Linkage Disequilibrium between the Human Leukocyte Antigen Class II Region of the Major Histocompatibility Complex and Graves’ Disease: Replication Using a Population Case Control and Family-Based Study¹

GRAVES’ disease (GD) is an autoimmune disease of the thyroid gland characterized by hyperthyroidism, diffuse goiter, and positive thyroid autoantibodies. Although the etiology is unknown, the disease process appears to be influenced by protective and permissive environments in genetically susceptible individuals (1). After genome-wide searches in type 1 diabetes and the identification of multiple susceptibility loci, it is likely that most autoimmune diseases will exhibit a polygenic mode of inheritance (2). Preliminary evidence for linkage on chromosome 14q31 (GD1) (3) and case control studies of the cytotoxic T lymphocyte-associated-4 (CTLA-4) gene (4, 5) in GD are in support of this. The concordance rate in GD in monozygotic twins is similar to that in type 1 diabetes (3, 4), and epidemiological data have shown that both diseases cluster within the same families and individuals (5). As the human leukocyte antigen (HLA) class II region is the major susceptibility locus for type 1 diabetes, accounting for 40% of familial clustering in the UK (λs = 3, the ratio of the expected proportion of affected sibling pairs sharing zero alleles identical by descent, 0.25, and the observed proportion) (2), it seems likely that it will play a similar role in GD. This region is an obvious candidate for a role in the genetic susceptibility to GD, as there is aberrant expression of the major histocompatibility complex (MHC) class II antigens on follicular cells (the target cell in the autoimmune process) and on activated lymphocytes in patients with disease.

Although classical linkage analysis has been successfully used to find major genes, its ability to detect genes of modest effect has been limited (9). This may explain why linkage analysis of the HLA region in GD in family-based studies (10–12) has failed to replicate case control data (13–18).

We have, therefore, examined the MHC HLA class II region in patients with GD using the alternative approach of linkage disequilibrium analysis in two independent United Kingdom datasets. The Transmission Disequilibrium Test (TDT) (19) was used in a family-based study to replicate the findings of a population-based case control study.

Subjects and Methods

Caucasian patients (with both parents and both grandparents of British or Irish origin) with GD were recruited from three large thyroid clinics in Birmingham and Bournemouth, UK. Graves’ patients were defined by the presence of biochemical hyperthyroidism together with 2 of the following criteria: diffuse goiter; a significant titer of microsomal, thyroglobulin, or TSH receptor autoantibodies; and the presence of dysthyroid eye disease. Taking all index cases together, 87% had diffuse goiter, and 95% had positive thyroid autoantibodies. Thyroid eye disease was present in 22.7% of the index cases classified using the NOSPEC classification, with eye disease defined as positive features in any of classes 2–6 (20). Eyelid signs alone were not considered indicative of thyroid eye disease. Microsomal and thyroglobulin antibodies were measured by gelatin particle agglutination (SERODIA-AMC and SERODIA-ATG, Fujirebio, Inc., Tokyo, Japan), and a titer of 1:100 was considered significant for both assays. TSH receptor autoantibody status was determined by a radioactive inhibition method (RSR Ltd., Cardiff, UK). A value of 10 U/L or more was deemed significant after comparison with 50 controls obtained from the local blood transfusion service. Ethnically matched controls with no history of autoimmune disease were bled at various sites, including the Blood Transfusion Service, Birmingham Heartlands Hospital, and the Queen Elizabeth Hospital. Simplex families were also recruited, where in most instances blood samples were obtained from the index case, both parents, and any unaffected siblings. All siblings were tested for thyroid function and autoantibody status, and any showing evidence of subclinical autoimmune thyroid disease were removed from the study before genotyping was performed. In total, DNA was obtained from 228 Graves’ index cases, 58 males (25%) and 170 females (75%), and 364 controls, 77 males (22%) and 287 females (78%), for the population association study and from 98 simplex families.

The study was approved by the respective local ethics committees, and all subjects gave informed, written consent.

Genotyping of datasets

DNA was prepared from 10 mL whole blood using the Nucleon Bacc II kit from Nucleon Biosciences (UK). The HLA-DRB1, DQB1, and DQA1 regions were amplified using the phototyping method of the PCR, as previously published (21). Primers were obtained from Oswel (UK) and R&D Systems (UK). Results were visualized on a 1% ethidium bromide-stained agarose gel under ultraviolet light.

Statistical analysis

Analysis of the case control data was performed using theχ ² test with 95% confidence limits (CL). All P values were corrected for the number of comparisons made, and P < 0.05 was considered significant. Odds ratios (ORs) were calculated by the method of Woolf with Haldane’s modification for small numbers (22).

