Longitudinal effects of FTO gene polymorphism on body composition, cardiorespiratory fitness, physical activity, inflammatory markers, and cardiovascular risk in children and adolescents. “The UP & DOWN study”

The role of polymorphism rs9939609 of the FTO gene has been related with fat mass and cardiovascular risk in adults, but it remains unclear in children and adolescents. Hence, the main aim of this study was to determine the FTO polymorphism effects on body composition, cardiorespiratory fitness (CRF), physical activity (PA), inflammatory markers, and cardiovascular risk both in cross-sectional analysis and after two-years of follow-up in children and adolescents. A total of 2129 participants were included in this study. The rs9939609 polymorphism was genotyped. Body composition measurements, CRF, and moderate- to-vigorous PA (MVPA) were determined at baseline and after two-year of follow-up. Moreover, plasma leptin and adiponectin were also determined as inflammatory markers. Furthermore, an index of cardiovascular disease risk factors (CVDRF-I) was calculated. Codominant (TT vs. TA vs. AA) and dominant (AA + AT vs. TT) models were applied for statistical analysis. The results showed a main effect of the FTO genotype on body composition measures in both first and third year ( p < 0.05), with lower adiposity in TT compared with AA


| INTRODUCTION
The accumulation of fat in childhood is related to health complications (i.e., increased blood pressure, hyperinsulinism, dyslipidaemia, and insulin resistance) that may persist into adulthood. 1The lack of physical activity (PA) and an excess in energy intake are the main factors that increase the prevalence of obesity and overweight in children and adolescents which is associated with lower levels of cardiorespiratory fitness (CRF) and increasing inflammatory markers, such as leptin and adiponectin, and cardiovascular disease (CVD) risk factors. 2,3However, there are many children who despite having a similar diet and PA, have different degrees of these parameters.In fact, there are more than 120 genes that affect the degree of fat accumulation, inflammatory markers, CRF, or CVD risk during life. 4 meta-analysis suggests that approximately 50%-70% of variation in the body mass index (BMI) is attributable to genetic differences. 5The fat mass and obesity-associated (FTO) gene encodes a 2-oxoglutarate-dependent nucleic acid demethylase, which catalyzes the demethylation of 3-methylthymine in single-stranded DNA with the production of succinate, formaldehyde, and carbon dioxide.This chemical and dynamic modification in RNA soon made it a target of study as an epitranscriptomic marker, playing an important role in the regulation of RNA processing.
FTO is ubiquitously expressed in human tissues but is highest in the hypothalamic nuclei that regulate energy balance.Indeed, FTO mRNA levels is also regulated by processes such as feeding and fasting. 6Genetic variants in the FTO gene are associated not only with human adiposity but also with metabolic syndrome and CVD. 7However, the relationship between FTO gene with CRF and PA remains unclear. 8Despite that, it has been shown that the possible genetic susceptibility by FTO variants on adiposity could be blunted through PA, 9 hence this variable should be taken into account in the analysis.
Among the different variants of FTO gene, the rs9939609 polymorphism (A/T) in intron one of FTO gene is one of the most important genetic factor for obesity susceptibility to date, with subjects carrying the A allele showing a higher BMI compared to the TT genotype at adult age. 10 It is important to note that most of these published studies on polymorphisms and adiposity have focused mainly on adult populations with contradictory results in children and adolescents' population.In addition, there are currently no longitudinal studies that determine the influence of the polymorphism of the FTO gene on the change of body composition, inflammatory markers, and CVD risk in childhood and adolescence taken into account CRF and PA.It has been shown that FTO gene may influence the CRF levels, 11 while it has little or any effect on physical activity between groups. 12ence, the main aim of this study was to determine cross-sectionally and longitudinally (2-year follow-up) the influence of the FTO A/T polymorphism (rs9939609) on body composition, CRF, PA, inflammatory markers, and CVD risk factors in children and adolescents.
We defend the hypotheses that (i) subjects carrying at least one A allele of the polymorphism rs9939609 of the FTO gene will show worse levels of body composition, CRF, inflammatory markers, and CVD risk factors in children and adolescents, and (ii) participants with the AA and AT genotypes in the rs9939609 polymorphism of the FTO gene will accumulate greater amount of body fat throughout the two-years of follow-up in children and adolescents with a concomitant rise of inflammatory markers and CVD and a lowered CRF.

