Protein-disulfide Isomerase Regulates the Thyroid Hormone Receptor-mediated Gene Expression via Redox Factor-1 through Thiol Reduction-Oxidation*

Background: Previously we found that overexpression of PDI suppressed the T3 response in GH3 cells. Results: Overexpression of wild-type PDI, but not C/A Mt, suppressed the T3 response. Interactions between Ref-1 and TRβ1 and Ref-1 and PDI were detected. Conclusion: PDI is involved in regulation of T3-mediated gene expression via Ref-1. Significance: Ref-1 contributes to TR function, and Ref-1 activity is regulated by PDI. Protein-disulfide isomerase (PDI) is a dithiol/disulfide oxidoreductase that regulates the redox state of proteins. We previously found that overexpression of PDI in rat pituitary tumor (GH3) cells suppresses 3,3′,5-triiodothyronine (T3)-stimulated growth hormone (GH) expression, suggesting the contribution of PDI to the T3-mediated gene expression via thyroid hormone receptor (TR). In the present study, we have clarified the mechanism of regulation by which TR function is regulated by PDI. Overexpression of wild-type but not redox-inactive mutant PDI suppressed the T3-induced GH expression, suggesting that the redox activity of PDI contributes to the suppression of GH. We considered that PDI regulates the redox state of the TR and focused on redox factor-1 (Ref-1) as a mediator of the redox regulation of TR by PDI. Interaction between Ref-1 and TRβ1 was detected. Overexpression of wild-type but not C64S Ref-1 facilitated the GH expression, suggesting that redox activity of Cys-64 in Ref-1 is involved in the TR-mediated gene expression. Moreover, PDI interacted with Ref-1 and changed the redox state of Ref-1, suggesting that PDI controls the redox state of Ref-1. Our studies suggested that Ref-1 contributes to TR-mediated gene expression and that the redox state of Ref-1 is regulated by PDI. Redox regulation of PDI via Ref-1 is a new aspect of PDI function.

growth hormone (GH) gene is activated via thyroid hormone receptors (TRs) in the presence of T 3 . TR is a nuclear receptor and acts as a transcription factor. In GH3 cells, TR binds to T 3 -resoponse element, which exists in the upstream region of the GH gene. We demonstrated that overexpression of PDI in GH3 cells suppresses the mRNA levels of GH. This result suggested that PDI contributes to the regulation of T 3 -mediated gene expression in GH3 cells (26). As described above, PDI is considered to serve as a reserve of T 3 . Thus, it is considered that overexpressed PDI traps T 3 and inhibits the activation of TR-mediated gene expression by T 3 . However, the mechanism by which GH expression is regulated by PDI is not clear.
In the present study, we clarified the mechanism of suppression of T 3 -stimulated GH expression by PDI-overexpression, and found that PDI regulates the T 3 -stimulated gene expression in GH3 cells via its catalytic activity but not its T 3 -binding capacity. We also found that redox factor1 (Ref-1), which changes the redox state of some transcription factors and regulates transcriptional activity, contributes to the regulation of GH expression by PDI. The regulation of the redox state of Ref-1 is a new aspect of PDI's functions.

EXPERIMENTAL PROCEDURES
Chemicals-Dithiothreitol (DTT) was purchased from Wako Pure Chemical Industries Ltd. (Osaka, Japan). Diamide, N-ethylmaleimide (NEM), methoxypolyethylene glycol (PEG) maleimide 5000, and T 3 were obtained from Sigma-Aldrich. All other chemicals used were purchased from Wako. T 3 was dissolved in 0.1 M NaOH at 10 mM to make stock solutions and stored at Ϫ20°C. 10 nM T 3 in NaOH solution (0.1 M) was added to the assay system or cell culture medium. In control samples, the same amount of NaOH was also added as a vehicle.
