As a result, lab mice thrived. But accurately reproducing a human mental illness in the tiny brain of a mouse is still hugely challenging. The basic structure of a mouse brain is mostly analogous to a human brain: They have a hippocampus, we have a hippocampus; they have a prefrontal cortex, we have a prefrontal cortex, albeit one that is much larger. We even share about 99 percent of their genes. But no one would mistake you for a mouse. The mouse is a nocturnal animal with poor eyesight, adapted to fear predators that strike from above. Mice are fundamentally alarmed by light, open spaces, and sudden movements. It is no surprise, then, that they manifest depression and anxiety differently than humans do, if they manifest such ailments at all. “You cannot mimic an entire human psyche in a mouse or a rat,” says 
Jacqueline Crawley, a behavioral neuroscientist at the National Institutes of Health (NIH) and the author of the seminal mouse modeler’s handbook, What’s Wrong With My Mouse? “We can model only one behavior at a time,” she says. “Mice aren’t a one-to-one correspondence to humans. But they are better than zero.”

Measuring a mouse’s behavioral traits, whether healthy or unhealthy, is a delicate and error-prone task. Each mouse strain captures only a snippet of the human picture. Depending on which element of personality is under study, a mouse handler puts mice through mazes to test memory, “stoplight” tasks to test attention, bar-pressing games to test impulsivity, and so on.

With a newly engineered mouse, you might begin with a broad battery of tests to compile an overview of its traits: low anxiety, average activity, high impulsivity, normal memory. Then you would focus on the specific feature of interest. If you are studying the human link between impulsivity and addiction, you might try another test of impulsivity: The mouse gets a treat for pressing a bar when a light turns green—unless it immediately turns red. Will he be able to stop himself on the red trials? No? Excellent. Time to offer him a swig of sugar water. If he guzzles that, he may still be a good model of an addiction-prone personality.

Sometimes the same batch of 15 mice can perform a series of different tests. But each test must be specially sequenced so that the stress of one test will not significantly influence performance on the next. In the end, dozens of mice might be needed to accurately summarize one strain’s personality. If this process seems terribly cumbersome and time-consuming, it is. Creating a useful mouse model of disease can take years. And it is all conducted in a hushed clean room, wearing paper clothes, with no food or beverages allowed. A careful researcher would not let a flustering stranger watch.

In a lab at the McGovern Institute for Brain Research at MIT, just a few black mice are brought forth for a demonstration. The lab’s resident mouse whisperer, newly minted neuroscientist Kimberly Maguschak, will run some tests she uses in her quest to build a better mouse model of anxiety and depression. She is masked and gowned like a surgeon. The mice are fairly agitated, rearing and sniffing the air. “Any tiny thing can alter their behavior,” Maguschak says. “I pick these mice up every day so that they’re used to me. But the night before you do an experiment, you don’t clean the cages. You don’t want them all anxious and running around. You don’t suddenly change your shampoo, and you never wear perfume.”

She smoothly catches a mouse’s tail and lifts him from the cage to a giant test tube filled with water. This is the six-minute, forced swim test, a standard measure of an animal’s willingness to struggle against a crummy environment.

The mouse scrabbles at the sides of the cylinder for a few seconds and performs the traditional perimeter survey, searching for a way out. From a mouse-eye view, he is trying to escape a threatening environment. Or not. “That’s odd,” Maguschak says. Her mouse has quit paddling; instead he is floating like a cork. Mice are excellent swimmers, and even better floaters. A few mouse turds settle to the bottom—a sign of stress. Something is wrong. Even a mouse engineered to exhibit behavioral despair would swim more “optimistically” than this. Then again, no mouse handler would expect normal behavior in such a bright, noisy room. Maguschak’s actual search for the source of human misery is conducted in silence, behind locked doors, and only familiar humans attend.

In the dark ages before genetically engineered mice, one of the more amusing traditions in drug testing was for a researcher to swallow a dose and prepare to take notes. Another approach was to give a test rat or mouse one drug to make it crazy—anxious, depressed, hyperactive—and then test the candidate drug quickly before the first one wore off. The goal was to see if your drug solved a problem, without gruesome side effects, before you took it to that final, fretful step: human trials.