| Participants
We analyzed both cross-sectional and longitudinal data from the UP & DOWN study. 13Baseline data (September 2011-June 2012) were collected from a sample of 2225 or AA+AT group.These differences were maintained after accounting for pubertal maturity, sex, age, VO 2 max, and MVPA.Moreover, lower leptin level was observed in TT compared to AA+AT group in the third year.An interaction in Gene*Time*Sex was found in height and neck circumference in dominant model (p = 0.047; p = 0.020, respectively).No differences were found in CRF, MVPA nor CVDRF-I between groups.Hence, homozygous TT allele could be a protective factor against weight gain from early childhood.

K E Y W O R D S
cardiovascular disease, childhood obesity, FTO, genotype healthy Spanish children and adolescents (Figure 1) attending schools in the Cádiz and Madrid regions belonging to grades 1st/4th (6-7 and 9-10 years, respectively) for children, and 7th/10th (12-13 and 15-17 years old, respectively) for adolescents (baseline characteristics in Table 1).Two years later (September 2013-June 2014), 2129 participants completed the follow-up measurements of the studied variables (4.31% drop out).
Parents of the participants were informed about the study aims and written informed consents were provided.In accordance with the Helsinki Declaration, the study protocol was accepted by the Ethics Committee of the Puerta del Hierro Hospital (Madrid, Spain), the Bioethics Committee of the Research Council (Madrid, Spain) and the Research Committee on Human Subjects of the University of Cádiz (Cádiz, Spain).

| Pubertal maturity
Self-reported assessment of pubertal development of children and adolescents was determined by classifying them into one of the five stages of pubertal maturity defined by Tanner and Whitehouse. 14For this purpose, participants were shown images with different levels of development (levels ranging from 1 to 5), so that the participant had to indicate which of all of them represented his or her current state of development.

| Body composition measurements
Height and weight were measured using a telescopic stature-measuring instrument (Type SECA 225; range, 60-200 cm; precision, 1 mm; Hamburg, Germany) and an electronic scale (Type SECA 861; range, 0.05-130 kg; precision, 0.05 kg; Hamburg, Germany), respectively.BMI was calculated as weight/height squared (kg/m 2 ).Triceps and subscapular skinfold thickness were measured on the nondominant side of the body with a Holtain caliper (range, 0-40 mm; precision, 0.2 mm), and percentage of fat mass was estimated using Slaughter equations. 15Waist circumference and neck circumference were measured with a nonelastic tape (SECA 200; range, 0-150 cm; precision, 1 mm).Two nonconsecutive measurements were carried out in all assessment and average was used in rearward analyses.
The waist-to-height ratio (WHtR) was calculated from the original body composition index (waist circumference/height).Fat mass and fat-free mass index were derived as body fat (kg) and fat-free mass (kg), respectively, and divided by squared height (m 2 ).Using fat mass %, and body mass, the fat mass index was calculated.The validity and reliability of these measures have been previously determined. 16In addition, adiposity according to BMI and Slaughter's formulas have been found to correlate well with DXA and are therefore considered an acceptable method for initial estimation of body fat. 17

| Cardiorespiratory fitness
In order to assess CRF, the 20-m shuttle run test was used.It is a valid and reliable test in children and adolescents 16 and is included in the ALPHA health-related physical fitness test battery. 18In summary, the participants must run between two lines 20-m apart at a pace set by a prerecorded sound signal.The initial speed was 8.5 km/h, and it increased by 0.5 km/h in each section.The test is concluded when the participant fails to reach the line on time twice in succession.Maximal oxygen consumption (VO 2 max [Y, ml kg −1 min −1 ]) was estimate using the Léger equation 19 from the maximal aerobic shuttle running speed (X 1 , km/h) and age (X 2 , year as the lower rounded integer) through: The regression equation of Leger et al. has been validated and can be used to estimate VO 2 max within an acceptable margin of error in this age group, and it is recommended to consider the 20-m shuttle run test to include in schoolchildren's physical fitness batteries. 20The continuous form of the variable was used for the main analyses.

