Glutathionylation of two cysteine residues in paired domain regulates DNA binding activity of Pax-8.

We reported that the first two cysteine residues out of three present in paired domain (PD), a DNA-binding domain, are responsible for redox regulation of Pax-8 DNA binding activity. We show that glutathionylation of these cysteines has a regulatory role in PD binding. Wild-type PD and its mutants with substitution of cysteine to serine were synthesized and named CCC, CSS, SCS, SSC, and SSS according to the positions of substituted cysteines. They were incubated in a buffer containing various ratios of GSH/GSSG and subjected to gel shift assay. Binding of CCC, CSS, and SCS was impaired with decreasing GSH/GSSG ratio, whereas that of SSC and SSS was not affected. Because [3H]glutathione was incorporated into CCC, CSS, and SCS, but not into SSC and SSS, the binding impairment was ascribed to glutathionylation of the redox-reactive cysteines. This oxidative inactivation of PD binding was reversed by a reductant dithiothreitol and by redox factor (Ref)-1 in vitro. To explore the glutathionylation in cells, Chinese hamster ovary cells overexpressing CSS and SCS were labeled with [35S]cysteine in the presence of cycloheximide. Immunoprecipitation with an antibody against PD revealed that treatment of the cells with an oxidant diamide induced the 35S incorporation into both mutants, suggesting the PD glutathionylation in cells. Since the two cysteine residues in PD are conserved in all Pax members, this novel posttranslational modification of PD would provide a new insight into molecular basis for modulation of Pax function.

Various cellular functions such as gene expression, signal transduction, and enzyme activity are modulated by intracellular redox potential (1). In the context of gene expression, it has been shown that the redox potential alters the DNA binding activity of a number of transcription factors, such as AP-1 and NF-B, through oxidoreductive modification of thiols on the redox-reactive cysteine residues that mostly reside in the DNA-binding domain (2)(3)(4)(5)(6). It has also been shown that such modification is often catalyzed by intracellular redox enzymes such as redox factor-1 (Ref-1) and thioredoxin (5,7,8).
Pax-8 belongs to a family of transcription factors that contain a paired domain (PD) 1 as a DNA-binding domain (9). Pax-8 stimulates expression of a set of thyroid-enriched proteins, thyroglobulin, thyroperoxidase, and sodium/iodide symporter (10,11). We previously reported that DNA binding of Pax-8 is reversibly regulated via a redox control mechanism: its binding is abolished by treatment with an oxidant diamide and restored by a reductant dithiothreitol (DTT) (12). Furthermore, we showed in rat thyroid FRTL-5 cells that thyrotropin increases Pax-8 binding probably by converting the oxidized form of Pax-8 to the reduced one and that this reduction of Pax-8 is accompanied by the increase in Ref-1 expression (12,13). In vitro, Ref-1 can restore binding of oxidized Pax-8. It was therefore speculated that Ref-1 is involved in the reduction of Pax-8 in thyroid cells. Consistent with our results, Tell et al. (14) also reported the involvement of Ref-1 in redox regulation of Pax-8.
Our recent study defined the redox-reactive cysteine residues in Pax-8 (15). Six cysteine residues are present in Pax-8; three residues are in PD located at the N-terminal region, and three are located in the C-terminal region (16). We demonstrated that oxidation of the first two cysteine residues, Cys-45 and Cys-57, in rat PD loses Pax-8 binding in vitro. We also demonstrated by transfection assay that the modification of these cysteine residues plays a role in vivo because Ref-1 is capable of enhancing transcriptional activity of wild-type Pax-8 in cells but not that of the mutant in which both Cys-45 and Cys-57 are substituted with serines. The mutant displays constitutive transcriptional activity. These observations indicate that the wild-type Pax-8 with no DNA binding ability possesses the oxidized Cys-45 and/or Cys-57. However, the nature of the oxidized cysteine residues remained to be elucidated.
