Genomic control of gene regulatory networks governing early human lineage decisions

摘要 预测基因组变异的影响需要定量建模来解构相互作用 并确定它们对基因调控网络的联合影响 (GRN)控制细胞状态和细胞功能。我们专注于控制早期人类发育的GRN 作为一个范例。可以说，哺乳动物发育过程中最重要的世系决定是 上胚层细胞退出多能状态(当细胞具有产生所有体细胞的潜力时的状态 细胞和生殖细胞)，并分化为三个初级生殖层之一，内胚层，中胚层， 和外胚层。这种多能状态和三倍体分化可以用培养的人类 胚胎干细胞(HESCs)。许多人的注意力都集中在维持 自我更新的多能性状态，但控制hESC三系分化的GRN在很大程度上仍然存在 未被开发的。我们之前进行了基因组规模的CRISPR/CAS筛查，以发现蛋白质编码基因 调节人类胚胎干细胞向最终内胚层的转变。基于基因组和遗传数据和 机器学习(GKM-支持向量机序列分析)，我们扩展了最初的简单两个转录因子(Tf) 模型转换为多TF协作模型。在这里，我们提出了一种检查hESC的综合方法 过渡到确定的内胚层、中胚层和神经外胚层胚层身份，以改善 GRN模型的泛化能力。我们将进行高密度的定量基因组和蛋白质组测量 时间分辨率和单元格分辨率。这些定量测量将与KEY的扰动相结合 GRN元素、核心TF及其目标增强子，以生成动态GRN模型。至 进一步提高我们新的GRN模型的精度，我们将在状态转换期间映射细胞轨迹 通过谱系追踪结合scRNA-seq.除了hESC引导的分化，生理学的 增强剂的相关性将在人类和小鼠类器官(原肠类)和小鼠中进一步询问。 胚胎。然后，我们将把创新的计算和算法方法应用到我们的多式联运中 用于生成GRN模型的实验数据，旨在学习 基因组变异对细胞和最终生物表型的贡献。开发GRN模型用于 多能性的退出和生殖层身份的获得涉及对细胞状态的动态建模 过渡，这不仅将使我们了解人类早期的发展，而且还可以作为 为胚胎发育中的生物转变构建可推广的GRN模型的基础， 成体组织的动态平衡和再生以及不适当的细胞命运转变 病理性疾病，如癌症。

ABSTRACT Predicting the impact of genomic variation requires quantitative modeling to deconstruct the interplay between multiple individual variants and to determine their combined effects on gene regulatory networks (GRNs) that control cell state and cell function. We focus on the GRNs that control early human development as a paradigm. Arguably the most important lineage decision during mammalian development is the decision of epiblast cells to exit the pluripotent state (a state when the cells have the potential to give rise to all somatic cells and germ cells), and differentiate into one of the three primary germ layers, the endoderm, mesoderm, and ectoderm. This pluripotent state and the trilineage differentiation can be captured using cultured human embryonic stem cells (hESCs). Much attention has focused on the GRNs underlying the maintenance of the self-renewing pluripotent state, but the GRNs governing hESC trilineage differentiation remain largely unexplored. We previously conducted genome-scale CRISPR/Cas screens to discover protein-coding genes that regulate the transition of hESCs to definitive endoderm. Based on the genomic and genetic data and machine learning (gkm-SVM sequence analysis), we expanded our initial simple two transcription factor (TF) model to a multiple TF cooperative model. Here we propose an integrative approach examining the hESC transition to definitive endoderm, mesoderm and neuroectoderm germ layer identities to improve the generalizability of GRN models. We will perform quantitative genomic and proteomic measurements with high temporal and single-cell resolution. These quantitative measurements will be combined with perturbation of key GRN elements, core TFs and their target enhancers, to inform the generation of dynamic GRN models. To further improve the precision of our new GRN models, we will map cell trajectories during state transitions through lineage tracing combined with scRNA-seq. Beyond hESC guided differentiation, the physiological relevance of enhancers will be further interrogated in human and mouse organoids (gastruloids) and mouse embryos. We will then apply innovative new computational and algorithmic methods to our multimodal experimental data to generate GRN models, aiming to learn generalizable principles underlying the contribution of genomic variants to cellular and ultimately organismal phenotypes. Developing GRN models for the exit of pluripotency and the acquisition of germ layer identities involves dynamic modeling of the cell state transition, which will not only inform our understanding of early human development, but can also serve as the basis for construction of generalizable GRN models for biological transitions during embryonic development, adult tissue homeostasis and regeneration as well as inappropriate cell fate transitions that occur in pathological conditions such as cancer.