Risk alleles for systemic lupus erythematosus in a large case-control collection and associations with clinical subphenotypes

Abstract

Systemic lupus erythematosus (SLE) is a genetically complex disease with heterogeneous clinical manifestations. Recent studies have greatly expanded the number of established SLE risk alleles, but the distribution of multiple risk alleles in cases versus controls and their relationship to subphenotypes have not been studied. We studied 22 SLE susceptibility polymorphisms with previous genome-wide evidence of association (p < 5 x 10⁻¹²⁸) in 1919 SLE cases from 9 independent Caucasian SLE case series and 4813 independent controls. The mean number of risk alleles in cases was 15.1 (SD 3.1) while the mean in controls was 13.1 (SD 2.8), with trend p = 4 x 10⁻⁸. We defined a genetic risk score (GRS) for SLE as the number of risk alleles with each weighted by the SLE risk odds ratio (OR). The OR for high-low GRS tertiles, adjusted for intra-European ancestry, sex, and parent study, was 4.4 (95% CI 3.8-5.1). We studied associations of individual SNPs and the GRS with clinical manifestations for the cases: age at diagnosis, the 11 American College of Rheumatology classification criteria, and double-stranded DNA antibody (anti-dsDNA) production. Six subphenotypes were significantly associated with the GRS, most notably anti-dsDNA (OR(high-low) = 2.36, p = 9e-9), the immunologic criterion (OR(high-low) = 2.23, p = 3e-7), and age at diagnosis (OR(high-low) = 1.45, p = 0.0060). Finally, we developed a subphenotype-specific GRS (sub-GRS) for each phenotype with more power to detect cumulative genetic associations. The sub-GRS was more strongly associated than any single SNP effect for 5 subphenotypes (the above plus hematologic disorder and oral ulcers), while single loci are more significantly associated with renal disease (HLA-DRB1, OR = 1.37, 95% CI 1.14-1.64) and arthritis (ITGAM, OR = 0.72, 95% CI 0.59-0.88). We did not observe significant associations for other subphenotypes, for individual loci or the sub-GRS. Thus our analysis categorizes SLE subphenotypes into three groups: those having cumulative, single, and no known genetic association with respect to the currently established SLE risk loci.

Figure 1. Distributions of the number of risk alleles and genetic risk score (GRS) by disease status, anti-dsDNA status, and age at diagnosis high-low tertiles.

A) the number of risk alleles in cases and controls; B) the GRS in cases and controls; C) the number of risk alleles in anti-dsDNA positive versus negative cases; D) the GRS in anti-dsDNA positive versus negative cases; E) the number of risk alleles in low versus high age at diagnosis tertiles; and F) the GRS in low versus high age at diagnosis tertiles.

Figure 2. Odds ratio of GRS intervals, adjusted for four principal components, two parent studies, and gender.

Sample sizes for each interval are shown below the graph.

Figure 3. ROC curves for four anti–dsDNA models.

Four models shown: 1) sex and disease duration alone, 2) adding top locus (STAT4) to first model, 3) adding GRS to first model, and 4) adding sub-GRS to first model.

Figure 4. Categorization of SLE subphenotypes by strongest association with currently known susceptibility loci: genetic risk score, single locus, or none.

Figure 5. Association of subphenotypes with sub-GRS candidates in study 1, by number of included SNPs.

P-values are for likelihood-ratio test of models with sub-GRS plus covariates vs. covariates alone.