Primary Role of the HLA Class II DRB1*0301 Allelein Graves Disease

Mahdi Zamani,1 Marijke Spaepen,1 Marie Bex,2 Roger Bouillon,2 and Jean-Jacques Cassiman1*1Center for Human Genetics, University of Leuven, Leuven, Belgium2Endocrinology Department, University of Leuven, Leuven, Belgium

The association of the Graves disease (GD)with HLA DR3 and DQA1*0501 in Cauca-sians has been described previously. Fromthese studies it could not be determinedwhether one specific locus was primarily in-volved. Using a case-control study design,we have examined the role of HLA class IIgene polymorphisms in the predispositionfor GD in a group of Belgian subjects. Wedemonstrated that both DRB1*0301 andDQA1*0501 alleles conferred significant sus-ceptibility in the DRB1*0301–DQA1*0501haplotype. The DRB1*0301 allele was theprimary susceptibility allele for GD, how-ever, because the susceptibility provided byDQA1*0501 was most likely due to it being inlinkage disequilibrium with DRB1*0301.The DRB1*0701/x and DQA1*0201/x geno-types and the DRB1*0701–DQA1*0201 haplo-type provided protection with an equal RRof 0.29. Predictive value calculationsshowed that testing for DRB1*0301 gave thehighest positive predictive value for GD infemales and males. This was, however, 10times higher in females and predicted a3.63% risk for a random female to developGD. Am. J. Med. Genet. 95:432–437, 2000.© 2000 Wiley-Liss, Inc.

KEY WORDS: Graves disease; HLA class II;risk assessment

INTRODUCTION

Graves disease (GD) is an organ-specific autoim-mune disease characterized by hyperthyroidism as aresult of the stimulation of the thyroid gland by thebinding of autoantibodies to the thyroid stimulatingimmunoglobulin (TSH) receptor. The disease is themost common cause of hyperthyroidism and, in con-trast to type I diabetes, occurs over a wide age rangewith a mean age at onset of 48 years.

HLA class II molecules that are composed of the geneproducts of three major genes DR, DQ and DP, play akey role in the immune response by binding peptideantigens and presenting them to T cell receptors. Thesegenes are highly polymorphic and as in many otherautoimmune diseases, associations of particular alleleswith GD have been described. Several studies showed apositive and a negative association of GD with the se-rologically defined HLA DR3 and DR5 groups respec-tively in Caucasian populations [Bech et al., 1977;Farid et al., 1979, Farid and Thompson, 1986; Payamiet al., 1989]. The HLA-DRB3*0101 allele was alsofound to be a susceptibility allele for GD [Boehm et al.,1992]. Recently, a significant positive association wasreported between GD and DQA1*0501, whereas no sig-nificant association was found between GD and DR3[Yanagawa et al., 1993, 1994]. Badenhoop et al. [1995]also showed that susceptibility for GD was conferred byDQA1*0501 but also by an arginine at position 52 ofthe DQA1 alleles. Protection against GD was providedby the DQB1*0602 allele. Barlow et al. [1996] reportedthat DQA1*0501 provides a greater RR than DR. Incontrast to these studies, Cuddihy and Bahn [1996],found in a case control study lack of an independentassociation between DQA1*0501 and GD and sug-gested that it was DR3 that provided susceptibility forGD. Linkage analysis revealed evidence for linkage ofGD to DR3 [Shields et al., 1994]. Roman et al. [1992],however, could not find evidence for linkage of GD toHLA antigens. Using restriction fragment length poly-morphism (RFLP) analysis in a case control study nospecific DR3-related haplotype extending into the DQregions could be shown to associate with GD [Weetmanand McCorkle, 1990]. The results of different studiestherefore are confusing and their clinical relevance re-mains doubtful.

We have studied the association of DG with HLA

Grant sponsor: Belgian Government; Grant number: Geconcer-teerde Acties; Grant number: 99/07; Grant sponsor: Interuniver-sity Network for Fundamental Research; Grant number: IUAPP4-17; Grant sponsor: Ministry of Culture and Higher Educationof Iran.

