Note to users. If you're seeing this message, it means that your browser cannot find this page's style/presentation instructions -- or possibly that you are using a browser that does not support current Web standards. Find out more about why this message is appearing, and what you can do to make your experience of our site the best it can be.

Science 12 January 2007:
Vol. 315. no. 5809, p. 187
DOI: 10.1126/science.1129402
Prev | Table of Contents | Next

Technical Comments

Comment on "A Common Genetic Variant Is Associated with Adult and Childhood Obesity"

Christian Dina,1* David Meyre,1 Chantal Samson,1 Jean Tichet,2 Michel Marre,3 Beatrice Jouret,4 Marie Aline Charles,5,6 Beverley Balkau,5,6 Philippe Froguel1,7

Herbert et al. (Reports, 14 April 2006, p. 279) reported an association between the INSIG2 gene variant rs7566605 and obesity in four sample populations, under a recessive model. We attempted to replicate this result in 10,265 Caucasian individuals, combining family-based, case-control, and general population studies, but found no support for a major role of this variant in obesity.

1 CNRS 8090-Institute of Biology, Pasteur Institute, Lille, France.
2 Institut Inter-Régional pour la Santé, Tours, France.
3 Department of Diabetology, Bichat Claude Bernard Hospital, Assistance Publique-Hôpitaux de Paris, France.
4 Institut National de la Santé et de la Recherche Médicale (INSERM) U563, Children's Hospital, Toulouse, France.
5 INSERM U780-IFR69, Villejuif, France.
6 Université Paris-Sud, Villejuif, France.
7 Department of Genomic Medicine, Imperial College London, UK.

* To whom correspondence should be addressed. E-mail: dina{at}good.ibl.fr

Herbert et al. (1) identified a common DNA variant 10 kb upstream of the INSIG2 gene associated with obesity in 9881 adults and children from different ethnic groups. Their study combined case-control, general population, and family studies. They concluded that variation in this gene, which had been found through whole-genome scans in families, may contribute to obesity under a recessive model. Association studies of genome screen results need to be confirmed by additional replication studies (2). We therefore genotyped the INSIG2 genetic variant rs7566605 in 10,265 subjects of French Caucasian descent. Our first data set was recruited for studying childhood obesity and comprised 449 families (F1, 2426 subjects) (3) with at least one sibling over the 97th percentile of body mass index (BMI) and 145 obese unrelated French children (CC, BMI over the 97th percentile). The second set was recruited for studying adult obesity (4, 5) and combined 386 families (F2, 1765 subjects), 350 obese (MO, BMI > 30) and 230 nonobese (CO, BMI < 27 kg/m2) unrelated French adults. The third data set was composed of families with offspring from a general (6) French population, the Fleurbaix Laventie Ville Santé (FLVS) study (287 families, 619 individuals). Finally, from the Données Epidémiologiques sur le Syndrome d'Insulino-Résistance (DESIR) cohort, we geno-typed a fourth data set of 4998 middle-aged unrelated French individuals who were followed for a period of 9 years (7). We performed association studies in four groups, as described below. The distribution of BMI according to genotype for each study is shown in Table 1. The minor allele frequency varied from 30% to 35% in our study populations, which is similar to the frequency reported by Herbert et al. (1). The genotypes were in Hardy-Weinberg equilibrium in each of our study populations.


