Mammalian transcriptional hotspots are enriched for tissue specific enhancers near cell type specific highly expressed genes and are predicted to act as transcriptional activator hubs

Abstract

Background: Transcriptional hotspots are defined as genomic regions bound by multiple factors. They have been identified recently as cell type specific enhancers regulating developmentally essential genes in many species such as worm, fly and humans. The in-depth analysis of hotspots across multiple cell types in same species still remains to be explored and can bring new biological insights.

Results: We therefore collected 108 transcription-related factor (TF) ChIP sequencing data sets in ten murine cell types and classified the peaks in each cell type in three groups according to binding occupancy as singletons (low-occupancy), combinatorials (mid-occupancy) and hotspots (high-occupancy). The peaks in the three groups clustered largely according to the occupancy, suggesting priming of genomic loci for mid occupancy irrespective of cell type. We then characterized hotspots for diverse structural functional properties. The genes neighbouring hotspots had a small overlap with hotspot genes in other cell types and were highly enriched for cell type specific function. Hotspots were enriched for sequence motifs of key TFs in that cell type and more than 90% of hotspots were occupied by pioneering factors. Though we did not find any sequence signature in the three groups, the H3K4me1 binding profile had bimodal peaks at hotspots, distinguishing hotspots from mono-modal H3K4me1 singletons. In ES cells, differentially expressed genes after perturbation of activators were enriched for hotspot genes suggesting hotspots primarily act as transcriptional activator hubs. Finally, we proposed that ES hotspots might be under control of SetDB1 and not DNMT for silencing.

Conclusion: Transcriptional hotspots are enriched for tissue specific enhancers near cell type specific highly expressed genes. In ES cells, they are predicted to act as transcriptional activator hubs and might be under SetDB1 control for silencing.

Figure 1

Combinatorial binding events overlap across multiple cell types. A. schematic diagram of a genomic region with genome tracks of six ChIP sequencing samples in a cell type marking singletons (bound by only one transcription-related factor), combinatorials (bound by few transcription-related factors) and hotspots (peaks bound by more than five transcription-related factors). B. A table listing all 10 cell types used along with the number of transcription factors C. bar graph of fraction of peaks in hotspots across 10 cell types. D. Heatmap of Pearson’s correlation coefficients of peaks in 30 sets (three groups of 10 cell types) with a hierarchical clustering tree showing that combinatorials clustered according to the groups. E. box plots for peak heights for the three groups (singletons – red, combinatorials - green, hotspots - blue) for T cells and macrophages.

Figure 2

Hotspots preferentially bind in cell type-specific gene neighbourhoods. A. A heatmap of Pearson’s correlation coefficients of genes in 30 sets (three groups of 10 cell types) with a hierarchical clustering tree showing that singletons and combinatorials clustered together in a tight cluster while hotspot gene sets were very distinct from each other. B. Fraction of ES super enhancers defined by Whyte et al. [11] in three groups showing hotspots significantly overlapped with super enhancers. C. Fraction of high Oct4 bound regions in Oct4+/− compared to Oct4+/+ [12] ES cells in three groups showing hotspots enriched for high Oct4 bound regions in Oct4+/− and also pluripotent genes D. Functional enrichments of hotspot genes in 10 cell types along with p values showing that hotspot genes were enriched for cell type specific genes.

Figure 3

Hotspots are enriched for sequence motifs of multiple transcription factors. A. Bar graphs of fraction of peaks with a given sequence motif in the three groups (singletons – red, combinatorials - green, hotspots - blue) with the name of the sequence motif along with the sequence in IUPAAC format for three representative motifs for each of B cells, erythroid and HPCs (see Additional file 1: Figure S1 for the complete list of motifs). B. A table of cell type, transcription-related factors analysed by ChIP sequencing, transcription-related factors binding to more than 90% of hotspot peaks and statistically significant known sequence motifs found in hotspots for B cells, erythroid and HPCs (see Additional file 1: Figure S2 for the complete table with 10 cell types).

Figure 4

Hotspots lie near highly expressed genes and are enriched for enhancers in nucleosome flanked regions. A. Boxplots for the peak height for H3K4me3, RNA polymerase II using ChIP sequencing as well as average RPKM value using RNA sequencing data in MEL cells for the three groups (singletons – red, combinatorials - green, hotspots - blue) showing higher peak height for all chromatin modifications in hotspots. B. Bar graph of fraction of VISTA enhancers [19] and Boxplots for the peak height for H3K4me1, H3K27ac using ChIP sequencing in the three groups in MEL cells concluding hotspots were enriched for enhancers. C. The average density plot of MEL H3K4me1 in the 10 kb window for each of the three groups showing hotspots had a bi-modal H3K4me1 peak while singletons have a mono-modal peak.

Figure 5

ES hotspots are preferentially occupied by activators. A. Overlap of genes in three groups in ES Ng dataset with up- and down-regulated genes upon Oct4 knockdown in ES cells showing hotspots were enriched for only down-regulated genes. B. Fraction of up- and down-regulated genes upon Dnmt and Set1db knock out in three groups in ES Ng dataset showing hotspot enrichment only in Set1db KO down-regulated genes.