Identification of the MLL2 Complex as a Coactivator for Estrogen Receptor α*

A novel estrogen receptor (ER)α coactivator complex, the MLL2 complex, which consists of MLL2, ASH2, RBQ3, and WDR5, was identified. ERα directly binds to the MLL2 complex through two LXXLL motifs in a region of MLL2 near the C terminus in a ligand-dependent manner. Disrupting the interaction between ERα and the MLL2 complex with small interfering RNAs specific against MLL2 or an MLL2 fragment representing the interacting region with ERα significantly inhibited the ERα transcription activity. The MLL2 complex was recruited on promoters of ERα target genes along with ERα upon estrogen stimulation. Inhibition of MLL2 expression decreased the estrogen-induced expression of ERα target genes cathepsin D and to a lesser extent pS2. In addition, MCF-7 cell growth was also inhibited by the depletion of MLL2. These results demonstrate that the ERα signaling pathway is critically dependent on its direct interaction with the MLL2 complex and suggest a central role for the MLL2 complex in the growth of ERα-positive cancer cells.

The biological effects of estrogen are mediated by estrogen receptors (ER) 2 in estrogen responsive tissues. There are two types of estrogen receptors, ER␣ and ER␤. The well studied ER␣ is involved in normal mammary gland development as well as breast cancer initiation and progress (1)(2)(3)(4). ER␣ has two transcriptional activation domains, the N-terminal activation domain AF-1 and the C-terminal activation domain AF-2. Upon estrogen binding, ER␣ undergoes a conformational change and regulates the expression of its target genes (5,6). ER␣, just as other nuclear receptors, requires coactivators and corepressors for its function. A large number of ER␣ coactivators, including the three members of the SRC-1 family (SRC-1, SRC-2/GRIP1/TIF2, and SRC-3/AIB1/ ACTR/pCID/RAC3/TRAM1) (7)(8)(9), CREB-binding protein (CBP/ p300), and TRAP220 (DRIP 205, PBP) (10,11), have been identified to date. Most of the coactivators interact with the AF-2 domain of ER␣ in a ligand-dependent manner. Some of these cofactors are intrinsic enzymes with the activity of acetyltransferase or methyltransferase or are able to recruit such enzymes (12)(13)(14) that modify histone composition of chromatin to make transcription factors accessible to specific regions of the genome. The varying patterns of histone modification are now referred to as a histone code and are proposed to be epigenetic markers for determining gene activation status (15). Some nuclear receptor coactivators (corepressors) are presented as multiprotein complexes, and these steady-state protein complexes probably act as functional units of nuclear receptor coregulators (16). ER␣ coactivator TRAP220/PBP exists in the multiprotein TRAP complex, which has molecular mass of ϳ2 MD and is composed of more than 30 subunits (17). The TRAP complex facilitates ER␣ actions by synergizing basal transcription machineries. The ER␣ coactivator PRIP (TRBP, TRAP250, NRC, and AIB3) is also demonstrated to stay in a massive steady-state complex ASCOM (18), which consists of MLL4, PRIP (ASC2), MLL2, ASH2, RBQ3 and ␣/␤-tubulins.
Mixed lineage leukemia (MLL, also termed ALL-1, HRX, and HTRX) is often involved in chromosome translocations in human acute leukemia and functionally essential for maintaining the expression of the HOX gene during embryo development (19,20). MLL is the mammalian homolog of the yeast SET1 protein, which is present as a SET1 complex harboring the methyltransferase to modify chromatin histone H3 (19,22). MLL is reported to be present within a large multiprotein complex composed of MLL1, ASH2, RBQ3, WDR5, and Menin, which is a product of the MEN1 tumor suppressor gene (23)(24)(25)(26). Despite extensive research work being done on MLL, little is known about the function of MLL2 (or ALR), a close homologue of MLL, which is a polypeptide of 5262 amino acids encoded by an mRNA that is over 18-kb long (27).
