Introduction
Obesity is becoming an increasingly important clinical and public health challenge worldwide and is associated with several comorbidities such as type 2 diabetes, cardiovascular diseases, metabolic syndrome and certain forms of cancer [
1‐
3]. The complex etiology of obesity reflects effects of genes and environment as well as their interactions [
4]. In this context, an understanding of the effects of environment and genes on obesity and also their interactions is important to provide a basis for determining the role they could have on the development and prevention of obesity.
In recent years, several independent genome-wide association (GWA) studies reported significant associations of common genetic variants near
INSIG2(rs7566605) [
5], in the
FTO gene (rs9939609) [
6‐
9] and near
MC4R (rs17782313, rs17700633) [
10] with body mass index (BMI) as a measure of general adiposity and for
MC4R rs12970134 with waist circumference as a measure of central adiposity [
11]. Associations of
FTO and obesity have been broadly replicated in various populations including Caucasians, Asians and African–American children [
7,
8,
12,
13], but could not be confirmed in an African population [
14]. Similarly, the association of the common variants near
MC4R with obesity were replicated in Caucasians [
15‐
19], but two recent studies showed inconsistent findings in African–Americans [
15,
19]. In contrast, the association of the common variant near
INSIG2 has not been consistently replicated [
20,
21].
The rise of overweight and obesity in youth is of particular concern in the current obesity epidemic. Many overweight and obese children go on to become obese adults [
22] and are at greater risk of developing cardiovascular disease and metabolic syndrome [
1,
2]. Thus greater insight into the development of adiposity from childhood into adulthood is needed. However, most of the above mentioned studies were cross-sectional and do not offer information on the impact of genetic susceptibility on interindividual differences in development of obesity over time. To the best of our knowledge, the few available longitudinal studies [
16,
23‐
25] have not investigated the influence of these common variants on growth curves for adiposity from childhood to adulthood, nor did they study the interaction of these common variants with ethnicity or gender.
The main purpose of the current study was to assess the effect of these common variants in
FTO, near
INSIG2 and near
MC4R on longitudinal development of general and central obesity from childhood to early adulthood in European–American (EA) and African–American (AA) youth available from the Georgia Cardiovascular (CV) Twin study [
26] and the Blood Pressure (BP) Stress Cohort study [
27,
28]. We further investigated whether the effects of these common variants on general and central obesity were moderated by ethnicity or gender.
Discussion
In the current study, we investigated the role of common variants in FTO and near the INSIG2 and MC4R loci identified through GWA studies on development of obesity from childhood to adulthood. We performed growth curve modeling in 1592 EA and AA youth from two longitudinal studies separately, and pooled results through meta-analysis. There were five main findings in this study. The first was that we replicated the effect of FTO rs9939609 on levels of BMI (increasing with 2.1% per A-allele), weight (increasing with 1.9% per A-allele) and waist circumference (increasing with 1.2% per A-allele) in the meta-analysis with explained variances of 0.31, 0.14 and 0.16, respectively. The second was that MC4R rs17782313 was significantly associated with triceps skinfold (increasing with 5.0% per C-allele). The third was that no significant interaction between any of the SNPs and ethnicity on any of the adiposity-related phenotype was observed. The fourth was that significant interactions between MC4R rs17700633 and gender on the level of subscapular, suprailiac skinfold and sum of skinfolds were observed. Finally, a significant effect of MC4R rs17700633 on increase of triceps skinfolds with age (i.e., slope) was observed.