Mahdi Zamani is currently at the Department of Human Ge-netics, School of Public Health, Tehran University of MedicalSciences, Tehran, Iran.

*Correspondence to: Prof. J.J. Cassiman, M.D., Ph.D., Centerfor Human Genetics, University of Leuven, Campus Gasthuis-berg, Herestraat 49, B-3000 Leuven, Belgium.

Received 1 October 1999; Accepted 18 August 2000

American Journal of Medical Genetics 95:432–437 (2000)

© 2000 Wiley-Liss, Inc.

DRB1, DRB3, DRB4, DRB5, DQA1 and DQB1 polymor-phisms at the genomic level in a Belgian population todetermine which alleles, genotypes or haplotypes, ifany, provided susceptibility or protection against GD.In addition, we studied the clinical applicability of ourresults by calculating the positive and negative pre-dictive values, sensitivity and specificity of the tests.The results will show that both DRB1 and DQA1 al-leles contribute to the susceptibility and protectionof GD, but that DRB1*0301 presumably provides pri-mary susceptibility and that therefore testing forDRB1*0301 gives the highest positive predictive valuefor GD in females and males.

MATERIALS AND METHODSPatients

A population of 101 unrelated Caucasians (81 fe-males/20 males) with Graves disease (GD) were stud-ied. All patients were from Belgian origin and wererecruited at the Endocrinology units of the UniversityHospital of Leuven, Belgium. Clinical hyperthyroidismwas confirmed by low serum TSH concentrations withelevated free thyroxin and triiodothyronine levels.Thirty-five patients had a history of recurrent TSI-positive hyperthyroidism. In the remaining 66 patientswith a single episode of hyperthyroidism blood sam-pling was performed at diagnosis or during follow-upafter ablative thyroid therapy. Graves disease was con-sidered to be the underlying cause if the thyrotoxicosiswas accompanied by at least two of the following: en-docrine ophthalmopathy, a diffusely enlarged thyroidgland, an increased fractional thyroidal uptake ofpertechnetate with homogeneous tracer distribution orthe presence of thyrotrophin binding antibodies. Thecontrol group of unrelated Caucasians from Belgianorigin consisted of 205 blood donors.

HLA Class II Typing With SSOs

The genotypes of 101 GD patients and 205 healthycontrols were determined for HLA-DRB1, DRB3,DRB4, DRB5, DQA1 and DQB1 by PCR and sequencespecific oligonucleotides (SSOs) (Innolipa, InnogeneticsN.V.).

The polymorphic second exons of the DRB, DQA1and DQB1 genes were amplified from genomic DNA bythe polymerase chain reaction (PCR). PCR amplifica-tion was performed in a final volume of 50 ml contain-ing 50 ng of genomic DNA, 0.2 mM of each primer, 50mM KCl, 1.5 mM MgCl2 (for DQB 1 mM MgCl2), 10 mMTris-HCl (pH 8.3), 0.1% gelatine, 200 mM of eachdATP, dCTP, dGTP, 180 mM dNTP (Pharmacia LKBBiotechnology, Uppsala, Sweden), 20 mM biotin-11-dUTP (Sigma, St. Louis, MO) and 1 unit of Thermusaquaticus (Taq) DNA polymerase (Perkin-Elmer Cetus,Norwalk, CT). The amplification was performed in aDNA Thermal Cycler (Gene Amp RCR System 9600,Perkin-Elmer Cetus). The DNA was initially denaturedat 93.3°C for 3 min 20 sec. This was followed by 30cycles as follows:1) 1 min at 93.3°C; 2) 20 sec at specificannealing temperature (55°C for DQA1, 53°C for DQB1and 55°C for DRB1); and 3) 16 sec at 70.8°C. A finalextension was carried out at 70.8°C for 10 min.

