Genetics of Thyroid Function and Disease

Abstract

Genetics play a prominent role in both determination of thyroid hormone and thyrotropin (TSH) concentrations, and susceptibility to autoimmune thyroid disease. Heritability studies have suggested that up to 67% of circulating thyroid hormone and TSH concentrations are genetically determined, suggesting a genetic basis for narrow intra-individual variation in levels, perhaps a genetic ‘set point’. The search for the genes responsible has revealed several candidates, including the genes for phosphodiesterase 8B, iodothyronine deiodinase 1, F-actin-capping protein subunit beta and the TSH receptor; however, each of these only contributes a small amount to the variability of hormone concentrations, suggesting that further genes and mechanisms of genetic influence are yet to be discovered. Some genes known to influence thyroid function, including iodothyronine deiodinase 2 and the TSH receptor, have been shown to influence a wide range of clinical and developmental phenotypes from bone health to neurological development and longevity; such observations will help us understand the complex action of thyroid hormones on individual tissues. Finally, autoimmune thyroid disease commonly runs in families, and the search for genes which increase susceptibility has identified several good candidates, particularly those involved in immune regulation and thyroid function. However, these genes alone account for only a small percentage of the current prevalence of these disorders. Although the advancement of genetic technology has led to many significant findings in the last decade or two, it is clear that we are only just beginning to understand the role of genetics in thyroid function and disease.

Introduction

Thyroid hormones play a vital role in normal human physiology with effects on almost all tissues to influence growth and development, maintain normal cognition, cardiovascular function, bone health, metabolism and energy balance. In recent times we have come to understand the important influence that genetics play in normal and abnormal thyroid function. This has led to a greater knowledge of the intricacies of thyroid hormone action, differences between individuals and resultant disease. Whilst the work in this exciting, rapidly expanding field is far from complete, this review aims to provide a summary of discoveries thus far in the genes responsible for normal thyroid physiology, the effect of common genetic variation on clinical phenotypes and the genetic basis of autoimmune thyroid disease. The genetics of thyroid cancer is not covered in this review as it has been recently addressed in this journal.¹

Heritability of Thyroid Hormones

It has been recognised for some time that circulating TSH, free thyroxine (free T4) and free tri-iodothyronine (free T3) concentrations in euthyroid individuals have a much greater inter-individual than intra-individual variation. Andersen et al. showed that the width of the individual 95% confidence interval for all three variables was approximately half that of the entire group.² As a result, although the population reference ranges for these parameters are wide, each individual appears to have their own set point within this. This has significant implications given that small changes in thyroid function, even within the population reference range, have been shown to have clinically detectable effects on phenotypes as varied as cholesterol,³ mood⁴ and longevity.⁵ Therefore at what point an individual started within the range is very important when one is trying to determine if an alteration in thyroid function has resulted in a clinical problem.

Several studies have attempted to estimate the genetic (vs environmental) contribution to this individual set point using either twin- or family-based study designs. Twin-based studies compare difference in variation of traits between pairs of monozygotic and dizygotic twins; those traits which have a strong genetic component should have greater correlation in the monozygotic twins. Family-based studies use a more complicated calculation based on information from larger family groups and expected inheritance. Studies on the heritability of thyroid function have provided wide ranging estimates (Table 1).^6–10 This is most likely due to differences in size and ethnicity of study populations and difference in study design. However, the larger studies using widely accepted statistical methods would suggest a strong genetic component, particularly in TSH set point (approximately 65%), but perhaps less so for free T4 and free T3 (both around 40–50%). These findings suggest that individual thyroid function set points are mainly genetically derived, however, the genes responsible have until recently not been known.

Table 1.

Heritability of TSH, free T4 and free T3^*.

Study	Number	TSH	free T4	free T3
Panicker et al.6	2124	65% (58–71)	39% (20–55)	23% (3–41)
Hansen et al.7	1380	64% (57–70)	65% (58–71)	64% (57–70)
Samollow et al.8	1011	32%	37%	67%
Martin et al.9	378	ns	32% (10–53)*	ns
Meikle et al.10	60		44%

^*Total T4 not studied. Heritability as percentage of variation explained; 95% confidence interval (where available) in brackets.

