Introduction

The genetic basis of autoimmune thyroid disease (AITD), which includes Graves' disease (GD) and Hashimoto's thyroiditis (HT), is thought to be polygenic (Gough 2000; Brix and Hegedus 1998). Several genetic factors associated with AITD have been tentatively identified by candidate-gene and genome-scanning approaches. They include HLA, cytotoxic T-lymphocyte-associated-4 (CTLA-4), and TSHR gene, and other chromosomal regions, such as 14q31 (GD-1), 20q11.2 (GD-2), Xq21.33–22 (GD-3), 13q32 (HT-1), 12q22, 18q21 (IDDM 6), and Xp11 (Farid et al. 1979; Uno et al. 1981; Yanagawa et al. 1995; Donner et al. 1997; Vaidya et al. 1999; Barbesino et al. 1998; Tomer et al. 1999; Vaidya et al. 2000; Sale et al. 1997; Akamizu et al. 2000). Evidence for AITD genes, however, varies among researchers or populations, and no consensus has been obtained.

Recently, Sakai and co-workers mapped susceptibility loci for HT to 8q23-q24 and AITD to 5q31-q33 by multipoint affected sib-pair linkage analysis in a collection of Japanese AITD families (Sakai et al. 2001). First, chromosome 8q23-q24, which encompasses the thyroglobulin gene, yielded suggestive evidence for linkage to HT with an MLS of 3.77 at D8S272. A recent report confirmed that the thyroglobulin gene is linked and associated with AITD (Tomer et al. 2002). Secondly, chromosome 5q31-q33 yielded suggestive evidence for linkage to AITD with an MLS of 3.14 at D5S436. Chromosome 5q31-q33 includes a cytokine gene cluster and other immune-related genes, and has been reported to be associated or linked to allergic diseases such as asthma and atopic dermatitis (Meyers et al. 1994; Postma et al. 1995; Lonjou et al. 2000). Considering that AITD is an autoimmune disease, this is an intriguing region, where susceptible genes may exist.

In this study we attempted to confirm evidence for the presence of susceptibility genes in or near 5q23-q33 by association analysis, using several microsatellite markers, including D5S436, the linked marker from the genome screen. In addition, we carried out subgroup analysis of Graves' patients based on clinical manifestations in different subgroups and analyzed if the patterns of their associations with these markers were different between subgroups.

Subjects and Methods

Study subjects

Four hundred and forty unrelated Japanese individuals with AITD (337 female, 103 male) were evaluated. In cases where more than one individual from a family was collected, data from only a single individual from the family was included in the genetic analysis. Patients were classed as having either GD (n=241), HT (n=139), primary hypothyroidism with TSH-binding inhibitory immunoglobulin (TBII) (n=35), or hypothyroidism without TBII (n=25). These patients were diagnosed and assayed for antibodies as described previously (Sale et al. 1997; Akamizu et al. 2000). As an unaffected control population, we have also collected an expanded set of controls, consisting of 218 unrelated Japanese individuals (89 male, 129 female). Informed consent was obtained from all the subjects used in this study and the present study was approved by the ethical committee of Kyoto University Graduate School of Medicine.

Genotyping loci at 5q23-q33

Loci at 5q15-q33 were genotyped using seven microsatellite markers derived from the ABI PRISM Linkage Mapping Set Version 2 (Table 1). PCR was performed using the GeneAmp System 9600 with standard protocols. The PCR products were analyzed by electrophoresis through a 6% denaturing polyacrylamide gel using an ABI 373 automated sequencer apparatus, and the resulting data were analyzed with Genescan software (V 1.2) (Akamizu et al. 2000).

Table 1. Chromosome localization, allele size range, and heterozygosity of 5q microsatellite markers in a Japanese population
Full size table

Association analyses

The frequencies of alleles of loci at 5q23-q33 in the Japanese were calculated from the control populations. Male and female frequencies were compared using the RXC program (George Carmody, Carleton University, Ottawa, Canada), which carries out Monte Carlo simulations (Roff 1989; Waye et al. 1994) to calculate the statistical significance of contingency tables. Because the numbers of male patients were too small, compared with those of females, and patients and controls were not matched for gender, we have presented the data and results for all patients (males + females) throughout the tables. Results from female and male data are noted in the text where their mention seemed relevant. In all cases the allele frequencies in the control male and female allele frequencies were not significantly different, and they were consequently pooled. Rare alleles, e.g., alleles with frequencies less than 0.014 in the population, were pooled for analysis as noted in Table 2. Association analyses between controls and patients were carried out using a contingency 2×2 table to calculate an odds ratio and χ2. For multiple-comparisons procedures, P values from χ2 analysis were corrected by the method of Bonferroni adjustment (Akamizu et al. 2000; Rosner 1990). P values less than 0.05/N (where N is the number of alleles of each marker shown in Table 2) were considered significant. Pooled alleles were considered as one. For example, N was 8 in D5S2115.

