The Human Type 2 Iodothyronine Deiodinase Is a Selenoprotein Highly Expressed in a Mesothelioma Cell Line*

Types 1 and 3 iodothyronine deiodinases are known to be selenocysteine-containing enzymes. Although a putative human type 2 iodothyronine deiodinase (D2) gene ( hDio2 ) encoding a similar selenoprotein has been identified, basal D2 activity is not selenium (Se)-dependent nor has D2 been labeled with 75 Se. A human mesothelioma cell line (MSTO-211H) has recently been shown to have (cid:1) 40-fold higher levels of hDio2 mRNA than mesothelial cells. Mesothelioma cell lysates activate thyroxine (T 4 ) to 3,5,3 (cid:1) -triiodothyronine with typical characteristics of D2 such as low K m (T 4 ), 1.3 n M , resistance to propylthiouracil, and a short half-life ( (cid:1) 30 min). D2 activity is (cid:1) 30-fold higher in Se-supplemented than in Se-depleted medium. An antiserum prepared against a peptide deduced from the Dio2 mRNA sequence precipitatesa 75 Seproteinofthepredicted31-kDasizefrom 75 Se-labeled mesothelioma cells. Bromoadenosine 3 (cid:1) 5 (cid:1) cyclic


Se-labeled mesothelioma cells. Bromoadenosine 35 cyclic monophosphate increases D2 activity and 75 Se-p31
ϳ2.5-fold whereas substrate (T 4 ) reduces both D2 activity and 75 Se-p31 ϳ2-3-fold. MG132 or lactacystin (10 M), inhibitors of the proteasome pathway by which D2 is degraded, increase both D2 activity and 75 Se-p31 3-4-fold and prevent the loss of D2 activity during cycloheximide or substrate (T 4 ) exposure. Immunocytochemical studies with affinitypurified anti-hD2 antibody show a Se-dependent increase in immunofluorescence. Thus, human D2 is encoded by hDio2 and is a member of the selenodeiodinase family accounting for its highly catalytic efficiency in T 4 activation.
A group of three specific deiodinases monodeiodinate thyroxine (T 4 ) 1 to 3,5,3Ј-triiodothyronine (T 3 ), the active thyroid hormone, or 3,3Ј,5Ј-triiodothyronine, an inactive metabolite. The first of these enzymes to be cloned was the type 1 iodothyronine deiodinase (D1), which revealed a rare structural characteristic, i.e. the presence of the selenocysteine (Sec) codon, UGA, in the active center of the enzyme conferring an ϳ200-fold higher catalytic efficiency than sulfur in the deiodination reaction (1).
It also revealed a requirement for a stem loop structure in the 3Ј-untranslated region, the Sec insertion sequence (SECIS) element (2), which is found in all eukaryotic selenoprotein mRNAs described to date (3).
The most recently cloned deiodinase is the type 2 (D2), a propylthiouracil (PTU)-resistant, low K m (T 4 ) obligate outer ring deiodinase that catalyzes T 4 to T 3 conversion (4,5). The major open reading frame of the hD2 cDNA encodes a putative ϳ31-kDa protein with high similarity of the Sec-containing active center of the putative D2 enzyme with those of D1 and the other member of this group, type 3 iodothyronine deiodinase (D3) (6). Furthermore, a SECIS element has been identified in the extreme 3Ј-untranslated region of both the human and highly homologous chicken Dio2 genes (7,8). Human, mouse, rat, chicken, and frog D2 all contain a putative in-frame UGA codon in the deduced amino acid sequence of the active center (position 133 in hD2) and therefore all are considered to be selenoenzymes (8).
Despite this evidence, there is vigorous disagreement that D2 is a selenoprotein, with some investigators claiming that Dio2 encodes a "virtual" or artificial selenoprotein that is not expressed in humans or rats (9 -14). Three major arguments support this position, namely 1) the failure of Se deficiency to reduce D2 catalytic activity either in vivo or in vitro (9,15,16), 2) the inability to identify a 75 Se-labeled protein of the expected size in cells expressing D2 (13), and 3) the inability to identify immunoreactive protein by Western analysis, immunoprecipitation (IP), or immunocytochemistry using antibodies prepared against peptides deduced from the sequence of the putative D2 mRNA (4). An alternative scenario put forward is that D2 is a large protein complex (200 kDa) containing one or more 29-kDa substrate-binding subunits (p29), an ϳ60-kDa cAMP-induced activation protein, and one or more catalytic subunits (12). Rat p29 was identified in astrocytes by N-bromoacetyl-T 4 labeling (17)(18)(19), has no enzymatic activity, and is highly similar to Dickkopf-3 (20). Although many of these apparent discrepancies could be explained by low expression of a highly efficient enzyme, reservations as to the identity of D2 are expressed even in recent reviews of the subject (21,22).