Common Genetic Variation

Even unrelated human subjects share about 99.9% of their genome. It has been estimated that 90% of the remaining variation is accounted for by approximately 10 million common single nucleotide polymorphisms (SNPs), single base changes spread throughout the genome. These are very useful in studying gene-phenotype associations as they occur commonly in the general population, and may either cause changes in gene function themselves, or more frequently are markers of nearby elements that do. Due to publicly available databases such as that generated through the human genome project and the International Haplotype Mapping project (HapMap),¹¹ a considerable amount of information on the location, functionality and inheritance of these SNPs is freely available. Advancements in genetic technology have enabled genotyping to be performed rapidly and cheaply on large numbers of subjects, further enhancing their usefulness. Methods used to identify associations between genes and thyroid phenotypes include candidate gene studies, genome-wide linkage studies, genome-wide association studies (GWAS) and whole genome sequencing.

In candidate gene studies, genes which are biologically suspected to play a role in thyroid hormone physiology are selected and studied by fine mapping of the gene.

For example, by understanding the clinical picture of autoimmune thyroid disease, one may suspect that the thyroglobulin and TSH receptor genes may play an important role and that they are therefore candidates worthy of closer study. Study of candidate genes generally requires fewer subjects to be studied and less genotyping than the other approaches, and is therefore likely to lead to fewer false positives. Another approach, genome-wide linkage analysis uses markers evenly spread across the whole genome. The closer two markers on the genome are to each other, the higher the likelihood that they will be inherited together, as there is less chance recombination events will occur between them. Therefore a marker close to a gene causing susceptibility to a disease will tend to be inherited with that disease; it will have a higher linkage score. When these scores are plotted across the genome (as logarithm of odds or LOD scores) the areas of high scores will suggest where the susceptibility gene is. The benefit of this technique is that no assumption is made concerning which genes are studied (all are studied); therefore, novel gene associations can be found. GWAS takes genome scanning further, whereby hundreds of thousands to millions of SNPs chosen to be evenly spread across the genome are genotyped in a single experiment in each individual using a pre-built chip. This enables much finer coverage of the genome than linkage analysis. Due to significant multiple testing issues and automation required of the genotyping, quality control levels are necessarily set high and significant P values for associations very low (usually around 5 x 10⁻⁷).¹² This method has been shown to be effective in discovering novel susceptibility genes which only result in small increases in risk. The drawbacks are large sample size required (including large numbers of subjects with the disease being studied), cost and high false positive rate. Whole genome sequencing goes one step further in analysing the whole of the genome rather then markers along it. This should give us a better understanding of the relationship between genes associated with diseases, as currently we mainly identify non-functional markers which are linked to disease alleles. However, cost issues and amounts of data are further multiplied by this technique, and it has not been used convincingly in this area as yet.

In order to understand the basis of many candidate gene studies and how genes related to thyroid hormone action can alter clinical phenotypes it is important to review the process by which thyroid hormones produce effects in many cells.