| Physical activity
Moderate-to-vigorous physical activity (MVPA) was assessed by accelerometers (GT1M, GT3X, and GT3X+ models, Actigraph TM, LLC, Fort Walton Beach, FL, USA), which have previously been shown to be technically reliable instruments for this purpose. 21ifferent studies have shown strong agreement between GT1M, GT3X, and GT3X + measurements, 22 so additional calibration was not considered necessary.
Participants wore the accelerometer for 7 days, which was firmly attached to the back of the hip with an elastic band, positioned as close as possible to the centre of gravity, and participants were instructed to wear it only during waking hours, removing it during water activities according to previously established procedures. 23The epoch was set at 10 s.To be considered a valid day, participants had to wear it for at least 10 h per day.Only participants whose accelerometer recorded at least three valid days were included in the analysis (14.30% failed).
Data were downloaded and analyzed using Actilife software (v.6.6.2Actigraph TM, Pensacola, FL, United States).PA was expressed as average intensity expressed in counts per minute (cpm) and amount of time (minutes/day) spent in MVPA.The following cutoff points were used to define the different activity intensities: light between 100 and 2000 cpm; moderate between 2001 and 4000 cpm; and vigorous >4000 cpm.These cutoff points for defining the intensity categories were selected in relation to other previous studies conducted with European children and adolescents. 24Time without wearing the accelerometer was defined as a 60-min period of zero counts and had an allocation for a maximum of two consecutive minutes (<100 cpm).In the case of sedentary habits, they were quantified using the accelerometer under the same conditions that we explained for the assessment of physical activity.Taking into account the cutoff points of the HELENA study, sedentary behaviors were considered to be those whose values were between 0 and 99 cpm.

| Blood sampling
Inflammatory markers, leptin, and adiponectin were obtained in plasma after the extraction of a fasting blood sample from the cubital vein (Immunoassay (xMAP Techonology) using a kit plex: YB0000002Y Bio-Plex Human Diabetes 3-Plex Assay and 171-A7003M Bio-plex Pro Human Diabetes Adiponectin Assay), in the morning at the schools attended by the participants.Glucose, triglycerides, and high-density lipoprotein cholesterol (HDL-c) were determined through colorimetric assay (AU2700 Olympus analyser).

| Systolic blood pressure
Blood pressure was measured using a validated digital automatic blood pressure monitor (OMRON M6, OMRON HEALTH CARE Co., Ltd., Kyoto, Japan).The measurement was performed with the participants sitting quietly for 5 min before taking the measurements, with the left arm supported and in an extended position.Two measurements were taken 1-2 min apart.If the first two readings differed by >5 mm Hg, an additional third measurement was performed, eliminating the higher value.Mean values for systolic blood pressure and diastolic blood pressure (mm Hg) were calculated separately.

| Clustered cardiovascular disease risk factors
An index of CVD risk factors (CVDRF-I) was calculated from the mean of the standardized values (by population and sex) of systolic blood pressure, waist circumference, glucose, triglycerides, and HDL-c (CVD risk factors).Due to those higher levels of HDL-c represent a lower risk of CVD, the standardized value of HDL-c was multiplied by (−1).Considering to the close relationship between BMI and waist circumference, another index of CVD risk factors (CVDRF-II) was calculated excluding waist circumference.The calculations were done separately by population (children and adolescents) and sex (boys and girls). 25

| Genetic sampling
Buccal cells were collected on swabs or FTA® cards and genomic deoxyribonucleic acid was extracted and purified according to standard phenol/chloroform procedures followed by alcohol precipitation.The allelic discrimination analysis was performed by predesigned Life Technologies TaqMan® SNP Genotyping Assays on demand for the fat mass and obesity associated (FTO) rs9939609 polymorphism (ID: C_30090620_10).Subsequently, a polymerase chain reaction (PCR) amplification was performed using a StepOne™ Real-Time PCR System (Life Technologies, Foster City, CA) with a denaturation stage at 95°C for 10 min, 50 cycles of denaturation at 92°C for 15 s, annealing/extension at 60°C for 1 min, and a final extension stage of 30 s at 60°C.

