The combined use of EFS, GPX2, and SPRR1A expression could distinguish favorable from poor clinical outcome among epithelial-like head and neck carcinoma subtypes.

Abstract

We aimed at identifying molecular markers predictive of clinical outcome in patients with head and neck cancer based on the expression profile of cells showing epithelial-like (EL) or mesenchymal-like (ML) phenotypes. We analyzed the association between EL and ML cells and migration, drug resistance, or tumor growth. The differential gene expression profile between cell types was used to build a model to stratify patients according to survival. EL cells were sensitive to cisplatin and cetuximab, showed low migration, and generated squamous differentiated tumors in mouse. A differential 93-gene expression signature between ML and EL cells was used to build a three-gene (EFS, GPX2, and SPRR1A) survival model by analyzing the RNA-seq data of the TCGA-HNSC project. Its prognostic value was confirmed in two independent cohorts. EFS, GPX2, and SPRR1A are prognostic markers able to distinguish clinical outcome among subtypes sharing an EL phenotype.