The biotin-incorporated PCR products were hybrid-ized at the appropriate temperatures to membrane-bound sequence specific oligonucleotides (SSOs).Positive hybridized probes were detected by a nonra-dioactive colorimetric method.

Statistical Analysis

The level of significance of allele or haplotype fre-quencies was assessed by the Fisher’s exact test[Fisher, 1960]. P-values were corrected for multipletesting by the use of the Bonferroni’ method [Dunn,1958, 1961]. Relative risk (odds ratio) was calculatedusing the method of Woolf [1995]: (Number of patientswith the specific allele/Number of patients without thisallele)/(Number of controls with the specific allele/Number of controls without this allele). Only P-valuesand relative risks (RR) were calculated for those allelesor haplotypes that were observed more than 10 timesin the total (patient and control) population. Signifi-cant results were retested using the Mantel-Haenszeltest (MH). Moreover, logistic regressions were per-formed to detect the contribution of different alleles inthe haplotype analysis.

The predictive values (PVs) measure the effective-ness of a diagnostic test. The PVs include the positivepredictive value (PPV), that represents the likelihoodfor a person with a positive test of developing or havingthe disease, and the negative predictive value (NPV)that gives the likelihood for a person with a negativetest of not developing the disease. These methods canbe applied for calculation of predictive values of tests incohort studies but not in case control studies. We haveadapted PVs formulas for case control studies as preva-lence corrected positive and negative predictive value(PcPPV, PcNPV) [Zamani and Cassiman, 1998]; PcPPV4 a/(a + b) 4 (PDT+ × PD)/(PDT+ × PD) + [PCT+ × (1 −PD)], PcNPV 4 d/(c + d) 4 (1 − PD)(1 − PCT+)/PD(1 −PDT+) + (1 − PD)(1 − PCT+). Where PD represents theprevalence of the disease in the general population;PDT+ the proportion of the patients with the positivetest, and PCT+ the proportion of the controls with apositive test.

RESULTSDistribution of DRB Alleles in GD Patients

Versus Controls

The distribution of 30 DRB1, 3 DRB3, DRB4 and 3DRB5 alleles in patients with the GD and normal con-trols is given in Table I. Compared to the controls, thefrequency of DRB1*0301 was significantly increased inpatients (corrected P value (pc) of 0.0004 for 24.7% vs.11.5; RR 4 2.53) and MH test was MH 4 17.27, pc 40.0003. The frequency of this allele in male GD pa-tients (0.25) was slightly higher than in female pa-tients (0.23). The DRB3*0101 allele showed an in-creased frequency in GD patients compared to controlswith borderline significant pc value of 0.034. The fre-quency of DRB1*1300 was slightly lower in patients(8.7% vs. 14.5% in controls), but this difference was notsignificant after correction for multiple testing. In theDRB1 gene the frequency of allele 0701 was signifi-cantly reduced in patients compared to the controls

DRB1 in Graves Disease 433

(4.6% vs. 13.5%; pc 4 0.0047, RR 4 0.31), providing aprotective effect against GD. No significant associationwas found between DRB4 or DRB5 and GD. The fre-quency of DRB1*0401 decreased in the patients witheye disease (3.2%) compared to the patients withouteye disease (9.67%); this difference was not significant.Other clinical symptoms showed no significant associa-tion with particular alleles, possibly due to the smallsize of the subgroups.

Distribution of DQA1 Alleles in GD PatientsVersus Controls

Table II presents the distribution of DQA1 allelesin GD patients and controls. The frequency ofDQA1*0501 was significantly increased in patientswith GD compared to the controls (42.5% vs. 27.1%; pc4 0.00074; RR 4 1.99); MH 4 14.65; pc 4 0.0001. Thefrequency of this allele in male GD patients (0.50) com-pared to females (0.40) was increased but this was notsignificant. The DQA1*0401 allele was slightly in-creased in patients (4.5% vs. 1.2%), but not signifi-cantly after Bonferroni correction. Further analysis ofDQA1 alleles showed a significant decrease in the fre-quency of the DQA1*0201 allele in the patients (4.5%vs. 13.4% in controls; pc 4 0.0022, RR 4 0.30). Analy-sis of the DQA1 alleles at the amino acid level showeda significant positive association of the alleles encodingArg at position 52 on the DQa1 chain (59% vs. 40.8% incontrols; pc 4 0.00003, RR 4 2.09) and a significantnegative association of the alleles encoding amino acidsother than Arg at this position (41% vs. 59.2% in con-trols; pc 4 0.00003, RR 4 0.47).

