Three deiodinase enzymes are involved in regulation of thyroid hormone availability

Three deiodinases regulate local and systemic availability of thyroid hormones, which are produced and secreted by the thyroid gland in higher vertebrates including humans. The physiological role of the deiodinases has been reviewed in recent editorials of this journal (1;2). Deiodinase enzymes are intracellular, integral membrane proteins, which either activate the prohormone L-thyroxine (T4) to the thyromimetically active 3,3’,5-triiodo-L-thyronine (T3), a reaction catalyzed by the type I (5’DI) and type II 5’-deiodinases (5’DII) , or which inactivate T4 or T3 by generation of iodothyronine metabolites such as 3,3’,5’-triiodo-L-thyronine (reverse T3, rT3) from T4 or 3,3’-diiodo-L-thyronine from T3, reactions catalyzed by the Type III 5-deiodinase (5DIII) and the type I 5’-deiodinase (3). The latter enzyme has limited, promiscuous substrate specificity. The type I and type III enzymes are also involved in further degradation of iodothyronines to other di- or mono-iodinated metabolites, which similar to rT3 and 3’3-T2 do not bind to T3-receptors (TRs), the ligand activated transcription factors. Sulfated T4 metabolites are solely deiodinated by the type I 5’-deiodinase (4).

Thyroid hormone availability is regulated in time, space and concentration

The three deiodinases show a remarkably specific expression pattern with respect to developmental stage, tissue and cell types, that provides the basis for a sensitive fine tuning of T3 provision to target cells in concert with the expression and activity of several specific transporters (5;6) for the charged amino acid-derived thyroid hormones. Recent data suggest distinct compartmentalization of deiodinase-, transporter- and TR-expressing cells, especially in the developing organisms but also in adult differentiated tissues and individuals (7;8). For example glial cells and tanycytes expressing the T3 generating 5’DII are localized adjacent to TR-positive neurons which also express the T3 degrading 5DIII (9). Similar cell type-specific and time-dependent distinct expression patterns and levels of activity have been observed for the three deiodinase enzymes during placental implantation of the embryo (7), in human placental layers (10;11), or during the amphibian metamorphosis (1). 5DIII expression is assumed to limit T3 availability and to prevent exposure of target cells to T3 inappropriate in concentration, location and time in developing and adult organisms (12). Thyroid hormone levels and deiodinase expression are also modified by several drugs as well as by hypo- and hyperthyroidism, which differently affect the three deiodinase enzymes (13). Apart from cellular proteins involved in modulating thyroid hormone availability, serum contains the highly specific thyroid hormone transport, distribution and binding proteins thyroxine-binding globulin (TBG), transthyretin (TTR), albumin and some lipoproteins, which have evolved to generate a buffer system for the highly hydrophobic iodothyronines, covering almost 6 orders of magnitude in their binding affinity (Ka for T4 range from 1010 to 105 M-1).