It has been established that reduced GSH is the most abundant non-protein thiol in mammalian cells and that it plays a pivotal role in the maintenance of cellular reducing environment and defense against oxidative stress (17,18). Glutathione is present in cells at millimolar concentrations, and thus the GSH/GSSG ratios constitute a major redox buffer in cytosol. In resting cells, the GSH/GSSG ratio ranges from 30:1 to 100:1, whereas its ratio in endoplasmic reticulum is shown to be 1:1 to 3:1, which may facilitate the formation of protein disulfide that is often essential for the structure and function of secreted proteins (19). In addition, a number of recent studies suggest that GSH/GSSG coupling is involved in modulation of protein function through reversible disulfide bond formation between protein cysteines and glutathione (glutathionylation).
Several mechanisms of glutathionylation have been proposed (20 -22). Under a lower ratio of GSH/GSSG, glutathionylation may occur with the increase in amount of GSSG, which can exchange disulfide with reactive protein thiol. Because GSH prevails over GSSG in normal cells, oxidation of the limited amount of GSH can dramatically lower GSH/GSSG ratio. Indeed, it has been shown that the GSH/GSSG ratio is changed in response to growth factors and cytokines (23,24). Other mechanisms may include chemical oxidants such as hydrogen peroxide or diamide, which serve as catalysts and promote direct disulfide formation between reactive protein thiol and GSH. Also, the mechanisms involving the formation of sulfenic acid or S-nitrosothiol as intermediate products followed by glutathionylation have been proposed.
As for the control of transcription factors by glutathionylation, it was reported that the conserved cysteine residue in the basic region, a DNA-binding domain, of c-Jun undergoes glutathionylation and that this glutathionylated c-Jun loses binding activity (25). It was also reported that glutathionylation of one cysteine residue in the DNA-binding domain of p50 NF-B inhibits its binding (21).
In the present study, we investigated the possible involvement of glutathionylation in redox regulation of the DNA binding activity of Pax-8 PD in vitro and in vivo. We provide evidence that a decrease in the GSH/GSSG ratio induces glutathionylation of Cys-45 and Cys-57, resulting in loss of DNA binding.

MATERIALS AND METHODS
Plasmid Construction, in Vitro Translation, and Protein Purification-Cloning of rat Pax-8 cDNA and preparation of the mutant cDNA (Pax146), which encodes only 146 amino acids from N-terminal and thus harbors only the PD, were described previously (12,15). Sitedirected mutagenesis was also described previously (12,15). The sequencing of the entire cDNAs verified the mutations. As shown in Fig.  1, wild-type and PD mutants were named CCC, CSS, SCS, SSC, and SSS, according to the positions of substituted cysteines. These cDNAs were ligated into pIVEX2.3-MCS vector (Roche Applied Science). This vector contains six His codons at the 3Ј-end of the cloning sites. In vitro transcription and translation were performed using the Rapid Translation System (RTS 500 instrument, Roche Applied Science). The Histagged PD proteins were then purified by using nickel chelate resin (His MicroSpin purification module, Amersham Biosciences). After removing imidazole by Ultra Free C3-GC filter (molecular weight cut-off, 10,000, Millipore, Bedford, MA), the proteins were stored at Ϫ80°C in the presence of 1 mM DTT. Protein contents were determined by Bradford method.
