Lab mice train themselves

Researchers at the Riken Brain Science Institute in Japan say they have a way to fully automate the training of mice used in behavioral neuroscience experiments. 

Their new system reduces the time it takes to train a single lab mouse from 15 hours to just 90 minutes – and doesn’t require human input.

Reporting in this week’s Nature Communications, a team led by Andrea Benucci describes an around-the-clock system in which mice put themselves through behavioural training tasks at will, without human intervention. Four or more mice per day are used, and the system has already been used to safely train 100.

Recording neural activity during animal behaviour is a cornerstone of modern brain research. But before data can be collected — in studies of vision or cognition, for example — the animals must be trained to do specific tasks, such as pushing a button to indicate a preference or demonstrate a memory. 

This training can take months and is a full-time job for one or more people. The animals can become stressed from being handled by experimenters. And because training techniques and experiment procedures vary from lab to lab, the quality of information being gathered suffers. 

“It is hard to compare data across labs and even within the same lab, and we waste a lot of time getting comparatively little data,” Benucci says.

In the new system, mice are trained to voluntarily place their heads into a restraining device in which they can be tested and examined. They enter the apparatus to receive liquid rewards for doing tasks that test various audio or visual responses. They rotate a small wheel with their front paws to indicate a decision — for example, whether or not they can hear a tone.

Benucci says the training apparatus is also equipped with a two-photon microscope, to allow researchers to examine the brains of trained mice at single-cell resolution engaging in complex behavioral tasks.

He adds that because mice learn to self-direct and become familiar with the system, and it is modular, the experimental possibilities extend beyond studying mouse behavior to real-time brain imaging and physiology. 

“Normally we see a decline in mouse performance or other incompatibilities when moving from highly trained behaviors to different types of experiments for brain recordings, but that doesn’t happen with our system,” he says. 

Please login to favourite this article.