Insulin gene variable number tandem repeat (INS VNTR) is a common genetic variation located in a highly polymorphic region 5′ of the human insulin gene (1). Because of its position, the relation between INS VNTR and the genetic risk for glucose metabolism has been intensively studied for a number of metabolic disorders. Initially, the class I allele was reported to be associated with higher risk of type 1 diabetes in UK families (1) and replication studies have been consistently and reliably reported (2). On the other hand, the class III allele of INS VNTR was suggested to associate with higher birth weight (3) and increased risk of type 2 diabetes (4). However, several studies failed to replicate association of class III allele with fetal growth and birth weight (5,6) or reported associations with lower birth weight (7). Replication studies of INS VNTR association with increased type 2 diabetes were also mainly unsuccessful (8,9), although a modest association was reported in a family-based study (10). The putative contribution of the INS VNTR in the genetic risk for obesity was first investigated in children. Le Stunff et al. reported strong evidence for family-based association of INS VNTR class I allele with a 1.8-fold increased risk of early-onset obesity when this allele is paternally inherited (11). An earlier study from the same group has suggested higher insulin levels in obese children carrying the class I allele (12). However, several studies of INS VNTR with body composition and insulin levels in cohorts of children were inconclusive (13,14,15). A large negative study has been reported in UK population of adults as well (16). Given the complexity of the INS VNTR heredity and the evidence for transmission distortion (TD) at this locus (17), the relevance of studies employing unrelated individuals has been questioned (18). We believe that the investigation in family-based populations is therefore more appropriate to assess the association of INS VNTR with obesity and insulin resistance background in a definite manner. In this study, we report transmission analyses of the INS VNTR in a large number of families including French and German obese, low birth weight (LBW), and unselected population-based families.
Previous studies have described genetic associations of the insulin gene variable number tandem repeat (INS VNTR) variant with childhood obesity and associated phenotypes. We aimed to assess the contribution of INS VNTR genotypes to childhood obesity and variance of insulin resistance, insulin secretion, and birth weight using family-based design. Participants were either French or German whites. We used transmission disequilibrium tests (TDTs) for assessing binary traits and quantitative pedigree disequilibrium tests for assessing continuous traits. In contrast to previous findings, we did not observe any familial association with childhood obesity (T = 50%, P = 0.77) in the 1,023 families tested. In French obese children, INS VNTR did not associate with fasting insulin levels (P = 0.23) and class I allele showed only borderline association with increased insulin secretion index at 30 min (P = 0.03). INS VNTR did not associate with birth weight in obese children (P = 0.98) and TDT analyses in 350 French families with history of low birth weight (LBW) showed no association with this condition (P = 0.92). In summary, our study, the largest performed so far, does not support the previously reported associations between INS VNTR and childhood obesity, insulin resistance, or birth weight, and does not suggest any major role for this variant in modulating these traits.
Transmission disequilibrium tests (TDTs) for the INS VNTR through rs689 genotypes are presented in Table 1. In contrast to the initial study (11), the class I allele neither showed significant TD in French obese families (T % = 48%, odds ratio (OR) = 0.90 95% confidence interval = (0.74–1.06), P = 0.28) nor in German obese families (T% = 52%, OR = 1.10 95% confidence interval = (0.93–1.26), P = 0.19). When the 1,023 obese families were analyzed together using a Mantel-Haenszel test, no TD was found either (T% = 50%, OR = 1.01 95% confidence interval = (0.90–1.14), P = 0.72) despite a statistical power estimated to 99.99% to detect the previously reported relative risk of 1.8 (11) and 80% power for an OR of at least 1.27. We did not observe a parent-of-origin effect for the INS VNTR transmission (paternal origin: P = 0.96; maternal origin: P = 0.67). We have then assessed the transmission of INS VNTR in unselected families to check for natural TD that was reported previously (17). We observed no TD for the class I allele in 448 unselected families (T% = 53%, P = 0.12). In these families, transmission of paternal origin of class I allele showed a trend of overtransmission that did not reach significance (T% = 57%, P = 0.055). Although we previously reported no significant association between INS VNTR and LBW using case-control analyses (5), we assessed whether the INS VNTR associated with LBW using family-based design. We genotyped parents of LBW siblings from the initial study where DNAs were available and analyzed TD in 350 LBW families. We observed no TD for the INS VNTR (T% = 50%, P = 0.92) and no parent-of-origin effect (paternal origin: T% = 50%, P = 0.88; maternal origin: T% = 50, P = 0.99) in accordance with our previous negative case-control finding.
Familial association of INS VNTR with quantitative traits was tested in French obese children using the quantitative pedigree disequilibrium test and the results are displayed in Table 2. Class I allele only showed nominal significant association with higher insulinogenic index at 30 min after an oral glucose tolerance test (OGTT) (P = 0.03). This association did not reach significance after correction for multiple testing (P = 0.27).