| Statistical analyses
Descriptive statistics are presented as mean and SD A chisquared test was used to test the Hardy Weinberg equilibrium (HWE).Kolmogorov-Smirnov test was used to examine for normal distribution.Cross-sectional analysis was, respectively, performed at 1st and 3rd year to avoid the loss of sample size with one-way ANOVA to compare main effects between groups: TT versus TA versus AA (homozygotes and heterozygote); or with T-student to compare the difference between dominant model groups: AA+AT versus TT.Pubertal maturity, sex, age, VO 2 max, and MVPA were used, individually and jointly, as covariates due to its relationship with body composition.Moreover, MVPA was analyzed after adjusting for wear time.
After adjusting for age firstly and all covariates jointly afterwards, a longitudinal analysis was performed using a 3 × 2 and 2 × 2 generalized linear mixed effect model for repeated measure analysis comparing two main effects: (a) the genotype: (3 groups) TT versus TA versus AA (homozygotes and heterozygote) or (2 groups) AA+AT versus.TT (dominant model), and (b) time: (2 moments) 1st-3rd Year with a genotype*time interaction.Moreover, time*gene*sex interactions were also analyzed.Bonferroni post hoc analysis was performed to identify differences.Statistical analyses were performed with SPSS version 21.0.for Windows (IBM Corp, Armonk, New York), with significance set at p < 0.05.

| RESULTS
Mean participants' characteristics in our population divided by genotype (TT, TA, and AA) are presented in Table 2.The HWE test showed that χ2 = 0.778, d.f.= 2, p = 0.377, so the population is consistent with HWE, and confirming that the allele types were randomly sampled.The expected frequencies were TT (p = 0.3481) n = 742, TA (p = 0.484) n = 1029, and AA (p = 0.168) n = 357.The distribution of rs9939609 genotype frequencies were 35.32% for TT, 47.44% for TA, and 17.24% for AA.Because age and sex covary with many features of body composition, it was important that the genotype distribution for FTO was not significantly biased with respect to age both the first year (F = 1.802, p = 0.165) and the third year (F = 0.604, p = 0.547), while there were no bias in genotype frequency with respect to sex (χ2 = 0.462, d.f.= 2, p = 0.794).

TA vs. AA)
Descriptive statistics are shown in Table 2 as mean ± S.D.There was a main effect of FTO genotype (TT vs. TA vs. AA) on different body composition measurements, both in the first and third year of evaluation (p < 0.05).However, there was not any interaction of time*gene or time*gene*sex.No significant differences were observed in CRF, PA, inflammatory markers, or CVD risk clusters between FTO groups.

| TT versus AA
In terms of body composition measurements, Bonferroni post hoc comparison showed that in both, first and third year.TT genotype compared to AA group had lower BMI (point estimate -0.Similar results about fat-free mass index were obtained after adjusting for age in the first year (point estimate 0.675, CI 0.028/1.322,p = 0.038) and for age or VO 2 max in the third year (point estimate 0.958, CI 0.119/1.797,p = 0.019; point estimate 0.952, CI 0.204/1.701,p = 0.007, respectively).There was no significant effect of FTO genotype neither, first nor third year, on body mass, neck circumference, VO 2 max, MVPA, leptin, adiponectin, CVDRFI, and CVDRFII (all p > 0.05).

| TA versus AA
Related to body composition measurements, post hoc comparison only showed that TA had lower subscapular skinfold than AA group in the first year (point estimate −1.035, CI −2.036/−0.034,p = 0.040) but this difference disappeared after adjusting for Tanner, sex, age, VO 2 max, and MVPA jointly (p > 0.05) but not individually (p < 0.05).Moreover, after adjusting for these variables, TA showed lower values than AA group in body mass and waist circumference in the first year (point estimate −1.537, CI −2.907/−0.168,p = 0.022; point estimate −1.182, CI −2.296/−0.068,p = 0.033; respectively).
TA also showed higher fat-free mass index in first year compared to AA group after adjusting by age (point estimate 0.667, CI 0.048/1.285,p = 0.030), and similar results were obtained after adjusting for VO 2 max in the first year (point estimate 0.644, CI 0.069/1.220,p = 0.022) and in the third year (point estimate 0.717, CI 0.004/1.429,p = 0.048).However, these differences were disappeared after adjusting for Tanner, sex, age, VO 2 max, and MVPA jointly.

