Estrogen Inhibits Transforming Growth Factor β Signaling by Promoting Smad2/3 Degradation*

Estrogen is a growth factor that stimulates cell proliferation. The effects of estrogen are mediated through the estrogen receptors, ERα and ERβ, which function as ligand-induced transcription factors and belong to the nuclear receptor superfamily. On the other hand, TGF-β acts as a cell growth inhibitor, and its signaling is transduced by Smads. Although a number of studies have been made on the cross-talk between estrogen/ERα and TGF-β/Smad signaling, whose molecular mechanisms remain to be determined. Here, we show that ERα inhibits TGF-β signaling by decreasing Smad protein levels. ERα-mediated reductions in Smad levels did not require the DNA binding ability of ERα, implying that ERα opposes the effects of TGF-β via a novel non-genomic mechanism. Our analysis revealed that ERα formed a protein complex with Smad and the ubiquitin ligase Smurf, and enhanced Smad ubiquitination and subsequent degradation in an estrogen-dependent manner. Our observations provide new insight into the molecular mechanisms governing the non-genomic functions of ERα.

LBDs (5). Helix 12, the C-terminal helix, forms the critical core (AD core) of AF-2 function for the receptor. This region plays an important role in the binding of coactivators to the ligand-bound receptor (6). Estrogen-dependent ER␣ transactivation is also regulated by the ubiquitin-proteasome pathway (7)(8)(9)(10). ER␣ and its coactivators cycle onto and off of estrogen-responsive promoters in a ligand-dependent manner (11)(12)(13). ER␣ is ubiquitinated after each round of transcription, facilitating release from the promoter; this event may be essential for subsequent ER␣-dependent transcription.
TGF-␤ superfamily (TGF-␤, activin, and bone morphogenic protein (BMP))/Smad signaling is reportedly regulated by estrogen/ER signaling. In MCF-7 cells, TGF-␤-induced transcription and migratory potential are inhibited by estrogen (31), and activin and estrogen signaling are reciprocally suppressed their signaling (32). Smad3-dependent transcription is inhibited by ER␣ through binding to Smad3, and the inhibition is abrogated by the expression of AP-1 transcription factors (33,34). A complex of Smad3/4 mediates the inhibition of ER␣mediated estrogenic activity of gene transcription in breast cancer cells (35). In addition, ER␣ interacts with Smad1, which is a downstream signal transducer of BMP signaling, to inhibit BMP-induced transcription (36,37). These observations sug-gest the regulatory interaction between Smad and ER␣; however, it remains poorly understood at the molecular level.
Here, we show that ER␣ inhibits TGF-␤ signaling by decreasing Smad protein levels. ER␣-mediated reductions in Smad levels did not require the DNA-binding ability of ER␣, implying that ER␣ opposes the effects of TGF-␤ via a novel non-genomic mechanism. Our analysis revealed that ER␣ formed a protein complex with Smad and the ubiquitin ligase Smurf, and enhanced Smad ubiquitination and subsequent degradation in an estrogen-dependent manner. Our observations provide new insight into the molecular mechanisms governing the nongenomic functions of ER␣.
Cell Culture and Transfection-MCF-7 and MDA-MB-231 breast cancer cells, and 293 cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS). Twenty-four hours before transfection, cells were shifted to phenol red-free DMEM containing 4% charcoal-stripped FBS. Transfection was performed with either PerFectin Transfection Reagent (Gene Therapy Systems) or TransFast Transfection Reagent (Promega) according to the manufacturers' protocols. Cells were treated with or without TGF-␤ (1 ng/ml) and estrogen (10 Ϫ8 M). Twenty-four hours after the addition of estrogen, cells were harvested and analyzed by Western blotting using appropriate antibodies.
