Pediatric Diabetes 2014: 15: 416–421doi: 10.1111/pedi.12105All rights reserved

© 2013 John Wiley & Sons A/S.Published by John Wiley & Sons Ltd.

Pediatric Diabetes

Original Article

No association between type 1 diabetesand genetic variation in vitamin D metabolismgenes: a Danish study

Thorsen SU, Mortensen HB, Carstensen B, Fenger M, Thuesen BH,Husemoen L, Bergholdt R, Brorsson C, Pociot F, Linneberg A, Svensson J.No association between type 1 diabetes and genetic variation in vitamin Dmetabolism genes: a Danish study.Pediatric Diabetes 2014: 15: 416–421.

Background: Vitamin D, certain single nucleotide polymorphisms (SNPs) inthe vitamin D-receptor (VDR) gene and vitamin D metabolism genes havebeen associated with type 1 diabetes (T1D).Objective: We wanted to examine if the most widely studied SNPs in genesimportant for production, transport, and action of vitamin D were associatedwith T1D or to circulating levels of vitamin D 25-hydroxyvitamin D[25(OH)D] in a juvenile Danish population.Methods: We genotyped eight SNPs in five vitamin D metabolism genes in1467 trios. 25(OH)D status were analyzed in 1803 children (907 patients and896 siblings).Results: We did not demonstrate association with T1D for SNPs in thefollowing genes: CYP27B1, VDR, GC, CYP2R1, DHCR7, and CYP24A1.Though, variants in the GC gene were significantly associated with 25(OH)Dlevels in the joint model.Conclusion: Some of the most examined SNPs in vitamin D metabolismgenes were not confirmed to be associated with T1D, though 25(OH) levelswere associated with variants in the GC gene.

Steffen U Thorsena, HenrikB Mortensena, BendixCarstensenb, MogensFengerc, Betina H Thuesend,Lotte Husemoend, RegineBergholdte, CarolineBrorssonf, FlemmingPociotf,g, Allan Linnebergd

and Jannet Svenssona

aHerlev University Hospital, Herlev,Denmark; bSteno Diabetes Center,Gentofte, Denmark; cDepartment ofClinical Biochemistry and MolecularBiology, Hvidovre Hospital, Hvidovre,Denmark; dResearch Centre forPrevention and Health, GlostrupUniversity Hospital, Glostrup,Denmark; eHagedorn ResearchInstitute, Gentofte, Denmark; fGlostrupResearch Institute, Glostrup UniversityHospital, Glostrup, Denmark; andgDepartment of Biomedical Science,University of Copenhagen,Copenhagen, Denmark

Key words: cholecalciferol – diabetesmellitus, type 1 – polymorphism,single nucleotide – receptors –vitamin D

Corresponding author: Steffen UllitzThorsen, MD,Herlev University Hospital,Herlev Ringvej 75,DK-2730 Herlev,Denmark.Tel: +45-22-28-8995;fax: +45-38-68-5012;e-mail: s.u.thorsen@gmail.com

Submitted 5 July 2013. Acceptedfor publication 1 November 2013

416

T1D and SNPs in vitamin D metabolism genes

Type 1 diabetes (T1D) is looked upon as amulti-factorial disease, where both polygenic and envi-ronmental factors are contributors. In search of envi-ronmental factors, a north-south gradient and seasonalvariation in incidence have linked ultraviolet radiation(UVR) and hence vitamin D (25-hydroxyvitamin D(25(OH)D)) to the pathogenesis of T1D (1).

So far the anti-autoimmune properties of vitamin Dhave mostly been studied in animal models (2, 3), whichagain have been supported by epidemiologic studiesthat have shown a decreased risk of T1D in childrengiven vitamin D supplementation in early childhood(4–6), and lower levels of 25(OH)D in newly diagnosedpatients compared to healthy controls (7–9). Noneof these studies included information on the genesinvolved in vitamin D metabolism.

T1D genetic determinants of utmost importance arefound within the human leukocyte antigen (HLA) classII locus though other non-HLA gene regions have beenidentified – among the vitamin D metabolism genes,the vitamin D-receptor (VDR) gene is the most studied(10, 11).

