Mid-lateral cerebellar complex spikes encode multiple independent reward-related signals during reinforcement learning

Reward-driven cerebellar climbing fiber activity influences both neural and behavioral learning

The cerebellum plays a key role in motor coordination and learning. In contrast with classical supervised learning models, recent work has revealed that CFs can signal reward-predictive information in some behaviors. This raises the question of whether CFs may also operate according to principles similar to those described by reinforcement learning models. To test how CFs operate during reward-guided behavior, and evaluate the role of reward-related CF activity in learning, we have measured CF responses in Purkinje cells of the lateral cerebellum during a Pavlovian task using 2-photon calcium imaging. Specifically, we have performed multi-stimulus experiments to determine whether CF activity meets the requirements of a reward prediction error (rPE) signal for transfer from an unexpected reward to a reward-predictive cue. We find that once CF activity is transferred to a conditioned stimulus, and there is no longer a response to reward, CFs cannot generate learned responses to a second conditioned stimulus that carries the same reward prediction. In addition, by expressing the inhibitory opsin GtACR2 in neurons of the inferior olive, and optically inhibiting these neurons across behavioral training at the time of unexpected reward, we find that the transfer of CF signals to the conditioned stimulus is impaired. Moreover, this optogenetic inhibition also impairs learning, resulting in a deficit in anticipatory lick timing. Together, these results indicate that CF signals can exhibit several characteristics in common with rPEs during reinforcement learning, and that the cerebellum can harness these learning signals to generate accurately timed motor behavior.

Cerebellar climbing fibers signal flexible, rapidly adapting reward predictions

Classical models of cerebellar computation posit that climbing fibers (CFs) operate according to supervised learning rules, correcting movements by signaling the occurrence of motor errors. However, recent findings suggest that in some behaviors, CF activity can exhibit features that resemble the instructional signals necessary for reinforcement learning, namely reward prediction errors (rPEs). Despite these initial observations, many key properties of reward-related CF responses remain unclear, thus limiting our understanding of how they operate to guide cerebellar learning. Here, we have measured the postsynaptic responses of CFs onto cerebellar Purkinje cells using two-photon calcium imaging to test how they respond to learned stimuli that either do or do not predict reward. We find that CFs can develop generalized responses to similar cues of the same modality, regardless of whether they are reward predictive. However, this generalization depends on temporal context, and does not extend across sensory modalities. Further, learned CF responses are flexible, and can be rapidly updated according to new reward contingencies. Together these results suggest that CFs can generate learned, reward-predictive responses that flexibly adapt to the current environment in a context-sensitive manner.

Consensus Paper: Cerebellum and Reward

Cerebellum is a key-structure for the modulation of motor, cognitive, social and affective functions, contributing to automatic behaviours through interactions with the cerebral cortex, basal ganglia and spinal cord. The predictive mechanisms used by the cerebellum cover not only sensorimotor functions but also reward-related tasks. Cerebellar circuits appear to encode temporal difference error and reward prediction error. From a chemical standpoint, cerebellar catecholamines modulate the rate of cerebellar-based cognitive learning, and mediate cerebellar contributions during complex behaviours. Reward processing and its associated emotions are tuned by the cerebellum which operates as a controller of adaptive homeostatic processes based on interoceptive and exteroceptive inputs. Lobules VI-VII/areas of the vermis are candidate regions for the cortico-subcortical signaling pathways associated with loss aversion and reward sensitivity, together with other nodes of the limbic circuitry. There is growing evidence that the cerebellum works as a hub of regional dysconnectivity across all mood states and that mental disorders involve the cerebellar circuitry, including mood and addiction disorders, and impaired eating behaviors where the cerebellum might be involved in longer time scales of prediction as compared to motor operations. Cerebellar patients exhibit aberrant social behaviour, showing aberrant impulsivity/compulsivity. The cerebellum is a master-piece of reward mechanisms, together with the striatum, ventral tegmental area (VTA) and prefrontal cortex (PFC). Critically, studies on reward processing reinforce our view that a fundamental role of the cerebellum is to construct internal models, perform predictions on the impact of future behaviour and compare what is predicted and what actually occurs.

Purkinje cells in Crus I and II encode the visual stimulus and the impending choice as monkeys learn a reinforcement based visuomotor association task

Visuomotor association involves linking an arbitrary visual cue to a well-learned movement. Transient inactivation of Crus I/II impairs primates' ability to learn new associations and delays motor responses without affecting the kinematics of the movement. The simple spikes of Purkinje cells in the Crus regions signal cognitive errors as monkeys learn to associate specific fractal stimuli with movements of the left or right hand. Here we show that as learning progresses, the simple spike activity of individual neurons becomes more selective for stimulus-response associations, with selectivity developing closer to the appearance of visual stimuli. Initially, most neurons respond to both associations, irrespective of the identity of the stimulus and the associated movement, but as learning advances, more neurons distinguish between specific stimulus-hand associations. Using a linear decoder, it was found that in early learning stages, the visual stimulus can be decoded only when the choice can also be decoded. As learning improves, the visual stimulus is decoded earlier than the choice. A simple model can replicate the observed simple spike signals and the monkeys' behavior in both the early and late learning stages.

