Receptor for advanced glycation end-products (RAGE) provides a link between genetic susceptibility and environmental factors in type 1 diabetes
- First Online:
- Cite this article as:
- Forbes, J.M., Söderlund, J., Yap, F.Y.T. et al. Diabetologia (2011) 54: 1032. doi:10.1007/s00125-011-2058-z
- 878 Downloads
This group of studies examines human genetic susceptibility conferred by the receptor for advanced glycation end-products (RAGE) in type 1 diabetes and investigates how this may interact with a western environment.
We analysed the AGER gene, using 13 tag SNPs, in 3,624 Finnish individuals from the FinnDiane study, followed by AGER associations with a high risk HLA genotype (DR3)-DQA1*05-DQB1*02/DRB1*0401-DQB1*0302 (n = 546; HLA-DR3/DR4), matched in healthy newborn infants from the Finnish Type 1 Diabetes Prediction and Prevention (DIPP) Study (n = 373) using allelic analysis. We also studied islets and circulating RAGE in NODLt mice.
The rs2070600 and rs17493811 polymorphisms predicted increased risk of type 1 diabetes, whereas the rs9469089 SNP was related to decreased risk, on a high risk HLA background. Children from the DIPP study also showed a decline in circulating soluble RAGE levels, at seroconversion to positivity for type 1 diabetes-associated autoantibodies. Islet RAGE and circulating soluble RAGE levels in prediabetic NODLt mice decreased over time and were prevented by the AGE lowering therapy alagebrium chloride. Alagebrium chloride also decreased the incidence of autoimmune diabetes and restored islet RAGE levels.
These studies suggest that inherited AGER gene polymorphisms may confer susceptibility to environmental insults. Declining circulating levels of soluble RAGE, before the development of overt diabetes, may also be predictive of clinical disease in children with high to medium risk HLA II backgrounds and this possibility warrants further investigation in a larger cohort.
KeywordsAdvanced glycationAlagebrium chlorideAutoimmune diabetesChildrenInsulinNODPolymorphism
Diabetes Prediction and Prevention Study
High mobility group protein B1
Minor allele frequencies
Receptor for advanced glycation end-products
Single nucleotide polymorphism
Type 1 diabetes occurs as the result of a complex disease process where genetic and environmental factors lead to an autoimmune response which, to date, remains incompletely defined . Although some major genetic determinants of type 1 diabetes, such as alleles of the major histocompatibility locus (HLA) at the HLA-DRB1 and DQB1 loci  and more recently the HLA-B*39 locus , have been identified, these only account for some 40–50% of the familial clustering. In addition, approximately 70% of monozygotic twins are discordant for the development of type 1 diabetes , which indicates that environmental triggers are also likely to be important. Indeed, the risk of developing type 1 diabetes is still increasing by 3–4% annually in developed nations , which remains unexplained.
One such environmental contributor to type 1 diabetes may be AGEs. Western diets provide excesses of these non-enzymatic modifications , which contribute flavour and colour to foodstuffs (e.g. roasted meat or coffee), while more recently AGE modifications also add functional properties, such as improved emulsification. Dietary AGEs are absorbed intestinally and are a prominent contributor to the body’s AGE pool , in addition to the natural ageing process , redox imbalances , non-diabetes related diseases including renal impairment  and hyperglycaemia . There is increasing evidence to suggest that AGEs also promote beta cell dysfunction [12, 13]. Furthermore, a reduced incidence of diabetes in a mouse model of autoimmune diabetes, the NOD mouse (NODLt) , has been reported following dietary restriction of AGEs. However, the specific mechanisms whereby insulin secretory pathways are damaged by AGEs are not known.
AGEs can exert their biological effects via receptors such as the receptor for advanced glycation end-products (RAGE) , although these compounds may have greater effects via modulation of RAGE production per se . RAGE is a multi-ligand receptor involved in host-pathogen defence, in addition to cellular apoptosis . Thus, a recent study has identified a blockade of the late stages of adoptively transferred autoimmune diabetes with the ‘decoy’ soluble RAGE receptor . In addition, a small molecule inhibitor of RAGE  or antibodies against another RAGE ligand, HMGB1 , have been shown to delay islet destruction in hyperglycaemic NODLt mice. Furthermore, the heritability of insulin resistance , another factor postulated to be contributing to the increasing rate of type 1 diabetes incidence , is associated with specific RAGE polymorphisms.