Electrophoretic Mobility Shift Assay (EMSA)-Procedures for EMSA were described previously (26). Oligo(C) containing a recognition site for Pax-8 in the C region of rat thyroglobulin promoter (10) was used as a probe. In vitro translation lysates or purified PD proteins were first treated with 3 mM DTT and then subjected to an Ultra Free C3-GC filter to remove excess DTT. Subsequent chemical oxidation was achieved by incubation with 1 mM diamide (Sigma). After the removal of excess diamide by the filtration, the samples were again incubated with 3 mM DTT. These lysates or purified PD proteins (final 30 nM) were incubated with 32 P-labeled oligo(C) in EMSA reaction buffer without DTT. The reaction mixture without dye was subjected to polyacrylamide gel electrophoresis. The gel was dried and exposed to Fujix bioimage analyzer (BAS 2000; Fuji Photo Film, Tokyo, Japan). In some experiments, the reduced PD proteins were incubated for 60 min at 25°C in EMSA reaction buffers containing various ratios of GSH/GSSG and then subjected to EMSA. Since two GSH molecules are oxidized into one GSSG, the total concentration of glutathione, GSH plus 2 GSSG, was kept at 3 mM. Therefore, GSH/GSSG ratios of 200, 20, 2, and 0. Metabolic Labeling and Immunoprecipitation-CSS and SCS cDNAs were cloned into mammalian expression vector pcDNA3.1/Myc-His (Invitrogen), and the plasmids were transiently transfected into CHO cells by using a Lipofectamine reagent (Invitrogen). After an 18-h incubation with the liposome-DNA solution, the reagent was replaced with a fresh culture medium. After an additional 24-h incubation, the medium was changed to Hanks' balanced salt solution containing 10 g/ml cycloheximide (Sigma). After a 60-min incubation, endogenous glutathione pool was labeled with 3.7 MBq of [ 35 S]cysteine (39.8 TBq/mmol, 370 MBq/ml, PerkinElmer Life Sciences) for 60 min. Cycloheximide was used to prevent incorporation of [ 35 S]cysteine into proteins and therefore to facilitate its incorporation into glutathione. The cells were then treated with 1 mM diamide for 5 min and harvested for immunoprecipitation. As a parallel experiment, the transfected cells were labeled with [ 35 S]methionine. The detailed procedure was reported previously (29). In brief, the cells were preincubated in methionine-free Eagle's minimal essential medium for 2 h and then incubated in minimal essential medium containing [ 35 S]methionine (Express 35 S-protein labeling mix, PerkinElmer Life Sciences) for 2 h. Immunoprecipitation was performed using anti-Myc antibody (Roche Diagnostics), normal rabbit serum (NRS), and GammaBind G beads (Amersham Biosciences). The precipitates were incubated for 5 min at 100°C in a gel loading buffer with or without 10 mM DTT and then subjected to 15% SDS-polyacrylamide gel electrophoresis. After fixation, gels were incubated in an enhancer for fluorography (Amplify, Amersham Biosciences), dried, and exposed to the BAS 2000 system. To evaluate the efficiency of transfection, the whole cell lysates prepared from the transfected cells or their immunoprecipitates with anti-Myc antibody were subjected to Western blot analysis using the same antibody. Procedures for Western blot analysis were described previously (30). The proteins were visualized using enhanced chemiluminescence reagents (Pierce).
Immunocytochemical Analysis-CHO cells transfected with CSS-or SCS-expressing plasmid were subjected to immunocytochemical analysis. The procedures were described previously (30). Anti-Myc antibody and Alexa Fluor 488-conjugated anti-rabbit IgG (Molecular Probes, Eugene, OR) were used to visualize CSS and SCS proteins in cells. To identify the nucleus, the cells were simultaneously stained with Hoechst 33258 (Molecular Probes). The images were obtained using an Axiophot 2 fluorescence microscope (Carl Zeiss, Oberkochen, Germany).

RESULTS
Oxidative Modification of Cys-45 or Cys-57 Impairs DNA Binding of PD-PD and its mutants that contain only one or no cysteine residue were synthesized as fusion proteins with His 6 by in vitro translation system (Fig. 1). The products were reduced with DTT and then subjected to EMSA. As shown in Fig. 2, the reduced CCC, CSS, SCS, SSC, and SSS possess similar DNA binding activities (lanes 1, 4, 7, 10, and 13). When they were treated with diamide, not only CCC but also CSS and SCS lost the binding (lanes 2, 5, and 8). In contrast, the oxida- tion did not affect the binding of SSC and SSS (lanes 11 and  14). These results demonstrate that oxidative modification of the thiol of a single cysteine residue, Cys-45 or Cys-57, leads to impaired binding of PD. Of note, this modification is reversible because subsequent reduction of the oxidized proteins by DTT entirely restored the binding (lanes 3, 6, and 9).