Construction of Rat PDI Fragment Peptides-Fragments of rat PDI and mutants (C/A Mt, ⌬a, and ⌬bЈ) were obtained as described previously (22,23,27). C/A Mt was a mutant in which cysteine residues at two catalytic active sites (Cys-55, Cys-58, Cys-399, and Cys-402) were substituted with alanine residues. The ⌬a and ⌬bЈ lack the a and bЈ domains, respectively. Mutant cDNAs were cut with BamHI and XhoI and ligated into the vector pcDNA3.1 (Invitrogen). TR␣ (GenBank TM accession number NM_001017960), TR␤1 (GenBank accession number J03819), TR␤2 (GenBank accession number NM_012672), Ref-1 (GenBank accession number NM_024148.1), and thiore-doxin 1 (TRX1) (GenBank accession number NM_053800) were constructed as follows. A polymerase chain reaction (PCR) was performed for each fragment cDNA with primer 1 and primer 2 (TR␣), primer 3 and primer 5 (TR␤1), primer 4 and primer 5 (TR␤1), or primer 6 and primer 9 (Ref-1), which are shown in Table 1. The cDNA obtained from total RNA of GH3 cells was used as a template. PCR was done with KOD polymerase with denaturation at 94°C for 2 min and then 30 cycles of 94°C for 30 s, 55°C for 90 s, and 68°C for 1 min. The cDNA of C64S Ref-1 was obtained by PCR with two steps. In the first PCR, nucleotide fragment 1 of C64S Ref-1 was amplified with primer 6 and primer 7. Nucleotide fragment 2 was amplified with primer 8 and primer 9 (for pcDNA) or 10 (for pQE80L). The two fragments were purified by electrophoresis with agarose gels. The second round of PCR was performed with fragments 1 and 2 and primer 6 and primer 9. The cDNAs of TR␣, TR␤1, and TR␤2 were cut with EcoRI and XhoI (for pCMV-HA) or BamHI and XhoI (for pcDNA-3ϫFLAG) and ligated into pCMV-HA (Clontech) or pcDNA-3ϫFLAG. The cDNA of the wild-type and C64S Ref-1 were cut with BamHI and XhoI (for pcDNA) or BamHI and SalI (for pQE-80L) and ligated into pcDNA and pQE-80L (Qiagen, Valencia, CA). The cDNA of TRX1 was amplified with primer 11 and primer 12, and the obtained fragment was cut with BamHI and XhoI and ligated into pcDNA vector.
Expression and Purification of Rat Histidine-tagged PDI and Ref-1-Expression and purification of rat histidine-tagged PDI was performed as described previously (27). BL21 (Novagen, Madison, WI) E. coli cells transformed with pQE-80L encoding a rat histidine-tagged wild-type and C64S Ref-1 were grown at 37°C in 2ϫ yeast extract-tryptone-rich medium containing 0.1 mg/ml ampicillin, and protein expression was induced by addition of 1.0 mM isopropyl thio-␤-O-galactoside. After additional cultivation for 4 h, E. coli cells were harvested and lysed in a lysis buffer (50 mM NaH 2 PO 4 , pH 7.5 containing 300 mM NaCl, 1.0 mg/ml lysozyme, and 20 mM imidazole) for 60 min at 4°C. The cell lysate was sonicated for 2 min. The lysate sample was solubilized by 0.5% n-dodecyl maltoside and centrifuged at 50,000 ϫ g for 30 min, and the supernatant was loaded onto a nickel-nitrilotriacetic acid-agarose column (Qiagen, Hilden, Germany). After the column was washed with lysis buffer, the protein was eluted with lysis buffer containing 250 mM imidazole.

Factor
Primer no. Sequence Cell Culture and Transfection-The rat pituitary tumor cell line GH3 was provided by the Health Science Research Resources Bank (cell number JCRB9047; Osaka, Japan). Cells were maintained in Ham's F-10 medium containing 15% horse serum and 2.5% fetal bovine serum and incubated at 37°C in a humidified atmosphere of 5% CO 2 . Overexpression of wildtype PDI was performed as described previously (26). Forced expression of C/A Mt, ⌬a, and ⌬bЈ was also performed using the same method as for overexpression of the wild type. Transient expression of the wild-type and C64S Ref-1 was performed as follows. The pCMV-HA containing Ref-1 cDNA was transfected into GH3 cells. The vector (2 g) was introduced into GH3 cells using HilyMax transfection reagent (Dojindo, Kumamoto, Japan) at 2 ϫ 10 5 cells/well in 3.5-cm dishes. Twentyfour hours later, the culture medium was replaced with Td medium containing test chemicals. After another 24 h, cells were harvested for use in each experiment. For hypoxic treatment, the cells were incubated in 1% O 2 , 5% CO 2 , and 90% N 2 (balanced with a modulator incubator chamber) for 6 h.
