Spinal interneuronal populations encode static hindlimb posture in the cat

Proprioceptive signals from primary afferents often correlate with single-joint states (but can also reflect multi-joint configuration), whereas neuronal populations in the cerebral cortex represent limb endpoint postures. Where and how this transformation emerges along the somatosensory axis from peripheral proprioceptive receptors remains unclear. We simultaneously recorded lumbosacral spinal neurons in two decerebrate, immobilized cats while a robotic device held the hindlimb at 16 static endpoint positions spanning hip-knee configurations. Using high-density multielectrode recordings, we asked how spinal populations encode static limb state. At the single-neuronal level, most neurons covaried with a single joint angle (hip or knee), a smaller subset showed combined modulation by both joints, and a small subset (which we tentatively term 'single-endpoint' neurons) showed pronounced firing modulation at a single hip-knee configuration near the boundary of the sampled workspace, with comparatively little modulation across adjacent postures. Population analyses revealed a low-dimensional structure: the first two principal components tracked knee and hip angles, and the population activity contained sufficient information to reconstruct limb endpoint position in body-centred coordinates. Together, these findings are consistent with a hierarchical organization whereby joint-based representations within spinal-interneuron populations could contribute to limb-centred representations in ascending proprioceptive pathways. The boundary preference of these neurons is consistent with either a categorical, posture-confined code or a continuous but strongly non-linear tuning peaking near the sampled boundary; the present dataset cannot distinguish these alternatives. If categorical, such signals could in principle provide spinal 'landmarks' for switching control modes or recalibrating proprioceptive population codes. KEY POINTS: We simultaneously recorded many lumbar spinal neurons in two decerebrate, immobilized cats while a robot held the hindlimb at 16 static positions to test how spinal populations encode posture. Many neurons varied with a single joint angle (hip or knee), a smaller subset showed combined hip-knee modulation and a small subset of neurons in each cat showed pronounced activity modulation at a single endpoint posture near the boundary of the sampled workspace. Population analyses revealed a low-dimensional structure: the first two principal components tracked knee and hip angles, while a third captured an activity pattern dominated by these boundary-preferring neurons. The recorded population activity contained sufficient information to reconstruct hindlimb endpoint posture in body-centred coordinates. These findings are consistent with a hierarchical organization whereby joint-based representations within spinal-interneuron populations could contribute to the emergence of limb-centred representations in the ascending proprioceptive pathways.