Abstracts

Sclerosis of the dentate gyrus, characterized by loss of hilar neurons and the formation of abnormal recurrent connections between granule cells, is a poorly understood consequence of repeated seizures in human patients and animal models of temporal lobe epilepsy. To determine the characteristics and consequences of network reorganization following sclerosis, we applied data-driven graph theory and computational modeling techniques. We assembled a full-scale network graph of the dentate gyrus representing 1,064,000 excitatory and inhibitory neurons and their synaptic connections as nodes and directed links based on data from the literature. Distributions of the synaptic connections were constrained by Gaussian fits to axonal distributions. The average path length (L) and clustering coefficient (C) of the dentate graph were calculated and compared to a randomly connected control graph. Large scale simulations of network activity were performed in a detailed network model that replicated the topological changes during sclerosis determined for the full dentate graph. The network model contained realistic multi-compartmental models of granule (50000), mossy (1500), basket (500) and HIPP (600) cells. The graph of the healthy dentate network had a relatively low L (high global connectivity) and high C (high local connectivity) characteristic of a [quot]small world[quot] structure (Watts [amp] Strogatz, Nature 1998). Increasing degrees of hilar neuron loss and mossy fiber sprouting (i.e. sclerosis) resulted in increased C at lower degrees of sclerosis, followed by a decrease. The L remained low until severe scleroris, indicating a high global connectivity, except after almost complete hilar cell loss. Compared to corresponding random networks, the small world features of the dentate network were progressively enhanced until severe sclerosis, with the loss of the last hilar neurons leading to a primarily locally connected, regular network structure with high L and C.
In the large scale network model, simulated responses to a single perforant path stimulation showed propagating epileptiform activity at degrees of sclerosis above 20%. The network activity reflected the changes in network topology with both duration and average firing of granule cells increased up to 80% and then decreased at 100% sclerosis. Further simulations revealed the respective contributions of hilar cell loss and mossy fiber sprouting to network hyperexcitability. Our findings demonstrate that dentate network reorganization following sclerosis is sufficicient to induce epileptiform activity, even in the absence of seizure-induced alterations in synaptic or intrinsic excitability. (Supported by NIH (NS35915) to I.S..)