Uncovering the biology of FTO

Abstract

Genome-wide association studies have revealed that SNPs in the first intron of FTO (Fat mass and Obesity related) are robustly associated with body mass index and obesity. Subsequently, it has become clear that this association with body weight, and increasingly food intake, is replicable across multiple populations and different age groups. However, to date, no conclusive link has been made between the risk alleles and FTO expression or its physiological role. FTO deficiency leads to a complex phenotype including postnatal mortality and growth retardation, pointing to some fundamental developmental role. Yet, the weight of evidence from a number of animal models where FTO expression has been perturbed indicates some role for FTO in energy homoeostasis. In addition, emerging data points to a role for FTO in the sensing of nutrients. In this review, we explore the in vivo and in vitro evidence detailing FTO's different faces and discuss how these might link to the regulation of body weight.

Introduction

Obesity is a severe problem facing most developed nations and many other emerging economies [1], with the World Health Organization estimating that there will be more than 2 billion overweight and 700 million obese adults in the world by 2015. It is an inescapable fact that the underlying cause of obesity is a result of consuming more energy than you burn, and the rapid changes in lifestyle and food availability over the past 30 years have undeniably driven this rise in obesity. The question that is more complex to answer however, is why some people eat more than others. Individuals respond differently to these ‘obesigenic’ environmental changes and it is increasingly clear that this variation in response has a very potent genetic element. Indeed, studies of body mass index (BMI) correlations of monozygotic, dizygotic, biological and adopted siblings reveal the heritability of fat mass to be between 30% and 70% [2–4]. Consequently, genetic approaches offer a powerful tool for characterising the molecular and physiological mechanisms of body weight control and allow us to understand how these may become ineffective in the obese state.

Over the past 15 years, insights from human and mouse genetics have illuminated multiple pathways within the brain that play a key role in the control of food intake (reviewed by [5]). We now know for example, that the hypothalamic leptin–melanocortin signalling pathway is central to the control of mammalian food intake, with genetic disruption of most components of the pathway resulting in severe obesity in both mouse and man [5]. However, these and other monogenic syndromes of severe obesity remain very rare, with the major burden of disease carried by those with “common obesity”.

Recent progress into the underpinnings of common obesity has been made with genome-wide association studies (GWAS), with 2007 marking the publication of the first obesity GWAS studies [6–8]. In the first study, SNPs in the first intron of FTO (fat mass and obesity related) were reported to be associated with BMI and obesity [7]. In the intervening ‘post GWAS years’, a number of meta-analyses following on from the FTO discovery, covering multiple GWAS and encompassing hundreds of thousands of individuals, have revealed a further 31 ‘polygenic’ obesity loci [9,10]. FTO however, remains the gene with the most robust association and greatest effect size. Yet, despite the significant amount of scientific man-hours invested over the past few years and steady progress being made in determining its biochemical function, understanding the molecular mechanism underlying FTO's association with obesity remains elusive. Here, we review the latest data on the increasingly complex biology of FTO and discuss how these could link to the regulation of body weight .

Discovery of FTO

The study of the genetic basis of complex traits and disease involves the consideration of multiple genetic and environmental factors, and critically, how these interact with each other. However, the identification of bona fide, replicable common genetic variants for a number of diseases such as type II diabetes and obesity proved problematic for many years. In contrast to diseases with a clear Mendelian pattern of inheritance, a major hurdle in studying common complex diseases is that they are likely to have a ‘polygenic’ aetiology, with multiple variations each having a subtle effect. Because the effect sizes of polygenic variants are very often small, early studies were woefully under-powered, both in terms of size of patient cohorts and the number of single nucleotide polymorphisms (SNPs) studied. The advent of technology required to interrogate hundreds of thousands (now more than a million) SNPs in parallel however, ushered in the GWAS era.

