NetClamp: conducting neural network rhythms with mathematics

我们所有的行为，从认识朋友，到记得把车停在哪里，到泡一杯茶，都是大脑中神经元网络协调行为的结果。每种行为都是由这些网络中特定的电活动模式定义的。理解这些模式是如何产生的是神经科学的关键问题之一。解决这个问题需要实验室神经科学家和理论神经科学家之间的合作。实验室里的神经科学家确定了单个神经元是如何电工作的，并确定了神经元是如何相互交流的。他们绘制出连接图，表示哪些神经元影响其他神经元的电活动。以理论为基础的神经科学家利用这些信息构建神经网络活动的数学描述，称为“模型”。模型预测网络中的连接图如何决定电活动的模式。他们表明，地图上的细微变化会对网络活动产生巨大影响。例如，活动可以从看似随机的转变为高度协调的，神经元活动的节奏变化类似于足球场的墨西哥波。在大脑的某些区域，协调的节奏是健康的。例如，我们的呼吸是由一个具有高度同步活动的大脑区域控制的。在其他情况下，如癫痫，高度同步性可导致癫痫发作。通过揭示连接图如何影响网络活动，数学模型可以成为神经科学家工具箱中有用的一部分。只有当我们能够证明它们与真实神经网络的实验结果相匹配时，模型才有用。这需要改变神经元之间的连接图。为此，在实验室工作的神经科学家可以打开和关闭神经元群，并且可以以相当粗糙的方式阻断网络中的大部分连接。然而，目前还没有实验方法以测试模型所需的微妙方式改变连接。这个项目是关于建立新技术来解决这个问题。为了做到这一点，我们将把最近开发的实验工具和模型集成到一个系统中。实验工具允许我们用数码相机测量神经元的电活动，并通过在每个神经元上照射特定颜色和强度的光来改变这种活动。我们系统的关键组成部分是，我们将这些工具与神经元之间连接图的数学模型结合起来。该模型将使用相机记录的电活动来计算网络中的每个神经元应该根据连接图接收哪些输入信号。然后，它将控制一个照明系统，该系统将发出光模式，将计算信号传递给每个神经元。通过这种方式，我们将能够以与数学模型相同的方式操纵神经元之间的连接图。该系统将使实验室神经科学家能够以一种全新的方式进行实验。他们将能够探索神经元之间的交流如何影响网络活动，与基于理论的神经科学家在模型中享受的自由相同。他们将能够直接测试有关连接图如何塑造活动模式的理论。这将是理解大脑如何创造行为的重要一步。我们的系统还将促进智能植入物的发展，以治疗帕金森症和癫痫等脑部疾病，这些疾病的特征是异常的网络节律。当药物无法改变这些节律时，医生可能会转向彻底侵入性的医疗程序，永久性地改变网络结构。然而，如果正确的连接可以在正确的时间改变，更微妙的治疗可能就足够了。未来的智能植入物将检测到异常活动何时开始，然后将光线照射到特定的神经元上，修改它们的连接，以恢复正常、健康的活动。

All of our behaviours, from recognising friends, to remembering where we parked our car, to making a cup a tea, result from the coordinated behaviour of networks of neurons in our brain. Each behaviour is defined by a specific pattern of electrical activity in these networks. Understanding how these patterns are generated is one of the key problems in neuroscience.Solving this problem requires cooperation between lab-based, and theory-based neuroscientists. Lab-based neuroscientists determine how individual neurons work electrically and determine how neurons communicate with each other. They create maps of connections which represent which neurons influence the electrical activity of which other neurons.Theory-based neuroscientists use this information to construct mathematical descriptions of neural network activity, called 'models'. Models predict how the map of connections in a network determines the patterns of electrical activity. They show that subtle changes in the map can have large effects on network activity. For example, activity can switch from seemingly random to highly coordinated, with rhythmic changes in neuron activity resembling a Mexican wave in football stadiums. In some brain regions, coordinated rhythms are healthy. For example, our breathing is controlled by a brain area with highly synchronised activity. In other contexts, such as epilepsy, high levels of synchrony can lead to seizures. By uncovering how maps of connections affect network activity, mathematical models can be a useful part of the neuroscientist toolkit. Models are only useful if we can show that they match results from experiments with real neural networks. This requires altering the map of connections between neurons. To this end, lab-based neuroscientists can turn on and off groups of neurons, and can block large fractions of the connections in a network in rather crude ways. However, there is currently no experimental way to alter the connections in the subtle way required for testing models.This project is about building new technology to address this. To do this, we will integrate recently developed experimental tools with models in a single system. The experimental tools allow us to measure electrical activity in neurons using a digital camera, and to alter this activity by shining light of specific colour and intensity on each neuron. The key component of our system is that we will combine these tools with a mathematical model of the connection map between neurons. The model will use the camera recordings of electrical activity to calculate what input signals each neuron in the network should receive according to the map of connections. It will then control an illumination system, which will shine light patterns to deliver the computed signals to each neuron. In this way, we will be able to manipulate the map of connections between neurons in the same way as our mathematical model.This system will enable lab-based neuroscientists to do experiments in a radically new way. They will be able to explore how communication between neurons affects network activity with the same freedom that theory-based neuroscientists enjoy with models. They will be able to directly test theories about how connection maps shape patterns of activity. This will be a significant step towards understanding how the brain creates behaviour.Our system will also enable the development of smart implants to treat brain diseases such as Parkinson's and epilepsy, which are characterised by abnormal network rhythms. When drugs fail in changing these rhythms, doctors may turn to drastically invasive medical procedures that permanently alter network structures. However, if the right connections can be altered at the right time, more subtle therapies might suffice. Future smart implants will detect when abnormal activity starts, then shine light to specific neurons to modify their connections to restore normal, healthy activity.