Dynamic Role of trans Regulation of Gene Expression in Relation to Complex Traits

Abstract

Identifying causal genetic variants and understanding their mechanisms of effect on traits remains a challenge in genome-wide association studies (GWASs). In particular, how genetic variants (i.e., trans-eQTLs) affect expression of remote genes (i.e., trans-eGenes) remains unknown. We hypothesized that some trans-eQTLs regulate expression of distant genes by altering the expression of nearby genes (cis-eGenes). Using published GWAS datasets with 39,165 single-nucleotide polymorphisms (SNPs) associated with 1,960 traits, we explored whole blood gene expression associations of trait-associated SNPs in 5,257 individuals from the Framingham Heart Study. We identified 2,350 trans-eQTLs (at p < 10^-7); more than 80% of them were found to have cis-associated eGenes. Mediation testing suggested that for 35% of trans-eQTL-trans-eGene pairs in different chromosomes and 90% pairs in the same chromosome, the disease-associated SNP may alter expression of the trans-eGene via cis-eGene expression. In addition, we identified 13 trans-eQTL hotspots, affecting from ten to hundreds of genes, suggesting the existence of master genetic regulators. Using causal inference testing, we searched causal variants across eight cardiometabolic traits (BMI, systolic and diastolic blood pressure, LDL cholesterol, HDL cholesterol, total cholesterol, triglycerides, and fasting blood glucose) and identified several cis-eGenes (ALDH2 for systolic and diastolic blood pressure, MCM6 and DARS for total cholesterol, and TRIB1 for triglycerides) that were causal mediators for the corresponding traits, as well as examples of trans-mediators (TAGAP for LDL cholesterol). The finding of extensive evidence of genome-wide mediation effects suggests a critical role of cryptic gene regulation underlying many disease traits.

Keywords: GWAS; cardiometabolic traits; causal variants; eQTLs; hotspots; mediation; trans.

Figure 1

Work Flow and Results Summary 42,271 SNPs associated with 1,960 traits were obtained from GRASP (at p ≤ 5 × 10⁻⁸). Whole blood samples were collected from 5,257 FHS participants. Genome-wide genotyping and mRNA expression levels were assayed. We correlated 39,165 GWAS SNPs (after filtering) with expression levels of 17,873 genes to identify expression quantitative trait loci (eQTLs). For SNPs having both local (cis) and remote (trans) regulation effects, we then tested whether the effect of trans-eQTLs was mediated through cis-eGenes. Finally, integrating genotype, gene expression, and phenotype data, we conducted causal inference testing to identify causal variants for eight cardiometabolic traits (BMI, systolic and diastolic blood pressure, LDL cholesterol, HDL cholesterol, total cholesterol, triglycerides, fasting blood glucose).

Figure 2

Mediation Mechanisms of eQTLs Genetic variants can affect traits through the following mechanisms: (1) missense SNP affects protein structure/function; (2) non-coding SNP affects gene expression (cis); (3) non-coding SNP affects remote (trans) gene expression directly or by (4) cis-eGene mediation of the trans-eQTL-trans-eGene association; or (5) reverse causality (trait has feedback effect on gene expression).

Figure 3

trans-eQTL Hotspots x axis denotes the chromosomal location of SNPs. y axis denotes the number of trans-eGenes.

Figure 4

cis- and trans-eQTLs in the ALDH2 Causal Module Gray boxes list traits associated with SNPs from GWASs, green boxes list SNPs, red box lists the cis-eGene, and blue boxes list trans-eGenes. Red edges represent cis-associations; green edges represent trans-associations.