SPBP Is a Phosphoserine-Specific Repressor of Estrogen Receptor α

Abstract

Multiple signaling pathways stimulate the activity of estrogen receptor α (ERα) by direct phosphorylation within its N-terminal activation function 1 (AF1). How phosphorylation affects AF1 activity remains poorly understood. We performed a phage display screen for human proteins that are exclusively recruited to the phosphorylated form of AF1 and found the stromelysin-1 platelet-derived growth factor-responsive element-binding protein (SPBP). In a purified system, SPBP bound only the in vitro-phosphorylated form of the ERα AF1 or the phosphoserine mimic S118E, and the interaction domain could be mapped to a 42-amino-acid fragment of SPBP. In cells, SPBP preferentially interacted with liganded and phosphorylated ERα. Functionally, SPBP behaved as a repressor of activated ERα, which extends its previously demonstrated roles as a DNA binding transactivation factor and coactivator of other transcription factors. By targeting the phosphorylated form of AF1, SPBP may contribute to attenuating and fine-tuning ERα activity. A functional consequence is that SPBP inhibits the proliferation of ERα-dependent but not ERα-independent breast cancer cell lines, mirroring a reported negative correlation with the ERα status of breast tumors.

Estrogen receptors (ER) α and β are members of the nuclear receptor superfamily and mediate the responses to estrogens as well as a variety of other extracellular signals by signaling cross talk (reviewed in references 18, 36, 38, and 40). As nuclear receptors, they contain all of the hallmark domains of receptors and DNA-bound transcription factors (see Fig. 1). Transcriptional activation is mediated primarily by the two activation functions, AF1 and AF2. AF2 is associated with the ligand binding domain, and its activity is induced by a ligand-induced allosteric rearrangement (reviewed in references 37 and 50), which allows the recruitment of coregulatory proteins (reviewed in references 19, 33, and 58).

FIG. 1.

Schematic representation of human ERα and its mutants and GST fusions. Functional and structural domains are indicated. AF1 and AF2 map to the A/B and E/F domains, respectively. The DNA binding domain (DBD) and HBD consist of domains C and E, respectively. Amino acid substitutions are given above the open boxes. Designations on the right refer to the proteins. wt, wild type.

AF1 resides in the very poorly conserved N-terminal domain of nuclear receptors. Consequently, nuclear receptors differ considerably with respect to AF1 activity and regulation. The activity of ERα AF1 is typically constrained in the context of the intact receptor. Upon activation of ERα (for example, by a cognate ligand), AF1 activity contributes to and synergizes with AF2 activity in a cell-specific fashion (34, 54, 56). A large number of factors have been reported to bind ERα AF1 (see http://www.picard.ch for an overview), but their relative importance and the mechanistic details are far less well understood than those for AF2. Importantly, the phosphorylation of certain serine residues, notably the major phosphorylation site S118 (1, 22, 23, 28; reviewed in reference 27), increases AF1 activity. Several kinases, including Erk1/2, Rsk1/2, Cdk2, Cdk7, and Akt, can phosphorylate one or several of these serine residues. This places ERα at the crossroads of several signaling pathways. These can further stimulate the AF1 activity of estrogen-activated ERα and in some cases can even mediate the activation of unliganded ERα by growth factors (1-3, 5, 6, 8, 21-23, 28, 31, 45; reviewed in references 27 and 40). How phosphorylation of AF1 modulates, and in particular activates, ERα activity remains enigmatic (see Discussion).

As a first step toward identifying the factors that modulate the phosphorylation-induced activity of AF1 and thereby influence cross talk between ERα and other signaling pathways, we have isolated a factor, SPBP (stromelysin-1 platelet-derived growth factor [PDGF]-responsive element-binding protein), whose recruitment to ERα is strictly phosphorylation dependent. This factor had initially been discovered as a PDGF-induced transcription factor that binds to the stromelysin-1 promoter (24, 46). The human version was also referred to subsequently as AR1 or transcription factor 20 (TCF20) (42). Our characterization of SPBP revealed that it acts as a phosphoserine-specific repressor of ERα. Although these findings do not help to explain the stimulatory effects of phosphorylation on AF1 activity, they point to SPBP as a factor that may play an interesting modulatory role after ERα activation.

MATERIALS AND METHODS

Plasmids.

