Overlap between BRD4 and H2Bub1 recruitment sites

Since BRD4 and H2Bub1 showed similar pattern of gene regulation, we next used ChIP-seq to determine whether BRD4 and H2Bub1 play direct roles in regulating transcription of target genes. No studies have analyzed the correlation between genome-wide recruitment of BRD4 and H2Bub1. Thus, to understand the mechanistic link between BRD4 and H2Bub1 in gene regulation, ChIP-sequencing of BRD4 and H2Bub1 was performed in MCF10A cells. Using Cis-regulatory Element Annotation System (CEAS), part of cistrome package (Liu et al., 2011a), the enrichment for BRD4 on specific genomic features (eg. Promoter regions, introns etc.) could be determined. It is estimated as the relative enrichment of ChIP regions in particular genomic feature with respect to the whole genome. As shown in Fig. 21, the pie chart depicts the distribution of BRD4 recruitment on various genomic

93 locations. It was observed that BRD4 is recruited to gene promoters and coding exons.

Fig. 21. BRD4 occupancy in the genome. Using CEAS, the enrichment of BRD4 at various genomic locations in Human hg19 was determined.

In order to determine the correlation between BRD4 recruitment and overall gene expression, genes were categorized based on their absolute expression levels seen in RNA-seq in control cells. Gene expression levels were based on RPKM (reads per kilobase of DNA per million reads) values where higher RPKM indicates higher expression (Mortazavi et al., 2008). The genes were then separated into categories : under 500 RPKM, 500-1500 RPKM, 1500-3000 RPKM, 3000-5000 RPKM and 10,000 RPKM where under 500 indicates the lowest expressed genes and 5000-10,000 indicates the highest expressed genes. Based on this classification, average BRD4 recruitment was determined via aggregate plot analysis (using CEAS) (Liu et

94 al., 2011a) to deduce the average signal profiles around transcriptional start sites of genes in the various gene expression-dependent groups. In this case, BRD4 recruitment near the transcription start site (TSS) (± 3kb relative to TSS) varied according to the expression levels of the genes (Fig. 22A). The highly expressed genes had more BRD4 recruitment whereas the lowly expressed genes had the least BRD4 recruitment. A similar gene expression-dependent pattern was observed for H2Bub1, further strengthening the correlation between the two (Fig. 22B). Since H2Bub1 is known to be an important histone modification coupled with transcription elongation and often associated with the transcribed region (Minsky et al., 2008b), H2Bub1 recruitment was also observed across the gene bodies (Fig. 22C). Similarly, the highly expressed genes possessed higher H2Bub1 while the lowly expressed genes had lower H2Bub1 levels in the transcribed regions. These results are consistent with the previously published data for the presence of H2Bub1 in the transcribed region (Minsky et al., 2008b). Based on the published ChIP-seq data for RNAPII (SRR488765) (Baillat et al., 2012), H3K9/14ac (SRR398030) and H3K4me3 (SRR398029) (Choe et al., 2012), aggregate plot analyses were performed on the genes classified as in Fig. 22A. Consistent with BRD4 and H2Bub1 recruitment, there was similar pattern for RNAPII, H3K9/14ac and H3K4me3 on the TSS of these genes correlating well with the expression of these genes (Fig. 22D).

95 Fig. 22. BRD4 and H2Bub1 occupancy correlate with the expression of the gene. Genes were sorted into various categories based on their RPKM values from RNA seq in siCont sample. Using CEAS, the average signal intensity at TSS (3kb upstream and downstream) for the proteins or histone modifications correlating with the expression levels was evaluated. (A-B) Average BRD4 and H2Bub1 recruitment around the TSS of the genes classified on the basis of expression. (C) H2Bub1 signal across the gene bodies correlating with the gene expression. (D) Using published data for RNAPII, H3K9/14ac and H3K4me3, their occupancy correlated with the determined gene expression list. The analysis was performed by Prof. Steven A. Johnsen, UKE, Hamburg.

Heatmap analysis diagrammatically depicts the recruitment signals of ChIP-seq at various genomic locations. The genomic locations are clustered together based on k-means clustering into clusters having similar patterns. The heatmaps illustrating the recruitment of BRD4, RNAPII, H3K4me3, H3K9/14ac and H2Bub1 near all known TSS (UCSC genes) were analyzed (Fig. 23A). Consistent with the above data, there was recruitment of these proteins and histone modifications to the same sites. To further validate the data from heatmap, BRD4 binding sites were classified into proximal (close to TSS) and distal (greater than 10kb away from any known TSS) sites. Aggregate plot analyses of average recruitment to proximal or distal sites for BRD4, H3K4me3 and H3K9/14ac showed increased recruitment to proximal sites, consistent with their promoter binding. Surprisingly, H2Bub1 as well as substantial fraction of BRD4 were highly recruited to the distal sites as compared to proximal sites (Fig. 23B). These observations strengthen the role of BRD4 in enhancer function and suggest a potential role for H2Bub1 in enhancer activity. To further

96 determine the role of BRD4 and H2Bub1 in enhancer function, the distribution of BRD4, H3K4me3, H3K4me1 (SRR398028) (Choe et al., 2012) and H2Bub1 on distal BRD4 binding sites were also plotted (Fig. 23C). As depicted in Fig. 23B, the aggregate plots of H2Bub1, H3K4me1 and BRD4, all indicate their presence at enhancers. However, heatmaps suggest these are different classes, marked by only H2Bub1 (transcribed region), both H3K4me3 and H3K4me1 (active genes) and absence of H3K4me3 but presence of H3K4me1, BRD4 and H2Bub1 (possible enhancer sites). Till date, H2Bub1 has been associated with transcription elongation and transcribed region but this novel function of H2Bub1 on enhancer has never been determined.

97 Fig. 23. Proximal and distal occupancy of BRD4 and H2Bub1. (A and C) Heatmap depicting the occupancy of BRD4 and H2Bub1 at the proximal and distal sites with k-means clustering where k=5.

(B) Aggregate plot analysis was performed to determine the BRD4 and H2Bub1 occupancy at proximal and distal sites (greater than 10kb away from any known TSS).

4.2.3 RNA Sequencing Reveals Regulation of Mammary Stem Cell Gene