GSH/GSSG-dependent Oxidation Inhibits DNA Binding of PD-To explore the possible involvement of glutathionylation in the modification of Cys-45 or Cys-57, the translation products were subjected to purification by nickel columns. To remove the imidazole, which was used as eluent, and to reconstitute fusion proteins in EMSA reaction buffer, the eluates were filtered with EMSA reaction buffer. The purified proteins were then incubated in the buffer containing various GSH/ GSSG ratios and subjected to EMSA. As shown in Fig. 3, a decrease in the GSH/GSSG ratio induced marked inhibition of the DNA binding activities of CCC, CSS, and SCS, and this inhibition was reversed by DTT. Although CCC possesses two redox-reactive cysteine residues, its sensitivity to GSH/GSSG ratios appeared to be similar to those of CSS and SCS. In contrast, a change in the GSH/GSSG ratio did not affect SSC and SSS binding. These results clearly demonstrate that GSH/ GSSG-dependent oxidation of Cys-45 or Cys-57 inhibits DNA binding of PD. Thus, the most likely oxidative modification is glutathionylation.
Lower Ratio of GSH/GSSG Promotes Incorporation of Glutathione into CSS and SCS-To validate GSH/GSSG-dependent glutathionylation of Cys-45 and Cys-57, [ 3 H]glutathione incorporation into PD mutants was evaluated. As shown in Fig.  4, when the GSH/GSSG ratio was 200, only low level incorporation was observed in CSS, SCS, SSC, and CCC. In contrast, when the ratio was set at 0.2, the pronounced incorporation was detected in CSS, SCS, and CCC but not SSC. Notably, the incorporation into CCC nearly doubled when compared with those into CSS and SCS, suggesting that the glutathionylation of Cys-45 and Cys-57 can occur simultaneously. These results demonstrate the GSH/GSSG-dependent glutathionylation of Cys-45 and Cys-57 in PD. Possible Glutathionylation of CSS and SCS in Cells-The above studies were conducted in vitro. We next investigated possible glutathionylation of CSS and SCS in cells by utilizing transfection, metabolic labeling, and immunoprecipitation methods. First, we evaluated the transfection efficiency. Whole cell lysates were prepared from CHO cells transfected with the plasmids expressing Myc-tagged CSS or SCS and were subjected to Western blot analysis using anti-Myc antibody. As shown in Fig. 6A, a single band with ϳ17 kDa was detected in both CSS and SCS lysates (lanes 3 and 4). In contrast, no band was detected in the lysate prepared from the cells transfected with a parent plasmid (pcDNA, lane 2). Since Myc-tagged CSS or SCS consists of 146 amino acids of PD plus 26 amino acids of Myc epitope, the molecular mass of 17 kDa is an expected mass for the fusion proteins. We next examined the efficiency of immunoprecipitation using 13 times more lysates than those

FIG. 2. Oxidation of either Cys-45 or Cys-57 by diamide abolishes PD DNA binding.
The in vitro translation products of CCC, CSS, SCS, SSC, and SSS were reduced by 3 mM DTT. After removal of excess DTT by the filtration, they were subjected to EMSA using oligo(C) as a probe. Chemical oxidation was performed by the incubation of these products with 1 mM diamide (lanes 2, 5, 8, 11, and 14). Subsequent reduction was achieved by the incubation of the oxidized products with 3 mM DTT after removal of excess diamide by the filtration (lanes  3, 6, 9, 12, and 15). Specific PD/oligo(C) complexes are indicated by closed arrowhead. used above. Immunoprecipitation and Western blot analysis demonstrated that the substantial amounts of CSS and SCS proteins could be precipitated and detected with anti-Myc antibody (lanes 8 and 10) but not with NRS (lanes 7 and 9). No protein was detected in the lysates from the pcDNA-transfected cells (lanes 5 and 6). These results led us to the following metabolic labeling experiment.