Thyroid Hormone Action

The thyroid hormones T4 and T3 are released into the circulation by the thyroid gland under stimulation of TSH from the anterior pituitary gland. T3 is the active hormone which can bind to thyroid hormone receptors in target cell nuclei. T4 must be de-iodinated to T3 to have nuclear effects, however, it is secreted in much larger amounts (thought to be approximately 14 times that of T3). Both thyroid hormones must be transported across lipid membranes into cells; this action is performed by thyroid hormone transporters. These are either specific for thyroid hormones or transport other peptides as well, and may be specific to certain tissues or found throughout the body. Examples of specific thyroid hormone transporters include the monocarboxylate 8 (MCT8) transporter¹³ and the organic-anion transporting polypeptide 1C1 (OATP1C1).¹⁴ Once inside cells, thyroid hormones can be deiodinated by the three iodothyronine deiodinases: D1, D2 and D3; their actions are shown in Figure 1. These either activate the hormones by changing T4 into T3 (D1 and D2), or effectively inactivate them by turning T4 into reverse T3 (rT3) and T3 into T2, with both unable to produce effects (D3). The deiodinases vary in their presence and activity within different tissues, and also in a temporal manner during development, allowing them to control thyroid hormone delivery to specific tissues. T3 then moves into the cell nucleus where it binds to thyroid hormone receptors (TRs), resulting in a change in formation and binding of the receptor which binds to DNA and causes transcription of thyroid responsive genes. The TR often binds as a heterodimer with retinoid X receptor (RXR) and its action is also influenced by co-regulator proteins which can bind once T3 is bound to the receptor. These again vary between tissues, as does the TR, of which there are two main types and several isoforms. A simplified version of what is actually a very complex pathway, varying significantly between tissues, is displayed in Figure 2.

Figure 1.

Action of the iodothyronine deiodinases. T4, thyroxine; T3, tri-iodothyronine; rT3, reverse T3; T2, di-iodothyronine; D1/D2/D3, iodothyronine deiodinases.

Figure 2.

Pathway of thyroid hormone action.TSH, thyroid stimulating hormone; T4, thyroxine; T3, tri-iodothyronine; rT3, reverse T3; T2, di-iodothyronine; D1/D2/D3, iodothyronine deiodinases; TR, thyroid hormone receptor; RXR, retinoid X receptor.

Genetics of Thyroid Function

Whilst it is clear from heritability studies that a significant proportion of TSH, free T4 and free T3 variation is genetically derived, the genes responsible for this remain largely undetermined, as discussed below. Thus far, polymorphisms within three genes have been shown to be associated with thyroid function in healthy subjects at genome-wide levels of significance: phosphodiesterase 8B (PDE8B), iodothyronine deiodinase 1 (DIO1) and F-actin-capping protein subunit beta (CAPZB). Furthermore, a polymorphism in the TSH receptor gene (TSHR), whilst not associated with TSH at genome wide significance levels, has been shown to have associations with thyroid function in multiple studies in different populations. Many other possible candidates have not been replicated, whilst many genes which we would expect to influence thyroid function, such as the thyroid hormone receptors (THRA and THRB) and MCT8, have not shown associations. This may be because the genes have not been sufficiently finely mapped, studies have not had enough power, there may not be functional polymorphisms within these genes or these polymorphisms may not be compatible with life.

Phosphodiesterase 8B

PDE8B found on chromosome 5 encodes a protein which catalyses the hydrolysis and inactivation of cyclic AMP (cAMP). Arnaud et al. performed a GWAS and discovered an A>G SNP (rs4704397) within this gene to be associated with circulating TSH concentrations, each copy of the rarer A allele conferring a mean increase of 0.13 mU/L TSH,¹⁵ translating to an increase of 0.26 mU/L for A homozygotes. The authors estimate this polymorphism is responsible for approximately 2.3% of TSH variation in their population. Whilst the original study did not contain free T4 or free T3 levels, three further studies have not shown any association of this SNP with free T4 or T3.^16–18 As phosphodiesterase 8B is found in the thyroid but not the pituitary, and given the importance of cAMP activity in TSH signalling, it is felt that the polymorphism may reduce cAMP activity in the thyroid, resulting in a decreased response of the thyroid gland to TSH stimulation, which leads to an increase of TSH set point for the same free T4 and free T3 levels (unconfirmed).

F-Actin-Capping Protein Subunit Beta (CAPZB)

Upstream of CAPZB another SNP, rs10917469, was discovered by GWAS to be associated with circulating TSH concentrations in healthy individuals.¹⁷ Found on chromosome 1, each of the rarer G allele is associated with lower mean TSH of approximately 0.16 mU/L and similar to PDE8B is not associated with free T4 or free T3 levels, suggesting it alters pituitary-thyroid set points. It is estimated this SNP is responsible for about 1.3% of the total variation of TSH. The mechanism by which it affects TSH is unknown. The F-actincapping protein subunit beta binds to the fast growing end of the actin filament, blocking the exchange of subunits and regulating growth.