Isolation of RNA and Reverse Transcription-PCR-GH3 cells were cultured in Td medium for 24 h, the culture medium was replaced with fresh Td medium containing 10 nM T 3 , and incubation was continued for 24 h. Total RNA was extracted from GH3 cells with Isogen (Nippon Gene, Toyama, Japan). A reaction mixture containing 1 g of RNA and 200 units of reverse transcriptase (Takara) was incubated according to the manufacturer's directions as follows: 10 min at 25°C, 60 min at 42°C, and then 10 min at 70°C to stop the reaction. The quantified real time polymerase chain reaction (qRT-PCR) was performed using a Thermal Cycler Dice Real Time System Single TP850 (Takara, Shiga, Japan). SYBR Primer Ex TaqII, 10 pmol of forward and reverse primers, and 1 g of cDNA were mixed, and RT-PCR was performed according to the manufacturer's instructions. PCR was conducted at 95°C for 10 s followed by 40 cycles of 95°C for 5 s and 60°C for 20 s. Primers for rat GH (GenBank accession number GQ890681) were 5Ј-CTGCTGA-CACCTACAAAGA-3Ј (sense) and 5Ј-CAGTGTGTGCCTA-GAAAGCA-3Ј (antisense). Primers for rat ␤-actin (GenBank accession number CAA24528) were 5Ј-CTACAATGAGCTG-CGTGTGG-3Ј (sense) and 5Ј-TGAGGTAGTCTGTCAGG-TCC-3Ј (antisense). Quantification was done by using the second derivative maximum method according to the manufacturer's directions.
Luciferase Reporter Assay-The luciferase reporter gene assay was performed using the vector pGL3, which includes the rat growth hormone 5Ј-flanking region (from Ϫ1803 to ϩ5) (GenBank accession number X12967) containing T 3 -response elements as reported (26). GH3 cells were transfected using GenePORTER2 reagent. In the case of HEK293T cells, pGL3 and pCMV-HA that included rat TR␣, TR␤1, or TR␤2 were co-transfected by the calcium phosphate method. Twenty-four hours after transfection, the culture medium was replaced with Td medium containing test chemicals. After another 24 h, cells were harvested to measure luciferase activities using the Dual-Luciferase assay system (Promega) according to the manufacturer's instructions. Transfection efficiencies were corrected with the internal control.
Induction of GH-GH3 cells were cultured in 3.5-cm dishes at 2 ϫ 10 5 cells/dish in Td medium for 24 h to eliminate endogenous T 3 . The culture medium was then replaced with fresh Td medium containing test chemicals, and 24 h later, the culture medium and cells were collected. Proteins in the culture medium were precipitated by acetone. The GH protein in the cells and medium was detected by Western blotting using an anti-GH monoclonal antibody (Chemicon International, Temecula, CA).
Immunoprecipitation-Immunoprecipitation of Ref-1 using GH3 cells was performed as follows. GH3 cells were treated with ice-cold PBS containing 20 mM NEM for 20 min and lysed in a lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl containing 0.5% Nonidet P-40 and 0.1% protease inhibitor mixture). Insoluble material was removed by centrifugation at 12,000 ϫ g for 10 min, and the supernatant was adjusted to 2 mg/ml by adding lysis buffer. Next, 500 l of supernatant was incubated with 2 l of anti Ref-1 antibody or unimmunized rabbit serum for 1 h, and 20 l of protein A-Sepharose (50% (w/v); GE Healthcare) in lysis buffer was added to this solution and incubated for 1 h at 4°C. The samples were centrifuged for 1 min at 8,000 ϫ g, and the supernatant was discarded. Precipitates were washed with wash buffer 1 (50 mM Tris-HCl, pH 7.5, 150 mM NaCl containing 0.5% Nonidet P-40) and wash buffer 2 (50 mM Tris-HCl, pH 7.5 containing 150 mM NaCl). Resulting pellets were heat-denatured with 30 l of buffer (125 mM Tris-HCl, pH 6.8 containing 4% SDS, 20% glycerol, and 10% 2-mercaptethanol) for 5 min at 95°C. Immunoprecipitation of TR␤1 using HEK293 cells was performed as follows. TR␤1 containing 3ϫFLAG tags was expressed in HEK293 cells. Cells were treated with ice-cold PBS containing 20 mM NEM for 20 min and lysed using IP buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl containing 0.5% Nonidet P-40 and 1% protease inhibitor mixture). Insoluble material was removed by centrifugation at 8,000 ϫ g for 10 min, and the supernatant was adjusted to 4 mg of protein/ml with lysis buffer. Then 500 l of supernatant was incubated with 2 l of anti-DYKDDDDY (FLAG) tag antibody (Wako) or unimmunized mouse serum for 1 h, and 20 l of protein G-Sepharose (50% (w/v); GE Healthcare) in lysis buffer was added to this solution and incubated for 1 h at 4°C. The samples were centrifuged for 1 min at 5,000 ϫ g, and the supernatant was discarded. Precipitates were washed using wash buffer 1 and wash buffer 2. Immunoprecipitation of wild-type or C64S mutant Ref-1 and PDI using HEK293 cells was performed as follows. Wild-type or C64S Ref-1 containing the HA tag was expressed in HEK293 cells, and immunoprecipitation was performed using anti-HA tag antibody (Clontech) or unimmunized mouse serum.
