NEW YORK (GenomeWeb News) – A team of American and German researchers has used a systems biology approach to integrate various genomic and clinical features of brain cancer. In the process, the team found a set of non-random mutations that often co-exist in cancer — and pulled out a handful of genes in the brain cancer network that were linked to patient survival.
The group's research appears in a pair of papers in today's issue of the Journal of the American Medical Association.
In the first of these studies, researchers used genomic and clinical data on more than 500 brain cancer tumors that had been collected at academic institutions and/or through The Cancer Genome Atlas, or TCGA. By generating networks that brought together gene alterations, copy number, and gene expression information, the team found a handful of highly connected "hub" genes as well as dozens of "hub-interacting" genes that tend to change in brain tumors.
When they looked more carefully at some of the genes in the network, the researchers detected seven genes that were independently associated with patient survival times. And together, they found, this seven-gene set could retrospectively distinguish between high- and low-risk patients — a finding that the team hopes will improve both brain cancer prognoses and treatment.
In a second JAMA paper, the team examined a pair of hub genes — ANXA7 and EGFR — in more detail, demonstrating that chromosome 10 losses that decrease ANXA7 levels correspond to a jump in EGFR levels that increase tumor aggressiveness and decrease survival times.
"We did this because we know that glioblastoma — or brain tumor disease — is very complex," Markus Bredel, director of Northwestern University's Brain Tumor Institute Research Program, who was lead author on the first paper and senior author on the second, told GenomeWeb Daily News.
Brain cancer is one of the most aggressive and deadly types of cancer. It also tends to be characterized by complicated genetic patterns affecting numerous genes on many different chromosomes. For instance, Bredel noted, researchers often find coincident chromosome 10 losses and chromosome 7 gains in a type of brain cancer called glioblastoma.
But Bredel and his colleagues suspected that the genetic and genomic changes did not occur at random. Instead, they hypothesized that there may be networks of related changes. To test this, they pulled together gene dosage, expression, mutation, and other data from 45 brain cancer patients treated at Stanford University — the team's initial discovery set.
They subsequently integrated information for a validation set including another 456 tumors from TCGA, the University of California at Los Angeles, the University of Texas MD Anderson Cancer Center, and elsewhere.
Overall, the researchers saw that many of the chromosomal areas tended to be altered in brain cancer, including regions on chromosomes 1, 7, 8, 10, 12, 13, 19, 20, and 22.
When they focused in on the genes involved and the types of changes to these genes, the team found that the genes fell into a network with distinct interactions and architecture, Bredel said. At the center of this network were genes with lots of connections, which he compared to the "mob bosses" of the brain tumor genetic landscape. These genes are "attractive targets for future therapies," Bredel said, because they tend to be central in the network.
Among the most interactive genes were some with known roles in tumor biology, the researchers noted, including the EGFR, MYC, and PTEN. By focusing on the most significant interactions, the team came up with a list of interactions involving hub and hub-interacting landscape genes that seem to have a "cooperatively tumorigenic relationship."
And seven of the 31 most intriguing landscape genes were independently associated with patient survival: POLD2, CYCS, MYC, AKR1C3, YME1L1, ANXA7, and PDCD4.
Together, the researchers found that this seven-gene set could retrospectively classify patients into sub-groups linked to survival times. Individuals who had alterations in between zero and two of the seven genes were classified as low risk, while those with five or more affected genes were considered high risk. Those in between were classified as high risk.
For instance, in the training set, which consisted of 189 TCGA glioblastomas, the researchers reported 49.24 deaths per 100 person-years in the low-risk genetic group but 79.56 deaths per 100 person-years in the high-risk group — a pattern that the team confirmed in three retrospective validation studies.
Based on these results, Bredel says he is optimistic that this type of approach could have clinical applications both for improving brain tumor classification methods (currently based on histology and clinical factors such as age) and guiding treatment decisions. He and his colleagues also hope their findings will spur the development of new therapies based on key brain cancer pathways.
Boris Pasche, director of the University of Alabama at Birmingham's hematology and oncology division, who was not involved in the study, agreed that the work reveals pathways that could make valuable drug targets. Pasche told GenomeWeb Daily News that he believes that down the road it will also be important to integrate other types of information — such as microRNA profiles and proteomic data — into cancer networks.
Pasche and HudsonAlpha Institute for Biotechnology President and Director Richard Myers wrote an editorial on the new brain cancer research, also appearing in today's JAMA. In it, the duo said that the "potential clinical implications of these findings are significant."
Bredel and his team are continuing to characterize the mechanism by which ANXA7 regulates EGFR — work that Bredel says the researchers plan to publish within a few months. In the future, Bredel said, they hope to do prospective clinical trials testing the clinical utility of the seven-gene set. They may also apply a similar genetic landscape approach to other aggressive types of cancer, such as ovarian and lung cancer, he added. Eventually, they may also create networks that account for both genetic and epigenetic changes in cancer cells.