skip to primary navigationskip to content
 

Parallel memory units for memory-based action selection

Supervisor: Yoshinori Aso

Animals learn predictive value of a sensory cue based on its temporal correlations with reward or punishment. This proposal seeks to understand the molecular and circuit mechanisms used to store information in parallel memory units, and how these memories are integrated to guide action selection. We will use the Drosophila mushroom body (MB) as a model system.

In the MB, sparse activity in the 2,000 Kenyon cells represents the identity of sensory stimuli. Reward and punishment activate distinct subsets of dopaminergic neurons to the MB. Along the parallel axonal fibers of Kenyon cells,  dopaminergic neurons and MB output neurons form 16 matched compartmental units.  Optogenetic activation experiments demonstrated that individual dopaminergic neurons independently write and update memories in each of these anatomically defined units with cell-type-specific rules. We found extensive differences in the rate of memory formation, decay dynamics, storage capacity and flexibility to learn new associations across different units. Thus individual memory units within the mushroom body store different information about the same learning event. Together, these memories cooperatively or competitively represent the predictive value of sensory cues.

By utilizing cutting edge methods including cell type specific genetic drivers, RNAseq, fully automated optogenetic behavioral assays, expansion microscopy combined with light-sheet microscope, we will now identify molecules and cell biological features that enable dopamine neurons to produce diverse forms of synaptic plasticity underlying distinct learning rules in different memory units. We will anatomically identify neurons that integrate information from parallel memory units in the MB, and make genetic drivers for them. Then, we will probe functions of these downstream neurons by calcium imaging and manipulating their activity while flies retrieve and integrate memories for action selection in virtual reality setup.

Relevant references

The neuronal architecture of the mushroom body provides a logic for associative learning, Y Aso, D Hattori, Y Yu, RM Johnston, NA Iyer, TTB Ngo, H Dionne, Elife 3, e04577

Dopaminergic neurons write and update memories with cell-type-specific rules, Y Aso, GM Rubin, Elife 5, e16135

Heterosynaptic plasticity underlies aversive olfactory learning in Drosophila, T Hige, Y Aso, MN Modi, GM Rubin, GC Turner, Neuron 88 (5), 985-998