Distribution of DQB1 Alleles in GD PatientsVersus Controls

The distribution of DQB1 allelic frequencies was notsignificantly different between the GD and controlgroups (not shown). The frequency of DQB1*0201 wasslightly higher in patients than controls (25.7% vs.20%), but this difference was not statistically signifi-cant. The frequency of DQB1*0604 was decreased inpatients compared to the controls (1% vs. 4.9%), butafter correction, this difference was not significant.

Haplotypes and Genotypes

The results of this analysis are summarized in TableIII. Both DRB1*0301/x and DQA1*0501/x genotypes

TABLE I. Distribution of the DRB Alleles in Patients andHealthy Controls*

Allele

GD(n 4 194)

Controls(n 4 408)

pc RR (CL)n Fr n Fr

DRB10101 15 0.077 40 0.0980102 0 00 7 0.0170103 3 0.015 5 0.0120301 48 0.247 47 0.115 0.00045 2.53 (1.62–3.91)a

0304 1 0.005 4 0.0090401 15 0.077 30 0.0730402 2 0.010 0 000403 1 0.005 1 0.0020404 3 0.015 9 0.0220405 1 0.005 1 0.020406 1 0.005 0 0.000408 0 00 4 0.0090701 9 0.046 55 0.135 0.0047 0.31 (0.16–0.65)0801 8 0.041 5 0.012 NS0802 1 0.005 0 000803 0 00 2 0.0040804 0 00 3 0.0070901 1 0.005 1 0.0021001 1 0.005 2 0.0041101 23 0.119 50 0.1221102 1 0.005 0 001103 3 0.015 1 0.0021104 1 0.005 0 001201 2 0.010 6 0.0141202 0 00 1 0.0021300 17 0.088 59 0.145 NS1303 0 00 2 0.0041401 7 0.036 8 0.0191500 27 0.139 55 0.1351600 3 0.015 10 0.024

DRB30101 47 0.242 64 0.156 0.034 1.72 (1.1–2.62)0200 54 0.278 87 0.213 NS0301 3 0.015 23 0.056 NS− 90 0.464 234 0.573 0.029 0.64 (0.46–0.91)

DRB40101 41 0.211 104 0.255− 153 0.789 304 0.745

DRB50101 25 0.129 50 0.1220102 1 0.005 5 0.0120200 2 0.010 10 0.024− 166 0.856 343 0.841

*pc, P-value of Fisher’s exact test with correction for multiple comparison;Fr, frequency; n, number of chromosomes; NS, not significant; −, no DRB3,4 or 5 alleles detected; RR, relative risk (odds ratio); CI, confidence limits.No homozygotes for the designated alleles in DRB3-DRB4-DRB5 werefound; null alleles are absent for DRB1. Alleles DRB1*0302, 1105, 1304,1306, 1402, 1403, 1404, 1405, 1408 and 1410 were not observed in eitherpopulation. DRB1*1501, 1502 and DRB1*1601, 1602 were grouped asDRB*1500 and DRB1*1600 respectively. The DRB1*1300 group containsthe alleles DRB1*1301, 1302 and 1305.aMH test 4 17.27, pc 4 0.00003.

