Obesity is a prevalent condition and a modifiable risk factor for various complex diseases, including diabetes and cardiovascular disease. Although the current World Health Organization (WHO) definitions for being overweight and obesity are body mass index (BMI) of ≥25-29.9 kg/m² and ≥30 kg/m², respectively, different BMI cut-off points of 23-24.9 kg/m² and ≥25 kg/m², respectively, were suggested for the Asian population [1,2]. In Korea, the prevalence of obesity (BMI ≥25 kg/m²) has dramatically increased over the past decade and the third Korea National Health & Nutrition Examination Survey reported that the overall prevalence of obesity in Korean adults was 30.6% (29.4% in women and 32.4% in men) in 2001 [3].

Adiponectin decreases body weight by increasing lipid oxidation in muscles and other organs such as the pancreas and liver [4]. The concentration of adiponectin in plasma was shown to be negatively correlated with BMI [5]. In humans, adiponectin is encoded by portions of exons 2 and 3 among the three exons of the adiponectin, C1Q and collagen domain-containing (ADIPOQ) gene (also denoted as APM1 or ACDC) located on chromosome 3q27 [6]. The 13 single nucleotide polymorphisms (SNPs) in the ADIPOQ gene were previously reported in Japanese and French populations [7]. Two SNPs located on exon 2 (+45T>G, a synonymous mutation, Gly15Gly) and intron 2 (+276G>T) of the ADIPOQ gene were reported to be associated with both plasma adiponectin concentrations and type 2 diabetes (T2D) in Japanese, German, and Italian populations [7-9]. The G allele of the SNP -11377C>G located in the promoter region was also shown to be associated with lower adiponectin levels and severe obesity among Danish women [10]. Neither the mechanism responsible for controlling the synthesis of adiponectin nor the regulation of ADIPOQ gene expression have been fully determined as yet. The effect of the ADIPOQ gene on the risk of obesity may vary according to ethnicity, age, and he degree of obesity across populations [11].

Thus, this study aimed to perform an analysis of the association between the ADIPOQ gene and BMI, an index of obesity, whilst considering plasma adiponectin levels in a Korean population. We defined three groups of two extremes for comparison according to a combination of BMI levels and plasma adiponectin concentrations after adjusting for age and sex in 986 healthy Koreans.

As shown in Table 1, 73.1% of the study population was composed of men (with a mean age of 49.5 yr old), and 26.9% of the population was female (mean: 51.2 yr old). The average BMI values were 24.8 and 23.6 kg/m² for men and women, respectively. The plasma adiponectin level was approximately 30% higher in women (9.1 µg/mL) than in men (6.2 µg/mL). No gender difference was observed in the homeostasis model assessment of insulin resistance (HOMA-IR) levels. As shown Figure 1, the numbers of cases and controls increased from G1 (55 cases, 71 controls) to G3 (259 cases, 272 controls).

The major alleles for each marker of -11377C>G, +45T>G, and +276G>T in the controls of the G3 group were C, T, and G, respectively. All three markers showed sufficient heterozygosity, and no evidence of deviation from HWE was observed in any marker (Table 2). Three SNPs were not found to be in strong linkage disequilibrium based on the values of r². Although individuals with the GG genotype of -11377C>G had a higher BMI and lower adiponectin concentration (25.1 kg/m² and 6.7 µg/mL) compared to individuals with the CC reference genotype (24.4 kg/m² and 7.1 µg/mL), the difference was not statistically significant (p=0.09) (Table 3).

In the present study, three SNPs located on the ADIPOQ gene, which have previously been studied for an association with obesity in other ethnic groups [16], were evaluated in healthy Koreans. The C, T, and G alleles were the major alleles at SNPs -11377C>G, +45T>G, and +276G>T, respectively, as shown in other studies, including non-diabetic Japanese and non-diabetic Koreans [7,17]. A common allele, C, of -11377C>G was strongly associated with a lower plasma adiponectin level, a higher BMI, or T2D in some studies [18,19], whereas a rare allele, G, was associated with the same phenotypes in other studies [10,20]. Similar trends toward inconsistent results across studies were observed for the +45T>G and +276G>T loci. The rare G allele at the +45T>G locus significantly increased the risk of obesity in non-diabetic Germans and in Japanese with T2D [7,8], while Loos et al. [21] recently reported that carriers of the rare GG homozygote showed a leaner phenotype than carriers of the common T allele. The G allele at the +276G>T locus was associated with lower plasma adiponectin concentrations and higher insulin resistance only in subjects with a high BMI (≥26.7) in non-diabetic Korean and Japanese populations [7,17]. Homozygotes for the haplotype T45-G276 had a higher body weight and waist circumference and lower plasma adiponectin concentrations in non-diabetic Italians and non-diabetic obese Koreans [17,22,23].

