Hormone-induced Translocation of Thyroid Hormone Receptors in Living Cells Visualized Using a Receptor Green Fluorescent Protein Chimera*

Thyroid hormone nuclear receptors (TRs) are ligand-dependent transcription factors that regulate growth, differentiation, and development. To understand the role of the hormone, 3,3′,5-triiodo-l-thyronine (T3), in the nuclear translocation and targeting of TRs to the regulatory sites in chromatin, we appended green fluorescent protein (GFP) to the human TR subtype β1 (TRβ1). The fusion of GFP to the amino terminus of TRβ1 protein did not alter T3 binding or transcriptional activities of the receptor. The subcellular localization of GFP-TRβ1 in living cells was visualized by laser-scanning confocal microscopy. In the presence of T3, the expressed GFP-TRβ1 was predominately localized in the nucleus, exhibiting a nuclear/cytoplasmic ratio of ∼5.5. No GFP-TRβ1 was detected in the nucleolus. In the absence of T3, more GFP-TRβ1 was present in the cytoplasm, exhibiting a nuclear/cytoplasmic ratio of ∼1.5. In these cells, cytoplasmic GFP-TRβ1 could be induced to enter the nucleus by T3. The T3-induced translocation was blocked when Lys184-Arg185 in domain D of TRβ1 was mutated to Ala184-Ala185. Furthermore, the inability of the mutant TR to translocate to the nucleus correlated with the loss of most of its transcriptional activity. These results suggest that TR functions may, in part, be regulated by T3-induced nuclear entry.

To function as transcription factors, TRs have to interact with transcriptional machinery in the nucleus. However, the process by which TRs are targeted to the nucleus is poorly understood. Before the genes encoding TRs were isolated, high affinity, low capacity T 3 binding sites were detected in the nuclear fractions of tissues and cultured cells by subcellular fractionation (5)(6)(7)(8)(9). Subsequently, when anti-TR antibodies became available, TRs were found only in the nuclei of fixed cells by immunocytochemistry and immunohistochemistry (10 -12). However, in these studies, dynamic cytoplasmic nuclear trafficking of TRs and the effect of T 3 on nuclear trafficking were not addressed. In the present study, we appended the green fluorescence protein (GFP) to the human TR subtype ␤1 (TR␤1) allowing direct examination of the nuclear transport of TRs in living cells. TR exhibited both constitutive nuclear localization and hormone-dependent nuclear localization. Furthermore, we identified a nuclear localization signal in D domain that mediates the T 3 -dependent cytoplasm-nucleus translocation.

Construction of the Plasmid Containing the cDNA Encoding GFP-TR
Fusion Protein-The plasmid pCI-nGFP-C656G containing the GFPglucocorticoid receptor fusion gene was digested with the restriction enzymes BssHI and EcoRI to release the glucocorticoid receptor cDNA (13). The plasmid containing the TR␤1 cloned from human placenta (14) was digested with NcoI and PstI or PstI and EcoRI to obtain the NcoI-PstI and PstI-EcoI DNA fragments, respectively. The two fragments were ligated onto the plasmid pCI-nGFP BssHII-EcoRI vector using a BssHI-NcoI adaptor. The resultant plasmid is denoted as pGFP-TR␤1.
Electrophoresis Gel Mobility Shift Assay-Electrophoresis gel mobility shift assay was carried out as described (15,16). Briefly, equal amounts of in vitro translated GFP-TR␤1 or GFP-TR mutants (TNT kit, Promega) were incubated with the [ 32 P]Lys-and Pal-TRE in the binding buffer (25 mM Hepes, pH 7.5, 5 mM MgCl 2 , 4 mM EDTA, 10 mM dithiothreitol, 0.11 mM NaCl, and 0.4 g ssDNA). In some experiments, RXR␤ was added in the reaction mixture (15,16). After incubation for 30 min at 25°C, the reaction mixture was loaded onto a 5% polyacrylamide gel and electrophoresed at 4°C for 2-3 h at a constant voltage of 250 V. The gel was dried and autoradiographed.