Genetic susceptibility to type 1 diabetes remains to be fully defined, in particular, the contribution of certain genotypes in the context of our increasingly Westernised environment. It is therefore unknown if a specific RAGE genotype may play a role in susceptibility to type 1 diabetes; we have now investigated this issue using a combination of human genetics, molecular biology and integrative pathophysiology.
FinnDiane/DIPP design and recruitment
Patient characteristics: patients with T1D taken from the FinnDiane study
Sex, male (%)
40.3 ± 9.9
Age at T1D diagnosis (years)
12.8 ± 7.6
Since AGER is located within the HLA class III region , we further analysed these AGER associations outlined above on a matched high risk HLA background. To achieve this, newborn infants (n = 373), who had the highest risk HLA genotype (DR3)-DQA1*05-DQB1*02/DRB1*0401-DQB1*0302 (from n = 546; HLA-DR3/DR4 ), were matched with unaffected newborn infants with the identical HLA genotype, from the Finnish Type 1 Diabetes Prediction and Prevention (DIPP) Study (n = 373) . A subgroup of children was also used for circulating soluble RAGE analyses. We included all children with the high risk HLA genotype identified from the DIPP study as controls. However, unfortunately some participants with the high risk HLA background had to be excluded because it was not possible to extract sufficient quantities of DNA for the analyses from the miniscule blood samples available. All children studied were from Finland and no other children were excluded from the present study. The ethics committees of all participating centres approved the study protocol and all patients or guardians gave written consent. The studies follow the Declaration of Helsinki.
DNA isolation and genotyping
DNA was isolated from whole blood with either a Puregene DNA Purification Kit (Gentra Systems, Minneapolis, MN, USA) according to the manufacturer’s protocol or with a phenol-chloroform protocol .
Tag SNPs for genotyping were selected from the HapMap Project’s NCBI build 34 and 35 for the CEU (Utah, USA residents with ancestry from northern and western Europe populations; n = 180), as defined by Gabriel et al.  to encompass the genetic variation of the entire AGER gene. Minor allele frequencies (MAFs) were required to be ≥0.05. Genotyping was performed with TaqMan SNP Genotyping Assays (Applied Biosystems, Foster City, CA, USA) with an ABI Prism 7900HT Sequence Detection System (Perkin-Elmer, Foster City, CA, USA) and the Sequenom multiplex platform (Sequenom, Hamburg, Germany) according to the manufacturers’ instructions.
HLA genotyping for the two high risk haplotypes was performed with a sequence specific hybridisation method using oligonucleotide probes specific for DQA1*05 and DQB1*02 (DR3-DQ2 haplotype) and DRB1*0401 and DQB1*0302 (DR4-DQ8 haplotype) as described previously .
Soluble RAGE analysis
A series of children from the DIPP study were analysed for plasma soluble RAGE. The group comprised 15 children (12 boys) who progressed to clinical type 1 diabetes during prospective observation (patients) and 15 autoantibody-negative controls (non-progressors) who remained non-diabetic and were matched with the patients for sex, date of birth and HLA genotype. Altogether 14 children carried the high risk genotype (DQA1*05-DQB1*02/DQB1*0302) and 16 patients carried moderate risk genotypes (HLADQB1*0302/x;×≠ DQB1*02, *0301, *0602 or *0603). All children provided three plasma samples. Among the patients, the first sample was obtained before seroconversion to autoantibody positivity at a mean age of 1.1 (range 0.3–3.0) years, the second sample soon after the seroconversion at a mean age of 2.4 (range 0.7–5.3) years and the third sample soon after the diagnosis of clinical type 1 diabetes at a mean age of 5.2 (range 2.0–8.9) years. In the control patients, the samples were obtained at the corresponding ages. Human plasma samples were assayed undiluted according to the manufacturer’s instructions contained within the Human RAGE ELISA (R&D Systems, Minneapolis, MN, USA). The inter-assay CV was 7.8% while the intra-assay CV was 5.7%. The limit of sensitivity was 4.12 pg/ml.
Hardy–Weinberg equilibrium (HWE) for SNPs was tested in the control individuals (n = 703), and which were considered to be in HWE when p > 0.05 using Haploview 4.0 (MIT/Harvard Broad Institute, Boston, MA, USA) . Testing for allelic associations was carried out with logistic regression adjusted for sex in SPSS 15.0 (Chicago, IL, USA).