5’DII knockout mice are viable and show remarkably mild phenotype

All three deiodinases belong to the small and exclusive class of selenocysteine-containing proteins (3). Selenocysteine, the 21st proteinogenic amino acid, located in the active site of these enzymes, is essential for their proper function (14). Recently, models of mechanism of reaction and inhibition by the antithyroid drug PTU have been proposed based on the active site selenocysteine of the 5’-deiodinase enzymes (15;16). So far no reports on major genetic defects associated with deiodinase mutants have been reported for humans but polymorphisms have been linked to altered thyroid hormone metabolism (see below). Knockout mouse models for the single deiodinase enzymes revealed markedly mild phenotypes (17), suggesting some redundancy and/or functional compensation by the unaffected family deiodinase members, thyroid hormone transporters and/or T3-receptors. This is somewhat reminiscent of many TR?-receptor mutants and some TR-transgenic mouse models, which also indicate a complex network of fine-tuned fail-save control mechanisms of thyroid hormone action even in the case of thyroid hormone resistance (18). In mice the locus of the 5DIII gene appears to be genomically imprinted and preferentially expressed from the paternal allele and might be linked to phenotypic abnormalities associated with uniparental disomy (19). Targeted disruption of the 5’DII gene in the mouse model reveals serum hormone constellations similar to those observed in pituitary resistance to thyroid hormone with elevated T4 and TSH, normal T3 and altered tissue levels of type 5DIII activities (17). In addition, an impaired thermoregulation has been observed, compatible with 5’DII expression in brown adipose tissue of rodents. No full report on the phenotype of knockout mouse models for 5’DI or cross breeding of the individual knockout mouse models has been published yet. Inbred mouse strains with impaired 5’DI expression due to CGT repeats in an altered promoter structure of this gene have normal serum TSH and T3, but elevated rT3 and free T4 levels (20), again indicating the potential of functional compensation (21) for impaired hepatic, renal and thyroidal expression of 5’DI, even under conditions of streptozotocin-induced diabetes in this mouse strain (22).

Non-thyroidal illness, the low-T3 syndrome and the deiodinases

Impaired hepatic production of T3 by 5’DI has been one of the hallmarks of this dazzling syndrome frequently encountered in several clinically relevant conditions such as starvation, infection, sepsis, major trauma, and chronic non-thyroidal illness but also after administration of several drugs (23;24). The exact molecular mechanism of decreased hepatic expression of 5’DI has been elusive for three decades of research. Therefore no rationale for thyroid hormone treatment of these patients was given, as indicated by the alternatively used term euthyroid sick syndrome (25). Only few studies actually showed decreased tissue T3 (and T4) levels in humans (26). Remarkably, impaired 5’DI expression has been mainly observed in the liver, while 5’DI activities in the kidney, thyroid and pituitary were rarely affected in most animal experimental models mimicking this human constellation. Selenium deficiency which, due to low serum selenium levels would limit hepatic expression of the selenoenzyme 5’DI could be ruled out as one potential factor contributing to the low T3 syndrome (27;28). Strong evidence for inhibition of 5’DI by proinflammatory cytokines, frequently elevated in serum of these patients, has been presented in several animals and cell culture models and in patients, and even strong inverse correlations to serum T3 levels in patients were demonstrated (24;29;30). Interleukins 1 and 6 (IL-1, Il-6), tumor necrosis factor a (TNF?), and Interferon ? inhibit 5’DI expression and activity by transcriptional and post-transcriptional mechanisms involving activation of the transcription factor NF-?B, limiting the amount of the nuclear receptor coactivator SRC1 or directly inhibiting 5’DI promoter activity (31). However, many of the later effects are not limited to hepatic 5’DI expression, but also inhibit the enzyme in the kidney, and the thyroid, or even stimulate 5’DI in the pituitary. This indicates a complex disturbance of the homeostasis of thyroid hormone deiodination pathways, an observation supported by studies in several mouse knockout models where individual components of the cytokine network were inactivated (32). A promising recent observation in critically ill patients provides novel insight into an additional mechanism contributing to the pathogenesis of the low T3 syndrome and also offers an explanation for the elevated rT3 levels found in these patients: Their liver re-expresses type III 5-deiodinase similar to the fetal and neonatal organ, while livers of healthy adult mammals including humans solely express 5’DI activity (33). Hypoxia has been suggested as the triggering signal for 5DIII re-expression, reminiscent of high 5DIII expression in some hemangioma (see below) (34).

