Automatic 3D Quantification of Synapse Distribution in Complex Dendritic Arbor

描述(申请人提供)：突触的亚细胞分布对神经系统的组装、功能和可塑性至关重要，并在神经系统的紊乱中发挥作用。然而，潜在的分子机制在很大程度上仍不清楚。虽然先进的多维图像与单细胞基因技术相结合，为在新的水平上理解突触的发育提供了前所未有的机会，但我们从大量三维图像中有效量化亚细胞突触的能力存在知识缺口。这是一个重大的问题，并阻碍了对突触发育的分子机制的大规模研究，特别是在具有复杂树枝的神经元--如哺乳动物的浦肯野细胞和果蝇的小叶板切向细胞(LPTC)--现有的方法不能对整个树突树产生完整或可靠的突触量化，也不能扩展到有效的遗传筛选。该项目的目标是通过提供工具，利用三维显微镜图像定量研究亚细胞突触的分布及其分子机制，从而弥合这一差距。具体地说，我们高度跨学科的团队将追求两个目标：(1)开发自动算法来分析和量化具有复杂树枝的神经元的整个树突树中突触的分布。突触密度的全面和客观描述将使突变模式的自动检测成为可能。(2)开发自动算法来分析和量化具有复杂树枝的神经元树的不同部分的突触分布。在不同的亚细胞位置进行有效的定量将有助于发现不同亚细胞部分的新调控因子。作为测试案例，我们将使用果蝇LPTC神经元的突触分布，它既适用于全基因组遗传筛选，也适用于单神经元分辨率的遗传操作。我们将开发可靠的方法从三维荧光共聚焦图像中表征抑制性GABA能和兴奋性胆碱能突触的密度。我们的算法将导致对控制抑制性和兴奋性突触的亚细胞分布的下一层次的机械理解，并使具有类似复杂性的其他类型的神经元能够进行广泛的定量分析。强大的多通道协同分析和机器学习方法将用于改进突触检测和亚细胞室提取，以克服3D共焦图像中的挑战，包括染色伪影和各向异性分辨率。算法将使用模型指导的方法来开发，该方法强调在遗传筛查期间对大容量3D图像的效率。将使用模式识别方法来加快对突触量化结果的校对。一种新的排序策略将适用于复杂树突状茎的神经元，以一种有功能意义的方式对亚细胞突触进行量化。该项目将生产一套开源、可扩展的工具，用于自动突触量化和校对，并具有友好的图形用户界面，以服务于神经科学界。

DESCRIPTION (provided by applicant): The subcellular distribution of synapses is critical for the assembly, function, and plasticity of the nervous system and plays a role in its disorders. Underlying molecular mechanisms, however, remain largely unknown. While advanced multidimensional images, in conjunction with single-cell genetic techniques, have afforded an unprecedented opportunity to understand synapse development at a new level, there is a knowledge gap in our capacity to effectively quantify subcellular synapses from large quantities of three-dimensional images. This is a significant problem and has hampered large-scale studies of the molecular mechanisms of synapse development, especially in neurons with complex arbor-such as Purkinje cells in mammals and lobula plate tangential cells (LPTC) in Drosophila-where existing approaches do not yield complete or robust synapse quantification for the entire dendritic tree and do not scale to efficient genetic screening. The objective of thi project is to bridge this gap by providing tools for quantitative investigation of subcellular synapse distribution and its molecular mechanisms using three-dimensional microscopy images. Specifically, our highly cross- disciplinary team will pursue two aims: (1) Develop automatic algorithms to analyze and quantify synapse distribution in the entire dendritic tree of neurons with complex arbor. Holistic and objective description of synapse density will enable automatic detection of mutant patterns. (2) Develop automatic algorithms to analyze and quantify synapse distribution in different parts of the entire dendritic tree of neurons with complex arbor. Efficient quantification at distinct subcellular locations will assist discovery of novel regulators for different subcellular parts. As a test case, we will use synapse distribution n Drosophila LPTC neurons, which are amenable to both genome-wide genetic screens and genetic manipulations with single-neuron resolution. We will develop reliable methods to characterize the density of inhibitory GABAergic and excitatory cholinergic synapses from three-dimensional fluorescence confocal images. Our algorithms will lead to the next level of mechanistic understanding that controls the subcellular distribution of inhibitory and excitatory synapses, and enable a wide range of quantitative analyses for other types of neurons with similar complexity. Powerful multichannel co-analysis and machine learning approaches will be used to improve synapse detection and subcellular compartment extraction for overcoming challenges in 3D confocal image, including staining artifacts and anisotropic resolution. Algorithms will be developed using a model-guided methodology that emphasizes efficiency for large volume 3D images during genetic screening. Pattern-recognition methods will be used to speed up proofreading of the synapse quantification results. A novel ordering strategy will be adapted for neurons of complex dendritic arbor to quantify subcellular synapses in a functionally meaningful way. The project will produce a set of open-source, extensible tools for automatic synapse quantification and proofreading, with friendly graphical-user interfaces, to serve the neuroscience community.