In models from three fly mushroom body connectomes, glomerulus input share gave a community input advantage, Kenyon cell co-convergence gave none, and real wiring gave no gain for 110 odors
We tested the proposal that the fly mushroom body is an associative fovea for food odors (Zheng et al. 2022). We used FlyWire v783 (left and right) and hemibrain v1.2 (right) projection neuron to Kenyon cell wiring. Against a null that keeps each neuron's connection count, a co-convergent community of 16 to 18 glomeruli occurred in each hemisphere. Thirteen glomeruli (DL2v, DM1-DM6, DP1l, DP1m, VA2, VA4, VM2, VM3) were in all three. A community defined in one hemisphere replicated in another in 4 of 6 tests at p = 0.0002 and more weakly from FlyWire to hemibrain (p = 0.038, 0.070). Glomerulus input shares agreed across hemispheres (Spearman 0.89-0.93). We then built sparse thresholded Kenyon cell models from real wiring and two nulls: a fixed-degree null that keeps input shares and removes co-convergence, and an equal-share null. With synthetic inputs, real wiring reduced the identification gap between community-only and other-only patterns relative to the equal-share null (z 2.75 and 4.41 in FlyWire, 0.48 in hemibrain; 7.09, 6.18, 3.51 with per-PN rate division). Relative to the fixed-degree null, it gave no change (z -1.67 to 0.24). With 110 measured odors (Hallem and Carlson 2006), real wiring and the fixed-degree null gave equal accuracy. Real wiring was less accurate than the equal-share null in 10 of 24 setting-hemisphere cases and more accurate in none. It had lower Kenyon cell dimensionality than each null in about half of the cases. Odors that drive the community more did not get a larger gain (48 tests, no positive test at p < 0.05). Thus, in these models, the community advantage comes from unequal input shares, not from co-convergence. Limitations: a simple rate model with assumed parameters, receptor data for only 23 glomeruli (none for DM1, VA2, VA4, DP1m), two of the three hemispheres from one fly, and no spatial bouton null.
Drosophila · mushroom body · Kenyon cells · projection neurons · connectome · FlyWire · hemibrain · olfaction · odor discrimination · null models · co-convergence · structure-function
References
14- reviewsdoi:10.1016/j.cub.2022.06.031
We test the associative fovea proposal of this work in three hemispheres with thresholded models and measured odors.
- usesdoi:10.5281/zenodo.10676866
FlyWire v783 proofread connections.
- usesdoi:10.1038/s41586-024-07558-y
FlyWire connectome.
- usesdoi:10.1038/s41586-024-07686-5
FlyWire cell type annotations and hemibrain type matching.
- usesdoi:10.7554/eLife.57443
Hemibrain v1.2 connectome.
- useshttps://storage.googleapis.com/hemibrain/v1.2/exported-traced-adjacencies-v1.2.tar.gz
Hemibrain traced adjacency export.
- usesdoi:10.1016/j.cell.2006.01.050
Receptor responses to 110 odors.
- usesdoi:10.1038/srep21841
DoOR database copies of receptor data, consensus matrix and receptor-glomerulus mapping.
- useshttps://github.com/ropensci/DoOR.data/tree/db323a496577c4b4a72b5c2fcd1859e07521ffb5
DoOR.data commit used.
- useshttps://github.com/flyconnectome/flywire_annotations/tree/a83b2776d60d5764cef36b927f5f9679c16c47a2
Annotation commit used.
- usesdoi:10.1016/j.neuron.2010.04.009
PN transform with divisive normalization.
- usesdoi:10.1038/ncomms5114
Curveball fixed-degree null algorithm.
- doi:10.1038/nature12063
Random convergence model of PN to KC wiring.
- doi:10.1016/j.neuron.2017.01.030
Dimensionality of random expansion layers.
Cited by
4 publicationsPublic catalog entries on Sciscape.
- In three fly antennal-lobe connectomes, the local-neuron inhibition pattern of a glomerulus did not explain its projection-neuron odour responses better than the pattern of other glomeruli
Earlier campaign wiring model test in the fly mushroom body.
- reviewsComplete channel loss covered all sampled recall failures in three selected fly calyx models
Prior odor identification test with noise and wiring controls; the current test uses fixed baseline templates and graded input suppression.
- reviewsWeight shuffles gave less mean code change after one input loss in three fly calyx models
Prior odor-identification test of observed and randomized fly mushroom-body topology. No code or results reused.
- Projection-neuron wiring in three fly brain hemispheres predicts lateral horn odour tuning better than shuffled wiring, but with mean r 0.16 against 0.46 between cells of the same type
Campaign test of wiring-based odour models in the mushroom body; we extend the question to the lateral horn.