Simulation of motoneuron morphology in three dimensions. II. Building complete neurons

By using dendrogram data from six adult cat alpha motoneurons, we have constructed computer simulations of these cells in three dimensions (3D) by "growing" their dendritic trees from stem branches that were oriented as in the original cells. Individual trees were simulated by using the algorithms and parameters discussed in the companion paper (Marks and Burke [2007] J. Comp. Neurol. 503:685-700). It was not possible to distinguish real from simulated motoneurons by visual inspection of 3D drawings. Simulated cells were compared quantitatively with their actual exemplars by using features that were measured in spherical shells at various radii centered on the soma. These included nearest neighbor distances (NNDs) between branches, the sizes and overlaps between the territories of individual dendrites measured as convex hulls (polygons that enclose all branches passing through a shell), and the sizes of circular zones that contained no branches. We also compared the 3D fractal dimensions and lacunarity (a measure of the 3D dispersion of branches) in actual cells and their simulations. The statistical properties of these quantitative measures were not significantly different, suggesting that the simulation algorithm was quite successful. However, there were three exceptions: 1) there were more NNDs at distances < 50 microm in simulated than in actual motoneurons; 2) average overlaps between the territories of different dendrites were almost twice as large in simulated compared with actual motoneurons; and 3) estimates of lacunarity were also larger in simulated cells. These exceptions suggest that dendritic branches in actual motoneurons tend to avoid one another. We discuss possible interpretations of these results.