The Human Estrogen Receptor-α Is a Ubiquitinated Protein Whose Stability Is Affected Differentially by Agonists, Antagonists, and Selective Estrogen Receptor Modulators*

The human estrogen receptor α-isoform (ERα) is a nuclear transcription factor that displays a complex pharmacology. In addition to classical agonists and antagonists, the transcriptional activity of ERα can be regulated by selective estrogen receptor modulators, a new class of drugs whose relative agonist/antagonist activity is determined by cell context. It has been demonstrated that the binding of different ligands to ERα results in the formation of unique ERα-ligand conformations. These conformations have been shown to influence ERα-cofactor binding and, therefore, have a profound impact on ERα pharmacology. In this study, we demonstrate that the nature of the bound ligand also influences the stability of ERα, revealing an additional mechanism by which the pharmacological activity of a compound is determined. Of note we found that although all ERα-ligand complexes can be ubiquitinated and degraded by the 26 S proteasome in vivo, the mechanisms by which they are targeted for proteolysis appear to be different. Specifically, for agonist-activated ERα, an inverse relationship between transcriptional activity and receptor stability was observed. This relationship does not extend to selective estrogen receptor modulators and pure antagonists. Instead, it appears that with these compounds, the determinant of receptor stability is the ligand-induced conformation of ERα. We conclude that the different conformational states adopted by ERα in the presence of different ligands influence transcriptional activity directly by regulating cofactor binding and indirectly by modulating receptor stability.

ER␣ 1 resides within the nuclei of target cells in an inactive form in the absence of hormone. Upon binding its cognate ligand estradiol, the receptor undergoes an activating conformational change permitting it to interact with specific cofactors and bind DNA response elements within target gene promoters (1,2). The DNA-bound receptor-ligand complex is then capable of either activating or repressing target gene transcription, depending on both the cell and the promoter context. The classical model of ER␣ action suggests that the role of agonists, such as estradiol, is that of a switch converting the receptor from an inactive to an active form. It now appears that ER␣ pharmacology is more complex, since it has been observed that different ER␣-ligands induce different changes in receptor conformation and that target cells can distinguish between these complexes (3)(4)(5). For instance, the anti-estrogen tamoxifen opposes estrogen action in the breast, whereas it manifests estrogenic activities in bone, the cardiovascular system, and the uterus. Reflecting its complex pharmacology, tamoxifen has recently been reclassified as a selective estrogen receptor modulator (SERM). Additional SERMs have been identified, such as raloxifene, GW5638, TSE424, lasofoxifene, and arzoxifene, each of which has distinct agonist/antagonist profiles (6). The challenge, therefore, has been to understand the mechanism(s) underlying SERM-mediated action and evaluate why these compounds distinguish themselves from pure agonists like estradiol.
Analysis of the crystal structure of ER␣ revealed that the conformation of the agonist-receptor complex is distinct from that formed in the presence of antagonists (7,8). Furthermore, we have used combinatorial phage display to identify a series of peptides whose ability to interact with ER␣ is regulated by the nature of the bound ligand. Using these probes, we have been able to show that, even among the SERMs, there are significant differences in the overall structure of the receptor (3). These findings lend support to the hypothesis that conformation is a primary regulator of ER␣-coactivator interactions. One of the first coactivators identified, SRC-1, has been shown to enhance estrogen-activated ER␣ transcriptional activity when overexpressed in target cells (9). In addition, it was observed that overexpression of SRC-1 enhances the partial agonist activity of 4OH-tamoxifen, whereas it has no effect on the antagonist activity of ICI 182,780 (10). The importance of transcription corepressors in ER␣ action was demonstrated in studies that showed that the partial agonist activity of tamoxifen could be suppressed by overexpressing N-CoR and SMRT in cultured cells (10,11). Interestingly, tamoxifen resistance in breast tumor explants (propagated in mice), has been shown to be correlated with a decrease in the expression level of N-CoR (12). Clearly, cofactor expression is a primary determinant of a cell's ability to distinguish among different classes of ligands.
Ligand binding, in addition to altering the conformation of the receptor, has been shown to influence the stability of the receptor. In particular, it has been shown that in the absence of ligand, the half-life of ER␣ is about 4 -5 h, whereas estradiol binding accelerates receptor degradation, reducing its half-life to ϳ3-4 h (13)(14)(15)(16)(17). In addition to estradiol, tamoxifen has been shown to stabilize ER␣ following long term treatment (5), whereas the ER␣-pure antagonist ICI 182,780 has been shown to decrease the half-life of the mouse ER␣ (17). Given these results, it seems likely that receptor degradation may be an important event that regulates the duration of response to an activating ligand.
