Trans-eQTLs reveal that independent genetic variants associated with a complex phenotype converge on intermediate genes, with a major role for the HLA.

Disease and trait-associated SNPs are enriched for both cis- and trans-eQTLs.17% of SNPs, present on common SNP platforms, affect gene expression levels in cis or trans (at FDR of 0.05). This is substantially different from 1,167 SNPs that have been found associated with traits or disease: 40.4% affect gene expression in cis, while 5.7% of these SNPs affect gene expression in trans. These eQTL SNPs significantly more often than expected map within the HLA (13.8% of cis-eQTLs, 47.8% of trans-eQTLs, extreme value distribution p<1.1×10−16).

Type 1 diabetes associated SNPs both affect genes in cis and in trans.

Human leukocyte antigen (HLA) trait-associated SNPs affect gene expression levels in trans.Thirty-two trait-associated SNPs that map within the HLA are trans-acting on other genes. Trans-genes are indicated in red. Peripheral blood co-expression (Pearson correlation coefficient r≥0.19, p<10−11) between genes is indicated in light grey. Several trans-genes are co-expressed with HLA genes.

Pairs of SNPs that cause the same phenotype more frequently than expected also affect the same downstream genes.Various pairs of unlinked SNPs cause the same phenotype but also converge on the same downstream genes. a) When using cis- and trans-eQTLs, identified when controlling FDR at 0.05, 7 unique pairs of SNPs cause the same phenotype but also affect the same downstream gene. When controlling the FDR at 0.50 for the trans-eQTLs, 18 unique pairs of SNPs show this convergence. b) This is significantly higher than expected, determined using 100 permutations. c) The SNPs that affect these downstream genes in most instances explain a proportion of the downstream gene expression variation that is substantially higher than what their effect is on the eventual phenotypes.

Trait-associated SNPs show convergence on multiple genes.For several traits different and unlinked SNPs affect the same genes in cis or trans. For beta thalassemia three different loci affect hemoglobin (HBG2) gene expression (one in cis, indicated with gray arrow, two in trans (at FDR 0.05), indicated with red arrows). For mean corpuscular volume (MCV) the same trans-effects on HBG2 (at FDR 0.05) exist, but convergence is also apparent on ESPN, VWCE, PDKZ1IP1 and RAP1GAP (at FDR 0.50). For mean platelet volume (MPV) numerous trans-effects on genes involved in blood coagulation were identified. Two MPV loci (rs12485738 on 3p26.3 and rs11602954 on 11p15.5) both affect GP9, F13A1 and C19orf33 (at FDR 0.05) and SAMD14, GNG11 and VCL (at FDR 0.50). Peripheral blood co-expression (Pearson correlation coefficient r≥0.19, p<1.0×10−11) between genes is indicated in light grey.