Abstract :
[en] Among all essential functions of sleep, research has highlighted its preponderant role in
memory formation and consolidation. At the cellular level, synaptic plasticity, the ability of
neurons to modify their connections, is the cellular mechanism responsible for learning and
memory. Therefore, sleep and synaptic plasticity are two linked concepts in the context of
memory consolidation.
A large amount of experimental and modeling work have extracted mechanisms of synap-
tic plasticity, such as e.g. spike-timing-dependent plasticity, that occur during wakefulness [Bi
and Poo, 1998]. But are they compatible with memory consolidation during sleep [Tononi,
2020; Born, 2016]?
Here, we tackle this question by focusing on the behavior of phenomenological models of
plasticity, i.e. mathematical models based on spike-timing, during sleep. To this end, we
build a 3-cell circuit composed of conductance-based models capable of reproducing faithfully
the neuronal activity switches observed during the sleep-wake cycle [Drion, 2018; Jacquerie,
2021]. Plasticity is taking place between a pre- and a postsynaptic cell and implemented using
well-known plasiticty rules from the literature: the pair-based [Abbott, 2000; Van Rossum,
2000] and triplet models [Pfister, 2006]. We show that those rules, which are successful in
reproducing experimental observation in waking state, are no longer compatible with memory
consolidation after a sleeping period. Interestingly, the outcome is not determined by the
synaptic plasiticity rule itself but rather by the choice of saturation bounds applied to the
synaptic connection. We show that modeling the effect of neuromodulators on the learning
rule during the sleep-wake cycle, for instance by switching to synaptic depression only during
sleep [Gonzalez-Rueda, 2018], permits to reconcile classical synaptic plasticity rules with
sleep-dependent memory consolidation
