We have examined the possibility of using high-throughput gene chips to examine the allele-specific nature of open chromatin using FAIRE (illustrated in ). The study identified a functional SNP, rs7120118, where the minor C allele is enriched in open chromatin and associated with increased HDL-C. Although the level of significance for HDL-C levels was adequate for a SNP with an a priori
hypothesis, this would be much lower than required for genome-wide significance, highlighting the importance of combining functional studies with GWAS to identify candidate SNPs for disease or trait associations, particularly those with lower effect sizes, rare SNPs or small cohorts. Indeed, examining a recent meta-analysis of lipid traits in >100,000 individuals, rs7120118 did show a strong association with HDL-C levels (p
, ) although this was not reported as significant in the study 
, perhaps due to the strong LD in the region, with the association signals covering >29 genes. We have shown that the minor allele is associated with increased NR1H3
gene expression in heart tissue and aortic adventitia, adding to a previous genome-wide study revealing a significant association with rs7120118 and gene expression of NR1H3
in lymphoblast cells 
. From this data it can be postulated that rs7120118 directly affects a long-range regulatory element (>15 kb from NR1H3
TSS) in a non-tissue-specific manner, altering gene expression and HDL-C levels.
Association of rs7120118 with HDL-C in More Than 100,000 Individuals.
The principle of allele-specific FAIRE was previously applied by Gaulton et al
to examine the functionality of a single type II diabetes (T2D) GWAS SNP in TCF7L2
. The authors used FAIRE-seq to determine global tissue-specific regions of open chromatin in pancreatic tissue, followed by TaqMan allelic discrimination to ascertain the effect of a single putative functional SNP on open chromatin. They found that the allele conferring increased risk of T2D and higher gene expression was also associated with enrichment for open chromatin. Although successfully demonstrating the use of FAIRE to identify a causal SNP from a GWAS, the use of TaqMan would not be applicable for examining a large number of potentially functional SNPs. FAIRE-gen, in contrast is only restricted by the number of SNPs that can fit on a genotyping chip.
The action of IL-1β on chromatin structure, a cytokine known to induce proliferation of EBV-transformed lymphoblasts 
, was examined in this study to reveal further potential allele-specific differences in open chromatin under different environmental conditions. For rs7120118, an allele-specific effect was observed in both unstimulated and IL-1β-stimulated cell lines, although the effects were stronger in the IL-1β stimulated samples. The action of IL-1β activates NF-κB, potentially altering expression of transcription factors that bind to the regulatory region surrounding rs7120118. Indeed, a nearby cluster of transcription factor binding sites determined by ChIP-seq includes a site for c-Jun binding (); the JUN
promoter contains several NF-κB binding sites (UCSC Genome Browser hg19/NCBI37) 
, which may explain this enhanced effect. It could be hypothesised that the C allele that favours open chromatin allows for preferential access for known, or as yet uncharacterised, transcription factors, which would act as an enhancer for NR1H3
gene expression, and increased HDL-C levels.
In contrast, a potential allelic effect was observed with the promoter SNP rs2167079 (in complete LD with rs7120118), only in unstimulated cells. IL-1β is known to reduce expression of NR1H3
in HK-2 cells 
, and it could be postulated that IL-1β may lead to chromatin remodelling and a decrease in open chromatin at the NR1H3
promoter in lymphoblasts, accounting for the lack of allelic effect in the IL-1β-stimulated cells. Alternatively, the modest allele-specific chromatin effects from the unstimulated cell lines could simply represent false-positive findings.
Haplotype structure may also affect local chromatin, particularly where more than one SNP occurs in the same region of open chromatin. We have examined the variation surrounding rs7120118 using HapMap-derived genotypes for the lymphoblasts used in the Metabochip study. No further SNPs at the locus provided additional haplotypic information for the effects on open chromatin, suggesting that rs7120118, rather than a haplotype, is responsible for this observation.
To assess the reproducibility of the FAIRE-gen methodology, the two Metabochip datasets were examined, considering the second IL-1β-treated study as a replicate. Examining the SNPs showing an allele-specific effect on open chromatin from the untreated samples, following Bonferroni correction (p<5.2×107
127), 100% were replicated in the treated sample with significance set at p<0.05, (91% replicated with Pc
116), indicating the sensitivity of the assay. The sensitivity and specificity of the assay to identify true functional variants can only be accurately determined by further functional analysis of each putative SNP. The smallest detectable difference in allele-specific open chromatin for the SNPs reaching genome-wide Bonferroni cut-off in the Metabochip was 10% (rs75106522).
One limitation with FAIRE-gen, as opposed to FAIRE-seq is the dependence of the gene chip to contain all relevant SNPs for the trait under examination. For the recent custom-designed chips which contain dense markers and aim to include all SNPs that tag GWAS-identified markers for diseases and related traits, such as the Illumina Metabochip and Immunochip, this is less of a problem. Future genotyping chips containing all common SNPs associated with diseases/traits could potentially resolve this drawback. For determining the location of potential causal SNPs from a number of SNPs acting as proxies, FAIRE-gen is only able to identify single allele-specific SNPs if other proxies are not located within the same region of open chromatin. This can be illustrated for rs7120118, where a nearby SNP, rs2279239, is located only 4.6 kb away, and close to the same region of open chromatin (Figure S2
). This SNP shows a similar trend for allelic-specificity, although somewhat reduced due to the distance from the putative causal SNP.
Since the assay includes data from SNPs that are not present in open chromatin, there may also be a number of false-positive associations from the methodology, where amplification from background (non-open) chromatin may, in theory, preferentially occur for one allele. For this reason, replication using FAIRE-gen or FAIRE-seq in a separate study, and in vitro methodologies would be desirable to confirm functionality.
In conclusion, FAIRE-gen shows promise as an economical, high-throughput method to enable targeted unbiased detection of allele-specific regulatory elements, which may help to refine GWAS disease-association signals to identify disease-causing variants.