In the FlyWire fly brain, the transmitter share of neuron input varied less among neurons of one cell type than matched input mixes; eight worm brains gave no clear result

We tested if wiring alone holds the transmitter composition of the input of a neuron near a cell-type value. In FlyWire v783 we used 2,299 target cell types and 40,253 neurons, with connections of 5 or more synapses. For each type, we found the variance across members of the input share from GABA, glutamate or ACh neurons. We compared it with nulls that move transmitter labels among input types of matched share and matched single-type variance. A placebo partition gave a mean null percentile of 0.49 to 0.52. The GABA or glutamate share gave 0.437 (95% CI 0.425 to 0.449). There were 172 types below the 5th null percentile and 79 above the 95th. The median dispersion was about 9% below the null. GABA alone gave 0.457 and glutamate alone 0.462. The effect stayed within hemispheres, with supertype or hemilineage input units, with a transmitter-confidence cut, and at a 10-synapse threshold (0.457). In 566 types, input pairs with the same transmitter but different hemilineages had share correlations 0.03 lower than pairs with different transmitters. Same-hemilineage pairs had higher correlations. Thus the effect is not due to sibling input types that replace each other. In 35 worm nerve-ring classes from 8 Witvliet brains, results were not clear (0.480, CI 0.389 to 0.570), and the worm placebo was not well calibrated. Limitations: structure only, predicted transmitters, synapse counts and not weights, one fly brain, dependent tests across types, and no mouse test.

connectome · FlyWire · Drosophila · C. elegans · excitation-inhibition balance · neurotransmitter · cell type · input composition · permutation test · hemilineage · structure-function