In 2007, a type 2 diabetes (T2D) GWAS identified multiple single nucleotide polymorphisms (SNPs) in the first intron of FTO associated with disease [7]. However the association between these FTO SNPs and T2D disappeared after adjustment for BMI. With increased weight being a risk factor for T2D, it suggested that these SNPs are actually associated with BMI [7]. Two subsequent reports, one a French study [6] and the second a Sardinian study [8], closely followed, confirming a robust association between FTO SNPs and body weight . Many other studies have now examined and confirmed the influence of FTO variants on BMI across multiple populations and age groups (reviewed by [11]).

When the components of energy balance are studied in more detail to tease out the mechanisms underlying the impact of these common genetic variants on human adiposity, it is clear that there is a major effect on increase in energy intake [12–17], with reduction in satiety [18–20] and possibly with increased daily fat intake [16,21]. In contrast, the evidence for involvement of mechanisms involving energy expenditure in FTO's effect on energy balance, either as altered basal energy utilisation [13,15,22–24] or physical activity [22,25–27] is, to date, less compelling.

From ‘Fatso’ to FTO

Fto was initially identified in 1999, prior to the GWAS era, as one of six contiguous genes encompassed by a naturally occurring 1.6 megabase deletion in a mouse model known as fused toes (Ft) [28]. The homozygous Ft mice were embryonic lethal with severe developmental malformation of the CNS, characterised by growth retardation and the absence of left–right symmetry [29]. Heterozygous mice show polydactyly (thus fused toes) on the fore limbs which is attributed to impaired programmed cell death [29,30]. In addition to Fto, the other five genes are Irx3, Irx5, Irx6, Fts and Ftm (now known as Rpgrip1l). Incidentally, before any link with obesity was established, Fto was originally a contraction of ‘Fatso’, a name so coined by scientists studying the fused toe deletion because it happened to be the largest of the six deleted genes. Soon after it was associated with a body weight phenotype however, the unintentionally and unfortunately named ‘Fatso’ rapidly transitioned to the less offensive and far more acceptable ‘FaT mass and Obesity related transcript’.

Biochemical role for FTO

Soon after the first report of FTO, we and others used a bioinformatics approach to show that human FTO has the highest sequence similarity with the E. coli DNA repair protein AlkB and its mammalian homologues ABH2 and 3, which all belong to the family of Fe²⁺ and 2-oxoglutarate (2-OG) dependent dioxygenases [31,32]. Members of the family are involved in various cellular processes including DNA repair, fatty acid metabolism and posttranslational modifications [33]. We and others have reported that in vitro, FTO is able to catalyse the demethylation of 3-methyl Thymidine and 3-methyl Uridine (3meU) in single-stranded DNA and RNA respectively, in the presence of Fe²⁺ and 2-OG [31,34]. X-ray crystallography of FTO reveals a protein composed of 2 domains: an N-terminal domain carrying a catalytic core and a C-terminal domain of unknown structural homology [35]. In the catalytic domain are five obligate amino acid residues found in all members of this enzyme superfamily; two residues, a histidine (H) and an aspartic acid (D), required for binding Fe²⁺; and three residues, an H and two arginines (R) separated by six amino acids, required for 2-OG binding [36]. Structural based sequence alignment showed that FTO has an extra region, referred to as Loop 1, which is highly conserved among FTOs from different species, but is absent from AlkB and ABH proteins. The crystal structure reveals that Loop 1 sterically hinders the unmethylated strand of DNA from gaining access to the substrate-binding site, explaining why FTO demonstrates no appreciable activity on double stranded nucleic acids, and suggesting that it has a substrate preference for methylated RNA over DNA [35].

Subsequently, Jia et al. [37] reported that N6 methyl Adenosine (6meA) in both DNA and RNA is yet another substrate of FTO, but with 50-fold greater affinity than 3meU. 3meU is found largely in ribosomal RNA (rRNA) [38], while 6meA is the most common modified nucleoside found in mRNA [39]. However, because rRNA accounts for the vast majority of total RNA, in absolute terms there is actually a hundred fold more 3meU than 6meA in any given cell. It is still not entirely clear if one or both of these modified bases are the endogenous substrate/s for FTO.

