Effect of Obesity-Linked FTO rs9939609 Variant on Physical Activity and Dietary Patterns in Physically Active Men and Women

Single nucleotide polymorphisms (SNPs) in the fat mass and obesity-associated (FTO) locus are associated with obesity, but lifestyle factors may modulate the obesity risk related to FTO. This study examined the physical activity and dietary patterns of 528 physically active white men and women (mean (SD): 34.9 (9.5) years, 26.6 (4.3) kg·m−2) carrying different risk variants of FTO SNP rs9939609. Sex, age, and anthropometric measurements (stature, body mass, and waist circumference) were self-reported using an online questionnaire, and body mass index and waist-to-height ratio were calculated. Physical activity and eating behaviour were assessed using the International Physical Activity Questionnaire (IPAQ) and Three-Factor Eating Questionnaire (TFEQ), respectively. Body mass, body mass index (BMI), waist circumference, and waist-to-height ratio were not significantly different between individuals expressing different FTO rs9939609 risk variants (all P ≥ 0.66). The cohort was physically active (4516 (3043) total MET min·week−1), although homozygous risk allele carriers (AA) displayed higher TFEQ cognitive restraint compared with nonrisk allele carriers (TT) (ES = 0.33 and P=0.03). In conclusion, obesity-related parameters were not different in physically active individuals expressing different risk variants of FTO rs9939609, although homozygous risk allele carriers exhibited higher cognitive restraint.


Introduction
Obesity is a major risk factor for several chronic diseases and represents a major health and economic burden on society [1]. e aetiology of obesity is multifactorial and is in uenced by complex interactions between environmental, lifestyle, and genetic factors [2]. Consequently, it is important to understand the interplay between these factors when designing strategies targeting the prevention of obesity.
e fat mass and obesity-associated gene (FTO) was the rst common variant identi ed by genome-wide association studies that in uences obesity risk [3]. Single nucleotide polymorphisms (SNPs) in intron 1 of FTO have been associated consistently with obesity risk across di erent ages and populations [4][5][6][7][8][9]. At FTO rs9939609 SNP, individuals homozygous for the risk allele (AA) weigh 3 kg more and have a 1.7-fold higher risk of obesity than those who do not carry a risk allele (TT) [3]. Studies have examined the e ect of FTO variants on regulators of energy homeostasis to elucidate the mechanisms in uencing FTO-mediated obesity risk. In this respect, evidence suggests that FTO may play a central role in the regulation of food intake [10,11]. is is supported by studies demonstrating that individuals homozygous for the risk allele exhibit reduced satiety, poor food choices, and increased energy consumption [12][13][14]. Conversely, FTO obesity SNPs have not been related to energy expenditure, with evidence suggesting that those carrying the risk allele do not show reduced basal metabolic rate [15] or physical activity levels [7,16,17].
Although the association between FTO and obesity risk is well established, lifestyle may modulate obesity risk related to FTO. Several studies have suggested that FTO-mediated body fatness may be attenuated in physically active individuals [18][19][20]. Indeed, a meta-analysis concluded that higher physical activity levels attenuate the in uence of FTO variation on obesity risk by 30% [21], and exercise interventions have demonstrated e cacy in promoting weight loss in FTO risk allele carriers [22,23]. However, it is not well understood how the body mass index-(BMI-) increasing in uence of FTO is attenuated in physically active individuals. An improved understanding of the di erences in dietary and physical activity patterns in variants of FTO rs9939609 SNP within a physically active cohort may provide a greater insight into the behaviours that o set FTOmediated obesity. erefore, the aim of this study was to examine physical activity and dietary habits in a sample of physically active men and women carrying di erent risk variants of FTO rs9939609 SNP. and were predominantly based in European countries or the United States. Full informed consent, via an online consent form, was obtained from participants before the study commenced. Participants were volunteers from a customer base that were known to be participating in a regular tness program but were not involved in elite sports. Participants classi ed themselves into one of fteen ethnic groups. To reduce variability in the cohort, individuals who identi ed themselves as non-mixed white descent (British, Irish, and other White background) were included in this study (see Supplementary Methods (available here) for ethnic background selection criteria). Of the 708 recruited, 1 had missing genotype data, 17 had missing demographic or anthropometric data, and 162 did not classify themselves as non-mixed white descent. erefore, results are presented for 528 participants (421 males and 107 females). Participant characteristics are presented in Table 1.

