| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Thyroid Division, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School (A.C.B., M.J.B., P.R.L.), Boston, Massachusetts 02115; Dipartimento di Endocrinologia ed Oncologia Molecolare e Clinica, Universita’ Federico II (D.S.), 80131 Naples, Italy; Institute of Experimental Medicine, Hungarian Academy of Sciences (B.G.), Budapest, H-1083 Hungary
Correspondence: Address all correspondence and requests for reprints to: P. Reed Larsen, M.D., Brigham and Women’s Hospital, 77 Avenue Louis Pasteur, Harvard Institutes of Medicine Building, Room 550, Boston, Massachusetts 02115. E-mail: rlarsen{at}rics.bwh.harvard.edu
 |
Abstract |
|---|
 Top Abstract I. Introduction and Historical...II. The Synthesis of...III. Specific Biological...IV. Summary of the...V. The Physiological Roles...VI. The Deiodinases in...VII. Effects of Genetic...VIII. Conclusions and Future...References |
|---|
I. Introduction and Historical Review
II. The Synthesis of Selenoproteins
A. Recoding UGA from STOP to selenocysteine (Sec)
B. Trans-acting factors are recruited by the Sec insertion sequence (SECIS) element to catalyze Sec incorporation
III. Specific Biological Properties
A. Type 1 iodothyronine deiodinase (D1)
B. Type 2 iodothyronine deiodinase (D2)
C. Type 3 iodothyronine deiodinase (D3)
IV. Summary of the Important Similarities and Differences in the Human Iodothyronine Selenodeiodinases
V. The Physiological Roles of the Selenodeiodinases
A. The critical role of D2 in feedback regulation of TSH secretion
B. T3 homeostasis
C. Embryonic development and metamorphosis
D. Maternal-fetal physiology
E. The essential role of D2 in adaptive thermogenesis
F. Summary
VI.The Deiodinases in Human Pathophysiology
A. Alterations in iodothyronine deiodination in the response to fasting or illness
B. D3 overexpression in hemangiomas causes consumptive hypothyroidism
C. D1 overexpression contributes to the relative excess of T3 production in hyperthyroidism
D. Effects of inhibition of deiodinase function during therapy with amiodarone
VII. Effects of Genetic Alterations in Deiodinase Expression
A. Effects of a spontaneous genetic deficiency in Dio1 gene expression
B. Effects of targeted disruption of the Dio2 gene
C. Isolated myocardial D2 overexpression causes cardiac thyrotoxicosis
VIII. Conclusions and Future Directions
|
I. Introduction and Historical Review |
|---|
|
TopAbstractI. Introduction and Historical...II. The Synthesis of...III. Specific Biological...IV. Summary of the...V. The Physiological Roles...VI. The Deiodinases in...VII. Effects of Genetic...VIII. Conclusions and Future...References |
|---|
and Ref. 2). Twenty years passed before the development of assays for quantitation of T3 in human serum reawakened interest in this activation step (3). Work over the subsequent two decades consisted primarily of documenting the presence of two different enzyme activities that catalyzed T4-to-T3 conversion, the types 1 and 2 iodothyronine deiodinases (D1 and D2, respectively), and the identification of an inner ring deiodinase, which can inactivate T4 or T3 (4, 5). D1 and D2 were first distinguished by the presence (D1) or absence (D2) of sensitivity to inhibition by 6-n-propyl-2-thiouracil [PTU (6, 7, 8, 9, 10)]. It is important to recognize that, due to the free rotation of the phenolic (outer) ring in the iodothyronine molecules, monodeiodination at the 5 or 3 positions of the tyrosyl ring are equivalent inner-ring deiodinations (IRD), and those of the 3' or 5' positions (phenolic ring) are equivalent outer-ring deiodinations (ORD). In this review we will refer to ORD and IRD as 5' and 5, respectively (Fig. 1).
|
). The complete cDNA sequences of the human, chicken, and mouse D2 cDNAs containing the critical mRNA structures required for Sec incorporation (see Section III.B) were subsequently identified (18, 19, 20).
|
The goal of this review is to place the exciting advances that have occurred in our understanding of the molecular biology of the iodothyronine deiodinases into a biochemical and physiological context. The reader is referred to several earlier reviews for a more detailed scientific background of concepts underlying much of the work to be discussed below (25, 26, 27, 28, 29). Although we will focus on new insights, these will be placed in the context of previous knowledge to allow presentation of a coherent picture of the role of the deiodinases in thyroid physiology. After a discussion of the mechanism of selenoprotein synthesis, we will review the specific molecular and cellular biological properties of the individual deiodinases. These provide the background for a discussion of their role in thyroid physiology and pathophysiology. Whenever possible, we have focused the discussion on the biological role of these enzymes in human physiology.
|
II. The Synthesis of Selenoproteins |
|---|
|
TopAbstractI. Introduction and Historical...II. The Synthesis of...III. Specific Biological...IV. Summary of the...V. The Physiological Roles...VI. The Deiodinases in...VII. Effects of Genetic...VIII. Conclusions and Future...References |
|---|
). Thus, a major unanticipated avenue of research had its beginnings with the identification of Sec in the active site of rat D1-the investigation of the requirements for and mechanism of Sec incorporation in eukaryotes.
