The Promoter Region of the Adiponectin Gene Is a Determinant in Modulating Insulin Sensitivity in Childhood Obesity


Endocrinology, Department of Clinical Sciences, University of Rome La Sapienza, Viale del Policlinico 155, 00161 Rome, Italy. E-mail:


We investigated the association of the −11,391G>A, −11,377G>C, +45T>G, and +276G>T adiponectin single-nucleotide polymorphisms (SNPs) and expected haplotypes with the insulin resistance (IR) state in overweight/obese children; by using the haplotype background analysis, we also assessed the effect of each SNP independently. GG genotype at the −11,391 locus was associated with higher fasting insulin levels and homeostasis model assessment-IR index and lower adiponectin levels compared with GA + AA genotypes (p = 0.01, 0.002, and 0.03, respectively). Those heterozygous and homozygous for G allele at the −11,377 locus showed higher fasting glucose (p = 0.001 for both), fasting insulin (p = 0.001 for both), homeostasis model assessment-IR index (p < 0.001 for both), and triglyceride levels (p = 0.02 and 0.03, respectively) and lower adiponectin levels (p = 0.002 and 0.02, respectively) compared with C homozygotes. The +45G carriers showed higher fasting and 2-hour glucose levels (p = 0.01 for both) and lower adiponectin levels (p = 0.02) compared with non-carriers. Haplotype analysis suggested that, considering the same haplotypic background, each of the three polymorphisms exerted an independent effect on investigated parameters. The −11,391G>A, −11,377C>G, and +45T>G SNPs are associated with IR syndrome in overweight/obese children; they independently influence the investigated variables. The effect of +45T>G SNP seems to be marginal compared with the promoter SNPs. The GGT haplotype is associated with the highest degree of IR.

Multiple mechanisms are thought to contribute to the pathogenesis of insulin resistance (IR).1 Among these, the role of adipose tissue, obesity, and genetic factors seem to be of great significance. Genes involved in adipose tissue metabolism can be considered possibly responsible for insulin sensitivity. Hormones and cytokines produced by adipose tissue have wide-ranging effects on food intake, energy expenditure, and lipid metabolism. Among these cytokines, adiponectin plays a role in modulating insulin sensitivity. Concentrations of plasma adiponectin have been shown to negatively relate with BMI (1) insulin (1,2,3), and triglyceride (TG) levels (1,4) and positively with high-density lipoprotein (HDL)-cholesterol in obese subjects (1).

Different single-nucleotide polymorphisms (SNPs) were localized in the adiponectin gene [adipocyte, C1q, and collagen domain containing (ACDC)]; these genetic variations were associated with risk of obesity, IR, type 2 diabetes (T2D), and metabolic factors (5,6,7,8,9); however, several studies performed so far on adults show conflicting results regarding the association between these polymorphisms and complex traits and on the identity of the allele involved (10). In view of these considerations, the aim of our study was to assess the association of the −11,391G>A, −11,377C>G, +45T>G, and +276G>T adiponectin SNPs and expected haplotypes with the metabolic abnormalities of IR in overweight/obese children and to distinguish the effect of each SNP independently on the investigated phenotypes. We investigated the association of these genetic variants with measures of insulin sensitivity. The influence of environmental factors in children may be less than the effect on adults. Hence, the search for genetic determinants for obesity-related traits in such a population could increase the chance of finding the responsible loci.

In children, −11,391 G/G showed higher fasting insulin levels and homeostasis model assessment (HOMA)-IR index compared with GA + AA carriers (p = 0.01 and 0.002, respectively). Children heterozygous and homozygous for the −11,377G allele showed higher fasting glucose (p = 0.001 and 0.001), fasting insulin levels (p = 0.001 and 0.001), HOMA-IR index (p = 0.0005 and 0.0004), and TG levels (p = 0.02 and 0.03) compared with C homozygous. Analysis of −11,377 SNP association with phenotypic traits using the three genotypes seems to display a codominant model. The IR parameters showed increasing trend from C/C homozygous to C/G heterozygous and to G/G homozygous even if no statistically significant difference was observed between C/G and G/G. The +45G carriers showed higher fasting glucose (p = 0.01) and 2-hour glucose levels (p = 0.01) compared with non-carriers (Table 1).

