Specialization and Abstraction for Forelimb Control Across Sensorimotor Cortex

Voluntary movement requires the coordinated action of distributed neural circuits to translate internal goals into the patterns of muscle activation that drive behavior. Sensorimotor cortex has long been implicated in this process, but determining how its different regions contribute has been difficult due to its densely recurrent structure. Here we make progress towards this goal by mapping how movement-related signals at multiple levels of abstraction are distributed across single neurons, populations, and cortical areas. To do so, we leveraged two-photon imaging during mouse forelimb reaching and high-fidelity markerless tracking of movement kinematics. First, we established that joint kinematics and abstract target-related signals could be decoded from mouse primary motor and somatosensory forelimb regions, with the two areas distinguished primarily by the temporal profile of abstract signals rather than by the fidelity of detailed movement information. Second, we expanded the experimental paradigm to record nearly 40,000 layer 2/3 neurons across the five main forelimb sensorimotor cortical areas. Characterizing the structure of single-cell responses revealed that interpretable activity features were widely distributed across cortex and often changed sharply at anatomical and somatotopic borders. Clustering single-cell responses revealed subpopulations of neurons with similar response profiles whose spatial distributions formed characteristic footprints that crossed anatomical boundaries. Third, we characterized the movement-related information present across areas and in these area-spanning subpopulations. Forelimb kinematic information was widespread but strongly concentrated at the border of primary motor and somatosensory cortex, while abstract action signals like target identity and target geometry were localized to a frontal footprint spanning primary and secondary motor cortex. Subpopulations were specialized for different levels of movement information, yet similar response profiles carried different amounts of information depending on area membership. Primary motor cortex emerged as an integration zone for high- and low- level representations: overlap in abstract and detailed representations emerged from mixed coding within cells from a single subpopulation. Finally, abstract target representations in frontal areas stabilized early in the behavior and displayed population- level structure suggesting that high-level goal information is preserved across movement epochs. Together, these findings describe an organizational scheme in which movement signals at different levels of abstraction are mixed across cortical areas, but hierarchically arranged across specialized, area-spanning subpopulations.