Fig 1.
Analysis flowchart of the study.
Table 1.
Characteristics of 119 Scandinavian men included in the mQTL analysis.
Fig 2.
Associations between SNPs and DNA methylation in human adipose tissue.
A genome-wide mQTL analysis in human adipose tissue was performed by associating SNPs with DNA methylation of CpG sites located in either cis (≤500 kb) or trans. Boxplots of (a) the top cis-mQTL, (b) the bottom cis-mQTL, and (c) the top cis-mQTL where the SNP introduces or removes a CpG site (CpG-SNP), showing significant associations between SNPs (genotype groups, x-axis) and DNA methylation of CpG sites (%, y-axis). (d-e) The frequency of associations (y-axis) is plotted in relation to the relative distance between SNPs and CpG sites (kb, x-axis) of significant cis-mQTLs. In (d) the full cis-mQTL distance of 500 kb is represented and the frequency of significant cis-mQTLs within each distance bin of 10kb are plotted, and, in (e) the region of 0-50kb is zoomed and the frequency of significant cis-mQTLs within in each distance bin of 1kb is plotted. (f) Histogram showing the strength of association (-log10 p-value, y-axis) in relation to distance between SNP and CpG site (kb, x-axis) of significant cis-mQTLs. The most frequent and strongest association signals of cis-mQTLs are shown within SNPs located close to CpG sites. (g-h) Boxplots of (g) the top trans-mQTL, and (h) the bottom trans-mQTL, showing significant associations between SNPs (genotype groups, x-axis) and DNA methylation of CpG sites (%, y-axis). pcorr, p-values have been corrected for multiple testing by a modified Bonferroni correction where the LD structure of SNPs is taken into account (see methods).
Table 2.
Number of significant mQTL results in human adipose tissue.
Fig 3.
Distribution of CpG sites of significant mQTLs in relation to genomic regions.
We examined the chromosomal and genomic distribution of CpG sites in significant mQTLs in human adipose tissue. By using chi-squared-tests, we determined whether the observed frequency of significant CpGs in cis- or trans-mQTLs differs from the frequency of all analyzed CpG sites for a particular genomic region. The histograms show the distributions of CpGs in relation to (a) chromosomes, (b) nearest gene, and (c) CpG islands. *Frequencies, significantly different (over-represented) from what expected by chance. #Frequencies, significantly different (under-represented) from what expected by chance. Genomic region in relation to nearest gene includes: TSS 1500 and TSS 200 (sites located 1500–201 or 200–0 bases upstream of the transcription start site (TSS) respectively), 5’UTR, 1st exon, gene body, 3’UTR and intergenic region (not mapped to any of the other regions). Genomic region in relation to CpG island includes: CpG island, shore (flanking region of CpG island, 0–2000 bp), shelf (flanking region of shore, 2000–4000 bp distant from CpG island) and open sea (not mapped to any of the other regions).
Fig 4.
mQTLs in adipose tissue capture reported disease loci.
Depiction of some identified mQTLs in adipose tissue of previously reported GWAS loci associated with obesity: (a) POMC / ADCY3, (b) GIRP, and (c) PARP4; lipid profiles, waist and metabolic syndrome: (d) CETP, (e) APOA5, (f) LEPR, (g) SORT1, (h) GCKR and (i) FADS2; and metabolic traits: (j) ACADS and (k) GRB10. ADCY3 locus and LEPR loci were identified through proxy SNPs based on LD.
Fig 5.
mQTLs affect gene expression in human adipose tissue.
Significant mQTL SNP-CpG pairs where the SNP also shows significant association with gene expression in adipose tissue. The boxplots represent some identified mQTL SNPs and associations of the same loci with mRNA expression: (a) CHRNA5, (b) G6PC2, (c) GPX7, (d) RPL27A, (e) THNSL2 and (f) ZFP57. Annotations for these mQTLs are included in S1 Table.
Table 3.
Number of significant eQTL results in human adipose tissue.
Fig 6.
mQTLs in human adipose tissue affect metabolic phenotypes.
The boxplots show significant mQTL SNPs associated with metabolic phenotypes in our study cohort with p<0.05, and associations of these loci with DNA methylation in adipose tissue for (a) rs2523453, (b) rs7205804, (c) rs11603334.
Fig 7.
mQTLs/eQTLs in human adipose tissue affect metabolic phenotypes.
Significant mQTL SNPs associated with both gene expression and a metabolic phenotype, with boxplots showing associations of some of these loci with DNA methylation, gene expression and metabolic traits: (a) rs619824, (b) rs7349405, (c) rs7210728, (d) rs176095, (e) rs529488.
Fig 8.
Possible relationship models between genotype as a causal factor (G), DNA methylation as the mediator factor (M) and metabolic phenotype as the phenotypic outcome (P).
Table 4.
Identified cis-mQTLs where DNA methylation potentially mediates the interactions between a genotype and a phenotype in human adipose tissue.