How do organisms use neural circuits to learn about their environment and then use that knowledge to guide behavior? The nervous system of C. elegans contains 302 neurons whose connectivity is fully mapped, simplifying the process of investigating neural circuits. These animals are faced with the common problem of how to best find new sources of food when they are transferred to a food-free plate. Their strategy is to produce many turns to stay in a small area when it believes food is nearby and to suppress those turns when it cannot find food. We have found that C. elegans' off-food search strategy is dependent on the size of the food patch it has most recently been exploring suggesting that it uses its prior experience to guide search behavior.
Using a dimensional reduction technique, we have identified the relevant information in the on-food behavior and in the structure of the bacterial patch that predicts off-food search. The information is decoded by two sensory neurons, ASI and ASK, whose responses to bacteria match the relevant on-food statistics. Downstream of these sensory neurons, we identify several interneurons required for learning: removal of any individual neuron prevents the animal from learning the patch size. We additionally identify a downstream dopaminergic neuron, CEP, which is crucial for learning. Dopamine interacts with the broader circuit via two D1-like receptors,
dop-1 and
dop-4 on sensory neurons and postsynaptic interneurons respectively. We also find that the CREB homologue,
crh-1, is required in the same interneurons as
dop-4. The neurons required for learning overlap with the neurons required to perform a search suggesting that the same network that generates behavior also contains within itself the ability to modify it.