Mechanisms of Memory Triage during Sleep

At first glance, sleep appears to be a maladaptive step of evolution. Sleep imposes a prolonged period of "inactivity", reducing the time an animal can spend on foraging, hunting, and finding/attracting mates. Furthermore, sleep is substantially more dangerous than staying awake, as the very sensory inputs protecting us from predation are suppressed. So, what then do we get in return? The short answer is that we do not know. It is clear that evolution has made sleep a biological requirement. Your brain will not let you voluntarily avoid sleeping more than a few nights, and involuntary sleep deprivation leads to chronic stress, fatigue, disorientation, poor health and even death. However, in spite of decades of research, we still do not fully understand why we have this biological need to sleep.

Sleep likely serves a multitude of roles benefiting from a prolonged "offline" state with reduced sensory inputs. One major hypothesized function is systems-level memory consolidation, a process by which new memories stored in the hippocampus are stabilized and transferred to neocortical brain regions for long-term storage. One key aspect of memory consolidation is the phenomenon of memory triage, by which salient or repeated experiences are more likely to be consolidated compared to irrelevant experiences. How the brain accomplishes this selective strengthening of important memories is unknown, and the focus of our research proposal.

Sequential patterns of neural activity occurring during an experience have been observed to spontaneously reactivate during sleep. This reactivated memory trace, referred to as replay, is postulated to be how the brain consolidates a memory, with the transfer of a memory residing in the hippocampus to other brain regions. We will examine how replay is modulated by an experience's salience, and hypothesize that more important experiences will replay more during sleep. Furthermore, we predict that cortical feedback plays a key role in modulating replay based on salience. To test this, we will inactivate cortical inputs to the hippocampus and examine whether the hippocampus no longer biases replay towards more salient experiences.

Approximately 70% of Brits sleep less than the 7-9 hours per night recommend by the medical community. Furthermore, increased social media use is correlated with poorer sleep quality, along with an increased occurrence of anxiety and depression. As the amount of "junk information" in our lives increases, and the amount we sleep decreases, there is a growing pressure on our memory triage system to effectively prioritize the consolidation of salient information, and an increased need to understand how this occurs. These issues are exasperated further in older individuals, as both sleep quality and duration decline with age. Our research will directly examine the underlying mechanisms of memory triage, and provide new ideas on how this process could be further optimised.

The hippocampus plays a critical role in encoding new episodic memories. However, after an episodic memory is initially stored, it undergoes systems-level memory consolidation, a sleep dependent process that stabilizes the memory and modifies its storage within the brain such that it is less reliant on the hippocampus for recall. When we sleep, our brain must prioritize which new memories are consolidated, giving a lower priority to behaviourally irrelevant memories or older memories that have already been consolidated. Data in humans clearly show that at least two main factors impact memory strength: repetition and salience. We will examine the role of salience and repetition in modulating hippocampal replay activity during sleep. Sleep replay, a phenomenon where neural sequences previously evoked by a behavioral episode are spontaneously reactivated during sleep, is postulated to be a mechanism responsible for memory consolidation. We hypothesize that more replay will occur during sleep for experiences that have been repeated or associated with a high value reward. Data in rodents also suggest that spontaneous cortical activity prior to replay, occurring during up-states during non-REM sleep, can predict which memory replays. We hypothesize that this is a form of cortical feedback, and is responsible for the memory triage-based prioritization of salient memories during consolidation. Our prediction is that suppression of medial entorhinal cortex, which will block a major cortical input to the hippocampus, will disrupt this bias. The consequence of this is that although we will still observe replay events, but will not see a difference in replay activity between high salience and low salience memories.

Academic impact- Despite its importance, the function of sleep remains an open question in science, and has garnered interest in a diverse number of scientific disciplines including molecular biology, evolutionary biology, clinical psychiatry, and neuroscience. We will attempt to reach this broad array of academic audiences through the dissemination of our work through open-access, peer-reviewed publications, scientific conferences, invited talks, and collaborations with other research labs. In addition, we plan to provide open access to our electrophysiological and behavioural data and analysis pipeline (using CNRS.org and Github), upon completion of our project, to facilitate this type of research in other labs, and to hopefully stimulate future collaborations. Our data set will be especially useful to computational neuroscientists modelling memory and reinforcement learning, and developing deep-learning neural networks.


Economic/technological impact-Our project relies heavily on a number of open-source tools, that are both substantially cheaper than commercially available alternatives, and provide greater flexibility for customization. The main limitation in these techniques being used is the lack of professional support. To help make these tools more widely available and useable across the UK, we plan to hold a small 3-day workshop at the end of the grant, where we will teach the main open-source methods in the lab. We have requested £2000 in funds to provide support for this workshop and additional support will be obtained from UCL. Using customized, open source methods provides more value for money in future RCUK grants, and provide a larger userbase to facilitate both any required troubleshooting as well as further development.


Societal impact- Approximately 70% of Brits sleep less than the 7-9 hours per night recommend by the medical community. These issues are exasperated further in older individuals, as both sleep quality and duration decline with age. To encourage individuals to sleep the recommended amount, it is important to convey what the function of sleep is, and the consequences of not sleeping enough. Through our public engagement, our research will help emphasize the importance of sleep, and promote better sleeping practices which has widespread benefits for societal health and wellbeing.