Ubc9 Is a Novel Modulator of the Induction Properties of Glucocorticoid Receptors*

The EC50 of agonists and the partial agonist activity of antagonists are crucial parameters for steroid hormone control of gene expression and endocrine therapies. These parameters have been shown to be modulated by a naturally occurring cis-acting element, called the glucocorticoid modulatory element (GME) that binds two proteins, GMEB-1 and -2. We now present evidence that the GMEBs contact Ubc9, which is the mammalian homolog of a yeast E2 ubiquitin-conjugating enzyme. Ubc9 also binds to glucocorticoid receptors (GRs). Ubc9 displays no intrinsic transactivation activity but modifies both the absolute amount of induced gene product and the fold induction by GR. With high concentrations of GR, added Ubc9 also reduces the EC50 of agonists and increases the partial agonist activity of antagonists in a manner that is independent of the ability of Ubc9 to transfer SUMO-1 (small ubiquitin-like modifier-1) to proteins. This new activity of Ubc9 requires only the ligand binding domain of GR and part of the hinge region. Interestingly, Ubc9 modulation of full-length GR transcriptional properties can be seen in the absence of a GME. This, though, is consistent with the GME acting by increasing the local concentration of Ubc9, which then activates a previously unobserved target in the transcriptional machinery. With high concentrations of Ubc9 and GR, Ubc9 binding to GR appears to be sufficient to permit Ubc9 to act independently of the GME.

A major role of steroid hormones in human physiology is to regulate gene expression during development, differentiation, and homeostasis. The mechanism used to achieve these objectives is basically the same for all members of the steroid receptor superfamily. Ligand enters the cell by passive diffusion and binds to an intracellular receptor protein. After a still poorly understood step called activation, the receptor steroid complexes acquire an increased affinity for DNA and bind to specific DNA sequences called hormone response elements. These DNA-bound receptor complexes recruit an ever increasing number of coactivators, corepressors, and comodulators, the combination of which interacts with the transcriptional machinery to increase or decrease the magnitude of gene transcription (1,2).
The levels of circulating steroids in mammals are usually well below that required for maximal induction of the regulated genes and much closer to that which is required, according to the dose-response curve, to afford half-maximal induction of most genes (i.e. the EC 50 ). This permits relatively small differences in steroid hormone concentration to cause major changes in the levels of gene activation or repression. Conversely, any process that modifies the position of the dose-response curve (or the value of the EC 50 ) will also affect the levels of gene expression (3,4). In extreme conditions, the dose-response curves of two regulated genes can be so different that a physiological concentration of steroid will cause full induction, or repression, of one gene and have no effect on the other. Thus, processes or factors that modify the dose-response curve or EC 50 represent an additional mechanism for achieving differential control of gene expression by steroid hormones in cells and organisms.
A major clinical application of steroid hormones involves antisteroids, which block the actions of the endogenous hormones. Antisteroids, or antagonists, have found widespread use in endocrine therapy of such conditions as breast cancer (5), pregnancy (6), and prostate cancer (7). However, a puzzling feature of antisteroids is that they often display residual amounts of agonist activity that can vary between genes and among cells (8 -10) for the same gene (10 -13). Clearly an understanding of the parameters that affect the partial agonist activity of antisteroids would be of major assistance in their clinical uses.
Several years ago, we described the first cis-acting element that could modify both the dose-response curve of agonist steroids and the partial agonist activity of antisteroids. This element, which we named the glucocorticoid modulatory element (GME), 1 is located at Ϫ3.6 kb of the rat liver tyrosine aminotransferase gene (14,15). The GME is fully active with a heterologous enhancer, promoter, and reporter gene in nonliver cells (14, 16 -21). The binding of a trans-acting factor to the GME correlates with biological activity in that mutations of the GME that reduce binding also inhibit the biological activity (14,16). Unexpectedly, the GME-binding factor (GMEB) was found to be a heteromeric complex of two new proteins, GMEB-1 and -2 (16,(22)(23)(24). Many of the properties of the GMEBs are consistent with their being key components in the modulation of GR transcriptional properties. The GMEBs possess intrinsic transactivation activity, they interact with each other in addition to GRs and CBP, they bind to DNA, they inhibit GR total transactivation, and they cause both a right shift in the dose-response curve of agonists and a decrease in * 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.
