Burst Ensemble Multiplexing: A Neural Code for Hierarchical Information Processing

We study hierarchical information processing in the cortex. Thick-tufted pyramidal neurons (TPNs) combine information ascending the cortical hierarchy with descending information. These cells are known to produce distinct bursts of action potentials and long dendritic action potentials in their dendrites. In the classical view, these distinct inputs are combined nonlinearly to give rise to a single firing rate output, which collapses all input streams into one. Here, we propose an alternative view, in which a single neural ensemble can represent multiple input streams simultaneously, by using a novel code that distinguishes single spikes and bursts of action potentials at the level of the ensemble. Using computational simulations constrained by experimental data, we first show that the electrophysiological properties of TPNs are well suited to generate such a multiplexed neural code. Secondly, an information theoretical analysis shows that this novel neural code maximizes information for short and sparse bursts, consistent with in vivo recordings. Finally, we show that the two inputs streams can be decoded by widespread neural microcircuits, which combine short-term plasticity with feedforward inhibition. We propose that multiplexed temporal codes could be advantageous within a hierarchical system to distinguish ascending and descending information.