A mesothelioma cell line, MSTO-211H, was recently shown to express large amounts of Dio2 mRNA by microarray analysis (accession number U53506), offering a potential system to resolve the issue of D2 identity (23).

EXPERIMENTAL PROCEDURES
Reagents-MG132 and lactacystin were obtained from Calbiochem (La Jolla, CA) and dissolved in Me 2 SO. MG132 is a reversible proteasome inhibitor. Lactacystin blocks proteasome activity by targeting the catalytic ␤-subunit and covalently inhibiting the chymotrypsin-and trypsin-like activities. T 4 , cycloheximide (CX), and 8-bromoadenosine 3Ј5Ј cyclic monophosphate (8-Br-cAMP) were from Sigma and were dissolved in 40 mM NaOH (T 4  Cell Culture and D2 Activity-Mesothelioma (MSTO-211H) and mesothelial (MeT-5A) cell lines were obtained from American Type Culture Collection (ATCC; Manassas, VA) and made available through Dr. Rihn. Cells were plated in 60-mm dishes and grown until confluence in RPMI or M-199, respectively, and supplemented with 0 -10% fetal bovine serum (FBS). Human embryonic kidney epithelial cells (HEK-293) cells were incubated in Dulbecco's modified Eagle's medium with 0 or 10% fetal bovine serum. Each experiment was performed with duplicate dishes for each condition, and control plates contained the respective vehicles, 0.2% Me 2 SO and/or 0.6 mM NaOH. At the appropriate times, cells were harvested and processed for D2 activity in the presence of 1 mM PTU as described previously (24). The results are reported as fmol T 4 deiodinated/mg protein⅐min Ϯ S.D.
Procedures for Tranfections-In some experiments D2 was transiently expressed in HEK-293. Cells were transfected by the CaPO 4 method as described previously (25) with plasmids containing wild type D2 (KD2-SelP), a D2 mutant in which the Sec-133 (but not Sec-266) was replaced by Cys (CysD2), or a CysD2 in which the SECIS element was deleted (CysD2⌬Xba).

75
Se Incorporation and D2 IP-This was performed as described previously (25) after cells were labeled with 4 -6 Ci of Na 2 [ 75 Se]O 3 / dish. The lysis buffer contained 25 mM Tris-HCl (pH 7.4), 300 mM NaCl, 1 mM CaCl 2 , 0.5% Triton X-100, 10 g/ml leupeptin, 0.2 units/ml aprotinin, 1 mM phenylmethylsulfonyl fluoride. The 2,000-rpm supernatant of each cell lysate was incubated for 16 h at 4°C with D2 rabbit antisera 85254 at final dilution of 1:100 (25). 30 l of protein G plus/protein A-agarose solution were added per tube and incubated under slow agitation for 2 h at 4°C. The IP pellets were washed once with lysis buffer and once with 25 mM Tris-HCl (pH 7.4), 140 mM NaCl, 1 mM CaCl 2 . Pellets were re-suspended in sample loading buffer and analyzed in 12% SDS-PAGE. In some experiments IP was processed in unlabeled cells to allow measurement of D2 activity.
Affinity Purification of D2 Antiserum-For the affinity purification of the antiserum 45618 (25) the keyhole limpet hemocyanin conjugate of the hD2 peptide (EVRSWLEKNFSKR; residues 253 to 265) used to immunize the rabbits was coupled to agarose resin using Amino-Link agarose gel (Pierce) following the manufacture's instructions. The column was regenerated by washing with 3 M sodium acetate (pH 4.5) and equilibrated with phosphate-buffered saline (PBS) prior to each cycle. The D2 antiserum (45618) was passed through the column five times and then washed 10 times with PBS. Bound antibodies were eluted with 3 M sodium acetate (pH 4.5), and eluates were dialyzed against PBS at 4°C for 2 days. This affinity-purified antiserum (AP-45618; 1:400) was used in the immunocytochemical and Western analysis of MSTO-211H or MeT-5A cells as described (26).
