A genomic scale investigation on enhancer-promoter interactions mediated by transcription factors.

Background Chromatin looping can model enhancer-promoter cooperation. This model considers that enhancers and promoters are close in the three dimensional space of the active chromatin, independently of the length of the separating nucleotide sequence. A chromatin loop is thought to be mediated by protein complexes mostly composed of transcription factors (TFs) bound to DNA and transcriptional cofactors. The interactions between these proteins (collectively referred to as “enhanceosome”) are responsible of RNA-polymerase II recruitment and transcriptional activation. Accordingly, enhanceosomes bind to DNA via transcription factors. Consequently, one may assume that part of the information needed for loop stabilization is stored in the same DNA molecule. Here we investigate whether co-occurrent pairs of transcription factors (TFPs) over paired enhancer-promoter sequences (EPPs) are important for loop formation and enhancer-promoter cooperation. Results We downloaded 702 experimentally validated developmental enhancers from VISTA [1] and the corresponding 621 promoters from UCSC, generating a database of 1011 EPPs. Transcription factor binding sites (TFBS) were predicted using profiles from Hocomoco [2]. A set of 2837 TFPs was derived, each composed by two TFs, one bound to the enhancer and the other to the promoter, respectively. With a binomial test, which positively scores TFPs found over EPPs, we found 364 TFPs involved in looping and generated their interaction network. We used the last release of the human interactome in StringDB [3] to compute the shortest paths among our paired TFs and we found that 93% of them are separated by zero to two proteins, corroborating our analysis. Furthermore, some of the proteins most frequently found in the shortest paths are part of the very few known enhancosomes (EP300, CREBBP). Conclusions The importance of co-occurrent TFs over single DNA sequences for transcription activation was known before. However, when cooperative looping mechanisms among enhancers and promoters are taken into account, co-occurrent TFs over both DNA sequences are important. Here we describe 364 TFPs that are likely to participate in looping formation. Our data are supported by: a) an enrichment analysis of the Biological Process Gene Ontology of the set of proteins including our TFPs and their bridging proteins in the human interactome, highlighting chromatin organization and organ development as very significant GO terms (qvalue≤10-18); b) the profile of expression transcription patterns (Gene Expression Atlas) is consistent with the simultaneous existence of the paired TFs in different tissues. References [1] Visel et al., VISTA Enhancer Browser--a database of tissue-specific human enhancers., Nucleic Acids Res. 2007 Jan;35(Database issue):D88-92. [2] Kulakovskiy et al., HOCOMOCO: a comprehensive collection of human transcription factor binding sites models., Nucleic Acids Res. 2013 Jan;41(Database issue):D195-202. [3] Jensen et al., STRING 8--a global view on proteins and their functional interactions in 630 organisms., Nucleic Acids Res. 2009 Jan;37(Database issue):D412-6