Shared Genetic Basis for Type 1 Diabetes, Islet Autoantibodies, and Autoantibodies Associated With Other Immune-Mediated Diseases in Families With Type 1 Diabetes

Type 1 diabetes (T1D) is a chronic disease that results from an immune-mediated destruction of the insulin-producing β-cells in the pancreatic islets of Langerhans. The specific β-cell destruction is caused by a complex interplay between multiple risk-conferring genes and environmental factors. To date, genome-wide association studies (GWAS) have identified more than 40 susceptibility loci for T1D (1). Except for the MHC region on chromosome 6p21, which contributes approximately 50% of the genetic risk, each of the loci identified through GWAS has modest individual effect on the total genetic risk for T1D. Autoantibodies directed against islet antigens are often detected at the onset of disease in affected individuals and are a distinctive signature of the autoimmune process that takes place at the initiation of the disease. Autoantibodies against insulin protein (IAA), IA-2 (IA-2A), GAD (GADA), and zinc transporter 8 (ZnT8A) have also been demonstrated to predict the onset of T1D in individuals at high genetic risk of the disease (2,3). Apart from the MHC region (4,5), very few studies have comprehensively studied the genetic causes of autoantibody development in T1D. Two GWAS have identified the FCRL3/1q23 locus as a susceptibility locus for IA-2A (6) and ZnT8 (7) positivity in a European population.

A number of genetic susceptibility loci are shared between T1D and other immune-mediated and autoimmune diseases (8), such as celiac disease, rheumatoid arthritis, and autoimmune thyroid disease (AITD). It is also known that some of these diseases commonly coexist in individuals with T1D and cosegregate in families with T1D. The Type 1 Diabetes Genetics Consortium (T1DGC) has collected the largest number of affected offspring families with T1D. In addition to genetic markers and measurements of islet autoantibodies (GADA, IA-2A, and ZnT8A), the T1DGC Autoantibody Workshop sample collection has also been investigated for the presence of other autoimmune diseases. Measurements of other autoimmune disease–associated autoantibodies in individuals with T1D were made available for analysis through the T1DGC Autoantibody Laboratory. In order to investigate the possible genetic overlap between T1D and autoantibody positivity, we analyzed the association between single nucleotide polymorphisms (SNPs) in 50 T1D risk loci identified through recent GWAS and GADA, IA-2A, and ZnT8A, as well as autoantibodies against thyroid peroxidase (TPOA) (detected in AITD), gastric parietal cells (PCA) (autoimmune gastritis), transglutaminase (TGA) (celiac disease), and 21-hydroxylase (21-OHA) (autoimmune hypoadrenalism).

A total of 7,077 T1D cases from 4,134 families were available from the T1DGC and collected by the four T1DGC regional networks. A total of 975 cases came from the Asia-Pacific Network, 2,480 from the European Network, 3,299 from the North American Network, and 323 from the U.K. Network. Caucasian (non-Hispanic white) ethnicity was self-reported in 84.0% of the cases, 10.4% were of Black or African American ethnicity, 5.4% were of Asian ethnicity, and 0.3% was of Native American, Native Alaskan, Hawaiian or Pacific Islander ethnicity. All T1DGC protocols and consents were approved by the individual institutional review boards. Informed consent (and assent) was obtained from members of all participating families. Inclusion criteria have been described previously (9). Fifty percent of the cases were female. The median age at diagnosis was 9 years (range 0–52 years), and the median disease duration at blood sampling was 7 years (range 0–63 years). Approximately 25% of the samples were taken within 3 years of diagnosis of T1D. A subset of cases were tested for ZnT8 (N = 1,504), with a disease duration less than 2 years with a median age at onset in those affected individuals of 11 years.

Autoantibodies were measured in serum that had been stored at −80°C. GADA, IA-2A, TPOA, TGA, and 21-OHA were measured by radiobinding assays (outlined in Akolkar et al. [10]) in two laboratories (Bristol, U.K., and Aurora, CO). PCA and ZnT8A were measured as described in Wenzlau et al. (11). Quality-control methods and results for the autoantibody measurements are described in Akolkar et al. (10).

