Jonathan A. Mitchell, Alessandra Chesi, Diana L. Cousminer, Shana E. McCormack, Heidi J. Kalkwarf, Joan M. Lappe, Vicente Gilsanz, Sharon E. Oberfield, John A. Shepherd, Andrea Kelly, Babette S. Zemel, Struan F.A. Grant

Osteoporosis is a complex disease with developmental origins. It is therefore important to understand the genetic contribution to pediatric areal bone mineral density (aBMD). Individual skeletal site phenotyping has been primarily used to identify pediatric aBMD loci. However, this approach is limited because there is a degree of aBMD discordance across skeletal sites. We, therefore, applied a novel multi-dimensional phenotyping approach to further understand the genetic regulation of pediatric aBMD. Our sample comprised a prospective, longitudinal cohort of 1,293 children of European ancestry (52% female; up to 7 annual measurements). Principal components analysis was applied to dual energy X-ray absorptiometry derived aBMD Z-scores for total hip, femoral neck, spine and distal radius to generate multi-dimensional aBMD phenotypes (i.e. principal component scores). We tested the association between a genetic score (percentage of bone lowering alleles at 63 loci) and each principal component. We also performed a genome wide association (GWAS), using the multi-ethnic baseline data (N = 1,885) to identify novel loci associated with these principal components. The first component (PC1) reflected a concordant phenotypic model of the skeleton (e.g., higher loading score = higher BMD across all sites). In contrast, PC2 was discordant for distal radius versus spine and hip aBMD, and PC3 was discordant for spine versus distal radius and hip aBMD. The genetic score was associated with PC1 (beta = -0.05, P = 3.9 × 10-10 ), but was not associated with discordant PC2 or PC3. Our GWAS discovered variation near CPED1 that associated with PC2 (rs67991850, P = 2.5 × 10-11 ) and near RAB11FIP5 (rs58649746, P = 4.8 × 10-9 ) that associated with PC3. In conclusion, an established bone fragility genetic summary score was associated with a concordant skeletal phenotype, but not discordant skeletal phenotypes. Novel associations were observed for the discordant multi-dimensional skeletal phenotypes that provide new biological insights into the developing skeleton. This article is protected by copyright. All rights reserved.