| Differences between FTO dominant model groups (AA + AT vs. TT)
Differences on body composition, cardiorespiratory fitness, physical activity, inflammatory markers, and cardiovascular risk by FTO dominant model groups are shown in Table 3 as mean ± SD The distribution of rs9939609 genotype frequencies were 64.68%AA+AT groups and 35.32% for TT.Comparing all individuals with allele A (AA+AT) with TT group, significant differences appear in Tanner in the first year (point estimate 0.128, CI 0.013/0.244,p = 0.029), while TT had lower subscapular skinfold and waist circumference in the third year (point estimate −0.943, CI −1.644/−0.241,p = 0.008; point estimate −0.964, CI −1.888/−0.039,p = 0.041).Moreover, BMI, triceps skinfold, fat mass, and fat mass index were also lower in TT group compared to AA+AT, both first and third year, even when was adjusted for Tanner, sex, age, VO 2 max, and MVPA (p < 0.05).TT group compared with AA+AT group showed also lower body mass, subscapular skinfold and waist circumference, both first and third year, only when was adjusted for Tanner, sex, age, VO 2 max, and MVPA (all p < 0.05).While there were differences in height and neck circumference, only in the first year (point estimate −0.668, CI −1.335/−0.002p = 0.049; point estimate −0.230, CI −0.418/−0.042,p = 0.016), and leptin levels, only in the third year (point estimate −1332.099,CI −2207.738/−456.459,p = 0.003).

| DISCUSSION
The main findings of this study indicate that within Spanish schoolchildren, the presence of the A allele, homozygous or heterozygous for rs9939609 polymorphism (FTO gene), was associated with higher levels of adiposity, and these differences were maintained after two-years of follow-up, even when were adjusted for Tanner, sex, age, VO 2 max, and MVPA.In concordance, leptin level was lower in TT group compared to the group with the presence of the A allele (AA+AT) in the third year.Moreover, a sexual dimorphism is also observed in the influence of rs9939609 polymorphism during two-years of follow-up on height and neck circumference.However, this genetic variant has little or any effect on CRF, MVPA, or CVD clusters in our sample.
][28] In this line, some authors have found a positive association between additional minor alleles (A) and BMI from 5.5 years of age onwards, 29  an influence of the gene starting at 12 years of age, 30 while Rutters et al. added to this date a decrease at 13-14 years that intensifies later at 17. 31 However, in another longitudinal study of 32 years of follow-up, it was observed that neither the A nor the T allele at any age was associated with a greater increase or decrease in BMI, suggesting that the variant studied does not seem to affect changes in BMI levels at any stage of life. 32t is important to consider the role that PA may play in the effect of the polymorphism on obesity.Andreasen et al. showed that physically inactive FTO rs9939609 homozygous A allele carriers had higher BMI compared to inactive homozygous T allele carriers. 33However, this difference between alleles in BMI was not found in physically active subjects.Hence, the influence of FTO rs9939609 polymorphism on BMI could be hidden by PA levels as described in a meta-analysis. 34However, our results showed similar PA among groups, hence the differences found in the FTO group is not due to physical activity or inactivity.
Consistent with PA levels, all groups showed similar CRF levels in our study.This fact is important because a low CRF level is frequently associated with adiposity and contributes as an individual risk factor for many obesityrelated comorbidities. 35Although other authors have determined that CRF does not modify the effect of FTO variation on BMI or waist circumference. 36he mechanism by which FTO polymorphisms might affect fat accumulation is unclear.Although several studies support the association between FTO and food intake, 37 a meta-analysis published with 177 330 participants found no significant interactions between the FTO variant and total dietary energy intake but did find a positive relationship between BMI-increasing allele of the FTO variant and higher dietary protein intake, providing insights into the potential link between FTO, dietary protein intake, and adiposity. 38ptin also plays an important role in metabolism stimulating fat oxidation in the cells and in the regulation of appetite in the hypothalamus. 39As with obesity, chronically elevated leptin levels affect its resistance, which is characterized by reduced satiety, excessive nutrient intake, and impaired fat oxidation capacity and consequently an increase in total body mass. 39In our study, the TT group showed lower leptin levels, which may mean a higher sensitivity to leptin and therefore an increase in satiety levels and an improvement in fat oxidation, contributing to explain the differences in adiposity observed in our results.In concordance, it has been shown that TT group had higher maximal fat oxidation during exercise compared with AA and AT groups. 40lthough at younger ages we found no differences in CVD risk, a meta-analysis in adults (45-67 years) confirmed that the rs9939609 variant of the FTO gene is significantly associated with an increased risk of CVD and that the genetic effect is not mediated by changes in BMI and other conventional CVD risk factors, 41 suggesting that the presence of the A allele influences CVD risk.Although it is not known at what time of life it appears, according to our findings, this polymorphism could affect CVD risk in adulthood due to the influence on adiposity and leptin levels from early ages.
It also seems unclear whether sex may affect the influence of the gene over time.Although our results showed no sexual dimorphism in the influence of the FTO polymorphism in the 2 years of follow-up, an influence of the gene has been previously observed from the age of 12 years mainly in girls, 30 probably due to the variation of the endocrinological changes that occur at puberty and the influence of the different sex hormones. 42However, in adults, some studies seem to relate the polymorphism to a greater influence in boys, which may be due to a different inheritance pattern in boys than in girls. 43This reveals the need to consider sex as an important factor modulating F I G U R E 2 Significative differences in time*gene*sex interaction in height and neck circumference after adjusting for age.
the genetic association of FTO with overweight and obesity, although we do not know at what age it occurs.
Given the above, longer-term follow-up studies are needed to examine at what ages and by what mechanisms the FTO genotype begins to influence fat mass and body composition.Longitudinal studies can shed light on the effect of the polymorphism on individuals over time.This would allow the study of this genetic marker in children and adolescents and its relationship with various phenotypic characteristics over time, while allowing a better understanding of the pathogenic mechanisms involved in obesity-associated diseases, being useful as a predictive element to buffer health problems related to these diseases.
Therefore, the TT genotype of the FTO gene could be a protective factor against weight gain from early childhood and could reduce the risk of obesity-associated comorbidities in individuals genetically predisposed to obesity.

