Additional file 2 of A low-input high resolution sequential chromatin immunoprecipitation method captures genome-wide dynamics of bivalent chromatin

Additional file 2: Figure S1. (A) Agarose DNA gel showing chromatin fragmented with MNase versus sonication (B) ChIP-qPCR for single (top) and reChIP (bottom) experiments using IgG (Invitrogen, black), H3K4me3 (CST 9751, green) and H3K27me3 (CST 9733, red) antibodies on sonicated chromatin. An active H3K4me3 region (Gapdh1), inactive H3K27me3 region (Meis2) and 6 bivalent regions are assayed, four of which are novel to this study. (C) Single ChIP-qPCR analysis using IgG (Invitrogen, left), H3K4me3 (Millipore 07-473, middle) and H3K27me3 (Active Motif 91167, right) antibodies with variable number of cells ranging from 1 million (1M), through to 2,000 (2K). Five primer sets were used including three H3K4me3 enriched regions (green) and two H3K27me3 enriched regions (red). Note that below 50K the background increases and specificity of enrichment is no longer detected. For this experiment a different H3K27me3 antibody was used to the sequenced reChIP datasets. (D, E) downsampling analysis for K4-K27 and K27-K4 datasets using the higher coverage replicate 2. Peaks were called separately for each downsampled dataset and classified according to Figure 3A. (D) Total peaks, (E) promoter peaks. (F, G) Downsampling analysis of independent total H3K4me3 and total H3K27me3 from (GSE135841) (7) showing in silico bivalent predictions atall peaks (G) and promoter peaks (H) at different simulated read depths. (H) Pseudocolour density scatterplot showing log2 CPM/bp enrichment at combined H3K4me3 and H3K4me3-IgG peaks for total H3K4me3 single ChIP (x-axis) compared to H3K4me3-IgG reChIP (y-axis). R=0.906. Data reanalysed from Mas et al. 2018 (14). Figure S2. (A) Schematic showing peak calling strategy for in silico bivalent peak prediction (B-F) Genome browser views of (B) H3K4me3-only, (C) high-confidence, (D) K4-biased, (E) K27-biased, and (F) low confidence genes showing H3K4me3 (green), total H3K27me3 (red) and K4-K27 (purple) and K27-K4 (blue) reChIP datasets. Height of peak represents CPM/bp. E-M represents data from independent 10 million cell total H3K4me3 and total H3K27me3 (GSE135841) (7). R1 and R2 are two independent biological replicates from this study. (G) number of bivalent promoters overlapping (dark grey) or not-overlapping (light grey) CpG islands for the four different classes. Genome-wide 51.97% promoters overlap a CpG island using these criteria (H) Overlap of non-promoter bivalent regions with candidate cis-regulatory regions (cCREs) from ENCODE for the four different classes. Figure S3. (A) Classification strategy for calling bivalent promoters in data from Mas et al. 2018 (14). Note only one replicate of the reChIP datasets were generated in this study. Numbers denote number of peaks/promoters at each step. (B) Enrichment of bivalent promoters shared between this study and Mas et al. 2018 (14) (top, n=3593) or unique to our study (bottom, n=1511) in Mas et al. 2018 (14) datasets. (C) log2 fold change in gene expression levels for pluripotency genes (top) and a random set of expressed non-bivalent genes (bottom) across 9 days of embryoid body differentiation. Each gene has been normalised separately across the time series. Top and bottom groups of genes are on different scales. Gene expression data reanalysed from (GSE135841). Figure S4. (A) Classification strategy for calling differential bivalent promoters between wild type (WT) and Dppa2/4 double knockout (DKO) clones. Numbers denote number of peaks/promoters at each step.

补充文件2：图S1。(A) 琼脂糖凝胶电泳结果，对比微球菌核酸酶（MNase）与超声破碎（sonication）处理的染色质片段。(B) 针对超声破碎染色质开展单重染色质免疫沉淀定量PCR（ChIP-qPCR）实验：采用免疫球蛋白G（IgG，Invitrogen，黑色）、H3K4me3抗体（CST 9751，绿色）及H3K27me3抗体（CST 9733，红色），分别进行单重实验（上图）与重染色质免疫沉淀（reChIP，下图）检测。本实验共检测1个活性H3K4me3区域（Gapdh1）、1个非活性H3K27me3区域（Meis2）以及6个二价区域，其中4个为本次研究首次报道。(C) 采用不同细胞量（从100万[1M]至2000[2K]）开展的单重ChIP-qPCR分析：分别使用IgG抗体（Invitrogen，左侧）、H3K4me3抗体（Millipore 07-473，中间）及H3K27me3抗体（Active Motif 91167，右侧）。实验共使用5对引物，其中3对富集H3K4me3区域（绿色）、2对富集H3K27me3区域（红色）。需注意，当细胞量低于5万（50K）时，背景信号升高，无法检测到富集特异性。本实验使用的H3K27me3抗体与后续测序的重ChIP数据集所用抗体不同。(D, E) 采用高覆盖度重复样本2开展的K4-K27与K27-K4数据集下采样分析：对每个下采样数据集单独进行峰识别，并参照图3A进行分类。(D) 总峰数统计；(E) 启动子区峰数统计。(F, G) 针对从GEO数据集GSE135841（文献7）中获取的独立总H3K4me3与总H3K27me3数据开展下采样分析，展示不同模拟测序深度下所有峰（G）及启动子区峰（H）的计算机模拟二价预测结果。(H) 伪彩色密度散点图，展示总H3K4me3单重ChIP（x轴）与H3K4me3-IgG重ChIP（y轴）在联合H3K4me3与H3K4me3-IgG峰处的log₂(CPM/bp)富集倍数，相关系数R=0.906。数据重新分析自Mas等人2018年的研究（文献14）。图S2。(A) 计算机模拟二价峰预测的峰识别策略示意图；(B-F) 基因组浏览器视图，分别展示(B) 仅H3K4me3、(C) 高置信度、(D) K4偏向型、(E) K27偏向型及(F) 低置信度基因的H3K4me3（绿色）、总H3K27me3（红色）、K4-K27（紫色）及K27-K4（蓝色）重ChIP数据集。峰高代表CPM/bp值。E-M代表从GEO数据集GSE135841（文献7）中获取的独立1000万细胞总H3K4me3与总H3K27me3数据。R1与R2为本研究的两个独立生物学重复。(G) 四类不同二价启动子分别与CpG岛重叠（深灰色）或不重叠（浅灰色）的数量统计。采用本标准，全基因组范围内51.97%的启动子与CpG岛重叠。(H) 四类不同非启动子区二价区域与ENCODE候选顺式调控区域（cCREs）的重叠统计。图S3。(A) 针对Mas等人2018年研究（文献14）的数据中二价启动子识别的分类策略。需注意本研究仅生成了重ChIP数据集的一个重复样本。各步骤标注的数字代表对应阶段的峰/启动子数量。(B) 本研究与Mas等人2018年研究（文献14）共享的二价启动子富集情况（上图，n=3593），以及本研究特有的二价启动子富集情况（下图，n=1511），分析均基于Mas等人2018年研究（文献14）的数据集。(C) 胚胎体分化9天过程中，多能性基因（上图）与随机选取的表达型非二价基因（下图）的基因表达水平log₂倍变化。每个基因均已在时间序列中单独完成标准化。上图与下图基因的数值标尺不同。基因表达数据重新分析自GEO数据集GSE135841。图S4。(A) 野生型（WT）与Dppa2/4双基因敲除（DKO）克隆之间差异二价启动子识别的分类策略。各步骤标注的数字代表对应阶段的峰/启动子数量。