Luciferase Assay-The 9ϫCAGA-Luc plasmid was co-transfected into 293 cells with expression vector encoding either wild-type or mutant ER␣. We co-transfected the ph-RL-TK vectors into all cell lines as a reference plasmid, which was used to normalize transfection efficiency. Twenty-four hours after transfection, we added either estrogen (10 Ϫ8 M) or vehicle alone (ethanol) to the cells with fresh medium containing 1% Charcoal-stripped FBS for an additional 24 h of incubation. Luciferase assays were performed using cell extracts according to the manufacturer's protocol (Promega). Individual transfections were assessed in triplicate and repeated at least three times.
Pulse Chase-MCF-7 cells were cultured in medium containing 4% charcoal-stripped FBS for 48 h, which was then replaced with fresh medium containing 4% FBS. After 24 h, fresh culture medium containing 0.2% charcoal-stripped FBS, TGF-␤ (1 ng/ml), and cycloheximide (100 g/ml) with or without estrogen (10 Ϫ8 M) was added. Cells were harvested at several time points after the addition of cycloheximide.
Ubiquitination Assay-293 cells were transiently co-transfected with vectors encoding HA-tagged ubiquitin (Ub) and Myc-tagged Smad2, Myc-tagged Smurf1, or ER␣ as indicated. Cells were cultured in the presence of MG132 (1 M) with or without estrogen (10 Ϫ8 M), and lysed in radioimmune precipitation assay buffer (RIPA). After clarification by centrifugation, soluble proteins were immunoprecipitated from the cell extracts using anti-Myc (Nacalai Tesque) or anti-HA (Sigma) agarose for 4 h at 4°C. The resin was washed in RIPA buffer, and immunoprecipitates were immunoblotted using anti-HA or anti-Myc antibodies.
GST Pulldown Assay-We prepared GST fusion protein constructs of full-length Smad2 or Smurf1 by subcloning the protein-coding regions into pGEX-4T-1 (Amersham Biosciences, Piscataway, NJ). Constructs were individually transformed into Escherichia coli BL21 cells; recombinant protein expression was induced for 6 h with 0.1 mM isopropyl-thio-␤-D-galactopyranoside. We used glutathione-Sepharose (GE) to purify the GST fusion proteins according to the manufacturer's instructions. In vitro translated, [ 35 S]methionine-labeled ER␣ constructs were prepared using the TNT coupled transcription/ translation system (Promega, Madison, WI). Five microliters of labeled protein was mixed with ϳ1 g of purified GST-Smad2 or GST-Smurf1 bound to glutathione-Sepharose (GE). This mixture was incubated for 1 h in GST pull-down buffer (50 mM We then determined the protein levels of Smad2/3, phosphorylated Smad2/3, ER␣, and ␤-actin by Western blotting using appropriate antibodies. B, MCF-7 cells were transfected with control or ER␣ siRNA and cultured in the presence or absence of TGF-␤ or estrogen. Phosphorylated Smad2, Smad3 levels were examined by Western blotting. C, expression plasmids encoding FLAG-tagged Smad2 and Smad3 were co-transfected into 293 cells with ER␣ and a constitutively active TGF-␤ type I receptor ALK5 TD expression plasmids. These cells were then incubated in the presence or absence of estrogen or MG132. The protein levels of Smad2, Smad3, pSmad2, pSmad3, and ER␣ were examined by Western blotting using the indicated antibodies. D, DNA binding ability of ER␣ was not required for induction of Smad degradation. FLAG-tagged Smad2, Smad3, and ALK5 TD expression plasmids were transfected into 293 cells with or without expression plasmids encoding ER␣ or ER␣(mC). These cells were then incubated in the presence or absence of estrogen and MG132. We then determined the protein levels of Smad2, Smad3, ER␣, and ER␣(mC) by Western blotting using protein-specific or anti-FLAG-M2 antibodies. E, estrogen facilitated Smads degradation. MCF-7 cells were cultured in medium containing cycloheximide (100 g/ml). Cells were harvested at the indicated times. Smad protein levels in untreated and estrogen-treated cells were examined by immunoblotting (left). The bands were quantified and represented by graphs (right). F, estrogen induced Smad2 ubiquitination. Plasmids encoding Myc-tagged Smad2, HA-tagged ubiquitin, and ALK5 TD were transfected into 293 cells in the presence or absence of an ER␣ expression plasmid. After culturing transfected cells in the presence of MG132 with or without estrogen, ubiquitinated proteins were immunoprecipitated with an anti-HA antibody. Ubiquitinated Smad was detected by immunoblotting these immunoprecipitates with an anti-Myc antibody (upper panel). To confirm Smad ubiquitination, Myc-tagged Smad2 was immunoprecipitated using an anti-Myc antibody; the ubiquitination of the immunoprecipitated protein was confirmed by immunoblotting with an anti-HA antibody (second panel). Smad2 and ER␣ were detected in whole cell lysates by immunoblotting using anti-Myc (third panel) and anti-ER␣ (fourth panel) antibodies, respectively.