A. Recoding UGA from STOP to selenocysteine (Sec)
1. Identification of the Sec insertion element (SECIS) element.
Type 1 deiodinase was only the second eukaryotic mRNA shown to encode a selenoprotein, the first being classical glutathione peroxidase (GPX). However, little was known about the mechanism of synthesis of selenoproteins in eukaryotes. Several prokaryotic selenoprotein cDNAs had been sequenced, and using these cDNAs in biochemical and genetic studies, the cis-acting sequences and trans-acting factors required for Sec incorporation in prokaryotes had been elucidated (31, 32). The cis-acting sequences consist of the Sec codon itself, UGA, and a specific RNA stem loop immediately downstream of the UGA codon. UGA is recognized in the vast majority of mRNAs as a STOP codon. Only in the presence of the stem-loop structure and trans-acting factors are UGA codons "recoded" to specify Sec instead.
The trans-acting factors identified in bacteria are encoded by genes designated Sec synthase (selA), elongation factor with mRNA stem-loop binding activity (selB), tRNA [Ser]Sec (selC), and selenophosphate synthase [selD (Table 1 and Refs. 31 and 33, 34, 35, 36, 37)]. Both selA and selD encode enzymes required for Sec biosynthesis, and selC encodes a unique tRNA possessing an anticodon complementary to UGA and a secondary structure that differs from all other tRNAs. selB encodes a protein with two distinct functional domains. The first, an elongation factor domain, recognizes selenocysteyl-tRNA via its unique structure and amino acid and delivers the tRNA to the ribosome. The second domain, a C-terminal extension, specifically binds the RNA stem loop downstream of the UGA codon in prokaryotic selenoprotein mRNAs. Thus, recruitment of the elongation factor via the RNA stem loop results in recoding of only the immediately adjacent UGA.
|
). Examination of the GPX sequence revealed the potential of its 3' UTR to form a similar stem loop. In addition, both the D1 and GPX stem loops contain three short regions of conserved nucleotide sequence. The property of these two elements to function interchangeably in conferring D1 expression led to the concept of the stem loop structure in the 3' UTR as necessary and sufficient for conferring Sec incorporation at UGA codons and as a feature whose presence would be a hallmark of selenoprotein mRNAs. It was thus termed the SECIS element (21).
|
). The minimal sequence required for SECIS function was defined, the boundaries of which correspond precisely with the conserved 5' A/GUGA and 3' GA sequences (46). This region was subsequently shown to form non-Watson-Crick base pairs: purine pairs between the GA at the 5' base of the stem (in the conserved A/GUGA sequence) and the GA at the 3' base, and pyrimidine pairs flanking these two (Fig. 3 and Refs. 49 and 50). Similar nonstandard base-pairing features have been shown to serve as binding sites for several sequence- and structure-specific RNA-binding proteins. Finally, there can be two alternative arrangements in the hairpin loop, designated form 1 or form 2 (51, 52, 53). In form 1 SECIS elements, including those of D1 and GPX, the conserved adenosines are contained in a simple open loop. However, the D2 and D3 SECIS elements are predicted to form additional secondary structure in this region, with the adenosines located in a bulged region (Fig. 3, Form 2). Additional mechanistic insights into SECIS function were gleaned from the D1-activity assay. These studies revealed that a SECIS element in the 3' UTR could direct incorporation at any upstream in-frame UGA codon, and at multiple UGAs within an mRNA, provided a minimal spacing requirement was met (42). It was further shown that increasing the spacing between UGA and SECIS element by the insertion of 1.5 kb had no effect on SECIS activity. However, deletions that narrowed the spacing between UGA and SECIS to less than approximately 60 nucleotides (nt) abolished Sec incorporation (46). This may be due to steric constraints between the complex of factors assembled at the SECIS element (see Section II.B) and the ribosome decoding the UGA codon. At the other extreme, the identification of the human D2 mRNA with a UGA to SECIS spacing of nearly 5 kb indicates the upper limits for this distance may be very large (18).
B. Trans-acting factors are recruited by the Sec insertion sequence (SECIS) element to catalyze Sec incorporation
Studies of Sec incorporation in prokaryotes lent considerable insight into investigation of this process in eukaryotes. Homologs of the selC (tRNA[Ser]Sec) gene were identified in many species (54). Homologs of selD have also been identified in several eukaryotic species, some of which contain two selD genes, one encoding a selenoenzyme (Table 1). This provides a Se-dependent autoregulatory step in Sec biosynthesis. Recently, candidate selA homologs have been identified and are currently under investigation. Finally, factors conferring the two functions of prokaryotic selB have recently been identified and characterized. Identification of the eukaryotic SECIS element as the essential feature required for conferring UGA recoding led to the proposal of a model whereby this element would recruit a factor or factors conferring Sec incorporation, analogous to prokaryotic selB (42). This in turn led to the search for such factors, employing RNA-protein interaction methods using wild-type and nonfunctional mutant SECIS elements to establish specificity. Nonetheless, progress on this front was slow (55, 56, 57). In parallel, progress in genomics provided databases in which to search for sequence homologs of prokaryotic selB. This resulted in identification of candidates in archaea (58), and subsequently, in lower and higher eukaryotes. Finally, these two lines of research converged with the identification and cloning of two factors crucial to Sec incorporation in eukaryotes. First, a SECIS-specific binding protein, termed SECIS binding protein 2 (SBP2), was purified and cloned and shown to function in Sec incorporation (59, 60). This factor is limiting in reticulocyte lysates and in the cultured cell lines examined. Addition of the factor in vitro or its expression in vivo increases the efficiency of Sec incorporation (60). Next, the elongation factor candidates in archaea and in mammals were shown to exhibit specificity for selenocysteyl-tRNA (22, 61, 62). The designation EFsec was proposed to reflect this specificity. A final piece of information begins to bring the puzzle together was the demonstration that the elongation factor interacts with the SECIS-binding protein (22). Thus, the two functions contained within a single prokaryotic factor, selB, recruitment by the prokaryotic equivalent of a SECIS element and selenocysteyl-tRNA-specific elongation factor activity, are distributed between two separate but interacting eukaryotic proteins.
