Maximizing survival when hungry: neural mechanisms for computing behavioural priorities

Hunger is a potent internal drive that can significantly change an animal's behavioural priorities. For example, hungry animals favour actions that increase the chance of finding food, but this comes with an elevated risk of predation. Moreover, the expression of non-essential behaviours (e.g. reproduction) is down-regulated as an energy-conserving strategy. Remarkably, hunger can also substantially change the way an animal responds to environmental cues; when well-fed, an ambiguous stimulus might be perceived as a threat but with increased hunger this may be ignored or even considered as a possible food cue. How does an animal integrate all this information and reach a consensus decision about which action - from its full behavioural repertoire - to select, and thus maximize its survival?

These computations must be solved by key interactions between the brain circuits that control each distinct behaviour. However, to understand these interactions is challenging; it requires an extensive knowledge of each circuit and a means to monitor all of them across the brain at the same time. In the mammalian nervous system, this is not possible but simpler animals must solve the exact same problems using less complex nervous systems that are highly-accessible for interrogation.

Here, we will use a remarkably well-understood invertebrate system, Lymnaea, whose six principal behaviours (feeding, locomotion, reproduction, withdrawal, respiration, heart control) have been extensively characterized down to the level of the individual identified neurons that control them. As such, this provides the opportunity to monitor the key survival-linked decision-making events 'online' as the system processes information about both its internal hunger state and cues arising from the environment.

To achieve this, we will exploit the latest advances in behaviour and brain recording approaches. Specifically, behaviours will be monitored using new machine-learning algorithms that can track animal body-parts, postures and units of behaviour (eg. feeding events) automatically. We will assay brain activity using a novel fluorescence imaging microscope developed in-lab to monitor neurons across the nervous system down to single cell level. We will also exploit commercial solutions that allow 100s-1000s of neurons to be recorded simultaneously over long-time periods.

We will first establish how this animal encodes information about its hunger-state across all the behaviour-generating neural circuits in the brain and then determine how these circuits interact to decide which action to select. Subsequently, we will examine how neural circuits are re-tuned such that the same input can drive completely different behaviours when hungry versus when fed; this remarkable shift in the perceived meaning of an input is a highly-adaptive mechanism for adjusting risk to suit an animal's current situation, but the neural basis for it is poorly understood. Using real-world natural predator cues, we will also test how animals compute a decision when faced with two conflicting threats: predation versus starvation. This will provide insight into the fundamental neural mechanisms controlling an animal's most immediate survival-linked decisions.

This topic has increasing significance as animals start to face major alterations to their habitat and food availability due to climate change and urbanization. This proposal aligns directly with the BBSRC responsive mode priorities '3Rs' by using a non-'protected' invertebrate species, 'Food, Nutrition and Health' through identifying integral cellular and network mechanisms involved in metabolic regulation and 'Data driven Biology' through our deep-learning behaviour-tracking approaches and novel whole-CNS neuronal activity readout strategies. The outputs from this work, which aim to provide a fundamental understanding of survival-linked decision-making, also have relevance to 'Systems Approaches to the Biosciences'.

Hunger is a potent motivator, fundamentally changing an animal's behaviour to satisfy its nutritional demands. For example, when food-deprived, animals down-regulate non-essential behaviours and favour actions for localizing and consuming food, accepting an increased risk of predation or other types of harm. As part of any action selection process, perceptual decisions about the value of competing inputs must also be made. This large-scale reprioritization of an animal's full behavioural repertoire requires the coordination of multiple distinct neural circuits across the brain but how this is achieved remains unclear.

Examining the neural basis of action selection demands readout of whole-CNS activity, but also neuron-level analysis of the networks involved in each behaviour. In complex mammalian brains this is not realistic, but simpler animals solve the same problems, using accessible nervous systems that make a full interrogation of decision-making events possible.

Here, we will exploit Lymnaea, a powerful experimental system for circuit analysis, whose six principal behaviours have been extensively characterized down to the level of individual identified neurons. Applying the latest advances in deep-learning posture-tracking, multi-electrode recording and whole-CNS imaging we will determine how behavioural prioritization is computed according to motivational state and adaptively modulated to maximize survival. In particular, we will elucidate how distinct circuits encode hunger-state and how these circuits interact to reach a consensus decision about which behaviours to select. We will also determine how perception is altered, such that a single type of input can activate distinct behaviours according to changes in hunger state. Finally, we will elucidate how threat-conflict is resolved at the neuronal level, examining the decision-making events that allow animals to select appropriate actions when faced with threats from both predation and starvation.

Academic Community. Our research will focus on how the CNS selects appropriate actions from an animal's full behavioural repertoire to maximize survival. It will inform neuroscientists working on similar control circuits in other systems but also benefit the wider academic research community where knowledge of mechanistic principles of neural circuit regulation are important. Findings will be published in high-profile peer-reviewed journals (eg. Nature, Neuron, Nature Neurosci, Nature Comms, Science Advances) as we do currently, and disseminated at international meetings. Together, these benefits will enhance the knowledge economy starting in 1-5 years, with relevance for worldwide academic advancement. Additionally, the research plan will use new, innovative technical approaches - deep-learning behaviour tracking, new applications of commercial solutions for readout of neuronal populations, and in-lab developed methods for whole-brain imaging. These will be beneficial for driving advances in methodology and understanding in many fields of neuroscience-related research; potential recipients include other academic research institutes, both nationally and globally. The work will also deliver and train highly-skilled researchers (PDRA, PhD students, MSc students, UGs) with expertise in data organization, analysis, oral communication, and formal scientific writing skills, relevant to many employment sectors and thus further the knowledge economy. We will also foster interdisciplinary connections through local talks in other university departments (eg. School of Physics and Engineering, Sussex Innovation Centre). The timecourse of this benefit will start from 2-3+ years

Commercial Private Sector and Economy. The findings will reveal the highly-parsimonious neuronal strategies that Lymnaea uses to integrate information, resolve conflicts and make real-world survival-linked decisions. This may inform efficient design principles relevant to AI architecture, robotics and engineering. We already have close collaborations with computational and AI labs and links with industry (for example, Google) which could potentially be used to help realize such impact. The application of new technologies may also have relevance for Industry Partners. Another component of our research impact will be the application of powerful technologies for neuronal population readout. TB and KS have collaborated in the development of a novel 2-photon mesoscope based on a unique optical principal ('divergent beam optics'). The promise of this platform in enabling whole-CNS imaging on a very limited budget (~£1000) will be developed further through this project and is a possible focus of engagement with science microscopy companies (eg. Scientifica, based near Brighton), looking to develop better imaging platforms, thus potentially supporting growth of commercial private sector companies with international reach. The timecourse of this benefit will start from 3+ years.

Wider Public. We anticipate that the ideas emerging from this work will have impact for society (years 2+). We will highlight the challenges for animals in the natural world to make key survival-linked decisions, and their increasing importance as environments and food availability are influenced by climate change and urbanization. The 3Rs message will also be promoted; beneficiaries include organizations working to protect animals (eg. the RSPCA). Communicating ideas about the regulation of food-intake, a strong theme in our research, will allow us to emphasize the importance of healthy eating, a message with possible future impact through enhancing quality of life and health. These are accessible ideas and we will publicise them, along with general interest research findings, through the Sussex Press Office, open lectures and demonstrations (eg. Brighton Science Festival), open days, open-labs, school visits and sixth-form work experience. Benefits will start from the beginning of the grant.