‡ To whom correspondence should be addressed: Steroid Hormones Section, Bldg. 8 the partial agonist activity of antagonists, presumably due to squelching (16,22,24). 2 Interestingly, GMEB-1 and -2 each possess a region of 93 amino acids that is 80% identical (24). Furthermore, several other molecules with transcriptional activity have been found to have a high degree of homology with this 93-amino acid region and thus may comprise a family of related molecules (24). The two best characterized members of this possible superfamily are Suppressin (25) and the Drosophila protein called DEAF-1 (26). Suppressin has been described as a global negative regulator of human cell proliferation, and especially the immune system, and appears to be the short form of a protein called NUDR (27). DEAF-1 is a transcriptional cofactor from Drosophila that is required for the activity of a 120-bp response element of the Deformed gene.
To further define the mechanism of action of the GME, we sought to identify those molecules that the GMEBs contact and could thus be involved in modulating GR transcriptional properties. In view of the known ability of the GMEBs to homo-and hetero-oligomerize (16), 2 it is possible that the GMEBs might also heterooligomerize with other members of the KDWK superfamily, such as Suppressin and DEAF-1. Alternatively, a currently unidentified protein may be involved. In this study, we examined Suppressin and DEAF-1 for their ability to interact with the GMEBs and used a yeast two-hybrid screen to isolate other potential interacting proteins. One of the proteins identified in the yeast two-hybrid screen is Ubc9, which is a human homolog of the E2 ubiquitin-conjugating enzymes of yeast. Ubc9 often conjugates an ubiquitin-like molecule, called small ubiquitin-like modifier-1 (SUMO-1), to proteins in vertebrate cells (28,29). Sumoylation of proteins can have a variety of effects, such as inhibiting the ubiquitin-mediated proteolysis of Mdm2 (30), increasing the nuclear translocation of RanGAP1 (31,32) and Dorsal (33), and regulating the transcriptional activity of androgen receptor (AR) (34) and HIPK2 (35). However, Ubc9 also displays many of the same activities with other proteins in a manner that is independent of sumoylation, for example enhancing the transcriptional activity of ETS1 (36) and AR (37) but repressing the activity of TEL (38) and promoting the nuclear translocation of Vsx-1 (39). Thus, Ubc9 appears to have both enzymatic and non-enzymatic effects in cells. Ubc9 also binds to a large number of proteins, possibly due to electrostatic interactions that are mediated by the presence of patches of both positive and negative charges on the surface of Ubc9 (40).
Our studies indicate that neither Suppressin nor a mammalian homolog of DEAF-1 are involved in GME action. Conversely, Ubc9 exhibits all of the characteristics expected for an intermediate that would participate in the ability of the GME to modulate the dose-response curve and partial agonist activity of GR complexes. These effects occur in the absence of sumoylation by Ubc9.
GAL/GR-525C was constructed using PCR amplification of pSVLGR with the 5Ј primer of 5Ј-CGGGATCCGTGGAGTCTCACAAGACACTT-CGGAAAATCCT-3Ј and the 3Ј primer of 5Ј-GCTCTAGATTCATTTTT-GATGAAACAGAAGCTTTTTGATA-3Ј. The resulting PCR products were purified using a Wizard direct gel purification kit, digested with BamHI and XbaI, and ligated into the pM (CLONTECH) vector using the same restriction sites. GAL/GR-407C was similarly prepared using the 5Ј primer of 5Ј-ATGGATCCGGTCAGTGTTTTCTAATG-3Ј and the 3Ј primer of 5Ј-TATCTAGAGTCATTTTTGATGAAACAG-3Ј. The PCR products were cut with BamHI and XhoI and ligated with pM vector at the BamHI/XhoI site. pPC86/Ubc9[1003] is the original clone isolated from the yeast twohybrid screen using GMEB-1 (in pDB-Leu) as the bait and an adult rat liver cDNA library (Invitrogen) in pPC86. The full-length rat Ubc9 (rUbc9) was excised as a SalI-NotI fragment and cloned into the corresponding sites of pCMVSport (Invitrogen) to give pCMVSport/rUbc9. The EcoRI-BamHI fragment of pCMVSport/Ubc9 was cloned into the corresponding sites of pSG5 to provide pSG5/rUbc9. GAL/GMEB-1/pDBleu-The open reading frame of GMEB-1 was PCR-amplified with Pfu polymerase for high fidelity using oligonucleotides SK200 (5Ј-CT GAA TTC ATG TCG ACG ATG GCT AAT GCA G-3Ј) and SK201 (5Ј-CCC TGC GGC CGC CAG TTA ATC CTC TAA GAC-3Ј) and pM/GMEB-1 as the template. The above oligonucleotides were designed to create SalI and NotI at 5Ј and 3Ј ends of the PCR fragment. The amplified fragment was cloned in the above sites of pDB-Leu. The clones obtained were sequenced to confirm the correct frame with Gal-4 DBD of the plasmid.
