Figure 1.
Overview of experimental design and data analysis.
(A) qPCR validation steps during ChIP-Seq library preparation using small and large scale protocols. Several regions known to be K27-enriched (positive controls) or not-enriched (negative controls) in MCF-7 cells were tested by qPCR using DNA templates before and after linker-mediated PCR amplification and after size-selection. Y-axis indicates enrichment relative to averaged negative controls. (B) Comparison of small-scale and standard ChIP-Seq experiments. Scatter plots depict the counts of ChIP-Seq reads for each gene. X and Y-axes indicate mapped read counts around promoter regions (+/− 5 kb from TSS) for each gene in small-scale and in standard experiment, respectively. (C) Schematic view of cell purification and sample processing. Red numbers indicated the number of independent samples in each data type. (D) Representative examples of genes known to be specifically expressed in luminal (blue) and stem (red) cells. Each bar represents a different sample. Y-axis indicates SAGE-Seq tag counts.
Figure 2.
Distinct histone methylation profiles of human CD44+ mammary epithelial progenitors and CD24+ differentiated luminal epithelial cells.
(A) Examples of histone modification patterns of genes known to be expressed in stem or luminal cells. Total aligned tag count in each ChIP-Seq library was scaled to 10 million, Y-axis shows tag counts averaged over a 10 bp window. (B) Comparison of genes enriched for the indicated histone marks in CD44+ and CD24+ cells. Venn diagram depicts the number of unique and overlapping genes. (C) Functional enrichment analysis of K27-enriched genes using DAVID Functional Annotation Tool.
Figure 3.
Associations between chromatin and cell type–specific expression patterns.
(A) Overall correlation between histone methylation (K4, K27) and gene expression in CD44+ progenitors from sample 1. Box plot shows distribution of gene expression of corresponding genes in each category. Red bar: median, box: interquartile ranges and whisker; most extreme value within 1.5 times of box length. (B) Bar chart shows the correlation between cell type–specific K27 enrichment and gene expression patterns. Blue and red: K27-enriched only in CD24+ and CD44+ cells, respectively, yellow and gray: K27 enriched in both and neither cell type. (C) Differences in K27 patterns between CD24+ and CD44+ cells are consistent in three individual samples and distinct from that of K4 profiles. Heatmap depicting unsupervised clustering of histone modification patterns of genes highly expressed in CD44+ (656 genes) and CD24+ (435 genes) cells. Blue color indicates the level of enrichment for the indicated histone modification based on the ranking of ChIP-Seq read counts for each gene in each sample. (D) Differentially expressed genes that are enriched (K27+) or not enriched (K27-) for K27 were analyzed for relative enrichment with the indicated protein classes (left panel) and for relative connectivity (right panel). X-axes indicate –log10 p-values for enrichment with the listed protein classes (left panel) and the number of overconnected objects, defined as proteins with higher than expected number of interactions, in each functional category within each group (right panel), respectively.
Figure 4.
Genome-wide H3K27me3 patterns of human mammary epithelial progenitor and differentiated luminal cells.
(A) Patterns of K27 enrichment and gene expression are mutually exclusive. Representative example of K27 distribution and gene expression in chr1. Data was analyzed using SICER algorithm [28] using 10 kb as window size. Significantly enriched regions and gene expression levels are plotted as colored lines across chromosome position. Red and blue lines represent K27 enrichment in CD44+ and CD24+ cell libraries from three different individuals, respectively. Gene expression levels are plotted across chromosome position. The height of line indicates the level of expression of corresponding gene. Orange and light blue: median expression in CD44+ and CD24+ samples, respectively. (B) Representative examples showing that distinct K27 distributions correlate with gene expression using the same plot as in panel A but at different scale. Red and blue dots indicate genes highly expressed (>2 fold difference) in CD44+ and CD24+ cells, respectively. Clear differences in K27 distribution between the two cell types are observed consistently in the regions where the selected genes are located. (C) Correlations between the number of K27 blocs and the number of genes in these blocs in CD44+ (left) and CD24+ (right) cells depending on setting K27 blocs at different sizes. X-axes indicate the threshold of length for defining K27 blocs, whereas y-axes show the number of K27 blocs and the number of genes within these blocs. (D) Fraction of genes with the indicated expression pattern located within K27 blocs in CD44+ and CD24+ cells.
