Population-level modulation of Purkinje cell complex spikes during behavioural adaptation to imposed sensorimotor error

Overview & Synopsis

The cerebellum's role in motor control and learning, as proposed by the Marr-Albus-Ito theory, emphasises its involvement in error detection and the refinement of motor commands using internal models. The classical model posits that the presence of sensorimotor error is signalled by complex spikes (CS), which drive long-term depression (LTD) at parallel fibre-Purkinje cell (PF-PC) synapses. However, a binary representation of error does not adequately explain observed behavioural adaptation. Established cerebellum-dependent learning paradigms consistently demonstrate trial-by-trial motor adaptation which requires information about error direction and size. We found populations of Purkinje cells (PC), in spatially-organised parasagittal bands, that exhibit synchronised CS modulation, encoding the sign and the magnitude of sensorimotor error. The finding of bidirectional CS modulation provides a plausible loss function to guide supervised learning. The established cerebellar-learning paradigms are not compatible with rodents and thus preclude the use of state-of-the-art genetic tools and recording techniques. To bridge this, we introduce the joystick paradigm, a novel behavioural task in rodents. Animals were trained to position a lick-port by controlling a joystick, and responded to perturbations of the lick-port introduced by the experimenter. We used 2-photon scanning laser microscopy (2PSLM) to concurrently record CS activity in PC expressing GCaMP8m while animals adapted their behaviour in response to novel sensorimotor dynamics.