Histone Demethylase JMJD2B Functions as a Co-Factor of Estrogen Receptor in Breast Cancer Proliferation and Mammary Gland Development

Estrogen is a key regulator of normal function of female reproductive system and plays a pivotal role in the development and progression of breast cancer. Here, we demonstrate that JMJD2B (also known as KDM4B) constitutes a key component of the estrogen signaling pathway. JMJD2B is expressed in a high proportion of human breast tumors, and that expression levels significantly correlate with estrogen receptor (ER) positivity. In addition, 17-beta-estradiol (E2) induces JMJD2B expression in an ERα dependent manner. JMJD2B interacts with ERα and components of the SWI/SNF-B chromatin remodeling complex. JMJD2B is recruited to ERα target sites, demethylates H3K9me3 and facilitates transcription of ER responsive genes including MYB, MYC and CCND1. As a consequence, knockdown of JMJD2B severely impairs estrogen-induced cell proliferation and the tumor formation capacity of breast cancer cells. Furthermore, Jmjd2b-deletion in mammary epithelial cells exhibits delayed mammary gland development in female mice. Taken together, these findings suggest an essential role for JMJD2B in the estrogen signaling, and identify JMJD2B as a potential therapeutic target in breast cancer.


Introduction
Estrogen plays an important role in normal physiology and several human diseases including breast cancer. The estrogen signaling pathway is a reliable therapeutic target for estrogenreceptor (ER) positive subtype of breast cancer. Understanding of how ER regulates transcription is key to overcoming resistance to existing selective ER modulators (SERMs) and identifying non-SERM targets suitable for novel therapeutic approaches [1].
The biological functions of estrogen are mediated through ER, which regulates transcription of ER target genes by binding to estrogen responsive elements (EREs) [2]. Liganded ER undergoes conformational changes which facilitate cofactor recruitment [3], and forms multi-subunit protein complexes [4,5,6]. Many of these co-regulators are enzymes that alter chromatin structure or control sequential transcriptional reactions [1,7], which include ATPdependent chromatin remodeling complexes (SWI/SNF) [8]. However, the molecular mechanisms of how co-regulators are recruited to the specific genes and how the integrated signals are transmitted to chromatin are not fully understood.
Genome-wide analyses of chromatin modifications have revealed a complex landscape of modified histones at transcription start sites (TSSs), distal regulatory elements and conserved sequences. In general, methylated H3K4 and H3K36 are associated with active transcription, whereas methylated H3K9, H3K27 and H4K20 are associated with gene silencing. H3K4me3 and H3K27me3 signals peak near TSSs. The relative presence of H3K4me3 and H3K27me3 has been shown to affect promoter activity [9,10,11,12,13]. H3K9me3 is generally linked to gene silencing since H3K9me3 is enriched in heterochromatin and inactive genes [13,14,15].
Abnormalities in the methylation of histones by histone methyltransferases have been implicated in various cancers [24]. Thus, it is possible that dysregulation of LSD1 or JmjC-domaincontaining histone demethylases could contribute to tumorigenesis. Here we characterize JMJD2B as a newly-appointed coregulator of ERa signaling in breast cancer growth and mammary gland development.

Ethics Statement
All experimental procedures were approved by the local Ethical Committee for Animal Experimentation of Ontario Cancer Institute (OCI), University Health Network (AUP#2031). The research projects that are approved by the local Ethical Committee for Animal Experimentation of OCI are operated in accordance with applicable Federal, Provincial, Municipal and Institutional regulations, the Policies and Guidelines of the Canadian Council on Animal Care and the Province of Ontario's Animals for Research Act. The animals are housed -accordance with applicable Federal, Provincial, Municipal and Institutional regulations, the Policies and Guidelines of the Canadian Council on Animal Care and the Province of Ontario's Animals for Research Act in the OCI Animal Facilities of the university. The Institute is committed to the highest ethical standards of care for animals used for the purpose of continued progress in the field of human medicine.