hSA/pSG5-The 2.3-kb hSA cDNA (Stratagene Liver catalog number 937224), cloned into the EcoRI and XhoI sites of pBluescript SK, was excised with EcoRI and KpnI and inserted into the pFLAG/CMV2 vector. The hSA insert was then excised from hSA/pFLAG/CMV2 using EcoRI and BamHI and ligated into the EcoRI and BamHI sites of pSG5 to yield hSA/pSG5.
as an internal control. Cells were incubated with plasmid DNA, Opti-MEM I, and LipofectAMINE reagent for 24 h, after which this mixture was replaced by the normal media (5% fetal calf serum, Dulbecco's modified Eagle's medium). The cells were induced with steroids 24 h after the start of transfection and then harvested after an additional 24 h. The cells were lysed and assayed for reporter gene activity using the Dual Luciferase assay reagents according to manufacturer's instruction (Promega, Madison, WI). Luciferase activity was measured by an EG & G Berthhold's luminometer (Microlumat LB 96 P).
Yeast Two-hybrid Assay-Small scale and library scale transformations of yeast Mav203 cells with GAL or VP16 constructs of GMEB-1 or -2 were performed using the competent and highly competent cells from Invitrogen as per the manufacturer's instructions. Yeast two-hybrid selection was conducted with GAL/GMEB-1 or -2 and an adult rat liver cDNA library fused to VP16 (Invitrogen). Plasmids were obtained from putative yeast clones according to the instructions for the Proquest two-hybrid system, and the DNA thus obtained was transformed into Escherichia coli (Electromax DH10B) by electroporation according to manufacturer's instructions. Quantitation of ␤-galactosidase activity was accomplished according to the protocol of the Hybrid Hunter version kit (Invitrogen).
Gel Shift Assays-Gel shift assays to detect the binding of GMEB-1, GMEB-2, DEAF-I, Suppressin, and HTC cytosol to a GME oligonucleotide were conducted as described (22). The sequences of the GME and mutant oligonucleotides (M1, M2, and M3) have been described (14). Gel shift assays with the deformed response element were conducted according to Gross et al. (26).
Sumolyation Assay-GMEB-1 or GMEB-2 and RanGAP were in vitro transcribed and translated in absence or presence of purified SUMO-1 or GST-SUMO-1 and Xenopus Ubc9 (gifts from Mary Dasso, NCI) (44). The products were run on a standard 10% SDS-PAGE gel.
Overexpression of GMEBs in Baculovirus-GMEB-1 and GMEB-2 fusion proteins with a FLAG tag at the NH 2 terminus were expressed in Sf9 insect cells using the BAC to BAC TM baculovirus expression system according to manufacturer's instructions (Invitrogen).
Pull-down Assay-M2-agarose beads (anti-FLAG monoclonal antibody conjugated to agarose, Sigma) were prepared by treating them with 100 mM glycine⅐HCl buffer (pH 2.5, 10 volumes) for 2-3 min, which was then neutralized with 1 M Tris⅐HCl (pH 8.0, 10 volumes). The beads were washed once with phosphate-buffered saline, pH 7.4 (BioWhittaker, Walkersville, MD) and once with extraction buffer (10% glycerol, 20 mM Tris⅐HCl, 0.4 mM EDTA, 0.2% Tween, 10 mM mercaptoethanol, and 500 mM KCl). Aliquots (100 l) of Sf9 cell extracts containing baculovirus-expressed FLAG-mUbc9 were incubated with the above prepared M2-agarose beads for 2-3 h at room temperature. The beads were then washed five times with extraction buffer followed by one wash with elution buffer (10% glycerol, 20 mM Tris⅐HCl, 0.4 mM EDTA, 0.2% Tween, 10 mM mercaptoethanol, and 250 mM KCl). The radioactive proteins (labeled with [ 35 S]methionine during in vitro translation in rabbit reticulocyte lysates) were added separately to the M2 beads bound with FLAG-Ubc9 and incubated at room temperature for 1 h and then at 4°C for 12-16 h. The beads were washed with elution buffer five times. Finally the beads were resuspended in 20 l of elution buffer and 20 l of 2ϫ SDS-PAGE buffer, boiled for 3-5 min, and loaded onto SDS-PAGE gels.
