Effect of Common Genetic Variants Associated with Type 2 Diabetes and Glycemic Traits on α- and β-cell Function and Insulin Action in Man

Transcription

1 Page 1 of 34 Diabetes Effect of Common Genetic Variants Associated with Type 2 Diabetes and Glycemic Traits on α- and β-cell Function and Insulin Action in Man Anna Jonsson 1,2, Claes Ladenvall 1, Tarunveer Singh Ahluwalia 1,2, Jasmina Kravic 1, Ulrika Krus 1, Jalal Taneera 1, Bo Isomaa 3,4, Tiinamaija Tuomi 3,5, Erik Renström 1, Leif Groop 1,6 and Valeriya Lyssenko 1,7 From the 1 Department of Clinical Sciences, Diabetes and Endocrinology, Lund University, Malmö, Sweden, Lund University Diabetes Centre, Malmö, Sweden, 2 Novo Nordisk Foundation Centre for Basic Metabolic Research, Section of Metabolic Genetics, Faculty of Health and Medical and Sciences, University of Copenhagen, Copenhagen, Denmark, 3 Folkhälsan Research Centre, Helsinki, Finland, 4 Department of Social Services and Health Care, Jakobstad, Finland, 5 Department of Medicine, Helsinki University Central Hospital, Helsinki, Finland, 6 Finnish Institute of Molecular Medicine (FIMM), Helsinki University, Helsinki, Finland, 7 Steno Diabetes Center A/S, Gentofte, Denmark Running title: Genetic variants, insulin and glucagon secretion Word count: Abstract 197, Main text 1,997 Items on display: 4 Figures, 8 Supplementary Tables, 5 Supplementary Figures Abbreviations SNP - single nucleotide polymorphism, OGTT - oral glucose tolerance test, FPG fasting plasma glucose, ISI - insulin sensitivity index, HOMA-IR - insulin resistance index, CIR - corrected insulin response, DI - disposition index. Address for correspondence Anna Jonsson Novo Nordisk Foundation Centre for Basic Metabolic Research Section of Metabolic Genetics Faculty of Health and Medical and Sciences University of Copenhagen DIKU Building Universitetsparken 1, 1st floor DK-2100 Copenhagen Ø Denmark Phone: jonsson@sund.ku.dk 1 Diabetes Publish Ahead of Print, published online April 4, 2013

2 Diabetes Page 2 of 34 Abstract Although meta-analyses of genome-wide association studies have identified more than 60 single nucleotide polymorphisms (SNPs) associated with type 2 diabetes and/or glycemic traits, there is little information whether these variants also affect α-cell function. The aim of the present study was to evaluate the effects of glycemia-associated genetic loci on islet function in vivo and in vitro. We studied 43 SNPs in 4,654 normoglycemic participants from the Finnish population-based PPP-Botnia study. Islet function was assessed, in vivo, by measuring insulin and glucagon concentrations during OGTT, and, in vitro, by measuring glucose stimulated insulin and glucagon secretion from human pancreatic islets. Carriers of risk variants in BCL11A, HHEX, ZBED3, HNF1A, IGF1 and NOTCH2 showed elevated, while those in CRY2, IGF2BP2, TSPAN8 and KCNJ11 decreased fasting and/or 2hr glucagon concentrations in vivo. Variants in BCL11A, TSPAN8, and NOTCH2 affected glucagon secretion both in vivo and in vitro. The variant was a clear outlier in the relationship analysis between insulin secretion and action, as well as between insulin, glucose and glucagon. Many of the genetic variants shown to be associated with type 2 diabetes or glycemic traits also exert pleiotropic in vivo and in vitro effects on islet function. 2

3 Page 3 of 34 Diabetes In the last few years genome wide association studies have substantially increased the knowledge of genetic variants predisposing to type 2 diabetes. Although the majority of these SNPs seem to influence insulin secretion, few if any studies have assessed their simultaneous effects on β- and α-cell function in vivo and in vitro. The aim of the present study was to provide a comprehensive evaluation of the effects of genetic loci associated with type 2 diabetes [1] and/or glucose and insulin levels [2] on islet function, in vivo, in a large well characterized population-based study from the western coast of Finland, the PPP-Botnia study, and, in vitro, in human pancreatic islets. Islet function was assessed by measuring insulin and glucagon concentrations during an oral glucose tolerance test (OGTT). Research Design and Methods Study population The PPP-Botnia Study (Prevalence, Prediction and Prevention of diabetes) is a populationbased study from the Botnia region of Western Finland. 4,654 non-diabetic subjects (fasting plasma glucose (FPG) <7.0 mmol/l and 2hr plasma glucose <11.1 mmol/l) were included in the current study [3] (Supplementary Table 1). Measurements Blood samples were drawn at 0, 30 and 120 minutes of the OGTT. Insulin was measured using a fluoroimmunometric assay (AutoDelfia, PerkinElmer, Waltham, MA, USA), and serum glucagon using radioimmunoassay (Millipore, St. Charles, MO, USA). Insulin sensitivity index (ISI) was calculated as 10,000/ (fasting P-glucose fasting P- insulin meanogtt glucose meanogtt insulin ) [4]. Insulin secretion was assessed as corrected insulin response during OGTT (CIR=100 insulin at 30 min/[glucose at 30 min (glucose 30 3

