The role of the thyroid gland in the activity of various organs and tissues within the body is controlled at several different levels: (1) the hypothalamus which exerts control over (2) the thyrotropes in the pituitary which in turn secrete a thyrotropin that stimulates the activity of (3) the thyroid gland. The thyroid gland synthesises, stores and releases (4) thyroid hormones, which are carried in the blood bound to carrier proteins, such as thyroxine-binding globulins, transthyretins or albumin. In the tissues, the thyroid hormones are transferred to cells by (5) membrane-bound thyroid hormone transporters where they are processed by (6) deiodinases before either being excreted from the cell or binding to specific (7) receptors in the nucleus, thus affecting the activity of that cell. This review will concentrate on the role of the deiodinases (6) in this cascade. In order to do so it will first briefly consider the features of thyroid hormones.

Thyroid hormones

Thyroid hormones are iodinated thyronines, synthesised and released from the thyroid gland of vertebrates, where they are picked up by the circulatory system and conveyed therein to all the tissues of the body. The thyroid gland primarily releases the fully iodinated form, thyroxine, commonly referred to as T4. Other less-iodinated thyronines, found in the circulatory system are generally the result of deiodination in the peripheral tissues. Thyronines have two benzene rings, both of which can incorporate two iodines. The removal of one iodine from the outer benzene ring is referred to as ‘outer ring deiodination’ (ORD) and similarly, removal of one iodine from the inner ring is referred to as ‘inner ring deiodination’ (IRD). ORD of T4 results in the formation of T3, 3,5,3′-triiodothyronine, which is a much more active form of thyroid hormone than T4, whereas IRD forms reverse T3 (rT3), which has no T4/T3 activity. Further inactivation to T2 is carried out by the same enzymes (Fig. 1). T2 in any of its forms (3,5-T2; 3′,5′-T2; 3,3′-T2) is the clearance form of thyronine (reviewed by Hulbert 2000).

Fig. 1
figure 1

A schematic of outer and inner ring deiodination of iodothyronines

Full size image

Deiodinating enzymes

There are three deiodinating enzymes (Table 1), first described in mammals (reviewed by Kohrle 1999; Bianco et al. 2002). These enzymes catalyse deiodination of thyroid hormones at the cellular level in extrathyroidal tissues. Type 1 (D1) was the first deiodinase enzyme to be described (Visser et al. 1976). It has both ORD and IRD activities and is primarily found in the liver, where it is responsible for converting rT3 to T2, by its IRD activity and for producing some T3 from T4 by its ORD activity. This T3 is considered to make up the bulk of T3 found in the circulatory system of mammals (Leonard and Kohrle 1996). A D1 enzyme has been found in the liver of all classes of amniote vertebrates. It appears to be also present in some teleost fish, where it is predominantly found in the kidney, not the liver (Sanders et al. 1999). A D1 enzyme has not been found in amphibians (Davey et al. 1995; Galton 1988; Becker et al. 1997) or lungfish (Sutija 2004). Molecular phylogenetics strongly support lungfish as the closest living sister group to the amphibians (Meyer and Zardoya 2003; Brinkmann et al. 2004). The absence of a D1 enzyme in both, living lungfish and amphibians, strongly suggests that the common ancestor of both these groups also did not have a D1 enzyme. Moreover, a D1 enzyme has not been found in living jawless vertebrates (lampreys: Eales et al. 1997, 2000). It has not yet been examined in cartilaginous jawed fish. Thus the D1 enzyme described in the kidney of the teleost fish, tilapia (Sanders et al. 1997; van der Geyten et al. 2001) is likely to be a result of parallel evolution between teleost fish and amniotes.

Table 1 Summary of the features of the three types of deiodinase enzymes
Full size table

In vivo studies of the hypothyroidal rat pituitary suggested the existence of another deiodinase enzyme with ORD activity (Silva and Larsen 1977). This led to the discovery of type 2 deiodinase (D2), which has only ORD activity. Its main role is to catalyse the conversion of T4 to the more active T3. This production of T3 at the cellular level is independent of circulating T3, which strongly suggests that the T3 produced as a result of D2 activity does not contribute to circulating T3 (Nguyen et al. 1998). D2 is expressed and active in all classes of vertebrates, with the possible exception of lizards in which all ORD activity may be via a D1 enzyme (Fenton and Valverde 2000). The most ancient of all vertebrates, the hagfish, has been described as having D2 enzyme activity in liver, intestine and muscle (McLeese et al. 2000). The other class of jawless vertebrates, lampreys, have highest D2 acivity in intestine, followed by liver (Eales et al. 1997; 2000). For most teleost fishes investigated, the highest D2 activity is found in the liver (Sanders et al. 1999). For lungfish also, the highest D2 activity was found in the liver. In fact, it was the only tissue that had sufficient measurable activity of D2 to allow the kinetic studies to be undertaken that confirmed this enzyme was D2 and not D1 (Sutija 2004). Mammalian liver, however, expresses exclusively D1. D2 has not been found in the liver of adult humans (Croteau et al. 1996), rats (Salvatore et al. 1996) or in amphibian tadpoles (Davey et al. 1995). Expression of D2 in mammals is localised mostly in the CNS (Crantz and Larsen 1980), brown adipose tissue (Leonard et al. 1983), placenta, thyroid, skin (Kaplan et al. 1988), pineal gland (Tanaka et al. 1986), retina (Buettner et al. 1998) and mammary gland, kidney, heart, spleen, prostate (Song et al. 2000).