Antibodies and primers
All antibodies and primers used are described in Procedure S1.

Database search for JMJD2B expression in breast cancers
The ONCOMINE database and gene microarray analysis tool, a repository for published complementary DNA microarray data (http://www.oncomine.org) [25,26], was searched to retrieve information on JMJD2B mRNA expression in human breast cancers. Statistical analysis of the differences in JMJD2B expression between ER-positive and ER-negative breast cancers was performed using ONCOMINE algorithms that allow for multiple comparisons among different studies.

Real-time RT-PCR
RNA was extracted from cells using an RNeasy Mini kit (Qiagen). Total cellular RNA was converted into cDNA by reverse transcription (Superscript III; Invitrogen Life Technologies) using random primers. PCR amplification was performed using Power SYBR Green qPCR SuperMix-UDG with ROX (Invitrogen Life Technologies) through 40 cycles of 95uC for 15 s and 60uC for 60 s using an Applied Biosystems PRISM 7900 Sequence Detection System.

Soft agar assay
ZR-75-1 cells, MCF-7 cells, or derivative cell lines were cultured in 12-well plates containing a bottom agar layer consisting of culture medium plus 0.7% agar and 2 ng/mL puromycin. The middle layer contained 10 5 cells in culture medium plus 0.35% agar and 1 ng/mL puromycin. Medium alone was added as the top layer to prevent desiccation of the agarose. Colonies were allowed to form for 14 days prior to visualization by crystal violet staining.

Mouse Xenograft Breast Cancer Models
Slow-release estradiol pellets were implanted subcutaneously into NIH-III mice (Crl:NIH-Lyst bg Foxn1 nu Btk xid , 7-week old, Charles River Laboratories) three days before tumor transplantation. 3610 6 ZR-75-1 cells grown in cell culture were suspended in 50 mL medium, mixed with 50 mL Matrigel, and injected subcutaneously into hind flanks. Tumors derived from injected cells were harvested two weeks after transplantation.

BrdU/7-AAD staining
For cell cycle analysis using BrdU and 7-amino-actinomycin D (7-AAD), cells were pulsed with 10 mM BrdU for 1 hr. The FITC BrdU flow kit (BD Biosciences) was used to detect BrdU. Fluoresence-activated cell sorting (FACS) analysis was performed on a FACSCalibur (Becton Dickinson), and data were analyzed with Cellquest or FlowJoe software.

Immunoprecipitation assay
Cells were washed with PBS, scraped off, and collected by centrifugation. The cells were suspended in hypotonic buffer (10 mM Tris-HCl (pH 7.4), 10 mM NaCl, 3 mM MgCl 2 , 0.5% NP-40) in the presence of protease inhibitors (Complete(R), Roche). After washing with PBS, the pellet was resuspended in 1 mL of NP-40 IP lysis buffer (50 mM TrisHCl (pH 7.6), 150 mM NaCl, 1% NP-40, 1 mM EDTA) containing protease inhibitors. The lysate was sonicated at 5% output for 20 sec. After centrifugation, the supernatant was collected and precleared by incubating with protein A magnetic beads (DynabeadsH, Invitrogen) at 4uC for 1 hr. Fresh protein A beads were washed with PBS-T (PBS+0.05% Tween) three times, resuspended in 100 mL PBS-T, and incubated with 5 mg of indicated antibodies at 4uC for 1 hr. The antibodycoated beads were then washed with PBS-T three times and incubated with 800 mL of the lysate at 4uC over night. The rest of the samples were kept for INPUT. After extensive wash, the beads were suspended in 80 mL of 26sample buffer and boiled at 95uC for 5 min to elute associated proteins. To elute JMJD2C, the beads were incubated with 40 mL of 8 M Urea for 15 min at 37uC, and then boiled with 40 mL of 46 sample buffer.

