LADA and T1D in Estonian population — Two different genetic risk profiles

Gene 497 (2012) 285–291

Contents lists available at SciVerse ScienceDirect

Gene journal homepage: www.elsevier.com/locate/gene

Short Communication

LADA and T1D in Estonian population — Two different genetic risk profiles Kalle Kisand ⁎, Raivo Uibo Immunology Group, IGMP, University of Tartu, Estonia

a r t i c l e

i n f o

Article history: Accepted 29 January 2012 Available online 3 February 2012 Keywords: Diabetes T1D T2D LADA Genetic

a b s t r a c t Aims/hypothesis: The aim of our study was to analyze combined impact of 17 polymorphisms at 8 gene regions previously shown to be associated with autoimmunity in diabetes. We hypothesized that the genetic predisposition is multiplicative and joint risk of different diabetic phenotypes forms by distinct combination of susceptibility loci. Methods: An ethnically homogenous population of Estonian origin, including 65 LADA patients, 154 patients with T1D, 260 patients with T2D and 229 non-diabetic controls, was genotyped for polymorphisms/ haplotypes in HLA-DQB1, insulin gene (rs689, rs3842729), PHTF1–PTPN22 region (rs2476601, rs6679677), CTLA4 region (rs231806, rs16840252, rs5742909, rs231775, rs3087243, rs2033171), ICOS region (rs10932037, rs4675379), CD25 (rs706778), CD226(rs763361), NAA25 (rs17696736). Results: As expected, the risk of T1D was consistently attributed by HLA-DQB1 haplotypes, but also by haplotypes of INS and PHTF1–PTPN22 region, and rs17696736 in NAA25. By contrast, LADA was associated only with T1D-protective HLA haplotypes and with two more frequent haplotypes of the CTLA4. It is of interest, that seldom CT haplotype of PHTF1–PTPN22 region carried the risk for autoantibody-negative T2D. The final best-fitted model for T1D genetic risk contained six gene regions (HLA-DQB1, INS, PHTF1, CTLA4 + 49, CD226 and NAA25) and for LADA only two (HLA-DQB1 and CTLA4 + 49). The AUCs of these models are 0.869 and 0.693, respectively. Conclusions: Classical T1D-risk haplotypes of HLA and some non-HLA loci describe quite well the genetic risk for T1D but not for LADA. The need of further studies should be stressed to discover the real risk factors for slower forms of autoimmune diabetes in adults. © 2012 Elsevier B.V. All rights reserved.

1. Introduction Diabetes mellitus is a heterogeneous group of disorders characterized by increased plasma glucose levels. The two major types of diabetes, type 1 diabetes (T1D) and type 2 diabetes (T2D), display markedly different pathophysiological processes. Most cases of T1D are considered to be autoimmune and characterized by the presence of autoantibodies such as islet cell antibody (ICA), GAD65 antibody (GADA), IA-2 antibody (IA2A), insulin autoantibody (IAA), or the recently described autoantibody to zinc-transporter 8. In contrast, the pathogenic features of T2D include insulin resistance in peripheral tissues, with relatively little or no insulin deficiency (American Diabetes Association, 2011). Although the specific aetiology is unknown, T2D is not associated with autoimmune destruction of pancreatic beta cells. However, in several studies, islet autoimmunity Abbreviations: AUC, Area under the curve; GADA, GAD65 antibody; LADA, Latent autoimmune diabetes in adults; IA2A, IA-2 antibody; ICA, Islet cell antibody; LD, Linkage disequilibrium; ROC, Receiver operating characteristic; SNP, Single-nucleotide polymorphism; T1D, Type 1 diabetes; T2D, Type 2 diabetes; VNTR, Variable number tandem repeat. ⁎ Corresponding author at: Immunology Group, IGMP, University of Tartu, Ravila str.19, Tartu 51014, Estonia. Tel.: + 372 7 374 233; fax: + 372 7 374 232. E-mail address: [email protected] (K. Kisand). 0378-1119/$ – see front matter © 2012 Elsevier B.V. All rights reserved. doi:10.1016/j.gene.2012.01.089

has been demonstrated at the time of diagnosis in 10–25% of patients with the T2D phenotype (Tuomi et al., 1999; Vatay et al., 2002; Zimmet et al., 1994). The clinical presentation of this specific subgroup, called latent autoimmune diabetes in adults (LADA), exhibits some features of each of the two main forms of the disease, earning the moniker “type 1.5 diabetes”(Grant et al., 2010; Schernthaner et al., 2001). To date, more than 50 separate chromosome regions have been implicated in T1D (Todd et al., 2007). Specific haplotypes of DRB1, DQA1, and DQB1 in the human leukocyte antigen (HLA) complex, variants in the insulin gene (INS), the cytotoxic T lymphocyte antigen-4 gene (CTLA4), and the protein tyrosine phosphatase, non-receptor type 22 gene (PTPN22) have been repeatedly shown to be associated with T1D susceptibility (Bennett et al., 1995; Bottini et al., 2004; Ueda et al., 2003). There is substantially less known about the genetic basis of LADA. Several studies looking at the role of the HLA locus and INS polymorphisms have suggested that LADA lies somewhere at the genetic intersection of T1D and T2D (Andersen et al., 2010; Cervin et al., 2008; Haller et al., 2007; Hosszufalusi et al., 2003; Kobayashi et al., 2006; Stenstrom et al., 2002; Tuomi et al., 1999; Vatay et al., 2002; Vella et al., 2005). Our earlier investigations in which we studied selected diabetes gene polymorphisms individually led us to similar conclusions (Douroudis et al., 2008, 2009a,b,c,d;