The following plasmids have been described previously: the ERα expression vector HEG0 (53); EREtkLuc (also called XETL), GREtkLuc, ER.GR, and GR.ER (2); pRL-CMV (Promega); and pcDNA3-HA and pcDNA3-HA-hSPBP (44). The latter served to express the hemagglutinin (HA)-tagged short splicing variant of human SPBP, which differs from the “long” form only by the 28 most C-terminal amino acids. Plasmids pCMV4-hAhR and pCMV4-hArnt were gifts from Lorenz Poellinger (25, 32), and pCMV5-hERβ was a gift from J.-Å. Gustafsson. Plasmids pSG5-ER(3×A) and pSG5-ER(3×E) were made by introducing point mutations into HEG0 by PCR to change the serine codons at positions 104, 106, and 118 to encode alanine and glutamic acid residues, respectively. pGEX2T-AF1, pGEX2T-AF1(S118E), and pGEX2T-AF1(3×E), for expression of the corresponding glutathione S-transferase (GST) fusion proteins, were made by inserting the coding sequences for the first 180 amino acids of human ERα, ER S118E, and ER(3×E), respectively, into the pGEX2T vector (Amersham Biosciences). The same fragments were subcloned into the pET32-Ek/Lic vector (Novagen) to produce His₆-tagged thioredoxin fusion proteins of AF1 and AF1(3×E). The construct for producing His₆-SPBP(1459-1615) was made by inserting the EcoRI/HindIII fragment of the T7 phage isolated in the screen into plasmid pET-15b (Novagen). The same vector was used to produce other His₆-tagged SPBP fragments. Different deletion versions of human SPBP for mammalian expression [HA-SPBP(1-1833), HA-SPBP(1-1458), HA-SPBP(1535-1615), and HA-SPBP(1459-1534)] were made by inserting the indicated coding sequences into plasmid pcDNA3-HA. SPBP fragment 1459-1615 was expressed with a different N-terminal HA tag (MQDLPGNDNSTAG) from the mammalian expression vector pC7 (2). Plasmid F-ER (which expresses human ERα with an N-terminal Flag tag) was made by inserting the coding sequences for ERα into the CKF expression vector (11). Plasmids pSCTEV-Gal93-AF1 and pSCTEV-Gal93-AF1(3×E), for expression of Gal4 fusion proteins, were made by inserting codons 82 to 152 of ERα into the pSCTEV-Gal93 vector (48). pC7-Flag-NCoR, which expresses NCoR with a double C-terminal Flag tag, was made by inserting the mouse NCoR (nuclear receptor corepressor) and Flag tag coding sequences into the mammalian expression vector pC7.

Antibodies and recombinant proteins.

The following primary antibodies were used: rabbit polyclonal antiserum (Gramsch Laboratories) raised against recombinant protein His₆-SPBP(1459-1615), mouse monoclonal antibodies against the Flag (antibody M2; Sigma) and His₆ (antibody His-1; Sigma) epitope tags, mouse monoclonal antibodies against ERα (antibodies F3 and ER17, gifts from Daniel Metzger and David F. Smith, respectively), and rabbit polyclonal antiserum against phosphoserine 118 of ERα (Cell Signaling). Two different HA tags and corresponding monoclonal antibodies were used. Unless otherwise indicated, the HA tag YPYDVPDYAHA (antibody HA.11; Babco) was used; the other HA tag was MQDLPGNDNSTAG (antibody anti-HA [described in reference 9]). Recombinant proteins were expressed in Escherichia coli and purified on glutathione-Sepharose (Amersham) or on TALON Metal Affinity Resin (Clontech) as directed by the manufacturer. After elution from the beads, proteins were first dialyzed against a solution containing 20 mM HEPES (pH 7.9), 100 mM KCl, 20% glycerol, 3 mM MgCl₂, 0.2 mM EDTA, and 0.5 mM dithiothreitol (DTT) and then stored at −80°C. For some experiments, the recombinant GST-AF1 fusion protein was in vitro phosphorylated by using p42 mitogen-activated protein (MAP) kinase (Erk2) (New England Biolabs) as directed by the manufacturer.

Phage display screen.

A T7Select human breast cDNA library (Novagen) was chosen for screening according to the manufacturer's protocol. As a bait, GST and His₆-thioredoxin fusion proteins of AF1(3×E) expressed in E. coli were used in alternating rounds of biopanning to eliminate the phages that bind nonspecifically to either glutathione or TALON beads. After seven rounds of biopanning, isolated plaques were used for PCR amplification and for sequencing. The phages from isolated plaques were tested for binding to GST, GST-AF1, and GST-AF1(3×E).

Cell culture, transfection, and luciferase assays.

HepG2 human hepatoma cells, 293T human embryonic kidney cells, and MCF7 human breast cancer cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum (FBS) at 37°C in a humidified atmosphere with 5% CO₂. MCF7-SH cells (which express 10 times more ERα than wild-type MCF7 cells) were cultured in phenol red-free Dulbecco's modified Eagle's medium with 5% charcoal-treated FBS. SkBr3 human breast cancer cells were grown in RPMI 1640 supplemented with 10% FBS. Cells were transfected by using the calcium phosphate coprecipitation method as described previously (12, 13). For luciferase assays, HepG2 cells were plated in six-well dishes and transfected at 50 to 70% confluency. To assay ERα-dependent reporter activity, the following amounts of plasmids were used: 500 ng of HEG0 or its mutated versions, 500 ng of XETL, 10 ng of pRL-CMV, and 1 μg of pcDNA3-HA-SPBP or its deletion variants. In all cases, transfections were compensated for with the empty vector pcDNA3-HA to ensure equal concentrations of plasmid DNA. Sixteen hours after transfection, cells were washed with Tris-buffered saline, and then fresh medium (without serum) without or with hormones or other factors was added. Cells were harvested 24 h later and assayed for luciferase activity by using the Dual-Luciferase Reporter Assay system (Promega). Firefly luciferase activities were standardized to that of the Renilla luciferase transfection control. Unless otherwise indicated, the data shown are averages of triplicate samples. The interaction between overexpressed ERα and SPBP was studied in 293T cells. Transfections were performed in 10-cm-diameter culture dishes at 50 to 70% confluency by using 7.5 μg of F-ER or HEG0 and 7.5 μg of pcDNA3-HA-SPBP. The cells were stimulated with a hormone 3 h before lysis.

