What is this 'risk variant' shared by 40% of people with Type 2 Diabetes?

I heard a surprising statistic the other day.  At least it was surprising to me.  The Oct 9 episode of the excellent BBC Radio 4 program "Inside Science" covered a new treatment for type 2 diabetes (T2D). The blurb about the segment said this:

In 2010, a particular gene variant was associated with around 40% of Type 2 diabetics - not directly causal, but this so-called 'risk variant' increases the chance of developing the condition if you have the wrong lifestyle.

The focus of the program was on what seems to be a promising new treatment targeted at people who carry a specific variant in the ADRA2A gene.  The variant inhibits the secretion of insulin from beta cells, and the treatment, yohimbine, seems to reverse that in people with this particular risk variant.  Yohimbine isn't a cure, or the only treatment that would be necessary for carriers of the variant, but it would be used in combination with other drugs.  It seems as though it could potentially be a good addition to control of type 2 diabetes in perhaps a large segment of the population.

But what I was really interested in was this 40% statistic.  It was mentioned but not discussed -- where did it come from, and what did it mean?  Do we now have a very significant explanation for the cause of type 2 diabetes?  If it actually accounts for such a large fraction of cases, why haven't we heard more about it, after so many big genomewide studies?   If not, in what sense is it a 'risk variant'?  And what does it mean that it's causal "if you have the wrong lifestyle"? Presumably some lifestyle risk factor such as energy imbalance or a dietary component interacts with the variant (whatever that means), but that's true of anyone with T2D, so is the causal pathway different in people with the variant?  Do people with the variant and T2D have a different disease in some sense than those without?  What is the frequency of the variant in people without T2D and should we expect it to be much lower than in those who have T2D?

I tried to run down the statistic.  I found a few 2010 papers that looked promising as the source, but I couldn't find it in any of them.  I Tweeted Adam Rutherford, the Inside Science presenter, and he kindly replied with the reference he had seen, a paper in the October 8 issue of Science ("Genotype-based treatment of type 2 diabetes with an a2A-adrenergic receptor antagonist," Tang et al.).  Indeed, the authors of the new, 2014 paper, write:

A genetic variant in ADRA2A was recently associated with defective b cell function (7). The finding represents the first exact disease mechanism for type 2 diabetes associated with a common (30% of the normal population; 40% of patients with type 2 diabetes) risk variant and provides an opportunity to examine the feasibility of a “pharma- cogenetic” approach to treat complex polygenic disorders like type 2 diabetes.

So, there were the numbers but not the actual source of the data.  Citation number 7 turned out to be one of the 2010 papers I'd already looked at, a paper in Science from the same group ("Overexpression of Alpha2A-Adrenergic Receptors Contributes to Type 2 Diabetes," Rosengren et al.) reporting overexpression of an ADRA2A variant associated with suppression of insulin secretion in rats, and humans, and "increased type 2 diabetes risk" in humans.  Rosengren et al. found that "...in a case-control material with 3740 nondiabetics and 2830 diabetics, rs553668 was associated with increased risk of T2D [recessive effect; odds ratio (OR) 1.42, confidence interval (CI) 1.01 to 1.99, P = 0.04]."  That is, risk is 42% higher in those with rs553668 than those without.  Ok, but that wasn't what I was looking for.

So I turned to the Supplemental information.  Here's the best I could do, but it's not the 30% and 40% I was looking for, either:

Table S2; Effects of SNPs on plasma glucose and serum insulin, Supplemental information, Rosengren et al, 2010

The 'A' allele is the 'risk variant', so here 28% of the cases and 24% of the controls have at least one copy, are either GA or AA at the chromosomal locus.  Not a large difference between the two groups, and not the figures I was looking for, either. I emailed the senior author of the 2010 paper last week to ask about these statistics, but haven't heard back.

I'm either overlooking something obvious, or it's not where I'm looking or it has somehow been misreported along the way.  But this is just increasing my curiosity.  What about this paper, reporting on a study of the ADRA2A association with T2D in Sweden?  SNP rs553668 is the 'risk variant' of interest, according to the previous papers I'd read.

SNP rs553668 was associated with T2D in men (odds ratio [OR] = 1.47; 95% confidence interval [CI] = 1.08–2.01; P = 0.015) but this association was lost after adjusting for age and for body mass index (BMI). Associations were also detected when comparing obese NGT and lean NGT subjects (OR = 1.49; 95% CI = 1.07–2.07; P = 0.017), and in obese (OR = 1.62; 95% CI = 1.06–2.49; P = 0.026), but not in lean T2D. In women, multiple logistic regression regarding SNP rs521674 demonstrated an increased OR of 7.61 (95% CI = 1.70–34.17; P = 0.008) for T2D when including age as a covariant. Correcting for BMI removed the significant association. When age was included in the model, association also found when obese T2D patients were compared with lean NGT subjects (P = 0.041). ADRA2A mRNA expression in human pancreatic islets was detectable, but with no statistically significant difference between the diabetic and the control groups.  [Highlighting is mine.]

So 'the' risk variant is associated with obesity in men, but it's a different variant in this gene in women that's associated with obesity.  Of course, obesity is associated with T2D, but that's a step removed from what the other papers are suggesting.

A meta-analysis in 2013 found that SNP rs553668 may be associated with T2D in Europeans, but no other ethnic groups.  But what about this statistic: GWAS have explained only ~10% of "heritability" of T2D, including ADRA2A because that chromosome region is covered by genome-spanning markers used in GWAS.  This doesn't seem to jibe with the idea that 40% of people with T2D have the ADRA2A 'risk variant.'  And of course if 30% of the healthy population has the risk variant, either they went on to develop T2D after the study was completed, or it's simply not a significant risk variant.

This is an important point: if a large fraction of the population carries the variant, it's not a signifiant risk variant by itself, and doesn't cause 40% of cases, even if it may turn out to be a useful variant to know about if you have T2D, in making treatment decisions.  The population risk of T2D is heavily dependent on lifestyles and very changeable over the years (it is rapidly becoming very much more common than even earlier in our own lifetimes).  But if we were to say that 8% of the population will get T2D, the current estimate, then of these, 40%, or 3.2% of the population has an involvement of this particular gene.  That may be important, but it doesn't explain the epidemic, and leaves unanswered questions.

I hoped to get to the bottom of this by the end of this post.  Instead, I remain confused.
-------------

*Update* Oct 26. Dr Rosengren replied to my email this afternoon.  He said that "the major allele frequency for the variant is in Table S1 in the Science paper. " That's this table:

Source

The risk variant is rs553668, for which the minor allele frequency is 0.143.

He also said that "T2D patients have a 40% higher frequency."  I have to say, this doesn't clarify things for me.

Additionally, he said that 'controls' with the variant are at higher risk of developing diabetes, which means, at least to me, that they aren't good controls.