ChIP assay
Cells were crosslinked in 1% formaldehyde for 10 min at room temperature. Cells were lysed and sonicated to obtain fragmented chromatin samples, which were immunoprecipitated using specific antibodies coupled to protein A Dynabeads (Invitrogen). The protein-DNA complexes were eluted from the beads and incubated at 65uC overnight to reverse protein-DNA crosslinks. Precipitated DNA was analyzed by real-time PCR. Refer to Procedure S1 for primer information. The full version of ChIP assay procedure is in Procedure S1.

Gene targeting in mice
A Jmjd2b-targeting vector was constructed in which the neo cassette was flanked by frt sequences, and Jmjd2b exon 5 was flanked by loxP sequences ( Figure S6A). Cre-mediated removal of exon 5 results in a frameshift and translation termination, and removal of H189 and E191 residues essential for demethylase activity. Homologous recombination in ES cells and germ line transmission were confirmed by Southern blot analysis ( Figures  S6B abd S6C). The full version of the gene targeting strategy is Procedures S1.

Isolation of mammary epithelial cells
Mammary fat pads were minced into paste, digested in collagenase/hyaluronidase solution (StemCell Technologies, 07912), and dissociated with dispase (StemCell Technologies, 07913). Samples were washed with Hanks' Balanced Salt Solution supplemented with 2% FBS and 2 mmol/L EDTA (HFE) and centrifuged at 3006g for 5 min. The pellet was washed with HFE and then passed through a 40 mm cell strainer to obtain a singlecell suspension. CD45 + Ter119 + and CD31 + cells were removed using the EasySep biotin selection kit (StemCell Technologies) to obtain Lin 2 cells.

