Critical periods emerge from homeostatic structural plasticity in a full-scale model of the developing cortical column
Rinke, S., Naveau, M., Wolf, F., and Markus Butz-Ostendorf, M. (2017). In: Van Ooyen, A., and Butz-Ostendorf, M., eds. The Rewiring Brain: A Computational Approach to Structural Plasticity in the Adult Brain. San Diego: Academic Press, pp. 177-201.
Abstract
Thanks to detailed, large-scale models of the cortical column, we have today a good understanding of how electrical activity propagates along the synaptic connectivity chains linking different cortical layers. However, how this connectivity is shaped in the first place remains an open question, as large-scale models so far have used fixed connectivity only.
With our model of structural plasticity (MSP) and the modeling approaches described in this chapter, it has become possible to explore how the reciprocal interaction of electrical activity, connectivity, and synaptogenesis may lay down connections between different cell types within and across cortical layers. In our model cortical column, we found that the connectivity generated showed remarkable similarity with the real anatomical connectivity of a cortical column. Furthermore, the course of network formation showed a phase of pronounced excitatory synapse formation, which we interpreted as a “critical period” during cortical column development. The timing of this critical period was set by the developing inhibition. The critical period opened as soon as the influence of inhibitory neurons on excitatory neurons grew strong and closed when the inhibitory neurons reached their set-point level of activity. This effect may explain why benzodiazepines, such as diazepam, can advance the opening of critical periods during brain development.
The large-scale model of the cortical column was simulated using the MSP framework in the spiking neuron simulator NEST (www.nest-initiative.org). To make MSP ready for the next generation of supercomputers, we also describe an approach for simulating MSP in networks consisting of even more than 106 neurons, with network size being limited merely by memory capacity.