We aimed to validate genetic variants as instruments for insulin resistance and secretion, to characterize their association with intermediate phenotypes, and to investigate their role in type 2 diabetes (T2D) risk among normal-weight, overweight, and obese individuals. We investigated the association of genetic scores with euglycemic-hyperinsulinemic clamp– and oral glucose tolerance test–based measures of insulin resistance and secretion and a range of metabolic measures in up to 18,565 individuals. We also studied their association with T2D risk among normal-weight, overweight, and obese individuals in up to 8,124 incident T2D cases. The insulin resistance score was associated with lower insulin sensitivity measured by M/I value (β in SDs per allele [95% CI], −0.03 [−0.04, −0.01]; P = 0.004). This score was associated with lower BMI (−0.01 [−0.01, −0.0]; P = 0.02) and gluteofemoral fat mass (−0.03 [−0.05, −0.02; P = 1.4 × 10−6) and with higher alanine transaminase (0.02 [0.01, 0.03]; P = 0.002) and γ-glutamyl transferase (0.02 [0.01, 0.03]; P = 0.001). While the secretion score had a stronger association with T2D in leaner individuals (Pinteraction = 0.001), we saw no difference in the association of the insulin resistance score with T2D among BMI or waist strata (Pinteraction > 0.31). While insulin resistance is often considered secondary to obesity, the association of the insulin resistance score with lower BMI and adiposity and with incident T2D even among individuals of normal weight highlights the role of insulin resistance and ectopic fat distribution in T2D, independently of body size.
Introduction
Type 2 diabetes (T2D) develops when insulin secretion is insufficient to maintain normoglycemia, often in the context of an obesity-induced increase in insulin demand, i.e., insulin resistance (1). Despite the importance of obesity as a risk factor for T2D, clinical heterogeneity exists in pathways leading to T2D. A recent report from the European Prospective Investigation into Cancer and Nutrition (EPIC)-InterAct study showed that over 10% of incident cases of T2D occurred among individuals of normal weight and over 50% occurred in individuals who were nonobese at baseline (2). It was also shown that waist circumference was associated with risk of T2D within BMI strata, suggesting that for a given BMI, the pattern of fat storage is an important determinant of T2D risk. Indeed, being overweight but with a high waist circumference made future risk of T2D comparable to that of obese individuals (2).
Most genetic variants associated with T2D are implicated in β-cell function (3). Recent studies revealed a stronger effect of these variants on T2D in lean individuals (4), highlighting the role of impaired insulin secretion in individuals who develop T2D in the absence of obesity. The relative role of insulin resistance in T2D, independent of obesity, has been more difficult to disentangle. This is partly attributable to the strong correlation between obesity and insulin resistance and also because gold standard measures are seldom feasible in large-scale prospective studies. Rare monogenic examples of severe insulin resistance in lean patients have been described (5). Among these syndromes, patients with lipodystrophy exhibit severe insulin resistance, metabolic dyslipidemia, and diabetes resulting from impaired adipose tissue function. However, the role of common genetic variants associated with insulin resistance in the etiology of T2D, particularly among nonobese individuals, remains poorly documented.
Initial evidence that genetic approaches can highlight specific etiological pathways comes from recent investigations showing that individuals carrying body fat–lowering alleles at the IRS1 locus are insulin resistant and have a higher risk of dyslipidemia, T2D, and coronary heart disease (6). As part of the Meta-Analyses of Glucose and Insulin-related traits Consortium (MAGIC), we have recently identified 19 single nucleotide polymorphisms (SNPs) associated with fasting insulin (including IRS1), 10 of which were also associated with a dyslipidemic profile suggestive of a role in insulin resistance (7,8). Such genetic variants allow the opportunity to investigate the correlates and consequences of lifelong genetic susceptibility to insulin resistance and/or insulin secretion independently of obesity (7–9).
This study therefore aimed to 1) validate the use of recently identified common genetic variants as specific markers for insulin resistance or secretion; 2) characterize associations between these variants and detailed metabolic measures, including measures of body size and fat mass and distribution; and 3) use these instruments to investigate the contributions of insulin resistance and secretion to the risk of T2D in normal-weight, overweight, and obese individuals.
