Cloning, Expression, and Functional Characterization of the Substrate Binding Subunit of Rat Type II Iodothyronine 5′-Deiodinase*

Type II iodothyronine 5′-deiodinase catalyzes the bioactivation of thyroid hormone in the brain. In astrocytes, this ∼200-kDa, membrane-bound enzyme is composed of at least one p29 subunit, an ∼60-kDa, cAMP-induced activation protein, and one or more uniden- tified catalytic subunit(s). Recently, an artificial type II-like selenodeiodinase was engineered by fusing two independent cDNAs together; however, no native type II selenodeiodinase polypeptide is translated in the brain or brown adipose tissue of rats. These data suggest that the native type II 5′-deiodinase in rat brain is unrelated to this artificial selenoprotein. In this report, we describe the cloning of the 29-kDa subunit (p29) of type II 5′-deiodinase from a λzapII cDNA library prepared from cAMP-induced astrocytes. The 3.3-kilobase (kb) cDNA encodes an ∼30-kDa, 277-amino acid long, hydrophobic protein lacking selenocysteine. Northern blot analysis showed that a 3.5-kb p29 mRNA was present in tissues showing type II 5′-deiodinase activity such as brain and cAMP-stimulated astrocytes. Domain-specific, anti-p29 antibodies specifically immunoprecipitated enzyme activity. Overexpression of exogenous p29 or a green fluorescence protein (GFP)-tagged p29 fusion protein led to a >100-fold increase in deiodinating activity in cAMP-stimulated astrocytes, and the increased activity was specifically immunoprecipitated by anti-GFP antibodies. Steady-state reaction kinetics of the enzyme in GFP-tagged p29-expressing astrocytes are identical to those of the native enzyme in brain. Direct injection of replication-deficient Ad5-p29GFP virus particles into the cerebral cortex of neonatal rats leads to a ∼2-fold increase in brain type II 5′-deiodinating activity. These data show 1) that the 3.3-kb p29 cDNA encodes an essential subunit of rat type II iodothyronine 5′-deiodinase and 2) identify the first non-selenocysteine containing subunit of the deiodinase family of enzymes.

The enzyme-catalyzed deiodination of thyroxine (T 4 ) 1 gener-ates the bioactive thyroid hormone, T 3 , and is the essential first step in the mechanism of thyroid hormone action. In the brain, type II iodothyronine 5Ј-deiodinase (D2) generates up to 75% of the T 3 found within the cell and plays a key role in regulating intracellular T 3 levels (1-8). Brain D2 is very short-lived in vivo and in cultured astrocytes with a t1 ⁄2 ranging from 10 to 20 min (9 -11), and cellular levels of the enzyme are dynamically regulated by both T 4 and 3,3Ј,5Ј-triiodothyronine (rT 3 ) but not T 3 (10,(12)(13)(14)(15).
D2 belongs to a family of membrane-bound deiodinating enzymes. D1 (EC 3.8.1.4) and D3 are composed of 27-to 30-kDa selenoprotein subunits encoded by different mRNAs. Each mRNA contains a bifunctional, in-frame UGA codon that signals either selenocysteine (SeC) insertion or translation arrest (stop codon) depending on the cellular content of selenium (16 -22) and the presence of a ϳ90-nucleotide-long stem loop sequence, the selenocysteine insertion sequence (SECIS), located in a 3Ј-untranslated region (UTR) (23)(24)(25). Unlike D1 and D3, the influence of selenium on D2 catalysis in vivo is unsettled; decreases in selenium intake that nearly eliminate D1 from the rat liver and kidney have only marginal effects on rat brain D2 activity (21, 26 -28).
The discovery that frogs lack the D1 enzyme and express a deiodinase similar to D2 led to the cloning of a ϳ1.5-kilobase (kb) cDNA D2-like selenodeiodinase (SeD2) with an SeC codon and a 3Ј-UTR-located SECIS element (29). Injection of in vitro synthesized SeD2 mRNA in frog oocytes produced an ϳ30-kDa selenoprotein with D2-like catalytic properties and modest amino acid homology to the mammalian D1 (26).
Mammalian homologs of the frog SeD2 cDNA were cloned from rat brown adipose tissue (BAT) (30) and the human thyroid gland (31), but both cDNA clones did not produce functional enzymes due to the lack of the essential SECIS. However, appending a heterologous SECIS to the 3Ј-UTR of these inert clones generated artificial cDNA chimeras (SeD2 SECIS ) that yielded catalytically active ϳ30-kDa selenoproteins in transient expression studies (30,31). These results were not unexpected, because the mammalian SeD2 cDNA(s) share significant homology with the catalytic core of D1 (30,31) and adding a SECIS to the 3Ј-UTR of any mRNA facilitates SeC incorporation at in-frame UGA(s) (32,33). Expressed sequence * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF245040.