Mammalian Two-hybrid Assay-The recommended procedure for the Mammalian Matchmaker two-hybrid assay kit (CLONTECH) was modified slightly by changing from a CAT reporter (pG5CAT) to a luciferase reporter pFRLuc (Stratagene). Smaller amounts of GAL4 and VP16 chimera DNAs were used (0.5 g) because of the higher sensitivity of the luciferase reporter.
Statistics-Unless otherwise noted, all experiments were performed in triplicate several times. The error bars in graphs of individual experiments correspond to the S.D. of the triplicate values. Best-fit curves (R 2 Ն 0.95) for each experiment to yield a single EC 50 value were obtained with KaleidaGraph (Synergy Software, Reading, PA). The values of n independent experiments are analyzed for statistical significance by the two-tailed Student's t test using the program "InStat 2.03" for Macintosh (GraphPad Software, San Diego, CA). In paired analyses, all comparisons in a given figure are performed at the same time with the same seeding of cells and then analyzed for n different experiments.
Given the lack of interaction with the GMEBs by other members of the KDWK family, we used chimeras of GAL-DBD and GMEB-1 or -2 to identify potential interacting proteins in a yeast two-hybrid screen. One clone containing an open reading frame of 528 nucleotides was isolated from an adult rat liver cDNA library that was identical to rat Ubc9. The VP16/ Ubc9 chimera interacts with both GAL-DBD/GMEB constructs, but not GAL-DBD, in yeast plate assays (data not shown). In solution assays, VP16/Ubc9 selectively bound more avidly to GMEB-1 than to GMEB-2 (Fig. 1).
The GAL/Ubc9 chimera has no intrinsic transactivation activity in a one-hybrid assay in mammalian (COS-7) cells (data not shown). Therefore, Ubc9 is not a candidate transcription coactivator.
Sumoylation Activity of Ubc9 -In view of the known ability of Ubc9 to add the ubiquitin-like molecule, SUMO-1, to proteins (28,29) such as RanGAP1 (31) covalent modification by Ubc9 of either GMEB-1 (lanes A-F) or GMEB-2 (lanes G-L) is observed in the presence of either SUMO-1 or GST/SUMO-1. Therefore, Ubc9 does not appear to sumoylate either GMEB-1 or -2. It should be noted that neither GMEB-1 nor GMEB-2 contain the consensus sequence for SUMO-1 modification that is found in many sumoylated proteins (47).
Interaction of Ubc9 with GMEBs and GR-Pull-down assays using immobilized, overexpressed FLAG/Ubc9 were used to obtain direct evidence for Ubc9 binding to GMEB-1 and -2 (Fig.  3A). We also examined the mutant Ubc9/C93S. Ubc9/C93S is unable to sumoylate proteins (29,48), because Cys-93 is required to form a thiolester bond with the COOH terminus of SUMO-1, which is a necessary step in sumoylation (38). Both GMEB-1 and -2 specifically bound to Ubc9 in a manner that is independent of Cys-93 (compare lanes 1, 2, 5, and 6 to lanes 9,  10, 17, and 18). Therefore, the ability of Ubc9 to sumoylate proteins does not appear to be necessary for its interaction with GMEB-1 or -2. Unexpectedly, Ubc9 also binds to GR (lane 3 versus lanes 11 and 19). rGR contains three consensus SUMO-1 modification sites of the sequence ⌿KXE (⌿ ϭ hydrophobic, X ϭ any amino acid) (47) at amino acids 296, 312, and 720. Nevertheless, GR binding to Ubc9 is undiminished with the mutant Ubc9/C93S (lane 7 versus lane 3). We suspect that the binding of Ubc9 to GR, like that to the GMEBs, is independent of the sumoylation activity of Ubc9 (Fig. 3A). This binding is specific, as seen from the lack of binding of CREB (lanes 4 and  9 versus lanes 16 and 20).
Mammalian two-hybrid assays were used to obtain evidence for an interaction of Ubc9 with GR in intact cells. A steroid-and Ubc9-dependent association between VP16/Ubc9 and a fusion of GAL-DBD with the full-length GR (GAL/GR ϭ pM/GR) was seen in CV-1 cells (Fig. 3B). Interestingly, the response with Ubc9 was equally robust for GR bound by the antiglucocorticoid Dex-Mes and the glucocorticoid Dex.
