Cholinergic and dopaminergic prediction error coding in associative learning

01/30/2020

Bálint Király^1, 2, Vivien Pillár¹, Balázs Hangya¹

¹ Lendület Laboratory of Systems Neuroscience, Department of Cellular and Network Neurobiology- Institute of Experimental Medicine, Budapest, Hungary

² Department of Biological Physics, Institute of Physics, Eötvös Loránd University, Budapest, Hungary

Acetylcholine and dopamine are neuromodulators that control the behaviour of diverse populations of neurons, influencing neural information processing at a variety of temporal scales. Specifically, neuromodulatory systems have a fundamental role in normal cognitive functions. Additionally, disorders of the¬¬¬ cholinergic and dopaminergic neurons underlie certain degenerative neurological conditions such as Alzheimer\’s and Parkinson’s diseases; therefore, a better understanding of these systems might lead to significant steps towards better treatments. Different neuromodulatory systems participate in overlapping cognitive processes and often represent similar behaviourally relevant variables (such as reward prediction error), hence their unique role in cognitive functions are not well characterized. To address this, it’s crucial to study them simultaneously under the same experimental conditions. In order to do this we crossed ChAT-Cre and DAT-Cre mice lines and used microdrives suitable for dual electrophysiological study and optogenetical identification of basal forebrain cholinergic and midbrain dopaminergic neurons. Next, we designed an auditory operant learning paradigm in which animals are required to form new auditory cue-outcome associations, and studied the firing activity of recorded neurons.Using an automatic clustering method, we found groups of neurons showing distinct characteristic firing patterns in the task. Across the two brain areas we found pairs of groups coding correlated information, including the cholinergic and dopaminergic groups. Our results suggest that both systems participate in prediction error coding but at different timescales: fast cholinergic firing following behavioral feedback was scaled with the absolute value of the prediction error while the somewhat slower dopaminergic response coded signed prediction error.