News Feature: What happens when lab animals go wild

Experiments on mice living in more natural habitats can deliver results dramatically different from those in traditional laboratories—with profound implications for biomedical science.

In the summer of 2015, a pioneering band of laboratory mice did something their ancestors hadn’t done for roughly 1,000 generations—they went outside.

Andrea Graham of Princeton University moved lab mice outdoors to test their susceptibility to nematodes in a more natural setting. Image courtesy of David Tricker (photographer).

It was hardly a trek into the wilderness. The 90 mice were fenced into pens, with feeding stations providing all the mouse chow they could eat and aluminum pie plates dangling over their heads to deter passing hawks. Still, it was a world away from their former home in the laboratory of Andrea Graham, an ecological and evolutionary immunologist at Princeton University. These mice could now roam around an area of roughly 180 square meters, feeling the dirt under their feet and rain on their backs.

Graham’s project aimed to show how a mouse’s natural environment affects its susceptibility to parasitic worms called nematodes, which live in the animal’s digestive tract. As with so many other conditions, from cancer and diabetes to Alzheimer’s disease and stroke, scientists study nematode infestations in mice to develop treatments for humans. Most of these studies take place in laboratory settings where researchers can control myriad complicating variables such as temperature, diet, and social interactions.

But researchers such as Graham are questioning whether this tactic is always the best approach. If the natural environment of a mouse—or a human—is itself a major factor affecting a disease or its treatment, studying it under strict lab conditions could skew the results. As pressure mounts for scientists to make mouse findings translatable to humans, a small but growing number of researchers are designing studies that use more natural experimental conditions. Their aim is not to replace traditional lab studies but rather to complement them with real-world context.

Some studies, for example, have already shown that experiments on mice in a barn-like setting can uncover a drug’s potential side effects, which traditional preclinical research had missed. And Graham’s results from the nematode study, which used a nematode-resistant strain of mouse, were “just eye-popping,” she says. A few weeks after she infected both indoor and outdoor mice with nematodes, the outdoor group harbored massive infections, with a mass of worms 100 times greater than the indoor group (1).

Researchers have good reason to establish a controlled environment, long a hallmark of scientific rigor. And as some start to bring their lab animals into wild territory, they grapple with many questions: What is the best way to mimic the wild? How should researchers monitor animals without disturbing their environment? And how much control can they really afford to give up before experiments become irreproducible or logistically and financially intractable? One thing is certain: Take mice out of the cage and the results can be fundamentally different.

Translation Trouble

Mice are still the human stand-in of choice in biomedicine. The vast majority of their genes have a counterpart with the same function in humans, and they suffer from many of the same diseases. Researchers can manipulate mouse genomes to emulate the human conditions. And they can make experiments more repeatable using inbred strains—lineages of genetically identical mice created through 20 or more generations of mating between sisters and brothers or parents and offspring (2).

Yet most successful therapeutic treatments in mice do not translate successfully into humans (3). “The ultimate goal is translatability,” says laboratory veterinarian Norman Peterson at MedImmune in Gaithersburg, MD, who serves as a board member of the North American 3Rs Collaborative, an organization dedicated to replacing, reducing, and refining the use of laboratory animals. “If an animal is in an unnatural environment and stressed and hormones and cytokines are out of whack, the question is: Is that the best model? In most cases it is not.”

In some ways, the life of a lab mouse is unnaturally stressful. Many argue, for example, that the standard temperature for mouse facilities (20–26 °C), the subject of one of several continuously updated recommendations contained in the National Research Council’s guide to lab-animal care and use (4), is comfortable for researchers but lower than optimal for mice (5, 6).

In other ways, the life of a lab mouse is unnaturally cushy—no competition for territory and all the food they can eat. “Most organisms function better under some level of eustress, the good stress,” says laboratory animal veterinarian Kathleen Pritchett-Corning of Harvard University in Cambridge, MA. “When you respect the animal’s natural behaviors, you get the best results out of those animals.”

Biologist Wayne Potts of the University of Utah in Salt Lake City has been encouraging natural behavior in his study animals since the late 1980s. As a postdoctoral fellow at the University of Florida, Potts wanted to understand how natural selection and sexual selection affect the major histocompatibility complex (MHC), a collection of genes that helps the immune system identify bacteria and other foreign substances. He built a barn-like structure to house wild strains of mice so he could watch these forces play out naturally, and he found that female mice preferred males with MHC genotypes different from their own (7). But he also realized that the “mouse barn” offered a powerful way to study mice under more natural conditions.