The turnover of many short lived transcription factors has been shown to be mediated by the ubiquitin-proteasome pathway (18), and the rate of degradation of these proteins appears to directly correlate with their transcriptional activity (19 -21). Although human ER␣ has been shown to be a substrate for ubiquitination in the presence of estradiol in vitro (22), in vivo evidence is still lacking. However, several studies have demonstrated that in the intact cell, estradiol-mediated ER␣ degradation occurs through the 26 S proteasome pathway (23,24). Although it has been speculated that recruitment of coactivators like SRC-1 may be a rate-limiting step in this pathway (21), the mechanism(s) by which estradiol-activated ER␣ is recognized by the 26 S proteasome remains to be determined. Furthermore, the observation that in addition to the agonist estradiol, the pure antagonist ICI 182,780, which when bound to ER␣ is unable to recruit any known coactivators, can induce a rapid degradation of ER␣ clearly indicates that other factors besides coactivator recruitment regulate ER␣ degradation. At variance with observations made in studies of other transcription factors, it appears therefore, that either transcriptional activity and ER␣ stability are not linked or that the mechanisms by which agonist-and antagonist-activated ER␣ are degraded are not the same. Consequently, in this study, we have performed a series of experiments to 1) evaluate the influence of ER␣ agonists, antagonists, and SERMs on the stability of the human ER␣ and 2) probe potential correlations between ligand-induced receptor stability and the relative agonist/antagonist activity of an individual compound. Information of this nature will help us to define the mechanisms underlying the agonist/antagonist activity of ER␣-ligands and will be useful in the design of screens for ER␣-ligands with unique pharmacological attributes.

EXPERIMENTAL PROCEDURES
Biochemicals-General laboratory reagents, 17␤-estradiol, 4OHtamoxifen, methionine, and phenol red-free tissue culture media were purchased from Sigma. ICI 182,780 was a gift from Dr. A. Wakeling (Zeneca Pharmaceuticals, Macclesfield, UK), and GW7604 was a gift from Dr. T. Willson (Glaxo Wellcome Research and Development, Research Triangle Park, NC). H222 (monoclonal antibody raised against human ER␣) was a gift from Dr. G. Greene (Ben May Institute, University of Chicago). ␤-Galactosidase antibody was purchased from Chemicon International, Inc. (Temecula, CA 92590), and the ERK1 antibody, anti-rat secondary antibodies, and normal IgGs were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). Other secondary antibodies used, Hybond-C Extra transfer membranes, and x-ray film were obtained from Amersham Pharmacia Biotech. 35 S-labeled methionine/cysteine was purchased from ICN Biomedicals Inc. (Irvine, CA). Protein G Plus/Protein A-agarose was purchased from Oncogene Research Products (Cambridge, MA). Ni 2ϩ -nitrilotriacetic acid-agarose beads were purchased from Qiagen (Valencia, CA). All transfection reagents and media were purchased from Life Technologies, Inc., and sera were purchased from Hyclone (Logan, UT).
Plasmid Constructs-The ER␣-⌬DBD construct was a gift from Dr. R. Bambara (University of Rochester). The insert within this construct was recloned into a pRST7 vector to be consistent with the rest of the ER␣ constructs used in this study. Expression vectors for the mouse AP-1 proteins c-Fos and c-Jun and the AP-1-responsive collagenase reporter gene (pCOL-Luc) have been described elsewhere (25).
Western Blot Analysis-MCF-7 cells were maintained in phenol redfree medium containing 8% charcoal-filtered serum for at least 24 h prior to induction with ligand. When HeLa cells were used for the expression of proteins, cells were transiently transfected as described below. All cell types were induced with ligand in phenol red-free culture media containing 8% charcoal-filtered serum for 4 h, and whole-cell extracts were prepared as described previously (26). Approximately 15 g of total protein was analyzed by SDS-PAGE. Proteins were transferred to nitrocellulose membrane and probed with monoclonal antibody H222. The amount of protein loaded has been normalized to ERK-1 protein in an endogenous setting and Escherichia coli ␤-galactosidase in an exogenous setting. Complexes were detected using ECL. Densitometric quantitation of ER␣ levels, relative to ␤-galactosidase levels or ERK-1, were performed using the ImageQuant software program (Molecular Dynamics, Inc., Sunnyvale, CA).
Pulse-Chase Analysis-MCF-7 cells were maintained in phenol redfree medium for at least 24 h prior to starvation for 1 h in growth medium lacking phenol red and methionine. Cells were radiolabeled with 200 Ci of 35 S-labeled methionine/cysteine for 4 h in the same medium used for starvation. Under conditions where cells had to be pretreated with lactacystin (20 M) or chloroquine (100 M), reagents were added into the labeling medium for incubation periods of 2.5 h or 30 min, respectively. Following incubation, cells were washed twice with PBS and incubated for 1 or 4 h in the presence of 10 nM solvent or ligand in phenol red-free normal culture medium containing 1 mM methionine and cysteine. Cells were lysed with 0.3 ml of buffer consisting of 50 mM HEPES, pH 7.0, 500 mM NaCl, 1% Nonidet P-40, and protease inhibitors as described previously (17). The amount of protein in each supernatant was measured by BCA protein assay and normalized for immunoprecipitation.
Immunoprecipitation-Equal amounts of cell lysates were precleared by incubating with the corresponding IgG, which had been preincubated with Protein G Plus/Protein A-agarose beads for 1 h at 4°C, followed by a further incubation with the beads alone for another 40 min. Following each of these incubations, the lysates were centrifuged at 10,000 ϫ g for 10 min. After the preclearing, lysates were incubated with specific primary antibody prebound to agarose beads for 1 h at 4°C. The beads were washed four times in high and low salt buffers (lysis buffer with 150 mM NaCl), resuspended in loading buffer, and then analyzed by SDS-PAGE.
