Introduction

The prevalence of antibodies to the protein tyrosine phosphatase islet antigen-2 (IA-2A) at presentation of type 1 diabetes is influenced by HLA-DRB1*04 [1], but other genes may be important. We previously reported that IA-2A at disease onset in children and adolescents did not vary with sex or age [2], but have now determined IA-2A and HLA-DRB1, -DQA1 and -DQB1 genotype in a larger dataset of patients, allowing us to investigate the influence of these alleles on IA-2A prevalence. We have also investigated the effect of these factors on the prevalence of antibodies to the juxtamembrane region of IA-2 (JMA), which have been associated with faster progression to clinical disease in at-risk children [3]. Identifying the important determinants of humoral autoimmunity to this major autoantigen should improve our understanding of the pathways leading to disease, allow refinement of prediction, and may suggest new targets for immunointervention.

Methods

Type 1 diabetes

Serum was obtained from 618 children and adolescents with recently diagnosed type 1 diabetes, recruited to the Bart’s–Oxford study of childhood diabetes between 1985 and 2002 [2]. All required insulin treatment from diagnosis, and were resident in the Oxford region of the UK. Median age at diagnosis was 11 years (range 0.8–20.9), and 352 were male. Samples were collected no later than 90 days after diagnosis (median 1 day) and within 30 days in 445. The study was approved by Local Research Ethics Committees.

Assay for IA-2A

Samples were assayed for IA-2A as previously described [2], using 35S-labelled IA-2ic (605–979). Results were expressed in arbitrary units derived from a standard curve. Samples with levels above the 97.5th percentile of 2,860 schoolchildren (0.9 units) were considered positive. The IA-2A assay had an inter-assay CV of 21% at both 0.7 and 1.7 units and achieved a laboratory-defined sensitivity of 58% with a specificity of 98% in the First Diabetes Antibody Standardization Program [4].

Assay for JMA

Binding to the juxtamembrane (JM) region of human IA-2 was investigated in 481 of the 484 IA-2A-positive samples. Samples were assayed as for IA-2A, but using a 35S-labelled chimera protein encoding the JM region (609–631) of IA-2 [3]. Results were expressed as an index against a serum containing antibodies specific to the JM region. The JMA threshold was defined as the level three SDs above the mean of 80 schoolchildren. The inter-assay CV of the JMA assay was 18.4% at 0.68 units and 24.2% at 0.31 units.

Genetic analysis

HLA class II genotyping was carried out on blood or mouth swab DNA using published methods for DNA extraction and HLA class II DRB1, DQA1 and DQB1 analysis by PCR with sequence-specific primers [5].

Statistical analysis

Proportions were compared using χ 2 testing. Binary logistic regression was used to determine the influence of age, sex and HLA haplotypes on IA-2A and JMA prevalence. Age at testing was modelled in four age bands. Results for haplotype comparisons are expressed as ORs with 95% CIs in comparison with patients not carrying the respective haplotypes. The influence of IA-2A levels on JMA prevalence was investigated by modelling IA-2A levels as quartiles, with results expressed in comparison with the highest quartile. All determinants found significant at the 10% level in univariate analysis were tested together in the final model. Analyses were performed using the Statistics Package for Social Sciences Version 12.0.1 (SPSS, Chicago, IL, USA). Results were considered significant at the 5% level.

Results

In this dataset, 78% of patients had IA-2A. There was no significant variation in IA-2A prevalence with age or sex (Table 1).

Table 1 IA-2A prevalence according to patient characteristics, HLA class II genotypes and DQA1 haplotypes
Full size table

IA-2A and HLA

The prevalence of IA-2A was strongly influenced by HLA genotype; prevalence was highest in patients carrying at least one HLA-DRB1*04-DQA1*0301 (385 of 450; 86%), DRB1*07-DQA1*(0201 or 0301) (58 of 64; 91%) or DRB1*09-DQA1*0301 haplotype (18 of 19; 95%). Only 69 of 132 (52%) patients carrying none of these haplotypes had IA-2A, compared with 322 of 391 (82%) patients with one and 93 of 95 (98%) with two of these haplotypes (p < 0.001). IA-2A prevalence was similar in the 364 heterozygous patients in whom DRB1*04 was in linkage with DQB1*0302 and the 37 patients in whom it was in linkage with DQB1*0301 (84 vs 89%, p = 0.39). Further, IA-2A prevalence was similar in DRB1*04 heterozygous patients carrying DRB1*0401 and those carrying other DRB1*04 subtypes (231 of 273 [85%] vs 109 of 131 [85%]).

Multiple logistic regression analysis showed that HLA-DRB1*04 (p < 0.001), DRB1*07 (p < 0.001) or DRB1*09 haplotypes (p = 0.001) were independent determinants of IA-2A prevalence (Fig. 1). The highest OR was seen in patients carrying two of these haplotypes in any combination (OR 42 [95% CI 10–180], p < 0.001) (Fig. 1).

