Integration of disease association and eQTL data using a Bayesian colocalisation approach highlights six candidate causal genes in immune-mediated diseases

Abstract

The genes and cells that mediate genetic associations identified through genome-wide association studies (GWAS) are only partially understood. Several studies that have investigated the genetic regulation of gene expression have shown that disease-associated variants are over-represented amongst expression quantitative trait loci (eQTL) variants. Evidence for colocalisation of eQTL and disease causal variants can suggest causal genes and cells for these genetic associations. Here, we used colocalisation analysis to investigate whether 595 genetic associations to ten immune-mediated diseases are consistent with a causal variant that regulates, in cis, gene expression in resting B cells, and in resting and stimulated monocytes. Previously published candidate causal genes were over-represented amongst genes exhibiting colocalisation (odds ratio > 1.5), and we identified evidence for colocalisation (posterior odds > 5) between cis eQTLs in at least one cell type and at least one disease for six genes: ADAM15, RGS1, CARD9, LTBR, CTSH and SYNGR1. We identified cell-specific effects, such as for CTSH, the expression of which in monocytes, but not in B cells, may mediate type 1 diabetes and narcolepsy associations in the chromosome 15q25.1 region. Our results demonstrate the utility of integrating genetic studies of disease and gene expression for highlighting causal genes and cell types.

Figure 1.

Overall results of colocalisation analyses. No evidence for association with either trait is found in most cases. Where association is observed, it is mostly with a single trait. Convincing evidence for colocalisation exists in a small proportion of genes. Shown are three posterior densities (PP₀: causal variant for neither trait—black; PP₁ + PP₂: causal variant for one trait—blue; PP₃ + PP₄: causal variant(s) for both traits—red) of all pairwise comparisons performed between the diseases and gene expression (unstimulated monocytes and B cells, stimulated monocytes after 24-h IFN, 2-h LPS and 24-h LPS).

Figure 2.

In the gene SYNGR1, expression of ILMN_1810875 is captured by almost all isoforms but that of ILMN_1727805 is captured by only one isoform. Shown is gene context plot of SYNGR1 depicting the positions of three probes.

Figure 3.

Effect of prior probability p₁₂ on the prior and posterior support for colocalisation in regions where both disease and eQTL show evidence for association (PP₃ + PP₄ > 0.8). We repeated the colocalisation analysis with all prior probabilities fixed as specified in Materials and Methods, except p₁₂ which varied between 10⁻⁷ and 10⁻⁵. For each pairwise analysis in which PP₃ + PP₄ > 0.8, we plotted the number of SNPs in that region against the relative posterior support for H₄, defined by PP₄/(PP₃ + PP₄) (green), and similarly the relative prior support (orange).