Cell Culture and Transient Transfection Assay-HeLa cells were maintained in minimum essential medium supplemented with 8% fetal calf serum. Cells were plated in 64-mm plates (for Western blots) or 24-well plates (for the measurement of transcription) 24 h prior to transfection. DNA was introduced into cells by transient transfection using Lipofectin. Briefly, 64-mm plate transfections were performed using 7.5 g of total DNA. For standard transfections, 250 ng of pCMV-␤GAL (normalization vector), 2500 ng of reporter (ERE 3 -TATA-Luc), and 2500 ng of receptor (pRST7-hER␣ (27), ER␣-LL (28), or ER␣-⌬DBD) were used. The total amount of DNA was brought up to 7.5 g with the control vector pBSII-KS (Stratagene). Cells were transfected for 3 h, at which time medium was replaced with phenol red-free culture medium containing 8% charcoal-filtered serum. Forty-eight h following transfection, cells were induced with ligand for 1 or 4 h and lysed as described above under Western blot analysis.
When performing transient transfection assays in 24-well plates to measure transcriptional activity, triplicate transfections were performed using 3 g of total DNA as described previously (5). Following the transfection, cells were incubated in the presence of ligand for 48 h, and subsequently lysed and assayed for luciferase and ␤-galactosidase activity as previously described (29).
Detection of ER␣ Ubiquitination in Vivo-HeLa cells were plated in 100-mm plates and transiently transfected with 15 g of total DNA. For standard transfections, 2500 ng of reporter (ERE 3 -TATA-Luc), 9000 ng of receptor (pRST7-hER␣), and 3000 ng of His 6 -tagged ubiquitin expression vector (pMT107) (30) were used. The total amount of DNA was brought up to 15 g with pBSII-KS. Forty-eight h following transfection, cells were treated with ligand for 4 h followed by lysing in lysis buffer containing 50 mM HEPES, pH 7, 150 mM NaCl, 1.5 mM MgCl 2 , 1 mM EGTA, 1% Triton X-100, 10% glycerol, and protease inhibitors as described previously (31). Insoluble material was removed by centrifugation. The protein concentration of the supernatants was measured, and equal amounts of protein were taken and denatured by boiling for 5 min in the presence of 2% SDS. The samples were then diluted with 11 volumes of lysis buffer, followed by the addition of Ni 2ϩ -nitrilotriacetic acid beads, with which they were rotated at 4°C for 4 h. The beads containing the His 6 -ubiquitin-tagged proteins were washed twice with the lysis buffer and twice with HNTG buffer (20 mM HEPES, pH 7.5, 300 mM NaCl, 1% Triton X-100, 10% glycerol, and 20 mM imidazole). Bound proteins were recovered by boiling in SDS-PAGE loading buffer and electrophoresis and detected by immunoblotting with the H222 antibody.
The linearity of these ECL responses was measured by constructing a standard curve with increasing concentrations of purified ER␣ and determining that all experimental densitometric values were in the linear range of the standard curve densitometric values. When calculating the ratios of ubiquitinated ER␣, input ER␣, the first ER␣ ubiq-uitin conjugate, was selected for calculations because it was the only degradation product that appeared consistently under the treatment conditions used.

The Estrogen Receptor Is a Short Lived Protein Whose Stability Is Influenced by the Nature of the Bound Ligand-Previ-
ous studies that examined the dynamics of ER␣ turnover have shown that the half-life of this receptor in both breast and uterine tissue, in the absence of ligand, is about 4 -5 h (13-16). Additionally, it has been established that upon binding estradiol or ICI 182,780, the half-life of mouse ER␣ decreases to ϳ3 and 0.5 h, respectively (17). However, a direct comparison of the influence of ligands on the stability of human ER␣ has not yet been performed. As an initial step, therefore, we evaluated the effect of short term treatment (4 h) with a variety of ER␣ ligands on the relative expression level of endogenous ER␣ within MCF-7 cells (human breast carcinoma), using Western immunoblot analysis. The ligands chosen for this study have been shown to have distinct ER␣ pharmacologies exhibiting a range of activities from pure agonist to pure antagonist activity (5). In this system, at physiologically relevant ligand concentrations (10 nM), we were able to demonstrate that endogenous ER␣ expression levels are differentially affected by the nature of the bound ligand ( Fig. 1, A and B). It was observed, for instance, that treatment with either 17-␤-estradiol, GW7604, or ICI 182,780 leads to a decrease in ER␣ levels, with ICI 182,780 being the most effective in this regard (Fig. 1A, lanes [1][2][3][4][5][6][7][8]. In contrast, treatment with tamoxifen appears to stabilize ER␣ and increases intracellular receptor levels above the basal level (Fig. 1A, lanes 1 and 2 and lanes 9 and 10).
It is likely that the ligands under study influence both the stability of ER␣ and the activity of the ER␣ promoter. Therefore, in order to separate the transcriptional from post-translational effects of exposure to ligands, we performed a pulsechase analysis of ER␣ stability in MCF-7 cells. Endogenous ER␣ in MCF-7 cells was radiolabeled with [ 35 S]methionine/ cysteine and subsequently chased in medium containing unlabeled methionine/cysteine and either solvent or 10 nM ligand for the indicated period. The relative levels of ER␣ protein were measured by immunoprecipitation followed by autoradiography. As shown in Fig. 1, C and D, relative to the unliganded receptor, estradiol, ICI 182,780, and GW7604 enhanced the rate of degradation of ER␣ within 1-4 h of treatment, whereas 4OH-tamoxifen decreased this rate of degradation. We conclude that ER␣ stability is influenced by the nature of the bound ligand following short term treatment, a finding that may explain the different pharmacological activities of the known ER␣ ligands.