286

K. Kisand, R. Uibo / Gene 497 (2012) 285–291

Haller et al., 2004, 2007). These studies did not take into account the continuous spectrum of diabetes phenotypes, the combinatory effect of alleles, or the exact role of extended haploblocks in diabetes risk. Here, we present an analysis of the individual and combined impact of polymorphisms of the HLA-DQ, INS, PHTF1, PTPN22, CTLA4, IL2RA, CD226 and NAA25 gene regions within a continuous spectrum of diabetes (T1D, LADA, and T2D). The selection of these gene regions was done using GWAS data (Todd et al., 2007) and our own preliminary data using ABI SNPlex genotyping 48-plex technology. We hypothesized that the genetic predisposition to diabetes is multiplicative and that the risk of a particular diabetic phenotype depends on a distinct combination of susceptibility loci. We believe that evaluation of genetic profiles will provide better predictive power for the development of diabetes. 2. Material and methods 2.1. Subjects An Estonian population including 229 non-diabetic controls (mean age 45.9 ± 14.5 y, 138 females), 154 patients with T1D (mean age at diagnosis 22.0 ± 14.3 y, 77 females), 65 LADA patients (mean age at diagnosis 54.5 ± 11.5 y, 40 females) and 260 patients with T2D (mean age at diagnosis 56.1 ± 10.1 y, 159 women) was studied. LADA patients were selected from a group with primary clinical diagnosis of T2D, who presented beta cell autoantibodies (ICA, IA2A, and/or GADA) in the serum and had no need for insulin treatment for at least 6 months post diagnosis (Haller et al., 2007). Diabetic patients were diagnosed between 1962 and 2003 in two pediatric and three adult inpatient endocrinology and diabetes units in Estonia (Haller et al., 2007; Rajasalu et al., 2007). The control group was comprised of 138 healthy blood donors and 91 hospitalized subjects who did not have diabetes as an accompanying illness. The study was approved by the Ethics Committee of the University of Tartu, and informed consent was obtained. 2.2. Genotyping Genotyping was performed as described previously (Douroudis et al., 2008, 2009a,d, 2010; Haller et al., 2007). Briefly, HLA-DQB1 typing was performed using the hybridization of lanthanide-labeled allelespecific oligonucleotide probes with a PCR amplified gene product from blood spots (DELFIA®, Wallac, Perkin-Elmer Life Sciences, Boston, MA) (Sjoroos et al., 1995). We used five sequence-specific oligonucleotide probes to identify DQB1 alleles (DQB1*0302, DQB1*02 (alternative alleles *0201, *0202 or *0203), DQB1*0602, DQB1*0603 and DQB1*0301) known to be associated with T1D. The detected alleles were classified into four groups of genotypes by susceptibility or protection against T1D: high-risk HLA-DQB1*02/ 0302; moderate-risk HLA-DQB1*0302/x (where x indicates *0302 or a non-defined allele); low-risk HLA-DQB1*0301/0302; *02/0301; *02/x; *0302/0602; *0302/0603 (where x indicates *02 or a nondefined allele); and decreased-risk HLA-DQB1*0301/x; x/x; *02/ 0602; *02/0603; *0602/x; *0603/x (where x indicates a non-defined allele) (Kukko et al., 2003). The single-nucleotide polymorphisms (SNPs) in exon 1 of the INS gene, −23 HphI A/T (rs689) and -2221MspI C/T (rs3842729) were determined by restriction fragment length polymorphism (RFLP) analysis. The genotyping of polymorphisms of PHTF1 (rs6679677), PTPN22 (rs2476601), CTLA4 region (rs231806, rs16840252, rs5742909, rs231775, rs3087243, rs2033171), ICOS region (rs10932037, rs4675379), IL2RA (rs706778), CD226 (rs763361), and NAA25 (rs17696736) was carried out using RFLP or TaqMan SNP genotyping assay (Applied Biosystems, Foster City, CA) on the ABI 7000 instrument (Applied Biosystems). A 98–100% genotyping success rate was obtained for all studied SNPs. The CTLA4 SNP rs2033171,

and ICOS region SNPs rs10932037 and rs4675379 were not genotyped in T2D patients. All investigated polymorphisms were in Hardy– Weinberg equilibrium (p > 0.05) except the ICOS rs2033171 SNP in patients with T1D (p = 0.006). 2.3. Statistical analysis The R (version 2.11.1) statistical computer program (The R Foundation for Statistical Computing, Boston, MA, http://www.r-project. org) was employed to obtain odds ratio (OR) values as well as Wald's confidence intervals (CI) for alleles, genotypes and haplotypes using logistic regression analysis. The package MASS was used for stepwise model selection by AIC for investigation of combinatory effects of genes. The association analysis with alleles and haplotypes was performed using haplo.stats (R) or the Haploview 4.2 version software (http://www.broad.mit.edu/mpg/haploview) (Barrett et al., 2005). One million permutations were carried out to estimate the significance of the results, correcting for the multiple loci tested. A p-value of ≤0.05 was considered significant for all analyses. Our sample set has the power to detect an odds ratio of at least 1.5 for T1D, 1.75 for LADA and 1.45 for T2D, assuming a significance level of 0.05, power of 0.8 and risk allele frequency >0.3 at multiplicative genetic model. 3. Results We analyzed haplotypes and polymorphic loci at 8 regions associated with diabetes. Table 1 lists the risk or protective haplotypes and genotypes for each type of diabetes. Seven of the investigated regions (HLA-DQB1, INS, PHTF1, PTPN22, CTLA4, CD226, and NAA25) were statistically significantly associated with T1D in the Estonian population. The highest risk for T1D was contributed by HLA haplotypes DQB1*02/0302 (OR = 6.69, 95% CI 2.77–16.13, p = 2.35 × 10 − 5) and DQB1*0302/x (OR = 3.48, 95% CI 1.56–7.77, p = 0.0023) (Table 1). Moreover, HLA haplotypes DQB1*0301/x; x/x; *02/0602; *02/0603; *0602/x; *0603/x contributed significant protection from T1D and the magnitude of this effect was comparable to the effect of the protective genotypes of INS (decreased-risk HLA OR = 0.21, 95% CI 0.12–0.36, p = 2.3 × 10 − 8 and TT-genotype of INS rs689 OR = 0.06, 95% CI 0.01–0.42, p = 0.005; Tables 1 and 2). It is of interest that the effect of INS was independent of HLA assessed by adjustment of logistic regression model by HLA (TT genotype of INS rs689 OR = 0.03, 95% CI 0–0.29, p = 0.0022; adjusted by HLA). None of the investigated HLA haplotypes or INS genotypes was associated with risk of autoantibody-negative T2D in this Estonian population. In this study, we found no PHTF1 or PTPN22 genotypes to be associated with LADA susceptibility, and a weak association between the PTPN22 CT genotype and T2D (OR = 1.56, 95% CI 1.04–2.35, p = 0.033; Table 2) became insignificant after adjustment of analysis by age. It is of interest that PHTF1 and PTPN22 genes were observed to carry risk for T1D. Furthermore, adjustment of the logistic regression model by HLA even increased the significance of the association of T1D with PHTF1 and PTPN22 (AA genotype of rs6679677 OR = 5.17, 95% CI 1.54–17.38, p = 0.0078; TT genotype of rs2476601 OR = 7.24, 95% CI 2.34–22.36, p = 0.0006; adjusted by HLA). In this study, we also analyzed the CTLA4-ICOS region and found that four of eight investigated polymorphisms in CTLA4-ICOS region were significantly and consistently associated with LADA (Table 2). The ancestral (major) allele A of rs231775, the new (minor) allele C of rs231806 and the new (minor) allele A of rs3087243 in CTLA4 gene conferred significant protection from LADA in the Estonian population. The association of CTLA4 with T1D was somewhat weaker as after adjustment for age only AA (and not AG) genotype of + 49A > G (OR = 0.49, 95%CI 0.25–0.95, p = 0.036; adjusted by age) and AG genotype of CT60G > A (OR = 0.58, 95%CI 0.35–0.96, p = 0.033; adjusted by age) showed statistically significant association with T1D. None of investigated polymorphisms in CTLA4 or ICOS