4 Diabetes Page 4 of 34 min-3.89)]) [5] or as disposition index, i.e. insulin secretion adjusted for insulin sensitivity (DI=CIR ISI). Genotyping Genotyping was performed either by matrix-assisted laser desorption-ionization time-of-flight mass spectrometry on the MassARRAY platform (Sequenom, San Diego, CA, USA) or by an allelic discrimination method with a TaqMan assay on the ABI 7900 platform (Applied Biosystems, Foster City, CA, USA). We obtained an average genotyping success rate of 98.1%, and the average concordance rate, based on 10,578 (6.5%) duplicate comparisons was 99.9%. Hardy-Weinberg equilibrium was fulfilled for all studied variants (P>0.05). Glucose-stimulated insulin and glucagon secretion in human pancreatic islets Glucose-stimulated insulin and glucagon secretion were performed in high (16.7 mmol/l) and low (1.0 mmol/l) glucose concentration in the medium as described previously [6] and were available from 56 non-diabetic human pancreatic islet donors. Given the limited number of donors for genetic analyses, the relationships between absolute genotypic means of insulin and glucagon concentrations and the 95% CI were plotted separately for different genotype carriers rather than contrasting mean differences in phenotypes between genotypes. Statistical Analysis Variables are presented as mean ± standard deviation (SD), and if not normally distributed as median [interquartile range (IQR)]. Genotype-phenotype correlations were studied using linear regression analyses adjusted for age, gender and BMI. Non-normally distributed variables were logarithmically (natural) transformed for analyses. All statistical analyses were 4

5 Page 5 of 34 Diabetes performed with IBM SPSS Statistics version 19.0 (IBM, Armonk, NY) and PLINK version 1.07 [7, 8]. Results Effect of SNPs on metabolic variables Insulin secretion We observed the strongest effect on reduced insulin secretion for the rs (CIR P= ; DI P= ). Additionally, we confirmed that genetic variants in CDKAL1 (CIR P= ; DI P= ), HHEX (CIR P= ; DI P=0.0021), GIPR (CIR P= ; DI P=0.0061), (CIR P=0.0010; DI P= ), KCNQ1 (CIR P=0.0032; DI P=0.0054), CDKN2A/2B (CIR P=0.016; DI P=0.0072), FADS1 (CIR P= ), IGF2BP2 (DI P=0.0013), JAZF1 (DI P=0.015), KCNJ11 (CIR P=0.019), PPARG (DI P=0.023), ZFAND6 (CIR P=0.026), CHCHD2P9 (DI P=0.039), ARAP1 (CIR P=0.039) and BCL11A (CIR P=0.048) were associated with reduced insulin secretion (Supplementary Table 2). Insulin action The variants in (P= ), CHCHD2P9 (P=0.011), FADS1 (P=0.013), PPARG (P=0.041) and CDKAL1 (P=0.042) were associated with reduced insulin sensitivity (Supplementary Table 3). Glucagon secretion To explore whether these genetic variants would also influence α-cell function, we analyzed glucagon concentrations during OGTT. The variants in BCL11A (P=0.0069), HHEX (P=0.0096), ZBED3 (P=0.033) and HNF1A (P=0.047) were associated with elevated, while variants in CRY2 (P=0.0021), IGF2BP2 (P=0.0036) and TSPAN8 (P=0.034) were associated with reduced fasting glucagon levels. Furthermore, variants in 5

6 Diabetes Page 6 of 34 IGF1 (P=0.0048), ZBED3 (P=0.0094) and NOTCH2 (P=0.016) were associated with increased and the variant in KCNJ11 (P=0.027) with decreased postprandial glucagon levels (Supplementary Table 4). Effect of SNPs on relationships between metabolic variables Insulin secretion vs. sensitivity (CIR vs. ISI) We plotted the effects of the tested genetic variants on CIR and ISI (Figure 1). Most SNPs were related to insulin secretion and sensitivity in a linear fashion, with SNPs in the, CDKAL1,, CDKN2A/2B and PPARG loci in the lower left quadrant, reflecting both impaired insulin secretion and action, with being the most extreme. In contrast variants in GIPR, FADS1, CAMK1D and TCF7L2 showed relatively high insulin sensitivity despite low insulin secretion. A genotypecentric analysis (Supplementary Figure 1) also showed that homozygous carriers of, CDKAL1 and risk genotypes had CIR values below the values expected from ISI values whereas TT carriers of TCF7L2 showed rather high insulin sensitivity. Glucose vs. insulin levels The variant in showed both high fasting glucose and insulin levels (Figure 2, Supplementary Table 5 and 6). At the same glucose level, CHCHD2P9 showed the highest and FADS1 the lowest fasting insulin concentration, while, at the same fasting insulin concentration, FTO showed the lowest and the highest fasting glucose levels. Homozygous carriers of the TCF7L2 risk variant had the lowest, whereas homozygous carriers of HMGA2 showed the highest insulin levels (Supplementary Figure 2). These differences were partially also seen at 2hr of the OGTT. Glucose vs. glucagon levels The highest fasting glucose level related to glucagon level was seen for (Figure 3, Supplementary Figure 3), which was also reflected by high fasting and 2hr insulin/glucose ratios (Supplementary Table 7, Supplementary Figures 4 and 6