Two independent studies have just been published describing the human and mouse 6meA modification landscape in a transcriptome-wide manner [40,41]. Both groups used antibodies against 6meA to immunoprecipitate transcripts that carry the modification and then subjected the ‘pulled-down’ material to next-generation RNA-seq. 6meAs, as it turns out, are common, are enriched near stop-codons, are highly conserved between mice and man, and crucially, are dynamically, developmentally and tissue specifically regulated [40,41]. Both these studies have far reaching implications as the appropriate presence (and presumably absence) of 6meA at specific mRNA sites appears to perform a fundamental role in regulating gene expression. In particular, CAP structures with 6meA are most efficiently translated as well as exerting varying effects on mRNA splicing and transport.

To date, only one enzyme, methyltransferase like 3 (METTL3), has been identified to catalyse the conversion of adenosine to 6meA [42]. Intriguingly, and of relevance to our current discussion, to date only one enzyme is known to catalyse the removal of this methyl group, and that is FTO [37]. In fact, Meyer and colleagues show that transient overexpression of FTO in HEK293 cells decreases the total amount of 6meA found in the transcriptome [41]. However, it is still unknown if FTO is the only enzyme able to demethylate 6meA. Certainly examination of the transcriptome of tissues from FTO KO mice could determine if there is an increased burden of 6meA. It is also still unclear as to how this demethylation activity of FTO is linked to human obesity.

FTO deficiency

Human FTO deficiency

How about the phenotype of FTO deficiency and what that tells us about FTO's physiological role? In 2009, a large consanguineous family of Palestinian Arab origin with nine members affected by a previously unknown polymalformation syndrome was reported, all of whom were homozygous for an arginine to glutamine change at position 316 (R316Q) in FTO [43]. Unfortunately for the affected individuals, because R316 is one of the obligate residues required for binding of 2-OG, the R316Q mutation completely abolishes FTO's demethylase activity. The syndromic characteristics observed in the affected individuals include postnatal growth retardation, severe functional brain deficits and microcephally, facial dysmorphism and cardiac abnormalities. Tragically, the severity of the phenotype was such that all affected children died within the first 30 months of life. Although the heterozygous carriers within this family did not have their clinical phenotype specifically studied, no overt obesity-related phenotypes were reported [43].

Mouse models of FTO deficiency

The global ablation of Fto expression in mice also results in high early mortality, although in contrast to humans without FTO, around 50% of homozygous mice do survive past weaning [44]. The phenotype of the surviving Fto null mice is complex. The cause of the postnatal lethality is still unknown, although the fact that they are born normally implies that implantation and embryonic development are not grossly affected. It appears that a complete ablation of Fto in mice causes retarded postnatal development [44], consistent with FTO deficiency in humans [43], supporting the hypothesis that Fto is involved in normal body development. They also suffer from a curious and as yet unexplained reduction in ambulatory movement [44]. So lack of FTO in both man and mouse does not appear to be compatible with a long or healthy existence.

A role for FTO in energy homoeostasis?

From the perspective of energy homoeostasis, the absolute food intake of Fto +/+ and Fto −/− mice is identical but when corrected for lean mass, Fto −/− have an increased food intake. Similarly Fto −/− mice have increased oxygen consumption when corrected for lean mass [44]. Additionally, Fto heterozygous (+/−) mice are resistant to high-fat diet induced obesity [44]. Together with a contemporaneous paper reporting that mice with a missense loss of function ENU (N-ethyl-N-nitrosourea) induced mutation in Fto display a lean phenotype [45], these initial studies pointed to some involvement of FTO in the control of energy balance, and that the association of FTO intron 1 SNPs with human obesity arises via functional effects on FTO rather than other genes in the region.

Perturbation of FTO expression in rodents

FTO is widely expressed in foetal and adult tissues in humans, mice and rats, with the highest expression in the brain [31]. Within the brain, Fto expression is relatively high in a number of hypothalamic nuclei: arcuate (ARC), paraventricular (PVN), dorsomedial (DMN), and ventromedial (VMN) nuclei, where control of energy homoeostasis is centred. Within the ARC, expression of Fto is bi-directionally regulated as a function of nutritional status; decreasing following a 48 h fast [31] and increasing after 10 weeks of exposure to a high fat diet [46]. In contrast, another report showed that Fto expression in the whole hypothalamus was up-regulated following 48 h of food deprivation in rats [47]. This disagreement is most likely due to the difference in the extent of food deprivation between the two species, as a 48 h fast in mice is far more severe than that for rats. This is evident from the altered fasting induced gene expression changes normally seen in the hypothalamus [47,48].