TABLE II. Distribution of the DQA1 Alleles in 100 PatientsWith Graves Disease (GD) and 205 Healthy Controls*

GD(n 4 200)

Controls(n 4 410)

pc RRn Fr n Fr

Alleles0101 25 0.125 62 0.1510102 33 0.165 88 0.215 NS0103 15 0.075 38 0.093 NS0201 9 0.045 55 0.134 0.0022 0.30 (0.16–0.64)0300 24 0.120 50 0.1220401 9 0.045 5 0.0120501 85 0.425 111 0.271 0.00074 1.99 (1.40–2.28)a

0601 0 0.0 1 0.002Arg52+ 118 0.59 167 0.408 3 × 10−5 2.09 (1.48–2.94)Arg52− 82 0.41 243 0.592 3 × 10−5 0.47 (0.33–0.66)

*pc, P-value of Fisher’s exact test with correction for multiple comparison;Fr, frequency; n, number of chromosomes; NS, not significant; −, no DRB3,4 or 5 alleles detected; RR, relative risk (odds ratio); CI, confidence limits.No homozygotes for the designated alleles in DRB3-DRB4-DRB5 werefound; null alleles are absent for DRB1. Alleles DRB1*0302, 1105, 1304,1306, 1402, 1403, 1404, 1405, 1408 and 1410 were not observed in eitherpopulation. DRB1*1501, 1502 and DRB1*1601, 1602 were grouped asDRB*1500 and DRB1*1600 respectively. The DRB1*1300 group containsthe alleles DRB1*1301, 1302 and 1305.aMH test 4 14.65, pc 4 0.0001. Null alleles are assumed to be absent.

434 Zamani et al.

provided susceptibility for GD with a RR of 2.64 and2.43 respectively.

Haplotype analysis revealed that the DRB1*0301-DQA1*0501–DQB1*0201 haplotype associated withGD ( 0.42 vs. 0.215, P 4 0.00019, RR 4 2.64). WhenDQB1*0201 was removed from this haplotype, theRR remained unchanged identifying DRB1*0301–DQA1*0501 as the susceptible haplotype (pc 40.0001). The DRB1*0301–DQA1*0501 haplotype andthe DRB1*0301/x genotype had an equal RR. Indeedthe DRB1*0301–DQA1*0501 haplotype identified allindividuals who were homozygous or heterozygous forDRB1*0301 allele. On the other hand the DRB1*0301–DQA1*0501− haplotype (i.e., without DQA1*0501) wasnot found in any of the 305 cases and controls exam-ined, whereas the DRB1*0301−–DQA1*0501 haplotype(i.e., without the DRB1*0301 allele) did not show anyassociation with GD (pc for MH 4 0.9) and its fre-quency in the cases and controls was very close (0.270vs. 0.263). This indicates that the DRB1*0301 allele is

presumably the primary susceptibility allele in thedevelopment of GD. The data also showed thatthe DRB1*0701–DQA1*0201 haplotype provided pro-tection against GD. This haplotype identified all in-dividuals who carried DRB1*0701/x and also allDQA1*0201/x individuals. The frequencies ofDRB1*0701–DQA1*0201 haplotype and of theDRB1*0701/x and DQA1*0201/x genotypes were thesame (0.090 vs. 0.254, P 4 0.00039, RR 4 0.29).

Clinical Applicability of Testing for HLA ClassII Genotypes

Prevalence corrected positive and negative predic-tive values (PcPPV, PcNPV), sensitivity and specificityof testing for HLA class II genotypes were calculatedfor GD. The prevalence of GD for calculating the pre-dictive values was taken as a 1.9% for a female popu-lation and as 0.19% for males [Larsen, 1992]. PcPPV oftesting for DRB*0301/x was 0.0363 for the females and

TABLE III. Genotype and Haplotype Analysis for DRB1, DQA1, and DQB1Genes in GD and Controls*

GD(n 4 100)

Controls(n 4 205)

P RR (CL)N Fr N Fr

GenotypesDRB1*0301/x 42 0.420 44 0.215 0.00019 2.64 (1.58–4.41)a

DRB1*0701/x 9 0.090 52 0.254 0.00039 0.29 (0.15–0.63)DQA1*0201/x 9 0.090 52 0.254 0.00039 0.29 (0.15–0.63)DQA1*0501/x 69 0.690 98 0.478 0.00033 2.43 (1.46–3.97)