Identifying ER␣ interacting complexes is essential to understanding how ER␣ controls gene transcription and how it is involved in the development of breast cancer. Here we describe the identification of a novel protein complex, the MLL2 complex, consisting of MLL2, ASH2, RBQ3, and WDR5, which is required for ligand-dependent ER␣ transactivation. Disruption of the complex repressed estrogen-dependent expression of ER␣ target genes and also inhibited growth of MCF-7 cells.
GST-MLL2 set proteins were produced in the Escherichia coli BL21 codon plus strain (Stratagene), and purified by glutathione-Sepharose beads 4B. The 172-amino-acid MLL2 fragment was released from the GST fusion protein by treatment with thrombin. Polyclonal rabbit antiserum against the C-terminal 172-amino-acid of MLL2 was developed by the Spring Valley Laboratory. Rabbit polyclonal anti-Ash2 and anti-RBQ3 antibodies were purchased from Bethyl Laboratories. Anti-ER␣ monoclonal antibody was purchased from Santa Cruz Biotechnology. Polyclonal rabbit anti-PRIP was generated in previous work (29). The polyclonal anti-WDR5 antiserum was a kind gift of Dr. W. Herr (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY).
Isolation of ER␣ AF2-associated Complexes-DU4475 cells, an ERnegative breast cancer cell line, were grown in a suspension culture by the National Cell Culture Center (Minneapolis, MN). DU4475 cell nuclear extracts were prepared as described (30) and precleaned by incubating with an excess amount of GST-Sepharose 4B beads. Immobilized GST-ER␣ (LBD) fusion proteins were preincubated for 1 h at 4°C in GST binding buffer (20 mM Tris-HCl, pH 7.9, 180 mM KCl, 0.2 mM EDTA, 0.05% Nonidet P-40, 0.5 mM phenylmethylsulfonyl fluoride, and 1 mM dithiothreitol) containing bovine serum albumin (1 mg/ml). Bead-immobilized proteins were then incubated at 4°C overnight with precleaned DU4475 nuclear extract in the presence or absence of 100 nM 17␤-estradiol (E2). After washing with GST wash buffer (GST binding buffer with 0.1% Nonidet P-40) four times, proteins were eluted with binding buffer containing 30 mM reduced glutathione. Elutants were concentrated and layered on top of a 4.5-ml linear 10 -40% glycerol gradient in 5-ml tubes. Following a 16-h centrifugation at 4°C using an SW40 rotor (Beckman) with a speed of 40,000 rpm, 200 l of solution from each fraction was collected from top to bottom. Ovalbumin (44 kDa), ␤-globulin (158 kDa), and thyroglobulin (667 kDa) were used as protein standards.
Immunoprecipitation-DU4475 nuclear extracts, 2.0 mg, were precleaned with polyclonal rabbit anti-PRIP for 2 h at 4°C and high speed centrifugation to remove precipitates. The supernatant was divided equally and immunoprecipitated by specific antibodies or control IgG in GST binding buffer (20 mM Tris-HCl, pH 7.9, 180 mM KCl, 0.2 mM EDTA, 0.05% Nonidet P-40, 0.5 mM phenylmethylsulfonyl fluoride, and 1 mM dithiothreitol). After extensive washing using the same buffer, the bound proteins were eluted by boiling in SDS-PAGE sample buffer, resolved by SDS-PAGE, transferred onto a nitrocellulose membrane, and subjected to Western blot analysis using antibodies as indicated.
GST Pull-down Assays-GST and GST fusion proteins were produced in the E. coli BL21 codon plus strain (Stratagene) and purified by glutathione-Sepharose beads 4B according to the manufacturer's instructions (Amersham Biosciences). Proteins were in vitro translated using TNT quick coupled in vitro transcription and translation kit (Promega) and labeled with [ 35 S]methionine. A 20-l volume of GST fusion protein bead slurry was incubated with 5 l of in vitro translated protein for 2 h in 500 l of NETN (20 mM Tris-HCl, pH 7.5, 100 mM KCl, 0.7 mM EDTA, 0.05% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride) in the presence or absence of 100 nM E2. After the beads were washed four times with binding buffer, 30 l of Laemmli protein loading buffer were added and boiled for 3 min. Samples were separated by SDS-PAGE, amplified, dried, and autoradiographed.
