Dynamical mechanisms of flexible pattern generation in spinal neural populations

Recent investigations into the coordinated activity of neuronal populations in motor circuits have revealed low-dimensional dynamical features hypothesized to support the generation of patterned motor output. However, such studies have typically relied on vast simplifications - trial-averaging, heavy smoothing - and are often far removed from muscle activity, the true output of the motor system. Thus, it is unclear whether the uncovered dynamical features are primarily useful for intuition-building, or reflect computational mechanisms that operate at the temporal precision of motor control. Here we investigated the neural population closest to motor output - spinal interneurons - and tested the degree and temporal precision with which low-dimensional spinal dynamical features reflect the generation of muscle activity. We studied lumbar intermediate zone interneuronal population activity and bilateral, multi-muscle intramuscular EMG recorded in two decerebrated T9 spinalized cats performing air-stepping. We uncovered dynamical features using AutoLFADS [1], an unsupervised deep learning method to infer latent dynamics from neural and muscle population recordings. Spinal population dynamics were highly predictive of multi-muscle activity on individual gait cycles on a millisecond timescale. However, the reverse was not true: the spinal activity was higher dimensional than muscle activity and contained features not directly related to muscle output. We hypothesized that spinal populations may use these higher dimensions to separate internal timing mechanisms from motor output. Specifically, we investigated extensor burst duration during the gait cycle, which varied step to step based on locomotion speed. We uncovered oscillatory dynamics within the spinal population activity, and found that the duration spent within the oscillator was highly predictive of single-gait cycle variations in extensor burst duration, precise to within tens of milliseconds. These results reveal low-dimensional dynamical features in spinal interneuron activity that may be integral in enabling flexible pattern generation and for precisely controlling timing variations in motor output.