Table 2. Polymorphism allele frequencies in Japanese AITD patients and controls. Frequencies in bold are significantly different in patients vs controls at P<0.05 in the method of Bonferroni adjustment; italics show that the initial frequencies are P<0.05 but become non-significant after the adjustment; n=total number alleles
Full size table

Tests for risk associated with specific alleles and interaction between AITD associated alleles

Alleles of markers that showed evidence of significant association with AITD were evaluated for the relative risk which they contributed to disease susceptibility. These evaluations were also carried out using a conventional 2×2 table to calculate an odds ratio and χ2.

Clinical manifestations of Graves' patients

Patients with GD were divided into subgroups based upon various clinical manifestations: goiter size, ophthalmopathy, and titer of anti-TSH receptor antibody. Goiter size and titer of anti-TSH receptor antibody were measured at onset or before treatment. The size of goiter was determined by measuring the transverse diameter. The patients were divided into two subgroups, large (≥5 cm) (n=119) or small (<5 cm) (n=78) goiter groups. Patients with TAO were defined by the presence of proptosis [American Thyroid Association (ATA) class 3] or higher ATA class (diplopia, corneal, and optic-nerve involvement) (Werner 1977). GD patients without and with ophthalmopathy were 111 and 88, respectively. For anti-TSH receptor antibodies, TBII (TSH binding inhibitor immunoglobulin) activity (Rees Smith et al. 1988; Endo et al. 1978) was measured, because TBII activity was most frequently measured at the onset of GD as an anti-TSH receptor antibody. GD patients with potent (≥50%) and weak (<50%) TBII were 64 and 90, respectively. Patients who responded well to anti-thyroid drugs and went into remission were classed as "excellent responders" (n=32), while those who needed anti-thyroid drugs (propylthiouracil or methimazol) after two years' treatment to maintain euthyroid function were classed as "poor responders" (n=119).

Results

Association studies

Table 1 shows chromosome localization, allele size range and heterozygosity of 5q microsatellite markers in a Japanese population. Six markers are located in 5q23-q33. D5S2011, D5S2090 and D5S2049 were not used in the previous study of Sakai et al. (2001). Table 2 shows the results of the association analysis of polymorphic alleles of these seven markers and AITD. Significant differences between the allele frequencies of cases and controls are reported on the basis of allele-specific tests of significance, using a χ2 statistic with one degree of freedom. Table 3 summarizes P values of the χ2 statistic. Frequencies of alleles of all these markers in the control Japanese males and females were not significantly different (P=0.12–0.98) and the data were pooled for analysis. We found significant allelic association between AITD and three of six markers. Of three markers, two showed significant allelic association with GD, but D5S2011 did not. For the D5S410 polymorphism, allele 2 (332 bp) was significantly decreased only in GD cases. HT and other hypothyroidisms did not show significant associations with any markers. Although several alleles of these markers showed P values less than 0.05 in association analyses (italic figures in Table 2), they were considered non-significant after correction by the method of Bonferroni adjustment for multiple-comparisons procedures.

Table 3. P values of chi square analysis in association study of Japanese AITD patients. P values in bold are significantly different in patients vs controls at P <0.05 in the method of Bonferroni adjustment.
Full size table

Risks associated with AITD susceptibility alleles

The increased risk for developing AITD in individuals with specific susceptibility alleles of six microsatellite markers was evaluated by calculating the odds ratios (i.e., risk ratios), as shown in Table 4. Allele 4 of D5S2115 showed increased risks of 1.5. Similarly, the reduced risk for developing AITD was evaluated by calculating the odds ratios. The reduced risks were 0.2–0.3.

Table 4. Risks associated with AITD susceptibility alleles
Full size table

Clinical manifestations of Graves' disease

To further examine association of GD with 5q markers, patients with GD were categorized into subgroups, based on clinical findings (Table 5). GD without ophthalmopathy and GD of "poor drug responders" were associated with most or all alleles that showed significant association with GD all. Patients with large goiter (≥5 cm) also showed similar tendency, although they associated with fewer alleles (two alleles). For the TBII activity, GD with weak TBII was associated with allele 4 of D5S410 that also showed significant association with GD of "excellent drug responders".

