An adult-based insulin resistance genetic risk score associates with insulin resistance, metabolic traits and altered fat distribution in Danish children and adolescents who are overweight or obese

An adult-based insulin resistance genetic risk score associates with insulin resistance,... Aims/hypothesis A genetic risk score (GRS) consisting of 53 insulin resistance variants (GRS ) was recently demonstrated to associate with insulin resistance in adults. We speculated that the GRS might already associate with insulin resistance during childhood, and we therefore aimed to investigate this in populations of Danish children and adolescents. Furthermore, we aimed to address whether the GRS associates with components of the metabolic syndrome and altered body composition in children and adolescents. Methods We examined a total of 689 children and adolescents who were overweight or obese and 675 children and adolescents from a population-based study. Anthropometric data, dual-energy x-ray absorptiometry scans, BP, fasting plasma glucose, fasting serum insulin and fasting plasma lipid measurements were obtained, and HOMA-IR was calculated. The GRS was examined for association with metabolic traits in children by linear regressions using an additive genetic model. Results In overweight/obese children and adolescents, the GRS associated with higher HOMA-IR (β = 0.109 ± 0.050 (SE); −2 −2 p =2.73×10 ), fasting plasma glucose (β = 0.010 ± 0.005 mmol/l; p =2.51×10 ) and systolic BP SD score (β =0.026 ± −2 −3 0.012; p =3.32× 10 ) as well as lower HDL-cholesterol (β = −0.008 ± 0.003 mmol/l; p =1.23× 10 ), total fat-mass percentage −3 −4 (β = −0.143 ± 0.054%; p = 9.15 × 10 ) and fat-mass percentage in the legs (β = −0.197 ± 0.055%; p = 4.09 × 10 ). In the population-based sample of children, the GRS only associated with lower HDL-cholesterol concentrations (β = −0.007 ± −2 0.003 mmol/l; p =1.79× 10 ). Anne-Sofie Graae and Mette Hollensted contributed equally to this study. Electronic supplementary material The online version of this article (https://doi.org/10.1007/s00125-018-4640-0) contains peer-reviewed but unedited supplementary material, which is available to authorised users. * Torben Hansen Center for Clinical Research and Disease Prevention, Bispebjerg and torben.hansen@sund.ku.dk Frederiksberg Hospital, The Capital Region, Copenhagen, Denmark Department of Clinical Experimental Research, Rigshospitalet, Section for Metabolic Receptology, Novo Nordisk Foundation Glostrup, Denmark Center for Basic Metabolic Research, Faculty of Health and Medical Department of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark Sciences, University of Copenhagen, Copenhagen, Denmark Section of Metabolic Genetics, Novo Nordisk Foundation Center for Steno Diabetes Center, Gentofte, Denmark Basic Metabolic Research, Faculty of Health and Medical Sciences, Faculty of Health Sciences, University of Southern Denmark, University of Copenhagen, Blegdamsvej 3B, DK-2200 Odense, Denmark Copenhagen N, Denmark Section of Systems Genomics, Department of Bio and Health The Children’s Obesity Clinic, Department of Pediatrics, Informatics, Technical University of Denmark, Kongens Copenhagen University Hospital Holbæk, Holbæk, Denmark Lyngby, Denmark 1770 Diabetologia (2018) 61:1769–1779 Conclusions/interpretation An adult-based GRS comprising 53 insulin resistance susceptibility SNPs associates with insulin resistance, markers of the metabolic syndrome and altered fat distribution in a sample of Danish children and adolescents who were overweight or obese. . . . . . . . Keywords Epidemiology Genetic association Genetic risk score Genetics Insulin resistance Insulin sensitivity Obesity Paediatric obesity Weight regulation Abbreviations populations, obesity strongly associates with alterations in DXA Dual-energy x-ray absorptiometry glucose metabolism, and impaired glucose metabolism and FDR False discovery rate IR are observed not only in adults, but also in a large fraction GRS Genetic risk score of children with obesity [3, 4]. Furthermore, as increased BMI GRS Genetic risk score comprising 53 SNPs known to during childhood strongly correlates with increased risk of associate with insulin resistance-related pheno- developing type 2 diabetes in adulthood [5], it is of great types in adults importance to identify risk factors potentially influencing or IR Insulin resistance even mediating the link between childhood obesity and adult SDS SD score type 2 diabetes. Examination of factors predisposing to IR in TCOCT The Children’s Obesity Clinic’sTreatment clinical subsets of children who are overweight or obese is thus Protocol particularly relevant. If individuals predisposed to IR are TDCOB The Danish Childhood Obesity Biobank identified early in life, targeted measures could potentially delay the development of IR, and thereby potentially the development of cardiovascular disease and type 2 diabetes Introduction later in life. A genetic component of the development of IR is evident Individuals diagnosed with the metabolic syndrome have an from genome-wide association studies in adults [6–8]. increased risk of developing cardiometabolic diseases such as Recently, 53 genetic loci associated with IR-related pheno- types, i.e. fasting insulin adjusted for BMI, and circulating cardiovascular disease and type 2 diabetes [1]. Insulin resis- tance (IR), a complex metabolic condition with both genetics concentrations of triacylglycerol and HDL-cholesterol, were identified through an integrative genomic approach [9]. A ge- and the environment as contributing factors, has been sug- gested to be the primary mediator of the metabolic syndrome netic risk score (GRS) comprising the 53 lead SNPs from the identified loci associated with IR (GRS )[9], as based on [1]. Obesity, especially visceral obesity [2], is a crucial factor in the development of IR, and the prevalence of obesity is rapidly measures from a euglycaemic–hyperinsulinaemic clamp, an insulin suppression test and an insulin sensitivity index from increasinginbothchildrenandadults.Inpaediatric Diabetologia (2018) 61:1769–1779 1771 a frequently sampled OGTT [10, 11]. Furthermore, the GRS as previously described [19, 20]. All participants without pre- associated with lower BMI, body fat percentage and leg, arm viously diagnosed diabetes underwent a standardised 75 g glu- and gynoid fat mass [9]. GRSs for type 2 diabetes are reported cose OGTT, from which participants were diagnosed with type to have a higher predictive value in younger individuals than 2 diabetes according to the WHO 1999 criteria. No individuals older ones [12–17], probably due to a higher impact of genetic with previously diagnosed or screen detected type 2 diabetes vs environmental factors in youth. The same concept may hold were included in the present study. true for a GRS for IR, which could potentially be used as a clinical tool in the identification of children for whom early Genotyping This was performed on 5255 participants from the intervention might be especially relevant. Previously, a GRS Inter99 cohort, using the Illumina HumanOmniExpress-24 relating to IR based on only ten SNPs was reported to have no v1.0_A and HumanOmniExpress-24 v1.1_A (Illumina, San association with IR in 1076 children with obesity [18], yet the Diego, CA, USA). Genotypes were called using the associations of a GRS comprising the newly identified 53 Genotyping module (version 1.9.4) of GenomeStudio soft- SNPs in a paediatric population remain unknown. ware (version 2011.1; Illumina). Only individuals having a The aims of this study were therefore to investigate the call rate ≥98% were included. Genotypes were phased using following in a sample of Danish children and adolescents Eagle on autosomes and Shapeit on chromosome X and im- who were overweight or obese, as well as in a population- puted in the Phase 3 1KG and HRC1.1 using the Michigan based sample: (1) the associations of the GRS with estimates imputation server (https://imputationserver.sph.umich.edu/ of IR phenotypes (fasting concentrations of insulin, index.html)[21]. All variants included in this study were in triacylglycerols and HDL-cholesterol) and HOMA-IR; (2) Hardy–Weinberg equilibrium (p >0.05). the SNP-specific effects on these phenotypes; and (3) the as- sociations of the GRS with other components of the meta- The Danish Childhood Obesity Biobank study bolic syndrome and body composition. The GRS was ini- population tially validated in an adult Danish population. Clinical data on Danish children and adolescents was obtained from The Danish Childhood Obesity Biobank (TDCOB; Methods ClinicalTrials.gov NCT00928473). Between March 2007 and March 2013, 1069 children and adolescents (aged 6–18 years) Inter99 study population who were overweight or obese were recruited from the Children’s Obesity Clinic, Department of Pediatrics, Clinical data for adults were obtained from the Inter99 study Copenhagen University Hospital Holbæk as part of the (ClinicalTrials.gov NCT00289237). The Inter99 study is a Children’s Obesity Clinic’s Treatment Protocol (TCOCT) population-based, randomised, non-pharmacological interven- [22](see Table 1 for clinical characteristics). In the following tion study for the prevention of ischaemic heart disease, con- sections, this study sample will be referred to as the TCOCT ducted by the Research Centre for Prevention and Health, sample. Overweight was defined as a BMI above the 90th Glostrup University Hospital, Glostrup, Denmark. Of 13,016 percentile for sex and age according to Danish BMI charts individuals (aged 30–60 years) randomly selected from the [23] (corresponding to a BMI SD score (SDS) >1.28). All Civil Registration System and invited to participate, 6784 measures included in this study were obtained at the first visit (52%) participated in baseline examinations. Detailed pheno- to the clinic, i.e. before treatment initiation. Between typic characteristics from Inter99 have previously been pub- September 2010 and March 2013, a population-based sample lished [19], and baseline characteristics are presented in Table 1. of 719 children and adolescents (6–18 years) were recruited from local schools and high schools (see Table 1 for clinical Anthropometric measurements While wearing light indoor characteristics). In the following sections, this study sample clothes and no shoes, height (cm) and weight (kg) were mea- will be referred to as the population-based control sample. sured, and BMI was calculated as weight (kg) divided by height squared (m ). Waist and hip circumference were mea- Anthropometric measurements With participants wearing sured in cm, and WHR was calculated as waist measurement light indoor clothes and no shoes, height was measured by a (cm) divided by hip measurement (cm). stadiometer (to the nearest 1 mm), and weight was measured on a digital scale (to the nearest 0.1 kg). BMI was calculated as BP This was measured using a mercury sphygmomanometer. the weight (kg) divided by the height squared (m ), and BMI SDS was calculated using the least mean squares method [24] Blood sampling Blood samples were drawn following a 12 h based on a Danish reference [23]. Waist circumference was overnight fast, and measures of insulin, blood glucose, HDL- measured at umbilical level in the upright position after exha- cholesterol, triacylglycerol and total cholesterol were obtained lation using a stretch-resistant tape (to the nearest 5 mm). The 1772 Diabetologia (2018) 61:1769–1779 Table 1 Clinical characteristics of participants Variable TDCOB Inter99 TCOCT sample Population-based control sample n Median (interquartile range) n Median (interquartile range) n Median (interquartile range) Sex (male/female) 2565/2690 306/383 – 266/409 – Age (years) 5255 45.1 (39.9–50.2) 689 11.65 (9.7–13.9) 675 12.54 (10.1–15.2) Biochemical measures HOMA-IR 5063 1.36 (0.9–2.0) 689 4.34 (2.8–6.4) 675 2.23 (1.7–6.4) Fasting serum insulin (pmol/l) 5065 34 (23–50) 689 115.00 (77.0–166.5) 675 61.24 (45.5–83.9) Fasting plasma