Affymetrix Microarray Analysis-Total RNA was prepared from samples using an RNeasy RNA isolation kit (Qiagen). The preparation of in vitro transcription products and the hybridization and scanning of the oligonucleotide array were performed according to Affymetrix protocols (2001 Affymetrix Genechip Technical Manual; Santa Clara, CA).
Invasion and Migration Assay-Suspensions (0.5 ml) containing 1.25 ϫ 10 5 cells (invasion assay) or 0.5 ϫ 10 5 cells (migration assay) were layered in the upper compartments of rehydrated Matrigel Invasion chambers or transwell chambers, respectively. These cells were then incubated for 20 h (invasion assay) or 12 h (migration assay) at 37°C with 0.75 ml of DMEM in the lower chambers. After incubation, cells on the upper surface of the filter were removed, and invading or migrating cells were fixed in methanol. Fixed cells were stained with crystal violet and counted under a microscope.
Statistical Analysis-Significance of differences was determined by Student's t test analyses, using Microsoft Excel.

RESULTS
Estrogen Abrogates TGF-␤-dependent Transcription-To examine the effects of estrogen on TGF-␤ signal transduction, we performed a transcription reporter assay. A reporter plasmid encoding a TGF-␤ response element (9ϫCAGA) was transfected into MCF-7 cells. TGF-␤ treatment enhances transcription from the reporter plasmid, which was profoundly inhibited by estrogen treatment (Fig. 1A, lanes 3 and 4). We also examined plasminogen activator inhibitor-1 (PAI-1) mRNA level, which is one of the target genes of TGF-␤/Smad signaling, using real-time RT-PCR. TGF-␤-induced PAI-1 mRNA expression level was significantly reduced by the treatment with estrogen (Fig. 1B). To examine whether estrogen inhibits TGF-␤ signaling via ER␣, we reduced the endogenous level of ER␣ in MCF-7 cells using siRNA. In ER␣-deficient cells, we did not observe an estrogen-dependent inhibitory effect on PAI-1 mRNA levels (Fig. 1C). Next, we used microarray analysis to determine the effects of estrogen on several endogenous TGF-␤-induced gene expression. This approach allowed us to compare the gene expression profiles of untreated (Fig. 1D, lane 1), TGF-␤treated (Fig. 1D, lane 2), and TGF-␤ and estrogen-treated MCF-7 cells (Fig. 1D, lane 3). Of the 54,675 genes represented on the microarray, the expression levels of 956 genes increased greater than 2-fold following TGF-␤ treatment (Fig. 1D, lane 2). Estrogen reduced the expression of 683 of these 956 genes upregulated by TGF-␤ treatment (Fig. 1D, lane 3). These data suggest that estrogen antagonizes TGF-␤-dependent transcriptional regulation via ER␣.