Of the 7,077 T1D cases with autoantibody measurements, a total of 6,556 were genotyped for 50 disease-risk SNPs selected from published GWAS loci. The SNPs were genotyped using the TaqMan 5′ nuclease assay (Applied Biosystems) according to the manufacturer’s protocol. All SNPs had a genotyping success rate >90%, with 45/50 SNPs >95%. The mean genotyping success rate per individual was 95.7%. Genotype frequencies were tested for deviation from Hardy-Weinberg equilibrium. Three SNPs (rs2476601 in PTPN22, rs689 in INS, and rs9388489 in CENPW) exhibited statistically significant deviations from Hardy-Weinberg expectations (P < 0.001). These SNPs were not excluded from analysis, as disease association can cause genotype frequencies to deviate in affected offspring families.

The statistical analyses were performed using the R (www.r-project.org) package functions. In R, we used the geepack (12) function to perform logistic regression analysis while controlling for family relatedness. This approach used the generalized estimating equations (GEE) (13) method. Family ID was used to identify clusters and an exchangeable working correlation matrix and robust variance was used to test for association using the Wald test. Autoantibody positivity was coded as a binary phenotype. In addition to family relatedness, the logistic regression models were adjusted for self-reported (primary) ethnicity. Ethnicity was coded as a factor with four levels (1 = Caucasian, 2 = black or African American, 3 = Asia, 4 = Native Indian/Alaskan or Pacific Islander). Covariates, including age at diabetes onset, duration of diabetes, and sex, were included in the models if significantly associated with the autoantibody trait. SNP genotypes were included in the model under the assumption of additive allelic effects, coded 0, 1, or 2 for the number of minor alleles at a site. The estimated odds ratios (ORs) for autoantibodies were calculated for a 10-year difference in onset and duration, except for the ZnT8A that was measured in subjects with duration of diabetes less than 2 years. A false discovery rate (FDR)-adjusted P value below 0.05 (FDR <0.05) was used as the threshold for statistical significance. Despite only 50 SNPs tested, the threshold for genome-wide significance is defined as P < 5.0 × 10⁻⁸.

Of the 6,556 T1D subjects who had genotyping data available, 44.5% were positive for GADA, 46.6% were positive for IA-2A, 25.6% were positive for TPOA, 20.0% were positive for PCA, 7.4% were positive for TGA, and 1.6% were positive for 21-OHA autoantibodies. In the subset with ZnT8A measured, 1,335 cases were genotyped and 58.2% were positive.

Positivity for GADA was associated with later age at diabetes onset (OR 1.90, P < 2.0 × 10⁻¹⁶), shorter duration of disease (OR 0.67, P < 2.0 × 10⁻¹⁶), and greater prevalence of female sex (OR 1.52, P = 5.6 × 10⁻¹⁶). Positivity for IA-2A and ZnT8A was only associated with a shorter duration of diabetes (IA-2A: OR 0.52, P < 2 × 10⁻¹⁶; ZnT8A: OR 0.76, P = 8.7 × 10⁻⁵). Positivity for TPOA and PCA was associated with a later age at onset (TPOA: OR 1.23, P = 1.0 × 10⁻⁷; PCA: OR 1.27, P = 1.5 × 10⁻⁸), longer diabetes duration (TPOA: OR 1.22, P = 4.4 × 10⁻¹²; PCA: OR 1.37, P < 2 × 10⁻¹⁶), and female sex (TPOA: OR 2.22, P < 2 × 10⁻¹⁶; PCA: OR 1.72, P < 2.0 × 10⁻¹⁶). TGA positivity was associated with an earlier age at onset (OR 0.56, P = 5.7 × 10⁻¹¹), shorter disease duration (OR 0.83, P = 1.3 × 10⁻³), and female sex (OR 1.31, P = 4.8 × 10⁻³). 21-OHA positivity was only associated with a longer diabetes duration (OR 1.31, P = 3.5 × 10⁻⁴).