Results
As a first step towards identifying signature differences between ER-positive and ER-negative breast cancers, we searched the ONCOMINE database [25][26] and found 19 suitable studies (Table S1). In most of these studies, expression of JMJD2B mRNA was higher in ER-positive cancers than in ER-negative cancers ( Figure 1A; refer to Table S1 for p-values). We examined protein levels of JMJD2 family members in breast cancer cell lines. JMJD2B expression was generally higher in ER-positive than in ER-negative lines ( Figure 1B). Furthermore, JMJD2B protein levels increased in response to E2 treatment in the ER-positive breast cancer cell line T-47D, but not in the ER-negative breast cancer cell line MDA-MB468 ( Figure 1C). Real-time RT-PCR analysis confirmed that JMJD2B mRNA, but not JMJD2A or JMJD2C mRNA, was induced upon E2 treatment ( Figure 1D).
To understand the role of JMJD2B in regulating cancer cell growth, siRNA was used to knock down JMJD2B expression in T-47D cells ( Figure S1A). At 72 hr post-siRNA transfection, we analyzed proliferation by BrdU/7-AAD staining and flow cytometry, and found that BrdU incorporation was substantially reduced in JMJD2B-depleted cells but not in control cells or JMJD2A-depleted cells (Figure 2A). To investigate the effects of prolonged JMJD2B depletion, we generated ZR-75-1 cells in which JMJD2B expression was knocked down using short hairpin RNA (shRNA; Figure S1D), and assessed the ability of these cells to form colonies in soft agar. JMJD2B-depleted ZR-75-1 cells formed fewer and smaller colonies relative to control ( Figure 2B). Reduced colony formation was also observed for MCF-7 cells expressing JMJD2B-targeted shRNA ( Figure S2).
We performed xenograft experiments in which NIH-III mice were subcutaneously implanted with slow-release estradiol pellets and injected with ZR-75-1 cells expressing control shRNA and shRNA against JMJD2B. The two cell lines were injected into opposing flanks of the same mouse. JMJD2B-depeleted cells gave rise to significantly smaller tumors than cells expressing control shRNA ( Figure 2C). These results demonstrate that JMJD2B positively regulates the proliferation of ER-positive breast cancer cell lines.
JMJD2C associates with androgen receptor and is required for its transcriptional activity [29]. Furthermore, knockdown of JMJD2C in prostate cancer cell lines impairs the response to androgen receptor ligand [29]. We, therefore, speculated that JMJD2B may be required for ER transcriptional activity. We cotransfected 293T cells with MYC-tagged ERa and FLAG-tagged JMJD2B, and observed that JMJD2B co-immunoprecipitated with ERa ( Figure 3A). We then generated a series of FLAG-tagged JMJD2B mutant proteins bearing carboxy-terminal deletions ( Figure 3B) and tested them for co-immunoprecipitation with ERa. The JMJD2B protein contains N-terminal JmjN and JmjC domains, a central Pro-rich domain, and C-terminal double PHD and double Tudor domains ( Figure 3B). We found that only the JmjC domain was required for the interaction with ERa ( Figures 3C and S3D). We then tested whether ERa interacts with a catalytically inactive JMJD2B mutant (DFeJMJD2B; Figure  S3F) which contains point mutations (H189Y and E191A) in the iron-binding region [30,31,32,33]. The data suggest that catalytic activity of JMJD2B does not affect the interaction with ERa ( Figure S3A). We found that the benzonase treatment did not influence the interaction between ERa and JMJD2B ( Figures S3B  and S3C). These data demonstrate that the ERa-JMJD2B interaction is DNA independent.
Given that the JmjC domain of JMJD2B mediates the interaction with ERa and that the JmjC domain is conserved among JMJD2 proteins, it is possible that JMJD2A and JMJD2C also interact with ERa. To address this issue, nuclear proteins were extracted from T-47D cells and co-immunoprecipitation experiments were performed with antibodies specific to endogenous proteins. Antibodies to JMJD2A or JMJD2B immunoprecipitated a substantial amount of ERa. In contrast, complexes between JMJD2C and ERa appeared to be less abundant ( Figure 3D). We also performed co-immunoprecipitation assay for the extracts treated with benzoase.
ERa recruits the SWI/SNF chromatin remodeling complex (BAF complex) to target genes upon estrogen stimulation. We found that JMJD2A, JMJD2B, and JMJD2C interacted with one of the core subunits of SWI/SNF complex, BRG1 ATPase ( Figure 3D). JMJD2A and JMJD2B also interacted with Poly-bromo1 but not ARID1B, whereas JMJD2C interacted with ARID1B and, to a lesser extent, Polybromo1 ( Figure 3D). Because Polybromo1 is specific to the SWI/SNF-B complex (P-BAF complex) and ARID1B is specific to the SWI/SNF complex, these results suggest that JMJD2A and JMJD2B associate specifically with the SWI/SNF-B complex whereas JMJD2C associates preferentially with the SWI/SNF complex. We further examined whether formation of these JMJD2B-containing complexes is estrogen dependent. Estrogen treatment resulted in nuclear accumulation of ERa and an increased association between ERa and JMJD2B ( Figures 3E and S3E). In contrast, estrogen stimulation did not affect the interaction between JMJD2B and subunits of the SWI/SNF complex ( Figures 3E and  S3E), suggesting that JMJD2B can interact with the SWI/SNF-B complex in the absence of estrogen stimulation. We could not detect an interaction between JMJD2B and p300 ( Figure 3E) which is supposedly recruited to ERa with kinetics distinct to those of the SWI/SNF complex [7].
We next assessed whether JMJD2B is required for activation of ER target genes. Induction of GREB1 in response to E2 was significantly reduced in JMJD2B-depleted T-47D cells and JMJD2B-depleted MCF-7 cells but not in JMJD2A-depleted cells ( Figure 4A). MYB oncogene is an ER target gene and required for ER-induced cell proliferation [34]. The induction of MYB in T-47D cells and MCF-7 cells was impaired by JMJD2B knockdown, whereas JMJD2A knockdown impaired MYB induction to a modest degree ( Figure 4A). These results are consistent with the observation that knockdown of JMJD2A had little effect on the proliferation of T-47D cells (Figure 2A). Collectively, these results indicate that JMJD2B is required for the full extent of ERa transcriptional activity and that the JMJD2A has only a limited role in the transcriptional activity of ERa despite a comparable affinity for ER and SWI/SNF-B complexes. Therefore, we focused our attention on the function of JMJD2B. Induction of other ER target genes including MYC, CCND1, and BCL-2 was also reduced in JMJD2B-depleted T-47D cells ( Figure 4B) and JMJD2Bdepeleted MCF-7 cells ( Figure S4A).
