Adrenal corticosteroids readily enter the brain and exert markedly diverse effects, including stress responses in the target neural cells via two receptor systems, the mineralocorticoid receptor (MR) and the glucocorticoid receptor (GR). It has been shown that the GR and MR are highly colocalized in the hippocampus. Given the differential action of the MR and GR in the hippocampal region, it is important to elucidate how these receptors interact with each other in response to corticosteroids. We investigated the heterodimerization of the MR and GR with green fluorescent protein-based fluorescence resonance energy transfer (FRET) microscopy in living cells with spatiotemporal manner. FRET was evaluated in three ways: (1) ratio imaging; (2) emission spectra; and (3) acceptor photobleaching. FRET analysis demonstrated that cyan fluorescent protein-GR and yellow fluorescent protein-MR form heterodimers after corticosterone (CORT) treatment both in the nucleus of cultured hippocampal neurons and COS-1 cells, whereas they do not form heterodimers in the cytoplasm. The content of the GR-MR heterodimer was higher at 10-6 m CORT than at 10-9 m CORT and reached a maximum level after 60 min of CORT treatment in both cultured hippocampal neurons and COS-1 cells. The distribution pattern of heterodimers in the nucleus of cultured hippocampal neurons was more restricted than that in COS-1 cells. The present study using mutant fusion proteins in nuclear localization signal showed that these corticosteroid receptors are not translocated into the nucleus in the form of heterodimers even after treatment with ligand and thus allow no heterodimerization to take place in the cytoplasm. These results obtained with FRET analyses give new insights into the sites, time course, and effects of ligand concentration on heterodimersization of the GR and MR.
Adrenal corticosteroids [cortisol in humans or corticosterone (CORT) in rodents] exert numerous effects in the CNS that regulate stress response, mood, learning and memory, and various neuroendocrine functions (McEwen et al., 1986; de Kloet, 1991; McEwen, 1991). CORT actions in the brain are mediated by the glucocorticoid receptor (GR) and mineralocorticoid receptor (MR), both of which are ligand-inducible transcriptional factors. The GR and MR show a high degree of colocalization in the hippocampus (Arriza et al., 1988; de Kloet et al., 1998). Because the MR has ∼10-fold higher affinity for CORT than the GR, hippocampal MR responds strongly to CORT (Krozowski and Funder, 1983; Beaumont and Fanestil, 1988; Rupprecht et al., 1993). Thus, in the hippocampus, this one compound, CORT, serves to regulate the two signaling pathways via the MR and GR (Reul and de Kloet, 1985; Pearce and Yamamoto, 1993; Kawata, 1995). The progressive activation of the MR at a low CORT concentration and additional activation of the GR when CORT levels increase can cause extreme alterations of neuronal integrity for responding to stress conditions (Magarinos et al., 1996; Gould et al., 1997) and neuronal functions, such as changes in neuronal excitability (Joels and de Kloet 1992), associated with changes in neuroendocrine regulation and behavior.
In studying these processes, the subcellular dynamics of both receptors are one of the most important issues. Several investigations have followed subcellular trafficking of MR and GR in living cells using green fluorescent protein (GFP) and its color variants (Htun et al., 1996; Fejes-Toth et al., 1998; Nishi et al., 2001; DeFranco, 2002). Furthermore, McNally et al. (2000) showed that the hormone-occupied GR undergoes a rapid exchange between chromatin and the nucleoplasmic compartment by using the fluorescent recovery after photobleaching technique (Reits and Neefjes, 2001). We focused on the spatiotemporal-specific interactions between GR and MR in living cells. It has been demonstrated that these receptors bind as homodimers to the same hormone response elements (HREs) (Umesono and Evans, 1989). However, physiological studies in various systems suggest that the GR and MR also functionally interact with each other (Gomez-Sanchez et al., 1990; Joels and de Kloet, 1992, 1994). Previous investigators reported that GR-MR heterodimerization takes place and indicated that functional interactions between GR and MR using reporter assay (Trapp et al., 1994; Liu et al., 1995; Savory et al., 2001). But these studies examined hetrodimerization only in whole-cell extracts, not in a spatiotemporal-specific manner maintaining intact cell structures. We, therefore, looked for heterodimerization of GR and MR in living cells by using GFP-based fluorescence resonance energy transfer (FRET) microscopy. Our findings indicated that the GR and MR can form heterodimers in the nucleus after ligand treatment. Cotransfection studies of cyan fluorescent protein (CFP)-GR mutant containing mutations in the domain of nuclear localization signal 1 (Savory et al., 1999) and wild-type yellow fluorescent protein (YFP)-MR have shown that the receptors do not form a heterodimer in the cytoplasm even after ligand treatment.
