A study has provided new insights on why sleep is so important for learning and memory – finding that neuronal activity patterns are reorganised during sleep to reflect those seen during memory recollection upon awakening.
Researchers from Professor Jozsef Csicsvari’s group at the Institute of Science and Technology Austria (ISTA) have probed the key roles of sleep stages in optimising memory recollection.
They wirelessly measured neuronal activity patterns in rat brains for up to 20 hours of sleep, considerably extending previously reported measurement times.
“We showed that the neuronal assemblies in the early stages of sleep reflect recently learned spatial memories. However, as sleep progresses, neuronal activity patterns gradually transform into those seen later, when the rats awaken and remember the locations of their food rewards,” said Csicsvari.
Mapping and remembering
Past work showed that a cortical brain area called the hippocampus is important both for exploring and maintaining routes in an environment (called spatial navigation), and for spatial learning.
Hippocampal neurons keep track of the animal’s location by firing at specific locations, thus forming a cognitive map of the environment. Animals use this map to navigate in space while updating it during learning. In this process, the reward locations play an instrumental role, becoming disproportionately represented on the animals’ cognitive map.
Following spatial learning, the hippocampus plays an important role in enhancing memory during sleep. It does so by reactivating recently learned memory traces. Previously, the Csicsvari group showed that the more often a specific reward location is reactivated during sleep, the better the animal remembered that location when they woke up.
On the other hand, when the team blocked the reactivation of a specific reward memory, the animals were unable to recall the respective location.
Reorganising neuronal patterns during sleep engraves memories
While scientists could so far only examine the reactivation of spatial memories in shorter sleep periods of two to four hours, the team now achieved such experiments during long overnight sleep.
Using wireless recordings, they monitored neuronal activity in the hippocampus for up to 20 hours while the rats rested and slept after a spatial learning paradigm.
“Our findings were unexpected. We showed that the activity patterns of neurons linked to the reward locations reorganised during the long sleep,” said ISTA PhD graduate Lars Bollmann, one of the study’s co-first authors.
When a given reward location was reactivated, not all the neurons that represented that location remained active in the entire sleep. While some did, the ISTA researchers called them a “stable subgroup”, others stopped firing during later sleep stages. But at the same time, a new group of neurons started to fire gradually.
“Most surprisingly, we showed that while the pattern of firing neurons in the early stages of sleep echoed the neuronal activity in the learning phase, this pattern later evolved to mirror the neuronal activity when the rats woke up and remembered where the rewards were located,” said Bollmann.
The team not only observed a drift in neuronal activity patterns during sleep in the frame of spatial learning but also linked it to the process of memory reactivation, shedding light on how sleep helps keep memories fresh. In addition, they showed that this reorganisation happens during non-rapid eye movement (non-REM) sleep, while REM sleep counteracts it.
Freeing neurons up for new memories?
What could be the role of this phenomenon, called “representational drift”, that occurs in sleep?
“We can only speculate in this regard,” said Csicsvari.
“It is possible that memory representations must be formed quickly during learning but that such representations are not optimal for long-term storage. Therefore, a process may take place in sleep that optimizes these representations in sleep to reduce brain resources to store a specific memory.”
In support of this hypothesis, the researchers observed that fewer neurons were linked to a given reward location after sleep than before. Therefore, some neurons get freed up to take in newer memories.
“Any new memories must find a way to be integrated into existing knowledge. Frequent repetitions of the new memories as well as partial change in the neuronal code may thus help optimise their integration into existing memory representations,” said Csicsvari.