HaplotypesDRB1–DQA1–DQB10301+–0501+–0201+ 42 0.420 44 0.215 0.00019 2.64 (1.58–4.41)0301+–0501+–.... 42 0.420 44 0.215 0.00019 2.64 (1.58–4.41)a

0301−–0501+ 27 0.270 54 0.263 NS0301+–0501− 0 0.0 0 0.00701+–0201+ 9 0.090 52 0.254 0.00039 0.29 (0.15–0.63)0701−–0201+ 0 0.0 0 0.00701+–0201− 0 0.0 0 0.0

*The haplotypes are inferred from the genotypes assuming the absence of a haplotype if thecorresponding homozygote was not observed. P, P-value of Fisher’s exact test; Fr, frequency;n, number of individuals; NS, not significant; +, presence of allele; −, absence of allele; RR,relative risk (odds ratio); CI, confidence limits; .... , any other.aLogistic regression analysis gives an odds ratio of 2.53 for the 0301–0501 haplotype and forthe 0301 allele.

TABLE IV. Prevalence Corrected Positive and Negatives Predictive Value, Sensitivity and Specificity of Testing for DRB1 andDQA1 Alleles*

GD females(n 4 81) vs. controls (n 4 204)

GD males(n 4 20) vs. controls (n 4 204)

GenotypeorHaplotype DRB1*0301/x DQA1*0501/x DRB1–DQA1–DQB1 DRB1*0301/x DQA1*0501/x DRB1–DQA1–DQB1

0301–0501–0201 0301–0501–0201GD 34 54 34 8 15 8Control 44 98 44 44 98 44RR (CL) 2.63 2.16 2.63 2.42 3.24 2.42

(1.5–4.5) (1.3–3.6) (1.5–4.5) (1–6.10) (1.2–8.1) (1–6.10)Sen 0.4198 0.6667 0.4198 0.400 0.750 0.400Spe 0.7843 0.5196 0.7843 0.7843 0.5196 0.7843PcPPV 0.0363 0.0262 0.0363 0.0035 0.0030 0.0035PcNPV 0.9859 0.9877 0.9859 0.9985 0.9991 0.9985

*Prevalence of Graves disease (GD) in the general population was taken as 1.9% for females and 0.19% for males. Sen, sensitivity; Spe, specificity; PcPPV,prevalence corrected positive predictive value; PcNPV, prevalence corrected negative predictive value.

DRB1 in Graves Disease 435

0.0035 for the males, whereas PcPPV of testing forDQA1*0501/x was 0.0262 for the females and 0.0030for the males. A PcPPV of 0.0363 for DRB*0301/x in afemale predicts that she has 3.63% risk to develop DG.The PcNPV of testing for DRB*0301/x was 0.9859 forfemale that means that if a female does not carry aDRB*0301/x genotype she has 98.59% chance not todevelop GD. In both sexes, the sensitivity of testing forthe DRB*0301/x was lower than for the DQA1*0501/xwhereas the testing for the DRB*0301/x was more spe-cific (Table 4). PcPPV, PcNPV, sensitivity and specific-ity of testing for the DRB1*0301–DQA1*0501–DQB1*0201 haplotype were exactly the same astesting for the DRB1*0301/x genotype.