GST pull-down experiments.

One microgram of GST fusion proteins [GST-AF1, GST-AF1(S118E), or GST-AF1(3×E)] was bound to glutathione beads in NETN buffer (20 mM Tris-HCl [pH 8], 100 mM NaCl, 0.5 mM EDTA, 0.5% NP-40). An equimolar amount of His-tagged SPBP was added and incubated with GST fusion proteins in a total volume of 200 μl of NETN buffer for 45 min at 4°C. The beads were washed six times with NETN buffer, and then protein complexes were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis. The results were visualized by immunoblotting with anti-His and ER17 antibodies.

Coimmunoprecipitation and ChIP assays.

Cells were collected and resuspended in immunoprecipitation buffer (20 mM Tris-HCl [pH 8], 100 mM NaCl, 10% glycerol, 0.1% NP-40, 1 mM monovanadate, 1 mM DTT, and protease inhibitors [Sigma]). DNA was sheared by 7 to 8 passages through a 25-gauge needle. Lysates were cleared by centrifugation, and protein concentrations were determined by the Bradford method. One milligram of lysates was incubated with Flag resin (anti-Flag M2 agarose affinity gel [Sigma]) or antibody F3 and protein A-Sepharose for 2 h at 4°C, and the mixture was then washed three times with immunoprecipitation buffer. Proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, followed by immunoblotting with anti-HA and either the anti-Flag or ER17 antibody, respectively. For coimmunoprecipitation of endogenous ERα and SPBP, MCF7-SH cells were grown in 15-cm-diameter plates and stimulated with estradiol (E2) 3 h prior to lysis. Cells were collected in cold buffer A (10 mM HEPES [pH 7.9], 1.5 mM MgCl₂, 10 mM KCl, 0.5 mM DTT, and protease inhibitors), centrifuged, and then resuspended in buffer A with 0.1% NP-40. The suspension was left on ice for 10 min to release cytoplasmic proteins. Nuclei were pelleted and resuspended in immunoprecipitation buffer. The rest of the protocol was carried out as described above. Four milligrams of lysates was used for immunoprecipitation, which was carried out with the anti-ERα antibody F3. Immunoblots were probed with anti-SPBP and anti-ER antibodies. Chromatin immunoprecipitation (ChIP) assays were performed as described previously (49). MCF7-SH cells were grown to confluency and stimulated where indicated with E2 for 3 h. Rabbit polyclonal antiserum against SPBP and preimmune serum (as a control) were used at 1:100 dilutions. The sequences for the PCR primers (pS2 and unsp.) have been published previously (43).

RNA interference experiments.

RNA interference experiments were conducted with NIH 3T3 cells as described previously (12, 13). To produce T7 small interfering RNAs (siRNAs) targeting SPBP nucleotides 5400 to 5419 (numbered relative to the start codon), the following DNA oligonucleotides were used: sense, 5′-AAGGAGCCCCCTGGACAGGGACTATAGTGAGTCGTATTA-3′; antisense, 5′-CGGTCCCTGTCCAGGGGGCTCCTATAGTGAGTCGTATTA-3′.

Colony formation assay.

Approximately 10,000 MCF7 or SkBr3 cells were transfected with 5 μg of the indicated pcDNA3 vectors, which carry the neomycin resistance gene, and 5 μg of a vector carrying the puromycin resistance gene. Plates were fixed with methanol and stained with methylene blue for visualization after 14 days of selection with puromycin (500 ng/ml) and the neomycin analogue G418 (500 μg/ml). Colonies with 20 or more cells were scored for quantitation.

RESULTS

Phage display screening.

We undertook a screening for proteins that specifically recognize and bind the phosphorylated version of the human ERα AF1 domain. To simplify our task, we generated a phosphoserine mimic mutant of AF1, designated 3×E, in which serines 104, 106, and 118 were replaced by glutamic acid (Fig. 1). Such mutants are known to have increased transcriptional activity (1) and to be recognized by phosphorylation-specific antibodies (6), just like the phosphorylated wild-type version. Using a T7 phage display library, we screened for human proteins that specifically bind the 3×E and not the wild-type version of AF1. These recombinant T7 phages expressed fusion proteins consisting of human proteins fused to a major capsid protein on their surface. While performing pilot experiments with the GST fusion protein GST-AF1(3×E) as the “bait” protein, we noticed that the high nonspecific binding of recombinant T7 phages to affinity resins such as glutathione Sepharose and/or control proteins could be overcome by alternating the bait protein with different tags and hence different affinity resins. With this approach, the screening of the human breast cDNA library yielded a population of phages with only two types of inserts. One of these recombinant phages clearly fulfilled our initial criterion of binding only the phosphoserine mimic 3×E and not the wild-type (unphosphorylated) AF1 or the GST tag by itself (Fig. 2). This phage turned out to contain an internal fragment of the cDNA of SPBP. This is the first time that SPBP has been linked to a nuclear receptor, although it has previously been described as a DNA binding transcription factor as well as a coactivator of several other transcription factors (24, 44, 46).