Type 3 deiodinase (D3) has only IRD activity, which means it is exclusively responsible for the conversion of T4 to rT3 and T3 to T2 and therefore prevents accumulation of T4 and T3 in extrathyroidal tissues. D3 is expressed and active in all classes of vertebrates where it usually has highest activity in the brain (Larsen and Berry 1995; Leonard and Kohrle 1996; St Germain and Galton 1997; Mol et al. 1997), but is also expressed in other tissues such as liver and gill in tilapia (teleost; Mol et al. 1997), and liver, kidney, muscle and intestine in lamprey (Eales et al. 1997). The activity of each of these enzymes can vary between tissues at different stages in the life cycle of an animal, which may be a means of gaining a precise control over a particular tissue’s response to thyroid hormone with time (e.g. Galton 1988).

Deiodinases as selenoproteins

Over the last 15 years, these three enzymes have been further characterised as selenoproteins, meaning that they contain a selenocysteine in their amino acid sequence. In particular, the selenocysteine is contained within the ‘active centre’ of the enzyme and is directly involved in catalytic reactions (reviewed by Bianco et al. 2002). The chemical characteristics of the selenocysteine selonol group compared to a cysteine thiol group are: (1) greater reactivity towards nucleophilic substrate and (2) ionization at physiological pH (Muller et al. 1994; Burk and Hill 1999). Thus selenoproteins have a higher rate of catalytic reaction than proteins containing cysteine. An interesting feature of selenocysteine is that it is coded for by UGA, which under normal circumstances is a stop codon. In order to turn this stop codon into the codon for selenocysteine, the mRNA for a selenoprotein has a specific secondary structure in the 3′ untranslated region, known as the SElenoCysteine Insertion Sequence (SECIS, Berry et al. 1991, 1993; Shen et al. 1993). To date, two forms of SECIS have been described (Fig. 2). They share several features: an open region, a core, a stem and a loop/bulge. At the 5′ end of the stem, there is AUGAN and NGAN in the 3′ end. The core typically has non-Watson-Crick base pairing, termed a quartet. The function of the SECIS element depends on conservation of the nucleotides in the core region and in the loop/bulge region and on the secondary structure of the entire element (Shen et al. 1995). Experimental mutation of the SECIS core bases aiming to disrupt the non-Watson-Crick interaction between them, or to mutate the adenosines from the apical loop (form 1) or bulge (form 2), resulted in a rapid decrease of SECIS activity during translation (Berry et al. 1993; Martin et al. 1996, 1998). In rats, experimental conversion of form 2 to a form 1 SECIS in selenoprotein P resulted in the loss of 98% of its activity; whereas conversion of form 1 to form 2 had little effect on D1 activity. These results were interpreted as suggesting that the form 2 SECIS is the evolutionary sequel to the form 1 SECIS (Fagegaltier et al. 2000). This conclusion is somewhat complicated by a recent study of SECIS elements in lungfish deiodinases. Lungfish D3 has the expected form 2 SECIS (Sutija et al. 2004), in line with all other vertebrate D3 enzymes described so far, whereas the lungfish D2 gene has a form 1 SECIS (Sutija et al. 2003), normally only present in the D1 gene, D2 usually presenting the second form. The presence of the form 1 SECIS for D2 in lungfish suggests that the D1 of amniotes may have evolved from a lungfish-type D2 gene, prior to it adopting a form 2 SECIS. Molecular studies of deiodinase genes in cartilaginous fishes should resolve this seeming dilemma.

Fig. 2
figure 2

Schematics of the two forms of known eukaryotic SECIS elements, showing their characteristic features: a conserved region of the SECIS core, AUGAN at the 5′ part of the stem, NGAN in the 3′ part of the stem and the conservation of two or three adenosines whose position depends on the form of the SECIS element. The form 1 SECIS element is characterised by the presence of a conserved SECIS core, a 9–11 base pair stem that separates the core from an apical adenosine loop, consisting of 10–14 nucleotides with 2 or 3 adenosines on the 5′ site of the loop. In the form 2 SECIS element, 2 or 3 conserved adenosines are positioned in the adenosine bulge (5′ site of the bulge), which is separated from the SECIS core by the lower stem. The 2–5 base pair upper stem separates the adenosine bulge from the loop that consists of 3–6 nucleotides. N represents a characteristic nucleotide; filled circle represents non-Watson-Crick base pairing