Modulatory Activity of Ubc9 in Intact Cells-The sequence to which the GMEBs bind (i.e. the GME (14,16,22,24)) modulates both the dose-response curve of glucocorticoid agonists and the partial agonist activity of antiglucocorticoids (14,17,18). Therefore, we asked if Ubc9 can further augment the modulatory activity of the GME. When low concentrations of GR plasmid (4 ng/well of a 24-well plate) are used so that the receptor is a limiting factor for transactivation (20,21,24,43,49), 150 ng of pSG5/Ubc9 plasmid causes a concentration-dependent increase in the total activity (2.5 Ϯ 0.4-fold, S.D., n ϭ 4, p ϭ 0.0044) and a decrease in the fold induction (0.51 Ϯ 0.11-fold, S.D., n ϭ 4, p ϭ 0.0033) for 1 M Dex with a GME-GREtkLUC reporter (data not shown). However, there is little effect of added Ubc9 on either the position of the dose-response curve or the partial agonist activity of antisteroids relative to a control vector expressing hSA (Fig. 4A).
We have previously established that the GME, and changing concentrations of GR, act through a common rate-limiting step or intermediate X that is saturated at high GR concentrations (20). Therefore, as long as GR is present at low concentrations, the presence of other intermediates downstream of X are kinetically invisible. To overcome this limitation, and to reveal the involvement of steps downstream of X, we asked whether the presence of saturating concentrations of GR (Ն90 ng (20)) would make a Ubc9-mediated step now become limiting and responsive to added Ubc9. Under these conditions, 150 ng of Ubc9 plasmid causes a 1.3 fold increase in total Dex activity and a concentration-dependent decrease in the fold induction (0.32 Ϯ 0.07-fold, S.D., n ϭ 5, p ϭ Ͻ 0.0001). More significantly, GMEB-2, and rGR were added to M2-agarose columns containing immobilized FLAG-tagged mUbc9 (wild type and C93S mutant) that had been overexpressed from baculovirus vectors in insect cells. The proteins remaining after washing were eluted, separated on SDS-polyacrylamide gels, and visualized as described under "Materials and Methods." In vitro translated CREB was used to assay nonspecific retention. The "10% Input" panel displays one-tenth of the in vitro translated, full-length proteins that were assayed. The "Mock" panel shows the amount of each protein that was retained in the absence of Ubc9. Similar results were obtained in two others experiments. See "Results" for details. B, interaction of Ubc9 and GR in mammalian two-hybrid assays. Triplicate dishes of CV-1 cells containing the indicated GAL4 and VP16 chimeras of mUbc9 were cotransfected with the FRLuc reporter and treated with EtOH Ϯ 1 M Dex or 1 M Dex-Mes. The induced levels of luciferase were determined as described under "Materials and Methods" and plotted Ϯ S.D. Similar results were obtained in a second experiment.
150 ng of Ubc9 now produces a 3.5 Ϯ 0.7-fold (S.D., n ϭ 3, p ϭ 0.023) left shift of the dose-response curve to a lower EC 50 and increases in the partial agonist activity of Dex-Mes (2.1 Ϯ 0.5-fold, S.D., n ϭ 5, p ϭ 0.01) and Dex-Ox (1.25 Ϯ 0.14-fold, S.D., n ϭ 4, p ϭ 0.34) (Fig. 4B). The lack of statistical significance for increases in Dex-Ox activity appears to be due to the fact that Dex-Ox already displays nearly full agonist activity (about 90%) with 90 ng of GR plasmid in the absence of Ubc9. In summary, these data indicate that Ubc9 has little effect on the dose-response curve and partial agonist activity when GR is limiting but does change GR induction properties when some factor other than GR becomes limiting. In contrast, other GR induction properties, such as the total transactivation, are affected by Ubc9 even when GR is not limiting. This suggests that the effects of Ubc9 on the total transactivation and the position of the dose-response curve occur via different mechanisms.