Direct attempts to find the native 30-kDa SeD2 polypeptide in rat tissues that express D2 failed, even though a functional SeD2 selenoprotein was easily identified in SeD2 SECIS -transfected C6 astrocytoma cells (32). Instead of a full-length SeD2, the native 7.5-kb SeD2 transcript encodes a catalytically inert, 15-kDa polypeptide of unknown function (32), indicating that the UGA triplet in the native SeD2 mRNA signals translation arrest rather than SeC insertion. These findings suggest that D2 activity in rat brain is unrelated to SeD2 (32).
In this report we describe the cloning of the cDNA (GenBank accession number AF245040) encoding the p29 subunit of the native D2 from cultured rat astrocytes. Expression of this cDNA both in cultured cells and in brain in vivo directly increases D2 activity. The cloned 3.3-kb cDNA (p29) is found in neurons in vivo and encodes an essential iodothyronine binding subunit of D2 with a deduced molecular mass of 30.7 kDa.

EXPERIMENTAL PROCEDURES
Materials-All reagents were of the highest purity commercially available. Restriction endonucleases and DNA-and RNA-modifying enzymes were purchased from New England BioLabs (Beverly, MA). A femtomole DNA sequencing kit was obtained from Promega (Madison, WI). The AdpREC shuttle vector and the replication-deficient Ad5-␤gal viral genome were gifts from T. Kowalik (University of Massachusetts Medical School); the replication-deficient Ad5-EGFP virus particles were a gift from X. Wang (Massachusetts General Hospital, Boston). [␣-35 S]dATP (3000 Ci/mmol) and [␣-32 P]dCTP (800 Ci/mmol) were purchased from NEN Life Science Products. [3Ј or 5Ј-L-125 I]rT 3 (ϳ2200 Ci/mmol) and [3Ј or 5Ј-L-125 I]T 4 (ϳ2200 Ci/mmol) were prepared by radioiodination of 3Ј,3-L-diiodothyronine and 3,3,5-L-triiodothyronine, respectively, as described previously (10). Synthetic oligonucleotides were prepared in-house or purchased from Life Technologies (Grand Island, NY). The zapII cDNA library kit, picoBlue immunoscreening kit, and Duralose membranes were purchased from Stratagene (La Jolla, CA). All iodothyronines were of the L configuration and were purchased from Henning Berlin GmbH. Dulbecco's modified Eagle's medium, antibiotics, Hanks' buffered salt solution, glucose, and trypsin were obtained from Life Technologies; supplemented bovine calf serum from HyClone Laboratories; dibutyryl cyclic AMP (bt 2 cAMP) and hydrocortisone from Sigma; acrylamide and N,NЈ-methylenebisacrylamide from U.S. Biochemical Corp.; ammonium persulfate and TEMED from Bio-Rad; and dithiothreitol from Calbiochem. Rabbit anti p29 antisera and anti-SeD2 antisera were documented previously (32,37,38) and as detailed below.
Cell Culture and D2 Induction-Rat astrocytes were prepared from 1-day-old neonatal cerebral cortex as described previously (39). Cells were grown in growth medium composed of Dulbecco's modified Eagle's medium, containing 15 mM sodium bicarbonate, 15 mM HEPES (pH 7.2), 33 mM glucose, 1 mM sodium pyruvate, 10% (v/v) calf serum, 50 milliunits/ml of penicillin, and 90 mg/ml streptomycin, in a humidified atmosphere of 5% CO 2 and 95% air at 37°C. Astrocytes used for cDNA library construction were subcultured from primary dispersions by seeding 3 ϫ 10 7 cells onto four 625-cm 2 plates. Cells were grown to confluence, and D2 activity was induced with 1 mM bt 2 cAMP and 100 nM hydrocortisone for 24 h before harvest as described previously (10,11,40). Cells were harvested by scraping and collected by centrifugating at 500 ϫ g for 10 min. All other astrocyte cultures were subcultured every 7-10 days by seeding 5 ϫ 10 4 cells/cm 2 into 25-cm 2 flasks. D2 activity was induced by adding 1 mM bt 2 cAMP and 100 nM hydrocortisone for 16 h as described previously (10,11,40). Culture media was changed three times weekly.
