Distinct neural state dynamics associated with motor state transitions in C. elegans

Locomotor behaviors are composed of ordered transitions between motor states required to produce a functional outcome. However, the circuit mechanisms by which the brain generates a sequential ordering of motor actions for constructive behavioral output, particularly in freely-behaving animals, is largely unknown. With 302 neurons within a well-defined connectome and a library of genetic tools, complex yet discernible behaviors, and recent advances in brain-wide imaging, the nematode Caenorhabditis elegans offers a unique opportunity to study how the brain generates behavior. To better understand the link between neural activity and behavior, we fit a data-driven dynamical systems model called a recurrent Switching Linear Dynamical Systems model (rSLDS) to data from recently published whole brain recordings of freely moving C. elegans on sparse food (Atanas et al., Cell 2023). We show the behaviors and dynamics recovered by rSLDS are in good alignment with canonical and quantitatively measured behaviors and broadly known neural activity associated with the given behaviors. We further show that the model identifies internal states that differentially express subclasses of the same motor state (e.g. different durations and transitions to reversals) and that these behaviors can be decoded from neural activity before they are expressed behaviorally. We identify the interneuron RIM as involved in the state that promotes sustained reversals and use the model to predict that the acute inhibition of RIM should thus suppress sustained reversals while still permitting short, intermittent reversals. This model is consistent with results from prior studies. We confirmed this experimentally through chemogenetic inhibition of RIM during spontaneous foraging and find that both reversal frequency and duration are reduced as predicted. These results not only identify new neural underpinnings to poorly understood behaviors, but more broadly provide evidence for internal states driving behaviors and behavioral transitions.