Here, we report for the first time the dynamic interaction between GR and MR in a spatiotemporal-specific manner using the FRET technique in living cultured hippocampal neurons and non-neural cells.
Materials and Methods
The 6RMR vector (provided by Dr. S. J. Watson, Mental Health Research Institute, University of Michigan Medical School, Ann Arbor, MI) containing the rat hippocampus MR cDNA was digested with SalI. The MR cDNA obtained was ligated in frame into the XhoI site in the multiple cloning site of pEGFPC1 or pEYFPC1 (Clontech, Palo Alto, CA) using a ligation kit (Takara, Kusatsu, Japan). The 6RGR vector (provided by Dr. K. R. Yamamoto, Department Biochemistry and Biophysics, University of California, San Francisco, CA) containing the rat liver GR cDNA was digested at the BamHI sites and subcloned into the pGEM-4 vector (Promega, Madison, WI), resulting in pGEM-4-GR. The GR cDNA with a truncated 5′ coding region was isolated from pGEM-4-GR by EaeIBamHI digestion and ligated in frame into the multiple cloning site (Bsp120I and BamHI) of pEGFPC1 or pECFPC1 (Clontech) using a ligation kit (Takara). The mutant of CFP-GR deficient in NL1 function of GR (CFP-GRNL1-) was created by mutating amino acids 513KKK515 of wild-type rat GR to 513NNN515 by site-directed mutagenesis according to a previous study (Savory et al., 2001). The mutagenesis of NL1 of GR was performed by using the QuickChange Site-Directed Mutagenesis kit (Stratagene, La Jolla, CA) according to the manufacturer's instructions, and these mutations were confirmed by DNA sequencing. The pUC-ER6 vector (provided by Dr. M. Muramatsu, Department of Biochemistry, Saitama Medical School, Saitama, Japan) containing rat estrogen receptor α (ERα) cDNA was introduced with XhoI site just upstream of the first ATG using the QuickChange Site-Directed Mutagenesis kit (Stratagene). After cutting with XhoI and EcoRI, the gene was subcloned into pEGFPC1 or pEYFPC1 vector. Positive control for a FRET experiment, a CFP-YFP construct was used in which CFP and YFP were tandemly linked via three glycine residues.
Cell culture and transfection
Dissociated hippocampal primary neuronal cultures were prepared from 18-d-old Sprague Dawley rat fetuses according to a previously reported method (Nishi et al., 1999). Briefly, the rat fetuses were removed from the placenta in a laminar flow hood and transferred to ice-cold dissecting solution (0.8% NaCl, 0.04% KCl, 0.006% Na2HPO4.12H2O, 0.003% KH2PO4, 0.5% glucose, 0.00012% phenol red, 0.0125% penicillin G, and 0.02% streptomycin). The isolated hippocampus was mechanically dissociated by triturating through a fire-polished glass pipette. The dissociated cells were plated on a 35 mm glass-bottomed dish precoated with 0.1 mg/ml polyethylenimine (Sigma, St. Louis, MO) at an initial plating density of 1 × 105 cells/well by adding 200 μl of the cell suspension to each well (glass-bottomed area of 0.78 cm2; Matsunami Glass, Tokyo, Japan). The cultures were maintained in complete neuronal medium (CNM) consisting of 92.5% (v/v) Eagle's MEM (Sigma), 1% (w/v) nonessential amino acids (Invitrogen, Gaithersburg, MD), 0.16% (w/v) glucose, and 5% (v/v) FCS (Sigma) in a CO2 incubator at 37°C with 5% CO2/95% air. COS-1 cells were maintained in DMEM (Sigma), without phenol red, supplemented with 10% FCS overnight in a 4-well multidish with 16 mm diameter (Nunc, Roskilde, Denmark) at an initial plating density of 2 × 104 cells/well in 400 μl of medium.
Plasmid DNA was transiently transfected into cells by a liposome-mediated method using LipofectAMINE PLUS (Invitrogen) according to the manufacturer's instructions. Hippocampal neuronal cells were cultured in CNM for 48 hr and then treated with 1 mm cytosine-b-Ad-arabinofuranoside for 24 hr to suppress the proliferation of glial cells. Hippocampal neuronal cells were cultured in serum-free medium (SFM) without steroids (MEM with 0.16% of glucose, 1% nonessential amino acids, 20 mm putrescine, 15 nm sodium selenite, 5 μg/ml insulin, and 100 μg/ml transferrin) for 48 hr before transfection. For COS-1 cells, the medium was replaced with SFM 2 hr before transfection. Cells were transfected with 200 μl of OPTI MEM (Invitrogen) containing 8 μl of LipofectAMINE solution and 200 ng of plasmid DNA per well of 1 × 105 cells for 5 hr at 37°C. For ligand stimulation, cells were washed in SFM and treated with 10-6 or 10-9 m CORT (Sigma). As a negative control, cells were treated with 10-8 m 17β-Estradiol.