TSH Receptor

Several smaller studies have independently found associations between TSHR SNPs and TSH concentrations, and some have shown associations with clinical phenotypes (see below), suggesting this is a real association. However, GWAS have not shown SNPs in TSHR to be associated with TSH concentrations at a high level of significance,^15,17 which raises doubts as to the validity of these associations. The most commonly studied SNP is rs1991517 (also known as Asp727Glu) which has been shown in three independent studies involving greater than 2600 individuals in total, to be associated with TSH.^19–21 The wild type AA genotype has a mean TSH between 0.13–0.30 mU/L higher than the rarer genotypes combined. One of the studies estimated overall contribution to TSH variation to be only 0.91%.²⁰ This SNP is not associated with free T4 or free T3 levels and is thought to influence pituitary-thyroid axis set points by changing the sensitivity of the TSH receptor to TSH.

As displayed in Table 2, together these three SNPs explain at most 4.5% of the variation in TSH concentrations.

Table 2.

Genes associated with baseline circulating TSH, free T4 or free T3 levels in individuals with no thyroid dysfunction.^*

Gene	SNP	Association	Effect size (95% CI)	% Variation
PDE8B	rs4704397	↑ TSH15	0.13 (0.07–0.19) mU/L	2.3
CAPZB	rs10917469	↓TSH17	0.16 (0.10–0.22) mU/L	1.3
TSHR	rs1991517	↓TSH19–21	0.13 to 0.30 mU/L	0.9
DIO1	rs2235544	↑free T3/T422	0.015 (0.009–0.021)
		↓free T422	0.77 (0.45–1.09) pmol/L	2.0

^*Association refers to the direction the rarer allele alters the named phenotype. % variation is the percentage of variation within the phenotype explained by the polymorphism.

Iodothyronine Deiodinase 1

The three iodothyronine deiodinases play an important role in thyroid hormone action and are likely to influence serum and local tissue concentrations of T4 and T3; their action is summarised in Figure 1. A candidate gene study of the genes using HapMap to identify SNPs representing most of the variation across genes discovered a SNP, rs2235544, in the gene coding for D1 (DIO1) associated with circulating free T3/free T4 ratio, free T4 and rT3 concentrations.²² The rarer C allele was associated with an increase in free T3/T4 ratio (approximately 0.01), a decrease in free T4 (0.6 pmol/L) and rT3 (0.2 pmol/L) levels, and a trend towards an increase in free T3 levels (0.5 pmol/L). This pattern suggests this allele confers improved function of D1, as Figure 1 shows that increased action would result in increased conversion of free T4 to T3 and rT3 to T2, resulting in lower T4 and rT3 and higher T3. The SNP has no association with TSH concentrations, most likely because it alters free T3 and free T4 levels in opposite directions.

Previous smaller studies had identified another SNP in DIO1 (rs11206244, also known as D1a-C/T) which was also associated with free T4, rT3 and free T3 concentrations.^19,23 These two SNPs are in strong linkage disequilibrium (commonly inherited together) with r² of 0.41; conditional analysis revealed the association was with rs2235544.²² This SNP occurs in intron 3 of DIO1 and functional studies will be required to determine how it affects function of D1. In this study none of the SNPs in DIO2 or DIO3 were found to be associated with thyroid hormone or TSH concentrations.²² This usually indicates that the SNPs studied did not influence gene function; however, one of the SNPs in DIO2, rs225014, appears to be functional in terms of clinical outcomes (see below). Another possibility is that minor alterations in the function of these genes do not affect circulating levels of thyroid hormones (perhaps due to compensatory mechanisms such as the action of D1), however, alterations in concentrations at a tissue level may still have clinical effects.

