Hidden signals may hold key to mechanism of memory

Rice engineer Caleb Kemere wins prestigious NIH grant to study how brain stores data

What are our brains up to when we sleep?

Rice University electrical and computer engineer Caleb Kemere will seek answers to that question with the help of a National Institutes of Health grant. A five-year R01 award — the most prestigious offered by the agency — will let Kemere and his colleagues advance a long-running study of how heretofore hidden signals in the brain facilitate the long-term storage of memories.

“Everybody sleeps, and everybody has memories,” he said. “So the connection between those two things is relevant to all of us.”

Caleb Kemere

The award to Kemere, an associate professor at Rice’s Brown School of Engineering, and co-principal investigator Kamran Diba, an associate professor at the University of Michigan Medical School’s Neural Circuits and Memory Lab, is for $1,700,000.

The collaborators reported last year they had found waves of firing neurons in the hippocampus and beyond when animals were active in a maze, moving on a track or just resting. Hidden Markov models, commonly used to study sequential patterns, helped show that data captured during rest appeared to represent a replay of “active” memories as they are being encoded.

The statistical models helped them recognize when resting signals represented the memory of an environment, even when the experiments were “unsupervised”; that is, none of the data directly correlated brain activity with physical activity.

Those experiments incorporated brief periods when the animals simply paused. In the next stage, funded through NIH’s National Institute of Neurological Disorders and Stroke, the researchers expect to go deeper to evaluate the role sleep plays in reorganizing information in the brain.

“Learning is the process in which you experience something and you store it; you build a model from what you learn,” Kemere said. “You can call these models ‘memory.’ There’s processing during the initial experience in the hippocampus, where there’s rapid storage. Then during subsequent periods, when you’re not actively getting new experience, when you’re awake or asleep, the patterns are reactivated.

“It seems like those reactivations are one of the primary mechanisms for this model building, or memory formation,” he said. “Reactivated patterns then lead to long-lasting changes in how the information is stored.”

The researchers want to identify critical time windows and neuronal activities during sleep that are important for storing and stabilizing information. To do so, Diba’s lab is monitoring rodents for 24 to 48 hours during active and sleep periods, gathering a mass of data that has to be parsed to sort out noise and give order to what they believe are the “true” signals involved in storing memories.

“We started out with the observation that noise increases during sleep,” he said. “Why is that? Are the true things are getting noisier? Are they being distributed across multiple environments? Or is the mapping between true things and neural activity getting noisier?”

He said the new round of experiments expose rodents to two novel environments before letting them sleep, instead of one. “This allows us to look at sleep patterns and see how the memories of these two different places interact. Do they merge? Do they stay separate? Do they separate in time, in discreet little chunks? We just don’t know yet.”

The researchers see great value in a better understanding of how sleep impacts lives. They noted in their proposal that “assumptions and deductions about the nature and purpose of sleep implicitly inform all manner of public policy, from the durations of shifts for hospital and relief workers, to morning start times of public schools.”

“We already know a lot about how important sleep is for memory and its likely impact on what happens when we learn multiple things at the same time,” Kemere said. “We know there can be challenges there; the memories will mix with each other.

“Understanding how that happens during sleep may give us some insight into how to teach multiple new concepts,” he said. “I can imagine a scenario where we would learn how best to structure the presentation of novel information.”