Full size image

The unusual method of coding for the key amino acid in the function of selenoprotein enzymes provides a means of modifying the translation of deiodinating enzymes in tissues. For example, splicing of the gene that interferes with the transcription of the SECIS can result in the non-insertion of a selenocysteine and the reading of the UGA code of the mRNA as a stop codon, thus severely truncating the translation of the protein. Again, we can turn to lungfish for an example. The D2 mRNA obtained from lungfish brain, unlike that from liver, was truncated in the 3′UTR in the region of the SECIS (Sutija et al. 2003). Thus, in the brain of the immature lungfish studied, the SECIS could not fully form and presumably would not function to translate the stop codon UGA of the mRNA as a selenocysteine. This would produce a severely truncated protein without a fully formed ‘active centre’. Therefore, it appears that the lungfish D2 gene transcript is alternatively spliced in a tissue-specific manner. Alternative splicing of a single pre-mRNA is recognised as a pathway that generates diverse protein products. The current view is that around 60% of human genes are alternatively spliced (Maniatis and Tasic 2002). Furthermore, alternative splicing is recognised as a regulatory process for gene expression. It can act as an on–off gene expression control for selenoproteins, in that it can establish a premature stop codon (Smith and Valcarel 2000) which results in a truncated protein. Alternative splicing of the D2 mRNA in the brain of immature lungfish would result in no conversion of T4 to the more active T3 in the brain at this stage of their life history, i.e. beyond differentiation of the lungfish brain pattern.

Metamorphosis and deiodinases

Deiodinating enzymes are a very important level of control over a tissue’s response to thyroid hormones during metabolism, reproduction, growth and development. In this review we focus on the last of these—development. The best-known example of thyroid control over development is metamorphosis, particularly amphibian metamorphosis (Alberch 1989). While many hormones are involved in amphibian metamorphosis, only thyroid hormones are essential for this developmental event to commence. As metamorphosis approaches, there is a change in activity between D3, which decreases, following high activity during larval stages, and D2, which increases in activity, following low activity during larval stages (Galton 1989). The end result is a decrease in degradation to T2 and an increase in circulating levels of the more active thyroid hormone, T3. This T3 increase induces an increase in thyroid hormone receptors in tissues about to undergo remodelling (Yaoita and Brown 1990). The whole process results in a rise in plasma thyroid hormone (TH) levels, which peak at metamorphosing climax (White and Nicoll 1981). By this stage most of the thyroid hormone receptors are TRβ as opposed to the TRα type that characterise the pre-metamorphic tadpole (Eliceiri and Brown 1994). The increase in circulating TH levels prior to metamorphic climax is essential for development, but most of the metamorphic changes happen during climax, which takes only a few days and, during which, T3 triggers transcription of the genes involved in amphibian tissue remodelling.

Not all tissue remodelling takes place at the same stage of development. Becker et al. (1997) proposed that expression of D2 and D3 genes is one of the parameters responsible for the differential timing of metamorphosis in peripheral tissues. D3 is the major inactivation enzyme with the role to protect tissues from excess thyroid hormone. Expression of D3 is high during early development. As metamorphosis approaches, the expression of D3 begins to decrease but it does so in some tissues before others. Those tissues retaining high expression of D3 are resistant to TH-induced metamorphic change (Berry et al. 1998a, b). Overexpression of D3 can block metamorphic change (Huang et al. 1999). D2, responsible for availability of the more active T3, reaches its maximum expression in each tissue just prior to its commencing metamorphic change (Becker et al. 1997). Thus the relative activity of D3 and D2 enzymes in peripheral tissues regulates T3 availability to a tissue, which in turn regulates the onset and control of metamorphic changes (Huang et al. 2001).

It seems likely that similar scenarios for the level of control exerted by deiodinases on the outcome of responses to THs for other developmental events, for growth, metabolism and reproduction would also occur.

Summary

Since the realisation that deiodinases are selenoproteins (Berry et al. 1991), there has been considerable attention given to this class of enzyme. This has led to the discovery of the SECIS element in the 5’UTR, which is responsible for the insertion of a selenocysteine in the active centre of the translated protein. So far two types of SECIS elements have been described, each type being specific to a particular deiodinase, of which there are three (D1, D2, D3). Most studies have used mammalian models, but some have looked at other groups of vertebrates with a view to elucidating the evolution of this group of enzymes. Recent studies on lungfish deiodinases have produced some tantalising information in this regard. For example, all amniote vertebrates studied have a form 2 SECIS associated with the translation of D2 and D3 and a form 1 SECIS associated with D1. Moreover, D1 is absent in amphibians and lungfish, the anamniote vertebrates most closely related to the amniotes. Lungfish D2 SECIS is of the form 1 type, suggesting that the D1 of amniotes may have been derived from a lungfish type D2. Lungfish D2 was also shown to be alternatively spliced in brain such that the SECIS was truncated which would lead to non-insertion of selenocysteine for which the codon would then be read as a ‘stop’ in the translated product. Such tissue specific alternative splicing would result in lowered conversion of T4 into T3 in the brain, thus exerting a specific control over this tissue’s response to thyroid hormones. Future research should reveal how widespread this level of control may be.