Address for correspondence:

Simon Pearce,
Institute for Human Genetics,
International Centre for Life,
Central Parkway,
Newcastle upon Tyne,
NE1 3BZ, UK.

Tel. 441912418674
Fax. 441912418666
s.h.s.pearce@ncl.ac.uk


Introduction

Despite one third of patients with autoimmune thyroid disease (AITD) having an affected first degree relative, and a twin study showing that around three quarters of the susceptibility to AITDs can be attributed to heritable factors (1), progress in elucidating the genetic basis of these common disorders has been frustratingly slow. This article aims to review the current state of knowledge, examine the reasons why progress in genetics has been slow, and to consider future developments.

Current state of knowledge

Major Histocompatibility Complex: It has been known for more than 30 years that certain Human Leukocyte Antigen (HLA) alleles, encoded within the Major Histocompatibility Complex (MHC) on chromosome 6p21, are over-represented in AITD patients (2). The exact allele or haplotype combinations that are associated with AITD depend upon the nature of thyroid disease (i.e. Graves’ disease [GD], Hashimoto’s thyroiditis or post-partum thyroiditis) and the ethnic origin of the patient group. In white populations of European origin (hereafter “whites”) the ‘DR3’ haplotype (HLA DRB1*0301-DQB1*0201-DQA1*0501) is typically found in about 50% of individuals with GD compared to about 25 to 30% of the background population. In patients with Hashimoto’s thyroiditis, HLA associations have also been found with the HLA DR4 and DR5 haplotypes, however the effects are less consistent than those found in GD. This may be due to the smaller size of the study populations or the relatively inconsistent definition of Hashimoto’s thyroiditis or autoimmune hypothyroidism (AH) between different studies. More recently, there have been attempts to fine-map the HLA disease associations in GD patient cohorts, using a sequence-based approach (3,4). These studies have confirmed that HLA-DRB1 alleles encoding an arginine residue at position 74 are most strongly associated with GD in whites (3,4). However, despite these studies involving substantial numbers of patients, the linkage disequilibrium in this region is sufficiently strong that effects of polymorphisms in nearby genes still cannot be excluded as having a significant contribution to disease aetiology.

We make two observations from these studies: Firstly, with an odds ratio (OR) for association of the DR3 haplotype with GD of around 2 in many studies, compared to autoimmune diseases such as type 1 diabetes or rheumatoid arthritis (OR for association at HLA between 3 and 4), there appears to be a lesser contribution of HLA to disease aetiology in AITD. This implies that non-HLA disease alleles have relatively stronger effects in AITDs than in other autoimmune conditions. Given the similar genetic contributions thought to be present in these common autoimmune conditions, this means it may be more straightforward to identify these non-HLA disease alleles through genetic means by studying AITD patients. Secondly, the HLA molecules serve to process and present peptide fragments as antigens to the T cell receptor, thus determining the antigenic specificity of an immune response. As the HLA associations are relatively heterogeneous in AITDs (different in distinct populations, and only present in 50% of patients), this may mean that there is not a single T cell epitope (for instance, on the TSH receptor) that can be identified as being critical for disease onset or progression in AITDs.

Cytotoxic T-lymphocyte antigen-4:  Alleles in the 3’ region of the Cytotoxic T-lymphocyte antigen-4(CTLA4) gene on chromosome 2q33 have been extensively associated with GD, and in fewer studies, with AH. CTLA4 is involved in the regulation of the costimulatory (“second”) signal that permits T lymphocyte activation following HLA-peptide antigen encounter, and so was a good candidate gene for autoimmunity. Initial associations of CTLA4 alleles with GD by Yanagawa and colleagues (5), were confirmed with a gamut of replication studies in many different GD populations (reviewed in 6). The true disease susceptibility allele at CTLA4 remains to be defined but probably lies within a 6kb region including the 3’untranslated region (UTR) of the gene (7). The susceptibility haplotype at CTLA4 is carried by about 50% of the healthy white population and its prevalence increases to 60% in subjects with GD, with an odds ratio for the most associated allele of about 1.5 (7). The mechanism by which these non-coding polymorphisms might modulate the immune response is still far from clear. One theory is that there is a circulating, soluble isoform of the CTLA4 protein that may be able to engage and occupy the CD28/CTLA4 receptors on antigen presenting cells (known as B7 molecules) and thereby modulate costimulatory signaling. Despite significant work in this area, this theory still lacks direct experimental evidence to support it (8). Another recent study has suggested a role for CTLA4 polymorphisms at an early stage of T cell differentiation and lineage commitment, with CTLA4 genotypes being shown to correlate with the number of circulating CD4⁺, CD25⁺ T regulatory lymphocytes (9). Furthermore, it is notable that the allele associations at CTLA4, and at the adjacent inducible costimulator (ICOS) locus, appear to be subtly different in other autoimmune diseases (eg. systemic lupus erythematosus), than with AITDs (10). There may be an unsuspected level of complexity underlying the mechanistic effect of CTLA4 variant(s), with a possibility that different variants and hence aetiological mechanisms contribute to different autoimmune conditions.

