An EV-Associated Gene Signature Correlates with Hypoxic Microenvironment and Predicts Recurrence in Lung Adenocarcinoma

Abstract

Extracellular vesicles (EVs) mediate intercellular communications in the tumor microenvironment and contribute to the aggressive phenomenon of cancers. Although EVs in body fluids are supposed to be ideal biomarkers for cancer diagnosis and prognosis, it remains difficult to distinguish the tumor-derived EVs from those released by other tissues. We hypothesized that analyzing the EV-related molecules in tumor tissues would help to estimate the prognostic value of tumor-specific EVs. Here, we investigate the expression of coding genes of proteins carried by small EVs (sEVs) in primary lung adenocarcinoma. Based on the protein-protein interaction network, we identified three network modules (3-PPI-Mod) as a signature that could predict recurrence. This signature was validated in three independent datasets and demonstrated better prognostic value than signature generated from gene expression alone. Meanwhile, the high-risk subgroup assigned by the signature could benefit from adjuvant chemotherapy, although it was not beneficial in unselected patients. Two out of three modules were enriched by proteins identified in sEVs from non-small-cell lung cancer cells. Furthermore, the two modules were remarkably correlated with intratumoral hypoxia score. These results suggest that the 3-PPI-Mod signature was enriched in tumor-derived sEVs and could serve as a prognostic and predictive biomarker for lung adenocarcinoma.