Roach EB, Hussain Shuler MG (2011) Cholinergic reinforcement and temporal learning in rodent visual cortex. Neuroscience 2011 Abstracts 608.16. Society for Neuroscience, Washington, DC.
Summary: The idea that neuromodulators act as reinforcement signals has an intricate scientific history, including the well-characterized analogue of prediction error relayed by midbrain dopamine neurons. Neuromodulators released from discrete nuclei are poised to broadcast to many brain regions at once, and so it is an appealing concept to investigate other neuromodulatory systems within a reinforcement learning framework. Reward timing activity, a neural reflection of operantly learned stimulus-reward intervals in the primary visual cortex (V1), offers a tractable in vivo model to examine the role of candidate neuromodulators in temporal reward learning. Reward timing was first characterized in rats trained to lick a delivery tube to receive water rewards, where stimulation to one eye indicated reward availability after x licks, while stimulation to the other eye required y licks. Simultaneously recorded activity in V1 indicated that single unit responses evolve from reporting only visual characteristics to showing persistent increased/decreased firing or peak activity corresponding to the time of anticipated reward. Individual neurons report one interval or the other, even those with binocular peri-stimulus responses, arguing that reward timing is learned locally within V1. Theoretical work suggests that the local expression of reward associated intervals requires an interaction between the visually-evoked network response and a reinforcement signal conveying the time of reward. Based on anatomical and neurophysiological evidence, we hypothesized that cholinergic input from the basal forebrain (BF) could provide such a reward signal to V1. To test its necessity, BF cholinergic innervation in V1 was lesioned — using the selective neurotoxin 192 IgG-saporin — prior to changing the experimental policy between cues and associated reward delays. This allowed an examination of two potential roles for BF cholinergic input: in expressing previously learned intervals and in acquiring information about new intervals. We found that neurons from saline-infused controls, but not lesioned animals, shifted as a population to report the new, behaviorally relevant intervals (Kolmogorov-Smirnov, p < 0.05). Importantly, neurons from lesioned animals continued to report the previously learned intervals, suggesting that BF cholinergic input is required to learn, but not express, reward timing. These results support the notion that acetylcholine released from BF afferents acts as a reinforcement signal that guides cortical network plasticity.
Related Products: 192-IgG-SAP (Cat. #IT-01)