Dopaminergic regulation of spatial learning

摘要 在以连接权重存储信息的神经网络中，在敏感度之间存在权衡 和稳定性。连接必须是塑料的，才能包含新信息，但如果塑料太好，则应存储 信息可能会被破坏。因此，如果大脑中的学习速度由一种 “何时学习”的信号随着当前新信息的可获得性而变化。在奖励学习中， 众所周知，多巴胺通过对奖赏的反应迅速上调突触可塑性来发挥这一功能 预测错误。这一建议的主要假设是，多巴胺还提供了一种学习的时间 空间学习的信号。在空间学习过程中，当有机体处于 在太空中移动。因此，我们假设空间学习是由多巴胺释放调制的，即 特别是与活跃的运动有关。这一想法之所以吸引人，是因为它可以解释为什么会这样 许多多巴胺神经元与运动是时间锁定的。这份提案概述了三个项目，都侧重于 果蝇大脑中空间导航的主要中枢--中央复合体的空间学习。在……里面 每个项目都有来自果蝇连接体的解剖学证据，这暗示着多巴胺的作用 神经元。此外，在每个项目中，已经有证据表明有问题的多巴胺神经元是活跃的。 当苍蝇在移动时。这激发了我们的假设，即多巴胺将运动与空间学习联系起来。 尽管这些项目在概念上是联系在一起的，但它们各自都专注于一种不同的多巴胺细胞类型，以及一种不同的 空间学习的形式。首先，我们将确定多巴胺如何调节空间位置线索的学习。 在头部方向系统中。其次，我们将研究多巴胺调节学习的假设。 关于头部方向系统中的旋转速度提示。第三，我们将研究一种假设 反馈电路集成了随时间变化的信息，以消除环境风向变化对 头部方向神经元。在所有三个项目中，我们都使用连接组分析和计算建模来 对大脑中的特定网络产生可测试的预测。然后，我们在活体内测试这些预测 当苍蝇在虚拟现实环境中导航时，钙成像和/或电生理学。我们的结果应该是 阐明了所有复杂物种中潜在的导航行为的基本机制，包括 环吸引子网络、Hebbian学习规则和反馈回路。广义地说，我们认为多巴胺 提供一个控制旋钮，用于调节这些机构的升降。因此，我们看到多巴胺能神经元 作为综合理解复杂认知过程中的网络动力学的切入点。

Summary In neural networks that store information in their connection weights, there is a tradeoff between sensitivity and stability. Connections must be plastic to incorporate new information, but if they are too plastic, stored information can be corrupted. Therefore, it would be useful if learning rates in the brain were regulated by a “when-to-learn” signal that varies with the current availability of new information. In reward learning, dopamine is known to serve this function, by rapidly upregulate synaptic plasticity in response to reward prediction errors. The overarching hypothesis of this proposal is that dopamine also provides a when-to-learn signal for spatial learning. During spatial learning, new information is generally available when an organism is moving through space. Thus, we hypothesize that spatial learning is modulated by dopamine release that is specifically linked to active movements. This idea is attractive because it can provide an explanation for why so many dopamine neurons are time-locked to movements. This proposal outlines three projects, all focusing on spatial learning in the central complex, the primary center for spatial navigation in the Drosophila brain. In each project, there is anatomical evidence from the Drosophila connectome that implies a role for dopamine neurons. Moreover, in each project, there is already evidence that the dopamine neurons in question are active when the fly is locomoting. This motivates our hypothesis that dopamine links movement to spatial learning. Although these projects are linked conceptually, they each focus on a distinct dopamine cell type, and a distinct form of spatial learning. First, we will determine how dopamine modulates learning about spatial position cues in the head direction system. Second, we will investigate the hypothesis that dopamine modulates learning about rotational velocity cues in the head direction system. Third, we will investigate the hypothesis that a feedback circuit integrates information over time to discount the influence of environmental wind shifts on head direction neurons. In all three projects, we use connectome analyses and computational modeling to generate testable predictions about specific networks in the brain. Then, we test these predictions using in vivo calcium imaging and/or electrophysiology as flies navigate in virtual reality environments. Our results should shed light on the fundamental mechanisms underlying navigation behaviors in all complex species, including ring attractor networks, Hebbian learning rules, and feedback loops. Broadly speaking, we think that dopamine provides a control knob for modulating these mechanisms up or down. As such, we see dopaminergic neurons as an entry point for an integrative understanding of network dynamics during complex cognitive processes.