Tuesday, September 18, 2012

Big deal over a big meal.....for some! The tiny bite of FTO and obesity

"Small genetic change has heavy consequences," says the headline on a PR release from the University of Queensland, one of genetic epidemiologist Peter Visscher's academic homes.  Further, the release goes on to say that "[o]ne small change to the DNA sequence can cause more weighty changes to the human body, according to a new study..."

Has the gene for obesity been identified perhaps?  Well, no.  Even the release doesn't actually claim that.  Rather, if you look at the actual paper, published online in Nature on September 16, the research group reports, from a meta-analysis of 38 genomewide association studies (GWAS) of variation in height and BMI, that they've found a SNP in the FTO gene that they say is responsible for variation in BMI, body mass index.  The FTO gene, or fat mass and obesity-associated protein, codes for an enzyme associated with regulation of food intake.  Variation in BMI in the group with the obesity-associated SNP variant is about 7% greater than those without, and people with the variant are, on average, about a whopping 0.5 kg heavier than people without.  Visscher, one of the co-authors on the study, says that this is important "because it demonstrates that genes can be found that affect trait variability."

That is, if there is phenotypic variability in a trait for which the genotype is presumed to be known, this could indicate a gene by environment interaction affecting the phenotype.  This is known as the "reaction norm" of the genotype: a tree will be tall and straight in low altitudes, but the same tree would be short and wide higher up, for example.

Visscher goes on to say that the study provides an indirect way to get a handle on genotype by environment interactions, although they did not directly measure environmental variables so can't say this definitively.  But, he says:
“For example, if the effect of a gene on weight is smaller in people who physically exercise than in people who do not, then this will lead to less variation among people with two copies of the weight decreasing variant.
This has no necessary connection to the environment, except if we understand that the rest of each person's genome is part of the FTO's environment.  The physically external environment, perhaps the internal bacterial environment, and the genomic environment are all involved.  This doesn't change the point of the story, however.

Unlike the press coverage of this study, the co-authors downplay the significance of their findings with respect to explaining variation in BMI or height, and report that they haven't actually tested or found any gene by environment interaction to explain the modest effect of the FTO SNP that they did find. Possibly it is the genomic variation that, in the context of one of the FTO variants, leads to more trait variation than the same genomic variation in persons with the 'normal' FTO variant.

The paper concludes:
Overall, our findings are consistent with a low heritability of phenotypic variability and no common genetic variants that account for a large proportion of variation in environmental or phenotypic variability. They also indicate an absence of widespread genotype-by-environment interaction effects, at least for height and obesity in humans and with interaction effects large enough to be detected in our study in which specific environmental factors were not identified. Nevertheless, the demonstration that individual genetic loci with effects on variability can be identified with sufficiently large sample sizes facilitates further study to understand the function and evolution of the genetic control of variation.
But this is (forgive us) culpably overstated!
Genes code for proteins and they interact with other things--often, they're catalysts which means they affect the rate of other reactions in the cell.  Or they interact with other proteins, in ways whose efficiency depends on how tightly or effectively they bind with each other, or with DNA, or other cell products.  Expression levels of a gene are similarly about quantity of effects.  This means, almost by definition, that genetic variation affects variation, depending on the other things most any gene has to interact with.  Further, most disease effects associated with genes (or their interaction with environments) affect the age pattern of onset of traits like body weight, or of disease.  Again, these are quantitative effects on variation.

Roughly before WWII, Native Americans and 'Hispanics' who are admixed with Native Americans and Europeans, had very little diabetes or gallbladder disease and had a more typical body shape.  Since 1950 or so, they have experienced a near pandemic of such disorders, with morbid and even lethal consequences.  And increased BMI is one of the well-known, classical effects of this (a topic on which Ken has written for many years).  There is no absence of GxE evidence!  One would say there was only if one is, as this study seems to be, only dealing with present-day populations.

The main author of this paper is a very capable scientist, working in Australia.  And surely he knows that the same kind of thing absolutely and in well-documented fashion, applies to the aboriginal population of Australia.

So, then, why is this paper significant?  According to the authors, it's because they've shown that individual alleles can be responsible for variation in phenotypic variation.  Important?  New?  You be the judge.

We should conclude on a more positive note.  We don't challenge the results, which seem perfectly reasonable: FTO, one of the clearest and most replicated obesity-related genes known, has only a tiny effect as the authors have noted, and the extra variation means that the variant SNP allele has little if any predictive value.  Yet genome-based prediction is what you're being sold by the science these days.

Our point is to challenge the  overstatement, which we think is part of a systematically misleading campaign to geneticize almost everything.

No comments: