Identification of Estrogen Receptor β as a SUMO-1 Target Reveals a Novel Phosphorylated Sumoylation Motif and Regulation by Glycogen Synthase Kinase 3β

Abstract

SUMO conjugation has emerged as a dynamic process in regulating protein function. Here we identify estrogen receptor β (ERβ) to be a new target of SUMO-1. ERβ SUMO-1 modification occurs on a unique nonconsensus sumoylation motif which becomes fully competent upon phosphorylation of its contained serine residue, which provides the essential negative charge for sumoylation. This process is further regulated by phosphorylation of additional adjacent serine residues by glycogen synthase kinase 3β (GSK3β), which maximizes ERβ sumoylation in response to hormone. SUMO-1 attachment prevents ERβ degradation by competing with ubiquitin at the same acceptor site and dictates ERβ transcriptional inhibition by altering estrogen-responsive target promoter occupancy and gene expression in breast cancer cells. These findings uncovered a novel phosphorylated sumoylation motif (pSuM), which consists of the sequence ψKXS (where ψ represents a large hydrophobic residue) and which is connected to a GSK3-activated extension that functions as a SUMO enhancer. This extended pSuM offers a valuable signature to predict SUMO substrates under protein kinase regulation.

INTRODUCTION

Sumoylation is a highly dynamic posttranslational process that consists of conjugation of the small ubiquitin-like modifier SUMO on target proteins (20). SUMO modification is involved in diverse aspects of protein function which are often linked to nuclear activities, such as DNA replication, genome stability, nuclear transport, and gene transcription (41). Despite a low and transient stoichiometric proportion of SUMO-modified proteins, an increasing number of substrates have been characterized mostly on the basis of the presence of a predicted minimal core SUMO consensus motif, ψKXE/D (where ψ represents a large hydrophobic residue), in which the lysine serves as the acceptor site to covalently link SUMO (35). The sequential enzymatic events leading to protein sumoylation closely resemble those of ubiquitination (13). The core SUMO motif is recognized by the unique 17β-estradiol (E2)-conjugating enzyme Ubc9, which upon transfer from E1-activating enzyme 1 (SAE1)/SAE2 conjugates SUMO onto a specific lysine residue. Although Ubc9 is sufficient to promote sumoylation, three classes of E3 ligases that consist of the nucleoporin RanBP2, the polycomb repressor Pc2, and the PIAS ligase family members have been described to facilitate the conjugation process. Sumoylation is reversible, as the SUMO tags can rapidly be cleaved from target proteins by the SENP family of sentrin-specific isopeptidases, which ensure the dynamic and appropriate maintenance of SUMO substrates (24). Recently, two different extensions following the ψKXE/D motif have been found to enhance substrate sumoylation for some targets: the phosphorylation-dependent sumoylation motif (PDSM), which consists of the sequence ψKXEXXpSP (21), and the negatively charged amino acid-dependent sumoylation motif (NDSM), which is characterized by a cluster of acidic residues (51). As both motifs share a feature of negatively charged residues to participate in SUMO modification, an important role for specific kinase signaling pathways becomes an attractive regulatory mechanism for sumoylation.

Estrogens play a pivotal role in reproductive physiology through direct interaction with estrogen receptor α (ERα) and ERβ, which belong to the nuclear hormone receptor family of ligand-activated transcription factors. However, disruption of ER transcriptional regulation is also associated with pathological events such as breast and endometrium cancers. While ERα is considered a strong predictive factor in endocrine therapy of reproductive cancers, the clinical value of ERβ is still debated, although recent evidence has associated ERβ with antitumorigenic properties and a favorable outcome in specific contexts (12, 18, 26, 39). Whereas the two receptors share obvious similarities in terms of structure, response to hormone, and overlapping target gene sets, subtype-specific effects on gene transcription, cell-based response, dimeric context, and kinase-dependent regulation have revealed that ERα and ERβ can exert distinct functions as well.

Transcriptional activation of ERα and ERβ by estrogen involves a multistep process consisting of ligand binding, dimerization, interaction with their cognate DNA estrogen response element, and combinatory recruitment of transcriptional coregulators. Such complex assembly is achieved in part by numerous posttranslational modifications that have been ascribed to the various components, providing extensive layers of specificity. For ERs, both ligand-induced and ligand-independent activation can be modulated by phosphorylation, in support of a significant role of kinase transduction pathways in differential signaling of ERα and ERβ (28, 39, 50). More precisely, a cluster of phosphorylation sites has been identified to regulate activation function 1 (AF-1) activity of ERβ and specific recruitment of coactivators SRC-1 and CBP and of ubiquitin ligase E6-AP, highlighting the key role of this modification in regulation of receptor activity and degradation (31, 36, 45). Sumoylation has recently been added as a new regulator in the fine-tuning of nuclear receptors and associated coregulator activities (48). Most nuclear receptors that have been reported to undergo sumoylation were shown to be inhibited by this process, through either an active repression, such as for androgen receptor (AR), progesterone receptor (PgR), and estrogen-related receptor α/γ (ERRα/γ) (1, 33, 46, 49), or a signal- and promoter-specific transrepression, such as for peroxisome proliferator-activated receptor γ (PPARγ) and liver X receptor α/β (LXRα/β) (14). SUMO modification was also reported for ERα in response to hormone but was linked to transcriptional activation of the receptor (42), providing a potential therapeutic perspective in ER-regulated breast carcinogenesis (23).

This study identifies ERβ to be a target of sumoylation through an unexpected SUMO-1 conjugation to a nonconsensus SUMO-accepting motif in which the usual aspartic acid residue flanking the acceptor lysine is replaced by a phosphorylated serine. This novel motif of SUMO target recognition, which we termed pSuM, for phosphorylated sumoylation motif, regulates various aspects of ERβ function, including receptor stability, hormonal transcriptional control, and chromatin binding.

MATERIALS AND METHODS

Plasmids.

pCMX expression plasmids coding for mouse and human full-length ERβ, ERβ without the first 64 amino acids (aa) (Δ1-64), N-terminally truncated CDEF (aa 215 to 530) ERβ, and hemagglutinin (HA)-tagged ubiquitin have been described previously (31, 38, 45). The human ERβ constructs with Lys-4 mutation to arginine, Ser-6 mutation to alanine, aspartic acid, or glutamic acid, Ser-8 and Ser-12 mutation to alanine, and the corresponding double mutations were all obtained by PCR mutagenesis using Pwo DNA polymerase (Roche) and verified by automated sequencing. Plasmids for green fluorescent protein (GFP)-tagged SUMO-1 (29), wild-type and catalytically inactive C603S SENP1 (3), and Ubc9 and its inactive C93S mutant (46) have been described. An untagged SUMO-1 construct was obtained by removing the GFP tag in pcDNA3.1 (Invitrogen).

Cell culture and transfection.

Human embryonic kidney 293 cells (293 cells) were routinely maintained in Dulbecco's modified Eagle's medium (DMEM; Sigma) supplemented with 5% fetal bovine serum (FBS). Human breast cancer Hs578t and MCF-7 cells were cultured in DMEM containing 10% FBS. For transient transfection, cells were seeded 16 to 20 h prior to transfection in phenol red-free DMEM supplemented with charcoal dextran-treated serum. Plasmid constructs were introduced into cells using the calcium phosphate precipitation method as described previously (31, 43). Treatments were added in fresh phenol red-free medium in the absence of serum for the time and with the concentration indicated previously, unless otherwise stated.

Luciferase assay.

Luciferase assays were done as described previously using an EREbLuc reporter (31). Cells were treated for 18 to 20 h with vehicle or 10 nM 17β-estradiol (E2), except in the dose-response curve analysis, in which increasing concentrations (10⁻¹⁶ to 10⁻⁵ M) were used. Data were statistically analyzed with GraphPad Prism (version 4) software. Values are normalized to the β-galactosidase activity and expressed as relative luciferase units (RLU) or fold response derived from at least three independent experiments performed in triplicate.

Cell lysates and immunoblotting.

Determination of ERβ cellular content was done by Western blot analysis essentially as described previously (31). Cells were lysed in phosphate-buffered saline (PBS) containing 1% Triton X-100, 0.5% deoxycholate acid, 0.1% SDS, 1 mM sodium orthovanadate, 1 mM sodium fluoride, 1 mM phenylmethylsulfonyl fluoride, and protease inhibitor cocktail (Roche) and subjected to Western blot analysis using an anti-ERβ antibody (Santa Cruz). Total protein loading was normalized using an anti-β-actin antibody (Abcam). Total endogenous glycogen synthase kinase 3β (GSK3β) and phosphorylated GSK3β were determined with anti-GSK3β and anti-phospho-GSK3β (Ser-9) antibodies (Cell Signaling), respectively.

In vitro sumoylation assay.

[³⁵S]Met-labeled ERβ wild type and variants were produced using the TNT T7-coupled reticulocyte lysate system (Promega) and subjected to in vitro sumoylation reaction. Typically, ERβ proteins were incubated at 37°C for 1 h with 150 ng of E1 enzyme (SAE1/SAE2), 1 μg of E2 enzyme (UbcH9), and 1 μg of SUMO-1 (Boston Biochem) in 50 mM Tris, pH 7.5, containing 5 mM MgCl₂ and 2.5 mM ATP. The reactions were stopped by adding an equal volume of Laemmli buffer, and the mixture was boiled. For in vitro phosphorylation, labeled ERβ and variants were incubated with Erk2 or with GSK3β in the presence of 0.2 mM ATP, according to the manufacturer (New England BioLabs). Samples were separated by SDS-PAGE and analyzed by fluorography.

Cellular sumoylation assay.

293 cells were transfected with the specified ERβ constructs along with GFP–SUMO-1 in the presence of Ubc9 when indicated. Dominant-negative variants of SUMO-1 (SUMO1ΔGG) and Ubc9 (C93S) were also used to inhibit sumoylation, as well as SENP1 and its inactive C603S mutated form. At 24 h after transfection, cells were treated with vehicle or 10 nM estradiol for 5 to 18 h in fresh medium and then harvested in 50 mM Tris, pH 7.5, containing 100 mM NaCl, 1% Triton X-100, 0.8% SDS, 5 mM EDTA, 20 mM N-ethylmaleimide, and protease/phosphatase inhibitors (Roche). Extracts were immunoprecipitated with an anti-ERβ antibody (Santa Cruz Biotech), and Western blot analysis was performed using antibodies against either GFP (Roche) or ERβ. Cells were also treated with protein kinase inhibitors for Mek (50 μM PD98059; BioMol), p38 (10 μM SB203580; BioMol), Jnk (10 μM SP600125; BioMol), Src (1 μM PP2; Calbiochem), protein kinase A (PKA; 100 μM H89; Calbiochem), phosphatidylinositol 3-kinase (PI3K; (10 nM LY294002; Alexis Biochemicals), and GSK3 [4 μM (2′Z,3′E)-6-bromoindirubin-3′-oxime (BIO); Alexis Biochemicals] to address their respective roles in ERβ sumoylation.

Ubiquitination assay.

The detection of ubiquitinated forms of ERβ was performed essentially as described previously (31). 293 cells were transfected with wild type or ERβ mutants along with HA-tagged ubiquitin with or without SUMO-1 plasmids. Cells were then treated with vehicle, 1 μM MG132 (Sigma), and/or 10 nM 17β-estradiol for 16 h and harvested for immunoprecipitation using an anti-ERβ antibody (Santa Cruz). Extracts were analyzed with an anti-HA antibody (12CA5), and input levels of ERβ were normalized.

Cycloheximide chase.

293 cells were transfected with wild-type and mutated ERβ. At 24 h after transfection, 50 μM cycloheximide (Sigma) was added in fresh medium and cells were treated with or without 10 nM 17β-estradiol. Cells were then harvested at the indicated time points, and ERβ steady-state levels were analyzed as previously described (31). Each result was derived from at least three separate experiments and expressed relative to β-actin levels.

Subcellular fractionation.

293 cells were transfected with ERβ wild type or mutants with site-specific mutations as specified, and at 24 h after transfection, cells were treated with 10 nM 17β-estradiol for 16 h in the presence of 1 μM MG132. Cells were washed in ice-cold PBS, and the cytoplasmic fraction (C) was obtained by repeated freeze-thaw cycles in 10 mM HEPES, pH 7.9, 1.5 mM MgCl₂, 10 mM KCl, 0.5 mM dithiothreitol, 20 mM N-ethylmaleimide, 20 U/ml RNasin in the presence of protease and phosphatase inhibitors (Roche). The nucleoplasmic fraction (N) was obtained by centrifugation, and nuclei were resuspended in 10 mM PIPES [piperazine-N,N′-bis(2-ethanesulfonic acid), pH 6.8], containing 100 mM NaCl, 300 mM sucrose, 3 mM MgCl₂, 0.5% Triton, 20 mM N-ethylmaleimide, 20 U/ml RNasin, and protease/phosphatase inhibitors. The soluble fraction (S) was obtained after centrifugation, and the chromatin-bound (CB) proteins were removed by extraction in freshly prepared digestion buffer containing 10 mM PIPES, pH 6.8, 1 mM CaCl₂, 50 mM NaCl, 300 mM sucrose, 3 mM MgCl₂, 0.5% Triton, 20 mM N-ethylmaleimide, 20 U/ml RNasin in the presence of 200 U/ml DNase I (Ambion). DNA digestion was terminated by adding 250 mM ammonium sulfate. The matrix fraction (M) was obtained by resuspending the final pellet in Laemmli buffer. Proteins of each fraction were resolved by SDS-PAGE and analyzed by Western blotting using an anti-ERβ antibody (Santa Cruz Biotech). The following antibodies of selective markers of each subcellular compartment were used to ensure the purity of the prepared fractions: anti-GAPDH (anti-glyceraldehyde-3-phosphate dehydrogenase; Santa Cruz Biotech), anti-lamin A/C (Cell Signaling), antinucleolin (Enzo Life Sciences), anti-DNA PolII (Santa Cruz Biotech), and anti-cytokeratin 5/8 (Santa Cruz Biotech).

RNA isolation and real-time PCR.

ER-negative Hs578t cells were transfected with ERβ wild type or variants as specified in the presence or absence of pcDNA-SUMO1. After 36 h of transfection, cells were treated with vehicle or 10 nM 17β-estradiol for 6 h in fresh medium. cDNA was prepared as described previously (34), and PCR amplification was done in a volume of 20 μl with 0.5 to 1 μl of the reverse transcription reaction mixture for 25 to 35 cycles (Fermentas). PCR products were analyzed on a StepOnePlus cycler (Applied Biosystems) and on gel (Alpha Innotech). Sequences of PCR primers are available on request. Values are derived from at least two separate experiments performed in triplicate and normalized to ribosomal protein RPLP0/36B4 expression.

