It has long been recognised that DNA elements located sometimes hundreds of kilobases away from gene promoters can influence their expression in cis. The most commonly explored class of these cis-regulatory elements are enhancers, which by definition contribute to activating gene expression. However, some elements have the opposite effect – they negatively affect gene expression in cis, and have therefore been named “silencer” elements. Silencers are completely unexplored – we lack a systematic study of not only where they are in the genome but also of what they are.
The authors produced two atlases of putative silencer elements that are cell-type specific, composed of ~1.5 million elements in the human genome, and ~1 million in mouse.
How were they identified? Assuming that silencers need to bind repressor factors and will therefore correspond to DNase hypersensitivity sites (DHS), the authors considered DHS from 82 human cell lines (ENCODE+Roadmap) or from 22 mouse cell lines (ENCODE) and then excluded all that overlapped with chromatin features of enhancers (H3K4me1), promoters (H3K4me3 and TSS) or insulator elements (CTCF binding). The remaining DHS were filtered for cell-type specificity and considered “putative silencer elements” – each cell type contains an average of tenths of thousand elements.
Characteristics of (putative) silencers
|Binding sites for repressor transcription factors (TFs)
||Enriched for binding of REST, YY1, ZBTB33, SUZ12 and EZH2 (based on ChIP-seq) and also for motifs of other repressor TFs
||Many silencers are hypermethylated (contrary to enhancers, which are normally un- or lowly methylated). On average, silencers are 75% methylated versus 43% for enhancers
|Chromatin signatures (Roadmap)
||Most silencers fall into the uncharacterised category (not annotated/lack of signal). Otherwise the three next most enriched categories correspond to weak transcription, weak polycomb and heterochromatin
||No or very poor conservation across 100 vertebrates analysed (enhancers are poorly conserved as well)
|Promoter interactions (promoter capture Hi-C)
||Majority of promoter interactions correspond to non-expressed or lowly expressed genes. Overall expression of all genes interacting with silencers is significantly lower than for genes interacting with enhancers.
|Disease traits (SNPs)
||25% of disease traits present are enriched at silencers, similar to other cis-regulatory elements
Validation of silencer elements: The authors performed MPRA/STARR-seq (click here for a review on high-throughput approaches to test cis-regulatory elements) with 7430 putative silencers, 20 “known” silencers, 20 known enhancers and 20 random regions. Median expression levels of the reporter promoter cloned together with putative silencers was similar to that of random regions and close to 1 (i.e. no change in promoter activity); 51% of putative silencers did induce lower levels of expression compared to controls – these were considered true silencers and motifs found de novo on these were closest to REST and YY1 known motifs.
Interestingly, the authors also found that a subset of silencers can act as enhancers (or as part of enhancer elements) in other cell types: 25-65% silencers (in human; mouse: 1-55%) showed enhancer chromatin features in at least one other cell type. In the cell types in which silencers have enhancer features, the expression of promoters with which they interact is significantly higher than in the cell lines where silencers were identified.
What I like about this preprint
Most of us have heard about enhancer elements, but the idea that some cis-regulatory elements might actually work as repressors is less well-appreciated. During my PhD project, I have come across such elements while trying to find enhancers – they even displayed enhancer signatures but then had a negative effect on target genes (in cis). I’m still trying to understand how they might be working. So these atlases of putative silencers represent a valuable starting point to begin exploring the biology of these elements.
Genetics is still in my view the golden standard to test and define cis-regulatory elements – it will thus be very important to genetically dissect these silencer elements at their endogenous loci and verify their role as cis-repressors of gene expression. Maybe the easiest approach would be to select an amenable cell type for genomic engineering and target the top ten strongest silencers (as identified with MPRA/STARR-seq) specific for that cell type.