Detection of Redox State by PEG Maleimide-Cells were washed with PBS containing 20 mM NEM, which blocks free SH, and incubated with PBS containing 20 mM NEM on ice for 20 min, and then cells were harvested and lysed with 50 mM Tris-HCl, pH 7.5 containing 1% SDS. Proteins were precipitated with ice-cold acetone. Precipitates were dissolved in 50 mM Tris-HCl, pH 8.0, and disulfide bonds in proteins were reduced by 10 mM tris(2-carboxyethyl)phosphine hydrochloride (TCEP). The proteins were incubated with 2 mM PEGmaleimide for 1 h at room temperature, and reactions were stopped by acetone precipitation. Then SDS-PAGE was performed on a 10% polyacrylamide gel. Proteins were transferred to nitrocellulose and detected using a polyclonal anti-PDI or -Ref-1 antibody. Detection of the redox state of recombinant wild-type Ref-1 and C64S Ref-1 was performed as follows. 5 M Ref-1 protein and 5 M PDI protein were co-incubated in 50 mM Tris-HCl, pH 7.5 for 30 min at room temperature. Then proteins were incubated with 20 mM NEM for 1 h at room temperature and precipitated with acetone to remove NEM. Precipitates were dissolved in 50 mM Tris-HCl, pH 8.0, and disulfide bonds in proteins were reduced by 10 mM TCEP. The proteins were incubated with 2 mM PEG-maleimide for 1 h at room temperature, and reactions were stopped by acetone precipitation.

PDI Suppresses GH Expression via Its Isomerase Activity-
Previously we demonstrated that overexpression of PDI suppressed the GH mRNA expression in GH3 cells (28). We also clarified that the suppression of GH mRNA was mediated by the GH promoter in a luciferase reporter gene assay using the upstream region of GH (28). In PDI-overexpressing cells, promoter activity was significantly suppressed.
PDI is a T 3 -binding protein and considered to act as a modulator of free T 3 levels. This raises the possibility that increases in PDI protein levels in GH3 cells suppress the TR-mediated GH expression by trapping the T 3. To test whether PDI acts as a modulator of T 3 , we prepared several mutants of PDI such as C/A Mt, ⌬a, and ⌬bЈ, and GH mRNA levels in cells overexpressing PDI mutants were investigated. We previously found that T 3 bound to the a and bЈ domains of PDI (23). Therefore, cDNAs of ⌬a and ⌬bЈ, which lack the a and bЈ domains, respectively, were prepared as deletion mutants of the T 3 -binding domain and expressed in GH3 cells. In addition, the cDNA of C/A Mt in which the cysteines of catalytic active sites were substituted by alanines was prepared and expressed in GH3 cells. These cells were treated with 10 nM T 3 , and GH mRNA levels were determined by real time quantitative PCR. Consistent with previous findings, the expression of GH mRNA was suppressed by the overexpression of wild-type PDI (Fig. 1A). It was also suppressed by forced expression of ⌬bЈ. Conversely, GH was not suppressed by forced expression of C/A Mt. These results suggested that isomerase activity rather than T 3 binding activity is involved in the regulation of the T 3 response. We previously found that knockdown of PDI did not increase GH expression (28). These results indicate that PDI does not suppress GH expression by trapping T 3 to inhibit TR-mediated gene expression. Overexpression of ⌬a did not suppress the GH expression despite the a domain binding to T 3 . This is because the a domain has a catalytic active site, and deletion of the a domain affected the isomerase activity.
