Re-examining the role of cerebellar climbing fibers in motor learning

Duke University, Durham NC

Investigators

Linked publications, trials & patents

Abstract

The cerebellum plays a key role in motor learning, and cerebellar damage can result in a wide range of motor deficits. However, despite decades of research designed to reveal how neural circuits in the cerebellum enable motor learning, it has remained unclear whether common learning mechanisms are engaged to enable learning across different behavioral paradigms. The leading model of cerebellar learning is based on the premise that climbing fiber inputs to the cerebellar cortex can instruct motor output through a supervised learning rule by signaling motor errors. However, many motor learning regimes do not allow for this type of deterministic instructional signals required for supervised learning. Instead, the correct association between sensory input and motor output must be learned through experience. I propose that a reinforcement learning model may provide a more robust explanation of cerebellar learning that accounts for the diverse relationship between sensory inputs and motor output encountered across behavioral regimes. To address whether the climbing fiber system operates according to a reinforcement learning model, I have developed a voluntary motor learning paradigm for head-fixed mice. In this paradigm, the mouse must make a forelimb movement with correct timing in relationship to a sensory cue in order to receive reward. When the sensory cue arrives with predictable timing, mice learn to adjust their motor output to anticipate the sensory cue. In this proposal, I will use modern optical techniques to record and manipulate climbing fiber activity to determine how it drives motor learning. In Aim 1 I will use a behavior based approach in conjunction with two forms of in vivo calcium imaging to test what information is carried by the climbing fibers during a behavioral regime in which no learning can occur. If cerebellar learning operates according to a reinforcement model, then this activity should reflect sensorimotor events that are predictive of task outcome. In Aim 2, I will test whether climbing fiber activity in response to task events observed in Aim 1 are innate or acquired when the animals initially learn the task. This will be done by monitoring climbing fiber activity in mice which are naïve to the sensorimotor association. In a reinforcement model, this association must be acquired, and therefore should be absent in naïve mice. Finally, in Aim 3 I will use an optogenetic strategy to test the role of climbing fiber activity in instructing changes in motor output during a behavioral regime in which learning can occur. These experiments will provide key evidence to evaluate the validity of the reinforcement model of cerebellar learning. This will be an important step in our understanding of how cerebellar circuits contribute to voluntary, associative sensorimotor learning.