We further investigated the molecular mechanism of ER␣mediated transcriptional suppression by transfecting the reporter plasmid (9ϫCAGA) into the human kidney cell line 293 in the presence of either a constitutively active TGF-␤ type I receptor (mutation of Thr-204 to Asp) or a catalytically inactive receptor (ALK5 KR; mutation of Lys-378 to Arg). ALK5 TD, but not ALK5 KR, induced transcription from the reporter plasmid promoter (Fig. 1E, lanes 2 and 3). Co-expression of ER␣ in ALK5 TD-expressing cells reduced transcriptional activity; further reduction was observed following estrogen treatment (Fig. 1E, lanes 5 and 6). We obtained similar results following co-transfection of ALK5 TD with ER␣(mC), a form of ER␣ bearing three amino acid substitutions in the DNAbinding domain (10) that eliminates the interaction of the receptor with DNA (Fig. 1E, lanes 7 and 8). These results suggest that inhibition of TGF-␤ signaling does not require ER␣ binding to estrogen-responsive elements (ERE) within the nucleus.
Estrogen Induces the Degradation of Smad and Smurf Proteins-As Smad proteins are key transducers in TGF-␤ signaling, it is possible that estrogen inhibits TGF-␤ signaling by preventing the recruitment of Smad to promoter region of TGF-␤ responsive genes. Therefore, we performed the chromatin immunoprecipitation experiment to examine whether estrogen interferes with DNA binding of Smad. As shown in supplemental Fig. S1, the binding of Smad3 on PAI-1 promoter was reduced by treatment with estrogen. These data suggest that ER␣ inhibits TGF-␤ signaling by reducing the recruitment of Smad protein on the promoter region. Thus, we examined the endogenous protein levels of Smad2 and -3 by Western blotting. In MCF-7 cells, the protein levels of both Smad2 and 3 were reduced by co-treat- ment with TGF-␤ and estrogen ( Fig. 2A). Following TGF-␤ receptor ligation, phosphorylated Smads translocate into the nucleus. In addition to reductions in total protein, phosphorylated Smad2 and 3 levels were significantly decreased by estrogen treatment (Fig. 2A). In contrast to estrogen, treatment with the pure ER antagonist ICI182,780, which attenuates receptor dimerization, effecting rapid degradation of the ER protein and inhibition of transcription, did not markedly affect Smad protein levels ( Fig. 2A). The protein levels of ER␣ were reduced in the presence of estrogen or ICI182,780 ( Fig. 2A). The estrogen-dependent reduction in Smad2 and 3 could be abrogated by treatment with the proteasome inhibitor MG132 (Fig. 2A, lane 7). In addition, Smad mRNA levels remained unchanged by estrogen treatment (data not shown). These results indicated that estrogen induces Smad proteolysis. To examine whether estrogen induces Smad degradation via ER␣, we reduced the endogenous levels of ER␣ in MCF-7 cells using siRNA. In ER␣-deficient cells, we did not observe an estrogen-dependent reduction of Smad protein levels (Fig. 2B), indicating that estrogen induces Smad proteolysis via ER␣.
Next, we transfected expression vectors encoding Smad and either ER␣ or ER␣(mC) into 293 cells. We observed estrogen-dependent degradation of both Smad and phosphorylated Smad proteins following co-expression of ER␣ (Fig.  2C). Consistent with the results of the transcription reporter assay (Fig. 1E), ER␣(mC) also induced Smad protein degradation (Fig. 2D). These data raise the possibility that Smad protein degradation does not require ER␣ binding to ERE within the nucleus and ER␣-dependent transcription. To obtain further evidence for this hypothesis, we inhibited ER␣-dependent transcription using transcriptional inhibitor, ␣amanitin, and tested its effects on estrogen-dependent Smad degradation. As expected, the degradation of Smad by estrogen and ER␣ was not affected by treatment with ␣-amanitin (supplemental Fig. S2). These findings indicate that ER␣dependent transcription was not necessary for the estrogen-dependent Smad degradation. We also examined the effects of ER␤ on TGF-␤ signaling and Smad degradation. As shown in supplemental Fig. S3, ER␤ also inhibited TGF-␤ signaling and enhanced Smad degradation in an estrogen-dependent manner, similar to ER␣.