Next, the protein levels of GH in wild-type PDI-, C/A Mt-, ⌬a-, and ⌬bЈ-expressing cells were investigated in the presence or absence of T 3 (Fig. 1, B and C). Because GH is a secretory protein, intracellular and extracellular protein levels were determined. Protein levels, like mRNA levels, were suppressed in wild-type PDI-and ⌬bЈ-overexpressing cells.
Effects of Changes in Redox State on GH Expression-The isomerase activity of PDI regulates the redox state of cysteine residues of proteins. We next investigated the effects of reducing and oxidizing agents on GH expression. DTT was used as a reducing agent, and diamide and hydrogen peroxide were used as oxidizing agents. GH3 cells were treated with these agents in the presence of T 3 , and expression levels of GH were determined. In the presence of diamide and hydrogen peroxide, the mRNA level of GH was suppressed, whereas it was not suppressed by DTT ( Fig. 2A). The protein level of GH was also decreased in the presence of diamide (Fig. 2B). These results suggested that a change of the redox state of the cell affects the T 3 -induced GH expression. Next, to evaluate the contribution of glutathione to GH expression, GH3 cells were treated with buthionine sulfoximine (BSO), which is an inhibitor of glutathione synthesis, and GH expression was investigated (Fig. 2C).
In the presence of 1 mM BSO, little suppression of GH expression was detected. Changes of the redox state by glutathione depletion affected the GH expression slightly.
Transcriptional  (29,30). In addition, Ref-1 regulates the transcriptional activity of estrogen receptor ␣. The redox state of Ref-1 is known to be regulated by thioredoxin or ERp57, which is one of the PDI family proteins (29,(31)(32)(33). Given these findings, we hypothesized that the redox state of TR is also regulated by Ref-1 and that PDI is involved in the regulation of Ref-1.
To test this hypothesis, the mRNA expression of GH in Ref-1-overexpressing cells was examined by qRT-PCR (Fig. 3). To reduce the cysteines of transcription factors, the cysteine 64 residue in the N-terminal domain of Ref-1 is reported to be important (34). Therefore, C64S Ref-1 in which the 64th residue was changed from cysteine to serine was also overexpressed. The mRNA level of GH was elevated by Ref-1 overexpression (Fig. 3A), whereas it was not changed by the overexpression of C64S Ref-1. These results suggested that Ref-1 is involved in transcriptional activation via TR and that cysteine at position 64 plays an important role in this activation. The protein level of GH also was increased by WT Ref-1 overexpression (Fig. 3B). Moreover, Ref-1 was overexpressed in PDI-overexpressing GH3 cells, and GH mRNA levels were investigated (Fig. 3C). Stimulation of GH expression by T 3 was suppressed by overexpression of PDI, and this suppression was recovered by overexpression of Ref-1. The redox state of Ref-1 has been proposed to be regulated by TRX1. Thus, we also investigated the effects of TRX1 overexpression on T 3 -mediated gene expression (Fig. 3D). As a result, TRX1 overexpression did not affect the GH expression, suggesting that TRX has little contribution to TR-mediated gene expression and that PDI mainly regulates the redox state of Ref-1 in GH3 cells.
