NEW YORK – Researchers have uncovered thousands of gene pairs whose expression levels could account for the tissue specificity of cancer driver genes and influence patients' survival as well as drug response.
While scientists have previously investigated interactions between pairs of genes in cancer, particularly ones that lead to synthetic lethality, a University of Maryland team has expanded on those studies to include additional types of gene pairs.
Using a computational strategy they devised, the researchers examined gene pairs whose joint expression is linked to patient survival, which they dubbed survival-associated pairwise gene expression states, or SPAGEs. As they reported today in Cell Reports, they uncovered nearly 72,000 SPAGEs that represent a dozen types of interactions within The Cancer Genome Atlas dataset. These interactions, they added, could explain why some cancer driver genes are tissue specific and could stratify breast cancer tumors into a number of survival-related subtypes.
"Relying on specific cancer vulnerabilities, such as a particular mutated gene's functional relationship with other genes, is potentially an effective approach to treating cancer," senior author Sridhar Hannenhalli, a professor of cell biology and molecular genetics at the University of Maryland, said in a statement
He and his colleagues developed a computation pipeline called SPAGE-finder to tease out these interactions from tumor transcriptomes. The pipeline first bins genes based on their activity level into low, medium, or high expression states, so that for each pair of genes there are nine possible co-activity states. It then screens the gene pairs for those that affect patient survival, either positively or negatively.
Using this approach, the researchers analyzed 5,288 tumors representing 18 different cancer types. In all, they identified 71,946 SPAGEs associated with patient survival. Many of these genes are involved in cell division and proliferation, processes that are themselves linked to cancer.
After further filtering, the researchers homed in on 1,704 SPAGEs — including 133 known cancer genes and 50 breast-cancer-specific driver genes — that involve neighboring genes in a protein interaction network.
These gene interactions, Hannenhalli noted, could influence why some driver genes lead to tumor development in some tissue types but not others because their interacting partners could be expressed at varying levels in different tissues.
The researchers also found some SPAGEs that were associated with response to drugs like gemcitabine, lomustine, and paclitaxel.
In particular, they uncovered SPAGEs in the TCGA cohort that are associated with paclitaxel response in head and neck cancers. One of these SPAGEs was involved in the inactivation of BCL and the overactivation of ITPR1, and negatively affected tumor fitness. Paclitaxel acts by inhibiting the proteins encoded by BCL2, TUBB1, and MAP, and the researchers noted that their analysis suggests BCL2 inhibition by paclitaxel has increased effectiveness when ITPR1 is highly expressed.
SPAGEs could also stratify tumors into subtypes, the researchers reported. For breast cancer, they used SPAGEs to group the 1,981 samples in the METABRIC dataset into nine clusters with distinct survival characteristics. While some of these clusters were in line with the five well-established clinically distinct breast cancer subtypes, others differed. These SPAGE-based groups, the researchers said, could represent an improved way to distinguish breast cancers with different survival prognosis and mutational profiles.
"Our work expands the potential scope of strategies, thus far restricted to synthetic lethality, by generalizing the concept of exploiting genetic interactions to include many other yet-unexplored types of gene-pair relationships," Hannenhalli said. "We believe this lays the foundation for using a computational method for identifying and studying additional types of genetic interactions in the future."
The next step, Hannenhalli added, is to work with other cancer researchers to examine therapies that target some of the gene pairs they identified in their study.