ChIP assay.

Chromatin immunoprecipitation (ChIP) assays were done as previously described (2, 40). Hs578t cells were transfected with wild-type ERβ and mutants of ERβ in the absence or presence of SUMO-1 and compared to mock-transfected cells. At 24 h after transfection, cells were treated with 10 nM 17β-estradiol for 40 min and then harvested. Primer pairs were designed to encompass proximal and distal estrogen response element (ERE) sites of estrogen-responsive promoters (40).

RNA interference.

Lentiviral knockdown of GSK3β was performed essentially as described previously (37) and monitored by Western blot analysis. A short hairpin RNA (shRNA) targeting luciferase (shLuc) was used as a negative control.

RESULTS

ERβ is modified by SUMO-1.

To determine whether ERβ is a substrate for SUMO, in vitro sumoylation was carried out on ERβ in the presence of ATP and purified components of the SUMO pathway, such as the SUMO E1-activating (SAE1/2) and E2-conjugating (Ubc9) enzymes. As shown in Fig. 1A, addition of all constituents revealed a specific higher-molecular-weight form, indicating that ERβ was sumoylated. We then tested the sumoylation of ERβ in cells by transfecting human embryonic kidney 293 cells with ERβ and GFP-tagged SUMO-1 plasmids. Western blot analysis of immunoprecipitated ERβ using anti-ERβ and anti-GFP antibodies, respectively, showed a single higher-molecular-weight band whose intensity was clearly enhanced by the presence of estrogen (Fig. 1B). The molecular weight of this band corresponds to the predicted shift induced by the covalent linkage of GFP-SUMO-1 to ERβ, suggesting that ERβ is sumoylated at a unique site. To further address the specificity of ERβ sumoylation, addition of the E2 SUMO conjugase Ubc9 potently enhanced the sumoylation of ERβ in cells, as opposed to that achieved with its catalytically inactive C93S mutant (Fig. 1C). Conversely, nuclear desumoylase SENP1 strongly abolished ERβ sumoylation, whereas the inactive SENP1 C603S mutant was inefficient (Fig. 1D). These results establish ERβ to be a specific SUMO-1 substrate under hormonal regulation.

Fig 1.

ERβ is SUMO-1 modified. (A) In vitro sumoylation of ERβ. [³⁵S]methionine-labeled ERβ was incubated in the presence or absence of purified human SAE1/2 and SUMO-1 proteins and subjected to in vitro sumoylation with Ubc9. Sumoylation was analyzed by SDS-PAGE and fluorography. (B) In vivo sumoylation of ERβ. 293 cells were transfected with ERβ and GFP–SUMO-1 expression plasmids, as indicated, and then treated with 10 nM estradiol or left untreated for 18 h. Immunoprecipitation (IP) was carried out with an anti-ERβ antibody, and bound proteins were analyzed by Western immunoblotting (IB) using a second anti-ERβ (top) or an anti-GFP (bottom) antibody. K, molecular weight (in thousands). (C and D) Immunoprecipitation assays as described for panel B, except that plasmids for wild-type Ubc9 or the ligase-deficient Ubc9 C93S mutant (C) or for wild-type SENP1 or the catalytically inactive SENP1 C603A mutant (D) were added in the transfections. In each case, cells were treated with 10 nM estradiol for 18 h prior to Western blot analysis with an anti-ERβ antibody. Input amounts of SUMO-1, Ubc9, and SENP1 determined by Western blot analysis are also indicated. Asterisks denote nonspecific signals.

A novel nonconsensus SUMO recognition motif dictates ERβ sumoylation at Lys-4.

We next wanted to identify the exact sumoylation site on ERβ, but inspection of the sequence of ERβ from mouse or human has not revealed any site corresponding to the core SUMO consensus ψKXE/D motif. In order to determine which region confers sumoylation potential, we thus performed a sumoylation assay in cells expressing N-terminal truncated forms of ERβ (Fig. 2A). While mouse full-length ERβ was efficiently sumoylated in the presence of estrogen, removal of the entire N-terminal AF-1 region (CDEFβ construct) completely abolished ERβ sumoylation, a result also observed with the human CDEFβ construct (see Fig. S1 in the supplemental material). This indicates that the AF-1-containing region is required for ERβ sumoylation. Removal of the first 64 amino acids of mouse ERβ (Δ1-64 construct) further established the proximal N-terminal region as being targeted by SUMO-1 (Fig. 2B).

Fig 2.

Identification of a novel nonconsensus SUMO conjugation motif in ERβ. (A) Schematic representation of full-length (549 aa), N-terminal truncated (Δ1-64), and CDEF (aa 164 to 549) forms of mouse ERβ. Also shown are transcriptional activation functions AF-1 and AF-2. (B) 293 cells were transfected with plasmids coding for HA-tagged mouse full-length ERβ and a truncated form of ERβ (Δ1-64) in the presence or absence of GFP–SUMO-1 plasmid. Cells were treated with 10 nM estradiol or vehicle for 18 h and harvested for immunoprecipitation with an anti-ERβ antibody. Extracts were analyzed for sumoylation by Western immunoblotting using an anti-HA antibody. (C) Sequence alignment for predicted SUMO motif located near the N-terminal end of mouse and human ERβ. The predicted SUMO site differs from the consensus SUMO motif and from the extended phosphorylation-dependent SUMO motif (PDSM) and the negatively charged amino acid-dependent SUMO motif (NDSM) with the replacement of the conserved aspartic acid residue by a serine residue (Ser-6 in human ERβ). (D) Estrogen promotes SUMO-1 attachment to Lys-4 of human ERβ and requires Ser-6. ERβ sumoylation performed with the wild type and mutants with various point mutations of human ERβ expressed in 293 cells. Immunoprecipitation and Western blot analysis were done as described for panel B. Immunoprecipitated samples were also analyzed with an anti-GFP antibody. (E) Estrogen-dependent sumoylation of ERβ variants compared to wild-type receptor. Sumoylation was carried out as described for panel D. Sumoylation was also assessed using an anti-GFP antibody (bottom). (F) In vitro sumoylation of [³⁵S]methionine-labeled ERβ and variants incubated with purified Ubc9 and SUMO-1 in the presence or absence of SAE1/2. Sumoylation was analyzed by SDS-PAGE and fluorography. (G) Sumoylation of ERβ wild type and Ser-6 mutants in response to RasV12 expression. Sumoylation was performed as described for panel D. (H) In vitro sumoylation of ERβ and S6A variant carried out as described for panel E, except that both proteins were also subjected to in vitro phosphorylation with purified Erk2 prior to sumoylation. Asterisks denote nonspecific signals.

