Adipose tissue eQTL meta-analysis reveals the contribution of allelic heterogeneity to gene expression regulation and cardiometabolic traits

Complete characterization of the genetic effects on gene expression is needed to elucidate tissue biology and the etiology of complex traits. Here, we analyzed 2,344 subcutaneous adipose tissue samples and identified 34K conditionally distinct expression quantitative trait locus (eQTL) signals in 18K genes. Over half of eQTL genes exhibited at least two eQTL signals. Compared to primary signals, non-primary signals had lower effect sizes, lower minor allele frequencies, and less promoter enrichment; they corresponded to genes with higher heritability and higher tolerance for loss of function. Colocalization of eQTL with conditionally distinct genome-wide association study signals for 28 cardiometabolic traits identified 3,605 eQTL signals for 1,861 genes. Inclusion of non-primary eQTL signals increased colocalized signals by 46%. Among 30 genes with ≥2 pairs of colocalized signals, 21 showed a mediating gene dosage effect on the trait. Thus, expanded eQTL identification reveals more mechanisms underlying complex traits and improves understanding of the complexity of gene expression regulation.


Figure S2. Distribution of ADIPOQ expression for each study.
Density plot of ADIPOQ expression in each study for (A) all studies and (B) FUSION and GTEx separated by sex.The x-axis is the counts per million normalized for trimmed mean of M values.The y-axis is the density.Each study is represented by a different color as indicated by the legend.The vertical black dashed line is at 150 CPM which is the threshold for retaining samples in eQTL analyses.In C, 163 points with a distance to TSS greater than 600 were excluded.
Absolute Effect Size (Beta) A.  A.

Figure S8. pLI scores of eQTL genes with 1 through 5 or more signals.
(A) Proportion of genes in TwinsUK for each gene expression quartile with a pLI score ≥ 0.9.Quartile 1 indicates the genes with the lowest expression and quartile 4 indicates genes with the highest expression.(B) Proportion of genes in TwinsUK with a pLI score < 0.9 or ≥ 0.9 separated by signal number for each gene expression quartile.The darkest blue are the genes without an eQTL signal and the lightest blue are genes with five or more eQTL signals.For each expression level quartile, the proportion of genes with multiple signals was substantially lower for genes with pLI ≥0.9 than for genes with pLI <0.9: this trend was particularly pronounced in the highest expression category.(C) Proportion of genes in TwinsUK for each eQTL signal number with a pLI score ≥ 0.9.Each set of bar graphs is for a different gene expression quartile.
Genes with 1-5+ conditionally distinct signals Proportion of genes in TwinsUK with pLI score ≥ 0.9

Figure S3 .
Figure S3.Comparison of effect sizes in sex-stratified meta-analyses.(A)Effect sizes of lead variants per gene identified in the female marginal eQTL meta-analysis looked up in the male marginal eQTL analysis.(B) Effect sizes of lead variants per gene identified in the male marginal eQTL meta-analysis looked up in the female marginal eQTL analysis.Each point is a variant-gene pair.The linear regression lines are shown in blue and the black diagonal line indicates slope = 1.R 2 values are the Pearson correlations.

Figure S4 .Figure S5 .
Figure S4.Overlap of adipose and blood eQTL signals.Horizontal stacked bar charts show the proportion of significant adipose eQTL signals that overlap with significant blood eQTL signals (FDR ≤ 0.05).The top bar shows all adipose eQTL signals, the middle bar shows only primary adipose eQTL signals, and the bottom bar shows non-primary adipose eQTL signals.The gray indicates the proportion of adipose eQTL signals with a shared blood eQTL (LD r 2 ≥ 0.2), the purple indicates the proportion of adipose eQTL signals not shared with a blood eQTL (LD r 2 < 0.2), and white indicates the eQTL genes in adipose not tested in blood.

Figure S6 .
Figure S6.Characteristics of eQTL signal variants.Violin plots with inset boxplots of the (A) absolute value of the effect sizes of lead variants, (B) MAF, and (C) distance of the lead variants to the gene TSS for the indicated signals in order of discovery.Each plot represents 5 subgroups which are genes with only one signal, genes with exactly 2 signals, genes with exactly 3 signals, genes with exactly 4 signals, and genes with 5 or more signals.For the boxplots, the center line represents the median value, the box limits represent the upper and lower quartiles, whiskers represent the 1.5x interquartile range, and the black circles represent outliers.The black lines connect the median values of each signal group.In C, 163 points with a distance to TSS greater than 600 were excluded.

Figure S7 .
Figure S7.Heritability of genes with an eQTL grouped by gene expression.(A) Boxplots of heritability for genes with zero to five or more eQTL signals.Boxplots are split into subgroups based on the quartile of expression for the gene.Lowly expressed genes are in quartile 1 shown in white while highly expressed genes are in quartile 4 shown in darkest blue.The center line represents the median value, the box limits represent the upper and lower quartiles, whiskers represent the 1.5x interquartile range, and the black circles represent outliers.The black line connects the median of the boxplots.(B) Boxplots of heritability for genes separated by gene expression quartile.The number shown represents the median heritability value.

Figure S10 .
Figure S10.Proportion of eQTL signals colocalized with GWAS signals.Proportion of eQTL signals colocalized with a GWAS signal out of the total number of eQTL signals separated by signal number.(A) Includes all eQTL signals and (B) separates the eQTL signals by quartiles of P-value strength.The top left are the least significant eQTL signals, the top right are the second least significant eQTL signals, the bottom left are the second most significant eQTL signals, and the bottom right are the most significant eQTL signals.The standard error is shown by the error bars.

Figure S11 .
Figure S11.Mendelian randomization using MRLocus for select allelic series.Plots of the effect size of the GWAS signals (y-axis) versus the effect size of the eQTL signals (x-axis) from MRLocus.Six eQTL gene-trait pairs are included with at least 2 signals colocalized with each other.Each point represents a colocalized eQTL signal with standard error bars.The solid blue lines represent the overall slope of the effect of gene on the trait, and dotted blue lines represent the confidence intervals.

Figure S12 .Figure S15 .Figure S16 .
Figure S12.WHRadjBMI GWAS and ZNRF3 eQTL conditional signal plots.LocusZoom plots for WHRadjBMI signal 1 conditioned on signal 2 (top left) and WHRadjBMI signal 2 conditioned on signal 1 (bottom left).Plots for ZNRF3 signal 1 conditioned on signal 2 (top right) and ZNRF3 signal 2 conditioned on signal 1 (bottom right).The dots are colored by LD with the lead variant.Both plots are colored by the GWAS lead variant represented by a purple diamond.

Gene expression levels in METSIM surgical biopsies and pLI scores of eQTL genes with 1 through 5 or more signals.
Proportion of genes with a pLI score ≥ 0.9 separated by signal number for each gene expression quartile.The darkest blue are the genes without an eQTL signal and the lightest blue are genes with five or more eQTL signals.(D) Proportion of genes for each eQTL signal number with a pLI score ≥ 0.9.Each set of bar graphs is for a different gene expression quartile.
(A) Proportion of genes with the specified number of eQTL signals separated by gene expression quartiles.Quartile 1 indicates the genes with the lowest expression.The darkest blue are the genes without an eQTL signal and the lightest blue are genes with five or more eQTL signals.(B) Proportion of genes for each gene expression quartile with a pLI score ≥ 0.9.Quartile 1 indicates the genes with the lowest expression.(C)