The VDR, a ligand-activated transcription factor,is located in most tissues, including the pancreasand immune system cells. 25(OH)D is primarilyhydroxylated in the kidneys by the CYP27B1 enzymeto form 1,25(OH)2D, which exerts its systemic effectsvia binding to the VDR. The discovery of certainimmune cells, e.g., the dendritic cell’s capability ofconverting 25(OH)D to its active form (1,25(OH)2D)via CYP27B1 (1α-hydroxylase) expression promoteshypotheses on the importance of auto and paracrinestimulation for a tolerogenic immune system – theseprocesses have been shown to rely on an optimal25(OH)D level (12).

The pleiotropic actions of vitamin D through itsreceptor have been proven to influence immune andother pathways (13). The role of single nucleotide poly-morphisms (SNPs) in the VDR gene, e.g., Fok1, Bsm1,Taq1, and Apa1 has been unclear. Two meta-analyseshave been conducted to elucidate an associationbetween T1D risk and the aforementioned SNPs, themost comprehensive meta-analysis found Bsm1 to beassociated with overall increased risk of T1D, but aftersubgroup analysis by ethnicity the association wasonly significant in the Asian population (10, 14).

A few studies have found that polymorphismsin the gene that encodes the enzyme CYP27B1increase the risk of developing T1D (15–17).Interestingly, a large genome-wide association studyfound that genetic variants at four loci – involved invitamin D metabolism – were associated with 25(OH)levels: GC/4p12 (rs2282679), a gc-globulin/vitaminD-binding protein; CYP2R1/11p15 (rs10741657), anenzyme that converts cholecalciferol (vitamin D3)and ergocalciferol (vitamin D2) into 25(OH)D;

DHCR7/11q12 (rs12785878), an enzyme that converts7-dehydrocholesterol to cholesterol and therebyremoves the substrate from the synthetic pathwayof 25(OH)D; and CYP24A1/20q13 (rs6013897), anenzyme that initiates degradation of 1,25(OH)2D) tobe associated with 25(OH)D levels (18).

This genetic association with 25(OH)D levels hassince been replicated in healthy individuals and threeof the four loci were also associated with the risk ofT1D (17).

On the basis of a relatively large population sampleof children newly diagnosed with T1D and theirsiblings, we have the following hypotheses:

i Vitamin D metabolism genotypes are associated withthe risk of developing T1D.

ii Certain vitamin D metabolism genotypes influence25(OH)D serum levels.

Methods

Study population

Data were obtained from a large population-basedregister of diabetic children called DanDiabKids – theregistry has an associated biobank. Established in1996, the register includes incidence data and clinicalinformation from more than 4000 newly diagnosedcases for individuals with T1D aged 0–18. The biobankassociated with the register contains blood samples onapproximately 75% of all children and their first-degreerelatives. Among the latter, there are more than 2300healthy siblings less than 18 yr of age.

About 1467 affected offspring trios of Danishorigin were genotyped – all patients with measured25(OH)D were included – independent of time ofblood sampling. Additionally, 358 extra trios wereavailable for genotyping and therefore included.

For 25(OH)D analysis, a random sample of 1077cases with blood sampling less than 3 months afteronset were chosen. The date of onset was definedas the date of first insulin injection, and diabetesduration thereafter was measured in months. A randomsample of 983 siblings from the biobank was matchedwith the cases for age, gender, month of sampling,and sample year representing the entire study period.Blood samples from the same family were taken within1 month in 89% of the families.

Of the 2060 children with available 25(OH)D, 88(two patients and 86 siblings) were excluded becauseof age above 18 yr. Material was missing for twopatients and one sibling. We excluded 166 patientswhere samples were taken within the first 2 days afteronset due to a dramatic downward shift in 25(OH)Dlevels after the first 2 days – longitudinal sampling wasnot performed. The association between the eight SNPs

Pediatric Diabetes 2014: 15: 416–421 417

Thorsen et al.

and 25(OH)D levels were tested in 1803 children (907patients and 896 siblings where 859 were females and944 were males). The children were aged 0–18 yr, witha mean of 10.6 yr (SEM 0.13) for patients and 9.9 yr(SEM 0.12) for siblings.