Genotype Analysis.
Participant DNA was obtained from saliva, which was collected via an Oragene DNA selfcollection kit (DNA Genotek Inc., Ottawa, ON, Canada) Values for age, anthropometric characteristics, physical activity levels, and eating behaviour are mean (SD) and were analysed between sex using linear mixed models. Values for central obesity and FTO rs99396909 genotype are frequency (%) and were analysed between sex using chi-square tests. a Central obesity was de ned as a waist circumference >88 cm for women and >102 cm for men. b Complete physical activity data available for n � 408 (female: n � 83; male: n � 325). * Signi cant di erence between females and males (linear mixed model P < 0.05). * * Signi cant di erence between females and males (chi-square test P < 0.05). sent from and returned to FitnessGenes Ltd. by post. DNA was extracted by LGC Genomics (Hertfordshire, UK) and genotyped using a KASP ™ genotyping assay for FTO rs9939609 SNP. Genotype frequency of FTO rs9939609 SNP was assessed using a goodness-of-t chi-square test and did not deviate from Hardy-Weinberg equilibrium (AA � 95, AT � 236, TT �197; P � 0.10).

Collection of Demographic and Anthropometric
Data. Participants self-reported their sex, age, stature, body mass, waist circumference, country of birth, and ethnicity using an online questionnaire. Sex, age, and ethnicity were con rmed by cross-referencing against customer information held by FitnessGenes Ltd. Participant stature, body mass, and waist circumference were used to calculate BMI and waist-to-height ratio ( Table 1). Central obesity was calculated as the percentage of participants that exceeded previously de ned thresholds for waist circumference (females > 88 cm and males > 102 cm) [24].

Physical Activity Levels.
Physical activity levels were measured using the validated short format of the International Physical Activity Questionnaire (IPAQ) [25]. e IPAQ assesses the frequency and duration of walking, and moderate-and vigorous-intensity physical activities performed in bouts lasting 10 minutes or more during the previous seven days. Total physical activity (MET (metabolic equivalent), minutes per week) was estimated by multiplying the number of minutes reported in each activity level by the speci c MET score for that activity (walking: 3.3, moderate intensity: 4.0, and vigorous intensity: 8.0 METs), and participants were classi ed in one of the three physical activity levels: low, moderate, or high (http://www.ipaq.ki.se).

Statistical Analyses.
Data were analysed using the IBM SPSS Statistics Software for Windows Version 21 (IBM, New York). Between-sex di erences in participant characteristics were examined using linear mixed models with one xed factor (sex). Linear mixed models, adjusted for age and sex, were used to examine between-genotype di erences in obesity-related parameters, physical activity levels, and eating behaviour with one xed factor (FTO rs99396909 genotype). Between-genotype di erences in all outcome measures were analysed using the additive genotype model (AA versus AT versus TT). Statistical power for this model was estimated using the CaTS Genome Power Calculator methods [27].
is method calculates the probability of outcomes per genotype, and thus the statistical power of the study based on the disease allele frequencies and prevalence of the disease within a population (i.e., the probability that a randomly sampled individual is obese or has a higher BMI) [27]. e calculation was performed using the genotype relative risk (GRR) for the additive genotype model (GRR � f 1 /f 0, where f 0 and f 1 are the probabilities of being a ected for individuals with 0 or 1 risk allele, resp.). For this study, a power of over 0.97 for a one-stage study was reported with a signi cance of 0.005. Where signi cant main e ects were found in the additive genotype model, post hoc analysis was performed using Holm-Bonferroni correction for multiple comparisons [28]. Exploratory analysis was also conducted to analyse potential between-genotype di erences in all outcome measures using the dominant model (risk allele carriers (AA/AT) versus homozygous nonrisk genotype (TT)) and the recessive model (nonrisk allele carriers (AT/TT) versus homozygous risk genotype (AA)). Di erences in categorical variables (central obesity and FTO genotype frequency) between sex and/or FTO genotype groups were analysed using chi-square tests. E ect sizes are used to supplement important ndings. An e ect size of 0.2 was considered the minimum important di erence in all outcome measures, 0.5 was moderate, and 0.8 was large [29]. Continuous variables are presented as mean (SD) and categorical variables as frequency (%). Statistical signi cance was accepted as P < 0.05.

