CellGraph: bioimaging software to rapidly identify bacterial genes responsible for modifications to host cell organelle morphology

摘要 结核病(TB)几千年来一直是人类的传染性疾病，而结核分枝杆菌(Mtb)再次成为单一感染源导致的头号死因。紧张的6至10个月的多药治疗过程，加上可能出现的不良副作用，是患者依从性和治疗完成的主要障碍。随之而来的多药耐药结核病(MDR-TB)和广泛耐药结核病(XDR-TB)病例的增加要求我们增加有效药物的储备，并呼吁开发新的治疗方法。在几千年的时间里，宿主和病原体已经进化出极大地影响感染结果的机制和关系。了解这些进化的相互作用及其对病原体清除或宿主病理的影响将引导合理开发有利于宿主保护性反应的新疗法。这些宿主导向疗法最近显示了针对结核分枝杆菌的有希望的结果，增强了现有抗分枝杆菌药物的累积效果或直接减少了细菌的复制。然而，我们对导致细菌生长增加或宿主免疫逃避的宿主细胞-病原体相互作用的了解有限，因此，缺乏机制知识阻碍了识别新型宿主导向药物的靶点的能力。目前识别结核分枝杆菌毒力因子和成像宿主细胞效应的方法缓慢而费力，通常无法同时将多个因素联系起来。 通过使用高通量、大规模的计算管道，我们可以快速有效地检测病原体感染期间宿主细胞细胞器形态的变化。海洋分枝杆菌是一种生物安全的2级细菌，会在鱼类和两栖动物中引起类似结核病的病理，并被用作结核分枝杆菌的替代品，以研究感染过程的各个方面。这个名为CellGraph的框架将量化Z堆叠显微镜图像和数字视频中细胞器的形状、数量和空间分布的变化，提高我们对细胞机制的理解，因为它们对环境做出反应。在我们的系统中，任何标记的亚细胞组件都可以被跟踪。该框架采用了一种新颖的方法，将亚蜂窝组件作为社交网络中的节点进行检查。描述细胞器的集合，如本研究中标记的线粒体，因为社会网络允许我们的框架研究细胞器的进化作为细胞组件互连的函数。除了量化全局信息，如数量和外观，我们的框架的方法还可以提供关于细胞器集合子集如何进化的更详细的局部反馈。预测会影响宿主线粒体形态的六个Mtb基因的海洋分枝杆菌同源基因，包括编码ESAT-6的rv3875，将被删除并评估线粒体形态表型。CellGraph将为未来的高通量计算管道奠定基础，能够对超大数据的细胞器时间演变进行快速定量分析，以高统计分辨率提供详细结果，并作为开放源码软件发布，供整个科学界用于其他应用和验证。

Abstract Tuberculosis (TB) has been a transmissible human disease for many thousands of years, and Mycobacterium tuberculosis (Mtb) is again the number one cause of death due to a single infectious agent. The intense 6- to 10-month process of multi-drug treatment, combined with the adverse side effects that can run the spectrum are major obstacles to patient compliance and therapy completion. The consequent increase in multidrug resistant TB (MDR-TB) and extensively drug resistant TB (XDR-TB) cases requires that we increase our arsenal of effective drugs and calls for the development of novel therapeutic approaches. Over the millennia, host and pathogen have evolved mechanisms and relationships that greatly influence the outcome of infection. Understanding these evolutionary interactions and their impact on pathogen clearance or host pathology will lead the way towards rational development of new therapeutics that favor a host protective response. These host-directed therapies have recently demonstrated promising results against Mtb, enhancing the cumulative effects of currently available anti-mycobacterial drugs or directly decreasing bacterial replication. However, our understanding of the host cell-pathogen interactions that lead to increased bacterial growth or host immune evasion is limited, and thus the ability to identify targets for novel host-directed drugs is hampered by a lack of mechanistic knowledge. Current methods for identifying Mtb virulence factors and imaging host cellular effects are slow and laborious with a general inability to simultaneously link multiple factors. Through the use of a high-throughput, large-scale computational pipeline, we can rapidly and effectively detect changes in the organellar morphology of host cells during infection with pathogens. Mycobacterium marinum, a biosafety level 2 bacterium, causes tuberculosis-like pathology in fish and amphibians and is used as a Mtb surrogate to study aspects of the infection process. The framework, CellGraph, will quantify changes in organellar shape, quantity, and spatial distribution over large sequences of Z-stack microscope images and digital videos, improving our understanding of cellular mechanisms as they respond to their environments. Any tagged subcellular component can be tracked within our system. This framework takes the novel approach of examining subcellular components as nodes in a social network. Characterizing ensembles of cellular machinery, such as tagged mitochondria in this study, as social networks allows our framework to study organellar evolution as a function of interconnectedness of cellular components. In addition to quantifying global information such as quantity and appearance, our framework's approach can also provide more detailed local feedback regarding how subsets of the organellar ensembles evolve. Mycobacterium marinum homologs of six of the Mtb genes predicted to impact host mitochondrial morphology, including rv3875 (encoding ESAT-6), will be deleted and assessed for mitochondrial morphology phenotypes. CellGraph will form the foundation for future high-throughput computational pipelines, enable rapid quantitative analysis of organellar temporal evolution for extremely large data, provide detailed results at high statistical resolutions, and be released as open source software that is available to the entire scientific community for additional applications and for validation.