Since we believe that fasting, a strong stimulus to eat, reduces Fto mRNA levels in the ARC, if Fto was having a direct action on the control of energy intake one would predict that reducing its expression would recapitulate the fasting state, thereby resulting in an increase in food intake. This does indeed seem to be the case. Tung et al. [46] knocked down and overexpressed Fto specifically in the ARC in rats, using adeno-associated virus (AAV) vectors coupled with stereotactic injections. When FTO is overexpressed, food intake is reduced. Conversely, reducing Fto expression increases food intake [46]. Unlike genetic mouse models, the observed phenotypic effects are transient. However it has proved advantageous for studying acute changes in eating behaviour of the ‘mutant’ rats, and this spatial and temporal specific approach allows us to observe only direct action of Fto.

The regional specific manipulation of Fto supports the notion that by acting specifically in the ARC of the hypothalamus, FTO itself can influence energy homoeostasis by having a direct effect on food intake. These findings are consistent with the robust association between the FTO risk alleles and an increase in food intake. They are also consistent with the relative hyperphagia observed in the Fto −/− mice [44], although both these and the ENU Fto mutant mice paradoxically have a lean phenotype [44,45]. This raises the possibility that although FTO in the ARC seems to play a role in energy intake, FTO in other regions and/or peripheral tissues (that would also be affected by the global loss of FTO) might be responsible for the increase in energy expenditure seen in both mouse models. Thus it is clear from studies of its global deficiency in both man and mouse discussed above, that while FTO, particularly its ability to bind 2-OG and (presumably) act as a demethylase, has a fundamental function during development, it also has a role specifically within the hypothalamus to regulate food intake.

Intriguingly, central nervous system (CNS) specific Fto deleted mice have now been generated [49]. Surprisingly, these brain-specific Fto deficient mice recapitulate the phenotype of the whole-body knock-outs, although this has yet to be exhaustively examined [49]. This does suggest that much of FTO's function, including its link to the regulation of energy homoeostasis (and in keeping with the observations by [46]), is mediated in the brain.

Overexpression of FTO

Church and colleagues have generated a ‘knock-in’ mouse model that carries one or two additional copies of Fto [50]. Mice overexpressing Fto showed a dramatic increase in food intake resulting in a marked increase in body weight and fat mass when they were fed either chow or a high-fat diet (HFD). There was no significant change in either energy expenditure or physical activity. Although the increase in weight with the overexpression of Fto seems consistent with Fto deficiency resulting in a ‘lean’ phenotype, the increase in food intake seen in these mice is not [50]. The authors go on to look at circulating leptin concentrations, and find that, curiously, at the age of 8 weeks, mice overexpressing Fto had reduced leptin levels after a 16 h overnight fast compared to wild type controls. It is fascinating that this phenomenon occurs, keeping in mind that these are obese mice, whose normal response would be to have increased circulating leptin, reflecting their increased fat mass. The authors comment that the hyperphagic phenotype is possibly due to an FTO dependent reduction of leptin, although no mechanism is invoked [50].

Could FTO be acting as a nutrient sensor?

As FTO is nutritionally regulated, and manipulating FTO levels in vivo appears to influence food intake [31,46], we have spent some time exploring the possibility of FTO acting as a nutrient sensor.