Lymphoid tyrosine phosphatase: An additional locus that has recently come to light is PTPN22, which encodes the lymphoid tyrosine phosphatase (LYP). LYP is a negative regulator of T cell antigen receptor (CD3) signaling. A coding polymorphism, arginine to tryptophan at codon 620, activates the LYP molecule, paradoxically causing more potent inhibition of the T cell antigen receptor (CD3) signaling kinases, following engagement with MHC-antigen (11). The tryptophan allele is carried by about 7% of healthy subjects in northern European white populations, but is over-represented in GD subjects with a prevalence of about 13% (12,13). The odds ratio for the effect of this allele in GD is about 1.8, but because of its comparative rarity (12,13), it contributes slightly less to overall population Graves' disease susceptibility than CTLA4. Interestingly, the effects of the variant tryptophan 620 allele are very population specific, as the allele is only present at substantial frequency in white populations, being essentially absent in individuals of Asian and African descent (14). The mechanism by which the tryptophan 620 LYP variant predisposes to autoimmunity is unknown, but it is possible that it may mediate less efficient “weeding out” of T cells bearing potentially autoreactive T cell receptors in thymic development, leading to an autoimmune proclivity in later life.

Thyroid antigens: Disease specific loci for GD have started to be identified in recent years. After a period of negative investigations into the TSH receptor gene, alleles of SNP markers have now been shown to have unequivocal association with GD in 2 distinct patient cohorts (15,16). The associated polymorphism lies within the regulatory regions of the extracellular domain of the receptor, and fine mapping studies are in progress to more fully define the disease associated allele and hence the mechanism for the disease effect (15,16). In contrast, several studies have shown weaker evidence for association of GD with polymorphisms in the thyroglobulin (Tg) gene (17,18). On aggregate, these studies of the Tg gene have not shown convincing evidence for association with GD (although the effect may be different in AH) (19). Tg is a huge, 48 exon, gene and further work, to define and test the enormous diversity of haplotypes is currently awaited.

CD40 and CD25: The B lymphocyte surface immunoreceptor CD40 was identified as a candidate gene falling under a peak of linkage identified in GD families on chromosome 20 (20). A SNP marker, encoding a change in the context of the initiation of translation codon of CD40 has shown a weak association with GD in some studies (21,22) but not in others (23,24). Overall, a metanalysis of the various studies confirms that CD40 polymorphism does not have a major influence of GD susceptibility, but an effect with an odds ratio of 1.2 or less cannot entirely be excluded (25). The CD25 gene encodes the a-chain of the interleukin-2 receptor, a key player in lymphocyte activation, and markers within this gene were found to be associated with T1D, with an odds ratio of about 1.3 (26). A replication case-control study using GD probands has confirmed a modest effect, with an odds ratio of 1.24 at the most associated marker (27). Although association at CD25 needs to be confirmed in a second AITD patient cohort before it is robust, it seems likely to be a further susceptibility locus making a small contribution towards Graves’ disease pathogenesis.