COS-1 cells cotransfected with YFP-MR and CFP-GR were scraped with ice-cold lysis buffer [50 mm Tris, 500 mm NaCl, 2 mm EDTA, 1% NP-40, 0.5% Triton-X, and Complete proteinase inhibitor mixture (Nakalai, Japan), pH 7.5] and centrifuged at 15,000 rpm for 15 min at 4°C. The supernatant was preincubated with pre-cleaned protein G-Sepharose (Pharmacia Biotech, Peapack, NJ). After centrifugation, the supernatant was incubated with anti-MR antibody (Ito et al., 2000) for 1 hr at 4°C and then protein G-Sepharose for 1 hr at 4°C with rocking. After washing the beads three times with lysis buffer, immunoprecipitated proteins were boiled for 5 min in SDS-PAGE sample buffer and subjected to immunoblotting analysis as described previously (Nishi et al., 1999). Briefly, proteins were separated by a 7.5% SDS-PAGE, and samples were electroblotted on polyvinylidene difluoride membranes (Immobilon-P; Millipore, Bedford, MA) by using a semi-dry blotting apparatus (Transblot-SD; Bio-Rad, Hercules, CA). The membranes were incubated with anti-GR (1:5000 dilution) (Morimoto et al., 1996) overnight at 4°C. Secondary goat anti-rabbit-HRP (Bio-Rad) was added at a 1:5000 dilution for 1 hr at room temperature. Blots were visualized using ECL (Amersham, Buckinghamshire, UK).
Time-lapse image acquisition and analysis
For the living cell imaging experiments, the culture medium was replaced with SFM buffered with 20 mm HEPES (Sigma), and the image acquisition was performed in a temperature-controlled room at 37°C. Images were acquired using a Quantix high-resolution, cooled CCD camera (Photometrics, Tucson, AZ) attached to a microscope (IXL70; Olympus, Tokyo, Japan) equipped with an epifluorescence attachment. For the observation of neurons, a 60× objective lens was used, whereas COS-1 cells were observed with a 40× objective lens. GFP fluorescence was observed with a 480 nm excitation filter, 515 nm emission filter, and a 505 nm dichroic mirror (Olympus); YFP fluorescence with a 500AF25 nm excitation filter, a 545AF35 nm emission filter, and a 525DRLP dichroic mirror (Omega Optical, Brattleboro, VT); and CFP fluorescence with a 440AF21 nm excitation filter, a 480AF30 nm emission filter, and a 455DRLP dichroic mirror (Omega Optical). Data were evaluated with image analysis software, MetaMorph and MetaFlour (Universal Imaging West Chester, PA). Confocal scanning laser microscopic images were collected with an LSM 510 confocal microscope (Zeiss, Jena, Germany) equipped with an argon ion laser, a helium-neon laser, and a ×63 oil immersion objective lens (numerical aperture, 1.4; Zeiss).
Protein-protein interactions of CFP-GR and YFP-MR were studied using FRET microscopy in COS-1 cells and cultured hippocampal neurons. FRET was evaluated in three ways: (1) ratio imaging; (2) emission spectra by the Emission Fingerprinting method using LSM 510 META (Zeiss); and (3) acceptor photobleaching. In all FRET experiments, cells showing nearly the same fluorescence intensity of donor and acceptor were selected for analysis.
Ratio imaging. For ratio imaging FRET microscopy, images were taken with the donor filter set for CFP as described above and with a FRET filter set (XF88; Omega Optical), which consisted of a 440AF21 excitation filter for the donor, a 455DRPL dichroic mirror, and a 535AF26 emission filter for the acceptor. Images were captured with both filter sets under identical conditions, and ratio images were calculated by dividing FRET (acceptor-filter image) by CFP (donor image) using MetaMorph software according to the manufacturer's instructions after appropriate background subtraction (i.e., background fluorescence was measured in a space in which no cell was present, and total fluorescence was then subtracted from background fluorescence) (Tanaka et al., 2003). Ratio images were pseudocolored, in which the red range indicated a high ratio and the blue range indicated a low ratio. To prevent the detection of false positive FRET images, the imaging conditions were adjusted to favor donor emission over acceptor emission. We have confirmed that the level of bleed-through of CFP and YFP in our filter sets was very low (Tanaka et al., 2003). The data shown are representative images obtained from three independent experiments.