Since then, Potts has conducted many experiments that he calls organismal performance assays (OPAs) in structures that function much like barns. He uses strains of genetically diverse house mice (Mus musculus) that breed outside of family groups, unlike classic inbred lab strains. Potts typically puts about 30 of these mice into 32-square-meter enclosures that are divided into territories by sheets of metal mesh. Deluxe territories boast a feeding station and a plastic bin for hiding, while less desirable areas lack the hideout. The mice can easily climb between territories, and males compete with each other up to 50 times a day to defend or win territories. Those who win territory also win mates.

After putting mice from a treatment group and a control group together in the barn, Potts judges the health effects of the treatment by the number of offspring sired by male mice from each group. “If the experimental manipulation reduces health, then in general they’re going to be less able to maintain territories, and so have fewer offspring,” Potts explains.

Results, Reinterpreted

In one case, Potts used an OPA to study the health impacts of sugar consumption (8). Previous studies had only found disease outcomes when lab animals were fed on diets containing far more sugar than humans typically ingest. What, Potts wondered, might be the outcome with more modest amounts? So for 26 weeks, he fed mice a diet that reflected what many Americans actually eat, with a quarter of their calories coming from fructose and glucose. When he subsequently put the mice in his barn-like enclosure, the effect was stark. The death rate of sugar-fed females was almost twice as high as those in the control group who had eaten a sugar-free diet. Meanwhile, sugar-fed males won fewer territories and produced one-quarter fewer offspring than control mice. The results strongly suggest that sugar in these amounts is bad for mammalian health, although the experiment could not explain exactly why.

This barn-based approach can also flag drug risks that traditional preclinical and clinical trials miss. Potts found that the statin drug, cerivastatin (Baycol), which was pulled from the market after being linked to skeletal muscle breakdown, caused males to produce 41% fewer offspring and females to produce 25% fewer offspring compared with control mice (9)—consequences that were not evident when researchers conducted traditional tests in lab animals (10).

But the barn is not foolproof. In a test of the recalled antiinflammatory drug, rofecoxib (Vioxx), Potts’ experimental design was no more sensitive to the increased risk of cardiovascular events than traditional preclinical tests (11). That may be because of sample size: Potts estimates that his study would have needed 9,000 mice to pick up a signal of the rare events.

In another series of studies, Potts used his mouse barn to explore a key question in genetics: Why is it that 10 to 15% of mouse genes can be disrupted without having any noticeable effect on the mouse? One possible explanation is that many genes code for similar functions, so if researchers delete one, another takes over the job. Indeed, Potts’ colleagues at the University of Utah, Mario Capecchi and Petr Tvrdik, now at the University of Virginia in Charlottesville, had found that two genes called Hoxa1 and Hoxb1, which help control brain stem development, were functionally redundant—replacing Hoxb1 with Hoxa1 had no effect on mice in the lab (12). But because they studied lab animals in cages, they could only check for major effects on traits such as locomotion and body size. “The subtle behavior alterations are impossible to pinpoint,” says Tvrdik. So the pair teamed up with Potts and discovered that in the barn males with this genetic manipulation produced 64% as many offspring as controls (13). Replacing Hoxa1 with Hoxb1 had similar effects (14).

“We like to think of the organismal performance assay as a high-throughput initial health screen,” says Potts. “We can detect most insults to health without having a priori knowledge of what the mechanism might be.”

Wayne Potts judges the health consequences of a treatment largely by its effect on how many offspring male mice produce. The mice compete for territories and mates in barn-like structures such as this one at the University of Utah. Image courtesy of Douglas Cornwall (University of Utah, Salt Lake City, UT).

A Spectrum of Wildness

Unlike Potts, Graham studies how traditional strains of inbred lab mice fare in wilder conditions. Aside from her “eye-popping” study on nematode-resistant mice, she also studied a mouse strain that’s commonly used as a model for nematode susceptibility, because it lacks the immune response necessary to expel the parasites. In the field, the susceptible strain and the resistant strain suffered from nematodes equally. “Between 10 and 20 days outdoors made that massive difference between the mouse genotypes pretty much go away,” she says.

Several factors could be responsible for this dramatic effect. Bacteria were already known to play a critical role in helping nematode-egg hatching (15), and “microbial diversity went through the roof” in her outdoor mice, says Graham. Secondly, fighting off a nematode infection requires the right sort of immune response. Unlike those in the lab, outdoor mice infected with nematodes tended to have elevated levels of the T cell Th1, known to be more effective at fighting off bacteria and viruses, rather than the T cell Th2, which is typically better at fighting off worm infections. Regardless of the underlying mechanism, Graham’s findings show that the environment plays a major role in nematode susceptibility—hence findings in the lab may not translate into the clinic.

Other researchers are taking smaller steps toward more natural conditions but still seeing dramatic effects. For example, David Masopust of the University of Minnesota in Minneapolis studied the effects of mixing pet-store mice with lab mice from pathogen-free facilities. Lab mice that survived exposure to their dirty roommates acquired immune systems with a greater abundance and variety of T cell types, more closely resembling the immune systems of adult humans (16).

