Genetics:
Protein Reverses Chromatin Engineering

Biological Chemistry:
Molecule Implicated in Transcription Termination

Structural Biology:
DNA Splicing Enzyme Observed in Action

Scientific Symposium:
Fashions Change in Modeling Disease

Chronic Periodontitis Differs at the Microbial Level in Populations Worldwide

Brain Structure for Reward and Punishment Smaller in Cocaine Addicts

Brigham Celebrates 50th Anniversary of First Human Organ Transplant

Beth Israel Assumes Academic Oversight of Mass. Mental Health Faculty

Joslin Names Conley Chairman of the Board

Academic Officer Tapped for HMI Dubai Project

Macklis Receives Javits Neuroscience Investigator Award

Global Citizen Award Goes to Bill Moyers

HMS Family Health Guide Published in Paperback

New Appointments to Full Professor

Honors and Advances

Community Celebrates a Child's First Laugh

Front Page

Fashions Change in Modeling Disease

"Whether a phenocopy is a pathocopy depends upon how tens of thousands of genes are behaving in a coordinated fashion over time."

Yet with each apparent advance, researchers must navigate confounding obstacles in moving discoveries from the laboratory to the clinic. On Nov. 17, more than 400 people met for a daylong symposium in the new research building amphitheater to compare notes on better ways to model human disease. The symposium was sponsored by HMS in collaboration with Charles River Laboratories. Nine speakers addressed how to illuminate disease mechanisms with new tools and techniques for in silico, in vitro, and in vivo modeling, including the use of human subjects.

Animal phenotypes can be deceptive, said Isaac Kohane, the Lawrence J. Henderson associate professor of pediatrics and health sciences and technology at Children's Hospital Boston and HMS. A mouse that seems to model a human disease may have very different physiology happening underneath the skin. Kohane and his colleagues are pursuing a time series analysis of the genes that are upregulated and downregulated in man and mouse to verify the biological validity of mouse models. They published their first validation of the approach in the March 15 Genes and Development. "Whether a phenocopy is a pathocopy depends upon how tens of thousands of genes are behaving in a coordinated fashion over time," Kohane said.

Other genomic tools can be used to make animal models more predictive of outcomes in human trials, said Richard Roman, professor of physiology and medicine at the Medical College of Wisconsin. "Ninety percent of drugs fail in clinical trials, half for efficacy and half for safety," Roman said. "All are efficacious and safe in animal studies. Genetically, rodents are 90 percent identical to humans. Why then do they have only 10 percent predictive power for responses in humans?"

Gene function approaches and the way scientists design experiments using them may be impeding progress, Roman said. The process is too slow, and human diseases are complex and multigenic. Looking for the clusters of genes that cause hypertension and kidney disease, Roman and his colleagues have developed "consomic" rat strains with single chromosome substitutions. Instead of knocking out single genes in a mouse to create a disease phenotype, they swap pieces of chromosomes between normal and diseased strains to find regions that contain genes that prevent disease in rats with a susceptible genetic background. With a new potential drug target, they can search drug libraries for a molecule that can regulate the gene expression and see if it can treat the disease in rats. They also have developed six genetically diverse rat strains to better detect toxicity of potential drugs, which may not be picked up by a single strain.

Disease modeling extends to the clinical responses of people to therapeutics. "We can use clinical trials as vehicles to understand basic disease pathology," said keynote speaker Marc Tessier-Lavigne, senior vice president for research drug discovery at Genentech in South San Francisco. "Drug candidates are powerful tools to probe disease mechanisms."

Even the most promising compound can fail at the final stage of clinical testing. That is when scientists need to stay the course to deconstruct the complexity of disease processes, advised Tessier-Lavigne. The drug erlotinib (Tarceva) is a case in point. Although it failed its initial efficacy trials in 2,000 people with lung cancer, subsets of these patients clearly benefited. The medication, for example, doubled the survival time of the lung cancer patients who had never smoked. Other promising findings came from two independent teams of HMS researchers. They discovered that mutations in the epidermal growth factor receptor determine whether tumors shrink in response to the closely related drug gefitinib (Iressa). (See Focus, May 14, 2004.) The mutations have helped identify some of the people who are likely to benefit, but other tumor characteristics may help explain the dramatic response in people who have never smoked.

One lesson from this work is that targeting the relevant patients is essential to successful prosecution of clinical trials, said Tessier-

Lavigne, a point underscored by the results of a later trial of patients with advanced or metastatic non- small cell lung cancer, in which erlotinib showed a survival benefit across all subjects. The day after the symposium, the Food and Drug Administration approved erlotinib for this indication.

--Carol Cruzan Morton