Transient Transfection Assays Using CAT Reporters-CV1 cells were transfected with TR␤1 or GFP-TR␤1 together with the TRE-containing reporter plasmids (each 0.2 g) using the LipofectAMINE transfection method according to the manufacturer's procedure. Twelve hours later, the medium was replaced with Dulbecco's modified Eagle's medium containing 10% T 3 -depleted serum for an additional 12 h. Fifteen hours before cells were harvested, T 3 (100 nM) was added to the appropriate dishes. The cells were lysed, and the CAT activity was analyzed as described (16). The CAT activity was normalized for the protein concentrations of the lysates.
Expression of GFP-TR Proteins in CV1 Cells for Evaluation of Cytoplasmic Nuclear Trafficking-CV1 cells were plated 24 h before transfection in Dulbecco's modified Eagle's medium containing 10% fetal calf serum at a density of 4.5 ϫ 10 4 cells in a Lab-Tek chambered cover glass cuvette (Nage Nunc International, Naperville, IL). Cells were transfected with the appropriate plasmids (0.2 g) using the LipofectAMINE transfection method described above. Fifteen hours later, the medium was replaced with Dulbecco's modified Eagle's medium containing 10% T 3 -depleted serum, and cells were further incubated for an additional 12 h. Cells were observed as described below.
Confocal Microscopy and Image Analysis-Microscopy was performed using a Zeiss LSM410 laser-scanning confocal microscope. All images were captured using a 100 ϫ 1.4 numerical aperture objective lens. Image analysis and determination of nuclear/cytoplasmic ratios were carried out using standard functions of the LSM410 software. The ratios were determined by averaging the intensities of 20 random regions within the nucleoplasm and cytoplasm of 50 -60 living cells. All images were scanned at a fixed gain and black level.

RESULTS
Hormone and DNA Binding Activities of GFP-TR␤1-It is known that the COOH-terminal region is essential for the function of TR␤1 (17), whereas the amino-terminal A/B domain is not critically involved in the hormone and transcriptional activity of TR␤1 (16,18). We therefore fused GFP to the amino terminus of TR␤1 (Fig. 1A). The GFP used was a variant derived from the jellyfish Aequorea victoria, containing a serine to threonine substitution at amino acid 65, which makes this variant not only a stronger chromophore but also more resistant to photobleaching than the wild-type GFP (20).
To be certain that the tagging of GFP to TR␤1 did not significantly alter the functional characteristics of TR␤1, we evaluated the hormone and DNA binding activities of GFP-TR␤1. GFP-TR␤1 was prepared by in vitro transcription/translation, and its binding to T 3 was assessed. No significant differences were detected in the competitive displacement curves from the binding of T 3 to TR␤1 or GFP-TR␤1 (data not shown). The dissociation constants (K d ) were determined to be 0.20 Ϯ 0.05 and 0.14 Ϯ 0.04 nM for TR␤1 and GFP-TR␤1, respectively, indicating that tagging the GFP to the amino terminus of TR␤1 did not affect the T 3 binding activity of TR␤1.
We further evaluated the DNA binding activity of GFP-TR␤1 by using two different TREs. Fig. 2A shows that GFP-TR␤1 bound to Lys-TRE not only as a homodimer (lane 3), but also as a heterodimer with the RXR␤ (lane 6). The binding, however, was weaker than the binding of the wild-type TR␤1 to Lys-TRE as compared with the homodimer and heterodimer shown in lanes 2 and 5, respectively. We also examined the binding of GFP-TR␤1 to Pal-TRE. Consistent with the much weaker binding of the wild-type TR␤1 to Pal-TRE as compared with Lys-TRE (Fig. 2B, lane 2), binding of GFP-TR␤1 to Pal-TRE was not detectable under the experimental conditions (Fig. 2B, lane 3). However, GFP-TR␤1 bound to Pal-TRE as a heterodimer with RXR␤ as shown in lane 6 even though its binding was weaker than the wild-type TR␤1 (lane 5).
