Epigenetic marks, especially DNA methylation, play an important role in cellular differentiation and gene regulation in adipose tissue 
. Moreover, genetic variants that influence DNA methylation may not only be a carrier for the inheritance of epigenetic marks 
, but also provide a biological mechanism by which SNPs act that have previously been shown to associate with disease phenotypes and/or downstream mRNA expression 
We first investigated whether DMH methylation associated with MetSyn status, BMI, age or gender. We find no associations of DMH methylation with MetSyn case control status or BMI. Changes in DNA methylation associated with BMI have been previously reported in a longitudinal study 
, but these effects might not have been observed in our study due to lower sample size. While we observed an association between methylation and gender, we find no association between methylation and age (Figure S1D
), which is in line with a previous report 
. The relatively modest sample size and narrow age distribution (SD
5 years, ) may impact power to detect age-related effects, which have been reported in the literature 
. However, our main objective of this study was meQTL analysis, where effect sizes are expected to be larger, and where we are better powered to detect significant associations (Supplementary Information S1)
, as our results also indicate.
The results of our meQTL study indicate that there are a number of cis
-meQTLs in abdominal SAT. Moreover, we observe that most cis
-meQTLs act over a much shorter distance than the arbitrarily defined cis
window of ±500 kb. In our replication study we attempted to replicate 19 out of 149 significant meQTL signals and find that five out of 19 DMH probe sets that reached significance in the primary study are also found to be significantly associated and directionally consistent in the replication study. The associations of methylation with SNPs in these five regions are consistent throughout the datasets, as each top cis-
meQTL SNP within a ±500 kb region also is an meQTL in the other data set (Table S5
). As methylation was measured on a different platform, the replication study also provided validation of the DMH method used in the primary study. There are three main aspects between the two platforms that may have an effect on the replication rate: (i) targeting of single vs. multiple CpGs, (ii) imperfect correlation between the DMH probe sets and Illumina 27k probes (within 1 kb) used as methylation proxies and (iii) methodological differences between the two assay techniques. Different genotyping scaffolds have been used and we use very stringent quality control measures pre-imputation, which means that even though different chips are used, the filtering on the minor allele count we do only allow high quality data, and variants with low allele frequency will be filtered out (Supplementary Information S1)
Importantly, in a recent report investing imputation accuracies using different methods and sizes of European reference panels, it was shown that for common variants with a minor allele frequency >5% the imputation accuracy performs similarly well 
. Given this we feel confident that our association results, both on direct genotypes as well as the imputation, are as robust as they can be to genotype errors. However, the fact that methylation associates with genetic variants consistently across the two studies despite these differences suggests that in these regions there is indeed genetic control of methylation in SAT.
No significant associations of the 149 meQTL SNPs with expression of cis
-mRNA transcripts are found, but association cannot be ruled out due to the relatively low statistical power in this study. Using similar assumption as above, and adjusting for 149 tests, we have 80% power to detect SNP-mRNA associations explaining 41% of variation in expression (Supplementary Information S1)
. Again, we are limited by the small sample size to detect associations of this magnitude.
Despite the current understanding that promoter methylation acts as a suppressor of mRNA translation, the observation that many meQTLs do not influence mRNA expression is consistent with the study in brain tissues by Gibbs et al.
. Bell et al.
, on the other hand observe a significant enrichment for eQTLs in the meQTL results. This shows that there is still a need for future studies to investigate the downstream biological effects of methylation and the role of meQTLs. Our results show a significant enrichment for meQTLs found in brain tissue 
, but not for those in HapMap LCLs 
, potentially due to the lower power in the latter study. Limited overlap between the DMH probe sets and the Illumina 27k microarray used in these studies, as well as tissue-specific differences, may explain the low overlap between the study results. Further research could reveal the true degree of meQTL tissue specificity by using larger sample sizes and consistent assay techniques.
We find that two cis
-mRNA’s significantly associated with meQTL DMH probe sets, encode for protein products that have previously been implicated in type 2 diabetes and MetSyn. TNFRSF11B
, also known as RANKL
, was previously characterized as an extracellular negative regulator of osteoprotegrin, which acts as a decoy receptor when secreted 
. A number of studies have found that both osteprotegrin and this TNF-superfamily protein have elevated serum levels in type 2 diabetes patients 
, and MetSyn 
is normally secreted by osteoblasts 
, also known as AST1
, is a liver transaminase that plays a role in amino acid metabolism, the urea cycle and the Krebs cycle 
. The GOT1
gene promoter has been shown to be regulated by glucocorticoids, cAMP and insulin 
. However, a role of these proteins in SAT has not been hypothesized or investigated previously.
Overall, we show for the first time that meQTLs are present in adipose tissue. This indicates a direct genetic influence on DNA methylation and also an indirect influence on the general molecular phenotype of adipose tissue. Defining the genetic influence on both gene expression and CpG methylation in abdominal adipose tissue can help towards characterising this type of tissue and understanding molecular pathways associated with obesity.