Expression cDNA Library Construction and Screening-Total RNA was prepared from bt 2 cAMP-stimulated rat astrocytes by the method of Chomczynski and Sacchi (41), and poly(A ϩ ) RNA was isolated by double oligo(dT) selection. The cAMP-stimulated rat astrocyte cDNA library was constructed in zapII according to the manufacturer's instructions. The primary cDNA library contained 1.6 ϫ 10 6 independent members with cDNA inserts ranging from ϳ1.2 to 7 kb. Approximately 1 ϫ 10 6 plaque-forming units were screened with a polyclonal antibody raised against purified p29 (37,38) using the picoBlue immunoscreening kit (Stratagene). Cross-reacting antibodies were removed by preadsorption of the anti-p29 antisera with phage-infected, Escherichia coli lysates, and the cleared anti-p29 antisera were used at a final dilution of 1:100. Positive clones were detected with an alkaline phosphatase-conjugated goat anti-rabbit IgG.
DNA Sequencing-Double stranded DNA sequencing was done by the dideoxynucleotide method of Sanger (42) using cycle sequencing and iterative primers. All sequence information was confirmed by sequencing both strands.
Production of Domain-specific Anti-p29 Antibodies-Rabbit antibodies were raised against a 21-amino acid, synthetic peptide corresponding to the C terminus of the deduced amino acid sequence of p29 (NH 2 -YAQEMAFEEATPVDSLGGEKI-COOH, see Fig. 1B). An N-terminal tyrosine residue was included to facilitate diaminobenzidine coupling to keyhole limpet hemocyanin, and for radioiodination. This domain-specific antibody was used for all studies except the initial expression cloning of p29. Where indicated, domain-specific anti-p29 IgG was purified by affinity chromatography using a p29 peptide-Affi-Gel affinity matrix. In brief, primary antiserum was adsorbed to the p29 peptide-Affi-Gel 10 matrix at 25°C, and unbound proteins were removed by repeated washes with 150 mM NaCl, 20 mM sodium phosphate buffer (pH 7.4). Anti-p29 IgG was eluted with 100 mM HAc, and eluates were monitored by adsorbance at 280 nm. Fractions containing antibody were neutralized by adding 0.1 volume of 1 M TRIS (pH 8.6), pooled, and stored frozen (Ϫ70°C) at ϳ1 mg/ml IgG until use.
Immunoprecipitation of D2 Activity (Pull-down Assay)-Immunoprecipitation of catalytically active D2 was done at 4°C for 1 h in a total volume of 100 l. Precipitation reactions containing from 50 to 100 units of detergent-solublized enzyme, 100 mM HEPES buffer (pH 7.4), 100 mM NaCl, 10 mM octyl glycoside, 10 l of immobilized rProtein A beads (RepliGen, Cambridge, MA), and either preimmune rabbit antisera or anti-p29 antisera Ϯ excess blocking peptide. Immune complexes were removed by centrifugation, and the D2 activity remaining in the clarified supernatant was determined as described previously (38). One unit of D2 activity equals the release of 1 fmol of iodide/h.
Construction of p29 Expression Plasmids-Replication-deficient adenovirus constructs (Ad5-p29 and Ad5-p29 GFP ) were created using the coding region of p29 cDNA alone and a green fluorescence protein (GFP)-tagged p29 chimera (p29 GFP ) formed by appending GFP to the C terminus of p29. In brief, the 825-bp FspI-HinfI fragment (containing the p29 coding sequence) was excised from pBSK-p29, blunted with Klenow, gel-purified, and ligated into the EcoRV site of the AdpREC shuttle vector and ligated in-frame into the blunted HindIII site of pEGFP N1 (CLONTECH, Palo Alto, CA). The fusion construct p29 GFP cDNA was excised with BamHI and NotI and ligated, into a BamHI-NotI-ended AdpREC shuttle vector. Both shuttle constructs (Ad-pRECp29 and AdpRECp29 GFP ) were linearized with PvuI and individually cotransfected with ClaI-XbaI-linearized Ad5-␤gal cDNA into HEK293 cells using Lipofectin according to manufacturer's instructions. Recombinant Ad5 virus particles containing either p29 or the p29 GFP fusion cDNA were identified by monitoring plaque formation, by immunoblot analysis using either anti-p29 or anti-GFP antisera (CLONTECH), and/or by monitoring the number of p29 GFP -positive HEK293 cells. Replication-deficient Ad5 virus particles were purified from HEK293 cell lysates by CsCl gradient centrifugation and stored at Ն10 9 virus particles/ml at Ϫ70°C until use.