Genetic Variation in Genes Influencing Thyroid Hormone Action and its Effect on Clinical Phenotypes

Whilst only a few genes have been found to influence thyroid function, it has become clear that genes involved in thyroid hormone action can have clinically detectable effects with no effect on circulating thyroid hormone concentrations. As can be seen in Figure 2, there are many elements which can affect the final binding of T3 to the TR in the cell nucleus, and therefore circulating (measureable) concentrations of thyroid hormones may be a poor reflection of individual tissue levels. This may be particularly pertinent in tissues such as the brain in which there are mechanisms in place to protect the local tissue from swings in circulating levels. In this section, genes which influence clinical phenotypes are discussed. These are more difficult to study, as SNPs generally result in small functional changes which may be difficult to detect using inaccurate clinical measures and small sample sizes. Therefore associations have only been shown in a few, well-studied genes. These are summarised in Table 3.

Table 3.

Summary of thyroid related genes associated with clinical phenotypes.

Gene	SNP	Phenotype
DIO2	rs225014/rs12885300	Osteoarthritis
	rs225010/rs225012	Mental retardation in iodine deficient areas
	rs225014	Psychological well-being on T4
	rs225014	Preference for T3/T4 over T4 therapy
	rs225014/rs12885300	Bipolar affective disorder
	rs225014	Hypertension/blood pressure
TSHR	rs1991517	Bone density
	rs1991517	Insulin resistance
	rs10149689/rs12050077	Longevity
OATP1C1	rs10770704	Psychological well-being on T4
DIO1	rs11206244/rs12095080	IGF-1

^*Phenotypes in italics have conflicting evidence.

DIO2 SNP Associations

One of the best examples of this principle is the association between the DIO2 SNP rs225014 and psychological well-being in subjects on thyroxine. It has been noted above that this SNP does not affect circulating hormone concentrations, however Panicker et al. found an association between rs225014 and psychological well-being as measured by general health questionnaire and hospital anxiety and depression scale (HADS) depression scores, the rarer C allele having an odds ratio (OR) of 1.49 for being a HADS depression case (P=0.01).²⁴ This finding has been replicated in a separate cohort from the HUNT 2 study, with the rare CC genotype having an OR of 1.56 for being a HADS anxiety case (P=0.03) (unpublished observations), and a trend towards an association (P=0.11) between rs225014 and impaired well-being has been observed in a third, smaller study.²⁵ In the initial study, Panicker et al. went on to show that the psychological deficits caused by this SNP in subjects on thyroxine replacement were to a certain extent reversible by changing them onto combination T4/T3 therapy.²⁴ Given that D2 is the only activating deiodinase in the human brain, this suggests that subjects with the functionally impaired rare genotype have resultant mild local hypothyroidism in the brain, resulting in impaired psychological well-being and furthermore that this can be overcome by putting them on combination replacement therapy. Whilst replication is required before this finding can be translated into clinical practice, other studies have shown an association between DIO2 SNPs and mental retardation in iodine deficient areas of China²⁶ and DIO2 and bipolar affective disorder,²⁷ adding to the evidence for an effect of this SNP on brain function.

There is also evidence for an association between DIO2 SNPs and osteoarthritis. This comes from a genome wide linkage scan by Meulenbelt et al. examining SNPs associated with generalised osteoarthritis which found an association with rs225014.²⁸ This SNP lies in the coding region of DIO2, at the site important for ubiquitin binding and therefore regulation of D2 turnover,²⁹ which is a possible mechanism by which it may affect D2 function. Replication of this association with osteoarthritis in several other cohorts revealed an association between a common haplotype of the minor allele of rs225014 and the common allele of rs12885300 (also referred to as D2-ORFa-Gly3Asp), also on DIO2. This haplotype is found with an OR 1.79 in females with hip osteoarthritis compared to normal controls.²⁸ It is known that D2 plays a role in the developing growth plate controlling local T3 levels, which is vital for cartilage development,³⁰ and the authors proposed that reduced D2 function in those with the haplotype may have resulted in poorer cartilage development and increased risk of osteoarthritis.