Serum samples have been stored at −80◦C duringthe entire study period (1997–2009). The study wasperformed according to the criteria of the HelsinkiII Declaration and was approved by the DanishNational Committee on Biomedical Research Ethics(H-KA-20070009). For all the patients, their parentsor guardians gave informed consent.

Genotyping

Eight SNPs in or near genes associated with themetabolism of vitamin D were genotyped usingpredesigned TaqMan SNP genotyping assays, and runon the TaqMan 7900 HT (Applied Biosystems, FosterCity, CA, USA) or the CFX384 (Bio-Rad) systems.The SNPs genotyped were Bsm1 (C>T) (rs1544410);Fok1 (C>T) (rs2228570); and Apa1 (A>C) (rs7975232)in VDR; rs4646536 (T>C) in CYP27B1; rs2282679(A>C) in GC; rs10741657 (G>A) near CYP2R1;rs12785878 (T>G) near DHCR7; and rs6013897(T>A) in CYP24A1.

In CYP27B1, two SNPs rs4646536 (+2838 (T>C)and rs10877012 (−1260 C>A) have previously beengenotyped. However, as they are in perfect linkagedisequilibrium (16), we only evaluated rs4646536.

25(OH)D and parathyroid hormone

Vitamin D status was measured as serum 25(OH)Dby high-performance liquid chromatography (19).Detection limit for 25(OH)D was 9.5 nmol/L, witha variance of 8%. Twenty-five children had a levelbelow detection limit and three had missing values.Parathyroid hormone (PTH) levels were measured bycompetitive chemiluminescent enzyme immunoassays(IMMULITE 2000 System; Siemens Healthcare

Diagnostics, Deerfield, IL, USA). Detection limit was0.33 pmol/L, with a variance of 6.6–10%. Ten childrenhad missing values and 589 were below detection limit.

Statistical methods

Association with T1D was evaluated by means ofthe transmission disequilibrium test (TDT) in plink

version 1.07 (20).In addition the impact of genotypes on 25(OH)D

and PTH levels were modeled based on a linear mixedmodel including a random effect and correlation withinfamilies using Proc Mixed in sas9.2. The model is fur-ther adjusted for sample year, age, and season (monthof sample). The levels of vitamin D and PTH were log-transformed before analysis to meet the assumption ofthe statistical model (normally distributed residuals).A p value of 0.05 was considered to be statisticallysignificant. Furthermore, the p value for the SNPsassociation was corrected for eight multiple tests usingthe Bonferroni method (8 SNPs pcorrected = 0.05/8 =0.00625).

Results

Genes

The eight SNPs we selected for examination are allcommon SNPs with minor allele frequencies rangingfrom 0.21 to 0.45. We did not demonstrate associationwith T1D – in the 1467 trios – for any of the followingSNPs: CYP27B1, Bsm1, Fok1, Apa1, GC, CYP2R1,DHCR7, and CYP24A1 [p (TDT) = 0.49, 0.22, 0.85,0.92, 0.90, 0.88, 0.21, and 0.65, respectively) (Table 1).All SNPs were in Hardy–Weinberg equilibrium. Therewas no evidence of a different distribution of genesaccording to gender.

25(OH)D levels and genes

Variants in the GC gene were significantly associatedwith 25(OH)D levels (p = 0.0003) even after correcting

Table 1. Results of transmission disequilibrium test (TDT) of association between single nucleotide polymorphisms (SNPs) invitamin D-related genes and risk of type 1 diabetes (T1D) in a Danish family cohort

SNP Chromosome Position Transmitted Untransmitted Odds ratio p value Gene

rs7975232 (A>C) 12 46525104 711 715 0.99 0.9156 VDR (Apa1)rs1544410 (C>T) 12 46526102 666 712 0.94 0.2153 VDR (Bsm1)rs2228570 (C>T) 12 46559162 671 678 0.99 0.8488 VDR (Fok1)rs4646536 (T>C) 12 56444255 623 648 0.96 0.4832 CYP27B1rs2282679 (A>C) 4 72608383 543 539 1.01 0.9032 GCrs10741657 (G>A) 11 14914878 734 728 1.01 0.8753 CYP2R1rs12785878 (T>G) 11 71167449 597 641 0.93 0.2111 DHCR7rs6013897 (T>A) 20 52742479 483 469 1.03 0.65 CYP24A1

Position: base pair position on genome build hg18; Transmitted: number of transmitted minor alleles; Untransmitted: numberof untransmitted minor alleles. Odds ratio and p value from the TDT test. Gene: SNP location in or near the closest gene.