glucose (mmol/l) 5251 5.4 (5.1–5.8) 689 5.10 (4.9–5.4) 675 4.97 (4.8–5.2) HbA (mmol/mol) 5249 39.9 (36.6–43.2) 686 34.00 (33.0–37.0) 636 35.00 (33.0–36.0) 1c HbA (%) 5249 5.8 (5.5–6.1) 686 5.3 (5.2–5.5) 636 5.4 (5.2–5.4) 1c Fasting plasma LDL-cholesterol (mmol/l) 5186 3.42 (2.8–4.1) 680 2.50 (2.1–3.0) 635 2.10 (1.8–2.5) Fasting plasma HDL-cholesterol (mmol/l) 5247 1.40 (1.2–1.7) 680 1.20 (1.0–1.4) 635 1.50 (1.3–1.7) Fasting plasma total cholesterol (mmol/l) 5246 5.4 (4.8–6.2) 680 4.20 (3.7–4.8) 635 3.90 (3.6–4.5) Fasting plasma triacylglycerol (mmol/l) 5251 1.0 (0.8–1.5) 680 0.90 (0.7–1.3) 635 0.60 (0.5–0.9) Anthropometrics BMI (kg/m ) 5253 25.39 (23.1–28.3) NA NA NA NA BMI SDS NA NA 689 2.92 (2.5–3.3) 675 0.34 (−0.4–1.0) WHR 5252 0.85 (0.8–0.9) 664 0.98 (0.9–1.0) 672 0.83 (0.8–0.9) BP Systolic BP (mmHg) 5254 130 (120–140) NA NA NA NA Diastolic BP (mmHg) 5253 80 (75–90) NA NA NA NA Systolic BP SDS NA NA 660 1.55 (0.8–2.6) 642 1.57 (0.8–2.6) Diastolic BP SDS NA NA 660 0.59 (0.1–1.2) 642 0.42 (−0.1–0.9) DXA Fat mass, total (%) NA NA 391 43.96 (40.4–47.6) 44 25.26 (22.2–31.5) Fat mass, torso (%) NA NA 391 45.18 (40.5–49.9) 44 20.27 (16.9–28.5) Fat mass, legs (%) NA NA 391 45.63 (42.3–48.8) 44 31.00 (27.5–36.6) Fat mass, arms (%) NA NA 391 46.02 (42.4–50.0) 44 31.76 (24.9–36.4) Diabetologia (2018) 61:1769–1779 1773 WHR was calculated as the waist measurement (cm) divided (electronic supplementary material [ESM] Table 1). All geno- by the hip measurement (cm). types were retrieved from the imputed dataset, and GRS con- struction was therefore based on genotype dosage informa- BP Systolic and diastolic BP were measured with an tion. We constructed an unweighted GRS by summing the oscillometric device (Omron 705IT; Omron Healthcare, number of IR phenotype-increasing alleles. In addition, we Kyoto, Japan) with the appropriate cuff size, as validated in constructed a weighted GRS by summing the number of IR children [25]. BP was measured three times on the right upper phenotype-increasing alleles weighted by the effect size of the arm after 5 min of rest. An average of the last two measurements variants on fasting insulin concentrations adjusted for BMI, as was used to calculate systolic and diastolic BP SDS based on reported in the validation study in adults [9], and normalised sex-, age- and height-specific American references [25]. by dividing by the sum of all effects, to make the two GRSs comparable. Similar results were obtained from the two Blood sampling Blood samples were drawn from an antecubital GRSs, and therefore only results from the unweighted GRS vein after an overnight fast. Whole-blood HbA was analysed (GRS ) are reported. 1c 53 on a Tosoh HPLC G8 analyser (Tosoh Corporation, Tokyo, Japan). Plasma glucose was measured on a Dimension Vista Statistical analyses 1500 Analyser (Siemens Healthcare, Erlangen, Germany), and serum insulin, plasma cholesterol, plasma HDL-cholesterol and Only children and adolescents from the TDCOB cohort with plasma triacylglycerol on a Cobas 6000 Analyser (Roche available information on HOMA-IR were included in our Diagnostics, Mannheim, Germany). analyses (n = 1364). For clinical characteristics of study participants included in the analyses, see Table 1. All statistical Dual-energy x-ray absorptiometry Measurements taken using analyses were performed with and without the inclusion of dual-energy x-ray absorptiometry (DXA) included fat mass in participants who had conditions or were receiving medication the arms, legs, torso and whole body. Measures were performed potentially influencing IR, such as long-term present or prior using a GE Lunar Prodigy (DF+10031 GE Healthcare, Little systemic use of steroid hormones (n = 29). Statistical analyses Chalfont, UK) until 14 October 2009 and thereafter using a GE were performed using R software (version 3.1.3; R Foundation Lunar iDXA (ME+200,179; GE Healthcare). for Statistical Computing, Boston, MA, USA). HOMA-IR was calculated as ([fasting plasma glucose (mmol/l)] × [fasting Genotyping DNA was extracted at LGC Genomics serum insulin (pmol/l)])/135 [28]. LDL-cholesterol was calcu- (Teddington, UK), and samples from all participants (n = lated according to the Friedewald formula: [LDL-cholesterol 1788) were genotyped using the Illumina Infinium (mmol/l) = total cholesterol (mmol/l) − HDL-cholesterol HumanCoreExome Beadchip (Illumina, San Diego, CA, (mmol/l) − triacylglycerol (mmol/l)/5] [29]. Associations USA) using Illumina’s HiScan system at the Novo Nordisk between the GRS and IR, metabolic traits and body composi- Foundation Center for Basic Metabolic Research’s laboratory, tion estimates were examined by linear regression using Symbion, Copenhagen, Denmark. Genotypes were called additive genetic models. Analyses were adjusted for sex and using the Genotyping module (version 1.9.4) of age where indicated, and all analyses of DXA measures were GenomeStudio software (version 2011.1; Illumina). We ex- adjusted for type of DXA scanner. Quantitative traits deviating cluded individuals who were duplicates or ethnic outliers, or from normal distribution were log-transformed (log ) to ensure had extreme inbreeding coefficients, mislabelled sex or a call a normal distribution as assumed in the model. For log- rate of <95%, leaving 1618 individuals. Additional genotypes transformed traits, the corresponding p values are reported. were imputed using the 1000 genomes phase 1 panel using Furthermore, clinically interpretable effect sizes and SEs from shapeit/IMPUTE2 pipeline (http://mathgen.stats.ox.ac.uk/ the analyses of untransformed traits are reported. Binominal impute/impute_v2.html)[26, 27], with only genotyped tests were performed to assess the directionality of SNP- variants that were not significant (p >0.05) in Hardy– specific effects. Differences in effect sizes between groups were Weinberg equilibrium tests. Only variants with a high assessed using a standard two-tailed t-test with β values and imputation quality (IMPUTE2 estimated R > 0.95) were kept. SEs for each group. Correction for multiple testing was performed using a false discovery rate (FDR) of 10% [30]. GRS construction Values of p < 0.05 were considered statistically significant. Genotypes were coded according to the number of IR- Ethical aspects increasing alleles based on 53 independent SNPs shown to associate with IR phenotypes (higher fasting insulin concen- Written informed consent was obtained from all participants. trations adjusted for BMI, lower HDL-cholesterol concentra- If they were younger than 18 years, informed oral consent was tions and higher triacylglycerol concentrations) in adults [9] given by the participant while the parents provided informed 1774 Diabetologia (2018) 61:1769–1779 Table 2 Association between GRS and IR and related metabolic traits Variable TCOCT sample Population-based control sample Difference in GRS effect size between n β ±SE p value p value n β ±SE p value p value groups (FDR 10%) (FDR 10%) p t test Biochemical measures −2 a,b a,b HOMA-IR 689 0.109 ± 0.050 2.73 × 10 * 0.09 675 0.021 ± 0.012 0.10 0.59 0.22 a,b a,b Fasting serum insulin (pmol/l) 689 2.250 ± 1.114 0.05 0.12 675 0.495 ± 0.307 0.11 0.59 0.27 −2 a,b a,b Fasting plasma glucose (mmol/l) 689 0.010 ± 0.005 2.51 × 10 * 0.09 675 0.003 ± 0.003 0.40 0.70 0.35 b b HbA (mmol/mol) 686 0.016 ± 0.028 0.58 0.66 636 0.005 ± 0.025 0.86 0.97 0.67 1c a,b a,b Fasting plasma LDL-cholesterol (mmol/l) 680 0.007 ± 0.007 0.37 0.46 635 0.004 ± 0.005 0.49 0.70 0.61 −3 a,b −3 −2 a,b Fasting plasma HDL-cholesterol (mmol/l) 680 −0.008 ± 0.003 1.23 × 10 ** 8.00 × 10 ** 635 −0.007 ± 0.003 1.79 × 10 * 0.29 0.68 a,b a,b Fasting plasma total cholesterol (mmol/l) 680 0.003 ± 0.007 0.73 0.78 635 −0.002 ± 0.006 0.55 0.70 0.54 a,b a,b Fasting plasma triacylglycerol (mmol/l) 680 0.007 ± 0.005 0.23 0.34 635 0.001 ± 0.003 0.94 0.97 0.46 Anthropometrics BM SDS 689 −0.007 ± 0.006 0.26 0.35 675 0.0003 ± 0.009 0.97 0.97 0.54 −5 −4 b b WHR 664 −6.267 × 10 ± 6.655 × 10 0.93 0.93 672 0.0004 ± 0.001 0.48 0.70 0.60 BP −2 Systolic BP SDS 660 0.026 ± 0.012 3.32 × 10*0.09 642 −0.011 ± 0.012 0.35 0.70 0.17 Diastolic BP SDS 660 0.013 ± 0.008 0.09 0.14 642 −0.007 ± 0.006 0.23 0.70 0.32 DXA −3 b −2 b Fat mass, total (%) 391 −0.143 ± 0.054 9.15 × 10 ** 4.80 × 10 *44 −0.140 ± 0.203 0.49 0.70 1.00 b b Fat mass, torso (%) 391 −0.116 ± 0.066 0.08 0.14 44 −0.156 ± 0.213 0.47 0.70 0.71 −4 b −3 b Fat mass, legs (%) 391 −0.197 ± 0.055 4.09 × 10 *** 6.40 × 10 ** 44 −0.133 ± 0.231 0.57 0.70 0.62 b b Fat mass, arms (%) 391 −0.116 ± 0.062 0.06 0.12 44 −0.149 ± 0.258 0.57 0.70 0.76 Results are shown for the unweighted GRS. Effect sizes and SEs were calculated using untransformed variables. Values for p were calculated using untransformed variables unless otherwise indicated. All DXA scan analyses were adjusted for type of scanner p value calculated using log-transformed (base 10) variables Analyses adjusted for age and sex *p < 0.05, **p < 0.01 and ***p < 0.001 Diabetologia (2018) 61:1769–1779 1775 written consent. The study was approved by the Danish Data SNP-specific associations with IR phenotypes Protection Agency (REG-06-2014), the Ethics Committee of Region Zealand, Denmark (SJ-104) and the Scientific Ethics When calculating the individual association of each SNP with Committee of the Capital Region of Denmark (KA98155). HOMA-IR, fasting insulin, HDL-cholesterol and triacylglyc- The study was performed in accordance with the Declaration erol, associations were identified for three, three, five and four of Helsinki 2013 and is registered at ClinicalTrials.gov SNPs, respectively, in the children with obesity, whereas as- (NCT00928473 and NCT00289237). sociations were identified for four, four, four and six SNPs, respectively, in the control population (Fig. 1,ESM Figs 1–3). Consistency of the SNP-specific directionally effect sizes was Results only observed for HDL-cholesterol in the children with obe- sity, with 38 out of 53 SNPs showing negative directional −3 Validation of the association between the GRS effects (p =2.19× 10 ; ESM Table 3). In contrast, consistent and IR phenotypes in Danish adults directional effects of the included SNPs were observed for all four traits examined in the adult Inter99 population (ESM As a means of validating the association of the GRS with IR Table 3). phenotypes in an adult Danish population, we used data from the Inter99 study population comprising 5255 non-diabetic Association between the GRS and metabolic traits Danish individuals (aged 30–60 years). The Inter99 popula- tion was not part of the study by Lotta et al [9], and Inter99 When investigating whether the GRS associated with addi- therefore constitutes an adult study sample suitable for vali- tional traits related to the metabolic syndrome, the GRS dating the association between the GRS and IR phenotypes. associated with higher fasting plasma glucose concentrations −2 In the Danish adults, the unweighted GRS associated with (β = 0.010 ± 0.005 mmol/l; p = 2.51 × 10 ) in the TCOCT fasting concentrations of insulin adjusted for BMI (β =0.014 sample (Table 2), whereas no association was observed in −7 ±0.003; p =2.25 ×10 ), triacylglycerol (β = 0.022 ± 0.003; the population-based control sample (p =0.40) (Table 2). −12 p = 1.14 × 10 ) and HDL-cholesterol (β = −0.018 ± 0.003; Although no associations between the GRS and BMI SDS, −10 p =4.86×10 ) (ESM Table 2) with effect sizes similar to WHR, HbA or diastolic BP SDS were found in either of the 1c those previously reported [9]. Furthermore, the unweighted populations, an association between the GRS and the GRS associated with HOMA-IR in the Danish adults (β = systolic BP SDS was identified in the TCOCT sample (β = −7 −2 0.014 ± 0.003; p =1.45×10 ) (ESM Table 2). Similar results 0.026 ± 0.012; p =3.32×10 )(Table 2). No difference in the were obtained for the weighted GRS . effect size of the GRS between the two study populations 53 53 was identified (Table 2). Association of the GRS with IR phenotypes in children and adolescents Association between