Graves' disease (GD) is an organ-specific autoimmune thyroid disorder characterized (with rare exceptions) by hyperthyroidism, various degrees of diffuse goitre (30% lack goitre), ophthalmopathy (clinically evident in the minority) and rarely pretibial myxedema. From this, it is already clear that GD covers a wide range of phenotypes. The hyperthyroidism in GD is due to the presence of antibodies binding to and stimulating the thyrotropin receptor. The etiology of GD is multifactorial with clinical disease developing on the basis of genetic susceptibility interacting with environmental factors (1, 2, 3) (Fig. 1).

Affecting 0.5 to 1% of the population with a 5-10:1 female-to-male preponderance, GD is one of the most common autoimmune disorders. Despite a vast number of studies of the genetics of GD the reasons why Graves' disease develop in certain individuals are unknown. In fact, no single gene or genetic marker has been shown to be either necessary or sufficient for the development of clinically overt GD (3). The same is certainly true for any suggested environmental factor (4).
In the following we will briefly highlight the current perception regarding the relative importance of genetic and environmental factors in the etiology of GD.

Is there a genetic component λ

A family history of thyroid disease can be obtained in up to 50% of patients with GD and family studies have repeatedly demonstrated a familial aggregation of GD (2). By assuming that the population prevalence at large equals the cumulative lifetime risk,
it is possible to make a rough estimate of the increased risk of disease for a relative given the presence of disease in the proband. With the background population frequency of 0.4-1.1% this implies that a sister has a 5.4-12.6 times increased risk of experiencing GD when her sibling already has the disease. For a brother the increased risk ranges from 1.2-7.4 times. This value for familial clustering is named lambda (λ). Whether this familial clustering is due to shared environment or a shared genetic predisposition can only be investigated in twins. Based on the assumption that the intrapair difference in environment is the same for MZ and DZ twins the fact that concordance rates have unanimously been found higher in MZ than in DZ twins can be taken as evidence of a genetic contribution to its etiology. In our two recent population based twin studies the probandwise concordance rates were 36% and 35% in MZ pairs versus 0% and 7% in DZ pairs, respectively, (5, 6). Our findings are supported by preliminary results from a population based twin study from California (7). The fact that the concordance rates for GD among MZ twins were well below 100%, clearly suggest that environmental factors are also important.

Estimation of the genetic contribution in GD

Heritability, which is independent of the disease prevalence, is considered the best indicator of the magnitude of the genetic influence. Using structural equation modelling, decomposing individual specific and shared genetic and environmental factors, we have found that 79% (95% CI 38-90%) of the liability to the development of GD is attributable to additive genetic factors. Individual specific environmental factors not shared by the twins explained the remaining 21% (95% CI 10-37%) (6). The above suggests that search for susceptibility genes could be worthwhile but gives no information on the number of possible genes.

Candidate genes in Graves' disease

The major histocompatibility complex region
Consistent associations between GD and the class II HLA allele DR3 on chromosome 6p21 have been reported in Caucasian populations since the 1970s (2). Case-control studies have shown an increased frequency in GD of D3B1*0304, DQB1*0201, DQB1*0301/4 and DQA1*0501. Strong linkage disequilibrium exists across the HLA region, therefore, it has generally been difficult to ascertain which allele exerts an independent effect due to lack of power (too amall data sets). However, DQA1*0501 seems to be an independent HLA susceptibility locus conferring a relative risk of 2.5-3.8 (2, 3, 8).
Population based case-control studies are sensitive to the detection of susceptibility loci exerting small effects. However, they lack specificity and an enormous number of subjects need to be investigated. To detect a gene conferring a relative risk for the development of GD of 2.0 a population of around 800 cases is needed to have an 80% power to achieve a result with a significance of p < 0.001 (3). Therefore, it has been difficult to provide evidence for linkage between the HLA region and GD (3).
Case-control studies of gene polymorphisms in the HLA class III region, which contains many genes encoding immune regulator proteins (including some of the cytokines) have yielded contradictory results. But neither tumor necrosis factor-b (TNF-b) nor TNF-a seem to be major susceptibility genes in GD. Neither does transporters associated with antigen processing (2, 3 8).
With a probandwise concordance rate of 30-40% in MZ twins and a 7-10% risk of GD in HLA identical siblings with an affected proband (2) it can be calculated that HLA at best has a moderate effect (around a fourth of the total genetic effect) and suggests that other genes contribute.

