Short-term synaptic plasticity and dendritic non-linearities support spike-based communication in the nervous system
Arguably the most salient property of neurons (with only a few exceptions) is that they communicate by spike-like action potentials. A little-appreciated consequence of spike-based communication is that it puts severe constraints on the communication bandwidth between neurons: spikes are a substantially impoverished representation of the fast-evolving, analog membrane potentials (or firing rates) of the neurons concerned that are normally considered to lie at the heart of computation. We have developed a theoretical framework suggesting that two ubiquitous properties of neuronal biophysics allowing the highly nonlinear integration of signals in individual neurons both in the temporal (short-term plasticity) and spatial domain (active dendritic processes) are in fact optimal adaptations to the informational bottleneck brought about by spike-based communication. A key prediction of this theory is that nonlinearities in a postsynaptic neuron must be matched to the statistics of activity patterns in the population of its presynaptic partners. Beside reproducing already existing experimental data available in the literature, our theory also accurately predicted the responses of two cortical cell types to patterned stimulation by two-photon glutamate uncaging. These results reveal a new computational principle underlying the nonlinear integrative properties of neurons by suggesting a functional link between cellular and systems-level properties of neural circuits.