In our study, the G allele of SNP -11377C>G (OR, 1.48; 95% CI, 1.13-1.94) and the T allele at the +45T>G locus were identified as being risk alleles for obesity, even after issues of study power and multiple testing were taken into account, where obesity was defined using both a higher BMI and a lower adiponectin concentration compared to the reference group. The T allele of the +276G>T locus was not statistically significant, but it increased the risk of obesity. Possible explanations for these inconsistent results across the studies include not only differences in population characteristics (e.g. demographic, genetic, and environmental factors, diagnostic criteria, etc.) but also the possibility that the configurations of genotypes or hapolotypes were mixed up in some studies. Minor allele frequencies of each SNP were different even from Japanese HapMap data (e.g. 28.9% of Japanese HapMap data vs. 24% of data in the current study for the SNP -11377C>G). Therefore, more prospective studies of a better design are necessary to confirm the effects of each variant on the development of obesity [17,24]. As shown in other Korean studies, linkage disequilibrium was not observed between SNPs -11377C>G, +45T>G or +276G>T, whereas two SNPs, +45T>G and +276G>T were in imperfect (r²=0.14) but complete linkage disequilibrium (D'=1) in our study [18,23].

Overall, a stricter definition of the phenotype increased the statistical power to detect evidence of an association in our study. Although significantly high ORs were observed in the CG haplotype of -11377G>C and +45T>G (OR, 4.02) and the GGT haplotype (OR, 11.52), these high ORs may have been the result of the relatively small sample size (Table 4). Our results replicated the reverse relationship between BMI and plasma adiponectin (Table 3) and the association between ADIPOQ gene variants and obesity previously reported, although this study did not describe a causal relationship, but associations (Tables 4, 5) [5,10].

Several potential limitations of our study should be mentioned. Firstly, BMI may not be a sufficient indicator for measuring the degree of obesity. Although regional differences in body fat affect the risk of metabolic abnormalities, and although ADIPOQ gene variants were suggested as modulating visceral fat accumulation in some studies, other studies detected no association between ADIPOQ gene variants and BMI [20]. Thus, other measurements of obesity such as the waist:hip ratio or waist circumference [10,25] might be better measures of abdominal obesity in the Korean population. Secondly, there may also be some interactions with other genes related to susceptibility to obesity. For instance, evidence for an interaction between SNPs in the promoters of ADIPOQ (-3971A>G) and ADIPOR1 (-3882T>C) was found [21]. Thirdly, the effects of environmental factors controlling the risk of obesity could not be evaluated in this study. Given the role of adiponectin in fatty acid oxidation, a reduction in plasma adiponectin levels may impact the ability to clear cellular adipose in obese individuals with an at-risk diplotype [20]. However, the effect of this risky polymorphism could be validly evaluated after controlling for environmental factors (e.g. physical activity, fried food intake, alcohol consumption, etc.), clinical characteristics (e.g. serum C-reactive protein, total cholesterol HDL ratio, blood pressure, etc.), and ethnicity [22,26]. Finally, the positive associations found in this study were based on the comparison made between a group of high BMI and low adiponectin subjects versus a group of low BMI and high adiponectin subjects. Thus, the statistical evidence might not reflect the effect of ADIPOQ gene variants on BMI and plasma adiponectin concentrations separately, but rather a joint consideration of both indices for obesity. Therefore, additional studies warrant validation of the relationship between the ADIPOQ gene and obesity identified in this study.

In conclusion, the SNPs identified here (i.e. -11377C>G and +45T>G) may assist in identifying subjects who are at a greater risk of obesity among healthy Koreans. Such information may provide a chance for susceptible individuals to prevent obesity and for obese individuals to determine a more efficient means of weight control. Furthermore, understanding the mechanisms by which obesity could be controlled would make it possible for us to better understand the metabolic complications of obesity.