Estradiol-, SERM-, and ICI 182,780-mediated Degradation of ER␣ Occur by Different Mechanisms-Recent studies have provided evidence for a link between the relative transcriptional activity of transcription factors and their rate of degradation; more activity equates with decreased protein stability (36,49). For instance, it has been shown that by altering the potency of transcriptional activity of c-Myc, its stability can be regulated (36). Based on these observations, we evaluated whether the ability of an ER␣-ligand complex to activate transcription is integrally linked to receptor stability. In addition, we addressed whether this relationship held for SERM-and ICI 182,780-activated ER␣, since these ligands have distinct ER␣ pharmacologies. As a first step, we compared the ligandinduced changes in the stability of ER␣-wt and a transcriptionally inactive ER␣-LL mutant (6,34). The ER␣-LL protein contains two mutations within helix 12 of the ligand binding domain, which have been shown to disrupt the p160 coactivator binding pocket within the receptor (Fig. 2A). The hormone-and DNA-binding properties of this mutant are the same as those of the wild type receptor; however, it is unable to activate transcription in the presence of estradiol (28, 32, 33). Thus, this mutation provides a starting point from which to evaluate the FIG. 1. Estrogen receptor is a short lived protein whose stability is influenced by the nature of the bound ligand. A, MCF-7 cells maintained in estradiol-free medium were induced with each ligand (10 nM) for 1 or 4 h. The relative amount of endogenous ER␣ present in whole cell extracts was analyzed by immunoblotting with an antibody directed against ER␣. Loading of protein has been normalized to endogenous ERK1 protein levels. B, normalized results summarized in graph form (normalization was performed by densitometric analysis of the intensity of each ER␣ and ERK1 band and by dividing each ER␣ value with the corresponding ERK1 value). In each case, a representative of three independent experiments is shown. C, MCF-7 cells were pulselabeled for 4 h with [ 35 S]methionine/cysteine and chased with medium containing 10 nM ligand and cold methionine/cysteine. At the times indicated, cells were lysed and immunoprecipitated with ER␣ antibodies and analyzed by SDS-PAGE and autoradiography. The asterisk denotes a nonspecific band pulled down by immunoprecipitation. D, pulse-chase results from three experiments are summarized in graph form. relationship between transcriptional activity and receptor stability.
The transcriptional activity and stability of the ER␣-LL and ER␣-wt proteins were assayed in transiently transfected HeLa cells, an ER␣-negative cell line. We chose to transiently overexpress ER␣ under the control of a heterologous SV40 promoter to rule out the potential influence of the ligands on the ER␣ promoter. The influence of ligands on the stability of these proteins was assayed using Western immunoblot analysis. As demonstrated in Fig. 2C, we were able to reconstitute the same pattern of ligand-induced ER␣ degradation in this system as was observed in the endogenous MCF-7 system (Fig. 1A). In the HeLa cell background, ER␣-LL displayed a minimal amount of transcriptional activity in the presence of estradiol (Fig. 2B). Under these conditions, although estradiol induced the degradation of the wild type ER␣, it was unable to induce the degradation of transcriptionally inactive ER␣-LL (Fig. 2, C and D,  lanes 1 and 2). These observations suggest the existence of a correlation between the ability to activate transcription and a decrease in receptor stability. This link, however, was not apparent when the analysis was extended to other ligands. In particular, although 4OH-tamoxifen exhibited minimal agonist activity on both ER␣ and ER␣-LL, the impact of these ligands on the two proteins was not identical. Most notably, 4OHtamoxifen-bound ER␣-LL was less stable than 4OH-tamoxifenbound ER␣ (Fig. 2, C and D, lanes 1 and 5). Similarly, a relationship between ICI 182,780 or GW7604 transcriptional activity and their relative impact on stability could not be observed.
The data presented here clearly indicate that the processes that regulate the stability of the ER␣-estradiol complex are different from those that affect pure antagonist-or SERMbound ER␣. This is an important result that suggests that other factors, in addition to the ability to activate transcription, are able to trigger the degradation of ER␣.
ER␣ Transcriptional Activity and Receptor Stability Are Linked-It was important to determine whether the observed differences in the degradation patterns of ER␣-LL and ER␣-wt, in the presence of both estradiol and 4OH-tamoxifen, are due to 1) structural alterations conferred by the mutations in helix 12 or 2) the inability of this mutant to interact with specific cofactor complexes as described above. For this purpose, we used VP16-ER␣ or VP16-ER␣-LL, chimeric receptors in which one of the transcriptional activation regions of the herpes simplex virus (VP16 domain) has been inserted into the amino terminus of full-length receptor (Fig. 3A). Previous studies have shown that the ER␣-VP16 chimera is able to constitutively activate transcription when delivered to DNA by any class of ER␣ ligand (4). Insertion of this potent transcriptional activation domain enables ER␣ to manifest strong transcriptional activity, although in a manner that does not require the indigenous ER␣ transcriptional activation domains. Using this strategy, we converted the transcriptionally inactive ER␣-LL into a potent transcription factor. The VP16-ER␣-LL mutant is unable to interact with cofactors such as SRC-1 (32,33). The use of VP16-ER␣-LL, in parallel with VP16-ER␣, permits the evaluation of 1) whether transcriptional activity and receptor stability are linked and 2) the contribution of ER␣-specific coactivators to the degradation induced by estradiol.