Table 1. BMI (kg/m2) distribution according to rs7566605 genotypes in French case-control data sets. Study populations: (A) DESIR cohort. Cases: adults having obese status (BMI ≥ 30 kg/m2) at least once in the four clinical examinations (over 9 years). Controls: nonobese adults (BMI < 30 kg/m2) in any of the four clinical examinations. (B) Adult obesity. Cases: obese adults (BMI ≥ 30 kg/m2) from families F2, recruited for one overweight (BMI > 27 kg/m2), one morbidly obese (BMI > 40 kg/m2), and unrelated morbid obese individuals (sibships and unrelated). Controls: nonobese individuals (BMI < 30 kg/m2) from the adult obesity familial set and additional controls (BMI < 27) recruited within the same study. (C) Parents of F1 and FLVS. Cases: obese parents (BMI ≥ 30 kg/m2) of the obese children of the F1 childhood obesity families and parents of the FLVS families. Controls: nonobese parents (BMI < 30 kg/m2) of the obese children of the F1 childhood obesity families and parents of the FLVS families. (D) Childhood obesity. Cases: obese children (BMI percentile ≥ 97th) from the childhood obesity sample (sibships of F1 and unrelated individuals). Controls: nonobese (BMI percentile < 90th) from the childhood obesity sample and from the FLVS child sample (sibships).
A. DESIR sample Mean* SD SE Proportion of individuals with: BMI < 18.5; 18.5 ≤ BMI < 25; 25 ≤ BMI < 30; 30 ≤ BMI < 35; 35 ≤ BMI < 40; BMI ≥ 40
Females
GG 24.09 4.13 0.12 0.0368; 0.6424; 0.2284; 0.0719; 0.0171; 0.0034
CG 24.00 4.16 0.13 0.0365; 0.6475; 0.2200; 0.0826; 0.0086; 0.0048
CC 23.87 4.00 0.25 0.0502; 0.6641; 0.1969; 0.0811; 0.0077; 0.0000
Males
GG 25.43 3.34 0.1 0.0093; 0.4875; 0.4123; 0.0817; 0.0084; 0.0009
CG 25.54 3.41 0.1 0.0046; 0.4745; 0.4279; 0.0803; 0.0109; 0.0018
CC 25.35 3.47 0.21 0.0038; 0.5267; 0.3893; 0.0611; 0.0115; 0.0076
B. Adult obesity Mean SD SE{dagger}
Females
GG 39.31 12.17 0.59 0.0093; 0.1422; 0.0862; 0.1538; 0.0839; 0.5245
CG 37.38 11.21 0.51 0.0083; 0.1801; 0.1077; 0.1118; 0.1284; 0.4638
CC 38.05 11.26 0.98 0.0076; 0.1450; 0.1145; 0.1374; 0.0916; 0.5038
Males
GG 34.46 10.97 0.75 0.0000; 0.1963; 0.2420; 0.1918; 0.0868; 0.2831
CG 35.44 11.43 0.72 0.0000; 0.2112; 0.1633; 0.1952; 0.0837; 0.3466
CC 34.26 10.52 1.38 0.0169; 0.1525; 0.2712; 0.1695; 0.0678; 0.3220
C. Parents of F1 and FLVS Mean SD SE
Females
GG 28.86 7.62 0.50 0.0044; 0.3772; 0.2412; 0.1667; 0.1272; 0.0833
CG 29.00 8.64 0.49 0.0083; 0.3734; 0.2697; 0.1577; 0.1037; 0.0871
CC 29.50 7.78 1.18 0.0000; 0.3256; 0.2791; 0.2093; 0.0698; 0.1163
Males
GG 28.21 5.37 0.33 0.0078; 0.2335; 0.5019; 0.1566; 0.0700; 0.0311
CG 28.22 5.12 0.35 0.0047; 0.2736; 0.4292; 0.1934; 0.0755; 0.0236
CC 29.18 5.51 0.85 0.0000; 0.2143; 0.4524; 0.1429; 0.1429; 0.0476
D. Childhood obesity{ddagger} Mean SD SE{dagger} Proportion of individuals with BMI in percentiles: BMI < 0.05; 0.05 ≤ BMI < 0.58; 0.58 ≤ BMI < 0.87; 0.87 ≤ BMI < 0.97; 0.97 ≤ BMI < 0.99; BMI ≥ 0.99
Females
GG 2.49 2.43 0.12 0.0106; 0.1941; 0.1223; 0.0372; 0.0878; 0.5479
CG 2.60 2.25 0.13 0.0153; 0.1713; 0.1437; 0.0367; 0.0489; 0.5841
CC 2.45 2.38 0.26 0.0172; 0.1897; 0.1207; 0.0172; 0.1207; 0.5345
Males
GG 2.40 2.27 0.13 0.0295; 0.1967; 0.1246; 0.0426; 0.0721; 0.5344
CG 2.21 2.35 0.13 0.0261; 0.2313; 0.1466; 0.0293; 0.0684; 0.4984
CC 2.67 2.17 0.26 0.0147; 0.1912; 0.1324; 0.0147; 0.0293; 0.6176

* Mean is given for the BMI at first examination.

{dagger} Standard errors are estimated through the GEE procedure implemented in the Stata 5.0 software (command xtgee).

{ddagger} For the childhood case-control study, the BMI distribution is shown by percentiles from a general population.

We conducted tests of association in the presence of linkage and tests of simple association. As shown in Table 2, we observed no overtransmission of the rs7566605 C allele to obese children (BMI > 97th percentile) or adults (BMI > 30) in the F1 and F2 family data (P = 0.84 and P = 0.76, respectively) (8). Similar results were observed for more severely obese children (BMI > 99th percentile, P = 0.61). Furthermore, no association with BMI, corrected for gender and age, was found in the FLVS sample [quantitative family-based association test (FBAT), P = 0.61].