The TDT (19) was used to assess linkage disequilibrium between the HLA susceptibility haplotype and disease in the family dataset. A significant excess of transmission frequency of the associated haplotype DRB1¹0304-DQB1¹02-DQA1¹0501 from parents heterozygous for that haplotype to affected offspring was taken as evidence of linkage disequilibrium. Comparisons of transmissions to unaffected offspring using the 2 × 2 test of heterogeneity were performed to exclude segregation distortion.

Results

Population-based case control study

Two hundred and twenty-eight Graves’ patients and 364 healthy control subjects were available to be genotyped at the DRB1, DQB1, and DQA1 loci. All patients and control subjects were successfully genotyped at all 3 loci.

The distribution of DRB1 alleles among patients and control subjects is summarized in Table 1. A significant increase in the frequency of the DRB1¹0304 allele was seen in patients compared to control subjects (47% vs. 24%, respectively; OR = 2.72; pc < 1.4 × 10⁻⁵).

The distribution of DQB1 alleles among patients and control subjects is summarized in Table 2. Significant increases in the frequency of DQB1¹02 and DQB1¹0301/4 were seen in patients compared to controls [58% vs. 46%, respectively (OR = 1.6; pc = 0.035), and 42% vs. 28%, respectively (OR = 1.86; pc < 3.5 × 10⁻³)].

The distribution of DQA1 alleles among patients and control subjects is summarized in Table 3. A significant increase in the frequency of DQA1¹0501 was seen in patients compared to controls (67% vs. 39%, respectively; OR = 3.16; pc < 7 × 10⁻⁶). Due to recent reports of the independent association of DQA1¹0501 with Graves’ disease, this allele was compared in DRB1¹0304 negative patients and control subjects. There was no evidence of an increase in the frequency of the DQA1¹0501 allele in DRB1¹0304-negative Graves’ patients (46%) compared to controls (54%; χ² = 0.62; P= NS). Linkage disequilibrium between DRB1¹0304 and DQA1¹0501 was confirmed by the increased frequency of the DQA1¹0501 allele in DRB1¹0304-positive (69%) compared to DRB1¹0304-negative (31%) Graves’ patients (χ² = 23.4; P < 1× 10⁻⁶).

Due to the tight linkage disequilibrium among DRB1¹0304, DQB1¹02, and DQA1¹0501, distribution of the haplotype DRB1¹0304-DQB1¹02-DQA1¹0501 was compared between patients and control subjects (Table 4). A significant increase in the frequency of this haplotype was seen in patients compared to control subjects (47% vs. 24% respectively; OR = 2.72; pc < 1.8 × 10⁻⁵).

Family-based studies

Transmission of the haplotype DRB1¹0304-DQB1¹02-DQA1¹0501 from heterozygous parents to affected and unaffected siblings was analyzed using a standard TDT, McNemar’s test (19). Of the 98 families available for study, 54 were informative for this haplotype. A significant increase in transmission (T) compared to nontransmission (NT) of the DRB1¹0304-DQB1¹02-DQA1¹0501 haplotype was seen in affected offspring [44 T (72%) vs. 17 NT (28%); χ² = 11.95; 1 d.f.; P = 0.0005; Table 5]. There was no preferential transmission of this haplotype to unaffected offspring (53%). The 2 × 2 test of heterogeneity comparing transmission of the DRB1¹0304-DQB1¹02-DQA1¹0501 haplotype to affected and unaffected offspring (χ² 4.1; 1 d.f.; P = 0.04) confirms that the significant excess of transmissions to affected offspring was not the result of segregation distortion.

To determine whether there was differential expression of the maternal or paternal haplotypes, we examined transmission of the parental DRB1¹0304-DQB1¹02-DQA1¹0501 haplotypes to affected offspring. There was no significant difference between transmission from fathers (23 T, 52%) and mothers (21 T, 48%) and, therefore, no evidence of parent-of- origin effects at HLA in this UK dataset. Because of the higher prevalence of GD among females, we also examined transmission of the DRB1¹0304-DQB1¹02-DQA1¹0501 haplotype by the sex status of the affected offspring. Again, no evidence for differential transmission to males (8 of 11, 80%) or females (32 of 43, 75%) was seen. Although the numbers were small, there is no evidence for a trend, and similar frequencies of the haplotype were seen in males (27 of 58, 46%) and females (80 of 170, 47%) in the case control study.

