Abstracts

Rationale: Absence seizures are possibly the best understood seizure type, due in large part to the excellent rodent models available. However the frequency of spike and wave discharge (SWD) in rodent models is 5-8Hz while in human it is around 3Hz. This raises concerns about the relevance of animal models to the human condition. Given the lack of a 3Hz animal model an alternative approach to understanding this frequency difference is to build biophysically detailed mathematical models of neural circuits that can then be simulated on computers. These models can be explored in ways that are impossible in the wet lab leading to the development of testable hypotheses. To link molecular level biophysics to large network dynamics, these models must necessarily be complex themselves thus requiring considerable computer resources to fully explore their potential. Methods: Our simulation is based on a previously published highly detailed model of a thalamocortical column (Traub, et al 2005). This model has been ported to the NEURON simulation environment and modified to run in parallel over large numbers of processors (code provided by M Hines). The code runs efficiently with up to 40 processors and takes 30-40 processor hours to simulate 1s of network time. We updated the model to include recently published anatomical data. The model realistically describes the electrophysiological and anatomical properties of important cells in the thalamocortical circuits of a single column. Cells modelled include principal cells in Layers 2/3, 3, 5, and 6, inhibitory interneurons in Layers 2/3 and 6, nucleus reticularis cells and thalamocortical relay cells. Model neurons have realistic morphology and membrane distribution of voltage and calcium gated conductances. GABA, AMPA and NMDA synapses were also modelled. We used Nimrod, a specialised parametric modelling system, to perform a fractional factorial parameter space search, a method by which only a fraction of all possible parameter combinations are evaluated. Approximately 30 parameters describing the magnitude of synaptic conductances and the strength of exogenous inputs were examined. Results: To represent sensory input, current injections were played into thalamocortical relay cells. To represent transcolumnar input, current injections were played into later 2/3 pyramidal cells and/or layer 5 pyramidal cells. From different combinations of three types of inputs, a range of behaviours were observed from high frequency gamma-like oscillations to oscillations with SWD-like frequencies. In states where low frequency oscillations were observed, increasing the current injections into layer 5 pyramidal cells could alter the frequency from 2 Hz to 8 Hz. Altering the relative levels of excitatory transmission (AMPA and NMDA) and inhibitory transmission (GABA) could also produce a large range of network behaviours for a given input. Conclusions: Different frequency of SWD observed between humans and rodents may be explained by different levels of input from nearby cortical columns during a seizure.