Spatial and network principles behind neural generation of locomotion

Generation of locomotion is a fundamental function of the spinal cord, yet the underlying principles remain unclear. In particular, the relationship between neuronal cell types, networks and functions has been difficult to establish ^1,2 . Here, we propose principles by which functions arise primarily from spatial features of the cord. First, we suggest that projections of distinct cell types constitute an asymmetrical “Mexican hat” topology, i.e. local excitation and surrounding inhibition with dissimilar length of projection along the rostro-caudal axis. Second, this projection topology constitutes the mechanism of rhythm- and pattern generation of mammalian locomotion. Third, the role of segregation of cell types in the transversal plane is for descending fibers to find appropriate targets. Modulation of these targets allows control of motor activity by adjusting the symmetry of the projection topology. We extract these principles via a model of the mouse spinal cord, where networks are constructed by probabilistic sampling of synaptic connections from cell-specific projection patterns, which are based on previous studies ^{3, 4} . The cell-type distributions are derived from single-cell RNA sequencing combined with spatial transcriptomics ⁵ . We find that essential aspects of locomotion are readily reproduced and controlled without requiring parameter optimization, and several experimental observations can now be explained mechanistically. Further, two main features are predicted: propagating bumps of neural activity during rhythmical activity and formation of static bumps during arrest and posture. Besides linking cell types, structure and function, we propose our approach as a new theoretical framework for motor control.