Emission spectra. For detecting emission spectral changes in FRET imaging, the Emission Fingerprinting method using confocal laser-scanning microscope LSM 510 META (Zeiss) was used. First, spectral signatures of the fluorescence within the specimen were captured by means of lambda stack acquisition with excitation at 458 nm and detection at 10 nm intervals from 458 through 596 nm using an HFT 458/543 dichroic mirror. Several regions of interest (ROI) with a diameter of 2 μm were then randomly selected for obtaining emission spectral patterns, and the mean ratio of fluorescence intensity of 527 nm and 474 nm was calculated from selected ROI at each time point after ligand addition (20 ROI per cell in 10 cells from three independent experiments). Because the levels of protein expression in each cell were not exactly the same, especially between donor and acceptor molecules, we normalized the fluorescence intensity in each cell by dividing the mean ratio of fluorescence intensity after ligand treatment by that before ligand treatment.
Acceptor photobleaching. For acceptor photobleaching, we used the confocal laser-scanning microscope (Kinoshita et al., 2001). Energy transfer was detected as an increase in donor fluorescence (CFP) after photobleaching of the acceptor molecules (YFP). The acceptor was photobleached by using a 514 nm laser for 1 min at maximum power (25 mW) after 60 and 90 min of 10-6 and 10-9 m CORT treatment, respectively. The cells were then subjected to emission spectral analysis as described above for detecting the changes in fluorescence intensity of the donor molecule. The increase in donor fluorescence intensity was shown as a percentage. We analyzed 10 cells from three independent experiments.
All results are expressed as means ± SEM values. The significance of differences between the mean of various groups was determined by Student's t test for independent samples.
Properties of fusion proteins
The mutants of CFP-GRNL1-, wild-type CFP-GR, YFP-MRNL-, and wild-type YFP-MR were transiently transfected to COS-1 cells and analyzed by immunoblotting. Fusion proteins of CFP-GRNL1- showed the expected molecular mass of 110 kDa, the same molecular mass as that of the wild-type protein, when transiently transfected into COS-1 cells (Fig. 1A). We thus confirmed that the DNA sequences of mutation sites in chimera construct were correct.
We have shown previously that GFP-GR (Nishi et al., 1999), GFP-MR, CFP-GR, YFP-MR (Nishi et al., 2001), CFP-ERα, and YFP-ERα (Matsuda et al., 2002) fusion proteins retained their normal expression patterns, expected molecular masses, and functional abilities when transiently transfected into COS-1 cells.
To look for heterodimerization of the GR and MR, we analyzed the direct binding of GR and MR by immunoprecipitation in COS-1 cells transiently coexpressing CFP-GR and YFP-MR. In transfected COS-1 cells treated with 10-6 M CORT, anti-MR antibody coimmunoprecipitated GR from cell extracts, as detected by immunoblotting with anti-GR antibody (Fig. 1B, lane 1). Coimmunoprecipitation did not occur in the absence of CORT (Fig. 1B, lane 2).
Positive and negative control for FRET analysis
We used the GFP-based FRET technique to examine direct GR-MR interactions in living cells. The low level of CFP and YFP bleed-through enabled us to selectively and accurately measure the amounts of CFP-GR and YFP-MR coexpressed in the same cell. Ratio images were presented as pseudocolor, in which the red range indicated a high ratio and the blue range indicated a low ratio. We set up the range of color hue between 1.4 (the highest) and 0.4 (the lowest). As a positive control, the ratio image of COS-1 cells transfected with a CFP-YFP construct, in which CFP and YFP were tandemly linked via three glycine residues, showed a strong red hue in both the cytoplasm and the nucleus, indicating a positive FRET sign (Fig. 2A). As a negative control, COS-1 cells cotransfected independently with CFP and YFP showed a green hue in both the cytoplasm and the nucleus, indicating a negative FRET sign (Fig. 2B). To semiquantitatively evaluate spectral changes in FRET experiments, we applied a new method of Emission Fingerprinting with LSM 510 META (see Materials and Methods). Figure 2C1 shows the typical spectral pattern obtained from a ROI in a COS-1 cell expressing CFP-YFP. The mean ratio of fluorescence intensity of 527 and 474 nm (the emission maximum of YFP and CFP, respectively) was 2.06 ± 0.22. We used this spectral pattern as a positive FRET sign in the following FRET analyses. To confirm the occurrence of FRET, we performed acceptor photobleaching, in which energy transfer was detected as an increase in donor (CFP) fluorescence after photobleaching of the acceptor (YFP) (Kinoshita et al., 2001; Mochizuki et al., 2001). YFP was photobleached with 514 nm laser at maximum power for 1 min, and then its spectral image was detected by creating emission spectra. When the same ROI used to create the spectrum shown in Figure 2C2 were photobleached, the mean reduction in acceptor fluorescence intensity was 27.3 ± 2.2% and the increase in donor fluorescence intensity was 22.5 ± 5.3% (see Fig. 6). In the negative control, in contrast, there was no prominent peak at 527 nm (Fig. 2D1), and the mean ratio of fluorescence intensity of 527 and 474 nm was 0.89 ± 0.08. Furthermore, no fluorescence recovery of the donor was detected after acceptor bleaching (Figs. 2D2, 6), indicating that FRET did not occur. In both the positive and negative controls, no spectral changes were observed after ligand treatment (data not shown).
