Intrinsic Spine Dynamics Are Critical for Recurrent Network Learning in Models With and Without Autism Spectrum Disorder

Abstract

It is often assumed that Hebbian synaptic plasticity forms a cell assembly, a mutually interacting group of neurons that encodes memory. However, in recurrently connected networks with pure Hebbian plasticity, cell assemblies typically diverge or fade under ongoing changes of synaptic strength. Previously assumed mechanisms that stabilize cell assemblies do not robustly reproduce the experimentally reported unimodal and long-tailed distribution of synaptic strengths. Here, we show that augmenting Hebbian plasticity with experimentally observed intrinsic spine dynamics can stabilize cell assemblies and reproduce the distribution of synaptic strengths. Moreover, we posit that strong intrinsic spine dynamics impair learning performance. Our theory explains how excessively strong spine dynamics, experimentally observed in several animal models of autism spectrum disorder, impair learning associations in the brain.

Drifting Assemblies for Persistent Memory

“…Previous computational studies addressing the impact of intrinsic synaptic uctuations focused often on the question how neural representations can nevertheless be stable. They proposed that a preserved core structure keeps neuronal activity stable [44,8,45,46], that the networks are retrained to counteract degradation [43,47,48] or that the ensembles of neurons storing memories are kept invariant via reactivations and unsupervised learning [49,18,25]. A few computational studies addressed how representations that change may emerge and partially also how the change's impact may be attenuated.…”

“…For faithful memory storage the ensemble of neurons forming an assembly is assumed to remain the same [12]. Previous theoretical analysis has carefully studied the formation and maintenance of such static neuronal assemblies [13,14,15,16,17,18,19]. In particular, it has been suggested that in presence of noisy, irregular spiking activity [14,20,15,16] and spontaneous synaptic changes [18] assemblies are preserved with the help of activity dependent synaptic plasticity.…”

Computational Roles of Intrinsic Synaptic Dynamics

“…Furthermore, a recent study [96] suggests that intrinsic dynamics, when interacting with extrinsic ones, could maintain memory by introducing bistability in a group of synaptic weights. In a recurrent network model of spiking neurons, their appropriate combination allowed formation and maintenance of cell assemblies (i.e., groups of strongly connected and coactive neurons) that encoded memory.…”

“…Figure 3 Normal and semi-logarithmic plots of spine sizes and their multi-stability. The data is from Figure 3E of the ref [96], in which synaptic weight w is proportional to spine size. imaging, that dendritic spine volumes fluctuated even after the total pharmacological blockade of neural activities in hippocampal slice cultures, and the fluctuations quantitatively account for the statistics of spine volumes using stochastic differential equations.…”

“…Several modeling works have also consistently demonstrated that the combination of intrinsic and extrinsic dynamics can lead to stable maintenance of memory [54,55,93]. In particular, a work by Humble et al [93] hinted that intrinsic dynamics are more than noise to be continually corrected by extrinsic dynamics.…”