Introduction

Type 1 diabetes mellitus and autoimmune thyroid diseases (AITD) are common autoimmune endocrine disorders [1]. Type 1 diabetes develops through a complex interaction of genetic and environmental factors and is characterised by immune-mediated destruction of the pancreatic islet beta cells [2]. Individuals with type 1 diabetes often develop other autoimmune diseases, of which AITD is the most frequent [1]. As childhood clinical hypothyroidism has an impact on growth and intellectual development as well as on glucose control in type 1 diabetes patients, screening for AITD may be important at the clinical diagnosis of type 1 diabetes and during follow-up.

The pathogenesis of AITD involves both humoral and cellular immune responses against the thyroid autoantigens thyroid peroxidase (TPO) and thyroglobulin (TG). These autoantibodies are produced in parallel to infiltration of T and B lymphocytes in the thyroid gland [3].

Autoantibodies to glutamic acid decarboxylase 65 (GADA), insulinoma-associated protein 2 (IA-2A) and insulin (IAA) are predictive and diagnostic markers of type 1 diabetes [4, 5]. Recently, autoantibodies to the zinc transporter 8 (ZnT8A) have been shown to contribute to the diagnostic sensitivity of diabetes [6–8]. Interestingly enough, ZnT8 is not only expressed in beta cells [9] but also in the cubical cells lining the thyroid follicle [10]. SLC30A8 encodes three amino acid variants of ZnT8A to amino acids at position 325 (arginine [ZnT8RA], tryptophan [ZnT8QA] or glutamine [ZnT8QA]), having possible consequences for the association with HLA-DQ genotype [8].

Both type 1 diabetes and AITD have a complex genetic background. The major genetic susceptibility of type 1 diabetes has been mapped to the HLA region on the short arm of chromosome 6, particularly the class II DQ and DR genes. DQA1*03:01-B1*03:02 (DQ8), DQA1*05:01-B1*02:01 (DQ2) or both haplotypes, confer the highest risk for type 1 diabetes [11]. It is of interest that GADA was associated with DQ2 and IAA and IA-2A with DQ8, while DQ2 was associated with a decreased risk for IA-2A [12]. Autoantibodies against TPO (TPOAb) and TG (TGAb) were also associated with HLA-DQ genotype [13, 14], although their relationships with specific HLA-DQ haplotypes were less obvious and would need further clarification in larger studies.

The paucity of population-based investigations of TPOAb and TGAb in incident children and adolescents newly diagnosed with type 1 diabetes would require large investigations such as the Better Diabetes Diagnosis (BDD) study in Sweden [15]. Since it is important to find AITD in children and adolescents, investigations of genetic and immunological associations may be needed to improve the diagnostic sensitivity and specificity for thyroid autoimmunity and clinical thyroid disease in patients with type 1 diabetes.

The aim of this study was to investigate Swedish patients with type 1 diabetes consecutively diagnosed at up to 18 years of age between May 2005 and October 2009 for: (1) the prevalence of TPOAb and TGAb; (2) the association between TPOAb or TGAb and islet autoantibodies; (3) the association between TPOAb or TGAb and HLA-DQ genotype; and (4) the relationship to thyroid-stimulating hormone (TSH) and free thyroxine (T4). We hypothesised that: (1) TPOAb and TGAb in children at the clinical onset of type 1 diabetes would be associated with autoantibodies to ZnT8 because they are also expressed in the thyroid; and (2) the large number of individuals studied would allow us to detect associations between HLA-DQ genotype and thyroid autoantibodies.

Methods

Study population

In Sweden, with 9.4 million inhabitants, all children and adolescents with type 1 diabetes are diagnosed and treated in 43 paediatric clinics. BDD is a prospective study including almost all children under the age of 18 with newly diagnosed diabetes in Sweden [15]. Patients with type 2 diabetes, MODY and neonatal and secondary diabetes were excluded. Blood samples were collected between May 2005 and October 2009 at onset of type 1 diabetes (n = 2,670) and analysed for HLA-DQ genotype and GADA, ZnT8RWQA, IAA and IA-2A as well as TPOAb, TGAb, TSH and free T4. After exclusion of 237 children with missing data, 2,433 remained (Fig. 1); 56% boys (n = 1,390) and 44% girls (n = 1,073), with a median age of 10.3 years (range 0.7–17.9).

Fig. 1
figure 1

Flow chart of the distribution of thyroid autoantibodies in the 2,670 patients newly diagnosed with type 1 diabetes

Full size image

Autoantibodies to TPO and TG

The analysis of TPOAb (kit No. L2KTO2) and TGAb (kit No. L2KTG2) were performed on serum samples on the Immulite 2000 analyzer according to the manufacturer’s instructions (Siemens Healthcare Diagnostics, Deerfield, IL, USA).