Immunocytochemistry and Confocal Analysis-This was performed as described previously (27). MSTO-211H and MeT-5A cells were grown directly in glass slides for 24 h, fixed with 3.7% paraformaldehyde in PBS (pH 7.4), permeabilized with 0.5% Triton X-100 in PBS for 10 min, and then blocked for 30 min with 1.0% anti-goat serum, 1.0% bovine serum albumin in PBS. Affinity-purified rabbit D2 antibody (AP-45618) was used at 1:500 followed by incubation with 1.25 g/ml goat antirabbit fluorescein isothiocyanate (Molecular Probes, Eugene OR). In the co-localization experiments, the affinity-purified anti-goat GRP78/BiP (dilution 1:20) antibody (Research Diagnostics, Inc., Flanders, NJ) was used simultaneously with anti-D2 antibody. This was followed by incubation with donkey anti-goat rhodamine F(abЈ) 2 fragment (Jackson ImmunoResearch Laboratories, West Grove, PA) and with goat antirabbit fluorescein isothiocyanate. No doublet is seen here, because this cDNA contains a Sec133Cys mutation, and the 75 Se labeling is restricted to Sec-266. CysD2⌬xba, which lacks a SECIS element, was used as a negative control. The film was exposed for 10 days. b, total cell lysates of MSTO-211 and MeT-5A cells incubated in media containing 10% FBS Ϯ Na 2 SeO 3 . The film was exposed for 1 week. c, IP pellets from the same samples shown in b. The film was exposed for 3 weeks. T 3 with kinetic properties typical of native D2, namely a K m for T 4 of ϳ1.0 nM and insensitivity to inhibition by 1 mM PTU as shown in the Lineweaver-Burk plot (Fig. 1A). Neither D1 nor D3 activities were detected nor did the transformed normal mesothelial cells (MeT-5A) express D2 activity (not shown). D2 activity in MSTO-211H cells was highly Se-dependent. Basal activity was decreased 4-fold by reducing the concentration of FBS containing endogenous selenoproteins (Fig. 1B). That these changes are specifically caused by Se deficiency and not by other components in FBS is evident from the fact that Se supplementation progressively increased D2 activity at each concentration of FBS, from 2-3-fold at 10% FBS up to ϳ30-fold when no FBS was present (Fig. 1B). The Lineweaver-Burk plots from cells incubated in the absence of FBS with or without 100 nM Na 2 SeO 3 for 24 h shows an ϳ16-fold increase in V max with no change in the K m (Fig. 1A). A dose-response curve of D2 activity versus Na 2 SeO 3 concentration indicated that the maximum stimulatory effect is obtained at 100 nM. A timeresponse curve showed that an effect of Se is present after 2 h, but the full effect requires 8 h (not shown). Similar effects of Se to increase D2 activity were observed in HEK-293 cells transiently expressing Sec-encoding D2 mRNA. Exposure to 100 nM Na 2 SeO 3 for 24 h increased D2 activity by ϳ10-fold suggesting this effect is primarily post-transcriptional (Fig. 1B).

Characterization
To confirm that the MSTO-211H cells express a selenoprotein containing a deduced D2 epitope, cells were labeled with Na 2 [ 75 Se]O 3 for 24 h and processed for D2 IP using an anti-hD2 peptide antiserum (85254) (25). Mesothelial (MeT-5A) cells were used as negative controls. Four 75 Se-labeled bands of ϳ70, 58, 25, and 18 kDa (but not 31 kDa) are present in the SDS-PAGE of total cell lysates of MSTO-211H or MeT-5A cells (Fig.  1C, b). These 75 Se-labeled proteins are increased by incubation with 100 nM Na 2 SeO 3 in MSTO-211H cells but are decreased in MeT-5A (Fig. 1C, b). SDS-PAGE analysis of the anti-D2 IP of the MSTO-211H cell lysate revealed an ϳ31-kDa 75 Se-labeled doublet band (p31) of the predicted size of hD2, the intensity of which is increased ϳ2-fold in the presence of 100 nM Na 2 SeO 3 (Fig. 1C, c and d). A doublet is seen when hD2 cDNA is transiently expressed because of the presence of a second UGA at the 3Ј-end of the open reading frame that can be translated either as a Sec or as a stop codon (28). The content of all other 75 Se-labeled proteins is reduced in the anti-D2 IP pellets (Fig.  1C, c). IP of non-labeled cell lysates reduced D2 activity ϳ20%, and ϳ 1 ⁄3 of this activity can be recovered from the washed IP pellet (Fig. 1D). Addition of the antiserum to the cell lysate did not inhibit D2 activity nor did exposure to preimmune serum or protein A-Sepharose (not shown). Despite these IP results, we were unable to consistently visualize D2 by Western blot analysis of MSTO-211H cell lysates using affinity-purified anti-D2 antibody (AP-45618).