Modulation of GR Activity by Ubc9 Does Not Require Protein Sumoylation-Ubiquitination of transactivation domains has recently been found to be essential for transcriptional activation by some factors (50). Thus, it seemed possible that the modulation of GR activity by Ubc9 might involve sumoylation of selected proteins, like RanGAP1, even though the GMEBs are not substrates (Fig. 2), and Ubc9 binding to the GMEBs, or GR, appears to be independent of Ubc9 sumoylation activity (Fig. 3A). We therefore asked whether the C93S mutant of Ubc9, which is defective for sumoylation (29,38,48), can still modify GR properties. As seen in Fig. 4C, elimination of the sumoylation activity of Ubc9 has little effect on the ability of Ubc9 to modulate the EC 50 of agonists or the partial agonist activity of antagonists. We conclude that these properties of Ubc9 are independent of its ability to sumoylate proteins.
GR Sequences Required for Modulation of Transactivation Properties by Ubc9 -To determine which domains of GR are needed to respond to the effects of added Ubc9, we first examined the properties of the rat GR lacking the amino-terminal 406 amino acids. This truncated GR (GR-407C; Fig. 5A) lacks the activation domain (AF1) that is present in the aminoterminal half of the receptor but maintains the DBD and ligand binding domain (LBD). Added Ubc9 still causes a left shift in the dose-response curve for Dex induction of the GMEGREtk-LUC reporter by GR-407C (average ϭ 1.67 Ϯ 0.09-fold; n ϭ 2) and increased partial agonist activity of Dex-Mes and Dex-Ox (Fig. 5B).
To answer whether the GR DBD is required for the above responses, we replaced the GR DBD, and part of the hinge region, with the GAL4-DBD to give the chimera GAL/GR-525C (Fig. 5A), which contains amino acids 525-795 of GR. We then assayed the ability of added Ubc9 to alter the properties of GAL/GR-525C induction of FRLuc, which is a luciferase reporter driven by five tandem repeats of the GAL4 binding site, UAS. As for the full-length GR and GR-407C, overexpressed Ubc9 produces a left shift of the dose-response curve for GAL/ GR-525C (Fig. 5C, average ϭ 2.0 Ϯ 0.5-fold, S.D., n ϭ 4, p ϭ 0.026) and an increase in the partial agonist activity of Dex-Mes (1.96 Ϯ 0.20-fold, S.D., n ϭ 4, p ϭ 0.0003). Furthermore, added Ubc9 increases both the total activity (4.6 Ϯ 1.8-fold, S.D., n ϭ 4, p ϭ 0.027) and the fold induction (1.77 Ϯ 0.39-fold, S.D., n ϭ 4, p ϭ 0.028) of GAL/GR-525C with 1 M Dex. Therefore, the region of rGR encompassing the COOH-terminal half of the hinge region and the LBD is sufficient to respond to Ubc9.
Role of GME in Mediating Ubc9 Modulation of GR Transactivation Properties-The ability of the GME to modulate GR transactivation properties is linked to the binding of the GMEBs to GME oligonucleotides (14,16). Because Ubc9 was isolated by its binding to GMEB-1 in a yeast two-hybrid screen, the modulation of GR transcriptional properties by Ubc9 might be expected to require the presence of the GME element in the reporter construct. However, the ability both of Ubc9 to interact with GR (Fig. 3) and of overexpressed Ubc9 to alter the properties for GAL/GR-525C induction of the FRLuc reporter, which does not contain a GME (Fig. 5B), suggests that the GME is not required for Ubc9 activity. This was confirmed by showing that Ubc9 affects a concentration-dependent left shift in the dose-response curve and increased partial agonist activity for Dex-Mes (averages ϭ 2.7 Ϯ 0.2-fold, S.D., n ϭ 3, p ϭ 0.0056, and 1.57 Ϯ 0.16-fold, S.D., n ϭ 3, p ϭ 0.024, respectively) with wild type GR on a simple GREtkLUC reporter (Fig.  6). Therefore, Ubc9 can modulate GR induction properties in the absence of the GME. DISCUSSION The purpose of this study was to identify proteins that participate in the modulation of GR transactivation properties by the GME. The GME is a cis-acting element that modulates both the dose-response curve of GR-agonist complexes and the partial agonist activity of GR-antagonist complexes (14 -21). These effects appear to be mediated by a heteromeric complex of two proteins, GMEB-1 and -2, that we have cloned and found to have limited homology with other members of the KDWK family of proteins (16,(22)(23)(24). Selected members of this family are unable to interact with the GMEBs and thus are unlikely to mediate the effects of the GME. However, a yeast two-hybrid assay was used to isolate a protein, Ubc9, which displays all of the criteria expected for an intermediary protein in the modulation of selected GR transcriptional properties. Ubc9 preferentially interacts with GMEB-1 in yeast two-hybrid assays. In pull-down assays with partially purified Ubc9, GMEB-1 and -2 are bound with apparently similar affinity. This discrepancy between whole cell and cell-free assays is not unusual (51) and probably reflects low affinity interactions that are not abundant in intact cells with low concentrations of proteins. Overexpression of Ubc9 in CV-1 cells has no effect on the GR dose-response curve when GR is present in low concentrations and is limiting. However, when GR is overexpressed to levels where GR is no longer limiting, Ubc9 now causes both a left shift in the dose-response curve for agonist complexes and an increase in the partial agonist activity of antagonist complexes. Therefore, Ubc9 has all of the characteristics of a transcriptional cofactor for GR (34 -38).
