Comprehensive analysis of single-cell and bulk RNA sequencing reveals postoperative progression markers for non-muscle invasive bladder cancer and predicts responses to immunotherapy

Abstract

Abstract Background Non-muscle-invasive bladder cancer (NMIBC) is renowned for its high recurrence, invasiveness, and poor prognosis. Consequently, developing new biomarkers for risk assessment and investigating innovative therapeutic targets postoperative in NMIBC patients are crucial to aid in treatment planning. Approaches Differential gene expression analysis was performed using multiple Gene Expression Omnibus (GEO) datasets to identify differentially expressed genes (DEGs) between NMIBC and normal tissue, as well as between NMIBC and muscle-invasive bladder cancer (MIBC). Functional enrichment analysis was conducted based on the DEGs identified. Subsequently, prognosis-related genes were selected using Kaplan–Meier (KM) analysis and Cox regression analysis. The Boruta algorithm was utilized to further screen for core DEGs related to postoperative progression in NMIBC based on the aforementioned prognosis-related genes. Single-cell and clinical correlation studies were performed to verify their expression across various stages of bladder cancer. To investigate the link between core genes and the immune microenvironment, single-sample gene set enrichment analysis (ssGSEA) was utilized, and Receiver Operating Characteristic (ROC) and KM analyses were performed to confirm predictive power for immune therapy outcomes. Machine learning (ML) models were constructed using the DepMap dataset to predict the efficacy of core gene inhibitors in treating bladder cancers. The prognostic performance of the core genes was evaluated using ROC curve analysis. An online prediction tool was developed based on the core genes to provide prognostic predictions. Finally, RT-qPCR, CCK-8, and Transwell assays were used to verify the pro-tumor effects of the GINS2 in bladder cancer. Results A total of 70 DEGs were identified, among which 11 prognostic genes were obtained through KM analysis, and an additional 8 prognostic genes were obtained through COX analysis. The Boruta algorithm selected AURKB, GINS2, and UHRF1 as the three core DEGs. Single-cell and clinical variable correlation analyses indicated that the core genes promoted the progression of bladder cancer. The analysis of immune infiltration revealed a strong positive association between the core genes and both activated CD4 T cells and Type 2 helper T cells. Two random forest (RF) models were constructed to effectively predict the treatment effect of bladder cancer after targeted inhibition of AURKB and GINS2. In addition, an online nomogram tool was developed to effectively predict the risk of postoperative progression in NMIBC patients undergoing TURBT. Finally, RT-qPCR, CCK8, and Transwell assays showed that GINS2 promoted the growth and progression of bladder cancer. Conclusion AURKB, GINS2, and UHRF1 have the potential to enhance postoperative management of NMIBC patients undergoing transurethral resection of bladder tumor (TURBT) and can predict immunotherapy response, establishing them as promising therapeutic targets.

Keywords

• Bladder cancer
• Biomarkers
• Machine learning
• Single-cell RNA sequencing