Abstracts

Rationale: A recent modeling study suggested that details of circuit wiring might have a significant effect on excitability in dentate gyrus (PMID 18375756). The neocortex is an area of great circuit complexity where 6 layers and multiple cell types permit many wiring permutations. We investigated a computer model of disinhibited neocortex in order to determine effects of altered inter- or intralaminar connectivity. Methods: Simulations were run in Neuron using event-driven rule-based units featuring various dynamical features thought to play a role in seizures: adaptation, bursting, depolarization blockade, voltage-sensitive NMDA conductance, etc. Baseline wiring was based on published models and anatomical studies, primarily but not exclusively from visual cortex. The basic network had 1,620 cells with 278,800 synapses, of which 223,200 were excitatory. Simulated shock was performed by stimulating between 60-80% of layer 4 and 6 pyramidal neurons to threshold. Alteration in path lengths was affected through sprouting of synaptic connections between and among specific cell populations. In addition to applying standard graph theoretic measures, we developed a new path-length measure for neuronal networks that uses a weighted edge distance proportional to propagation delay/excitatory synaptic weight. We used afterdischarge (AD) duration as a measure of network excitability. Results: 100 randomly wired networks produced ADs between 170 and 330 ms. Decreasing pathlength (PL) and increasing clustering coefficient (CC) correlated weakly with duration (-0.3 and 0.3 respectively). 20 randomly wired networks were then rewired through focal wiring changes between or within individual cortical layers, producing 520 networks. AD duration correlated weakly with decreasing PL (-0.36), and moderately with CC (0.56) or decreasing weighted path length (-0.59). We examined specific pathways to determine their role in hyperexcitability. Connections with and between layers 4 and 2/3 were primary determinants of AD duration. Layer 4 intralaminar connectivity probability above ∼15% was required for AD production (duration increased from 28.3±.3ms to 193±2ms; n=20). Increasing layer 4→2/3 connection probability resulted in a sharp AD increase at 25% connectivity (180±16ms to 230±35ms;n=20). Increasing layer 2/3→4 recurrent connectivity at ∼18% resulted in mild increase in AD from 192±22ms to 213ms±30ms. Increasing layer 2/3 connectivity probability at ∼23% produced AD increases from 198ms±32ms to 230ms±35ms. Other changes in connection probabilities produced small or inconsistent changes in AD. Conclusions: The basic topological interpretations of a network provided by standard graph theoretic measures are correlated moderately with disinhibited neocortical excitability. Slightly improved correlation is provided by using a weighted pathlength measure that makes use of synaptic delay and strength. The circuit appears to be acutely sensitive to alterations in wiring affecting layer 4, partly because layer 4 was the primary stimulation site in these simulations.