The Role of the Cerebellum in Learning to Predict Reward: Evidence from Cerebellar Ataxia

Recent findings in animals have challenged the traditional view of the cerebellum solely as the site of motor control, suggesting that the cerebellum may also be important for learning to predict reward from trial-and-error feedback. Yet, evidence for the role of the cerebellum in reward learning in humans is lacking. Moreover, open questions remain about which specific aspects of reward learning the cerebellum may contribute to. Here we address this gap through an investigation of multiple forms of reward learning in individuals with cerebellum dysfunction, represented by cerebellar ataxia cases. Nineteen participants with cerebellar ataxia and 57 age- and sex-matched healthy controls completed two separate tasks that required learning about reward contingencies from trial-and-error. To probe the selectivity of reward learning processes, the tasks differed in their underlying structure: while one task measured incremental reward learning ability alone, the other allowed participants to use an alternative learning strategy based on episodic memory alongside incremental reward learning. We found that individuals with cerebellar ataxia were profoundly impaired at reward learning from trial-and-error feedback on both tasks, but retained the ability to learn to predict reward based on episodic memory. These findings provide evidence from humans for a specific and necessary role for the cerebellum in incremental learning of reward associations based on reinforcement. More broadly, the findings suggest that alongside its role in motor learning, the cerebellum likely operates in concert with the basal ganglia to support reinforcement learning from reward.

Reinforcement Motor Learning After Cerebellar Damage Is Related to State Estimation

Recent work showed that individuals with cerebellar degeneration could leverage intact reinforcement learning (RL) to alter their movement. However, there was marked inter-individual variability in learning, and the factors underlying it were unclear. Cerebellum-dependent sensory prediction may contribute to RL in motor contexts by enhancing body state estimates, which are necessary to solve the credit-assignment problem. The objective of this study was to test the relationship between the predictive component of state estimation and RL in individuals with cerebellar degeneration. Individuals with cerebellar degeneration and neurotypical control participants completed two tasks: an RL task that required them to alter the angle of reaching movements and a state estimation task that tested the somatosensory perception of active and passive movement. The state estimation task permitted the calculation of the active benefit shown by each participant, which is thought to reflect the cerebellum-dependent predictive component of state estimation. We found that the cerebellar and control groups showed similar magnitudes of learning with reinforcement and active benefit on average, but there was substantial variability across individuals. Using multiple regression, we assessed potential predictors of RL. Our analysis included active benefit, somatosensory acuity, clinical ataxia severity, movement variability, movement speed, and age. We found a significant relationship in which greater active benefit predicted better learning with reinforcement in the cerebellar, but not the control group. No other variables showed significant relationships with learning. Overall, our results support the hypothesis that the integrity of sensory prediction is a strong predictor of RL after cerebellar damage.

A cerebro-cerebellar network for learning visuomotor associations

Consensus is rapidly building to support a role for the cerebellum beyond motor function, but its contributions to non-motor learning remain poorly understood. Here, we provide behavioral, anatomical and computational evidence to demonstrate a causal role for the primate posterior lateral cerebellum in learning new visuomotor associations. Reversible inactivation of the posterior lateral cerebellum of male monkeys impeded the learning of new visuomotor associations, but had no effect on movement parameters, or on well-practiced performance of the same task. Using retrograde transneuronal transport of rabies virus, we identified a distinct cerebro-cerebellar network linking Purkinje cells in the posterior lateral cerebellum with a region of the prefrontal cortex that is critical in learning visuomotor associations. Together, these results demonstrate a causal role for the primate posterior lateral cerebellum in non-motor, reinforcement learning.

Activating Lobule VI PCTH+-Med Pathway in Cerebellum Blocks the Acquisition of Methamphetamine Conditioned Place Preference in Mice

Cerebellum has been implicated in drug addiction; however, its underlying cellular populations and neuronal circuitry remain largely unknown. In the current study, we identified a neural pathway from tyrosine hydroxylase (TH)-positive Purkinje cells (PCTH+) in cerebellar lobule VI to calcium/calmodulin-dependent protein kinase II (CaMKII)-positive glutamatergic neurons in the medial cerebellar nucleus (MedCaMKII), forming the lobule VI PCTH+-MedCaMKII pathway in male mice. In naive male mice, inhibition of PCTH+ neurons activated Med neurons. During conditioned place preference (CPP) training, exposure to methamphetamine (METH) inhibited lobule VI PCTH+ neurons while excited MedCaMKII neurons in mice. Silencing MedCaMKII using a tetanus toxin light chain (tettox) suppressed the acquisition of METH CPP in mice but resulted in motor coordination deficits in naive mice. In contrast, activating lobule VI PCTH+ terminals within Med inhibited the activity of Med neurons and subsequently blocked the acquisition of METH CPP in mice without affecting motor coordination, locomotor activity, and sucrose reinforcements in naive mice. Our findings identified a novel lobule VI PCTH+-MedCaMKII pathway within the cerebellum and explored its role in mediating the acquisition of METH-preferred behaviors.

