Structural plasticity of circuits in cortical neuropil

Learning and memory formation in the brain depend on the plasticity of neural circuits. In the adult and developing cerebral cortex, this plasticity can result from the formation and elimination of dendritic spines. New synaptic contacts appear in the neuropil where the gaps between axonal and dendritic branches can be bridged by dendritic spines. Such sites are termed potential synapses. Here, we describe a theoretical framework for the analysis of spine remodeling plasticity. We provide a quantitative description of two models of spine remodeling in which the presence of a bouton is either required or not for the formation of a new synapse. We derive expressions for the density of potential synapses in the neuropil, the connectivity fraction, which is the ratio of actual to potential synapses, and the number of structurally different circuits attainable with spine remodeling. We calculate these parameters in mouse occipital cortex, rat CA1, monkey V1, and human temporal cortex. We find that, on average, a dendritic spine can choose among 4-7 potential targets in rodents and 10-20 potential targets in primates. The potential of neuropil for structural circuit remodeling is highest in rat CA1 (7.1-8.6 bits/mum(3)) and lowest in monkey V1 (1.3-1.5 bits/mum(3)). We also evaluate the lower bound of neuron selectivity in the choice of synaptic partners. Postsynaptic excitatory neurons in rodents make synaptic contacts with >21-30% of presynaptic axons encountered with new spine growth. Primate neurons appear to be more selective, making synaptic connections with >7-15% of encountered axons.