The role of obesity-related genetic loci in insulin sensitivity


Tove Fall, Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, 171 77 Stockholm, Sweden.


Diabet. Med. 29, e62–e66 (2012)


Aims  Despite rapid advancements and many new diabetes susceptibility loci found in the past few years, few genetic variants associated with insulin sensitivity have been described, potentially attributable to the lack of larger cohorts examined with gold standard methods for insulin sensitivity assessment. There is a strong link between obesity and insulin sensitivity, and we hypothesized that known obesity susceptibility loci may act via effects on insulin sensitivity.

Methods  A cohort of 71-year-old men without diabetes (Uppsala Longitudinal Study of Adult Men) underwent a euglycaemic–hyperinsulinaemic clamp and genotyping for genetic variants representing 32 loci recently reported to be associated with BMI (n = 926). The effect of these loci on the insulin sensitivity index (M/I ratio) was examined using linear regression. An in silico replication was performed in publically available data for the three top single-nucleotide polymorphisms from the Meta-Analyses of Glucose and Insulin-related traits Consortium analyses of homeostasis model assessment of insulin resistance (n = 37 037).

Results  Three loci (SH2B1, MTCH2 and NEGR1) were associated with decreased insulin sensitivity at a nominal significance (P ≤ 0.05) after adjustment for BMI, but did not hold for multiple comparison correction. SH2B1 rs7359397 was also associated with homeostasis model assessment of insulin resistance in the Meta-Analyses of Glucose and Insulin-related traits Consortium data set (P = 3.9 × 10–3).

Conclusions  Our study supports earlier reports of SH2B1 to be of importance in insulin sensitivity and, in addition, suggests potential roles of NEGR1 and MTCH2.


homeostasis model assessment of insulin resistance


Meta-Analyses of Glucose and Insulin-related traits Consortium


ratio, insulin sensitivity index


single-nucleotide polymorphism


Uppsala Longitudinal Study of Adult Men


Decreased peripheral sensitivity to insulin (i.e. insulin resistance) and pancreatic β-cell dysfunction are the two main physiological features of Type 2 diabetes and, thus, increased understanding of these features are of great importance for unravelling pathways leading to this disorder. Genome-wide association studies have been very successful in identifying common genetic Type 2 diabetes loci and, to date, 44 such loci have been identified [1]. However, the vast majority of these loci have been associated with β-cell dysfunction or are of unknown function, whereas the number of loci associated with insulin sensitivity is unexpectedly low, prompting further identification. The heritability of insulin sensitivity, measured with intravenous methods, has been estimated to be 0.2–0.6 [2]. In the large genome-wide association studies consortia, the homeostasis model assessment of insulin resistance (HOMA-IR) has been widely used for quantifying insulin sensitivity in spite of its modest correlation (r2 ∼ 0.3 to 0.5) to gold standard methods [3]. An alternative approach for finding insulin sensitivity loci is to study genetic associations of established susceptibility loci for closely related traits.

Obesity and insulin sensitivity are correlated, and share pathways including inflammation and fatty acid abundance [4]. The known number of genetic variations altering the risk of obesity is increasing rapidly, with each variant having small effects on BMI. Speliotes et al. recently published a meta-analysis of ∼250 000 individuals with identification of 18 new single-nucleotide polymorphisms (SNPs) and confirmation of 14 already identified loci [5].

Considering that obesity and insulin resistance are highly correlated, we hypothesized that some loci that have been discovered as BMI-susceptibility loci may primarily act through an effect on insulin sensitivity. Hence, the objective of the present study was to evaluate the effect of BMI-susceptibility loci on euglycaemic–hyperinsulinaemic clamp-derived insulin sensitivity in a community-based cohort of elderly men.

Patients and methods

Study sample

In 1970, all men born between 1920 and 1924 and residing in Uppsala, Sweden, were invited to participate in a health survey, the Uppsala Longitudinal Study of Adult Men (ULSAM; At baseline, 2841 men were invited and 82% (n = 2322) agreed to participate. The subjects were then re-invited for examinations at regular intervals. At age 71 years, 1681 men were invited, of whom 73% (n = 1221) participated.

Eligible for the present study were those individuals that participated in the euglycaemic–hyperinsulinaemic clamp study and who provided DNA (n = 1071). We excluded individuals with Type 2 diabetes (n = 144), or low genotyping call rate (< 90%, n = 1).