Cell Culture and Transfection-A total of 2 ϫ 10 5 MCF-7 cells were seeded in 6-well plates 24 h before transfection. Cells were transfected with 1.25 g of ERE-Luc, 20 ng of pCDNA3.1-ER␣, 2 g of siRNA or plasmid as indicated, and 0.1 g of pCMV␤ (Clontech) using Lipofectamine 2000 reagent (Invitrogen). The cells were split equally into two different wells on the next day, and the cell medium was changed to phenol red-free modified Eagle's medium containing 10% of charcoal stripped FBS. The cells were treated with or without 100 nM E2 overnight and lysed for luciferase and ␤-galactosidase activity assays (Tropix, Bedford, MA). Three independent transfections were performed for each assay.
Chromatin Immunoprecipitation-MCF-7 cells were maintained in phenol red-free modified Eagle's medium supplemented with 10% charcoal-dextran-stripped fetal bovine serum. After the hormone treatment for 45 min, cells were cross-linked with 1% formaldehyde at room temperature for 10 min. After the cells were collected, chromatin immunoprecipitation was performed as described (31). Briefly, nuclei were prepared by incubating the cells in cell lysis buffer (10 mM Tris, pH 8.0, 10 mM NaCl, 0.2% Nonidet P-40) containing protease inhibitors (phenylmethylsulfonyl fluoride, pepstatin, aprotinin, and leupeptin) for 10 min on ice. Nuclei were precipitated and lysed with nuclear lysis buffer (50 mM Tris, pH 8.1, 10 mM EDTA, 1% SDS, phosphatase and protease inhibitors) for 10 min on ice. The chromatin was sheared by sonication to an average size of 500 -1000 base pairs. Soluble chromatin was diluted 10-fold with dilution buffer (16.7 mM Tris-Cl, pH 8.1, 1.1% Triton X-100, 1.2 mM EDTA, 167 mM NaCl) and precleared with salmon sperm DNA-blocked preimmune IgG protein A beads. Immunoprecipitation was performed by incubating the precleared cell lysate with spe-FIGURE 1. A, isolation of the MLL2 complex. DU4475 cell nuclear extracts were pulled down with glutathione-Sepharose 4B immobilized GST-ER␣ (LBD) in the presence or absence of E2. The binding proteins were eluted from beads with glutathione and fractioned by glycerol gradient centrifugation. The fractions with a molecular mass over ϳ600 kDa were combined, resolved by SDS-PAGE, visualized by silver stain, and subjected to tandem mass spectrometry analysis. The components of the MLL2 complex identified by tandem mass spectrometry analysis are indicated next to the corresponding bands. The asterisks denote the GST-ER␣. B, Western blot analysis of the same samples analyzed by tandem mass spectrometry (A). Specific antibodies direct against indicated proteins were used in immunoblotting.
Cell Proliferation Assay-MCF-7 cells were plated at a density of 5 ϫ 10 5 cells/well in 6-well plates. The cells were transfected with 2 g of siRNAs using Lipofectamine 2000 reagent on the following day. After 24 h, the cells were trypsinized, counted, and seeded into 24-well plates with a density of 2 ϫ 10 4 cells/well in fresh phenol red-free modified Eagle's medium with 10% fetal bovine serum and 100 nM E2. Following every 2 days, viable cells were counted after trypan blue staining.

Isolation of the MLL2 Complex
Interacting with ER␣ AF2 Domain-To isolate potential protein complexes that facilitate the liganddependent transcription activity of ER␣, the GST-fused ER␣ ligand binding domain GST-ER␣ (LBD) was used to pull down its interacting proteins from the DU4475 cell line nuclear extracts in either the presence or absence of E2. Glycerol gradient segmentation was subsequently performed to collect the high molecular mass fractions (Ͼ600 kDa), which were subjected to tandem mass spectrometry analyses. As expected, we detected the TRAP/PBP complex, which was abundant and consisted of more than 30 subunits, three members of SRC-1 family, PRIP, and CBP/P300. In addition, mass spectrometry identified multiple peptides from the MLL-related protein MLL2 (ALR-1) (GenBank TM accession number 4505197), ASH2 (accession number 4009336), retinoblastoma-binding protein RBQ-3 (accession number 755750), and WD repeat domain 5 protein (WDR5) (accession number 12804457) (Fig. 1A), which likely formed a new protein complex. The presence of MLL2, ASH2, RBQ-3, and WDR5 in the proteins pulled down by GST-ER␣ (LBD) was confirmed by Western blot (Fig. 1B).