Research Design and Methods
Cohort Characteristics
Up to 1,374 (range by phenotype [Nrange], 1,136–1,374) individuals without diabetes who attended phase 3 of the Medical Research Council (MRC) Ely study (10) had relevant phenotypic measurements and genotyping from the Illumina Cardio-MetaboChip (MetaboChip). Up to 4,322 individuals from the Fenland study (11) without diabetes and with MetaboChip genotyping (Nrange = 2,618–2,973) or imputed into 1,000 genomes (12) using Impute from the Affymetrix 5.0 genotyping chip (Nrange, 1,223–1,357) were included. The definition of regional compartments in dual-energy X-ray absorptiometry (DXA) data are shown in Supplementary Fig. 1. We performed sensitivity analyses subtracting the gynoid component from leg estimates to avoid double counting of gynoid mass. Up to 1,031 (Nrange, 923–1,031) individuals from the Relationship Between Insulin Sensitivity and Cardiovascular Disease Risk (RISC) study (13) who underwent a euglycemic-hyperinsulinemic clamp were included. Genotyping was performed at KBioscience and imputed into 1,000 genomes using MaCH and Minimac. Up to 909 (Nrange, 884–909) nondiabetic participants from the Uppsala Longitudinal Study of Adult Men (ULSAM) study (14) were included and had genotyping available from MetaboChip. Participants were of European ancestry. Participant characteristics and measurement availability for each study are shown in Table 1, while the number of participants included in each analysis is also shown in Figs. 1–3.
. | Study . | |||||
---|---|---|---|---|---|---|
. | MRC-Ely . | RISC . | ULSAM . | Fenland (MetaboChip) . | Fenland (genome-wide) . | EPIC-InterAct subcohort . |
N (male/female) | 1,374 (629/745) | 1,031 (453/578) | 907 (907/0) | 2,978 (1,403/1,575) | 1,357 (597/760) | 16,154 (6,111/10,043) |
Age (years) | 60.9 (9.2) | 43.96 (8.37) | 70.98 (0.59) | 47.0 (7.1) | 45.0 (7.3) | 52.4 (9.2) |
Height (cm) | 167.5 (8.9) | 170.91 (9.32) | 175.12 (5.95) | 170.6 (9.5) | 169.8 (9.3) | 166.2 (9.3) |
BMI (kg/m2) | 27.03 (4.68) | 25.47 (4.01) | 25.97 (3.22) | 26.7 (4.9) | 27.1 (4.9) | 26.0 (4.2) |
Waist circumference (cm) | 92.57 (13.2) | 86.46 (12.70) | 93.85 (9.12) | 90.5 (13.4) | 92.2 (13.6) | 86.4 (12.7) |
M/I | 0.14 (0.07) | 5.43 (2.42) | ||||
Matsuda index | 7.09 (4.60) | 6.42 (1.86) | 4.30 (2.66) | |||
Insulinogenic index (pmolinsulin/mmolglucose) | 115.70 (108.0) | 30.42 (19.86) | ||||
Fasting glucose (mmol/L) | 4.97 (0.54) | 5.05 (0.56) | 5.34 (0.53) | 4.8 (0.71) | 4.89 (0.61) | |
Fasting insulin (pmol/L) | 55.84 (34.64) | 34.4 (18.68) | 72.41 (40.70) | 45.1 (11.2) | 46.4 (32.9) |
. | Study . | |||||
---|---|---|---|---|---|---|
. | MRC-Ely . | RISC . | ULSAM . | Fenland (MetaboChip) . | Fenland (genome-wide) . | EPIC-InterAct subcohort . |
N (male/female) | 1,374 (629/745) | 1,031 (453/578) | 907 (907/0) | 2,978 (1,403/1,575) | 1,357 (597/760) | 16,154 (6,111/10,043) |
Age (years) | 60.9 (9.2) | 43.96 (8.37) | 70.98 (0.59) | 47.0 (7.1) | 45.0 (7.3) | 52.4 (9.2) |
Height (cm) | 167.5 (8.9) | 170.91 (9.32) | 175.12 (5.95) | 170.6 (9.5) | 169.8 (9.3) | 166.2 (9.3) |
BMI (kg/m2) | 27.03 (4.68) | 25.47 (4.01) | 25.97 (3.22) | 26.7 (4.9) | 27.1 (4.9) | 26.0 (4.2) |