Ref-1 appeared to contribute to the TR-mediated GH expression. In GH3 cells, there are three variants of TRs: TR␣, TR␤1, and TR␤2. Therefore, we investigated which is important for the expression of GH. TR␣, TR␤1, or TR␤2 was overexpressed in GH3 and HEK293 cells, and the promoter activity of GH was determined by luciferase reporter assay. We also investigated the effects of diamide, which suppressed the expression of GH mRNA, on the activation of promoter activity of GH by these TRs. In GH3 cells, promoter activity was activated by T 3 , and the T 3 -induced activity was suppressed in the presence of 100 M diamide (Fig. 3E). In GH3 cells overexpressing TR␣ and TR␤1, the promoter activity of GH was enhanced, suggesting the contribution of TR␣ and TR␤1 to the T 3 -mediated activation of promoter activity. Moreover, in cells overexpressing TR␤1, promoter activity was suppressed by diamide, suggesting that transcriptional activity of TR␤1 is sensitive to the redox state. On the other hand, TR␤2 did not enhance the T 3 -medited promoter activity. To eliminate the contribution of endogenous TRs to GH promoter activity, we also performed the same experiment using HEK293 cells, which do not have endogenous activation of promoter activity of GH. Enhancement of GH promoter activity was seen in TR␣and TR␤1-overexpressing HEK293 cells, suggesting the contribution of TR␣ and TR␤1 to the activation of promoter activity (Fig. 3F). Moreover, the promoter activity was suppressed by diamide-treated TR␤1-overexpressing HEK293 cells the same as in GH3 cells, suggesting that the transcriptional activity of TR␤1 is redox-sensitive.
We next examined whether TR␤1 interacts with Ref-1. TR␤1 containing a 3ϫFLAG tag was expressed in HEK293 cells, and immunoprecipitation was performed. When a FLAG tag anti-FIGURE 1. GH levels in GH3 cells overexpressing PDI mutants. A, GH3 cells overexpressing PDI mutants were cultured for 24 h in the presence or absence of 10 nM T 3 . Total RNA was isolated from three culture plates, and qRT-PCR was performed. GH mRNA levels of mock-transfected cells without T 3 were set at 1.0. The overexpression of PDI mutants in GH3 cells was checked by immunoblotting with an anti-PDI antibody (upper panel). B and C, GH3 cells overexpressing PDI mutants were cultured for 48 h in the presence or absence of 10 nM T 3 . Total cellular protein (B) or medium protein (C) was collected in three culture plates, and immunoblotting was performed. GH protein levels of mock-transfected cells without T 3 were set at 1.0. Values are expressed as the mean Ϯ S.D. (error bars) for three replicates. **, p Ͻ 0.01; *, p Ͻ 0.05, significantly different from mock cells without T 3 . ##, p Ͻ 0.01; #, p Ͻ 0.05, significantly different from mock cells treated with T 3 .
body was used, Ref-1 was immunoprecipitated with TR␤1 ( Fig.  3, G and H). When unimmunized mouse serum was used, Ref-1 was not precipitated. These results suggested that Ref-1 interacts with TR␤1 in HEK293 cells.
PDI Interacts with Ref-1-In general, the redox state of Ref-1 is known to be regulated by a thioredoxin and thioredoxin reductase system. Grillo et al. (31) suggested that ERp57, which is a member of the PDI family and has a domain architecture similar to that of PDI, also interacts with Ref-1 and regulates its redox state.
To investigate whether PDI interacts with Ref-1, immunoprecipitation was performed. We predicted that PDI interacts with Ref-1 via a disulfide bond; therefore, the cells were treated with NEM, which binds with free thiol groups and stops further reaction (internal disulfide formation) to catch the reaction intermediate  (Fig. 4C). The same results were obtained by immunoprecipitation using HEK293 cells (Fig. 4, D and E).