Table 1.  Clinical and biochemical parameters of children according to adiponectin SNPs
 11,391 G>A (A allele 0.10)11,377 C>G (G allele 0.27)
 G/G(1) (mean ± SD)G/A(2) (mean ± SD)A/A(3) (mean ± SD)p, ANOVA1 vs. 2 + 3C/C(1) (mean ± SD)C/G(2) (mean ± SD)G/G(3) (mean ± SD)p, ANOVA1 vs. 22 vs. 31 vs. 3
n (Boys/girls)219 (101/108)49 (23/26)2 (0/2)  148 (67/81)97 (45/52)25 (12/13)    
Age (years)10.49 ± 1.9810.48 ± 2.0210.38 ± 3.11NSNS10.44 ± 2.0810.62 ± 1.9910.62 ± 2.14NSNSNSNS
Systolic BP (mm Hg)110.77 ± 9.1117.11 ± 10.04103.75 ± 13.77NSNS114.84 ± 9.9111.3 ± 10.00109.29 ± 9.07NSNSNSNS
Diastolic BP (mm Hg)74.25 ± 7.0876.37 ± 7.8076.25 ± 9.31NSNS74.13 ± 7.0275.71 ± 7.3276.61 ± 7.42NSNSNSNS
SDS-BMI2.64 ± 0.653.23 ± 0.671.87 ± 0.53NSNS2.81 ± 0.572.94 ± 0.571.94 ± 0.37NSNSNSNS
Fasting glucose (mM)4.77 ± 0.544.73 ± 0.644.34 ± 0.67NSNS4.63 ± 0.584.88 ± 0.535.07 ± 0.63<0.00010.001NS0.001
2-hour OGTT glucose (mM)6.73 ± 0.986.68 ± 1.045.15 ± 0.67NSNS6.71 ± 0.996.75 ± 0.996.45 ± 0.66NSNSNSNS
Fasting insulin (mU/mL)26.3 ± 12.521.15 ± 12.019.77 ± ± 10.7926.9 ± 10.229.7 ± 10.20.00010.001NS0.001
2-hour insulin (mU/mL)93.3 ± 46.893.3 ± 44.193.7 ± 44.2NSNS93.7 ± 42.892.05 ± 42.292.55 ± 42.2NSNSNSNS
HOMA-IR (arbitrary units)5.56 ± 2.204.49 ± 2.103.71 ± 2.070.0050.0024.74 ± 2.515.91 ± 2.66.73 ± 2.6<0.00010.0005NS0.0004
CT (mM)4.27 ± 0.884.31 ± 0.684.46 ± 0.84NSNS4.2 ± 0.724.4 ± 0.844.5 ± 0.86NSNSNSNS
HDL-cholesterol (mM)1.29 ± 0.331.27 ± 0.290.99 ± 0.22NSNS1.29 ± 0.31.31 ± 0.261.36 ± 0.29NSNSNSNS
LDL-cholesterol (mM)2.61 ± 0.692.55 ± 0.602.8 ± 0.46NSNS2.55 ± 0.572.62 ± 0.632.88 ± 0.63NSNSNSNS
TGs (mM)1.11 ± 0.460.99 ± 0.571.18 ± 0.5NSNS1.08 ± 0.421.22 ± 0.511.28 ± 0.550.020.02NS0.03
Serum adiponectin levels (ng/mL)*17.85 ± 7.3220.12 ± 8.110.0320.55 ± 7.717.85 ± 7.6616.88 ± 7.80.0040.002NS0.03

A decrease in adiponectin levels was observed for individuals with −11,391 G/G compared with individuals with G/A (p = 0.03) and for −11,377G and +45G carriers compared with non-carriers (p = 0.02 for both). The relation between SNPs and IR phenotypes does not persist after adjusting for adiponectin levels (p > 0.05). No statistically significant difference was observed between +276 SNP and the investigated variables. We observed different adiponectin levels between boys and girls (16 vs. 20.5 ng/μL; p = 0.01). p values for interactions between genotypes and sex are not statistically significant.