Waist circumference (cm) | 92.57 (13.2) | 86.46 (12.70) | 93.85 (9.12) | 90.5 (13.4) | 92.2 (13.6) | 86.4 (12.7) |
M/I | 0.14 (0.07) | 5.43 (2.42) | ||||
Matsuda index | 7.09 (4.60) | 6.42 (1.86) | 4.30 (2.66) | |||
Insulinogenic index (pmolinsulin/mmolglucose) | 115.70 (108.0) | 30.42 (19.86) | ||||
Fasting glucose (mmol/L) | 4.97 (0.54) | 5.05 (0.56) | 5.34 (0.53) | 4.8 (0.71) | 4.89 (0.61) | |
Fasting insulin (pmol/L) | 55.84 (34.64) | 34.4 (18.68) | 72.41 (40.70) | 45.1 (11.2) | 46.4 (32.9) |
Data are n and mean (SD). M/I in RISC is reported in (mmol/kg fat-free mass/min/nmol/L)/1,000, while M/I in ULSAM is reported in mg/kg body weight/min/mU/l*100.
We tested associations between genetic risk scores and incident diabetes in the EPIC-InterAct study (15), a case-cohort study nested within EPIC cohorts that includes 12,403 incident cases of T2D and a subcohort of 16,154 individuals (including 778 randomly selected incident T2D cases). A maximum of 18,676 participants (8,136 incident cases, 10,540 noncases) had genotypes available from the MetaboChip (n = 9,361) or Illumina 660W-Quad Chip (n = 9,290) imputed into 1,000 genomes and were included in the current study. Up to 10,923 participants (Nrange, 10,029–10,923) from the EPIC-InterAct subcohort were also included in quantitative trait analyses (Table 1).
All participants gave written informed consent, and studies were approved by local ethics committees and the Internal Review Board of the International Agency for Research on Cancer.
Genetic Risk Scores
We created unweighted (i.e., per allele) genetic risk scores for insulin resistance and impaired insulin secretion using effect alleles defined from the literature as shown in Supplementary Table 1. The insulin resistance genetic score comprised variants associated with fasting insulin in recent meta-analyses (8). In order to improve specificity, we restricted the insulin resistance score to the 10 variants showing association (P < 0.05) with lower HDL and higher triglycerides (8,16), a hallmark of common insulin resistance. This excluded TCF7L2, associated principally with insulin secretion (17), and FTO, whose effect on insulin levels was entirely mediated by BMI (8). Variants included were those in or near the IRS1, GRB14, ARL15, PPARG, PEPD, ANKRD55/MAP3K1, PDGFC, LYPLAL1, RSPO3, and FAM13A1 genes (Supplementary Table 1). For the insulin secretion score, from loci associated with T2D and related traits (18–21), we undertook literature searches to identify SNPs showing an association with impaired early insulin secretion. In addition, we investigated the literature to identify additional candidate genes associated with early insulin secretion. Up to 21 variants associated with lower early insulin secretion were included in the insulin secretion score. Where SNPs were missing, we included a proxy where available (Supplementary Table 1), and where no proxy was available, we did not impute missing SNPs. For each SNP, the reason for inclusion in the score and its availability in each study is shown in Supplementary Table 1. The genetic score distributions in each study are shown for the insulin secretion and insulin resistance scores in Supplementary Fig. 2A and B, respectively.