To demonstrate that MLL2, ASH2, RBQ3, and WDR5 indeed formed a complex, we generated a rabbit polyclonal antibody against the C-terminal 172 amino acids of mouse MLL2 that specifically recognizes MLL2 of both mouse and human origin. The DU4475 cell nuclear extract was precipitated with anti-MLL2, and the precipitates were then examined by Western blot using antibodies against ASH2, RBQ3, and WDR5. As shown in Fig. 2A, anti-MLL2 but not the control IgG was able to precipitate ASH2, RBQ3, and WDR5, suggesting that MLL2 interacted with ASH2, RBQ3, and WDR5. Similarly, we demonstrated that ASH2 could be coprecipitated with RBQ3, WDR5, and MLL2, whereas RBQ3 was coimmunoprecipitated with ASH2, WDR5, and MLL2 ( Fig. 2A). These results suggested that MLL2, ASH2, RBQ3, and WDR5 were in one complex, which we designated as the MLL2 complex. A, verification of the components of the MLL2 complex. Nuclear extract of DU4475 cells was immunodepleted of PRIP and subjected to immunoprecipitation (IP) using antibodies specific for components of the MLL2 complex including ASH2, MLL2, and RBQ3, respectively. The immunoprecipitates were separated by SDS-PAGE and then immunoblotted with the antibodies indicated on the right of the panels. B, MLL2 complex and ASCOM complex are distinct. DU4475 cell nuclear extracts were immunodepleted with an excess amount of protein A bound anti-PRIP or preimmune rabbit IgG. The equal amounts of proteins from the supernatants after depletion were separated by SDS-PAGE and then immunoblotted with indicated antibodies.
ASCOM consisting of MLL4, PRIP (ASC2), MLL2, ASH2, RBQ3, and ␣/␤-tubulins (18) is a complex that shares components with the MLL2 complex. However, the results from mass spectrometry indicated that the MLL2 complex and ASCOM complex belong to two distinct complexes. This was evident as the abundance of MLL2 in the pull-down proteins as revealed by mass spectrometry was 20 times more than that of PRIP, and no MLL4 was detected by mass spectrometry. In addition, WDR5 is present in MLL2 complex but not in ASCOM complex. To confirm that the MLL2 protein complex is not an ASCOM complex, we performed an immunodepletion in which we incubated DU4475 nuclear extract with an excess amount of protein A-bound anti-PRIP. As a control, a same amount of DU4475 nuclear extract was immunodepleted with preimmune rabbit IgG. The remaining proteins from both groups were subjected to Western blot to compare relative abundances of individual proteins. As shown in Fig. 2B, although PRIP was almost completely cleared by precleaning with anti-PRIP, AHS2, RBQ3, and MLL2 proteins remained at similar levels to those in the preimmune rabbit IgG immunodepletion group, indicating that the vast majority of these proteins did not associate with PRIP in DU4475 nuclear protein.
These results suggest that the MLL2 complex was a novel protein complex distinct from the ASCOM complex.