418 Pediatric Diabetes 2014: 15: 416–421

T1D and SNPs in vitamin D metabolism genes

Table 2. SNPs from vitamin D metabolism genes and25(OH)D concentration (nmol/L)

Median Q1 Q3

Case Case Case

Patient Sibling Patient Sibling Patient Sibling

GenotypesGC (A>C)

AA 61.00 59.42 43.00 39.62 88.34 88.10AC 56.39 57.00 35.41 35.00 90.00 87.00CC 51.51 47.00 32.00 30.00 85.00 73.00

DHCR7 (T>G)TT 61.00 59.23 41.00 38.75 85.38 87.24TG 58.00 56.58 35.41 35.00 92.00 86.00GG 56.40 51.71 40.00 32.00 74.08 92.70

Bsm1 (C>T)CC 56.90 57.77 38.04 37.00 83.88 84.09CT 60.06 58.00 39.00 38.00 91.00 89.00TT 57.00 57.00 38.00 32.00 87.00 81.00

Fok1 (C>T)CC 57.07 58.92 37.51 38.38 91.00 87.00CT 59.83 56.65 39.81 36.00 85.00 88.10TT 62.00 53.79 39.50 34.75 100.00 78.00

Apa1 (A>C)AA 59.00 51.71 40.00 32.38 89.00 85.04AC 59.59 59.51 37.19 40.21 89.08 88.10CC 58.00 56.11 36.99 37.95 84.19 84.09

CYP27B1 (T>C)TT 60.00 59.45 39.00 36.27 93.76 91.67TC 58.00 56.00 37.51 37.50 82.59 84.92CC 60.00 52.72 39.22 35.41 87.87 71.41

CYP2R1 (G>A)GG 58.00 52.00 39.81 35.00 89.00 83.00GA 59.00 59.40 39.00 39.55 87.87 91.60AA 60.60 58.09 35.95 33.56 88.34 83.00

CYP24A1 (T>A)TT 58.70 58.00 37.51 37.50 85.16 89.00TA 59.83 59.00 39.22 36.17 90.25 83.00AA 58.00 44.00 39.70 31.50 103.10 69.00

for multiple testing. 25(OH)D levels being 21.3% (CI95% 6.8; 37.9%) higher in carriers homozygous for themajor allele and 7.6% (CI 95% 5.4; 22.3%) higher inheterozygous carriers when compared to the carriershomozygous for the minor allele in the adjusted model,whereas none of the other genes were associated with25(OH)D levels (Table 2).

Age and gender

We found a linear decrease of 2.0% [(CI 95% 1.2; 2.8);p < 0.001] per year in 25(OH)D levels and an increaseof 1.4% per year [(CI 95% 0.1; 2.7%); p = 0.04] in PTHin both patients and siblings. The increase in PTHdeviated slightly from a linear increase but we havechosen to report the linear factor for simplicity. Therewas no difference in 25(OH)D between gender.

Seasonal variation

25(OH)D levels varied significantly with months ofsampling, having the highest levels from July untilSeptember (p < 0.001). There was no difference in seas-onal variation between patients and siblings (p = 0.57).

Discussion

The aim of our study was to actively participate inthe on-going examination of whether or not commonSNPs in the VDR gene or genes involved in vitamin Dtransport (GC), hydroxylation (CYP27B1, CYP2R1,and CYP24A1), or cholesterol synthesis (DHCR7)influence the risk of developing T1D. Furthermore,we wanted to contribute to a more thorough under-standing of how – or if – such SNPs could influence25(OH)D homeostasis.

In our population of Danish children with newlydiagnosed T1D, we could not confirm previous findingsof an association between T1D and any of theeight tested SNPs. Though we found a significantassociation between genetic variation in the GC locusand 25(OH)D levels.