A model emerging from these results is depicted in Fig. 4. According to this scheme, the SECIS element recruits SBP2, an event that could theoretically occur in the nucleus as soon as this region is transcribed. The SECIS-SBP2 complex could then recruit the EFsec-tRNA complex and deliver it to the ribosome in the coding region. Because the SECIS element is located in the 3' UTR in eukaryotes, not in the coding region as in prokaryotes, it obviates the need for dissociation and reassociation of the SECIS-SBP2 complex with each incorporation cycle. A scheme such as this could potentially allow rapid reformation of SECIS-SBP2-EFsec-tRNA complexes from the two individual RNA-protein complexes after each EFsec-tRNA delivery cycle, i.e., "reloading" for the next approaching ribosome. This would also be advantageous in the translation of a protein containing multiple Sec residues, such as selenoprotein P (63, 64).
|
|
III. Specific Biological Properties |
|---|
|
TopAbstractI. Introduction and Historical...II. The Synthesis of...III. Specific Biological...IV. Summary of the...V. The Physiological Roles...VI. The Deiodinases in...VII. Effects of Genetic...VIII. Conclusions and Future...References |
|---|
and Ref. 68).
1. Dio1 gene structure, chromosomal localization, mRNA and protein characteristics, and tissue distribution.
a. Gene structure and chromosomal localization.
The elucidation of the Dio1 gene structure was derived from studies comparing a polymorphism in the Dio1 gene between the C57/BL6J and C3H/HeJ mouse strains (69, 70). The human gene is found on chromosome 1 p32–p33, in a region syntenic with mouse chromosome 4, the location of mouse Dio1 (71). The mouse and human Dio1 genes consist of four exons. The transcription start site is approximately 25 nt upstream of the initiator methionine. The UGA (Sec) codon is in exon 2, and the UAG (STOP) codon and the SECIS element are in the 953-nt fourth exon. The coding sequences of the mouse and rat D1 proteins are virtually identical. Both contain a Sec residue at position 126 (70).
b. D1 mRNA and protein characteristics.
The complete cDNA sequences have been determined for rat, human, mouse, dog, chicken and tilapia D1 proteins (11, 70, 72, 73, 74, 75). The mRNA sizes are about 2–2.1 kb and all contain a UGA codon in the region encoding the active center, which is highly conserved among species (Fig. 2). The cDNA encodes a protein of about 27 kDa that is highly similar in size (26–30 kDa) and sequence with a few informative exceptions (76). Depending on the detergent used, the molecular mass of the solubilized wild-type enzymes is about 50–60 kDa, suggesting that it may be a homodimer, although it is not yet certain that homodimerization is required for its catalytic activity (see Section III.A.3 and Refs. 77, 78, 79, 80).
c. Tissue distribution.
By Northern analysis, D1 is expressed in many tissues of most vertebrates but not in amphibia (27, 81, 82). In the rat, these include liver, kidney, central nervous system (CNS), pituitary, thyroid gland, intestine, and placenta. In humans, D1 activity is notably absent from the CNS but is present in liver, kidney, thyroid, and pituitary and mRNA in circulating mononuclear cells by RT-PCR (83, 84).
2. Subcellular localization and topology.
The D1 monomer is a type 1 integral membrane protein oriented with a 12-amino acid NH2-terminal extension in the endoplasmic reticulum (ER) lumen and a single transmembrane domain exiting the ER at about position 36 (Fig. 5 and Ref. 85). The hydrophobic nature of the NH2 terminus suggests that this portion of the molecule is an uncleaved signal recognition sequence and incorporates both signal and STOP-transfer functions. The transmembrane domains of other proteins, such as 17
- hydroxylase (P450-17) or D3, cannot substitute for the NH2 terminus of D1 even though these permit synthesis of a membrane protein. This orientation is in agreement with earlier studies showing that gentle trypsinization of kidney microsomes caused both loss of enzyme activity and N-bromoacetyl (BrAc)T3 labeling (86, 87, 88). Studies of the in vitro-translated Sec126Cys mutant of rat D1 show that, although the NH2-terminal and transmembrane portions of the enzyme are not catalytically active, their sequence is critical because even minimal exchanges of amino acids in the transmembrane domain reduce the efficiency of its transient expression. These mutations do not affect the catalytic function of the protein that is successfully synthesized (85).
|
and Ref. 92). Furthermore, when either COOH- or NH2-terminal FLAG-tagged D1 is transiently expressed in HEK293 cells that are then subjected to limited permeabilization of the plasma membrane with digitonin, the FLAG tag is visualized at the plasma membrane, even though BiP cannot be visualized. This observation confirms the earlier assignment of D1 to plasma membrane of kidney and thyroid cells. A preliminary report using transiently expressed green fluorescent protein-tagged D1 and D2 predicted an ER location for both enzymes (93). The reason for the discrepancies with the above-mentioned results with respect to the subcellular location of D1 is not clear (92). The location of D1 in primary hepatocytes remains to be determined.
