Mapping of glucocorticoid receptor DNA binding domain surfaces contributing to transrepression of NF-kappa B and induction of apoptosis.

Glucocorticoids (GCs) function, in part, through the ability of the glucocorticoid receptor (GR) to activate gene expression and in part through the transrepression of AP-1 and NF-kappaB. Here we characterize the effect of GR DNA binding domain (DBD) mutations, previously analyzed for changes in the ability to activate gene expression or transrepress AP-1. We have identified a GR mutant capable of distinguishing between transrepression of NF-kappaB and AP-1. Using circular dichroism spectroscopy, we show that this mutation does not appreciably alter GR DBD conformation, suggesting that functional differences between the mutant and wild type protein are the result of an alteration of a specific interaction surface. These data suggest that transrepression of NF-kappaB and AP-1 occurs through distinct protein-protein interactions and argue against the hypothesis that transrepression occurs through competition for a single coactivator protein. Introduction of these mutations into GC-resistant CEM lymphoblastic T cells restored dexamethasone (DEX)-mediated apoptosis as did wild type GR regardless of whether these mutants were transrepression or activation defective. Thus, DEX-mediated apoptosis in transformed T cells is more complex than originally appreciated.

sequently, it was discovered that GR could inhibit the activity of the transcription factor, AP-1, in the absence of a GC-responsive element (GRE) (3)(4)(5)(6). This ability to inhibit directly a transcription factor activity in the absence of a GRE was termed transrepression and was shown to involve a direct physical interaction between GR and AP-1. GCs were also found to inhibit the activity of the transcription factor NF-B (7)(8)(9)(10)(11). As NF-B and AP-1 together regulate genes involved in inflammation and immunosuppression, these discoveries allowed the construction of a model for the therapeutic properties of GCs involving both activation and transrepression (12,13).
Footprinting studies of the AP-1-driven collagenase promoter after GC treatment showed no change in protein occupancy, suggesting that the GR/AP-1 interaction functions to promote a restructuring of the composition of factors binding to the promoter rather than a simple inhibition of promoter occupancy (14). Domain mapping studies identified the zinc finger DNA binding domain (DBD) of GR as essential for transrepression of AP-1, suggesting that the GR DBD is responsible for both transrepression and activation (4 -6). DNA binding and the subsequent activation of gene expression require the dimerization of GR and its binding to a palindromic GRE (15). Careful dissection of the GR DBD, through mutagenesis, demonstrated that disruption of the D loop within the second zinc finger could abolish the ability of GR to dimerize and, consequently, inhibit GC-mediated activation (16). Others (17,18) have shown that mutations in this region, while inhibiting GR-mediated activation, did not affect transrepression of AP-1. Further mutations within the first zinc finger of the GR DBD were identified that disrupted transrepression of AP-1 but did not inhibit the ability of GR to transactivate through a GRE (18), suggesting that activation and transrepression are separable phenomena.
Evidence in support of GR-mediated inhibition of NF-B as one mechanism for the clinical actions of GCs has been generated in various systems (19,20). Transrepression of NF-B by GCs has turned out to be mediated by a number of separate mechanisms, however. NF-B comprises dimers of a protein family that shares homology with c-Rel, including relA (p65), RelB, c-Rel, p50, and p52. Classic NF-B is a dimer of RelA and p50. Most forms of NF-B are held in the cytoplasm by members of the IB family. NF-B can be activated through multiple signaling pathways through the activation of the IB kinase (IKK) and the phosphorylation of IBs. Initial reports of GC-mediated transrepression of NF-B included data demonstrating a physical interaction between NF-B subunits and GR similar to that shown for AP-1 (7,8,10). Subsequently, we and others (9,11) have demonstrated that GCs could induce transcription of the NF-B inhibitor, IB␣. Most recently, it was demonstrated that both NF-B and GR can compete for interactions with coactivators such as p300 and CREB-binding protein (CBP), which are present in limited amounts (21)(22)(23). Interestingly, however, a number of conflicting reports have emerged demonstrating systems in which GCs transrepress NF-B in the absence of IB␣ induction (24 -26). In addition, others have reported that GC-mediated transrepression can occur also when the coactivators p300 and SRC-1 are overexpressed (27). Thus, the molecular basis for GC-mediated transrepression is controversial and may, to some extent, be tissue-specific.