Our main finding is the lack of replication of association between the INS VNTR class I allele and childhood obesity in two independent family-based and well-powered samples (80% power to detect an OR as low as 1.27 and 97% power to detect an OR ∼1.40, as the one we have reported for the FTO variants in a family-based design (19)). To our knowledge, our study is the first to have assessed the initial association with childhood obesity reported by Le Stunff et al. (11) using the same design in a larger number of families. Several studies attempted to replicate the associations of the INS VNTR with obesity and insulin resistance indexes in populations of children and reported controversial findings (Table 3). An additional study carried out in a sample overlapping with the obese children population studied initially (11,12) reported a lack of association of INS VNTR with the insulinogenic index (20) (Table 3). In an Italian obese children population, class I allele associated with insulin levels during an OGTT and the insulinogenic index but not with BMI (15) (Table 3). In families from the Fleurbaix-Laventie Ville Santé study, we have reported an overtransmission of class I allele in overweight children. In the light of this study, this finding is likely to be a false-positive association observed in a small cohort (n = 431) that only included few informative families with overweight children (n = 14) (13) (Table 3). In this population, INS VNTR did not associate with fasting insulin (13) (Table 3). In a large birth cohort (the Avon Longitudinal Study of Pregnancy and Childhood, n = 947), class I allele showed unexpected associations with lower fat mass and lower BMI and had no significant effect on fasting insulin levels or the insulinogenic index (14). However, significant interaction with weight gain for association with insulin secretion was observed in the class I allele carriers (14). A series of possible explanations for these controversial findings have been advanced, mainly population admixture and low variability of insulin levels and BMI that might have lowered the statistical power in the cohort samples (18). We believe that both arguments can be excluded to explain the lack of replication in our study because (i) we have used TDT analyses known to be robust to population stratification and (ii) we have analyzed insulin levels in obese children as in the initial study (12) and observed no significant association with insulin resistance or insulin-secretion indexes. Obese children from our families have rather a more severe phenotype as most of them have a BMI >99th percentile (90% in the French sample, 70.5% in the German sample) vs. only 22% in the initial study (105 of the 458 studied). Nevertheless, the variability of insulin levels in our obese sample is still reasonably high (mean = 13.8 U/l ± 9.9 U/l, variance = 98.1 U/l, min = 1.0 U/l, max = 80.0 U/l). The lack of quality control of genotyping is an important (and often neglected) source of erroneous results in genetic association studies (21). PCR-restriction fragment length polymorphism genotyping technique was used in the initial study from Le Stunff et al. and in all studies of the INS VNTR in obese children, except the Fleurbaix-Laventie Ville Santé study (Table 3). In a study where methodological quality for several genotyping techniques was assessed, the reproducibility of PCR-restriction fragment length polymorphism genotyping data has been questioned as this method is highly subjective (22). In contrast to previous studies, we have used a highly specific genotyping techniques (concordance between LightTyper, Taqman, and Sequenom techniques and direct sequencing (genotyping gold standard) is >99.9%, D.M., unpublished data). We have also performed several genotyping quality controls as regenotyping with exclusion of families with genotype incompatibilities, which allow us to rule out significant genotyping errors confidently. Because of the molecular complexity of the INS VNTR, most studies, including ours, have assessed associations indirectly by genotyping the −23Hph1 polymorphism (rs689), which is in complete linkage disequilibrium with INS VNTR in 99% of whites (1). A recent study where a refinement of the genotyping is proposed suggests that a subclass allele of INS VNTR class I named ID allele showed stronger associations with fasting insulin levels in obese children (23). It is noteworthy that the transmission of this subclass allele was not reported in the childhood obesity trios of this study (23). Genotype refinement of the INS VNTR would probably provide a better assessment of this variant's effect, given that a difference in enrichment in the ID allele in obese populations could lead to inconsistent findings.
In summary, our study did not support any association between INS VNTR and childhood obesity, insulin resistance, or insulin secretion thereby suggesting no major role of this variant at least in the white genetic background.
In this study, 1,023 obese families were analyzed. We analyzed DNA from 449 nuclear French families with at least one obese offspring and both parents, all whites recruited in the Pediatric Endocrine Unit of Jeanne de Flandre Hospital, Lille or through national media campaign. Obese children were defined as BMI >97th percentile for age and gender in French population of children (317 girls/309 boys; BMI = 29.1 ± 6.3 kg/m2; age = 11.2 ± 3.1 years). German population was 574 families, each comprising at least one obese child or adolescent with a BMI ≥90th percentile and both biological parents. Altogether, 834 obese children and adolescents (464 girls/371 boys; BMI = 30.9 ± 6.0 kg/m2; age = 14.1 ± 3.8 years).
