"Off-context" gene expression in lung cancer identifies a group of metastatic-prone tumors

An unexplored consequence of epigenetic alterations associated with cancer is the ectopic expression of tissue-restricted genes. Here, a new strategy was developed to decipher genome-wide expression data in search for these “off-context” gene activations, which consisted first, in identifying a large number of tissue-specific genes normally epigenetically silenced in most somatic cells and second, in using them as cancer biomarkers on an “on/off” basis. Applying this concept to analyze whole-genome transcriptome data in lung cancer, we discovered a specific group of 26 genes whose expression was a strong and independent predictor of poor prognosis in our cohort of 293 lung tumours, as well as in two independent external populations. In addition, these 26 classifying genes enabled us to isolate a homogenous group of metastatic-prone highly aggressive tumours, whose characteristic gene expression profile revealed a high proliferative potential combined to a significant decrease in immune and signaling functions. This work illustrates a new approach for a personalized management of cancer, with applications to any cancer type.