Here, we have begun to address this problem by introducing mutations into the GR DNA binding domain and analyzing the * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ Supported through grants from the Arthritis Foundation and the American Diabetes Association. To whom correspondence should be addressed. Tel.: 303-315-194; Fax: 303-315-0274; E-mail: robert. scheinman@uchsc.edu. 1 The abbreviations used are: GC, glucocorticoids; GR, GC receptor; GRE, GC-responsive element; AP-1, activating protein 1; DBD, DNA binding domain; GST, gluthathione S-transferase; DMEM, Dulbecco's minimal essential medium; DEX, dexamethasone; CAT, chloramphenicol acetyltransferase; PAGE, polyacrylamide gel electrophoresis. effect of these mutations on protein conformation and on transrepression of NF-B. Using this structural information, we are able to propose that transrepression of NF-B and AP-1 is mediated through different interaction surfaces within the GR DBD. Our data argue against the hypothesis that transrepression of both AP-1 and NF-B is mediated by competition with GR for a common factor.
Circular Dichroism-Wild type and mutant GRDBD GST fusion proteins were expressed in DH5␣ and purified by binding to glutathione-Sepharose 4B beads, and the GST domain was cleaved by thrombin digestion (Amersham Pharmacia Biotech) according to the manufacturer's instructions. To remove contaminating bacterial DNA, the GR DBDs were then incubated with 0.2 volumes of Q-Sepharose (Amersham Pharmacia Biotech) at room temperature for 4 h and dialyzed against 2 liters of phosphate-buffered saline overnight.
Far-UV CD spectra were recorded at 20°C in a thermostat-controlled cell holder using an 0.1-cm strain-free quartz cell in an Aviv model 62DS circular dichroism spectropolarimeter. Data was collected from 260 to 180 nm at 0.5-nm intervals using a bandwidth of 1.5 nm and an averaging time of 5 s for each point. The protein concentration was determined by measuring absorbency at 280 nm (A 280 ). The final spectrum for each mutant was generated by averaging the spectra collected from several independent purifications of each protein. Thermal melting curves for each mutant were generated by monitoring ellipticity at 222 nm as a function of increasing temperature between 20 and 95°C at a rate of 1.5°C/min, using a step size of 1°C and an averaging time of 5 s at each temperature.
Cell Culture, Transfection of Cell Lines, and Reporter and Apoptosis Assays-COS-7, CEM C7-14 (GC-sensitive), and CEM C1-15 (GCresistant) cells were maintained at 37°C in a 5% CO 2 humidified atmosphere in Dulbecco's minimal essential medium (DMEM) supplemented with 10% fetal calf serum, 50 g/ml penicillin, and 50 g/ml streptomycin. Transient transfection assays for transrepression were carried out using the LipofectAMINE Plus reagent according to the manufacturer's instructions. For the NF-B transrepression assays, 100-mm plates of 80% confluent COS-7 cells were transfected with 23 g of DNA including 10 g of 3XMHC-luc NF-B reporter construct, 1 g of ␤-galactosidase expression plasmid along with 1 g of RelA and either 10 g of wild type GR or mutant GR expression plasmid. DEX was added to a final concentration of 10 Ϫ6 M at 3 h after the transfection. The cells were harvested 48 h later, and luciferase activity was measured by combining 40 l of cell extract with 160 l of luciferase assay buffer (100 mM KPO 4 , pH 7.8, 15 mM MgSO 4 , 5 mM ATP, 1 mM dithiothreitol). The ␤-galactosidase assay was carried out as described by Rosenthal (28). For the AP-1 transrepression assays, 6-well plates of 80% confluent COS-7 cells were transfected with 1 g of 5XAP1-TATAcat, 100 ng of ␤-galactosidase expression plasmid along with 100 ng of either wild type GR or mutant GR expression plasmid. Endogenous AP-1 activity was induced 3 h post-transfection with 80 ng/ml phorbol 12-myristate 13-acetate in the presence or absence of 10 Ϫ6 M DEX. CAT activity was measured using a CAT enzyme-linked immunosorbent assay kit according to the manufacturer's instructions (Roche Molecular Biochemicals). Relative transrepression was calculated as the percent of decrease in ␤-galactosidase normalized reporter activity after DEX treatment and then normalized to transrepression of either NF-B or AP-1 activity, respectively, by wild type GR. The average transrepression mediated by wild type GR is defined as 1.0.
CEM C1-15 cells were stably transfected by electroporation. Aliquots of 2 ϫ 10 6 cells were suspended in 0.5 ml of DMEM containing 1 mM sodium pyruvate (Life Technologies, Inc.). Using 0.4-cm electroporation cuvettes (Invitrogen), samples were electroporated at a capacitance of 960 microfarads and 50 volts with 5 g of pCI-neo (Promega) and 20 g of wild type GR or GR mutants. Cells were then removed and split into three parts, and each part was cultured in 2 ml of supplemented DMEM for 72 h. The cells were then fed with medium supplemented with G418 (450 g/ml or 900 g/ml, Gemini Bio-Products) for 3 month. The medium was changed every 3-4 days. Apoptosis was assayed morphologically using the acridine orange/ethidium bromide double dye method (29).