When comparing the sequence alignment of ERβ isoforms in this region (Fig. 2C), we found a putative nonconsensus SUMO attachment site corresponding to the sequence IKNS, which bears a serine residue instead of the usual glutamic acid residue of the SUMO ψKXE consensus motif. This raises the striking possibility that phosphorylation of Ser-6 could confer the necessary negative charge for SUMO modification to occur. If true, then Lys-4 (Lys-23 in mouse ERβ) in the IKNS sequence would serve as a SUMO attachment site. To test this hypothesis, a K4R mutant in which Lys-4 was mutated to arginine in human ERβ was assessed for sumoylation, and as predicted, the estrogen-dependent sumoylation signal obtained for wild-type ERβ was lost when using the K4R mutant (Fig. 2D, lanes 3 and 4, and E). To assess whether Lys-4 is a direct target of SUMO-1 attachment, in vitro sumoylation was performed, which demonstrated a strongly impaired sumoylation of the K4R mutant compared to the wild-type receptor (Fig. 2F). These results identify Lys-4 to be a SUMO accepting site in ERβ.

Interestingly, mutation of Ser-6 to alanine (S6A) without affecting Lys-4 also prevented ERβ sumoylation in cells (Fig. 2D, lane 5, and E) and in vitro (Fig. 2F), indicating that Ser-6 is essential for ERβ sumoylation and further suggesting that phosphorylation of this residue might be required. This hypothesis was supported by adding a negative charge at this position with the substitution of Ser-6 with an aspartic acid (S6D) or glutamic acid (S6E) residue, which contributed to strongly enhance SUMO-1 attachment compared to wild-type ERβ (Fig. 2D, lanes 7 and 9, and E and F). The S6E mutant was shown to be more effective since this modification creates a perfect SUMO ψKXE motif which is functional in sumoylating ERβ. When Lys-4 was mutated in the context of Ser-6 substitution to acidic residues (the K4R and S6D and the K4R and S6E double mutants), the ERβ sumoylation signal was lost, indicating again the essential role of Lys-4 as a SUMO acceptor site (Fig. 2D and F). Furthermore, the S6D and S6E mutations did not promote sumoylation to other sites since no additional SUMO-modified signals were observed under these conditions. While sumoylation of S6E was enhanced by estrogen, it was also observed in the absence of hormone, demonstrating the potential of Ser-6 to mediate ligand-independent ERβ sumoylation (Fig. 2E). However, attempts to generate a phosphorylation-specific antibody to directly support Ser-6 phosphorylation were unsuccessful due to poor epitope immunogenicity and a lack of selectivity with surrounding serine residues. We then addressed the SUMO recruiting signal that would be responsible for ERβ Ser-6 phosphorylation. The effect of constitutive Ras activation on ERβ sumoylation was determined on the basis of the fact that Ser-6 is contained within a consensus sequence for proline-directed kinases, such as the mitogen-activated protein kinase (MAPK)/Erk family members. As predicted, expression of the constitutive RasV12 mutant markedly enhanced ERβ sumoylation, whereas mutants with Ser-6 substitutions remained unresponsive (Fig. 2G). In addition, Ser-6 was also involved in promoting sumoylation of ERβ in vitro in response to Erk-mediated phosphorylation (Fig. 2H). These results indicate a contributing role of the MAPK pathway in promoting ERβ sumoylation through Ser-6. Therefore, the SUMO-1-recruiting IKNS sequence represents a nonconsensus SUMO site that appears to function through phosphorylation to then mimic a classic consensus SUMO motif. In that respect, this site is different from the reported phosphorylation-dependent SUMO motif PDSM and from the negatively charged amino acid-dependent SUMO motif NDSM (Fig. 2C) (21, 51). We therefore termed this newly identified sumoylation site pSuM, for phosphorylated sumoylation motif.

Sumoylation of pSuM represses ERβ transcriptional activity.

We next investigated whether ERβ transcriptional potential was modulated by SUMO modification. Using reporter assays with a luciferase gene under the control of an estrogen-responsive element (EREbLuc), we observed that sumoylation-deficient K4R (26-fold) and S6A (13-fold) ERβ mutants displayed a greater transcriptional response to estrogen than wild-type receptor (8.5-fold), suggesting that SUMO inhibits ERβ activity in the context of hormone (Fig. 3A). In line with such an interpretation, the SUMO-enhancing S6E mutation clearly impaired the response to estrogen (2.5-fold), while the response of the S6D mutant (7.5-fold) was less affected compared to the one of wild-type ERβ (Fig. 3A), consistent with their respective levels of sumoylation (Fig. 2G). Thus, the extent to which these mutants respond to estrogen inversely correlates with their ability to undergo sumoylation. This is especially emphasized with the SUMO-defective K4R mutant, in which the inhibitory potential of the S6E mutation could not abrogate its increased capacity to respond to estrogen (Fig. 3B). These results indicate that SUMO conjugation to Lys-4 results in ERβ transcriptional repression.

Fig 3.

ERβ transcriptional potential is reduced by sumoylation. (A) Transcriptional activity of wild type and mutants with site-specific mutations of ERβ in response to estrogen and SUMO-1. 293 cells were transfected with an EREbLuc reporter gene along with each ERβ construct (100 ng) and increasing amounts of SUMO-1 (0, 25, 50, 100, and 200 ng). Cells were then treated with 10 nM estradiol for 18 h and harvested for luciferase activity measurement. (B) Reporter gene assay performed as described for panel A for the indicated ERβ mutants with point mutations, except that SUMO-1 was omitted. Luciferase values were normalized to β-galactosidase activity and expressed as a percentage of the estradiol response of wild-type receptor, which was set at 100%. (C) ERβ-specific inhibition of progesterone receptor (PgR) target gene expression by SUMO-1 requires Lys-4. Real-time PCR analysis of PgR expression in ER-negative breast cancer Hs578t cells expressing ERβ wild type or mutants with site-specific mutations. Cells were transfected or not with the SUMO-1 construct and treated with 10 nM estradiol for 6 h. Results were normalized to RPLP0 expression.

The inhibitory role of SUMO on ERβ activity was further addressed in cells transfected with increasing amounts of SUMO-1, which caused a dose-dependent reduction in the ERβ response to estrogen of nearly 3-fold compared to that for cells not transfected with SUMO-1 (Fig. 3A). Such an inhibitory role of SUMO-1 on ERβ activity was severely impaired using the SUMO-defective K4R and S6A mutants, which showed a weak but not significant reduction in activity. In the same line, the SUMO-competent S6D and S6E mutants still retained the inhibitory response to SUMO-1 expression in this context (Fig. 3A). This is consistent with a selective role of pSuM in the inhibition of ERβ by SUMO-1.

To establish whether the integrity of ERβ pSuM has an impact on estrogen-responsive gene expression, we performed quantitative reverse transcription-PCR analysis on human breast carcinoma Hs578t cells expressing ERβ wild type or SUMO-related mutants. Stable expression of ERβ in Hs578t cells conferred estrogen responsiveness on known ER target genes, such as progesterone receptor (PgR), trefoil factor TFF1, and cytochrome c oxidase Cox7A2L (38). Indeed, expression of wild-type ERβ contributed to enhancement of the expression of these target genes, with values ranging from 2- to 2.5-fold in response to estrogen, as opposed to expression in parental ERβ-negative Hs578t cells, which remained unresponsive (Fig. 3C; see Fig. S2 in the supplemental material). Interestingly, greater responses to estrogen were achieved in the SUMO-defective K4R-expressing cells, supporting the negative role of sumoylation on ERβ activity. Consistent with this, while addition of SUMO-1 strongly abolished wild-type ERβ regulation of PgR, TFF1, and Cox7A2L genes by estrogen, it had no significant effect on ERβ K4R activation (Fig. 3C; see Fig. S2 in the supplemental material). In addition, SUMO-competent S6E expression exhibited an impaired gene response to estrogen compared to wild-type ERβ, consistent with the poor activity of the S6E mutant (Fig. 3A). These results support a functional role of pSuM sumoylation in downregulating ERβ transcriptional activity and hormone-responsive gene expression.

