Credit: Stockbyte Royalty Free Photos

A continuously expanding list of genes are proposed as putative driver genes in various cancers, but confirmation and investigation of their roles require functional analyses. Two new studies take advantage of RNA interference (RNAi) screens to identify and characterize driver genes in breast cancer.

In breast cancer, the most frequent genetic lesions are somatic copy number alterations (CNAs). The affected regions often harbour driver genes with causal roles in cancer; however, identifying these driver genes among the excess of co-altered passenger genes in each chromosomal region is a challenge.

Sanchez-Garcia and Villagrasa et al. first identified amplified chromosomal regions in 785 breast cancer genomes from The Cancer Genome Atlas (TCGA), using an algorithm that accounts for the local somatic CNA rate to increase detection sensitivity. To identify constituent driver genes in these amplified regions, they devised a machine-learning tool called Helios that integrates additional available data on the constituent genes, such as point mutation frequency, expression levels and results of in vitro genome-wide RNAi screens. The RNAi screen information is based on the concept of oncogene addiction, in which a cell can become 'addicted' to the overexpression of a functionally relevant driver oncogene. Such oncogene addiction was inferred when RNAi-mediated knockdown of a gene across a panel of cell lines caused reduced viability specifically in the cell lines that overexpressed that gene.

Highlighting the sensitivity of this integrative approach, 9 of the top 10 Helios-identified genes were known breast cancer driver genes. Focusing on amplified regions without known driver oncogenes, the researchers then tested 12 Helios-identified candidates. Ten of these scored for anchorage-independent growth (an assay for driver oncogene properties) when overexpressed in the MCF10A breast cell line. Furthermore, follow-up of one of the candidates, remodelling and spacing factor 1 (RSF1), confirmed a role in tumour growth and invasion in vivo in mice.

In a separate study, Wang and Fu et al. also investigated the co-occurring alteration of genes in cancer genomes but focused on combinations of different driver genes. By mining TCGA data from breast cancer samples, the team identified 67 known or putative driver genes based on cancer-associated somatic point mutations, CNAs or gene expression changes.

From this set of genes the investigators carried out RNAi screens in MCF10A cells, which involved single-gene knockdown and combinatorial knockdown of 1,508 pairs of these genes. Cellular phenotypic consequences were assessed by automated microscopy, such as cell number as a measure of cell viability. The researchers scored genetic interactions (that is, epistasis) when the phenotypes of double knockdown differed from expectations based on each single knockdown. They found that genetic interactions were extensive: 66 of the 67 genes had at least one significant interaction with one of the other genes, and most of these interactions had not been described previously.

Subsequently, the team organized these interacting genes into functional clusters and networks. Moreover, from clinical data they showed that for various genetic interactions, the co-occurrence of lesions in both genes was associated with reduced patient survival. This finding emphasizes the value of studying combinations of cancer gene alterations that occur in patients, rather than just single gene alterations.

The approaches described in both projects have the potential to expand the range of drug targets for cancer therapy. Novel cancer driver genes — as well as the genes that they functionally interact with or rely on — might reveal unappreciated vulnerabilities in cancer cells that can be therapeutically targeted.