Quantitative traits were only analyzed in French children. During the OGTT, overnight fasting children received 1 g glucose/kg if subject's weight was < 50 kg or 75 g glucose if subject's weight was ≥ 50 kg. Plasma glucose was measured at 0, 30, 60, 90, and 120 min using the glucose oxidase procedure and insulin levels were measured using double-antibody radioimmunoassays. Insulin sensitivity was assessed by fasting insulin levels, homeostasis model assessment of insulin resistance (calculated as fasting insulin/(22.5e-ln(fasting glucose)) and ISI (calculated as ISI = 10,000/SQRT(G0 × I0 × G_mean_OGTT × I_mean_OGTT)). Insulin secretion was assessed by the insulinogenic index as (30-min insulin−fasting insulin)/(30-min glucose−fasting glucose). Birth weight was available in 464 obese children. Fat mass was assessed by dual energy X-ray absorptiometry.
Total unselected families were 448 and included 286 pedigrees from the Fleurbaix-Laventie Ville Santé II study (211 females/226 males; BMI = 18.8 ± 3.3 kg/m2; age = 13.5 ± 2.5 years) (13) and 162 pedigrees from the Haguenau cohort study (unselected pedigrees for obesity whose offspring birth weight was between the 25th and the 75th percentiles; 159 females/146 males; BMI = 22.5 ± 4.3 kg/m2; age = 22.1 ± 3.9 years). LBW families were 350 families with offspring birth weight under the 10th percentile (314 females/289 males; BMI = 22.6 ± 3.8 kg/m2; age = 22.1 ± 3.9 years).
All adult participants and parents of children signed informed consent. The genetic study was approved by ethical committees in Lille for childhood obesity and Fleurbaix-Laventie Ville Santé studies, in the University of Paris (St Louis, Paris) for the Haguenau study and by the ethics committees of the Universities of Marburg and Duisburg-Essen for the German studies.
The number of repeats of the INS VNTR are not randomly distributed and occur in a short repeat (class I: from 26 to −63 repeats), intermediate repeat (class II: ∼80 repeats), and a long repeat (class III: from 141 to 209 repeats) (1). In whites, class II repeat is rare and class I and class III frequencies are ∼0.70 and 0.30, respectively. We assessed INS VNTR variant associations through the rs689 located in intron 1, in almost complete linkage disequilibrium (r2 = 0.99) in whites (1). We used LightTyper technology (Roche, Meylan, France) to genotype French DNAs as described previously (5). Genotyping was performed two times in 250 DNAs and genotyping concordance rate was >99%. German DNAs were genotyped using matrix-assisted laser desorption/ionization-time of flight mass spectrometry of allele-specific primer extension products (Sequenom, San Diego, CA). Single-nucleotide polymorphism assays were designed using the SpectroDesigner software (Sequenom). Genotype compatibilities were checked within all families using PedCheck (24) before performing statistical analyses.
Statistical power was determined using the Quanto program. For association with childhood obesity, we used TDT which compares the number of transmissions of the at-risk allele, from heterozygous parent to affected offspring, to its expectation (50%). A McNemar χ2 test assesses the significance. We used the TDT implemented in unphased package (25) in the French families and the STATA package gamenu (http:www-gene.cimr.cam.ac.ukclaytonsoftwarestata) in the German families. The ORs 95% confidence interval was calculated using the link http:www.dimensionresearch.comresourcescalculatorsconf_prop.html. We used a Cochrane test to check for the lack of heterogeneity between the French and the German samples (all alleles P = 0.09, paternal alleles P = 0.06, maternal alleles P = 0.47) and a Mante-Haenszel test to assess the transmission in the combined sample of 1,023 families. We used the QTDT implemented in unphased to analyze residuals for fasting insulin levels homeostasis model assessment of insulin resistance, insulinogenic index, and fasting glucose levels after adjustment for age, gender, and BMI and residuals for birth weight after adjustment for gender and gestational age. P < 0.05 was considered statistically significant. We used Bonferroni correction to adjust for multiple testing analyses in obese families (number of tests = 9; new threshold P = 0.005).
This study was partly supported by the Diabomics grant from the Agence Nationale de Recherche and a grant from the “Fondacoeur” association. We thank C. Dina for statistical advice, M. Deweirder, and F. Allegaert for DNA extraction and S. Gaget for databases management. We are indebted to all families that participated in this study. The research in the German obesity families was supported by the European Union (FP6 LSHMCT-2003-503041) and the Federal Ministry of Education and Research (NGFN2 01GS0482, 01GS0483).
The authors declared no conflict of interest.