We genotyped rs2943641C > T, located 500 kb downstream of IRS1, in 2389 prevalent or incident T2D patients and 6494 controls from two prospective and three case studies based in UK and found evidence for an association of the minor rs2943641T allele with T2D protection. This allele was associated with lower fasting insulin and HOMA-IR index in middle-aged participants of the WHII study and with lower post-load insulin after OGTT in young adults of the EARSII study. In silico analysis with follow-up genotyping also identified that the minor allele of the IRS1 promoter variant rs6725556A > G showed association with reduced T2D risk (OR per G-allele: 0.82, 95%CI: 0.69–0.96, p = 0.015).
Rung and colleagues 
identified rs2943641 as a T2D susceptibility locus in a multistage association study across 14,051 French and Danish individuals (6258 cases and 7793 controls) and showed strong association of the major C-allele with increased risk of T2D (OR: 1.19, 95%CI: 1.13–1.25, p
= 9.3 × 10−12
). This result is equivalent to OR per T-allele: 0.84, 95%CI: 0.80–0.88. Our findings in these UK studies are consistent with an association of rs2943641T with 6% decreased risk of T2D (OR per T-allele: 0.94, 95%CI: 0.87–1.03, p
= 0.18). This association became statistically significant when analyses were repeated with additional adjustment for BMI (overall OR: 0.88; 95%CI: 0.80–0.96, p
= 0.006), although since there was no relationship of this SNP with BMI, and GWAS of genetic variants influencing BMI, obesity and related phenotypes have not identified IRS1
as a BMI related gene 
, the mechanism of this is unclear.
Notably, data from the recently published DIAGRAM meta-analysis 
identified a different SNP (rs7578326A > G) adjacent to rs2943641 to be associated with T2D (OR per A-allele: 1.11, 95%CI: 1.08–1.13, p
= 5.4 × 10−20
; 42,542 cases and 98,912 controls). The two SNPs lie ~73 kb apart and are in strong LD (r2
= 0.79 in HapMap CEU), and therefore this finding provides further confirmation of the previously reported signal. Moreover, using data from up to 46,186 non-diabetic subjects from the Meta-Analyses of Glucose and Insulin-related traits Consortium the authors reported the risk allele to be associated with higher fasting insulin 
, consistent with a primary effect on insulin action.
Rung and colleagues 
also examined the effect of rs2943641 on diabetes-related quantitative traits in three independent cohorts with normoglycemic individuals of Finnish, French and Danish origin (n
= 14,358) and found that the diabetogenic rs2943641C allele was associated with higher fasting insulin and HOMA-IR indices, but not with fasting glucose levels. In middle-aged Danes, the C-allele was also associated with higher insulin levels after OGTT 
. Concordant with those findings, we found an association of the minor rs2943641T allele with lower fasting insulin (p
= 0.04) and HOMA-IR (p =
0.03) in middle-aged individuals of WHII study, and no association with fasting glucose levels. We have also taken this study forward by examining the association of this variant with insulin levels after an OGTT. The T-allele was associated with lower post-load insulin levels in the healthy young males of EARSII (13.3% lower AUCinsulin
= 0.003), but it was not associated with fasting insulin or HOMA-IR, which implies that a significant association of rs2943641T with lower fasting insulin levels may be evident or become established only later in life.
The importance of IRS1
in insulin signaling has been confirmed in studies showing that this gene is associated with peripheral insulin sensitivity as well as in the regulation of insulin secretion [21,22]
and a functional IRS1
variant (Gly972Arg, rs1801278) has been related to T2D risk 
, although some studies have failed to replicate this [24,25]
. Rs1801278 was present on the 50K-chip 
and in WHII it did not show significant association with T2D risk (OR: 1.20, 95%CI: 0.88–1.63, p
= 0.25). Morini and colleagues 
in their meta-analysis of 32 studies suggest that when analysis took into account age of onset of disease (data from 14 studies) there was evidence that this variant was more strongly associated with risk in the tertile of those who had early-onset disease. Our results support this (Supplementary Table 9
) but our study was not powered to find a significant interaction (p
In our exploration of T2D risk as a function of 23 polymorphic IRS1
variants on the HumanCVD BeadChip using data from WHII [12,15]
, with follow-up genotyping in other study cohorts, we found evidence for a possible independent effect on risk of a genetic variant in the 5′-flanking region of IRS1
(rs6725556; −3538A > G), although no test met our prespecified criteria of p
< 0.01 for statistical significance. Specifically, the G-allele of rs6725556 was associated with 18% lower risk of T2D in a meta-analysis of individual participant data from all UK-study cohorts (p
= 0.015). This variant, together with rs2943641 near IRS1
, was independently associated with T2D risk in WHII using a variable selection model with adjustment for age, gender and BMI (OR: 0.50, 95%CI: 0.33–0.78, p
= 0.002 and OR: 0.82, 95%CI: 0.69–0.99, p
= 0.04, respectively) and appeared to have an additive effect on risk. No corrections have been made for multiple comparisons and so these effects should be interpreted cautiously. Rs6725556 and rs2943641 lie 573.5 kb apart, and show no LD (r2
= 0.0 in WHII) and although they were both associated with risk of T2D, rs6725556 was not associated with fasting and post-load insulin levels and HOMA-IR. Interestingly, transcription factor binding site analysis using the MatInspector software tool (http://www.genomatix.de/products/MatInspector/
) revealed that rs6725556G abolishes a PAX-2/5/8 binding site. The paired-box (PAX) gene family encodes a group of transcription factors that have emerged as important regulators of organogenesis in all species 
and PAX2 has been shown to be expressed in endocrine pancreas where one of its functions may be the regulation of pancreatic hormone genes 
. This could be of relevance in the pathogenesis of diabetes and other endocrine disorders; however, whether rs6725556 is indeed a functional polymorphism affecting IRS1
expression needs to be proven in future functional studies. Moreover, we acknowledge that these results are preliminary and that replication of our findings in independent cohorts is essential. We also acknowledge that a limitation of our study is that it is underpowered to detect an association in the Indian Asian cohort. We only have 24% power to detect the association found by Rung and colleagues 
(OR = 0.84) for rs2943641. However, for the Whites we have 99% power to detect a OR of 0.84. If we take account of multiple comparisons for the 6 traits (Supplementary Table 4
) we would still have 94% power.
In summary, this report confirms the association of the major C-allele of rs2943641 near IRS1 with increased risk of T2D, fasting- and glucose-stimulated hyperinsulinemia and impaired insulin sensitivity. Our data also suggest that rs2943641 and an IRS1 putative promoter variant (rs6725556) may independently influence T2D risk, although further studies with larger cohorts are needed to confirm the etiological SNPs and to analyze their interactions in different populations.