Myelin-induced gain control in nonlinear neural networks

Abstract

Abstract Myelin surrounds axonal membranes to increase the conduction velocity of nerve impulses and thus reduce communication delays in neural signaling. Changes in myelination alter the distribution of delays in neural circuits, but the implications for their operation are poorly understood. We present a joint computational and non-linear dynamical method to explain how myelin-induced changes in axonal conduction velocity impact the firing rate statistics and spectral response properties of recurrent neural networks. Using a network of spiking neurons with distributed conduction delays driven by a spatially homogeneous noise, we combined probabilistic and mean field approaches. These reveal that myelin implements a gain control mechanism while stabilizing neural dynamics away from oscillatory regimes. The effect of myelin-induced changes in conduction velocity on network dynamics was found to be more pronounced in presence of correlated stochastic stimuli. Further, computational and theoretical power spectral analyses reveal a paradoxical effect where the loss of myelin promotes oscillatory responses to broadband time-varying stimuli. Taken together, our findings show that myelination can play a fundamental role in neural computation and its impairment in myelin pathologies such as epilepsy and multiple sclerosis.