|
and Ref. 92 ; see below). This differential subcellular localization of D1 and D2 may explain why there is such a minimal contribution of the T3 generated by D1 to the intranuclear T3 in contrast to the large fraction of D2-generated T3 to this compartment (94, 95). Early studies of rat kidney or liver D1 suggested the possibility that it was dimerized with a second protein, giving it a molecular mass of approximately 54–55 kDa (78, 96). It was not clear whether the enzyme was present as a homodimer or was bound to a protein of similar size. As mentioned, transient expression of a D1 enzyme in which the transmembrane domain has been deleted does not permit synthesis of functional protein (85). However, recent studies indicate that synthesis of a functional D1 enzyme lacking the NH2-terminal transmembrane domain can occur if an intact, but catalytically inactive, D1 protein is also expressed in the cell (80). This suggests that homodimerization can occur between the cytosolic portions of D1 and that it is only necessary for one member of the homodimer to have a membrane anchor to permit the successful synthesis of functional D1. The amounts of intact D1 protein used to trap active, NH2-terminal-deleted D1 were about 10-fold in excess of the quantity of the truncated protein. Accordingly, it is not certain whether homodimerization is required for enzyme function or whether dimerization and successful synthesis of a monomer without a transmembrane domain can occur when D1 is produced in large amounts.
3. Enzymatic properties and structure-activity relationships.
Studies with both endogenous and recombinant enzymes indicate that the deiodination reaction catalyzed by D1 follows ping-pong kinetics with two substrates, the first being the iodothyronine, and the second being an endogenous intracellular thiol cofactor (30, 65, 66, 97, 98, 99). The first half-reaction deiodinates the iodothyronine leading to the formation of a putative selenoleyl iodide intermediate (Fig. 7). This is then reduced by an as yet unidentified intracellular thiol cofactor regenerating the enzyme. As indicated, PTU inhibits D1-catalyzed deiodination by competing with the putative thiol cosubstrate to form an essentially irreversible Enzyme-Se-S-PTU dead-end complex.
|
and Ref. 30). This is similar to effects of this substitution in other selenoenzymes, such as formate dehydrogenase in bacteria (100). It should be noted, however, that the efficiency of the translation of the Cys126 is 50–100 times higher than SecD1 due to the inefficiency of selenoprotein translation in eukaryotes (see Section II and Ref. 30). Interestingly, a recent paper illustrates that deiodination of T4 to T3 with ping-pong kinetics can occur with a synthetic D1 enzyme. This protein was generated from a mouse monoclonal anti-T4 antibody in which the OH groups of the four seryl residues of the variable region of the light chain were chemically replaced by Se, thus forming four Sec residues (101). The reaction was inhibited by PTU, which competed with dithiothreitol (DTT) just as in D1. This remarkable result serves to emphasize the importance of Sec in the deiodination reaction.
|
and Refs. 98 and 102). That this reaction is specific for the active-center Sec residue is shown by the protection of D1 from inactivation by iodothyronine substrates. Alternatively, a similar alkylation-based inactivation occurs using iodothyronine derivatives containing BrAcT3, rT3, or T4 (Fig. 8). If these iodothyronines are labeled with 125I, the protein will be covalently labeled (103, 104). The BrAc-iodothyronine derivatives also specifically label other iodothyronine binding proteins (e.g., T4-binding globulin, transthyretin, and TR), but also nonspecifically "alkylate" nonthyroid hormone binding proteins in microsomal preparations of rat liver such as protein disulfide isomerase (104). The specificity of the D1 labeling is established by the fact that D1 substrates block labeling of the 27-kDa D1 monomer but not that of other proteins (Fig. 8 and Refs. 25 , 89 , and 105). The blockade of this labeling by substrates or competitive inhibitors, such as gold thioglucose (GTG), confirms that BrAcT3 or T4 interacts with the substrate binding site. Gold is a competitive inhibitor (with iodothyronine) of the deiodination reaction, presumably interacting with the Se in the active center (11, 30). PTU does not block D1 labeling with BrAcT4 unless D1 is first exposed to substrate because the first half-reaction is required for formation of the Enzyme-Se-PTU complex (106). The Sec126Cys mutant D1 can also be labeled by BrAcT4, although the protection afforded by GTG is considerably less potent in agreement with the approximately 100-fold decrease in sensitivity of the Cys D1 mutant-catalyzed deiodination reaction to GTG (107). In fact, the loss of sensitivity to GTG of the Sec126Cys mutant and the relative insensitivity of D2 to inhibition by GTG led to the speculation that D2 would not prove to be a selenoprotein (108). This was based on the assumption that GTG would be as effective an inhibitor of any selenoenzyme as it was of GPX and D1 (11, 109). In fact, this assumption was not valid because both endogenous and recombinant D2 are 100-fold less sensitive to GTG than is D1. Furthermore, the kinetics are noncompetitive, implying that this inhibition is not due to interaction of GTG with the active center (110).
|
). The only exception to this is in the tilapia D1, in which proline replaces serine at position 128 (75). Because this is also characteristic of the PTU-insensitive D2 and D3 enzymes (Fig. 2), site-directed mutagenesis was used to replace this proline with serine (75). However, PTU sensitivity was not restored by this substitution, indicating that the explanation for the PTU insensitivity of tilapia D1 lies elsewhere in the protein sequence.