K. Kisand, R. Uibo / Gene 497 (2012) 285–291

287

Table 1 Haplotypes associated with different forms of diabetes. Haplotypes

Controls Forms of diabetes T1D n (%)

HLA haplotypes High risk 7 (3) Moderate risk Low risk Decreased risk

n (%)

39 (25) 10 (4) 29 (19) 66 (29) 55 (36) 146 (64) 25 (16)

LADA

T2D

OR(95%CI)

p-value/permuted pvalueb

n (%)

OR(95%CI)

p-value/permuted pvalueb

n (%)

6.69(2.77– 16.13) 3.48(1.56– 7.77) Reference

2.35 × 10− 5

5 (8)

NS

10 (4)

0.0023

7 (11) 27 (42) 25 (39)

1.75(0.51– 5.9) 1.71(0.59– 4.96) Reference

0.21(0.12– 0.36)

INS haplotypes (rs689/INS MspI)a AC 157(69) 114 Reference (74) TT 54(24) 29(19) 0.37(0.25– 0.55) TC 16 (7) 8(5) 0.67(0.23– 0.89)

2.3 × 10

−8

3.55 × 10− 5/0.0003 0.0239 /NS

PHTF1–PTPN22 haplotypes (rs6679677/rs2476601)a CC 197(86) 114 Reference (74) AT 31.5 40(26) 2.04(1.423.27 × 10− 5/0.0003 (13.8) 2.93) CT 0.5 (0.2) -

46 (71) 14 (22) 5(7)

NS -

p-value/permuted pvalueb

1.47(0.53– 4.11) 13 (5) 1.34(0.55– 3.28) 64 (25) Reference

NS NS -

0.42(0.23– 0.78)

0.0057

171 (66)

1.21(0.80– 1.82)

NS

Reference

-

Reference

-

0.65(0.27– 3.97) 0.95 (0.28– 4.7)

NS

184 (71) 60(23)

NS

NS

16(6)

0.77(0.53– 1.10) 0.92(0.50– 1.68)

-

Reference

-

1.16(0.64– 1.95) 3.06(1.43– 4.96)

NS

88(34)

Reference

-

NS

31(12)

1.05(0.78– 1.41) 0.94(0.56– 1.58) 1.07(0.69– 1.65)

NS

NS

112 (43) 18(7)

56 Reference (86) 8(13) 1(0–∞)

NS

208 (80) 41(16)

-

-

-

5(2)

Reference

-

CTLA4 haplotypes (rs231806/rs16840252/rs5742909/rs231775/rs3087243)a CCCAA 78(34) 49(32) Reference 14 (22) GCCGG 93(41) 75(49) 1.39(0.99– NS 38 1.96) (59) GTCAG 16(7) 9(6) 1.00(0.54– NS 4(5) 1.83) GTTAG 25(11) 17(11) 1.21(0.73– NS 5(8) 2.02)

OR(95%CI)

2.45(1.46– 4.09) 1.14(0.46– 2.80) 1.23(0.55– 2.76)

3 × 10

−4

/0.003

NS

9.52 × 10− 5/0.0003

NS NS

a

Rare haplotypes (b 2%) not shown; bNS: p-value > 0.05.

genes were found to be associated with risk of T2D in Estonian population. In addition to the association between T1D and the TT genotype of the CD226 rs763361 SNP, no association was revealed with LADA or T2D. Similarly, the GG and AG genotypes of the NAA25 rs17696736 SNP were associated with T1D but not LADA or T2D (Table 2). The rs706778 SNP of CD25 (IL2RA), commonly referred as confirmed genetic locus of T1D, was not associated with risk of diabetes in this Estonian population. We also analyzed the association of individual alleles of 16 different polymorphisms with each type of diabetes and give the correction for multiple testing (permutation analysis). Individual alleles in five genes: INS, PHTF1, PTPN22, CD226, and NAA25 were statistically significantly associated with T1D (Table 2). It is important to stress that all these associations, except CD226, remained significant after correction of p-values for multiple tested markers by 106 permutation. The four alleles associated with risk for LADA, on the other hand, were variations within the CTLA4 gene (Table 2). Permutation analysis rendered one of the CTLA4 polymorphisms (rs2033171) insignificant. For antibodynegative T2D, only the T allele of the PTPN22 rs2476601 SNP was found to be associated (p= 0.019), but disappeared after permutation. Haplotype analysis revealed that the investigated polymorphisms grouped into three tight haploblocks in INS gene, in the PHTF1– PTPN22 region, and in CTLA4 gene. Much lower linkage disequilibrium (LD) was detected between SNPs in ICOS region (Supplementary Fig. 1). We investigated the association of these haplotypes with each type of diabetes and found that T1D risk is significantly associated with the INS and the PHTF1–PTPN22 haploblocks, but not with CTLA4 regions (Table 1). The association of T1D with two common haplotypes of the

