Meta‐analysis of genome‐wide studies identifies WNT16 and ESR1 SNPs associated with bone mineral density in premenopausal women

Daniel L Koller, Hou‐Feng Zheng, David Karasik, Laura Yerges‐Armstrong, Ching‐Ti Liu, Fiona McGuigan, John P Kemp, Sylvie Giroux, Dongbing Lai, Howard J Edenberg, Munro Peacock, Stefan A Czerwinski, Audrey C Choh, George McMahon, Beate St Pourcain, Nicholas J Timpson, Debbie A Lawlor, David M Evans, Bradford Towne, John Blangero, Melanie A Carless, Candace Kammerer, David Goltzman, Christopher S Kovacs, Jerilynn C Prior, Tim D Spector, Francois Rousseau, Jon H Tobias, Kristina Akesson, Michael J Econs, Braxton D Mitchell, J Brent Richards, Douglas P Kiel, Tatiana Foroud

Previous genome-wide association studies (GWAS) have identified common variants in genes associated with variation in bone mineral density (BMD), although most have been carried out in combined samples of older women and men. Meta-analyses of these results have identified numerous single-nucleotide polymorphisms (SNPs) of modest effect at genome-wide significance levels in genes involved in both bone formation and resorption, as well as other pathways. We performed a meta-analysis restricted to premenopausal white women from four cohorts (n = 4061 women, aged 20 to 45 years) to identify genes influencing peak bone mass at the lumbar spine and femoral neck. After imputation, age- and weight-adjusted bone-mineral density (BMD) values were tested for association with each SNP. Association of an SNP in the WNT16 gene (rs3801387; p = 1.7 × 10(-9) ) and multiple SNPs in the ESR1/C6orf97 region (rs4870044; p = 1.3 × 10(-8) ) achieved genome-wide significance levels for lumbar spine BMD. These SNPs, along with others demonstrating suggestive evidence of association, were then tested for association in seven replication cohorts that included premenopausal women of European, Hispanic-American, and African-American descent (combined n = 5597 for femoral neck; n = 4744 for lumbar spine). When the data from the discovery and replication cohorts were analyzed jointly, the evidence was more significant (WNT16 joint p = 1.3 × 10(-11) ; ESR1/C6orf97 joint p = 1.4 × 10(-10) ). Multiple independent association signals were observed with spine BMD at the ESR1 region after conditioning on the primary signal. Analyses of femoral neck BMD also supported association with SNPs in WNT16 and ESR1/C6orf97 (p < 1 × 10(-5) ). Our results confirm that several of the genes contributing to BMD variation across a broad age range in both sexes have effects of similar magnitude on BMD of the spine in premenopausal women. These data support the hypothesis that variants in these genes of known skeletal function also affect BMD during the premenopausal period.