Why Have Dendrites? A Computational Perspective

This chapter concerns how dendritic trees may contribute to the information processing functions of individual neurons in the CNS. Before tackling this fascinating question, though, it is useful to address two background issues. First, could dendrites exist for reasons other than to enhance a cell's computing functions per se? If we focus on their spatial extent, we might infer that dendrites exist to increase the receptive surface area of a neuron, which they do by 10 or 20 fold. This makes room for a much larger number of synaptic connections between neurons, which seems advantageous. If we focus on their highly branched morphology, we might infer that dendrites are optimally shaped (per unit length) to extract information-bearing signals from the surrounding neuropil, analogous to root systems that are optimized for the extraction of water and nutrients from surrounding soil (Bejan 2000; Chklovskii, Schikorski et al. 2002). If we focus on their anatomically distinct subregions (e.g., apical vs. basal subtrees), we might infer dendrites exist so that different input pathways can be physically segregated on the surface of the cell (Shepherd 1998; McBain and Fisahn 2001). This could allow different classes of presynaptic terminals to be targeted by different modulatory substances (Patil, Linster et al. 1998), to be influenced by different classes of inhibitory interneurons (McBain and Fisahn 2001), or to be subjected to different synaptic learning rules (Golding, Staff et al. 2002; Froemke, Poo et al. 2005). Thus, at the very least, dendritic trees are likely to be multi-purpose devices which, apart from their information processing functions, play various physical roles in the early development and day-to-day operation of the neural circuit.