Analysis of five chronic inflammatory diseases identifies 27 new associations and highlights disease-specific patterns at shared loci

Abstract

We simultaneously investigated the genetic landscape of ankylosing spondylitis, Crohn's disease, psoriasis, primary sclerosing cholangitis and ulcerative colitis to investigate pleiotropy and the relationship between these clinically related diseases. Using high-density genotype data from more than 86,000 individuals of European ancestry, we identified 244 independent multidisease signals, including 27 new genome-wide significant susceptibility loci and 3 unreported shared risk loci. Complex pleiotropy was supported when contrasting multidisease signals with expression data sets from human, rat and mouse together with epigenetic and expressed enhancer profiles. The comorbidities among the five immune diseases were best explained by biological pleiotropy rather than heterogeneity (a subgroup of cases genetically identical to those with another disease, possibly owing to diagnostic misclassification, molecular subtypes or excessive comorbidity). In particular, the strong comorbidity between primary sclerosing cholangitis and inflammatory bowel disease is likely the result of a unique disease, which is genetically distinct from classical inflammatory bowel disease phenotypes.

Figure 1. 27 novel genome-wide significant disease associations (P_disease<5×10⁻⁸) for ankylosing spondylitis (AS), Crohn's disease (CD), psoriasis (PS), primary sclerosing cholangitis (PSC) and ulcerative colitis (UC)

Single disease analyses were performed only on SNPs that achieved P_SBM<5×10⁻⁷ in the primary (unconditioned) cross-disease subset-based association meta-analysis (SBM) approach (see Main Text). We identified 17 novel genome-wide significant susceptibility loci for AS, 6 loci for CD and 4 loci for PSC (Supplementary table 2). Corresponding P-values and ORs for each novel association are shown separately for each disease. With this, the number of known AS, IBD, and PSC risk loci increased to 48, 206, and 20, respectively. For 22 out of 27 gws associations, lead SNPs from the SBM approach (P_SBM<5×10⁻⁸) and the single disease lookups (P_SBM<5×10⁻⁷ and P_disease<5×10⁻⁸) are identical, in five instances we have different lead SNPs between SBM and the single disease analyses (Supplementary table 2). −log₁₀ P-value: −log₁₀ P-values (P_disease) from Immunochip analysis (Supplementary Table 2) with regard to the physical location of markers; direction of triangle denotes direction of disease-individual effect; OR: odds ratio from the five single disease vs. control subsearches (OR(disease) in Supplementary Table 2). Large circles denote nominal significant disease-individual P-values (P_disease<0.05); CAF cases/controls: case/control minor allele frequency; If available, the nearest gene within 10kb of the variant is depicted.

Figure 2. Heritability explained per risk variant from 244 independent multi-disease association signals identified through cross-disease subset-based association meta-analysis

(a) Cumulative fraction of explained variance in disease liability (heritability) for each disease (see Methods). The loci are ordered from largest to smallest individual contribution. In total, 7.38, 10.30, 5.72, 2.53, and 5.95 percent of the heritability of ankylosing spondylitis (AS), Crohn's disease (CD), psoriasis (PS), primary sclerosing cholangitis (PSC) and ulcerative colitis (UC), respectively, is explained by the 169 loci outside the extended MHC region (for a maximum of 244 independent signals from Supplementary Table 3a). When adding known risk alleles from the major histocompatibility complex (MHC)^- to the 169 loci, the cumulative variance increases to 27.82, 10.88, 12.20, 5.48 and 7.66 percent for AS, CD, PS, PSC and UC, respectively. (b) Example of a pair-wise comparison of heritability explained per risk variant between PS and UC. See Supplementary Figure 6 for all ten pair-wise comparisons. Even if disease-associations account for approximately the same amount of variance explained in disease liability, e.g. as seen here for PS and UC, the pattern of sharing is complex in terms of size and direction of effect. Each box represents an independently associated SNP for the given disease. The size of each box is proportional to the amount of heritability for that variant. The colors of the boxes denote whether the difference in variance explained is due to different direction of effect (risk versus protective), significant heterogeneity of odds ratios (P<0.01) or both.

Figure 3. Identification of drug targeting genes from a core disease protein-protein interaction network

We evaluated the potential role of genes in drug discovery by linking genes from a core protein-protein-interaction (PPI) network (Supplementary Figure 9) to drugs using Drugbank (see Methods). All drugs were selected based on evidence from phase I/II/III randomized clinical trials (RCTs) or published animal studies. Nine drug target genes overlap with the genes from the core network. (a) Connections between biological genes from core PPI (red), and drugs (blue) used for treatment of AS, CD, PS, PSC and UC (yellow). (b) Connections between biological genes from core PPI (red), and drugs (blue) used for treatment of other inflammatory disease and traits (yellow).

Figure 4. Estimation of Immunochip-wide pleiotropy (excluding the MHC region) between the five diseases under study

Proportion of genetic variance in liability (SNP-based heritability) and proportion of genetic covariance in liability between diseases (SNP-based coheritability) with 95% error bars (see Supplementary Table 15a; estimates including MHC SNPs see Supplementary Table 15b). Disease-associated SNP markers from Supplementary Table 3a (at a maximum of 244 independent signals from 169 non-MHC risk loci, see also Supplementary Figure 6) explain 42.2% of AS-, 79.63% of CD-, 39.6% of PS-, 29% of PSC- and 55% of UC-Immunochip-wide SNP-heritability (excluding the MHC region) on the liability scale, respectively. As the Immunochip densely tags common variants but at the cost of losing genome-wide coverage, the estimated SNP-heritabilities are lower bounds for genome-wide SNP-heritabilities.