7 Page 7 of 34 Diabetes 5). In line with previous reports, the TCF7L2 variant showed low glucagon levels [9] and homozygous T-allele carriers showed high fasting and 2hr glucose levels, despite low insulin/glucagon ratio. Glucagon vs. insulin levels in human pancreatic islets The risk allele carriers of variants in CAMK1D, TCF7L2, GIPR, TSPAN8 and KLF14 showed lower, while the risk allele carriers of CDKAL1, BCL11A, CRY2, FADS1, NOTCH2, HMGA2 and showed higher glucagon levels than expected from the 95% CI of the relationship between mean values of insulin and glucagon concentrations measured at normal glucose levels. At high glucose, carriers of the risk alleles at DGKB, GIPR, JAZF1, CDKN2A/2B, HMGA2 and NOTCH2 displayed high glucagon levels, while those of TCF7L2, CAMK1D, KLF14, FTO and ZBED3 showed lower glucagon levels (Figure 4). Discussion In the present study, we demonstrate that the effects of some of the genetic variants associated with type 2 diabetes and/or glycemic traits seem to involve the crosstalk between hormone secretion from the β- and α-cells. The novel findings in the present study were that we observed associations of variants in 10 loci with glucagon secretion, of which 6 loci (HHEX, IGF2BP2, CDKAL1, FADS1, and CDKN2A/2B) were also associated with reduced insulin secretion. Effects of variants on the relationship between CIR and ISI were similar to those observed by Grarup et al. [10], but we also showed that and CDKAL1 were clear outliers with strong effects on both impaired insulin secretion and sensitivity. These analyses used betavalues which reflect fractions of the standard deviations. We therefore extended the analyses 7

8 Diabetes Page 8 of 34 by analyzing the relationship between the absolute values of CIR and ISI in different genotype carriers, which allowed us to obtain a quantitative effect of SNPs. These analyses differ from those presented by Grarup et al. as we now related the effect of the homozygous risk genotypes to the mean values of a reference population instead of expressing them as allelic changes in fraction of standard deviations. It can, of course, be questioned whether homozygosity for risk alleles would best describe their effect on this relationship, but it has the advantage that it allows us to show the absolute effects of the variants. Importantly, we show that many of the genetic variants seem to affect the relationship between insulin/glucagon secretion and blood glucose levels both at fasting and 2hr during OGTT. This would imply that in the risk carriers, reduced early insulin secretion in conjunction with impaired suppression of glucagon secretion and/or action would lead to excessive delivery of glucose into the circulation. An interesting finding was that, again, was different from the rest of the variants and showed elevated glucose despite high insulin/glucagon ratio. There is no obvious explanation to these findings apart from that the ratio between insulin and glucagon may be more important than the individual insulin and glucagon levels for the glucose-raising effect of. Also, is mostly expressed in β-cells, while MTNR1A predominating in α-cells. Until now, no variant in the MTNR1A gene has been associated with type 2 diabetes or elevated glucose levels. In keeping with a previous study [11], the rs variant in TCF7L2 was associated with lower glucagon values both in vivo and in vitro. In contrast to some previous studies [12, 13] we did not observe a significant reduction in fasting nor glucose-stimulated insulin secretion for this variant in the current study. The PPP-study is a population-based study including people aged years, most of whom had normal glucose tolerance. In the previous studies the effect on insulin secretion was clearly stronger in hyperglycemic individuals or in those 8

9 Page 9 of 34 Diabetes who subsequently developed diabetes. The same was seen in the present study; if we restrict the analysis to hyperglycemic individuals (FPG >5.5 mmol/l), we see a clear effect of the TCF7L2 SNP on reduced insulin secretion (N=1,440, beta (0.035), P=0.002). This effect might be too small to fully explain the glucose-increasing effect of the SNP, nor is it easy to see an effect of reduced glucagon levels. One possibility is that the TCF7L2 variant has effects on hepatic glucose production as previously shown [13-15] which cannot be explained by effects of glucagon on the liver. From the figures relating the mean values of fasting and 2hr insulin/glucagon with fasting and 2hr glucose concentrations (Supplementary Figure 6), it was obvious that homozygous carriers of HMGA2 differed most from the others. Thus, while carriers of two HMGA2 risk alleles showed increased insulin secretion and decreased insulin sensitivity, possibly protecting them from type 2 diabetes, they demonstrated low insulin/ glucagon, which in turn led to elevated glucose. In vitro, homozygous risk allele carriers showed reduced suppressive effect of high glucose on glucagon secretion. HMGA2 encodes for a non-histone protein with structural DNA-binding domains and may act as a transcriptional regulating factor. The current data suggest that it may involve a complex interplay between islet function and insulin action. The present study also provides some novel insight for the potential mechanism by which the BCL11A variant could increase the risk of type 2 diabetes. BCL11A rs was originally identified as a risk variant for type 2 diabetes in the DIAGRAM+ meta-analysis [1]. The protein encoded by BCL11A is a zinc-finger protein with highest expression in the fetal brain and the adult central nervous system. While its biological function is not fully elucidated, it might be involved in the signal transduction from cell to cell or in the neuronal regulation of hormonal secretion. Association of variants in BCL11A with reduced insulin secretion has 9