Kinetic studies have recently been performed using a rat D1 enzyme in which the vicinal Cys at position 124 of the active center was replaced by the alanine (Ala) found in D2 (Fig. 2) to test whether this residue is involved in catalysis by D1 (111, 112). The rat Cys124Ala D1 protein had a 10- to 15-fold higher apparent Michaelis-Menten constant (Km) for DTT than wild type, suggesting that the SH group of this Cys residue was involved in the interaction with the second substrate. However, the maximum velocity (Vmax) and Km of the C124A mutant was not significantly different, although there was a 2-fold increase in the Ki for PTU. This supported a reaction mechanism for the D1 enzyme in which DTT interacts with the vicinal Cys to facilitate reduction of the oxidized Se in the active center (111). However, this mechanism, as is the case for that shown in Fig. 7, must remain speculative due to the lack of structural information.
In addition, Cys194 in D1 is conserved in all three deiodinase classes, suggesting an important role for this residue (Fig. 9). Replacement of this residue in D1 with Ala caused a modest increase in the Km and decrease in Vmax for rT3 (112). Interestingly, neither the Cys124Ala nor the Cys194Ala mutations affected the rate of deiodination in cells transiently expressing these mutant D1 enzymes, suggesting that the increase in the Km of the Cys124 mutant for DTT and the decreased Vmax observed in vitro are not rate limiting in vivo (112). This could occur if reactivation of D1 by an endogenous thiol cofactor is very slow or does not occur in vivo (see below).
|
Comparisons of the D1 enzymes of different species have led to the recognition of other structurally important amino acids. For example, the phenylalanine at position 65 is critically important for 5' deiodination of rT3 and 3,3'-diiodothyronine sulfate (T2S) but not for deiodination of substrates with two iodines on the inner ring (73, 116). This was demonstrated by selective mutagenesis of the dog D1, which has an approximately 30-fold higher Km for rT3 than human or rat enzymes and contains a leucine at this position. It suggests a specific interaction of the inner ring of rT3 and 3,3'-T2S with Phe65, possibly through 
- interactions of the two aromatic rings, which is permitted by the absence of the bulky I atom at position 5.
More detailed studies of the effects of various differences in amino acid sequences between the human and dog D1 enzymes on substrate specificity have led to further insights into structural-activity correlations. Despite the roughly 20-fold higher Vmax/Km ratio for 5' deiodination of rT3 by human than dog D1, the Vmax/Km ratios for IRD of T3S and 3,3'-T2S were comparable for the two D1 enzymes (116). This indicated the major decrease in catalytic activity toward rT3, due primarily to the Phe65Leu substitution in canine D1, does not affect IRD of these sulfated substrates. Although the reinsertion of the missing TGMTR peptide (residues 48–52 of human or rat D1) into dog D1 does not enhance rT3 deiodination, it causes a marked decrease in the ORD of T2S. This could be due to interference with the interaction of the SO4 group with the active center. Nonetheless, taken together, these results lead to the surprising conclusion that these five residues are not critical to D1 function.
There are four histidine (His) residues in rat D1. Early studies showed that modification of one or more of these by diethylpyrocarbonate or rose bengal caused marked inhibition of deiodination (117). Systematic site-directed mutagenesis of these residues showed that His158 is critical for normal enzyme structure, whereas mutagenesis of His174 to glutamine or asparagine causes a 20- to 100-fold increase in the Km for rT3 (118). Subsequent comparisons of the dog, mouse, and human D1 proteins with the rat D1 shows that only these two His residues (158 and 174) are conserved in all four species (11, 70, 73).
The necessity for these indirect approaches to structure-activity correlations reflects the fact that the selenodeiodinase enzymes are integral membrane proteins, and thus, their crystallization in an active form is quite challenging. It is possible, for example, that the role of His174 is to maintain the reducing environment for the Se active center, a conclusion that is not apparent from inspection of its linear sequence.
4. Regulation of D1 synthesis.
a. Thyroid hormone.
There are a number of substances, agents, or conditions that can influence the rate of D1 synthesis, the most potent being thyroid hormone (Table 3 and Refs. 11 , 70 , and 119, 120, 121, 122). Thyroid hormone-induced increases in D1 activity and/or mRNA are well documented in rats, mice, and humans (11, 121). This is due to increased transcription, which in the human Dio1 gene can be attributed to the presence of two thyroid hormone response elements (TREs) in the 5'-flanking region (FR) of the gene (Fig. 10 and Refs. 123, 124, 125). One of these, TRE-2, is a typical direct repeat with 4 bp separating the RXR-T3 TR binding half-sites (DR+4). It is formed due to a polymorphism in an Alu sequence and is present 660 nt 5' to the transcription start site [TSS (125)]. TRE-1 is an unusual element in which two TR-binding octameric half-sites are separated by 10 bp. Both of the octamers binding TR have a pyrimidine at their most 5' position, this being the highest affinity TR-binding DNA half-site (123). Both TREs contribute to the response of the human Dio1 promoter, and methylation interference binding studies show that the unconventional TRE-1 binds TR but not RXR (123). Studies in TR-knockout mice indicate that TRß is primarily responsible for T3-mediated D1 stimulation (126). Given these results, one would expect that the T3 responsiveness of the human Dio1 gene would be most obvious in patients with thyrotoxicosis, such as in Graves’ disease. In fact, semiquantitative PCR of D1 mRNA in human peripheral blood mononuclear cells demonstrates it is increased in proportion to the degree of hyperthyroidism (84). As discussed below, this can explain the marked increase in PTU-sensitive plasma T3 production in patients with hyperthyroidism (127).
|
|
b. RA.