INS gene (AC and TT) was very strong and remained significant after correction of p-values for multiple testing (10 6 permutations). Tight LD (D′ = 1 and r2 =1) between the polymorphisms in the PHTF1–PTPN22 region makes it difficult to determine which polymorphisms are directly responsible for the observed risk of T1D in this large 73 kb haploblock, but both common haplotypes CC and AT showed highly significant association (p= 3.27 × 10− 5) with the disease and the associations remained significant after permutation analysis (p= 0.0003) (Table 1). By contrast, the risk of development of LADA was conferred solely by CTLA4 haploblock (Table 1). Statistically significant risk was carried by the most frequent haplotype of the CTLA haploblock GCCGG (OR = 2.45, 95%CI 1.46–4.09, p-value = 2.4 × 10 − 4) and this association remained significant after permutation (p = 0.0055). Two haplotypes (CC and CT) of the PHTF1-PTPN22 haploblock were associated with T2D. The association of the CT haplotype with T2D was more significant (p = 9.52 × 10 − 5) and sustained after permutation (p = 0.0005). However, because this haplotype occurred quite infrequently (2.1%), the impact of it would be quite small at the population level. To test the combinatory effect of the investigated genes on diabetes risk, we performed stepwise model selection by AIC (package MASS in R-project). The initial model for T1D contained all polymorphisms, which showed significant association with the disease in single marker association analysis (HLA haplotypes, INS, PTPN22, PHTF1, CTLA4 MH30, CTLA4 + 49, CTLA4 CT60, CTBC217, CD226 and NAA25). As a result of that analysis, the polymorphisms of PTPN22, CTLA4 MH30, CTLA4 CT60, and CTBC217 were removed from the model and the final best-fit model for T1D contained six gene regions: HLA,

288

K. Kisand, R. Uibo / Gene 497 (2012) 285–291

Table 2 SNP associated with different forms of diabetes. Polymorphism

Controls N (%)

Forms of diabetes T1D

LADA

T2D

n (%)

OR(95%CI)

p-value/permuted p-value

n (%)

OR(95%CI)

p-value/permuted p-value

n (%)

OR(95%CI)

p-value/permuted p-value

IL2RA rs706778a A 207 (45) G 251 (55)

145(47) 163(53)

Reference 0.93 (0.69–1.24)

NS

66(52) 64(48)

Reference 0.80 (0.54–1.18)

NS

204(40) 312(60)

Reference 1.26 (0.98–1.63)

NS

INS HphI A/T (rs689) AA 106 (46) AT 105(46)

106(70) 44 (29)

1.2 × 10− 4

38 (59) 23 (36)

135 (52) 109 (42)

1 (1)

0.005

3 (5)

NS

14 (6)

A T

317() 141()

256(85) 46(15)

1.16 × 10− 6/ 1.4 × 10− 5

99(77) 29(23)

NS

379(73) 137(27)

Reference 0.82 (0.56–1.18) 0.61 (0.29–1.28) Reference 0.81 (0.62–1.07)

NS

18 (8)

Reference 0.61 (0.34–1.10) 0.46 (0.13–1.67) Reference 0.66 (0.42–1.04)

NS

TT

Reference 0.42 (0.27–0.65) 0.06 (0.01–0.42) Reference 0.40 (0.28–0.59)

Reference 0.49 ( 0.31–0.77) -

0.0019

45 (69) 18 (28)

NS

158 (61) 91 (35)

2 (3)

NS

11 (4)

Reference 0.83 (0.57–1.20) 0.74 (0.32–1.74) Reference 0.86 (0.64–1.16)

NS

-

Reference 0.58 (0.31–1.06) 0.47(0.10–2.20)

INS MspI C/T (rs3842729) CC 128 (56) CT 89 (39)

115(75) 39 (25)

TT

12 (5)

0 (0)

C T

346(76) 112(24)

269(87) 39(13)

Reference 0.45 (0.30–0.6)

5.8 × 10− 5/ 6 × 10− 4

108(83) 22(17)

Reference 0.63 (0.38–1.04)

NS

407(78) 113(22)

PHTF1 rs6679677 CC 172 (75) AC 51 (22)

88 (57) 53 (35)

0.0027

48 (75) 15 (23)

180 (69) 76 (29)

13 (8)

Reference 1.76 (0.20–15.83) 1.67 (0.20–14.25) 1.21 (0.85–1.72) Reference

NS

AA

Reference 2.03 (1.28–3.23) 4.23(1.56–11.52)

NS

4 (2)

NS

84(16)

-

436(84)

NS

175 (67) 81 (31)

NS

4 (2)

NS

418(80) 102(20)

0.36 (0.12–1.09) 0.44 (0.24–0.81) Reference 0.52 (0.33–0.83) Reference

NS

33 (13)

0.0076

115 (45)

0.005/ 0.0554

108 (42) 181(35)

-

331(65)

6 (3)

A

63(14)

79(26)

C

395(86)

229(74)

0.0047

1 (2) −5

2.16 (1.50-3.13) Reference

3.3 × 10 / 4 × 10− 4 -

17(13)

Reference 2.03 (1.28–3.23) 4.23 (1.56–11.52) Reference 2.16 (1.50–3.13)

0.0027

49 (75) 15 (23)

0.0047

1 (2)

3.3 × 10− 5/ 4 × 10-4

113(87) 17(13)

0.79 (0.40–1.56) 0.66 (0.43–1.02) Reference 0.80 (0.59–1.09) Reference

NS

4 (6)

NS

21 (32)

NS

40 (62) 29(22)

-

101(78)

111(87)

PTPN22 rs2476601 CC 172 (75) CT 51(22)

88 (57) 53 (34)

TT

6 (3)

13 (9)

C T

394(86) 64(14)

229(74) 79(26)

CTLA4 MH30 rs231806 CC 26 (11)