10 Diabetes Page 10 of 34 previously been reported [16]. Here we report novel associations of a BCL11A variant with increased glucagon levels in vivo. Additionally, in vitro, homozygous risk allele carriers of the BCL11A variant had glucagon levels above the expected 95% CI of the relationship between insulin and glucagon secretion in human pancreatic islets. Together, results from the present paper and previous reports suggest that this variant might predispose to increased risk of type 2 diabetes via reduced insulin and increased glucagon secretion. In conclusion, we here provide a comprehensive description of how common genetic variants associated with type 2 diabetes and/or glycemic traits influence insulin and glucagon secretion in vivo and in vitro in human pancreatic islets. The data clearly emphasize the need to consider effects on α-cell function when studying mechanisms by which these variants increase risk of type 2 diabetes. Acknowledgements Studies at LUDC, Malmö were supported by grants from the Swedish Research Council, including a Linné grant (No ) and a Strategic Research Grant EXODIAB (Dnr ), the Heart and Lung Foundation, the Swedish Diabetes Research Society, the European Community's Seventh Framework Programme grant ENGAGE (Health-F ) and CEED3 (223211), the Diabetes Programme at the Lund University, the Påhlsson Foundation, the Crafoord Foundation, the Knut and Alice Wallenberg Foundation, Novo Nordisk Foundation and the European Foundation for the Study of Diabetes (EFSD). The PPP-Botnia study were supported by grants from the Sigrid Juselius Foundation, Folkhälsan Research Foundation, the Finnish Diabetes Research Society, the Signe and Ane Gyllenberg Foundation, Swedish Cultural Foundation in Finland, Ollqvist Foundation, Ministry of Education of Finland, Foundation for Life and Health in Finland, Jakobstad Hospital, Medical Society of Finland, Närpes Research Foundation and the Vasa and Närpes Health centers. Human islets were provided by the Nordic network for clinical islets transplantation by the courtesy of Dr. Olle Korsgren, Uppsala, Sweden. We thank the patients for their participation and the Botnia Study Group for clinically studying the patients. No potential conflicts of interest relevant to this article were reported. Anna Jonsson participated in genotyping and study design, researched and interpreted the data and wrote the manuscript. Claes Ladenvall, Tarunveer Singh Ahluwalia and Jasmina Kravic researched the data. Ulrika Krus, Jalal Taneera and Erik Renström were responsible for 10

11 Page 11 of 34 Diabetes mrna analyses and the insulin and glucagon secretion measurements from the human pancreatic islets. Bo Isomaa and Tiinamaija Tuomi collected and phenotyped the PPP-Botnia study population, Leif Groop and Valeriya Lyssenko designed and supervised the study, interpreted the data and contributed to discussion. All co-authors reviewed and approved the final version of the manuscript. Prof. Leif Groop is the guarantor of this work and, as such, had full access to the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. 11

12 Diabetes Page 12 of 34 References 1. Voight, B.F., et al., Twelve type 2 diabetes susceptibility loci identified through largescale association analysis. Nat Genet, (7): p Dupuis, J., et al., New genetic loci implicated in fasting glucose homeostasis and their impact on type 2 diabetes risk. Nat Genet, (2): p Isomaa, B., et al., A family history of diabetes is associated with reduced physical fitness in the Prevalence, Prediction and Prevention of Diabetes (PPP)-Botnia study. Diabetologia, (8): p Matsuda, M. and R.A. DeFronzo, Insulin sensitivity indices obtained from oral glucose tolerance testing: comparison with the euglycemic insulin clamp. Diabetes Care, (9): p Hanson, R.L., et al., Evaluation of simple indices of insulin sensitivity and insulin secretion for use in epidemiologic studies. Am J Epidemiol, (2): p Jonsson, A., et al., A variant in the KCNQ1 gene predicts future type 2 diabetes and mediates impaired insulin secretion. Diabetes, (10): p Whole genome association analysis toolset. [cited 2010; Available from: 8. Purcell, S., et al., PLINK: a tool set for whole-genome association and populationbased linkage analyses. Am J Hum Genet, (3): p Alibegovic, A.C., et al., The T-allele of TCF7L2 rs associates with a reduced compensation of insulin secretion for insulin resistance induced by 9 days of bed rest. Diabetes, (4): p Grarup, N., T. Sparso, and T. Hansen, Physiologic characterization of type 2 diabetesrelated loci. Curr Diab Rep, (6): p Pilgaard, K., et al., The T allele of rs TCF7L2 is associated with impaired insulinotropic action of incretin hormones, reduced 24 h profiles of plasma insulin and glucagon, and increased hepatic glucose production in young healthy men. Diabetologia, (7): p Lyssenko, V., et al., Clinical risk factors, DNA variants, and the development of type 2 diabetes. N Engl J Med, (21): p Lyssenko, V., et al., Mechanisms by which common variants in the TCF7L2 gene increase risk of type 2 diabetes. J Clin Invest, (8): p Boj, S.F., et al., Diabetes risk gene and Wnt effector Tcf7l2/TCF4 controls hepatic response to perinatal and adult metabolic demand. Cell, (7): p Norton, L., et al., Chromatin occupancy of transcription factor 7-like 2 (TCF7L2) and its role in hepatic glucose metabolism. Diabetologia, (12): p Simonis-Bik, A.M., et al., Gene variants in the novel type 2 diabetes loci CDC123/CAMK1D, THADA, ADAMTS9, BCL11A, and affect different aspects of pancreatic beta-cell function. Diabetes, (1): p