Due to the strong linkage disequilibrium between SNPs in the HLA region and the HLA class II loci, it was necessary to exclude an effect of the HLA on the associations. Therefore, we matched patients and references for the high risk HLA genotype (DR3)-DQA1*05-DQB1*02/DRB1*0401-DQB1*0302. Logistic regression was used to evaluate allelic odds ratios.
The issue of sufficient power to detect association between AGER SNPs and type 1 diabetes was evaluated using the Genetic Power Calculator (http://pngu.mgh.harvard.edu/∼purcell/gpc/; ) with the module of case–control for discrete traits. Bonferroni correction for multiple comparisons was applied to the FinnDiane cohort under the conditions of 13 SNPs, α level 0.05, and three performed tests. Human soluble RAGE concentrations were analysed by repeated measures ANOVA.
We next studied islet and circulating RAGE production in a spontaneous experimental model which resembles type 1 diabetes, the NODLt mouse, since the genetic loci thought to be responsible for susceptibility to diabetes in these mice include the AGER gene sequence [31, 32]. NODLt or NODScid mice fed standard mouse chow (containing the AGE, carboxymethyl lysine (CML): 93.39 nmol [mol lysine]–1 [100 mg]–1) were studied prediabetes at 4, 8 and 12 weeks of age (n = 6 per group). In addition, groups of 4-week-old NODLt mice were randomised to receive either no treatment (vehicle) or the AGE-lowering therapy alagebrium chloride (1 mg kg–1 day–1 i.p.) and followed for 6 weeks (prediabetes, n = 10 per group). For incidence studies, littermate NODLt mice (n = 20 per group) were studied until the diagnosis of diabetes was confirmed or until day 200 of age, with one group randomised to receive alagebrium chloride (1 mg kg–1 day–1 i.p.) for the study duration from week 4 of age. Mice were given ad libitum access to food and water and maintained on 12 h dark–light cycles. Plasma glucose was measured by autoanalyser (Beckman Coulter LX20PRO, Brea, CA, USA) and plasma insulin concentrations were measured by radioimmunoassay (Linco Research, MO, USA).
Circulating RAGE and AGE ELISA
Soluble RAGE in mouse serum (RAGE sandwich ELISA kit, R&D systems, Minneapolis, MN, USA) was assayed undiluted according to the manufacturer’s instructions. The inter-assay CV was 6.7% and the intra-assay CV was 5.9%. The limit of sensitivity of the assay was 1 pg/ml. The AGE, CML, was assessed in serum or plasma using an in-house ELISA. The limit of detection of the assay was 8.0 nmol/mol lysine. The inter-assay CV was 7.3%. The intra-assay coefficient of variation was 5.5%.
Formalin fixed paraffin sections  were incubated with primary antibody (AGE , 1:1,000; goat anti-RAGE 1:2,500, Chemicon, Temecula, CA, USA) overnight at 4°C. Tissue sections were consecutively stained with biotinylated IgG for 10 min and avidin–biotin horseradish peroxidase complex for 15 min (Vectastain ABC Elite kit, Vector Laboratories, Burlingame, CA, USA) before 3,3′-diaminobenzidine tetrahydrochloride treatment (DAB; Sigma Chemical Co., St Louis, MI, USA). Negative control sections omitted the primary antibody. Positive control tissues were also included. Quantitation of islet immunostaining was completed by computer-aided densitometry (Image Pro Plus 6.0, Media Cybernetics, Bethesda, MD, USA) where all islets were analysed (×100) and results expressed as proportional to the area of positive staining.
Real-time reverse transcription-polymerase chain reaction
RNA obtained from pancreatic islet tissue collected immediately was later used to synthesise cDNA with the Superscript First strand synthesis system for RT-PCR (Gibco BRL, Grand Island, NY, USA). Gene expression for AGER was analysed by real-time quantitative RT-PCR performed with the TaqMan system based on real-time detection of accumulated fluorescence (ABI Prism 7700, Perkin-Elmer) normalised to 18 S as previously . For mouse Ager (NM_007425), the probe was 6-FAM CACAGCCCGGATTG-MGB, the forward primer was 5′-GCTGTAGCTGGTGGTCAGAACA-3′ and the reverse primer was 5′-CCCCTTACAGCTTAGCACAAGTG-3′.