GFP-TR␤1 Is Transcriptionally Active-To further characterize the functional properties of GFP-TR␤1, we examined the transcriptional activity of GFP-TR␤1 using the CAT reporter system. CV1 cells were co-transfected with the GFP-TR␤1 or the wild-type TR␤1 expression plasmid with the TRE-containing CAT expression vectors. Fig. 3A shows that in using Lys-TRE, the transcriptional activity of GFP-TR␤1 was indistinguishable from that of TR␤1 in that the extent of T 3 -dependent transcriptional activity was not significantly different (bar 6 versus 4), and it also similarly repressed the basal transcriptional activity in the absence of T 3 (bars 5 and 3 versus bar 1). Bars 1 and 2 are controls to indicate the basal transcriptional activity. Similar results were also obtained when Pal-TRE was used. Bars 4 and 6 of Fig. 3B indicate that no significant differences were detected in the T 3 -dependent transactivation activities between the wild-type TR␤1 and GFP-TR␤1. Taken together, these results indicate that GFP-TR␤1 was transcriptionally competent and is a valid probe to study the intracellular movement of TR␤1.
T 3 -induced Nuclear Translocation of GFP-TR␤1-Using fixed cultured cells, T 3 -bound TRs were shown to be localized in the nucleus (12). To understand whether T 3 regulates the nuclear translocation of TRs in living cells, CV1 cells transfected with the GFP-TR␤1 expression vector were grown in T 3 -containing or T 3 -depleted medium, and the expressed GFP-TR␤1 was visualized by confocal microscopy. We determined the intracellular distribution of GFP-TR␤1 by measuring the intensities of 20 random regions in the cytoplasm and nucleus of 50 -60 living cells. The plot of frequency of cells with nuclear/ cytoplasmic ratios is shown in Fig. 4. Fig. 4A shows that in the absence of T 3 , cells had nuclear/cytoplasmic ratios in the range of 1-4 with the highest frequency of cells having the ratio of 2. However, no cell with a ratio higher than 4 was detected. Including the cells exhibiting only cytoplasmic localization, the average nuclear/cytoplasmic ratio was calculated to be ϳ1.5. In contrast, in the presence of T 3 (Fig. 4B), most of the cells had higher nuclear/cytoplasmic ratios, ranging from 2-20, with an average ratio of ϳ5.5. Representative cells showing localization of GFP-TR␤1 in both the nucleus and cytoplasm in the absence of T 3 are shown in Fig. 5A and those in the presence of T 3 showing localization in the nucleus are shown in Fig. 5B. The distribution profiles shown in Fig. 4 were independent of the expression levels of GFP-TR␤1. We had also determined the distribution in cells transfected with a 10-fold higher concentration of GFP-TR␤1 plasmid. When a higher concentration of plasmid was transfected, GFP-TR␤1 was expressed in more cells and the intensity was higher. However, no significant differences in the nuclear/cytoplasmic ratios were detected (data not shown).
The GFP-TR␤1, which was localized in the cytoplasm, could be induced to translocate into the nucleus by T 3 in a time-dependent manner. Fig. 5B shows that the GFP-TR␤1, which was in the cytoplasm shown in Fig. 5A, was translocated into the nucleus after 30 min at 37°C. Fig. 5C shows another representative cell in which GFP-TR␤1 was detected in both the nucleus and the cytoplasm in the absence of T 3 . However, as shown in Fig. 5D, L-thyronine, which is a biologically inactive analog of T 3 , did not induce the translocation of GFP-TR␤1. The distribution of GFP-TR␤1 in the cell shown in Fig. 5D was the same as that in Fig. 5C. Thus, although a major fraction of GFP-TR␤1 is "constitutively" localized in the nucleus, GFP-TR␤1 can also respond to hormone to undergo nuclear translocation.