Immunocytochemistry-Cells were seeded onto poly-D-lysine (10 g/ ml)-coated glass coverslips (22 ϫ 22 mm) and grown for 1-4 days. cAMP-stimulated astrocytes expressing D2 activity were treated with 10 mM colchicine for 30 min to depolymerize the microtubular network and relax the cell borders. Cells were then fixed with 4% paraformaldehyde and permeabilized with 0.1% Triton (v/v) in phosphate-buffered saline. Where indicated, rat tissues were fixed with 4% paraformaldehyde in phosphate-buffered saline and embedded in paraffin, and ϳ6-m sections were prepared. P29 expression was identified using affinity-purified, domain-specific, anti-p29 antibodies for 2 h at 4°C (0.1-1 g/ml). The specificity of antibody labeling was determined using peptide-blocked IgG prepared by preincubating the antibody with a 100-fold molar excess of blocking peptide for 60 min at room temperature. Immune complexes were visualized with either Texas Red-conjugated rabbit IgG, or horseradish peroxidase-conjugated rabbit IgG as indicated, and the stained sections were examined with epifluorescence illumination (Texas Red) or standard illumination using a Zeiss Axioskop microscope equipped with an Olympus OM4 camera. Photomicrographs shown are representative of 20 -30 independent fields.
All experiments were performed at least three times. Statistical analysis was done by using Student's t test.

RESULTS
Cloning of the 29-kDa Substrate Binding Subunit of Brain D2 (p29)-Polyclonal anti-p29 antibodies (anti-p29) raised against the purified, affinity-labeled 29-kDa subunit of rat astrocyte D2 (37,38) were used to screen approximately 10 6 clones from a cAMP-stimulated astrocyte zapII cDNA library. Eight positive plaques were identified, all from a single mRNA species, and all contained a cDNA insert of ϳ3 kb. Clone 11.24, with a 3.3-kb insert, was used in all subsequent work. The nucleotide and deduced amino acid sequence of the ϳ3.3-kb p29 cDNA is shown in Fig. 1A, and a limited restriction map is depicted in Fig. 1B. Two consensus Kozak translation initiation sequences (43,44) are present; the first beginning at nucleotide 223 and the second at nucleotide 319, and yield open reading frames extending to nucleotide 1045, which encodes proteins of 277 and 245 residues, respectively, with molecular masses of ϳ30.5 and 27.2 kDa. Both deduced proteins have pIs of 4.5.
Cell-free translation under context-dependent translation initiation conditions (43,44) was done to confirm the assignment of the initiator methionine. Shown in Fig. 2 are the p29 immunoreactive polypeptides programmed by cell-free translation of in vitro synthesized, full-length p29 mRNA. As expected, a doublet of ϳ30 and 27 kDa translation products was synthesized confirming the assignment of nucleotide 223 as the most 5Ј authentic translation start site.
Tissue Distribution of the p29 mRNA and Cellular Localization of p29 Polypeptide-Northern blot analysis was done to determine the tissue distribution of the p29 mRNA. As shown in Fig. 3, the 3.5-kb p29 mRNA was expressed in rat tissues that contain D2 activity such as brain (B) and bt 2 cAMP-stimulated astrocytes (A), but not in tissues lacking this isozyme such as liver (L), kidney (K), and skeletal muscle (M). Consistent with earlier work that showed that p29 was present in unstimulated astrocytes lacking D2 activity (37), the p29 mRNA was also present in unstimulated astrocytes at levels similar to those found in the cAMP-stimulated cells.
The tissue distribution of native p29 polypeptide and its relationship to catalytically active D2 was then determined using a domain-specific, anti-p29 antibody raised against the deduced sequence of the C terminus of p29. As shown in Fig. 4, immunoblots of untreated astrocytes (A), bt 2 cAMP-stimulated astrocytes (Aϩ), and cerebral cortex (CC) revealed a doublet of specific, immunoreactive p29 protein, indicating that both AUG codons initiate translation in cell culture and in vivo. To establish the relationship between the cloned p29 and native D2, domain-specific, anti-p29 antisera were used to deplete p29 from detergent-soluble D2 preparations and measured the antibody-dependent loss of p29 on D2 activity. Detergent-soluble D2 activity was prepared from cAMP-stimulated astrocytes and SeD2 cells, and from microsomal fractions isolated from the BAT and the cerebral cortex of hypothyroid rats and incubated with anti-p29 antisera (1:100 dilution) Ϯ 100-fold molar excess of blocking peptide. The quantity of residual, non-adsorbed D2 activity was then determined. As shown in Fig. 5, immune depletion of p29 resulted in the loss of 60 -95% (p Ͻ 0.02 for glial and BAT, p Ͻ 0.002 for brain versus the peptideblocked controls) of the D2 activity from all native enzyme preparations. No antibody-dependent loss of D2 activity was observed from the extracts prepared from the engineered SeD2 cell. In control experiments, Ͼ98% of the soluble D2 activity remained when excess blocking peptide was present, indicating the anti-p29 antisera did not directly inhibit the catalytic reaction. These findings establish that the domain-specific, anti-p29 antisera specifically recognizes the native, catalytically active D2 and confirm that p29 is a subunit of the native D2.
