HomeNewsGene Expression Data Matches Old Drugs to New Uses


Gene Expression Data Matches Old Drugs to New Uses


An HHMI-funded researcher has used online gene expression profiles to match old drugs with diseases in need of treatments.

Singles aren’t the only ones using online profiles to find new partners. A Howard Hughes Medical Institute-funded researcher has used online gene expression profiles to match old drugs with diseases in need of treatments.

This novel approach uses a new computer program to compare gene expression patterns from diseased cells with those from cells treated with individual drugs. When the program finds a gene pattern that matches, it suggests a potential drug-disease pair. And so far, the team is two-for-two when they tested proposed matches in animal models: an over-the-counter heartburn medicine slowed lung cancer growth in mice and an anti-seizure drug controlled inflammatory bowel disease.

“We always hope our experiments will work, but I’m still unexpectedly happy that this approach worked for us twice,” says Atul Butte, an HHMI Early Career Physician Scientist awardee at Stanford University School of Medicine. The work was published in two papers on August 17, 2011, in Science Translational Medicine.

Using molecular similarities with our new method gives us a leg up, compared with looking at patient responses. We have a higher level of resolution for both disease biology and drug effects, so we can match them better.

Atul Butte

The idea of repurposing existing drugs that are already known to be safe isn’t new. “The crude classic example is to see a patient have a side effect and then use that side effect to guide a new use,” Butte says. He points out that some blockbuster drugs were found this way, including Viagra. Until now, however, there have been few strategies to systematically test a large number of diseases and drugs simultaneously. That means the repurposing has been haphazard, even though it is an opportunity to keep health care costs down. “Using molecular similarities with our new method gives us a leg up, compared with looking at patient responses,” Butte says. “We have a higher level of resolution for both disease biology and drug effects, so we can match them better.”

Butte started thinking about the approach in 2004, as a recently-trained physician who returned to graduate school at Harvard Medical School and Massachusetts Institute of Technology in Boston. Biologists were rapidly generating gene expression profiles from a large variety of cell types and disease states, then depositing the data in publicly-available databases on the web. “As the collection grew, I realized we could start to build a taxonomy of diseases, to see what diseases looked like the others in terms of what genes were expressed,” says Butte, who worked as a software engineer at both Apple and Microsoft before getting his M.D. and Ph.D.

The comparative approach became an even better opportunity after Todd Golub, an HHMI investigator at the Broad Institute in Cambridge, and his colleagues started the Connectivity Map. Golub and colleagues have been pioneers in sharing their molecular research measurements on the internet, and their database of shows how gene expression patterns change when cells are treated with small molecule drugs—including many FDA-approved drugs.

“We came up this idea that if we now had the gene expression data for so many diseases and for all of those drugs, we could find therapeutic matches,” Butte says. Building from Golub’s work, Butte reasoned that if a given disease drove expression of a set of genes in one direction and a drug drove their expression in the opposite direction, then the drug might be able to reverse or mitigate the biological pathways underlying that disease.

The overall project, described in the two Science Translational Medicine papers, compared gene expression profiles from 100 diseases and 164 drugs and small molecules. The team uncovered potential therapeutics for 53 diseases and more than 1,300 potential therapeutic pairs. A quick scan of the pairs showed some well-established disease treatments, indicating the method was working. For example, the program matched inflammatory bowel disease and prednisolone, a steroid that is a standard treatment for the autoimmune disorder.

Unexpectedly, a drug used to control epilepsy, topiramate, scored even higher as a potential therapy for inflammatory bowel disease than the existing treatment. To find out if the in silico match would actually work in a biological setting, the team tested topiramate in a rat model of inflammatory bowel disease. Both topiramate and prednisolone reduced the animals’ symptoms compared to untreated animals, and a post-mortem analysis showed that topiramate, like prednisolone, reduced colon tissue swelling and damage, compared with the control animals. The team is now working on more pre-clinical experiments testing topiramate as a therapy for inflammatory bowel disease and related conditions, such as Crohn’s disease and colitis. If the therapy succeeds in these experiments and subsequent clinical trials, patients may be able to avoid the severe side-effects associated with regular corticosteroid use—such as high blood sugar, calcium loss from bone, and increased risk of infections—while maintaining the benefits of therapy.

Perhaps even more surprising to Butte was a predicted match of cimetidine (Tagamet) and lung adenocarcinoma. Cimetidine is sold as a treatment for heartburn and ulcers because it reduces the amount of acid produced by the stomach. But the gene expression program predicted that it was as good a match to treat lung cancer as an existing treatment called gefitinib (Iressa). Based on the prediction, the team implanted human lung cancer cells in mice and watched what happened when the animals were treated with cimetidine, a known chemotherapy called doxorubicin that is commonly used in lung cancer, or a saline control. After 12 days, the tumors in mice treated with cimetidine had grown 2.3-fold, compared with 2-fold growth in animals treated with doxorubicin and 3.3-fold in animals treated with saline. The results suggest that cimetidine does have some dose-dependent anti-tumor activity, though not as much as the standard therapy. But Butte thinks the data are a good starting place, and that his team might be able to improve cimetidine’s activity, perhaps by altering its dose and treatment schedule or by combining it with other agents.

Butte says his team has a lot more work to do before cimetidine could be used as a chemotherapeutic, but at least the drug is now on the map as a drug to consider. “Cimetidine is available over the counter,” Butte says. “If it could really be useful as a chemotherapeutic for lung cancer that would be an amazing story for improving patient care.”

For More Information

Jim Keeley
[ 301-215-8858 ]