RA increases the concentration of D1 in human thyroid carcinoma cell lines (129). This can be accounted for by the TREs in the human Dio1 gene that also respond to RA (Fig. 10 and Refs. 124 , 125 , and 130).
c. Glucocorticoids.
Acute administration of glucocorticoids to humans or rats decreases the ratio of circulating T3 to T4, implying that these agents block T4-to-T3 conversion (131, 132). Hepatic D1 in liver homogenates decreases in dexamethasone-treated rats, but in spheroid cultures of primary rat hepatocytes, glucocorticoids enhance the induction of the D1 mRNA induced by T3. This occurs despite the fact that dexamethasone alone causes only a modest increase in D1 activity (133). The dexamethasone-induced increase in D1 mRNA was blocked by pretreatment of the cells with cycloheximide, indicating that ongoing protein synthesis is required for this effect. In rats, the fall in T3 that follows the administration of dexamethasone may be explained by a decrease in the plasma T3 production rate and in the fractional conversion of T4 to T3 (134, 135). However, more recent studies in humans indicate that D3 activity is induced by dexamethasone, and the acute decrease in serum T3 that follows a high dose of glucocorticoids may be due to an increase in D3-mediated T3 clearance via 5 deiodination (136).
d. Gonadal steroids.
Although no direct studies of effects of gonadal steroids on D1 activity have been performed, D1 activities are higher in male than in female rat liver, and this difference is eliminated by gonadectomy (137, 138). However, there are no gender-related differences in D1 content in the kidney.
e. GH.
Treatment of euthyroid adults with GH increases the ratio of plasma T3 to T4 and reduces that of rT3 to T4 (139). The mechanism for this is peripheral because it is found in T4-replaced individuals with central hypothyroidism. It could be a consequence of enhanced D1 activity or due to a reduction in D3, analogous to the effect of GH to reduce D3 activity in chicken liver (140, 141).
f. cAMP.
Studies in the FRTL5 rat thyroid cell line have shown a 3-fold increase in D1 mRNA induced by TSH, which is replicated by (Bu)2cAMP or forskolin. The effects of these agonists were additive to that of T3, the combination resulting in a 5-fold stimulation relative to control (142). This could not be explained by an alteration in D1 mRNA disappearance rate, and the effect was blocked by cycloheximide, indicating that persistent protein synthesis is required for the effect. The mechanism for the stimulation of rat Dio1 transcription by cAMP has not been elucidated.
g. Cytokines.
IL-1, IL-6, TNF, and other cytokines have been postulated as potential mediators of the alterations in thyroid function that occur during severe illness (143, 144, 145). TNF, IL-1ß, and interferon 
decrease D1 activity and mRNA in FRTL5 cells, although TGFß has no effect (142). The effects of TNF have been examined in hepatocytes and HepG2 cells with contradictory results. TNF decreased the T3-stimulated D1 mRNA in HepG2 cells (146). This effect is blocked by dominant-negative nuclear factor 
B (NF-B) coexpression and also by inhibition of the TNF-induced activation of NF-B by clarithromycin, suggesting that it is related to the TNF-induced increase in NF-B. NF-B impairs the function of a number of hormonal ligand-directed transcriptional stimulators, although no direct interaction between TR and NF-B has been demonstrated. In a second study in dispersed rat hepatocytes, IL-1ß and IL-6 blocked the T3 induction of D1 mRNA and activity but TNF had no effect (147). The T3 effect with IL-1ß was rescued by coexpression of the nuclear steroid receptor coactivator (SRC-1) but not by cAMP response element binding protein-binding protein or cAMP response element binding protein-binding protein-associated factor. Because IL-1 does not affect the amounts of SRC-1 in the hepatocytes, the effect was attributed to competition between IL-1 and T3-stimulated transcriptional events for limiting quantities of SRC-1. This was supported by evidence that IL-1 and IL-6 reduced T3 induction of Spot-14 and malic enzyme mRNA as well as D1, and that SRC-1 coexpression also rescued these as well as IL-1-suppressed, glucocorticoid-induced mouse mammary tumor virus promoter activity.
The differences in the effects of TNF in these two studies could be due to differences in the experimental paradigms. Despite this, both studies suggest that one mechanism for an acute decrease in D1 expression during illness could be competition for limited amounts of one or more transcription factors that are rate limiting for both cytokine- and T3- dependent transcriptional events. IL-1ß stimulates D1 and D2 and TNF stimulates D2 in rat pituitary cells and GH-3 cells (148). If this occurs in the thyrotrophs, it could act to reduce TSH synthesis and release in severe illness.
h. Se deficiency.
The decrease in hepatic D1 activity in liver of Se-deficient rats and the demonstration that D1 could be labeled with 75Se were the first clues that this trace element was critical to the function of this enzyme (12, 13, 14, 149). However, the effects of Se deficiency are complex due to a combination of factors. First, Se retention during dietary deficiency differs among different tissues high in brain, pituitary, thyroid, adrenals, and gonads. In contrast, dietary Se deficiency rapidly reduces the Se content of plasma, liver, skeletal muscle, and heart (150, 151, 152). Thus, the effect of Se deficiency on the synthesis of intracellular selenoproteins, such as the selenodeiodinases and GPX, will depend on the tissue being examined. For example, in rats with Se deficiency, thyroidal D1 activity is preserved, whereas that in the liver drops precipitously (152). In the intact rat, Se deficiency is generally associated with an increase in serum T4 concentrations but little change in serum T3 concentrations, effects that are analogous to the situation in the D1-deficient C3H mouse (153). Se deficiency also decreases D1 in kidney but this is accompanied by a decrease in D1 mRNA, which does not occur in the liver (154). Se deficiency is observed in patients receiving diets that are restricted in protein content, such as those given for phenylketonuria, and has also been found in elderly patients (155, 156, 157, 158). In Se-deficient humans, the serum T4 and T4 to T3 ratios are mildly elevated, but TSH is normal. In one endemic goiter region in Africa, there is an accompanying Se deficiency (159, 160). When Se was resupplied to these iodine-deficient individuals, there was a deterioration of thyroid function as evidenced by an increase in TSH and a reduction in serum T3, suggesting that the reduction in D1 during Se deficiency can protect against iodine deficiency, presumably by reducing the IRD of T4, T3, or T3S (161, 162).
