Inhibitory synapses as a cellular substrate for long-term memory storage
Information to be preserved over long time periods (weeks to years) is thought to be encoded by synaptic modifications in the cortex. As cortical networks are largely excitatory, learning and memory have been traditionally associated with the plasticity of the excitatory synapses. Consistently with this notion, experiments have exposed significant re-organization of the excitatory connectivity in cortical networks upon learning. Experiments, however, also show that (i) excitatory synapses exhibit substantial volatility in the absence of learning, and (ii) their strength is only partly determined by neuronal activity. These observations challenge the notion that excitatory synapses support long-term storage.
Here, we quantified spontaneous synaptic re-organization by chronically imaging thousands of dendritic spines in the mouse auditory cortex, and investigated its consequences in a biologically-constrained model of cortical network. We show that the ongoing pattern of firing rates in the network is primarily determined by the inhibitory sub-network, despite the fact that the majority of neurons and of synapses are excitatory. This is a direct consequence of the differences in the distributions of the firing rates of excitatory and inhibitory neurons. Thus, patterns of firing rates are robust against the substantial remodeling of the excitatory connectivity, which preserves the overall distribution of connections. By contrast, transient changes in the distribution of excitatory connections, associated with learning-induced plasticity, have a considerable effect on the patterns of network activity. Next, we investigated long-term storage performances in the model network when information is encoded either in the excitatory or in the inhibitory connectivity. We show that the storage capacity of the inhibitory sub-network is significantly larger than that of the excitatory sub-network, and that the information stored in the inhibitory connectivity is robust against the levels of experimentally-observed volatility in the excitatory synapses.
Taken together, these results suggest that the excitatory sub-network is responsible for the rapid, transient encoding of the information, whereas the inhibitory sub-network is responsible for its long-term storage.