CTLA-4 and other non-HLA markers
The cytotoxic T-lymphocyte-associated antigen 4 (CTLA-4), on chromosome 2q33, is a critical molecule in the activation of T-cells by antigens. CTLA-4 activation leads to T-cell suppression and the possible augmentation of immune responses and development of autoimmunity (3, 9).
Several studies on CTLA-4 polymorphisms have reported weak but positive associations with GD in various ethnic groups including Caucasians, Japanese and Hong-Kong Chinese (9). The risk, however, is weak, the relative risk being < 3 and clearly non-specific for GD since this polymorphism is also seen in autoimmune hypothyroidism, type 1 diabetes mellitus and Addison's disease (3, 9). It is currently debated whether CTLA-4 polymorphisms are specifically associated with TAO. The fact that linkage of CTLA-4 with GD has been reported, that this linkage becomes stronger when AITD families (not just GD) are included, as well as linkage of the CTLA-4 region with thyroid antibody production, suggests that CTLA-4 is an important susceptibility gene for the development of autoimmunity in general (3, 9).
A vast number of non-HLA markers have been investigated in association studies. These include the immunoglobulin heavy chain region (IgH), the T-cell receptor, the interleukin-1-receptor antagonist (IL-1-RA), the TSH-receptor gene (TSH-R), the thyroid peroxidase gene (TPO), the thyroglobulin gene (Tg), the insulin-dependent diabetes mellitus 2 gene (IDDM2), the large multifunctional proteosome genes (LMP2 and LMP7), the IL-4 promotor region, IL-1a, IL-1b, IL-4 receptor, IL-6, IL-10, transforming growth factor-b (TGF-b), estrogen receptor b gene (ER-b), vitamin D receptor gene (VDR) and autoimmune regulator 1 gene (AIRE 1). The literature reports of conflicting findings most likely due to inadequate size and poorly matched cases and controls. More important, no confirmation from family based studies (e.g. linkage) is available.

Genetic markers showing evidence of linkage
Recently, Davies and colleagues have identified new loci including GD-1, GD-2 and GD-3. The GD-1 locus on chromosome 14q31 is located within 2cM of the multinodular goitre-1-gene (MNG-1) raising the possibility that they are in fact the same, conferring susceptibility to both disorders (10). This region is interesting since it also contains other thyroid autoimmunity candidate genes, including the genes for TSH-R, IgH, T-cell receptor α, IDDM11 and ER-b. However, no linkage has been demonstrated at any of these loci.

Using the same 53 multiplex multigenerational families the same group have identified GD-2 (chromosome 20q11.2 (11). Interesting genes such as interleukin-6 nuclear factor (NF-IL6) map to this region. This gene encodes a transcription protein that binds to several regulatory regions of other cytokine genes.
Linkage has also been reported on the X-chromosome (Xq21.33-22) (GD-3) possibly explaining the female predisposition to GD (12). In recent UK studies GD-2 was linked to GD while GD-1 and GD-3 were not (13). The finding of a susceptibility locus on chromosome 18q21 awaits confirmation (14).

Is there an environmental component?

The following facts point towards the importance of environmental factors in the etiology of GD: 1) Even with more that 25 years of follow-up the probandwise concordance rates for GD are no higher than 30-40% in MZ twins (2, 6). 2) Specific candidate genes associated with GD are also present in a high proportion of healthy subjects (3). 3) There is a considerable regional variation in the prevalence of GD (15).

If so, how large is its effect?

In the only available study of heritability of GD, using two large population based twin cohorts, we have estimated that 79% of the liability to develop GD is attributable to additive genetic factors (heritability) whereas environmental factors explain the remaining 21% (6). It is, however, important to point out that none of the two components cause disease alone.

Can the environmental factors be identified?

A number of environmental factors have been associated with the development of GD. Much as is the case with genetic markers (4). Is it possible to distinguish causal from non-causal associations? Not with certainty. However, when attempting this, it is generally accepted that strength, consistency, specificity, temporality, dose-response, and biological plausibility of an association should be considered (16). In the following we will summarize our interpretation of the existing evidence for or against a causal relationship between specific environmental exposures and GD.