With the specificity of the anti-p29 antibody documented by immunoblot and immunoprecipitation studies, the presence of p29 in D2-containing tissues was evaluated using affinity-purified, anti-p29 IgG. As expected, abundant, specific p29 immu- nostaining was found in all tissues expressing high levels of native D2 such as hypothyroid BAT, hypothyroid cerebral cortex, and the bt 2 cAMP-stimulated astrocyte. Only nonspecific staining was observed in the SeD2 cell (32), an astrocytoma cell line constitutively expressing an engineered SeD2 SECIS chimera (Fig. 6). Immunohistochemical staining of rat brain slices showed that p29-positive cells were scattered throughout the cerebral cortex and that the immunoreactive protein was localized to neurofilament-positive (data not shown), multiprocessed cells with a morphology typical of neurons in both euthyroid and hypothyroid brain (Fig. 7). Unlike cultured astrocytes, glial fibrillary acidic protein-positive cells in the cerebral cortex lacked immunoreactive p29. These data suggest that p29 is primarily expressed in neurons in vivo, and are consistent with earlier work done in dispersed brain cell cultures (45).
Effects of Overexpression of Exogenous p29 on D2 Activity in Cell Culture and in Vivo-Because earlier work (37) showed that the generation of catalytically active D2 in astrocytes requires at least two components, p29 and one or more cAMPinduced proteins, we used two replication-deficient adenovirus constructs (Ad5-p29 and Ad5-p29 GFP ) to introduce exogenous p29 in rat astrocytes and examined the effects of p29 overexpression on D2 activity. As expected, in the absence of cAMP stimulation, overexpression of a p29 GFP chimera yielded abundant GFP-positive cells (data not shown) but no D2 activity (Fig. 8A). However, after cyclic AMP stimulation, both the Ad5-p29-and the Ad5-p29 GFP -infected astrocytes showed a dramatic increase in D2 activity over that in control, Ad5-GFPinfected cells (Fig. 8A). To determine if the elevated D2 activity observed in p29 GFP -expressing astrocytes was due to synthesis of the GFP-tagged p29, anti-GFP antibodies were used to immune-deplete the fusion protein from detergent-soluble cell extracts, and the effects of the loss of the exogenous p29 GFP on D2 activity were examined. Anti-GFP IgG immunoprecipitated 76 Ϯ 12% of the p29 GFP fusion protein, as judged by immunoblot analysis (data not shown), and reduced D2 activity by a parallel 75 Ϯ 6% (Fig. 8B). On the contrary, the loss of GFP  Kozak (44). Newly synthesized proteins were immunoprecipitated with 1 g of affinity-purified, anti-p29 IgG, and immune complexes were collected on rProtein A-Sepharose (RepliGen, Cambridge, MA) beads. Immune complexes were eluted by heating to 100°C in 2ϫ Laemmli sample buffer containing 10 mM 2-mercaptoethanol. Eluates and the original translation reaction were resolved on 12.5% SDS-PAGE gels under reducing conditions and dried, and an exposure was made to Kodak X-Omat AR5 radiographic film for 24 h at Ϫ70°C.

FIG. 3. Northern blot analysis of the p29 content in selected rat tissues.
Total RNA was isolated from cAMP-stimulated astrocytes (Aϩ), unstimulated astrocytes (A), cerebral cortex (B), liver (L), kidney (K), and skeletal muscle (M) as described under "Experimental Procedures." Northern blot analysis was performed by standard techniques using 1.2% formaldehyde/agarose gels and 10 g of total RNA. Blots were washed to high stringency (30 mM NaCl, 3 mM sodium citrate, 1 mg/ml SDS) at 65°C for 15 min. Hybridization signals were visualized using Kodak X-Omat AR5 radiographic film. The 32 P-labeled p29 cDNA probe was prepared from the ϳ825-base pair FspI-HinfI fragment (nucleotides 218 -1043, see Fig. 1B 4. Immunoblot of native p29 in untreated (A), cAMP-stimulated astrocytes (A؉), and rat cerebral cortex (CC). 25-g aliquots of cell lysate and cerebral cortex protein were separated on a 12.5% SDS-PAGE gel under reducing conditions, transferred by electroblotting to Immobilon-P (Millipore, Bedford MA), and probed with a 1:1500 dilution (final) of domain-specific anti-p29 antisera, Ϯ10 g of blocking peptide. Immune complexes were identified using goat, antirabbit IgG conjugated to horseradish peroxidase and chemiluminescence detection (Lumiglo, Kirkegaard & Perry, Gaithersburg, MD).