GRS and measures of fat deposition All statistical tests were performed with and without the in- clusion of individuals (n = 29) who had conditions or were In the TCOCT sample, we observed an association between receiving medication potentially influencing IR, but the re- the GRS and lower total fat-mass percentage (β = −0.143 ± −3 sults obtained did not differ (data not shown). Only results 0.054%; p =9.15× 10 )(Table 2). We also investigated fat- based on the inclusion of these individuals are thus provided. mass percentages for specific parts of the body, such as the In the TCOCT sample, the GRS associated with HOMA-IR arms, legs and torso. In the TCOCT sample, the GRS asso- 53 53 −2 (β = 0.109 ± 0.050; p =2.73×10 ); however, no association ciated with reduced leg fat-mass percentage (β = −0.197 ± −4 was observed in the population-based control sample (p = 0.055%; p =4.09× 10 )(Table 2). In the population-based 0.10) (Table 2). The GRS was inversely associated with control sample, no associations between GRS and measures 53 53 HDL-cholesterol in both the TCOCT sample (β = −0.008 ± of fat deposition were observed, yet the effect size of the −3 0.003 mmol/l; p =1.23×10 ) and the population-based con- GRS did not differ between the examined study populations, −2 trol sample (β = −0.007 ± 0.003 mmol/l; p =1.79×10 ) as determined by a two-tailed t test (Table 2). (Table 2). The GRS did not associate with concentrations of fasting insulin or triacylglycerol in either population (p > 0.05) (Table 2). Despite the fact that the observed associations Discussion of the GRS were greater in the TCOCT sample, no statisti- cally significant difference in the effect size of the GRS in In this study, we sought to investigate whether the associations the two groups was identified, as determined by a two-tailed t of a GRS previously associated with IR phenotypes (fasting test (Table 2). insulin adjusted for BMI, HDL-cholesterol and 1776 Diabetologia (2018) 61:1769–1779 SYN2/PPARG rs308971 (G) *† Fig. 1 SNP-specific associations ARL15/FST rs4865796 (A) * with HOMA-IR in children from ZMYND15 rs754814 (T) * the TCOCT sample (black lines) MEOX2 rs17169104 (G) and population-based control MYO1F rs4804311 (A) sample (grey lines). For each DMRTA2 rs17386142 (C) SNP, the name of the nearest FAM13A rs3822072 (A) gene, rs number and risk allele as RSPO3 rs2745353 (T) reported [9] are provided. *p < PIK3R1 rs4976033 (G) EBF1 rs2434612 (G) † 0.05 for the association between MAP2K7 rs4804833 (A) the given SNP and HOMA-IR in CPEB4 rs966544 (G) the TCOCT sample. p <0.05 for CCDC92/DNAH10 rs7973683 (C) the association between the given ANKRD55 rs459193 (G) SNP and HOMA-IR in the TNFAIP8 rs1045241 (C) population-based sample KLF14 rs972283 (G) PPP1R3B rs2126259 (T) PDGFC rs6822892 (A) ADCY5 rs9881942 (A) DNM3 rs9425291 (A) COBLL1/GRB14 rs10195252 (T) RNU5F-1/LYPLAL1 rs4846565 (G) † INSR rs8101064 (T) MCC rs6887914 (C) PLA2G6 rs132985 (C) LINC02537 rs6937438 (A) PEPD rs731839 (G) † LIPG rs7227237 (C) MACROD1 rs11231693 (A) MACF1 rs683135 (A) DOK7 rs2699429 (C) MSL2 rs645040 (T) LINC01625 rs3861397 (G) CSF1 rs11577194 (T) LPL rs1011685 (C) MIR548H3 rs498313 (A) C15orf54 rs7176058 (A) TMEM110-MUSTN1 rs11130329 (A) USP37 rs492400 (T) ANKS1A rs12525532 (T) EYA1 rs4738141 (G) KLHL18 rs295449 (A) IRS1 rs2943645 (T) ANKRD10 rs7323406 (A) COL6A4P1 rs3864041 (T) NRBF2 rs10995441 (G) TRIB1 rs7005992 (C) ITPR2 rs718314 (G) L3MBTL3 rs9492443 (C) EYA2 rs6066149 (G) CEP68 rs2249105 (A) ADPGK-AS1 rs8032586 (C) ATF7IP rs17402950 (G) -0.05 -0.025 0 0.025 0.05 Protecting Predisposing β value and 95% CI per risk allele triacylglycerol) in adults [9] are already evident during child- associated with higher HOMA-IR, but no association was hood and adolescence. We initially demonstrated that the identified in the population-based control sample. GRS also associated with these phenotypes and the Previously, the 53 SNPs included in the GRS have all been HOMA-IR in Danish adults. We then proceeded to examine independently associated with IR in adults [9]. In the current whether these associations between the GRS and IR pheno- study, however, only a few of the 53 SNPs displayed associ- types could be identified in two samples of Danish children ations with either HOMA-IR, fasting insulin, HDL- and adolescents: one sample that was population-based, and cholesterol or triacylglycerol in both Danish children and the other comprising children and adolescents who were over- adults, a discrepancy which may be due to the larger statistical weight/obese, thereby representing a clinical subset with ele- power of the original study [9]. Although the majority of the vated risk of obesity-induced IR. In the TCOCT sample com- included loci exhibited same-directional effect sizes, com- prising overweight/obese children and adolescents, the GRS pared with the reports made in the discovery study [9], formal 53 Diabetologia (2018) 61:1769–1779 1777 tests of directional effects only reached statistical significance circumference in adults [9]; however, no associations for HDL-cholesterol in the children with overweight/obesity, between the GRS and these traits were identified in either and for all the examined traits in the Danish adults. This same- of our child populations. directional effect was more evident in the Danish adult popu- When examining the DXA-derived measures of fat de- lation than the child populations, suggesting that the associa- position, the GRS showed an association with lower total tion of the SNPs may be stronger in adulthood than childhood fat percentage and leg fat-mass percentage, corresponding and may thus be age-dependent. Nevertheless, as the current to the findings in adults [9]. IR, lower fat mass and ectopic study has only limited power to detect SNP-specific associa- fat deposition associate with the GRS in adults [9]. As the tions, larger study populations of both children and adults GRS in our study associates with both HOMA-IR and seem necessary to elucidate whether the included SNPs do lower fat-mass percentage, it is likely that the GRS may in fact display age-dependent effects. associate with ectopic lipid deposition, a potential cause of In a recent study by Morandi et al comprising 1076 chil- IR, in children as well. A more detailed analysis of lipid dren with obesity (mean age 11.4 years) and 1265 young accumulation in liver and skeletal muscles is needed to val- adults with normal weight (mean age 21.1 years), no associa- idate this hypothesis. tions between a GRS comprising ten IR-associated SNPs and Although we observed more associations between the OGTT-derived traits (fasting insulin, fasting glucose and GRS and IR-related phenotypes in the sample of children HOMA-IR) were identified [18]. This discrepancy in relation with obesity, the identified effect estimates of the GRS in to our results may be explained by the difference in the SNPs the two groups of children did not differ. With our statistical included in the GRS. Morandi et al used in their GRS ten power, we can therefore not claim that the associations of SNPs originally chosen by Vassy et al [17] because of their the GRS differ between our two populations. association with HOMA-IR, higher fasting insulin or IR- Nevertheless, our results still indicate that the GRS has a related traits such as lower HDL-cholesterol or higher triacyl- greater association in children with obesity, compared with glycerol, BMI or WHR in earlier publications. The GRS a population-based sample of children. Collectively, our used in the current study consists of 53 SNPs used by Lotta results obtained in Danish children, adolescents and adults et al to create their GRS [9]. By employing a meta-analysis, suggest that the association of the GRS comprising the 53 Lotta et al included up to 188,577 individuals and identified SNPs are mediated via metabolic stresses such as obesity 630 SNPs within 53 loci associating with higher fasting insu- and ageing. Our findings indicate a different effect size of lin, lower HDL-cholesterol and higher triacylglycerol. The 53 the individual SNPs in the GRS in populations comprising lead SNPs from these loci were then compiled into the GRS. children and adolescents compared with adults. Potentially, In the same study, the association between the GRS and IR the SNP-specific associations are modified by obesity, age- in adults was validated based on previously obtained measures ing and other exposures. Our findings correspond well with from a euglycemic–hyperinsulinemic clamp or insulin sup- previous studies reporting age-dependent effects of loci as- pression test in 2764 adults, or by insulin sensitivity index in sociating with metabolic traits such as BMI and obesity 4769 individuals [10, 11] . The better validated and higher [31–33]. An age-dependent variation of the effect of the number of SNPs in the GRS used in our study may explain individual SNPs in the GRS complicates the interpretation 53 53 why the GRS , in contrast to the earlier publication [18], of data and calls for the identification of age-specific trait associates with IR despite our smaller study population. loci and subsequent construction of age- and trait-specific We found that the GRS associated with other traits of the GRSs. metabolic syndrome, namely lower HDL-cholesterol concen- Although our results must be validated in larger paediatric trations and higher systolic BP in the TCOCT sample. study populations, our findings could have clinical implica- Interestingly, the GRS was inversely associated with HDL- tions, as they suggest that children with an increased risk of IR cholesterol level in the population-based control children as may benefit from preventive and therapeutic lifestyle and/or well. In contrast, there were no associations between the pharmacological treatment approaches aiming to reduce obe- GRS and higher fasting insulin and triacylglycerol concen- sity to prevent the development of IR-associated cardiometa- trations in either population, even though associations bolic risks. Furthermore, our findings indicate that the GRS with these traits, together with HDL-cholesterol, were has potential as a clinical marker to aid the identification of originally used to identify the susceptibility SNPs in adults. children with a higher risk of IR than that mediated via obesity Unfortunately, concentrations of fasting plasma lipids and alone. In adults, the GRS strongly associated with increased BP were not investigated by Lotta et al [9]. Furthermore, in risk of developing type 2 diabetes [9], yet it remains unknown the TCOCT sample, the GRS associated with higher whether children with a high genetic burden, as assessed by fasting plasma glucose, in consistency with the association the GRS , also have an increased risk of developing diabetes identified in adults [9]. The GRS has previously been and/or diabetes-related comorbidities later in life. Future stud- associated with lower BMI and higher WHR/hip ies examining the associations of the GRS in larger 53 1778 Diabetologia (2018) 61:1769–1779 Acknowledgements The authors wish to thank all children and adoles- populations and across a lifespan could potentially help to cents who participated in the present study, as well as their families. elucidate whether the GRS could be used as a clinical tool Additionally, we wish to thank O. Troest, B. Holløse, F. Pinar (the that would, during childhood and adolescence, already enable Children’s Obesity Clinic, Department of Pediatrics, Copenhagen the identification of individuals with an increased risk of IR University Hospital Holbæk, Denmark) for careful laboratory assistance. Finally, we thank A. Forman, T. H. Lorentzen and G. J. Klavsen for their and ultimately type 2 diabetes. As such, the GRS ,or even an dedicated laboratory assistance, P. Sandbeck for data management, G. improved GRS comprising other or additional SNPs with val- Lademann for secretarial support, and T. F. Toldsted for grant manage- idated strong associations with IR phenotypes in