Childhood obesity is an increasing public health problem all over the world, including many low- and middle-income countries. This complex disease is a major cause of morbidity and mortality; a higher risk of obesity is associated with premature death and disability in adulthood. According to the World Health Organization (WHO), 1.6 billion adults (aged 15 years and older) were overweight and 400 million were obese in 2005. By 2015, these figures are predicted to rise to 2.3 billion overweight adults and over 700 million obese adults [1].

Characteristics of the Western lifestyle such as excessive energy intake and low levels of physical activity largely influence the increasing prevalence of obesity worldwide. In addition to environmental factors, genetic factors also clearly contribute to obesity-related phenotypes, with heritability estimates ranging from over 50% to 60% for body mass index (BMI) [2].

A recent Genome Wide Association study (GWAs) for type 2 diabetes mellitus (T2DM) susceptibility genes identified common variants in the FTO gene (fat mass and obesity associated) that predispose adults to diabetes through an effect on BMI [3]. Several single nucleotide polymorphisms (SNPs) have been described in this gene, but a T-to-A change in intron 1 (rs9939609) is the most widely investigated and is consistently associated with obesity-related phenotypes, mainly body mass index (BMI). Approximately 63% of all individuals of European origin carry at least one A - risk allele of this FTO variant, 16% being A/A homozygotes [4]. The risk of being obese correlated to each A allele ranges from 1.35 to 1.76 [5].

Of the many tissues in which the FTO gene is highly expressed, the hypothalamic nuclei are of particular interest because the nuclei control energy balance and are influenced by nutritional signals [6, 7]. However, the exact role of the FTO gene product in the regulation of human energy metabolism is still unclear. While there is some evidence indicating that the FTO gene product functions during early infancy [3], many questions persist regarding the timing of gene function and what systems are affected.

Multiple studies have associated FTO polymorphisms with obesity in different European populations, thereby making this gene one of the most important loci implicated in polygenic obesity thus far [8]. However, the contribution of the FTO common variants to obesity is controversial in African American, Asian (Chinese Han and native Oceanic) and Hispanic populations [9–13]. To our knowledge, there are no studies in Brazil addressing the effect of FTO variation on BMI in children.

In the present study, we analyzed the association between the FTO rs9939609 gene variant and predisposition to weight gain by profiling excessive accumulation of body mass and diet composition in a cohort of children followed from the time of birth until the age of 8 years old. The major findings were then replicated in another independent sample of children.

The total sample was stratified between whites and non-whites. No statistically significant differences were observed between these groups for allelic and genotypic frequencies (Table 1; P = 0.410 and P = 0.273, respectively). No differences were observed regarding either anthropometric or dietary measurements among ethnicities, also. So, the individuals from both ethnicities were grouped in further analyses.

When we analyzed the diet composition, we found a significant difference in the total energy intake from lipids between the three genotypes (P = 0.014). The Tukey post-hoc test revealed that the T/T genotype carriers had a higher mean of lipids intake than the T/A genotype (P = 0.013). When we adjusted the energy intake from lipids for BMI, this association was no longer significant (P = 0.055; data not shown). The total energy intake from carbohydrates was not different between genotypes.

Likewise, at 8 years of age we observed a statistically significant difference in the mean BMI Z-score among the FTO genotypes (P = 0.011; Table 2). When the Tukey test was applied, we found that the carriers of the A/A genotype had higher means of BMI than the T/A carriers (P = 0.014). For the other anthropometric parameters at the age of 8, we observed that the sum of skinfolds showed marginally significant differences among the FTO genotypes. The percentage of BMI variation between the ages of 4 to 8 also showed significant differences between the genotypes (P = 0.015). The variables related to the dietary profile showed no significant differences between the genotypes at the age of 8.

The Sapucaia sample was further divided in children (4 to 10 years old, n = 293) and adolescents (11 to 18 years old, n = 322) and the anthropometric measurements were compared among genotypes in these subgroups. There were no significant differences in BMI Z-score, sum of skinfolds or waist circumference in children. In adolescents, the only statistical significant difference was observed in the sum of skinfolds (mean ± SD, T/T = 30.7 ± 14.5 mm; T/A = 34.4 ± 14.8 mm; A/A = 36.2 ± 14.5 mm; P = 0.020), although for BMI Z-score, the P-value was 0.065. This loss of significance can be due to the smaller sample size when the sample is reduced to about half of its size to perform these analyses.

In the present study, we confirmed the association of the FTO variant rs9939609 with body mass index and related phenotypes in Brazilian children and adolescents. To our knowledge, this is the first study regarding this gene and phenotypes related to BMI in Latin American children and adolescents.

The frequency of the minor (A) allele found in Brazilian children was 40%, similar to those reported in European cohorts of children and adolescents [3, 21, 23] and similar to those reported for Brazilian adults from the same geographic region as ours [24, 25].

After the first report identifying the association between the FTO gene and obesity-related phenotypes [3], several studies have confirmed the association between rs9939609 and increased BMI in populations of European children and adults [23, 26–31] while negative results were found in some studies regarding populations of Oceanic [11], African [32, 33] and Asian ancestries [10]. However, other studies have found significant associations in these populations [34–41]. The conflicting results among different ethnic populations are likely the result of varying degrees of linkage disequilibrium between SNPs, which suggests that the underlying causative variant is being tagged differently by rs9939609 in these populations [9, 23, 32] or that there are different gene-environment interactions. The Brazilian population is, in general, highly admixed with contributions from Europeans, Africans and Native Americans to its gene pool. The Southern Brazilian population, which the present study targeted, differs from this general pattern; its inhabitants are mainly of European descent, with an African contribution estimated at approximately 8 to 13% [42]. The Native American contribution is fairly low. Our results show that the rs9939609 SNP is consistently associated with obesity-related phenotypes in this population, even as early as 4 years old. Although this SNP is located in an intron and several studies have shown a strong genetic disequilibrium in this gene region, it is not possible at this time to rule out rs9939609 as a functional variant [4]. Rather, it is necessary to refine the strength and exact nature of the genetic signal that is functionally related to obesity or related traits [4].

