Self-organized reactivation maintains and reinforces memories despite synaptic turnover

Abstract

Long-term memories are believed to be stored in the synapses of cortical neuronal networks. However, recent experiments report continuous creation and removal of cortical synapses, which raises the question how memories can survive on such a variable substrate. Here, we study the formation and retention of associative memory in a computational model based on Hebbian cell assemblies in the presence of both synaptic and structural plasticity. During rest periods, such as may occur during sleep, the assemblies reactivate spontaneously, reinforcing memories against ongoing synapse removal and replacement. Brief daily reactivations during rest-periods suffice to not only maintain the assemblies, but even strengthen them, and improve pattern completion, consistent with offline memory gains observed experimentally. While the connectivity inside memory representations is strengthened during rest phases, connections in the rest of the network decay and vanish thus reconciling apparently conflicting hypotheses of the influence of sleep on cortical connectivity.

Homeostatic control of synaptic rewiring in recurrent networks induces the formation of stable memory engrams

“…In the case discussed in Section 2.2, in contrast, the memories formed are "silent" (elsewhere classified as "transient" [38]), very different from the persistent activity usually considered in working memory tasks (elsewhere classified as "persistent", or "dynamic" [38]). The latter type of activity seems to be consistent with Hebbian plasticity models [10,30,46], and one may wonder what is the relation of our model with these alternative models.…”

“…Recently, [10] showed that Hebbian structural plasticity could be the force behind memory consolidation through a process of stabilization of connectivity, which is based on the existence of an attractive fixed point in the plastic network structure. In our model, because of the decay of the slow manifold, memories are never permanent, and repeated stimulation is necessary to stabilize them.…”

“…Campbell's theorem allows us compute the mean [35]. As a consequence of the Central Limit Theorem, the amplitude distribution of the calcium signal is approximately Gaussian N (µ Ca , σ 2 Ca ), provided the mean spike rate is much larger than the inverse time constant of the calcium signal r i (t) We are actually interested in the equilibrium rates of free and negative elements in Equation 10. These rates are rectified versions of the shifted calcium trace, and in order to calculate them we resort to the ergodic theorem and to an adiabatic approximation.…”

Drifting Assemblies for Persistent Memory

“…In living animals long-term recordings observed both complete [23] and, more often, incomplete structural remodeling [24]. We will nd drifting assemblies also in absence of remodeling, suggesting that they occur in intermediate cases as well, for example when turnover is less random through weight dependence or incomplete [3,24,25].…”

Optimal synaptic dynamics for memory maintenance in the presence of noise

“…Another hypothesis is that any memory degradation due to fluctuations is counteracted by restorative, systematic plasticity processes that allow circuits to continually 'relearn' stored associations. The source of information for the systematic plasticity term could come from an external reinforcement signal Kappel et al (2018), from interactions with other circuits Acker et al (2018), or spontaneous, network-level reactivation events Fauth and van Rossum (2019). A final possibility is that we rarely observe behavioural states in which an animal is not learning, and that unobserved behavioural changes account for apparent fluctuations in brain connectivity in any given experiment.…”