News & AnalysisToxicology

'Humanized' Mouse Detects Deadly Drug Side Effects

See allHide authors and affiliations

Science  18 Apr 2014:
Vol. 344, Issue 6181, pp. 244-245
DOI: 10.1126/science.344.6181.244
Recipe for a chimeric mouse.

Researchers first kill mouse liver cells, and then transplant human hepatocytes.


After five patients died in a small clinical trial of an experimental hepatitis B drug in 1993, high-level committees scrambled to work out what went wrong. In the end, they blamed not the researchers but the lab animals. Earlier studies, including several in animals, simply failed to show that the drug was toxic.

Now, researchers say they have developed an animal model that would have detected the drug's toxicity: an engineered mouse with a humanlike liver. They say it could serve as a versatile testbed for liver toxicity, a worrisome side effect of many drugs. "The study is a tour de force," says Jay Hoofnagle, director of the Liver Disease Research Branch at the National Institute of Diabetes and Digestive and Kidney Diseases in Bethesda, Maryland. Hoofnagle was not involved in the work, although his division partly funded it. He led the infamous 1993 trial, which tested a hepatitis B drug known as fialuridine (FIAU).

That phase II study was designed to evaluate different doses of FIAU in 15 people who had chronic hepatitis B infections. The drug caused severe lactic acidosis in seven patients; two died quickly, and five received liver transplants, but only two survived. Earlier experiments in mice, rats, dogs, and monkeys had tested FIAU at various doses and found no detectable harm. Similarly, small human studies conducted by other researchers had not tied any serious adverse events to the drug. Expert committees convened separately by the U.S. National Institutes of Health (Science, 10 June 1994, p. 1530) and the Institute of Medicine (Science, 24 March 1995, p. 1759) each determined that Hoofnagle and collaborators had acted appropriately in conducting their trial.

Gary Peltz, a rheumatologist at Stanford University School of Medicine in California, heads one of several groups that have engineered mice to have "humanized" livers. Other groups have primarily used their models to investigate fundamental questions about metabolism, and some have reported drug effects. But none has taken the toxicology model as far as the FIAU study that Peltz and co-authors published in the 15 April edition of PLOS Medicine. This is "the first definitive example of a human-specific, drug-induced toxicity that could be detected in chimeric mice, but not in conventional rodent, dog, or monkey toxicity studies," the researchers assert. (One of the co-authors, Stanford's Jeffrey Glenn, has collaborated with Hoofnagle on hepatitis studies.)

Peltz's team destroys the liver cells in an immunodeficient mouse strain called NOG by stitching in a transgene of an enzyme, thymidine kinase, from herpes simplex virus-1. Then they give the mice the drug ganciclovir, which becomes active only in the presence of this enzyme, selectively killing mouse liver cells but not otherwise harming the animals. Finally, they transplant human liver cells into the mice, which they dubbed TK-NOG. The resulting mice live for up to 8 months, and their livers metabolize drugs in much the same way as a human liver.

For their new study, Peltz and co-workers wanted to demonstrate the mouse's value for drug testing. "I called a number of people and asked what was the best example of a drug where preclinical toxicology failed to predict what would happen in humans," says Peltz, who formerly ran genetics and genomics research at Roche's now-closed Palo Alto, California, lab. FIAU seemed a perfect fit. (After the FIAU disaster, a study in woodchucks infected with their own version of hepatitis B did detect the drug's toxicity, but woodchucks are rare in toxicology labs.)

In the new study, Peltz's team compared various doses of FIAU in the TK-NOG mice, using NOG mice that retained their mouse livers as controls. The first experiment, which gave the mice a high dose of FIAU, began on a Monday. Peltz thought it would take a few months before any toxicities showed up in the TK-NOG animals. "On Wednesday, [the lab] called and they said, 'You may want to stop this. The mice are really sick,' " Peltz remembers. "It was the happiest I've ever been about a sick mouse." One mouse had died and 14 others had acute liver failure. FIAU had no effect on the NOG mice. Lower doses of FIAU caused illness or serious liver abnormalities in the TK-NOG mice, and the severity of the effects increased with the dose—a key indication, the authors argue, that the drug caused the toxicities. Different doses of a safe, widely celebrated hepatitis C drug, sofosbuvir (Science, 13 December 2013, p. 1302), did not harm the TK-NOG mice.

Markus Grompe, a pediatrician at Oregon Health & Science University in Portland, says the new work is a "solid report" and "good proof of principle." But Grompe, who has a company that markets his own chimeric mice, says he expects all the existing mouse models with humanized livers, including his own, would have performed equally well. Peltz, who says "at present" he has "no financial interest" in his mice, insists that other chimeric mice have ongoing liver toxicities caused by the process used to engineer them; the two researchers slugged it out in the May and August 2013 issues of Trends in Pharmacological Sciences.

Charles Rice, who studies hepatitis C virus at Rockefeller University in New York City, calls Peltz's study "interesting" but says "critical controls" are missing. He says the immunodeficiency in the TK-NOG mice could have contributed to FIAU's effects. To rule this out, Rice says he'd like to see FIAU tested in TK-NOG mice reconstituted with liver cells from rats, dogs, and monkeys.

Peltz says the team will test other drugs that, like FIAU, did not show toxicity in animal studies but caused harm in humans. He'd also like to create chimeric mice with human liver cells from different genetic backgrounds to see if results from the mice could help link drug effects to specific populations.

Hoofnagle says the humanized mice might help answer a fundamental question about FIAU: Why did it cause harm in the first place? The drug exists in two different structures, or isomers, and some evidence suggests only one is dangerous to humans. Says Hoofnagle: "This model might be an elegant means to sort out these issues."

Navigate This Article