The linkage disequilibrium (LD) calculated among the four polymorphisms (data not shown) is similar to that observed in another white population (15). All different haplotypes were estimated; only the ones with a frequency higher than 5% are shown in the tables. Haplotype background analysis showed that each of the three SNPs exerted an independent effect on investigated parameters. This analysis enabled the investigation of the effect of each polymorphism considering the same haplotypic background (between two corresponding haplotypes that only differ at the position of the investigated polymorphism). The −11,391G variant (GCT vs. ACT haplotype) was associated with higher fasting insulin levels, HOMA-IR index, and lower adiponectin levels, whereas the −11,377G variant was also associated with glucose levels (GCT vs. GGT haplotype) (Table 2). We found an increase in glucose levels for the +45G allele (GCT vs. GCG haplotype). Moreover, we observed that the GGT haplotype was associated with higher fasting insulin levels, HOMA-IR index, and lower adiponectin levels for all comparisons made and with higher fasting glucose levels compared with the other haplotypes, except GCG (Table 2). Moreover, Table 3 shows that the GG haplotype is strongly associated with lower adiponectin levels and higher degree of IR.

Table 2.  Expected phenotypic mean according to the four most frequent estimated haplotypes for SNPs −11,391, −11,377, and +45 of the adiponectin gene in overweight/obese children
Haplotype frequenciesGCT [0.52 (CI)]GGT* [0.27 (CI)]GCG [0.10 (CI)]ACT [0.09 (CI)]
  • SNP, single nucleotide polymorphism; OGTT, oral glucose tolerance test; HOMA-IR, homeostasis model assessment of insulin resistance; CT, total cholesterol; HDL, high-density lipoprotein; TG, triglyceride; SDS, standard deviation score.

  • p values are adjusted for SDS-BMI and Tanner stage.

  • *

    GGT vs. all comparisons, p ≤ 0.01 for fasting insulin, HOMA-IR index, and adiponectin levels; for fasting glucose, GGT vs. all comparisons except vs. GCG, p < 0.05.

  • GGT vs. GCT, p < 0.0001; GCT vs. GCG, p = 0.007.

  • GGT vs. GCT, p = 0.007; GCT vs. ACT, p = 0.02.

  • §

    GGT vs. GCT, p < 0.0001; GCT vs. ACT, p = 0.01.

  • Values are log10 transformed; performed in 150 children.

Fasting glucose (mM)2.22 (2.14 to 2.30)2.53 (2.42 to 2.64)2.49 (2.3 to 2.68)2.12 (1.91 to 2.33)
2-hour OGTT glucose (mM)3.28 (3.13 to 3.42)3.41 (3.22 to 3.6)3.15 (2.87 to 3.43)3.2 (2.89 to 3.51)
Fasting insulin (mU/mL)12.31 (10.8 to 13.7)16.11 (13.5 to 18.7)11.80 (8.4 to 15.2)8.12 (5.7 to 10.5)
2-hour insulin (mU/mL)44.28 (38.7 to 49.9)44 (36.5 to 51.4)43 (30.7 to 55.3)42.7 (30.1 to 55.6)
HOMA-IR§2.51 (2.25 to 2.76)3.45 (3.05 to 3.85)2.53 (1.98 to 3.08)1.71 (1.28 to 2.14)
CT (mM)2.07 (1.94 to 2.2)2.17 (1.98 to 2.36)2.11 (1.79 to 2.43)2.11 (1.76 to 2.46)
HDL-cholesterol (mM)0.55 (0.51 to 0.59)0.61 (0.55 to 0.67)0.51 (0.43 to 0.59)0.52 (0.43 to 0.61)
TGs (mM)0.56 (0.49 to 0.64)0.58 (0.48 to 0.68)0.52 (0.36 to 0.68)0.57 (0.38 to 0.76)
Serum adiponectin levels (ng/mL)1.03 (0.98 to 1.08)0.88 (0.82 to 0.95)1.02 (0.9 to 1.14)1.07 (0.93 to 1.2)
Table 3.  Expected phenotypic mean of IR parameters according to estimated haplotypes of promoter SNPs in overweight/obese children
Haplotype frequenciesGC [0.58 (CI)]GG [0.31 (CI)]AC [0.10 (CI)]
  • IR, insulin resistance; SNP, single nucleotide polymorphism; CI, confidence interval; HOMA-IR, homeostasis model assessment of IR; SDS, standard deviation score.

  • GG vs. all comparisons, p < 0.01; GC vs. AC, p < 0.01; p values are adjusted for SDS-BMI and Tanner stage.

  • *

    Values are log10 transformed; performed in 150 children.