Initially we assessed the transcriptional activities of VP16-ER␣ and VP16-ER␣-LL on an estrogen-responsive promoter in transiently transfected HeLa cells. The activity of the GAL4-VP16 on a Gal4-responsive promoter was measured as a control (34). The stabilities of these chimeric proteins in HeLa cells were assessed following a 4-h ligand treatment. In this context, when bound to estradiol, both VP16-ER␣ and VP16-ER␣-LL displayed agonist activity (Fig. 3B), and both were subject to agonist-induced degradation (Fig. 3, C and D, lanes 1 and 2). Identical results were obtained using an ER␣-VP16 chimera, in which the VP16 fusion is at the carboxyl terminus of ER␣, ruling out fusion-related artifacts (data not shown). These data suggest that the event that targets estradiol-activated ER␣ for degradation is the overall transcriptional activity and not the recruitment of ER␣-specific coactivators. Loading of protein has been normalized to ␤-galactosidase protein levels produced from a cotransfected expression vector. In each case, a representative of three independent experiments is shown.
Four hydroxytamoxifen displayed agonist activity when bound to either VP-16-ER␣ or VP16-ER␣-LL proteins (Fig. 3B). However, a relationship between transcriptional activity and stability was not observed. Specifically, in this background, tamoxifen stabilized both ER␣-wt (transcriptionally inactive) and VP16-ER␣ (transcriptionally active) (Figs. 2C and 3C, lanes 1 and 5); yet it facilitated degradation of ER␣-LL (minimally active) and VP16-ER␣-LL (transcriptionally active), where helix 12 has been disrupted ( Figs. 2D and 3D, lanes 1  and 5). This observation suggests that, independent of its potential to activate transcription, a structural component of the receptor may be involved in the regulation of the stability of ER␣.
GW7604-and ICI 182,780-bound receptors were degraded under all four conditions tested regardless of the receptor's potential to activate transcription and the structure of helix 12. These results clearly uncouple the requirements necessary for degradation of the ER␣-estradiol complex from complexes formed in the presence of 4OH-tamoxifen, ICI 182,780, and GW7604 as well as the requirements necessary for receptor degradation in the presence of 4OH-tamoxifen from ICI 182,780 and GW 7604.
Influence of DNA Binding on the Ligand-mediated Degradation of ER␣-Based on the observation that transcriptional activity and receptor stability are linked in the presence of estradiol, we hypothesized that since DNA binding facilitates transcriptional initiation, processes involved in DNA binding should also influence the stability of receptor-ligand complexes. To test this possibility, we utilized an ER␣ mutant ER␣-⌬DBD, in which the DNA binding domain has been deleted in a manner that prevents it from directly interacting with estrogen response elements (Fig. 4A). Initially, we evaluated the transcriptional activity of this mutant on both a classical estrogen response element pathway, where transcription is mediated by ER␣ binding to DNA, and on an AP-1 pathway, where transcription is mediated by ER␣ binding to AP-1 proteins that are tethered to an AP-1-responsive element (35,36). A parallel analysis was carried out to evaluate the stability of ER␣-⌬DBD by Western immunoblot analysis following a 4-h ligand treatment. Under these conditions, as expected, this mutant was unable to elicit agonist activity (Fig. 4, B and C). Reflecting our previous result with the ER␣-LL mutant, we found that when estradiol-bound ER␣ was prevented from interacting with DNA either directly (through estrogen response elements) or indirectly (through AP-1), it was no longer susceptible to degradation (Fig. 4, D and E, lanes 1 and 2). In contrast, GW7604 and ICI 182,780 still induced the degradation of ER␣. Interestingly, 4OH-tamoxifen induced degradation of the receptor, suggesting that DNA binding is a process necessary for the stabilization of the receptor when occupied by this ligand.
Degradation of ER␣ Is Mediated through the 26 S Proteasome Pathway-Degradation of cellular proteins is carried out predominantly by the 26 S proteasome-or lysosome-mediated pathways. Previous studies have indicated that estradiol-induced ER␣ degradation may occur through the 26 S proteasome-mediated pathway (22)(23)(24), whereas ICI 182,780-induced degradation may occur through the lysosome-mediated pathway (37). Given our observations that the stability of each ER␣-ligand complex is not the same and that each ligand influences the stability through distinct mechanisms, we speculated that each proteolytic pathway may contribute differently to ER␣ pharmacology. Hence, blocking each proteolytic pathway and assaying the impact of these treatments on ER␣ levels in the presence of each ligand would allow the assessment of the role of these proteolytic systems on the stability of each ER␣-ligand complex. The effect of blocking these degradation pathways on ER␣ expression and stability in MCF-7 cells was evaluated by Western immunoblotting (Fig. 5A) and pulsechase analysis (Fig. 5, B and C), respectively. Lactacystin was used to block proteasome-mediated proteolysis (38), whereas chloroquine was used to block lysosome-mediated proteolysis (39). The assays were performed essentially as described above, except that cells were pretreated with the indicated inhibitor prior to induction with ligand (40). Inhibition of proteasome function greatly reduced the rate of degradation of ER␣ under all treatment conditions (Fig. 5, A-C). Inhibition of lysosome activity had no significant effect on the rate of receptor degradation mediated by any ligand. Similar results were observed when the assay was repeated in Ishikawa cells (human endometrial adenocarcinoma cells), suggesting that ER␣ is degraded by the pro -FIG. 3. ER␣ stability and transcriptional activity are linked. A, schematic of the VP16-ER␣/ER␣-LL chimeric receptor. B, HeLa cells were transiently cotransfected with either an estrogenresponsive reporter or a Gal4-responsive reporter, along with expression vectors for either ER␣-wt, VP16-ER␣, VP16-ER␣-LL, or Gal4-VP16 and a normalizing ␤-galactosidase expression vector as indicated. Following transfection, cells were treated with 10 nM ligand or solvent as indicated and subsequently assayed for luciferase and ␤-galactosidase activity. C and D, HeLa cells were transiently transfected as above. Forty-eight h later, the cells were treated with ligand for 4 h, and the levels of VP16-ER␣ (C) or VP16-ER␣-LL (D) in whole-cell extracts were measured by immunoblotting and densitometry. Loading of protein was normalized to exogenous ␤-galactosidase protein levels. In each case, a representative of three independent experiments is shown.