Although several studies have reliably confirmed, to varying degrees, the association between the variation at FTO, BMI and related phenotypes [43], there is still some controversy regarding when FTO gene function begins to take effect. In our research, the FTO gene was not associated with any variable for 1-year-old children. We were only able to find an association of the A/A genotype of rs9939609 and a higher BMI Z-score by the age of 4. This finding agrees with a recent meta-analysis performed on 8 cohorts of European ancestry [43]. In that study, the authors suggested that the variation at FTO is associated with a shift in the timing of adiposity rebound; the additive effects of the A allele accelerated developmental age by approximately 2.4% per allele, which corresponded to an early adiposity rebound and a higher BMI later in childhood. According to these data, our longitudinal analysis has shown a greater BMI variation between the ages of 4 and 8 in the A/A subjects.

When we followed the same sample of children until 8 years old, we again found an association between the FTO genotypes and the BMI; the A/A homozygotes had higher mean BMI Z-scores than the children carrying one copy of the rs9939609 variant. The A/A genotype was also marginally associated with increased subcutaneous fat, evaluated through the sum of skinfolds. These findings are corroborated by previous childhood [31] and adolescents [44] studies. Furthermore, these results were replicated in an independent sample of schoolchildren between 5 and 18 years of age. In this second sample, the carriers of the T/A genotype possessed a higher BMI Z-score and more skinfolds than the T/T genotype. However, in these older children, the T/A and A/A individuals showed similar means of these two parameters, which diverges somewhat from the results in the first cohort, where the A/A and T/A genotype groups were different. Previous works indicate an additive effect of the A-allele [3, 22, 43, 45]. In all cross-sectional and longitudinal analyses performed, it can be observed that the significant differences regarding BMI are between the T/A and A/A genotypes in the cohort of São Leopoldo, which suggest that, in these young children, there is an recessive effect of the A allele on BMI. However, we cannot exclude the possibility that the difference between the A/A and T/T genotypes may have not been observed due to the moderate sample size evaluated herein or to some environmental or gene-gene interaction that cannot have been addressed with the current approach.

It has been hypothesized that the FTO gene has been associated with obesity because it can influence energy homeostasis by having a direct effect on food intake in animal models [46]. A study recently published by Larder et al. (2011) [47] suggests that the increased consumed energy in risk allele carriers is due to a major preference for energy-dense foods, specifically those with a higher fat content, rather than a general increase in the amount of food consumed. However, our data do not seem to corroborate this hypothesis. When we analyzed dietary composition, we observed that the A/A children consumed less energy from lipids by the age of 4. We hypothesized that this reduction in fat consumption in the 4-year-old children with the A/A genotype could be the result of greater parental diet control because this group possessed a higher BMI. In fact, when we adjusted the energy intake from lipids for BMI, the former association was no longer significant. Accordingly, at 8 years old, the association between the dietary composition and the FTO genotypes in our sample disappears and, unfortunately, in the independent sample the dietary patterns were not available. As stated previously, the exact role of the FTO gene product in the regulation of human energy metabolism is still unclear, and our findings do not suggest a higher total energy intake in the A-allele carriers.

It is known that the FTO is expressed in the hypothalamus, mainly in the arcuate nucleus, which is responsible for controlling appetitive behavior [6, 48]. The FTO variants may also influence macronutrient choice because the arcuate nuclei contains key neuronal populations that make up the central melanocortin pathway, and studies in mice have shown that Fto is expressed within proopiomelanocortin neurons. A reduction in central melanocortin tone has been linked to an increase dietary preference for fat [49]. Thus, it is possible that FTO variants are able, either directly or indirectly, to modulate hypothalamic melanocortin activity [47].

A recent study in mice reported the relationship of FTO and the epigenetic process. The FTO gene has been proposed as a transcriptional coativator that enhances the transactivation potential of the CCAAT/enhancer binding proteins (C/EBPs) from unmethylated as well as methylation-inhibited gene promoters, which suggests a role in the epigenetic regulation of development and maintenance of fat tissue [50]. While the exact function of the FTO gene remains to be clarified, it becomes difficult to speculate the mechanism underlying these findings.

There are some limitations to our study. First, the sample size is lower than in other studies; although even under these conditions, we were able to replicate the previous associations of FTO SNPs and increased BMI in children and adolescents. Another limitation is that our replication sample data were cross-sectional and detailed food intake data were not available, which precludes deeper analyses in this sample. Furthermore, only 127 (20.7%) of the children of the Sapucaia sample were in the same age range of the São Leopoldo cohort. On the other hand, relevant data are brought by this second sample exactly due to the fact that it comprehends children and adolescents of a higher age and allows us to infer if that the findings in younger children can be extended until the age 18.

Overall, our work presents new data regarding associations of the FTO gene in a previously unexplored population and confirms worldwide data regarding the influence of this gene on the genetic susceptibility for an increased BMI and specific adiposity measures. Furthermore, we were able to show, through a longitudinal analysis that this effect begins around the age of 3 to 4. The exact role of the FTO gene in body weight regulation as well as the functionality of the rs9939609 variant remain to be elucidated, but this genetic association is, undoubtedly, the most consistent finding to date.