17 (11)

CG

110 (48)

60(39)

GG C

93 (41) 161(35)

77 (50) 94(31)

G

297(65)

214(69)

CTLA4-1147 rs16840252a C 369(81) T 89(19)

253(82) 55(18)

Reference 0.90 (0.62–1.31)

NS

113(87) 17(13)

Reference 0.62 (0.36–1.09)

NS

CTLA4-318 rs5742909a C 403(88) T 55(12)

273(89) 35(11)

Reference 0.94 (0.60–1.47)

NS

120(92) 10(8)

Reference 0.61 (0.30–1.23)

CTLA4 + 49 rs231775 AA 68 (30)

40 (26)

0.044

9 (14)

AG

124 (54)

74 (48)

0.029

35 (54)

GG A

37 (16) 260(57)

40 (26) 154(50)

0.54 (0.30–0.99) 0.55 (0.32–0.94) Reference 0.76 (0.57–1.02)

NS

21(32) 52(41)

0.23 (0.10–0.56) 0.50 (0.26–0.96) Reference 0.52 (0.35–0.78)

Reference 1.03 (0.53–1.99) 0.59 (0.07–4.98) Reference 0.94 (0.53–1.68)

Reference 2.24 (0.60–8.32) 1.57 (0.44–5.66) 0.96 (0.54–1.71) Reference

Reference 1.56 (1.04–2.35) 0.66 (0.18–2.36) Reference 1.50 (1.07–2.11)

NS NS

NS NS

NS NS NS -

0.033 NS 0.019/NS

1.09 (0.61–1.96) 0.90 (0.61–1.32) Reference 1.00 (0.77–1.30) Reference

NS

421(81) 99(19)

Reference 0.97 (0.71–1.34)

NS

NS

456(88) 64(12)

Reference 1.03 (0.70–1.51)

NS

0.0012

81 (31)

NS

0.036

129 (50)

0.0012/0.0157

50 (19) 290(56)

0.88 (0.52–1.50) 0.77 (0.47–1.26) Reference 0.96 (0.75–1.24)

NS NS -

NS NS

K. Kisand, R. Uibo / Gene 497 (2012) 285–291

289

Table 2 (continued) Polymorphism

Controls N (%)

Forms of diabetes T1D

LADA

T2D

n (%)

OR(95%CI)

p-value/permuted p-value

n (%)

OR(95%CI)

p-value/permuted p-value

n (%)

OR(95%CI)

p-value/permuted p-value

198(43)

154(50)

Reference

-

76(59)

Reference

-

230(44)

Reference

-

CTLA4 CT60 rs3087243 AA 29 (13)

19 (12)

NS

4 (6)

34 (13)

61 (40)

0.013

24 (37)

0.0077

123 (47)

GG A

82 (36) 176(38)

74 (48) 100(32)

NS

37 (57) 32(25)

0.0036/0.0434

103 (40) 191(37)

G

282(62)

208(68)

-

98(75)

-

329(63)

0.93 (0.53–1.66) 0.83 (0.56–1.22) Reference 0.93 (0.72–1.21) Reference

NS

118 (51)

0.31 (0.10–0.93) 0.45 (0.25–0.81) Reference 0.52 (0.34–0.81) Reference

0.037

AG

0.73 (0.38–1.40) 0.57 (0.37–0.89) Reference 0.77 (0.57–1.04) Reference

Reference 0.48 (0.30–0.76) 0.70 (0.39–1.26) Reference 1.30 (0.97–1.75)

0.0018

6 (9) 30 (46)

0.034

-

-

-

NS

29 (45)

Reference 0.53 (0.29–0.95) 0.31(0.12–0.81)

0.017

-

-

-

NS

42(32) 88(68)

Reference 1.74 (1.16–2.63)

0.0076/NS

-

-

-

G

CTLA4 CTBC217_1 rs2033171 TT 42 (18) 30 (19) CT 124 (54) 60 (39) CC

63 (28)

64 (42)

C T

208 250

120 188

NS NS -

ICOS rs10932037a C 401(88) T 57(12)

276(90) 32(10)

Reference 0.82 (0.52–1.29)

NS

115(88) 15(12)

Reference 0.92 (0.50–1.68)

NS

-

-

-

ICOS rs4675379a C 30(7) G 428(93)

29(9) 279(91)

Reference 0.67 (0.40–1.15)

NS

13(10) 117(90)

Reference 0.63 (0.32–1.25)

NS

-

-

-

CD226 rs763361 CC 76 (33) CT 118 (52)

36 (23) 79 (51)

NS

24 (38) 27 (43)

80 (31) 132 (51)

39 (25)

0.0056

12 (19)

NS

46 (18)

C T

268 190

151 157

0.0068/NS

75(60) 51(40)

NS

292(57) 224(43)

Reference 1.06 (0.71–1.59) 1.25 (0.73–2.14) Reference 1.10 (0.85–1.42)

NS

35 (15)

Reference 0.72 (0.39–1.35) 1.09 (0.49–2.42) Reference 0.98 (0.65–1.46)

NS

TT

Reference 1.41 (0.87–2.30) 2.35 (1.29–4.31) Reference 1.49 (1.12–2.00)

Reference 1.76 (1.10–2.84) 2.39 (1.29–4.43) Reference 1.54 (1.15–2.07)

0.019

30 (48) 21 (34)

NS

91 (36) 109 (43)

11 (18)

NS

54 (21)

0.0036/ 0.0409

81(65) 43(35)

NS

291(57) 217(43)

Reference 0.97 (0.65–1.44) 1.63 (0.96–2.76) Reference 1.23 (0.95–1.59)

NS

0.0057

Reference 0.57 (0.30–1.06) 1.01 (0.45–2.25) Reference 0.87 (0.58–1.32)

NAA25(C12orf30) rs17696736 AA 88 (38) 38 (25) AG 109 (48) 83 (54) GG

32 (14)

33(21)

A G

284(62) 174(38)

159(52) 149(48)

NS NS

NS NS

NS: p-value > 0.05; aGenotypes not shown if p-values are NS.

