Machine learning-based integration develops a stress response stated T cell (Tstr)-related score for predicting outcomes in clear cell renal cell carcinoma

Abstract

Background: Establishment of a reliable prognostic model and identification of novel biomarkers are urgently needed to develop precise therapy strategies for clear cell renal cell carcinoma (ccRCC). Stress response stated T cells (Tstr) are a new T-cell subtype, which are related to poor disease stage and immunotherapy response in various cancers.

Methods: 10 machine-learning algorithms and their combinations were applied in this work. A stable Tstr-related score (TCs) was constructed to predict the outcomes and PD-1 blockade treatment response in ccRCC patients. A nomogram based on TCs for personalized prediction of patient prognosis was constructed. Functional enrichment analysis and TimiGP algorithm were used to explore the underlying role of Tstr in ccRCC. The key TCs-related gene was identified by comprehensive analysis, and the bioinformatics results were verified by immunohistochemistry using a tissue microarray.

Results: A robust TCs was constructed and validated in four independent cohorts. TCs accurately predicted the prognosis and PD-1 blockade treatment response in ccRCC patients. The novel nomogram was able to precisely predict the outcomes of ccRCC patients. The underlying biological process of Tstr was related to acute inflammatory response and acute-phase response. Mast cells were identified to be involved in the role of Tstr as a protective factor in ccRCC. TNFS13B was shown to be the key TCs-related gene, which was an independent predictor of unfavorable prognosis. The protein expression analysis of TNFSF13B was consistent with the mRNA analysis results. High expression of TNFSF13B was associated with poor response to PD-1 blockade treatment.

Conclusions: This study provides a Tstr cell-related score for predicting outcomes and PD-1 blockade therapy response in ccRCC. Tstr cells may exert their pro-tumoral role in ccRCC, acting against mast cells, in the acute inflammatory tumor microenvironment. TNFSF13B could serve as a key biomarker related to TCs.

Keywords: Bioinformatics; Clear cell renal cell carcinoma; Machine learning; Prognosis; TNFSF13B; stress response stated T cells.