The impact of stochastic ion channel activity depends on neuronal morphology

It is well known that the way a neuron integrates its synaptic inputs depends both on its morphology and its dendritic distribution of ion channels. A great deal is also known about the stochastic gating of individual ion channels. However, up to now it has been difficult to incorporate all of these aspects into unified models of single cells. Most studies either use deterministic conductances to abstract the ensemble ion channel activity, or examine stochastic ion channel behaviour in simplified cell morphologies. To investigate the impact of stochastic ion channel gating on membrane noise in realistic cell morphologies, we have used a new simulation system, PSICS, which has been designed to enable efficient simulation of models that incorporate all of these details (http://www.psics.org).

To first investigate the relative contributions of different ion channel populations to membrane noise, we simulated a simple isopotential patch of membrane with stochastic ion channels governed by Hodgkin-Huxley kinetics (after Chow & White, 1996). We find that channels which have larger open probabilities, such as K channels at resting potential in the Hodgkin-Huxley model, contribute more to membrane potential fluctuations, and even to spontaneous action potentials, than channels that have small open probabilities.

We then assessed the role of morphology in determining intrinsic electrical noise by applying an identical suite of ion channels (Mainen & Sejnowski, 1996) to several reconstructed morphologies of different cell types downloaded from an online database (http://www.neuromorpho.org). By systematically varying the resting potential in each cell and recording the voltage fluctuations in multiple dendritic locations, we determine qualitative relationships between intrinsic membrane noise and dendritic morphology. These fluctuations can be substantial (standard deviation ~ 1mV) and depend on factors such as dendritic branching pattern, diameter, location and the level of depolarisation. This work also demonstrates the feasibility of running full stochastic models of electrical activity in single neurons using a new simulator, PSICS.