The Nuclear Localization Signal of TR␤1 Is Located in Domain D-Based on the studies of many nuclear proteins including steroid receptors, the consensus sequence mediating the nuclear translocation has been determined to consist of two clusters of 2-3 basic amino acids separated by two nonbasic amino acids ((K/R)(K/R)XX(K/R)(K/R)(K/R)) (21). Analysis of the amino acid sequence of TR␤1 identified a possible nuclear localization signal in domain D (Lys 184 -Arg 185 -Leu 186 -Ala 187 -Lys 188 -Arg 189 -Lys 190 ). To verify that this indeed mediated the nuclear translocation for TR␤1, we mutated the Lys 184 -Arg 185 to AA. The mutated TR␤1 was fused to GFP to yield GFP-TR2A whose expression was under the same cytomegalovirus promoter as in GFP-TR␤1 (Fig. 1B). In living CV1 cells, which had been cultured in the presence of T 3 , two different distribution patterns of GFP-TR2A were observed. Approximately 20 -30% of the cell population had GFP-TR2A equally distributed in the nucleus and cytoplasm. However, about 70 -80% of the cell population had most of the GFP-TR2A localized in the cytoplasm, indicating the inability of GFP-TR2A to translocate despite the presence of T 3 . Representative cells are shown in Fig. 6B. This is in contrast to GFP-TR␤1 in that all GFP-TR␤1 was located in the nucleus in the presence of T 3 (Figs. 6A and 5B). Panels C and D of Fig. 6 are the corresponding phase contrast micrographs for the cells shown in panels A and B, respectively. We also mutated Lys 188 -Arg 189 -Lys 190 to AAA to obtain GFP-TR5A (Ala 184 -Ala 185 -Leu 186 -Ala 187 -Ala 188 -Ala 189 -Ala 190 ). These additional mutations only slightly increased the cell population having the receptors localized in the cytoplasm (80 -90%). These results indicate that Lys 184 -Arg 185 -Leu 186 -Ala 187 -Lys 188 -Arg 189 -Lys 190 was one of the nuclear localization signals for TR␤1. Furthermore, these observations suggest that Lys 184 -Arg 185 was functionally more important than Lys 188 -Arg 189 -Lys 190 in mediating the translocation of TR␤1.
Correlation of T 3 -induced Nuclear Translocation with the Transcription Activity of TR␤1-Because a subpopulation of GFP-TR␤1 was induced by T 3 to translocate from cytoplasm into the nucleus, we ascertained whether the hormone-induced nuclear translocation is functionally relevant by comparing the transactivation activities between GFP-TR␤1 and GFP-TR2A.

FIG. 3. Comparison of the transcriptional activity of GFP-TR␤1 and GFP-TR2A mediated by Lys-TRE (A) or Pal-TRE (B) in CV1 cells.
The expression vector of TR␤1, GFP-TR␤1, or GFP-TR2A (0.2 g) was co-transfected with the CAT reporter plasmid (0.2 g) into CV1 cells as described under "Materials and Methods." Cell lysates were prepared, and the CAT activity was determined. The CAT activity was normalized to the protein concentration in the lysates. Data are expressed as mean Ϯ S.E. (n ϭ 6), each with duplicates.

FIG. 4. Nuclear/cytoplasmic ratios of GFP-TR␤1 in living cells
in the presence or absence of T 3 . CV1 cells were transfected with the expression vector of GFP-TR␤1 (2 g) as described under "Materials and Methods." Fifteen hours later, the medium was replaced with Dulbecco's modified Eagle's medium containing T 3 -depleted serum with or without 100 nM T 3 . The intensities and intracellular distribution of GFP-TR␤1 were determined after cells were incubated for an additional 12 h. The resulting nuclear/cytoplasmic ratios were measured as described under "Materials and Methods." Bar 8 of Fig. 3 shows that ϳ75 and ϳ90% of T 3 -dependent transactivation activity of GFP-TR2A was lost as compared with that of GFP-TR␤1 (Fig. 3, A and B, bar 4) on Lys-and Pal-TRE, respectively.
The lower T 3 -dependent transactivation activity of GFP-TR2A could be due to its loss of T 3 binding because of mutations, a lower protein expression level in the transfected cells, an inability to bind to TREs, and/or an impairment in nuclear translocation. Therefore, we evaluated these possibilities. We determined the T 3 binding activity of GFP-TR2A. The competitive displacement curve for the binding of GFP-TR2A to T 3 was indistinguishable from those of TR␤1 and GFP-TR␤1 (data not shown). The K d was determined to be 0.1 Ϯ 0.06 nM, which was not significantly different from that of TR␤1 and GFP-TR␤1 (0.20 Ϯ 0.05 and 0.14 Ϯ 0.04 nM, respectively). These data indicate that the lower transactivation activity of GFP-TR2A was not because of its inability to bind T 3 . We then evaluated protein expression levels of GFP-TR␤1 and GFP-TR2A by Western blotting. As shown in Fig. 7, lanes 3 and 4, the expression levels of the two proteins as detected by monoclonal anti-TR␤1 antibody, C4, were quite similar. The identical protein expression levels were further confirmed by using anti-GFP antibodies (Fig. 7, lanes 6 and 7). Therefore, the lower transactivation activity of GFP-TR2A was not because of a lower expression level for the chimeric receptor.
