Network and circuit dynamics supporting dopaminergic control of sudden insight learning

Project Summary Learning about environmental stimuli that predict meaningful outcomes such as rewards is crucial for the survival of an organism. Cue-reward learning has historically been described by the “learning curve”, which appears gradual in group-averaged behavioral data from many animals. However, it was shown nearly two decades ago that learning curves from individual animals are not as gradual, but instead display an abrupt appearance of behavioral learning. Such abrupt learning has been described as reflecting a sudden insight into task structure. Despite this demonstration, little work has focused on understanding the distributed brain networks that support such sudden behavioral learning of cue-reward associations. Here, we focus on the network and single cell changes that precede both conditioned behavioral responding and dopamine release in nucleus accumbens and during cue-reward learning. While many studies focus on the role of cue evoked dopamine release in behavioral responding, our approach is fundamentally different as it identifies signals upstream of dopaminergic learning and/or behavioral responding. Based on theoretical models of reinforcement learning, we hypothesize that such signals will be conveyed by brain regions that maintain a timeline of experience, thereby providing the template from which cue-reward associations can be extracted by DA neurons to guide learning. Based on published and preliminary data, we will test the hypothesis that a network spanning lateral entorhinal cortex (LEC), hippocampal CA1, and orbitofrontal cortex (OFC) guide DA signaling during acquisition of cue-reward associations. To test this hypothesis, we will use a combination of large-scale two-photon calcium imaging, high-density electrophysiology with Neuropixels probes, dopamine sensor fiber photometry, behavior measurements and circuit dissection approaches around abrupt cue-reward learning in individual mice. We will analyze the data using a host of approaches to quantify latency of neural learning in each region and identify the causal flow of information between the different regions. We will also analyze the nature of learned representations in neurons that precede DA signaling, which will allow a distinction between theories of associative learning. Collectively, this proposal aims to close a fundamental gap in our understanding of the neural mechanisms guiding insight learning.