-1 cells; (E) Wound-healing assays showing that CCNF overexpression accelerated RCC cell migration; (F) In vivo xenograft models revealed that CCNF overexpression significantly promoted tumor growth in nude mice (n = 5 per group); (G) GSEA analysis showing that CCNF overexpression was positively associated with activation of the mTOR signaling pathway; (H) WB validated that CCNF overexpression upregulated phosphorylation of mTOR in RCC cells. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ns, non-significant

Collectively, these in vitro findings support a pro-tumor role for CCNF overexpression in RCC cells. To further assess its effect in vivo, we established a subcutaneous xenograft model using Caki-1 cells stably overexpressing CCNF in BALB/c nude mice. Tumors in the OE-CCNF group exhibited significantly increased growth, volume, and weight compared with those in the NC group (p < 0.05), supporting the tumor-promoting role of CCNF in vivo (Fig. 9F; Table S4). To investigate potential downstream pathways, we analyzed RNA-seq data from the WCH-FHRCC cohort and observed enrichment patterns suggestive of mTOR pathway involvement (Fig. 9G). In line with this, Western blot analysis showed elevated p-mTOR expression in the CCNF overexpression group (Fig. 9H). These data suggest a potential association between CCNF overexpression and mTOR pathway activation.

Discussion

In this study, we conducted a comprehensive transcriptomic and multi-cohort analysis of MC–associated molecular patterns in RCC, aiming to delineate the biological and clinical significance of mitotic dysregulation in tumor progression and therapeutic response. Our findings provide novel insights into the genomic and immunologic features that underpin MC activation in RCC, confirming that aberrant cell-cycle control and mitotic stress are not merely proliferative hallmarks but rather active drivers of tumor aggressiveness. We demonstrated that MC-related transcriptional alterations are closely linked to copy-number instability and activation of oncogenic pathways, highlighting their contribution to tumor evolution. Moreover, we established a robust MC-associated risk model that stratifies patients by prognosis and is associated with outcomes in immunotherapy-treated cohorts, and we identified distinct molecular subgroups characterized by divergent proliferative and immune microenvironmental states. Collectively, these findings suggest that mitotic catastrophe–associated transcriptional programs are linked to RCC progression and immunotherapy-related outcomes, potentially within Treg-enriched immune states; however, a direct effect on immunotherapy responsiveness remains to be established.

We first observed that mitosis-related signaling pathways are broadly activated in RCC; however, the precise biological mechanisms through which mitotic dysregulation contributes to tumor progression remain insufficiently defined. To address this knowledge gap, we systematically analyzed four independent RCC transcriptomic cohorts comparing tumor and adjacent normal tissues and complemented these findings with an illustrative spatial transcriptomic analysis of a representative RCC sample. Collectively, these multi-cohort analyses consistently demonstrated that mitotic activity is markedly upregulated in RCC, highlighting its central role in tumor progression and biological aggressiveness. Evasion of mitosis represents a key pathway in oncogenesis. MC is a tumor-suppressive mechanism that precedes and is distinct from apoptosis, necrosis, or senescence, ultimately leading to the elimination of cells with mitotic defects and genomic instability [20]. Notably, MC also has the potential to overcome tumor drug resistance [14]. The prognostic significance of MCRGs has been validated in breast, hepatocellular, and bladder cancers, but remains poorly defined in RCC [15–17].

We then identified an MCRS by machine learning. High-MCRS tumors were characterized by enrichment of acute inflammatory and angiogenesis-related pathways, consistent with prior evidence linking heightened angiogenic and inflammatory signaling to aggressive tumor progression across multiple malignancies, including RCC. Clinically, the MCRS-high group exhibited more advanced clinicopathologic features and adverse imaging characteristics, consistent with its association with poorer prognosis. Tumor microenvironment profiling further showed increased Treg infiltration in high-MCRS tumors, and MCRS was positively correlated with the immune checkpoints LAG3 and TIGIT, suggesting a microenvironment characterized by greater Treg abundance and higher checkpoint expression. Collectively, these findings support an association between mitotic catastrophe–related transcriptional programs and RCC progression, and indicate a potential link to immunotherapy-related outcomes. LAG3 and TIGIT may therefore warrant further investigation in patients with high MCRS.

In recent years, the treatment paradigm for renal cell carcinoma (RCC) has shifted from tyrosine kinase inhibitor (TKI) monotherapy to combination regimens incorporating a TKI and an immune checkpoint inhibitor (ICI), markedly prolonging survival in patients with metastatic RCC (mRCC) [21]. However, not all patients derive clinical benefit from systemic therapy. Our results suggest that the MCRS model is a reliable prognostic indicator of overall survival in RCC and may also be associated with clinical outcomes among patients receiving immune checkpoint inhibitor (ICI)-based therapy. However, these findings should not be interpreted as evidence of true predictive value for ICI benefit.

Mutation profiling further supported the biological basis of the high-risk phenotype. Enrichment of TP53 mutations suggests impaired post-mitotic checkpoint control, permitting the survival of genomically unstable clones following mitotic errors and weakening the tumor-suppressive barrier of MC [22, 23]. Increased mutation rates in BAP1 and SETD2, both critical regulators of chromatin integrity and genome maintenance, further indicate increased replication stress and chromosomal instability—conditions closely linked to mitotic dysfunction [24, 25]. The higher prevalence of mTOR pathway alterations may further promote unchecked cell-cycle progression and tolerance to mitotic defects [26, 27]. Together with elevated TMB, these findings depict a genomically unstable state characterized by defective mitotic surveillance and insufficient elimination of aberrant cells, aligning mutational landscapes with MC-related transcriptional activation and adverse clinical outcomes.