FIG. 5. Immunoprecipitation of catalytically active D2 using domain-specific anti-p29 IgG. Detergent-soluble, catalytically active D2 was prepared as described under "Experimental Procedures." Clarified extracts were incubated, in triplicate, in a total volume of 100 l with 0.5 g of affinity-purified, domain-specific anti-p29 IgG, in the absence (□) or presence (o) of 10 g of C terminus (p29) blocking peptide, and 10 l of rProtein A-Sepharose beads. D2 activity was determined in the antibody-clarified extracts. Data are reported as the means Ϯ S.E., n ϭ 3. *, p Ͻ 0.02; §, p Ͻ 0.002 when compared with the ϩpeptide control. alone from detergent extracts of control, Ad5-GFP-infected, cAMP-stimulated astrocytes had no effect on D2 activity. To ensure that the p29 GFP fusion protein could be affinity-labeled with N-bromoacetyl-L-thyroxine (BrAcT 4 ), control, Ad5-GFPinfected cells, and Ad5-p29 GFP -infected cells were stimulated with cAMP and affinity-labeled with 0.2 nM BrAc[ 125 I]T 4 (4000 cpm/fmol) as described previously (38,46). Detergent extracts containing ϳ 2000 cpm of protein-bound BrAc[ 125 I]T 4 were then incubated with 1 g of anti-GFP IgG, and immune complexes were then collected on protein A beads. As shown in Fig.  8C, the p29 GFP fusion protein accounted for almost 40% of the protein-bound affinity label, whereas little, if any, affinity label was associated with GFP alone. Because 30 -40% of proteinbound BrAcT 4 is associated with native p29 in cell lysates of cAMP-stimulated astrocytes (38,46), these findings indicate that the D2 activity present in p29 GFP -expressing astrocytes is due to the exogenous p29 fusion protein.
As shown in Fig. 9 the quantity of functional D2 in p29 GFPexpressing cells was directly related to the quantity of fusion protein synthesized. Stepwise increases in the number of Ad5-p29 GFP virus particles added to the astrocyte monolayer led to a progressive increase in D2 activity and a proportional increase to the number of GFP-positive cells that was directly related to the quantity of immunoreactive p29 GFP expressed per cell. At multiplicity of infection (m.o.i.) Ն 6, more than 98% of the astrocytes expressed the p29 GFP fusion protein, and D2 activity was Ͼ100-fold that in uninfected controls.
We then examined the catalytic properties of D2 activity in p29 GFP -expressing astrocytes using steady-state reaction kinetics in the presence of 1 mM propylthiouracil with both rT 3 and T 4 as substrates (39,47). D2 activity in cAMP-stimulated astrocytes expressing the exogenous p29 GFP showed a converging set of lines (data not shown) consistent with the sequential, two-substrate reaction expected for this isozyme using rT 3 as the substrate (39,47). Limiting K m values determined from secondary replots of the data yielded a K a for rT 3 of 6.7 nM, and K b for dithiothreitol of 18 mM, that are in close agreement with those of the native D2 (39,47). The V f for the D2 activity in p29 GFP -expressing astrocytes was 59,000 units/mg of protein. T 4 was an excellent competitive inhibitor of rT 3 deiodination catalyzed by the D2 activity in p29 GFP -expressing astrocytes with a K i of 3.1 nM, in close agreement with previous results (39,47) and the K a for T 4 of 4.1 nM determined from secondary replots using T 4 as the substrate.
The very short, biological half-life of native D2 in brain is another unique characteristic of this enzyme (10,12,15). As detailed in Table I, the biological half-life of D2 in cycloheximide-blocked cells (10, 11) was 18 min in both cAMP-stimulated astrocytes expressing GFP alone (control cells) and in cAMPstimulated, p29 GFP -expressing cells, in close agreement with that determined previously for native D2 in cAMP-stimulated astrocytes (10,11). Assuming steady-state expression of the p29 GFP fusion protein, production rates of catalytically active D2 in the p29 GFP -expressing cells and in the GFP-expressing astrocytes were 1200 and 8 units/min, respectively, indicating a Ͼ150-fold increase in D2 synthesis in the p29 GFP -expressing astrocytes. Thus, two of the key distinguishing characteristics of the native D2 enzyme, (i) sequential reaction kinetics with K m for iodothyronine in the nanomolar range, and (ii) a short biological half-life, are also properties of the D2 activity resulting from overexpression of the p29 GFP fusion protein.
