Spatial and temporal tumor-immune co-evolution and interactions that model lung adenocarcinoma development

项目摘要 肺腺癌（LUAD）是肺癌中最常见的亚型，占大多数癌症 死亡早期检测的改善增加了在早期病理阶段诊断的LUAD的数量， 从而制定了治疗这一不断增长的患者亚群的策略。阻碍这些进步是一个非常糟糕的 了解推动LUAD发展的早期事件，从而指导 拦截而LUAD患者的正常肺上皮显示肿瘤相关分子， 和炎症变化，目前尚不清楚为什么LUAD在肺的特定区域内发展。而 肺在生态学上富含许多细胞群， 过程中，我们仍然不知道如何个别细胞群体的性质和作用，如上皮细胞， 和免疫亚群，共同进化和相互作用，以激发LUAD的发展，从一个特定的生态位在肺。 在我们的初步努力中，我们通过多区域单细胞测序发现了 多个细胞子集和状态的性质和转录组特征（例如，促肿瘤免疫抑制 表型）跨越宏观空间，这样细胞生态系统和免疫细胞受体库被 LUAD和邻近正常区域之间的相似性高于更远的正常部位。此外， 空间特性沿着匹配的人正常肺的病理连续体逐渐丰富， 到癌前病变，再到侵入性LUAD我们的初步发现激发了这样一种假设， 以及上皮细胞和免疫细胞的表达程序、特性和相互作用的时间演变 从正常和癌前肺中建立LUAD早期发展模型。在目标1中，我们将研究LUAD 并通过单细胞RNA匹配与肿瘤具有确定空间邻近度的多区域正常组织 和免疫受体测序结合肿瘤中突变的分析，以建立单细胞 LUAD和免疫协同进化的空间图。空间调制的细胞性质和状态，然后将 用于喂养和训练机器学习模型，该模型描绘了肺部生态系统的LUAD发展。在 目标2：我们将沿着正常和免疫的病理连续体， 以及识别由早期免疫调节的细胞状态和性质 干预我们将使用小鼠的时间信息，沿着人类匹配的正常肺组织， 肿瘤前病变和侵入性LUAD，以迭代验证和微调我们机器的性能 学习模型，以及时描绘正常和癌前肺的LUAD发展。结束时 通过这些研究，我们将建立新的模型，可靠地描绘LUAD在空间和时间上的演变。通过提供 LUAD开发的一个可访问的数据门户地图集，我们还希望我们的研究将提供可扩展的 为科学界制定治疗这种致命疾病的新战略提供了路线图。

PROJECT SUMMARY Lung adenocarcinoma (LUAD) is the most frequent subtype of lung cancer and accounts for most cancer deaths. Improved early detection has increased the number of LUADs diagnosed at earlier pathological stages, thus warranting strategies to treat this growing patient subpopulation. Thwarting these advances is a very poor understanding of early events that drive LUAD development and that thus would guide ideal approaches for interception. While normal lung epithelia of LUAD patients were shown to display tumor-pertinent molecular and inflammatory changes, it is not clear why a LUAD develops within a particular region in the lung. Whereas the lung is ecologically rich with many cell populations that partake in both physiological and pathological processes, we still do not know how the properties and roles of individual cell populations, such as epithelial and immune subsets, co-evolve and interact to instigate LUAD development from a specific niche in the lung. In our preliminary efforts, we found by multi-region single-cell sequencing remarkable evolution of the properties and transcriptomic features of multiple cell subsets and states (e.g., protumor immunosuppressive phenotypes) across macro-space, such that cellular ecosystems and immune cell receptor repertoires were more similar among LUADs and adjacent normal regions than with more distant normal sites. Also, such spatial properties were progressively enriched along the pathologic continuum of matched human normal lung, to preneoplasias, up to invasive LUADs. Our preliminary findings motivate the hypothesis that geospatially and temporally evolving expression programs, properties, and interplay of epithelial and immune cells model early development of LUAD from the normal and premalignant lung. In Aim 1, we will study LUADs and matched multi-region normal tissues with defined spatial proximities from the tumors by single-cell RNA and immune receptor sequencing in conjunction with analysis of mutations in the tumors to establish single-cell maps of LUAD and immune co-evolution in space. Spatially modulated cell properties and states will then be used to feed and train a machine learning model that portrays LUAD development from the lung ecosystem. In Aim 2, we will single-cell decode tumor-immune co-evolution along the pathologic continuum of normal and premalignant lung to LUAD as well as identify cell states and properties that are modulated by early immune intervention. We will use temporal information in mice, along with human matched normal lung tissues, preneoplastic lesions, and invasive LUADs, to iteratively validate and fine-tune the performance of our machine learning model to portray LUAD development in time from the normal and premalignant lung. At the end of our studies, we will have built new models that reliably portray LUAD evolution in space and time. By providing an atlas of LUAD development in an accessible data portal, we also expect that our study will offer scalable roadmaps for the scientific community to develop new strategies for treatment of this fatal disease.