Statistical Analysis
In order to meta-analyze data from multiple studies centrally, each study first natural log transformed and standardized the phenotype, such that for each variable, the mean was equal to zero and SD equal to one. Each study then fit linear regression models on each of these outcomes using the genetic risk scores as exposures, adjusted for age and sex (with and without adjustment for BMI). Genetic risk scores were unweighted, and effect sizes were expressed per fasting insulin–raising or insulin secretion–lowering allele, respectively. We investigated the association of these risk scores with euglycemic-hyperinsulinemic clamp (22) and oral glucose tolerance test (OGTT)-based measures of insulin sensitivity and secretion (23,24). We also investigated the associations of scores with glycemia and insulinemia during the OGTT, lipids, BMI, waist and hip circumferences, and body fat percentage (assessed by bioimpedance in RISC and by DXA in Fenland). We performed fixed-effect, inverse-variance weighted meta-analyses using Stata SE-12.1 software (StataCorp LP, College Station, TX). Associations with T2D in the EPIC-InterAct study were investigated using Prentice-weighted Cox regression with age as the underlying time variable, adjusted for age at entry (to account for potential cohort effects), sex, and center of recruitment. BMI strata were defined by World Health Organization cutoffs, and waist circumference strata were defined by sex-specific tertiles. Interactions of risk scores with BMI and waist circumference were tested by including the product term of risk scores and BMI categories or waist circumference tertiles. Effect sizes were expressed as hazard ratios (HRs) per risk allele.
Results
Validation of Genetic Risk Scores: Associations With Insulin Sensitivity and Secretion
The insulin resistance score was associated with lower whole-body insulin sensitivity based on the M/I value from euglycemic-hyperinsulinemic clamps (β in SDs per allele [95% CI], −0.03 [−0.04, −0.01]; P = 0.004) (Fig. 1A) and with lower Matsuda index (β = −0.03 [−0.05, −0.02]; P = 2.2 × 10−5) calculated from frequently sampled OGTTs (Fig. 1A). The score was not associated with insulinogenic index but was associated with higher insulin levels throughout the OGTT (P < 0.001) and higher levels of glycemia, albeit only statistically significant for 2-h glucose (Fig. 1A).
In contrast, the insulin secretion score was associated with lower insulinogenic index (−0.05 [−0.06, −0.04]; P = 2.1 × 10−14) and lower 30-min insulin levels (−0.05 [−0.06, −0.03]; P = 3.2 × 10−13) but showed no associations with any of the measures of insulin resistance, including M/I (0.00 [−0.01, 0.02]; P = 0.59), Matsuda index (0.00 [−0.01, 0.02]; P = 0.40), or fasting insulin (0.00 [−0.01, 0.01]; P = 0.95) (Fig. 1B). Unlike the insulin resistance score, associations of the insulin secretion score with lower postchallenge insulin were accompanied by higher glucose levels at all time points (P values < 1.7 × 10−4).
Associations With Detailed Anthropometric and Metabolic Traits
The insulin resistance score was strongly associated with both higher triglycerides (0.03 [0.02, 0.03]; P = 3.5 × 10−20) and lower HDL cholesterol (−0.02 [−0.03, −0.02]; P = 1.6 × 10−14). It was also associated with lower BMI (−0.01 [−0.01, −0.00]; P = 0.02), smaller hip circumference (−0.01 [−0.02, −0.01]; P = 4.4 × 10−4) and lower body fat percentage (−0.01 [−0.02, −0.00]; P = 0.02) (Fig. 1A). The insulin resistance score was also associated with lower BMI when we restricted analyses to incident cases of T2D in the EPIC-InterAct study (N = 7,577; −0.02 [−0.03, −0.01]; P = 0.001). Further investigation of detailed anthropometric measures obtained by DXA in Fenland participants highlighted inverse associations of the score with fat mass in different body compartments (Supplementary Fig. 1). The strongest associations were observed for leg (−0.03 [−0.05, −0.02]; P = 1.4 × 10−6) and gynoid (−0.03 [−0.04, −0.02]; P = 9.9 × 10−6) fat mass. These associations remained after excluding PPARG and IRS1 variants from the genetic score (leg [−0.03 (−0.04, −0.01); P = 1.2 × 10−4] and gynoid [−0.03 (−0.04, −0.01), P = 3.4 × 10−4] fat mass). Sensitivity analysis on leg fat mass, removing the gynoid region, showed that the association remained highly significant (−0.03 [−0.05, −0.02]; P = 9.2 × 10−7). For these associations, we saw similar magnitudes of association in men and women, which were statistically significant (P < 0.05) in both sexes. We also saw an association with arm fat mass (−0.02 [−0.03, −0.00]; P = 0.02), but not with lean mass measurements (Fig. 2).
