Analysis of the Clustered Protocadherin (cPcdh) Locus in Human Pluripotent Stem and Derived Cells [scRNA-seq]

In the mammalian genome, the clustered protocadherin (cPcdh) locus is a paradigm of stochastic gene expression with the potential to express a different cPcdh combination in every neuron. Here, we report a limited version established during the transition from the naive to the primed states of human cell pluripotency that reduces by orders of magnitude the combinatorial potential in the cPcdh locus. It is a chromatin-based mechanism that increases the frequency of stochastic selection of a subset of cPcdh promoters upon differentiation to fetal-like neurons in monolayers, organoids, or the rat spinal cord. Signs of similar preferential selections can be observed in the brain throughout fetal development, disappearing after birth, but not in a condition of delayed maturation such as Down Syndrome. We therefore propose that pluripotent cells impose a pattern of limited cPcdh diversity on neurons that increases the likelihood of cPcdh repetition until these cells acquire adult-like maturation. Overall design: Lines/sublines: The hiPSC1-8 sublines were generated from single cells of the parent CVB hiPSC line. Six of these sublines were genetically modified (hiPSC2-7) and two were cases of failed genome editing (hiPSC1 and 8); please see Supplementary Table S1 in the associated publication for more details. Initially, these modifications should be irrelevant to the study of the cPcdh locus. The. hiPSC1 1.1-6 sublines were generated from the hiPSC1 subline and all underwent a process of genome editing, which again should be irrelevant to the study of the cPcdh locus (see Supplementary Table S1 in the associated publication for more details). The hiPSC 1.7-1.9 sublines were also generated from single cells of the parent CVB hiPSC line without undergoing any genome editing protocol. The hiPSC 1.13-15 sublines were generated from single cells of the parent CVI hiPSC line also without undergoing any genome editing protocol. The parent CVB and CVI lines were generated in the same reprogramming process and corresponded to two independent colonies. The hESC 1.7-1.9 sublines were generated from single cells of the parent HUES9 hESC line without undergoing any genome editing protocol. The HUES9 hESC 1.8 1.1-8 sublines were generated from single cells of the HUES9 1.8 subline after a process of primed-to-naive-to-reprimed conversion. All hiPSC/hESC lines/sublines were cultured in the primed state unless indicated (when indicated, some lines/sublines were cultured under naive-inducing conditions (5iLA protocol, and when returned to the primed state, the conditions were referred to as 're-primed'). For mouse cells, the 46C mESC 1.1-1.6 sublines were generated from single cells of the parent 46C mESC line. The R1 mESC 1.4/1.7 and 2.10/2.13 sublines were generated from single cells of the parent R1 mESC line. The R1 EpiSC 1.1-1.15 sublines were generated from single cells of the parent R1 mESC line after conversion to EpiSCs. The parental R1 mESC and EpiSC populations have also been included here. ChIP-seq data: All ChIP-seq experiments were performed in pluripotent stem cells (hiPSCs/hESCs/mESCs/EpiSCs). Most profiles were based on H3K4me3, but many others also on CTCF, Rad1, H3K4me2, H3K9me3, H3K9ac, H2A.Z, REST/NRSF, SIN3A, and JMJD2A. RNA-seq data: triplicate differentiation replicates of neurons generated from NPCs derived from hiPSC1-8 (n=24); cortical neurons generated from hiPSC1/3/5 (n=3); 10-month-old cortical organoids generated from the CVB line (n=1, pool of organoids); and primed, naive, and re-primed HUES9 1.8 subline (n=1 each). scRNA-seq data (SMART-seq+NEXTERA; full transcript): 79 single cells of neurons generated from NPCs of the hiPSC1 line (54 to 71), hiPSC2 (1 to 51), hiPSC6 (88 to 96), and hiPSC7 (72 to 87). The libraries were sequenced twice, in one case the libraries from all cells were pooled before cleaning and measuring DNA concentration, whereas in the other case the libraries were pooled after cleaning. and measuring DNA concentration (so equal amounts were mixed).