childhood, ment (all from the Section of Metabolic Genetics, Novo Nordisk could potentially be one of the first steps towards personalised Foundation Center for Basic Metabolic Research, Faculty of Health and Medical Sciences, University of Copenhagen, Denmark). intervention programmes aiming to minimise the occurrence of cardiovascular events and preterm death associated with Data availability The datasets generated during and/or analysed during early-onset type 2 diabetes [34]. the current study are available from the corresponding author on reason- Our study is limited by the relatively small study sample, able request. which reduces the statistical power. Furthermore, analyses Funding The study was funded by Innovation Fund Denmark (grants were not adjusted for stage of puberty but only for age. 0603-00484B and 0603-00457B), the Danish Diabetes Academy and Nevertheless, a very detailed phenotypic characterisation the Region Zealand Health Scientific Research Foundation. This study was available for the included study population, enabling the is part of the TARGET (The Impact of our Genomes on Individual detailed analysis of several traits related to IR and fat deposi- Treatment Response in Obese Children, http://target.ku.dk)and BioChild (Genetics and Systems Biology of Childhood Obesity in India tion. Furthermore, data for the examined traits were available and Denmark, http://biochild.ku.dk) consortia studies, as well as The for two populations of children with similar age spans yet Danish Childhood Obesity Biobank. The Novo Nordisk Foundation different ranges of BMI SDS, enabling an evaluation of the Center for Basic Metabolic Research is an independent research centre effect of obesity on the effect of the GRS. It should be noted at the University of Copenhagen, partially funded by an unrestricted donation from the Novo Nordisk Foundation (www.metabol.ku.dk). that the two populations of children were selected in different The Inter99 study was initiated by T. Jørgensen (PI; Research Centre ways: the children with obesity were highly selected accord- for Prevention and Health, Glostrup University Hospital, Glostrup, ing to their BMI and age, whereas the children from the Denmark), K. Borch-Johnsen (co-PI; Steno Diabetes Center A/S, Gentofte, Denmark), T. Thomsen (Research Centre for Prevention and population-based group were selected only according to age. Health, Glostrup University Hospital, Glostrup, Denmark) and H. Ibsen This discrepancy in the selection of study participants may (Division of Cardiology, Holbæk University Hospital, Holbæk, potentially affect our findings. Denmark). The present steering committee comprises T. Jørgensen and In conclusion, we investigated whether the GRS asso- 53 C. Pisinger (Research Centre for Prevention and Health, Glostrup University Hospital, Glostrup, Denmark). ciates with IR phenotypes and HOMA-IR in both children and adolescents who are overweight or obese, and in a Duality of interest The authors declare that there is no duality of interest population-based control sample. A GRS associating with associated with this manuscript. IR in children could help to identify children predisposed to IR. In overweight or obese children and adolescents, the 53 Contribution statement ASG, MH, JTK, TRHN, AL, MEJ, JCH and TH contributed to the conception and design of the study as well as acquisi- SNPs cumulatively associate with IR. The results indicate tion of data. MH, EVRA, NG, HNK, OP and TH planned and performed that children who have a genetic predisposition to IR, as the acquisition of genotypes, and TMS, EVRA and YM constructed the assessed by the 53 SNPs, will have a higher risk of devel- GRSs. MH, JR, MØJ, NG and TH planned and performed the statistical oping IR if they become overweight or obese. However, as analyses, while ASG, MH, JR, JTK, TRHN, NG, BH and TH interpreted the data. ASG and MH wrote the initial draft, while all authors contrib- no difference between the effects size of the GRS in the uted to the critical revision of the draft. The final draft was commented two groups of children could be identified, we cannot with upon and approved by all authors. TH is the guarantor of this work. certainty conclude that obesity is essential for the associa- tion between HOMA-IR and the GRS . This hypothesis Open Access This article is distributed under the terms of the Creative needs to be verified in a larger population. The identifica- Commons Attribution 4.0 International License (http:// tion of additional SNPs displaying strong associations with creativecommons.org/licenses/by/4.0/), which permits unrestricted use, IR-related phenotypes during childhood would increase the distribution, and reproduction in any medium, provided you give appro- priate credit to the original author(s) and the source, provide a link to the clinical impact of the GRS and allow the identification of Creative Commons license, and indicate if changes were made. children predisposed to IR. Treatment strategies targeted against factors important for the development of IR, such References as obesity, could be developed specifically for predisposed children. Furthermore, our study showed that fat percentage 1. International Diabetes Federation (2006) The IDF consensus world- in the body extremities was inversely associated with wide definition of the Metabolic Syndrome. Available from http:// GRS in the children and adolescents who are overweight www.idf.org/e-library/consensus-statements/60-idfconsensus- or obese, suggesting that impaired capacity to store fat in worldwide-definitionof-the-metabolic-syndrome. Accessed 17 Sep peripheral compartments increases the risk of IR. 2017 Diabetologia (2018) 61:1769–1779 1779 2. Roberts CK, Hevener AL, Bernard RJ (2014) Metabolic Syndrome Traits, in Obese and Normal-Weight Youths. J Clin Endocrinol Metab 101:4244–4250 and Insulin Resistance: Underlying Causes and Modification by Exercise Training. Compr Physiol 3:1–58 19. Banasik K, Hollensted M, Andersson E et al (2012) The effect of 3. Sinha R, Fisch G, Teague B et al (2002) Prevalence of impaired FOXA2rs1209523 on glucose-related phenotypes and risk of type 2 glucose tolerance among children and adolescents with marked diabetes in Danish individuals. BMC Med Genet 13:10 obesity. N Engl J Med 346:802–810 20. Jørgensen T, Borch-Johnsen K, Thomsen TF et al (2003) A ran- 4. Kloppenborg JT, Fonvig CE, Nielsen TRH et al (2017) Impaired domized non-pharmacological intervention study for prevention of fasting glucose and the metabolic profile in Danish children and ischaemic heart disease: baseline results Inter99. Eur J Cardiovasc adolescents with normal weight, overweight, or obesity. Pediatr Prev Rehabil 10:377–386 Diabetes 1–10 21. McCarthy S, Das S, Kretzschmar W et al (2016) A reference panel 5. Zimmermann E, Bjerregaard LG, Gamborg M et al (2017) of 64,976 haplotypes for genotype imputation. Nat Genet 48:1279– Childhood body mass index and development of type 2 diabetes throughout adult life—A large-scale danish cohort study. Obesity 22. Holm J-C, Gamborg M, Bille DS et al (2011) Chronic care treat- 25:965–971 ment of obese children and adolescents. Int J Pediatr Obes 6:188– 6. Dupuis J, Langenberg C, Prokopenko I et al (2010) New genetic loci implicated in fasting glucose homeostasis and their impact on 23. Nysom K, Mølgaard C, Hutchings B, Michaelsen KF (2001) Body type 2 diabetes risk. Nat Genet 42:105–116 mass index of 0 to 45-y-old Danes: reference values and compari- 7. Manning AK, Hivert M-F, Scott RA et al (2012) A genome-wide son with published European reference values. Int J Obes Relat approach accounting for body mass index identifies genetic variants Metab Disord 25:177–184 influencing fasting glycemic traits and insulin resistance. Nat Genet 24. Cole TJ, Green PJ (1992) Smoothing reference centile curves: The 44:659–669 lms method and penalized likelihood. Stat Med 11:1305–1319 8. Scott RA, Lagou V, Welch RP et al (2012) Large-scale association 25. National High Blood Pressure Education Program Working Group analyses identify new loci influencing glycemic traits and provide on High Blood Pressure in Children and Adolescents (2004) The insight into the underlying biological pathways. Nat Genet 44:991– fourth report on the diagnosis, evaluation, and treatment of high blood pressure in children and adolescents. Pediatrics 114:555–576 9. Lotta LA, Gulati P, Day FR et al (2017) Integrative genomic anal- 26. Delaneau O, Marchini J, 1000 Genomes Project Consortium (2014) ysis implicates limited peripheral adipose storage capacity in the Integrating sequence and array data to create an improved 1000 pathogenesis of human insulin resistance. Nat Genet 49:17–26 Genomes Project haplotype reference panel. Nat Commun 5:3934 10. Prokopenko I, Poon W, Mägi R et al (2014) A central role for 27. Howie BN, Donnelly P, Marchini J (2009) A flexible and accurate GRB10 in regulation of islet function in man. PLoS Genet 10: genotype imputation method for the next generation of genome- e1004235 wide association studies. PLoS Genet 5:e1000529 11. Knowles JW, Xie W, Zhang Z et al (2016) Identification and vali- 28. Matthews DR, Hosker JP, Rudenski AS et al (1985) Homeostasis dation of N-acetyltransferase 2 as an insulin sensitivity gene. J Clin model assessment: insulin resistance and beta-cell function from Invest 126:403 fasting plasma glucose and insulin concentrations in man. 12. Lyssenko V, Jonsson A, Almgren P et al (2008) Clinical risk factors, Diabetologia 28:412–419 DNA variants, and the development of type 2 diabetes. N Engl J 29. Friedewald WT, Levy RI, Fredrickson DS (1972) Estimation of the Med 359:2220–2232 concentration of low-density lipoprotein cholesterol in plasma, with- 13. De Miguel-Yanes JM, Shrader P, Pencina MJ et al (2011) Genetic out use of the preparative ultracentrifuge. Clin Chem 18:499–502 risk reclassification for type 2 diabetes by age below or above 50 30. Benjamini Y, Hochberg Y (1995) Controlling the false discovery years using 40 type 2 diabetes risk single nucleotide polymor- Rate: a practical and powerful approach to multiple testing. R Stat phisms. Diabetes Care 34:121–125 Soc 57:289–300 14. Vassy JL, Durant NH, Kabagambe EK et al (2012) A genotype risk 31. Felix JF, Bradfield JP, Monnereau C et al (2016) Genome-wide score predicts type 2 diabetes from young adulthood: The CARDIA association analysis identifies three new susceptibility loci for child- study. Diabetologia 55:2604–2612 hood body mass index. Hum Mol Genet 25:389–403 15. Vassy JL, Dasmahapatra P, Meigs JB et al (2012) Genotype predic- 32. Bradfield J, Taal H, Timpson N (2012) A genome-wide association tion of adult type 2 diabetes from adolescence in a multiracial pop- meta-analysis identifies new childhood obesity loci. Nat Genet 44: ulation. Pediatrics 130:1235–1242 526–531 16. Walford GA, Porneala BC, Dauriz M et al (2014) Metabolite traits 33. Winkler TW, Justice AE, Graff M et al (2015) The influence of age and genetic risk provide complementary information for the predic- and sex on genetic associations with adult body size and shape: a tion of future type 2 diabetes. Diabetes Care 37:2508–2514 large-scale genome-wide interaction study. PLoS Genet 11:1–42 17. Vassy JL, Hivert MF, Porneala B et al (2014) Polygenic type 2 diabetes prediction at the limit of common variant detection. 34. Zoungas S, Woodward M, Li Q et al (2014) Impact of age, age at Diabetes 63:2172–2182 diagnosis and duration of diabetes on the risk of macrovascular and 18. Morandi A, Bonnefond A, Lobbens S et al (2016) Associations microvascular complications and death in type 2 diabetes. Between Type 2 Diabetes-Related Genetic Scores and Metabolic Diabetologia 57:2465–2474 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Diabetologia Springer Journals

An adult-based insulin resistance genetic risk score associates with insulin resistance, metabolic traits and altered fat distribution in Danish children and adolescents who are overweight or obese

Free
11 pages
Loading next page...
