Correspondence:
Pia Skov Hansen
The Danish Twin Registry, Epidemiology, Institute of Public Health,
University of Southern Denmark, Odense
JB Winsløwsvej 9B
DK – 5000 Odense C
Denmark
Phone: + 45 65 50 33 76
fax: + 45 65 90 69 38
e-mail: piaskovhansen@dadlnet.dk


Introduction (Thyroid related phenotypes)
Despite intensive research, the mechanisms by which overt thyroid diseases develop are incompletely understood (1-3). Twin studies point toward a strong genetic influence in the aetiology of autoimmune thyroid disease (4-6) as well as simple goitre (7). Thus, in autoimmune thyroid disease, genetic susceptibility in combination with external factors (i.e. dietary iodine, cigarette smoking and infections) is believed to initiate the autoimmune response to thyroid antigens (1). Likewise, simple goitre seems to develop on the basis of genetic susceptibility interacting with environmental triggers, the two most important being inadequate iodine intake and cigarette smoking (3).
Health and disease are often defined by a continuum of biochemical and physiological measures (8). Despite distinct reference intervals, the definition of normality is not straightforward. There is a continuum from being disease-free and healthy without symptoms to having overt thyroid disease. Asymptomatic thyroid enlargement or presence of thyroid antibodies in euthyroid subjects illustrate this point. Another example is subclinical or mild thyroid dysfunction, which is defined by a condition with serum TSH levels outside the reference range and levels of free T4 and free T3 within the reference range (9). Studying, in healthy individuals, the regulation of biochemical and physiological measures related to the thyroid might be essential in understanding the pathways that eventually lead to thyroid disease. When individuals have been exposed to a disease status for a period, the physiology changes, and the original aetiological components may become invisible. Moreover, according to the endophenotypic approach, it could be useful to decompose a complex phenotype into a set of variables that might represent more basic processes (10,11). The understanding of the aetiology of overt thyroid disease may be enhanced by investigating underlying quantitative biochemical and physiological measures related to the thyroid.

Several thyroid-related measures/phenotypes exist, and different aspects of the thyroid homeostasis can be visualized and analysed. Investigations of parameters such as thyroid function, thyroid size, thyroid morphology and presence of thyroid autoantibodies are crucial in the diagnosis of thyroid disease. These phenotypes illustrate the thyroid homeostasis from different angles and each represents a small piece of a puzzle. Consequently, the serum TSH, free T4 and free T3 levels represent measures of the thyroid function set-point – reflected from different angels. The regulation of thyroid function is of special relevance, because the thyroid-specific genes may be partly involved in the genetic influence affecting most overt thyroid diseases.

Analyzing the causes of variation
Studies of the biological variation in thyroid function tests have shown that the intra-individual variation in serum T4, serum T3 and TSH levels, in healthy subjects, is narrow compared to the interindividual variation and the laboratory reference ranges (12-18). The distribution of serum T4 values in one individual is approximately half the width of the population-based reference range (15,16). This is compatible with a unique thyroid function set-point in each individual.
The sources of the interindividual differences in thyroid function are many. Ultimately, all those effects arise from genetic and environmental sources. Only a few family and twin studies have characterized the relative importance of the factors associated with thyroid function in healthy individuals. In general, previous studies suggest a low genetic influence on TSH and free T4 levels, the heritability estimates ranging from 0.32-0.44 (19-21). In the study by Samollow et al. (21), the heritability for TSH was only 0.32 – with an apparent difference between males and females. Excluding individuals with elevated serum TSH values (TSH>4.5 mIU/liter), the heritability estimate for serum TSH dropped to 0.20 and the gender difference disappeared, emphasizing the importance of an exact phenotype definition. In contrast, the heritability estimate for serum free T3 levels has been established to be higher - 0.67. Serum TSH, T3 and T4 depend on each other, and it therefore seems unlikely that the genetic influence is much different across the phenotypes that reflect the thyroid function set-point. It is important to keep in mind that heritability estimates are population specific. But actually, these studies are hampered by problems such as inadequate sample sizes (19,20,22), without specification of confidence intervals for the main results (20), crude statistical methods (20,22), and problems with phenotype- and zygosity definitions (22) making it difficult to interpret the results.
In our view, these limitations have been overcome in our recent twin studies. Based on well-defined phenotypes in a large study population of twins, we have found that intraclass correlations for serum TSH, free T4 and T3 concentrations were consistently higher for MZ than for DZ twin pairs indicating a strong genetic influence (23). Using structural equation model fitting (24,25) we have established that about 65% of the variation in serum TSH level is explained by genetic influences (23). Almost identical heritability estimates was found for serum free T4 (65%) and free T3 (64%) levels (23).

Reflections regarding the molecular basis of the estimated genetic influence
The interpretation of the above heritability estimates is not straightforward. The understanding of these results needs to be linked to the molecular knowledge of the regulation of thyroid function. The thyroid hormone homeostasis is achieved through the close interconnection of multiple, redundant mechanisms. Serum levels of TSH, free T4, and free T3 form part of a delicate feed back mechanism, and the variables are tightly interconnected. The circulating serum concentrations represent the product of balanced rates of secretion and metabolism of each of the variables along the hypothalamic-pituitary-thyroid-axis. Thus, the estimate of genetic effects includes genetic influences from many different levels and stages, and it is necessary to consider and describe the complete system to understand the estimated genetic influence.
A gene is a candidate gene if its product is known to be part of a relevant biochemical pathway to the phenotype in question. Many of the potential genes involved in the regulation of thyroid homeostasis are known from molecular studies (26-37). Moreover, the work identifying the specific genetic markers within these thyroid hormone pathway genes, associated with the circulating thyroid hormone phenotypes, has already been initiated. In the studies by Peeters et al. (38,39) distinct single nucleotide polymorphisms (SNP’s) in three of the major thyroid hormone pathway genes (deiodinase type 1, deiodinase type 2 and TSHR) have been found to be significantly associated with plasma TSH and thyroid hormone levels in healthy individuals. On the other hand, the relative role of these distinct genes and specific polymorphisms in the biochemical pathways is unknown, so far. Trying to quantify the importance of these distinct genes and specific genetic markers is essential, because it would be an important step in trying to understand how the genes actually work together. With the advent of sophisticated statistical tools and molecular genetics it is possible to decompose the genetic variance into contributions from individual genetic loci, quantitative trait loci (QTL) (40). The ability to detect a QTL is a complex function of the effect size of the QTL itself (i.e. the proportion of the total phenotypic variance attributable to the QTL), the study design, sample size, and the characteristics of the genetic data.