| Perspective
Although all groups showed similar levels of CRF and MVPA, the presence of the A allele, homozygous or heterozygous for the rs9939609 polymorphism (FTO gene), was associated with higher levels of adiposity from baseline to 2 years of follow-up in schoolchildren.In concordance, the TT group showed lower leptin levels, which may mean a better sensitivity to leptin, increasing both satiety and fat oxidation.
One of the keys to avoid morbidities associated with childhood obesity both at this stage and in the adult phase is early detection.School age and adolescence are crucial stages for the configuration of eating habits and other lifestyles that will persist into adulthood.
These findings would make it possible in the future to predict individuals genetically predisposed to suffer from overweight/obesity and, therefore, to design intervention plans in critical periods of childhood and adolescent development and thus mitigate the problems derived from overweight in adulthood.

| LIMITATIONS AND STRENGTHS
The main limitations of our work are as follows (i) the lack of the evaluation of energy intake as important variable of fat gain, (ii) we do not have much ethnic variability, a factor that can have an important influence, and (iii) body composition measurements such as fat mass have been estimated from anthropometric measurements, without a more accurate method such as dual-energy X-ray absorptiometry (DXA).However, it has important strengths such as having many body composition variables, combined with CRF and MVPA, as well as blood and CVD measurements in a large sample.In addition, all these measurements have been studied longitudinally after two years of follow-up, providing novel results with several measurements taken into account objective MVPA and CRF as cofounders.
other authors have observed T A B L E 3 Differences of FTO genotypes by dominant model (AA+AT vs. TT) on body composition, cardiorespiratory fitness, physical activity, inflammatory markers, and cardiovascular risk with the cross-sectional and longitudinal data. 24 Baseline characteristics divided by sex.
T A B L E 1Note: All data mean ± SD Abbreviations: BMI, body mass index; CVDRF, cardiovascular disease risk factor index; CVDRF-II, cardiovascular disease risk factor index (excluding waist circumference); MVPA, moderate-to-vigorous physical activity.*p < 0.05 significant difference.