Abstract

Animals have to constantly reassess the reliability of learned information to optimize their behaviour. When acquired knowledge turns out to be inaccurate the underlying memory needs to be re-evaluated. Some of the basic principles of these memory update processes seem to be conserved from insects to mammals.

We have started to identify the neural operations underpinning re-evaluation of memory in Drosophila. A lack of accuracy of prediction during the retrieval of olfactory memory can either lead to a cycle of memory destabilisation and re-stabilisation, an update process called reconsolidation, or direct new learning of an opposing memory, a phenomenon know as extinction. Each process involves specific parts of the dopaminergic circuitry innervating the flies’ memory cenre, the mushroom body. During reconsolidiation, the re-stabilization of the memory requires activity of a specific set of mushroom body output neurons, which sequentially recruit distinct groups of dopaminergic neurons. In memory extinction a learned odour that proves an unreliable predictor has its associated valence nullified by establishing a new opposing memory. Based on the involvement of the respective dopaminergic system, we proposed that the omission of reward is learned as punishment, whereas the absence of expected punishment functions as reward. In vivo imaging reveals the co-existence of the two memory traces, the original memory and a parallel extinction memory. Guided by light and ultrastructural anatomy we uncovered a mechanism of how these two memory traces are integrated to guide behaviour. Together, these data establish that recurrent connectivity between valence-coding output neurons and dopaminergic neurons support error driven memory re-evaluation.