Neural connectivity of a computational map for fly flight control.

Nervous systems rely on sensory feature maps, where the tuning of neighboring neurons for some ethologically-relevant parameter varies systematically, to control behavior. Such maps can be organized topographically or based on some computational principle. However, it is unclear how the central organization of a sensory system corresponds to the functional logic of the motor system. This problem is exemplified by insect flight, where sub-millisecond modifications in wing-steering muscle activity are necessary for stability and maneuverability. Although the muscles that control wing motion are anatomically and functionally stratified into distinct motor modules, comparatively little is known about the architecture of the sensory circuits that regulate their precise firing times. Here, we leverage an existing volume of an adult female VNC of the fruit fly to reconstruct the complete population of afferents in the haltere-nature's only biological "gyroscope"-and their synaptic partners. We morphometrically classify these neurons into distinct subtypes and design split-GAL4 lines that help us determine the peripheral locations from which each subtype originates. We find that each subtype, rather than originating from the same anatomical location, is comprised of multiple regions on the haltere. We then trace the flow of rapid mechanosensory feedback from the peripheral haltere receptors to the central motor circuits that control wing kinematics. Our work demonstrates how a sensory system's connectivity patterns construct a neural map that may facilitate rapid processing by the motor system.