Prospective and retrospective learning in orbitofrontal cortex

PROJECT SUMMARY Animals learn to predict rewards to maximize their fitness. Rewards often follow environmental cues and/or actions performed by animals. How do animals learn and remember the associations between cues/actions and salient outcomes such as rewards? One possibility is that they remember prospective associations, i.e., how often does reward follow a specific cue/action? Another possibility is that they remember retrospective associations, i.e., how often does a specific cue/action precede reward? Though these possibilities may sound similar, they are in fact dissociable, and likely have different behavioral functions. The common view of learning and memory is that animals only acquire prospective associations. Accordingly, it is well known that many neurons in the brain encode prospective associations. Nevertheless, whether the brain also stores retrospective associations was unknown. We recently showed that different neuronal subpopulations in the mouse ventral/medial orbitofrontal cortex (vmOFC), a key regulator of reward learning, encode prospective and retrospective associations between a cue and reward. Along with other recent findings, this demonstrated that the brain stores both prospective and retrospective associations in memory. Nevertheless, how these memories manifest in neuronal activity during the course of learning and whether these memories are encoded by different neural circuits to differentially control behavior are unknown. Here, we propose to investigate this overarching question through three specific aims. First, how do neurons acquire activity representing prospective and retrospective cue-reward associations? Is such activity acquired in distinct directions or the same direction for the two associations (i.e., forward from cue to reward and/or backward from reward to cue)? We will address this question using two-photon calcium imaging to longitudinally track activity of the same neurons over days of behavioral learning. This will allow us to study activity acquisition of individual neurons over behavioral learning. These experiments will determine whether blocking specific directions of learning may block the formation of specific types of memories. Second, can cue-action-reward learning differentially recruit prospective or retrospective memories in vmOFC based on the training history of an animal? We will address this question by biasing animals to either a prospective or a retrospective learning strategy during behavioral training. Throughout such learning, we will record the activity of the same vmOFC neurons. These experiments will determine whether different individuals can acquire different memories during the same behavior. Lastly, do different vmOFC output circuits to the ventral tegmental area (VTA) or the nucleus accumbens (NAc), both key regulators of learning, encode distinct prospective or retrospective memories to differentially control behavior? We will study this question using projection-specific imaging and optogenetics to test the causal function of these different circuits. Overall, these experiments will uncover novel mechanisms by which OFC circuits regulate reward learning.