CHO cells were transfected with the plasmid expressing CSS or SCS, and the glutathione pool was labeled with [ 35 S]cysteine in the presence of cycloheximide. The cells were then exposed to oxidative stress by diamide. As a parallel experiment, the transfected cells were labeled with [ 35 S]methionine. As shown in Fig.  6B, when the cells were labeled with [ 35 S]methionine, CSS and SCS proteins with a molecular mass of 17 kDa were precipitated with anti-Myc antibody (lanes 1 and 2), whereas no corresponding protein was detected with NRS (data not shown). Also, no corresponding protein was precipitated with anti-Myc antibody in a non-reducing condition, when the pcDNA-transfected cells were labeled with [ 35 S]cysteine, and then treated with diamide (lane 3). In contrast, 17-kDa protein was detected with anti-Myc antibody in a non-reducing condition, when the CSS-expressing cells were labeled with [ 35 S]cysteine, and then treated with diamide (lane 7). However, the corresponding protein disappeared, when the precipitate was reduced by DTT (lane 8) or in the absence of diamide treatment (lane 4). Also, no protein was detected by the immunoprecipitation with NRS (lanes 5 and 6). The similar results were obtained by the experiment with SCS (lanes 9 -13). Together, these results strongly suggest that CSS and SCS form a complex with [ 35 S]glutathione via reversible disulfide bond in cells under oxidative stress with diamide, although 35 S incorporation into the mutants may imply not only the incorporation of [ 35 S]glutathione but also that of [ 35 S]cysteine itself (S-thiolation).
We next examined subcellular localization of Myc-tagged CSS 42 h after transfection when the above metabolic labeling experiments were conducted. Fig. 7, A and C, show the staining of CSSand pcDNA-transfected CHO cells with anti-Myc antibody, respectively. Fig. 7, B and D, are the nuclear staining with Hoechst 33258. It was clearly shown that Myc-tagged CSS was dominantly present in the nucleus; two nuclei out of four were stained with anti-Myc antibody (Fig. 7, A and B). The two non-stained nuclei would represent the absence of CSS-expressing plasmid in the cells. When the cells were transfected with pcDNA, no remarkable staining with anti-Myc antibody was detected in nucleus and cytoplasm (C and D). The similar results were obtained by the experiment with SCS (data not shown). These observations are consistent with the previous report showing that the endogenous Pax-8 is detected in the nucleus of dog thyrocytes (31) and with the report demonstrating the presence of nuclear translocation signal in PD (32). These results indicate that glutathionylation of Pax PD may occur in the nucleus. DISCUSSION Our previous studies defined Cys-45 and Cys-57 in PD as redox-reactive cysteine residues that are responsible for the reversible oxidative inactivation of Pax-8 binding (15). The present study expanded this finding. Using CSS and SCS mutants, it was clearly demonstrated that oxidation of either one of the cysteine residues results in loss of binding (Fig. 2).
To prove the glutathionylation of the cysteine residues, we investigated GSH/GSSG-dependent glutathionylation using purified PD mutants. A decrease in the GSH/GSSG ratio induced the inhibition of CSS and SCS binding, whereas SSC and SSS binding was not affected (Fig. 3). In addition, [ 3 H]glutathione was only incorporated into CSS and SCS (Fig. 4). These results strongly indicated that GSH/GSSG-dependent glutathionylation of Cys-45 and Cys-57 induces oxidative inactivation of PD. Since the corresponding cysteines in Pax-6 and Drosophila paired (prd) are shown to contact with the DNA backbone (33,34), glutathionylation may directly interfere with the DNA contact of PD. Accordingly, it was reported that the mutation of Cys-57 in human Pax-8, which corresponds to rat Cys-57, to tyrosine abolished DNA binding and led to congenital hypothyroidism (35), also indicating that the large side group of tyrosine may interfere with DNA binding. Note that the glutathionylation-dependent inhibition of PD binding was reversed by DTT, indicating the reversible control of PD by glutathione (Fig. 3). Furthermore, it was suggested that glutathionylation of CSS and SCS occurs in cells in response to treatment with 1 mM diamide for 5 min (Fig. 6). Since it was reported that treatment of retina pigmental epithelium cells with 0.5 mM diamide for 5 min induced a marked decrease in the GSH/GSSG ratio to less than 1 (36), the glutathionylation of CSS and SCS in cells may occur in a GSH/GSSG-dependent manner as it does in vitro.