In order to investigate the possibility that lower levels of gluteofemoral fat mass might limit subcutaneous fat storage and hence increase ectopic fat deposition, we also investigated the association of the insulin resistance score with estimates of liver damage. The score was associated with both higher alanine transaminase (ALT; 0.02 [0.01, 0.03]; P = 0.002) and γ-glutamyl transferase (γ-GT; 0.02 [0.01, 0.03]; P = 0.001). The insulin resistance score was not associated with self-reported alcohol intake (P = 0.66), and associations with liver enzymes were unchanged after adjustment alcohol intake (ALT, 0.02 [0.01, 0.04], P = 0.002; γ-GT, 0.02 [0.01, 0.04], P = 0.003).
The insulin secretion score was not associated with triglycerides or HDL cholesterol (P > 0.1) or with any of the anthropometric traits (P > 0.18) (Fig. 1B). The secretion score was nominally associated with higher android fat mass (P = 0.04) but with no other parameters in the DXA data. We saw a weak association of the insulin secretion score with higher levels of ALT (0.01 [0.00, 0.02]; P = 0.02), but not γ-GT (0.00 [−0.00, 0.02]; P = 0.24).
Associations with T2D
Both the insulin secretion (HR [95% CI], 1.09 [1.07, 1.11]; P = 8.0 × 10−30) and resistance scores (1.08 [1.06, 1.10]; P = 4.0 × 10−15) were associated with incident T2D (Fig. 3). To investigate the relative importance of genetically predicted insulin resistance and secretion on T2D incidence at different levels of BMI, we examined the effect of the score on incident T2D in normal-weight, overweight, and obese individuals and found no difference in associations between strata (normal weight HR = 1.07 [1.04, 1.11], P = 1.3 × 10−4; overweight HR = 1.08 [1.05, 1.10], P = 4.7 × 10−9; obese HR = 1.06 [1.03, 1.09], P = 1.0 × 10−4; Pinteraction = 0.58) (Fig. 3). There was also no difference between strata of waist circumference (Pinteraction = 0.31) (Fig. 3). In contrast, the insulin secretion score showed an interaction with waist circumference on the risk of T2D (Pinteraction = 0.001). The association was stronger in individuals with smaller waist circumference than in those with large waist circumference (lowest third HR = 1.13 [1.10, 1.16], P = 1.6 × 10−10; middle third HR = 1.12 [1.09, 1.15], P = 9.5 × 10−23; highest third HR = 1.07 [1.04, 1.09], P = 9.0 × 10−9). There was a similar but nonstatistically significant trend for BMI (Pinteraction = 0.07), with a tendency toward stronger associations in leaner compared with obese individuals (Fig. 3).
Discussion
While rare monogenic examples of insulin resistance highlight the causal role of inadequate subcutaneous adipose tissue in the etiology of cardiometabolic disease, the causal role of impaired adipose expandability and ectopic lipid accumulation in “common” cardiometabolic disease remains largely unproven. We observe that a genetic score for insulin resistance displays a pattern of association (lower subcutaneous adipose tissue and T2D) similar to that observed in monogenic forms of lipodystrophy, implicating a role for inadequate capacity to store surplus lipids in the etiology of T2D. Furthermore, we show that these genetic scores are associated with incident T2D even in individuals of normal weight, highlighting the role of impaired adipose expandability in T2D independently of BMI.
Validation of the Genetic Risk Scores
We found that the genetic risk score comprising variants previously associated with fasting insulin was associated with euglycemic-hyperinsulinemic clamp–based insulin sensitivity. Furthermore, a genetic score comprising variants previously associated with early insulin secretion was strongly associated with 30-min insulin and insulinogenic index, but not with insulin sensitivity. These associations validated the utility of these risk scores as specific and sensitive genetic instruments to understand the role of both insulin resistance and β-cell dysfunction in the etiology of diabetes and other disease outcomes.