Next, we investigated whether cysteine residues contribute to the PDI/Ref-1 interaction. Interaction between PDI and C64S Ref-1 was tested by immunoprecipitation (Fig. 4, F and  G). HA-tagged wild type or C64S Ref-1 was expressed in the cell, and immunoprecipitation was performed using anti-HA tag antibody. Expression levels in cells and immunoprecipitated amounts of Ref-1 were approximately equal between wild type and C64S (Fig. 4F). However, by substitution of Cys-64, the amount of co-precipitated PDI with Ref-1 was increased (Fig.  4G). These results suggest that PDI is trapped by C64S Ref- The reaction with PEG maleimide increased the protein mass by 5 kDa, which was determined by Western blotting. In this experiment, before treatment with PEG maleimide, cell extract was treated with NEM, and disulfide bonds were reduced by TCEP. Therefore, PEG maleimide was incubated with cysteines that form disulfide bonds. When cells were treated with DTT before being treated with NEM, upshifted bands of PDI (Fig. 5A) and Ref-1 (Fig. 5B) were not detected. When PDI or Ref-1 was treated with DTT, PDI and Ref-1 were in the reduced state. Conversely, when cells were treated with diamide, several upshifted bands appeared, indicating that PDI and Ref-1 are in the oxidized state. As shown in Fig. 5A, upon immunoblotting using PDI antibody, several bands were detected, suggesting that both reduced and oxidized forms of PDI exist in GH3 cells. T 3 did not change the redox state of PDI. As shown in Fig. 5B, with immunoblotting using Ref-1 antibody, two bands were mainly detected. The redox state of Ref-1 was not changed by T 3 treatment. When an internal disulfide bond is present in Ref-1, Ref-1 contains two SH groups that react with two PEG-maleimide (5-kDa) molecules, and a band upshifted by 10 kDa will be observed on SDS-PAGE. In the case of an external disulfide bond, one SH is formed by TCEP, and a band upshifted by 5 kDa will be observed. In the in vivo experiment shown in Fig. 5B, a band of Ref-1 upshifted by 5 kDa was detected, and the intensity of this band was increased by overexpression of PDI. We cannot explain why a band upshifted by 10 kDa was not detected. One possibility is that PDI or another protein binds to Ref-1 by a disulfide bond. Thus, we performed an in vitro experiment.
PDI Oxidizes the Cysteine Residue of Ref-1-Next, changes in the redox state of wild-type Ref-1 and C64S Ref-1 by PDI were investigated using PEG maleimide (Fig. 5, C and D). Purified Ref-1 and C64S Ref-1 were incubated with PDI, and then free cysteine residues were blocked by NEM. Internal and external We also tested the effects of glutathione on the redox state of Ref-1 using PEG maleimide because GH expression was only slightly suppressed by BSO (Fig. 2C). Purified Ref-1 was incubated with 2 mM glutathione containing a mixture of reduced glutathione (GSH) and oxidized glutathione (GSSG) in different ratios. The redox state of Ref-1 was not changed at 2GSH/ GSSG ratios between 100 and 10 ( Fig. 5E), suggesting that glutathione did not directly react with Ref-1. Upper bands seen in the 2GSH/GSSG Ͻ1 lane are considered to be an artificial polymer of Ref-1. These results suggested that the suppression of GH expression in GH3 cells (Fig. 2C) is an indirect effect of BSO and that glutathione depletion affected the T 3 response via PDI.
Overexpression of PDI Also Affects the HIF-1␣-regulated Gene Expression-Ref-1 regulates the redox state of many transcription factors. Ref-1 also regulates the redox state of HIF-1␣. We investigated whether PDI overexpression also affects HIF-1␣regulated gene expression (Fig. 6). In hypoxic conditions, HIF-1␣ activates the expression of target genes such as erythropoietin, vascular endothelial growth factor (VEGF), and glu-

DISCUSSION
PDI had been considered to serve as a high capacity reservoir of hormones to modulate the T 3 concentration in cells because PDI localizes to the endoplasmic reticulum at high concentration (200 M) (37) and has lower affinity for T 3 than T 3 receptors (18). Therefore, we predicted that overexpressed PDI traps T 3 , leading to a decrease in the binding of T 3 to TR. Unexpectedly, however, in this study, we found that PDI does not suppress the response to T 3 by acting as a reservoir of T 3 . Actually, overexpression of C/A Mt PDI did not change the response to T 3 , although C/A Mt has almost the same capacity to bind to T 3 as the wild type (22). On the other hand, overexpression of ⌬bЈ suppressed GH expression despite the deletion of the T 3 -binding domain. These results indicated that PDI does not suppress the expression of GH by acting as a reservoir of T 3 in GH3 cells, but the isomerase activity of PDI is involved in the regulation of T 3 -mediated gene expression.
PDI regulates the response to T 3 via its catalytic activity, suggesting that PDI regulates the gene expression of GH by alternating the redox state of TR. In fact, the activities of several transcription factors, including hormone receptors, are regulated by the reduction of cysteine residues. Estrogen receptor ␣ is one of the transcription factors regulated by a change in redox state of a cysteine residue (38). Moreover, the redox state of estrogen receptor ␣ is regulated by Ref-1. Ref-1 enhances the interaction of estrogen receptor ␣ with estrogen-response elements in DNA and activates the expression of downstream genes. Tell et al. (39) reported that thyroid hormone-stimulated factor 1 is also regulated by a change in the redox state of its cysteine residue. Reduction of Cys-87 in thyroid hormone-stimulated factor 1 by Ref-1 activates binding to DNA. Ref-1 also stimulates the DNA binding activity of several transcription factors, including p53, c-Myb, AP-1, NF-B, and HIF-1␣, by the reduction of their cysteine residues (29,38,40,41).