Fluctuation in 75 Se-p31 Levels Parallel D2 Activity in MSTO-211H Cells-The hD2 gene contains a canonical cAMP-response element-binding protein-binding site about 90 base pairs 5Ј to the most 5Ј-transcription start site (29). D2 activity increases in cells treated for 6 h with 8-Br-cAMP in a dose-dependent fashion ( Fig. 2A). To test whether there is a correlation between 8-Br-cAMP-induced D2 activity and p31, cells treated with 1 mM 8-Br-cAMP were labeled with 75 Se and processed for D2 IP. The intensity of the 75 Se-p31 band increased ϳ2.5-fold in 8-Br-cAMP-treated cells (Fig. 2F). No effects of insulin (0.1-5 M) or tumor necrosis factor-␣ (0.01-10 ng/ml) on D2 activity were detected (data not shown).
Both endogenous and transiently expressed D2 have a shortlife (Ͻ1 h) that is reduced further by exposure to substrates such as 3,3Ј,5Ј-triiodothyronine or T 4 (24, 25, 30 -32). This is explained by ubiquitin conjugation and subsequent proteasomal degradation of the ubiquitinated D2 (25,26). Treatment of cells with 100 M CX for 1 h resulted in a rapid loss of D2 activity, compatible with a D2 half-life of Ͻ30 min (Fig. 2B). The decrease in D2 activity was prevented by the proteasome inhibitor, MG132 (10 M) (Fig. 2B). Furthermore, exposure to MG132 increased D2 activity up to ϳ4-fold, with similar results obtained with 10 M lactacystin (Fig. 2C). Remarkably, anti-D2 precipitable 75 Se-p31, but not other labeled proteins, was also specifically increased by treatment with MG132 (Fig.  2F). Exposure to substrate (0.01-10 M T 4 in 10% FBS) for 1 h caused a concentration-dependent reduction in D2 activity of up to 10-fold (Fig. 2D), which again was blocked by MG132 (Fig. 2E). This was accompanied by a reduction in the anti-D2 precipitable 75 Se-p31 (Fig. 2F).
MSTO-211H and MeT-5A cells were also analyzed by immunocytochemistry using an affinity-purified anti-D2 antibody (AP-45618). In MSTO-211H cells incubated in serum-free medium for 24 h, the cytoplasm showed mild staining for D2 (Fig. 3A). Treat- , and E (10 M T 4 ). Film was exposed for 3 weeks. This experiment was repeated twice with identical results. In panels A-E values are the mean Ϯ S.D. of two cell plates, and each experiment was repeated two to five times. ment with 100 nM Na 2 SeO 3 for 24 h substantially increased the intensity of the staining without changing the distribution pattern (Fig. 3B). No fluorescence was detected in MSTO-211H cells incubated with secondary antibody alone (Fig. 3C) or in Na 2 SeO 3 -treated MeT-5A cells incubated with affinity-purified anti-D2 antibody (Fig. 3D). To study the subcellular distribution of endogenously expressed D2, MSTO-211H cells were stained with D2 antibody (Fig. 3F) and co-stained with antibody raised against the endoplasmic reticulum resident protein GRP78/BiP (Fig. 3G), shown previously to colocalize with transiently expressed D2 in HEK-293 and neuroblastoma cells (27). Colocalization was confirmed by superimposition of confocal images (Fig. 3H).

DISCUSSION
The Se-supplemented MSTO-211H cells have the highest D2 activity reported to date in a human tissue (ϳ140 fmol/min/mg protein), ϳ40% higher than in Graves' thyroid tissue (33). This high D2 expression permits the unequivocal demonstration that endogenous hD2 is a selenoprotein encoded by the Dio2 gene (5). The present results satisfy a number of important criteria establishing the identity of hD2 as the product of the Dio2 gene. Furthermore, they allow confirmation that the endogenous human enzyme behaves as predicted from a number of studies using transiently expressed protein.
The first important criterion to establish D2 as a selenoprotein is to demonstrate its Se dependence. Past studies showed that the elevated D2 in cerebral cortex of thyroidectomized rats is not reduced by Se deficiency (15) and that astroglial cells depleted of Se for 3-5 days did not have lower basal D2 activity (9). Although there was about 50% reduction of the D2 activity in cAMP-treated primary astrocyte cultures after 7 days of Se deprivation, basal D2 activity was still not affected (34). However, the central nervous system is well known to retain Se with high efficiency, which could explain why D2 synthesis is not impaired by Se deficiency (35).