Autoantibodies to zinc T8 transporter variants

The radiobinding assay for all three variants, ZnT8R, ZnT8W and ZnT8Q, were performed as previously described [8]. The results were expressed in arbitrary units derived from in-house positive and negative standard samples. Intra-assay CV for the ZnT8RA was 6%, ZnT8WA 5% and ZnT8QA 4%. Inter-assay CV for ZnT8RA was 7%, ZnT8WA 8% and ZnT8QA 10%. Our laboratory has been validated in Diabetes Autoantibody Standardization Program (DASP) 2011 with 50% study sensitivity and 100% study specificity for ZnT8RA, 46% study sensitivity and 100% study specificity for ZnT8WA and 38% study sensitivity and 100% study specificity for ZnT8QA [16]. Cut-off values were ZnT8RA ≥75, ZnT8WA ≥75 and ZnT8QA ≥100 U/ml.

Autoantibodies to GAD65 and IA-2

The radiobinding assays for GADA and IA-2A were carried out as detailed elsewhere [15]. GADA and IA-2A levels were expressed as U/ml derived from the WHO standard 97/550 [17]. Samples were considered positive if GADA levels were above 50 U/ml and IA-2A levels above 10 U/ml. The intra-assay CV for duplicates in the GAD65A assay was 7% and in the IA-2A, 11%. In the DASP 2009, GAD65A showed workshop sensitivity of 68% and specificity of 99% and IA-2A sensitivity of 60% and specificity of 99% [18].

Autoantibodies to insulin

The radiobinding assay for IAA was carried out as described [19]. The results were expressed in arbitrary units derived from in-house positive and negative standard samples. Samples were considered positive if IAA levels were above 1 RU. Intra-assay CV was 6% and inter-assay CV 13%. Our laboratory has been validated in DASP 2010 with 26% workshop sensitivity and 100% workshop specificity.

HLA genoty**

HLA-DQB1 and -DQA1 alleles were typed by sequence-specific oligonucleotide probes on a dried blood spot used directly for PCR amplification of DQA1 and DQB1 alleles using a DELFIA Hybridization assay (Perkin Elmer, Boston, MA, USA) as described [20]. The first set of probes defines the presence of HLA-DQB1*02, 03:02, 03:01, 06:02, 06:03 and 06:04 (DQ6.4). The second set of probes defines the presence of additional DQB1 alleles. HLA-DQA1 probes define the DQA1*02:01, 03:01 and 05:01 alleles. We compared patient HLA frequencies with 2,000 newborn children in the Diabetes Prediction in Skåne (DiPiS) study [21, 22].

Serum TSH and free T4

The analysis of TSH and free T4 (fT4) was performed on serum samples using the Immulite 2000 analyser according to the manufacturer’s instructions (Siemens Healthcare Diagnostics). As mild transient TSH derangements are common at clinical onset of type 1 diabetes, the following values were used for hypothyroid disease: TSH >8 mU/l and fT4 <12 pmol/l and for hyperthyroid disease: TSH <0.01 mU/l and fT4 >22 pmol/l.

Statistical methods

Statistical analyses were performed using SPSS statistical software (version 18.0; SPSS, Chicago, IL, USA). The Mann–Whitney test was used to compare differences between two independent groups. Differences in proportions between groups were tested using the χ 2 test or Fisher’s exact test when appropriate and ORs with 95% CI were calculated. All p values for the HLA-DQ gene associations were corrected for multiple comparisons (p c). p < 0.05 was considered significant.

Results

TPO and TG autoantibodies in relation to age and sex

At the time of diagnosis, a total of 12.3% (299/2,433) of the patients were positive for either TPOAb (10.1%; 245/2,433), TGAb (8%; 195/2,433) or both (5.8%; 141/2,433). There were 44% girls (1,073/2,433) among the 2,433 patients but among the TPOAb-positive patients, 60.8% were girls (p < 0.001). Among the patients with TGAb, 57.9% (113/195) were girls (p < 0.001). Among patients positive for both TPOAb and TGAb, 62% (87/141) were girls (p < 0.001). The girl-to-boy ratio was 1.65:1 in TPOAb-positive patients, 1.72:1 in TGAb-positive patients and 1.68:1 in patients positive for TPOAb, TGAb or both.

Children negative for TPOAb and TGAb had a median age of 9.9 years (range 0.7–17.9). Compared with the thyroid-autoantibody-negative children, those positive for both TPOAb (median age 12.2 years; range 1.4–17.9; p < 0.0005) and TGAb (median age 12.1 years; range 2.2–17.9; p < 0.0005) were older. The children positive for both TPOAb and TGAb were also older (median age 12.4 years; range 2.5–17.9; p < 0.0005).

