Integrated analysis of single-cell and bulk RNA sequencing data reveals the association between hypoxic tumor cells and exhausted T cells in predicting immune therapy response

Abstract

Continuous stimulation of tumor neoantigens and various cytokines in the tumor microenvironment leads to T cell dysfunction, but the specific mechanisms by which these key factors are distributed among different cell subpopulations and how they affect patient outcomes and treatment response are incompletely characterized.By integrating single-cell and bulk sequencing data of non-small cell lung cancer patients, we constructed a clinical outcome-associated T cell exhaustion signature. We discovered a significant association between the T cell exhaustion state and tumor cell hypoxia. Hypoxic malignant cells were significantly correlated with the proportion of exhausted T cells, and they co-occurred in patients at advanced stage. By analyzing the ligand-receptor interactions between these two cell states, we observed that T cells were recruited towards tumor cells through production of chemokines such as CXCL16-CXCR6 axis and CCL3/CCL4/CCL5-CCR5 axis. Based on 15 immune checkpoint blockade (ICB)-treatment cohorts, we constructed an interaction signature that can be used to predict the response to immune checkpoint blockade therapy. Among genes composed of the signature, CXCR6 alone has similarly high prediction efficacy (Area Under Curve (AUC) = 1, 0.89 and 0.73 for GSE126044, GSE135222 and GSE93157, respectively) with the signature and thus could serve as a potential biomarker for predicting immunotherapy response.Together, we have discovered and validated a significant association between exhausted T cells and hypoxic malignant cells, elucidating key interaction factors that significantly associated with response to immunotherapy.