The modern mouse model is somewhat more precise, owing to knockout technology. The power of the knockout is particularly evident with mental disorders in which a single gene has a large, obvious effect. The autistic-like mouse is a good example. In 2007 a French team discovered a disrupted gene called SHANK3 in a small subset of people with autism. Maguschak’s boss at the McGovern Institute, Guoping Feng, promptly created a mouse that also had a disrupted version of SHANK3. Now he is exploring the mouse’s brain for clues as to how this particular type of autism disrupts behavior, and how to treat it.

A video filmed at the McGovern Institute shows how dramatically the damaged SHANK3 gene can alter an animal’s behavior. Two mice sit in a cage. One was bred with Feng’s altered gene. They look like average mice but their actions give them away. The unaltered mouse sits motionless against the wall, instinctively avoiding open spaces where winged predators could strike from above. That’s typical behavior. The mouse with a defective SHANK3 gene, by contrast, hugs the opposite wall, and he bends to a repetitive task: He paws mechanically at a spot on the floor.

Is this mouse really autistic? His symptoms may come closer to mimicking the human disease than most mouse models of mental illness, because the gene involved has such a powerful effect. That is unusual. Disorders like depression and schizophrenia are each linked to hundreds of genes. No one gene is likely to make much difference.

But genes are only one part of the story. Other clues to human mental health can be found in the neural circuits of mouse brains. By tracing the wiring that connects one brain region to the next, researchers hope to develop more precisely targeted medications. Many vintage psychiatric drugs, such as Valium, Ritalin, and antipsychotics, were stumbled upon rather than tailor-made to solve a problem. As a result, they are too broad: They affect more than one type of receptor, on more than one kind of nerve cell, in more than one part of the brain. Many patients decide the cure is not worth the many side effects.

A few corridors away from Chesler’s office at Jackson Lab is neuroscientist Zhong-wei Zhang, a man on the hunt for the impaired circuits that might give rise to autism. He wants to know what causes social messages to stall in an animal’s brain. On his bench, 50 pounds of microscope magnifies a translucent shaving of mouse brain. A hair of a diode feeds electrical pulses into one side of a single cell; an electrode on the other side records that cell’s response. If it is shorting out like a bad lamp, Zhang could add a drug to the liquid in which the brain slice floats. Better now? Worse? “The brain as tissue is very normal—it’s like a piece of tofu,” Zhang says. “But the complexity is because it has such a large number of components interacting.”

Those interactions are his key interest. Even when each component of a brain works well, if the connections between regions are missing a crucial protein or chemical messenger, autism could result. Under the giant microscope, Zhang tests the connections like an electrician, navigating by a spiral-bound atlas of the mouse brain. Each slice in the atlas is a Swiss cheese of discrete little regions. And mouse brains are relatively simple. The human brain is vastly more complex, not to mention far less accessible to scalpel-wielding researchers.

With mice, Zhang never has to worry about supply running low. He can always order more brains. Mice are champion breeders, capable of producing 10 to 15 offspring a litter and about one litter every month. There is just one snag: Many psychiatric diseases in humans may well result from circuitry found only in humans.

Mice may be the best models we have of psychiatric disorders, but best does not mean great, or even decent. Gerald Dawson, founder and chief scientific officer of P1Vital, a pharmaceutical consulting firm in the United Kingdom, had his heart broken by the mouse mismatch. In the late 1990s, when he was working at a British division of the pharmaceutical giant Merck, mice ruled the world of drug discovery. You would create a mouse model of attention deficit/hyperactivity disorder (ADHD) or depression and dose the rodent with molecules carefully designed to close one cell receptor or open another.