Figure 5.
Changes in chromatin state and cell type–specific gene expression patterns.
(A) Number of genes for each of the four possible chromatin states (i.e., bivalent – purple, K4 only – orange, K27 only – green, and neither – gray) in the indicated three cell types. (B) Potentially interesting differences in chromatin patterns. Bar chart shows associations between changes in chromatin-state and gene expression patterns. Each row indicates the type of chromatin-state change and the number of genes in each category. Blue and red: genes highly expressed in CD24+ and CD44+ cells, respectively (≥2-fold change). Yellow and gray: genes with ≤2-fold difference between CD24+ and CD44+ cells and with low/no detectable expression in either cell type, respectively. (C) Representative examples of genes in each category. ChIP-Seq tag counts for K4 and K27 modification in hES, CD44+, and CD24+ cells are shown. Total aligned tag count was scaled to 10 million and tag counts were averaged over a 10 bp window. (D) Functional enrichment analysis of genes within each chromatin pattern category (left panel) and the number of overconnected objects in each functional category within each group (right panel). X-axis indicates –log10 p-values for enrichment with the indicated protein class (left panel) and the number of overconnected objects (right panel), respectively. Definitions are the same as described in Figure 3D.
Figure 6.
Combined view of DNA methylation and gene expression patterns.
(A) Associations between fraction of genes highly expressed in CD24+ or CD44+ cells and the location of DMRs (p<10−5) in the two cell types relative to TSS. Y-axes show fraction of genes in the four different gene expression groups (i.e. CD24-high, CD44-high, no difference, and not expressed) relative to the location of DMRs hypermethylated in CD24+ (blue line) or CD44+ (red line) cells and all MSDK sites used as control. (B) Differentially (≥2-fold difference) expressed genes are enriched in DMRs. Bar charts show associations between genes with indicated DMR and gene expression patterns as described in panel A. Genes in each gene set have DMR (indicated left side) within promoter region (−5 kb from TSS to +2 kb, left panel) or in gene body (+2 kb to end, right panel). The number of MSDK sites and associated genes in each group is indicated. We used four different cut-offs for DMRs, 2, 5,10, 20 (-log10 p-value) from top to bottom (black triangle). Randomly picked MSDK sites did not show any enrichment pattern. (C) Correlation between mean gene expression levels in relation to promoter and gene body methylation in CD24+ and CD44+ cells. Red stars mark statistically significant differences relative to all MSDK sites. (D) Functional enrichment analysis (left panel) of genes associated with promoter and gene body DMRs in CD24+ and CD44+ cells and the number of overconnected objects in each functional category within each group (right panel). X-axis indicates –log10 p-values for enrichment with the indicated protein class (left panel) and the number of objects (right panel), respectively. Definitions are the same as described in Figure 3D.
Figure 7.
Integrated view of genome-wide gene expression and DNA and histone methylation patterns.
(A) Genomic regions (+/− 5 kb and +/− 0 kb from DMR for K27 and K4, respectively) associated with DMRs in one cell type (e.g., CD44+ cells) are enriched for K4 or K27 mark in the other (e.g., CD24+ cells). Bar chart shows observed/expected ratio of the indicated MSDK sites with the designated K27 and K4 patterns between CD44+ and CD24+ cells. (B) Associations between gene expression and histone and DNA methylation patterns. Y-axis shows the –log10 p-value of enrichment for genes with the indicated expression and histone modification pattern in gene body and promoter DMRs. Orange line indicates –log10 (p value) of statistical significance, numbers 1–4 mark significantly enriched patterns. (C) Schematic models depicting possible changes in DNA methylation and K27 enrichment during CD44+ to CD24+ cell differentiation and their effect on gene expression based on data presented in panel B. Examples of genes within each group are listed. White (unmethylated) and black (methylated) circles indicate potential DNA methylation sites (i.e., CpG) in the promoter and gene body, blue and orange ovals represent lack and presence of K27 mark, respectively. Red and dashed green arrows indicate increased and decreased gene expression, respectively.