Since Frayling et al. [
9] first reported the significant associations between the
FTO variant rs9939609 and obesity-related phenotypes such as BMI, weight, waist circumference, %BF and skinfolds in both children and adults, these associations have been replicated in Europeans [
7,
8,
24,
25,
39] and Asians [
12] through both cross-sectional and longitudinal studies. In our longitudinal studies, we found that each copy of the
FTO rs9939609 A allele was significantly associated with BMI with 2.1% increase per allele. For example, within a population with average BMI of 25 kg/m
2, the approximate per A-allele effect is ~0.5 kg/m
2, which is similar to the effect that Frayling reported in UK children at the age of 11 years (0.4 kg/m
2) and a little higher than that in Finnish children at the age of 14 years (0.1 kg/m
2) [
9]. A cross-sectional sample of 1,600 subjects has 80% power at an alpha of 0.05 to detect an effect size of the SNP of 0.5% of the trait variance. Thus, our meta-analysis of two cohorts with repeated measures will have had sufficient power to detect even smaller effects. We found that the variance in BMI explained by rs9939609 was 0.31%, which is lower than the previously reported ~1.0% and probably due to sample variability [
9]. In addition, we replicated the association of
FTO rs9939609 with weight and waist circumference, which is consistent with Frayling et al.’s findings [
9]. The effect of
FTO rs9939609 on sum of skinfolds did not reach significance. This could be due to a more inaccurate measurement procedure compared to BMI, or that the measures represent slightly different aspects of adiposity. Furthermore, several longitudinal studies have been conducted to investigate the effect of rs9939609 on the development of obesity over time [
24,
25,
40]. However, the results of the rs9939609 by age interaction were inconsistent. In the MRC National Survey of Health and Development (NSHD) cohort, a longitudinal birth cohort from 1946 and followed through to age 53 years, significant genotype-by-age interaction was observed for BMI, showing the association strengthened between age 2 and 20 years at a rate of 0.007 sex-specific standard deviation scores (SDS) (95% CI 0.003–0.01) per A-allele per year, reached a peak at age 20 years, and thereafter weakened with age from 20 to 53 years (rate of decline: −0.003SDS per A-allele per year). And similar pattern was observed for weight [
40]. An increasing effect of genotype over time (
P = 0.047) was observed in 1886 middle-aged females from the Cebu Longitudinal Health and Nutrition Survey (CLHNS) cohort [
25] with a maximum of 8 measurements spanning 22 years. However, the interaction between rs9939609 and age was not significant in either a female cohort of about 2,200 nurses followed up from 1976 to 2002 (average age at entry: 44 years) or a male cohort of about 3,500 Health professional followed up from 1986 to 2002 (average age at entry: 56 years) (
P = 0.08 and 0.20, respectively) [
24], although in the male cohort the genetic effects appeared to decrease at older age. In our longitudinal study in youth, we did not find any significant interaction between rs9939609 and age for any of the obesity-related phenotypes. That is, rs9939609 did not affect the slope of the growth curves of these traits. Although the association between rs9939609 and adiposity has been broadly replicated in Europeans [
7,
8,
24,
39] and Asians [
12], it could not be confirmed in an African population [
14]. In our study including both EA and AA youth, we did not find any significant interaction between rs9939609 and ethnicity on any adiposity-related phenotype. We also did not find any significant interaction between rs9939609 and gender, which was consistent with Hardy et al.’s finding of no differences between males and females in the genetic associations with body size across the life course [
40].