FIG. 2.

Isolation of SPBP as a specific interactor with the phosphoserine mimic mutant of ERα AF1 by a T7 phage display screen. (A) Specific binding of the recombinant T7 phage expressing a portion of SPBP to the phosphoserine mimic of ERα AF1. Binding to GST alone, GST fused to wild-type (and unphosphorylated) AF1, and GST-AF1(3×E), containing the three changes S104E, S106E, and S118E, was monitored by determining the number of bound infectious phage particles by a plaque assay. (B) Schematic representation of SPBP. Previously known functional domains include an N-terminal transactivation domain, a DNA-binding domain (DBD) containing an AT-hook motif, and a C-terminal zinc finger domain (see the text). The portion of SPBP that was expressed on the surfaces of AF1(3×E)-specific T7 phage particles encompasses amino acids 1459 to 1615.

SPBP specifically binds phosphorylated AF1 in vitro.

To confirm the specificity of the interaction, we analyzed it in vitro with purified recombinant proteins. The portion of SPBP that we had isolated with the phage display screen was produced as a tagged recombinant protein in E. coli (Fig. 3A). The GST pull-down experiment for which results are shown in Fig. 3B confirmed that SPBP bound only AF1(3×E), not GST alone or unphosphorylated wild-type AF1. Moreover, it demonstrated that replacing the major MAP kinase phosphorylation site, S118, by glutamic acid was sufficient, even within the context of the entire N-terminal domain encompassing 180 amino acids. This suggested that S104 and S106 might play only minor roles, if any. Most importantly, the in vitro phosphorylation of wild-type AF1 by the MAP kinase Erk2 conferred SPBP binding activity, supporting our initial assumption that the 3×E mutant mimics the phosphorylated status of AF1.

FIG. 3.

Interaction domain of SPBP for phosphorylated AF1. (A) Schematic representation of the SPBP truncations tested. Gray boxes indicate the retained portions. The longest fragment (amino acids 1459 to 1615) corresponds to the fragment originally isolated. The sequence of the shortest interacting fragment is shown. (B) The SPBP fragment encompassing amino acids 1459 to 1615 binds to in vitro-phosphorylated AF1 (indicated by a circled P) and to phosphoserine mimic mutants of AF1 [AF1(S118E) and AF1(3×E)] but not to wild-type AF1. Following the GST pull-down experiments, SPBP and the different GST-AF1 fusion proteins were revealed by immunoblotting with antibodies directed against the His₆ tag and the ERα N terminus, respectively. Input, 10% of the amount used for the GST pull-down. (C) Further mapping of ERα binding. The experiment is the same as that for panel B, except that equal binding of the GST-AF1 proteins is visualized by Ponceau S staining of the membrane after transfer.

Relatively few types of phosphoserine binding motifs are known (60), and none of these is recognizable in SPBP (data not shown). We therefore further delimited the portion of SPBP that was able to recognize phosphorylated AF1. Progressively truncated SPBP fragments (Fig. 3A and C) were tested by GST pull-down experiments. A C-terminal degradation product of the SPBP fragment encompassing amino acids 1535 to 1615 (Fig. 3C, left panel, second lane) served as a convenient internal negative control for the GST pull-down experiment (Fig. 3C, right panel). Specific binding to the phosphoserine mimic 3×E was retained by the very C-terminal end of the initially isolated ERα interaction domain of SPBP. These 42 relatively basic (predicted pI, 11.7) amino acids, which are fully conserved between mouse and human, represent the first interaction domain to recognize the phosphorylated AF1 domain with such specificity.

SPBP interacts with activated ERα in vivo.

To ascertain that the interactions we had observed in vitro with fragments of SPBP and ERα were representative of interactions of the full-length proteins in living cells, we performed three types of experiments: immunoprecipitation experiments with tagged or untagged overexpressed proteins and with endogenous proteins and ChIP experiments. Exogenous proteins were transiently expressed in 293T cells, which also afforded the opportunity to assess the dependence of the interaction on the phosphorylation status of AF1 or on the corresponding serine mutants. Cell extracts were subjected to immunoprecipitation with an anti-Flag (Fig. 4A) or anti-ERα (Fig. 4B) antibody. In Fig. 4A, extracts from cells expressing exogenous ERα without a tag were used as a negative control. The immunoblot of Fig. 4A shows that exogenously expressed wild-type Flag-tagged ERα and SPBP interacted in an estradiol-dependent fashion. As before, in the in vitro experiment with recombinant proteins, we compared the interactions of SPBP with the serine mutants 3×E and 3×A overexpressed in vivo (Fig. 4B). Here again, interaction could be observed only with the phosphoserine mimic mutant 3×E (even upon extended exposure to compensate for the somewhat lower expression of the 3×A mutant), and as with wild-type ERα, this interaction was strictly ligand dependent and could be elicited by either E2 or the antiestrogen hydroxytamoxifen (OHT), which blocks AF2 while allowing AF1 to function. These experiments indicated that the phosphorylation of S118 (or the genetic mimicry thereof) was also a requirement for the interaction in vivo, since S118 is known to become hyperphosphorylated in the presence of hormone (1). To provide more-direct evidence for this notion, we performed a reciprocal coimmunoprecipitation experiment and probed for phosphoserine 118 with a specific antibody. As can be seen in Fig. 4C, ERα molecules that coprecipitated with the SPBP fragment (amino acids 1459 to 1615) were indeed enriched in phosphoserine 118.