Genome-wide genomics

Genome wide linkage: Genome-wide linkage scans have been completed in several patient cohorts with AITDs including Han Chinese, Japanese, the Old Order Amish and two studies using patients of mixed white origins (20, 28-32). The majority of these studies have used small “nuclear” families with 2 or more members each with AITD, most frequently a sib-pair. However, with one exception, families where one affected individual had GD and another AH have generally been included, with subanalysis of GD-only and AH-only families. Many chromosomal regions have been linked to AITD using this strategy including 1q36, 2q36, 5q11, 5q31-q33, 8q24, 11p15, 12q22, 13q32, 14p11, 17q21, 18p11, 19q13, 20q11 and Xq21 (20, 28-32). However, for relatively few of these loci has the evidence in favour of linkage (LOD score) exceeded the 3.6 threshold for declaring definite linkage in a complex trait (33), with most of these linkage peaks being in the 2.0 to 3.0 range. It is also disturbing that the MHC, CTLA4 and PTPN22 loci, where association is unequivocally established, have not by-and-large, also been found to show linkage with GD or AH in these studies. This remains an unexplained anomaly, as one would anticipate linkage to be found at these loci. Furthermore, although linkages to the long arm of chromosome 5 (5q31-q33) and 8q24 have been replicated in two patient cohorts each (29,30), the vast majority of these linkage peaks remain unreplicated and hence unconfirmed. After the first linkage studies, involving 100 to 130 AITD families (20,28-31), this failure to replicate both linkage peaks and disease associations was felt to be owing to a lack of power in the linkage analysis, probably combined with allelic heterogeneity between individuals with GD and AH. However, the most recent study used 1119 AITD relative pairs from 558 AITD families and thus power alone is unlikely to be responsible for these discrepant results (32).

Squaring the linkage circle: A combination of several factors, including power, is likely to explain the failure to replicate both previous linkages and known associations. Firstly, an initial study reporting linkage of a genetic variant(s) with disease may be a relatively extreme result and the follow-up study will generally need to be much larger to stand a chance of replicating a true linkage (34). Secondly, it is implicit that with a genome-wide design that there will be a ‘false-discovery rate’ due to multiple testing and the threshold of over 2 for reporting ‘suggestive linkage’ has certainly proven to be insufficiently stringent (33), if judged by the criteria of replication. In addition, all of these genome-wide linkage studies have been carried out using microsatellite (short tandem repeat) markers, which are genotyped by measuring their size, and most have alleles every two base-pairs, which may be difficult to separate. Even with pedigree checking of inheritance, rejection of any incomplete genotype data and duplicate genotyping of a proportion of families, some spurious results may result from genotyping errors. It is notable that almost all research labs have abandoned the use of such markers for mapping over the last 5 years. Also, with families derived from a broad ethnic background (two of the studies used white families from 4 or more different countries) (20,28,32), there can difficulty in defining the background allele frequencies from which the linkage scores are derived. In other words, there may be inflation of linkage scores when an apparently rare allele in the population is found to co-segregate with disease in a proportion of families who may be ethnically distinct. This may be a particular issue with autoimmune diseases where immune response alleles have been under recent natural selection and may have quite different frequencies even in populations that are superficially similar (e.g. PTPN22, ref 14). Lastly, there are theoretical reasons why linkage analysis might be a less sensitive method than association studies for detecting loci with relatively small genetic effects (35).

The future

Although candidate gene studies have had modest success, it remains disappointing that no novel AITD disease susceptibility gene has been robustly identified by reverse genetic approaches, despite a substantial amount of effort from several research groups over the last 10 years. The advent of SNP genotyping chips has made rapid high-fidelity genotyping of numerous markers in thousands of individuals a reality and genome-wide association studies are now underway using dense marker maps of 300,000 SNPs or more in substantial AITD cohorts. These studies may be the ultimate genomic experiment in AITD and will no doubt provide new information, which may be helpful in understanding pathogenesis, in disease prediction or in design of novel compounds to ameliorate disease. These genome-wide association analyses will have even greater issues with multiple testing than the equivalent linkage studies, and will also have practical limitations in being insensitive to rare aetiological variants, or those variants in recombination ‘hotspots’ with little surrounding linkage disequilibrium (36). Furthermore, although copy number variation (CNV) is firmly established and mapped across the genome (37), its association with disease is just starting to become apparent. It is already clear that CNV has a role in other autoimmune and infectious disorders (38-40) and it could well also be important in AITDs.