So when Dawson set out to eliminate the drowsiness from anxiety drugs, he naturally turned to mice. The class of drugs he wanted 
to modify, benzodiazepines such as Valium, Xanax, Ativan, and Klonopin, target the GABAa system. To grossly simplify, that system’s mandate is to put the brakes on nerve firing: It slows things down. So GABAa drugs help address problems like anxiety attacks and seizures. As with most neurotransmitters, the GABAa system is so evolutionarily ancient that it has diversified to serve many purposes. Hence the brain has six different GABAa receptor types, presumably to perform six different jobs. Dawson had a suspicion that the sleepiness side effect originated from just one of those six receptors. If he could determine which one, corporate chemists could design a molecule that would avoid activating it. He began to make mice.

One by one, he manipulated the receptor genes, breeding a new line of mice each time. With each new strain, he would administer the equivalent of a tiny Valium. If the animals grew drowsy, he knew he had not yet knocked out the right receptor. Knocking out receptor 1 made little difference. Receptor 3 proved too hard to knock out. Receptor 5 seemed to account for the amnesia that people (and mice) experience when they take anxiety drugs. Targeting receptor 2, Dawson identified a chemical that reduced a mouse’s startle response—a measure of anxiety—without impairing its ability to balance atop a rotating rod. Success!

Or so he thought. “When these compounds went into humans, they turned out to be just as sedating as the original drugs,” Dawson sighs. “It happens very rarely that a researcher gets to go through the whole process with a chemical, from mouse to man.” Normally the many steps are farmed out hither and yon, and no one feels the pride of parenting a new drug. He came so close. He got to test the drug on people. “And they fell asleep.”

Dawson blames the mice. “There’s not enough predictability in animal research. A lot of pharma companies are getting out.” Dawson moved on too. His company offers drug companies a new animal model for testing their drugs: humans. He will assemble a group of people to voluntarily try an existing drug for a new application before a pharmaceutical company embarks on a bigger, more costly human trial.

In a way, his shift hearkens to the backroom reality of drug testing. “I’ve tried lots of things: scopolamine, benzodiazepines, antipsychotics,” Dawson says. “We take lots of existing drugs too, to see if they have other applications. That’s very common.” He chuckles at the blockbuster success of the anti-narcoleptic drug modafinil, better known as Provigil. It is now widely taken off-label to boost alertness and acuity. “Modafinil has $2.4 billion in sales a year,” Dawson says. “There is not that much narcolepsy around.” The journal Nature reported recently that modafinil and other “cognitive enhancers” are in particular demand on the “ivory market” of academia.

But for all Dawson’s frustration with mice, the rodents did yield a couple of interesting drug leads. That receptor 5 implicated in the amnesia side effect? An experimental chemical that blocked its action created temporary geniuses: Mice on it were whizzes in the Morris water maze. A drug company is testing the compound to treat people with Down syndrome. And in the process of trying to eliminate drowsiness, Dawson and his team homed in on one of the chemical switches that cause mammals to go to sleep. Ambien locks onto that switch, associated with receptor 1, and sends you off to slumber.

Today, mice are the undisputed top animal for research on mental health therapies. The Foundation of Biomedical Research estimates that U.S. scientists use between 20 million and 30 million mice a year. Jackson Laboratory alone distributes 2.9 million to 19,000 investigators in more than 800 institutions in at least 50 countries. But the rodent’s reign is under threat. Many neuroscientists are turning back to the long-ignored rats. Last year researchers finally found a way to engineer a knockout rat, breeding a strain free of a gene that controls breast cancer in humans. Within one year a major supplier was offering two different rat models of schizophrenia, two of Alzheimer’s, and six of Parkinson’s disease—in addition to cancer and other disease models. Some think the mouse’s day is done.

Loyalists like Elissa Chesler are standing by their mice. They blame the tests, which subject mice to very unnatural tasks, such as pressing a bar for food. What’s more, many of the standard protocols for assessing behavior in rodents were developed for rats and were merely adjusted to accommodate mice. If the mouse-to-human translation is breaking down, perhaps it’s because we’re expecting mice to speak rat. “The hobbies and interests of rats are different from the hobbies and interests of mice,” Chesler says, with a hint of exasperation. “A lot of experiments depend on an animal pressing a bar. That is not what mice naturally spend their time on. That is not a hobby of theirs.”