DISCUSSION

In the present study we examined the association ofthe DRB, DQA1 and DQB1 alleles genotypes and hap-lotypes at the genomic level. In agreement with thepreviously established association between GD and theserologically defined DR3 specificity, we found that theDRB1*0301 allele was the allele that provided thehighest risk of all DRB alleles (including alleles ofDRB1, DRB3, DRB4 and DRB5 genes). In agreementwith Boehm et al. [1992] we also found a weak signifi-cant association between DRB3*0101 and GD with bor-derline significance. This could be due, however, tolinkage disequilibrium between the DRB1*0301 andDRB3*0101 alleles. In contrast to Payami et al. [1989],who found a significant decrease in the frequency of theDR5 specificity we could not confirm this at the geno-mic level. Instead we found, that the frequency of theDRB*0701 allele was significantly reduced in GD pa-tients compared to controls. At the DQA1 locus,DQA1*0501 and an arginine at position 52 of the DQa1chain [DQa1Arg52+] conferred susceptibility for GD.This confirms the findings of Badenhoop et al. [1995].In addition, we found that DQA1*0201 and the absenceof an arginine at position 52 of the DQa1 chain(DQa1Arg52−) provided significant protection againstGD. Because DQA1*0501 encodes an arginine at posi-tion 52 and has almost the same RR as DQa1Arg52+, andbecause DQA1*0201 encodes an amino acid other thanarginine at position 52 and provides an even higherprotection against GD, the susceptibility and protectiveeffect of DQa1Arg52+ and of DQa1Arg52− could resultfrom the effect of the DQA1*0501 and of theDQA1*0201 alleles respectively.

In the DQB1 locus no statistically significant asso-ciation was found between this locus and GD and theprotective effect of DQB1*0602 against GD found byBadenhoop et al. [1995] was not confirmed by the pre-sent study. At this locus, the frequencies of theDQB1*0201 and 0604 alleles respectively increasedand decreased in the GD group, but these differenceswere not significant after correction for multiple test-ing.

Haplotype analysis suggested that the DRB1*0301allele conferred the primary susceptibility to GD andthat the association of DQA1*0501 could be explainedby it being in linkage disequilibrium with DRB1*0301.This contrasts with the findings of Yanagawa et al.

[1993] who suggested that it was the DQA1*0501 allelethat was the primary susceptibility allele.

DRB1*0701-DQA1*0201 was identified as the pro-tective haplotype. DRB1*0701 and DQA1*0201 are instrong linkage disequilibrium and all DRB1*0701/x in-dividuals as well as all DQA1*0201/x individuals car-ried DRB1*0701–DQA1*0201 and they had an equalRR of 0.29. It was therefore not possible to determinewhich allele had the primary role in the protection.

The positive predictive values for DRB1*0301/x(0.0363 in the females and 0.0035 in the males) furthershowed that testing for DRB1*0301 in a random femalecan predict a 3.63% risk to develop GD.

It has been reported that the frequency of DR4 islower in GD patients with eye disease compared to GDpatient without eye disease [Frecker et al., 1998]. Inour data the frequency of DRB1*0401 was lower in GDpatients who had eye disease than in those who did nothave eye disease (3.2 vs. 9.67%), but this was not sta-tistically significant that might be due to the smallnumber of cases. No other association between DRB orDQ alleles and clinical manifestations such as thyroid-specific autoantibodies or eye symptoms could be de-tected.

In conclusion, our studies have identified theDRB1*0301 allele as the major HLA class II allele as-sociated with GD. Whether these results hold true forother populations and how class II molecules carryingthese alleles contribute to the pathogenesis of the dis-ease remains to be determined.

ACKNOWLEDGMENTS

This work was supported by a grant ‘GeconcerteerdeActies’ (99/07) from the Belgian Government and bythe IUAP P4-17 Interuniversity Network for Funda-mental Research sponsored by the Belgian Govern-ment (1997–2001). We acknowledge support from theMinistry of Culture and Higher Education of Iran.

REFERENCES

Badenhoop K, Walfish PG, Rau H, Fischer S, Nicolay A, Bogner U,Schleusener H, Usadel KH. 1995. Susceptibility and resistance allelesof human leukocyte antigen (HLA) DQA1 and HLA DQB1 are shared inendocrine autoimmune disease. J Clin Endocrinol Metab 80:2112–2117.

Barlow ABT, Wheatcroft N, Watson P, Weetman AP. 1996. Association ofHLA-DQA1*0501 with Graves disease in English Caucasian men andwomen. J Clin Endocrinol 44:73–77.