We confirmed that estrogen enhances Smad degradation using pulse-chase experiments. In the absence of estrogen, the half-lives of the Smad proteins exceeded 7.5 h (Fig. 2E). Estrogen treatment reduced the Smad protein halflives to less than 4.5 h (Fig. 2E). We also examined the effect of estrogen on Smad2 ubiquitination. Exogenous expression of ER␣ resulted in broad bands on Western blots, which were consistent with ubiquitin-conjugated Smad2 (Fig. 2F). These broad bands intensified in samples treated with estrogen (Fig. 2F), suggesting that estrogen enhances Smad2 ubiquitination.
Smurf1 and -2 are known as an E3-type ubiquitin ligase for Smad proteins. Therefore, we next determined the protein levels of Smurf. Interestingly, treatment with both TGF-␤ and estrogen reduced Smurf protein levels (Fig. 3A). This reduction was inhibited by MG132 (Fig. 3A, lane 5), suggesting Smurf levels were also decreased via proteasomal degradation in an estrogen-dependent manner. We next transfected into 293 cells either Smurf1 or Smurf1 CA, a mutant form of the protein bearing an amino acid substitution that abolishes ubiquitin ligase activity, in the presence or absence of ER␣. Steady state levels of Smurf1 were reduced by ER␣ expression (Fig. 3B, lane  3 of upper panel); this reduction was enhanced by estrogen

. ER␣ inhibits TGF-␤ signaling via Smad degradation.
A and B, Smurf1 enhanced estrogen-dependent Smad degradation. Plasmids encoding FLAG-tagged Smad2, Smad3, and ALK5 TD were transfected with or without those encoding ER␣, HA-tagged Smurf1, or HA-tagged Smurf1 CA. After culturing transfected cells with or without estrogen (10 Ϫ8 M) or MG132 (10 Ϫ6 M), we evaluated the indicated protein levels by Western blotting using appropriate antibodies. C, Smurf1 enhanced Smad2 ubiquitination. Vectors encoding Myctagged Smad2, HA-tagged ubiquitin, and ALK5 TD were transfected into 293 cells with or without plasmids encoding ER␣, Smurf1, or Smurf1 CA. After incubation in the presence of MG132 with or without estrogen, Myc-tagged Smad2 or HA-tagged ubiquitin was immunoprecipitated with anti-Myc or anti-HA antibodies, respectively. The ubiquitination status of Smad2 was assessed by Western blotting probed with anti-HA antibody or anti-Myc antibody, respectively. D, ubiquitin-proteasome pathway was involved in ER␣-mediated inhibition of TGF-␤-dependent transcription. Expression plasmids encoding ALK5 KR or ALK5 TD, the 9xCAGA luciferase reporter plasmid, and pRSV␤GAL were transfected into 293 cells with or without plasmids encoding ER␣ or Smurf1 CA. Transfected cells were cultured in the presence or absence of MG132 prior to examination of cell extracts by luciferase assay. E, stably ER␣-expressing MDA-MB-231 cells were transfected with control or the mixture of Smurf1 and Smurf2 siRNAs and cultured in the presence or absence of estrogen. Smad2, Smad3, and Smurf1 protein levels were examined by Western blotting. treatment (Fig. 3B, lane 4 of upper panel). In contrast, Smurf1 CA levels were unaffected by ER␣ co-expression (Fig. 3B, lanes  3 and 4 of lower panel), suggesting that estrogen-dependent Smurf1 degradation requires its ubiquitin ligase activity. We next examined the effect of estrogen on Smurf1 ubiquitination. The broad band representing multiple ubiquitin-conjugated Smurf1 products intensified following ER␣ expression (Fig. 3C,  lane 4); estrogen treatment enhanced the intensity of this signal (Fig. 3C, lane 5), indicating that estrogen-bound ER␣ enhances Smurf1 ubiquitination.