In MSTO-211H cells, on the other hand, Se depletion for 24 h reduced basal D2 activity by ϳ4-fold. This decrease in D2 activity was specifically reversed by Se supplementation (Fig.  1, A and B). In fact, addition of Se increased D2 activity by ϳ30-fold in Se-depleted cells, in a dose-and time-dependent fashion, a phenomenon that also occurs in HEK-293 cells transiently expressing D2 (Fig. 1, A and B). Selenium depletion of MSTO-211H cells also markedly reduces the levels of another selenoprotein, glutathione peroxidase (36).
A second criterion met by the present studies is the 75 Se labeling and immunological recognition of D2. A recent study could not identify the 75 Se-D2 protein by IP using antibodies that were raised against the COOH terminus of full-length rat D2 or against the catalytic core of the enzyme (13). This is not the case in the present studies (see Fig. 1C and Fig. 2F). Aside from potential differences in the antisera per se, which might be an issue, the D2 activity in rat tissue sonicates or stimulated primary astrocyte cultures are only ϳ10% of that in MSTO-211 cells. Because of the higher D2 expression, we can demonstrate IP of D2 activity, as well as of a 75 Se-labeled protein of the predicted size (31 kDa) from MSTO-211H cells (Fig. 1C). This is the first demonstration of endogenous 75 Se-D2. The 75 Se-labeled 31-kDa protein is increased by Na 2 SeO 3 , mirroring changes in D2 activity (Fig. 1C). A correlation between changes in D2 activity and the amount of 75 Se-p31 is also evident during treatment with proteasome inhibitors (see below), 8-Br-cAMP, or D2 substrate (Fig. 2). Additional immunological identity criteria were obtained by immunocytochemical studies using the affinity-purified anti-hD2 antibody. Staining was only present in MSTO-211H cells and was markedly increased by Se supplementation (Fig. 3,  A-D). Furthermore, D2 co-localized with the endoplasmic reticulum-specific marker BiP as found previously in HEK-293 and neuroblastoma cells transiently expressing hD2 (Fig. 3, E-H) (27). Despite successful IP of the D2 protein and activity, we were unable to visualize D2 by Western blot analysis presumably because of low antibody affinity for denatured protein, low D2 expression, or both.
A short half-life and the acceleration of endogenous D2 degradation by exposure of cells to substrates observed in pituitary tumor cells and in cells transiently expressing D2 are reproduced in these experiments. A short half-life (Ͻ1 h) is characteristic of D2 in all cells studied (11,24,31) and is explained by the fact that D2 is the target of selective ubiquitination and proteolysis by the proteasomal system (24 -26). This is also the case in MSTO-211H cells, in which D2 activity and 75 Se-D2 protein levels were extremely sensitive to inhibitors of the proteasomal pathway (Fig. 2, B-E). It is remarkable that in GH4C1 rat pituitary tumor and transfected HEK-293 cells, treatment with proteasomal inhibitors for 1 h increases D2 activity by only ϳ20% (24,25) whereas in the present study a similar treatment resulted in a 3-4-fold increase in activity (Fig. 2C). This can be explained by a more rapid turnover of the enzyme in mesothelioma cells. Furthermore, as in pituitary tumor cells and transfected HEK-293 cells, exposure to T 4 , the preferred D2 substrate, accelerates the turnover, increasing ubiquitination and subsequent proteolysis, a phenomenon that can be specifically inhibited by treatment with proteasomal inhibitors (Fig. 2, D-E). These results demonstrate that the targeting of D2 to the ubiquitin-proteasomal system and the acceleration of its degradation are characteristic of this protein, regardless of the cell type in which it is present or whether it is transiently expressed or endogenous.
In conclusion, the evidence developed in the present study unequivocally establishes that endogenous hD2 is a short-lived selenoprotein, the product of the human Dio2 gene, as well as the 90 -95% homologous frog, chicken, mouse, and rat proteins. These findings are also consistent with preliminary data reported recently that targeted disruption of the mouse Dio2 gene eliminates D2 activity in brain, pituitary gland, and brown adipose tissue (37). From a broader perspective it is interesting to note that this is the second example of high levels of a selenodeiodinase in human tumor cells derived from a tissue that normally does not express deiodinase, the other being the high levels of D3 in hemangiomas (38). It will be of interest to determine whether primary mesotheliomas also express high levels of D2 and what, if any, effects of increased rate of T 4 to T 3 conversion might have in these cells.