INS, PHTF1, CTLA4 +49, CD226, and NAA25. From the initial model for LADA containing HLA, CTLA4 MH30, CTLA4 +49, CTLA4 CT60, and CTBC217, the stepwise model selection analysis excluded CTLA4 MH30, CTLA4 CT60 and CTBC217, so the final model for LADA contained only two polymorphisms: HLA haplotype and CTLA4 + 49. Another way to assess the predictive utility of a combination of genetic risk markers is the ROC curve (Janssens et al., 2004). The ROC curves of different combinations of genes for T1D and LADA are presented in Fig. 1. In general, including more susceptibility loci into the model increased the predictive value for the disease as assessed by an increasing an area under the curve (AUC). The ROC curve for T1D using HLA alone gave an AUC of 0.798. The ROC for LADA, including only HLA, conferred remarkably lower predictive value (AUC = 0.644). Adding other non-HLA loci of best-fit model into the analysis increased the AUC to 0.868 for T1D. For LADA, however, combining the CTLA4 +49 locus with HLA haplotype improved the AUC

only slightly (0.693). Interestingly, the model for T1D containing only non-HLA loci also gave a high predictive value (AUC = 0.732) (Fig. 1B). 4. Discussion We studied haplotypes and polymorphic loci at 8 regions associated with different forms of diabetes. We found that the genetic markers associated with T1D differed from those associated with LADA. Indeed, T1D was consistently associated with HLA-DQB1 haplotypes and SNPs of four genes (INS, PHTF1, CD226 or NAA25). Unexpectedly, we found no statistically significant association between the classical risk haplotypes of HLA and LADA. In LADA, statistically significant association was selectively detected with the protective haplotypes of HLA-DQB1. This finding is in accordance with earlier studies suggesting that the increased frequency of protective HLA genotypes is a specific feature of

290

K. Kisand, R. Uibo / Gene 497 (2012) 285–291

Fig. 1. (A) ROC curve of different models for T1D. The area under the curve was 0.868 for best-fit model (HLA+ INS VNTR+ PHTF1+ CTLA4(+49) +CD226+ NAA25) (black), 0.798 for HLA only (green), and 0.732 for non-HLA genes only (INS VNTR+PHTF1+CTLA4+CD226+ NAA25) (red); (B) ROC curve of different models for LADA. The area under the curve was 0.693 for best-fit model (HLA+ CTLA4(+49)) (black), 0.644 for HLA only (green), 0.602 for CTLA4 only (red).

patients with slow-progressing T1D (Graham et al., 1999; Tuomi et al., 1999). However, the results of current study contradict with opinion that the patterns of susceptibility at the HLA-DQB1 loci in LADA are similar to those reported for T1D and insisting the hypothesis that autoimmune diabetes occurring in adults is an age-related extension of the pathophysiological process presenting as childhood-onset T1D (Desai et al., 2007; Hosszufalusi et al., 2003; Vatay et al., 2002). A possible reason for such a discrepancy could be due to the inconsistency of the diagnostic criteria for LADA (Kobayashi et al., 2006). Another explanation could be that distinct interactions of genes (epistasis) and environment may affect phenotypes differently in different populations. Of note, not only HLA, but other proven non-HLA T1D susceptibility loci (INS, PHTF1, PTPN22 and NAA25) showed no significant association with LADA in this Estonian population. This suggests that genetic risk of autoimmunity in LADA is determined by some other not yet revealed genetic loci. Indeed, in the Estonian population, the only gene associated with risk of LADA other than HLA was the CTLA4. We observed the sustained association between LADA and CTLA4 on single SNP as well as haplotype level, even after correction for multiple markers by permutation. Our results are in concordance with the data published by Caputo et al. (Caputo et al., 2005, 2007), who identified CTLA4 SNP +49A/G but not −318C/T as a susceptibility marker for LADA. Moreover, the frequency of the heterozygous A/G genotype in LADA patients was significantly increased compared to T1D (Caputo et al., 2005). Until recently, few studies have systematically investigated the complex interaction of multiple diabetes-associated genetic polymorphisms (Bjornvold et al., 2008). The first data supporting the existence of gene–gene interactions between HLA-DR and INS genes were published by Julier et al. (Julier et al., 1991). These authors noticed that in multiplex families, the diabetes-associated alleles of INS VNTR were transmitted preferentially to HLA-DR4-positive diabetic offspring from heterozygous parents. Future investigators, including us, stated that the association is not restricted to HLA-DR4-positive individuals and that the INS genotype confers T1D susceptibility independently of HLA-DR (Bain et al., 1992; Haller et al., 2007; Metcalfe et al., 1995; van der Auwera et al., 1993). The present study supports the idea of independent association of these two gene regions with diabetes. We could detect no association between HLA haplotype and INS VNTR in logistic regression analysis in cases or controls. Moreover, both gene regions were included into the best fit model for risk

assessment of T1D as significant additive risk markers. Therefore, our data do not support the results of previous studies which showed that the relative risk conferred by INS gene polymorphism is more prominent in the low-risk HLA (Metcalfe et al., 1995; Motzo et al., 2004) or in the high-risk HLA-DR diabetic patients (Julier et al., 1991). We also studied the effects of interactions between HLA and other genes on diabetes susceptibility. Using multiple logistic regression with stepwise model selection, we found two very different best-fit models for T1D and LADA. From our data, the best model describing genetic susceptibility for T1D should contain information about polymorphisms in HLA-DQB1 haplotype, INS (VNTR), PHTF1 (or PTPN22), CTLA4, CD226, and NAA25 genes. The best model for LADA should include HLA and CTLA4. The AUC of these two models are quite different (0.869 for T1D and only 0.693 for LADA). In general, our results of ROC analysis for T1D are similar to previously published data (Bjornvold et al., 2008). It is true that HLA alone could account for most of the prediction of T1D. However, even if the individual effect of each non-HLA locus is small, then the total improvement of the predictive value could be quite remarkable in our cohorts. In the case of LADA, it is possible that many susceptibility loci have not yet been discovered, but when included into future models will allow better predictions of LADA. Alternatively, the genetic risk for LADA may be lower compared to T1D, and environmental risk factors may contribute most to the development of this slower form of autoimmune diabetes. Interestingly, the proportion of high-risk HLA genotypes has decreased among patients diagnosed with T1D during recent decades compared with those diagnosed a few decades earlier (Gillespie et al., 2004; Hermann et al., 2003). This change is associated with a greatly increased disease incidence during the past decades and indicates that greater environmental pressure affects the penetrance of T1D in genotypes with lower HLA risk (Ilonen and Hermann, 2010). One limitation of this study is the relatively small sample size, nevertheless our given sample set has enough power to detect an odds ratios of at least 1.45–1.75. Ascertainment of individuals with LADA is difficult since this diagnosis has not yet been added to the WHO International Classification of Diseases (ICD-10, http://www. who.int/entity/classifications/icd/en) or into the criteria established by the Expert Committee on the Diagnosis and Classification of Diabetes Mellitus (American diabetes association, 2011). However, in increasing numbers of studies, LADA patients have been distinguished from the cohort of T2D according to the presence of diabetesspecific autoantibodies (Cervin et al., 2008; Zimmet et al., 1994).