 
/lp/springer_journal/an-adult-based-insulin-resistance-genetic-risk-score-associates-with-0hzMbDZb0h
Publisher
Springer Berlin Heidelberg
Copyright
Copyright © 2018 by The Author(s)
Subject
Medicine & Public Health; Internal Medicine; Metabolic Diseases; Human Physiology
ISSN
0012-186X
eISSN
1432-0428
D.O.I.
10.1007/s00125-018-4640-0
Publisher site
See Article on Publisher Site

Abstract

Aims/hypothesis A genetic risk score (GRS) consisting of 53 insulin resistance variants (GRS ) was recently demonstrated to associate with insulin resistance in adults. We speculated that the GRS might already associate with insulin resistance during childhood, and we therefore aimed to investigate this in populations of Danish children and adolescents. Furthermore, we aimed to address whether the GRS associates with components of the metabolic syndrome and altered body composition in children and adolescents. Methods We examined a total of 689 children and adolescents who were overweight or obese and 675 children and adolescents from a population-based study. Anthropometric data, dual-energy x-ray absorptiometry scans, BP, fasting plasma glucose, fasting serum insulin and fasting plasma lipid measurements were obtained, and HOMA-IR was calculated. The GRS was examined for association with metabolic traits in children by linear regressions using an additive genetic model. Results In overweight/obese children and adolescents, the GRS associated with higher HOMA-IR (β = 0.109 ± 0.050 (SE); −2 −2 p =2.73×10 ), fasting plasma glucose (β = 0.010 ± 0.005 mmol/l; p =2.51×10 ) and systolic BP SD score (β =0.026 ± −2 −3 0.012; p =3.32× 10 ) as well as lower HDL-cholesterol (β = −0.008 ± 0.003 mmol/l; p =1.23× 10 ), total fat-mass percentage −3 −4 (β = −0.143 ± 0.054%; p = 9.15 × 10 ) and fat-mass percentage in the legs (β = −0.197 ± 0.055%; p = 4.09 × 10 ). In the population-based sample of children, the GRS only associated with lower HDL-cholesterol concentrations (β = −0.007 ± −2 0.003 mmol/l; p =1.79× 10 ). Anne-Sofie Graae and Mette Hollensted contributed equally to this study. Electronic supplementary material The online version of this article (https://doi.org/10.1007/s00125-018-4640-0) contains peer-reviewed but unedited supplementary material, which is available to authorised users. * Torben Hansen Center for Clinical Research and Disease Prevention, Bispebjerg and torben.hansen@sund.ku.dk Frederiksberg Hospital, The Capital Region, Copenhagen, Denmark Department of Clinical Experimental Research, Rigshospitalet, Section for Metabolic Receptology, Novo Nordisk Foundation Glostrup, Denmark Center for Basic Metabolic Research, Faculty of Health and Medical Department of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark Sciences, University of Copenhagen, Copenhagen, Denmark Section of Metabolic Genetics, Novo Nordisk Foundation Center for Steno Diabetes Center, Gentofte, Denmark Basic Metabolic Research, Faculty of Health and Medical Sciences, Faculty of Health Sciences, University of Southern Denmark, University of Copenhagen, Blegdamsvej 3B, DK-2200 Odense, Denmark Copenhagen N, Denmark Section of Systems Genomics, Department of Bio and Health The Children’s Obesity Clinic, Department of Pediatrics, Informatics, Technical University of Denmark, Kongens Copenhagen University Hospital Holbæk, Holbæk, Denmark Lyngby, Denmark 1770 Diabetologia (2018) 61:1769–1779 Conclusions/interpretation An adult-based GRS comprising 53 insulin resistance susceptibility SNPs associates with insulin resistance, markers of the metabolic syndrome and altered fat distribution in a sample of Danish children and adolescents who were overweight or obese. . . . . . . . Keywords Epidemiology Genetic association Genetic risk score Genetics Insulin resistance Insulin sensitivity Obesity Paediatric obesity Weight regulation Abbreviations populations, obesity strongly associates with alterations in DXA Dual-energy x-ray absorptiometry glucose metabolism, and impaired glucose metabolism and FDR False discovery rate IR are observed not only in adults, but also in a large fraction GRS Genetic risk score of children with obesity [3, 4]. Furthermore, as increased BMI GRS Genetic risk score comprising 53 SNPs known to during childhood strongly correlates with increased risk of associate with insulin resistance-related pheno- developing type 2 diabetes in adulthood [5], it is of great types in adults importance to identify risk factors potentially influencing or IR Insulin resistance even mediating the link between childhood obesity and adult SDS SD score type 2 diabetes. Examination of factors predisposing to IR in TCOCT The Children’s Obesity Clinic’sTreatment clinical subsets of children who are overweight or obese is thus Protocol particularly relevant. If individuals predisposed to IR are TDCOB The Danish Childhood Obesity Biobank identified early in life, targeted measures could potentially delay the development of IR, and thereby potentially the development of cardiovascular disease and type 2 diabetes Introduction later in life. A genetic component of the development of IR is evident Individuals diagnosed with the metabolic syndrome have an from genome-wide association studies in adults [6–8]. increased risk of developing cardiometabolic diseases such as Recently, 53 genetic loci associated with IR-related pheno- types, i.e. fasting insulin adjusted for BMI, and circulating cardiovascular disease and type 2 diabetes [1]. Insulin resis- tance (IR), a complex metabolic condition with both genetics concentrations of triacylglycerol and HDL-cholesterol, were identified through an integrative genomic approach [9]. A ge- and the environment as contributing factors, has been sug- gested to be the primary mediator of the metabolic syndrome netic risk score (GRS) comprising the 53 lead SNPs from the identified loci associated with IR (GRS )[9], as based on [1]. Obesity, especially visceral obesity [2], is a crucial factor in the development of IR, and the prevalence of obesity is rapidly measures from a euglycaemic–hyperinsulinaemic clamp, an insulin suppression test and an insulin sensitivity index from increasinginbothchildrenandadults.Inpaediatric Diabetologia (2018) 61:1769–1779 1771 a frequently sampled OGTT [10, 11]. Furthermore, the GRS as previously described [19, 20]. All participants without pre- associated with lower BMI, body fat percentage and leg, arm viously diagnosed diabetes underwent a standardised 75 g glu- and gynoid fat mass [9]. GRSs for type 2 diabetes are reported cose OGTT, from which participants were diagnosed with type to have a higher predictive value in younger individuals than 2 diabetes according to the WHO 1999 criteria. No individuals older ones [12–17], probably due to a higher impact of genetic with previously diagnosed or screen detected type 2 diabetes vs environmental factors in youth. The same concept may hold were included in the present study. true for a GRS for IR, which could potentially be used as a clinical tool in the identification of children for whom early Genotyping This was performed on 5255 participants from the intervention might be especially relevant. Previously, a GRS Inter99 cohort, using the Illumina HumanOmniExpress-24 relating to IR based on only ten SNPs was reported to have no v1.0_A and HumanOmniExpress-24 v1.1_A (Illumina, San association with IR in 1076 children with obesity [18], yet the Diego, CA, USA). Genotypes were called using the associations of a GRS comprising the newly identified 53 Genotyping module (version 1.9.4) of GenomeStudio soft- SNPs in a paediatric population remain unknown. ware (version 2011.1; Illumina). Only individuals having a The aims of this study were therefore to investigate the call rate ≥98% were included. Genotypes were phased using following in a sample of Danish children and adolescents Eagle on autosomes and Shapeit on chromosome X and im- who were overweight or obese, as well as in a population- puted in the Phase 3 1KG and HRC1.1 using the Michigan based sample: (1) the associations of the GRS with estimates imputation server (https://imputationserver.sph.umich.edu/ of IR phenotypes (fasting concentrations of insulin, index.html)[21]. All variants included in this study were in triacylglycerols and HDL-cholesterol) and HOMA-IR; (2) Hardy–Weinberg equilibrium (p >0.05). the SNP-specific effects on these phenotypes; and (3) the as- sociations of the GRS with other components of the meta- The Danish Childhood Obesity Biobank study bolic syndrome and body composition. The GRS was ini- population tially validated in an adult Danish population. Clinical data on Danish children and adolescents was obtained from The Danish Childhood Obesity Biobank (TDCOB; Methods ClinicalTrials.gov NCT00928473). Between March 2007 and March 2013, 1069 children and adolescents (aged 6–18 years) Inter99 study population who were overweight or obese were recruited from the Children’s Obesity Clinic, Department of Pediatrics, Clinical data for adults were obtained from the Inter99 study Copenhagen University Hospital Holbæk as part of the (ClinicalTrials.gov NCT00289237). The Inter99 study is a Children’s Obesity Clinic’s Treatment Protocol (TCOCT) population-based, randomised, non-pharmacological interven- [22](see Table 1 for clinical characteristics). In the following tion study for the prevention of ischaemic heart disease, con- sections, this study sample will be referred to as the TCOCT ducted by the Research Centre for Prevention and Health, sample. Overweight was defined as a BMI above the 90th Glostrup University Hospital, Glostrup, Denmark. Of 13,016 percentile for sex and age according to Danish BMI charts individuals (aged 30–60 years) randomly selected from the [23] (corresponding to a BMI SD score (SDS) >1.28). All Civil Registration System and invited to participate, 6784 measures included in this study were obtained at the first visit (52%) participated in baseline examinations. Detailed pheno- to the clinic, i.e. before treatment initiation. Between typic characteristics from Inter99 have previously been pub- September 2010 and March 2013, a population-based sample lished [19], and baseline characteristics are presented in Table 1. of 719 children and adolescents (6–18 years) were recruited from local schools and high schools (see Table 1 for clinical Anthropometric measurements While wearing light indoor characteristics). In the following sections, this study sample clothes and no shoes, height (cm) and weight (kg) were mea- will be referred to as the population-based control sample. sured, and BMI was calculated as weight (kg) divided by height squared (m ). Waist and hip circumference were mea- Anthropometric measurements With participants wearing sured in cm, and WHR was calculated as waist measurement light indoor clothes and no shoes, height was measured by a (cm) divided by hip measurement (cm). stadiometer (to the nearest 1 mm), and weight was measured on a digital scale (to the nearest 0.1 kg). BMI was calculated as BP This was measured using a mercury sphygmomanometer. the weight (kg) divided by the height squared (m ), and BMI SDS was calculated using the least mean squares method [24] Blood sampling Blood samples were drawn following a 12 h based on a Danish reference [23]. Waist circumference was overnight fast, and measures of insulin, blood glucose, HDL- measured at umbilical level in the upright position after exha- cholesterol, triacylglycerol and total cholesterol were obtained lation using a stretch-resistant tape (to the nearest 5 mm). The 1772 Diabetologia (2018) 61:1769–1779 Table 1 Clinical characteristics of participants Variable TDCOB Inter99 TCOCT sample Population-based control sample n Median (interquartile range) n Median (interquartile range) n Median (interquartile range) Sex (male/female) 2565/2690 306/383 – 266/409 – Age (years) 5255 45.1 (39.9–50.2) 689 11.65 (9.7–13.9) 675 12.54 (10.1–15.2) Biochemical measures HOMA-IR 5063 1.36 (0.9–2.0) 689 4.34 (2.8–6.4) 675 