Introduction

Interrelated multifactorial diseases like obesity, cardiovascular diseases, metabolic syndrome, type 2 diabetes and cancer have emerged as major diseases globally with severe effects on health care costs. In order to better understand these diseases, genome-wide association studies were started to unravel these complex diseases by identifying candidate genes. In obesity, common variants in the Fat mass and obesity associated (FTO) gene were found to have the highest effect on body weight [1–3]. Homozygote risk-allele carriers (AA) of the FTO single nucleotide polymorphism (SNP) rs9939609 showed a 1.67-fold increased risk to develop obesity in comparison to non-risk-allele carriers (TT) [2] which could be reproduced for several FTO SNPs in different ethnic populations [4–6]. Additionally, murine loss and gain of function analyses confirmed FTO’s role in body weight regulation [7, 8]. Nevertheless, FTO’s cellular function as a N6-methyladenosine RNA demethylase [9] remains poorly understood. Despite its ubiquitous expression in different organs with the greatest expression in the hypothalamus, an area important for controlling energy metabolism and food intake [2, 10, 11], data on FTO expression in relation to experimentally altered nutritional states have been conflicting [10, 12, 13].

Since most GWAS can only detect a locus but not a gene the question arose whether FTO is “the sole” obesity gene. A potential candidate in close 5’ proximity to FTO would be RPGRIP1L. Initial studies identified a transcription factor, possibly regulating FTO and RPGRIP1L expression by binding to SNPs within FTO [14, 15]. However, several other studies could not confirm an association of FTO SNPs with expression of RPGRIP1L [16, 17] or altered Rpgrip1l expression in Fto-deficient mice [8]. An interaction of the obesity-associated intronic region of FTO and the promoter sequence of Iroqouis 3 (IRX3) has recently been identified [18]. Furthermore, FTO SNPs could be associated with expression of IRX3 in human brains, while Irx3-deficient mice showed a phenotype similar to Fto-deficient mice. These results suggest that IRX3 could be a functional long-range target of FTO SNPs. However, analyses in human brain tissues have been restricted to the cerebellum and have not focused on the hypothalamus. Thus, further studies are necessary to answer the question if either FTO or IRX3, or both are “the” obesity gene(s).

Obesity is a major risk factor for the development of cardiovascular diseases. For this very reason the question arose whether FTO genotype might bear a direct risk for cardiovascular diseases. First data on this have been inconsistent: showing either a BMI-independent [19, 20] or a BMI-dependent [21, 22] association of FTO with cardiovascular diseases. Furthermore, only few studies referred to a specific cardiovascular disease.

Among cardiovascular diseases aortic valve stenosis (AVS) is one of the most prominent in the Western world [23]. The probability of developing AVS increases with age and male gender [24]. In 2013 one third of German cardiac surgical procedures were on heart valves with 72% due to AVS alone [25]. Severity increases with symptomatic AVS leading to a death rate of more than 50% when immediate aortic valve replacement is not performed [26]. There are no medical therapies known to slow down or avert the progression of AVS with molecular mechanisms being poorly understood. Although familial clustering of individuals has been reported [27], little is known about genetic factors contributing to the development of AVS.

It remains unclear whether increased BMI leads to an association of FTO with cardiovascular diseases or if FTO has a direct impact. Because AVS is one of the most prominent cardiovascular diseases the aim of this study was to determine a possible association of FTO with AVS. A better understanding of how FTO and IRX3 impacts AVS is a first step towards a better treatment of AVS in the future.

Methods

Results

Baseline Data of AVS Cases and KORA Controls

Demographic and clinical data of AVS cases and KORA controls are summarized in Table 1. As expected AVS cases were older with increased prevalence for diabetes mellitus and hypertension (all were p<0.001). Nevertheless, AVS cases had significantly lower BMI and blood lipid levels (BMI and cholesterol levels p<0.001, triglycerides p = 0.005). Similar results were obtained by others when comparing baseline data of patients with acute coronary syndrome and controls representing a random population in this study [19]. Appearance of peripheral artery disease and smoking status did not differ between AVS cases and KORA controls (p = 0.511 and p = 0.381 respectively).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Demographic and Clinical Data of AVS Cases and KORA Controls.

https://doi.org/10.1371/journal.pone.0139419.t001

FTO rs9939609 Genotype Frequencies of AVS Cases and KORA Controls

In the KORA controls, the genotypic frequency of FTO rs9939609 did not deviate from HWE (p = 0.277, S2 Table), whereas a significant deviation was found in the AVS cases (p = 0.001).

