Inhibitory engrams in learning and memory consolidation

The human brain is the most complex system in the known universe. Yet despite this complexity, and in contrast to other systems such as cars or computers, the brain does not need to be regularly serviced. Instead, the brain has a remarkable ability to undergo modification without compromising previously learned information. This reflects a finely tuned trade-off between plasticity and stability. Arguably this trade-off ensures we have the flexibility to meet the demands of an ever-changing environment (plasticity) while also protecting memories from interference (stability).

This trade-off between plasticity and stability is thought to be set by homeostatic mechanisms. However, the details of these mechanisms remain poorly understood, particularly in relation to behavioural read-outs of learning. This is in part reflected in the performance of artificial neural networks, which typically forget past learning if trained on new tasks, giving rise to 'catastrophic forgetting' which has emerged as one of the main challenges facing artificial intelligence.

Previous studies in both animals, humans and computational models suggest stability within the brain is restored after new learning by establishing a balance between excitatory and inhibitory activity. Specifically, while new learning is thought to first induce plasticity at excitatory connections, this leads to an increase in overall activity which must later be stabilized by matched changes in inhibitory connections. Here, we will investigate this homeostatic mechanism in the human brain. In our experiments, volunteers will acquire new memories by learning sets of associations between pictures and symbols. We will then measure changes in the inhibitory component of a memory using neuroimaging techniques that involve non-invasive Magnetic Resonance Imaging.

First, we will investigate mechanisms that control the formation of matched inhibitory connections after new learning. Specifically, we will ask how brain activity during rest after a learning session serves to build matched inhibitory connections. Second, we will investigate the circumstances under which this homeostatic mechanism is disrupted, leading to instability and disturbance in memory. To this end, we will use a single-dose of a non-harmful drug to mimic natural changes in our brain chemistry under stress. Third, we will assess the adaptive advantage associated with having a transient window of memory instability immediately after new learning. We will test whether we can promote generalization of shared features across different memories when this window of instability is prolonged.

Together these studies will reveal mechanistic insight into how the human brain regulates a finely tuned trade-off between plasticity and stability. In doing so, these studies will provide an important basis from which to establish how memory distortions arise in psychological and neurological disorders.

Nervous systems face a fundamental problem: how do neural circuits undergo modification to meet the demands of an ever-changing environment, while not compromising previously learned information? In other words, the brain must set a trade-off between plasticity and stability that ensures new learning does not interfere with memories from the past. This trade-off is thought to be set by homeostatic mechanisms. However, the details of these mechanisms remain poorly understood in relation to behavioural read-outs of learning. A recent hypothesis proposes that following changes in cortical excitatory synaptic strength, homeostasis is maintained by matched increases in inhibitory synaptic strength, otherwise termed 'inhibitory engrams'. These inhibitory engrams may help preserve neural coding schemes that rely upon a tight coupling between excitatory and inhibitory transmission.

Here, across Work Packages 1-3, we will investigate how inhibitory engrams form, together with their effect on behavioural readouts of learning. In our past research, we have developed techniques to index the formation of excitatory and inhibitory engrams in healthy human volunteers, using non-invasive functional Magnetic Resonance Imaging, combined with non-invasive brain stimulation and Magnetic Resonance Spectroscopy. We will use these techniques to measure the strength of inhibitory engrams after new learning. In Work Package 1, we will investigate the mechanism that drives formation of inhibitory engrams, with a focus on the role of offline memory reactivation. In Work Package 2 and 3, we will then disturb and delay the formation of inhibitory engrams, to prolong the period of neural instability after new learning. This will be achieved using two double-blind placebo-controlled drug studies designed to manipulate cortical GABA. Following drug intake, in Work Package 2 we will measure the effect on memory interference, while in Work Package 3 we will assess the effect on generalization.