Genomic Tools Identify Profiles of Gene Activity Underlying Cancer

Researchers have developed a new gene-profiling technique that can define complex patterns of gene activity in cancerous tumors having different genetic origins.

By developing a more complete picture of the subtle differences in gene activity associated with cancer-causing pathways, the method might ultimately allow physicians to tailor treatments to individual patients and better predict disease outcomes, said the researchers.

“The gene profiles allow us to be much more precise,” said study co-author Joseph Nevins, Ph.D., a Howard Hughes Medical Institute Investigator and James B. Duke Professor of molecular genetics and microbiology at Duke University Medical Center. “Now, we can find subtleties among tumors that we can’t see under the microscope and also uncover the biology to better understand the basis of the disease.” The method highlights the power of genomics and statistics to move toward a personalized approach to medicine, Nevins said.

Mutations in individual cancer-causing genes, called oncogenes, elicit a cascade of changes in the activity of hundreds of other interacting genes — either increasing or decreasing their activity — explained study lead author Erich Huang, M.D., who will begin his surgical residency this summer at Duke University Medical Center.

Rather than focusing on single genes in isolation, the new method provides a more global view of the many shifts in gene activity characteristic of cancers having a specific genetic cause, the Duke team said.

The study, a collaboration between the Duke Institute for Genome Sciences and Policy (IGSP) and Georgetown University’s Lombardi Cancer Center in Washington, D.C., will be published online in Nature Genetics on May 19, 2003. The new results build upon the Duke team’s earlier work, published in the May 10, 2003, issue of The Lancet, which found that a similar genetic profiling method could predict the spread and recurrence of breast cancer with a high degree of accuracy.

“We’re just beginning to see the predictive value of gene expression data that reflects gene activity in determining how a tumor will fare under a particular therapy,” said study co-author Mike West, Ph.D., Arts & Sciences professor of statistics and decision sciences at Duke University. This new study focuses on specific gene networks known to contribute to a wide variety of human cancers including those of the breast, lung, and colon.

The team first manipulated normal mouse cells to cause an increase in the activity of single cancer-causing genes, known as RAS and MYC. “These genes are prototypical oncogenes,” said Huang. “They were therefore natural choices for study.” Increased oncogene activity leads the genes that govern cell growth to become deregulated. In cancerous tissue, one or more of these pathways are often deregulated, triggering the uncontrolled cell growth characteristic of the disease.

Next, the researchers used DNA microarray, or gene chip, technology to measure global increases and decreases in gene activity that resulted from changes in the individual oncogenes. Gene chip technology works by measuring the level of molecules known as messenger RNA (mRNA)—the result of a gene’s activity. The resulting mRNA is translated into proteins that carry out the myriad functions of the cell. The researchers used gene chips that could measure the activity of over 10,000 genes simultaneously.

The team then analyzed information from the microarrays to explore changes in the activity of genes and identify clusters of genes, which the Duke team has dubbed metagenes, whose increase or decrease in mRNA levels differed between normal cells and those with increased activity in particular cancer-related genes.

These metagene profiles were then used in a statistical model to predict the activity of other genes controlled by the oncogenes. They found that their genetic model accurately identified cells in which the pathways were active.

Similarly, additional profiles characteristic of another key set of related proteins, known as E2F, also accurately discriminated between cells with an excess of either E2F1, E2F2 or E2F3 proteins — a notable accomplishment given that the three proteins largely overlap in terms of their function, the researchers said.

In a groundbreaking application of the method, Huang and his colleagues then tested how well these “genetic signatures” could distinguish between cancers, by measuring the metagenes’ ability to identify cancerous cells having defects in a single oncogene, MYC or RAS.

Testing the genetic profiles against tumors in mice was the final and biggest challenge, said Huang, and the results were impressive: the method accurately identified the genetic basis of mammary tumors in transgenic mice as a defect in either the oncogene MYC or RAS.

“Tumor tissue has lots of heterogeneity,” added Huang, who said that tumors can include normal cells, as well as blood vessels and other noncancerous tissues. “The models were able to filter through all that noise and predict whether a tumor was aberrant for one oncogene or another,” he said.

“The power of this analysis is shown by its ability to accurately … predict tumors arising from deregulation of specific oncogenes,” the researchers wrote. The next step, they said, will be to identify patterns of relationship between gene profiles and the outcome of disease and to use that information to make educated decisions about cancer treatments.

Other Duke members of the team were Seiicha Ishida, Ph.D., Jennifer Pittman, Ph.D., Holly Dressman, Ph.D., Andrea Bild, Ph.D., and Mark Kloos, Ph.D. Lombardi Comprehensive Cancer Center collaborators included Mark D’Amico, Ph.D., and Richard Pestell, Ph.D.