Enhancer (sort of) Journal Club: Detection and replication of epistasis influencing transcription in humans

I thought I would write a short journal club on a paper that I really enjoyed reading recently. It isn't as directly related to enhancers as previous papers I've covered, but I think it has some interesting implications for the field. The paper is called "Detection and replication of epistasis influencing transcription in humans" and comes from Peter Visscher's lab at the University of Queensland. The original paper can be found at: http://www.nature.com/nature/journal/vaop/ncurrent/full/nature13005.html and the abstract is as follows:

Epistasis is the phenomenon whereby one polymorphism's effect on a trait depends on other polymorphisms present in the genome. The extent to which epistasis influences complex traits and contributes to their variation is a fundamental question in evolution and human genetics. Although often demonstrated in artificial gene manipulation studies in model organisms, and some examples have been reported in other species, few examples exist for epistasis among natural polymorphisms in human traits. Its absence from empirical findings may simply be due to low incidence in the genetic control of complex traits, but an alternative view is that it has previously been too technically challenging to detect owing to statistical and computational issues. Here we show, using advanced computation and a gene expression study design, that many instances of epistasis are found between common single nucleotide polymorphisms (SNPs). In a cohort of 846 individuals with 7,339 gene expression levels measured in peripheral blood, we found 501 significant pairwise interactions between common SNPs influencing the expression of 238 genes (P < 2.91 × 10(-16)). Replication of these interactions in two independent data sets showed both concordance of direction of epistatic effects (P = 5.56 × 10(-31)) and enrichment of interaction P values, with 30 being significant at a conservative threshold of P < 9.98 × 10(-5). Forty-four of the genetic interactions are located within 5 megabases of regions of known physical chromosome interactions (P = 1.8 × 10(-10)). Epistatic networks of three SNPs or more influence the expression levels of 129 genes, whereby one cis-acting SNP is modulated by several trans-acting SNPs. For example, MBNL1 is influenced by an additive effect at rs13069559, which itself is masked by trans-SNPs on 14 different chromosomes, with nearly identical genotype-phenotype maps for each cis-trans interaction. This study presents the first evidence, to our knowledge, for many instances of segregating common polymorphisms interacting to influence human traits.

The study is a sort of spiritual cousin of the recent explosion of Genome Wide Association Studies (GWAS). In a normal GWAS study, one measures variations in DNA sequence (generally single nucleotide changes, or SNPs) and looks for correlations to an aspect of phenotype, for example susceptibility to a disease, or expression of a certain gene. Generally, these studies only have the statistical power to ask whether a certain sequence variant has an independent effect on the outcome - i.e. does having an A instead of a T at position x lead to an increased chance of heart disease? Recently, it has become clear that many of the sequence variants identified in these studies may actually be changes in the sequence of cis-regulatory regions such as enhancers (for a nice review on this, see Sakabe, Savic and Nobrega 2012 - http://genomebiology.com/2012/13/1/238). In this paper, the authors are looking for pairwise SNP interactions associated with heritable gene expression. In other words, "having an A instead of a T at position x increases expression of this gene, but only if you also have a C instead of a G at position y".

What they find in the course of their study is (for me) super interesting. They find 501 of these pairwise interactions, where 26 are cases in which both SNPs lie within 1Mb of the promoter (the authors call these cis-cis acting), 13 are cases in which both SNPs are farther than 1Mb from the promoter (trans-trans) and 462 are cases in which one of the SNPs is within 1Mb and the other is not (cis-trans). Of course, the caveat with this and all studies using SNP arrays is that the SNP you observe may not have a functional effect, and may only be genetically linked to a functional sequence variant by virtue of being close by it in the genome.

I find it very striking that they find so many interactions over 1Mb away. One clear interpretation of the data is that the trans acting SNPs are actually some form of distal enhancer or repressor. Indeed, the authors present evidence that these SNPs are significantly enriched within 5mb of long-range interactions detected by Hi-C. If this is true, it would involve a huge expansion of the set of currently known ultra-long range regulatory elements, as in the current state of knowledge only a handful of enhancers that act over distances greater than 1Mb have been characterized. One other clear interpretation is that these SNPs actually control the expression of other genes, and it is the gene product that regulates the activity of the cis-SNP. It would be interesting to test whether SNPs which did not show an independent effect on the expression of the gene to which they are epistatic had independent effects on the expression of any other gene. My reading of the paper is that they did not do this (although I could be missing something), but in either case they only analyzed probes that mapped to RefSeq genes and would therefore be missing any regulation that might be mediated through non coding RNA expression.

In any case, it is interesting that at least some of the cis-trans interactions they detect may be enhancer/promoter pairs. In this case, the implication would be that the effect of a particular enhancer sequence can actually be dependent on the sequence of the promoter that it controls. Currently, many enhancers are identified or verified through luciferase assays, where we test the ability of a DNA sequence to affect the transcription of the luciferase gene. I think this finding certainly highlights that these assays may be more sensitive to the choice of promoter than perhaps was previously thought. Conversely, they identify a large number of cases where the trans-SNP has a masking effect on the cis-SNP. In other words there is a cis-SNP which is associated with decreased expression of the gene, but the effect is not observed in the presence of the trans-SNP. This possibly points to some redundancy in the system, by which these regulatory elements serve to maintain the correct expression level of the gene in the presence of deleterious promoter mutations. This would mean that one could not observe the effect of these regulatory elements by deleting them.

All in all, I think this is a really nice paper and certainly highlights a lot of issues that deserve some more thought and research.