Pancreatic islets were isolated as previously described . Briefly, 10 ml of cold Hanks balanced salt solution (HBBS) containing 0.75 mg/ml collagenase type V (Sigma Chemical Co.) was injected into the bile duct. The pancreas was digested at 37°C for 10–20 min and then disrupted and filtered through a 500 μm mesh. The pancreatic islets were separated from exocrine tissue by histopaque density gradient, for which the islets were suspended in histopaque 1.119 g/l, followed by layering of histopaque 1.083 g/l and histopaque 1.077 g/l (Sigma). The hand-picked islets were then used for flow cytometry as outlined below and RNA extraction . For flow cytometry, purified islets were washed (10 mM EDTA in Hanks’ balanced salt solution) and stained using goat anti-RAGE (Santa Cruz Biotechnology, Santa Cruz, CA, USA) followed by rabbit anti-goat IgG, F(ab′)2, FITC conjugated (Chemicon International, Temecula, CA, USA).
Mouse results are expressed as mean ± standard deviation unless otherwise specified. All analyses of rodent data with more than two groups were performed by ANOVA followed by post hoc analysis with Tukey’s test. Student’s t test or Mann–Whitney analyses were used to compare two groups where appropriate and are labelled as such. A logrank (Mantel–Cox) test was used to analyse diabetes incidence in NOD mice. A value of p < 0.05 was considered to be statistically significant (GraphPad Prism; GraphPad Software, San Diego, CA, USA).
AGER gene polymorphisms and type 1 diabetes susceptibility
SNP associations between AGER and type 1 diabetes: analysis of genomic DNA from patients with T1D from the FinnDiane cohort and from newborn infants with the highest risk HLA genotype (DR3)-DQA1*05-DQB1*02/DRB1*0401-DQB1*0302 obtained as part of the Finnish Type 1 DIPP
Position NCBI35 Chr 6 (bp)
The FinnDiane cohort
OR (95% CI)b
Sample size (n)
1.82 × 10−11 c
1.45 × 10−39 c
6.30 × 10−10 c
1.27 × 10−23 c
2.82 × 10−9 c
2.76 × 10−9 c
5.89 × 10−5 c
2.71 × 10−6 c
1.53 × 10−4 c
3.03 × 10−16 c
3.65 × 10−5
1.36 × 10−30 c
6.34 × 10−4 c
1.23 × 10−8 c
We performed allelic analyses, which identified that three SNPs (rs2070600, rs9469089 and rs17493811) were still associated with type 1 diabetes (p = 0.017, 3.65 × 10−5, and 0.031, respectively) in HLA-matched children from the DIPP (Table 2). In particular, the rs2070600 and rs17493811 polymorphisms predicted increased risk of type 1 diabetes (ORs 1.452 and 1.518, respectively) whereas rs9469089 was related to decreased risk (OR 0.423) in the high risk HLA setting. However, using less stringent HLA adjustment, i.e. analysing those individuals with the risk alleles DQB1*02 and/or DQB1*0302 (type 1 diabetes patients n = 2,660 and controls n = 294), all AGER SNPs apart from rs204993 remained associated with type 1 diabetes following HLA matching (data not shown).
Circulating soluble RAGE concentrations
Islet RAGE and circulating soluble RAGE decline prediabetes in NODLt mice
Metabolic variables for NOD mouse groups
2.6 ± 0.1
18.3 ± 1.3
2.2 ± 0.1
23.7 ± 1.7*
2.4 ± 0.1
27.7 ± 1.0*†
2.8 ± 0.3
18.1 ± 1.6
3.1 ± 0.1
23.4 ± 1.6*
2.8 ± 0.1
24.0 ± 1.3*‡
Many chronic diseases are likely to be polygenic, including type 1 diabetes, where less than 50% of the familial clustering can be attributed to currently reported genotypes. This highlights the importance of discovering other candidate genes and environmental contributors. Within the present study, we have identified three polymorphisms in the AGER gene, which encodes for the proteins RAGE and soluble RAGE, that may partly explain the genetic susceptibility to type 1 diabetes in humans, even in individuals with a high risk HLA class II background. Of particular interest is the finding that a temporal decline in circulating concentrations of the protective isoform of RAGE (termed soluble RAGE) at seroconversion to produce autoantibodies, predicted the clinical onset of type 1 diabetes in patients within high to moderate risk class II HLA genotypes. These findings are reminiscent of changes in NODLt mice, in whom seroconversion appears concurrently with a decrease in circulating soluble RAGE concentrations . There was also an inverse relationship between circulating CML (an AGE) and either soluble RAGE concentrations or islet RAGE expression seen in both NODLt and NODScid mice. Interestingly, despite elevated circulating AGE levels, islet AGE concentrations were also lower in NODLt mice compared with NODScid mice, suggesting that RAGE may be involved in the cellular uptake or intracellular production of AGEs in pancreatic islet cells.