To determine whether JMJD2B is required for cellular responses to estrogen, we stimulated JMJD2B-depleted T-47D and MCF-7 cells with E2 and measured BrdU incorporation at 72 hr post-transfection. E2 stimulated the proliferation of control cells but had a limited effect on the proliferation of JMJD2Bdepleted cells (Figures 4D and S4D), indicating that JMJD2B is required for the proliferative response to estrogen. Since the induction of MYB, MYC, and CCND1 genes were reduced in JMJD2B-depleted cells upon E2 treatment, impairment of G1/S transition appears to contribute the defective proliferation. It has been recently reported that JMJD2B-depletion also influences genes required for G2/M transition [35].
We next analyzed the interaction of ERa with target gene loci by performing chromatin immunoprecipitation (ChIP). We selected candidate ER binding sites according to the results of a published genome-wide analysis [36] and confirmed the binding of ERa to ER target genes including MYB, JMJD2B, and GREB1 ( Figure 5A). Recently, it has been reported that most estrogen receptor binding sites are distal to transcription start sites (TSSs). Indeed, ER binds to MYB and GREB1 loci more than 20 kb downstream of their respective TSSs ( Figure S5A). The interaction of ERa with these sites increased 45 min after E2 stimulation and then declined 4 hr later ( Figure 5A). We also observed marked increase of JMJD2B binding to these sites following E2 stimulation ( Figure 5A). Notably, ERa and JMJD2B were both detected at the ERa binding site in the JMJD2B locus, suggesting that JMJD2B regulates itself at the transcriptional level in concert with ERa, which is consistent with the previous genome wide analysis of ERa binding site [37].
We next addressed the demethylation of H3K9me3 at ER binding sites. We observed decreased H3K9me3 at ER binding sites following E2 stimulation in control but not in JMJD2Bdepleted T-47D cells ( Figure 5B), suggesting that JMJD2B regulates the demethylation of H3K9me3 at ER binding sites. H3 enrichment at MYB ERE appears to decrease over the time after E2 addition but H3 displacement during the time window we have analyzed is not statistically significant whereas H3K9me3/ H3 ratio at MYB ERE decreases to approximately one-third in the presence of E2 ( Figure S5C). We also examined RNA polymerase II (RNAPII) levels at upstream of the ER binding site in MYB gene. The RNAPII levels were markedly increased in control T47D cells in response to E2 treatment. Conversely, the induction of RNAPII levels was significantly impaired in JMJD2B-depleted cells ( Figure 5C). These findings further support a role of JMJB2B in ER regulated transcription.  To further study the involvement of JMJD2B in the demethylation of H3K9me3 at an ER target gene [34], we performed ChIP on 12 regions within MYB locus. Loading of ERa occurred almost exclusively at the region identified above ( Figure 5D). In control T-47D cells, H3K9me3 levels were reduced following E2 treatment at several regions of the MYB locus, including the ER binding site. In contrast, chromatin-bound H3K9me3 was not reduced in JMJD2B-depleted cells ( Figure 5D). We also performed ChIP for ERa. We found that JMJD2B depletion reduced ERa enrichment ( Figure 5D), suggesting that JMJD2B depletion reduces recruitment of ERa or stability of ERa complex.
Taken together, in response to E2 stimulation, JMJD2B is recruited to the ER binding site and demethylates H3K9me3 in the surrounding region, thus facilitating gene induction. In some cases, levels of chromatin-bound H3K9me3 were higher in JMJD2B-depleted cells ( Figure 5B and 5D). Histone methyltransferases has been proposed to associate with ERa to establish a basal repressive state at ER targets [21]. It is possible that the increase in H3K9me3 levels was caused by the loss of the competing demethylase activity of JMJD2B.
To determine whether JMJD2B is required for the proliferation of normal epithelial cells under physiological conditions, we generated mice carrying a conditional allele of Jmjd2b that can be removed in mammary epithelial cells (MECs) by expression of Cre recombinase under the control of the MMTV promoter (Jmjd2b flox/flox ;MMTV-Cre mice; Figure S6). We chose to flank Jmjd2b exon 5 with loxP sequences because this exon encodes a fragment containing H189 and E191 residues that are essential for iron-binding and demethylase activity [30,31,32,33]. We analyzed whole-mount specimens of mammary fat pads and noticed that mammary gland development was delayed in Jmjd2b flox/flox ; MMTV-Cre mice ( Figure 6A). Relative to control mice, Jmjd2b flox/flox ;MTV-Cre mice exhibited reduced branching in the mammary gland ( Figure 6A), which is consistent with a previous report that ERa is required for ductal morphogenesis [38].
We next examined gene expression in isolated MECs. We isolated CD45-Ter119-CD31-(Lin2) MECs [39,40,41], and found that mRNA levels of Myb, CyclinD1 and c-myc were significantly reduced in Jmjd2b flox/flox ;MMTV-Cre mice compared to Jmjd2b flox/flox mice ( Figure 6B). These data further support that the defective mammary gland development is due to the impairment of the ER target genes induction. However, further study is required to clarify whether Jmjd2b-deficiency primarily leads to cell proliferation defect, apoptosis, or both.
To further assess impact of Jmjd2b-deficiency on proliferation of MECs, the isolated Jmjd2b flox/flox MECs were infected with either Mock-GFP (control) or iCre-IRES-GFP retrovirus, and then cultured in vitro. Western blotting confirmed that Jmjd2b protein was absent from cells infected with Cre-expressing virus ( Figure 6C). FACS was used to monitor the fraction of GFP-positive cells. The proportion of GFP-positive control cells did not change over time, whereas that of iCre-infected, GFP-positive cells gradually decreased, suggesting a role for Jmjd2b in the proliferation of mammary epithelial cells ( Figure 6D).
It is noteworthy that we have obtained viable Jmjd2b2/2 mice without gross abnormalities (data not shown), indicating that Jmjd2b is not a general regulator of cell proliferation. These observations further support our findings that Jmjd2b has tissue and signal specific functions.