A euglycaemic–hyperinsulinaemic clamp was used to assess insulin sensitivity according to the procedure described by DeFronzo et al. [6], but with a higher insulin infusion rate to suppress liver glucose output more completely [56 vs. 40 mU min–1 (m2)–1]. The insulin sensitivity index (M/I ratio) was calculated and represents the amount of glucose metabolized per unit of plasma insulin and was given in 100 × mg kg−1 min−1 mU−1 l−1. The study was approved by the Ethics committee of Uppsala University, Faculty of Medicine. All participants gave their written informed consent.


The ULSAM cohort has undergone prior genotyping on the Human Cardio-Metabo beadchip (Metabochip;, which is designed to interrogate ∼200 000 markers of interest for cardiovascular and metabolic diseases. For this study, we extracted the genotypes representing the loci reported by Speliotes et al. [5] from available Metabochip data. In 23/32 SNPs, the lead SNP was available. For the remaining nine loci, the proxy SNP with the highest r2 with the lead SNP according to HapMap2 release 24 ( was selected (only including proxies with r2 ≥ 0.7). For the nine proxy SNPs, we used phased CEU haplotypes from the HapMap project release 24, build 36 to identify the allele corresponding to the BMI-increasing allele reported by Speliotes et al. [5]. All SNPs had a genotyping call rate > 99% and exact Hardy–Weinberg equilibrium P-values > 0.005.

Statistical analysis

The insulin sensitivity index followed the normal distribution. We assessed the effect of the 32 SNPs using linear regression in additive models. Each locus was evaluated separately in age-adjusted, as well as in age- and BMI-adjusted models. To adjust for multiple comparisons, a permutation procedure (10 000 permutations) and a Bonferroni correction for multiple testing were performed.

The study had a power of 75, 32 and 2.5% to detect a SNP that significantly increased the variance explained (r2) of M/I in the BMI-adjusted models with at least 1, 0.5 and 0.1%, respectively, calculated at an alpha level of 0.0016 (0.05/32) and sample size of 926. Standardized residuals were calculated for the models with nominally significant SNP effects and plotted against predicted values to evaluate equality of variances. The residuals were also evaluated graphically for normality, which were deemed acceptable. Moreover, in order to assess the additive linear effect of genotype, residuals were plotted over genotype. We also tested for non-linearity by fitting a model with fixed effects for having no, one or two risk alleles, and thereafter testing whether the regression coefficients for two risk alleles differed from double the risk of having one risk allele. No test showed evidence of deviance from linearity.


We performed an in silico replication by extracting results for the SNPs significantly associated with M/I ratio adjusted for BMI in our study from publically available results from the Meta-Analyses of Glucose and Insulin-related traits Consortium (MAGIC) analyses of HOMA-IR (n = 37 037; [7]. All analyses were performed using the statistical software packages PLINK v1.07 ( and Stata 11 (StataCorp., College Station, TX, USA).


Sample characteristics for the 926 individuals who were included in the final analysis are shown in Table 1.

Table 1.  Clinical characteristics and measures of glucose homeostasis (n = 926)*
  1. *Results are expressed as means (standard deviations) or percentages.

  2. HOMA-IR, homeostasis model assessment of insulin resistance; M/I ratio, insulin sensitivity index.

Age71 ± 0.59
BMI (kg/m2)26 ± 3.2
Waist–hip ratio0.94 ± 0.05
Fasting plasma glucose (mmol/l)5.4 ± 0.54
Fasting plasma insulin (mU/l)12.1 ± 6.7
HOMA-IR2.9 ± 1.7
M/I ratio (100 × mg kg–1 min–1 mU–1 l–1)5.4 ± 2.4
Creatinine (μmol/l)93 ± 13
Smoking, n (%)192 (21%)
Systolic blood pressure (mmHg)146 ± 18
Anti-hypertensive treatment, n (%)305 (33%)
Total cholesterol (mmol/l)5.9 ± 1.0
HDL cholesterol (mmol/l)1.3 ± 0.3
Lipid-lowering treatment, n (%)81 (9%)
Myocardial infarct or angina prior to investigation, n (%)101 (11%)

Six SNPs (closest genes SH2B1, MTCH2, CADM2, FANCL, NEGR1 and TMEM160) showed nominally significant associations in the age-adjusted models. Three of these SNPs (closest genes NEGR1, MTCH2 and SH2B1) also showed inverse associations with insulin sensitivity after adjusting for BMI. None of these SNPs, however, remained significant after adjusting for multiple testing (Table 2). The variance explained for M/I of these three SNPs were 0.2–0.6% when added to a model including age and BMI. The BMI-increasing allele of rs7359397 (SH2B1) (3.9 × 10–3) significantly increased HOMA-IR in the MAGIC data set. No significant association was seen for the two variants in or near the NEGR1 and MTCH loci.

