The spatial reconstructions of cerebellar Purkinje cells and hippocampal pyramidal neurons of mice and spinal motoneurons of frogs (n=5–5 neurons) were used to create passive segmental cable models of these cells by the NEURON simulator (Duke University, USA). Sinusoidal (f=50 Hz) and steady-state current inputs were modelled at thousands of dendritic sites per neuron to map the whole dendritic surface, and log attenuations of somatopetally propagating PSPs were computed between dendritic points and the soma in response to the different current inputs. Neurons were modelled both with homogeneous and inhomogeneous soma-dendritic membranes, where somatic and dendritic specific membrane resistances were equal (Rms=Rmd, homogeneous model) or different (Rms<Rmd, inhomogeneous model with a step increase in Rm at the soma-dendritic junction). Log attenuations of PSPs and path distances of dendritic input sites from the soma were normalized in each neuron and percentages of dendritic surface areas with the same distance and log attenuation ranges were summed up. These data were visualized as colour coded morphoelectrotonic matrices (MEM) and similarities between these matrices of neurons were measured by the sums of absolute differences between corresponding pairs of matrix elements. Finally, cluster analysis of four sets of MEMs of neurons (matrices computed in the homogeneous and inhomogeneous membrane models with steady-state and sinusoidal current inputs) was performed based on the above definition of similarity. Cluster analyses of MEMs grouped all matrices correctly by neuron types if MEMs were computed in homogeneous soma-dendrite models with steady-state current inputs. In all other cases (inhomogeneous models with both steady-state and sinusoidal current inputs and homogeneous models with sinusoidal current inputs) 14 out of 15 neurons were correctly classified based on their MEMs. We conclude that morphoelectrotonic matrices are sensitive tools to map the architecture of the whole dendritic surface and they can discriminate between different neurons with high confidence.

This work was supported by the grants OTKA K67747 and ETT 025/2006.