teasome-mediated proteolytic pathway in different cells (data not shown).
ER␣ Is a Ubiquitinated Protein in Vivo-We have shown that ER␣ degradation takes place predominately through the 26 S proteasome-mediated pathway. The majority of the proteins degraded in this pathway are tagged with ubiquitin, a covalent modification that marks specific proteins for proteolysis. Previous studies have determined that the pattern and the extent to which a protein is ubiquitinated determine the rate with which that substrate is degraded (41)(42)(43). An initial indication that human ER␣ is a substrate for ubiquitination has been suggested from an in vitro ubiquitination study performed by Nawaz et al. (22). Therefore, we analyzed whether ER␣ is ubiquitinated in vivo and, if so, if the pattern of ubiquitination is modulated by the bound ligand. Although several strategies were employed to detect endogenous ubiquitinated ER␣ in MCF-7 cells, we were able to detect higher molecular weight band formation, indicative of ubiquitination, only when cells were treated with ICI 182,780 in the presence of lactacystin (Fig. 6A). We speculate that the inability to detect these degradation intermediates in the presence of other ligands could be due to the inherent lability and low abundance of these degradation products. Furthermore, since ubiquitin is a conserved protein across species, development of specific antibodies against ubiquitin has been challenging. We overcame these problems by transiently transfecting HeLa cells with the expression plasmids for ER␣ and a hexahistidine (His 6 )-tagged ubiquitin (30). The transfected cells were subsequently treated with ligand for 4 h prior to lysis. Duplicate incubations were performed in the presence of lactacystin to block the proteasome-mediated proteolysis and to enhance the ability to study the ER␣-ubiquitin conjugates, if any. His 6 -ubiquitin-bound protein conjugates were isolated using a Ni 2ϩ -nitrilotriacetic acid chromatography column as described previously (30). The eluates from the column were subjected to SDS-PAGE and immunoblotted with an anti-ER␣ antibody.
The accumulation of higher molecular weight ER␣-ubiquitin conjugates was observed only when both exogenous ER␣ and His 6 -ubiquitin were present (Fig. 6B, lanes 5-14). Although a small fraction of ER␣ was able to nonspecifically stick to the nickel column, higher molecular weight ubiquitinated forms of ER␣ were not detected in the absence of exogenously expressed ubiquitin (Fig. 6B, lanes 1 and 2) (nonspecific binding of ER␣ to the column appears to be mediated by the histidine residues within the protein and not through the zinc fingers in the DNA binding domain). Lactacystin treatment increased the yield of these high molecular weight forms of ER␣, confirming that the aggregates were ubiquitin conjugates. Interestingly, unliganded receptor appeared to be highly ubiquitinated (Fig. 6B,  lanes 5 and 6). Since the unliganded ER␣ has a short half-life, it is likely that ubiquitination plays an important role in predetermining the half-life of ER␣. Another interesting observation was that the input ER␣ levels (Fig. 6C), did not directly correlate with the amount of ER␣-ubiquitin conjugate formed under each treatment condition (Fig. 6B). For instance, although the ER␣ levels in cells following treatment with ICI 182,780 are lower than that recovered from 4OH-tamoxifentreated cells (lanes 11 and 13), the intensity of the first higher molecular weight band (lowest state of ubiquitination) relative to ER␣ appears to be similar. This discrepancy suggested a ligand-dependent variation in the extent of ubiquitination. Therefore, we calculated the ratio of the major ER␣-ubiquitin conjugate formed (since it is the only band consistent across all treatment conditions) to the corresponding input ER␣ band, using densitometry. These values indicated that ICI 182,780bound ER␣ is the most heavily ubiquitinated (Fig. 6D), whereas 4OH-tamoxifen-bound ER␣ was the least ubiquitinated. This observation correlates well with our initial obser- FIG. 4. Influence of DNA binding on the stability of ER␣-ligand complexes. A, schematic of the ER␣-⌬DBD mutant. B, HeLa cells were transiently cotransfected with an estrogen-responsive reporter, along with expression vectors for ER␣-wt and ␤-galactosidase. C, HepG2 cells were transiently cotransfected with an AP-1-responsive collagenase reporter along with expression vectors for ER␣-⌬DBD, c-Fos, c-Jun, and ␤-galactosidase. Following transfection, cells were treated with 10 nM ligand or solvent as indicated. Subsequently, the transfected cells were assayed for luciferase and ␤-galactosidase activity. D and E, HeLa cells were transiently cotransfected with an estrogen-responsive promoter along with expression vectors for either ER␣-wt (D) or ER␣-⌬DBD mutant (E) and a normalizing ␤-galactosidase expression vector. Forty-eight h later, the cells were treated with ligand for 4 h, and the levels of ER␣-wt or ER␣-⌬DBD mutant in whole cell extracts were measured by immunoblotting and densitometry. Loading of protein was normalized to exogenous ␤-galactosidase protein levels. In each case, a representative of three independent experiments is shown. vation that ICI 182,780 is the most efficient inducer of ER␣ degradation and that 4OH-tamoxifen is the least efficient. The ability of 4OH-tamoxifen to decrease the basal level of ubiquitination is likely to explain why this ligand stabilizes the receptor. Furthermore, based on our observation that when bound to ER-LL, 4OH-tamoxifen decreases receptor levels, we compared the level of ER-LL ubiquitination in the absence and presence of 4OH-tamoxifen (Fig. 6, E and F). Not surprisingly, we were able to show that the degree of ubiquitination of ER-LL was increased substantially in the presence of tamoxifen.