Effects of p29 GFP Expression on D2 Activity in the Cerebral Cortex of 12-Day-Old Rats-Finally, the relationship between p29 and native D2 activity in vivo was examined by introducing the GFP-tagged p29 in one cerebral hemisphere of neonatal rats using the replication-deficient Ad5-p29 GFP virus. Ad5-p29 GFP or control Ad5-GFP (ϳ10 6 virus particles) was injected into the left cerebral hemisphere of 4-day-old neonatal rats, and D2 activity was determined in homogenates of both the injected (left) and uninjected (right) hemispheres on day 12. As shown in Fig. 10, expression of p29 GFP (panel A) led to a 2-fold increase in native D2 activity in the left hemisphere (panel B). Neither p29 GFP (panel A) and nor increase in D2 activity (panel B) was found in either the contralateral cerebral hemisphere or in the cerebral hemispheres expressing the control GFP protein. These data confirm that p29 is an essential subunit of the native D2 enzyme in rat brain in vivo. FIG. 6. Immunocytochemistry of p29 in cultured cells and selected rat tissues. Cells and tissues were fixed and permeabilized, and the p29 protein was identified with 0.2 g/ml domain-specific, affinitypurified anti-p29 IgG Ϯ 10 g of C terminus (p29) blocking peptide. Immune complexes were visualized using a Texas Red-conjugated goat anti-rabbit IgG. Magnification, ϫ 400. FIG. 7. Immunocytochemistry of p29 in tissue slices of the rat cerebral cortex. Tissue slices of rat cerebral cortex (temporal lobe) were prepared as described under "Experimental Procedures" and as detailed in the legend to Fig. 6. Immune complexes were visualized using a horseradish peroxidase-conjugated, goat anti-rabbit IgG and 3,3Ј,5,5Ј-tetramethylbenzidine (Kirkegaard & Perry, Gaithersburg, MD). Bar ϭ 50 m.

DISCUSSION
Intracellular thyroid hormone deiodination regulates the levels of bioactive T 3 in the central nervous system. D2 serves as the source of T 3 in the brain and differs from the D1 isozyme in substrate specificity and inhibitor profiles (47-50), physiochemical properties (37,40,51), trace mineral requirements (21,27,32,(52)(53)(54), and regulation (3,8,10,14,37,55,56). In this report, we identify and characterize an essential substrate binding subunit of this key enzyme and the first non-selenocysteine-containing subunit of the deiodinase family of enzymes.
D2 was first identified by affinity labeling with BrAcT 4 , an alkylating analog of the D2 substrate (46). This substrate analog selectively labeled a 29-kDa protein (p29) with all of the properties of the substrate binding subunit of D2 (46). Both T 4 and rT 3 specifically block the selective labeling of p29 and the affinity label-dependent loss of D2 activity, whereas the product, T 3 , had no effect on either affinity label incorporation or BrAcT 4 -dependent inhibition of D2 activity (46). Rate inactivation studies confirmed that accumulation of BrAcT 4 by p29 led directly to D2 inactivation (46), indicating that p29 was an essential subunit of the enzyme (14,38,57).
Expression cloning yielded the p29 cDNA from an astrocyte cDNA library. This 29-kDa protein is encoded by a 3.5-kb mRNA that is found in all native tissues expressing D2 activity. Domain-specific, anti-p29 antibodies raised against synthetic peptides based on the deduced amino acid sequence of the C terminus of the p29 cDNA, selectively immune-depleted native D2 activity from detergent extracts of both the brain and BAT of rats. Importantly, cAMP stimulation was required be- FIG. 8. Effects of expression of exogenous GFP, p29, and p29 GFP on D2 activity in cultured astrocytes. A, effects on D2 activity in unstimulated (open bars) and cAMP-stimulated (filled bars) rat astrocytes. Triplicate flasks (25 cm 2 ) of confluent astrocytes were infected with 1 ϫ 10 6 replication-deficient Ad5-GFP, Ad5-p29, or Ad5-p29 GFP virus particles for 60 min. Culture medium was then replaced with growth medium, and the cells were grown for 24 h. As indicated in the solid bars, cells were stimulated with 1 mM bt 2 cAMP and 100 nM hydrocortisone for 16 h in complete growth medium. Cells were then harvested by scraping and collected by centrifugation, and D2 activity was determined, in triplicate, using ϳ25 g of cell lysate per assay tube. Data are reported as the means of closely agreeing (Ϯ10%) triplicate flasks. B, immunoprecipitation of D2 activity in cAMP-stimulated astrocytes expressing GFP or the p29 GFP fusion protein. Two 75-cm 2 flasks of confluent astrocytes were infected with 5 ϫ 10 6 Ad5-GFP or Ad5-p29 GFP virus particles and stimulated with 1 mM bt 2 cAMPas detailed above. Detergent extracts of the cell pellets were prepared as described under "Experimental Procedures," and ϳ50 g of soluble cell protein was incubated with 1 g of normal rabbit IgG (□) or 1 g of anti-GFP IgG (o) (CLONTECH) for 60 min at 4°C. D2 activity after removing the immune complexes was then determined as