Zapping just 20 neurons made mice perform a learned movement, a Nature study shows

Briefly activating about 20 carefully chosen neurons in a mouse’s motor cortex was enough to set off a learned movement, researchers at the University of California San Diego reported this week. The mouse brain holds roughly 70 million neurons, so the trigger amounted to a vanishingly small slice of the tissue. Senior author Takaki Komiyama called that ratio surprising.

Free AI newsletter: The Deep View is the go-to daily newsletter for 600,000+ founders, knowledge workers and professionals who want to stay up to date with artificial intelligence.

Get The Deep View free →

We may earn a commission if you sign up.

The experiment, published on October 14 in the journal Nature, speaks to a long-running question in neuroscience: whether the choreographed bursts of activity seen in the motor cortex during a practiced skill actually command the movement, or merely accompany it.

Holographic optogenetics aimed at early-onset neurons

Komiyama, a professor in the Department of Neurobiology in the School of Biological Sciences and the Department of Neurosciences in the School of Medicine, worked with graduate student Qiyu Chen, the lead author, and a team that included Bin Yu, Soyoung Chae, Zijing Tan, Assaf Ramot and Johnatan Aljadeff. The project was started by An Wu, an assistant project scientist in the Komiyama lab who died in a fire in Montreal in 2023, according to the university’s announcement of the study.

Two techniques made the work possible. Two-photon calcium imaging recorded the activity of hundreds of neurons in fine detail while mice performed a learned task, and holographic optogenetics let the team write light-driven activity back into chosen cells. Chen said the combination made it possible to read the activity of hundreds of neurons and write new activity into that population, and added that it offers a way to observe neurons during a specific behavior as it happens.

Not every neuron in the motor cortex is a good lever. The team targeted what it calls early-onset neurons, cells that become active just before a movement begins, and compared them with neurons that switch on after the movement has started. The earlier group mattered: stimulating the early-onset neurons triggered movement, while the later group did not do the job.

A lever-pressing task and the state of the cortex

An open preprint from the same lab on bioRxiv describes the behavioral setup as a lever-pressing task in mice. Briefly and synchronously stimulating about 20 neurons that had become active before movement onset in voluntary trials produced movements resembling the learned one, and also evoked activity patterns in other motor-cortex neurons that resembled the voluntary pattern. The stimulated cells acted like a starter that pulled a much larger network into its learned sequence.

Success was not guaranteed on any given trial. According to the preprint, trial-to-trial variation in the evoked population activity matched variation in the movements it produced, and the initial state of the population just before stimulation predicted that variation. In some trials the learned activity appeared without any visible movement, which suggests the pattern can be triggered internally. Stimulation also failed when mice were disengaged from the task, so the same 20 cells did not work like a universal switch.

The finding also fits a pattern in Komiyama’s earlier work. A 2014 Nature study led by his group showed that the motor cortex is not a fixed set of switches carrying out instructions from elsewhere. Andrew Peters, the first author, said the animal generates “a whole new set of activity in the motor cortex to make that movement” over the course of learning. A 2025 Simons Foundation report on later Nature work covered the motor thalamus, a pathway that reorganizes as mice master a skill.

From mouse motor cortex to paralysis and stroke research

The practical hope is a long way off but concrete. The primary motor cortex organizes and carries out intentional movements, so a method that identifies a few cells able to ignite a learned action could inform brain-computer interfaces for people recovering from stroke or living with paralysis. The researchers also suggest that understanding how movement-related neurons coordinate could help produce specific movements needed for daily activities.

The announcement goes further and floats a more speculative idea, that imprinting particular brain activity might someday help promote learning and restore memory. The release does not report the number of mice, the number of trials or success rates, so the strength of the effect is best judged from the published paper rather than from the summary. The work was supported by the National Institutes of Health through four R01 grants, by National Science Foundation grant 2024776, and by a Simons Collaboration on the Global Brain Pilot Award.

Next, the lab plans a follow-up study on how the key neurons communicate with one another, the missing piece in explaining why 20 cells can recruit so many others. Until that wiring is mapped, the result stands as evidence that learned movements in the mouse motor cortex are actively generated by the circuit, and that a handful of well-timed cells can start the sequence.

This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.