2.23 (1.7–6.4) Fasting serum insulin (pmol/l) 5065 34 (23–50) 689 115.00 (77.0–166.5) 675 61.24 (45.5–83.9) Fasting plasma glucose (mmol/l) 5251 5.4 (5.1–5.8) 689 5.10 (4.9–5.4) 675 4.97 (4.8–5.2) HbA (mmol/mol) 5249 39.9 (36.6–43.2) 686 34.00 (33.0–37.0) 636 35.00 (33.0–36.0) 1c HbA (%) 5249 5.8 (5.5–6.1) 686 5.3 (5.2–5.5) 636 5.4 (5.2–5.4) 1c Fasting plasma LDL-cholesterol (mmol/l) 5186 3.42 (2.8–4.1) 680 2.50 (2.1–3.0) 635 2.10 (1.8–2.5) Fasting plasma HDL-cholesterol (mmol/l) 5247 1.40 (1.2–1.7) 680 1.20 (1.0–1.4) 635 1.50 (1.3–1.7) Fasting plasma total cholesterol (mmol/l) 5246 5.4 (4.8–6.2) 680 4.20 (3.7–4.8) 635 3.90 (3.6–4.5) Fasting plasma triacylglycerol (mmol/l) 5251 1.0 (0.8–1.5) 680 0.90 (0.7–1.3) 635 0.60 (0.5–0.9) Anthropometrics BMI (kg/m ) 5253 25.39 (23.1–28.3) NA NA NA NA BMI SDS NA NA 689 2.92 (2.5–3.3) 675 0.34 (−0.4–1.0) WHR 5252 0.85 (0.8–0.9) 664 0.98 (0.9–1.0) 672 0.83 (0.8–0.9) BP Systolic BP (mmHg) 5254 130 (120–140) NA NA NA NA Diastolic BP (mmHg) 5253 80 (75–90) NA NA NA NA Systolic BP SDS NA NA 660 1.55 (0.8–2.6) 642 1.57 (0.8–2.6) Diastolic BP SDS NA NA 660 0.59 (0.1–1.2) 642 0.42 (−0.1–0.9) DXA Fat mass, total (%) NA NA 391 43.96 (40.4–47.6) 44 25.26 (22.2–31.5) Fat mass, torso (%) NA NA 391 45.18 (40.5–49.9) 44 20.27 (16.9–28.5) Fat mass, legs (%) NA NA 391 45.63 (42.3–48.8) 44 31.00 (27.5–36.6) Fat mass, arms (%) NA NA 391 46.02 (42.4–50.0) 44 31.76 (24.9–36.4) Diabetologia (2018) 61:1769–1779 1773 WHR was calculated as the waist measurement (cm) divided (electronic supplementary material [ESM] Table 1). All geno- by the hip measurement (cm). types were retrieved from the imputed dataset, and GRS con- struction was therefore based on genotype dosage informa- BP Systolic and diastolic BP were measured with an tion. We constructed an unweighted GRS by summing the oscillometric device (Omron 705IT; Omron Healthcare, number of IR phenotype-increasing alleles. In addition, we Kyoto, Japan) with the appropriate cuff size, as validated in constructed a weighted GRS by summing the number of IR children [25]. BP was measured three times on the right upper phenotype-increasing alleles weighted by the effect size of the arm after 5 min of rest. An average of the last two measurements variants on fasting insulin concentrations adjusted for BMI, as was used to calculate systolic and diastolic BP SDS based on reported in the validation study in adults [9], and normalised sex-, age- and height-specific American references [25]. by dividing by the sum of all effects, to make the two GRSs comparable. Similar results were obtained from the two Blood sampling Blood samples were drawn from an antecubital GRSs, and therefore only results from the unweighted GRS vein after an overnight fast. Whole-blood HbA was analysed (GRS ) are reported. 1c 53 on a Tosoh HPLC G8 analyser (Tosoh Corporation, Tokyo, Japan). Plasma glucose was measured on a Dimension Vista Statistical analyses 1500 Analyser (Siemens Healthcare, Erlangen, Germany), and serum insulin, plasma cholesterol, plasma HDL-cholesterol and Only children and adolescents from the TDCOB cohort with plasma triacylglycerol on a Cobas 6000 Analyser (Roche available information on HOMA-IR were included in our Diagnostics, Mannheim, Germany). analyses (n = 1364). For clinical characteristics of study participants included in the analyses, see Table 1. All statistical Dual-energy x-ray absorptiometry Measurements taken using analyses were performed with and without the inclusion of dual-energy x-ray absorptiometry (DXA) included fat mass in participants who had conditions or were receiving medication the arms, legs, torso and whole body. Measures were performed potentially influencing IR, such as long-term present or prior using a GE Lunar Prodigy (DF+10031 GE Healthcare, Little systemic use of steroid hormones (n = 29). Statistical analyses Chalfont, UK) until 14 October 2009 and thereafter using a GE were performed using R software (version 3.1.3; R Foundation Lunar iDXA (ME+200,179; GE Healthcare). for Statistical Computing, Boston, MA, USA). HOMA-IR was calculated as ([fasting plasma glucose (mmol/l)] × [fasting Genotyping DNA was extracted at LGC Genomics serum insulin (pmol/l)])/135 [28]. LDL-cholesterol was calcu- (Teddington, UK), and samples from all participants (n = lated according to the Friedewald formula: [LDL-cholesterol 1788) were genotyped using the Illumina Infinium (mmol/l) = total cholesterol (mmol/l) − HDL-cholesterol HumanCoreExome Beadchip (Illumina, San Diego, CA, (mmol/l) − triacylglycerol (mmol/l)/5] [29]. Associations USA) using Illumina’s HiScan system at the Novo Nordisk between the GRS and IR, metabolic traits and body composi- Foundation Center for Basic Metabolic Research’s laboratory, tion estimates were examined by linear regression using Symbion, Copenhagen, Denmark. Genotypes were called additive genetic models. Analyses were adjusted for sex and using the Genotyping module (version 1.9.4) of age where indicated, and all analyses of DXA measures were GenomeStudio software (version 2011.1; Illumina). We ex- adjusted for type of DXA scanner. Quantitative traits deviating cluded individuals who were duplicates or ethnic outliers, or from normal distribution were log-transformed (log ) to ensure had extreme inbreeding coefficients, mislabelled sex or a call a normal distribution as assumed in the model. For log- rate of <95%, leaving 1618 individuals. Additional genotypes transformed traits, the corresponding p values are reported. were imputed using the 1000 genomes phase 1 panel using Furthermore, clinically interpretable effect sizes and SEs from shapeit/IMPUTE2 pipeline (http://mathgen.stats.ox.ac.uk/ the analyses of untransformed traits are reported. Binominal impute/impute_v2.html)[26, 27], with only genotyped tests were performed to assess the directionality of SNP- variants that were not significant (p >0.05) in Hardy– specific effects. Differences in effect sizes between groups were Weinberg equilibrium tests. Only variants with a high assessed using a standard two-tailed t-test with β values and imputation quality (IMPUTE2 estimated R > 0.95) were kept. SEs for each group. Correction for multiple testing was performed using a false discovery rate (FDR) of 10% [30]. GRS construction Values of p < 0.05 were considered statistically significant. Genotypes were coded according to the number of IR- Ethical aspects increasing alleles based on 53 independent SNPs shown to associate with IR phenotypes (higher fasting insulin concen- Written informed consent was obtained from all participants. trations adjusted for BMI, lower HDL-cholesterol concentra- If they were younger than 18 years, informed oral consent was tions and higher triacylglycerol concentrations) in adults [9] given by the participant while the parents provided informed 1774 Diabetologia (2018) 61:1769–1779 Table 2 Association between GRS and IR and related metabolic traits Variable TCOCT sample Population-based control sample Difference in GRS effect size between n β ±SE p value p value n β ±SE p value p value groups (FDR 10%) (FDR 10%) p t test Biochemical measures −2 a,b a,b HOMA-IR 689 0.109 ± 0.050 2.73 × 10 * 0.09 675 0.021 ± 0.012 0.10 0.59 0.22 a,b a,b Fasting serum insulin (pmol/l) 689 2.250 ± 1.114 0.05 0.12 675 0.495 ± 0.307 0.11 0.59 0.27 −2 a,b a,b Fasting plasma glucose (mmol/l) 689 0.010 ± 0.005 2.51 × 10 * 0.09 675 0.003 ± 0.003 0.40 0.70 0.35 b b HbA (mmol/mol) 686 0.016 ± 0.028 0.58 0.66 636 0.005 ± 0.025 0.86 0.97 0.67 1c a,b a,b Fasting plasma LDL-cholesterol (mmol/l) 680 0.007 ± 0.007 0.37 0.46 635 0.004 ± 0.005 0.49 0.70 0.61 −3 a,b −3 −2 a,b Fasting plasma HDL-cholesterol (mmol/l) 680 −0.008 ± 0.003 1.23 × 10 ** 8.00 × 10 ** 635 −0.007 ± 0.003 1.79 × 10 * 0.29 0.68 a,b a,b Fasting plasma total cholesterol (mmol/l) 680 0.003 ± 0.007 0.73 0.78 635 −0.002 ± 0.006 0.55 0.70 0.54 a,b a,b Fasting plasma triacylglycerol (mmol/l) 680 0.007 ± 0.005 0.23 0.34 635 0.001 ± 0.003 0.94 0.97 0.46 Anthropometrics BM SDS 689 −0.007 ± 0.006 0.26 0.35 675 0.0003 ± 0.009 0.97 0.97 0.54 −5 −4 b b WHR 664 −6.267 × 10 ± 6.655 × 10 0.93 0.93 672 0.0004 ± 0.001 0.48 0.70 0.60 BP −2 Systolic BP SDS 660 0.026 ± 0.012 3.32 × 10*0.09 642 −0.011 ± 0.012 0.35 0.70 0.17 Diastolic BP SDS 660 0.013 ± 0.008 0.09 0.14 642 −0.007 ± 0.006 0.23 0.70 0.32 DXA −3 b −2 b Fat mass, total (%) 391 −0.143 ± 0.054 9.15 × 10 ** 4.80 × 10 *44 −0.140 ± 0.203 0.49 0.70 1.00 b b Fat mass, torso (%) 391 −0.116 ± 0.066 0.08 0.14 44 −0.156 ± 0.213 0.47 0.70 0.71 −4 b −3 b Fat mass, legs (%) 391 −0.197 ± 0.055 4.09 × 10 *** 6.40 × 10 ** 44 −0.133 ± 0.231 0.57 0.70 0.62 b b Fat mass, arms (%) 391 −0.116 ± 0.062 0.06 0.12 44 −0.149 ± 0.258 0.57 0.70 0.76 Results are shown for the unweighted GRS. Effect sizes and SEs were calculated using untransformed variables. Values for p were calculated using untransformed variables unless otherwise indicated. All DXA scan analyses were adjusted for type of scanner p value calculated using log-transformed (base 10) variables Analyses adjusted for age and sex *p < 0.05, **p < 0.01 and ***p < 0.001 Diabetologia (2018) 61:1769–1779 1775 written consent. The study was approved by the Danish Data SNP-specific associations with IR phenotypes Protection Agency (REG-06-2014), the Ethics Committee of Region Zealand, Denmark (SJ-104) and the Scientific Ethics When calculating the individual association of each SNP with Committee of the Capital Region of Denmark (KA98155). HOMA-IR, fasting insulin, HDL-cholesterol and triacylglyc- The study was performed in accordance with the Declaration erol, associations were identified for three, three, five and four of Helsinki 2013 and is registered at ClinicalTrials.gov SNPs, respectively, in the children with obesity, whereas as- (NCT00928473 and NCT00289237). sociations were identified for four, four, four and six SNPs, respectively, in the control population (Fig. 1,ESM Figs 1–3). Consistency of the SNP-specific directionally effect sizes was Results only observed for HDL-cholesterol in the children with obe- sity, with 38 out of 53 SNPs showing negative directional −3 Validation of the association between the GRS effects (p =2.19× 10 ; ESM Table 3). In contrast, consistent and IR phenotypes in Danish adults directional effects of the included SNPs were observed for all four traits examined in the adult Inter99 population (ESM As a means of validating the association of the GRS with IR Table 3). phenotypes in an adult Danish population, we used data from the Inter99 study population comprising 5255 non-diabetic Association between the GRS and metabolic traits Danish individuals (aged 30–60 years). The Inter99 popula- tion was not part of the study by Lotta et al [9], and Inter99 When investigating whether the GRS associated with addi- therefore constitutes an adult study sample suitable for vali- tional traits related to the metabolic syndrome, the GRS dating the association between the GRS and IR phenotypes. associated with higher fasting plasma glucose concentrations −2 In the Danish adults, the unweighted GRS associated with (β = 0.010 ± 0.005 mmol/l; p = 2.51 × 10 ) in the TCOCT fasting concentrations of insulin adjusted for BMI (β =0.014 sample (Table 2), whereas no association was observed in −7 ±0.003; p =2.25 ×10 ), triacylglycerol (β = 0.022 ± 0.003; the population-based control sample (p =0.40) (Table 2). −12 p = 1.14 × 10 ) and HDL-cholesterol (β = −0.018 ± 0.003; Although no associations between the GRS and BMI SDS, −10 p =4.86×10 ) (ESM Table 2) with effect sizes similar to WHR, HbA or diastolic BP SDS were found in either of the 1c those previously reported [9]. Furthermore, the unweighted populations, an association between the GRS and the GRS associated with HOMA-IR in the Danish adults (β = systolic BP SDS was identified in the TCOCT sample (β = −7 −2 0.014 ± 0.003; p =1.45×10 ) (ESM Table 2). Similar results 0.026 ± 0.012; p =3.32×10 )(Table 2). No difference in the were obtained for the weighted GRS . effect size of the GRS