Connectivity and computations in higher-order olfactory neurons in Drosophila

Understanding how odors are encoded in the brain is of fundamental importance to neurobiology. The first two stages of olfactory information processing have been relatively well studied in both vertebrates and invertebrates. However, the organizational principles of higher order olfactory representations remain poorly understood. Neurons in the first relay of the olfactory system segregate into glomeruli, each corresponding to an odorant receptor. Higher-order neurons can receive input from multiple glomeruli, but it is not clear how they integrate their inputs and generate stimulus selectivity.;In the fruitfly Drosophila melanogaster and other insects, there are two higher-order olfactory brain regions---the lateral horn and the mushroom body. These areas are thought to be functionally specialized in how they encode odorants and what mechanisms they use. One proposal suggests that the mushroom body performs associative computations and underlies olfactory learning capabilities, whereas the lateral horn is proposed to perform stereotyped computations that underlie unlearned odor-driven behaviors. While the mushroom body has received considerable attention, we know little about how lateral horn neurons (LHNs) integrate information across glomeruli and how they respond to odorants.;Here we show that LHNs receive input from sparse and stereotyped combinations of glomeruli that are coactivated by odors, and that certain combinations of glomeruli are over-represented. We show that one morphological LHN type is broadly tuned and sums input from multiple glomeruli. These neurons have a broader dynamic range than their individual glomerular inputs do. By contrast, a second morphological type is narrowly tuned and receives prominent odor-selective inhibition through both direct and indirect pathways. We show that this wiring scheme confers increased selectivity. Our findings show that the connectivity and stimulus selectivity of LHNs contrasts with the properties of mushroom body neurons. These differences support the notion that the lateral horn performs stereotyped computations that could mediate unlearned behaviors. Our findings also show that LHNs perform a greater diversity of computational operations than previously thought. We also present, in chapter four, new techniques we are developing for comprehensive and high-throughput mapping of the inputs to individual lateral horn neurons.