Discussion
Intensive studies of how nuclear receptors transmit ligand binding signals have identified several distinct multiprotein complexes. These complexes are recruited to promoters and enhancers, and modulate higher-order chromatin and nucleosomal structures leading to activation or repression of target genes. Genome-wide analysis of ERa binding sites has been performed by numerous groups and potential ERE sites have been mapped in promoter, enhancer and intronic regions; however, how the distal regulatory EREs function and what are co-factors at these sites remain to be elucidated.
Here, we describe indispensable roles for an H3K9 histone demetylase, JMJD2B, in ER signaling. JMJD2B was required for the full induction of ER target genes that function in cell proliferation and survival pathways ( Figure 4). Consequently, inactivation of JMJD2B severely impaired the growth of ERpositive breast cancer and development of normal breast tissues in vivo (Figures 2 and 6). These findings are consistent with a recent report that JMJD2B depletion causes decrease in cell proliferation and colony formation capacity of ER-positive breast cancer cells [35].
Further, our studies address the mechanisms by which JMJD2B acts as a co-regulator in ER signaling. We observed that once ligand-bound ERa translocated to the nucleus, JMJD2B was recruited to distal ER binding sites of target genes concomitant with a decrease in chromatin-bound H3K9me3 levels ( Figure 5). In agreement with these findings, depletion of JMJD2B severely impaired the demethylation of H3K9me3 in ER-binding regions upon E2 treatment ( Figure 5). In addition to ERa, JMJD2B also associated with members of a chromatin remodeling complex, SWI/SNF-B, which is required for transactivation by nuclear receptors [42]. We therefore propose a model in which JMJD2B at ER binding sites removes the repressive histone marks, which may also set up docking sites for enzymes and transcription factors that remodel chromatin or otherwise modify gene-control regions.

