Multifactor dimensionality reduction reveals antagonistic epistasis between ESR1 and ESR2 with BMD variation in Southern Chinese Women

Abstract

Osteoporosis is a complex disease, which is affected by both genetic and environmental factors. Epistasis is thought to be a critical genetic architecture underlying complex disease, which allows genetic buffering, i.e. to stabilize the trait through the interaction of polymorphisms in different loci in the gene network. A recent study demonstrated that multifactor dimensionality reduction (MDR) is one of the best ways to study epistasis. In this study, we applied MDR in studying the epistasis between two closely related candidate genes for BMD variation, ESRl and ESR2. To study the relation of ESRl and ESRZ polymorphisms in hip BMD variation in southern Chinese women, 827 subjects with either low BMD (defined by a BMD Z score less than - 1.28 at either the femoral neck or total hip) or high BMD (score greater than +I) were studied. A novel CA repeat polymorphism (D6S440) in ESRl and 8 SNPs in ESRZ were genotyped. We previously reported CA repeat size of I8 was most discriminative for hip BMD in a population cohort. Multifactor dimensionality reduction was applied to test the epistatic effect, with subsequent confirmed by binary logistic regression (at least one allele >I8 CA repeats vs 48 repeats) with adjustment of age, height and weight. Carriers of 18 CA repeats of ESRl were associated with higher BMD at the femoral neck (OR: 2.4, p=O.OOl) or hip (OR: 3.2, p <0.001). Variant allele of T-1213C of ESR2 was associated with lower BMD at the femoral neck (OR: I .7. ~0.042) and variant allele of T- 1213C and ~3742641 was associated with low BMD at the total hip (OR: 2.2, p=O.OI I and OR: 2.2, e.013 respectively). MDR revealed significant epistasis between 18 CA repeats of ESRl with several polymorphisms of ESR2 for 2-way, 3-way and 4-way interactions (testing accuracy > 50%). These results were confirmed by binary logistic regression model. Interestingly, when 18 CA repeat interacted with rs944052, rs3742641 and rs928554 of ESR2, the OR resulted from gene-gene interaction was 1.0 (p0.038) at femoral neck and I .I (p=0.004) at hip. The epistasis between ESRl and ESRZ significantly altered the single gene effect of either ESRl or ESRZ. We have demonstrated the epistasis between genes in the same gene network significantly altered the single gene effect. Testing for gene epistasis in association with BMD variation is a powerful approach to understand the pathophysiology of low bone mass in the general population. [https://asbmr.onlinelibrary.wiley.com/doi/full/10.1002/jbmr.5650221407]