Modification of Lys-4 by SUMO regulates ERβ degradation.

ERβ transcriptional competence was reported to be closely associated with its proteasome-directed degradation, supporting a means by which target cells can limit their response to hormonal activating signals (31). Given the regulation of ERβ activity by SUMO-1, we thus addressed the role of sumoylation on ERβ stability. Figure 4A shows that the steady-state levels of ERβ in the presence of hormone were increased with the SUMO-inducing S6D and S6E mutations, whereas the S6A mutation had no significant effect, suggesting that ERβ stability can be enhanced by sumoylation. More strikingly, disruption of the SUMO Lys-4 acceptor site (K4R) alone or in combination with the S6D and S6E mutations contributed to a further increase in ERβ accumulation, as if the absence of SUMO-1 was also stabilizing ERβ. In an attempt to better understand this intriguing relationship, increasing SUMO-1 expression caused the accumulation of wild-type ERβ, an effect not observed with the K4R mutant (Fig. 4B). Similarly, the SUMO-competent S6D and S6E forms of ERβ also accumulated in response to SUMO-1 expression, as opposed to the S6A mutant, which did not again indicating that sumoylation of ERβ might confer a stabilizing effect. The fact that K4R levels remained unresponsive to SUMO-1 also agrees with such an interpretation, and the enhanced accumulation of K4R over that of the wild type seen in the absence or presence of hormone (Fig. 4A and B) might depend on receptor turnover. As such, cycloheximide chase experiments revealed that disruption of Lys-4 almost completely abrogated ERβ degradation in the absence of estrogen compared to the results for the wild type (half-life [t_1/2] = 9 h) (Fig. 4C). However, as for wild-type ERβ (t_1/2 = 6 h), the addition of estrogen increased the ERβ K4R turnover rate (t_1/2 > 10 h), indicating that this mutant still exhibited hormone-dependent degradation. These results imply that while sumoylation of Lys-4 appears to protect ERβ from degradation, targeting of Lys-4 might not represent the unique mechanism involved in the regulation of ERβ degradation by hormone.

Fig 4.

SUMO-1 modulates ERβ stability. (A) Steady-state levels of wild-type (WT) ERβ and pSuM mutants of ERβ with site-specific mutations. Extracts of 293 cells transfected with the various ERβ constructs and treated with 10 nM estradiol for 18 h were subjected to Western blot analysis with an anti-ERβ antibody. Protein loading was monitored with β-actin content. (B) Western blot analysis as described for panel A, except that increasing amounts of SUMO-1 (0-, 0.5-, 1.0-, and 2.0-fold relative to each ERβ construct) were added in the transfections. β-Actin was used as a loading control. (C) The K4R mutation confers stability to ERβ. (Top) Cycloheximide chase experiment using 293 cells expressing ERβ wild type or K4R mutant. Cells were treated with 50 μM cycloheximide and lysed at the indicated times for Western blot analysis using anti-ERβ antibody. β-Actin was used as a loading control. (Bottom) Quantification of signal intensity derived from three separate chase experiments. Results are normalized to β-actin content and expressed as the percent change from time zero, which was set at 100%.

SUMO-1 competes with ubiquitin at Lys-4 of the core pSuM sequence of ERβ.

Based on the apparent ability of sumoylation to protect ERβ from degradation through modification of the pSuM IKNS sequence and the known role of ubiquitin chain addition to lysine residues to conduct protein for degradation by the 26S proteasome, we next tested the interesting possibility that Lys-4 might also serve as a target for ERβ ubiquitination. As expected, ERβ ubiquitination was increased in cells transfected with ubiquitin and treated with estrogen and MG132, a commonly used 26S proteasome inhibitor (Fig. 5A). However, substitution of Lys-4 under these conditions markedly reduced the ubiquitination of ERβ, revealing that Lys-4 is an important regulatory site for ERβ ubiquitination. Regarding the apparent shared regulation of Lys-4 in ERβ ubiquitination and sumoylation, we then tested whether ERβ ubiquitination was affected by SUMO modification at Lys-4. Interestingly, SUMO-1 expression induced a marked decrease in ERβ ubiquitination, an effect that required SUMO-1 attachment, since a conjugation-defective SUMO1ΔGG mutant had a minor effect (Fig. 5B). In addition, SUMO-1 contributed to a decrease in the polyubiquitination of the SUMO-responsive S6D and S6E mutants, while no inhibition was observed for the SUMO-deficient S6A mutant (Fig. 5C; see Fig. S3 in the supplemental material). These results identify the pSuM sequence to be a convergent motif of regulation for both ERβ sumoylation and ubiquitination processes.

Fig 5.

SUMO-1 competes with ubiquitin at Lys-4 of ERβ pSuM. (A) ERβ Lys-4 is a target of ubiquitin. Ubiquitination assay of 293 cells transfected with ERβ wild type or K4R mutant in the absence or presence of ubiquitin. Cells were treated with 1 μM MG132 for 16 h, and extracts were subjected to immunoprecipitation with anti-ERβ antibody. Western blot analysis was performed using a different anti-ERβ antibody. (B and C) SUMO-1 competes with ubiquitin at Lys-4 of ERβ. Cells were transfected with wild type (B) or pSuM mutants (C) of ERβ in the presence or absence of ubiquitin and SUMO-1 constructs. A nonconjugatable SUMO-1ΔGG mutant was also used to inhibit sumoylation. Cell extracts were processed as described for panel A.

Sumoylation alters ERβ interaction with chromatin.

Our results depict Lys-4 to be an important site in regulating ERβ modifications and activity, but it remains uncertain how SUMO-1 can repress ERβ transcriptional activity by hormone while promoting its accumulation through competition with receptor ubiquitination. To further address the role of Lys-4 in the ERβ response to hormone and when ERβ sumoylation is impaired, we determined whether SUMO-deficient mutations affected ERβ sensitivity to hormone. In a luciferase assay, the K4R and S6A SUMO-deficient mutants were more responsive to increasing concentrations of estrogen than wild-type receptor, with greater activation found for the K4R mutant (Fig. 6A), similar to the results shown in Fig. 3A. When these data were plotted using a nonlinear regression transformation to measure estrogen sensitivity, we observed that disruption of Lys-4 or Ser-6 had no effect on the extent of the ERβ response to hormone, with a 50% effective concentration of 0.05 nM obtained in each case (Fig. 6B), showing that when ERβ sumoylation is impaired, its sensitivity to estrogen is not altered.

Fig 6.