FIG. 4.

Full-length SPBP and ERα interact in vivo in the presence of ligand. (A) SPBP and ERα overexpressed in 293T cells coimmunoprecipitate. 293T cells were transfected with appropriate expression vectors for expression of Flag-tagged or wild-type ERα and HA-tagged full-length SPBP; then they were either left untreated or treated with 100 nM E2 for 3 h as indicated. Following immunoprecipitation (IP) with anti-Flag resin, Flag-ERα and HA-SPBP were revealed by immunoblotting with antibodies directed against the tags. (B) SPBP interacts only with the phosphoserine mimic ER(3×E), not with ERα with alanine residues blocking phosphorylation [ER(3×A)]. 3×E and 3×A were coexpressed with HA-SPBP in 293T cells and immunoprecipitated with the anti-ERα monoclonal antibody F3. The immunoblots were developed with antibodies against HA and ERα (antibody ER17) in the top and bottom panels, respectively. (C) SPBP preferentially coprecipitates phosphorylated ERα. An HA-tagged SPBP fragment encompassing amino acids 1459 to 1615 was coexpressed with wild-type ERα. Following IP with the anti-HA antibody (9), coprecipitating total ERα (middle panel) and ERα with phosphorylated S118 (top panel) were revealed with antibody ER17 and phosphoserine 118-specific antiserum, respectively. Two percent of the total lysate used for the IP was loaded as input.

To demonstrate interaction between the endogenous proteins, we immunoprecipitated ERα complexes from an MCF7 breast cancer cell variant expressing elevated levels of endogenous ERα and revealed SPBP with a polyclonal antiserum (Fig. 5A). As expected, SPBP coprecipitated with ERα only in the presence of hormone. Next we performed a ChIP assay to assess whether endogenous SPBP was recruited to an ERα target promoter. As can be seen in Fig. 5B, more of the pS2 promoter fragment was coprecipitated with the anti-SPBP antiserum than with the preimmune serum, and association with the pS2 promoter was stimulated by hormone. (There was a basal level of SPBP at the promoter in the absence of E2, which may have been due to some SPBP being bound in an ERα-independent fashion.) In view of the phosphorylation dependence of the interaction, it seems less likely that a low level of SPBP became loosely bound to unliganded ERα that is known to cycle on and off this promoter (35, 43). Thus, endogenous SPBP was recruited to hormone-stimulated endogenous ERα, presumably through phosphoserine 118.

FIG. 5.

Interaction of endogenous SPBP and ERα. (A) Coimmunoprecipitation of endogenous ERα and SPBP. Extracts were prepared from confluent cultures of MCF7-SH cells, stimulated where indicated with E2 for 3 h. Immunoprecipitation (IP) was performed with the anti-ERα antibody F3, and SPBP was revealed by immunoblotting with the polyclonal anti-SPBP antiserum. Overexpressed HA-SPBP was run alongside as a molecular weight marker for the endogenous SPBP. (B) SPBP is recruited to the pS2 promoter upon stimulation with E2. The ChIP assay was performed with MCF7-SH cells treated as described above. IP was performed with preimmune serum (Ctrl) or with the polyclonal anti-SPBP antiserum (α-SPBP). An intergenic fragment (unsp.) was amplified by PCR as a control for the specific pS2 promoter fragment. For the quantitation shown in the graph, data from two independent experiments were averaged. They are standardized to the control band (unsp.) and to the control IP (Ctrl).

SPBP represses ERα through phosphorylation-activated AF1.

The functional significance of the interaction for ERα activity was addressed by the next series of experiments. Whereas interaction assays were easier to perform with 293T cells, HepG2 cells, known to have robust AF1 activity, seemed more appropriate for functional assays. In most cell types, both AF1 and the hormone binding domain (HBD)-associated AF2 contribute to the transcriptional activity of ERα. Because we expected SPBP to function through AF1, and in order to avoid a complicating contribution of AF2, we began by testing the effects of SPBP overexpression on ERα activity in the presence of OHT, which would activate only AF1. The phosphoserine mimic ER(3×E) has considerably higher activity in a transactivation assay than the nonphosphorylatable mutant ER(3×A) (see schemes in Fig. 1). Interestingly, SPBP specifically repressed ER(3×E), whereas ER(3×A) was resistant to this inhibitory effect and served as a negative control (Fig. 6A). In the presence of OHT, the AF1-mediated activity of ERα can be further stimulated by peptide growth factors such as epidermal growth factor (EGF) (discussed in references 4 and 40), and this EGF-stimulated activity was also repressed by SPBP (Fig. 6B). Wild-type ERα stimulated either by estradiol or by EGF in the absence of cognate ligand was repressed by overexpressed SPBP (Fig. 6C and D), although the repression was weaker in the presence of estradiol, presumably because AF2 was activated by the cognate agonist and was not directly affected by SPBP. Further support for the conclusion that SPBP inhibited ERα activity via the activated version of AF1 (see our unpublished data at our website [http://www.picard.ch]) comes from experiments showing that SPBP repressed (i) AF1(3×E) fused to the Gal4 DNA binding domain, (ii) the ERα AF1 (and DNA binding domain) fused to the HBD of the glucocorticoid receptor (GR) but not the reciprocal chimera, and (iii) ERα activated by the dioxin receptor. The dioxin receptor is known to be able to activate the unliganded ERα by association with AF1 (39). The specificity of the repression is also underlined by the fact that SPBP failed to affect ERβ (see http://www.picard.ch).

