Transforming a head direction signal into a goal-oriented steering command

To navigate, we must continuously estimate the direction we are headed in, and we must use this information to guide our path toward our goal (Schöne, 2014). Direction estimation is accomplished by ring attractor networks in the head direction system (Knierim et al., 2012, Hulse et al., 2020). However, we do not understand how the sense of direction is used to guide action. Drosophila connectome analyses (Rayshubskiy et al., 2020, Hulse et al., 2021) recently revealed two cell types (PFL2 and PFL3) that connect the head direction system to the locomotor system. These populations also receive almost all the same inputs, including those that may constitute a goal signal. Here we show how both cell types combine an allocentric head direction signal with an internal goal signal to produce an egocentric motor drive. We used 2-photon calcium imaging and electrophysiology to monitor their activity as flies navigated in a virtual reality environment toward a goal stored in memory. Strikingly, PFL2 and PFL3 populations are both modulated by deviation from the goal direction, but with opposite signs. The amplitude of PFL2 activity is highest when the fly is oriented away from its goal; chemogenetically activating these cells destabilizes the current orientation and drives turning. By contrast, total PFL3 activity is highest around the goal; these cells generate directional turning to correct small deviations from the goal. Our data support a model where the goal is stored as a sinusoidal pattern whose phase represents direction, and whose amplitude represents salience. Variations in the amplitude of this internal goal signal can explain observed transitions between goal-oriented navigation and exploration. Together, these results show how the sense of direction is used for feedback control of locomotion.