Association of Genetic Risk Scores With Other Metabolic Traits
The insulin resistance score was associated with lower BMI, hip circumference, and body fat percentage and particularly with lower gynoid and leg fat mass (Fig. 2), adipose tissue depots considered protective against the complications of ectopic fat deposition (25). A prevailing hypothesis for the pathogenesis of insulin resistance proposes that the capacity of adipose tissue to expand in the face of sustained positive energy balance is finite and that exceeding this limit results in lipid storage in tissues less well adapted to this need (26). This phenomenon of ectopic lipid accumulation has been strongly associated with insulin resistance in multiple studies, and plausible, albeit still largely unproven, explanations exist linking this lipid accumulation to impaired insulin action (lipotoxicity) (27). Lipodystrophic disorders are characterized by a primary lack of adipose tissue and present an extreme example of a mismatch between the need and capacity to store surplus lipids. These extremely rare disorders are associated with particularly severe ectopic fat accumulation, dyslipidemia, insulin resistance, and diabetes. Recent reports suggests novel forms of lipodystrophy that may be more common (28). Here we report evidence that common genetic variants show associations similar to those observed in rare, monogenic lipodystrophies (Fig. 2) and highlight a potential role a mismatch between the need and capacity to store surplus lipids in “common” metabolic disease. Associations with elevated ALT and γ-GT are indicative of hepatic fat deposition, consistent with ectopic lipid accumulation. While elevated ALT and γ-GT are associated with fatty liver (29), they can be elevated in response to other forms of liver injury or disease (30), including alcohol consumption, medication, or hepatitis. While we could not exclude the possibility that the insulin resistance score was associated with higher ALT and γ-GT via mechanisms other than fatty liver, associations with ALT and γ-GT were unchanged after adjustment for self-reported alcohol intake and other forms of disease sufficiently rare that in this healthy middle-aged population, we consider them unlikely to explain our findings.
Associations With Incident Disease
While obesity is a major risk factor for insulin resistance and T2D, there is considerable interindividual variation in the metabolic response to obesity, with some individuals apparently protected from the typical consequences of obesity (31). It has previously been shown that insulin-sensitive obese individuals have lower visceral and hepatic fat content than insulin-resistant obese individuals, as well as a lower intima-media thickness (32), further implicating ectopic fat deposition as a determinant of the metabolic consequences of obesity. Despite negative confounding by BMI, we saw an association of the insulin resistance score with incident T2D. As the adipose tissue of obese individuals is placed under greater demand for fat storage, we hypothesized that the insulin resistance score would have a larger effect on T2D risk in these individuals than in normal-weight individuals. However, the insulin resistance score was associated with incident T2D even in normal-weight individuals and those with the lowest waist circumference (Fig. 3), with similar effect sizes to obese individuals. This suggests that the relationship between adipose expandability and positive energy balance is not subject to a threshold effect but may result in degrees of ectopic fat accumulation even in people with a “normal” BMI. This is reminiscent of what is observed in South Asian individuals who have been reported to have higher visceral fat and exacerbated metabolic consequences for a given BMI (33). While the role of β-cell function has been highlighted in the etiology of T2D among lean individuals (4), our findings also highlight the role of impaired adipose expandability and insulin resistance in T2D in lean individuals.
Recent analyses have highlighted the causal role of increased adiposity in impaired cardiometabolic health (34), and we now highlight a causal link between insulin resistance and incident disease, completely independent of BMI. The obesity epidemic is heavily implicated in driving the increased incidence of metabolic disease (35,36). However, while there is some suggestion that hyperinsulinemia can be a cause and consequence of obesity (37), we observe that genetically predicted insulin resistance and hyperinsulinemia are associated with lower adiposity (Fig. 2). However, as we restricted our genetic score to those variants associated with dyslipidemia (to improve specificity), we cannot exclude the possibility that primary insulin resistance of another form could cause obesity or display different associations with metabolic traits or disease. While insulin resistance is strongly associated with obesity and the secular trends in obesity raise concerns about the growing consequences of insulin resistance (35), of the 19 loci recently found to be associated with fasting insulin levels, only one (FTO) was mediated entirely by higher BMI, highlighting the role of other pathways in the etiology of insulin resistance. Furthermore, the association of 10 of the 19 SNPs with dyslipidemia (indicative of postreceptor-mediated insulin resistance [38]) implicate this as a prevalent form of common insulin resistance.
