Advanced development and validation of an in vitro platform to phenotype brain metastatic tumor cells using artificial intelligence

摘要 从原发肿瘤部位转移到大脑是晚期乳腺癌最致命的并发症。 全世界大约20%的乳腺癌患者都经历过这种情况。目前还没有办法 检测肿瘤是否有脑转移潜力，是否没有预测未来成功转移的标志，因此 没有针对任何涉及的过程的治疗方法。这些差距很难弥合，因为缺乏 一种可以通过根据大脑对癌细胞进行分类来阐明其潜在机制的技术 转移潜能。目前的体内小鼠模型表现为转移缓慢，并且没有 捕捉单细胞形态和动力学的能力；目前的体外模型寿命短， 由于缺乏软件支持，我们提出了一种芯片上的体外血脑利基(µm-BBN)来测量 癌细胞与正常细胞的表型差异以及癌细胞在转移过程中的表型差异 模型利基，并为它们分配脑转移潜能。此外，我们建议捕获 通过BBN过境以作进一步分析。 该系统由微米BBN、集成的压电泵和控制器、自动表型 软件和分类算法。微米-BBN有两个腔室，形成一个血管(人脑 内皮细胞和脑基质(ECM和正常人星形胶质细胞(NHA)、小胶质细胞和周细胞) 由涂有Matrigel的5微米多孔膜分离。其目标是在该设备中培养癌细胞 使用集成泵最多可使用20天。在前人工作的基础上，细胞表型和 患者细胞库的迁移行为将被记录下来。表型指标的生成 单个癌细胞、微转移和肿瘤微环境将使自动分析成为可能 肿瘤细胞的转移特征。培养后，细胞将被回收并固定或分析以创建 一种多组体读数，能够为单个细胞和亚细胞提供免费的细胞和分子签名 转移性癌细胞群体。 这项工作将有助于更好地理解脑转移的潜在机制， 在它的下游，有一套更强大的靶向通路，用于预防脑转移。

Abstract Metastasis from the primary tumor site to the brain is the most lethal complication of advanced breast cancer and is experienced by approximately 20% of breast cancers worldwide. There is at present no approach to detect if a tumor has brain metastatic potential, no markers that predict successful future metastasis, and thus no therapies to target any of the processes involved. These gaps are difficult to bridge due to a lack of technology that can elucidate the underlying mechanisms by classifying a cancer cell based on its brain metastatic potential. Current in vivo murine models are slow to manifest metastasis and do not have the capability of capturing single cell morphology and dynamics; and current in vitro models are short lived and lack software support therefore, we propose an in vitro blood brain niche (µm-BBN) on-a-chip to measure the phenotypic differences between cancer cells and normal cells and amongst cancer cells as they transit through the model niche and to assign them a brain metastatic potential. Moreover, we propose capturing the cells that transit through the BBN for further analysis. This system is composed of the µm-BBN, an integrated piezo pump and controller, automated phenotyping software and a classification algorithm. The µm-BBN has two chambers which form a vessel (human brain endothelial cells and brain stroma (ECM and Normal Human Astrocytes (NHA), Microglia and Pericytes) separated by a 5µm porous membrane coated with Matrigel. The goal is to culture cancer cells in the device for up to 20 days using the integrated pump. Expanding on previous work, the cellular phenotype and migratory behavior of a library of patient cells will be recorded. Generation of phenotypic measures for individual cancer cells, micro-metastasis and the tumor micro-environment will enable automated profiling of the metastatic signature of tumor cells. After culture the cells will be recovered and fixed or analyzed to create a multi-omic readout enabling a complimentary cellular and molecular signature for single cells and sub- populations of metastatic cancer cells. This work will enable improved understanding of the underlying mechanisms of brain metastasis and, downstream from it, in a more robust set of targetable pathways for prevention of brain metastasis.