Matter of context: Revealing the circuit architecture of internal brain state influence on behaviour

我的目标是了解大脑内部正在进行的活动模式是如何塑造它处理信息和控制行为的方式的。人类和动物的行为不仅仅是一套“自动”反射。相反，我们对感官输入（如看到食物或听到电话铃声）的反应方式取决于多种环境因素，如情绪状态、一天中的时间以及我们的饱足程度或警觉程度。现代神经科学在理解报告这些因素的大脑系统方面取得了重要进展。例如，多巴胺能和血清素能系统作为奖励信号，由于它们在塑造正常行为和精神疾病方面的重要性，已经被广泛研究。然而，主要的挑战仍然是了解多种大脑通路如何共同作用来调节感觉处理和行为。在很大程度上，这是由于大脑的大小和复杂性，这使得无法同时测量涉及的许多脑细胞。通过在伦敦大学学院神经科学、生理和药理学系建立一个新的研究项目，我计划采取一种新的方法来解决这个问题。我的策略将最先进的成像技术与数据驱动生物学和计算建模相结合，这是一种实验上具有优势的模式生物——斑马鱼幼虫——这是BBSRC确定的关键研究途径。斑马鱼幼体特别适合同时追踪多个大脑结构的活动。这种微小的动物（3.5毫米长）几乎是完全透明的，当鱼执行一系列可识别的行为（如狩猎和躲避）时，可以使用荧光显微镜对其小大脑进行非侵入性监测。重要的是，由于鱼类与包括人类在内的所有其他脊椎动物共享大脑回路，许多这些行为受到诸如饥饿或警觉性等环境因素的影响。为了研究分布式大脑网络是如何共同作用来塑造行为的，我将使用尖端的“光片显微镜”来单独跟踪斑马鱼的8万个神经元的活动。在此过程中，我将改变环境因素来操纵饱腹感、警觉性和其他环境因素。破译由此产生的数据集将是一项复杂的工作，堪比通过同时倾听伦敦金融城（City of London） 10万名金融员工中的每一个人的声音，来提取对市场动态的洞察。有价值的见解的潜力是巨大的，但在理解大量数据和找到最丰富的信息来源方面的挑战也是巨大的。为了应对这一挑战，我将使用循环神经网络——一种类似于自动语音识别的现代机器学习算法。它将使我能够识别神经元，这些神经元可以预测动物是否可能对特定的视觉线索做出反应，甚至在刺激出现之前。这样的细胞是传递上下文信息的良好候选者，我的计算模型将解决它们如何协同工作以集体影响行为。为了验证我的假设，我将使用先进的“光遗传学方法”，利用光直接控制大脑活动，并检查由此产生的对大脑其他部位活动和鱼行为的影响。最终，这些发现将揭示与环境和经验相关的神经活动如何结合起来影响基本的大脑功能。因为所有脊椎动物都拥有相同的基本大脑结构，我的实验发现很可能揭示出适用于包括人类在内的许多物种的原理。因此，与BBSRC支持世界级健康基础生物科学的优先事项一致，该项目将为我们理解健康的大脑如何产生行为提供重大进展。从长远来看，这可能有助于更好地理解疾病期间大脑功能是如何被破坏的。

My aim is to understand how ongoing internal activity patterns within the brain shape the way it processes information and controls behaviour. Human and animal behaviour is not merely a set of 'automatic' reflexes. Rather, the way we respond to sensory inputs such as the sight of food or the sound of a phone ringing depends on multiple contextual factors such as emotional state, time of day and how satiated or alert we are. Modern neuroscience has made important progress towards understanding the brain systems that report these factors. For instance, the dopaminergic and serotonergic systems that signal reward have been extensively studied due to their importance in shaping normal behaviour as well as psychiatric disorders. However, major challenges remain in terms of understanding how multiple brain pathways act together to modulate sensory processing and behaviour. To a large extent this is due to the size and complexity of the brain which precludes simultaneous measurement of the many brain cells involved. By establishing a new research programme in the Department of Neuroscience, Physiology & Pharmacology at UCL, I plan to take a novel approach to tackle this problem. My strategy combines state-of-the-art imaging in an experimentally advantageous model organism - the larval zebrafish - with data-driven biology and computational modelling: key research avenues identified by the BBSRC.Zebrafish larvae are particularly well suited for simultaneously tracking activity in multiple brain structures. This tiny animal (3.5 mm long) is almost perfectly transparent, allowing its small brain to be monitored non-invasively using fluorescent microscopy while the fish performs a range of recognizable behaviours such as hunting and avoidance. Importantly, many of these behaviours are influenced by contextual factors such as hunger or alertness, by virtue of brain circuits fish share with all other vertebrates, including humans. To study how distributed brain networks work together to shape behaviour, I will use cutting-edge "light-sheet microscopy" to individually track the activity of each of the zebrafish's 80,000 neurons. While doing so, I will alter environmental factors to manipulate satiety, alertness and other contextual elements. Deciphering the resulting dataset will be a complex endeavour, comparable to extracting insights into market dynamics by simultaneously listening to each and every one of the 100,000 finance employees in the City of London. The potential for valuable insights is enormous, but so is the challenge in making sense of the massive amount of data and finding the most informative sources. To meet this challenge, I will use recurrent neural networks - a modern machine-learning algorithm akin to the one that powers automated speech recognition. It will enable me to identify neurons that can predict if the animal is likely to respond to a specific visual cue, even before the stimulus is presented. Such cells are good candidates for signalling contextual information and my computational modelling will resolve how they work together to collectively influence behaviour. To test my hypotheses, I will use advanced "optogenetic methods" to directly control brain activity using light and examine the resulting effects on activity elsewhere in the brain and on the behaviour of the fish.Ultimately, these findings will shed new light on how neural activity related to context and experience combine to influence fundamental brain function. Because all vertebrates possess the same basic brain plan, my experimental findings are likely to reveal principles that apply to many species, including humans. Thus, in line with the BBSRC's priority of supporting world-class basic bioscience for health, this project will provide a major advance in our understanding of how the healthy brain produces behaviour. In the longer term, this could underpin greater understanding of how brain function is disrupted during disease.