There have been numerous studies of the effects of Se deficiency on thyroid status in rats, as researchers have attempted to determine the role of hepatic and renal D1 in plasma T3 production. There is general agreement that hepatic D1 is markedly reduced by Se deficiency but that thyroid and pituitary D1 are not (12, 13, 150, 152). Unexpectedly, there is little change in serum T3 despite 10–20% increases in serum T4 in intact animals, and serum rT3 and T3S are generally increased (163, 164, 165). These results argue that hepatic and renal D1 make minimal contributions to plasma T3, but the results are confounded by the roughly 40% contribution of thyroidal T3 secretion to this pool in the rat (see Section V.B).
To eliminate this problem, the effects of Se deficiency have also been examined in thyroidectomized T4-replaced rats. Such studies are analogous to those performed with PTU in which approximately 50% inhibition of extrathyroidal T4-to-T3 conversion is found (6, 166, 167, 168). The results of the Se-depletion studies are conflicting. In one, Se deficiency caused no decrease in plasma T3 despite a greater than 93% decrease in hepatic D1 (163). This result led to the conclusion that the thyroid gland must be the major source of circulating T3 in the rat. However, that conclusion did not take into account the contribution of D2-catalyzed T4-to-T3 conversion in tissues resistant to Se depletion to plasma T3 (169). In a later study in T4-replaced rats, an approximately 25% decrease in serum T3 and 32% decrease in total T3 production in Se-deficient rats was found, similar to the effects of PTU in the same study (170). The reasons for the more modest decreases in the serum T3 during PTU treatment in this study than the 60% typically observed are not clear. The overall conclusion of these studies is that thyroidal T3 secretion provides about 40% of the plasma T3 in the rat and that approximately 50% of extrathyroidal T4-to-T3 conversion is catalyzed by D1. This is consistent with estimates of approximately equal contributions of D1 (PTU sensitive) and D2 (PTU resistant) to extrathyroidal T3 production in rats generated by sophisticated kinetic techniques (169).
i. D1 expression is reduced in fetal tissues.
It is well recognized that the serum T3/T4 ratio in the fetus and newborn is quite low relative to infants even a few hours older (171). This is likely due to the high hepatic D3 expression in the human fetus, together with placental D3 expression. Hepatic D3 expression disappears in late fetal life (172). The abrupt increase in T3 levels that occurs in the first hours after delivery and the higher ratio of T3 to T4 that is maintained thereafter is probably, then, due to a combination of factors: a rapid increase in TSH inducing both T3 and T4 secretion from the thyroid, and the absence of the placenta. There may also be increases in D1 and D2 (173). Changes in deiodinase activity have been investigated in the developing rat, and in general, D1 activity is low in all tissues of the fetal rat. It begins to appear soon after birth in the intestine, liver, kidney, cerebrum, cerebellum, and gonads (174). D1 activity is higher in the skin of the newborn rat than in the 2-wk-old or adult rat, in which it is virtually undetectable. D1 is the major deiodinase activity in liver, kidney, and intestine at all stages of life in the rat, and these tissues presumably are the most active in the PTU-sensitive conversion of T4 to T3. Because the age-related differences are also apparent using RT-PCR measurements of rat D1 mRNA, they arise at a pretranslational level. The mechanism for the age-related effects on D1 expression is unknown. The physiological purpose served by the low D1 activity in the fetus is thought to be to reduce circulating T3, thus permitting the changes in intracellular T3 to be determined by the developmentally programmed changes in D2 and D3 activities (Ref. 173 ; see Section V).
j. Nutritional influences on D1 expression.
A decrease in the concentration of circulating T3 relative to that of T4 and an increase in rT3 concentrations in fasting humans was one of the earliest indications that the peripheral metabolism of thyroid hormones in humans could be modulated by physiological or pathophysiological events (175). Similar changes are apparent in the acutely ill patient (176, 177). Thyroidal secretion accounts for only about 20% of T3 production in humans (2). Therefore, the acute reduction in serum T3 during fasting or illness to less than 50% of its baseline concentration must derive, at least in part, from impaired T4-to-T3 conversion by D1 or D2 or by an increased T3 clearance by D3. Analysis of the effects of illness and fasting are discussed further in Section VI, but data in studies with rodents are reviewed here.