Pritchett-Corning argues that such results are pushing science toward a paradigm shift in experimental design (17) and that Potts’ OPA approach is a key step in the right direction. “He’s turned the whole thing on its head and looked for translational and actual biological effects, rather than regarding mice as the furry test tube,” she says.

Potts believes it will take a cultural shift for biomedical researchers to consider conducting studies in more natural arenas. One big reason why: Researchers often focus on nailing down the molecular mechanisms underlying a disease, whereas the mouse barn is most useful for showing effects at the level of the whole mouse. Yet Potts contends the approaches are complementary. Once the barn raises a red flag about a treatment, researchers can use traditional techniques to search for the underlying molecular basis. Or, Graham suggests, researchers might also develop treatments in a controlled lab and then test them in a more natural environment.

Of course, there are good reasons why researchers typically stick to traditional laboratory settings. “The essence of research is often being able to control the particular variable that you are studying,” notes neuropharmacologist James Barrett of Drexel University in Philadelphia, who is the editor-in-chief of the Handbook of Experimental Pharmacology. “You can really only do that under highly controlled conditions where you know the strain of the animals, the conditions under which they exist, their diet, the daylight cycle, the humidity in the room—all of those variables are known to exert sometimes subtle and sometimes profound effects on what it is you are studying.” Let go of some of this control, and it becomes more difficult to repeat the experiment and produce the same results, Barrett notes.

But reproducibility, while important, can come at the cost of relevance when studies are conducted under narrow experimental conditions on inbred strains of mice, counters surgeon and medical researcher Ronald Tompkins, founding director of the Center for Surgery, Innovation and Bioengineering at Massachusetts General Hospital in Boston. “The human is way more complex—by genetic diversity, and by environmental diversity,” he notes. In general, he argues that researchers studying inbred mouse populations should offer a caveat in their publications: wilder settings and outbred mice could yield different results.

Tompkins says researchers studying genetically diverse mice under more natural conditions are on the right track. But he contends that attempts to apply mouse findings to humans deserve even more scrutiny. In 2013, he and colleagues presented a cautionary tale, a controversial study suggesting major shortcomings when applying mouse-model genomics to human inflammatory disease (18). “Mice aren’t men,” he says.

Making Better Models

Those researchers who do embark on seminatural experiments face a suite of logistical challenges that their institutions may not be ready to accommodate. “Universities have great animal facilities for the most part, but they are not prepared to build barns for people. It requires a big push on the end of the investigator,” says evolutionary geneticist Beth Dumont of the Jackson Laboratory in Bar Harbor, ME. Space is one of the biggest challenges. “To build a seminatural enclosure, you’re going to take a room that could have held a couple of hundred to 1,000 mice down to 40,” says Pritchett-Corning.

Dumont says that classic inbred strains of mice are also much easier to work with than wild mice or even inbred mouse strains whose ancestors were collected from the wild only about 20 generations back. These classic strains are less jumpy, easier to handle, and reproduce readily in the lab. “Handlers have inadvertently selected for them to be docile,” she says.

It’s also difficult to administer separate treatments to different groups when they are all in the same enclosure. Potts calls the results from his sugar study “conservative” because both his control and sugar-fed mice, once transferred to the barn after their initial feeding, all had the same high-sugar diet, drawing from feeding stations on their own. “We had, and still have, no way to keep the mice on their respective diets when they are actually doing the competition in the barn,” he says.

Looking to investigate the effect of a more natural environment on the brain, Graham’s colleague, Princeton neuroscientist Elizabeth Gould, encountered her own difficulties. Gould, who explores how the adult mammalian brain produces new neurons in response to different stimuli, studied a group of mice she borrowed from Graham’s nematode research. “We were very interested in what kinds of changes in brain growth we would see in animals that were living outside in an enriched environment, with stresses like needing to forage for food and deal with weather,” she says. Gould found that outdoor mice had more new neurons compared with mice kept under standard laboratory conditions, results that she plans to publish soon. But to conduct standard cognitive and anxiety tests on the animals, Gould had to capture them and bring them back into her lab—a stressful episode for the mice that could itself affect the results. “It didn’t make the findings useless, but it’s something that you have to keep in mind,” she says.

Even in the most well-crafted experimental designs, the mice may not experience all the environmental variables that could affect disease. And in many cases, researchers are still exploring which of many natural variables—from microbes and weather to diet and social interactions—actually really matter for a study. “It’s early days,” says Graham. “Maybe in another 5 or 10 years there will be a succinct statement we can make about the important differences between lab and field.”

“We’re not saying that mouse barns are going to solve every problem,” adds Potts. “Just that they will solve many that the conventional methods are missing.”