FRET analysis in COS-1 cells
Time-lapse ratio images of COS-1 cells coexpressing CFP-GR and YFP-MR revealed that both receptors resided predominantly in the cytoplasm in the absence of ligand, and a green hue diffusely distributed in the cytoplasm indicated a negative FRET sign (Fig. 3A,C). Although both receptors began nuclear translocation after ligand addition, the ratio image in the cytoplasm at 5 min after treatment with 10-6 m CORT still showed mostly green hue, indicating that the receptors did not form heterodimers in the cytoplasm even after ligand binding (Fig. 3A). After 30 min, more than half of the receptors resided in the nucleus, and a red hue was detected in the nucleus. After 60 min, however, the density of red hue in the nucleus became higher and a red hue was homogeneously distributed throughout the nucleus, indicating an increase in the amount of heterodimerization of CFP-GR and YFP-MR throughout the entire nucleus (Fig. 3A). The brightness of color hue after 60 min was also higher than that after 30 min. Cells cotransfected with CFP-GR and YFP-ERα or with YFP-MR and CFP-ERα did not produce a FRET sign when treated with CORT (data not shown). Figure 3B1 shows a typical spectral pattern obtained in the cytoplasm of COS-1 cells coexpressing CFP-GR and YFP-MR at 0, 5, 30, and 60 min after 10-6 m CORT. The peak at 474 nm was a little higher than that at 527 nm in the case of 0 and 5 min, as was the case in the negative control shown in Figure 2D1. The fluorescence intensity detected at 30 and 60 min in the cytoplasm was very low, because most of the receptors were accumulated in the nucleus at these time points. The mean ratios of fluorescence intensity of 527 and 474 nm detected from the cytoplasm in each time point are shown in Figure 5A. Figure 3B2 shows a typical spectral pattern obtained in the nucleus of COS-1 cells coexpressing CFP-GR and YFP-MR at 0, 5, 30, and 60 min after 10-6 m CORT. In contrast to the case of cytoplasm, emission spectra in the nucleus after treatment with 10-6 m CORT for 30 and 60 min showed the same pattern as those observed in the positive control in which a prominent peak was detected at 527 nm (Fig. 2C1). The mean ratios of fluorescence intensity of 527 and 474 nm detected from the nucleus in each time point are shown in Figure 5B. The mean ratio of fluorescence intensity of 527 and 474 nm gradually increased after ligand treatment in the nucleus and reached a maximum level at 60 min. In contrast, the mean ratios in the cytoplasm were not significantly changed after ligand treatment and showed a low level of ∼0.8, similar to that of negative control indicating that FRET was not taken place in the cytoplasm. When the same ROI was subjected to acceptor bleaching at 60 min after ligand treatment, the acceptor fluorescence peak at 527 nm decreased by 36 ± 2.4%, whereas the donor fluorescence peak at 474 nm increased by 27 ± 7.6%, confirming that FRET had actually taken place (Figs. 3B3, 6).
To investigate the effects of CORT concentration on heterodimer formation, we treated the cells with a lower concentration of CORT, 10-9 m. Under this condition, the density of red hue after ligand treatment was less than the case of 10-6 m CORT in ratio image (Fig. 3C), and the mean ratios of fluorescence intensity of 527 and 474 nm detected from the nucleus showed lower values than those detected at 10-6 m CORT (Fig. 5B). Significant difference in the mean ratio between 10-6 and 10-9 m was obtained after 60 min of CORT treatment. These results indicated that the amount of GR-MR heterodimer was higher at 10-6 m CORT than at 10-9 m CORT.
