Association of both peroxisome proliferator-activated receptor, gene-gene interactions and the body mass index

Abstract

To investigate the association of ten single nucleotide polymorphisms (SNPs) in the peroxisome proliferator-activated receptor (α, δ, γ) with obesity and the additional role of a gene-gene interaction among 10 SNPs. Participants were recruited within the framework of the PMMJS (Prevention of Multiple Metabolic Disorders and Metabolic Syndrome in Jiangsu Province)-cohort-population-survey in the urban community of Jiangsu province, China. 820 subjects (513 non obese subjects, 307 obese subjects) were randomly selected and no individuals were related to each other. Ten SNPs (rs135539, rs4253778, rs1800206, rs2016520, rs9794, rs10865710, rs1805192, rs709158, rs3856806, rs4684847) were selected from the HapMap database, which covered PPARα, PPARδ and PPARγ. Logistic regression model was used to examine the association between ten SNPs in the PPARs and obesity. Odds ratios (OR) and 95% confident interval (95%CI) were calculated. Interactions were explored by using the Generalized Multifactor Dimensionality Reduction (GMDR). A group of 820 participants (mean age was 50.05± 9.41) was involved. The frequency of the mutant alleles of rs2016520 in obese populations was less than that in non-obese populations (26% vs. 33%, P < 0.01). The frequency of the mutant alleles of rs10865710 in obese populations was more than that in non-obese populations (37% vs. 31%, P = 0.01). C allele carriers had a significantly lower obesity occurrence than TT homozygotes [OR (95%CI) = 0.63 (0.47 - 0.84)] for rs2016520 but no significant association was observed between other SNP and incident obesity. GMDR analysis showed a significant gene-gene interaction among rs2016520, rs9794 and rs10865170 for the three-dimension models (P = 0.0010), in which prediction accuracy was 0.5834 and cross-validation consistency was 9/10. It also showed a significant gene-gene interactions between rs2016520 and rs10865170 in all the two-dimensional models (P = 0.0010), in which predictive accuracy was 0.5746 and cross-validation consistency was 9/10. Our data showed that rs2016520 was associated with lower obesity risk, as well as interactions among rs2016520, rs9794 and rs10865170 on incident obesity.