13 Page 13 of 34 Diabetes CIR CHCHD2P9 BCL11A HNF1A DUSP9(m) VPS13C ADAMTS9 FTO KCNQ1* HMGA2 KLF14 IGF1 PROX1 TSPAN8 GLIS3 ZBED3 SLC2A2 THADA CAMK1D PRC1 DUSP9(f) MADD TP53INP1 PPARG WFS1 TCF7L2 SLC30A8 CRY2 ADRA2A JAZF1 DGKB IGF2BP2 ARAP1 KCNJ11 ZFAND6 KCNQ1 CDKAL1 CDKN2A/2B NOTCH2 HHEX GIPR FADS ISI Figure 1. Graphic representation of effects of genetic variants on relationship between insulin secretion and insulin sensitivity in non-diabetic participants in the PPP-Botnia study. Effect sizes are betas from linear regression adjusted for age, gender and BMI. Whereas most genetic variants either increased insulin secretion or insulin sensitivity, those in, CDKAL1 and were located to the lower left quadrant, reflecting both impaired insulin secretion and insulin sensitivity, with being the most extreme. In contrast, GIPR and FADS1 showed relatively high insulin sensitivity despite low insulin secretion. KCNQ1 represents rs , while KCNQ1* represents rs Black circles P<0.05; White circles P>0.05. A 0.15 B GIPR CDKN2A/2B CHCHD2P9 CDKAL1 Fasting glucose MADD CDKAL1 THADA CDKN2A/2B CRY2 PPARG DGKB KCNQ1 JAZF1 HNF1A IGF2BP2 TCF7L2 FADS1 GLIS3 HMGA2 KCNQ1* WFS1 TP53INP1 ARAP1 BCL11A HHEX ZFAND6 KCNJ11 IGF1 ZBED3 TSPAN8 SLC30A8 VPS13C DUSP9(f) SLC2A2 ADRA2A CAMK1D DUSP9(m) ADAMTS9 PRC1 NOTCH2 GIPR KLF14 PROX1 CHCHD2P9 2hr glucose KCNQ1 VPS13C FTO ZBED3 MADD TP53INP1 KCNQ1* KLF14 TCF7L2 TSPAN8 HMGA2 SLC2A2 JAZF1 IGF2BP2 CAMK1D SLC30A8 CRY2 PPARG ADRA2A HNF1A KCNJ11 DGKB DUSP9(f) BCL11A ARAP1 ADAMTS9 IGF1 HHEX PROX1 NOTCH2 ZFAND6 WFS1 THADA DUSP9(m) GLIS3 PRC1 FADS1 FTO Fasting insulin hr insulin Figure 2. Graphic representation of effects of genetic variants on relationship between glucose and insulin levels during OGTT in non-diabetic participants of the PPP-Botnia study. Effect sizes are betas from linear regression adjusted for age, gender and BMI. KCNQ1 represents rs , while KCNQ1* represents rs Black circles P<0.05; White circles P>0.05. Panel A shows the relationship between fasting insulin and fasting glucose. showed the highest effect on fasting glucose despite elevated insulin levels. CHCHD2P9 showed increased, while FADS1 showed decreased fasting insulin. Panel B shows the relationship between 2hr insulin and 2hr glucose., CDKAL1, ZBED3 and all showed high postprandial glucose levels despite elevated insulin levels. 13

14 Diabetes Page 14 of 34 A 0.15 B 0.15 Fasting glucose IGF2BP2 MADD THADA CHCHD2P9 CDKN2A/2B CDKAL1 DGKB PPARG CRY2 FADS1 JAZF1 KCNQ1 GLIS3 HNF1A HMGA2 TP53INP1 KCNQ1* HHEX TCF7L2 ARAP1 TSPAN8 WFS1 ZFAND6 SLC30A8ADRA2A IGF1 ZBED3 BCL11A SLC2A2 KCNJ11 DUSP9(f) VPS13C CAMK1D PRC1 ADAMTS9 DUSP9(m) NOTCH2 GIPR PROX1 KLF14 FTO 2hr glucose CDKN2A/2B CDKAL1 CHCHD2P9 GIPR VPS13C KCNQ1 FTO TP53INP1 MADD KCNQ1* JAZF1 TCF7L2 KLF14 TSPAN8 CAMK1D SLC2A2 IGF2BP2 HMGA2 CRY2 SLC30A8 ADRA2A PPARG HNF1A KCNJ11 DUSP9(f) DGKB ADAMTS9 BCL11A ARAP1 PROX1 HHEX DUSP9(m) WFS1 GLIS3 PRC1 ZFAND6 THADA FADS1 ZBED3 IGF1 NOTCH Fasting glucagon hr glucagon Figure 3. Graphic representation of effects of genetic variants on relationship between glucose and glucagon levels during OGTT in non-diabetic participants of the PPP-Botnia study. Effect sizes are betas from linear regression adjusted for age, gender and BMI. KCNQ1 represents rs , while KCNQ1* represents rs Black circles P<0.05; White circles P>0.05. Panel A shows the relationship between fasting glucagon and fasting glucose. The highest fasting glucose level related to glucagon level was seen for. IGF2BP2 and CRY2 showed reduced glucagon secretion, while BCL11A and HHEX were associated with increased fasting glucagon. Panel B shows the relationship between 2hr glucagon and 2hr glucose. ZBED3, IGF1 and NOTCH2 showed elevated, while KCNJ11 showed decreased postprandial glucagon secretion. 14