A decline in circulating soluble RAGE has also been shown in children during active autoimmune disease, such as Kawasaki disease or systemic onset juvenile idiopathic arthritis . Decreases in plasma soluble RAGE were also seen in samples from children who progressed to type 1 diabetes at seroconversion to autoantibodies, compared with samples obtained before seroconversion. At the diagnosis of diabetes, those individuals who progressed to type 1 diabetes in the DIPP study had an increase in plasma soluble RAGE, probably as the result of active inflammation, as has been seen previously in individuals with type 2 diabetes . Therefore, although tantalising, the changes in soluble RAGE seen within the present study in children who progressed to type 1 diabetes need to be confirmed in larger patient cohorts. In addition, since soluble RAGE concentrations were not measured in the individuals from whom we have the AGER genotype data, Mendelian randomisation was not possible in the present populations.
The genetic loci thought to be responsible for susceptibility to autoimmune diabetes in NODLt mice include the AGER gene sequence encoding RAGE proteins [31, 32]. The presence of the serine residue within the G82S polymorphism, identified using AGER rs2070600 in humans, has previously been associated with lower  or elevated  circulating soluble RAGE concentrations. Interestingly, this SNP was also associated with a higher risk of type 1 diabetes in our human cohorts in the present study. In our human studies, it is difficult to increase the numbers of patients and controls studied, given the key requirement to control for high risk HLA genotypes to delineate the association between AGER for type 1 diabetes. It is nevertheless intriguing that, even in such a small HLA-matched sample, there were clear genotypic differences and changes in circulating soluble RAGE concentrations between those with type 1 diabetes and those without. However, this finding now needs to be replicated in larger human populations.
One cannot discount, however, the possibility that changes in the proteases responsible for the cleavage of membrane RAGE to soluble RAGE  may also be altering the circulating soluble RAGE pool in our diabetic individuals and mice, independent of changes in the AGER gene. Indeed, this possibility should be further investigated in subsequent studies, in particular given the effects of AGEs on circulating RAGE concentrations in NODLt mice observed in the present study. Interestingly, prediabetic NODLt mice in the present study also had a transient decline in islet RAGE expression at seroconversion to autoantibody positivity which was not seen in NODScid mice. Furthermore, we have determined that circulating soluble RAGE concentrations correlate with islet RAGE expression. In the future, it would be desirable to confirm these findings using a NODLt mouse with an islet specific knockout of the gene encoding RAGE (AGER). It would also have been highly desirable to confirm a loss of islet RAGE expression in human participants with type 1 diabetes, although we have shown that decreases in circulating soluble RAGE concentrations in children with medium to high risk HLA II genetic backgrounds may be predictive of this loss in islet RAGE expression. Therefore, at this stage we can only speculate as to the association of RAGE with type 1 diabetes.
Some previous studies in adoptively transferred autoimmune diabetes and syngeneic islet transplants in hyperglycaemic NODLt mice have shown benefits from the interruption of RAGE ligand binding with small molecule inhibitors of RAGE  or exogenous soluble RAGE . In adoptively transferred diabetes , however, it could only be concluded that RAGE ligands may be involved in the differentiation of T cells to a mature phenotype. Indeed, NODScid mice also have neither T nor B cells, providing further evidence that RAGE expression on T cells, such as Th1 [18, 19], may also be important in the pathogenesis of diabetes, given that NODScid mice do not develop autoimmune diabetes.
Prediabetic NODLt mice in the present study showed increases in circulating RAGE ligands, AGEs, that were not seen with respect to another RAGE ligand, HMGB1. This was surprising given the previous association of HMGB1 to autoimmune diabetes in NODLt mice . However, another ligand of RAGE is the amylin polypeptide, which is co-produced with insulin by the pancreatic beta cells in response to elevation of plasma glucose . Moreover, there is speculation that insulin may also bind to RAGE, since there is evidence of amyloidosis composed of iatrogenic A-Ins type amyloid . Indeed, other investigators have suggested that proinflammatory RAGE and HLA II DRB1 polymorphisms may synergise to activate the immune response in diabetes complications . However, whether this affects the immune system during the development of autoimmune diabetes or type 1 diabetes remains to be determined.
