Population-level encoding of somatosensation in mouse sensorimotor cortex

Somatosensation constructs the body's dynamic sense of state and allows for dexterous and precise movements. Encoding somatosensation via similar computational principles as motor commands would provide a simple and direct mechanism for sensorimotor integration. Here, we tested the hypothesis that neural encoding of somatosensory information from the limbs shares the known structure of motor commands, including low dimensionality, encoding of specific kinematic variables during limb movements, orthogonal representations of bilateral limbs, and conserved representations across animals. To address this question, we used two-photon imaging to record the activity of thousands of neurons in layers 2/3 of sensorimotor cortex of eight anesthetized mice during passive single-limb deflection. We additionally analyzed responses to passive single-limb movements in eight awake mice, sourced from an open dataset. In both datasets, we computed the principal components of the neural population to overcome the heterogeneous responses of single neurons. In support of our hypothesis, we found that a small fraction of principal components explained a large fraction of response variance. These low-dimensional representations of limb movements were well conserved across animals, including the orthogonal representations of ipsilateral and contralateral limbs. Finally, neural populations encoded information about endpoints of limb movements in addition to dynamic changes in joint angles during movements. Together, these results demonstrate that population-level encoding of somatosensory information in mouse sensorimotor cortex is structured to facilitate sensorimotor integration across the brain.NEW & NOTEWORTHY We provide evidence that neural encoding of somatosensory limb inputs mirrors well-known encoding principles of motor commands. Population-level analysis of neural activity during passive limb movements in mice revealed a conserved and low-dimensional structure with orthogonal representations of ipsilateral and contralateral limbs. Neural populations also encoded endpoint positions in addition to joint angle changes during limb movements. These results demonstrate somatosensory representations are organized to facilitate sensorimotor integration, which can inform the development of neural prostheses.