Table 5. Association study of Graves' subgroups with 5q markers. P values in bold are significantly different in patients vs controls at P<0.05 in the method of Bonferroni adjustment
Full size table

Discussion

We found significant association between AITD/GD and several microsatellite markers located in 5q23-33. None of these markers were not associated with HT or other hypothyroidism. Although it might be possible that analysis of HT or other hypothyroidism has less power than that of AITD or GD because HT or other hypothyroidism groups are fewer in number, this finding supports the previously published linkage analysis evidence that chromosome 5q31-33 contains AITD/GD susceptibility gene(s), but not HT (Sakai et al. 2001). Although alleles of D5S436 showed a strong linkage with AITD and GD in the previous genome-wide screen, allelic association with them became non-significant after the correction by the method of Bonferroni adjustment. These observations suggest that D5S2115 and/or D5S410 may be more likely locations of AITD and GD susceptibility genes than D5S436 in our population. Considering that genetic distance between D5S2115 and D5S410 is approximately 20 cM, it is unlikely that the two markers are both in linkage disequilibrium with a single variation of the putative susceptibility gene of AITD and GD. Instead, they might suggest that multiple independent susceptibility genes exist within 5q23-33.

Chromosome 5q23-33 contains the human cytokine gene cluster (5q31.1), containing genes for interleukin (IL)-3, IL-4, IL-5, IL-9, IL-13, granulocyte/macrophage colony-stimulating factor (GM-CSF) and interferon regulatory factor 1 (IRF-1) (van Leeuwen et al. 1989; Rioux et al. 1998; Wills-Karp 2000; Le Beau et al. 1993). Other immune-related molecules, such as β2 adrenergic receptor (5q31-q32) and CSF receptor 1 (CSF1R) (5q33-q35), are also localized to this region (Mirow et al. 1994; Mansur et al. 1998). IL-4, IL-13, IL-5 and IL-9, which are the T-helper type 2 (Th2) cytokine gene cluster, have been linked to asthma or related phenotypes in many studies (Meyers et al. 1994; Postma et al. 1995; Lonjou et al. 2000; Wills-Karp 2000). The fact that GD tended to show a stronger association with 5q markers than HT both in the previous (D5S436) and the present study (D5S2115) may reflect that GD is thought to be a Th2 disease, while HT is likely a Th1 disease (Heuer et al. 1996; Yamada et al. 2000). Indeed, the association of GD with IL-4 has been reported (Hunt et al. 2000).

When patients with GD were stratified based on clinical manifestations, patterns of allelic association with these markers were different between subgroups. This finding suggests that GD is heterogeneous and that different susceptibility genes might contribute to GD phenotypes. "GD without ophthalomopathy" and "GD of poor drug response" were both associated with the same several alleles, although the clinical significance of this association pattern is not clear. "GD with large goiter" also tended to show a similar association pattern to these subgroups. This finding may reflect the fact that GD patients with large goiter are harder to become in remission by ATD treatment than those with small goiter (Young et al. 1988; Laurberg et al. 1986). The finding that "GD with weak anti-TSH receptor antibody" and "GD of excellent drug response" associated with the same allele (allele 4 of D5S410) also corresponds to the clinical observation that patients with weak anti-TSH receptor antibody tended to easily become in remission (Young et al. 1988; Michelangeli et al. 1998). Thus, these association patterns between subgroups may help to clarify the genetic basis that contributes to forming the phenotypes of GD.

In the present study, we presented the data and results for all patients (males + females), since the numbers of male patients were too small compared with that of females, patients and controls were not matched for gender, and frequencies of alleles of markers in the control males and females were not significantly different. In fact, when we analyzed the data using female subjects alone, we found the same association for most of strongly associated alleles, though weaker in some alleles (data not shown).

Through association analysis using several polymorphic markers, we have confirmed the presence of susceptibility genes in or near the 5q31-q33 region previously identified by linkage analysis to contain a susceptibility locus for AITD by the genome-wide screening in Japanese. Since this region contains a cytokine gene cluster, it will be quite interesting to perform further analysis: (1) to narrow the susceptibility region by using 50–100 kb-spaced polymorphic markers spanning this region; (2) to find single-nucleotide polymorphism of candidate genes located in this region, which are not only associated or linked, but also functionally related to AITD.