Importantly, lowering AGEs using alagebrium chloride in the present study not only increased circulating soluble RAGE concentrations in NODLt mice at seroconversion to autoantibodies, but also reduced the incidence of autoimmune diabetes. Taken together, one could speculate that it is important therapeutically to prevent ligand-induced chronic modulation of RAGE to protect islet function. Also of interest was the fact that it is unlikely that increases in plasma glucose are driving the elevations in plasma AGEs seen within the present study, as is classically seen in diabetes complications [33, 49, 50], since elevations in circulating AGEs by 8 weeks of age were also evident in NODScid mice, which do not develop diabetes. These results raise the possibility that precipitation of diabetes could occur as the result of either genetic or environmental modulation of RAGE.
Another limitation of the present study is that so-called high risk HLA haplotypes are ‘losing power’ in the younger age groups developing type 1 diabetes . However, FinnDiane is collecting data on adult type 1 diabetic patients, rather than paediatric patients, and therefore the mean duration of type 1 diabetes is relatively high (24.4 years, or median of 24.2 years), meaning that less than 10% of the patients studied were born in the late 1970s and 1980s, when the HLA trend mentioned above was primarily observed. Furthermore, given that the controls for the FinnDiane study were recruited through the Finnish blood bank, we have little epidemiological data relating to these individuals.
Thus, in summary, the data presented here show that inherited AGER gene polymorphisms should be considered as novel contributors to susceptibility to type 1 diabetes. This group of studies also suggests that excesses of environmental factors such as AGEs might modulate changes in islet and circulating RAGE expression, which may contribute to insulin secretory defects and ultimately the development of overt diabetes. In addition, declining circulating levels of soluble RAGE, which binds excesses of AGEs, at seroconversion before the development of overt diabetes, may warrant investigation in larger human cohorts as a predictor of clinical type 1 diabetes. Thus, we not only propose a novel mechanism for the development of diabetes, but also suggest a potential treatment, the AGE lowering therapy alagebrium chloride, currently under clinical investigation, albeit for other medical indications. Furthermore, we believe that these results challenge current thoughts in this area, by hypothesising that genetic or environmental declines in RAGE may be important for the development of type 1 diabetes. These findings are also likely to have implications for the pathogenesis of other chronic diseases in which this receptor is considered to play a pivotal role.
We are indebted to M Parkkonen and R Sallinen, University of Helsinki, Helsinki, Finland, who contributed to the genotyping of the FinnDiane patients. The authors would also like to thank K Gilbert, V Thallas and M Arnstein, Baker IDI Heart and Diabetes Institute, Melbourne, Australia, for their technical expertise. Thanks also to A Blair for assistance with rat and mouse IV/IPGTTs. This work was completed with support from an innovative grant from the Juvenile Diabetes Research Foundation (JDRF; 5-2010-163). J Forbes is a JDRF Career Development Fellow. S Andrikopoulos is a recipient of an RD Wright Fellowship from the National Health and Medical Research Council of Australia (NHMRC) of Australia. M Cooper is a JDRF Scholar and an NHMRC Australia Fellow. The FinnDiane is funded by the Folkhälsan Research Foundation, Wilhelm and Else Stockmann Foundation, Sigrid Juselius Foundation, European Commission (LSHB-CT-2003-503364, LSHB-CT-2006-037681), Academy of Finland (214335 and 124280 to M.W.), Medicinska understödsföreningen Liv och Hälsa, Signe and Ane Gyllenberg Foundation, Waldemar von Frenkell Foundation and Governmental Grants for Health Sciences Research. We acknowledge all the physicians and nurses at each participating centre for their invaluable role in patient recruitment, collection of samples and data (see ESM). The DIPP Study is supported by grants from the Juvenile Diabetes Research Foundation (grants 4-1998-274, 4-1999-731, 4-2001-435), Emil Aaltonen Foundation, Academy of Finland, Jalmari and Rauha Ahokas Foundation, Diabetes Research Foundation, Finland, Signe and Ane Gyllenberg Foundation, Yrjö Jahnsson Foundation, Sigrid Juselius Foundation, Novo Nordisk Foundation, Päivikki and Sakari Sohlberg Foundation, and Special Research Funds for the Oulu, Tampere and Turku University Hospitals.
Duality of interest
The authors declare that there is no duality of interest associated with this manuscript.