The common FTOvariant rs9939609 is not associated with BMI in a longitudinal study on a cohort of Swedish men born 1920-1924

Today, about one in three adult can be classified as obese based on objectively measured weight and height. The increase started after the World War II, escalated in the seventies and the obesity rates have roughly tripled in the past 20 years. In Sweden one in hundred is considered morbidly obese today compared to one in thousand in the early 1970s [1–4].

This worldwide rapid increase in the prevalence of obesity is much due to environmental factors, such as a sedentary lifestyle and develops from an imbalance between energy ingested and expended [5]. It is however also widely accepted that obesity is under strong genetic control, explaining about 30% to 70% of the variation in BMI [6–8]. Recently, genome-wide association studies have led to rapid progress in our understanding of the genetic basis of various common diseases and a new candidate gene for obesity has been identified, the fat mass and obesity associated gene, FTO [9–14]. The first association between FTO and human obesity was found by Frayling et al. in early 2007 [10] in a genome wide association study of diabetes and has been positively replicated in additional studies involving several different ethnicities [15–21] including Swedish subjects [22, 23]. These studies all show that subjects who are homozygous for the risk allele weigh about 3 kg more compared to those without the allele and individuals with the risk allele have about 1.5-fold increased risk of having obesity.

Most of the previous studies on the FTO gene variants are cross-sectional, leaving the longitudinal pattern of the associations between obesity and age-specific genetic effects poorly studied. Furthermore, the genetic effects on lifestyle factors and the effect in populations not affected by an obesogenic environment have not been thoroughly investigated. There have also been additional studies indicating that the genetic effects on BMI may be dependent on physical activity [24–26]. Four studies have so far analyzed if the effect of FTO on obesity in adults may be changed upon physical activity [23, 27–29] with somewhat contrasting results. The aim of the present study was to address the issues of BMI and physical activity by assessing the genetic effects in a longitudinal cohort of adult men, the ULSAM cohort. The subjects were all born between 1920 and 1924, investigated at the age of 50 and reinvestigated at the ages of 60, 70, 77 and 82 years of age. The data available in Uppsala Longitudinal Study of Adult Men (ULSAM) allowed us to, compared to previous studies, investigate BMI differences at older ages and also highlight that the environment may play a fundamental role in the genetic contribution to increased BMI.

We further investigated the association between rs9939609 and BMI as well as other measurements of obesity such as waist circumference and skinfold thickness at baseline for all subjects as well as BMI and waist circumference at 60, 70, 77 and 82 years of age. No association at any age was found between the minor A-allele and BMI, or any of the other obesity measurements (Table 3, Figure 1).

The longitudinal pattern of the associations between obesity and age-specific genetic effects was analyzed in 32-years follow up. Although not significant, we observed a trend toward decreasing differences in BMI according to genotype at older ages, as was also observed by [36](Figure 1A). We also investigated the longitudinal changes in BMI from the age of 50 years to 82 years (Figure 1B). We observed an increase in BMI levels from 50-77 years, with the highest increase between 60-70 years, and a decrease in BMI from 77-82. However, neither the A or T allele at any age were associated with a higher increase or decrease in BMI which means that the variant does not seem to affect the changes in BMI levels at any stage in older life.

From a medical questionnaire we had access to self-reported leisure time physical activity at age 50 years which divided the subjects into four categories. 15% of all subjects were considered sedentary, 36% had moderate physical activity, 44% had regular physical activity and 5% were considered athletic, and as expected subjects that were considered athletic had lower BMI levels compared to the less active subjects (p < 0.01). The subjects, based on the physical activity, were further stratified according to the FTO rs9939609 genotype (Figure 2A). Physical activity influence the effect of FTO rs9939609 on BMI (P_interaction = 0.039) across the four different categories. The BMI per FTO risk allele was higher in sedentary active (0.33 per risk allele, SD 0.39) and moderate active individuals (0.13 per risk allele, SD 0.21) compared with the BMI among individuals with regular activity (-0.25 per risk allele, SD 0.18) and athletic individuals (-0.29 per risk allele, SD 0.47). The difference in BMI per allele in each category was however not significant different (P > 0.05 for all categories) (Figure 2B).

