Fine-mapping the HOXB region detects common variants tagging a rare coding allele: evidence for synthetic association in prostate cancer

Abstract

The HOXB13 gene has been implicated in prostate cancer (PrCa) susceptibility. We performed a high resolution fine-mapping analysis to comprehensively evaluate the association between common genetic variation across the HOXB genetic locus at 17q21 and PrCa risk. This involved genotyping 700 SNPs using a custom Illumina iSelect array (iCOGS) followed by imputation of 3195 SNPs in 20,440 PrCa cases and 21,469 controls in The PRACTICAL consortium. We identified a cluster of highly correlated common variants situated within or closely upstream of HOXB13 that were significantly associated with PrCa risk, described by rs117576373 (OR 1.30, P = 2.62×10(-14)). Additional genotyping, conditional regression and haplotype analyses indicated that the newly identified common variants tag a rare, partially correlated coding variant in the HOXB13 gene (G84E, rs138213197), which has been identified recently as a moderate penetrance PrCa susceptibility allele. The potential for GWAS associations detected through common SNPs to be driven by rare causal variants with higher relative risks has long been proposed; however, to our knowledge this is the first experimental evidence for this phenomenon of synthetic association contributing to cancer susceptibility.

Figure 1. Results of the HOXB locus fine-mapping analysis.

Upper Panel – Regional association plot of SNPs at the HOXB13 locus. Association data from the iCOGS dataset of 20,440 PrCa cases and 21,469 controls are shown with genotyped SNPs in red and imputed SNPs in green. The Bonferroni-adjusted level of significance is denoted by the red line. The G84E variant rs138213197 was genotyped in a smaller subset of 5500 PrCa cases and 4923 controls and is marked by the blue rectangle. Also indicated are the position of genes within this interval and the location of neighbouring recombination hotspots. Middle Panel – Intersection between the 5 SNPs significantly associated with PrCa and putative functional elements identified by the ENCODE project or regions of mammalian sequence conservation by PhyloP. Lower Panel – Pairwise Linkage Disequilibrium values for the 5 SNPs significantly associated with PrCa. r² values are shown in grey and D′ in red.