Unexpectedly, Ubc9 also interacts with GR both in pull-down assays and mammalian two-hybrid assays. Furthermore, Ubc9 is able to modulate GR induction properties independent of the presence of a GME. This is most likely due to the binding of Ubc9 directly to GR (Fig. 3). Ubc9 is known to transfer SUMO-1 to a variety of substrates (28,29,47). However, no sumoylation of the GMEBs was observed (Fig. 2), Ubc9 sumoylation activity is not required for the binding of either the GMEBs or GR (Fig.  3A), and the modulation of GR transcriptional properties does not require protein sumoylation by Ubc9 (Fig. 4C). This ability of Ubc9 to modulate the induction parameters of a steroid receptor represents a new activity of Ubc9.
Ubc9 has no intrinsic transactivation activity on a GAL4regulated reporter when fused to a GAL-DBD. By this criterion, Ubc9 is neither a coactivator nor a corepressor. However, Ubc9 does increase the total transactivation elicited by GR from GR-regulated reporters and decrease the fold induction, due to greater effects on basal level activity. Thus, Ubc9 qualifies as a transcriptional adaptor. Interestingly, these effects on fold induction and total activity are independent of GR concentration, while the ability of Ubc9 to modulate the dose-response curve and partial agonist activity of antisteroids is seen only with high and saturating concentrations of GR. This indicates that the effects of Ubc9 on the magnitude of GR transactivation are mechanistically distinct from those on the GR dose-response curve and partial agonist activity. In these respects, Ubc9 is reminiscent of other factors (e.g. TIF2, CBP, SMRT, NCoR, GME) that alter the dose-response curve and partial agonist activity of GR com-plexes in a manner that is independent of their effects on total activity (19 -21, 43, 49, 52).
Ubc9 has been found to interact with a variety of transcription factors, such as HIPK2 (35), ETS1 (36), and TEL (38), but reports of effects on steroid receptors are much less common. The first report was by Gottlicher et al. (53), who briefly mentioned that Ubc9 interacts with the glucocorticoid receptor but did not describe any details or biological consequences. More recently, Poukka et al. (34,37) found that Ubc9 interacts with AR via the nuclear translocation sequence in a mammalian two-hybrid assay and pull-down assays and that Ubc9 overexpression increased transactivation by AR in the presence of steroid in a manner that can be, depending upon the condi-tions, both independent (37) and dependent (34) on the SUMO-1 ligating activity of Ubc9. The effect of Ubc9 on the total transactivation of GR is the same as that seen with AR. However, it is not yet known if Ubc9 also affects the doseresponse curve and partial agonist activity of AR complexes. Furthermore, in marked contrast to the AR, the nuclear translocation sequence at amino acids 496 -517 of the rat GR is not required for any of the responses of GR to Ubc9; amino acids 525-795 are sufficient (Fig. 5C). In addition, Ubc9 binding to GR does not appear to involve sumoylation (Fig. 3A), although putative sumoylation consensus sequences (47) exist in rGR. These differences may derive from several properties of AR that diverge from GR, including a relatively inactive AF-2 domain in AR, the reduced binding of coactivators to the AR LBD, and significant interactions between the NH 2 -and COOH-terminal domains of AR (54 -58).
