BEYOND RANDOMNESS: BIASED CONNECTIVITY SHAPES ODOR DISCRIMINATION IN THE <EM>DROSOPHILA</EM> MUSHROOM BODY

The insect mushroom body (MB)—center for olfactory associative learning—utilises pattern separation to facilitate odor discrimination. Within MB calyx, odors are encoded by Kenyon Cells (KCs), which receive combinatorial input from olfactory projection neurons (PNs) (A). We examined the mechanisms underlying odor discrimination in the Drosophila MB using EM-based connectome analysis, functional imaging, and modeling. Our analyses confirmed that PN-to-KC connections are not entirely random as usually assumed; instead, different KC types (e.g. αβ and γ) sample the olfactory space in a type-specific biased manner (B). Specifically, we showed that αβ KCs preferentially sample food-odor-responsive PNs, while γ KCs instead exhibit an input bias from reproduction-odor-responsive PNs.
Functional imaging of KC somata in response to ethologically relevant odors further demonstrated that αβ KCs can form separable groups of representations of food-, and reproduction-related odors (C). Within the category of food odors, the odor representations were more dispersed in KCs’ activity space, potentially supporting discrimination among food cues. In contrast, γ KCs displayed no separation across odor categories.
To assess the functional impact of these biased connections, we developed a network model incorporating type-specific PN-to-KC connectivity optimized by our experimental data. Simulations revealed that biased connections in αβ KCs increased discrimination accuracy between food odors compared to models assuming random input patterns (D). Together, our results suggest that αβ KCs oversample food-odor-responsive PNs, facilitating separation of various odor groups as well as discriminating between food-related odors, whereas γ KCs, as generalists, contribute to more similar, less categorized odor representations.