Single‑cell and bulk RNA sequencing identifies T cell marker genes to predict the prognosis of ovarian cancer

Abstract

Ovarian cancer, with high mortality and often late diagnosis, shows high recurrence despite treatment. The variable effectiveness of immunotherapy highlights the urgent need for personalized, advanced therapeutic strategies. To investigate T-cell marker genes, single-cell RNA-sequencing (scRNA-seq) data were sourced from the Gene Expression Omnibus (GEO) database. Additionally, bulk RNA-sequencing data along with clinical information from ovarian cancer patients were retrieved from the Cancer Genome Atlas (TCGA) database to establish a prognostic signature. This study involved survival analysis to evaluate associations between different risk groups, and explored cellular communication and relevant pathway analyses, including metabolic pathways. We identified 41 genes showing varied expression between two T-cell subclusters, marking subcluster 0 with CCL5 and GZMA, and attributing the rest to subcluster 1. Re-annotation using canonical T-cell lineage and exhaustion markers confirmed subcluster 0 as activated cytotoxic CD8+T cells (CCL5+GZMA+), while subcluster 1 showed an immediate-early/stress-response gene signature rather than a canonical exhausted phenotype. These markers delineate four prognostic groups within the TCGA OV dataset, with T-cluster 2 exhibiting the poorest survival, in contrast to T-cluster 3, which shows the best. We developed a six-gene T-cell scoring model (DUSP2, ATF3, NR4A1, RGS2, RGS1, TNFAIP3), with the low-score group showing significantly better overall survival (HR 0.74, 95% CI 0.55–0.99, P = 0.048) and progression-free survival (HR 0.69, 95% CI 0.51–0.93, P = 0.012). Moreover, cellular and metabolic pathway analyses underscored the importance of CCL, IL2 and MGMT pathways in these subclusters. The study identifies CCL5 as a biomarker for T-cell subtypes in ovarian cancer using scRNA-seq and bulk RNA-seq data. A T-cell scoring model based on six genes predicts survival and progression rates, and provides incremental prognostic information beyond conventional clinicopathological factors (multivariable Cox: OS HR = 1.32, P = 0.028; PFS HR = 1.38, P = 0.009). This model could be a candidate prognostic indicator and CCL5 is a candidate biomarker and potential therapeutic target that warrants further functional validation in ovarian cancer.