15 Page 15 of 34 Diabetes A Homozygous non-risk Heterozygous Homozygous risk 25 JAZF1 25 HHEX HNF1A 25 Glucagon (pg/islet/h) SLC2A2 ARAP1 HHEX MADD CDKN2A/2B FTO TP53INP1 WFS1 SLC30A8 CRY2 VPS13C CHCHD2P9 ADRA2A Glucagon (pg/islet/h) GIPR CAMK1D PROX1 MADD DGKB TP53INP1 THADA NOTCH2 Glucagon (pg/islet/h) CDKAL1 HMGA2 NOTCH2 FADS1 BCL11A CRY2 IGF2BP2 KLF14 TSPAN8 CDKN2A/2B JAZF1 FTO 5 ZFAND6 5 5 GIPR TCF7L2 THADA PPARG HMGA2 CDKN2A/2B CAMK1D B Insulin (ng/islet/h) Homozygous non-risk Insulin (ng/islet/h) Heterozygous Insulin (ng/islet/h) 25 Homozygous risk ARAP1 JAZF1 MADD HHEX HNF1A 20 TP53INP1 FTO CDKN2A/2B 20 PROX1 CAMK1D GIPR 20 DGKB GIPR Glucagon (pg/islet/h) KCNQ1 KCNQ1* GIPR HHEX KCNJ11 DGKB ADAMTS9 CHCHD2P9 Glucagon (pg/islet/h) ARAP1 ADRA2A DGKB THADA TP53INP1 MADD Glucagon (pg/islet/h) TP53INP1 WFS1 IGF2BP2 KCNJ11 CDKN2A/2B JAZF1 NOTCH2 HMGA2 CHCHD2P9 BCL11A PROX1 5 SLC2A2 ADRA2A 5 PPARG HMGA2 NOTCH2 5 ZBED3 KLF14 CAMK1D TCF7L2 THADA PPARG CDKN2A/2B FTO Insulin (ng/islet/h) Insulin (ng/islet/h) Insulin (ng/islet/h) Figure 4. Relationship between glucagon and insulin secretion in human pancreatic islets at low and high glucose level. Panel A shows secretion at low glucose (1 mmol/l). The risk allele carriers of variants in CAMK1D, TCF7L2, GIPR, TSPAN8 and KLF14 showed lower glucagon secretion, while the risk allele carriers of CDKAL1, BCL11A, CRY2, FADS1, NOTCH2, HMGA2 and showed glucagon levels above the expected 95% CI of the relationship between insulin and glucagon secretion. Panel B shows secretion at high glucose (16.7 mmol/l). The risk allele carriers of DGKB, GIPR, JAZF1, CDKN2A/2B, HMGA2 and NOTCH2 remained high glucagon secretion, while those of FTO, CAMK1D, TCF7L2, KLF14 and ZBED3 had low glucagon. 15

16 Diabetes Page 16 of 34 72x62mm (600 x 600 DPI)

17 Page 17 of 34 Diabetes 1593x701mm (96 x 96 DPI)

18 Diabetes Page 18 of x701mm (96 x 96 DPI)

19 Page 19 of 34 Diabetes 1599x913mm (96 x 96 DPI)

20 Diabetes Page 20 of 34 Supplementary Tables and Figures Tables Supplementary Table 1. Characteristics of the PPP-Botnia study participants. Supplementary Table 2. Effect of genetic variants on insulin secretion in non-diabetic participants of the PPP-Botnia study. Supplementary Table 3. Effect of genetic variants on BMI and insulin sensitivity in non-diabetic participants of the PPP-Botnia study. Supplementary Table 4. Effect of genetic variants on glucagon levels during OGTT in non-diabetic participants of the PPP-Botnia study. Supplementary Table 5. Effect of genetic variants on glucose levels during OGTT in non-diabetic participants of the PPP-Botnia study. Supplementary Table 6. Effect of genetic variants on insulin levels during OGTT in non-diabetic participants of the PPP-Botnia study. Supplementary Table 7. Effect of genetic variants on insulin to glucagon ratios during OGTT in nondiabetic participants of the PPP-Botnia study. Supplementary Table 8. Overview of investigated genetic variants. Figures Supplementary Figure 1. Relationship between insulin secretion (CIR) and insulin sensitivity (ISI). Supplementary Figure 2. Relationship between glucose and insulin levels during OGTT in 4,654 nondiabetic participants of the PPP-Botnia study. Supplementary Figure 3. Relationship between glucose and glucagon levels during OGTT in 4,654 non-diabetic participants of the PPP-Botnia study. Supplementary Figure 4. Graphic representation of effects of genetic variants on relationship between glucose and insulin to glucagon ratios during OGTT in non-diabetic participants of the PPP-Botnia study. Supplementary Figure 5. Relationship between glucose levels and insulin to glucagon ratios during OGTT in 4,654 non-diabetic participants of the PPP-Botnia study.

21 Page 21 of 34 Diabetes Supplementary Table 1. Characteristics of the PPP- Botnia study participants. Number (m/f) 4,654 (2,159/2,495) Age (years) 47.9 ± 15.2 BMI (kg/m 2 ) 26.2 ± 4.3 Insulin sensitivity index 147 [112] Corrected insulin response 150 [154] Disposition index 21,450 [23,450] Data are mean ± SD or median [IQR].

22 Diabetes Page 22 of 34 Supplementary Table 2. Effect of genetic variants on insulin secretion in non-diabetic participants of the PPP-Botnia study. Gene SNP CIR Beta SE P-value DI Beta SE P-value ADAMTS9 rs rs ADRA2A rs ARAP1 rs BCL11A rs CAMK1D rs CDKAL1 rs CDKN2A/2B rs CHCHD2P9 rs CRY2 rs DGKB rs DUSP9 (female) rs DUSP9 (male) rs FADS1 rs FTO rs rs GIPR rs GLIS3 rs HHEX rs HMGA2 rs HNF1A rs IGF1 rs IGF2BP2 rs JAZF1 rs KCNJ11 rs KCNQ1 rs KCNQ1 rs KLF14 rs MADD rs rs NOTCH2 rs PPARG rs PRC1 rs PROX1 rs SLC2A2 rs SLC30A8 rs TCF7L2 rs THADA rs TP53INP1 rs TSPAN8 rs VPS13C rs WFS1 rs ZBED3 rs ZFAND6 rs Beta and SE from linear regression analysis adjusted for age, gender and BMI denotes the effect size by each risk allele (additive model) on phenotype. Analyses of DUSP9 rs are stratified by gender and adjusted for age and BMI.

