Diversity among plastic synapses: Defining pathway specific eTrace competence in reinforced cortical learning

Johns Hopkins University, Baltimore MD

Investigators

Abstract

Project Summary Reinforcement learning is a highly successful framework that guides learning through reward and punishment. Standard Hebbian plasticity cannot account for how animals learn through delayed reinforcement as reward/punishment received after a behavior must specifically modify relevant synapses which were active seconds to minutes before. A theoretical solution to this discrepancy suggests the generation of transient synaptic tags, referred to as eligibility traces (eTraces), which decay after neural activity. The theory further posits that such transient and “silent” eTraces are then converted to express synaptic plasticity upon release of chemical neuromodulators conveying reward information. The first experimental demonstration of cortical eTraces by the Kirkwood lab was in the visual cortex, at the feedforward layer 4 → layer 2/3 connection, where neuromodulators acting through G-protein coupled receptors (GPCRs) anchored at the postsynaptic density (PSD) convert eTraces into synaptic plasticity. Recently, the Kirkwood lab found that neuromodulator effects are restricted to L4 → L2/3 synapses and are not effective at L2/3 → L2/3 inputs. In this application, I aim to test the hypothesis that eTraces are pathway specific even within a single postsynaptic neuron. Furthermore, I hypothesize that GPCRs involved in converting eTraces are differentially expressed at postsynaptic compartments to define which pathways are competent for eTrace-mediated synaptic plasticity. I aim to test the hypothesis, using electrophysiology, that within a single layer 2/3 pyramidal cell eTraces are restricted to the feedforward pathway (Aim I). Additionally, I will investigate the presence of eTrace-specific GPCRs at each of these pathways to confirm my prediction for a molecular definition of eTrace competence (Aim II). Lastly, I will test the predictions of input-specific eTraces using several models of in vivo plasticity, which produce bidirectional plasticity specific to the layer 4 → layer 2/3 and/or the layer 2/3 → layer 2/3 pathways. Results from this work will molecularly define reinforcement competent synapses and show that these are a subset of all plastic synapses, demonstrating previously undescribed synaptic diversity that is relevant for fully understanding how neural circuits learn. Finally, this knowledge may allow the development of methods to transplant reinforcement learning with the goal of facilitating functional recovery and learning in defined regions of the brain.