Enhancer Journal Club: Locus-specific editing of histone modifications at endogenous enhancers.

Moving away from enhancer/promoter looping for a little while, I came across this nice paper from the Bernstein lab at the Howard Hughes Medical Institute. This paper is about assigning enhancers to their cognate genes, and also has some interesting implications for the causality of histone modifications on active enhancers.

A non-technical summary is available at ScienceGist: http://sciencegist.com/doi/10.1038/nbt.2701

The abstract for the paper is as follows:

Mammalian gene regulation is dependent on tissue-specific enhancers that can act across large distances to influence transcriptional activity. Mapping experiments have identified hundreds of thousands of putative enhancers whose functionality is supported by cell type–specific chromatin signatures and striking enrichments for disease-associated sequence variants. However, these studies did not address the in vivo functions of the putative elements or their chromatin states and did not determine which genes, if any, a given enhancer regulates. Here we present a strategy to investigate endogenous regulatory elements by selectively altering their chromatin state using programmable reagents. Transcription activator–like (TAL) effector repeat domains fused to the LSD1 histone demethylase efficiently remove enhancer-associated chromatin modifications from target loci, without affecting control regions. We find that inactivation of enhancer chromatin by these fusion proteins frequently causes downregulation of proximal genes, revealing enhancer target genes. Our study demonstrates the potential of epigenome editing tools to characterize an important class of functional genomic elements.

First of all, I think the technology and methodology of the paper is extremely exciting. Since custom designed DNA binding proteins started to become common for genome editing, it has only been a matter of time before people start using the techniques to do some really cool molecular biology, and I think we are going to see some really exciting work in this area over the next decade or so. Papers using synthetic fusion proteins to study enhancer mechanisms in more molecular detail than was previously possible have started to trickle out recently, my favourite being Gert Blobel's demonstration that enhancer/promoter looping can induce gene expression (http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3372860/).

In this paper, Mendenhall et al. fuse an engineered TAL effector DNA binding domain to the chromatin modifier LSD1. By designing the TALE domain to bind to a predicted enhancer sequence, they can do in vivo targeting of H3K4 demethylase activity. This is great for enhancers as H3K4me1 and H3K4me2 are two of the marks commonly used to identify putative enhancer regions which are active in any specific cell line. For 40 tested LSD1/TALE fusion constructs targeted to different enhancers, they find that 26 show a more than 2 fold reduction in either H3K4me2 or H3K27Ac. This is a nice validation that the method appears to be reasonably efficient, but one interesting point here is that the effect on H3K27 acetylation seems to be larger than that on H3K4me2. This most likely indicates that LSD1 is recruiting additional proteins with HDAC activity, but the possibility remains that H3K4 methylation may lie upstream of this acetylation.

Having established that they are able to 'edit' the histone modifications at 26 putative enhancer loci, they then choose 9 of their constructs and look for changes in local gene expression using "a modified RNA-seq procedure termed 3' digital gene expression (3'DGE)". Of these nine, they found significantly downregulated genes in four cases. The paper is a little confusing here, as the main figure legend says the results are for "the closest expressed upstream and downstream genes", but the methods say they examined the three closest upsteam and three closest downstream genes, and the supplementary figure actually shows between 10 and 25 genes measured per targetting construct. In any case, there is plenty of literature showing that enhancers can act over really quite large distances and since they measured expression from the whole transcriptome it would be nice to see, for example, the expression of all genes within a megabase. Perhaps this might explain why they only identify candidates in four out of nine cases.

All in all, I really like this paper. It's nice to see a mechanistic identification of enhancer target genes, although some would argue that targeting eRNAs with an shRNA would have the same effect and would be potentially quicker/cheaper for identifying enhancer targets on a wider scale. I think the paper would benefit from really nailing down the order of causality - since the LSD1 construct seems to be interfering with H3K27 acetylation, it seems likely that the observed effects depend upon recruitment of other factors, at least in part. The question is which effects are due to the change in H3K4 demethylation and which are due to recruitment of extra factors by LSD1. Here I think it would have been really nice to have a catalytically inactive LSD1 fused to a TALE domain. If you saw no effect on gene expression with this construct, I think you would have very strong evidence that H3K4me2 actually plays a causal role in determining enhancer activity.

Original paper available at: http://www.nature.com/nbt/journal/vaop/ncurrent/full/nbt.2701.html

As always, let me know what you think about this paper in the comments!