Synaptic and Circuit Mechanisms of Olfactory Processing

这个项目旨在填补我们在算法层面上对神经计算的理解与我们的神经网络之间的差距。 在细胞机制水平上理解相同的计算-即，突触，通道，以及 单个神经元之间的连接。对简单网络的研究在填补这一空白方面尤其有力。在这 建议，我们将集中在两个相对简单的网络，果蝇触角叶和侧角。触角 叶是嗅觉系统的第一个脑区，侧角是接收大部分嗅觉信号的脑区。 触角叶轴突投射。触角叶是一个有用的模型，用于研究细胞机制的两个 基本的神经计算、增益控制和时间滤波。侧角是研究非线性动力学的一个有用的模型。 模式识别的细胞机制-在这种情况下，触角叶肾小球的活动模式。我们将 研究三个主要问题，每个问题都与这些网络中的细胞元素和计算之间的关系有关。首先，为什么触角叶中的抑制性中间神经元如此多样？这些中间神经元选择性地对气味浓度的增加或减少（ON或OFF细胞）或特定的气味脉冲重复率（快或慢细胞）做出反应，并且它们也对不同范围内的气味浓度敏感。为了验证不同的中间神经元具有不同的计算功能的假设，我们将使用大规模连续切片EM结合体内电生理学和光遗传学。第二，侧角神经元如何对嗅球的空间进行采样？每个侧角神经元平均从大约4个肾小球（总共50个肾小球）接收前馈兴奋，但突触后侧角神经元的总数远远小于可能的肾小球组合的数量，这就提出了一个问题，即在每个侧角中以刻板的方式连接在一起的肾小球组合可能有什么特别之处。为了检验存在强的统计学差异来控制哪些肾小球连接在一起的假设，我们将使用2 P介导的已识别肾小球的光遗传学刺激，结合来自突触后侧角神经元的全细胞记录。第三，侧角神经元如何整合它们的突触输入？如果模式识别涉及多个非线性步骤，它可能会更强大，但我们不知道这种非线性实际上会在细胞机制水平上实现。为了验证侧角突触整合涉及多个非线性元件的假设，我们将使用体内电压成像和体内全细胞记录，以及有针对性的突触抑制扰动。总的来说，这些研究应该大大促进我们对基本神经计算基础的细胞和突触构建块的理解。我们的长期目标是在多个粒度级别上对这些简单网络有一个相当完整的理解。我们希望我们的发现能够产生可检验的假设和概念方法，这将加速理解更复杂网络的进展。

This project aims to fill the gap between our understanding of neural computations at the algorithmic level and our understanding of the same computations at the level of cellular mechanisms – i.e., synapses, channels, and patterns of connectivity between individual neurons. Studies of simple networks are especially powerful in filling this gap. In this proposal, we will focus on two relatively simple networks, the Drosophila antennal lobe and lateral horn. The antennal lobe is the first brain region of the olfactory system, and the lateral horn is the brain region that receives the majority of antennal lobe axonal projections. The antennal lobe is a useful model for studying the cellular mechanisms of two fundamental neural computations, gain control and temporal filtering. The lateral horn is a useful model for studying the cellular mechanisms of pattern recognition – in this case, patterns of activity across antennal lobe glomeruli. We will investigate three main questions, each relating to the relationship between cellular elements and computations within these networks. First, why are inhibitory interneurons in the antennal lobe so diverse? These interneurons respond selectively to odor concentration increases or decreases (ON or OFF cells), or particular odor pulse repetition rates (fast or slow cells), and they are also sensitive to odor concentration over different ranges. To test the hypothesis that different interneurons have distinct computational functions, we will use large-scale serial section EM in combination with in vivo electrophysiology and optogenetics. Second, how do lateral horn neurons sample the space of olfactory glomeruli? Each lateral horn neuron receives feedforward excitation from ~4 glomeruli on average (out of 50 glomeruli in total), but the total number of postsynaptic lateral horn neurons is much smaller than the number of possible glomerular combinations, raising the question of what might be special about the glomerular combinations that actually wire together in a stereotyped fashion in every lateral horn. To test the hypothesis that there are strong statistical regularities governing which glomeruli wire together, we will use 2P-mediated optogenetic stimulation of identified glomeruli, in combination with whole cell recordings from postsynaptic lateral horn neurons. Third, how do lateral horn neurons integrate their synaptic inputs? Pattern recognition can be more powerful if it involves multiple nonlinear steps, but we do not know such nonlinearities would actually be implemented at the level of cellular mechanism. To test the hypothesis that lateral horn synaptic integration involves multiple nonlinear elements, we will use both in vivo voltage imaging and in vivo whole cell recordings, together with targeted perturbations of synaptic inhibition. As a whole, these studies should substantially advance our understanding of the cellular and synaptic building-blocks underlying fundamental neural computations. Our long-term goal is to develop a fairly complete understanding of these simple networks on multiple levels of granularity. We expect our discoveries to yield testable hypotheses and conceptual approaches which will accelerate progress in understanding more complex networks.