Tumor-intrinsic signatures of esophageal adenocarcinoma associate with cellular phenotypes and responses to therapy

Identifying a mesenchymal molecular subtype at the tissue-level for patient stratification has posed challenges due to non-tumor cells interfering with gene expression data. Here, we conducted gene expression profiling from 186 esophageal cancer samples and applied non-negative matrix factorization (NMF) to segregate tumor cellular compartments and identify tumor-intrinsic features.

Tumor-intrinsic gene expression, together with single-nucleus RNA sequencing analysis, revealed the existence of two distinct signatures defining subtypes; intestinal-like and mesenchymal-like. The latter was enriched after neoadjuvant chemoradiation treatment and during disease progression. In a separate cohort of RNA-sequenced metastatic EAC, the mesenchymal-like subtype was found to predict poor response to treatment.