A computational modeling approach to investigate energy-costs of linearly-summed synaptic activation in single neurons

Linear additivity of synaptic input is a pervasive assumption for computations performed by individual neurons. Bernander et al. (1994) first pointed out that, in a passive neuronal model, the inherent sublinear additivity of excitatory synaptic input could be linearized with the inclusion of voltage-dependent currents. However, the biophysical mechanisms needed to produce linear summation in such a manner may add to the overall metabolic cost of neural processing. The benefits may therefore be outweighed by the energy-costs. Based on in-vivo intracellular recordings, three dendritic voltage-dependent conductances seem to be of interest: a persistent sodium conductance with associated current I_NaP; a hyperpolarization-activated mixed-ion conductance with current I_h; and a potassium conductance with current I_A. Each of these voltage-dependent currents linearizes a particular range of synaptic excitation. Using a multi-compartment leaky single neuron model,  combinations of these conductances were also examined, and many are found to produce linearization over extended ranges. Regarding the energy-costs, comparison to a purely passive model reveals that some models display minimal or even no additional costs.

Work performed in primary collaboration with Dr. W.B Levy at UVA.