Connecting internal representations to behavior in the Drosophila navigation system.

Abstract

The act of navigating through an environment requires the ability to relate one’s movements to changes in location relative to one’s surroundings. However, an individual’s actions are inherently framed respective to the body (body-centric) and so cannot directly inform or be updated by representations that are anchored relative to the external world (world-centric). Here, we describe two studies that examine how this problem is solved by the navigation system of Drosophila melanogaster. First, we show that a population known as hΔB neurons encodes the translational velocity of the fly in world-centric coordinates and that this property emerges due to the converged input of PFNd and PFNv neurons onto this population. PFNd and PFNv neurons conjunctively encode world-centric head direction and body-centric velocity with opposing directional preferences. We use electrophysiology to characterize such conjunctive tuning in PFNd neurons and show a multiplicative interaction between head direction and velocity at the single-cell level. We further confirm that upstream SpsP and LNO2 neurons provide this velocity information through the graded release of inhibition. This work provides an example of how body-centric movement signals are transformed to produce an internal representation of world-centric motion. The second study focuses on the opposite problem: how world-centric representations are transformed to produce body-centric movement commands. We focus on the PFL2 and PFL3 populations, both of which directly connect the head direction and premotor systems in Drosophila. Using a combination of modeling, electrophysiology, calcium imaging, and iontophoresis we show that these neurons receive 120 ̊shifted copies of the current head direction and a common copy of the goal head direction vectors. The agreement between these vectors is then determined by a non-linear transformation of the signals. By converging onto shared downstream targets, these populations produce steering commands to maintain the fly at its goal head direction, with the PFL3 populations creating the steering drive itself and the PFL2 population modifying the gain of this drive to enhance the precision of the resulting behavior.