FRET analysis in cultured hippocampal neurons
We performed the same experiments in the cultured hippocampal neurons. In the absence of ligand, both receptors resided mainly in the cytoplasm, and a green hue was mostly observed in the ratio image, indicating no heterodimerization (Fig. 4A). There was still very few signs of heterodimerization in the cytoplasm at 5 min after treatment with 10-6 m CORT. After 30 min, a red hue was detected in the nucleus. The density of the red hue increased at 60 min after CORT treatment. Furthermore, the distribution of red hue in the nucleus was more heterogeneous and located in more restricted regions compared with COS-1 cells (Fig. 4A). Figure 4B1 shows a typical spectral pattern obtained in the cytoplasm of cultured hippocampal neurons coexpressing CFP-GR and YFP-MR at 0, 5, 30, and 60 min after 10-6 m CORT. The peak at 474 nm was a little higher than that at 527 nm in the case of 0 and 5 min, as was the case in the negative control shown in Figure 2D1. The fluorescent intensity detected at 30 min and 60 min in the cytoplasm was very low, because most of the receptors were accumulated in the nucleus at these time points. The mean ratios of fluorescence intensity of 527 and 474 nm detected from the cytoplasm in each time point were shown in Figure 5A. Figure 4B2 shows a typical spectral pattern obtained in the nucleus of cultured hippocampal neurons coexpressing CFP-GR and YFP-MR at 0, 5, 30, and 60 min after 10-6 m CORT. In contrast to the case of cytoplasm, emission spectra in the nucleus after treatment with 10-6 m CORT for 30 and 60 min showed the same pattern as those observed in the positive control in which a prominent peak was detected at 527 nm (Fig. 2C1). The mean ratios of fluorescence intensity of 527 and 474 nm detected from the nucleus in each time point are shown in Figure 5B. The mean ratio of fluorescence intensity of 527 and 474 nm gradually increased after ligand treatment in the nucleus and reached a maximum level at 60 min, whereas the mean ratios in the cytoplasm were not significantly changed after ligand treatment and showed low levels similar to that of the negative control, indicating that FRET did not occur in the cytoplasm. When the same ROI detected at 60 min after ligand treatment was subjected to acceptor bleaching, the acceptor fluorescence peak at 527 nm decreased by 32 ± 2.4%, whereas the donor fluorescence peak at 474 nm increased by 20 ± 6%, verifying that FRET had actually taken place (Figs. 4B3, 6). In the case of 10-9 m CORT, we detected less red hue (Fig. 4C) and lower mean ratio values with spectral analysis (Fig. 5B) in the nucleus compared with those detected at 10-6 m. A significant difference in the mean ratio between 10-6 and 10-9 m was obtained after 60 min of CORT treatment. These results indicated that the amount of GR-MR heterodimer in the cultured hippocampal neurons was higher at 10-6 m CORT than at 10-9 m CORT.
To compensate for different levels of protein expression in different cells, we divided the mean ratio of fluorescence intensity of 527 and 474 nm at the indicated times after ligand addition by the mean ratio of fluorescence intensity of 527 and 474 nm before ligand treatment in each cell. The result, which we call the normalized ratio, reflected more precise changes in fluorescence intensity before and after ligand treatment (Table 1). The normalized ratio at 10-6 m CORT is higher than that at 10-9 m CORT in both COS-1 cells and cultured hippocampal neurons. Compared with COS-1 cells, the increase in the ratio of cultured hippocampal neurons was smaller at both concentrations of CORT.
Subcellular trafficking analysis with nuclear localization signal mutant
To confirm our observations, we used the mutant CFP-GRNL1- that lacked nuclear translocation ability. When this construct was singly transfected into COS-1 cells, they resided in the cytoplasm in the absence and the presence of ligand (data not shown). When CFP-GRNL1- was cotransfected with wild-type YFP-MR, wild-type YFP-MR was completely translocated to the nucleus 60 min after 10-6 m CORT treatment, whereas mutated CFP-GRNL1- showed very little nuclear translocation (Fig. 7A). The ratio image of FRET analysis showed that only a very low FRET sign was being detected in COS cells coexpressing CFP-GRNL1- and wild-type YFP-MR at 5 min after treatment with CORT. Because the dimerization domains of mutant construct are intact (Savory et al., 2001), these results suggest that GR and MR do not form heterodimers in the cytoplasm even after ligand treatment.
