EN
BACKGROUND AND AIMS: Reinforcement-based learning drives behavior towards actions with highest perceived outcome value. It’s essential features are the ability to associate actions or stimuli with rewards, discounting of the delay or probability of the rewards and balance between exploitation of known rewarded actions against exploration of new possibilities. Here we investigate how disrupting NMDA receptor-dependent signaling in the brain’s dopamine systems affects reinforcement learning. METHODS: Genetically modified mice with selective inactivation of NMDA receptors on dopaminergic or dopaminoceptive neurons were generated using the CreERT2/loxP system. Behavior of control and mutant mice was assessed in tasks involving instrumental or Pavlovian learning as well as discounting of reward probability and delay. RESULTS: Inactivation of NMDA receptors on dopaminergic neurons impaired the acquisition of conditioned reinforcement, even though it had no general effect on associative learning. Conversely, in mice with inactivation of NMDA receptors in dopaminoceptive neurons, an opposite phenotype was observed: deficits in associative learning but normal conditioned reinforcement. Interestingly, the effects of the mutations on performance in probabilistic reversal or discounting was limited. CONCLUSIONS: These results show discrete functions of dopamine signaling in control of reinforcement learning. Mutations in either dopaminergic or dopaminoceptive neurons selectively affected conditioned reinforcement or associative learning, respectively.