Insects are ideally suited for studying fundamental mechanisms of nervous system function. Here, we propose a basic circuit mechanism in the insect mushroom body (MB) that establishes prediction error coding through the integration of present and expected reinforcement. It employs a circuit motif with feed-forward and plastic converging connections from the Kenyon cell (KC) population to a MB output neuron (MBON) and feedback from the MBON to a dopaminergic neuron (DAN) via an inhibitory interneuron. This motif has been described in the connectome of adult [1] and larval [2]fruit fly.

In a systems level model approach to sensory-motor transformation we integrate olfactory processing in the antennal lobe, sparse stimulus representation in the KC population, prediction error coding in the KC-MBON-DAN motif, presynaptic plasticity in the KC-MBON synapse and postsynaptic homeostasis in the MBON. This model accounts for the experimentally observed features of acquisition and extinction of associative memories during appetitive and aversive learning in dependence on the magnitudes of odor and reinforcer as well as on the relative timing between sensory and reinforcing stimuli. To allow for the comparison of our simulation results with behavioral data we model learning induced plasticity over the full time course of behavioral experiments, both in the adult [3] and larva (unpublished). In the larva we entail a detailed locomotory model [4] allowing us to simulate artificial agents in virtual experiments.

Funded by DFG-FOR 2705 (grant 403329959 to MN, BG) and BMBF within CRCNS project DrosoExpect (to MN, BG). PS received stipend with DFG-RTG 1960 (grant 233886668 to MN).

[1] Eichler et al. (2017) Nature, doi.org/10.1038/nature23455
[2] Eschbach et al. (2020) Nat Neurosci, doi.org/10.1038/s41593-020-0607-9
[3] Springer M, Nawrot MP (2021) eNeuro 8(3), doi.org/10.1523/ENEURO.0549-20.2021
[4] Sakagiannis P, Jürgensen AM, Nawrot MP (2021) bioRxiv, doi.org/10.1101/2021.07.07.451470