We reported previously that the GR and MR are rapidly translocated from the cytoplasm to the nucleus in neurons and non-neural cells in a ligand-dependent manner (Nishi et al., 1999, 2001). Recently, we also clarified that importin α and β are involved in the nuclear translocation of GR and MR and the spatiotemporal relationships of these molecules in living cells using the FRET technique (Tanaka et al., 2003; M. Tanaka, M. Nishi, M. Kawata, unpublished observations).
In the present study, we focused on the dynamic interaction between the GR and MR in living cells. Heterodimerization between transcription factors is not uncommon and seems to increase the level of functional diversity (Forman and Samuels, 1990; Power et al., 1992). Likewise, the formation of heterodimers between members of the nuclear receptor superfamily is a common property. Interactions have been reported between the retinoic X receptor and the retinoic acid receptor, and between the vitamin D receptor and the thyroid receptor (Kliewer et al., 1992). The same could be true for the case of GR and MR. Previous molecular biological studies have indicated that in cells expressing only one of the receptors, transcriptional regulation from HREs, many of which are imperfect inverted hexanucleotide repeats, is mediated by receptor homodimers (Dahlman et al., 1989; Umesono and Evans, 1989; Forman et al., 1995). However, physiological studies in various systems also suggest that the GR and MR functionally interact with each other (Gomez-Sanchez et al., 1990; Joels and de Kloet, 1992). Biogenetic evidence demonstrated that the GR (Umesono and Evans, 1989) and the MR (Govindan et al., 1991) form homodimers through a dimer interface within their zinc finger regions (ZFRs), and these receptors share complete sequence identity in this ZFR dimer interface, suggesting that this region might mediate heterodimerization as well (Liu et al., 1995). To look for such an interaction in a spatiotemporal-specific manner, we conducted a FRET analysis coupled with a new technique called spectral imaging fluorescence microscopy (Lansford et al., 2001; Hiraoka et al., 2002) to compensate for varying levels of protein expression. This technique allowed us to detect spectral changes in fluorescence in living cells and to address several controversial points of intermolecular FRET (Miyawaki and Tsien, 2000). We calculated the mean ratios of fluorescence intensity of acceptor and donor emission maximum wavelengths, 527 and 474 nm, respectively. We divided the mean ratio of fluorescence intensity of 527 and 474 nm at the indicated times after ligand addition by that detected before ligand treatment. The result, which we call the normalized ratio, reflected more precise changes in fluorescence intensity before and after ligand treatment. By using these methods, we were able to demonstrate that CFP-GR and YFP-MR directly interact with each other in the nucleus but not in the cytoplasm, after treatment with CORT, both in COS-1 cells and cultured hippocampal neurons. The degree of heterodimer formation gradually increased in the nucleus and reached a maximum level after 60 min of CORT treatment in both cell types. These results suggest that heterodimer formation depends on the content of GR and MR in the nucleus, because both receptors are completely accumulated in the nucleus after 60 min. Because there is the possibility that the various functions of the GR and MR may reflect the differences in their affinity for the common ligand, CORT, we then investigated whether GR and MR heterodimer formation is affected by varying concentrations of CORT.
Particularly in structures such as the hippocampus, where both GR and MR are coexpressed in the same cells (van Steensel et al., 1996), heterodimerization of these receptors may have a decisive influence on the regulation of corticosteroid-responsive genes in the brain. We used two different concentrations, 10-6 m CORT, which is much higher than physiological concentrations, and 10-9 m, which is between the Kd values of the MR and GR. We found that the amount of heterodimer of CFP-GR and YFP-MR detected at 10-6 m was higher than that at 10-9 m. These results suggest that MR, with a 10-fold higher affinity than GR, may predominantly form homodimers at lower concentrations, whereas at higher concentrations mimicking stressful conditions, the predominance of GR increases to promote the incidence of heterodimerization. Our findings are consistent with the previous demonstrations that MR is predominantly active at lower concentrations to exploit tonic influences, whereas the additional occupancy of GR with higher levels of CORT mediates the feedback actions to restore disturbances of homeostasis (Magarinos et al., 1996; de Kloet et al., 1998). We also observed that GR-MR heterodimers were distributed in more restricted regions of the nucleus in cultured hippocampal neurons than in COS-1 cells. Because endogenous GR and MR are expressed in cultured hippocampal neurons (Bohn et al., 1994), but not in COS-1 cells (Nishi et al., 1999), the heterodimers of GR and MR in cultured hippocampal neurons may present a more physiological distribution pattern, restricting to HREs of targeting genes, than in the case of COS-1 cells.