Discussion

HLA class II associations with GD have been reported in many case control studies, particularly associations of DR3 and DQA1¹0501 (13–18). However, these results have been inconsistent, and linkage between GD and HLA remained to be confirmed in family-based studies (10–12). Studies reporting evidence against linkage support the suggestion that the HLA region is exerting a modest effect on the expression of the disease which can not easily be detected by classical linkage analysis. In the current study, we confirmed associations of the disease with the alleles DR3 (DRB1¹0304) and DQA1¹0501, with the latter conferring the highest risk of disease (OR = 3.16; 95% CL = 2.23–4.46). These results are in agreement with those seen in previous case control studies (15, 17, 23–27). Lack of an independent association of DQA1¹0501 with disease, despite the increased OR, may be the result of the allele being in linkage disequilibrium with an unknown non-HLA gene that has a stronger association with disease. However, a more likely explanation is linkage disequilibrium between DQA1¹0501 and DR5 (DRB1¹11 and DRB1¹12). All 42 Graves’ patients with DRB1¹11 also carried the DQA1¹0501 allele, suggesting that the increase in DQA1¹0501 in DRB1¹0304-negative patients is due to linkage disequilibrium with DRB1¹11. We also found weak association of the DQB alleles, DQB1¹02 and DQB1¹0301/4, with disease, implying that association with disease was secondary to that conferred by DRB1 and DQA1 alleles. All of the 107 Graves’ patients that were DRB1¹0304-DQA1¹0501 positive carried the DQB1¹02 allele, suggesting that the increase in DQB1¹02 is due to linkage disequilibrium with DRB1¹0304-DQA1¹0501.

Due to the increased frequencies of the alleles DRB1¹0304, DQB1¹02, and DQA1¹0501, the strong linkage disequilibrium between them, and their lack of independent association with disease, we investigated the frequency of the haplotype DRB1¹0304-DQB1¹02-DQA1¹0501 in our cohort of Graves’ patients and control subjects. An increase in the frequency of this haplotype was seen in patients compared to control subjects and was consistent with the allele frequency data (OR = 2.72; confidence limits = 1.91–3.87; pc = 1.8 × 10⁻⁵). This result suggests that this haplotype is conferring genetic susceptibility to Graves’ disease in our population.

Although the case control approach has been frequently used to test candidate genes (28), it is widely accepted that differences in allele frequencies between a diseased population and normal control subjects may arise for reasons other than linkage disequilibrium between alleles of the marker and alleles of the disease locus. False positives can occur as the result of random chance events, a common occurrence in small datasets, or to population stratification (29). To minimize random chance events, it is important to use adequately sized datasets; to eliminate population stratification, it is crucial for susceptibility genes to be examined in family-based studies.

Classical linkage analysis has identified major genes in complex diseases, although its ability to detect genes of modest effect has proved limited (9). The TDT is a more powerful approach for detecting susceptibility loci that may be missed by classical linkage analysis (9). The detection of such susceptibility loci is important because, although likely to have small effects, the magnitude of their attributable risk may be large, as they are frequent in the general population. Moreover, by using single affected individuals and parents, the TDT does not require families with multiple affected siblings. This approach is particularly valuable, therefore, in diseases of late onset, such as autoimmune thyroid disease.

In the present study we used the TDT, which has been considered to be a test of linkage in the presence of linkage disequilibrium (19). In an independent family dataset, we examined the frequency of transmission of DRB1¹0304-DQB1¹02-DQA1¹0501 from parents heterozygous for the haplotype. We found a significant preferential transmission of the DRB1¹0304-DQB1¹02-DQA1¹0501 haplotype to affected Graves’ index cases (72%), providing evidence of linkage disequilibrium.

To overcome the confounding problem of subclinical or undiagnosed autoimmune thyroid disease in unaffected siblings, all had thyroid function and thyroid antibody status measured. Eight of 100 unaffected siblings (8%) were found to have evidence of subclinical autoimmune hypothyroidism, including a raised TSH receptor level and positive thyroid antibodies, and were excluded from our analysis. The transmission of the disease haplotype DRB1¹0304-DQB1¹02-DQA1¹0501 to unaffected siblings (25 transmissions, 53%) was reassuring and demonstrated that well characterized unaffected siblings can be used as controls for TDT analysis in GD. Inclusion of siblings with subclinical undiagnosed autoimmune thyroid disease as unaffected subjects, however, had no significant effect on the family results (data not shown). The 2 × 2 test of heterogeneity, performed between affected and unaffected offspring, ruled out segregation distortion as an alternative explanation to linkage disequilibrium of DRB1¹0304-DQB1¹02-DQA1¹0501 with GD.

In conclusion, we have shown for the first time in a family-based study that the MHC HLA class II region on chromosome 6p is in linkage disequilibrium with GD in Caucasians in the United Kingdom. The TDT in the family study provides strong confirmatory evidence for linkage disequilibrium between the haplotype DRB1¹0304-DQB1¹02-DQA1¹0501 and disease found in the case control study. No independent association of DQA1¹0501 was seen over and above linkage disequilibrium with DR alleles, suggesting that genetic susceptibility to GD cannot be attributed to any one allele.

Acknowledgements

We thank John Todd for his critical appraisal of the manuscript, and our colleagues for help in recruiting patients.