Our previous studies of the mechanisms by which both the GME and changing concentrations of GR or coactivator (TIF2) can modulate GR transcriptional properties (i.e. the dose-response curve and partial agonist activity of antisteroids) indicated the involvement of a rate-limiting step, or intermediate, X. This conclusion was based on the combined observations of 1) a limit to the increase of both GR transcription properties in the presence of high concentrations of GR (90 ng) or TIF2 plasmids and 2) the inability of one component (GME, GR, or TIF2) to cause any further increase when added to saturating concentrations of a second component (20). In contrast, Ubc9 is able to modify the GR dose-response curve only in the presence of saturating concentrations of GR (Fig. 4). This requirement of high concentrations of GR to permit the actions of Ubc9 is strong kinetic evidence that Ubc9 exerts its effects without proceeding through the intermediate X (Fig. 7A). The observation that Ubc9 alters GR activities even in the absence of the GME (Fig. 6) further argues that the modulation of GR transactivation properties by Ubc9 can proceed via a pathway that initially involves steps or intermediates other than X (Fig. 7A). Thus, there appear to be at least two different cellular mechanisms for manipulating these induction properties of GR. This diversity would be very beneficial for the differential control of gene expression by the concentration of circulating hormone during development, differentiation, and homeostasis. A 10fold shift in the dose-response curve is sufficient to increase the induction by the usually subsaturating concentrations of steroid from 10 to 50% of the maximal response. Similarly, endocrine therapies would benefit from changes in the partial agonist activity of antisteroids that could cause near normal levels of induction, or prevent induction entirely, depending on the direction of the changes.
The present data indicate that Ubc9 interacts with both GR and the GMEBs. Furthermore, Ubc9 modulates GR induction properties independently of the presence of the GME. We propose that a major function of the GME is to increase the local concentration of Ubc9 by binding to GMEB-1, and probably GMEB-2. Upon binding to the GMEBs, additional contacts with GR and other factors of the transcriptional machinery (such as the unidentified factor "Y" in Fig. 7B) would increase the net binding affinity of Ubc9. With a high enough concen-tration of GR and Ubc9, the presence of the GME (and GMEBs) would not be necessary for efficient recruitment of Ubc9. We further propose that Ubc9 interacts with Y at low GR concen-FIG. 7. Proposed model for modulation of GR properties by Ubc9. A, new pathway for Ubc9 action. Previous data established that GME, GR, and coactivators modulate GR properties via a rate-limiting step, or intermediate, X (20). The inability of excess GR to inhibit Ubc9 activity indicates that Ubc9 affects steps downstream of X. See "Discussion" for details. B, proposed molecular model for interactions of Ubc9, GR, and GME with various components of the transcriptional machinery to affect different steps in the model in A. See "Discussion" for details. trations to mediate the effects of the GME (and GMEBs). Unfortunately, the importance of Ubc9 for the actions of low GR and the GME cannot be confirmed directly by the addition of more Ubc9. This is because the relevant interaction of Ubc9 with Y occurs downstream of the rate-limiting step X and thus is kinetically invisible. The presence of this step can be inferred, though, from the ability of excess GMEB-1 and/or -2 to cause squelching. Due to the ability of the GMEBs to bind Ubc9, high concentrations of the GMEBs would be expected to form soluble complexes with Ubc9, thereby decreasing the amount of Ubc9 present in the GRE/GME-bound complex of Fig. 7B and causing a right shift in the dose-response curve and a decrease in the partial agonist activity of antisteroids. This is precisely what is observed (24). However, this proposed interaction of Ubc9 with Y can be made to be detectable with saturating concentrations of GR, which afford conditions where X is no longer rate-limiting. Under these new conditions, the involvement of an interaction between Ubc9 and Y can now be seen upon adding additional Ubc9 (Figs. 4, B and C, and 6). These multiple interactions of Ubc9 are reminiscent of those seen with larger adapter proteins such as CBP, which interacts with both steroid receptors and coactivators (for SRC-1 and receptors (59 -61)), and PCAF, which binds receptors, CBP, and coactivators (62,63).
In summary, we have identified Ubc9 as a protein that is capable of modifying the dose-response curve and partial agonist activity of GR. Ubc9 was isolated on the basis of its binding to GMEB-1 but also can bind to GMEB-2 and to GR. Ubc9 modulates GR induction properties by a mechanism that does not require either GR sequences upstream of the COOH-terminal half of the hinge region or sumoylation by Ubc9 and appears to proceed via a different pathway from that which has been defined for other modulators (i.e. the rate-limiting step X that is employed by GR, GME, and TIF2). These findings expand the mechanistic possibilities by which differential control of gene expression can be achieved within a cell. They also indicate that Ubc9 is not just a component of scavenger complexes of disposable proteins but is also an active participant of that portion of the transcriptional machinery that is modulated by GR. It will be interesting to see if these new activities of Ubc9 similarly modulate the induction properties of other steroid receptors.