ER␣ Forms a Protein Complex with Smad and Smurf and Induces Their Simultaneous Degradation-Next, we assessed the effect of Smurf1 expression on the estrogen-dependent degradation of Smad2 and 3. Under condition in which Smurf1 expression exerted little effect on Smad2 and -3 protein levels in the absence of ER␣ (Fig. 4A, lanes 3 and 4); co-expression with ER␣ enhanced the reduction of Smad2 and -3 in the presence of estrogen (Fig. 4A,  lanes 6 and 8). Smurf1 also enhanced estrogen-dependent ER␣ down-regulation. In contrast, Smurf1 CA expression inhibited the estrogen-dependent reduction in ER␣ and Smad levels (Fig. 4A,  lanes 9 and 10). The reduction in these proteins was abrogated by treatment with MG132 (Fig. 4B). These results confirmed the participation of Smurf1 in ER␣-mediated Smad degradation. Whereas co-expression of Smurf1 with ER␣ and Smad2 enhanced estrogen-dependent Smad2 ubiquitination (Fig. 4C, lanes 7 and 8), Smurf1 CA expression reduced ubiquitinated Smad2 levels (Fig.  4C, lane 9), indicating that Smurf1 functions in estrogen-mediated Smad ubiquitination. Smurf2 exerted similar effects to those seen for Smurf1 on Smad ubiquitination and degradation (data not shown). Therefore, we next attempted to determine the effect of Smurf1 expression on the estrogen-dependent inhibition of TGF-␤-induced transcription using a reporter assay. Estrogen-dependent suppression of reporter transcription was abrogated by either Smurf1 CA overexpression or MG132 treatment (Fig. 4D). As both of these treatments inhibit Smad degradation, it is likely that Smad protein degradation mediates ER␣-induced inhibition of TGF-␤ signaling. To confirm that ER␣ inhibits TGF-␤ signaling by inducing Smad degradation, we reduced endogenous levels of Smurf1 and Smurf2 using siRNA. In Smurf-deficient cells, we did not observe estrogen-dependent Smad degradation (Fig. 4E).
Estrogen treatment induced the simultaneous degradation of Smad2, Smurf1, and ER␣ in cells co-expressing these proteins (Fig. 4A, lane 8), suggesting that ER␣, Smad2, and Smurf1 form a complex within the nucleus. We examined the interactions between these proteins by co-immunoprecipitation from 293 cells transfected with ER␣ and either FLAGtagged Smad2 or 3 (Fig. 5A) or Myc-tagged Smurf1 CA (Fig.  5B). ER␣ was detected in anti-FLAG and anti-Myc immunoprecipitates by protein immunoblotting with antibodies against ER␣ (Fig. 5, A and B), suggesting that Smad2, Smad3, and Smurf1 interact with ER␣. These interactions were detected in both the absence and presence of estrogen. To confirm the interaction of endogenous proteins, we immunoprecipitated proteins from MCF-7 cell extracts using an anti-ER␣ antibody. By immunoblotting using antibodies against either Smad2 or Smurf1, we determined that the precipitated proteins included Smad2 and Smurf1 in the presence of TGF-␤ (Fig. 5C), confirming an interaction between ER␣ and these proteins in vivo.
To investigate if ER␣ forms a ternary complex containing both Smad and Smurf, we transfected ER␣, FLAG-tagged Smad2, and Myc-tagged Smurf1 into 293 cells. Proteins were initially immunoprecipitated with an anti-FLAG antibody. After elution with excess FLAG peptide, we performed a secondary immunoprecipitation with anti-Myc antibody. Immunoblotting of the precipitate with an anti-ER␣ antibody demonstrated the presence of ER␣ in the final immunoprecipitate (Fig. 5D), suggesting that ER␣ formed a ternary complex with both Smad and Smurf. To show further evidence that ER␣ forms a ternary complex with Smad and Smurf, and enhances simultaneous ubiquitination and degradation of these proteins, we examined if the ER␣ ubiquitination level is enhanced by the co-expression of Smad and Smurf in an estrogen-dependent manner. As shown in supplemental Fig. S4, ubiquitinated ER␣ were significantly increased by co-expression of Smad2 and Smurf1. These data indicate that ER␣ formed a ternary complex with both Smad and Smurf, and induces simultaneous degradation of these proteins to inhibit TGF-␤ pathways.