Bech K, Lumholtz B, Nerup J, Thomsen M, Platz P, Ryder LP, SvejgaardA, Siersbaek-Nielsen K, Hansen JM, Larsen JH. 1977. HLA antigens inGraves disease. Acta Endocrinol 86:510–516.

Boehm BO, Schoffling K, Kuhnl P, Schifferdecker E, Schumm-Draeger PM,Usadel KH. 1992. A heterozygosity effect of HLA-DRB3 gene alleles inCaucasians with Graves disease. In: Tsuji K, Aizawa M, Sasazuki T,editors. HLA 1991: proceedings of the eleventh International Histocom-patibility Workshop and Conference, held in Yokohama, Japan, 6–13November, 1991. New York: Oxford University Press. p 713–722.

Cuddihy RM, Bahn RS. 1996. Lack of an independent association betweenthe human leukocyte antigen allele DQA1*0501 and Graves disease. JClin Endocrinol Metab 81:847–849.

Dunn OJ. 1958. Estimation of the means of dependent variables. Ann MathStat 29:1095–1111.

Dunn OJ. 1961. Multiple comparisons among means. Am J Stat Assoc56:52–64.

Farid NP, Sampson L, Noel EP, Barnard JM, Mandeville R, Larsen B,

436 Zamani et al.

Marshall WH, Carter ND. 1979. A study of human leukocyte D locusrelated antigens in Graves disease. J Clin Invest 63:108–113.

Farid NP, Thompson C. 1986. HLA and autoimmune endocrine disease1985. Mol Biol Med 3:85–97.

Fisher RA. 1960. The design of experiments. Edinburgh: Oliver & Boyd.258 p.

Frecker M, Mercer G, Skanes VM, Farid NR. 1998. Major histocompatibil-ity complex (MHC) factors predisposing to and protecting againstGraves eye disease. Autoimmunity 1:307–315.

Larsen RP. 1992. Thyroid. In: Wyngaarden JB, Smith LH, Bennett JC,editors. Cecil textbook of medicine, 19th ed. Philadelphia: W.B. Saun-ders Co. p 1248–1271.

Payami H, Joe S, Farid NR, Stenszky V, Chan SH, Yeo PP, Cheah JS,Thomson G. 1989. Relative predispositional effects (RPEs) of markeralleles with disease: HLA-DR alleles and graves disease. Am J HumGenet 45:541–546.

Roman SH, Greenberg D, Rubinstein P, Wallenstein S, Davies TF.1992.Genetic of autoimmune thyroid disease: lack of evidence for linkage toHLA within families. J Clin Endocrinol Metab 74:496–503.

Shields DC, Ratanachaiyavong S, McGregor AM, Collins A, Morton NE.1994. Combined segregation and linkage analysis of Graves diseasewith a thyroid autoantibody diathesis. Am J Hum Genet 55:540–554.

Weetman AP, McCorkle. 1990. Evidence against extended DR3-relatedhaplotypes in Graves disease. J Immunogenetics 17:403–407.

Woolf B. 1955. On estimating the relation between blood group and dis-ease. Ann Hum Genet 19:251–253.

Yanagawa T, Mangklabruks A, DeGroot LJ. 1994. Strong association be-tween HLA-DQA1*0501 and Graves disease in a male Caucasian popu-lation. J Clin Endocrinol Metab. 79:227–229.

Yanagawa T, Mangklabruks A, Chang YB, Okamoto Y, Fisfalen ME, Cur-ran PG, DeGroot LJ. 1993. Human histocompatibility leukocyte anti-gen DQA1*0501 allele associated with genetic susceptibility to Gravesdisease in a Caucasian population. J Clin Endocrinol Metab. 79:1569–1574.

Zamani M, Cassiman JJ. 1998. Reevaluation of the importance of polymor-phic HLA class II alleles and amino acids in the susceptibility of indi-viduals to type I diabetes in different populations. Am J Med Genet76:183–194.

DRB1 in Graves Disease 437