23 Page 23 of 34 Diabetes Supplementary Table 3. Effect of genetic variants on BMI and insulin sensitivity in nondiabetic participants of the PPP-Botnia study. Gene SNP BMI Beta SE P-value ISI Beta SE P-value ADAMTS9 rs rs ADRA2A rs ARAP1 rs BCL11A rs CAMK1D rs CDKAL1 rs CDKN2A/2B rs CHCHD2P9 rs CRY2 rs DGKB rs DUSP9 (female) rs DUSP9 (male) rs FADS1 rs FTO rs rs GIPR rs GLIS3 rs HHEX rs HMGA2 rs HNF1A rs IGF1 rs IGF2BP2 rs JAZF1 rs KCNJ11 rs KCNQ1 rs KCNQ1 rs KLF14 rs MADD rs rs NOTCH2 rs PPARG rs PRC1 rs PROX1 rs SLC2A2 rs SLC30A8 rs TCF7L2 rs THADA rs TP53INP1 rs TSPAN8 rs VPS13C rs WFS1 rs ZBED3 rs ZFAND6 rs Beta and SE from linear regression analysis adjusted for age, gender and BMI (for ISI) denotes the effect size by each risk allele (additive model) on phenotype. Analyses of DUSP9 rs are stratified by gender and adjusted for age and BMI (for ISI).

24 Diabetes Page 24 of 34 Supplementary Table 4. Effect of genetic variants on glucagon levels during OGTT in non-diabetic participants of the PPP-Botnia study. Gene SNP Fasting glucagon Beta SE P-value 2hr glucagon Beta SE P- ADAMTS9 rs rs ADRA2A rs ARAP1 rs BCL11A rs CAMK1D rs CDKAL1 rs CDKN2A/2B rs CHCHD2P9 rs CRY2 rs DGKB rs DUSP9 (female) rs DUSP9 (male) rs FADS1 rs FTO rs rs GIPR rs GLIS3 rs HHEX rs HMGA2 rs HNF1A rs IGF1 rs IGF2BP2 rs JAZF1 rs KCNJ11 rs KCNQ1 rs KCNQ1 rs KLF14 rs MADD rs rs NOTCH2 rs PPARG rs PRC1 rs PROX1 rs SLC2A2 rs SLC30A8 rs TCF7L2 rs THADA rs TP53INP1 rs TSPAN8 rs VPS13C rs WFS1 rs ZBED3 rs ZFAND6 rs Beta and SE from linear regression analysis adjusted for age, gender and BMI denotes the effect size by each risk allele (additive model) on phenotype. Analyses of DUSP9 rs are stratified by gender and adjusted for age and BMI.

25 Page 25 of 34 Diabetes Supplementary Table 5. Effect of genetic variants on glucose levels during OGTT in nondiabetic participants of the PPP-Botnia study. Gene SNP Fasting glucose Beta SE P-value 2hr glucose Beta SE P-value ADAMTS9 rs rs ADRA2A rs ARAP1 rs BCL11A rs CAMK1D rs CDKAL1 rs CDKN2A/2B rs CHCHD2P9 rs CRY2 rs DGKB rs DUSP9 (female) rs DUSP9 (male) rs FADS1 rs FTO rs rs GIPR rs GLIS3 rs HHEX rs HMGA2 rs HNF1A rs IGF1 rs IGF2BP2 rs JAZF1 rs KCNJ11 rs KCNQ1 rs KCNQ1 rs KLF14 rs MADD rs rs NOTCH2 rs PPARG rs PRC1 rs PROX1 rs SLC2A2 rs SLC30A8 rs TCF7L2 rs THADA rs TP53INP1 rs TSPAN8 rs VPS13C rs WFS1 rs ZBED3 rs ZFAND6 rs Beta and SE from linear regression analysis adjusted for age, gender and BMI denotes the effect size by each risk allele (additive model) on phenotype. Analyses of DUSP9 rs are stratified by gender and adjusted for age and BMI.

26 Diabetes Page 26 of 34 Supplementary Table 6. Effect of genetic variants on insulin levels during OGTT in nondiabetic participants of the PPP-Botnia study. Gene SNP Fasting insulin Beta SE P-value 2hr insulin Beta SE P-value ADAMTS9 rs rs ADRA2A rs ARAP1 rs BCL11A rs CAMK1D rs CDKAL1 rs CDKN2A/2B rs CHCHD2P9 rs CRY2 rs DGKB rs DUSP9 (female) rs DUSP9 (male) rs FADS1 rs FTO rs rs GIPR rs GLIS3 rs HHEX rs HMGA2 rs HNF1A rs IGF1 rs IGF2BP2 rs JAZF1 rs KCNJ11 rs KCNQ1 rs KCNQ1 rs KLF14 rs MADD rs rs NOTCH2 rs PPARG rs PRC1 rs PROX1 rs SLC2A2 rs SLC30A8 rs TCF7L2 rs THADA rs TP53INP1 rs TSPAN8 rs VPS13C rs WFS1 rs ZBED3 rs ZFAND6 rs Beta and SE from linear regression analysis adjusted for age, gender and BMI denotes the effect size by each risk allele (additive model) on phenotype. Analyses of DUSP9 rs are stratified by gender and adjusted for age and BMI.

27 Page 27 of 34 Diabetes Supplementary Table 7. Effect of genetic variants on insulin to glucagon ratios during OGTT in non-diabetic participants of the PPP-Botnia study. Fasting Insulin/glucagon 2hr Insulin/glucagon Gene SNP Beta SE P-value Beta SE P-value ADAMTS9 rs rs ADRA2A rs ARAP1 rs BCL11A rs CAMK1D rs CDKAL1 rs CDKN2A/2B rs CHCHD2P9 rs CRY2 rs DGKB rs DUSP9 (female) rs DUSP9 (male) rs FADS1 rs FTO rs rs GIPR rs GLIS3 rs HHEX rs HMGA2 rs HNF1A rs IGF1 rs IGF2BP2 rs JAZF1 rs KCNJ11 rs KCNQ1 rs KCNQ1 rs KLF14 rs MADD rs rs NOTCH2 rs PPARG rs PRC1 rs PROX1 rs SLC2A2 rs SLC30A8 rs TCF7L2 rs THADA rs TP53INP1 rs TSPAN8 rs VPS13C rs WFS1 rs ZBED3 rs ZFAND6 rs Beta and SE from linear regression analysis adjusted for age, gender and BMI denotes the effect size by each risk allele (additive model) on phenotype. Analyses of DUSP9 rs are stratified by gender and adjusted for age and BMI.