The results obtained by FRET experiments were supported by trafficking analyses using nuclear translocation mutant of GR, in which the dimerization domains remained intact. Our results suggest that heterodimer formation between the GR and MR is required for corticosteroid-mediated transcription in the nuclear region but not essentially involved in the regulation of nuclear translocation of these receptors. Activated GR or MR may block the activity of other transcription factors, such as activator protein-1 and nuclear factor κB, by direct protein-protein interaction rather than by dimer formation (Ray and Prefontaine, 1994; Guardiola-Diaz et al., 1996; Webster and Cidlowski, 1999). For such interactions to occur, the receptors may have the flexibility to exist as monomers in the nucleus under certain conditions. In contrast to our findings, Savory et al. (2001) reported that the GR and MR formed both heterodimers and homodimers through alternative dimerization interfaces in the cytoplasm. They used myc-tagged GRNL1- mutants, examined receptor redistribution in fixed COS-7 cells, and then detected GR immunoreactivity using an anti-myc antibody. In their experiments, wild-type receptor and mutated receptor constructs were transfected at least 4:1, whereas equal amounts were transfected in the present study. Although the discrepancy between their results and ours may be attributed to the different detection systems and expression conditions of the receptors, the reason for this discrepancy remains unclear.
The physiological significance of the formation of GR-MR heterodimers has been proposed from the colocalization of these receptors in a variety of tissues and cells (Bohn et al., 1994; de Kloet et al., 1998). Hence, having two types of receptors may allow a more flexible response to widely varying corticosteroid concentrations that may be present under physiological and pathological conditions (Trapp and Holsber, 1996). When both GR and MR are expressed in the same cell like hippocampal neurons, their relative concentrations and the ligand concentration will define the proportion of each receptor dimer with its particular DNA-binding and transcriptional activities. The availability of a variety of corticosteroid receptor dimers gives the cell the potential to provide a more finely orchestrated regulation of corticosteroid-responsive genes than the previous model of corticosteroid action based on homodimerization (Evans and Arriza, 1989), although the real functional role of heterodimerization in vivo remains controversial. Trapp et al. (1994) indicated that GR and MR activate transcription synergistically through heterodimerization, whereas Liu et al. (1995) reported inhibitory regulation through heterodimerization. A recent study showed the strong expression of 5-HT1A receptors in MR-enriched hippocampal CA1 pyramidal cells and the cooperative interactions between GR and MR at the 5-HT1A-negative glucocorticoid response element (Ou et al., 2001). These findings suggest that the heterodimerization of these receptors is the key mechanism for inhibitory regulation by corticosteroid of the 5-HT1A receptor gene in the brain (Ou et al., 2001). The explanation for this opposite effect and the role of heterodimerization is still ambiguous, and the differences in transfection conditions, cell types, reporters, or promoters must be taken into consideration. Transcriptional cofactors are also another very important player to consider when evaluating the transcription activities of the GR and MR (Amazit et al., 2003), which were not discussed in the previous studies reporting heterodimer of GR and MR.
In conclusion, FRET analysis clearly demonstrated that GR and MR physically interact with each other in the nucleus of living cultured hippocampal neurons and non-neural cells after CORT treatment. These results indicate that heterodimerization of GR and MR may provide a spatiotemporally specific system for regulating gene expression in response to various cellular environments such as fluctuating CORT concentrations. We demonstrated valuable information that GR and MR form heterodimers in the nucleus after ligand treatment and that this heterodimer formation was more efficient at higher concentrations of CORT, suggesting that GR-MR heterodimers play a more important role under stress conditions. Additional studies will be required to elucidate whether the functional significance of our in vitro observations, including the differences between 10-6 and 10-9 m CORT, also hold true in vivo. Understanding the flexible behavior of corticosteroid receptors, particularly in the structures such as hippocampus that are mainly involved in the regulation of stress, mood, learning, and memory, could be useful for the management of stress.
This work was supported in part by Grants-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science, and Technology to M.N. and M.K. We thank Dr. Satoshi Kida for providing plasmids of CFP-YFP. We are grateful to Dr. Yasunori Hayashi for valuable discussions.
Correspondence should be addressed to Dr. Mayumi Nishi, Department of Anatomy and Neurobiology, Kyoto Prefectural University of Medicine, Kawaramachi Hirokoji, Kamigyo-ku, Kyoto 602-8566, Japan. E-mail:.
Copyright © 2004 Society for Neuroscience 0270-6474/04/244918-10$15.00/0