Table 2.  The effect of 32 obesity loci on the insulin sensitivity index (M/I ratio) measured by euglycaemic–hyperinsulinaemic clamp in a cohort of 71-year-old men without diabetes (n = 926)
GeneAlleleAge-adjusted modelAge- and BMI-adjusted model
Per allele change in M/I ratio P-value*Per allele change in M/I ratio P-value*
SNP reported by Speliotes et al. [5]Proxy (r2)Closest geneChrPosition b36/hg18 (bp)EffectOtherEAF†β (SEM)Age-adjustedEMP2‡β (sem)Age- and BMI-adjustedEMP2‡
  1. *Bonferroni corrected P-values (32 tests) were 0.35 and 0.70 for the age-adjusted association of rs7359397 and rs3817334, respectively, and P = 1 for all other associations.

  2. †EAF, risk allele frequency.

  3. ‡EMP2, empirical P-value 2.

rs7359397Original SNP SH2B1 1628793160TC0.44–0.29 (0.11)0.010.30–0.19 (0.09)0.050.81
rs3817334Original SNP MTCH2 1147607569TC0.40–0.26 (0.11)0.020.50–0.21 (0.1)0.030.62
rs13078807rs9852859 (1) CADM2 385925031CT0.19–0.31 (0.15)0.030.65–0.16 (0.12)0.191
rs887912Original SNP FANCL 259156381TC0.270.26 (0.13)0.040.770.20 (0.11)0.060.87
rs2815752Original SNP NEGR1 172585028AG0.59–0.22 (0.11)0.050.79–0.22 (0.1)0.020.56
rs3810291rs2303108 (1) TMEM160 1952281735CT0.710.24 (0.12)0.050.810.18 (0.1)0.070.92
rs10767664rs2030323 (1) BDNF 1127685115CA0.790.22 (0.14)0.120.980.17 (0.12)0.151
rs2241423Original SNP MAP2K5 1565873892GA0.79–0.20 (0.14)0.161–0.16 (0.12)0.171
rs12444979Original SNP GPRC5B 1619841101CT0.870.20 (0.17)0.2410.20 (0.14)0.171
rs713586rs10182181 (1) RBJ 225003800GA0.470.11 (0.11)0.3110.09 (0.09)0.341
rs1558902Original SNP FTO 1652361075AT0.39–0.11 (0.11)0.331–0.05 (0.1)0.581
rs9816226Original SNP ETV5 31.87E+08TA0.85–0.14 (0.16)0.401–0.07 (0.14)0.591
rs543874Original SNP SEC16B 11.76E+08GA0.21–0.10 (0.14)0.461–0.08 (0.12)0.481
rs4836133rs6864049 (1) ZNF608 51.24E+08GA0.52–0.07 (0.11)0.531–0.03 (0.09)0.721
rs4929949Original SNP RPL27A 118561169CT0.50–0.07 (0.11)0.5510.01 (0.1)0.931
rs571312Original SNP MC4R 1855990749AC0.25–0.08 (0.13)0.5610 (0.11)0.971
rs10968576Original SNP LRRN6C 928404339GA0.320.06 (0.12)0.6110.04 (0.1)0.711
rs2890652rs17834293 (0.71) LRP1B 21.43E+08CT0.110.09 (0.18)0.6110.22 (0.15)0.151
rs7138803Original SNP FAIM2 1248533735AG0.430.06 (0.12)0.6110.09 (0.1)0.341
rs206936Original SNP NUDT3 634410847GA0.200.07 (0.14)0.621–0.02 (0.12)0.901
rs13107325Original SNP SLC39A8 41.03E+08TC0.04–0.13 (0.3)0.651–0.1 (0.25)0.681
rs11847697Original SNP PRKD1 1429584863TC0.04–0.13 (0.3)0.6710.02 (0.26)0.941
rs4771122rs1006353 (0.80) MTIF3 1326945269AG0.270.05 (0.13)0.6910.02 (0.11)0.841
rs10938397Original SNP GNPDA2 444877284GA0.410.03 (0.11)0.8110.06 (0.1)0.511
rs2867125Original SNP TMEM18 2612827CT0.840.03 (0.16)0.8310.06 (0.13)0.671
rs987237Original SNP TFAP2B 650911009GA0.20–0.03 (0.14)0.8410.04 (0.12)0.771
rs10150332rs17109256 (1) NRXN3 1479009746AG0.220.02 (0.14)0.871–0.06 (0.11)0.581
rs1514175Original SNP TNNI3K 174764232AG0.42–0.01 (0.12)0.9210.06 (0.1)0.531
rs29941Original SNP KCTD15 1939001372GA0.68–0.01 (0.12)0.921–0.03 (0.1)0.751
rs2112347Original SNP FLJ35779 575050998TG0.630.01 (0.12)0.931–0.08 (0.1)0.401
rs2287019Original SNP QPCTL 1950894012CT0.780 (0.14)0.981–0.04 (0.12)0.761
rs1555543rs11165643 (1) PTBP2 196696685TC0.570 (0.12)11–0.01 (0.1)0.911


