Scientists demonstrate new method for discovering cancer gene function

Using a new approach for dissecting the complicated interactions among many genes, scientists at Dana-Farber Cancer Institute have discovered how a common cancer gene works in tandem with another gene to spur the unchecked growth of cells. The researchers say the technique was so useful in solving a longstanding puzzle that it may expedite the discovery of other such gene interactions that lead to cancer, and could accelerate the development of new cancer drugs.

The report in the Aug. 8 issue of Cell describes how the method was used in identifying what additional genes are affected by the common oncogene, cyclin D1, when it makes too much of its normal protein. By combining two types of data and applying a powerful statistical tool, the researchers pinpointed an unsuspected gene, C/EBP-beta, as a key mediator of cyclin D1 action.

Justin Lamb, PhD, a molecular biologist at Dana-Farber and lead author of the report, said the previous experimental efforts had failed to identify genetic accomplices of cyclin D1 in triggering cancer. “We didn’t know what cyclin D1 was interacting with,” said Lamb.

Sridhar Ramaswamy, MD, second author of the paper and the developer of a human tumor database used in the experiment, said that even when similar cancer gene interactions had been discovered in the laboratory, “the open question has been whether what is found in the test tube is what really happens in human tumors. The patterns of gene expression we found in human tumors corroborated our in vitro finding, and I think that the paper represents a proof of concept of the approach.”

The research was carried out in the laboratory of Mark E. Ewen, PhD, senior author of the paper.

Cancer is now viewed as a set of diseases fundamentally caused by mutations in genes that normally regulate all aspects of cellular function. Dozens of mutated oncogenes are known: some overproduce a growth signal to produce runaway cell division, while others fail to exert their normal check on excessive growth.

Oncogenes, however, don’t act alone. Their aberrant signals are relayed through other genes that interact in complex cascades, akin to a bucket brigade used to fight a fire. Lab procedures used to sort out which genes (actually, the proteins they produce) connect with each other rely on cancer cells in culture. This is not necessarily comparable to cells from a human patient. These studies are also time-consuming and require highly skilled scientists and technicians.

Lamb sought a way to combine, and perhaps shorten, this molecular biology approach with another strategy – one which would make use of data from the Global Cancer Map, a database established by Ramaswamy of information from hundreds of samples of human tumors and normal tissue.

For one arm of the experiment, the researchers artificially overexpressed cyclin D1 in human breast cancer cells in a test tube culture. They collected the genetic output of those cells, in the form of the RNA “messages” made by the thousands of genes in the cancer cells. The levels of some of these messages would have been altered by the action of the overexpressed cyclin D1 gene. To identify those messages, and the genes that produced them, the scientists placed RNA from the cancer cells on gene chips, or microarrays, which measure the activity of thousands of genes at once.

The gene chips identified 21 genes that were spurred into high gear by cyclin D1, making them prime suspects for being “downstream targets” of cyclin D1 overexpression. The scientists could then compare this 21 gene “expression signature” with gene activity in actual human tumors containing an overactive cyclin D1 gene. “The question was, could we find that same set of genes in human tumor samples,” said Lamb. “If so, we would know if they were relevant in human cancer.”

The activity of different genes in 190 human tumor samples was already profiled in the Global Cancer Map, and, using a statistical tool called the Kolmogorov-Smirnov metric, the scientists showed that the 21 genes identified by the gene chips correlated closely with human tumors having high expression levels of cyclin D1.

Finally, the scientists used the Kolmogorov-Smirnov tool again, this time to sort among the 16,000 different genes in tumors in the cancer map database, ranking each one of the 16,000 by its similarity in expression pattern to those in the 21-gene expression signature discovered in the laboratory cancer cells. The process can be likened to the results of an Internet search engine that ranks each “hit” by how closely it matches the keywords in the search.

It was this “data-mining” process that turned up the C/EBP-beta gene, which makes a protein known as a transcription factor, as a frequent co-conspirator with the genes in the cyclin D1 expression signature. The data strongly suggest, say the scientists, that the C/EBP-beta is involved in regulating genes affected by cyclin D1 overexpression, and is therefore likely a principal participant in a previously-unappreciated mechanism of cyclin D1 action.

“By inference,” says Ramaswamy, “C/EBP-beta is required for cyclin D1 to exert its effect in human cancer.” If so, the gene might be added to the list of targets in cancer cells that could be attacked with highly specific drugs.

Contact: Bill Schaller, william_schaller@dfci.harvard.edu