Long-range neuronal projections: circuit blueprint or stochastic targeting? Rigorous classification of brain-wide axonal reconstructions

摘要(项目摘要) 长期以来，哺乳动物大脑中神经元的分类一直是深入研究的焦点 神经科学。神经元被广泛认为是神经的基本计算元素 它们的形态、生理和分子特性的广泛多样性可能提供 对它们的功能和参与疾病的关键洞察力。尤其是长程轴突投射。 网络连接的典型决定因素，提供了蜂窝组织之间的关键联系 和电路架构。汇聚显微成像、基因标记和 直到最近，算法开发才使大规模、全脑的高通量收集成为可能 大脑主动性下的轴突重建。使用原则性统计战略，拟议的项目 利用这种信息丰富的资源，从大脑的每个区域严格识别所有的“投射神经元 类型“，具有客观上不同的解剖靶向模式。因此，此应用程序将直接测试 开创性的假设是，单个神经元的轴突轨迹遵循特定的协调计划，如 反对在地区联系的限制下随意瞄准。虽然这种数据驱动的分类 必须取决于现有的数字跟踪，我们将部署我们的完整分析工作流作为自动化 公共云服务器上的管道，不仅允许免费的社区访问，还允许不断完善 随着更多的数据集变得可用，由此产生的知识。此外，我们的方法允许量化 估计每一类神经元的种群大小及其唯一的路径分布 从SOMA到每个目标的距离，定义了信息传输的基本时间动力学。我们 也将确定不同类型的投射神经元在树突形态上是否不同，提供了重要的 有关输入处理是否针对预期输出进行了专门调整的线索。此外，我们还将延长 这一创新的方法将其应用于由随机核酸获得的互补数据集 条形码、激光捕获显微解剖和测序，产生了更大的样本大小，以换取 解剖分辨率较低。最后但同样重要的是，我们将把发现的轴突投影模式建模为 一种新的人工神经网络设计(“投影网”)，系统地探索它们可能的选择性 在学习和记忆的稳健性和性能方面的优势。因此，实现这些目标将量化 第一次，相关的单神经元基序勾勒出哺乳动物中枢的功能蓝图 神经系统和相关的损伤，为公共健康的长期利益。

ABSTRACT (PROJECT SUMMARY) The classification of neurons in the mammalian brain has long been a focus of intensive investigation in neuroscience. Neurons are widely recognized as the fundamental computational elements of the nervous system, and the broad diversity of their morphological, physiological, and molecular properties may provide crucial insights into their function and involvement in disease. Long-range axonal projections, in particular, are the quintessential determinants of network connectivity, providing a key nexus between cellular organization and circuit architecture. Converging technological breakthroughs in microscopic imaging, genetic labeling, and algorithmic development have only recently enabled the high-throughput collection of large-scale, whole-brain axonal reconstructions under the BRAIN initiative. Using a principled statistical strategy, the proposed project leverages such information-rich resources to rigorously identify, from each brain region, all “projection neuron types” with objectively distinct patterns of anatomical targeting. This application will thus directly test the seminal hypothesis that the axonal trajectories of individual neurons follow specific coordination plans as opposed to aiming randomly within the constraints of regional connections. While this data-driven classification necessarily depends on the existing digital tracings, we will deploy our full analysis workflow as an automated pipeline on public cloud servers, allowing not only free community access, but also continuous refinement of the resulting knowledge as more datasets become available. Moreover, our approach allows the quantitative estimation of the population size of every separate neuron class as well as its unique distribution of path distances from the soma to each target, defining the basic temporal dynamics of information transmission. We will also determine if different projection neuron types vary in their dendritic morphology, providing an important clue as to whether input processing is specifically tuned for the intended outputs. Furthermore, we will extend this innovative methodology by applying it to complementary datasets obtained by stochastic nucleic acid barcoding, laser capture microdissection, and sequencing, yielding far greater sample sizes in exchange for lower anatomical resolution. Last but not least, we will model the discovered axonal projection patterns into a novel artificial neural network design (“projectron nets”) to systematically explore their possible selective advantages in learning and memory robustness and performance. Achieving these goals will thus quantify, for the first time, the relevant single-neuron motifs to outline the functional blueprint of the mammalian central nervous system and related impairments for the long-lasting benefit of public health.