Deiodinases in tumors

5’DI:
5’DI is a sensitive marker for the differentiation stage of several epithelial tumors. E.g. while 5’DI is highly active in the healthy thyroid gland, its activity is reduced in differentiatied thyroid carcinomas and very low or even undetectable in anaplastic thyroid carcinoma or corresponding cell lines, accompanied by a decrease in the expression of the 27 kDa substrate binding catalytic subunit (35). Using DNA microarray technology, Huang et al. (34) demonstrated a down-regulation of 5´DI mRNA in papillary thyroid carcinoma vs. normal thyroid and Barden et al. (36) in follicular thyroid carcinoma vs. follicular adenoma. Expression of the deiodinase gene also loses responsiveness to physiologic stimulation by TSH and T3, but remains sensitive to retinoic acid in differentiated thyroid carcinoma cell lines, an effect not observed in normal thyroid cell lines, but detectable also in the hepatocarcinoma cell line HepG2 (37). In anaplastic lines, also RA responsiveness is lost.
5’DI expression is also reduced in prostate carcinoma (38), in clear cell carcinoma of the kidney, provided that these tumors do not derive from a cell type that does not express 5'DI (39), and in liver adenocarcinoma, although only one sample was compared to normal liver (40). In the mammary gland of the rat, 5’DI is up-regulated during lactation by suckling-triggered prolactin and ?-adrenergic stimulation (41). In 1-methyl-1-nitrosourea (MNU)-induced mammary cancer of female Sprague-Dawley rats, 5'DI activity was at least two orders of magnitude higher than in normal non-lactating mammary gland (42). In striking correspondence to other epithelial carcinoma cell lines, 5’DI gets responsive to RA in the human mammary carcinoma line MCF-7, whereas regulation by physiologic stimuli, T3 or the ?-adrenergic agonist isoproterenol, is abolished. In the more dedifferentiated breast cancer cell line MDA-MDB-231, retinoic acid stimulation is not observed (43).
Taken together, these results show a tight linkage of 5’DI expression and regulation to the differentiation stage of various epithelia, suggesting a central role for the enzyme in the proper function of these tissues.

5’DII:
Disturbed activities of 5’DI and 5’DII in the pituitary gland would impair hormonal feedback regulation. As reported by Baur et al. (44), 5’DII activity was higher than 5’DI activity in 40 of 43 adenomas and 2 of 3 normal pituitaries, whereby in 15 tumors, only 5’DII and no 5’DI activity at all was observed. In three adenomas and one normal pituitary, there were higher 5’DI than 5’DII levels. Tannahill et al. (45) reported that pituitary tumors expressed 2.6-fold higher 5’DII and 6.5-fold higher 5DIII mRNA levels as compared to normal pituitaries. 5’DI transcripts were only detectable in a minority of the tumors, and in these cases, overexpression was determined. A correlation between deiodinase mRNAs and enzyme activities that was observed in normal pituitaries was lost in the tumors.
Both 5’DI and 5’DII mRNAs and enzyme expression may be enhanced in benign thyroid diseases, i. e. Graves’ disease and thyroid adenomas (46-48).
Huang et al. (49) also reported down-regulation of 5’DII mRNA in PTC, and 5’DII acitivity is increased in thyroid adenoma, but reduced in papillary thyroid carcinoma (50). Kim et al. (51) recently described increased T3 formation from either endogenous or plasma-derived T4, resulting in a persistently increased ratio of serum T3 to T4, which most probably was due to 5’DII expression in follicular thyroid carcinomas. This was deduced from kinetic parameters and from the observation that the elevated serum T3 was not decreased after administration of the 5’DI inhibitor PTU, respectively. A remarkably high expression of 5’DII has been observed in the mesothelioma cell line MSTO-211H, facilitating the further characterization of this enzyme as a selenoprotein (52). It will be of interest, if primary mesotheliomas also express 5’DII; the transformed mesothelial cell line MeT-5A does not.

5DIII:
Hepatic or cutaneous hemangioma affects 5 to 10 % of one-year-old children. These tumors often express very high levels of 5DIII which mediate enhanced turnover of T3 that cannot be balanced by the synthetic capacity of the thyroid (“consumptive hypothyroidism”). In an adult patient, increased 5DIII activity in a hemangioma caused subclinical hypothyroidism (34;53).