28 Diabetes Page 28 of 34 Supplementary Table 8. Overview of investigated genetic variants. Gene Chr SNP RA RAF Identified by Reference association to ADAMTS9 3 3 rs rs C A Type 2 diabetes Fasting glucose Zeggini et al. [1] Dupuis et al. [2] ADRA2A 10 rs G 0.88 Fasting glucose Dupuis et al. [2] ARAP1 11 rs A 0.79 Type 2 diabetes Voight et al. [3] BCL11A 2 rs A 0.45 Type 2 diabetes Voight et al. [3] CAMK1D 10 rs G 0.21 Type 2 diabetes Zeggini et al. [1] CDKAL1 6 rs C 0.31 Type 2 diabetes Saxena et al [4], Zeggini et al. [5], CDKN2A/2B 9 rs T 0.84 Type 2 diabetes Scott et al. [6] Saxena et al [4], Zeggini et al. [5], Scott et al. [6] CHCHD2P9 9 rs C 0.86 Type 2 diabetes Voight et al. [3] CRY2 11 rs A 0.52 Fasting glucose Dupuis et al. [2] DGKB 7 rs T 0.50 Fasting glucose Dupuis et al. [2] DUSP9 X rs G 0.25 Type 2 diabetes Voight et al. [3] FADS1 11 rs T 0.59 Fasting glucose Dupuis et al. [2] FTO 16 rs A 0.38 BMI Frayling et al. [7] 7 rs A 0.12 Fasting glucose Dupuis et al. [2] GIPR 19 rs A hour glucose Saxena et al. [8] GLIS3 9 rs A 0.50 Fasting glucose Dupuis et al. [2] HHEX 10 rs G 0.53 Type 2 diabetes Sladek et al. [9] HMGA2 12 rs C 0.07 Type 2 diabetes Voight et al. [3] HNF1A 12 rs T 0.26 Type 2 diabetes Voight et al. [3] IGF1 12 rs35767 G 0.79 Fasting insulin Dupuis et al. [2] IGF2BP2 3 rs A 0.30 Type 2 diabetes Saxena et al [4], Zeggini et al. [5], Scott et al. [6] JAZF1 7 rs A 0.47 Type 2 diabetes Zeggini et al. [1] KCNJ11 11 rs5219 T 0.49 Type 2 diabetes Gloyn et al. [10] KCNQ1 11 rs C 0.49 Type 2 diabetes Yasuda et al. [11], Unoki et al. [12] KCNQ1 11 rs G 0.49 Type 2 diabetes Voight et al. [3] KLF14 7 rs G 0.51 Type 2 diabetes Voight et al. [3] MADD 11 rs A 0.81 Fasting glucose Dupuis et al. [2] 11 rs G 0.30 Fasting glucose Prokopenko et al. [13], Lyssenko et al.[14], Bouatia-Naji et al. [15] NOTCH2 1 rs T 0.14 Type 2 diabetes Zeggini et al. [1] PPARG 3 rs C 0.86 Type 2 diabetes Altshuler et al. [16] PRC1 15 rs A 0.31 Type 2 diabetes Voight et al. [3] PROX1 1 rs C 0.48 Fasting glucose Dupuis et al. [2] SLC2A2 3 rs T 0.87 Fasting glucose Dupuis et al. [2] SLC30A8 8 rs C 0.60 Type 2 diabetes Sladek et al. [9] TCF7L2 10 rs T 0.17 Type 2 diabetes Grant et al. [17] THADA 2 rs T 0.94 Type 2 diabetes Zeggini et al. [1] TP53INP1 8 rs T 0.48 Type 2 diabetes Voight et al. [3] TSPAN8 12 rs G 0.23 Type 2 diabetes Zeggini et al. [1] VPS13C 15 rs G hour glucose Saxena et al. [8] WFS1 4 rs G 0.57 Type 2 diabetes Sandhu et al. [18] ZBED3 2 rs G 0.25 Type 2 diabetes Voight et al. [3] ZFAND6 15 rs G 0.67 Type 2 diabetes Voight et al. [3]

29 Page 29 of 34 Diabetes Homozygous non-risk Heterozygous Homozygous risk ADRA2A PPARG MADD HMGA DUSP9(f) CIR IGF1 SLC2A2 DUSP9(m) ADAMTS9 CHCHD2P9 CIR CIR ZBED3 IGF2BP2 PRC1 DUSP9(m) CDKAL1 GIPR TSPAN8 TCF7L THADA ISI ISI ISI Supplementary Figure 1. Relationship between insulin secretion (CIR) and insulin sensitivity (ISI). Whereas most SNPs showed effects which kept the relationship between ISI and CIR within the confidence interval of 4,654 non-diabetic subjects, there were some obvious outliers., CDKAL1 and risk genotype carriers had CIR values below the values expected from the ISI values, whereas TT-carriers of TCF7L2 showed rather high insulin sensitivity. N=4,654 non-diabetic participant of the Botnia-PPP study. 10