Our findings are supported by a meta-analysis thatwas conducted on 42 case–control-type and family-transmission-type studies from 1997 to 2005, whichfocused mainly on juvenile T1D (14). Notably, we onlyreanalyzed data on the three most-studied SNPs in theVDR gene, and there are over 200 other SNPs in thesame gene (14).

Contrary to the study performed in a British popula-tion (17), we could not confirm an association betweenT1D and DHCR7, CYP2R1, CYP27B1 and CYP24A1.Ethnic differences in genetic polymorphisms and geneexpression must also be taken into account for discrep-ancy in results between countries. This was shown ina recently conducted large meta-analysis, which foundthat the Bsm1 polymorphism was only significantlyassociated with T1D risk in people with Asian descentafter population stratification (10). It has been pro-posed that the SNPs associated with T1D might not becausal variants, but rather be in linkage disequilibriumwith these variants – ethnicity might cause a differentdistribution of these yet unknown SNPs (18).

Paradoxically, the most associated locus with25(OH)D levels, GC, in the aforementioned Britishstudy showed no significant effect on T1D risk (17).We also replicated the association between the GClocus and 25(OH) levels, but found no associationwith T1D.

Our study consists of a homogeneous populationwhere blood sampling was performed within 3 monthsafter onset, which is definitely a strength in regard to25(OH)D levels. It can be argued that the TDT analysisdoes not have the same power as a case–control

Pediatric Diabetes 2014: 15: 416–421 419

Thorsen et al.

association analysis to detect a difference between T1Dand SNPs.

We confirm a variation with season and have previ-ously shown variation with year, thereby highlightingthe importance of sampling patients and controls notonly by season but also by year – the latter due tochange in ‘hours of bright sunshine’ (21). Matching onage should also be considered as we found decreasinglevels in older children – as have others (22).

The intensive study of the vitamin D metabolismgenes is driven by the hypothesis that vitamin Dhas an important role in the pathogenesis of T1D.Epidemiologic studies have shown that T1D patientshave lower levels of 25(OH)D around or years afteronset when compared to healthy controls (7–9, 17). Inour previous study, we could not confirm these findings(21). Moreover, we did not find lower 25(OH)Dlevels in the most recent cohorts in our study period(1996–2009), which does not support the hypothesisthat vitamin D is associated to the increasing incidenceof T1D in Denmark – at least not around onset (21).‘The window of opportunities’ for optimal levels of25(OH)D could be found in utero or in the first yearsof life when the immune system is being ‘educated’ tobecome self-tolerant (23, 24). Further research in thisfield is anticipated with great interest.

Conclusion

In summary, the hypothesis that a different distributionof SNPs from vitamin D metabolism genes is associatedwith T1D was not confirmed by our study. Though anassociation between genetic variation in the GC locusand 25(OH)D levels was confirmed.

Acknowledgement

We greatly acknowledge the members of the Danish StudyGroup for Diabetes in children and adolescents for collectingthe material.

Conflict of interest

The authors declare no conflicts of interest.

References

1. Ponsonby A-L, Pezic A, Ellis J et al. Variation inassociations between allelic variants of the vitamin Dreceptor gene and onset of type 1 diabetes mellitusby ambient winter ultraviolet radiation levels: a meta-regression analysis. Am J Epidemiol 2008: 168: 358–365.

2. Takiishi T, Van Belle T, Gysemans C, Mathieu

C. Effects of vitamin D on antigen-specific and non-antigen-specific immune modulation: relevance for type1 diabetes. Pediatr Diabetes 2013: 14: 81–89.

3. Wolden-Kirk H, Overbergh L, Christesen HT,Brusgaard K, Mathieu C. Vitamin D and diabetes:

its importance for beta cell and immune function. MolCell Endocrinol 2011: 347 (1–2): 106–120.

4. Substudy TE, Group S. Vitamin D supplement in earlychildhood and risk for type I. Diabetologia 1999: 42:51–54.

5. Hypponen E, Laara E, Reunanen A, Jarvelin MR,Virtanen SM. Intake of vitamin D and risk of type1 diabetes: a birth-cohort study. Lancet 2001: 358:1500–1503.

6. Stene LC, Joner G. Use of cod liver oil duringthe first year of life is associated with lower risk ofchildhood-onset type 1 diabetes: a large, population-based, case–control study. Am J Clin Nutr 2003: 78:1128–1134.