FTO is not a sensor for 2-OG

As mentioned above, FTO requires 2-OG, a key intermediate in the citric acid cycle, as a co-substrate in order to be catalytically active. Because FTO is implicated in obesity and is nutritionally regulated within the brain, it was certainly plausible that FTO could act as a sensor for intracellular metabolism. Previous assays of FTO function were all end-point assays that relied upon either radioactive tracers or HPLC detection of products. Although accurate, the problem is that both these approaches require large amounts of purified FTO enzyme, are time-consuming and are of limited utility in measuring enzyme kinetics. We developed a high-throughput, fluorescence, RNase cleavage assay to measure the demethylation activity of FTO on 3meU [51]. As 2-OG acts as a co-substrate, we could then apply our assay to examine the enzyme kinetics of FTO with regards to its usage of 2-OG, arriving at a 2-OG KM value for FTO of 2.88 μM [51]. Since typical intracellular concentrations are measured to be more than 10-fold higher, around 50–100  μM [52], it is clear that although 2-OG is required for FTO activity, it is unlikely that FTO's physiological role is to sense 2-OG .

FTO levels are regulated by availability of essential amino acids

In another study, we utilised an in vitro approach to determine which nutrients could regulate FTO levels at a cellular level. Using mouse and human cell-lines, we demonstrate that both glucose and total amino-acid deprivation regulates FTO expression [53]. In particular, we have found that FTO mRNA and protein levels are dramatically down-regulated by total amino-acid deprivation in mouse hypothalamic N46 cells, mouse embryonic fibroblasts (MEFs) and in human HEK293 cells. The drop rate of Fto mRNA is faster than its rate of natural degradation, pointing to regulation at the transcriptional level, which is reversible upon amino-acid replacement. Strikingly, this down-regulation was seen only with essential amino-acid deficiency. These data suggest that FTO might play a role in the sensing of essential amino-acid availability [53]. It is probably unlikely that FTO is the ‘sensor’ in and of itself because of the late response and dramatically dynamic nature of its expression upon nutrients limitations. However, a role for FTO in the pathways that sense cellular nutritional availability would be congruent with its association with obesity as well as its regulation by dietary challenges in the whole organism.

The different faces of FTO

So what is the current state of play with FTO? What is clear is that overwhelming genetic evidence points to SNPs in FTO being associated with body weight, behaviour and increasingly appetitive behaviour. However to date, we do not know if or how the FTO risk alleles are influencing the FTO protein. Considering the intronic location of all the FTO obesity-related SNPs (spanning across ∼40 kb), they are unlikely to cause functional mutations. Instead, the SNPs are more likely to be playing a transcriptional regulatory role, either to up- or down-regulate FTO expression. In addition, it is also possible that the SNPs could be involved in posttranscriptional modifications, which could result with various FTO splice variants and isoforms of different functions.

We know that FTO is able to demethylate at least two different RNA nucleosides, 3meU and 6meA, and that the consequences of not being able to perform this enzymatic activity are severe in both man and mouse. Although far from straightforward, studies where FTO levels have been perturbed in animal models, genetically or otherwise, appear to support the notion that FTO plays a role in the control of energy homoeostasis, making it likely that the association with body weight is down to some function of FTO.

A question to consider is whether FTO, given its ubiquitous expression, can ever be considered a viable pharmaceutical target? The severity of the phenotype seen in human and murine FTO deficiency clearly points to some fundamental role, particularly in early postnatal development [43,44]. Yet, we have been able to use it to influence food intake by discretely manipulating its expression in certain regions of the brain [46]. We hypothesise that while FTO clearly has a broader biological function, it also has a role specifically within the hypothalamus to regulate food intake. So there is the age-old problem of selectivity and specificity.

Nutritional regulation of FTO both in the whole organism and on a cellular level in response to essential AA availability makes sense and forms a possible link between FTO and regulation of body weight. Are either of these phenomena coupled to the ability of FTO to demethylate that appears to play such a fundamental developmental role? Or is the ‘nutritional sensing’ aspect of FTO a distinct function entirely? These are questions that will have to be urgently tackled, if we are to make inroads into understanding the biology of this first ‘post GWAS’ obesity gene. Understanding the biology of FTO and its downstream actions could potentially reveal novel therapeutic targets in our battle against the increasing epidemic of obesity. We believe that our efforts to date in turning the statistical association of FTO into a deeper understanding of its biology could form a template for how we approach the many other GWAS obesity genes of unknown function.

Conflict of interest

None declared.