Single nucleotide polymorphisms (SNPs) in deiodinase genes

SNPs in the 5’DI gene associated with altered thyroid hormone metabolism:
So far no deiodinase mutations have been discussed to be involved in thyroid diseases. However, several SNPs have recently been described for the 5’DI gene (D1a-C/T at position 785 and D1b-A/G at position 1814 of the 5’DI cDNA sequence) that were associated with plasma TSH and iodothyronine levels in a population of 158 healthy persons (54). Interestingly, these SNPs affect the 3’-untranslated region in the vicinity of the selenocysteine insertion element which facilitates the insertion of selenocysteine, encoded by a UGA codon, into the catalytic center of the selenoenzyme 5’DI. The T allele of D1a was very frequent and associated with higher plasma rT3 levels, with a corresponding 33 % increase per allele copy, higher plasma rT3/T4 ratios and lower plasma T3/rT3 ratios. The A allele of D1b correlated with decreased plasma rT3 levels, decreased plasma rT3/T4 ratios and increased plasma T3/rT3 ratios, although this was not significant due to the low frequency of this polymorphism. Of the three haplotype alleles observed in this population, namely 1: aCbA, 2: aT-bA and 3: aC-bG, haplotype 2 was positively correlated with rT3, rT3/T4 and negatively with T3/rT3 ratios, whereas haplotype 3 was negatively correlated with rT3, rT3/T4 and positively with T3/rT3, but this was not significant. Haplotype 1 did not show any correlation to serum hormone levels.
A T/G polymorphism in the 3’-untranslated region of DIII also was not correlated with plasma hormone levels.
Thus, SNPs in 5’DI may well be the cause for subtle variations in thyroid hormone levels, which nevertheless can have consequences for quality of life, cognition, heart rate and other physiological processes under the control of thyroid hormones and for set points of endocrine feedback regulation.

5’DII as a candidate gene for Syndrome X:
A SNP within the protein coding region of the 5’DII gene was recently described, featuring either Thr or Ala at amino acid 92. As 5’DII contributes to the feedback-regulation by thyroid hormone in the thyroid gland, a correlation with serum TSH levels is conceivable, but was not observed in the study mentioned above (54).
However, in the context of the so-called Syndrome X, the complex of obesity, hypertension, insulin resistance and glucose intolerance/diabetes, this polymorphism may be of importance (55). The allele frequency of this SNP was 0.35 in a population of morbidly obese caucasians, 0.72 in Pima Indians and 0.42 in Mexican-Americans. Association studies in 135 nondiabetic women showed that the Ala variant was correlated with a lower glucose disposal rate, and a trend to higher fasting insulin levels was also observed. In a large study group of 972 nondiabetic subjects recruited for energy balance studies, there was no correlation of the Ala variant with body weight or body mass index. However, carriers of an additional Trp64Arg polymorphism affecting the ?3-adrenergic receptor, showed significantly higher weight and BMI.
These data strongly suggest an association of the Thr92Ala mutation in the 5’DII gene with glucose intolerance and diabetes, suggesting a role of local T3-production by 5’DII in the regulation of energy metabolism. This would result in a vicious circle: reduced T3 leads to reduced expression of GLUT4 in insulin-sensitive and 5’DII expressing tissues such as skeletal muscle or adipose tissue. As 5’DII is stimulated by cAMP via a cAMP-responsive element in its promoter, this condition might be further aggravated via reduced stimulation by a variant ?3-adrenergic receptor, leading to a decrease in cAMP-stimulated expression of an already functionally impaired 5’DII enzyme.
The association between the Thr92Ala SNP and glucose tolerance/diabetes, however, is challenged by the observation that the Thr92Ala 5’DII variant, expressed in vitro, did not show any difference in the kinetic parameters of T4 deiodination as compared to the wildtype enzyme. If the discussed correlation can be explained, e. g., by association with another polymorphism, will have to be clarified in further studies.

Table: Clinically relevant factors regulating and modulating deiodinase function