A Small, Deep Brain Structure Helps Mice Hold a Thought

A mouse gets one whiff of an odor, or one puff of air across its whiskers, and then nothing, for as long as a minute. A thin sheet of neurons called the claustrum, buried deep in the brain, helps mice hold a sensory cue in mind, because turning its activity down impaired their memory of the cue. Then a second cue arrives and the mouse has to say whether it matches the first, with a sip of water riding on the answer. Getting that right takes working memory, a job nobody has pinned on a single brain region.

Working memory is the brain’s scratch pad, the stretch of a few seconds (tens of seconds, in these mice) during which you hold a phone number or the start of a sentence after it has gone, long enough to compare it with whatever comes next. It leans on the prefrontal cortex, the thalamus, the sensory areas and more besides, a sprawling committee in which no one can say exactly who does the holding.

Alan Carleton, Ivan Rodriguez and colleagues at the University of Geneva worked with 61 mice on two versions of that test, one using odors and one using whisker puffs. Their peer-reviewed report in Nature Communications describes waits lasting from 5 seconds to a full minute. The mice’s edge over guessing faded smoothly with the wait. It halved about every 20 seconds by the median, or 30 by the mean, whether the cue was a smell or a touch.

The claustrum is an odd suspect: a sliver of tissue lodged deep in the brain, thin enough to miss. “Its neurons are connected to almost all regions of the cortex, which puts it in a good position to receive and influence information from very different networks,” says Carleton.

A Trace That Outlasts the Wait

The team threaded a miniature microscope into 7 mice and tracked calcium signals, a proxy for firing, in 383 claustrum neurons during the smell task and 424 during the touch task. About 25% of the neurons fired up when the cue arrived, about 60% were busiest during the wait, and the rest were suppressed for much of the trial. A classifier (a program that learns to sort patterns) could name the original cue from that activity out to 30 seconds, and did much worse on trials the mice got wrong.

“Our live recordings of claustrum neuron activity showed that a trace of the initial information persisted throughout the entire waiting period,” says Rodriguez, a professor of genetics and evolution at Geneva. “Far from being a static memory, neuronal activity changed over time. Nevertheless, the identity of the stimulus remained sufficiently present to be decoded up to 30 seconds later.”

Turning the Claustrum Down

Activity that tracks a memory could still be a bystander, so the team turned the claustrum down. A drug-activated receptor (chemogenetics) let them dial neurons down for a session, cutting performance by about 40% in groups of 13 smell-task and 11 touch-task mice, while control mice given the same drug were unaffected. Light-sensitive proteins (optogenetics) let them go further and silence neurons at chosen moments, which cut performance by about 40 to 60%, and by up to about 80% in individual mice. The timing did not matter: light during the cue, the early wait, the late wait or the target hurt about equally, for smell and touch alike. Light on mice carrying only a fluorescent marker did nothing.

Perhaps the mice simply could not smell or feel the cue? They could: in separate tests of telling two odors, or two touches, apart, the same light left scores untouched, and licking, pupil size and body movement showed no detectable change.

Other Regions Carry Part of the Burden

Even so, the damage was partial, and the university’s release, which says performance was halved, sits inside that range but flattens it: about 40% with the drug method, 40 to 60% with light. The tools reached at most about 50% of the claustrum and some memory survived, so other regions evidently carry part of the load.

A separate team, working with freely moving mice in a T-maze, found that silencing a related set of neurons in and around the claustrum impaired memory during the wait but improved performance at decision time. The Geneva team suggests the clash may come from the task design or from exactly which neurons were hit, and the question is open. All of it happened in head-fixed mice answering two kinds of cue, a long way from a person juggling a sentence. The release points to attention deficit hyperactivity disorder and schizophrenia, where working memory runs weak, but nothing in these experiments tested either.

The next job, says Carleton, is to work out “how the claustrum communicates with cortical networks, and how this interaction preserves the fidelity of the representations held in working memory.” Until then, the authors’ working guess is that it helps steady activity across the cortex rather than storing anything itself.

FAQ