Fasting glucose (mM)2.17 (2.09 to 2.25)2.48 (2.36 to 2.60)1.4 (1.23 to 1.57)
Fasting insulin (mU/mL)11.9 (10.6 to 13.2)16.10 (13.5 to 18.7)9.7 (6.63 to 12.77)
HOMA-IR2.2 (1.93 to 2.47)3.27 (2.7 to 3.84)1.78 (1.19 to 2.37)
Serum adiponectin levels (ng/mL)*1.02 (0.97 to 1.07)0.86 (0.80 to 0.91)1.16 (1.02 to 1.28)

To our knowledge, this is the first study concerning the relationship between genetic variants of ACDC and clinical and biochemical parameters performed on childhood obesity. We observed that the promoter region variants were significantly associated with lower insulin sensitivity, that the +45T>G variant was associated with higher fasting glucose and 2-hour glucose levels, and that the three polymorphisms influenced serum adiponectin levels.

The LD observed between SNPs and different functional effects of the haplotypes are potential sources of bias when studying multiple SNPs of the same gene. Therefore, we performed the haplotype background analysis underlining that, although −11,391G>A, −11,377C>G, and +45T>G SNPs were in LD, they independently influenced the IR phenotypes. The effect of +45T>G seemed to be, at best, marginal compared with the promoter SNPs.

A major finding of the present study is the association of the promoter SNP haplotype with lower adiponectin levels and insulin sensitivity in overweight/obese Italian children. Consistent with our results, Vasseur et al. (15,16) found that the −11,391G/−11,377G haplotype is associated with low plasma adiponectin levels and T2D, even though they failed to detect an association with IR index.

This haplotype or an unknown nearby functional variant in LD could decrease adiponectin levels and, consequently, insulin sensitivity. Results obtained on IR phenotypes after adjusting for adiponectin levels supported this conclusion. Thus, the influence of the ACDC SNPs on IR may be fully explained by their relation with adiponectin levels. These data corroborate those obtained in adult obese subjects from different ethnic groups (5,6,15).

Two Japanese studies on the association of the promoter region with T2D obtained opposite results (6,17). Two prospective studies detected associations between +45 SNP and the development of hyperglycemia and T2D (18,19). Accordingly, we observed that the same SNP plays a role in modulating glucose levels.

Conflicting findings between previous studies can be due to several factors such as insufficient power, the study population heterogeneity (lean, obese) and type of study, or LD between SNPs. It also can be partly attributed to differences in allelic association with disease phenotype in various populations. The differences in SNPs allele frequencies of the adiponectin gene in various populations are consistent with these notions (5,6,7,15,20).

The limitations of our study include the possibility of false positive associations, small sample size, given the allele frequency of the four alleles, and finally the lack of normal-weight population. In the present study, we have demonstrated that the polymorphisms of adiponectin gene were associated with IR phenotypes in overweight/obese children. These results concord with those previously reported in adults (5,6,7,8,9). In conclusion, our observations suggest an important and independent effect of the SNPs in the promoter region and, to a lesser extent, of +45G variant in modulating insulin sensitivity in childhood obesity. The GGT haplotype is associated with the highest degree of IR. The genetic variants at the ACDC promoter region, which show a close cis relationship, were associated with adiponectin levels.

Research Methods and Procedures

Two hundred-seventy overweight/obese children (124 boys and 146 girls) were consecutively enrolled by the Center for Nutrition and Dietetics at the Department of Pediatrics, La Sapienza University of Rome. Exclusion criteria were previous diagnosis of type 2 diabetes and/or endocrine diseases.

Parents gave their written informed consent to participate in the study after being informed of its nature. The study protocol was approved by the Ethical Committee of the La Sapienza University of Rome. Tanner staging was performed by a pediatrician in recruited children considering the criteria previously established by Tanner et al. (11).

The degree of obesity was quantified using Cole's least mean square method, which normalizes the skewed distribution BMI and expresses BMI as a standard deviation score (SDS) (12). Clinical [height, weight, BMI, and blood pressure (BP)] and fasting biochemical features (fasting glucose, fasting insulin levels, and lipid profile) were evaluated at entry. Furthermore, flavored glucose in a dose of 1.75 g/kg body weight (up to a maximum of 75 grams) was given orally, and blood samples were obtained for measurements of plasma glucose and insulin at 0 and 120 minutes.