After adjusting for covariates, self-reported ethnicity also had an effect on autoantibody positivity. Compared with Caucasians, individuals of black or African American ethnicity had a slightly higher risk of positivity for GADA (OR 1.29, P = 0.013) but significantly lower risk of positivity for TPOA (OR 0.31, P = 1.4 × 10⁻¹²) and TGA (OR 0.36, P = 1.0 × 10⁻⁴). Individuals of Asian ethnicity had a significantly lower risk of positivity for IA-2A (OR 0.27, P < 2.0 × 10⁻¹⁶) and ZnT8A (OR 0.29, P = 3.7 × 10⁻⁶) compared with Caucasians. Although the Native American/Alaskan or Pacific Islander ancestry group was small (only 20 subjects), subjects of these ethnicities had higher risk for TGA positivity (OR 4.71, P = 5.2 × 10⁻³) compared with Caucasians.

The significant associations with T1D-associated SNPs for the seven autoantibodies are shown in Table 1. The logistic regression models were adjusted for family relatedness, self-reported ethnicity, and significant covariates. Eighteen of the 50 loci exhibited significant (FDR <5%) evidence of association with at least one autoantibody. Two associated MHC region SNPs, rs2187668 and rs7454108, are proxies for the predisposing HLA-DRB1*03 (DR3) and HLA-DRB1*04 (DR4) alleles, respectively (14,15). Both SNPs were associated strongly with autoantibody positivity. HLA-DR3 (rs2187668-A) was positively associated with GADA and TGA but negatively associated with IA-2A. HLA-DR4 (rs7454108-C) was positively associated with IA-2A, ZnT8A, PCA, TPOA, and 21-OHA.

For those T1D-associated SNPs not in the MHC region, only a SNP in the IFIH1/2q24 locus was significantly associated with GADA positivity. The major (A) allele of IFIH1 rs1990760 confers both risk of T1D and positivity for GADA. In contrast, six SNPs in non-MHC loci were associated with IA-2A positivity, residing in PTPN22/1p13, IL10/1q32, IFIH1/2q24, CTLA4/2q33, BACH2/6q15, and IL27/16p11. For SNPs associated with IA-2A positivity, the allele associated with risk of T1D was associated with protection from IA-2A positivity. The only exception was the A allele of CTLA4 rs3087243, which confers protection from T1D and is associated with IA-2A positivity. The ZnT8A was tested in 1,335 subjects and only the CTSH/15q25 SNP was associated with positivity for ZnT8A among the non-MHC SNPs.

Positivity for TPOA was detected in ∼25% of cases, whereas PCA was detected in ∼20% of cases. Positivity for PCA and TPOA was associated with a SNP in IFIH1/2q24, with the T1D risk-conferring major allele (A) of rs1990760 positively associated with both autoantibodies. A SNP in PTPN22/1p13 locus demonstrated association with TPOA and PCA, as did a SNP in SH2B3/12q24. For both PTPN22 and SH2B3 loci, the T1D risk allele also was associated with positivity for TPOA and PCA. In addition to the shared loci, TPOA was associated with SNPs in four additional non-MHC loci—BACH2/6q15, CD69/12p13, ERBB3/12q13, and SIRPG/20p13. Non-MHC SNP associations with PCA were observed in SKAP2/7p15, IL2RA/10p15, INS/11p15, and UBASH3A/21q22. For all SNPs in these T1D risk loci, the risk allele was associated with autoantibody positivity. The only exception was in the INS locus for which the T1D INS rs689 protective allele (T) was associated with positivity for PCA.

Positivity for TGA was detected in 482 cases and the only significantly associated non-MHC SNP was found in the T1D risk locus CTLA4/2q33. The protective minor allele (A) of CTLA4 rs3087243 was associated with positivity for TGA. Positivity for 21-OHA was only detected in 104 cases, yet SNPs in two non-MHC loci demonstrated significant association, PTPN22/1p13 and IL2/4q27, both of which conferred risk of T1D and autoantibody positivity (Fig. 1).

We demonstrate that there are genetic components that contribute to the risk of T1D and, at the same time, contribute to the development of autoantibodies associated with immune-mediated diseases. These shared genetic factors are replicated from the previous GWAS of autoantibody positivity in European subjects, such as the association for IA-2A with SNPs in the IL27 and IFIH1 loci; the association of TPOA with SNPs in the PTPN22, BACH2 and SH2B3 loci; and the association of PCA with SNPs in the IFIH1 locus (6). At the same time, our results do not replicate other loci reported in the GWAS, yet they suggest novel loci that could contribute to autoantibody positivity and therefore should be investigated in independent populations.