ER␣ Interacts with the MLL2 Complex through MLL2 in a Ligand-dependent Fashion-To address which protein of the MLL2 complex interacts directly with ER␣, we carried out an in vitro pull-down assay with each component of the complex and GST-ER␣ (LBD). As shown in Fig. 3A, no detectable interactions were seen between GST-ER␣ (LBD) and full-length ASH2, RBQ3, or WDR5 in the presence or absence of estrogen. As MLL2 is a huge protein with 5262 amino acids, we tested those regions containing LXXLL motifs, which are required for the interaction between nuclear receptor and coactivators in a ligand-dependent fashion. GST-ER␣ (LBD) efficiently bound to a [ 35 S]methionine-labeled fragment of MLL2 (amino acids 4167-4780) containing two LXXLL motifs only in the presence of estrogen. When this region was divided into two fragments (amino acids 4167-4471 and amino acids 4471-4780) with each containing one LXXLL motif, the in vitro binding assay demonstrated that both fragments interacted with estrogen-bound GST-ER␣ (LBD). When the two LXXLL motifs were mutated, the fragment no longer interacted with GST-ER␣ (LBD) (Fig.  3B). The other two MLL2 regions (amino acids 2377-2835 and amino acids 3783-3905), each carrying two LXXLL motifs, showed no interactions with GST-ER␣ (LBD) (Fig. 3B). These results suggest that ER␣ binds to the MLL2 complex through two LXXLL motifs in MLL2  (located between amino acids 4167 and 4780) in the presence of estrogen. MLL2 protein contains a highly conserved SET domain, which was found to strongly interact with ASH2 but not with RBQ3 and WDR5 by GST pull-down assay (Fig. 3C ).
MLL2 Complex Is Needed for ER␣ Transcriptional Activity-We next sought to determine whether the observed interaction between ER␣ and the MLL2 complex is functionally relevant in the ER␣ signaling pathway. A pool of four siRNA target against human MLL2 (siMLL2) was employed to deplete intracellular MLL2, whereas a pool of four small interfering RNA with no specific target (siControl) was used as a control. As seen in Fig. 4A, transfection of MLL2 siRNA in MCF-7 cells led to ϳ70% reduction of estrogen-dependent activation of the ERE-TK-Luc reporter gene (Fig. 4A), suggesting that the ER␣ signaling pathway requires MLL2.
Given that ER␣ interacts with the MLL2 complex via two LXXLL motifs in MLL2, disruption of this interaction could be achieved by overexpressing the truncated MLL2 fragment containing these two LXXLL motifs. Cotransfection of ER␣ and ERE-TK-Luciferase with the truncated MLL2 fragment (4167-4780 amino acids in MCF-7 cells) caused a 43% reduction of estrogen-dependent transactivation of ER␣ compared with the control (Fig. 4B), whereas the truncated MLL2 carrying mutated LXXLL motifs did not decrease the activity of ER␣, further demonstrating the importance of MLL2 complex in the transcriptional activity of ER␣.

MLL2 Is Required for the Expression of Endogenous Estrogen-inducible
Genes-To investigate if the MLL2 complex is recruited to liganded ER␣ in the nuclei of living cells, we performed chromatin immunoprecipitation assay with the promoters of endogenous estrogen-responsive target genes cathepsin D and pS2. The antibodies against ASH2, RBQ3, and MLL2, along with the antibody against ER␣ could efficiently precipitate the estrogen-responsive regions of cathepsin D and pS2 (Fig. 5A) only when MCF-7 cells were treated with estrogen. These results suggest that estrogen treatment induced the recruitment of ER␣ and the MLL2 complex onto the promoter regions of estrogen-responsive target genes, which contain the estrogen-responsive element (ERE) in MCF-7 cells.
The effect of MLL2 depletion on the expression of endogenous ER␣ target genes was assessed by quantitative real time reverse transcriptase-PCR following siRNA transfection in MCF-7 cells. TaqMan probes for human MLL2, cathepsin D, pS2, and ␤-actin were used in the experiments, and all data were normalized over endogenous ␤-actin levels. As seen in Fig. 5B, siMLL2 transfection led to nearly an 80% reduction in the MLL2 mRNA level compared with MLL2 mRNA levels in cells transfected with siControl, indicating the efficiency of the transfection method and the specificity of siMLL2. Consistent with the chromatin immunoprecipitation experiment, estrogen-dependent activation of cathepsin D mRNA expression was significantly reduced to ϳ50% of the control level by siMLL2 transfection (Fig. 5B). We also observed a consistent, but to a lesser extent, reduction of pS2 mRNA, suggesting that MLL2 is differentially required for the mRNA expression of different estrogen-responsive ER␣ target genes.