K. Kisand, R. Uibo / Gene 497 (2012) 285–291

We also want to stress the feasibility to evaluate continuous spectrum of diabetes (T1D-LADA-T2D) in the same population with similar ethnic, genetic and environmental risk background. We believe that our results of specifying more precisely the genetic profiles of the subtypes of diabetes will provide further clues for genetic (functional) studies and encourage scientific community searching other (nongenetic) risk factors for LADA. In conclusion, this study has shown that classical genetic risk markers of T1D (HLA high risk haplotypes and polymorphisms of INS and PTPN22 gene) are poor predictors of LADA and that further studies are necessary to discover additional risk factors (including \non-genetic) for this form of autoimmune diabetes. Our results provide some clues for functional studies and should encourage scientific community searching molecular mechanisms for LADA. We also want to stress that future studies considering diabetes as a spectrum disorder and much more complex than simply T1D and T2D, should compare simultaneously all forms of diabetes within the same population (influenced by similar genetic and environmental risk factors) and not concentrate solely on the most clear and distinctive diagnostic groups. Supplementary materials related to this article can be found online at doi:10.1016/j.gene.2012.01.089. Conflict of interest statement The authors declare no conflict of interest associated with this manuscript. Acknowledgments We thank M. Möls for advice in statistical analysis, Dr. V. Nemvalts and Dr. T. Rajasalu for collecting patients' material, K. HallerKikkatalo, E. Prans, and K. Douroudis for genotyping and for help in preparing patient data, prof. J. Ilonen for advice in preparing manuscript. This study was supported by the Estonian Science Foundation grant no. 7749, partly by the Estonian Ministry of Education and Research (Targeted project no SF0180035s08) and by the EU Regional Developmental Fund. Both authors participated in the conception, design and coordinating the study, in the analysis and interpretation of data, as well as drafting and approval the article. References American diabetes association, 2011. Diagnosis and classification of diabetes mellitus. Diabetes Care 34 (Suppl 1), S62–S69. Andersen, M.K., et al., 2010. Latent autoimmune diabetes in adults differs genetically from classical type 1 diabetes diagnosed after the age of 35 years. Diabetes Care 33 (9), 2062–2064. Bain, S.C., et al., 1992. Insulin gene region-encoded susceptibility to type 1 diabetes is not restricted to HLA-DR4-positive individuals. Nat. Genet. 2 (3), 212–215. Barrett, J.C., Fry, B., Maller, J., Daly, M.J., 2005. Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics 21 (2), 263–265. Bennett, S.T., et al., 1995. Susceptibility to human type 1 diabetes at IDDM2 is determined by tandem repeat variation at the insulin gene minisatellite locus. Nat. Genet. 9 (3), 284–292. Bjornvold, M., et al., 2008. Joint effects of HLA, INS, PTPN22 and CTLA4 genes on the risk of type 1 diabetes. Diabetologia 51 (4), 589–596. Bottini, N., et al., 2004. A functional variant of lymphoid tyrosine phosphatase is associated with type I diabetes. Nat. Genet. 36 (4), 337–338. Caputo, M., et al., 2005. Cytotoxic T lymphocyte antigen 4 heterozygous codon 49 A/G dimorphism is associated to latent autoimmune diabetes in adults (LADA). Autoimmunity 38 (4), 277–281. Caputo, M., Cerrone, G.E., Mazza, C., Cedola, N., Targovnik, H.M., Gustavo, D.F., 2007. No evidence of association of CTLA-4–318 C/T, 159 C/T, 3′ STR and SUMO4 163 AG polymorphism with autoimmune diabetes. Immunol. Invest. 36 (3), 259–270. Cervin, C., et al., 2008. Genetic similarities between latent autoimmune diabetes in adults, type 1 diabetes, and type 2 diabetes. Diabetes 57 (5), 1433–1437. Desai, M., et al., 2007. An association analysis of the HLA gene region in latent autoimmune diabetes in adults. Diabetologia 50 (1), 68–73.