Iodine
Iodine is essential for the biosynthesis of thyroid hormone and influences thyroid growth and function. Perhaps the best evidence that we have for iodine affecting autoimmune thyroid disease comes from the study of animal models. Although the processes are complex, in general, iodine deficiency attenuates, while iodine excess accelerates autoimmunity in predisposed individuals (17). In humans, epidemiological surveys have over and over demonstrated that differences in prevalence and/or incidence of overt GD in different parts of the world closely mimic the magnitude of the iodine intake, with GD being more prevalent in areas with the highest iodine intake (15, 18). Furthermore, increases in iodine intake in iodine deficient regions generally increases the incidence of GD (19).
The course of GD is also affected by iodine. Thus patients given iodine supplementation after discontinuing antithyroid drug therapy (ATD) are more likely to relapse than those not given iodine. Also the response to ATD in GD is more rapid and the dose required to control the disease is smaller in iodine deficient areas than in iodine replete areas (20). Furthermore, remission rate in GD is inversely related to level of iodine intake as evidenced, at least in part, by the differences in remission rates between the US and Europe. Recurrence rate following thyroidectomy for GD is also related to level of iodine intake. The mechanisms are unclarified. However, the observations that iodine enhances the activity of lymphocytes that have been primed by thyroid-specific antigens, and the increase in TSHRab titres in relation to administration of excess iodine in GD patients are undoubtedly of importance.

Overall, the data - although the mechanisms remain to be clarified - show consistency, temporality, dose-response and the observations are biologically plausible. This suggests a causal relationship.

Smoking
It has long been recognized that smoking has a number of immunological effects involving both the humoral and cellular components of the immune response. As far as GD is concerned smoking has repeatedly been associated with an increased risk of developing GD and especially Graves' ophthalmopathy (TAO) (4). This is despite major differences in study designs, size of study populations, definitions of smokers and non-smokers, iodine intake, and methods for evaluating thyroid function and the degree of TAO. Since smoking, in a recent study (21), was - even after adjusting for stressful life events, daily hassles, social support and coping skills - still an independent risk factor for GD it is not likely that this association is due to confounding factors. It is worth noting that non-autoimmune hyperthyroidism has not been found associated with smoking (4).
Most studies including our own (4, 22) demonstrate a temporal relation with debut of GD 10 or more years after commencement of smoking. This goes for GD with or without TAO (23). A strong dose-response effect has been demonstrated with the relative risk for TAO increasing parallel to the current number of daily cigarettes smoked (23). In twins concordant for smoking but discordant for GD, the twins with GD smoked significantly more than their healthy co-twins (22) strengthening the evidence for a dose-response relationship. Whether giving up smoking is beneficial (in this context!) is still a matter of debate (23). As with iodine, smoking seems to affect treatment outcome in GD. The response to treatment of GD with or without TAO is better in non-smokers than in smokers (24).
The mechanisms by which smoking contribute to the development of GD, with or without TAO, in susceptible individuals is unclarified. However, the suggested mechanisms include: 1) A direct irritative effect of cigarette smoke. 2) Nicotine mediated stimulation of the sympathetic nervous system. 3) Alteration of the structure of the TSH receptor, making it more immunogenic and, 4) An influence on the cellular immune response, changing the profile of secreted cytokines.
From the above, it is evident that the association between smoking and GD (and especially TAO) is consistent, temporal, with a clear dose-response pattern and, finally, biologically plausible. These features strongly suggest a causal relationship.

Stressful life events
From the very first descriptions of GD stressful life events have been suggested in the etiology of GD (25). In a number of studies from e.g. Sweden, UK, Italy, Hong-Kong and Yugoslavia (25, 26), comparing cases and controls, largely similar findings of significantly higher negative life event scores in subjects with GD than in controls have been reported. Stressful life events were still strongly associated with GD in a recent Japanese study (21) where confounding factors such as daily hassles, social support, coping skills, smoking and drinking habits were adjusted for. When studied, the relative risk of GD increased as life event scores increased reflecting a dose-response relationship (21, 26). The major drawback of all the studies is that they are retrospective increasing their vulnerability especially to recall bias which to some extent is present in them all. Therefore, it is most difficult, if not impossible, to determine which came first the stress or the disease. The suggested mechanisms involve stress related stimulation of the hypothalamic-pituitary-adrenocortical axis ultimately leading to alterations in the profile of cytokine secretion.
Thus, while the association between stress and GD is consistent, strong, seemingly dose-dependent and biologically plausible, the lack of prospective data precludes firm conclusions as to the temporal sequence. Therefore, the question of causality remains unanswered.

Infections
It has long been recognized that infectious agents may induce thyroid autoimmunity by mechanisms such as inducing alterations/modifications of self-antigens, mimicking self-antigens, superantigen induced T-cell activation, and inducing expression of HLA molecules on thyroid cells, making the proposed causality biologically plausible (27)
Although Bech and colleagues found antibodies against Yersinia enterocolitica (serotype 3) significantly more often in subjects with GD than in controls (28) subsequent retrospective case-control studies have been contradictory (27). Thus, in spite of being biologically plausible the lack of consistency, specificity and temporality, in addition to lack of prospective data suggests that there is no evidence for a causal relationship between Yersinia enterocolitica infection and GD.
Using the same arguments, no convincing evidence exists as to the vast number of other infectious agents, such as e.g. influenza B virus, retroviruses, human foamy virus, coxsackie B virus, and mycoplasma species, being of etiological importance in GD (27).