FIG. 6.

SPBP specifically represses the transcriptional activity of ERα through activated AF1. (A) SPBP represses the AF1 activity of the phosphoserine mimic mutant ER(3×E) but not that of ERα with alanine residues blocking phosphorylation [ER(3×A)]. (B) SPBP represses EGF-stimulated ERα AF1 activity. Data are averages of duplicates of representative experiments. wt, wild type. (C) SPBP inhibits E2-activated ERα. (D) SPBP represses ERα activated by EGF in the absence of estrogen. HepG2 cells were used for the experiments represented by panels A through D. (E) Depletion of endogenous SPBP from NIH 3T3 cells by RNA interference increases the activity of ER(3×E) AF1. NIH 3T3 cells were treated with serum and PDGF to induce SPBP expression. For the experiments represented by panels A, B, and E, OHT was added to block AF2, allowing only AF1 to be monitored. (F) Immunoblot confirming successful RNA interference with expression of endogenous SPBP in NIH 3T3 cells. Lanes 1 and 2, cells transfected with GFP siRNA and SPBP siRNA, respectively. (G) Immunoblots with anti-HA (top) and anti-ERα (bottom) antibodies showing that overexpression of SPBP does not affect ERα expression.

Using RNA interference, we set out to confirm the inhibitory role of the endogenous SPBP for ERα AF1 activity. For this experiment we used NIH 3T3 cells, because expression of SPBP is known to be stimulated by growth factors in these cells (46), allowing a more efficient knockdown. Figure 6E shows that cotransfection of SPBP siRNAs directed against SPBP mRNA led to a small but significant increase in ER(3×E) activity compared to that with a control siRNA. The effectiveness of the SPBP knockdown was ascertained by immunoblotting (Fig. 6F). This suggested that endogenous SPBP dampens ERα activity.

Importantly, SPBP did not affect the levels of the ERα protein (Fig. 6G), and hence, SPBP must have inhibited ERα activity at target promoters either by interfering with the access of other AF1-directed factors or by recruiting additional repressor proteins.

Interaction is not sufficient for repression.

Toward establishing the mechanism, we began a preliminary characterization of the domains of SPBP that were required for repression; we also began to explore the involvement of histone deacetylases (HDACs) and of the known nuclear receptor corepressor NCoR. Truncation mutants of SPBP were tested for their ability to repress the phosphoserine mimic ER(3×E). Whereas a mutant with a small C-terminal truncation, lacking the evolutionarily conserved C-terminal zinc finger domain, still repressed ER(3×E), a mutant lacking the C-terminal 480 amino acids, and therefore the entire ERα interaction domain (amino acids 1574 to 1615), was defective (Fig. 7). Likewise, a small fragment encompassing only the interaction domain (amino acids 1535 to 1615) was not sufficient. Further experiments will be necessary to narrow down the repressor domain(s) of SPBP, which could be located either between the N-terminal activation domain and the ERα interaction domain or just C-terminally of the latter.

FIG. 7.

The ERα interaction domain of SPBP is necessary but not sufficient for repression. (A) Schematic representation of the SPBP truncations tested. Gray boxes indicate the retained portions. All variants were made with an N-terminal HA tag. (B) Only full-length SPBP and a mutant with a small C-terminal truncation repress ER(3×E). Transactivation assays were carried out with HepG2 cells as for Fig. 6. Luciferase activities were standardized to that of the pcDNA3-HA vector control. Boxed bands above the bar graph are from an immunoblot with anti-HA antibodies to show that all SPBP variants are expressed at similar levels (but vastly different sizes!) in HepG2 cells.

A role for additional factors was supported by the following results. The inhibition of ER(3×E) by SPBP could be partially relieved by the addition of the HDAC inhibitor trichostatin A (TSA), whereas TSA had no effect on ER(3×E) in the absence of SPBP. This suggested that HDACs are involved in mediating the repression by SPBP. Moreover, coexpression of NCoR further accentuated the repression by full-length but not truncated SPBP (Fig. 8).

FIG. 8.

