A recurrent circuit links antagonistic cerebellar modules during associative motor learning

Abstract

The neural architecture of the cerebellum is thought to be specialized for performing supervised learning: specific error-related climbing fiber inputs are used to teach sensorimotor associations to small ensembles of Purkinje cells located in functionally distinct modules that operate independently of each other in a purely feedforward manner. Here, we test whether the basic operation of the cerebellum complies with this basic architecture in mice that learned a simple sensorimotor association during eyeblink conditioning. By recording Purkinje cells in different modules and testing whether their responses rely on recurrent circuits, our results reveal three operational principles about the functional organization of the cerebellum that stand in stark contrast to the conventional view: (1) Antagonistic organization, (2) Recurrent network dynamics, and (3) Intermodular communication. We propose that the neural architecture of the cerebellum implements these three operational principles to achieve optimal performance and solve a number of problems in motor control.

Neuronal dynamics of cerebellum and medial prefrontal cortex in adaptive motor timing

“…Previous studies have suggested that the CR-related activities of PCs in animals transitioning from DEC to TEC paradigms share remarkable similarities and can be explained by a common inverse firing rate model 6 . However, more recent research challenges this notion, revealing that both facilitatory and suppressive modulations occur in PCs during DEC 50 . Consistently, we observed both task-related facilitation and suppression in the IpN neurons of both DEC- and TEC-trained mice.…”

Emergence of syntax and word prediction in an artificial neural circuit of the cerebellum

“…Our cANN contains three layers-an input layer (granule cells), a middle layer (Purkinje cells), and an output layer (nuclei neurons) -and consists of a conventionally reported feedforward pathway and a recently identified recurrent pathway that starts from the Purkinje cells (Fig. 1b) [48][49][50][51][52][53][54] . We designed the inputs/outputs of the cANN (Fig.…”

“…Our cANN is highly consistent with a wealth of anatomical and physiological knowledge. While conventional local circuit models of the cerebellum are feedforward circuits [60][61][62][63][64][65] , recent reports have demonstrated the existence and functional significance of a recurrent connection starting from the Purkinje cells [48][49][50][51][52][53][54][66][67][68] . Since both next-word prediction and syntactic processing are inaccurate in a cANN without the recurrent pathway (data not shown) and conventional feedforward models (Supplementary Fig.…”

“…To replicate the local circuitry of the cerebellum, the cANN comprised three layers-an input layer (granule cells), a middle layer (Purkinje cells), and an output layer (cerebellar nuclei neurons)with a recurrent pathway from the output layer to the input layer (Fig. 1b Ohmae et al, 2021). The next-word prediction circuit was designed according to previous proposals, where words are fed sequentially into the circuit then, at each step, the circuit outputs a prediction of the next word, and information about the correct word information is given used as a teaching signal to improve future predictions.…”

Mesoscale simulations predict the role of synergistic cerebellar plasticity during classical eyeblink conditioning

“…Whereas PCs in the downbound microzones show prominent velocity-related responses, those in the adjacent upbound microzones show an increase in activity that is antagonistic to the response of the downbound PCs and is more proportional to the position of the eyelid during the CR [28]. Whether the downbound and upbound PCs correspond perfectly or only grosso-modo to the zebrin-negative (Z-) and zebrin-positive (Z+) microzones, respectively, remains to be shown [26,[28][29][30].…”