Association between TPO or TG autoantibodies and islet autoantibodies

TPOAb were found to be associated with GADA (p < 0.0001) (Table 1). In addition, TPOAb were associated with ZnT8A as one group (p = 0.039) (Table 1). When analysing the three individual ZnT8RAs it was found that this association was partly explained by a significant association between TPOAb and ZnT8RA (p = 0.036) (Table 1), while there was no significant association with either ZnT8WA (p = 0.48) or ZnT8QA (p = 0.38). In contrast, TPOAb were negatively associated with IAA (p = 0.030) (Table 1).

Table 1 Association between islet autoantibodies and autoantibodies against either TPO or TG
Full size table

TGAb were also positively associated with GADA (p < 0.0001) (Table 1) as well as with ZnT8A as one group (p = 0.015) but no association was found between TGAb and any of the individual variants of ZnT8A or IAA. Neither TGAb nor TPOAb were associated with IA-2A.

TPOAb and TGAb in association with HLA

TGAb but not TPOAb showed a possible association with genotypes containing DQA1*05:01-B1*02:01 (p = 0.014; p c = 0.098) (Table 2). In contrast, a negative association was found between TPOAb and genotypes containing DQA1*01:01-B1*05:01 (p = 0.004; p c = 0.012). The DQ8/5.1 genotype also showed a weak negative association with TPOAb (p = 0.019; p c = 0.057) (Table 2). TGAb also tended to be negatively associated with genotypes containing the DQ5.1 haplotype (p = 0.01; p c = 0.07), in addition to the DQ8/5.1 genotype (p = 0.0095; p c = 0.067) (Table 2).

Table 2 Association between HLA-DQ genotypea and either TPOAb or TGAb
Full size table

TPOAb, TGAb or both, as an indication of thyroid autoimmunity, showed a negative association with genotypes containing DQA1*01:01-B1*05:01 (p = 0.002; p c = 0.014) and a possible weak association with the genotype DQ8/5.1 (p = 0.0085; p c = 0.060).

Association of TSH and fT4 with thyroid autoantibodies

Patients with TPOAb, TGAb or both, were more likely to have clinical thyroid disease at diagnosis of type 1 diabetes compared with thyroid-autoantibody-negative patients (p < 0.001) (Table 3). The data in Table 3 suggest that of the patients studied, there may be six with hypothyroid disease and five with hyperthyroid disease (0.45% of the entire study group).

Table 3 Association between thyroid autoantibodies and possible clinical thyroid disease based on analyses of TSH and free T4 at onset of type 1 diabetes
Full size table

Discussion

In this study of 2,433 children and adolescents with type 1 diabetes, we found a prevalence of 12.3% TPOAb, TGAb or both, already at the time of diagnosis of the diabetes. While the association between TPOAb or TGAb and GADA has been described previously [1, 23], it was a novel finding that both types of thyroid autoantibodies were positively associated with ZnT8A. The negative association between TPOAb and IAA was also a novel observation. Our large cohort made it possible to confirm the negative association between HLA-DQ5.1 and either TPOAb, TGAb or both, as previously reported in 285 children with type 1 diabetes [13]. As TPOAb and TGAb, considered together, were weakly negatively associated with the DQ8/5.1 genotype, it cannot be excluded that DQ8 also contributes to protection from thyroid autoimmunity. Our study therefore indicates that primarily HLA-DQ5.1, but perhaps also DQ8, protects children with type 1 diabetes against develo** AITD. These results and the association between AITD and DQ2, as earlier described [24, 25], draw attention to the probable importance of HLA-DQ alleles for the risk of AITD in type 1 diabetes. However, the associations between thyroid autoimmunity and HLA-DQ were weak and did not fully explain the co-occurrence of islet and thyroid autoimmunity.

In this large population-based study of almost all children up to 18 years of age diagnosed with type 1 diabetes in Sweden between May 2005 and October 2009 it was possible to perform a detailed analysis of the associations between thyroid disease markers, islet autoantibodies and HLA genotypes. Since all children were analysed for all three variants of ZnT8A at diagnosis, we were able to reveal associations between the three variants of ZnT8A and thyroid autoantibodies in a way that has previously not been described.

A possible weakness to our analysis is that the results were only based upon the sampling at onset of type 1 diabetes. First, it cannot be ruled out that a few children had already been diagnosed and treated for AITD. The frequency of AITD in newly diagnosed type 1 diabetes children therefore remains to be determined. Second, it has not been possible to determine conversion to positive thyroid autoantibodies during follow-up after diagnosis. Hence, the frequency of AITD in patients with type 1 diabetes was likely to be underestimated.