Sumoylation modulates ERβ chromatin interaction without altering hormone sensitivity. (A and B) pSuM regulates the capacity of ERβ to respond to hormone but not its sensitivity. 293 cells expressing ERβ wild type or pSuM mutants K4R and S6A were cotransfected with an ERE-bLuc reporter gene and treated with increasing concentrations of estradiol for 18 h. Cells were then harvested for luciferase activity measurement, and the results were normalized to β-galactosidase activity. Average values derived from at least three separate experiments are indicated as relative luciferase units (RLU) for each ERβ construct (A) and as a percentage of the maximal response of each variant to estrogen (B) using a nonlinear regression curve-fitting transformation. (C and D) SUMO-1 affects nuclear ERβ content. Cells were transfected with the various ERβ constructs with or without SUMO-1, and protein extracts were separated into cytoplasmic (C) and nuclear (N) fractions and validated by Western blot analysis for respective GAPDH and lamin A markers (C). Both C and N fractions were also normalized to their respective marker and analyzed by Western blotting with anti-ERβ antibody (D). Prior to harvest, cells were treated with 1 μM MG132 and 10 nM estradiol for 16 h. (E and F) SUMO-1 decreases chromatin-bound ERβ. Nuclear extracts of cells transfected as described for panel D were separated into soluble (S), chromatin-bound (CB), and matrix (M) fractions validated by Western blotting for their respective markers (E). Nuclear fractions were also normalized to their respective marker and analyzed with anti-ERβ antibody (F). (G) SUMO-1 reduced ERβ promoter occupancy through pSuM. ChIP analysis on Hs578t cells expressing ERβ wild type or pSuM mutants with site-specific mutations and transfected or not with the SUMO-1 construct. Cells were treated with 10 nM estradiol for 40 min prior to harvest. Quantified ChIP signals were corrected to their corresponding inputs and expressed as fold change relative to wild-type ERβ. PgR prom, proximal promoter; PgR enh, distal enhancer.

We then evaluated whether ERβ subcellular localization was affected by sumoylation. In fluorescence microscopy, ERβ was mostly found in the nuclear fraction in cells, and no obvious relocalization of ERβ to the cytoplasm was observed in response to SUMO-1 addition (see Fig. S4 in the supplemental material). Using Western blot analysis of subcellular compartments, SUMO-1 expression was found to enhance ERβ accumulation in the nuclear fraction compared to the cytosol (Fig. 6C and D). An enrichment in the nuclear fraction was also observed for the SUMO-competent S6D mutant and was observed to a greater extent for the S6E mutant, even in the absence of SUMO-1 addition, indicating a favored clustering of sumoylated ERβ to the nucleus with no apparent transition to the cytoplasm. To better define the intranuclear localization, nuclear extracts were separated into soluble (S), chromatin-bound (CB) nucleoplasm, and insoluble matrix (M) fractions (Fig. 6E). When compared in terms of the relative proportion of total fractions, ERβ content was reduced in the CB fraction in response to SUMO-1 expression (17% versus 4% in the absence of SUMO-1), as opposed to the contents for the K4R (19 versus 15%) and S6A (14 versus 15%) SUMO-deficient mutants (Fig. 6F). Interestingly, the SUMO-competent S6D (19 versus 6%) and S6E (6 versus 3%) mutants also exhibited a release from the CB fraction in response to SUMO-1, supporting a reduced ability of sumoylated ERβ to reside in the chromatin fraction. Consistent with this, the S6E mutant already showed a weak occupation in the CB fraction in the absence of SUMO-1 addition, similar to the result observed for wild-type ERβ with SUMO-1 (compare the lower left with the upper right panels in Fig. 6F). These results indicate that sumoylation regulates ERβ intranuclear movement, in line with its transcriptional potential to respond in a SUMO-dependent context.

To add more specificity to the role of sumoylation in the regulation of ERβ binding to chromatin, we performed ChIP assays on ER-regulated promoters using human Hs578t breast cancer cells. Expression of SUMO-1 decreased ERβ recruitment to the estrogen response element (ERE) of the proximal promoter region of the TFF1, PgR, and Cox7A2L genes, with the reductions ranging from 30 to 80% compared to the result for the control (Fig. 6G). This effect of SUMO was also observed on ERβ recruitment to more distal estrogen-regulated binding sites, as shown for the PgR enhancer region (40), suggesting a similar cis-regulated role of SUMO on proximal and distal binding of ERβ. The K4R mutation induced a potent recruitment of ERβ to these promoter regions, with values reaching 4.8-fold enrichment occupancy on the TFF1 promoter compared to wild-type receptor, whereas SUMO-competent S6D and S6E variants had markedly reduced abilities to associate with these regions (Fig. 6G). These results are consistent with a negative effect of Lys-4 sumoylation on ERβ transcriptional potential and on the importance of pSuM integrity in altering the ERβ interaction with cis-binding elements, thereby supporting the inhibitory outcome of SUMO-1 on estrogen-dependent gene regulation by ERβ.

SUMO modification at the pSuM site of ERβ is enhanced by GSK3.

As suggested by the results shown in Fig. 2G and H, proline-directed kinases of the MAPK family are likely to be involved in promoting ERβ sumoylation through Ser-6 of pSuM. In an attempt to better characterize the incoming phosphorylation signal that functions to enhance pSuM sumoylation, we tested the effects of various kinase inhibitors on hormone-dependent ERβ sumoylation. Inhibition of Erk1/2, p38, and Jnk slightly decreased the sumoylation of ERβ, suggesting that all three MAPK members might participate, and inhibition of Src tyrosine kinase and of cyclic AMP-dependent kinase PKA had weak effects (Fig. 7A). However, a striking decrease in ERβ SUMO-1 conjugation was observed with the inhibition of GSK3 with BIO (Fig. 7A) and with GSK3β knockdown (Fig. 7B) in cells, indicating that GSK3β is involved in promoting ERβ sumoylation. Ser-6 is not contained within a consensus GSK3 phosphorylation site in ERβ, and sumoylation of S6D and S6E mutants was also decreased with GSK3 inhibition (Fig. 7C), suggesting that other putative phosphorylation sites might be required for GSK3-induced sumoylation of ERβ.

Fig 7.

Sumoylation of ERβ is dependent on GSK3 activity. (A) GSK3 regulates ERβ sumoylation. The effect of protein kinase inhibitors for Mek (PD98059), p38 (SB203580), Jnk (SP600125), Src (PP2), PKA (H89), and GSK3 (BIO) was assessed on ERβ sumoylation in 293 cells transfected with wild-type ERβ and GFP–SUMO-1. Cells were left untreated or treated with inhibitors and 10 nM estradiol for 6 h and 5 h, respectively. (B) Cells were transfected as described for panel A in the presence or absence of Ubc9 and then treated with 10 nM estradiol. Knockdown of GSK3β was achieved by lentiviral infection using GSK3β shRNA (shGSK3β). shLuc was used as a negative control. Also shown are input levels of SUMO-1 and endogenous amounts of GSK3β determined by Western blot analysis. (C) Effect of GSK3 inhibition on sumoylation of ERβ SUMO-competent S6D and S6E mutants. Cells were treated as described for panel A. Asterisks denote nonspecific signals.

GSK3 enhances estrogen-dependent sumoylation of ERβ through an extended pSuM.

