Project Summary All animals need to detect threats in their environment to survive. Objects on a collision course cast expanding shadows on the retina (i.e., looming) that elicit innate defensive responses from insects to humans. During the previous award of this grant, we discovered that a retinal interneuron, the VGLUT3-expressing amacrine cell (VG3-AC), detects looming and drives innate defensive responses in mice. Here, we follow up on this discovery to understand how dendritic processing gives rise to feature-selective responses of VG3-ACs (Aim 1), and how VG3-ACs use dual transmitters (glutamate and glycine) to generate divergent feature representations downstream and guide behavior (Aim 2). Dendritic processing and dual transmission are features of subcellular modularity, which we propose as an organizing principle of interneurons. To explore subcellular modularity, we developed methods to combine two-photon calcium imaging and serial-section electron microscopy in the same tissue (i.e., functional connectomics). In Aim 1, we will combine functional connectomics with computational modeling and cell-type-specific genetic manipulations to test the hypotheses that synaptic inhibition and arbor morphology compartmentalize VG3-AC dendrites and that dendritic compartmentalization generates looming- selective responses. In Aim 2, we combine functional connectomics, optogenetics, and cell-type-specific genetic manipulations, to test the hypotheses that VG3-ACs use glutamate and glycine to communicate their responses with opposite sign to two categories of ganglion cells and that this target-specific use of dual transmitters generates divergent representations of looming in the retinal output. We know little about how retinal processing relates to visual processing in the brain and behavior. To fill this gap in our knowledge, we have established projection-specific large-scale recordings from retinal ganglion cells, large-scale recordings from subcortical ganglion cell targets, and behavioral assays. This allows us to track how looming signals of VG3-AC dendrites are transformed across subsequent stages of processing to guide behavior. In Aim 1, we will test the hypotheses that downstream neurons lose their feature selectivity and that innate defensive responses generalize to non- threatening stimuli when dendritic processing of VG3-ACs is disrupted (i.e., when local processing becomes global). In Aim 2, we will test the hypothesis that VG3-ACs use glutamate and glycine to generate impressed- and suppressed-by-looming responses in two categories of ganglion cells and that these ganglion cells converge in the superior colliculus to drive defensive behaviors and control the contrast gain of these responses, respectively. Together, our studies will provide insights into the specifics and general principles of interneuron organization, mechanisms, and functions and bridge that gap in our understanding from retinal processing to visual processing in the...