We conclude that thyroid autoantibodies are common in children diagnosed with type 1 diabetes. Our prevalence of thyroid autoantibodies at clinical onset of type 1 diabetes was similar to findings in some previous studies [26, 27], but in other studies a prevalence as high as 15.8–29% has been reported. However, in these studies, the patients were tested at follow-up and not at the time of diagnosis [1, 13, 14, 28–30], which may explain the higher prevalence. Furthermore, only few children are diagnosed with AITD at the onset of type 1 diabetes [23, 26, 31], while the risk of AITD increases during follow-up of patients with type 1 diabetes. It has been estimated that 10–33% of adult patients with long-term type 1 diabetes have been diagnosed with clinical AITD [29, 32–34] and in patients followed from diagnosis TPOAb predict development of future thyroid disease [35]. It is important to diagnose childhood clinical hypothyroidism due to AITD as it has an impact on growth, intellectual development and metabolic control. Since the prevalence of AITD is higher in individuals with type 1 diabetes, screening may be important both at onset and during follow-up.

As in previous reports, girls and children at older age are more likely to have thyroid autoantibodies [1, 13, 26, 27, 29, 33]. This correlates with reports involving adults wherein women with type 1 diabetes are more prone to autoimmune thyroiditis than men [34].

We could confirm that GADA positivity was associated with both TPOAb and TGAb in children with type 1 diabetes [1, 23]. The negative association between TPOAb and IAA can be explained by the young age of those children positive for IAA [6]. Our study is the first to reveal an association between thyroid autoantibodies and ZnT8RWQA as one group in children at diabetes onset. This association can partly be explained by a significant association between TPOAb and ZnT8RA, one of the two most common variants of the ZnT8A [6]. The association between thyroid autoantibodies and autoantibodies against ZnT8A is interesting since the zinc transporter is abundant in the cubical cells lining the thyroid follicle [10]. It would therefore be interesting to find out whether ZnT8A is found in patients with AITD without type 1 diabetes.

The findings of previous studies and our study suggest that HLA genotype may be important for the risk of develo** both type 1 diabetes and AITD. An association between AITD and HLA-DR3 [36], HLA-DR4 [24, 37] and DRB1*04-DQB1*03:01 [38] has been reported. In addition, HLA-DR3 and DR3/DQB1*03:02 has also been reported to contribute to both type 1 diabetes and AITD in a study of families with both diseases, while DR4 was associated with type 1 diabetes but not with AITD [25, 28, 39]. Others have failed to find an association between HLA and TPOAb [1]. Our findings, that DQ5.1 was protective for TPOAb and TGAb, confirm the results of a previous study [13]. Additionally, our finding of a weak association between TGAb and DQ2 in part confirms previously reported associations between AITD and DQ2 [24, 25]. However, our data were not consistent with an association between AITD and HLA-DQB1*03:02 (DQ8) [13, 24, 37]. The differences in results could partly be explained by the fact that our study was primarily focused on thyroid autoimmunity and not on clinical thyroid disease. It cannot be excluded that there may be differences in genetics between AITD and TPOAb and TGAb, similar to the relation between HLA and GADA and IA-2A on one hand and type 1 diabetes on the other [40]. Taken together, the association between HLA and AITD seem to be weaker than between HLA and type 1 diabetes despite these diseases often occurring coherently.

The high-risk genotype for type 1 diabetes, DQ2/8, predominates in children diagnosed under the age of 6 years [13, 37]. Individuals with thyroid autoantibodies at onset of type 1 diabetes diagnosis are often older and less frequently carry the high-risk DQ2/8 genotype. Our finding of an association between HLA-DQ2 and thyroid autoimmunity may explain some of the joint genetic susceptibility and family clustering of type 1 diabetes and AITD [25]. However, other non-HLA genes, as well as environmental factors, may contribute to the covariance of AITD and type 1 diabetes. Future research is needed to reveal these relationships as well as to determine how many of our patients with thyroid autoantibodies at onset of type 1 diabetes will develop AITD during follow-up.

In conclusion, children with type 1 diabetes were more likely to develop thyroid autoantibodies. At risk were girls and children with positive GADA or ZnT8RA. HLA-DQ5.1 including DQ8/5.1 seemed to protect against thyroid autoimmunity. However, except for the negative association between DQ5.1 and thyroid autoimmunity, HLA-DQ genotype did not significantly affect the risk of thyroid autoimmunity in childhood type 1 diabetes.