Parallel, but Dissociable, Processing in Discrete Corticostriatal Inputs Encodes Skill Learning

Changes in cortical and striatal function underlie the transition from novel actions to refined motor skills. How discrete, anatomically defined corticostriatal projections function in vivo to encode skill learning remains unclear. Using novel fiber photometry approaches to assess real-time activity of associative inputs from medial prefrontal cortex to dorsomedial striatum and sensorimotor inputs from motor cortex to dorsolateral striatum, we show that associative and sensorimotor inputs co-engage early in action learning and disengage in a dissociable manner as actions are refined. Disengagement of associative, but not sensorimotor, inputs predicts individual differences in subsequent skill learning. Divergent somatic and presynaptic engagement in both projections during early action learning suggests potential learning-related in vivo modulation of presynaptic corticostriatal function. These findings reveal parallel processing within associative and sensorimotor circuits that challenges and refines existing views of corticostriatal function and expose neuronal projection- and compartment-specific activity dynamics that encode and predict action learning. •Associative and sensorimotor inputs are co-engaged during early action learning•Inputs disengage in a dissociable manner as actions are refined into skills•Associative input disengagement predicts subsequent skill learning•Somatic and presynaptic engagement diverges early in action learning Kupferschmidt et al. probe real-time activity dynamics of associative and sensorimotor cortical projections to striatum during skill learning. They reveal substantive co-engagement and dissociable disengagement of the two pathways across learning and find that associative pathway disengagement predicts subsequent skill learning.