Here, we found that GH expression stimulated by TR is also regulated by Ref-1. A luciferase reporter assay using an upstream region of the GH gene containing T 3 -resoponse element showed that TR␣ and TR␤1 had the ability to activate the GH promoter activity in the presence of T 3 and that TR␤1mediated promoter activity was suppressed by an oxidizing agent. These results suggested that the function of TR␤1 has redox sensitivity and that Ref-1 may contribute to the TR␤1mediated gene expression. Moreover, by immunoprecipitation, an interaction between TR␤1 and Ref-1 was detected in HEK293 cells. Thus, it is suggested that the transcriptional activity of TR␤1 is regulated by Ref-1. Oxidizing agents suppressed the T 3 -regulated gene expression. T 3 acts to up-regulate respiratory and metabolic gene expression via TR-dependent transcriptional activity and to elevate O 2 consumption. This action leads to the production of reactive oxygen species. In addition, Ref-1 is reported to be controlled in its subcellular localization by the cellular redox status. From these findings, regulation of the transcriptional activity of TR by the redox state may suppress the overproduction of reactive oxygen species by activation of calorigenesis, and Ref-1 may act as a key factor in this regulation.
The redox state of Ref-1 is known to be regulated by the TRX and thioredoxin reductase system. TRX associates with Ref-1 through its catalytic active cysteines and reduces a particular cysteine residue of Ref-1. TRX regulates the activation of transcription factors via the reduction of a cysteine residue of Ref-1. In this study, we found that PDI also contributes to the change in the redox state of Ref-1. We also found that overexpression of PDI facilitates the oxidation (disulfide bond formation) of Ref-1 in cells. Oxidation of Ref-1 by PDI was seen in the assay in vitro as well. These results suggested that PDI inhibits the activation of Ref-1 by oxidizing cysteine residues. This action toward Ref-1 is the opposite of that of TRX action toward Ref-1. Inactivation of Ref-1 by PDI is considered to lead to the suppression of GH expression. The valance of the redox state of Ref-1 is considered to be regulated by TRX-like proteins, TRX, ERp57, and PDI. In this study, we found that overexpression of TRX1 did not affect the GH expression. Also, the expression level of TRX1 is very low in the brain (42). Actually, we could not detect the band of endogenous TRX by immunoblotting in GH3 cells (Fig. 3D). Therefore, in the case of regulation of TR by Ref-1 in GH3 cells, contributions of TRX1 are considered to be low. In GH3 cells, the main regulator of Ref-1 is considered to be PDI.
Cys-65 of human Ref-1 (Cys-64 of rat Ref-1) is a critical amino acid for redox activity (34). This cysteine residue is also important for the redox regulation of estrogen receptors. Also, mutation of Cys-65 of human Ref-1 affects cell proliferation (35). In this study, we found that the mutation of Cys-64 eliminated the facilitation of GH expression via TR. These results suggested that PDI changes the redox state of Cys-64 in Ref-1.
Mutation of Cys-64 in Ref-1 increased PDI⅐Ref-1 complex formation by stopping further disulfide formation between PDI and Ref-1. Cys-65 of human Ref-1 forms a disulfide bond with Cys-93 (34). By substitution of Cys-64, PDI was considered to be trapped by Ref-1 via a disulfide bond with another cysteine as an intermediate of the thiol-disulfide conversion reaction. Vascotto et al. (35) reported that a higher amount of ERp57, which regulates the redox state of Ref-1, was immunoprecipitated with C65S Ref-1 than with wild type because of trapping of ERp57 by formation of a disulfide bond with Cys-93.
Moreover, not only TR-mediated gene expression but also HIF-1␣-mediated gene expression was affected by PDI overexpression. In addition, interaction between PDI and Ref-1 was seen not only in GH3 cells but also in HEK293 cells. Therefore, PDI may act as a Ref