A genome-wide association study identifies susceptibility loci for ovarian cancer at 2q31 and 8q24

Abstract

Ovarian cancer accounts for more deaths than all other gynecological cancers combined. To identify common low-penetrance ovarian cancer susceptibility genes, we conducted a genome-wide association study of 507,094 SNPs in 1,768 individuals with ovarian cancer (cases) and 2,354 controls, with follow up of 21,955 SNPs in 4,162 cases and 4,810 controls, leading to the identification of a confirmed susceptibility locus at 9p22 (in BNC2). Here, we report on nine additional candidate loci (defined as having P ≤ 10⁻⁴) identified after stratifying cases by histology, which we genotyped in an additional 4,353 cases and 6,021 controls. We confirmed two new susceptibility loci with P ≤ 5 × 10⁻⁸ (8q24, P = 8.0 × 10⁻¹⁵ and 2q31, P = 3.8 × 10⁻¹⁴) and identified two additional loci that approached genome-wide significance (3q25, P = 7.1 × 10⁻⁸ and 17q21, P = 1.4 × 10⁻⁷). The associations of these loci with serous ovarian cancer were generally stronger than with other cancer subtypes. Analysis of HOXD1, MYC, TIPARP and SKAP1 at these loci and of BNC2 at 9p22 supports a functional role for these genes in ovarian cancer development.

Figure 1

Genomic architecture of the four novel ovarian cancer susceptibility regions identified from the EOC-GWAS. Key: Clear triangles- the location of SNP with the most statistically significant association at each locus; Grey triangles- the location of SNPs correlated with the associated SNP with r²>0.8 (a) The 8q24.21 locus: the most significant SNP (rs10088218) lies >700kb distal to CMYC in an otherwise ‘gene desert’. Lower resolution map: rs10088218 is located with respect to 13 other SNPs (black triangles) significantly associated with susceptibility to prostate cancer (PrC), colorectal cancer (CrC), breast cancer (BrC), urinary bladder cancer (UBC) and renal disease (ReD). Higher magnification map: rs10088218 is located with respect to 10 other HapMap SNPs correlated with rs10088218 (b) The most significant SNP at 3q25.31 (rs2665390) lies within TiPARP, the only gene within a ∼200kb region spanning this SNP. Higher resolution map: rs2665390 location with respect to the only other highly correlated HapMap SNP, which is also in TiPARP. (c) 2q31.1 contains the HOXD gene family. The most significantly associated SNP (rs2072590) lies in a non-coding region ∼5kb distal to HOXD3 and ∼10kb proximal to HOXD1. Higher resolution map: The location of rs2072590 with respect to two correlated SNPs, one 3′ of HOXD3, the other distal to HOXD4. (d) 17q21.31 contains rs9303542, which is located in the intron of SKAP1, which lies distal to the HOXB family of transcription factors. This SNP is correlated with several other SNPs all located in SKAP1.

Figure 2

Gene expression analysis for five candidate EOC susceptibility genes: (a) BNC2 (9p22); (b) TiPARP (3q25); (c) CMYC expression (8q24); (d) HOXD1 (2q31), and (e) SKAP1 (17q21). Two different types of analysis are shown: the first panel of each pair shows real time RT-PCR analysis of each gene comparing gene expression between 48 POE and 24 OC cell lines. Gene expression is shown was normalised against b-actin. Expression was also normalized against GAPDH, which gave similar results (supplementary figure 2). Error bars sow standard deviation. The second panel of each pair shows expression microarray data from an in vitro stepwise model of neoplastic transformation POE cells. Briefly, gene expression in immortalised POE (iOE) cells, verses iOE^cmyc cells overexpressing CMYC, verses iOE^cmyc.kras cells overexpressing CMYC and mutant KRAS. (a) BNC2 expression is significantly lower in OC cell lines verses normal POE cells for both endogenous controls. In the model of neoplastic transformation, BNC2 expression decreases significantly with each additional oncogenic event, as the cells acquire a more neoplastic phenotype. (b) TiPARP shows significantly reduced expression in OC cells compared to POE cells for both endogenous controls; TiPARP expression also decreases sequentially in the neoplastic transformation model. (c) CMYC expression is significantly increased in OC compared to POE cells for both endogenous controls and increases in the transforming cell line model. (d) Expression of HOXD3 and HOXD1 was evaluated. HOXD3 expression was too low to draw reliable conclusions. HOXD1 increased in OC compared POE cells, suggesting activation in OC development. (e) SKAP1 expression was significantly greater in OC compared to POE cells. A trend towards increased expression was also observed in progressively transforming iOE cells.