The role of cerebellum in dopamine neuron reward prediction error coding

Making and evaluating predictions is a fundamental feature of brain function. All animals learn from experience to optimize their behaviour for survival, yet we don't know how key 'prediction nodes' in the brain talk to each other, or indeed what the conversation is. Consider the familiar jingle of an ice cream van; our brains have learnt through experience that this sound signals the availability of ice cream and how good the treat will be, guiding our decision to approach, or avoid it. The cerebellum and ventral tegmental area (VTA) are critical hubs for prediction, but until now have been considered as independent systems. New evidence reveals prominent connections from the cerebellum to the VTA that can influence reward-seeking behaviour and thus 1) redefine our view of the function of these systems and 2) indicate that predictions may be coordinated throughout the brain. The goal of this proposal is to provide a detailed circuit-level characterisation of the role of the cerebellum in the encoding of reward prediction error in the VTA. To achieve this, we will first map the organisation of cerebellar input onto dopaminergic and non-dopaminergic neurons within the VTA. Second, we will define the nature of the cerebellar signal to the VTA by recording from projection neurons during reward prediction behaviour. Third, we will selectively disrupt these inputs during behaviour to establish how the cerebellum shapes reward prediction error coding in the VTA. This work is essential to develop an integrated understanding of predictive coding and reward processing in the brain. Our findings will further the development of biologically plausible learning algorithms, and identification of targets for the maintenance and restoration of reward processing systems during ageing.

We will combine anatomical mapping with advanced in vivo recording in wild-type and genetically modified mice to determine how the cerebellum influences encoding of reward predictions in VTA. In the first objective, we will determine the identity of cerebellar-recipient neurons in VTA (i.e. which cells receive cerebellar input) by injecting anterograde-acting viral tracers and opsins into the cerebellar nuclei. We will perform stereological counts to quantify cellular targets in VTA and use in vitro patch clamp and in vivo juxtacellular recordings to characterise the synaptic properties and functional identity of cerebellar-recipient VTA neurons. In the second objective, we will characterise the functional activity of VTA-projecting neurons in the cerebellum (i.e. what information is conveyed from cerebellum to VTA). To do this, we will transduce channelrhodopsin in VTA-targeting cerebellar neurons via retrograde-acting viral vectors and make electrophysiological recordings from 'opto-tagged' neurons in cerebellar nuclei during performance of an associative learning task. In the final objective, we will define the influence of cerebellum on reward prediction error coding in VTA (i.e. how does cerebellum contribute to VTA predictive processing). We will transduce inhibitory opsins into the cerebellar nuclei and locally silence axon terminals in VTA during specific components of an associative learning task. We will directly test the hypothesis that cerebellum contributes timing and error signals important for the prediction and evaluation of rewards by VTA dopamine neurons.

Our aim is to uncover mechanisms of reward-based learning and predictive processing in the brain. These are fundamental biological processes that govern much of human behaviour and thus have enormous long-term consequences for individuals and society. The following groups will be directly impacted by this proposal:

The global research community will benefit from the generation of new conceptual knowledge. By publishing in high profile journals and presenting at international conferences, we will enhance the excellence of UK science. As part of the proposal we will be generating very large volumes of anatomical, functional and behavioural data. We will make this data freely available, providing a new resource for quantitative and computational neuroscientists. Determining the mechanisms of reward/prediction processing in model systems will be important for the life sciences industry in developing rational therapeutic strategies for a range of mental health disorders. Investigation into how the cerebellum shapes the firing of dopamine neurons may well yield potential targets for therapeutic intervention (e.g. for addiction or Parkinson's disease). Open access publications and participation at national/international scientific meetings will ensure that the benefits of this research are realised.

The named postdoc will benefit from training in cutting-edge recording technologies and analysis that will position him to apply for independent positions and thus build capacity of in vivo skills in the UK.

Dissemination of our work will promote the public's understanding of, and engagement with, science. We will explain the latest research from this project to school children at visits and open days in a framework of "how we learn" and to other members of the public (e.g. at the Bristol Neuroscience Festival http://www.bristol.ac.uk/neuroscience/bnf/). The team will regularly give talks locally (e.g. Bristol neuroscience or computational neuroscience seminars) and at departmental seminars at other universities. As such this work will be of benefit to local colleagues and collaborators as well as those further afield.