Importantly, the degree of ubiquitination of estradiol-bound ER␣ was not significantly different from that of the unliganded ER␣. Similarly, the extent to which GW7604-bound ER␣ was ubiquitinated was not significantly different from basal level. This observation clearly uncouples the mechanism by which GW7604 mediates degradation from that of ICI 182,780 and suggests that other factors besides ubiquitination and transcriptional activation can influence the rate at which ER␣ degradation occurs. DISCUSSION Within the framework of the classical models of ER␣ action, where the role of an agonist was that of an all or nothing switch, it was difficult to understand how different ligands, upon binding the same receptor, are able to manifest distinct pharmacologies. It is now apparent, however, that the "on/off" model is oversimplified and that the receptor undergoes different structural alterations upon binding different ligands impacting which cofactors bind to the receptor. In support of this hypothesis, we recently identified a surface on ER␣ that is presented upon tamoxifen binding and showed that peptides that bind to this surface block tamoxifen partial agonist activity but not the activity of other SERMs (44). Clearly therefore, receptor conformation dictates its cofactor preference, an activity that determines the relative agonist/antagonist activity of ER␣-ligands. Of late, much attention has been focused on identifying the cofactors that interact with ER␣ in the presence of different ligands. In this study, however, we demonstrate that ER␣ stability is affected by the nature of the bound ligand, an activity that probably reflects subtle alterations in receptor conformation.
One of the novel findings of this study was that the rate of degradation of estradiol-occupied ER␣ appeared to directly correlate with transcriptional activity; this relationship was not apparent for the SERMs or pure antagonists tested. We conclude from this observation that the degradation of ER␣-estradiol, ER␣-SERM, and ER␣-pure antagonist complexes may not occur in the same manner. Further dissection of these pathways revealed that ER␣ is a ubiquitinated protein in the intact cell and that the extent to which it is ubiquitinated is not the same in the presence of all ligands. When SERMs were examined in a similar manner, we observed that the complex formed in the presence of tamoxifen was the most stable and that this reflected a hypoubiquitination of ER␣. Further, the complex formed in the presence of ICI 182,780 was the least stable, reflecting a hyperubiquitination of the receptor. It appears, therefore, at least with respect to tamoxifen and ICI 182,780, that receptor stability and its degree of ubiquitination are related. Since the influence of these two ligands on the receptor stability is not directly linked to their transcriptional activity, it is likely that it is the ligand-induced conformational changes in ER␣ that influence its degradation by modulating its interaction with components of the degradation machinery. The differences in the stability of the ER␣-tamoxifen and ER␣-GW5638 complexes are interesting in view of the structural similarity of these two triphenylethylene-derived anti-estrogens. This finding supports those of other studies from our laboratory, which indicate that subtle differences in the structure of a ligand can lead to profound differences in ER␣ pharmacology.
The ability of these ligands to modulate intracellular levels of ER␣ in a differential manner is likely to be physiologically important. For instance, the observation that 4OH-tamoxifen binding stabilizes ER␣ may have important clinical implications. Specifically, long term treatment of ER␣-positive breast cancers with tamoxifen invariably leads to the development of resistance. Indeed, some tumors actively switch from being inhibited by tamoxifen to requiring tamoxifen for growth (45). Although the cellular mechanisms responsible for this change are not known, it is speculated that 1) administration of tamoxifen leads to the selection of a subpopulation of cells, within the tumor, that recognize tamoxifen as an agonist and/or 2) tamoxifen-activated ER␣ may interact in an ectopic manner with a transcription factor(s) that is not normally involved in ER␣-signaling, thereby converting it from an antagonist to an agonist in the breast (44). We propose that the ability of ta-moxifen to elevate ER␣ levels in breast cancer cells is likely to be important also. For instance, elevations in the intracellular levels of cAMP have been shown to be sufficient to activate ER␣ in the absence of ligand or in the presence of 4OH-tamoxifen (46). Thus, independent of ligand status, the presence of receptor alone is sufficient for target gene regulation under certain circumstances. Although not addressed in our studies, it is intuitive that ligand-independent activation of ER␣ will be very sensitive to changes in intracellular levels of the receptor. Interestingly, it has been shown that tamoxifen-resistant breast cancers can be effectively treated with either ICI 182,780 (47) or GW5638 (GW7604 prodrug), both of which reduce intracellular ER␣ levels (48).
