Abstract :
[en] Brains comprise well-defined circuits responsible for our behavior. Nonetheless, they are not rigidly wired and are adaptable by experience, a process critical for memory and learning. Neurons adapt their connections with each other thanks to synaptic plasticity.
Simultaneously, brains process incoming information through fluctuations in neuronal rhythmic activities, each defining brain states. Switches in brain states during wake-sleep cycle are described by a neuronal population shift from active to oscillatory state, at the network level. Zooming at the cellular level, it corresponds to a transition from tonic to burst. Neuromodulators orchestrate the switch. They refer to signaling molecules that reversibly change the functional properties of neurons or synapses.
Altogether, memory is impacted during sleep through a phenomenon called sleep-dependent memory consolidation. Experimental results show a down-selection mechanism ie. strong (resp. weak) connections established during wakefulness are preserved (resp. decreased) during sleep. However, its underlying physiological mechanisms remain unclear.
We built a cortical network to study the evolution of synaptic weights during switches in brain states, by using a conductance-based model. The synaptic weights are modified by a synaptic plasticity rule. We have shown that maintaining a synaptic rule parametrized on experimental data acquired in spiking regime leads to a homeostatic reset during burst. All the connectivity strengths converge towards the same basal value meaning that strong weights acquired during learning are converging at the same stable fixed point as weak weights. The reset occurs whatever the rule category, ie. spike-time dependent plasticity rules (such as STDP or triplet) and calcium-based rules [Graupner,2016].
Here, we developed an innovative sleep-dependent plasticity rule. It governs long-lasting change such as the number of AMPA receptors (AMPAr) available. We take advantage of the homeostatic reset. In the beginning of the night, AMPAr changes with a rate proportional to the synaptic weight acquired during the learning phase: strong (resp. weak) weights increase (resp. decrease) the number of AMPAr. Reaching the end of the night, the synaptic weight has converged towards its reset value stabilizing the change in AMPAr. Learning is well transferred, and the network can encode new information. This creative sleep-dependent plasticity rule avoids brain saturation. It is compatible with homeostasis and synaptic drift observed in memory engram.
