Neural circuits for computing dopamine prediction errors

Harvard Medical School, Boston MA

Investigators

Linked publications & trials

Abstract

As we encounter stimuli in the environment, we make predictions about their value. When outcomes fail to match these predictions, we update our predictions to reflect new experience. This learning process is thought to involve midbrain dopamine (DA) neurons, which encode prediction errors, or the difference between actual and expected reward. Besides contributing to learning, prediction errors may lie at the core of many mental illnesses, including addiction (where drug rewards are overvalued) and depression (where previously pleasurable activities no longer feel rewarding). Although much work has gone into characterizing the responses of DA neurons to reward and punishment, less is known about how DA neurons calculate these responses, and in particular, how predictions are conveyed. Recently, it was suggested that GABAergic neurons in the ventral tegmental area might relay information to neighboring DA neurons about the predicted value of cues. By sustaining activity from the onset of a reward-predicting cue to the time of expected reward, these inhibitory neurons might provide DA neurons with information they need to calculate prediction errors. This project will use optogenetics, electrophysiology, and reversible lesions to test this theory. In the first part, the inhibitory expectation signal will be eliminated on a trial-by-trial basis while the activity of DA neurons is recorded. Without the expectation signal, DA neurons should respond robustly to rewards, even when they are fully predicted. In the second part, the origin of the expectation signal will be probed through reversible inactivation of the orbitofrontal cortex (OFC). If the OFC helps drive the GABA expectation signal, then inactivating OFC should eliminate this signal in GABA cells, and cause DA neurons to respond in the same way to expected and unexpected reward. Together, these experiments will shed light on how dopamine neurons calculate prediction errors, and identify possible points of vulnerability in this system.