The lower transactivation activity of GFP-TR2A was also not because of its inability to bind to TREs. Lanes 6 and 7 of Fig. 2A show that GFP-TR2A bound to Lys-TRE as a heterodimer with RXR␤ stronger than GFP-TR␤1. On Pal-TRE, GFP-TR2A heterodimerized with RXR␤ equally well as GFP-TR␤1 (Fig. 2B,  lane 7 versus 6). Consistent with much weaker homodimeric binding of TR␤1 to Pal-TRE (15), binding of GFP-TR␤1 and GFP-TR2A to Pal-TRE was not detectable under the experimental conditions (Fig. 2B, lanes 3 and 4). Taken together, these results strongly suggest that the reduction of the transactivation activity of GFP-TR2A was most likely because of failure of cytoplasm to nucleus translocation of GFP-TR2A. DISCUSSION The present study demonstrated that the GFP-TR␤1 localized in the cytoplasm was induced to translocate into the nucleus by T 3 . We identified a nuclear localization signal that mediated the T 3 -induced nuclear translocation. The mechanism by which this signal responded to T 3 for nuclear import is not clear. This signal with the sequence of Lys 184 -Arg 185 -Leu 186 -Ala 187 -Lys 188 -Arg 189 -His 190 is located in the "A-box," which is part of a long ␣-helix (A-helix) in domain D (22). This structure was determined from the RXR␣/TR␤ DNA binding domain heterodimer complexed with a direct repeat TRE. At present, it is unknown whether the structure of this A-helix in the context of the intact TR␤1 undergoes changes upon the binding of T 3 to domain E because no crystallographic studies of the entire TR␤1 are yet available. However, it is reasonable to assume that this A-helix in domain D could undergo T 3induced conformational changes as it is clearly demonstrated that dramatic structural changes occur in the ligand binding domain E of TR␤1 when bound to T 3 (23,24). The T 3 -induced changes of A-helix in the context of the intact TR␤1 could expose the nuclear localization signal to become more accessible to the receptors that bind the nuclear localization signal (25,26), thereby providing additional regulation of the transcriptional activity of TR␤1. Conformational changes and protein folding have been implicated in the translocation of several steroid receptors and are thought to involve molecular chaper- ones. Hsp90 and a number of factors associate with the glucocorticoid receptor shortly after translation (27) and are displaced from the receptor during ligand activation and nuclear translocation (28). Stable complexes between the heat shock proteins and members of the thyroid receptor and retinoic acid receptor families have not been described (27), although Yamamoto and colleagues (29) reported that Hsp90 expression is required for retinoic acid receptors and RXR function in yeast. Thus it appears that the role of the heat shock chaperones in signal transduction by nuclear receptors may be more complex than simple association with the receptors in the cytoplasm. Moreover, whether Hsp90 plays a direct or indirect role in the translocation of TRs is totally unknown.
Consistent with the previous studies using fixed cells (12) and subcellular fractionation (5-9), we found that TRs are localized in the nuclei in the presence of T 3 . However, in the absence of T 3 , we found that the intracellular localization of TRs was heterogeneous with respect to nuclear/cytoplasmic ratios because of the presence of GFP-TR␤1 in the cytoplasm of some cells. This observation is in contrast to the previous reports in which high affinity and low capacity T 3 binding sites (presumably TRs) were not detected in the cytosolic fractions of tissues and GH3 cells by radiolabeled hormone binding studies (5)(6)(7)(8)(9). There are several possible explanations to account for the different observations. First, subcellular fractionation of tissues and cells could lead to the loss of low amounts of the high affinity T 3 binding sites present in the cytosol. Second, the tissues might still contain a sufficiently high level of endogenous T 3 so that all high affinity binding sites are already located in the nucleus. Third, the radiodetection method may not be sensitive enough to detect the low level of TRs in the cytosol.