Interestingly, a highly conserved GSK3α/β recognition motif corresponding to the consensus (S/T)XXX(S/T)P sequence was found immediately adjacent to the pSuM sequence, in which Ser-8 and Ser-12 of human ERβ are included (Fig. 8A). We thus tested the possibility that these sites might confer responsiveness to GSK3 in order to regulate ERβ sumoylation. Indeed, while they were still able to maintain low levels of SUMO-1 conjugation, possibly through their intact pSuM, the S8A and S12A mutants and the S8 and S12A double mutant remained mostly unresponsive to GSK3 inhibition (Fig. 8B). In addition, phosphorylation of ERβ by GSK3 promoted its sumoylation in vitro, which was impaired with Ser-8 and -12 substitutions (Fig. 8C). These results identify Ser-8 and Ser-12 to be regulatory sites of ERβ sumoylation under GSK3 control and define an extended regulatory motif immediately flanking the pSuM which is implicated in the hormone-dependent sumoylation of ERβ. The fact that both Ser-8 and Ser-12 are indispensable to achieve maximal sumoylation in response to hormone is in line with the model described for most GSK3 substrates (6), in which in this case Ser-12 of ERβ would serve as the priming phosphorylation site to conduct the subsequent phosphorylation of Ser-8 by GSK3. The importance of these two serine residues is also emphasized by their highly conserved status along with the pSuM sequence among mammals and other vertebrates expressing ERβ (Fig. 8A).

Fig 8.

GSK3 enhances estrogen-dependent sumoylation of ERβ through an extended pSuM. (A) Conserved homology sequence alignment of the pSuM ψKXS sequences among ERβ species (h, human; m, mouse; r, rat; rh, rhesus monkey; ov, Ovis aries; bi, Sturnus vulgaris; x, Xenopus) and other transcription regulators. The position of the first amino acid residue of each sequence is indicated in parentheses. Highly conserved residues are shaded. Also shown is the homology between putative phosphorylated residues (Ser-8 and Ser-12 in human ERβ) contained in the GSK3 consensus motif extension. (B) Ser-8 and Ser-12 participate in ERβ sumoylation. Sumoylation assay in 293 cells transfected with the ERβ wild type or S8A or S12A mutant or the S8 and S12A double mutant. Cells were left untreated or treated with 4 μM GSK3 inhibitor BIO and 10 nM estradiol for 6 h and 5 h, respectively. (C) In vitro sumoylation of [³⁵S]methionine-labeled ERβ and Ser-8 and -12 variants incubated with purified Ubc9 and SUMO-1 in the presence or absence of SAE1/2. When indicated, ERβ proteins were also subjected to in vitro phosphorylation with purified GSK3β. Sumoylation was analyzed by SDS-PAGE and fluorography. (D) GSK3β activity is induced by estrogen. 293 cells were transfected with ERβ and treated with 10 nM estradiol for the indicated times. Endogenous phosphorylated GSK3β levels were determined by Western blot analysis using a phosphorylation-specific antibody against the inhibitory Ser-9 phosphorylation site of GSK3β. Extracts were also analyzed for total GSK3β content. (E) Serum decreases GSK3β activity through Akt. Western blot analysis performed as described for panel D with serum-depleted cells treated with 10 nM estradiol, 5% FBS, and 10 nM Akt inhibitor LY294002 for 5 h. (F) Sumoylation of ERβ is decreased by serum and Akt. Sumoylation of ERβ was performed as described for panel B, and cells were treated with estrogen, serum, and Akt inhibitor as described for panel E. (G) Sumoylation of endogenous ERβ is dependent on GSK3β activity. ER-positive breast cancer MCF-7 cells were transfected with GFP–SUMO-1 and then treated with 10 nM estradiol for 16 h in the presence or absence of 6 μM GSK3 inhibitor BIO. ERβ was immunoprecipitated and analyzed with an anti-GFP antibody by Western blotting. Input ERβ levels were also analyzed for control loading.

In order to further understand how GSK3 contributes to ERβ sumoylation in the presence of estrogen, we next investigated whether GSK3 activity might itself be modulated by estrogen. GSK3α and GSK3β function as constitutive enzymes in resting cells, and it is upon Akt-dependent phosphorylation of their respective homologous Ser-21 and Ser-9 residues that they become inactivated (6). We thus analyzed GSK3 phosphorylation using a phosphorylation-specific Ser-9 GSK3β antibody in Western blot analysis and found that estrogen induces Ser-9 dephosphorylation over a 3-h period of treatment, supporting activation of GSK3β (Fig. 8D). ERβ was able to translate the effects of estrogen on GSK3β phosphorylation at as soon as 5 min of treatment, possibly referring to a nongenomic action of ERβ. Interestingly, SUMO-1 expression also increases GSK3β activity by reducing Ser-9 phosphorylation even in the absence of estrogen (Fig. 8E). This effect was alleviated with serum addition through activation of Akt, consistent with the role of Akt as a negative regulator of GSK3 activity. Such a decrease in GSK3β activity by serum correlates with a reduction in ERβ SUMO-1 conjugation, which was relieved with Akt inhibition (Fig. 8F). GSK3 inhibition also impaired ERβ sumoylation in ER-positive breast cancer MCF-7 cells, supporting a role of GSK3 in mediating ERβ sumoylation in an endogenous context (Fig. 8G). Altogether, these results indicate an important role of GSK3β in stimulation of the sumoylation of ERβ in the context of hormone and identify a regulatory extension to the nonconsensus pSuM SUMO recruiting motif that enables ERβ sumoylation in a phosphorylation-dependent manner.

DISCUSSION

In this study, we identify nuclear receptor ERβ to be a sumoylated target that undergoes conjugation of SUMO-1 to the accepting lysine 4 embedded into a ψKXS motif that diverges from the core consensus ψKXE sequence. This atypical SUMO recognition motif that we termed pSuM is located in the unstructured AF-1 domain of ERβ and is made competent for proper SUMO modification through the negatively charged environment provided by the phosphorylation of Ser-6, which seems sufficient to re-create a genuine SUMO consensus motif. The initial determination of the classic ψKXE SUMO conjugation motif had been made possible after comparing the sequence flanking the acceptor lysine residues located within the various SUMO targets originally identified, such as RanGAP1, PML, IκB, and p53, among others (13, 35). Structural studies have revealed that both the bulky aliphatic and acidic side chains of the ψKXE motif are needed to interact with Ubc9 in order to protrude the lysine into the catalytic pocket of the E2 enzyme (4). Whether phosphorylation of pSuM confers a stable recognized structure similar to the ones proposed for proliferating cell nuclear antigen (PCNA) (22) and E2-25K/Hip2 (32) or, rather, whether it might represent a different mechanism in recruiting and positioning of Ubc9 remains to be clarified. However, the negative charge provided by addition of phosphate seems indispensable and certainly represents an attractive mechanism in regulating SUMO modification at the level of the target itself.