Sensorimotor-linked reward modulates smooth pursuit eye movements in monkeys

Reward is essential for shaping behavior. Using sensory cues to imply forthcoming rewards, previous studies have demonstrated powerful effects of rewards on behavior. Nevertheless, the impact of reward on the sensorimotor transformation, particularly when reward is linked to behavior remains uncertain. In this study, we investigated how reward modulates smooth pursuit eye movements in monkeys. Three distinct associations between reward and eye movements were conducted in independent blocks. Results indicated that reward increased eye velocity during the steady-state pursuit, rather than during the initiation. The influence depended on the particular association between behavior and reward: a faster eye velocity was linked with reward. Neither rewarding slower eye movements nor randomizing rewards had a significant effect on behavior. The findings support the existence of distinct mechanisms involved in the initiation and steady-state phases of pursuit, and contribute to a deeper understanding of how reward interacts with these two periods of pursuit.

Dynamic organization of cerebellar climbing fiber response and synchrony in multiple functional components reduces dimensions for reinforcement learning

Cerebellar climbing fibers convey diverse signals, but how they are organized in the compartmental structure of the cerebellar cortex during learning remains largely unclear. We analyzed a large amount of coordinate-localized two-photon imaging data from cerebellar Crus II in mice undergoing 'Go/No-go' reinforcement learning. Tensor component analysis revealed that a majority of climbing fiber inputs to Purkinje cells were reduced to only four functional components, corresponding to accurate timing control of motor initiation related to a Go cue, cognitive error-based learning, reward processing, and inhibition of erroneous behaviors after a No-go cue. Changes in neural activities during learning of the first two components were correlated with corresponding changes in timing control and error learning across animals, indirectly suggesting causal relationships. Spatial distribution of these components coincided well with boundaries of Aldolase-C/zebrin II expression in Purkinje cells, whereas several components are mixed in single neurons. Synchronization within individual components was bidirectionally regulated according to specific task contexts and learning stages. These findings suggest that, in close collaborations with other brain regions including the inferior olive nucleus, the cerebellum, based on anatomical compartments, reduces dimensions of the learning space by dynamically organizing multiple functional components, a feature that may inspire new-generation AI designs.

Mixed Selectivity in the Cerebellar Purkinje-Cell Response during Visuomotor Association Learning

Although the cerebellum has been traditionally considered to be exclusively involved in motor control, recent anatomic and clinical studies show that it also has a role in reward-processing. However, the way in which the movement-related and the reward-related neural activity interact at the level of the cerebellar cortex and contribute toward learning is still unclear. Here, we studied the simple spike activity of Purkinje cells in the mid-lateral cerebellum when 2 male monkeys learned to associate a right or left-hand movement with one of two visual symbolic cues. These cells had distinctly different discharge patterns between an overtrained symbol-hand association and a novel symbol-hand association, responding in association with the movement of both hands, although the kinematics of the movement did not change between the two conditions. The activity change was not related to the pattern of the visual symbols, the movement kinematics, the monkeys' reaction times, or the novelty of the visual symbols. The simple spike activity changed throughout the learning process, but the concurrent complex spikes did not instruct that change. Although these neurons also have reward-related activity, the reward-related and movement-related signals were independent. We suggest that this mixed selectivity may facilitate the flexible learning of difficult reinforcement learning problems.SIGNIFICANCE STATEMENT The cerebellum receives both motor-related and reward-related information. However, it is unclear how these two signals interact at the level of cerebellar cortex and contribute to learning nonmotor skills. Here we show that in the mid-lateral cerebellum, the reward information is encoded independently from the motor information such that during reward-based learning, only the reward information carried by the Purkinje cells inform learning while the motor information remains unchanged with learning.

Reward signals in the cerebellum: origins, targets, and functional implications

The cerebellum has long been proposed to play a role in cognitive function, although this has remained controversial. This idea has received renewed support with the recent discovery that signals associated with reward can be observed in the cerebellar circuitry, particularly in goal-directed learning tasks involving an interplay between the cerebellar cortex, basal ganglia, and cerebral cortex. Remarkably, a wide range of reward contingencies-including reward expectation, delivery, size, and omission-can be encoded by specific circuit elements in a manner that reflects the microzonal organization of the cerebellar cortex. The facts that reward signals have been observed in both the mossy fiber and climbing fiber input pathways to the cerebellar cortex and that their convergence may trigger plasticity in Purkinje cells suggest that these interactions may be crucial for the role of the cerebellar cortex in learned behavior. These findings strengthen the emerging consensus that the cerebellum plays a pivotal role in shaping cognitive processing and suggest that the cerebellum may combine both supervised learning and reinforcement learning to optimize goal-directed action. We make specific predictions about how cerebellar circuits can work in concert with the basal ganglia to guide different stages of learning.