Circuit Mechanisms Underlying Persistent Activity in a Neural Integrator

摘要 短期记忆功能通常是通过持续活动来支持的，持续的 记忆刺激抵消后神经元群体的反应。这种形式的 活动是包括导航、运动控制和决策在内的各种任务的基础。经典 机械论理论通过强烈假设的模型将这种活动理想化。 同质神经元群体，只编码一个变量，并完美地生成 稳定的活动模式。这与最近的研究结果形成了鲜明对比，这些研究表明，真实生物中的神经元 记忆网络表现出对多个刺激属性的多路编码，这些属性随时间变化 整个人群的反应，以及对上下文的依赖。在这里，我们将介绍电路 通过实验-理论结合研究这种多样性在功能中的机制和作用 接近。实验是在斑马鱼幼体凝视的短期记忆回路中进行的 一种控制系统，通过精确地将眼睛保持在视觉上来帮助稳定视力 目标。利用该系统的定量精度和实验可操作性， 我们结合了全电路、突触分辨率解剖和全电路记录 细胞分辨率下的活动扰动。在目标1中，我们将这些数据组合成一个模型 神经元与实验记录的神经元以一对一的方式映射的系统。这 使我们能够推断不同解剖学、基因型别、 和功能细胞类别，并形成对这些相互作用如何管理电路功能的预测。 在目标2中，我们使用3D细胞分辨率光学成像和刺激神经元的扰动 改进我们的模型并测试模型预测的活动。在目标3中，我们扩大了我们的能力以形成 通过开发和应用来精确描述细胞类内部和之间的相互作用 记忆回路中神经元的3D抑制。总之，这项工作有望极大地 扩大我们对持续性细胞类型多样性的电路机制和作用的理解 激发、短期记忆和运动控制。

Abstract Short-term memory function is commonly supported through persistent activity, the sustained response of populations of neurons following the offset of a memorized stimulus. This form of activity underlies diverse tasks including navigation, motor control, and decision-making. Classic mechanistic theories have idealized such activity through models that assume strongly homogeneous populations of neurons that encode only a single variable and generate perfectly stable patterns of activity. This contrasts with recent work showing that neurons in real biological memory networks exhibit multiplexed encoding of multiple stimulus attributes, temporally varying responses across the population, and context dependence. Here we address the circuit mechanisms and role of this diversity in function through a combined experimental-theoretical approach. Experiments are conducted in a short-term memory circuit of the larval zebrafish gaze control system that contributes to stable vision by precisely maintaining the eyes on a visual target. Taking advantage of the quantitative precision and experimental tractability of this system, we combine whole-circuit, synapse-resolution anatomy with circuit-wide recordings and perturbations of activity at cellular resolution. In Aim 1, we combine these data into a model of the system in which neurons map in a one-to-one manner with experimentally recorded neurons. This enables us to infer the interactions within and between neurons of different anatomical, genotypic, and functional cell classes and form predictions for how these interactions govern circuit function. In Aim 2, we use 3D cellular resolution optical imaging and stimulating perturbations of neuronal activity to refine our model and test model predictions. In Aim 3, we expand our capacity to form precise characterizations of within and between cell-class interactions by developing and applying 3D suppression of neurons across the memory circuit. Altogether, this work promises to greatly expand our understanding of the circuit mechanisms and role of cell type diversity in persistent firing, short-term memory, and motor control.