Decision Making and Learning in Neuronal Networks

An important and achievable objective of 21st century neuroscience is to explain how the brain generates behaviour and moreover how it makes choices between alternative courses of future action. Indeed decision-making is surely one of the most important and conspicuous of the functions of the brain. The brain is continually orchestrating a series of decisions; choosing which particular action or behaviour to initiate next - to go to the library, eat lunch, to walk to the station or to avoid a threat, for instance. Clearly some of these choices are influenced primarily by high level cognitive processes involving the rational evaluation of the consequences of alternative courses of action. We also make lower level decisions without much rational thought: we eat when hungry for example. Sensory stimuli can also affect choice and decision-making without the need for conscious reasoning: in response to a threatening stimulus we may stop eating and withdraw from danger, for example. An important feature of such decisions is that they are not fixed or strictly reflexive in nature but rather are subject to the influence of experience. So for example we may learn from experience that a threat is not so threatening after all, and decide to continue eating in spite of it. How switching between different behaviours is achieved in neuronal terms remains unresolved, especially in complex mammalian systems in which cognitive processes are involved. The problem is that there is a wide gap between our current understanding of behavioural decision-making on the one hand and the neuronal mechanisms underlying it on the other. Occupying this conceptual gap are enormous numbers of neurons organized into hyper-complex networks. Ultimately it is these neural networks - their activity, the intrinsic properties of their components, interactions and susceptibility to experience - that are responsible for adaptive behavioural decisions. But the complexity of neural networks in the mammalian brain defies attempts at detailed analysis of their operation. In systems involving billions of neurons it is simply not possible to investigate neural networks with single neuron resolution. But this is the level of detail required to derive fundamentally new insights into how neural circuits are functionally configured to make decisions. We aim to achieve a step-change in our understanding of behavioural choice by substantially reducing the difficulty by conducting our experiments on an invertebrate model system, namely the mollusc Lymnaea stagnalis. Our subject possesses a relatively simple brain with thousands, not billions of neurons. Nonetheless, although its decisions are not cognitive, it has a substantial behavioural repertoire, makes adaptive choices between alternative behaviours and is able learn from experience. Hence Lymnaea is complex enough to be instructive and interesting, yet simple enough to be susceptible to a detailed, multi-neuron systems-level analysis. In short, Lymnaea represents an ideal model system for advancing our understanding of how decision-making neural circuits are configured. By analyzing the neuron-by-neuron circuitry involved in decision making in this system, we will make a major contribution to understanding fundamental design principles of neural networks underlying decision making in more complex systems in which single neuron resolution is far more difficult.

To understand how the CNS of the mollusk Lymnaea makes adaptive behavioural choices we will analyse underlying neural networks while behaviours are performed in semi-intact preparations. We will focus primarily on feeding and withdrawal - two behaviours that are already well characterized at the level of neuronal circuitry. Initially we will analyse the functions of neurons that coordinate the interactions between the two networks since these are likely to be key components of the decision-making circuitry. This basic characterization of the decision-making circuitry (Objective 1) will be accompanied by an analysis of how learning influences decision-making (Objective 2). Our experiments will involve integrating intracellular microelectrode recording from known identified neurons, up to 6 simultaneously, with multi-electrode array (MEA) recording from many more neurons widely distributed in the CNS. We have developed techniques that allow us to alter intrinsic neuronal properties while behaviours are on-going in semi-intact preparations. This 'active observer' strategy will employ novel dynamic-clamp microelectrode technology and stimulation through selected channels of the MEA. The influence of experience on decision making will be investigated using a number of associative learning paradigms that we have developed. These include food reward and taste aversion conditioning paradigms, which in combination will allow us to investigate how choices within a single network (feeding) are affected. We will also use conditioning that reverse behavioural decisions; turning an initially appetitive stimulus leading to feeding into an aversive stimulus triggering withdrawal. These experiments will allow us to analyse how decisions between the feeding and withdrawal networks are modified by experience. A benefit of our neuron resolution analysis of decision-making will be a better understanding of lower resolution data from vertebrate studies.

If the grant is awarded our research will benefit (outside the academic research community) the a) public sector, b) the commercial private sector, and c) the wider public - as indicated below: a) Impact on Public Sector: The benefit in this sector is likely to be on public health and will arise from the long-term impact of our work on understanding how the brain makes decisions and implements actions. There are a number of conditions such as depression, Parkinson's and dementia that affect our intentionality and thence our ability to implement actions. At the fundamental level of neural networks and neurotransmitters this is a very poorly understood area of significant clinical importance. Indeed considerable costs are associated with less than fully effective treatments for these conditions. Our research, using a model invertebrate system, will answer some of the most fundamental questions about the brain's control of actions - what neural architectures are involved, which neurotransmitter systems are involved? This research, carried out at a high level of neuronal resolution, will help the interpretation of much lower resolution information available to clinicians dealing with affected patients. Of long-term clinical benefit will be our research resolving the role of monoamines on the effects of learning in decision-making neural circuits. This is potentially of significant benefit in the health sector because abnormal monoamine signalling is though to underlie a number of disabling conditions affecting intentionality and monoamine systems are the target for intervention and treatment. b) Impact on Commercial Private Sector: Practical benefits to industry at the interface between the biomedical sciences and the physical sciences will include new design principles for robotic devices, insights and methods regarding the functional organization of complex technological networks and new tools for the management of large-scale multi-level datasets. The pharmaceutical sector for example will benefit from the clearer understanding how molecular signalling systems alter the properties of neural networks in the context of memory formation. The company Eli Lilly for instance has funded a project in our laboratory investigating the regulation of cholinergic function in the Lymnaea CNS. More recently the following companies have expressed an interest in our work on simple model nervous systems and are potential collaboration partners in our proposed project: Gedeon Richter Pharmaceuticals (for in vitro electrophysiological pre-screening of candidate compounds of memory enhancer drugs) and in the IT sector, NaturalMotion Ltd, NeuroRobotice Ltd, MultiChannelSystems MCS GmbH. c) Impact on Wider Public and Issues of Public Concern: We already contribute to the public understanding of science through engagement with local School 6th forms, popular articles and books explaining neuroscience research. Also our research on invertebrate model systems addresses an area of considerable public concern, namely animal experimentation and the use of animals in bioscience research. Our research reduces the need for animal experiments because we use an invertebrate model of more complex and sentient organisms. By conducting our experiments on an organism not covered by the A(SP)A 1986, our research will have a positive impact on the principles of the 3Rs - Replacement, Refinement and Reduction, principles adopted by all research councils and major charitable funding bodies.