Other Associations

The following additional associations have also been observed, but require further validation or replication before firm conclusions can be made. The TSHR SNP rs1991517 was reported to be associated with bone density (above that which would be expected of its effect on circulating TSH)²¹ and with insulin resistance³¹ (this may relate to the presence of the TSH receptor on adipose cells).³² Two other SNPs in TSHR, rs10149689 and rs12050077, were associated with longevity and present in higher frequency in Ashkenazi Jew centenarians and their offspring than controls,³³ adding to the argument that rising TSH with age is protective.^5,34,35 DIO1 SNPs are associated with IGF-1 and clinical markers of growth hormone sufficiency;³⁶ this may relate to the high liver activity of D1.

There is conflicting evidence for an association with rs10770704 in OATP1C1 and increased fatigue and depression in subjects on thyroxine.³⁷ This was not replicated by the author’s work in a larger cohort on T4 (unpublished data). In addition, studies on the DIO2 SNP rs225014 and blood pressure are also conflicting, either showing³⁸ or not showing an association with blood pressure.^39,40 have shown³⁸ and not shown associations.^39,40 Finally, although DIO2 SNPs were initially thought to be associated with type 2 diabetes mellitus and insulin resistance,^31,39,41 subsequent larger studies (combined number subjects >10,000) have shown no association.^42–44

Genetics of Autoimmune Thyroid Disease

Autoimmune thyroid disease (ATD) is one of the commonest endocrine and autoimmune diseases. Data from the Tayside area of the UK suggests the prevalence of hyperthyroidism in 2001 in adult females was 1.32% and in males 0.25%, whilst hypothyroidism was present in 5.46% and 0.95%, and furthermore, this was steadily increasing.⁴⁵ Most of this is secondary to ATD, namely Graves’ disease (GD) and Hashimoto’s thyroiditis (HT). ATD in the clinical setting has always been considered familial; a quarter to a third of subjects with GD have a first degree relative with GD or HT,⁴⁶ and twin studies suggest development of GD is mostly (79%) due to genetic factors.⁴⁷ Therefore, much work has been undertaken to determine the genetic basis for this condition, including candidate gene studies and GWAS. Initial findings were disappointing, and whilst some susceptibility genes have now been discovered, they have not fulfilled the promise of such a clearly familial condition.

Although GD and HT are clinically very different, it appears they share a common initial aetiology whereby thyroid reactive T cells escape tolerance and infiltrate the thyroid, therefore it should not surprising that they have some shared genetic susceptibility. In GD however, lymphocytic infiltration goes on to activation of TSH receptor B cells which release stimulating antibodies resulting in hyperthyroidism, whilst in HT the lymphocytic infiltration results in apoptosis of thyroid cells and hypothyroidism. Therefore on this basis they can also be expected to have specific susceptibility genes.

Many of the earlier candidate gene studies focused on the genes involved in regulation of the immune system, and indeed most of the current known susceptibility genes come from this area. These include genes involved in the interaction between antigen presenting cells and T cells; this interaction is responsible for the activation of the T cells required for the immune processes which result in autoimmune thyroid disease. These include the human leukocyte antigen (HLA) alleles in the major histocompatibility complex (MHC), lymphoid tyrosine phosphatase (PTPN22), cytotoxic T-lymphocyte antigen-4 (CTLA4), FC receptor-like genes (FCRLs), interferon-induced helicase-1 (IFIH1) and B lymphocyte surface immunoreceptor CD40. There are also regulatory T cell genes involved, including the genes for the α-chain of the interleukin-2 receptor (CD25) and FOXP3, as well as thyroid-specific genes including thyroglobulin (Tg) and the TSHR. Genes currently thought to increase susceptibility to GD and HT are listed in Table 4 along with the increased risk they confer. Only a few convincing associations will be discussed here to show how they affect susceptibility.

Table 4.