The control of gene expression
Genetic factors do not act independently. The current molecular knowledge clearly indicates that the genes are modulating the effect of each other. The complex networks of interacting pathways and regulatory feedback mechanisms coordinate and regulate multiple functions, and they are likely to be the norm rather than the exception. The understanding of these genetic regulatory pathways is, however, in its infancy (41,42). In addition, the genetic background acts together with the exposures of external environmental factors, which modify the regulation of thyroid function.
A variety of mechanisms are involved. The interactions between genotype and the environment may change over time and gene expression may be context dependent. Besides, it is possible that the effects of environmental exposures accumulate throughout life. But most deny, or choose to ignore these issues, when designing, carrying out, and reporting genetic studies of complex traits and diseases. It is crucial to consider genes and environmental risk factors together. Most likely, the effects of environmental factors, such as iodine intake, influence the expression of distinct polymorphisms. It is necessary to consider genetic and environmental risk factors together to obtain information about the environmental control of gene expression, and how the effects of distinct environmental factors influence the expression of specific polymorphisms. Such investigations could be done by incorporating information of measured genetic markers located in thyroid related target genes into specific models, and quantify the proportion of the total phenotypic variance in serum levels of TSH, free T4 and free T3 attributable to specific polymorphisms. Information on specific covariates (factors such as age, gender, iodine intake, and cigarette smoking) could be added into the models. The continuous rapid development of genetic epidemiological approaches is promising and has permitted more sophisticated analyses. As an example, it is possible to extend the classical twin design by incorporating gene-environment-interaction effects as well as the effects of gene-gene interactions (epistasis) (43,44). A huge challenge is epigenetic changes, such as DNA-methylation, which enable cells to respond to environmental signals without having to alter the DNA itself (45).

A multivariate framework
The studies of thyroid hormone homeostasis have focused on the genetic-environmental analysis of one phenotype at a time, a univariate (one variable) approach. Clearly this is an oversimplification. A multivariate framework serves to distinguish whether a certain phenotype has variation that is shared or common, with other phenotypes in the model, or whether variation exists that is not being shared – i.e. unique. Both common and specific variation is valuable information. It is possible to investigate the genetic overlap between different phenotypes using multivariate models (46). When genes have influence on more than one trait it is called pleiotropy (47) As an example, it would be possible to address whether a set of “common” genetic and environmental factors influence serum levels of TSH as well as free T4 and/or free T3, or whether there is a set of “specific” genetic and environmental factors that are unique to e.g. serum TSH regulation. The major advantage compared to separate univariate analyses is greater statistical power.

A hierarchical model for thyroid homeostasis
When trying to describe and understand the complexity of a biological system, it might be useful to apply a model which reflects the system. This can be illustrated by applying a hierarchical model. In such a model the genes are assigned to the basement level. Biochemical and physiological traits are located at intermediate levels and the clinical signs - the clinical endpoints - are located at the top. Studying a trait at a lower level as compared to the clinical endpoint located at a more gross phenotypic level may contain some advantages. There are likely to be fewer genes involved in intermediate phenotypes, and the effect of environmental exposure may be less important.


Figure 1
A simplified schematic diagram of the biological complexity assuming a hierachical organisation of thyroid related phenotypes. The inverted triangle on the far right side of the figure represents the likely diminishing effect of environmental factors at lower and lower levels of a biochemical and physiological hierarchy.

According to a hierarchical model the regulation of thyroid function could be regarded as a low or intermediate level phenotype (Figure 1). Several of the genes involved in regulation of thyroid function are known, and it is possible that this system is close to the effect of the genes. Nevertheless, it is important to bear in mind that the regulation of thyroid function is connected and modulated by several other thyroid phenotypes/variables such as thyroid volume, thyroid morphology, and presence of thyroid antibodies. In addition, these systems are connected to other subsystems as well. Variation in thyroid hormone levels could be intimately connected with variation in health-related characteristics such as regulation of blood glucose, growth hormone levels, lipid metabolism, the immune system, and many more. It would be useful trying to connect these phenotypes, although such connections are complicated by a plethora of possibilities.
Whether or not this theoretical thinking is true, a hierarchical model does visualize the complexity, and it makes one realize that mechanisms in one subsystem affects other subsystems. This is important to take into consideration when applying simple aetiological models to biological systems.

Conclusion
Studying the regulation of biochemical and physiological measures related to the thyroid, in healthy individuals, might be essential in understanding the pathways that eventually lead to thyroid disease. It has been established that the thyroid function set-point (reflected by serum TSH, free T4 and free T3 levels) is under tight genetic control. However, the detailed understanding of this genetic influence is still an illusive goal. Research which combines molecular and physiological studies with clinical, epidemiological and statistical knowledge is necessary in order to increase our understanding of the regulation of thyroid function.