NCoR and HDACs contribute to repression by SPBP. (A) TSA relieves SPBP repression of ER(3×E) activity. For the transactivation assays, ER(3×E) and SPBP were coexpressed in HepG2 cells in the presence of OHT and with or without 100 ng of TSA/ml. The reporter activity obtained with ER(3×E) without TSA, expressed as the ratio of activity in the presence of SPBP to activity in the absence of SPBP (see inset), was set at 1. (B) NCoR stimulates repression of ER(3×E) activity by full-length SPBP. Luciferase activities were standardized to that of ER(3×E) alone (in the presence of OHT).

SPBP inhibits the proliferation of ERα-dependent MCF7 cells.

We hypothesized that, if SPBP represses the transcriptional activity of ERα, it should be possible to demonstrate an inhibitory effect of SPBP in a more complex physiological system. MCF7 human breast cancer cells are a powerful model for estrogen-dependent breast cancer, since they depend on the activation of ERα (for example, by hormones) for efficient proliferation. Repression by SPBP was assessed by determining colony formation upon selection for stable transformants with two different selectable markers. Compared to that of the empty vector control, cotransfection of the SPBP expression vector dramatically reduced both basal and estradiol-induced proliferation in MCF7 cells. The SPBP ERα interaction domain by itself failed to inhibit proliferation, demonstrating that the mere binding to ERα AF1 is not sufficient. The inhibitory effect of SPBP on colony formation was specific for the ERα-dependent cell line, since ERα-negative and -independent SkBr3 breast cancer cells were not affected, even though SPBP could be expressed to similar levels in these two cell lines (Fig. 9).

FIG. 9.

SPBP suppresses the proliferation of ERα-dependent breast cancer cells. Colony formation assays were performed with ERα-positive MCF7 and ERα-negative SKBr3 cells. (A) Photos of representative tissue culture plates with fixed and stained colonies. (B) Quantitation of representative colony formation assays. Cells were transfected with the indicated pcDNA3 vectors, which carry the neomycin resistance gene, and a vector carrying the puromycin resistance gene. Colonies with 20 or more cells were scored. (C) Immunoblot of SPBP transiently expressed in MCF7 and SkBr3 cells. Equal amounts of total protein were loaded and probed with an antibody to the HA tag.

DISCUSSION

We have demonstrated that SPBP acts as a repressor of activated ERα. What makes this factor unique, compared to the plethora of other ERα-interacting proteins, is that it targets the N-terminal activation function AF1 by associating exclusively with its phosphorylated form. The role of SPBP may be to attenuate the AF1-mediated ERα activity to allow only a transient activation by signaling pathways that stimulate AF1 activity by hyperphosphorylation. Cell-specific expression levels or the availability and activity of SPBP might contribute to the cell specificity of AF1 activity.

Phosphorylation-dependent interaction of SPBP with ERα AF1.

The unique effects of SPBP must be considered in light of the known effects of phosphorylation on ERα AF1 activity. Despite the well-established stimulatory role of the phosphorylation of S118, the mechanism of this stimulation is still not understood. The only protein that has been reported to be recruited more tightly to phosphorylated ERα AF1 is the p68 RNA helicase (15), but the stimulation of ERα activity is relatively weak and cell specific. Moreover, it is not strictly S118 dependent, and it is absent in HeLa cells, a cell line in which signaling cross talk and phosphorylation-stimulated AF1 activity have been established (see, for example, references 1 and 2). Whereas the functional effects on ERα of several other nuclear receptor coactivators, such as SRC1, CBP, and SRA, also depend on S118, albeit to varying degrees, the binding of these factors to AF1 does not appear to be S118 dependent (10, 14).

The specificity of the SPBP interaction with phosphorylated AF1 is perhaps most dramatically illustrated by the fact that the ERα interaction domain of SPBP interacted exclusively with the phosphorylated form of AF1 (or the phosphoserine mimic with a glutamic acid residue replacing serine 118) in the absence of any other protein. In cells, SPBP interacted exclusively with the phosphoserine mimic 3×E; did not interact with AF1 when phosphorylation was blocked by replacement of the serines with alanines, as in the 3×A mutant; and preferentially coprecipitated wild-type ERα carrying a phosphate at S118. SPBP also repressed ERα in vivo only under conditions where AF1 could be phosphorylated, for example, in the presence of estradiol and/or growth factors, or when the 3×E mutation mimicked phosphorylation. The 3×A mutant, either as full-length ERα or as an AF1 fusion to another DNA binding domain, was resistant to repression by SPBP. As a result of this requirement, SPBP associates with and inhibits ERα only after ERα has been activated, either by its cognate ligand estradiol or by ligand-independent pathways involving phosphorylation of S118. This is strongly supported by our immunoprecipitation experiments and ChIP assays, which showed a ligand dependence of the interaction of the endogenous proteins.

The only other proteins known to interact specifically with phosphorylated AF1 are two phosphatases with broad specificity, protein phosphatases 2A and 5. After docking elsewhere on ERα, they dephosphorylate S118 and thereby inhibit ERα activity (20, 29). SPBP differs fundamentally from these and other ERα repressor proteins, such as LCoR (17) and Brca1 (16), in that it binds activated ERα in a phosphorylation-dependent manner.