To minimize redundancy and identify core prognostic drivers, we applied SurvSHAP(t) to quantify time-dependent gene importance, advancing the integration of machine learning–based interpretability into transcriptomic modeling. CCNF emerged as the dominant contributor. To validate the robustness and clinical applicability of the model, we conducted both in vitro and in vivo functional experiments and validated it in clinical specimens, with a focus on CCNF. Immunofluorescence confirmed increased CCNF expression in tumor cells, and functional assays showed that CCNF overexpression promoted RCC cell proliferation, migration, and invasion in vitro and in vivo, and influenced apoptosis. Transcriptomic profiling and Western blot analyses suggested a potential association between CCNF overexpression and mTOR pathway activation. These findings are generally consistent with previous reports [28].

Single-cell RNA sequencing further showed that CCNF was enriched in proliferative tumor subsets, co-expressing mitosis-related markers (SPP1, MKI67, TOP2A, CENPF, MCM6), and was also detectable in tumor epithelial cells and T cells, linking it to both tumor-intrinsic biology and the immune microenvironment. As a core component of the SCF ubiquitin ligase complex, CCNF regulates proteostasis during DNA replication and mitotic progression, and its dysregulation may promote replication stress and genomic instability [29]. Consistently, CCNF-high tumors displayed an immune-dysfunctional microenvironment. Importantly, high CCNF expression was associated with inferior outcomes in both the PD-L1 inhibitor–treated IMmotion151 cohort and the PD-1 inhibitor–treated CheckMate cohort. These findings support the potential relevance of CCNF in the immunotherapy setting; further research is required to establish CCNF as a predictive biomarker for reduced response to immune checkpoint blockade therapy [30, 31].

While these findings are encouraging, several limitations merit consideration. First, our institutional dataset comprised only 54 patients, which may limit the statistical robustness. Second, the immunotherapy-related implications of CCNF were primarily derived from bioinformatic analyses without functional validation in experimental immuno-oncology models. Future work will involve larger, prospective cohorts to validate model performance and single-cell sequencing of additional RCC samples to strengthen the reliability of our findings.

Conclusion

In summary, we constructed a novel risk model based on nine MC-related genes (CCNF, PLK1, FOXM1, BIRC5, EXO1, SPHK1, DDX21, TACC3 and CDC25C) that demonstrates robust performance in prognostic stratification and is associated with immune infiltration patterns and outcomes in immunotherapy-treated ccRCC patients. Additionally, the model trained on TCGA-KIRC achieved excellent predictive performance on our own FHRCC validation set. Overall, these findings indicate that this MC-based model achieved excellent prognostic performance in clear cell renal cell carcinoma and showed preliminary external applicability in the WCH-FHRCC cohort. Using SHAP, we identified the core gene CCNF in the model and verified its role in RCC through in vitro and in vivo experiments, further supporting the biological relevance of the MC-related model.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

RCC

Renal cell carcinoma

ccRCC

Clear cell renal cell carcinoma

FHRCC

Fumarate hydratase–deficient renal cell carcinoma

TFE3–tRCC–TFE3

Rearranged renal cell carcinoma

MC

Mitotic catastrophe

MCRS

Mitotic catastrophe–related score

MCRGs

Mitotic catastrophe–related genes

OS

Overall survival

PFS

Progression–free survival

DFS

Disease–free survival

SNV

Single–nucleotide variant

CNV

Copy number variation

DEGs

Differentially expressed genes

GO

Gene Ontology

KEGG

Kyoto Encyclopedia of Genes and Genomes

ssGSEA

Single–sample gene set enrichment analysis

scRNA

seq–Single–cell RNA sequence

ROC

Receiver operating characteristic

AUC

Area under the curve

RSF

Random survival forest

LASSO

Least absolute shrinkage and selection operator

C–index

Concordance index

SHAP

SHapley Additive exPlanations

TME

Tumor microenvironment

ICIs

Immune checkpoint inhibitors

PD-1

Programmed cell death protein 1

PD-L1

Programmed death–ligand 1

CTLA–4

Cytotoxic T–lymphocyte–associated protein 4

Treg

Regulatory T cell

Experimental techniques

IHC

Immunohistochemistry

IF

Immunofluorescence

WB

Western blot

EdU

5-ethynyl-2′-deoxyuridine

Author contributions

ZLW conceived and designed the study, performed data acquisition and analysis, conducted experimental work, and drafted the manuscript. XH participated in data interpretation and manuscript drafting. ZYL contributed to experimental procedures and data analysis. FC was responsible for imaging data acquisition and analysis and contributed to manuscript preparation. XYW participated in data collection and manuscript drafting. HZ supervised the study and provided critical intellectual input.All authors read and approved the final manuscript.

Funding

This study was not supported by any external funding sources.

Data availability

All data generated or analysed during this study are included in this published article method section. The datasheets generated during the analyzation in the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The studies were conducted in accordance with the Declaration of Helsinki and the International Conference on Harmonization (ICH) Guidelines for Good Clinical Practice. Approval for the use of patient samples (No. 2021 − 161) was granted by the Ethics Committee of West China Hospital. Written informed consent was obtained from all participating patients. All animal experiments were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) of West China Hospital (No. 20240510004) and were conducted in accordance with the approved protocol and relevant national and institutional guidelines for the care and use of laboratory animals. According to the approved protocol, the maximal permitted tumor burden was 1,500 mm³ (or a maximal diameter of 15 mm).

Consent for publication

All participants provided written informed consent for the publication of anonymized data and images. The consent forms are held by the authors’ institution and are available for review by the journal upon reasonable request.

Tumor size

At no time did tumor size exceed the approved maximum.

Competing interests

The authors declare no competing interests.