Project 2: Measuring and modeling the tumor and immune microenvironment before and during therapy and at the time of drug resistance

项目概要-项目2（AIM 5）测量和建模肿瘤和免疫 治疗前后的微环境项目2的总体目标是确定 肿瘤及其微环境使其对单独或组合的ICI或靶向疗法有反应。我们 将以单细胞分辨率收集关于肿瘤的身份、状态和物理排列的定量数据， 基质和免疫细胞以及构成肿瘤微环境的可溶性和ECM成分 （TME）。这将使用标准福尔马林固定的高度多路复用荧光成像来实现， 石蜡包埋（FFPE）组织和肿瘤样品结合单细胞RNA测序（scRNA）。到 提供洞察变量之间的因果关系，我们将分析在不同点收集的样本 在时间上，最常见的是治疗前和治疗中以及耐药疾病时的活检。的情况下 ICI诱导的皮肤毒性，我们将进行局部干预性研究（例如，维甲酸治疗） 然后进行活检以确定治疗的有效性并测试关于免疫的特定假设。 细胞内稳态。在这个项目中的大部分工作将是假设生成，并将 与项目1和3中的细胞和小鼠假设检验研究紧密结合。 目标5.1将集中在实验和计算方法获得20-60通道图像从 使用基于组织的循环免疫组织化学技术检测福尔马林固定的石蜡包埋（FFPE）组织和肿瘤样本 免疫荧光（t-CycIF）。目标5.2将整合高维t-CycIF成像和单细胞RNA 测序以生成关于单细胞水平上的肿瘤、基质和免疫细胞的组成和状态的数据。 分辨率（“深部肿瘤表型”）。目的5.3将使用深度表型分析BRAFV 600 E的肿瘤 接受BRAF和MEK抑制剂治疗的患者或接受ICI治疗的患者，无论BRAF突变如何 status.在治疗前和治疗期间以及进展时收集的活检将用于识别 恶性细胞、TME和免疫细胞组群的变化与以下疾病相关，并可能预测以下疾病： 治疗反应和耐药性。目的5.4将分析ICI对皮肤驻留T细胞的影响， 将不良反应与类似的特发性疾病进行比较;分析对类维生素A的局部反应 类固醇将为皮肤免疫稳态提供新的见解。目标5.5将确定特征 与脑肿瘤中ICI的异常反应相关（并最终预测），并提供以下数据： 这些生物标志物可以在贝叶斯适应性临床试验中进行评估。目标5.6将调查 三阴性乳腺癌中免疫浸润与内在或药物诱导的基因组不稳定性之间的关系 （TNBC）。目标5.7将整合肿瘤表型、药物干预和临床反应的数据， 一系列有监督和无监督的机器学习方法，包括基于网络先验的方法， 并且还使用多视图学习框架将scRNA转录组学与t-CycIF图像数据相关联。

PROJECT SUMMARY – PROJECT 2 (AIM 5) Measuring and modeling the tumor and immune microenvironment before and after therapy. The overall goal of Project 2 is to determine which features of a tumor and its microenvironment make it responsive to ICIs or targeted therapies alone or in combination. We will collect quantitative data at single cell resolution on the identities, states and physical arrangement of tumor, stromal and immune cells and on soluble and ECM components that comprise the tumor microenvironment (TME). This will be accomplished using highly multiplexed fluorescence imaging of standard formalin fixed, paraffin-embedded (FFPE) tissue and tumor samples combined with single cell RNA sequencing (scRNA). To provide insight into causal relationships among variables, we will analyze samples collected at different points in time, most commonly biopsies prior to and on therapy, and at the time of drug-resistant disease. In the case of ICI-induced skin toxicities, we will perform localized interventional studies (e.g. treatment with retinoids) followed by biopsies to determine the effectiveness of treatment and to test specific hypotheses about immune cell homeostasis in skin, respectively. Much of the work in this project will be hypothesis generating and will be tightly integrated with hypothesis testing studies in cells and mice in Projects 1 and 3. Aim 5.1 will focus on experimental and computational methods for obtaining 20-60 channel images from formalin-fixed, paraffin embedded (FFPE) tissue and tumor samples using tissue-based cyclic immunofluorescence (t-CycIF). Aim 5.2 will integrate high dimensional t-CycIF imaging and single cell RNA sequencing to generate data on the composition and states of tumor, stromal and immune cells at single-cell resolution (“deep tumor phenotypes”). Aim 5.3 will use deep phenotyping to analyze tumors from BRAFV600E patients treated with BRAF and MEK inhibitors or patients treated with ICIs irrespective of the BRAF mutation status. Biopsies collected before and during therapy, and at the time of progression, will be used to identify changes in the malignant cells, TME and immune cell cohort associated with, and potentially predictive of, therapeutic response and drug resistance. Aim 5.4 will analyze the effects of ICIs on skin-resident T-cells and compare adverse responses to the idiopathic conditions they resemble; analysis of local responses to retinoids and steroids will provide new insight into immune homeostasis in the skin. Aim 5.5 will identify features associated with (and ultimately predictive of) exceptional response to ICIs in brain tumors and provide data on biomarkers that can be evaluated in Bayesian adaptive clinical trials. Aim 5.6 will investigate the connection between immune infiltration and intrinsic or drug-induced genomic instability in triple negative breast cancers (TNBC). Aim 5.7 will integrate data on tumor phenotypes, drug interventions and clinical responses using a range of supervised and unsupervised machine-learning methods, including methods based on network priors, and also link scRNA transcriptomics with t-CycIF image data using a multi-view learning framework.