UCLA

Cancer is characterized by wide-spread dysregulation of gene-expression. Much of this dysregulation is well-understood, reflecting disease subtypes, signatures of the tumour microenvironment and changes in cancer driver genes. But cancers also exhibit rare, extreme gene-expression states (XEGs) that appear as outliers relative to a population of cells or individuals. Using a novel outlier detection algorithm, we performed the first comprehensive of extreme gene expression in cancer in a cohort of 4,934 breast tumours. XEGs were unexpectedly common: about two thirds of tumours harboured at least one. Their frequency varied by cancer subtype, and they were associated with patient survival within subtypes. XEGs were observable at all levels of the central dogma: RNA XEGs were observed in matched protein data, and fully a third of XEGs were associated with specific somatic mutations or aberrant DNA methylation. XEGs observed in patients were recapitulated in cell-lines. Both genetic and pharmacologic inhibition of outlier genes reduced cancer cell fitness in cell-lines with the XEG event, but not in those without. Extreme gene-expression is common in breast cancer, and can provide a targetable selective advantage to tumour cells.