Recently, several studies demonstrated that a change in intracellular redox potential influences thyroid cell function. Our previous study showed that the oxidative stress induced by redox-active copper attenuates thyroperoxidase and Pax-8 expression and stimulates cell proliferation (37). Lonigro et al. (38) demonstrated that a decline in intracellular GSH level by an inhibitor for ␥-glutamylcysteine synthetase is associated with decreased thyroglobulin expression and Pax-8 binding in FRTL-5 cells. This observation is of particular interest for us because the GSH/GSSG-dependent oxidative inactivation of Pax-8 could account for the decreased thyroglobulin expression in the cells with a low level of GSH.
To date, a number of proteins such as protein phosphatase 1B, protein phosphatase 2A, creatine kinase, protein kinase C, protein kinase A, thioredoxin, NF-1, c-Jun, and p50 NF-B have been proven to be glutathionylated in vitro and/or in vivo. Furthermore, recent studies demonstrated that a large number of proteins undergo glutathionylation in response to oxidative stress (39,40). Thus, the role of glutathionylation has been considered to protect critical cysteine residue from irreversible modification such as sulfinic or sulfonic acid by oxidative stress.
Reactive oxygen species such as superoxide, hydrogen peroxide, and hydroxyl radical are generated in cells as a result of normal aerobic metabolism as well as oxidative stress. Thyroid follicular cells are unique in that they produce hydrogen peroxide for thyroid hormone synthesis by a thyroid-specific system, and its generation is increased by thyrotropin (41). Therefore, one of the plausible roles of glutathionylation of Pax-8 PD may be the protection of Cys-45 and Cys-57 against oxidative damage by reactive oxygen species. On the other hand, we and others demonstrate that thyrotropin stimulates expression of various anti-oxidative, reducing enzymes such as Ref-1, thioredoxin, and peroxiredoxin (12,13,42,43). These proteins serve as a reactive oxygen species-scavenger and/or as a modulator for redox-sensitive proteins. Indeed, this study showed that Ref-1 can enzymatically reduce the glutathionylated PD and restore its binding (Fig. 5). Therefore, the present findings of oxidative inactivation of Pax-8 by glutathionylation and its reversal by Ref-1 may provide a molecular basis for thyrotropindependent activation of Pax-8 binding.
How does Ref-1 reduce the glutathionylated PD? It was demonstrated that the reduction of two cysteine residues (Cys-65 and Cys-93) in human Ref-1 is required for reducing c-Jun (4). Hydrogen derived from these cysteines would be provided to c-Jun because the binding of oxidized c-Jun was restored by the addition of only reduced Ref-1, where no other proteins nor reducing equivalents were present. A similar observation was reported by others (44). Therefore, it is likely that Ref-1 itself releases hydrogen and enzymatically provides it to glutathionylated PD. In our assay, 20 M Ref-1 was an excess over 30 nM CSS and SCS, suggesting that Ref-1 can reduce most of glutathionylated PD present in the assay, without the reduction of the oxidized Ref-1.
In conclusion, the cysteine residues corresponding to Cys-45 and Cys-57 and the surrounding amino acid sequences are well conserved in all Pax family members (9). This may imply that the thiols of the conserved cysteine residues are highly reactive, as are those of Cys-45 and Cys-57 in Pax-8 PD. Thus, the glutathionylation, a novel posttranslational modification for PD, would provide a new insight into the molecular basis for the modulation of Pax function.