Cholinergic neuronal responses to probabilistic outcome-predicting stimuli follow a weighed, unsigned prediction error model and anticipate behavioral responses

Basal forebrain cholinergic neurons (BFCNs) play an important role in associative learning, suggesting that BFCNs may participate in processing sensory stimuli that predict future outcomes. However, little is known about how BFCNs respond to outcome-predictive sensory cues and the impact of outcome probabilities on BFCN responses has not been explored. Therefore, we performed bulk calcium imaging and recorded spiking output of identified cholinergic neurons from the basal forebrain of mice performing a probabilistic Pavlovian cued outcome task that allowed us to control the predictive strength of cue stimuli. BFCNs responded strongly to sensory cues predicting likely reward, while little response was observed for cues that were rarely paired with reward. Reward delivery led to the activation of BFCNs, with less expected rewards eliciting a stronger response, while air puff punishments also evoked positive-going responses from BFCNs. We propose that BFCNs differentially weigh predictions of positive and negative reinforcement, reflecting divergent relative salience of forecasting appetitive and aversive outcomes, in accordance with a simple reinforcement learning model of a weighed, unsigned prediction error. Finally, the extent of cholinergic activation after cue stimuli predicted subsequent decision speed, suggesting that the expectation-gated cholinergic firing is instructive to reward-seeking behaviors.