Individuals with Substance Use Disorders (SUD) often exhibit preferences for risky, uncertain outcomes and demonstrate inflexible learning. The neural mechanisms of adaptive learning under uncertainty, however, have been predominantly investigated in human and nonhuman primates with limited capacity for microcircuit measurements and/or systems-level causal manipulations. Guided by mechanistic computational models, here we propose precise systems-level manipulation of interactions between multiple brain areas and imaging of stable neuronal ensembles in rodents, to reveal circuit- level mechanisms underlying adaptive learning. Previous studies point to the role of basolateral amygdala (BLA) and several interconnected subregions of the prefrontal cortex (PFC), including orbitofrontal (OFC) and anterior cingulate cortex (ACC), in adaptive behavior. Based on our recent modeling and experimental work, we hypothesize that OFC and ACC provide stable representations of stimulus-outcome (state values) and action-outcome associations (action values), respectively. BLA uses input from OFC and ACC to estimate volatility in both state and action values to destabilize these representations, which in turn enables faster adjustments in response to real changes in the environment. To test this hypothesis we will chemogenetically inhibit projection neurons between these regions during probabilistic reversal learning of state and action values, and record neuronal ensemble activity via calcium imaging in each cortical region during learning. The results from manipulating different pathways and high temporal- and spatial- resolution of calcium imaging data will be used to identify the relative strength and types of projections, and the representations of reward value in order to refine our models and subsequently develop circuit-level, spiking network models for adaptive learning under uncertainty. Given that SUDs are characterized by uncertain, rapidly-changing, and often extreme reward environments, our proposed aims are pertinent to clinical observations in SUDs and especially deficits in behavioral adjustments. Altogether, using a combination of detailed computational modeling and a sophisticated experimental approach, we will reveal the contributions of cortico- amygdalar circuits to adaptive behavior under uncertainty.
Recent advances in computational psychiatry have revealed failures in using models of the reward environment to flexibly change behavior in certain neuropsychiatric conditions. Results from our project will lead to improved systems-level understanding of such inflexibility in SUDs and of the precise roles of involved brain areas for better, more effective therapeutic targeting in the future.