Principal findings

The principal findings of our study were that SNPs in or near NEGR1, MTCH2 and SH2B1 were inversely related with insulin sensitivity (i.e. increased insulin resistance) in addition to their effect in increasing BMI.The BMI-increasing allele of the SH2B1 locus was also associated with increased insulin resistance (HOMA-IR) in the MAGIC data set, which further supports the validity of our findings. The lack of replication of associations of the MCTH2 and NEGR1 loci could be attributable to a true null effect, or the quite low correlation of HOMA-IR to gold standard measurements of insulin sensitivity. The associations of the 32 BMI susceptibility loci to insulin sensitivity have not been reported previously, except for a study on rs1558902 (FTO), where no association with HOMA-IR was found [8].

Potential mechanisms

Obesity has been hypothesized to cause decreased insulin sensitivity through pathways including inflammation and endoplasmic reticulum stress, as well as decreased glucose uptake linked to fatty acid abundance [4]. Genetic loci primarily affecting insulin sensitivity may have been mapped as BMI-susceptibility loci in large meta-analyses of genome-wide association studies because of the high phenotypic correlations of these two traits, or because of pleiotropy, as has been suggested in the case of SH2B1 rs7359397 [9]. This SNP is located < 200 bases downstream of the SH2B1 gene, which encodes an adaptor protein expressed both in the central nervous system and in peripheral tissues. The neuronal form is shown to regulate BMI by enhancing leptin sensitivity. In vivo studies have shown the peripheral SH2B1 to be an insulin sensitizer, which binds to insulin receptors as well as insulin receptor substrate (IRS)-1 and 2 [9]. Speliotes et al. identified SH2B1 rs7359397 as being associated with BMI and reported this variant to be associated with a decreased expression of SH2B1 in adipose tissue and omental fat. The present report is the first population-based study to report association to insulin sensitivity. However, this signal is located in a broad linkage disequilibrium block spanning several genes, and the same study also reported the variant to be associated with altered expression of SULT1A1 and SULT1A2 in subcutaneous and omental fat, so the causal gene in this locus is still to be determined [5].

The second SNP indicated in this study (rs2815752) is tagging a copy number variation close to NEGR1, neuronal growth factor 1, which has not earlier been indicated in insulin sensitivity. This protein is highly expressed in the brain and stimulates neuronal outgrowth [10]. However, carriers of rs2815752 BMI-increasing alleles also have an increased expression of NEGR1 in peripheral blood, which indicates that the protein may have unknown functions outside the brain [5].

The third SNP (rs3817334) is located within an intron of MTCH2, which encodes a putative mitochondrial carrier protein that is believed to have a function in cellular apoptosis. It is not clear if the association to obesity is mediated by actions in the brain or in peripheral tissue, but mitochondrial processes are important players in the cell metabolism and response to insulin. Speliotes et al. reported carriers of the BMI-increasing allele of this SNP to have altered expression of MCTH2 in subcutaneous fat [4].

Strength and limitations

The main strength of our study is the large, well-characterized cohort, tested with an intravenous gold standard method for insulin sensitivity. The main limitations are: (1) limited power to find associations with strict multiple-testing correction, because of the modest number of observations and small effect sizes; (2) lack of a replication sample with clamp information; (3) the inference to women, other age and ethnic groups is unknown; and (4) as in any genetic association study, the SNPs chosen are usually not the causal variants, but are merely tagging such variants, which also may affect other genes than those given as names of the loci studied.


In conclusion, the present study supports earlier reports of SH2B1 to be of potential importance in insulin sensitivity, and suggests a role of NEGR1 and MTCH2. Further studies are needed to identify additional loci involved in insulin sensitivity pathophysiology.

Competing interests

Nothing to declare.


Genotyping was performed by the SNP&SEQ Technology Platform in Uppsala (, which is supported by Uppsala University, Uppsala University Hospital, Science for Life Laboratory – Uppsala and the Swedish Research Council (contracts 80576801 and 70374401). This research was supported through funds from The European Community’s Seventh Framework Programme (FP7/2007-2013), ENGAGE Consortium, grant agreement HEALTH-F4-2007-201413.