Fig. 1
figure 1

ORs (diamonds) and 95% CIs (horizontal lines) for independent determinants of IA-2 antibody positivity derived from binary logistic regression analysis for 618 patients. Dependence of IA-2A prevalence on HLA-DRB1 haplotypes as covariates was tested in Model A for patients carrying one or two DRB1*04 haplotypes, at least one DRB1*07 haplotype or at least one DRB1*09 haplotype when tested together, and in model B for patients carrying one or two DRB1*04, DRB1*07 or DRB1*09 haplotypes in any combination compared with patients not carrying the respective haplotypes a p < 0.001, b p = 0.001

Full size image

JMA

JMA were found in 202 of 481 individuals tested (42%). There was no significant variation in JMA prevalence with age or sex (Electronic supplementary material [ESM] Table 1). JMA were less common in patients with haplotypes including DQB1*02 alleles. Only 12 of 38 (32%) IA-2A-positive patients carrying two DRB1*03-DQB1*02 or DRB1*07-DQB1*02 haplotypes in any combination were JMA positive and 100 of 277 (36%) with one of these haplotypes compared with 90 of 166 (54%) with none (p < 0.001). JMA prevalence also increased with IA-2A quartile (p < 0.001), from 31% in the lowest to 56% in the highest (ESM Table 1).

Multiple logistic regression confirmed that JMA in IA-2A-positive patients were negatively associated with DRB1*03-DQB1*02 or DRB1*07-DQB1*02 haplotypes (OR, two haplotypes, 0.43, 95% CI 0.2–0.92; OR, one haplotype, 0.48, 95% CI 0.32–0.71, p = 0.001) and positively associated with IA-2A quartile (OR, 1st, 0.38, 95% CI 0.22–0.65; OR, 2nd, 0.39, 95% CI 0.23–0.66; OR, 3rd, 0.71, 95% CI 0.42–1.19, p < 0.001).

Discussion

Our major finding was that the prevalence of IA-2A at disease onset in patients under 21 years is strongly associated with HLA-DRB1*07 and HLA-DRB1*09 as well as HLA-DRB1*04. IA-2A prevalence was highest in patients carrying at least one HLA-DRB1*04-DQA1*0301, DRB1*07-DQA1*(0201 or 0301) or DRB1*09-DQA1*0301 haplotype. The effect increased with the number of these haplotypes, and only two of 95 patients carrying any two of these haplotypes did not have IA-2A. Antibodies to the JM epitope of IA-2 were more common in IA-2A-positive patients with higher levels of IA-2A. JMA prevalence was also influenced by HLA class II haplotypes, being lower in those patients carrying DRB1*03-DQB1*02 or DRB1*07-DQB1*02.

The DRB1*04, DRB1*07 or DRB1*09 alleles belong to the DR53 extended haplotype [6] and share several unique characteristics, including a second expressed DRB locus, DRB4, which can play a role in presentation of autoantigens to T cells [7]. These DRB1 haplotypes also show similarities at HLA-DQA1. In our patients, both DRB1*04 and DRB1*09 were in strong linkage disequilibrium with DQA1*0301, while DRB1*07 was in strong linkage disequilibrium with DQA1*0201, which has close homology with DQA1*0301, particularly in exons 2 and 3. These exons encode the extracellular portion of the HLA-DQA1 molecule which, in combination with DQB1, forms the antigen recognition site. Further, the interaction between the HLA-DQ8 haplotype and an immunodominant insulin peptide, suggests that residues located in the P1 and P9 pockets (\({\text{Glu}}^{31\alpha } \) and \({\text{Ile}}^{72\alpha } \), respectively) shared between DQA1*0301 and DQA1*0201 may be important for antigen presentation [8]. Strong linkage means we cannot discriminate effects at HLA-DRB1 and HLA-DQA1, but our findings are consistent with a major genetic determinant of humoral autoimmunity to IA-2 being carried by HLA-DQA1. Modulation of the immune response to IA-2 may therefore be an effective approach to diabetes prevention in children carrying DQA1*0301 or DQA1*0201 in the context of not only DRB1*04, but also DRB1*07 or DRB1*09 haplotypes. By including some genotypes currently categorised as low-risk, this strategy would be applicable to more children than one based on DRB1*04 haplotypes alone.

IA-2 antibodies have been associated with DRB1*09 in the Japanese [9], but this allele is uncommon in Europeans. The strong association of IA-2A with DRB1*07 haplotypes in patients with type 1 diabetes has not been recognised previously, possibly because of their lower frequency in comparison with DRB1*04 haplotypes. Indeed, DRB1*07 haplotypes are considered neutral or protective in the UK population [5]. Of our patients, only 12 had the genotype DQA1*0201-DQB1*0201 (DRB1*07) combined with DQA1*0501-DQB1*0201 (DRB1*03), but 11 (92%) of these had IA-2A despite being associated with a lower prevalence of multiple antibodies and diabetes risk in relatives [10]. In contrast, only 23 of 49 (47%) patients homozygous for DRB1*03 haplotypes had IA-2A (p = 0.008) despite the much higher diabetes risk associated with this genotype [5].

The lower prevalence of antibodies directed to the JM epitope in IA-2A-positive patients carrying DQB1*02 is unreported. DQB1*02 is carried on DRB1*03 and *07 haplotypes. Presumably, increased prevalence of IA-2A in patients with DRB1*07 haplotypes is because of antibodies directed mainly to the other major epitopes of IA-2 found in the protein tyrosine phosphatase domain [3]. The increased JMA prevalence at higher IA-2A levels suggests increased epitope spreading in those with a more vigorous humoral response.

In conclusion, HLA class II alleles strongly influence humoral autoimmunity to IA-2 in patients at disease onset. IA-2A prevalence was highest in patients carrying DQA1*0301 or the closely related DQA1*0201, suggesting that the antibody response to IA-2 may be determined by DQA1 alleles.