MLL2 Is Required for Estrogen-dependent Growth of MCF-7 Cells-ER␣ regulates mammary gland development and is involved in the growth of ER-positive breast tumor cells. Because silencing of MLL2 by siRNA inhibited ER␣-mediated transcription in MCF-7 cells, we sought to determine whether estrogen-dependent growth of MCF-7 cells was affected by depletion of MLL2. In comparison with the transfection of control siRNA, MLL2 siRNA treatment significantly inhibited the estrogen-dependent growth of MCF-7 cells (Fig. 6), suggesting that MLL2 complex is involved in estrogendependent growth of breast cancer cells.

DISCUSSION
In an effort to identify novel molecules involved in ER␣ mediated signal transduction in mammary gland epithelium, we employed GST-ER␣ (LBD) to capture its interacting proteins from human breast carcinoma cell line DU4475 nuclear extract, followed by glycerol gradient centrifugation to fractionate associated proteins according to their molecular masses. We identified the MLL2 complex, which contains MLL2, ASH2, RBQ3, and WDR5. The MLL2 complex is a novel Set1like complex in mammalian cells (23)(24)(25)34). The yeast Set1-like complex COMPASS, containing proteins homologous to ASH2 and Trithorax, is involved in chromatin remodeling through the methylation of histone H3 at lysine residue 4. Although yeast only contains one Set1 complex, humans appear to have multiple Set1-like complexes, including the MLL1 complex and ASCOM. A major feature of the MLL1 complex is that it contains tumor suppressor menin, which should not be a component of MLL2 complex as it was not detected by mass spectrometry in our proteins pull downs with GST-ER␣ (LBD). The MLL2 complex is closely related to ASCOM as they share three components, but we demonstrated that they are distinct complexes. As MLL2 and ASCOM were isolated from different cells, it is possible that different cells possess different Set1-like complex for the ER␣ signaling pathway.
ER␣ strongly binds to the MLL2 protein through two LXXLL motifs in the region (amino acids 4167-4780) but does not bind to the other three components of the complex, indicating the MLL2 as a key component linking the MLL2 complex with ER␣. Full-length MLL2 protein, structurally similar to MLL, contains a SET domain and five PHD fingers. Although SET domains have been shown to exhibit histone meth-yltransferase activity (23,35), we were unable to demonstrate that the SET domain from MLL2 possessed this activity (data not shown). The function of the PHD domains in MLL family proteins remains to be understood. MLL2 also contains an N-terminal Ring finger domain, often harboring the ubiquitin ligases (36,37), which is not found in the MLL molecule. It is well established that the activation of ER␣ is coupled with its degradation (38,39,40). Hence, it is intriguing to question whether the MLL2 complex participates in the ligand-induced degradation of ER␣ through adding ubiquitin to ER␣.
Because of the gigantic molecular size of MLL2, which spans over 18 kb of mRNA, it is difficult to obtain its full-length cDNA for performing cotransfection to evaluate the effect of overexpression of MLL2 on ER␣ transcriptional activity. Indeed, silencing MLL2 by MLL2-specific siRNA in MCF-7 cell did impair ER␣ transcription activity. In addition, the MLL2 fragment, which only contains the region necessary for binding to ER␣, acted as a potent dominant-negative inhibitor of the ER␣ transactivation. Just as the MLL1 complex, which is recruited to the HOX gene promoter (41), we observed that components of the MLL2 complex were recruited onto the promoters of ER␣ target genes directly after ligand induction in MCF-7 cells. Consistent with this observation, suppression of MLL2 by its specific siRNA decreased estrogen-induced expression of ER␣ target genes. Suppression of MLL2 also inhibited estrogen-dependent growth of MCF-7 cells. Taken together, these results strongly suggest a key role of the MLL2 coactivator complex in ER␣-mediated signal transduction pathways. It should be noted that certain coactivators such as SRC-3 and PBP are amplified and overexpressed in some breast cancers (42,43). Besides, the MLL2 gene was mapped to chromosome band 12q12-13 (27), a region involved in duplications and translocations associated with cancers (21,44). Therefore it is tempting to propose that the MLL2 complex may be involved in the pathogenesis of breast neoplastic diseases.