Five non-MHC T1D risk loci (IFIH1, PTPN22, SH2B3, BACH2, and CTLA4) contained SNPs that were consistently and robustly associated with positivity for more than one autoantibody. These loci are known to confer risk of several autoimmune disorders, in varying degrees, including T1D (1), AITD (6,16), and celiac disease (17,18). The IFIH1 SNP rs1990760 was associated with positivity for PCA, TPOA, and GADA, yet it was associated with protection from IA-2A. IFIH1 has not previously been associated with the risk of autoimmune gastritis, but association has been reported in AITD (19), although convincing evidence is lacking from large GWAS (20,21). IFIH1 encodes a viral dsRNA-activated apoptosis protein with a role in host protection and clearance of virus infected cells (22). Enterovirus infections have been linked to T1D in epidemiological studies (23) and IFIH1 has, therefore, been suggested as a link between viral triggers and autoimmunity in T1D (24). The fact that SNPs in IFIH1 are associated with autoimmune diseases with no established link to virus infections and the strong effect of IFIHI on autoantibody positivity demonstrated in this study suggest that IFIH1 may have additional functions specific for autoimmunity. In pancreatic β-cells, IFIH1 has been shown to regulate local release of inflammatory mediators after dsRNA stimulation and contribute to apoptosis (25). Thus, IFIH1 could have cell-specific function that relates the immune system to selective target cells and contributes to autoimmunity.

PTPN22, SH2B3, BACH2, and CTLA4 all have reported function associated with the adaptive immune response, such as regulation of T-cell receptor (TCR) signaling, T-cell activation, or the antibody response. The PTPN22 SNP was associated with positivity for TPOA, PCA, and 21-OHA but with protection from IA-2A positivity. The PTPN22 gene encodes a lymphoid-specific intracellular phosphatase involved in the negative regulation of T-cell activation (26). The risk-associated allele is a gain-of-function variant that controversially results in a reduction of TCR signaling and activation (27). This same variant has been associated with the risk of AITD (28) and autoimmune hypoadrenalism (29,30) but not with autoimmune gastritis. Although IFIH1 and PTPN22 have relatively modest effects on the risk of positivity for islet autoantibodies, they both contribute to later development of markers for autoimmune hypoadrenalism, AITD, and autoimmune gastritis in T1D, especially in females.

SH2B3 SNP was associated with positivity for both TPOA and PCA; however, in our analyses, there was only marginal significance of the association with GADA (6) (P = 0.07). The SH2B3 gene encodes an adaptor protein involved in T-cell activation by regulating TCR signaling (31). The SH2B3 locus also confers susceptibility to celiac disease (18) but has not been associated with AITD or autoimmune gastritis.

A SNP in BACH2 was associated with positivity for TPOA but with protection from IA-2A positivity, a finding that has not been reported previously. BACH2 encodes a transcriptional repressor/activator involved in regulation of antibody response (32) and has, among others, been associated with the risk of AITD (20) and celiac disease (17,18). Recently, BACH2 was also identified as a regulator of β-cell apoptosis (33).

The SNP in the CTLA4 locus was associated with IA-2A and TGA, results that have not been reported previously. Variants in the CTLA4 locus have been associated consistently with susceptibility to other immune-mediated conditions, including celiac disease (18) and AITD (34,35). The CTLA4 gene encodes a major negative regulator of T-cell activation (36). The major allele (G) of the CTLA4 rs3087243 SNP confers risk of T1D and celiac disease and is the only risk allele that was associated with positivity for IA-2A; however, it was associated with protection from TGA autoantibodies. This differential effect highlights the degree of complexity in which the exact pathogenic mechanisms may differ between autoimmune diseases.

Eleven non-MHC SNPs demonstrated significant association with only a single autoantibody. For IA-2A, we replicate the association with IL27 (6) and also report an association with IL10. For TPOA, we report novel associations with SNPs in CD69, ERBB3, and SIRPG. PCA positivity was associated with SNPs in SKAP2, IL2RA, INS, and UBASH3A. 21-OHA positivity was associated with SNPs in IL2, while positivity of ZnT8A was associated with a SNP in CTSH. Most of these genes have been associated with immune regulation, e.g., T-cell activation and signaling. The single-autoantibody specificity association with single SNPs/genes also suggests target organ-specific mechanisms of autoimmunity.