To that end, Howard Eichenbaum, a Boston University neurobiologist, is tailoring personality tests specifically to mice. Doubting that mice are truly less intelligent than rats, he builds experiments that capitalize on a mouse’s strengths. Mice are nose driven, so instead of asking them to memorize mazes, he asks them to memorize odors. And mice love to dig, so he incorporates that instead of bar pressing.

By mixing kitchen spices with playground sand, then burying a treat in some spices but not others, Eichenbaum created a mouse IQ test he calls Rock, Paper, Scissors (but that could more logically be called Pepper, Paprika, Ginger). The reason that even adult humans pause during that game is that the rules slide: Logic holds that if Rock beats Scissors, and Scissors beats Paper, then Rock should also beat Paper. It takes time to learn the real rules. In the rodent version of the game, mice learn that the scent of pepper beats paprika, paprika beats ginger, and ginger beats pepper, all in a single trial. And they remembered it a week later. “I couldn’t remember the rule a week later,” Eichenbaum says. “Working with olfaction, we found out they’re even smarter than rats in some ways.” Now he has a stress-free way to study the physiology of memory.

At NIH, Jacqueline Crawley is investigating another novel way to read a mouse’s state of mind. Mice can chatter in voices pitched above human hearing. She is tuning in to that stream of 
microsqueaks to tease out differences between healthy mice and those made to model autism. The challenge is that no matter how well we get to know mice, their problems will always be at best a very rough analogue of our own.

“We can get closer to monitoring mice in a less artificial way,” Chesler says. “But mice aren’t laboratory reagents. They are beings.” Beings with stressful lives and fragile brains. Beings surprisingly like us, yes, but also beings with minds of their own.

Hannah Holmes has stepped on dinosaur eggshells in Mongolia and dived a mile and a half below the Pacific Ocean. Her latest book is Quirk: Brain Science Makes Sense of Your Peculiar Personality.

The Wrong Diagnosis?


A mouse model of schizophrenia casts doubt on psychiatric definitions.


Neuroscientist Mark Geyer is a schizophrenia man by trade, always on the lookout for mouse strains to model the reality-twisting disease. So he was intrigued when a colleague at the University of California, San Diego, noticed mutated mice making odd-looking nests. These mice weren’t literally schizophrenic, but they were undoubtedly off-kilter, and their lack of interest in nest building mimicked the social apathy of human schizophrenics. Thinking they may have hit upon a useful gene that could reveal disease pathways, his group searched to see if the mouse mutation could also be found in schizophrenic humans. It was. But somewhat surprisingly, it was strongly associated with bipolar disorder as well.

This double whammy from a single gene highlights a flawed but dominant view of psychiatric disease. From autism to schizophrenia to depression, psychiatrists tend to see mental disorders in black and white: Either you meet the diagnosis or you do not. “We force disorders into a dichotomy based on the necessity of treatment,” says UCSD psychiatrist John Kelsoe. But if a single gene can underlie two diagnostic categories, then perhaps the behavior (lack of interest in nest building, for instance) and the underlying biochemical anomalies should matter most.

That, at least, is the paradigm shift coming to psychiatry in the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders, or DSM-5, to be published in 2013. After long, heated deliberations, the manual is expected to change its psychiatric definitions to acknowledge the full spectrum of manifestations of symptoms such as mania, which might range from normal bouts of hyperproductivity to full-out disabling manic episodes.

Psychiatrists still are not at the point at which they can diagnose patients by analyzing their brain chemistry, but the field is moving in that direction. A forward-looking initiative called the Research Domain Criteria project, financed by the National Institute of Mental Health, promotes clinical research to match biochemical and genetic indicators of mental illness with treatments. “Increasingly we realized that animal research didn’t connect well to our clinical research,” says Bruce Cuthbert, head of the initiative. The solution, however, isn’t simply finding better mouse models; it also involves redefining diseases to include the physiological underpinnings that mice illuminate so well.  —Katie Palmer