A roles for ratios? How ion channel densities could define neuronal firing properties

In Xenopus laevis hatchling tadpoles, locomotion is produced and sustained though a central pattern generating network comprising of the neurons in the brain and spinal cord. Experimental data has shown that these neurons can be subdivided into classes based on morphology and electrical firing characteristics which are found grouped in columns along the rostrocaudal axis. Voltage clamp traces from disassociated neurons and in-vivo recordings have shown the existence of sodium, calcium and potassium currents. In this work, we investigate whether the distinct firing characteristics of each neuron type can be explained by adjusting the ratio of channel densities of a small number of ion channels in a single compartmental neuron model. In particular, we investigate the circumstances under which a neuron firing only once in response to a step current injection, up to twice rheobase, can be turned into a multiple-spiking neuron simply by reducing the density of slow potassium channels; a phenomenon which has been observed in pharmacological studies. Hogkin-Huxley style models of individual ion channels were built based on voltage clamp data and estimates made of the channel densities for each ion channel in a typical neuron. Next, large parameter sweeps around these channel densities were run; at each point in the parameter space, a model neuron was built and tested by injecting current steps of various magnitudes. The traces were classified automatically and the parameter space summarised by simplifying each trace to a set of key features, such as the number of spikes and the spike heights. Simulation administration and analysis code was written in python, simulations distributed over a cluster of computers and run in NEURON and simulation information stored in MySQL and sqlite databases. Based on the observed voltage traces, regions of the parameter space were found which defined single-spike or multiple-spike neurons. In this poster we will discuss the results from this parameter sweep, and the tools and techniques used to run it.