Note that estrogen-dependent Smad degradation of exogenously expressed Smad proteins occurred even in the absence FIGURE 5. ER␣, Smad, and Smurf form a ternary complex. A, ER␣ associated with Smad. Expression plasmids encoding FLAG-tagged Smad2 or Smad3 were co-transfected with or without an ER␣ expression plasmid into 293 cells. Smad proteins were immunoprecipitated from the cell extracts using an anti-FLAG-M2 antibody. ER␣ was detected in precipitates by Western blotting using an antibody against ER␣. *, antibody. B, ER␣ associated with Smurf. We co-transfected 293 cells with a Myc-tagged Smurf1 CA expression plasmid in the presence or absence of an ER␣ expression plasmid. Smurf1 CA protein was immunoprecipitated from cell lysates using an anti-Myc antibody. ER␣ was detected in precipitates by Western blotting using an antibody against ER␣. C, endogenous ER␣, Smad, and Smurf proteins formed a complex. After lysis of MCF-7 cells, proteins were immunoprecipitated from cell lysates using an anti-ER␣ antibody in the absence or presence of TGF-␤ with or without estrogen (10 Ϫ8 M). Precipitates were examined by Western blotting using antibodies against Smad2 and Smurf1. D, ER␣ formed a ternary complex with Smad and Smurf. FLAG-tagged Smad2 and/or Myc-tagged Smurf1 expression plasmids were co-transfected into 293 cells with or without an ER␣ expression plasmid. Smad2 was immunoprecipitated from cell lysates using an anti-FLAG-M2 antibody. Proteins were eluted from the beads using FLAG peptide, then re-immunoprecipitated with an anti-Myc antibody. ER␣ was detected in the precipitate by Western blotting using an antibody against ER␣. of ALK5 TD in 293 cells (supplemental Fig. S5; left), indicating that TGF-␤-dependent Smad phosphorylation is not necessary for its degradation. Immunohistochemical analysis of 293 cells revealed that Smad2 localized in the nucleus in the absence of ALK5 TD (supplemental Fig. S5; right), further indicating that the co-localization of Smad, Smurf, and ER␣ is necessary for ternary complex formation and subsequent Smad degradation.
As phosphorylated Smad2 and 3 bind to Smad4 and the complex translocates into the nucleus to stimulate the transcription of target genes, we examined whether ER␣ interacts with Smad4 (supplemental Fig. S6). In contrast to Smad2 and 3, Smad4 was not co-immunoprecipitated with ER␣. In addition, we found that the protein levels of Smad4 did not change by treatment with estrogen (supplemental Fig. S7). These data suggest that Smad4 is absent from the complex containing ER␣, R-Smad, and Smurf in the nucleus.
The Degradation of Smad and Inhibition of TGF-␤-dependent Transcription and Cancer Invasion Are Mediated by Identical Regions within ER␣-To identify the regions of ER␣ responsible for Smad degradation, we generated several truncations of ER␣ (Fig. 6A). GST pull-down experiments using this series of truncated ER␣ proteins demonstrated that the C region alone was sufficient to bind both Smad2 and Smurf1 (Fig. 6B). ER␣ABC and ER␣CDEF were both able to bind Smad2 and Smurf1 and subsequently stimulate Smad2 degradation (Fig. 6, B and C). In contrast, ER␣AB and ER␣DEF, neither of which could bind Smad or Smurf, had no effect on Smad protein levels, indicating that the interaction of ER␣, Smad, and Smurf is necessary to induce degradation. Whereas both ER␣C and ER␣CDE(⌬AD) were able to bind Smad and Smurf, expression of either construct could not induce degradation of Smad (Fig. 6, B and C). These results imply that the binding of ER␣ to Smad and Smurf is necessary, but not sufficient, for degradation. All ER␣ deletion mutants capable of inducing Smad2 degradation contained the transactivation domain (Fig. 6, B and C). As ER␣C and ER␣CDE(⌬AD) did not possess transcriptional activity, coactivator recruitment to the receptor may be necessary to induce Smad degradation. Next, we tested the effect of individual ER␣ deletion mutants on TGF-␤-dependent transcription. We co-transfected truncated forms of ER␣ into 293 cells with a plasmid encoding ALK5 TD and the 9ϫCAGA reporter plasmid. The ER␣ mutants capable of inducing Smad protein degradation also abrogated TGF-␤-dependent transcription (Fig. 6D), indicating that ER␣-induced inhibition of TGF-␤ signaling arises from Smad degradation.