291

Douroudis, K., et al., 2008. Protein tyrosine phosphatase non-receptor type 22 gene variants at position 1858 are associated with type 1 and type 2 diabetes in Estonian population. Tissue Antigens 72 (5), 425–430. Douroudis, K., et al., 2009a. Association of CTLA4 but not ICOS polymorphisms with type 1 diabetes in two populations with different disease rates. Hum. Immunol. 70 (7), 536–539. Douroudis, K., Nemvalts, V., Rajasalu, T., Kisand, K., Uibo, R., 2009b. The CD226 gene in susceptibility of type 1 diabetes. Tissue Antigens 74 (5), 417–419. Douroudis, K., Prans, E., Kisand, K., Nemvalts, V., Uibo, R., 2009c. Cytotoxic Tlymphocyte antigen 4 gene polymorphisms are associated with latent autoimmune diabetes in adults. Clin. Chim. Acta 403 (1–2), 226–228. Douroudis, K., Prans, E., Uibo, R., 2009d. CTLA-4 promoter polymorphisms are associated with latent autoimmune diabetes in adults. Hum. Immunol. 70 (11), 921–924. Douroudis, K., Kisand, K., Nemvalts, V., Rajasalu, T., Uibo, R., 2010. Allelic variants in the PHTF1-PTPN22, C12orf30 and CD226 regions as candidate susceptibility factors for the type 1 diabetes in the Estonian population. BMC Med. Genet. 11, 11. Gillespie, K.M., et al., 2004. The rising incidence of childhood type 1 diabetes and reduced contribution of high-risk HLA haplotypes. Lancet 364 (9446), 1699–1700. Graham, J., et al., 1999. Negative association between type 1 diabetes and HLA DQB1*0602-DQA1*0102 is attenuated with age at onset. Swedish Childhood Diabetes Study Group. Eur. J. Immunogenet. 26 (2–3), 117–127. Grant, S.F., Hakonarson, H., Schwartz, S., 2010. Can the genetics of type 1 and type 2 diabetes shed light on the genetics of latent autoimmune diabetes in adults? Endocr. Rev. 31 (2), 183–193. Haller, K., Kisand, K., Nemvalts, V., Laine, A.P., Ilonen, J., Uibo, R., 2004. Type 1 diabetes is insulin −2221 MspI and CTLA-4 + 49 A/G polymorphism dependent. Eur. J. Clin. Invest. 34 (8), 543–548. Haller, K., et al., 2007. Insulin gene VNTR, CTLA-4 + 49A/G and HLA-DQB1 alleles distinguish latent autoimmune diabetes in adults from type 1 diabetes and from type 2 diabetes group. Tissue Antigens 69 (2), 121–127. Hermann, R., et al., 2003. Temporal changes in the frequencies of HLA genotypes in patients with type 1 diabetes — indication of an increased environmental pressure? Diabetologia 46 (3), 420–425. Hosszufalusi, N., et al., 2003. Similar genetic features and different islet cell autoantibody pattern of latent autoimmune diabetes in adults (LADA) compared with adult-onset type 1 diabetes with rapid progression. Diabetes Care 26 (2), 452–457. Ilonen, J., Hermann, R., 2010. Novel gene associations in type 1 diabetes. Curr. Diab. Rep. 10 (5), 338–344. Janssens, A.C., Pardo, M.C., Steyerberg, E.W., van Duijn, C.M., 2004. Revisiting the clinical validity of multiplex genetic testing in complex diseases. Am. J. Hum. Genet. 74 (3), 585–588 author reply 588–9. Julier, C., et al., 1991. Insulin-IGF2 region on chromosome 11p encodes a gene implicated in HLA-DR4-dependent diabetes susceptibility. Nature 354 (6349), 155–159. Kobayashi, T., et al., 2006. Immunopathological and genetic features in slowly progressive insulin-dependent diabetes mellitus and latent autoimmune diabetes in adults. Ann. N. Y. Acad. Sci. 1079, 60–66. Kukko, M., et al., 2003. Signs of beta-cell autoimmunity and HLA-defined diabetes susceptibility in the Finnish population: the sib cohort from the Type 1 Diabetes Prediction and Prevention Study. Diabetologia 46 (1), 65–70. Metcalfe, K.A., Hitman, G.A., Fennessy, M.J., McCarthy, M.I., Tuomilehto, J., TuomilehtoWolf, E., 1995. In Finland insulin gene region encoded susceptibility to IDDM exerts maximum effect when there is low HLA-DR associated risk. DiMe (Childhood Diabetes in Finland) Study Group. Diabetologia 38 (10), 1223–1229. Motzo, C., et al., 2004. Heterogeneity in the magnitude of the insulin gene effect on HLA risk in type 1 diabetes. Diabetes 53 (12), 3286–3291. Rajasalu, T., et al., 2007. Insulin VNTR I/III genotype is associated with autoantibodies against glutamic acid decarboxylase in newly diagnosed type 1 diabetes. Diabetes Metab. Res. Rev. 23 (7), 567–571. Schernthaner, G., Hink, S., Kopp, H.P., Muzyka, B., Streit, G., Kroiss, A., 2001. Progress in the characterization of slowly progressive autoimmune diabetes in adult patients (LADA or type 1.5 diabetes). Exp. Clin. Endocrinol. Diabetes 109 (Suppl 2), S94–S108. Sjoroos, M., Iitia, A., Ilonen, J., Reijonen, H., Lovgren, T., 1995. Triple-label hybridization assay for type-1 diabetes-related HLA alleles. Biotechniques 18 (5), 870–877. Stenstrom, G., Berger, B., Borg, H., Fernlund, P., Dorman, J.S., Sundkvist, G., 2002. HLADQ genotypes in classic type 1 diabetes and in latent autoimmune diabetes of the adult. Am. J. Epidemiol. 156 (9), 787–796. Todd, J.A., et al., 2007. Robust associations of four new chromosome regions from genome-wide analyses of type 1 diabetes. Nat. Genet. 39 (7), 857–864. Tuomi, T., et al., 1999. Clinical and genetic characteristics of type 2 diabetes with and without GAD antibodies. Diabetes 48 (1), 150–157. Ueda, H., et al., 2003. Association of the T-cell regulatory gene CTLA4 with susceptibility to autoimmune disease. Nature 423 (6939), 506–511. van der Auwera, B.J., Heimberg, H., Schrevens, A.F., van Waeyenberge, C., Flament, J., Schuit, F.C., 1993. 5′ insulin gene polymorphism confers risk to IDDM independently of HLA class II susceptibility. Diabetes 42 (6), 851–854. Vatay, A., et al., 2002. Differences in the genetic background of latent autoimmune diabetes in adults (LADA) and type 1 diabetes mellitus. Immunol. Lett. 84 (2), 109–115. Vella, A., et al., 2005. Localization of a type 1 diabetes locus in the IL2RA/CD25 region by use of tag single-nucleotide polymorphisms. Am. J. Hum. Genet. 76 (5), 773–779. Zimmet, P.Z., et al., 1994. Latent autoimmune diabetes mellitus in adults (LADA): the role of antibodies to glutamic acid decarboxylase in diagnosis and prediction of insulin dependency. Diabet. Med. 11 (3), 299–303.