Abstract
Background Pinpointing causal risk alleles, carried on extended risk haplotypes, at susceptibility loci for common disease is challenging. We present a novel strategy to prioritise tag-SNPs at risk loci with the highest likelihood of biological relevance for laboratory functional studies. This strategy involves trans-ancestral mapping of risk-haplotypes, MAF-exclusion-mapping of discordant variants, followed by co-localisation of tag-SNPs with epigenetic annotation/identification of allele-specific transcription factor (TF) binding sites, using data from public databases such as RoadMap, ENCODE and HaploRegv4, as a surrogate for causality.
We illustrate the utility of this approach at the Ikaros zinc finger TF IKZF3, with a 194 kb associated European (EUR) SLE haplotype carrying 282 tag-SNPs, extending over multiple genes from the IKZF3 3 flanking-region into the upstream-region of ORMDL3.
Methods Align IKZF3 control haplotypes from our large EUR GWAS and five 1000G super-populations. Compare EUR-African MAF to delineate discordant tag-SNPs. Use SLE ImmunoChip data to exclude variants exhibiting >3% MAF and not associated (p>0.05) in African-Americans (AA). Undertake co-localisation analysis, using multiple tools (Coloc-stats webserver), to prioritise risk alleles overlapping DNAse I hotspots and chromatin modifications characterising active enhancers and/or promoters in blood cell-types (RoadMap data). Additional evidence of potential function was sought for chromatin looping (3D Genome Browser) and differential TF binding (HaploRegv4).
Results Trans-ancestral mapping reduced the risk-haplotype by 47% to 101 kb and the tag-SNPs by 28% to 140. 26 tag-SNPs exhibited association in both EUR and AA SLE ImmunoChip cohorts. All 26 variants reside in regions of open chromatin in LCLs. 15 of the tag-SNPs lie within regions involved in chromatin-looping events, bringing together the full-length promoter and I1 of a shorter IKZF3 isoform. We discovered allele-specific binding of Fox family members to risk alleles in the promoter (rs111678394) and variants in I1 of the shorter isoform: rs113730542 and rs112876941.
Conclusions We hypothesise that allele-specific binding of Fox TFs to risk alleles in the promoter and/or I1 of the shorter isoform IKZF3 may independently modulate the expression of both isoforms. Since the shorter isoform lacks the four zinc fingers responsible for DNA binding of Aiolos (E4–6), the shorter isoform is unable to directly bind DNA. Heterodimerisation between the two isoforms may therefore sequester the full-length isoform in a biologically inactive form. If Fox TFs stabilise chromatin loops within IKZF3 in an allele-specific manner, this may provide a mechanism where risk alleles influence the activity/availability of IKZF3 and promote autoimmunity.
Funding Source(s): Versus Arthritis, formerly Arthritis Research UK project grant (20265)
