A neural geometry for forelimb proprioception in the cervical spinal cord

Precise, real-time somatosensory feedback is essential for coordinated movement, yet how the nervous system constructs a kinematic representation of the limb from this feedback is unclear. Here, using single-unit electrophysiology in the cervical spinal cords of awake and anesthetized mice, we show that spinal neurons generate a precise, ordered, low-dimensional representation of forelimb kinematic state from muscle and tendon afferent input. This representation arises from two molecularly distinct, developmentally defined classes of dorsal horn neurons. One class encodes limb speed, position and direction; the other fires phasically, with precise, radially periodic firing fields that tile kinematic space. Transient perturbation of muscle and tendon afferents in mice reaching to spatial targets causes endpoint errors predicted by deficits in the neural code. These findings show that spinal networks, one synapse removed from the periphery, are sufficient to construct a precise representation of forelimb movement.