Other environmental factors
Certain drugs, in particular lithium (29) and amiodarone (30) have been suggested as possible environmental risk factors for GD. Based on the previous argumentation/criteria causality is not evidenced. Certainly, the same holds for alfa and beta interferon. Most recently, pulsed monoclonal antibody treatment of patients with multiple sclerosis has given exciting insight into the possibility of monoclonal antibodies against lymphocytes causing therapeutic modulation of the immune response, permitting the generation of antibody-mediated thyroid autoimmunity, in this case GD (31).
The possibility of an adverse intrauterine environment seems highly unlikely based on the findings in our recent population-based twin control study where no effect of any birth characteristic on the risk of developing GD could be demonstrated (32).

Conclusions

It is widely accepted that Graves' disease is a complex disease as is e.g. IDDM. That is, the clinical phenotype represents the net effect of all the contributing environmental, endogenous and genetic factors. It has been difficult to separate environmental influences from genetic susceptibility. Although the HLA and CTLA-4 gene region are well established susceptibility loci the magnitude of their contributions and the possible interaction with well established environmental factors such as dietary iodine intake and cigarette smoking is at large unclarified. Future studies should focus on adequate characterisation of phenotypes and control groups, better account of environmental factors and much larger cohorts. In addition to employing genome-wide linkage analysis and allelic association analysis of candidate genes (33, 34) it is imperative to study gene-environment interactions (35).

Acknowledgments

The authors gratefully acknowledge the economic support of The Agnes and Knut Mørk Foundation, The Clinical Research Institute, University of Southern Denmark and The Novo Nordisk Foundation.

References

1. Weetman AP.
Graves' disease
N. Engl. J. Med. 343: 1236-1248, 2000.

2. Brix TH, Kyvik KO, Hegedüs L.
What is the evidence of genetic factors in the etiology of Graves' disease? A brief review.
Thyroid 8: 727-734, 1998.

3. Gough SCL.
The genetics of Graves' disease.
Endocr. Metab. Clin. N. Am. 29: 255-266, 2000.

4. Brix TH, Hegedüs L.
Genetic predisposition versus environmental factors in autoimmune thyroid disease.
In: Peter F, Wiersinga W, Hostalek U (eds.). The thyroid and environment, Stuttgart, Schattauer, p.105-119, 2000.

5. Brix TH, Christensen K, Holm NV, Harvald B, Hegedüs L.
A population based study of Graves' disease in Danish twins.
Clin. Endocrinol. 48: 397-400, 1998.

6. Brix TH, Kyvik KO, Christensen K, Hegedüs L.
Evidence for a major role of heredity in Graves' disease - a population based study of two Danish twin cohorts.
J. Clin. Endocrinol. Metab. 86: 930-934, 2001.

7. Ringold DA, Nicoloff JT, Kesler H, Davis H, Mack T.
Graves' disease, genetics and environment: revised estimates of the concordance rates in monozygotic and dizygotic twin populations. The proceedings of the 72nd Annual Meeting of the American Thyroid Association. 1999 (abstract 13).

8. Davies TF.
Autoimmune thyroid disease genes come in many styles and colors.
J. Clin. Endocrinol. Metab. 83: 3391-3393, 1998.

9. Tomer Y.
Unraveling the genetic susceptibility to autoimmune thyroid diseases: CTLA-4 takes the stage.
Thyroid 11: 167-169, 2001.

10. Tomer Y, Barbesino G, Greenberg DA, et al.
Linkage analysis of candidate genes in autoimmune thyroid disease. III. Detailed analysis of chromosome 14 localizes Graves' disease -1 (GD-1) close to multinodular goiter - 1 (MNG-1).
J. Clin. Endocrinol. Metab. 83: 4321-4327, 1998.

11. Tomer Y, Barbesino G, Greenberg DA, et al.
A new Graves' disease - susceptibility locus maps to chromosome 20q11.2.
Am. J. Hum. Genet. 63: 1749-1756, 1998.

12. Barbesino G, Tomer Y, Concepcion ES, et al.
Linkage analysis of candidate genes in autoimmune thyroid disease. II. Selected gender-related genes and the X-chromosome.
J. Clin. Endocrinol. Metab. 83: 3290-3295, 1998.