A novel phosphoserine binding domain?

The in vitro results indicated that the phosphorylation of S118 might be sufficient for recognition by SPBP, but further analyses are needed to provide formal proof and to dissect the contribution of the surrounding peptide sequence. The latter undoubtedly contributes, as illustrated by the fact that SPBP does not affect the AF1 activity of other nuclear receptors such as ERβ and GR (see our unpublished data at http://www.picard.ch), despite the fact that their AF1 domains are phosphorylated on serines as well (26, 55). Note that the high degree of sequence conservation among members of the nuclear receptor family does not extend to AF1. We cannot exclude the possibility that S118 phosphorylation induces a conformational change in AF1, which in turn is recognized by SPBP. However, if the 42-amino-acid interaction domain of SPBP (Fig. 3A) directly recognized and bound AF1 through phosphorylated S118, it would represent a novel class of phosphoserine binding motifs. It would be distinct from other known phosphoserine-binding domains such as the WW, WD40, FHA, and 14-3-3 domains (reviewed in reference 60). In addition, the SPBP peptide could prove to be a powerful tool for structural studies of AF1. Indeed, the N-terminal domains of both ERα and ERβ appear to be unstructured in solution and have resisted structural analyses (58, 59). The observation that the SPBP peptide bound phosphorylated AF1 with high specificity suggests that it recognizes or even induces a stable structure.

SPBP is both a repressor and an activator.

Previous reports about SPBP described it as a ubiquitously expressed DNA binding transcription factor as well as a coactivator of other transcription factors. It was first discovered as a transcription factor that binds and activates the stromelysin-1 promoter (46). Later it was found to stimulate Sp1, c-Jun, Ets-1, and Pax-6 (44).

Several functional domains, typical of transcription factors, have been mapped. A transactivation domain lies at the N terminus, whereas a minimal DNA binding domain of the AT-hook type has been localized just N-terminally of the ERα interaction domain (44). Interestingly, a somewhat larger domain encompassing this AT-hook and the ERα interaction domain (44) binds DNA as a dimer and with higher affinity. The very C-terminal PHD/LAP/ZNF2 zinc finger, contained in both the long and short splicing variants, is required for stimulating a subset of the transcription factors mentioned above(44) and for restraining the interaction of SPBP with the RING finger protein SNURF/RNF4; the latter interaction requires sequences just C-terminal of the ERα interaction domain (30).

Given that the size of SPBP is more than 1,900 amino acids, it undoubtedly contains additional functional domains, including those for repression of ERα AF1. Indeed, we have demonstrated both by transactivation and by proliferation assays that by itself the ERα interaction domain of SPBP did not suffice, despite its theoretical potential to interfere sterically with the binding of other factors to AF1. It is worth considering that such a steric interference might contribute without being sufficient. A meaningful experimental confirmation will have to await the identification of the critical positive factors that are specifically recruited to phosphorylated AF1. SPBP might use different sets of domains in a combinatorial fashion to interact with DNA and to stimulate or to repress other transcription factors. Since ERα is also known to interact with Sp1 (41), c-Jun (51), Ets-1 (52), and SNURF (47), it is conceivable that SPBP takes part in or even promotes the formation of tripartite complexes. The functional output of SPBP-containing binary and tertiary complexes will depend on the nature of yet other factors recruited to these DNA-tethered complexes. In the case of the repression of ERα by SPBP, our preliminary evidence suggests a role for NCoR and HDACs.

Physiological significance.

As a physiological correlate of our in vitro binding and in vivo transactivation experiments, we showed that the overexpression of SPBP inhibited the proliferation of an ERα-dependent breast cancer cell line. We speculate that SPBP may be part of an attenuation mechanism that fine-tunes the magnitude and/or the duration of ERα activity. Considering the opposing effects of SPBP on ERα, on the one hand, and on other growth-promoting factors such as c-Jun, on the other, one might predict that SPBP levels in breast tumors might depend on ER status. Intriguingly, such an inverse correlation between SPBP (TCF20) and ERα is apparent at the mRNA level in a recent microarray analysis of breast tumor samples (57). In addition to this categorization according to ER status, overexpression of SPBP could also distinguish between two different types of tamoxifen-resistant breast tumors. In contrast to the ERα-negative tumors mentioned above, those ERα-positive tumors that are tamoxifen resistant because of elevated AF1 activity (for example, due to increased signaling cross talk with growth factors) would be expected to have lower SPBP levels.

Another possible role for SPBP might be in contributing to the organ-specific activity pattern of ERs during the estrous cycle. Maggi and collaborators (7) have monitored ER activity with an ER reporter mouse. Surprisingly, while reporter activity in reproductive organs is synchronized with estrogen levels, ER-dependent activity in nonreproductive organs, such as the liver, is equally evident, but shifted in time. This suggests that, in nonreproductive organs, ERs are activated by estrogen-independent signaling pathways. Why ERs in reproductive tissues are resistant to the same signals is not known, but it is tempting to speculate that growth factors might activate both the ERs and SPBP expression in nonreproductive and reproductive organs, respectively. In the latter organs, SPBP might then dampen the growth factor-induced ERα activity.