A gene-expression signature defines a subtype of Stomach Adenocarcinomas with low levels of Claudins and a high ratio of NF-YA long/NF-YA short splicing variants

Background: Claudin-3, Claudin-4, and Claudin-7 are expressed on the surface of epithelial cells. Their absence in neoplastic cells of epithelial origin is an aggressiveness marker in different cancers. The NF-YA gene codes for the Nuclear-Transcription-Factor-Y-Subunit-A, which is overexpressed in various tumors. In tumors, the relative ratio of the two major NF-YA alternative splicing isoforms, NF-YA long and NF-YA short, is associated with a mesenchymal phenotype and a poor prognosis. Based on a high NF-YA long/NF-YA short ratio, we generated a 158-gene signature that is common to Claudinlow Breast Carcinomas (BRCA) and Stomach Adenocarcinomas (STAD). Methods: To better classify STAD Claudinlow tumors, we employed a hierarchical clustering approach based on our 158-gene signature to classify STAD into a Claudinlow subgroup. We tested the classification potential of our signature in TCGA as well as in two additional datasets of tumors. We used the deep-learning DeepCC tool and the 158-gene signature to classify the STAD cell lines available in the CCLE platform. Obtained data were validated with qRT-PCR and Western blots. Results: The 158-gene signature resulted in the selection of a STAD subgroup with effective Claudinlow expression and with a high NF-YA long/NF-YA short ratio. This Claudinlow subgroup was separated from the EMT subgroup of STAD and it is characterized by poor clinical outcome. We identified nine Claudinlow STAD cell lines with a high NF-YA long/NF-Y short ratio and validated the expression of selected markers. Conclusions: Our work supports the notion that three overlapping features—low expression of Claudin-3/4/7, high NF-YA long/NF-YA short ratio and a 158-gene signature—mark a specific subset of STAD characterized by mesenchymal features and poor prognosis.