between the two study populations 53 53 was identified (Table 2). Association of the GRS with IR phenotypes in children and adolescents Association between GRS and measures of fat deposition All statistical tests were performed with and without the in- clusion of individuals (n = 29) who had conditions or were In the TCOCT sample, we observed an association between receiving medication potentially influencing IR, but the re- the GRS and lower total fat-mass percentage (β = −0.143 ± −3 sults obtained did not differ (data not shown). Only results 0.054%; p =9.15× 10 )(Table 2). We also investigated fat- based on the inclusion of these individuals are thus provided. mass percentages for specific parts of the body, such as the In the TCOCT sample, the GRS associated with HOMA-IR arms, legs and torso. In the TCOCT sample, the GRS asso- 53 53 −2 (β = 0.109 ± 0.050; p =2.73×10 ); however, no association ciated with reduced leg fat-mass percentage (β = −0.197 ± −4 was observed in the population-based control sample (p = 0.055%; p =4.09× 10 )(Table 2). In the population-based 0.10) (Table 2). The GRS was inversely associated with control sample, no associations between GRS and measures 53 53 HDL-cholesterol in both the TCOCT sample (β = −0.008 ± of fat deposition were observed, yet the effect size of the −3 0.003 mmol/l; p =1.23×10 ) and the population-based con- GRS did not differ between the examined study populations, −2 trol sample (β = −0.007 ± 0.003 mmol/l; p =1.79×10 ) as determined by a two-tailed t test (Table 2). (Table 2). The GRS did not associate with concentrations of fasting insulin or triacylglycerol in either population (p > 0.05) (Table 2). Despite the fact that the observed associations Discussion of the GRS were greater in the TCOCT sample, no statisti- cally significant difference in the effect size of the GRS in In this study, we sought to investigate whether the associations the two groups was identified, as determined by a two-tailed t of a GRS previously associated with IR phenotypes (fasting test (Table 2). insulin adjusted for BMI, HDL-cholesterol and 1776 Diabetologia (2018) 61:1769–1779 SYN2/PPARG rs308971 (G) *† Fig. 1 SNP-specific associations ARL15/FST rs4865796 (A) * with HOMA-IR in children from ZMYND15 rs754814 (T) * the TCOCT sample (black lines) MEOX2 rs17169104 (G) and population-based control MYO1F rs4804311 (A) sample (grey lines). For each DMRTA2 rs17386142 (C) SNP, the name of the nearest FAM13A rs3822072 (A) gene, rs number and risk allele as RSPO3 rs2745353 (T) reported [9] are provided. *p < PIK3R1 rs4976033 (G) EBF1 rs2434612 (G) † 0.05 for the association between MAP2K7 rs4804833 (A) the given SNP and HOMA-IR in CPEB4 rs966544 (G) the TCOCT sample. p <0.05 for CCDC92/DNAH10 rs7973683 (C) the association between the given ANKRD55 rs459193 (G) SNP and HOMA-IR in the TNFAIP8 rs1045241 (C) population-based sample KLF14 rs972283 (G) PPP1R3B rs2126259 (T) PDGFC rs6822892 (A) ADCY5 rs9881942 (A) DNM3 rs9425291 (A) COBLL1/GRB14 rs10195252 (T) RNU5F-1/LYPLAL1 rs4846565 (G) † INSR rs8101064 (T) MCC rs6887914 (C) PLA2G6 rs132985 (C) LINC02537 rs6937438 (A) PEPD rs731839 (G) † LIPG rs7227237 (C) MACROD1 rs11231693 (A) MACF1 rs683135 (A) DOK7 rs2699429 (C) MSL2 rs645040 (T) LINC01625 rs3861397 (G) CSF1 rs11577194 (T) LPL rs1011685 (C) MIR548H3 rs498313 (A) C15orf54 rs7176058 (A) TMEM110-MUSTN1 rs11130329 (A) USP37 rs492400 (T) ANKS1A rs12525532 (T) EYA1 rs4738141 (G) KLHL18 rs295449 (A) IRS1 rs2943645 (T) ANKRD10 rs7323406 (A) COL6A4P1 rs3864041 (T) NRBF2 rs10995441 (G) TRIB1 rs7005992 (C) ITPR2 rs718314 (G) L3MBTL3 rs9492443 (C) EYA2 rs6066149 (G) CEP68 rs2249105 (A) ADPGK-AS1 rs8032586 (C) ATF7IP rs17402950 (G) -0.05 -0.025 0 0.025 0.05 Protecting Predisposing β value and 95% CI per risk allele triacylglycerol) in adults [9] are already evident during child- associated with higher HOMA-IR, but no association was hood and adolescence. We initially demonstrated that the identified in the population-based control sample. GRS also associated with these phenotypes and the Previously, the 53 SNPs included in the GRS have all been HOMA-IR in Danish adults. We then proceeded to examine independently associated with IR in adults [9]. In the current whether these associations between the GRS and IR pheno- study, however, only a few of the 53 SNPs displayed associ- types could be identified in two samples of Danish children ations with either HOMA-IR, fasting insulin, HDL- and adolescents: one sample that was population-based, and cholesterol or triacylglycerol in both Danish children and the other comprising children and adolescents who were over- adults, a discrepancy which may be due to the larger statistical weight/obese, thereby representing a clinical subset with ele- power of the original study [9]. Although the majority of the vated risk of obesity-induced IR. In the TCOCT sample com- included loci exhibited same-directional effect sizes, com- prising overweight/obese children and adolescents, the GRS pared with the reports made in the discovery study [9], formal 53 Diabetologia (2018) 61:1769–1779 1777 tests of directional effects only reached statistical significance circumference in adults [9]; however, no associations for HDL-cholesterol in the children with overweight/obesity, between the GRS and these traits were identified in either and for all the examined traits in the Danish adults. This same- of our child populations. directional effect was more evident in the Danish adult popu- When examining the DXA-derived measures of fat de- lation than the child populations, suggesting that the associa- position, the GRS showed an association with lower total tion of the SNPs may be stronger in adulthood than childhood fat percentage and leg fat-mass percentage, corresponding and may thus be age-dependent. Nevertheless, as the current to the findings in adults [9]. IR, lower fat mass and ectopic study has only limited power to detect SNP-specific associa- fat deposition associate with the GRS in adults [9]. As the tions, larger study populations of both children and adults GRS in our study associates with both HOMA-IR and seem necessary to elucidate whether the included SNPs do lower fat-mass percentage, it is likely that the GRS may in fact display age-dependent effects. associate with ectopic lipid deposition, a potential cause of In a recent study by Morandi et al comprising 1076 chil- IR, in children as well. A more detailed analysis of lipid dren with obesity (mean age 11.4 years) and 1265 young accumulation in liver and skeletal muscles is needed to val- adults with normal weight (mean age 21.1 years), no associa- idate this hypothesis. tions between a GRS comprising ten IR-associated SNPs and Although we observed more associations between the OGTT-derived traits (fasting insulin, fasting glucose and GRS and IR-related phenotypes in the sample of children HOMA-IR) were identified [18]. This discrepancy in relation with obesity, the identified effect estimates of the GRS in to our results may be explained by the difference in the SNPs the two groups of children did not differ. With our statistical included in the GRS. Morandi et al used in their GRS ten power, we can therefore not claim that the associations of SNPs originally chosen by Vassy et al [17] because of their the GRS differ between our two populations. association with HOMA-IR, higher fasting insulin or IR- Nevertheless, our results still indicate that the GRS has a related traits such as lower HDL-cholesterol or higher triacyl- greater association in children with obesity, compared with glycerol, BMI or WHR in earlier publications. The GRS a population-based sample of children. Collectively, our used in the current study consists of 53 SNPs used by Lotta results obtained in Danish children, adolescents and adults et al to create their GRS [9]. By employing a meta-analysis, suggest that the association of the GRS comprising the 53 Lotta et al included up to 188,577 individuals and identified SNPs are mediated via metabolic stresses such as obesity 630 SNPs within 53 loci associating with higher fasting insu- and ageing. Our findings indicate a different effect size of lin, lower HDL-cholesterol and higher triacylglycerol. The 53 the individual SNPs in the GRS in populations comprising lead SNPs from these loci were then compiled into the GRS. children and adolescents compared with adults. Potentially, In the same study, the association between the GRS and IR the SNP-specific associations are modified by obesity, age- in adults was validated based on previously obtained measures ing and other exposures. Our findings correspond well with from a euglycemic–hyperinsulinemic clamp or insulin sup- previous studies reporting age-dependent effects of loci as- pression test in 2764 adults, or by insulin sensitivity index in sociating with metabolic traits such as BMI and obesity 4769 individuals [10, 11] . The better validated and higher [31–33]. An age-dependent variation of the effect of the number of SNPs in the GRS used in our study may explain individual SNPs in the GRS complicates the interpretation 53 53 why the GRS , in contrast to the earlier publication [18], of data and calls for the identification of age-specific trait associates with IR despite our smaller study population. loci and subsequent construction of age- and trait-specific We found that the GRS associated with other traits of the GRSs. metabolic syndrome, namely lower HDL-cholesterol concen- Although our results must be validated in larger paediatric trations and higher systolic BP in the TCOCT sample. study populations, our findings could have clinical implica- Interestingly, the GRS was inversely associated with HDL- tions, as they suggest that children with an increased risk of IR cholesterol level in the population-based control children as may benefit from preventive and therapeutic lifestyle and/or well. In contrast, there were no associations between the pharmacological treatment approaches aiming to reduce obe- GRS and higher fasting insulin and triacylglycerol concen- sity to prevent the development of IR-associated cardiometa- trations in either population, even though associations bolic risks. Furthermore, our findings indicate that the GRS with these traits, together with HDL-cholesterol, were has potential as a clinical marker to aid the identification of originally used to identify the susceptibility SNPs in adults. children with a higher risk of IR than that mediated via obesity Unfortunately, concentrations of fasting plasma lipids and alone. In adults, the GRS strongly associated with increased BP were not investigated by Lotta et al [9]. Furthermore, in risk of developing type 2 diabetes [9], yet it remains unknown the TCOCT sample, the GRS associated with higher whether children with a high genetic burden, as assessed by fasting plasma glucose, in consistency with the association the GRS , also have an increased risk of developing diabetes identified in adults [9]. The GRS has previously been and/or diabetes-related comorbidities later in life. Future stud- associated with lower BMI and higher WHR/hip ies examining the associations of the GRS in larger 53 1778 Diabetologia (2018) 61:1769–1779 Acknowledgements The authors wish to thank all children and adoles- populations and across a lifespan could potentially help to cents who participated in the present study, as well as their families. elucidate whether the GRS could be used as a clinical tool Additionally, we wish to thank O. Troest, B. Holløse, F. Pinar (the that would, during childhood and adolescence, already enable Children’s Obesity Clinic, Department of Pediatrics, Copenhagen the identification of individuals with an increased risk of IR University Hospital Holbæk, Denmark) for careful laboratory assistance. Finally, we thank A. Forman, T. H. Lorentzen and G. J. Klavsen for their and ultimately type 2 diabetes. As such, the GRS ,or even an dedicated laboratory assistance, P. Sandbeck for data management, G. improved