Table 2. Results of family-based and case-control analyses. Meta-analyses were performed on the French study populations alone and together with the study populations reported by Herbert et al. Tr, transmitted alleles; Non-TR, nontransmitted alleles.
Study Design Informative families Tr Non-Tr Test Recessive model Additive model
A. Test of association and linkage
Child obesity F1 Family 54 152 153 FBAT Z = 0.20 Z = 0.13
P = 0.84 P = 0.89
FLVS Family 31 - - Quantitative Z = 0.52 Z = 0.22
FBAT P = 0.61 P = 0.82
Adult obesity F2 Family 47 62 72 FBAT Z = -0.29 Z = 0.44
P = 0.76 P = 0.62
Study Design Genotyped Obese Controls Statistic Recessive model General model*
B. Independent case-control studies on adult obesity
DESIR Case-control 4998 905 4093 Logistic regression OR = 0.86 [0.68-1.11] {chi}22df = 1.35
P = 0.25 P = 0.49
Adult obesity Case-control 1572 1076 496 Logistic regression{dagger} OR = 0.93 [0.66-1.29] {chi}22df = 1.37
P = 0.67 P = 0.50
Parents (F1 and FLVS) Case-control 1023 329 694 Logistic regression OR = 1.18 [0.69-2.03] {chi}22df = 1.16
P = 0.61 P = 0.58
French population meta-analysis Mantel-Haenszel OR = 0.93 [0.77-1.12] Fixed effects
P = 0.61
Overall meta-analysis Mantel-Haenszel OR = 1.10 [0.98-1.23] Fixed effects
P = 0.10
C. Case-control study on childhood obesity
Children Case-control 1531 912 532 Logistic regression{dagger} OR = 1.11 [0.72-1.69] {chi}22df = 1.14
P = 0.67 P = 0.56
D. Test of association with the quantitative trait BMI
DESIR Cohort 4998 - - Linear regression{ddagger} ß = -0.17 [-0.5-0.17] {chi}22df = 2.14
P = 0.32 P = 0.34
FLVS parents Cohort 342 - - Linear regression ß = 0.36 [-1.39-2.12]
P = 0.68 P = 0.72
FLVS children Family 1138 - - Linear regression{dagger} ß = 0.27 [-0.08-0.62] {chi}22df = 3.42
P = 0.13 P = 0.18

* The general model included two variables for the single-nucleotide polymorphism genotype. The first one (0,1,2) is the number of C alleles, and the second reports whether the genotype is heterozygous (0) or not (1).

{dagger} GEEs (corrected for age and sex). Familial correlation was accounted for by using a sandwich estimator of the variance and exchangeable correlation.

{ddagger} Mixed model (corrected for age and sex) on four time points, every 3 years.

We performed association studies in four groups, described in Table 1, using the General estimating equation (GEE) method (9). This method allows logistic or linear regression analysis in clustered data (10), and thus pooling of related and unrelated individuals. The first case-control study was defined within the DESIR population (Table 1A). A second case-control study used adult cases and controls of the familial study F2 and the unrelated individuals from MO and CO (Table 1B). A third analysis was performed on the parents of the childhood obesity data set (F1) and parents of the FLVS study (Table 1C). The results of these three analyses of adult obesity were pooled in a meta-analysis. A BMI of 30 kg/m2 waschosenasa cutoff to define cases and controls, as in (1). The fourth study was on all the children, from F1, FLVS, and CC (Table 1D). Because allele frequencies in this study are correlated with those of the third case-control set, this study was not used in the meta-analysis.

As shown in Table 2, no significant association was observed between rs7566605 and obesity under a recessive model in any of the individual studies. When we combined the three independent analyses of adult obesity, we also detected no association with obesity [OR = 0.93 (0.77 to 1.12), P = 0.61] under the recessive model. No association was found under additive and general models. Moreover, no effect on BMI was shown within the DESIR cohort in a linear mixed model (P = 0.32) with up to four observations per individual, in children (P = 0.13), or in parents (P = 0.68) of the FLVS study (Table 2D).

Interestingly, the negative results found in the DESIR cohort are similar to those observed in the Nurses Health Study cohort in (1). Under the suggested recessive mode of inheritance, we would have 80% power to detect an increase of 0.6 baseline BMI units by genotype, the lowest effect shown in the KORA study (1) using Quanto software (11). The absence of replication in this French study population is therefore unlikely to be due to low power.

In conclusion, in a data set as large as that of Herbert et al. (1), we found no effect of the INSIG2 intronic variant on the risk of adult obesity or childhood obesity either in case-control or family-based designs. Our design combined extreme cases and general populations, which allowed testing of the effect both on a continuous trait and/or on morbid obesity. Furthermore, combining our results with the published case-control odds ratios (1) in a meta-analysis results in a global nonsignificant OR under a recessive model [1.10 (0.98 to 1.23), P = 0.10]. Although a major contribution of INSIG2 rs7566605 to the genetic risk of obesity in the West European population is unlikely, it remains possible that INSIG2 contributes to BMI variation in other ethnic groups.


References and Notes

Received for publication 1 May 2006. Accepted for publication 15 November 2006.


THIS ARTICLE HAS BEEN CITED BY OTHER ARTICLES:
A common VLDLR polymorphism interacts with APOE genotype in the prediction of carotid artery disease risk.
D. C. Crawford, A. S. Nord, M. D. Badzioch, J. Ranchalis, L. A. McKinstry, M. Ahearn, C. Bertucci, C. Shephard, M. Wong, M. J. Rieder, et al. (2008)
J. Lipid Res. 49, 588-596
   Abstract »    Full Text »    PDF »
Evidence of an Influence of a Polymorphism Near the INSIG2 on Weight Loss During a Lifestyle Intervention in Obese Children and Adolescents.
T. Reinehr, A. Hinney, T. T. Nguyen, and J. Hebebrand (2008)
Diabetes 57, 623-626
   Abstract »    Full Text »    PDF »


To Advertise     Find Products

ADVERTISEMENT

Featured Jobs

Science. ISSN 0036-8075 (print), 1095-9203 (online)