Northern Blot Analysis-Total RNA was isolated from cells with TRIzol Regent (Life Technologies, Inc.) per the manufacturer's instruction. RNA integrity and loading consistency were assessed by visual inspection of ribosomal RNA bands after agarose gel electrophoresis and ethidium bromide staining. 20 g of RNA was separated on a 1% formaldehyde-containing agarose gels, transferred to Zeta-probe membranes (Bio-Rad) by standard capillary blotting, and cross-linked by UV irradiation. The cDNA probe for IB␣ was labeled with [␣-32 P]dCTP using a random primer DNA labeling kit (Bio-Rad). The filter was washed twice in 40 mM NaPO 4 , 1 mM EDTA, 1% SDS, and data visualized by PhosphorImager.
Western Blot Analysis-Whole cell lysates were prepared, and 12.5 g of protein from each lysate was electrophoresed through 10% SDS-PAGE and transferred to nitrocellulose by electroblotting (Life Technologies, Inc.

Mutations in the GR DBD Do Not Cause Major Changes in
Conformation-Previously, we showed that GC-mediated transrepression of NF-B requires the DBD but not the 1 activation domain (10). To further dissect this requirement, we tested the DBD mutants S425G and L436V, shown to block transrepression of AP-1 (18), and N454D/A458T, a D loop mutation that blocks GR dimerization and activation of transcription. We began by analyzing the effect of these mutations on GR conformation to determine whether the mutations affected one specific region of the GR DBD or whether the effects of these mutations were spread throughout the domain. This control experiment was critical for the appropriate structural interpretation of the subsequent functional data; however, to our knowledge, it had never been performed in the context of studying GR-mediated transrepression. To this end, we expressed wild type and mutant GR DBD domains as GST fusion proteins in Escherichia coli. The GR DBD proteins were purified, GST domains cleaved, and contaminating DNA removed as described under "Experimental Procedures." The extinction coefficient (⑀ 280 ) of 5470 M Ϫ1 cm Ϫ1 was determined by the method of Gill and von Hipple (30). SDS-PAGE analysis of the mutants generated an apparent molecular mass of 10 kDa, consistent with the theoretical molecular weights of our GR DBD proteins (data not shown).
The structures of wild type and mutant GR DBDs were studied by CD spectroscopy. Fig. 1 shows representative far-UV CD spectra of the DBDs. The strong negative maxima at 222 and 208 nm indicate that the DBDs possess an amount of ␣-helical structure consistent with earlier studies of this domain done by others using NMR spectroscopy and x-ray diffraction (31,32). By visual inspection, the spectra for the S425G mutant appears to deviate slightly in intensity at the wavelengths representing ␣-helix, 208 and 222 nm. However, Transrepression by GR DBD Mutants 2330 measurements of the percentage of ␣-helical structure from three independent preparations, as measured by deconvolution analysis of the spectra, showed that these deviations were not statistically significant (Fig. 1, inset). These data indicate that within the error of the measurement, there are no obvious secondary structural differences between the wild type DBD and any of the mutants. To further test the structural similarity of these DBDs, thermal melting scans were collected on the wild type and each of the mutants. Although the transition was not thermodynamically reversible, the apparent melting temperatures (T m ) for all of the mutants were equivalent within error, indicating that the relative stability for the wild type DBD and the mutants are indistinguishable (data not shown).
In conclusion, CD spectroscopy indicates that no global conformational changes exist in any of the mutants, and any functional changes that result from these mutations must be caused by either specific changes in the nature of the amino acids in an interaction interface region or by subtle local conformational changes of involving only several residues.
Analysis of Transrepression-The mutations S425G, L436V, and N454D/A458T were introduced into our GR expression vector, pYCGR, as described under "Experimental Procedures." We then tested their ability to transrepress NF-B and AP-1 by transient transfection followed by treatment with DEX. The ability of these GR mutants to transrepress NF-B and AP-1 reporter constructs is shown in Fig. 2. We found that the mutations L436V and N454D/A458T had no effect on the ability of GR to transrepress NF-B, whereas the S425G mutation greatly reduced the ability of GR to transrepress NF-B. N454D/A458T-mediated transrepression of AP-1 is consistent with the observation that GR can transrepress AP-1 as a monomer. Here we show that this property of GR can be extended to transrepression of NF-B. In our hands, S425G and L436V both transrepress AP-1. These results are different from that reported by Heck et al. (18), which may be because of differences in the cell lines used by our respective laboratories. Additionally, these differences may be due to amino acid differences caused by the introduction of restriction sites flanking the DBD in their human GR expression plasmid (18). Given that our experiments were performed in one cell line with the same GR constructs, we can state that the requirements for transrepression of AP-1 and NF-B are not identical in that the S425G mutation can discriminate between these two functions. The serine to glycine substitution at position 425 removes an hydroxyl group, possibly altering hydrogen bonding patterns between the zinc finger domain and other proteins, suggesting that this hydrogen bonding pattern is important for the transrepression of NF-B but not AP-1. Serine 425 is not thought to hydrogen bond either with DNA or within the zinc finger. X-ray crystallography data indicate that this amino acid is solventaccessible (32), consistent with the possibility of its contribution to an intermolecular interaction interface. These data indicate that multiple elements within the GR DBD contribute to transrepression and argue against the hypothesis that transrepression primarily involves competition for a common factor, as this would predict that transrepression of NF-B and AP-1 would not be separable phenomena.

