A framework for analyzing converging feedforward and cortical-bulbar feedback dynamics in target detection from complex odor scenes

该项目利用光学成像和光遗传策略的最新进展来监控工作中的大脑。具体地说，该项目专注于了解行为正常的小鼠嗅觉系统中上行和下行(反馈)活动模式之间的相互作用。在这里，研究人员不仅关注整合信息并将信息从鼻子传输到大脑的嗅觉模块，还确定了高级大脑区域，即嗅觉皮质，在循环的处理环路中如何相互作用。这一策略使研究人员能够评估感觉输入是如何通过反馈由大脑内部状态塑造的。此外，研究人员通过将尖端的实验方法--光学成像和光遗传策略--与新的计算模型结合起来，在两种方法的界面上工作，从而产生非互斥的可测试预测。调查者确定这些前馈-反馈环路是否有助于注意状态、气味识别的提取或与传入气味相关的预测和错误信号的广播。通过国际合作，实验技术与最先进的数据分析相辅相成，这些数据分析表征了沿着高维轨迹的神经元种群动力学，并测量了活动模式、特征时间尺度、模式相互作用和作为行为功能的协调的发生。此外，该项目为来自美国和罗马尼亚的学生和博士后实习生提供了通过参与国际合作来扩展他们的实验和计算技能的机会。系统神经科学的一个中心目标是根据控制它们的神经回路来描述行为。这对哺乳动物的大脑构成了严峻的挑战，因为行为被认为依赖于广泛分布的前馈以及自上而下的反馈神经表征，这些在技术上很难在大范围内监控，也很难在细胞分辨率下进行操作。该项目建立在首席研究员的最新实验结果和国际合作者开发的气味识别新算法的基础上。具体地说，该项目探索了嗅觉的精细结构，并测试了反馈服务于以下三种机制中的一种或多种的中心假设：预测性编码、吸引子生成或注意以增强行为相关刺激的辨别能力。在参与嗅觉辨别强迫选择任务和情境反转学习任务的小鼠中，a)皮质-球部反馈和b)嗅球输出神经元的动力学受到监测，反馈通过中间神经元间接作用于这些神经元，并随后受到细胞分辨率的调节。嗅球皮质反馈的可逆光遗传局部抑制与数百个神经元的同步双光子共振扫描成像(100赫兹)相结合。为了解决所提出的反馈作用，具体的实验设计与机器学习工具和动态系统分析相结合。

This project makes use of recent advances in optical imaging and optogenetic strategies to monitor the brain at work. Specifically, the project is focused on understanding the interplay between ascending and descending (feedback) activity patterns in the olfactory system of behaving mice. Here, the investigator does not simply focus on the olfactory sensory module that integrates and transmits information from the nose to the brain but determines how higher brain areas, namely, the olfactory cortex, interact in the recurrent processing loop. This strategy enables the investigator to evaluate how sensory inputs are shaped by internal brain states via feedback. Furthermore, the investigator works at the interface of two approaches by combining cutting-edge experimental approaches--optical imaging and optogenetic strategies-- with novel computational models that give rise to non-mutually exclusive testable predictions. The investigator determines whether these feedforward-feedback loops contribute to attention states, extraction of odor identity, or broadcasting of predictions and error signals related to the incoming odorants. Experimental techniques are complemented, through an international collaboration, with state-of-the-art data analysis that characterizes neuronal population dynamics along high-dimensional trajectories and measures occurrence of activity patterns, characteristic timescales, patterns interaction, and coordination as a function of behavior. Additionally, the project provides opportunities for students and postdoctoral trainees from the USA and Romania to expand their experimental and computational skills through their participation in the international collaboration.A central goal of systems neuroscience is to describe behaviors in terms of the neuronal circuits that control them. This constitutes a steep challenge in the mammalian brain, because behaviors are thought to rely on widely distributed feedforward, as well as top-down feedback neural representations, which are technically difficult to monitor at large scales and manipulate at cellular resolution. The project builds on recent experimental results from the lead investigator and novel algorithms for odor identification developed by the international collaborator. Specifically, the project probes the fine structure of olfactory perception and tests the central hypothesis that feedback serves one or more of the following three mechanisms: predictive coding, attractor generation, or attention to enhance the discriminability of behaviorally relevant stimuli. The dynamics of: a) cortical-bulbar feedback, and b) olfactory bulb output neurons on which feedback acts indirectly via interneurons are monitored and subsequently modulated with cellular resolution in mice engaged in olfactory discrimination forced-choice tasks and contextual reversal learning tasks. Reversible optogenetic local suppression of cortical feedback in the olfactory bulb is combined with simultaneous two-photon resonant scanning imaging (100 Hz) of hundreds of neurons. To address the proposed feedback roles, specific experimental design is combined with machine learning tools and dynamical systems analysis.