GRS comprising other or additional SNPs with val- Lademann for secretarial support, and T. F. Toldsted for grant manage- idated strong associations with IR phenotypes in childhood, ment (all from the Section of Metabolic Genetics, Novo Nordisk could potentially be one of the first steps towards personalised Foundation Center for Basic Metabolic Research, Faculty of Health and Medical Sciences, University of Copenhagen, Denmark). intervention programmes aiming to minimise the occurrence of cardiovascular events and preterm death associated with Data availability The datasets generated during and/or analysed during early-onset type 2 diabetes [34]. the current study are available from the corresponding author on reason- Our study is limited by the relatively small study sample, able request. which reduces the statistical power. Furthermore, analyses Funding The study was funded by Innovation Fund Denmark (grants were not adjusted for stage of puberty but only for age. 0603-00484B and 0603-00457B), the Danish Diabetes Academy and Nevertheless, a very detailed phenotypic characterisation the Region Zealand Health Scientific Research Foundation. This study was available for the included study population, enabling the is part of the TARGET (The Impact of our Genomes on Individual detailed analysis of several traits related to IR and fat deposi- Treatment Response in Obese Children, http://target.ku.dk)and BioChild (Genetics and Systems Biology of Childhood Obesity in India tion. Furthermore, data for the examined traits were available and Denmark, http://biochild.ku.dk) consortia studies, as well as The for two populations of children with similar age spans yet Danish Childhood Obesity Biobank. The Novo Nordisk Foundation different ranges of BMI SDS, enabling an evaluation of the Center for Basic Metabolic Research is an independent research centre effect of obesity on the effect of the GRS. It should be noted at the University of Copenhagen, partially funded by an unrestricted donation from the Novo Nordisk Foundation (www.metabol.ku.dk). that the two populations of children were selected in different The Inter99 study was initiated by T. Jørgensen (PI; Research Centre ways: the children with obesity were highly selected accord- for Prevention and Health, Glostrup University Hospital, Glostrup, ing to their BMI and age, whereas the children from the Denmark), K. Borch-Johnsen (co-PI; Steno Diabetes Center A/S, Gentofte, Denmark), T. Thomsen (Research Centre for Prevention and population-based group were selected only according to age. Health, Glostrup University Hospital, Glostrup, Denmark) and H. Ibsen This discrepancy in the selection of study participants may (Division of Cardiology, Holbæk University Hospital, Holbæk, potentially affect our findings. Denmark). The present steering committee comprises T. Jørgensen and In conclusion, we investigated whether the GRS asso- 53 C. Pisinger (Research Centre for Prevention and Health, Glostrup University Hospital, Glostrup, Denmark). ciates with IR phenotypes and HOMA-IR in both children and adolescents who are overweight or obese, and in a Duality of interest The authors declare that there is no duality of interest population-based control sample. A GRS associating with associated with this manuscript. IR in children could help to identify children predisposed to IR. In overweight or obese children and adolescents, the 53 Contribution statement ASG, MH, JTK, TRHN, AL, MEJ, JCH and TH contributed to the conception and design of the study as well as acquisi- SNPs cumulatively associate with IR. The results indicate tion of data. MH, EVRA, NG, HNK, OP and TH planned and performed that children who have a genetic predisposition to IR, as the acquisition of genotypes, and TMS, EVRA and YM constructed the assessed by the 53 SNPs, will have a higher risk of devel- GRSs. MH, JR, MØJ, NG and TH planned and performed the statistical oping IR if they become overweight or obese. However, as analyses, while ASG, MH, JR, JTK, TRHN, NG, BH and TH interpreted the data. ASG and MH wrote the initial draft, while all authors contrib- no difference between the effects size of the GRS in the uted to the critical revision of the draft. The final draft was commented two groups of children could be identified, we cannot with upon and approved by all authors. TH is the guarantor of this work. certainty conclude that obesity is essential for the associa- tion between HOMA-IR and the GRS . This hypothesis Open Access This article is distributed under the terms of the Creative needs to be verified in a larger population. The identifica- Commons Attribution 4.0 International License (http:// tion of additional SNPs displaying strong associations with creativecommons.org/licenses/by/4.0/), which permits unrestricted use, IR-related phenotypes during childhood would increase the distribution, and reproduction in any medium, provided you give appro- priate credit to the original author(s) and the source, provide a link to the clinical impact of the GRS and allow the identification of Creative Commons license, and indicate if changes were made. children predisposed to IR. Treatment strategies targeted against factors important for the development of IR, such References as obesity, could be developed specifically for predisposed children. Furthermore, our study showed that fat percentage 1. International Diabetes Federation (2006) The IDF consensus world- in the body extremities was inversely associated with wide definition of the Metabolic Syndrome. Available from http:// GRS in the children and adolescents who are overweight www.idf.org/e-library/consensus-statements/60-idfconsensus- or obese, suggesting that impaired capacity to store fat in worldwide-definitionof-the-metabolic-syndrome. Accessed 17 Sep peripheral compartments increases the risk of IR. 2017 Diabetologia (2018) 61:1769–1779 1779 2. Roberts CK, Hevener AL, Bernard RJ (2014) Metabolic Syndrome Traits, in Obese and Normal-Weight Youths. J Clin Endocrinol Metab 101:4244–4250 and Insulin Resistance: Underlying Causes and Modification by Exercise Training. Compr Physiol 3:1–58 19. Banasik K, Hollensted M, Andersson E et al (2012) The effect of 3. Sinha R, Fisch G, Teague B et al (2002) Prevalence of impaired FOXA2rs1209523 on glucose-related phenotypes and risk of type 2 glucose tolerance among children and adolescents with marked diabetes in Danish individuals. BMC Med Genet 13:10 obesity. N Engl J Med 346:802–810 20. Jørgensen T, Borch-Johnsen K, Thomsen TF et al (2003) A ran- 4. Kloppenborg JT, Fonvig CE, Nielsen TRH et al (2017) Impaired domized non-pharmacological intervention study for prevention of fasting glucose and the metabolic profile in Danish children and ischaemic heart disease: baseline results Inter99. Eur J Cardiovasc adolescents with normal weight, overweight, or obesity. Pediatr Prev Rehabil 10:377–386 Diabetes 1–10 21. McCarthy S, Das S, Kretzschmar W et al (2016) A reference panel 5. Zimmermann E, Bjerregaard LG, Gamborg M et al (2017) of 64,976 haplotypes for genotype imputation. Nat Genet 48:1279– Childhood body mass index and development of type 2 diabetes throughout adult life—A large-scale danish cohort study. Obesity 22. Holm J-C, Gamborg M, Bille DS et al (2011) Chronic care treat- 25:965–971 ment of obese children and adolescents. Int J Pediatr Obes 6:188– 6. Dupuis J, Langenberg C, Prokopenko I et al (2010) New genetic loci implicated in fasting glucose homeostasis and their impact on 23. Nysom K, Mølgaard C, Hutchings B, Michaelsen KF (2001) Body type 2 diabetes risk. Nat Genet 42:105–116 mass index of 0 to 45-y-old Danes: reference values and compari- 7. Manning AK, Hivert M-F, Scott RA et al (2012) A genome-wide son with published European reference values. Int J Obes Relat approach accounting for body mass index identifies genetic variants Metab Disord 25:177–184 influencing fasting glycemic traits and insulin resistance. Nat Genet 24. Cole TJ, Green PJ (1992) Smoothing reference centile curves: The 44:659–669 lms method and penalized likelihood. Stat Med 11:1305–1319 8. Scott RA, Lagou V, Welch RP et al (2012) Large-scale association 25. National High Blood Pressure Education Program Working Group analyses identify new loci influencing glycemic traits and provide on High Blood Pressure in Children and Adolescents (2004) The insight into the underlying biological pathways. Nat Genet 44:991– fourth report on the diagnosis, evaluation, and treatment of high blood pressure in children and adolescents. Pediatrics 114:555–576 9. Lotta LA, Gulati P, Day FR et al (2017) Integrative genomic anal- 26. Delaneau O, Marchini J, 1000 Genomes Project Consortium (2014) ysis implicates limited peripheral adipose storage capacity in the Integrating sequence and array data to create an improved 1000 pathogenesis of human insulin resistance. Nat Genet 49:17–26 Genomes Project haplotype reference panel. Nat Commun 5:3934 10. Prokopenko I, Poon W, Mägi R et al (2014) A central role for 27. Howie BN, Donnelly P, Marchini J (2009) A flexible and accurate GRB10 in regulation of islet function in man. PLoS Genet 10: genotype imputation method for the next generation of genome- e1004235 wide association studies. PLoS Genet 5:e1000529 11. Knowles JW, Xie W, Zhang Z et al (2016) Identification and vali- 28. Matthews DR, Hosker JP, Rudenski AS et al (1985) Homeostasis dation of N-acetyltransferase 2 as an insulin sensitivity gene. J Clin model assessment: insulin resistance and beta-cell function from Invest 126:403 fasting plasma glucose and insulin concentrations in man. 12. Lyssenko V, Jonsson A, Almgren P et al (2008) Clinical risk factors, Diabetologia 28:412–419 DNA variants, and the development of type 2 diabetes. N Engl J 29. Friedewald WT, Levy RI, Fredrickson DS (1972) Estimation of the Med 359:2220–2232 concentration of low-density lipoprotein cholesterol in plasma, with- 13. De Miguel-Yanes JM, Shrader P, Pencina MJ et al (2011) Genetic out use of the preparative ultracentrifuge. Clin Chem 18:499–502 risk reclassification for type 2 diabetes by age below or above 50 30. Benjamini Y, Hochberg Y (1995) Controlling the false discovery years using 40 type 2 diabetes risk single nucleotide polymor- Rate: a practical and powerful approach to multiple testing. R Stat phisms. Diabetes Care 34:121–125 Soc 57:289–300 14. Vassy JL, Durant NH, Kabagambe EK et al (2012) A genotype risk 31. Felix JF, Bradfield JP, Monnereau C et al (2016) Genome-wide score predicts type 2 diabetes from young adulthood: The CARDIA association analysis identifies three new susceptibility loci for child- study. Diabetologia 55:2604–2612 hood body mass index. Hum Mol Genet 25:389–403 15. Vassy JL, Dasmahapatra P, Meigs JB et al (2012) Genotype predic- 32. Bradfield J, Taal H, Timpson N (2012) A genome-wide association tion of adult type 2 diabetes from adolescence in a multiracial pop- meta-analysis identifies new childhood obesity loci. Nat Genet 44: ulation. Pediatrics 130:1235–1242 526–531 16. Walford GA, Porneala BC, Dauriz M et al (2014) Metabolite traits 33. Winkler TW, Justice AE, Graff M et al (2015) The influence of age and genetic risk provide complementary information for the predic- and sex on genetic associations with adult body size and shape: a tion of future type 2 diabetes. Diabetes Care 37:2508–2514 large-scale genome-wide interaction study. PLoS Genet 11:1–42 17. Vassy JL, Hivert MF, Porneala B et al (2014) Polygenic type 2 diabetes prediction at the limit of common variant detection. 34. Zoungas S, Woodward M, Li Q et al (2014) Impact of age, age at Diabetes 63:2172–2182 diagnosis and duration of diabetes on the risk of macrovascular and 18. Morandi A, Bonnefond A, Lobbens S et al (2016) Associations microvascular complications and death in type 2 diabetes. Between Type 2 Diabetes-Related Genetic Scores and Metabolic Diabetologia 57:2465–2474

Journal

DiabetologiaSpringer Journals

Published: May 31, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off