Control of locomotor statistics by contralateral inhibition in a pre-motor center

During foraging, changes in sensory cues such as food odor drive changes in locomotor statistics, shifting animal behavior from dispersal to local search. While cue-evoked changes in locomotor statistics have been observed across species, the neural circuit mechanisms that govern locomotor statistics remain unclear. Here, we explore the neural encoding of locomotor statistics in Drosophila melanogaster. Analysis of walking trajectories following loss of an attractive food odor revealed multiple changes in locomotor statistics relative to baseline: a shift towards lower forward and higher angular velocities, and an increase in the autocorrelation of angular velocity. These changes drive localized search that persists for seconds after odor loss. To account for these changes, we constructed a physiologically-inspired computational model of locomotor control, consisting of units whose sums and differences govern forward and angular velocity. Unit activity depends on a connectivity matrix that includes contralateral inhibition between opposing turn-promoting units. This model reproduces the velocity statistics observed in our data, and smoothly shifts from dispersal to local search with increasing contralateral inhibition. Our model thus suggests that contralateral inhibition in pre-motor circuits may play a key role in shaping locomotor statistics. To test this model, we built genetic lines labeling two populations of contralaterally projecting inhibitory (glutamatergic) neurons in the fly pre-motor center, the Lateral Accessory Lobe. Activation of one line decreased forward velocity, increased angular velocity, and widened the angular velocity auto-correlogram, consistent with model predictions. In the second line, activation dramatically decreased forward velocity while silencing increase forward velocity. An updated model with tonic inhibitory drive can reproduce this second role for inhibition. Together, our data and model provide insight into how population activity in a pre-motor center can shape locomotor statistics to produce adaptive behavior.