There has been a considerable increase in obesity from 6% in 1963 to 11% in 1993 in Sweden [2, 37]. This increase has continued and there has been a twofold increase during the last two decades [3, 4, 38]. The men included in the ULSAM cohort were all examined 1970-74 at the age of 50 years, when the prevalence of moderate obesity were very low, 0.8% compared to 3.8% in 2000-2005 [4]. Moreover, these men have lower BMI levels compared to the cohorts previously used to study the FTO gene. The mean BMI was 24.5 in the ULSAM cohort compared to the mean BMI of 26.5 in the cohorts of older adult reported by Frayling et al. [9]. The ULSAM cohort provides thus an important opportunity to study genetic components in a lean population that lived in a less obesogenic environment compared with today. Our results indicate that the impact of FTO seems less prominent in this lean population since there was no significant difference in BMI for the carriers of the risk allele, not at baseline or at any of the other ages that we studied. Neither did we find any association for other measurements of obesity such as waist circumference and skinfold thickness at any age. It is however important that these findings are replicated in several lean populations in order to be able to draw any conclusions.

We further analyzed the longitudinal pattern of the associations between BMI and FTO variants. So far, Frayling et al. [9] and Qi et al. [36] have observed a trend toward decreasing associations between FTO variants and BMI at older age. Qi et al. further showed that this was found in men, whereas the associations were constant across different age groups in women [36]. Although we did not reach significance, our data support these findings. The small effect of rs9939609 on BMI in our cohort at baseline was not seen at all at older ages. Thus, it appears that the association between the SNP and obesity risk declined with older age especially in men.

Studies on BMI changes in relation to physical activity and genetic factors have shown that both play a significant independent role in weight changes. However, there has been additional evidence of a gene-environment interaction in men, indicating that genetic effects on weight gain may be more highly dependent on physical activity level [24–26]. Four studies have so far analyzed if the effect of FTO on obesity in adults may be changed upon physical activity [27, 28, 23, 29] with somewhat contrasting results. Andreasen et al. [27] demonstrated that the impact of the FTO rs9939609 genotype was influenced by the habitual level of physical activity in a population based study on Danish subjects with mean BMI of 26 kg/m² . These subjects were divided according to self-reported physically activity into passive, light or medium physically active and hard or very hard physically active, and stratified according to FTO genotype. The study showed that physical inactive homozygous FTO rs9939609 A-allele carriers had around 2 kg/m² higher BMI compared to inactive homozygous T-allele carriers [27]. These findings were supported by Cauchi et al [29] and Rampersaud et al. [28] for other FTO variants. Both showed that the differences in BMI were larger in physical inactive subjects while no difference was found at all in physical active subjects. However, a large study by Jonsson et al. [23] on Swedish adults did not find any interaction between rs9939609 and physical activity on BMI. The biggest differences between these studies is that the Swedish subjects had a mean BMI of 24.3 kg/m² and were leaner compared to the other cohorts and physical activity was only based on two categories, active or not active. Our results showed the presence of an interaction and support the suggestion that the increased levels of BMI related to FTO variants may be less prominent if the persons are physical active. Our interaction is however weaker compared to the other studies and the effect size is smaller. The lower effect size in our study compared to previous studies may be due to a leaner cohort. We have furthermore older subjects compared to other studies which may affect the interaction and weaken the effect of physical activity. Furthermore, since the categories of physical activity in our study have been assessed by questionnaire under- or over-reporting may occur which may also influence the result. Our finding is however consistent with previous studies and suggests that inherited factors significantly influence body weight in sedentary subjects and that in physically more active individuals the effect of the FTO gene seem to diminished, further suggesting that high physical activity would be particularly beneficial in those genetically predisposed to obesity due to the FTO gene. As the gene expression might vary due to differences in the environment it is interesting to study the effect of FTO in a population with lower mean BMI than has previously been studied. Furthermore, several of the susceptibility genes to common diseases do not have a primary causal role in the predisposition to disease without considering environmental factors. Rather, susceptibility genes may act as responders, or modifiers, to factors such as stress, environment, as well as physical inactivity. Fisher et al. [39] has also demonstrated that loss of the mouse Fto gene product reduces adiposity, and that even moderate reduction in Fto expression protects from diet-induced obesity which makes it tempting to speculate that physical activity may reduce the FTO gene expression and influence the interacting genes that in turn modify energy homeostasis.

Here we set out to examine the association of the FTO variant rs9939609 with BMI in a longitudinal cohort of adult men. We could not find any association between FTO variant and obesity or obesity measurements at any of the ages studied. We further analyzed the effect of physical activity on BMI levels and found an interaction between rs9939609 and level of activity and saw that inactive people carriers of the A-allele had higher BMI-values than non-carriers. This, together with our lack of association with BMI in our lean population, indicate that FTO seem to have a greater effect on those that are less physical active and with higher BMI. Thus, these results are an example of environmental influences and that inactivity is a potential vital environmental trigger for disease susceptibility dependent on the FTO genotype.