The use of GFP-labeled TRs to study the trafficking of TRs in living cells clearly offers significant advantages over the subcellular fractionation approach. However, as with all overexpression studies with GFP labeling (e.g. studies of intracellular trafficking of GFP-labeled steroid hormone receptors), other interpretations exist. First, GFP may alter the activity of TRs or its compartmentalization. Second, transport factors may be limiting or insufficient for the overexpressed TRs. This would  7. Analysis of the expression of GFP-TR␤1 and GFP-TR2A proteins in CV1 by Western blotting. An aliquot of lysates (60 g) from cells transfected with TR␤1, GFP-TR␤1, or GFP-TR2A plasmid as described in Fig. 3 was analyzed by Western blotting using monoclonal anti-TR␤1 antibody, C4 (lanes 2, 3, and 4), or monoclonal antibody against GFP (lanes 5, 6, and 7). seem unlikely, because essentially complete nuclear localization is achieved in all cells when ligand is added. The simplest interpretation of the current studies is that the GFP-TR␤1 reflects the behavior of the endogenous receptor. This interpretation is strongly supported by the findings that the hormone binding, DNA binding, and transactivation activities of GFP-TR␤1 are not significantly different from those of the unlabeled TR␤1.
The heterogeneous distribution of TRs with respect to nuclear/cytoplasmic ratios in living cells could be the result of additional levels of control such as the cell cycle. We have observed that synchronization of cells by serum starvation resulted in a more uniform distribution of GFP-TR␤1. 2 T 3 is known to stimulate cell growth (30). In a normal cultured cell population, cells are not synchronized and are in a different stage of cell cycles. In the absence of T 3 , cells may preferentially "arrest" at a particular cell cycle, which necessitates the retention and/or redistribution of TRs in the cytoplasm to properly regulate the activity of TRs. This notion is supported by the correlation of the loss of transcriptional activity with the retention of mutant TR␤1 in the cytoplasm because of the mutation of the nuclear localization signals (GFP-TR2A). The cell cycle-dependent subcellular distribution is not without precedent. E2F-4, a transcription factor involved in cell cycle progression, has been shown to regulate its activity by changing the nuclear/cytoplasmic distribution ratios during cell cycles (31). Furthermore, it has been shown that the retention of glucocorticoid receptors in the nuclei is not only affected by the hormone but also by the cell cycles. Glucocorticoid receptors are efficiently retained within nuclei following hormone-dependent nuclear translocation in the G 0 and S phase of the synchronized cells, but not in cells synchronized in the G 2 phase (32).
The intracellular localization of the transfected rat TR␣1 has been studied by immunocytochemistry in fixed cells (19). In this study, similar to our findings, Lys 135 -Arg 136 -Lys 137 (corresponding to Lys 188 -Arg 189 -Lys 190 in TR␤1) was identified to be one of the nuclear localization signals. However, in contrast to our findings, ligand was not found to affect the localization of the transfected rat TR␣1. There are several possible explanations for this discrepancy. The fixation procedures required for immunolocalization may cause major disruption of cellular architecture, causing a perturbation of the true in vivo receptor distribution. Also, the epitopes necessary for antibody recognition in a specific subcellular compartment may not be accessible in fixed cells to the large antibody complexes. Finally, the hormone-free conditions used for observing the localization of the rat TR␣1 in the studies were not clearly defined. It is entirely possible that the hormone-free conditions could still contain some level of thyroid hormones, which might explain the lack of detection of the cytoplasmic localization.
The findings that TR␤1 could be induced by T 3 to translocate to the nucleus in a subpopulation of cells added another dimension in the regulation of TR actions. Thus, in addition to the types of TREs and the co-regulatory proteins, which are known to regulate transcriptional activity of TRs, the present study demonstrates that the hormone-dependent nuclear trafficking could play an important regulatory role in the functions of TRs.