The accuracy of genotyping was confirmed with two additional SNPs in linkage disequilibrium with FTO rs9939609 (S3 Table). The genotypic distribution of FTO rs9939609 differed significantly between AVS cases and KORA controls (p = 0.004; OR 0.614, 95% CI 0.433–0.868 for TA genotype; OR 1.042, 95% CI 0.672–1.614 for AA genotype; Table 2 and S4 Table). AVS cases showed a higher proportion of the high-risk genotype (AA) compared to KORA controls (21.7% versus 16.8%, Table 2). However, this difference was not significant in the recessive model (p = 0.102, Table 2). The TA genotype was associated with a lower risk for developing AVS and significantly fewer TA individuals were found in the AVS population than in the KORA control population in the overdominant model (39.3% versus 51.7%, p = 0.001, OR 0.605, 95% CI 0.443–0.824, Table 2 and S4 Table). In a logistic regression controlling for age, BMI, diabetes mellitus and hypertension the association of TA with lower risk of developing AVS remained (p = 0.030, OR 0.675, 95% CI 0.472–0.962, Table 2 and S4 Table), whereas significance in the co-dominant model was lost (p = 0.092, Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. FTO rs9939609 Genotype Frequencies in AVS Cases and KORA Controls.

https://doi.org/10.1371/journal.pone.0139419.t002

FTO rs9939609 Genotype Frequencies of Male and Female AVS Cases and KORA Controls

FTO rs9939609 genotype frequencies of males and females were analyzed separately due to known gender differences for developing cardiovascular diseases, especially for AVS. A significant deviation from HWE in genotypic frequencies of FTO rs9939609 persisted in male AVS cases, but was not observed in female AVS cases (p = 0.001 and p = 0.516 respectively, S2 Table).

The genotypic frequency of FTO rs9939609 differed significantly in males (p = 0.005; OR 0.631, 95% CI 0.407–0.976 for TA genotype; OR 1.327, 95% CI 0.784–2.250 for AA genotype; not adjusted for age, BMI, diabetes mellitus and hypertension; Tables 2 and 3) likewise after adjustment for these factors (p = 0.030; OR 0.578, 95% CI 0.357–0.930 for TA genotype; OR 1.058, 95% CI 0.598–1.867 for AA genotype), but not in females (p = 0.170, Table 2). The percentage of the AA genotype in male AVS cases was elevated compared to KORA controls in the recessive model (26.2% versus 17.0%, p = 0.017, OR 1.727, 95% CI 1.087–2.747, Tables 2 and 3). However, results were non-significant after adjustment for co-variables (p = 0.157, OR 1.438, 95% CI 0.867–2.379). Evidence of heterozygote advantage persisted in male AVS cases in the adjusted overdominant model (37.9% versus 52.0%, p = 0.008, OR 0.565, 95% CI 0.369–0.861).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Odds Ratios (OR) and 95% Confidence Intervals (CI) of FTO rs9939609 Between Male AVS Cases and KORA Controls.

https://doi.org/10.1371/journal.pone.0139419.t003

Discussion

With this case-control study of participants residing in Germany, we demonstrate for the first time an association between FTO rs9939609 and AVS, exclusively restricted to men. Strong association of FTO with AVS was found in the overdominant model, with lower risk of developing AVS in TA heterozygotes. This effect was significantly higher than the association in AA carriers independent of age, BMI, diabetes mellitus and hypertension.

With three different approaches we minimized possibilities for false positive and negative results of an association of FTO with AVS. First, KORA controls were in HWE. Second, the power of the study was sufficient. Third, genotyping mistakes have been excluded via genotyping with two additional SNPs in linkage disequilibrium with FTO rs9939609, namely rs8050136 and rs17817449 [1, 2]. Furthermore, multiple association studies used the test of HWE to exclude possible error rates of SNP genotyping for the cases. In contrast Lee proposed screening the genome for disease-susceptibility genes by testing deviation from HWE in a gene bank of any disease [37] as supported by others [34–36]. Therefore we argue that association of FTO with AVS is evident by significant deviation from HWE in AVS cases.

Of all AVS study participants, 65% were male, which reflects the gender-specific proportions observed in the largest German national database for surgical and interventional procedures on the aortic valve and underlines the higher risk of AVS for men in comparison to women [38]. Surprisingly, the association of FTO with AVS was restricted to men. Such gender differences in FTO association studies have been reported previously. For example, variation at the SNP rs9939609 was associated with BMI in females, but not males, affecting children, adolescents and adults of different ethnic origin [39–42]. A study on FTO rs17817449 revealed an interaction between SNP and physical activity on BMI in males but not females [43]. Differences in hormone levels, fat mass storage or physical activity could contribute to the differences in association studies for FTO. Due to these results we propose to identify gender differences by first analyzing both genders together and then separately in association studies for FTO. With this approach missing a significant effect of FTO due to analyses of non-affected gender by chance could be avoided. These considerations could further explain previous conflicting results on an association of FTO with cardiovascular diseases in which separate gender analyses were intentionally not performed [19, 21, 44]. Nevertheless, we were able to reveal a gender specific role of FTO in the context of therapeutically relevant AVS.