Participant Characteristics.
On average, participants reported engaging in a total of 4516 (3043) MET minutes of activity per week, and the majority of participants were classi ed in the high physical activity category of the IPAQ (high 75.2%, moderate 19.4%, and low 5.4%), con rming the cohort were physically active (Table 1). Moreover, participants reported engaging in, on average, 3366 METmin·week −1 of moderate-to vigorous-intensity physical activity (Table 1), which is considerably higher than the average levels of moderate-to vigorous-intensity physical activity reported in adults across European countries (range 45-960 METmin·week −1 ) [30]. Participants reported the following reasons for engaging in physical activity: muscle building training n � 276 (52%); fat loss training n � 168 (32%); strength training n � 40 (8%); power training n � 20 (4%); endurance training n � 17 (3%); and no response n � 7 (1%).
Females were signi cantly older (ES � 0.24, P � 0.02), had a higher frequency of central obesity (odds ratio � 6.31, P < 0.001), and exhibited higher TFEQ disinhibition scores (ES � 0.42, P < 0.001) compared with males (Table 1). Body mass, BMI, and waist circumference were signi cantly lower in females than males (all ES ≥ 0.38, P ≤ 0.001), but waist-toheight ratio was similar between the sexes (ES � 0.09, P � 0.40) ( Table 1). No signi cant di erences were seen between the sexes for any measure of physical activity (all ES ≤ 0.13, P ≥ 0.28) or FTO genotype frequency (P � 0.12) ( Table 1).  (1st and 3rd quartiles). e whiskers represent the minimum and maximum observations. Linear mixed models, adjusted for age and sex, were used to examine between-genotype di erences. No signi cant di erences were seen in obesity-related parameters between FTO rs9939609 genotypes. 4 Journal of Obesity

FTO rs9939609 Genotype and Obesity-Related
Parameters. Obesity-related parameters in men and women carrying di erent risk variants of FTO rs9939609 SNP are displayed in Figure 1 and Table 2. Linear mixed models for the additive genotype model revealed no signi cant di erences in body mass, BMI, waist circumference, or waist-to-height ratio across FTO rs9939609 genotype groups (all P ≥ 0.66; Table 2). e minor allele frequency is analogous with that expected (0.4, A), suggesting this is not due to an unrepresentative population. Exploratory analysis revealed no signi cant di erences in obesity-related parameters when analysed in dominant (all ES ≤ 0.13, P ≥ 0.37) or recessive (all ES ≤ 0.08, P ≥ 0.68) allele models (Supplementary Table 1).

FTO rs9939609 Genotype and Physical Activity
Levels. Complete physical activity data were available for 408 participants (AA n � 74, AT n � 177, TT n � 157; female: n � 83 and male: n � 325). e physical activity levels in men and women carrying di erent risk variants of FTO rs9939609 SNP are displayed in Table 3. Linear mixed models for the additive genotype model revealed no signi cant di erences in physical activity levels across FTO rs9939609 genotype groups (all P ≥ 0.10). In the exploratory analysis, the dominant allele model revealed total MET minutes per week (ES � 0.22, P � 0.03, Supplementary

FTO rs9939609 Genotype and Eating Behaviours.
Eating behaviours in men and women carrying di erent risk variants of FTO rs9939609 SNP are displayed in Table 4. Linear   mixed models for the additive genotype model identi ed a signi cant main e ect for cognitive restraint score across FTO genotype groups (P � 0.03). Post hoc analysis of between-group di erences revealed that the cognitive restraint score was higher in homozygous A allele carriers than AT (ES � 0.25, P � 0.07) and TT (ES � 0.33, P � 0.03) genotypes; AT and TT genotypes were similar (ES � 0.06, P � 0.49). No di erences in disinhibition or hunger scores were seen across FTO genotype groups (both P ≥ 0.38).
In the exploratory analysis, the dominant allele model (AA/AT versus TT) revealed no signi cant di erences in cognitive restraint, disinhibition, or hunger scores between FTO genotype groups (all ES ≤ 0.14, P ≥ 0.12). e recessive allele model was consistent with that seen in the additive genotype model, in that the cognitive restraint score was higher in homozygous A allele carriers (AA) than T allele carriers (AT/TT) (ES � 0.28, P � 0.01; Supplementary