Sumoylation has been shown to be enhanced by phosphorylation, for example, in HSF-1 and -4 (21), GATA1 (7), ERRα and -γ (21, 46, 49), and MEF2 (17). In most cases, such regulation occurs through a PDSM motif, which differs from pSuM by its phosphorylated serine, which is not located directly within the core sumoylation motif but is located in the downstream extension (21). Similarly, a cluster of acidic residues flanking the ψKXE site was shown to facilitate sumoylation of Elk-1, PML, and LRH-1/NR5A2 (51). The proposed mechanism to enhance sumoylation through these two extended motifs referred to their common property of adding a negatively charged environment to promote substrate recognition with the Ubc9 basic patch (30). Here, we identify GSK3 to be an important regulator of sumoylation through the phosphorylation-dependent activation of a regulatory extension to pSuM, which is itself under kinase regulation. This dual relationship is depicted in Fig. 9. In this model, we define an extended pSuM corresponding to the sequence ψKXSPS(S)XXSP, in which the phosphorylation of adjacent serine residues at positions 8 and 12 of ERβ brings additional negative charges to potentiate the sumoylation process initiated at pSuM. Interestingly, the extended pSuM is highly conserved between ERβ species (Fig. 8A), which suggests that GSK3β is an essential regulator of pSuM-targeted sumoylation of ERβ across vertebrates. On the basis of the known penchant of GSK3 for phosphorylation-primed substrates, it is believed that phosphorylation of Ser-12 is primarily mediated by another kinase that targets ERβ. With Ser-12 being part of a Ser-Pro motif, the interesting possibility that proline-directed kinases of the MAPK family might participate, as also suggested for Ser-6 within pSuM, is highly expected, especially on the basis of their prominent role in regulation of the AF-1 activity of ERβ (36). Whether the same kinase would promote Ser-6 and Ser-12 phosphorylation, therefore enabling pSuM phosphorylation and activation by GSK3β, remains to be determined. Moreover, we do not know if Ser-9, which is also highly conserved in the pSuM extension of ERβ-expressing species (Fig. 8A), participates as well in such a buildup of negative charge addition under GSK3β action. Clearly, further investigation is needed to clarify these aspects.

Fig 9.

Proposed model for ERβ regulation by SUMO-1. ERβ is SUMO-1 modified on Lys-4 of the pSuM sequence (red) located at the N-terminal end of the AF-1 domain. This sumoylation can occur in a ligand-independent manner but is highly favored by hormone, which results in optimal Ser-6 phosphorylation by MAPK, providing the necessary negative charge in pSuM for SUMO-1 attachment. This process is further enhanced through presumably nongenomic effects of ERβ, resulting in GSK3β activation and phosphorylation of Ser-8 (underlined) immediately flanking pSuM. It is not known whether Ser-12 can be phosphorylated by MAPK, but such phosphorylation is thought to serve as a priming event for GSK3β action. Modification of ERβ by SUMO then regulates diverse aspects of ERβ functions, including receptor stability through competition with ubiquitin on the same accepting site, intranuclear localization, and transcriptional repression by disrupting ERβ interaction with active chromatin.

While investigating more closely the role of GSK3β in ERβ sumoylation, we observed that GSK3β was rapidly activated by estrogen and that ERβ itself can translate these effects, providing a mechanism by which ERβ SUMO addition is under hormonal regulation. These findings also suggest that GSK3 activity can be regulated by nongenomic actions of ERβ in response to estrogen. Initially confined to glycogen metabolism, GSK3 has turned out to have a key role in many cellular signaling pathways, including responses to insulin, Wnt, growth factors, and nutrients, among others, a finding which confers on GSK3 a multitasking role and the implication of its role in various pathologies (6, 10). GSK3β is active in resting cells through constitutive Tyr-216 phosphorylation, and PI3K/Akt-activating stimuli such as insulin can cause phosphorylation at Ser-9 and inactivation of GSK3β (8). The mechanism by which nongenomic activation of ERβ might contribute to Ser-9 dephosphorylation and GSK3β activation in our study is still uncertain, but recent findings in rodent brain capillaries and breast cancer cells have shown that sustained ERβ activation by extended exposure to estradiol resulted in PTEN activation and decreased GSK3 phosphorylation (19, 27). Interestingly, part of the mechanism involved might also refer to the ERβ response to Akt activation, as we demonstrated that activation of the PI3K/Akt pathway downregulated ERβ degradation and activity in breast cancer cells (38). This indicates that ERβ might play an important role in regulating GSK3β activity in estrogen-responsive cells.

Sumoylation of transcription factors has mainly been associated with their transcriptional attenuation or repression (15), and our results indicate that it is as such with ERβ. Interestingly, ERα is one of the few examples for which sumoylation was shown to exert transcriptional activation (5, 42). This uncovers sumoylation as a key regulator for both isoforms, but with such apparently opposite effects between each subtype, it might have clear implications on ER-regulated gene expression, especially in the context of breast cancer cells that express both forms. For example, it has been proposed that ERβ can interfere with ERα-mediated oncogenic proliferation of breast cancer cells, therefore providing a tumor suppressor-like activity of ERβ dependent on cellular context (12, 18, 26). Whether sumoylation of ERβ might affect such activity remains an interesting issue to resolve, but with the prominent role of ERβ AF-1 activity in modulation of ERα-mediated transcription (16, 47) and given that ERβ sumoylation appears to be restricted to the AF-1 domain, these findings support an integrated role for sumoylation in regulating ER hormonal actions and biological functions.

Our results revealed that SUMO-1 conjugation affected ERβ protein turnover by competing with ubiquitin addition in a process dependent on Lys-4, protecting ERβ against proteasomal degradation. This proposes that Lys-4 is a common and shared accepting residue for both modifications and suggests a mutually exclusive site that determines the ERβ response to incoming signals, but direct evidence of ubiquitin addition to Lys-4 remains to be ascertained. Since phosphorylation of pSuM is a determinant required for ERβ sumoylation, it is interesting to note that ubiquitination of ERβ was also shown to be induced by kinase-dependent signals targeting the AF-1 domain (31). This implies that ERβ sumoylation and ubiquitination are regulated in a dynamic fashion, possibly impairing each other if they depend on different somewhat overlapping recruiting signals, and/or are regulated by the modified target itself. Consistent with this, our observations suggest that as long as pSuM is activated by phosphorylation, ERβ remains more competent for sumoylation between the two reversible processes, suggesting a shift in the equilibrium on the side of the SUMO-ERβ pool in response to regulated kinase activating signals. For this reason, sumoylation might constitute a protective step for ERβ degradation by the proteasome, which possibly competes with the role of ERβ ubiquitination in achieving hormone-induced gene expression (31). This proposed mechanism is functionally analogous to the one utilized by IκBα, except that in this case, phosphorylation inhibited IκBα sumoylation (9). It is likely that although transient, each modification may label ERβ for different transcriptional outcomes, depending on Lys-4 availability. Alternative functions in DNA repair have been reported for PCNA, which also exhibits sumoylation and ubiquitination on the same site (22). Thus, our results demonstrate that this shared role of Lys-4 constitutes an important regulated step in ERβ turnover, chromatin binding, and modulation of transcription.

Our search for other potential targets that contain the minimal ψKXS pSuM sequence has identified several proteins involved in a variety of functions and therefore might not be stringent enough to predict phosphorylation-dependent sumoylated substrates. However, a homology search for pSuM with the GSK3 extension conferred more specificity, as it selected a restricted number of proteins, which, interestingly, were mostly associated with gene transcription, including GATA-4, SREBP-2, NF-IL-3, and also Nor-1, which belongs to the nuclear receptor superfamily (Fig. 8A), suggesting that a shared mechanism coordinates various gene expression programs. Of interest, is the identification of SREBP-1 and -2 as targets of GSK3, with several sites involved, some of which remain to be identified and which mediate SREBP ubiquitin-dependent turnover (25, 44). Also of interest is the identification of Ser-183, contained in the predicted pSuM of NF-IL-3, as a phosphorylation site involved in NF-IL-3 degradation (11). Although the functionality of the predicted extended pSuM in these targets and others has yet to be determined, our findings uncovered a novel and highly regulated SUMO motif, which expands the dynamic nature of the sumoylation process and adds to the intricate mechanisms used by incoming signals to regulate gene transcription.