A neuronal network generating flexible locomotor behaviour in a simple vertebrate: studies on function and embryonic self-assembly

神经系统如何让动物行为？从人脑中大量的神经元（超过20亿）来看，这里的挑战是显而易见的。不太明显的是神经系统结构的微小规模，许多神经元的直径只有0.01毫米。大小和复杂性的问题导致了对蜗牛和鱿鱼等简单动物的研究，这些动物具有复杂的行为，但神经元数量较少，通常较大。神经系统的另一个显著特征是，它们必须在胚胎发育期间迅速自我组装，以允许有助于生存的早期反应。他们是怎么做到的？最近关于发育的遗传学工作强调了基本特征，这些特征是线虫、果蝇和像我们这样的脊椎动物等各种动物所共有的。对胚胎的详细研究表明，脊椎动物的神经系统有一个共同的计划，特别是在像脊髓这样的核心部分。因此，我们可以研究神经系统如何在最简单的脊椎动物中发育和发挥作用。我们研究的是刚孵化的2天大的青蛙蝌蚪。虽然只有5毫米长，神经元不到2000个，但它们在被触摸时会游泳，在被捕食者抓住时会挣扎，当它们撞到东西时会停下来并附着：这些行为有助于生存。利用早期蝌蚪神经系统的简单性和我们设计的新方法，我们已经记录了大多数类型的神经元控制运动的活动，现在有一个独特的详细图片的神经元回路的游泳和挣扎。在计算机科学家和数学家的合作下，我们建立了这些电路的简化模型，并揭示了关键的操作原理。这项理论工作强调了从蜗牛到哺乳动物控制运动的神经元回路的共性，但也揭示了我们知识的空白。除了询问早期神经元回路是如何工作的，我们还想知道它们是如何发展的。我们发现蝌蚪神经元之间的连接不是很特异。广义上说，它们的神经纤维只是在生长过程中接触它们遇到的神经元。由于对感官刺激做出反应的神经元位于顶部，而控制肌肉的神经元位于底部，因此靠近顶部生长的神经元将连接到靠近底部生长的神经元。我们的数学模型表明，非常简单的规则可以指导神经纤维生长，形成神经元回路，当受到刺激时，能够产生游泳活动。在这项研究中，我们将询问早期蝌蚪神经系统的有序结构是否允许功能性神经元回路响应于已知控制神经纤维沿着和围绕神经系统生长的3种化学梯度进行自组装。为了回答这个问题，我们需要更多的电子记录来确定不同的皮肤刺激是如何启动游泳和挣扎的。我们还需要关于每种神经元类型的形态学的更详细的信息。我们的研究应该揭示神经系统如何“决定”启动运动。使用形态学信息，我们将建立一个神经元生长的数学模型，以生成不同类型的神经元相互之间的突触连接，从而自组装神经元回路。使用生理信息，我们将建立模型，其中不同的神经元连接到功能电路中，我们可以刺激这些功能电路，以发现它们如何产生产生游泳和挣扎的神经元活动。我们的最终目标是看看简单的生长规则是否可以允许神经元网络的自组装，这些神经元网络可以“决定”何时以及如何对感官刺激做出反应，表现得像蝌蚪一样。通过制作一个“虚拟蝌蚪”，它的运动由我们的网络控制，我们实际上可以观察它们产生的行为。如果成功，我们的研究将为理解更成熟、更复杂的神经系统控制运动的方式及其发展奠定基础。

How do nervous systems allow animals to behave? The challenge here is immediately clear from the vast numbers of neurons (over 2 billion) in a human brain. Less obvious is the minute scale of nervous systems construction with many neurons only 0.01 mm in diameter. Problems of size and complexity have led to the study of simpler animals like snails and squid which have complex behaviour but many fewer, often larger, neurons. A further remarkable feature of nervous systems is that they must self-assemble rapidly during embryonic development to allow early responses that aid survival. How do they do this? Recent genetic work on development has emphasised fundamental features, common to animals as diverse as nematode worms, fruit flies and vertebrates like us. Detailed research on embryos has shown that vertebrate nervous systems share a common plan particularly in core parts like the spinal cord. We can therefore investigate how nervous systems develop and function in the simplest vertebrates. We study just-hatched, 2 day old frog tadpoles. While only 5 mm long and with less than 2000 neurons, they will swim when touched, struggle when grasped by a predator, and stop and attach when they bump into things: behaviour that aids survival. Exploiting the simplicity of the early tadpole nervous system and new methods that we devised, we have recorded activity from most types of neuron controlling movement and now have a uniquely detailed picture of the neuronal circuits for swimming and struggling. In collaboration with computer scientists and mathematicians, we built simplified models of these circuits and uncovered key principals of operation. This theoretical work emphasised commonality in neuronal circuits controlling movement, from snails to mammals, but revealed gaps in our knowledge. As well as asking how early neuronal circuits work we also want to know how they develop. We found that connections between tadpole neurons are not very specific. Broadly, their nerve fibres simply contact the neurons they encounter as they grow. Since neurons responding to sensory stimuli lie at the top and neurons controlling muscles lie at the bottom, those growing near the top will connect to different neurons to those growing near the bottom. Our mathematical models showed that very simple rules could direct nerve fibre growth to form neuronal circuits able to generate swimming activity when stimulated. In this study we will ask whether the ordered structure of the early tadpole nervous system allows functional neuronal circuits to self-assemble in response to 3 chemical gradients known to control growth of nerve fibres along and around the nervous system. To answer this question we need many more electrical recordings to establish exactly how swimming and struggling are initiated by different skin stimuli. We also need more detailed information on morphology for each neuron type. Our study should reveal how nervous systems 'decide' to initiate movement. Using the morphological information we will build a mathematical model of neuron growth to generate the synaptic connections that different types of neuron make with each other to self-assemble neuronal circuits. Using the physiological information we will build models where the different neurons are connected into functional circuits which we can stimulate to find how they generate the neuronal activity that produces swimming and struggling. Our ultimate aim is to see if simple growth rules can allow the self-assembly of neuronal networks which can 'decide' when and how to respond to sensory stimuli, behaving like a tadpole. By making a 'virtual tadpole' whose movements are controlled by our networks we can actually watch them producing behaviour. If successful, our study will lay a foundation for understanding the way more mature, complex nervous systems control movements and how they develop.