Yang Lab

Drosophila learning and decision-making

We recently started to explore the genetic and circuit basis of operant learning in Drosophila. In particular, we are focusing on uncovering genes and circuit mechanisms that lead to superior capacity to learn, using a new optogenetics-assisted approach we have developed. Three important developments laid the foundation for this approach:

Discovery of the presumed MFB equivalent in flies
We believe to have discovered the Drosophila equivalent of the mammalian medial forebrain bundle (MFB). Using a GAL4-based screen, we found a group of neurons whose optogenetic activation is profoundly attractive to flies. Flies with CsChrimson (a red-shifted channelrhodopsin) expressed in these neurons will incessantly chase after red light (Movie 1). Importantly, their "interest to pursue" stimulation by red light is robust at single-animal level, does not attenuate over extended periods of time, and supersedes their interest in pursuing food and sex.

Discovery that optogenetic stimulation of presumed MFB equivalent can support operant learning in flies
Using optogenetic activation of the presumed MFB equivalent as "reward," we have found that Drosophila appear to be able to learn the causal relationship between their actions and the consequences. Using a closed-loop tracking and light-delivery system we developed (SkinnerTrax), we found that flies can learn a simple operant rule through trial and error so as to maximize receipts of stimulation of the presumed MFB equivalent (Movie 2). Again, this learning is robust at single-animal level and develops quickly, thus allowing us to start probing its underlying genes and circuits. Our preliminary results suggest this form of learning and classical learning likely require different sets of genes and circuit components.

Development of a high-throughpt optogenetics-assisted learning paradigm
We are developing a high-throughput apparatus for SkinnerTrax that allows "teaching" 40 individual flies simultaneously (Movie 3) and plan to produce at least 10 such apparatus so that we can examine learning of 400 individual flies per day. Armed with these apparatus, we will set out to pursue two goals. First, we will define the critical circuit components that enable this form of learning, taking advantage of the large collection of neuron-targeting tools available for Drosophila. Second, we will search for mutants -- by using both a forward genetics and a candidate-based approach -- that are outliers in learning. Specifically, we aim to find mutants whose learning capacity is significantly higher than that of WT flies.

Movie 1 Drosophila with CsChrimson expressed in presumed MFB equivalent chasing after red light. Sped up 6-fold.

Movie 2 Blind Drosophila with CsChrimson expressed in presumed MFB equivalent get rewarded whenever left (experimental) fly enters the circle. Left fly learned the reward rule quite well. Right fly is yoked control (identical in genotype but fly's actions are not linked to reward).

Movie 3 Twenty Drosophila with CsChrimson expressed in presumed MFB equivalent get trained in half (one side) of our high-throughput apparatus.