As our cross-sectional analyses were restricted to individuals without diagnosed diabetes, we considered the possibility that the insulin resistance score association with lower BMI and adiposity was as a result of a truncation effect, i.e., participants with both higher BMI and higher genetic predisposition to insulin resistance had a higher risk of T2D and were preferentially excluded from the sample (3). However, we observed the same association of the score with lower BMI in the incident cases of T2D in the EPIC-InterAct study, suggesting that this association is not wholly attributable to truncation effects.
A limitation of our approach is that we cannot ascribe a specific direction to the associations of the score with insulin resistance and adiposity. For example, while these loci were among the top signals in a genome-wide association study of fasting insulin, it is unclear whether they have a primary association with insulin resistance or with adipocyte function. Indeed, a variant near IRS1 is included in this list, and while absence of IRS1 is known to result in insulin resistance through impaired insulin signal transduction (39), IRS1 also influences adipocyte differentiation (40). Indeed, IRS1 was previously associated with body fat percentage (6), where the allele associated with higher body fat percentage was associated with a favorable metabolic profile, including lower risk of T2D and cardiovascular disease. Here we see the same pattern of association for our genetic risk score and thereby highlight a number of genetic variants that may be influential in the ability to store surplus lipids optimally. The association with insulin resistance was apparently paradoxically accompanied by lower body fat percentage and lower BMI. However, these results are paradoxical only when considered in the context of the observational epidemiological association between higher adiposity and insulin resistance. Our results, in combination with observations from individuals with monogenic lipodystrophy, suggest that these loci may have primary effects on subcutaneous adipocyte function, which then results in insulin resistance via ectopic lipid deposition. While we perform analyses using a combined genetic score, our findings implicate each of these loci in the etiology of insulin resistance and body fat distribution. This conclusion is supported by findings in an accompanying article (41) that independently identify the same variants in our score as being associated with a “monogenic lipodystrophy-like” phenotype using a hypothesis-free clustering approach. The inclusions of PPARG and IRS1 in the insulin resistance score further highlight the likely role of adipocyte function in their associations with insulin resistance. However, even after removing PPARG and IRS1 variants from the insulin resistance score, we observed consistent associations with body fat distribution. Furthermore, LYPLAL1, GRB14, and RSPO3 have been associated with waist-to-hip ratio at genome-wide levels of significance (42).
Conclusions
Genetic scores for insulin resistance and secretion based on common variants are valid tools to study the role of these features in a range of disease processes. In particular, while insulin resistance as a cause of T2D is largely considered to be a consequence of obesity, we highlight the role of polygenic insulin resistance in the development of T2D independent of body size. Furthermore, the association of these variants with lower subcutaneous fat mass and suggestion of ectopic fat deposition highlights the role of impaired adipose expandability and body fat distribution in T2D even among lean individuals.
Article Information
Acknowledgments. The authors are grateful to all the volunteers and to the staff of St. Mary’s Street Surgery; Ely; the study team; all the volunteers for their time and help; the general practitioners and practice staff for assistance with recruitment; the Fenland Study Investigators; the Fenland study coordination team; the Epidemiology Field, Data, and Laboratory teams; all EPIC participants and staff for their contribution to the study; the laboratory team at the MRC Epidemiology Unit for sample management; Nicola Kerrison of the MRC Epidemiology Unit for data management; the participants of the Spanish EPIC cohort for their contribution to the study; as well as to the team of trained nurses who participated in the recruitment. The RISC Study Project Management Board is B.B., F. Bonnet, S.W. Coppack, J.M.D., E.F., A. Golay, A. Mari, A. Natali, J. Petrie, and M.W.
