20 March 2015

The association of markers in FTO (“Fat mass and obesity associated”) with BMI and related phenotypes is one of the most secure GWAS results. The association seems solid, there’s a mouse model, and a plausible functional story. Even the name sounds pretty convincing. On the other hand, there’s lots of uncertainty about the function of FTO, and the details of the association. In fact, Smemo et al (2014) suggested that in fact FTO was not functionally involved at all. From the reaction of people in the community1, it seemed like many of them had been sceptical of FTO for a while. In general it seems like a good case study for difficulties of interpreting associations and translating GWAS results into biologically relevant information. I must confess I was a bit confused about the whole thing, so I thought I would try and organise my thoughts.

###History of the name The symbol FTO originally comes from the Ft locus in mouse. A large deletion at this locus leads to a variety of developmental phenotypes including fused toes, hence the symbol. The deletion actually involves six genes. The first of these genes to be cloned was given the name Fatso (Fto) not because of any bodyweight phenotype, but because the gene itself is relatively large (~250kb; Peters et al. 1999). The other five genes at the locus are Ftl, Ftm and the Iroqois2 B cluster Irx3 (more about this later), Irx4 and Irx5.

Through a rather amazing coincidence, in the first generation of large GWAS, the human FTO locus showed up as the strongest signal of association for BMI, as well as type 2 diabetes (Frayling et al 2007, WTCCC 2007). Apparently the people involved realised that you probably shouldn’t have an obesity gene called “Fatso” and renamed it to “Fat mass and obesity associated”, which is currently the official name in both human and mouse (Fischer et al. 2008).

###Mouse phenotypes

The original mouse Ft deletion is recessive lethal. Heterozygotes have the eponymous fused toe phenotype, as well as thyroid hyperplasy, left-right asymmetry and a number of other developmental phenotypes. Targeted deletion of Fto showed that Fto-/- mice show postnatal growth retardation and reduced body mass although there was no clear effect of Fto-/+ genotype (Fisher et al. 2009).

###Association results

Many follow up studies have confirmed the association between SNPs in the first intron of FTO and BMI. In populations with European ancestry the effect seems to be to increase BMI by around 0.39kg/m^2. It explains about 0.34% of the variance in the trait, which is pretty good for a single locus. The effect seems to replicate in populations with Asian ancestry, but possibly not in those with African ancestry (see Loos and Yeo 2014 for a review of the association results). Overall the association seems to be about as robust as any I’ve seen. It seems that the association with BMI (Again, reviewed by Loos & Yeo 2014) works by increasing calorific intake, not by decreasing physical activity. It seems to be the case therefore, that the association operates by changing behaviour rather than, say, some metabolic pathway. Some support for this comes from the observation that in adult mice, Fto is largely brain expressed (Gerken et al. 2007).

###Is it really FTO?

However, it turns out that obesity associated SNPs are not eQTLs for FTO expression in humans. Of course, this doesn’t mean that they aren’t doing something else to the protein but Smemo et al. (2014) make a convincing case that in fact, these SNPs are influencing BMI through IRX3 and not FTO. Briefly:

  • Obesity associated SNPs are associated with IRX3 expression in the human brain. BMI-increasing SNPs increase expression of IRX3.
  • Irx3 interacts physically with the Fto locus in mouse brains (using 4C).
  • Irx3 KO mice have reduced adiposity and have less weight gain than wild type on a high fat diet. This seems much closer to the human phenotype than the Fto KO mice. As Smemo et al. observe, 20% of gene knockouts in mice lead to alterations in body size, mass or growth.


This seems like a pretty good case study for the difficulty of interpreting association results.

  • The association looks good and, unlike many GWAS hits, the most associated SNPs are actually inside a gene! However, it turns out to be quite likely that the effect is mediated through another gene entirely. Let’s be very careful in assigning gene names to associations3.
  • eQTL work was very helpful in suggesting an answer here. Perhaps we should try and do some (very) large eQTL studies, from different tissues, in well-phenotyped cohorts.
  • Mouse phenotypes can be quite misleading, particularly for things like growth and development traits, and particularly when we know what we are looking for in advance. We should be very careful when we link mouse and human phenotypes.
  • Should we accept that IRX3 is the answer? Maybe we should stay sceptical (once bitten…). What about humans with LOF mutations in IRX3? This model suggests that they (but not those with LOF mutations in FTO) should be protected against obesity.

###References Boissel S et al. Loss-of-function mutation in the dioxygenase-encoding FTO gene causes severe growth retardation and multiple malformations. Am J Hum Genet 2009

Fischer J et al. On the history of Fto., Obes Facts 2008.

Fischer J, et al. Inactivation of the Fto gene protects from obesity. Nature 2009

Frayling T et al. A Common Variant in the FTO Gene Is Associated with Body Mass Index and Predisposes to Childhood and Adult Obesity, Science 2007.

Gerken T et al The Obesity-Associated FTO Gene Encodes a 2-Oxoglutarate-Dependent Nucleic Acid Demethylase Science 2007

Loos R & Yeo G The bigger picture of FTO: the first GWAS-identified obesity gene., Nat Rev Gen 2014.

Peters T et al. Cloning of Fatso (Fto), a novel gene deleted by the Fused toes (Ft) mouse mutation. Mamm Genome 1999

Smemo S et al. Obesity-associated variants within FTO form long-range functional connections with IRX3 Nature 2014

Wellcome Trust Case Control Consortium Genome-wide association study of 14,000 cases of seven common diseases and 3,000 shared controls Nature 2007


  1. The Obesity/anthropometric GWAS community, that is

  2. Itself one of those silly drosophila gene names

  3. See DISC1, for example