NEW YORK (GenomeWeb News) – In a paper appearing online today in the journal Cancer Cell, members of The Cancer Genome Atlas Research Network reported that they identified and characterized four subtypes of the brain cancer glioblastoma multiforme.
After finding the subtypes using gene expression profiling, the team integrated data on everything from somatic mutations and copy number changes to clinical characteristics and treatment outcomes to help them better understand each sub-group. In the process, they found subtype-specific differences in patient outcomes following aggressive treatment, suggesting these classifications may eventually help guide GBM treatments.
"This comprehensive genomic and genetic-based classification of GBM should lay the groundwork from an improved molecular understanding of GBM pathway signaling that could ultimately result in personalized therapies for groups of patients with GBM," senior author Neil Hayes, a hematology and oncology researcher at the University of North Carolina's Lineberger Comprehensive Cancer Center, said in a statement.
The National Cancer Institute and National Human Genome Research Institute launched a three-year Cancer Genome Atlas pilot project in late 2005. The Cancer Genome Atlas, or TCGA, effort has since expanded to include more than 20 different types of cancer.
The first work from TCGA was published in the fall of 2008, when researchers reported in Nature on copy number, gene expression, and DNA methylation changes in more than 200 GBM samples and sequenced 601 genes in samples from 91 GBM patients. Results from that and other studies suggest three genes — EGFR, NF1, and PIK3R1 — are often mutated in GBM.
For the current study, TCGA researchers integrated gene expression data generated from three platforms for 200 GBM samples and two normal brain tissue samples, identifying 1,740 genes in four clusters that showed consistent but variable expression with the three platforms.
In their subsequent analyses, they narrowed in an 840-gene signature — verified in independent data sets — for classifying the tumors into these four groups.
Next, the team attempted to integrate gene expression and other types of genomic data for the four GBM subtypes, which they dubbed proneural, neural, classical, and mesenchymal. To do this, they assessed copy number data for 170 samples as well as sequence data for 601 genes in 116 samples.
For instance, the researchers reported that samples belonging to the classical GBM subtype all shared a chromosome 7 amplification and a chromosome 10 deletion. This subtype also tended to carry amplifications affecting the EGFR gene as well as other EGFR mutations. Mutations in the TP53 gene, on the other hand, rarely appeared in the classical subtype.
In contrast, tumors in the proneural subtype frequently harbored TP53 mutations. This subtype also tended to contain mutations and amplifications involving the PDGFRA gene, along with point mutations in IDH1 and expression changes involving several developmental and cell cycle related genes.
The mesenchymal subtype was characterized by deletions affecting an NF1 containing region on chromosome 17 called 17q11.2. Tumors in this group also tend to have elevated CHI3L1 and MET expression.
As its name suggests, the neural subtype contained tumors with alterations and expression changes affecting neuron-related genes, including GABRA1 and SLC12A5. Along with GBM samples, the normal brain samples also fell into this group.
The team also started exploring how expression patterns in each subtype compared with those in various brain cell types — work that they say may ultimately provide information about the cellular origin of each subtype.
"We discovered a bundle of events that unequivocally occur almost exclusively within a subtype," Hayes said. "These are critical events in the history of the tumor's development and spread, and evidence is increasing that they may relate to the initial formation of the tumors."
And by looking at clinical features, survival times, and treatment outcomes associated with each subtype, the researchers found evidence suggesting aggressive treatments that benefit individuals with classical, mesenchymal, and neural subtypes don't improve outcomes for those with proneural GBM — a subtype that frequently includes tumors from younger patients.
The aggressive treatment, which involved concurrent chemotherapy and radiotherapy or three chemotherapy cycles, significantly improved outcomes in the classical and mesenchymal sub-groups and tended to improve outcomes in the neural sub-group but did not seem to benefit the proneural group.
"This comprehensive genomic- and genetic-based classification of GBM should lay the groundwork for an improved molecular understanding of GBM pathway signaling that could ultimately result in personalized therapies for groups of patients with GBM," the researchers concluded.
"TCGA is mobilizing the entire cancer community to find new strategies in detecting and treating cancer faster," National Institutes of Health Director Francis Collins said in a statement. "These findings are just a hint of what we expect to result from the comprehensive data generated by TCGA over the next few years."
TCGA data is available to members of the research community through the Cancer Genome Atlas Data Portal.