Abstract
Recurrent attractor networks are widely thought to form the basis of working memory(1-3), but how stable attractor activity can be rapidly switched on and off is unclear(4-7). Here we investigate how stability and rapid switching can be combined in a discrete recurrent circuit of the fly navigation center(8). hΔK and PFG neurons are recurrently connected in a ring structure. Using in vivo imaging, we find that these two populations exhibit shared persistent bump activity that turns on with odor and terminates at the end of a goal-directed upwind run. Using whole-cell recordings, we show that persistence in hΔK depends on recurrent signalling, and that hΔK receives slow recurrent excitation and fast inhibition from its synaptic partners. Computational modeling reveals that this combination of slow excitation with fast inhibition yields stable and tuneable persistent attractor dynamics. Next we examine the mechanisms that allow this activity bump to be rapidly turned on and off. We find that while the PFG bump tracks heading continuously, the hΔK bump is suppressed during turns, and only becomes active during straight goal-directed runs. We can reproduce these dynamics in our model by using inhibition to dynamically uncouple activity in hΔK from PFG. When hΔK is inhibited, PFG neurons follow their inputs from the compass system; when hΔK is disinhibited, recurrent interactions lock this input into place, forming a heading memory. Consistent with this model, we find that inhibitory inputs onto hΔK increase during turns and are suppressed during odor input and goal-directed upwind runs. Our work reveals how disinhibition can serve as a gate to rapidly write an ongoing measurement to a recurrent memory circuit.