CNS*2026 Halifax

Introduction
In Drosophila melanogaster, associative learning occurs in the mushroom body, where synapses between Kenyon cells (KCs) and mushroom body output neurons (MBONs)[1] are modulated by dopaminergic neurons (DANs) that convey reinforcement signals. KC–MBON–DAN circuits form parallel functional units that influence behaviour [2], and distinct learning properties across compartments allow appetitive and aversive memories for the same stimulus to coexist [3]. However, the interaction between the compartments is not fully understood. Here we implemented a mushroom body network model having parallel MBON units with different time scales and valences to investigate how the interactions between these units help in shaping different behaviours.


Methods
We propose a network model consisting of a KC layer for odor representation, multiple MBONs with short-term (STM) and long-term (LTM) memory, and a set of DANs representing unconditioned stimuli. Synaptic weight between KC and MBON depends on the relative timing between KC and DAN activity. We have implemented cross valence inhibitory modulation and hierarchical interaction between LTM and STM, where strong LTM activity can positively influence STM compartment activity. Behavioural readout is determined by the relative firing rates of the MBON population encoding opposite valences.


Results
The model could reproduce the core features of associative learning including first order conditioning. Parallel MBONs allow associations with opposite valence to coexist for the same stimuli. Due to the presence of cross valence inhibitions, the model can exhibit valence shifting during sequential experiences of opposing reinforcements or when relative influence of other compartment changes over time. Hierarchical LTM-STM interactions further enable second order conditioning, producing a short-term memory.


Discussion
Our results indicate that interactions between parallel memory units can produce flexible behaviour even with a relatively simple plasticity rule. The hierarchical interaction between memory units of different timescales allows stable long-term memories to influence short term memories, which helps in adapting to a dynamic environment. These results highlight how network architecture of the mushroom body can support flexible yet stable behaviours. 
\n


References

1. Waddell, S. (2013). Reinforcement signalling in Drosophila; dopamine does it all after all. Current Opinion in Neurobiology, 23(3), 324–329. https://doi.org/10.1016/j.conb.2013.01.005
2. Aso, Y., Hattori, D., Yu, Y., Johnston, R. M., Iyer, N. A., Ngo, T., Dionne, H., Abbott, L., Axel, R., Tanimoto, H., & Rubin, G. M. (2014). The neuronal architecture of the mushroom body provides a logic for associative learning. eLife, 3, e04577. https://doi.org/10.7554/elife.04577 
3. Aso, Y., & Rubin, G. M. (2016). Dopaminergic neurons write and update memories with cell-type-specific rules. ELife, 5. https://doi.org/10.7554/eLife.16135 

Acknowledgement
This work is supported by the Centre for High Impact Neuroscience and Translational Applications (CHINTA), TCG CREST. I sincerely thank Dr. C. Sivaraju for his valuable discussions and encouragement.