It is well known that TGF-␤ promotes cellular migration and invasion of breast cancer cells by increasing several target gene expressions. Therefore, we tested the effects of estrogen on TGF-␤induced migration and invasion. As shown in Fig. 7A, the migration of ER␣-expressing MDA-MB-231 cells was increased by treatment with TGF-␤, and significantly reduced by co-treatment with estrogen. Invasion of these cells was also reduced by co-treatment with estrogen (Fig.  7B). The expression of ER␣ mutants capable of reducing TGF-␤ signaling decreased the migratory and invasive potential of MDA-MB-231 cells (Fig. 7, A and B). In contrast, the ER␣ mutants incapable of reducing TGF-␤ signaling had no effect on migration or invasiveness (Figs. 6, B-D and 7, A and B). These findings indicate that estrogen/ER␣ inhibits cellular migration and invasion via inhibition of TGF-␤ signaling. Taken together, our results indicated that ER␣ forms a ternary complex with Smad and Smurf, enhances ubiquitination and degradation of Smad proteins, and inhibits TGF-␤ signaling pathway (Fig. 8).

DISCUSSION
In summary, ER␣ forms a protein complex with Smad and Smurf, and induces simultaneous degradation of these proteins to inhibit TGF-␤ pathways in an estrogen-dependent manner. As estrogen-dependent suppression was abrogated by inhibiting Smad degradation, it is likely that Smad protein degradation mediates ER␣-induced inhibition of TGF-␤ signaling. This ER␣-mediated degradation and inhibition of TGF-␤-dependent transcription did not require the DNA binding ability of ER␣, and are mediated by identical regions within ER␣, implying that ER␣ opposes the effects of TGF-␤ via a novel non-genomic mechanism.
To the best of our knowledge, this is the first study showing that ER␣ enhances the degradation of Smad proteins via a ubiquitin-proteasome pathway. ER␣-dependent Smad degradation and inhibition of TGF-␤-induced transcription did not require the binding of ER␣ to an ERE. We have already identified several proteins that are degraded via an ER-dependent pathway. 4 In addition, it was reported that aryl hydrocarbon receptor (AhR) induces the degradation of ER␣ (39). Considering these results together, ER␣-dependent protein degradation appears to be a novel non-genomic function of ER␣. Our data indicate that ER␣ non-genomically induces Smad degradation and that Smad degradation makes a major contribution to the ER␣-mediated inhibition of TGF-␤ signaling.
The dependence of ER␣-mediated Smad degradation on the presence of the ER␣ transactivation domain raises the possibility that coactivator binding to the receptor may be necessary for degradation. Several recent reports suggest that Smad is acetylated by CBP/p300 and that the acetylation status of Smad affects its ubiquitination and degradation. Overexpression of mutated CBP, which does not possess an acetyltransferase activity, inhibits estrogen-dependent Smad and ER␣ degradation (data not shown), supporting a role for coactivators in degradation. Therefore, we determined the acetylation status of Smad and Smurf using antibodies specific for acetylated proteins. The acetylation status of these proteins, however, was not altered by estrogen treatment (data not shown). A ubiquitinconjugating enzyme (E2) also may be part of the coactivator complex (40). One report showed that an E2, UbcH7, interacts with SRC-1 to regulate the transcriptional activity of NRs (41). Such an interaction between ER␣ and an E2 may be necessary for Smad degradation. Here, we first demonstrated the novel non-genomic ER␣ function, by which TGF-␤ signaling is inhibited via Smad degradation. Our observations provide new insights into the non-genomic functions of ER␣.