Work

A Programmed Cell Death Protein 1 Pathway-Derived Prognostic Transcriptomic Signature in Bladder Cancer: Development in TCGA-BLCA and Evaluation in Independent Cohorts

Year: 2026

Type: article

Abstract: Immune checkpoint inhibitors (ICIs) targeting programmed cell death protein 1 (PD-1) and programmed death-ligand 1 (PD-L1) have changed the treatment of advanced bladder cancer, but only a minority of patients benefit. We asked whether expression of PD-1 pathway genes carries prognostic information in bladder cancer. Sixty-eight genes annotated to the PD-1 signaling, CD28 co-stimulation, and cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) inhibitory signaling pathways in Reactome were examined in The Cancer Genome Atlas bladder urothelial carcinoma cohort (TCGA-BLCA). Genes differentially expressed between tumor and normal tissue were entered into a least absolute shrinkage and selection operator (LASSO) Cox model, yielding a seven-gene risk score (FYN, GRAP2, TRIB3, MAP3K8, AKT3, CD274, and CD80). In TCGA-BLCA (n = 404), the risk score was associated with overall survival independently of age, sex, and stage (hazard ratio [HR] per standard deviation 1.24, 95% confidence interval [CI] 1.11–1.39) and modestly improved discrimination over clinical variables (concordance index [C-index] 0.630 to 0.664). With coefficients and standardization fixed in advance, the score was not associated with cancer-specific survival in GSE13507 (n = 165; HR per standard deviation 1.21, 95% CI 0.83–1.77) or with outcome in four further independent cohorts, including an atezolizumab-treated cohort; in a second, smaller ICI-treated cohort it was associated with shorter survival. Gene selection was unstable under bootstrap resampling, and the signature’s associations with transcriptome-inferred immune indices were limited. This PD-1 pathway-derived signature should therefore be regarded as exploratory; its prognostic value outside TCGA-BLCA is not supported by the available evidence.

Source: Current Issues in Molecular Biology

DOI: 10.3390/cimb48101056