7. Pozzilli P, Manfrini S, Picardi A et al. Low levels of25 − hydroxyvitamin D 3 and 1, 25 − dihydroxyvitaminD 3 in patients with newly diagnosed type 1 diabetes.Horm Metab Res 2005: 37: 680–683.

8. Borkar VV, Devidayal , Verma S, Bhalla AK. Lowlevels of vitamin D in North Indian children with newlydiagnosed type 1 diabetes. Pediatr Diabetes 2010: 11:345–350.

9. Littorin B, Blom P, Scholin A et al. Lower levelsof plasma 25-hydroxyvitamin D among young adults atdiagnosis of autoimmune type 1 diabetes compared withcontrol subjects: results from the nationwide DiabetesIncidence Study in Sweden (DISS). Diabetologia 2006:49: 2847–2852.

10. Zhang J, Li W, Liu J et al. Polymorphisms in thevitamin D receptor gene and type 1 diabetes mellitusrisk: an update by meta-analysis. Mol Cell Endocrinol2012: 355: 135–142.

11. Todd JA. Etiology of type 1 diabetes. Immunity 2010:32: 457–467.

12. Hewison M. An update on vitamin D and humanimmunity. Clin Endocrinol 2012: 76: 315–325.

13. Ramagopalan SV, Heger A, Berlanga AJ et al.A ChIP-seq defined genome-wide map of vitaminD receptor binding: Associations with disease andevolution A ChIP-seq defined genome-wide map ofvitamin D receptor binding: associations with diseaseand evolution. Genome Res 2010: 20: 1352–1360.

14. Guo S-W, Magnuson VL, Schiller JJ, Wang X,Wu Y, Ghosh S. Meta-analysis of vitamin D receptorpolymorphisms and type 1 diabetes: a HuGE review ofgenetic association studies. Am J Epidemiol 2006: 164:711–724.

15. Lopez ER, Regulla K, Pani MA, Krause M, Usadel

K-H, Badenhoop K. CYP27B1 polymorphisms variantsare associated with type 1 diabetes mellitus in Germans.J Steroid Biochem Mol Biol 2004: 89–90: 155–157.

16. Bailey R, Cooper JD, Zeitels L et al. Association ofthe Vitamin D metabolism gene CYP27B1 with type 1diabetes. Diabetes 2007: 56: 2616–2626.

17. Cooper JD, Smyth DJ, Walker NM et al. Inheritedvariation in vitamin D genes is associated withpredisposition to autoimmune disease type 1 diabetes.Diabetes 2011: 60: 1624–1631.

18. Wang TJ, Zhang F, Richards JB et al. Commongenetic determinants of vitamin D insufficiency: agenome-wide association study. Lancet 2011: 376:180–188.

420 Pediatric Diabetes 2014: 15: 416–421

T1D and SNPs in vitamin D metabolism genes

19. Thuesen B, Husemoen L, Fenger M et al. Determinantsof vitamin D status in a general population of Danishadults. Bone 2012: 50: 605–610.

20. Purcell S, Neale B, Todd-Brown K et al. PLINK: atool set for whole-genome association and population-based linkage analyses. Am J Hum Genet 2007: 81:559–575.

21. Thorsen SU, Mortensen HB, Carstensen B et al. Nodifference in vitamin d levels between children newlydiagnosed with type 1 diabetes and their healthy siblings:a 13-year nationwide Danish study. Diabetes Care 2013:36: e157–e158.

22. Turer CB, Lin H, Flores G. Prevalence of vitamin Ddeficiency among overweight and obese US children.Pediatrics 2013: 131: e152–e161.

23. Zipitis CS, Akobeng AK. Vitamin D supplementationin early childhood and risk of type 1 diabetes: asystematic review and meta-analysis. Arch Dis Child2008: 93: 512–517.

24. Sørensen IM, Joner G, Jenum PA, Eskild A,Torjesen PA, Stene LC. Maternal serum levels of25-hydroxy-vitamin D during pregnancy and risk oftype 1 diabetes in the offspring. Diabetes 2012: 61:175–178.

Pediatric Diabetes 2014: 15: 416–421 421