Beau, Maxime J.R; (2025) Dissecting cerebellar computation using high-density extracellular electrophysiology during behaviour. Doctoral thesis (Ph.D), UCL (University College London).

Text Beau_10205137_Thesis_sigs_removed.pdf Access restricted to UCL open access staff until 1 March 2026. Download (131MB)

Abstract

Understanding motor control is a fundamental goal in neuroscience. The cerebellum is a key brain region involved in motor control, but the computational role of the cerebellar circuitry remains poorly understood. Classifying neurons into distinct cell-types makes the challenge of understanding neural function more tractable by enabling targeted investigations. In this thesis, I report experimental and analytical developments I performed in high-density electrophysiology and optogenetics to enable cell-type classification and yield novel insights into the function of core cell types of the cerebellar circuit. I revealed insights into how information is transmitted between Purkinje cells, the sole output of the cerebellar cortex, and neurons in the deep cerebellar nuclei, the sole output of the cerebellum, in mice performing motor tasks. I also study the teaching signals conveyed by climbing fibres during reward processing, which might support learning in motor and non-motor tasks. First, I pioneered the use of Neuropixels probes in the cerebellum, and also tested the second generation of Neuropixels probes in combination with optogenetic stimulation of genetically defined cell types. To analyze Neuropixels data, I developed the first Python package which provides a toolkit for processing Neuropixels recordings, NeuroPyxels. Second, I combined optogenetic tagging with Neuropixels recordings and pharmacology to assemble a ground truth library of the major cell types of the cerebellar cortex: mossy fibres, Golgi cells, molecular layer interneurons, and Purkinje cell (PC) simple and complex spikes. This library enabled me to build a deep learning classifier with semi-supervised strategies capable of predicting the identity of these five cell types with high accuracy in mice and monkeys. This approach provides a general roadmap for cell-type identification in extracellular recordings across the brain. Third, I investigated long-range information transmission in the cerebellum by simultaneously recording PCs and their monosynaptically connected downstream targets, the cerebellar nuclear cells (NCs), in mice engaging in motor behaviours. By recording from neighbouring PCs, I showed that PC populations exhibit high levels of millisecond-precise synchrony in awake mice. To detect monosynaptically connected PC-NC pairs in these recordings, I developed a general method to identify putative monosynaptic connections in extracellular recordings of synchronous neural populations. I showed that PCs can drive oscillatory spiking in downstream NCs. I also found that converging PCs can inhibit NCs to saturating levels and inhibit NCs supralinearly if they fire within ~1 ms of each other, suggesting that PC synchrony can play a role in strengthening corticonuclear inhibition. Finally, I used Neuropixels recordings in a sensorimotor integration task to examine how climbing fibre inputs to the cerebellum convey reward-reactive and predictive signals during operant and classical conditioning, revealing parallels between reinforcement learning and the teaching signals broadcast to the cerebellum from the inferior olive. Together, these experiments leverage state-of-the-art high-density electrophysiology and machine learning to shed light on the neural basis of cerebellar function. My results suggest that millisecond synchrony is an important determinant of signalling from the cerebellar cortex to their downstream targets in vivo and that climbing fibre input could convey signals related to reinforcement learning.

Download activity - last month

Export as XMLCSVJSON

Download activity - last 12 months

Export as CSVXMLJSON

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item