We have demonstrated that the rate of ER␣ degradation in the presence of estradiol directly correlates with transcriptional activity. In general, our conclusions are similar to those of others (21), with the exception that we have been able to show that it is overall transcriptional activity of ER␣, and not the availability of a protein interaction surface in the C-terminal helix 12, that determines ER␣ stability. This was demonstrated by showing that mutations within transcriptional activation domain 2 (ER␣-AF-2) of the receptor, which negatively impacted transcriptional activation in the presence of estradiol, increased the stability of the receptor. However, when a heterologous VP16 acidic activator was placed on these mutants, transcriptional activity was restored, and the receptor stability mirrored the wild-type receptor. This observation suggests that it is not AF-2 per se that is FIG. 6. Ubiquitination of ER␣ is influenced by ligands. A, MCF-7 cells were pretreated with lactacystin (20 M) for 3 h followed by treatment with each ligand for 4 h. The amount of endogenous ER␣ present in whole cell extracts was analyzed by immunoblotting with the ER␣ antibody. Loading of protein has been normalized to endogenous ERK1 protein levels. Results obtained following solvent or ICI 182,780 treatment are shown here. B-D, HeLa cells were transiently transfected with expression vectors for ER␣-wt and/or His 6 -tagged ubiquitin as indicated. Forty-eight h later, cells were treated with ligand for 4 h. Duplicate experiments were performed in the presence of lactacystin where the cells were pretreated with 20 M lactacystin for 3 h and treated with each ligand or solvent for 4 h. Cells were then lysed and were either loaded directly (lysate) or subjected to Ni 2ϩ -nitrilotriacetic acid purification followed by loading onto SDS-PAGE (nickel column eluate). His 6 -ubiquitin-ER␣ complexes were visualized by immunoblotting with an antibody directed against ER␣. B, nickel column eluate subjected to SDS-PAGE followed by immunoblotting with an anti-ER␣ antibody. C, 1% of lysate prior to enrichment through the nickel column. D, densitometric quantitation of the ratio of the first ER␣-ubiquitin degradation intermediate (in the absence of lactacystin, as denoted by an asterisk) to the corresponding ER␣ levels present in the lysate in graph form. E, B-D were repeated with the ER␣-LL construct in the presence and absence of 4OH-tamoxifen. The asterisk denotes the ER␣-LL-ubiquitin conjugate formed. recognized by the degradation machinery but that it is transcriptional activity as a whole that triggers degradation. Furthermore, the observation that the level of ubiquitination of unliganded and estradiol-bound receptor ER␣ is similar suggests that ubiquitination is necessary but not sufficient to permit estradiolinduced degradation of ER␣. Based on these findings, we suggest two mechanisms by which transcriptional activity may influence receptor degradation: 1) the transcription-initiation complex recruits subunits of the proteasome, thereby localizing the degradation machinery with the target, and/or 2) agonist-induced receptor is recruited to loci, where active transcription takes place and where components of the proteasome reside. Indeed, several proteins that are involved in the ubiquitin proteasome pathway, such as E3 ubiquitin ligase RSP5/RPF1 (49), mSiah2 (50), 26 S proteasome regulatory subunit SUG1 (51,52), and ubiquitinconjugating enzyme 9 (53), have been shown to interact with steroid receptors or steroid receptor-specific cofactors. Further experimentation is necessary, however, to determine if it is the formation of a transcriptional activation complex, initiation, or the reinitiation of transcription that triggers the degradation of ER␣ when bound to estradiol.
The differences in the susceptibility of an ER␣-ligand complex to degradation may be modulated in part by receptor compartmentalization. Recent studies have demonstrated that ER␣ is distributed in a reticular pattern within the nuclei in the absence of ligand (54). The addition of either estradiol or 4OH-tamoxifen results in a rapid and dramatic redistribution of ER␣ into a punctate pattern, whereas the addition of ICI 182,780 results in trapping of ER␣ in the cytoplasm (37,54). Since the distribution of proteasome factors within the cytoplasm and the nucleus is distinct (55,56), it is possible that these distinct ER␣-ligand complexes interact with distinct proteasome complexes that degrade them at different rates.
In conclusion, we have demonstrated that ER␣ ligands, upon binding the receptor, are able to differentially alter the stability of the receptor. Whereas estradiol-induced degradation is correlated with transcriptional activation, other factors are responsible for regulating the stability of SERM-and pure antagonist-occupied receptor. We speculate that the differences in stability observed in the presence of different SERMs and the pure antagonist reflect subtle differences in ER␣ conformation induced by these ligands, which regulate the interaction of the receptor with different components of the proteosome. It is possible that ER␣ degradation occurs by different mechanisms in different cells depending on the tissue-specific distribution of components that mediate degradation. If this turns out to be the case, then this activity may be exploited in the development of new classes of tissue-selective ER␣ ligands.