Scientists usually order laboratory mice online, but immunologist David Masopust went to more trouble. While doing research years ago at Emory University, he drove to a barn several hours away to trap the rodents himself. He suspected that commercially provided lab mice were missing some key immune cells because they had inexperienced immune systems—a result of being raised in extremely hygienic facilities. Masopust, now a professor at the University of Minnesota, went on to formally test his suspicion over the course of a decade and has found that it was correct: lab mice used by the scientific community and pharmaceutical world to test drugs and vaccines for human diseases are in some ways poor models of the human adult immune system.

As published this spring in the journal Nature, Masopust and his colleagues discovered that mice raised in germ-free facilities had immune systems that looked more like those of human babies than adults, as judged by the types of immune cells present and the genes that were active in those cells. For example, memory CD8+ T cells that serve as first responders to infection were virtually undetectable in adult lab mice but clearly present in barn mice and mice from pet stores. “We've ‘known’ this, but it's good to finally see it proven,” says Purvesh Khatri, a computational systems immunologist at Stanford University who was not involved in the study.

What is more, when the researchers housed “clean” lab mice with “dirty” pet store mice (which carried germs), about a fifth of the lab mice died of infections within a few months. The mice that survived, however, developed more robust immune repertoires, and the gene activity of their immune cells shifted to resemble those of adult humans. In follow-up experiments, those mice fought off bacterial infections just as well as mice vaccinated against the pathogens.

These results suggest that having lab mice share space with animals from the wild or from pet stores could give researchers a more realistic view of disease progression and treatment responses in human adults. Additionally, by showing that lab mice fail to model key immune features, the study could partially explain why therapeutics tested in animals often fail in human trials. “Variables that matter in the real world aren't present in controlled experiments,” Khatri explains.