Every year unforeseen toxicity scuttles new drugs, sometimes even after they've received FDA approval. A major reason is that it's difficult to predict the response of the liver, where drug toxicity often shows up.

"Liver toxicity issues are the primary reason for drug recall and withdrawal," according to Yvonne Dragan, director of systems toxicology at the FDA's National Center for Toxicological Research.

So it's not surprising that mimicking liver toxicity in the lab would be the centerpiece of efforts to predict a drug's toxicity early in the development process.

In the past, such screenings have been difficult because actual liver cells, once they are taken from the body, stop acting like liver cells within a day. But now Sangeeta Bhatia, an M.D., medical engineer, and associate professor at MIT, has found a way to keep liver cells doing their job for weeks. What's more, she has developed a method for arranging them in multiple test wells, allowing drug researchers to screen multiple compounds at the same time.

Because the liver cells function for such a long time, they can be used to test for chronic toxicity, which is caused by low-level exposure to a compound over time. "That's something you can't do in other platforms currently in use," says Bhatia.

Bhatia's innovation was to develop a way of organizing the liver cells using photolithography, the same process used to create computer microprocessors. First, a pattern is created in test wells for a layer of proteins. Then liver cells are applied, which stick to the proteins, replicating the patterns. Finally, supportive cells are added, filling in around the liver cells.

Using this technique, Bhatia was able to experiment with many different configurations of the two cell types, until she found the one that kept the liver cells functioning for weeks.

That is, she found the pattern that works for rat liver cells. As it turns out, "human cells are completely different," says Bhatia. Applying a technique called "soft lithography," which uses a polymer stencil to pattern the proteins, Bhatia found patterns that let human liver cells function for a month. Then she tested to see if these liver cells could correctly predict the effect that drugs would have on a human liver.