From microscale structure to population coding of normal and learned behavior

Abstract This study aims to understand how the ensemble activity and network architecture of a neuronal population guides natural and learned behavior. The model system is the midbrain localization pathway of the owl. Ensemble recordings, microcircuit analysis, behavioral measurements and computational modeling will be used to analyze the neural representation of auditory space and the head-orienting movement driven by it. The compact volume of tissue commanding this behavior makes a complete understanding of information processing tractable with high-throughput electrophysiological and microanatomical methods. How information about sound location is readout to guide orienting behaviors has not been demonstrated in any species. This project has the potential to fill this gap. Aim 1 will investigate the relationship between orienting behavior and activity in the neuronal population representing auditory space, in which frontal space is overrepresented. The hypothesis is based on recent work showing that sound localization can be explained by statistical inference, computed by integrating activity across the entire population. Microelectrode arrays (MEAs) will be used to map the activity of the population upon presentation of sounds. Population decoders will be constructed to determine how the population activity is readout to drive behavior. In Aim 2, the network architecture supporting the activity pattern will be studied with light and electron microscopy. Network models will combine the data to explain how connectivity and cellular computations result in the population activity and correlated firing that drives behavior. When auditory-visual cues are modified, the midbrain representation of auditory space adapts over time, and consequently drives a learned behavior. Aim 3 will directly examine this link. MEA recordings, microcircuit analysis and behavioral measurements will be made in owls adapted to prismatic spectacles. Population decoders will be used to test the hypothesis that population activity in the learned condition maintains a non-uniform population code with an overrepresentation of frontal space. Network models will be used to examine how local re-wiring may explain changes in the distribution of activity across the population. This would be the first time that neural activity and network architecture underlying sound localization are approached from the complete-population down to single-cell level, before and after learning. This integrative approach holds potential for understanding principles of population coding, plasticity and learning that operate across species and brain circuits.