Early studies of D1 activities in livers from fasted rats suggested that an impairment of T4-to-T3 conversion might be a consequence of a reduction in the thiol cofactor, which serves as the cosubstrate for D1 catalyzed T4-to-T3 conversion (120, 178). Despite three decades of research, this cofactor has not been identified. The rat has been used extensively as a model for the effects of fasting and illness on T4-to-T3 conversion in humans. Unfortunately, the young adult rat is a poor model for the effects of starvation in humans because of the low body fat content. Serum T3 does fall rapidly in the 8-wk-old fasted rat but so also does serum T4 (179). Surprisingly, despite reduced D1 in rat liver, the total body conversion of T4 to T3 in the rat is not reduced by fasting (180, 181). It is also well recognized that fasting is a severe stress to the 8-wk-old rat, which rapidly loses protein nitrogen during the first few days and succumbs after approximately 5 d (179). This is associated with a marked reduction in serum TSH, T4, and T3 concentrations, i.e., central hypothyroidism probably in part due to leptin deficiency (182). In contrast, if 16-wk-old rats with larger fat stores are fasted, the nitrogen loss due to protein catabolism is delayed and serum T4 falls less rapidly and to a lesser degree, allowing studies up to 10 d of starvation. Even more impressive is the effect of prefeeding rats with high-fat diets to induce obesity before starvation. Under these circumstances, urinary nitrogen loss is markedly reduced during the period of fasting, serum T4 and T3 concentrations fall less then 30% over the first 5 d, and serum T3 concentrations actually increase somewhat between 10 and 20 d of starvation (179).
This pattern differs markedly from that observed in starved humans, in whom circulating T3 concentrations decrease rapidly to about 50% of control and remain low for up to 3 wk of fasting (183). Thus, not only does the central hypothyroidism of the acutely fasted, nonobese rat not resemble the pathophysiology of the human, it appears that even when this is prevented by providing increased fat stores to this normally lean animal, the pattern of changes in serum T3 and T4 in the circulation do not match those seen in humans. One must conclude then that studies in the rat (and probably mouse) are not likely to shed much light on the acute pathophysiological changes in thyroid hormone deiodination in fasting humans.
The reduction in D1 activity in the liver of the fasted rat is in part due to a decrease in D1 protein at a pretranslational level (184, 185). This can be prevented by maintaining a euthyroid status in these animals and, thus, presumably reflects the effect of the stress-induced central hypothyroidism. Similarly, the reduced serum T3 concentrations in the diabetic rat can also be explained on the basis of decreased D1 mRNA in both kidney and liver but, in this case, the effect is reversed by insulin administration (185).
5. Regulation of D1 inactivation/degradation.
Studies with protein synthesis inhibitors have indicated that the half-life of D1 in intact or transiently transfected cells is greater than 12 h (186, 187, 188). The inactivation and subsequent degradation of D1 is enhanced by substrates such as iopanoic acid or rT3 (186). Substrate-induced inactivation is blocked by PTU, indicating that deiodination is required for the effect (187). The substrate-induced inactivation process consists of two phases. The early phase can be reversed by incubation of microsomes with high DTT concentrations. If longer times, e.g., 6 h, are allowed to pass between substrate exposure and incubation with DTT, recovery of D1 activity is much less complete, indicating that an irreversible change has occurred (187). Studies with transiently expressed D1 tagged with a FLAG epitope confirm the substrate-induced acceleration of D1 inactivation. There was no associated decrease in D1 protein, nor was there ubiquitination of D1 such as occurs with D2 (188). It is not certain whether the inactivated D1 can be reactivated in vivo. If it cannot, maintenance of D1 activity would require continued synthesis of D1.
B. Type 2 iodothyronine deiodinase (D2)
D2 is an obligate outer ring selenodeiodinase that catalyzes the conversion of T4 to T3 and rT3 to 3,3'-T2. D2 has a Km for T4 in the nanomolar range under in vitro conditions in the presence of 20 mM DTT. The Km in vivo is similar, given results in HEK293 or COS cells transiently expressing human D2 (189). As the most recently cloned of the three deiodinases, our knowledge as to its properties and function is still accumulating rapidly. For example, D2 was known to be particularly important in the brain, producing more than 75% of the nuclear T3 in the rat cerebral cortex (190). The presence of D2 activity in human skeletal muscle, unexpected from studies in rats, provides a plausible source for a significant amount of the extrathyroidally generated plasma T3 (110). Earlier results suggesting an important posttranslational regulation by substrate have been explained by the demonstration that D2 undergoes selective proteolysis via the ubiquitin-proteasome pathway. This pathway is markedly accelerated by interaction with substrate (188, 191, 192). The identification of the mouse Dio2 gene has led to the generation of the first deiodinase-knockout mouse, allowing potential new insights into the physiological role of D2 (24).
1. Dio2 gene structure, chromosomal localization, mRNA and protein characteristics, and tissue distribution.
a. Gene structure and chromosomal localization.
The Dio2 gene is present as a single copy located on the long arm of the 14th human chromosome in position 14q24.3 (193, 194). It is about 15 kb in size, and the coding region is divided into two exons by an approximately 7.4-kb intron. The exon/intron junction is located in codon 75 and is found at an identical position in the human and mouse Dio2 genes (20, 193, 195). For the human gene, there are three TSS, 707, 31, and 24 nt 5' to the initiator ATG. The longest 5' UTR of the human D2 mRNA contains an approximately 300-bp intron that can be alternatively spliced (Fig. 11 and Ref. 195). Other splicing variants have also been identified involving the coding region (196). The human, mouse, and rat Dio2 5'-FR have been isolated and functionally characterized. All contain a functional cAMP responsive element (CRE), but only human Dio2 has thyroid transcription factor-1 (TTF-1) binding sites (Fig. 12 and Refs. 195 , 197 , and 198).
|
|
The rat and human D2 mRNA are approximately 7.5 kb, and the chicken D2 mRNA is approximately 6 kb (17, 18, 19, 110). The ap