Funding. The MRC-Ely Study was funded by the MRC (MC_U106179471) and Diabetes UK. The Fenland study is funded by the MRC (MC_U106179471) and Wellcome Trust. D.B.S. and R.K.S. are funded by the Wellcome Trust, the U.K. National Institute for Health Research (NIHR) Cambridge Biomedical Research Centre, and the MRC Centre for Obesity and Related Metabolic Disease. Genotyping in ULSAM was performed by the SNP&SEQ Technology Platform in Uppsala (www.genotyping.se), which is supported by Uppsala University; Uppsala University Hospital; Science for Life Laboratory, Uppsala; and the Swedish Research Council (contracts 80576801 and 70374401). The RISC Study was supported by European Union grant QLG1-CT-2001-01252. Funding for the EPIC-InterAct project was provided by the EU FP6 program (grant number LSHM_CT_2006_037197). In addition, EPIC-InterAct investigators acknowledge funding from the following agencies: the Swedish Research Council, the Swedish Diabetes Association, and the Swedish Heart-Lung Foundation (P.W.F.); the Swedish Research Council (L.C.G.); and the Health Research Fund (FIS) of the Spanish Ministry of Health, Murcia Regional Government (6236) (N.S. and L.A.). R.K. acknowledges funding from German Cancer Aid and the German Ministry of Research. T.J.K. acknowledges funding from Cancer Research UK. P.M.N. acknowledges funding from the Swedish Research Council. K.O. acknowledges funding from the Danish Cancer Society. S.P. acknowledges funding from Compagnia di San Paolo. J.R.Q. acknowledges funding from the Asturias Regional Government. O.R. acknowledges funding from the Västerboten County Council. A.M.W.S. and D.L.v.d.A. acknowledge funding from the Dutch Ministry of Public Health, Welfare, and Sports; the Netherlands Cancer Registry; LK Research Funds; Dutch Prevention Funds; Dutch ZON (Zorg Onderzoek Nederland); the World Cancer Research Fund; and Statistics Netherlands. R.T. acknowledges funding from AIRE-ONLUS Ragusa, AVIS-Ragusa, and the Sicilian Regional Government. Verification of diabetes cases was additionally funded by NL Agency grant IGE05012 and an incentive grant from the board of the UMC Utrecht (I.S.). I.B. acknowledges funding from Wellcome Trust grant 098051 and U.K. NIHR Cambridge Biomedical Research Centre. M.I.M. acknowledges funding from InterAct, Wellcome Trust (083270/Z/07/Z), and MRC (G0601261). E.R. acknowledges funding from Imperial College Biomedical Research.
Duality of Interest. I.B. and her spouse own stock in GlaxoSmithKline and Incyte. The RISC Study was supported by AstraZeneca. InterAct investigators acknowledge funding from Novo Nordisk (P.W.F.). No other potential conflicts of interest relevant to this article were reported.
Author Contributions. R.A.S. wrote the first draft of the manuscript, performed study-level analyses, researched/provided data, and reviewed and revised/approved the manuscript. T.F., D.P., A.B., and S.J.S. performed study-level analyses, researched/provided data, and reviewed and revised/approved the manuscript. L.A., B.B., A.B., I.B., H.B., F.C.-C., F.L.C., J.M.D., G.F., E.F., N.G.F., P.W.F., D.G., V.G., S.G., L.C.G., R.K., T.J.K., T.K., L.A.L., P.M.N., K.O., D.P., S.P., J.R.Q., O.R., N.R., C.S., N.S., M.-J.S., A.S., N.S., I.S., A.M.W.S., M.B.S., A.T., R.T., D.L.v.d.A., H.Y., M.I.M., R.K.S., E.R., M.W., E.I., and T.M.F. researched/provided data and reviewed and revised/approved the manuscript. D.B.S., C.L., and N.J.W. wrote the first draft of the manuscript, researched/provided data, and reviewed and revised/approved the manuscript. R.A.S is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Prior Presentation. Parts of this study were presented in poster form at the 74th Scientific Sessions of the American Diabetes Association, San Francisco, CA, 12–17 June 2014.