detailed in the legend to Fig. 5. Data are reported as means Ϯ S.E. (n ϭ 11). C, BrAc[ 125 I]T 4 labeling of GFP or p29 GFP in cAMP-stimulated astrocytes. Triplicate flasks (25 cm 2 ) of confluent astrocytes were infected with replication-deficient Ad5 constructs as described above. After cAMP stimulation, washed cell monolayers were incubated with 500 l of 0.2 nM BrAc[ 125 I]T 4 (4000 cpm/fmol) in serum-free growth medium for 20 min, washed free of unincorporated affinity label, and the cells were collected by centrifugation. Immune precipitation was done with 2000 cpm of protein-bound BrAc[ 125 I]T 4 and normal rabbit IgG (1 g) or anti-GFP IgG (1 g) as described above. After 60 min at 4°C, the Immunobeads were washed a total of 5 times by resuspension in 100 volumes of 100 mM NaCl, 100 mM HEPES buffer (pH 7.4) containing 10 mM octyl glycoside, and the final pellets were counted in a well type ␥ counter. □, normal rabbit IgG (1 g); o, anti-GFP IgG (1 g). Data are reported as the means Ϯ S.E. (n ϭ 3).  fore either exogenous p29 or the GFP-tagged p29 increased D2 activity in astrocytes. This finding is consistent with prior work showing that a ϳ60-kDa, cAMP-induced protein was required to produce a catalytically active enzyme (37) and suggests that composition of D2 is more complex than that of the other deiodinases. Both its short biological half-life and steady-state kinetic analysis confirmed that the enhanced D2 activity present in p29 GFP -expressing cells was identical to that of the native enzyme in vivo, and that overexpression of p29 GFP fusion protein directly increased D2 activity in the cerebral cortex of neonatal rats locally infected with a replication-deficient adenovirus carrying the p29 GFP fusion protein cDNA. This is the first demonstration of an increase in functional D2 activity in the brain resulting from the expression of a exogenous cDNA encoding a D2 subunit. The ability of the p29 cDNA to directly increase brain D2 activity, together with the properties of the D2 activity generated by expression of exogenous p29 in cAMPstimulated astrocytes confirms that this cDNA encodes an essential subunit of native D2 in rodents. The identification and molecular characterization of D2 has been a goal of many laboratories, because of its key role in maintaining levels of T 3 in the brain. In 1995, the cloning of a propylthiouracil-insensitive selenodeiodinase from frog skin (29) and the subsequent cloning of mammalian homologs (30, 31, 58) served as a turning point and shifted focus away from the BrAcT 4 -labeled p29 subunit of D2 (14, 15, 37, 38, 46, 57). Despite the promise of these clones, no native, full-length SeD2 protein has been found in any mammalian tissue (32), even though in situ hybridization localized the abundant SeD2 gene product to tanicytes (59) and to cerebrocortical astrocytes (60,61). Several attempts have been made to reconcile the differences between the properties of SeD2 and the BrAcT 4 affinitylabeled enzyme (31,(62)(63)(64). Courtin and coworkers (62) used BrAcT 4 labeling of p29 in astrocytes to re-evaluate the seleniumdependent nature of D2 (52,53), even though earlier work showed that SeD2 selenoprotein is poorly labeled by BrAcT 4 (58) and unrelated to p29 (32).
The cloning of the first non-selenoprotein subunit of D2 allows this polypeptide to be used for the characterization of the remaining D2 subunit(s), especially the 60-kDa cAMPinduced subunit required to assemble a catalytically active enzyme and to direct D2 to the plasma membrane (37). Together with the well characterized, T 4 -dependent intracellular trafficking of the native p29 polypeptide, the p29 GFP fusion protein provides a readily visualized reporter that can monitor intracellular trafficking of D2 in real time and allow the molecular events that regulate D2 levels in vivo to be characterized. Preliminary results indicate that the p29 GFP fusion protein shows thyroid hormone-dependent endocytosis 2 with properties identical to those described for the D2 activity and native p29. Thus, the cloning of the p29 subunit of native D2 provides a key enzyme polypeptide for the characterization of the non-genomic events mediating D2 regulation and the first non-selenocysteine member of the deiodinase family of enzymes.
FIG. 10. Effects of p29 GFP on brain D2 activity in vivo. A, quantity of immunoreactive p29 GFP and GFP in the cerebral cortex of 12day-old rats. Five 4-day-old rats received intracerebral injections (3 l, 1 ϫ 10 6 Ad5 virus particles) administered through the fontanels into the left cerebral hemisphere. Animals were killed on postnatal day 12, and homogenates were prepared of the left and right cerebral hemispheres. Detergent extracts were prepared as detailed in the legend to Fig. 5, and ϳ100 g of solubilized protein was separated on a 10% SDS-PAGE gel. The GFP-immunoreactive proteins were identified as described in