Discussion
e primary nding from the present study was that obesityrelated parameters were not di erent in physically active individuals carrying di erent risk variants of FTO rs9939609 SNP. Furthermore, FTO rs9939609 homozygous A allele carriers exhibited higher cognitive restraint than nonrisk allele carriers, and exploratory analysis also showed higher levels of physical activity in A allele carriers compared with nonrisk allele carriers. e consensus of evidence suggests that FTO risk alleles are associated with elevated body weight across di erent ages and populations, with each minor risk allele increasing BMI and obesity risk by 0.25-0.39 kg·m −2 and 1.18-1.27 fold, respectively [9]. However, the most extensively studied rs9939609 SNP had no in uence on obesity-related parameters in the current study. e participants in our cohort were physically active, and the average levels of reported total and moderate-to vigorous-intensity physical activity were considerably higher in comparison with both the general population [30] and individuals in previous FTO studies [19][20][21][22][23]. Several studies have demonstrated, through self-reported questionnaires [31,32] and objective physical activity devices [19,33], that obesity-related traits associated with FTO risk alleles are attenuated in individuals with higher physical activity levels. erefore, it is possible that the high physical activity levels within our cohort may have diminished any di erences in obesity-related parameters between FTO rs9939609 genotypes. Intriguingly, despite being a highly active group when taken as a whole, the exploratory analysis using the dominant model demonstrated that individuals with the A allele reported higher total and vigorous physical activity levels compared with TT individuals. It is possible, therefore, that this di erence in physical activity patterns between FTO genotypes may further o set the adiposity-increasing e ect of FTO within this cohort, but additional work is required to examine this systematically. Nonetheless, this elevation in activity was small, and these di erences in physical activity levels between FTO genotypes con ict with previous studies suggesting that FTO genotype does not impact on physical activity levels [7,31]. Contradictions between the current study and other evidence may be due to the high levels of intensive physical activity of individuals within the current study. Although Berentzen et al. [16] reported no e ect of the FTO variant on physical activity in individuals classi ed by the authors as highly active (more than 4 h moderate physical activity a week), there is a lack of studies that have examined the e ect of FTO genotype on activity levels in cohorts with even greater levels of physical activity. Consequently, further studies, including longitudinal investigations, are required to examine FTO-mediated di erences in physical activity among individuals with varying activity status. In addition, given that much of the data on FTO-mediated di erences in physical activity levels are reliant on self-reported questionnaire data, additional scienti c enquiry using objective physical activity measures, such as accelerometers, is required.
It has been postulated that FTO-mediated predisposition to weight gain and obesity may be modi ed by dietary behaviours. Research suggests that FTO may play a signicant role in the regulation of satiety and food intake [10,11]. Furthermore, it has been reported in previous observational studies that the risk variant of FTO is associated with lower cognitive restraint and higher disinhibition and hunger, which may be indicative of poorer eating behaviours [34,35]. e present study extends these ndings by demonstrating that physically active individuals homozygous for FTO rs9939609 risk allele display higher cognitive restraint but similar disinhibition and hunger scores compared with nonrisk allele carriers. e potential bene t of cognitive restraint on the propensity to body weight changes is highlighted by studies demonstrating a negative association between cognitive restraint and markers of adiposity (e.g., body mass and BMI) [36,37]. However, such associations are not reported universally [38], and cognitive restraint tends to be higher in overweight than in healthy weight individuals [39]. Additionally, disinhibition has been proposed as a stronger predictor of BMI [40]. Nevertheless, increased dietary restraint during weight loss has been identi ed as a signi cant predictor of successful weight loss maintenance [41]. erefore, it is possible that the higher cognitive restraint observed in FTO risk allele carriers in the present study may o set FTO-mediated obesity risk in physically active individuals; but further work is required to con rm this chronically. e current study is not without limitations. e observations in the present study are limited by the crosssectional design, which precludes the ability to establish causality. Future longitudinal studies are required to establish systematically whether higher physical activity levels and cognitive restraint may diminish FTO-mediated predisposition to weight gain. In addition, the anthropometric, physical activity, and eating behaviour data were selfmeasured and self-reported using an online survey, which may have led to problems with participant measurement error, recall, and bias. Additional work is therefore required using more direct and objective measures. Furthermore, one of the primary measures of body fatness in the present study is BMI. Despite its frequent use in large-scale population studies, BMI does not take into account di erences between fat and fat-free mass such as muscle and bone [42]. Given the current cohort reported high levels of total activity, with the majority indicating that their primary tness goal was muscle building (52%), it is possible that variations in body mass (and hence BMI) could be largely due to higher muscle mass. is may explain why the average BMI among individuals in the current data set falls within the overweight category. Consequently, future studies using more direct measures of body adiposity are required to replicate the current study's ndings. Finally, statistical power was limited in the dominant and recessive models analysed in this paper. However, the results from these exploratory models allow for preliminary inferences that may guide a larger population-based study.

Conclusion
Within a physically active cohort, risk allele carriers of FTO rs9939609 exhibited greater physical activity levels and cognitive restraint than nonrisk allele carriers despite demonstrating similar adiposity-related measures. Future studies with repeated and objective measurements are required to further investigate physical activity and dietary behaviours that underscore the e ects of FTO on obesity risk.

Conflicts of Interest
Nathan R. West, Nicola C. Hanson, Samantha E. Decombel, and Stuart J. Grice work for FitnessGenes Ltd.