Discrimination learning and neural dynamics in gustatory cortical circuits

About This Grant

PROJECT SUMMARY Discrimination learning (DL) – that is, learning to produce different behavioral responses to sensory stimuli sharing some similarities but predicting distinct outcomes – is critical for survival. Different neural mechanisms have been proposed for DL, with some models focused on enhancing sensory representations in primary sensory areas, and others emphasizing a sharpened ability to interpret sensory information in “higher order” areas to guide decisions. The rodent gustatory cortex (GC), with its involvement in processing both sensory and decision-related signals, offers the opportunity to integrate these two views, providing a unifying network mechanism for DL. We hypothesize that such mechanism is mediated by metastability, a dynamic mode of cortical activity marked by transient patterns of coordinated spiking across neurons. Sequences of metastable states have been extensively described in GC and modeled though a specific network architecture featuring clusters of spiking neurons. This proposal will test the hypothesis that DL shapes both sensory and decision- making-related processes in GC through the development and modulation of metastable sequences. Aim 1 will rely on behavioral training on a taste-mixture 2-alternative choice task (2AC), chronic, high-density electrophysiological recordings, and Hidden Markov Model (HMM) analysis to study gustatory DL and monitor plastic changes of metastable activity in ensemble of neurons from the GC of behaving mice. The experiments will test the hypothesis that DL shapes activity associated with sensory and decisional processes in GC through the modulation of metastable sequences. Aim 2 will rely on machine learning to generate a class of spiking neuron networks capable of capturing the DL-induced changes in the 2AC experiments of Aim 1. These modeling studies will test the hypothesis that DL-mediated changes in metastable dynamics are the result of specific reconfigurations of network architecture and provide testable predictions on how different architectures would respond to optogenetic perturbations. Aim 3 will use a perturbative approach combined with electrophysiological recordings to test the predictions from the modeling studies in Aim 2. In addition to testing the hypothesis that the enhanced behavioral performance seen in DL is causally linked to changes in metastable dynamics, the experiments in this aim will help selecting the network model of DL that best captures responses to perturbations. Altogether, the proposed experiments will establish a new approach for studying the network mechanisms of DL. In addition to enhancing our understanding of gustatory DL, per se a very relevant topic that has so far garnered little attention, the proposed research will advance our knowledge of how metastable dynamics can mediate DL in the context of sensorimotor tasks, and unveil the network architectures that can support these changes.