13. Imrie H, Vaidya B, Perros P, et al.
Evidence for a Graves' disease susceptibility locus at chromosome Xp11 in a United Kingdom population.
J. Clin. Endocrinol. Metab. 86: 626-630, 2001.

14. Vaidya B, Imrie H, Perros P, et al.
Evidence for a new Graves' disease susceptibility locus at chromosome 18q21.
Am. J. Hum. Genet. 66: 1710-1714, 2000.

15. Vanderpump MPJ, Tunbridge WMG.
The epidemiology of thyroid diseases. In: Braverman LE, Utiger RD, eds. Werner and Ingbar's The Thyroid: a fundamental and clinical text. Philadelphia: Lippincott-Raven 467-474, 2000.

16. Hill AB.
The environment and disease: association or causation?
Proc. R. Soc. Med. 58: 295-300, 1965.

17. Ruwhof C, Drexhage HA.
Iodine and thyroid autoimmune disease in animal models.
Thyroid 11: 427-436, 2001

18. Wang C, Crapo LM.
The epidemiology of thyroid disease and implications for screening.
Endocrinol. Metab. Clin. N. Am. 26: 189-218, 1997.

19. Stanbury JB, Ermans AE, Bourdoux P, et al.
Iodine-induced hyperthyroidism: occurrence and epidemiology.
Thyroid 8: 83-100, 1998.

20. Benker G, Vitti P, Kahaly G, et al.
Response to methimazole in Graves' disease. The European multicenter study group.
Clin. Endocrinol. 43: 257-263, 1995.

21. Yoshiuchi K, Kumano H, Nomura S, et al.
Stressful life events and smoking were associated with Graves' disease in women, but not in men.
Psychosom. Med. 60: 182-185, 1998.

22. Brix TH, Hansen PS, Kyvik KO, Hegedüs L.
Cigarette smoking and risk of clinically overt thyroid disease: a population based twin case-control study.
Arch. Intern. Med. 160: 661-666, 2000.

23. Pfeilschifter J, Ziegler R.
Smoking and endocrine ophthalmopathy: impact of smoking severity and current vs. lifetime cigarette consumption.
Clin. Endocrinol. 45: 477-481, 1996.

24. Bartalena L, Marcocci C, Tanda ML, et al.
Cigarette smoking and treatment outcomes in Graves' ophthalmopathy.
Ann. Intern. Med.. 129: 632-635, 1998.

25. Leclére JF.
Stress and autoimmune thyroid disease. In: Peter F, Wiersinga W, Hostalek U, (eds.) The thyroid and environment. Stuttgart, Schattauer, p. 155-162, 2000.

26. Winsa B, Adami H-O, Bergström R et al.
Stressful life events and Graves' disease.
Lancet 338: 1475-1479, 1991.

27. Tomer Y, Davies T.
Infection, thyroid disease, and autoimmunity.
Endocr. Rev. 14: 107-121, 1993.

28. Bech K, Nerup J, Larsen JH.
Yersinia enterocolitica infection and thyroid diseases.
Acta Endocrinol. 84: 87-92, 1977.

29. Barclay ML, Brownlie BEW, Turner JG, Wells JE.
Lithium associated thyrotoxicosis: a report of 14 cases, with statistical analysis of incidence.
Clin. Endocrinol. 40: 759-764, 1994.

30. Trip MD, Wiersinga WM, Plomp TA.
Incidence, predictability, and pathogenesis of amiodarone-induced thyrotoxicosis and hypothyroidism.
Am. J. Med. 91: 507-511, 1991.

31. Coles AJ, Wing M, Smith S, et al.
Pulsed monoclonal antibody treatment and autoimmune thyroid disease in multiple sclerosis.
Lancet 354: 1691-1695, 1999.

32 .Brix TH, Kyvik KO, Hegedüs L.
Low birth weight is not associated with clinically overt thyroid disease: a population based twin case-control study.
Clin. Endocrinol. 53: 171-176, 2000.

33. Allahabadia A, Gough SCL.
The different approaches to the genetic analysis of autoimmune thyroid disease.
J. Endocr. 163: 7-13, 1999.

34. Gambaro G, Anglani F, D'Angelo A.
Association studies of genetic polymorphisms and complex disease.
Lancet 355: 308-311, 2000.

35. Gauderman WJ, Siegmund KD.
Gene-environment interaction and affected sib pair linkage analysis.
Hum. Hered. 52: 34-46, 2001