Susceptibility genes for autoimmune thyroid disease.^*

Gene	OR for Graves’ disease*	OR for Hashimoto’s*
MHC - DR3 haplotype	2.0	1.8
MHC - HLA-C*07	1.6
PTPN22	1.7	1.8
CTLA4	2.0	1.8
FCRL	1.2
IFIH1	1.1
CD40	1.1
CD25	1.2
TSHR	1.3

^*Odds ratios (OR) are estimation of increased risk of disease per susceptibility allele. Where several studies have provided different results, the OR are estimations only.

The MHC located on chromosome 6p21 is highly polymorphic and encodes the HLA proteins. As with type 1 diabetes, certain HLA alleles are more common in subjects with autoimmune thyroid disease, although the risk conferred appears lower in thyroid disease than in type 1 diabetes and is ethnically variable. The strongest risk for GD in white European derived populations is the ‘DR3’ haplotype (HLA DRB1*0301-DQB1*0201-DQA1*0501) which is found in about half of all subjects with GD but only 25–30% of controls; it confers an OR of approximately 2.0.^48,49 Sequencing studies have shown that an arginine at position 74 of HLA-DR beta chain changes the pocket in which the peptide is bound and confers extra risk.^50,51 The HLA-C region also makes an independent contribution to risk of GD.⁵² For HD the results have not been consistent, with several DR haplotypes showing mildly increased risk compared to controls.^53–57 The HLA markers process and present peptides to the T cell receptor for determination of the immune response. The variation between populations and strengths of markers suggests that there may be multiple antigens which may cause the initial event.

Lymphoid tyrosine phosphatase is a negative regulator of T cell antigen receptor (CD3) signalling, and a SNP in its gene, PTPN22, at codon 620 increases its activity resulting in more inhibition of CD3 signalling after engaging an MHC antigen.⁵⁸ This allele confers an increased risk (OR 1.7) for GD, however, its contribution to overall population susceptibility is quite low.⁵⁹ It is also found with increased frequency in HT with OR 1.8.⁶⁰

CTLA4 is another cell surface receptor involved in regulation of T cell activation. Several studies in different GD populations have shown an association with several CTLA4 alleles.^61–66 The strongest OR is approximately 2.0 and CTLA4 susceptibility haplotypes are carried by about 60% of subjects with GD.^63,66 CTLA4 also appears to confer susceptibility to HT (OR 1.8) and elevation of thyroid antibodies with no thyroid disease.⁶⁵

As shown in Table 4 the overall contribution of these genes to GD and HT susceptibility remains low, which is disappointing when one considers the amount of work and resources put into the area. As with other conditions, these findings have led to the suggestion that to develop autoimmune thyroid disease one must have several of the susceptibility alleles, each contributing a small increased risk, however, this is not feasible given the frequency of these alleles in the population (and therefore low likelihood of combinations) against the prevalence of disease. Therefore, other mechanisms have been proposed, including gene-gene interaction and genetic heterogeneity (whereby some genes have a much higher risk in certain populations). The lack of novel susceptibility genes found from GWAS may be due to lack of numbers, particularly cases.

Summary and Future Directions

Improved technology in genetic investigations have already provided us with a wealth of information on the genetic basis for ‘normal’ thyroid function, autoimmune thyroid disease and the influence of thyroid genes on clinical phenotypes. Early optimistic expectations that all genes responsible would be rapidly discovered have had to be reined in, and researchers across many fields are searching for the reasons this has not happened. The need for larger sample sizes and collaboration between groups with access to large cohorts has now been understood, and these studies will undoubtedly discover further genes. Furthermore, whole genome sequencing may provide more information as other types of genetic variation, such as copy number variants may also be found to play a role. In addition, influences on accessibility of genes to transcription may be at work. However, what we have already discovered has increased our understanding of normal thyroid hormone action and physiology, and we are beginning to understand the complex origins of autoimmune thyroid disease. Important for future work is the need to replicate the early findings presented above and perform functional studies to identify the true associations and the mechanisms behind them. These mechanisms will increase our understanding of thyroid physiology and identify therapeutic targets.