An association of FTO with AVS resulted in a reduced risk for developing AVS in men assuming the overdominant model. To our knowledge, this finding has not been discussed in any other genetic association study of FTO with cardiovascular diseases. The most prominent example of heterozygote advantage is a protection from severe symptoms of malaria in heterozygote carriers of sickle-cell anemia whereas carriers of the homozygote risk-allele suffer from severe symptoms due to sickle-cell disease [45]. So far, three other FTO association studies have found an effect of heterosis. One study analyzed the effect of FTO genotype with fat-related traits in an experimental pig population [46]. Fat weight, fat depth and fat area of different fat storages were associated with the SNP in heterozygotes. In a second study, rs9939609 was associated with BMI in European girls but not boys, whereas heterozygote boys profited from lower BMI and other obesity-related traits [39]. The third study identified an association of rs9939609 with cardiovascular diseases in males [20]. Taking a closer look at the results reveals a positive effect of heterosis with higher significance in different adjusted models than the association of the homozygote risk-allele genotype, although this was not discussed in the publication. The importance of the specific type of cardiovascular disease analyzed for an association with FTO should not be underestimated. According to the study [20], it was unclear whether study participants suffered from cardiovascular diseases similar in their etiology to AVS or whether they specifically suffered from AVS. Moreover, the proportion of the respective subgroups were not reported.

Whether TA carriers are at lower risk for AVS due to a broader range of FTO expression in different cells or tissues in comparison to homozygotes remains unclear. One possible explanation could be a deleterious effect of too little or too much FTO expression within the heart or aortic valve leaflets. So far, differences of FTO expression, dependent on the risk-allele genotype, were found in unspliced nuclear RNA of human fibroblasts [16]. In these studies, FTO expression for the risk-allele was elevated compared to the non-risk-allele; whether FTO SNPs are directly connected to FTO expression and if this is the case in aortic valve leaflets remains unknown. A third explanation for heterozygote advantage could be even the involvement of an additional factor [47, 48] or an interaction of the IRX3 promoter with the obesity-associated SNP region of FTO [18]. In this study, array expression data revealed that IRX3 was highly expressed within the heart at levels approximately 8-fold higher than the median expression across several human tissues. In comparison IRX3 expression within the cerebellum was approximately 50% lower than the median expression. Thus, a role of IRX3 in altering the development of AVS should be considered for future molecular analyses. In terms of function, IRX3 has been shown to be a direct repressor of transcription of Connexin 43 (CX43) in human hearts [49]. Based on these findings, an impact of FTO SNPs on CX43 may be suggested which goes beyond the altered expression of CX43 found in left ventricular hypertrophy in AVS patients [50]. Furthermore, gender specific differences in expression of IRX3 and CX43 have been described [51]. Thus, we suggest that IRX3 as a target of FTO SNPs may lead to altered CX43 expression, influencing progression of AVS. Nevertheless, further experiments are needed to test these predictions.

Although, rs9939609 was associated with AVS for all genetic models applied in males (especially in the subpopulation of 50–70 year old AVS cases and KORA controls), after logistic regression, the AA genotype failed to be associated with a higher risk of developing AVS. Sensitivity analyses revealed an impact of diabetes mellitus on the risk of AA carriers to develop AVS, supporting the effect of diabetes on the progression of AVS as previously suggested [52–54]. However, heterozygous males profited from a reduced risk of developing AVS, despite the considerable number of other risk factors identified so far. Interestingly, heterozygote advantage in European boys, uncovered by Jacobsson et al., was not only restricted to the BMI. Male individuals of the TA genotype showed decreased fasting serum insulin levels and an increased insulin sensitivity, whereas the SNP rs9939609 was associated with higher plasma glucose levels in girls [39].

With different approaches we minimized the chance for false results of an association of FTO with AVS. Nevertheless, an important limitation of this study is the lack of a second case cohort. Thus, false positive and negative results cannot be completely eliminated. Furthermore, we used a random control population where we cannot exclude the appearance of asymptomatic AVS in a few KORA controls due to the lack of routine echocardiography. Up to 2.8% of KORA controls could suffer from undetected AVS [55]. However, the effect of heterozygote advantage with a reduced risk to develop AVS should be even more pronounced with “true” controls. To test this and to overcome the limitation of a missing second case cohort, additional clinical studies with large cohorts of different ethnic origin who underwent routine echocardiography are needed. Further, an extended evaluation of the metabolic status, e.g. by standardized oral glucose tolerance test, due to observations in European boys [39] would more clearly delineate the disease etiology. Finally, in this study the primary inclusion criteria was the presence of AVS at the time of admission to the hospital for cardiac surgery on the aortic valve. Hence, patients with bicuspid aortic valve disease (n = 24) and patients with rheumatic valvular disease (n = 2) have been included in this study. Despite the prospective nature of this study, reliable longitudinal data are missing, making a detailed analysis with a focus on these sub-cohorts difficult.

In summary, in our case-control study we showed a gender specific association of FTO with AVS independent of BMI, diabetes mellitus and other co-variables. Heterozygote risk-allele carriers profit from a reduced risk to develop AVS. So far, not much is known about the genetics causing severe symptomatic AVS leading to urgent aortic valve replacement. We suggest that “a bit” of the FTO risk-allele is important to keep the balance of a relatively healthy aortic valve presumably through altered expression of IRX3 within the heart. This study provides a first step towards an understanding of the genetics basis of AVS and the complexity of action of FTO variants on different diseases.

Supporting Information

Acknowledgments

We would like to thank Torsten Seehaus, Laura Rose, Brant E. Isakson and Nenja Krüger for giving advice in manuscript preparation.