HLA-associated alleles in the MHC are, not surprisingly, the major determinants of the risk for autoimmune disease as well as autoantibody development. Although the coverage of the MHC region in these data were limited, we replicated the association for HLA-DR3 with GADA positivity as well as the positive association of HLA-DR4 with IA-2A (4,5,37). Positivity for ZnT8A was associated with HLA-DR4 but not HLA-DR3, which might reflect a true lack of association or the lower power given the smaller subset of individuals that were tested. All nonislet autoantibodies demonstrated a positive association with HLA-DR4, with the exception of TGA that was associated with HLA-DR3. The HLA-DR3-DQ2 haplotype is strongly associated with celiac disease (38) and is required for presentation of dietary wheat epitopes that have been modified by tissue transglutaminase to T cells in the intestine (39). For TPOA and PCA, the association with SNPs in the MHC is weaker than that observed for SNPs in PTPN22 and IFIH1. This interesting result may be explained by the genotyped SNPs failing to “tag” the most important HLA alleles. While risk of AITD and autoimmune hypoadrenalism is primarily associated with HLA-DR3 (40–42), autoantibody positivity was strongly associated with HLA-DR4, suggesting a difference in genetic control between disease susceptibility and development of autoimmunity.

Our study has lower power to detect associations of T1D-associated SNPs with 21-OHA, TGA, and ZnT8A due to the relatively low number of cases that were autoantibody positive. We cannot eliminate the possibility of an association with other T1D risk loci for these autoantibodies, especially for TGA, given the documented overlap between the genetic risk for T1D and celiac disease. Further, a majority of the subjects in this study come from a multiethnic collection of affected offspring trios. Although the data analysis controlled for relatedness among siblings and primary ethnicity, it is possible that the test statistics are inflated by residual stratification in the data. The effect sizes of the replicated loci, however, correspond with those reported in the European case-control cohort (6), suggesting that the test statistic inflation may be low. The lack of correspondence for some loci could also be explained by differences in effect sizes in a family-based cohort compared with a case-control–based cohort, as has previously been suggested for T1D (1).

These analyses are consistent with recent reports that demonstrate a relationship between alleles that confer risk of autoimmune disease (e.g., T1D) and IA-2A positivity. IA-2A positivity was associated consistently with the protective alleles for all SNPs in T1D loci, with the exception of CTLA4, and exhibited an inverse relationship with risk of positivity for other autoantibodies. This paradoxical relationship is also apparent for variants in the MHC region, where the HLA-DR3 and HLA-DR4 alleles confer strong protection versus risk of IA-2A, whereas the heterozygous DR3/DR4 genotype confers the highest risk of T1D. These findings support that the islet autoantibodies are not, in themselves, pathogenic in T1D but are a consequence of the immune-mediated destruction of the β-cells. The association of FCRL3 SNPs with IA-2A and ZnT8A but not with T1D supports this observation (6). In conclusion, the major loci reported here associate with multiple autoantibodies. Interestingly, four of the five loci showing significant association with multiple autoantibody positivity are also associated with seronegative autoinflammatory conditions, but here the main genetic signal is discordant with that of T1D (PTPN22, IFIH1, and BACH2 for inflammatory bowel disease and SH2B3 for ankylosing spondylitis) (43). These results highlight putative causal defects in T-cell activation and regulatory T cells, which may lead to defective control over B-cell tolerance and thereby contribute to initiation as well as maintenance of autoimmunity.

Funding. This study used resources provided by the T1DGC, a collaborative clinical study sponsored by the National Institute of Diabetes and Digestive and Kidney Diseases, the National Institute of Allergy and Infectious Diseases, the National Human Genome Research Institute, the Eunice Kennedy Shriver National Institute of Child Health and Human Development, and JDRF and supported by U01 DK062418. Parts of the project were funded by the National Institutes of Health (1-DP3-DK-085678-01).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. C.A.B. and F.P. researched data and wrote and edited the manuscript. F.P. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.