在哺乳动物基因组中，成簇原钙粘蛋白（clustered protocadherin, cPcdh）基因座是随机基因表达的经典模型，其赋予每个神经元表达独特cPcdh组合的潜能。本研究报道了一种在人类细胞多能性从初始态（naive）向激活态（primed）转变过程中建立的受限表达模式，该模式可将cPcdh基因座的组合表达潜能降低数个数量级。该机制基于染色质层面，可在单层培养、类器官或大鼠脊髓中向胎儿样神经元分化时，提升特定cPcdh启动子子集的随机选择频率。在整个胎儿发育阶段的大脑中，可观察到类似的偏好性选择现象，该现象于出生后消失，但在唐氏综合征（Down Syndrome）这类成熟延迟的病症中并未消失。因此我们推测，多能细胞会为神经元赋予受限的cPcdh多样性模式，从而提高cPcdh重复表达的概率，直至这些细胞达到类成体成熟状态。

实验设计：
细胞系与亚系：hiPSC1-8亚系均由亲本CVB诱导多能干细胞（human induced pluripotent stem cell, hiPSC）系的单细胞克隆获得。其中6个亚系（hiPSC2-7）完成了基因编辑，2个亚系（hiPSC1、8）的基因编辑未成功；详细信息请参见相关论文的补充表S1。初始实验中，上述基因编辑操作与cPcdh基因座的研究无关。
hiPSC1 1.1-6亚系均由hiPSC1亚系的单细胞克隆获得，所有亚系均进行了基因编辑，该操作同样与cPcdh基因座的研究无关（详细信息参见相关论文的补充表S1）。
hiPSC 1.7-1.9亚系均由亲本CVB hiPSC系的单细胞克隆获得，未进行任何基因编辑操作。
hiPSC 1.13-15亚系均由亲本CVI hiPSC系的单细胞克隆获得，同样未进行任何基因编辑操作。
亲本CVB与CVI hiPSC系均通过同一重编程流程获得，分别来自两个独立的细胞集落。
hESC 1.7-1.9亚系均由亲本HUES9人类胚胎干细胞（human embryonic stem cell, hESC）系的单细胞克隆获得，未进行任何基因编辑操作。
HUES9 hESC 1.8 1.1-8亚系则是将HUES9 1.8亚系经“激活态→初始态→再激活态”转化后，通过单细胞克隆获得。
所有hiPSC/hESC细胞系与亚系均在激活态多能性培养条件下维持，除非另有说明：当需使用初始态培养条件时，采用5iLA诱导方案；当细胞从初始态重新转回激活态时，该培养条件被称为“再激活态（re-primed）”。
针对小鼠细胞：46C小鼠胚胎干细胞（mouse embryonic stem cell, mESC）1.1-1.6亚系均由亲本46C mESC系的单细胞克隆获得。
R1 mESC 1.4/1.7与2.10/2.13亚系均由亲本R1 mESC系的单细胞克隆获得。
R1外胚层干细胞（epiblast stem cell, EpiSC）1.1-1.15亚系则是将亲本R1 mESC系转化为EpiSCs后，通过单细胞克隆获得。
本研究同时纳入了亲本R1 mESC与EpiSC细胞群。

ChIP测序（ChIP-seq）数据：所有ChIP-seq实验均以多能干细胞（hiPSCs/hESCs/mESCs/EpiSCs）为材料。绝大多数测序图谱基于H3K4me3标记，其余部分则分别针对CTCF、Rad1、H3K4me2、H3K9me3、H3K9ac、H2A.Z、REST/NRSF、SIN3A及JMJD2A等靶标。

RNA测序（RNA-seq）数据：包含24组由hiPSC1-8来源的神经前体细胞（neural progenitor cell, NPC）分化得到的神经元生物学重复样本（n=24）；3组由hiPSC1/3/5来源的皮层神经元生物学重复样本（n=3）；1组由CVB系构建的10月龄皮层类器官样本（混合多个类器官，n=1）；以及分别来自HUES9 1.8亚系的激活态、初始态与再激活态样本各1组（n=1 each）。

单细胞RNA测序（single-cell RNA sequencing, scRNA-seq）数据（采用SMART-seq+NEXTERA建库方案，覆盖完整转录本）：共计79个单细胞样本，分别来自hiPSC1系（54~71号细胞）、hiPSC2系（1~51号细胞）、hiPSC6系（88~96号细胞）及hiPSC7系（72~87号细胞）的NPC分化神经元。本研究分两轮进行测序：第一轮将所有细胞的文库在纯化与定量DNA浓度前进行混合；第二轮则在纯化与定量DNA浓度后进行混合，以保证各样本的投料量一致。