Since Loos et al. [
10] reported association between common variants near
MC4R and fat mass, weight and BMI in both children and adults, several studies have replicated the finding in European populations [
15‐
18]. In our longitudinal studies, we found the C allele carriers of rs17782313 showed higher weight, BMI and waist circumference compared to TT homozygotes, but the association did not reach statistical significance. However, we did find a significant effect of rs17782313 on triceps skinfolds level, and a borderline significant effect of rs177882313 on sum of skinfolds. The C allele at rs177882313 was found to be positively associated with height in European adults [
10,
17]. On the contrary, we found rs17782313 was negatively associated with height in youth, and stratified analysis showed significant negative association only in AAs (β = −0.9,
P = 0.035). Loos et al. [
10] also found the per-C allele effect of rs17782313 on BMI in children was about twice that observed in adults. In the NSHD cohort, significant genotype-by-age interaction was observed for weight, showing the association strengthened between birth and 20 years at a rate of 0.005 SDS (95% CI 0.001–0.008) per C-allele per year, reached a peak at age 20 years, and thereafter weakened with age (rate of decline: −0.002SDS per C-allele per year) [
40]. However, no significant genotype-by-age interaction was observed for BMI in the NSHD cohort [
40]. In our longitudinal study, we did not find significant interaction between rs17782313 and age on any of the obesity-related phenotypes, which is in line with findings for BMI in the NSHD cohort [
40]. A recent paper could not replicate the association of the common variants near
MC4R and obesity in African-Americans [
15]. In the current study including both EAs and AAs, we did not find any significant interaction between rs17782313 and ethnicity on any of these obesity-traits. Findings on the interaction between rs17782313 and gender on adiposity in previous studies are inconsistent [
40‐
42]. In 3885 Swedish adults, rs17782313 showed significant interaction with gender (
P
inter = 0.02), with the association limited to females (
P = 0.003) [
42]. On the contrary, a study in French adults found stronger association of rs17782313 with adiposity and fat mass deposition in males than in females (
P = 0.003 and 0.03, respectively) [
41]. In the NSHD cohort, no significant interaction between rs17782313 and gender for adiposity phenotypes was observed (
P > 0.05) [
40]. In our study, rs17782313 significantly interacted with gender on weight, and the stratified analyses showed that neither of the associations was significant. Further investigations are needed to ultimately clarify whether the effect of rs17782313 on obesity is modified by gender.
Loos et al. [
10] found that the association between rs17700633 and BMI seemed to be dependent on that of rs17782313. We did not observe any significant association between rs17700633 and weight, BMI or waist circumference, which is inconsistent with findings in a middle aged Danish population [
17]. We found significant interaction between rs17700633 and age on triceps skinfolds, which means rs17700633 affected the slope of the triceps growth curve. The A allele carriers showed a steeper increase in triceps skinfold per unit increase of age as compared to the GG homozygotes. However, we cannot exclude the possibility that this significant interaction is a chance finding. We observed significant interaction between rs17700633 and gender for subscapular, suprailiac and sum of skinfolds, with significant associations limited to males. These findings indicate that the effect of rs17700633 on skinfolds might be modified by gender.
The association of
INSIG2 rs7566605 with BMI [
5], has not been consistently replicated [
20,
21]. In our study, we did not find any significant association between rs7566605 and any obesity-related phenotype levels, nor did we find any significant interaction with ethnicity, gender or age.
A major strength of our study is the longitudinal design of both cohorts with up to 16 assessments in one of these. Longitudinal studies offer information on determinants of interindividual differences in the development of obesity over time. These two cohorts thus allowed us to investigate the influence of common genetic variants in growth curve modeling to obtain better insight into the development of obesity from childhood into adulthood. Longitudinal designs have superior power to detect genetic effects on the trait levels compared to cross-sectional studies, because measurement of subjects at multiple time points provides much better precision in determining the phenotypes [
43,
44]. As such, longitudinal studies such as this one require a fraction of the sample size of cross-sectional studies to achieve the same power. Another strength is the involvement of AA as well as EA youth, and the investigation of a potential interaction of these common variants with ethnicity. Several limitations of our study need to be mentioned as well. In these two studies, we used BMI and skinfolds as measures of general adiposity and waist circumference as a measure of central adiposity, which are convenient and low cost measurements, while more accurate measurements of percent body fat for general adiposity and visceral adipose tissue for visceral adiposity were not available. In these two studies, sexual maturation was not assessed, thus it could not be incorporated as a covariate.
In summary, in two longitudinal studies of EA and AA youth, we replicated the effect of FTO and common variants near MC4R on general and central adiposity. These effects were similar for EAs and AAs. Furthermore, these common variants did not affect the slope of adiposity development from childhood to adulthood with one exception. Common variants for obesity identified in genome-wide association studies have detectable but modest effects on growth curves for adiposity in African– and European–American youth.