Serum total cholesterol (CT) and TG levels were determined with the Technicon RA-1000 Autoanalyzer; low-density lipoprotein (LDL) was measured using the Friedewald formula: LDL = CT − HDL − (TG/5). Glucose levels were determined with the glucose oxidase method (Autoanalyzer; Beckman Coulter, Fullerton, CA). Serum insulin was measured by radioimmunoassay (insulin kit; Adaltis, Bologna, Italy). IR was estimated according to the HOMA-IR (13). Serum adiponectin levels were performed in 150 children and determined by radioimmunoassay (LINCO Research Inc., St. Louis, MO).

Genotyping of Polymorphisms

The −11,391G>A, −11,377C>G, +45T>G, and +276G>T SNPs were genotyped using the fluorogenic 5′ nuclease assay application of the ABI PRISM 7900 HT Sequence Detection System (ABI, Foster City, CA). The conditions for Taqman reaction were as follows: 95 °C for 10 minutes and 35 cycles of 95 °C for 15 seconds and 60 °C for 1 minute.

Statistical Analysis

The genotypic frequencies of the SNPs investigated were in agreement with Hardy-Weinberg equilibrium (p = 0.6 for −11,391G>A, p = 0.41 for −11,377C>G, p = 0.98 for +45T>G, and p = 0.49 +276G>T SNPs).

Statistical analysis was performed using the SPSS/PC statistical program (version 13 for Windows; SPSS, Inc., Chicago, IL). We evaluated and presented data according to the three different genotypes of each SNP without assuming any specific genetic model. ANOVA analysis was performed to obtain initial overall p values. For −11,391 and +45 SNPs, statistical analysis was performed taking into account A and G carriers, respectively, compared with non-carriers considering the small sample size of AA and GG homozygotes. The effect of the polymorphisms on quantitative variables was investigated using multiple linear regression analysis. The values were adjusted for BMI-SDS and pubertal stage; each polymorphism was introduced as a dichotomous variable in the analysis. We applied Bonferroni correction of a factor two for +11,391 and +45 SNPs and of a factor four for −11,377 SNP. In addition, the effect of the interaction between genotypes and sex on IR parameters and adiponectin levels was included in the model.

Data for insulin, TGs, HOMA-IR, and adiponectin levels were log10 transformed to normalize their distribution. The frequencies haplotypes and the LD matrix based on the D′ parameter were estimated using Testing Haplotype Effects in Association Studies (THESIAS) software. The software is designed to perform haplotype-based association analysis in unrelated individuals. THESIAS is based on the maximum likelihood model described by Trégouet et al. (14) and is linked to the standard error of the mean algorithm. THESIAS also allowed the simultaneous estimation of haplotype frequencies and their associated effects on the phenotype of interest. The estimate haplotype effects to the levels of biochemical variables are approximately one-half of the overall phenotypic mean. We set the haplotype GCT like the intercept of the model. We adjusted the crude effect of the haplotype, taking into account BMI-SDS and pubertal stage.

To distinguish the effect of each SNP independently, we performed the analysis named haplotype background (using THESIAS software), which enabled us to investigate the effect of each polymorphism considering the same haplotype background (between two corresponding haplotypes that only differ at the position of the investigated polymorphism). Considering the lack of association of the SNP +276 with the phenotypes tested and to obtain a greater statistical power to test the influence of the haplotypes on all parameters, we performed the haplotype analysis using only the following three loci haplotypes: −11,391, −11,377, and +45.

Bootstrap analysis of the minimum p values across all tests performed revealed that five of 1000 replicates (0.005) were less than the best nominal p value observed in our studies. Therefore, we can reject the global null hypothesis: no SNP or haplotype is associated with any traits investigated.


This work was supported, in part, by the Ministry of Health (Grant ICS 030.6/RF00-49) and by the Ministry of University and Research (Grant MIUR 2003 2003061834_004).


  • 1

    Nonstandard abbreviations: IR, insulin resistance; TG, triglyceride; HDL, high-density lipoprotein; SNP, single nucleotide polymorphism; ACDC, adipocyte, C1q, and collagen domain containing; T2D, type 2 diabetes; HOMA, homeostasis model assessment; LD, linkage disequilibrium; SDS, standard deviation score; BP, blood pressure; CT, total cholesterol; LDL, low-density lipoprotein; THESIAS, Testing Haplotype Effects in Association Studies.

  • The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.