Society for Neuroscience, Journal of Neuroscience, 19(33), p. 8159-8171, 2013
DOI: 10.1523/jneurosci.0118-13.2013
Full text: Download
Dopamine (DA) cells have been suggested to signal discrepancies between expected and actual rewards in reinforcement learning. DA cells in the ventral tegmental area (VTA) receive direct projections from the medial prefrontal cortex (mPFC), a structure that is known as one of the brain areas that represent expected future rewards. To investigate whether the mPFC contributes to generating reward prediction error signals of DA cells, we recorded VTA cells from rats foraging for different amounts of reward in a spatial working memory task. Our results showed that DA cells initially responded after the acquisition of rewards, but over training, they exhibited phasic responses when rats detected sensory cues originating from the rewards before obtaining them. We also observed two separate groups of non-DA cells that were activated in expectation of upcoming rewards or during reward consumption. Bilateral injections of muscimol, a GABAA agonist, into the mPFC significantly decreased the non-DA activity that encoded reward expectation. By contrast, the same manipulation of the mPFC elevated DA responses to reward-predicting cues. However, neither DA nor non-DA responses that were elicited after reward acquisition were affected by mPFC inactivation. These results suggest that the mPFC provides the information about expected rewards to the VTA, and its functional loss elevates DA responses to reward-predicting cues by altering expectations about forthcoming rewards.