DEX-induced Apoptosis Is Restored Both by GR Mutants Deficient in Dimerization and by GR Mutants Deficient in
Transrepression-GCs induce apoptosis in both thymocytes and transformed T cells. To analyze the properties of these GR mutants in the context of T cell apoptosis, we used previously described sister human acute lymphoblastic leukemic T cell lines, GC-sensitive CEM C7-14 and GC-resistant CEM C1-15 (33). We cotransfected CEM C1-15 cultures with pCI-neo along with an expression plasmid encoding either wild type GR or GR, containing one of the various GR mutants described above, and selected stable transfectants by culturing in G418. Although endogenous (nonfunctional) GR migrated identically to transfected GR on SDS-PAGE, we found that GR immunoreactivity was significantly higher in transfected lines, indicating that our transgenes were expressed (Fig. 3A, inset). To investigate whether any of our GR mutants could restore GC sensitivity to apoptosis, our transfected CEM cell lines were treated with DEX for 48 h, and apoptosis was determined morphologically as described under "Experimental Procedures." As expected, DEX induced apoptosis in ϳ25% of the C14 line after 48 h, with little apoptosis in the C15 line (Fig. 3A). Wild type GR restored DEX sensitivity to the C15 line as did the dimerization mutant, N454D/A458T, a result consistent with the results of Helmberg et al. (17). Interestingly, both the S425G and L436V mutants also restored DEX sensitivity to the C15 line. Statistical analysis of apoptosis levels indicated no significant differences in apoptosis between wild type GR transfected C15 cells and C15 cells transfected with S425G, L436V, or N454D/A458T GR mutants.
These data can be interpreted in several ways. As S425G is unable to transrepress NF-B, and yet restores apoptosis, it would suggest that transrepression of NF-B is not required for this process. Conversely, as N454D/A458T is unable to interact efficiently with DNA, and yet restores apoptosis, this would argue that GR-mediated gene expression is not required for this process. One potential resolution of this apparent paradox is to consider the possibility that several independent pathways leading to apoptosis are initiated by GCs.
If, however, N455D/A458T is capable of activating the expression of a subset of GC-responsive genes, an alternative explanation for the data would be that it is this subset of genes that regulates apoptosis. As S425G and L436V are capable of activating gene expression, this alternative hypothesis is consistent with our observations. Evidence supporting this idea was generated by Rogatsky et al. (34), who showed that when a GR dimerization mutant is introduced into GR negative SAOS cells, proliferation is blocked through induction of p21 gene expression. However, evidence against this idea was also generated by directing a D loop mutation to murine embryonic stem cells and establishing a dimerization-deficient GR-expressing mouse (35). This group found that thymocytes derived from these mice were insensitive to DEX-mediated apoptosis.
Induction of IB␣ Is Independent of GR-mediated Transrepression in CEM C1-15 T Cells-Given the above arguments, we wanted to test the ability of our mutant GR lines to activate transcription of an endogenous gene. Because of the interest in GC-mediated regulation of IB␣ expression, we chose to examine the ability of these mutants to regulate IB␣ mRNA abundance by Northern blot analysis. Endogenous IB␣ mRNA expression was unresponsive to DEX treatment in CEM C1-15 cultures lacking functional GR (Fig. 3B). In turn, IB␣ mRNA expression was highly responsive to DEX treatment in CEM C7-14 cultures that contained functional GR. Expression of the wild type GR transgene in CEM C1-15 cells restored the ability of DEX to induce IB␣ mRNA, as expected. In addition, the S425G and L436V mutant transgenes were capable of restoring the ability of DEX induction of IB␣ mRNA. In comparison, the N454D/A458T double mutant was unable to restore DEX induction of IB␣ mRNA, indicating that the IB␣ gene is not part of the subset of GC-responsive genes activated by this mutant.
To our knowledge, we are the first group to analyze the structure of our GR mutants and systematically compare NF-B and AP-1 in the study of GR-mediated transrepression. Through this work, we have identified a region of the first zinc finger, which is differentially involved in transrepression of NF-B and with AP-1. Furthermore, we have shown that DEXmediated apoptosis is dependent solely neither on transrepression nor on activation. Rather, our data suggest that GC-mediated apoptosis in CEM cells is mediated potentially through several independent pathways.