Dopaminergic neurons with distinct projection patterns and physiological properties compose memory subsystems in a brain. However, it is poorly understood whether or how they interact during complex learning. Here, we identify a feedforward circuit formed between dopamine subsystems and show that it is essential for second-order conditioning, an ethologically important form of higher-order associative learning. The Drosophila mushroom body comprises a series of dopaminergic compartments, each of which exhibits distinct memory dynamics. We find that a slow and stable memory compartment can serve as an effective “teacher” by instructing other faster and transient memory compartments via a single key interneuron, which we identify by connectome analysis and neurotransmitter prediction. This excitatory interneuron acquires enhanced response to reward-predicting odor after first-order conditioning and, upon activation, evokes dopamine release in the “student” compartments. These hierarchical connections between dopamine subsystems explain distinct properties of first- and second-order memory long known by behavioral psychologists.

To survive, animals must convert sensory information into appropriate behaviours. Vision is a common sense for locating ethologically relevant stimuli and guiding motor responses. How circuitry converts object location in retinal coordinates to movement direction in body coordinates remains largely unknown. Here we show through behaviour, physiology, anatomy and connectomics in Drosophila that visuomotor transformation occurs by conversion of topographic maps formed by the dendrites of feature-detecting visual projection neurons (VPNs) into synaptic weight gradients of VPN outputs onto central brain neurons. We demonstrate how this gradient motif transforms the anteroposterior location of a visual looming stimulus into the fly's directional escape. Specifically, we discover that two neurons postsynaptic to a looming-responsive VPN type promote opposite takeoff directions. Opposite synaptic weight gradients onto these neurons from looming VPNs in different visual field regions convert localized looming threats into correctly oriented escapes. For a second looming-responsive VPN type, we demonstrate graded responses along the dorsoventral axis. We show that this synaptic gradient motif generalizes across all 20 primary VPN cell types and most often arises without VPN axon topography. Synaptic gradients may thus be a general mechanism for conveying spatial features of sensory information into directed motor outputs.