Diversity among JMJD2 family proteins
Recent studies have found that JMJD2 family proteins can target methylated H3K9 and H3K36 with some target specificity [43], suggesting that JMJD2 members have both common and distinct biological functions. Biochemically, JMJD2A/KDM4A and JMJD2C/KDM4C can demethylate H3K9me3 and H3K36me3 [33,43,44], whereas JMJD2B/KDM4B is likely more Figure 5. JMJD2B is recruited to the ER binding site; H3K9me3 is demethylated at the ER binding site. (A) Chromatin immunoprecipitation (ChIP) assay of ERa and JMJD2B at indicated gene loci. T-47D cells were steroid-depleted for 96 hr and treated with E2 for 45 min or 4 hr. mRNA levels of JMJD2B and the indicated ER target genes in the corresponding samples are in Figure S5B. (B) ChIP assay of H3K9me3 at indicated gene loci. T-47D cells expressing control shRNA or shRNA against JMJD2B were steroid-depleted for 96 hr and treated with E2 for 4 hr. Results were normalized to input values, and expressed relative to untreated controls (set at 1). (C) ChIP assay of RNAPII in MYB gene. T-47D cells expressing control shRNA or JMJD2B shRNA were treated as in (B). Results were normalized to input values, and expressed relative to untreated controls (set at 1). (D) Presence of ERa and H3K9me3 within the MYB locus in T-47D cells. A region previously shown to be devoid of ERa was chosen as a negative control site (ER-neg) [36]. Cells were treated as in (B). H3K9me3 ChIP signals were normalized to input chromatin and expressed relative to ChIP signals of the negative control site (set at 1). X-axis represents distance from TSS. (A-C) ChIP samples were quantified by real-time PCR. Data represent mean 6 s.d. of triplicates. Representative data from three independent trials. **p,0.01; *p,0.05. ER binding sites assessed in this study are shown in Figure S5. doi:10.1371/journal.pone.0017830.g005 specific for H3K9me3 in vitro [43] although overexpression of JMJD2B demethylates H3K36me3 in cells [31]. Functionally, JMJD2A regulates the balance of H3K9me3 and H3K36me3 required for maintaining genomic stability in germ cells [43]. JMJD2B has been shown to antagonize H3K9me3 at heterochromatin [31]. JMJD2C promotes AR signaling [29]. Our data identify a specific role for JMJD2B in ER signaling. Among the JMJD2 family members we tested, JMJD2A and 2B exhibited robust interactions with ER ( Figure 3); however, in contrast to depletion of JMJD2B, depletion of JMJD2A caused only a marginal defect in ER target gene induction (Figure 4).
JMJD2A and JMJD2B, although expressed, were not found to be involved in AR signaling [29]. Our data strongly suggest that JMJD2C (SWI/SNF-A) and JMJD2B (SWI/SNF-B) associate with different remodeling complexes. Estrogen may also further enhance JMJD2B expression to drive the vicious cycle. These findings might explain the specificity of ligand-receptor signaling and the mutually exclusive pathological implications involving JMJD2B and JMJD2C. Further investigation will be required to elucidate the signal and target specificity of each JMJD2 family member.

Roles of histone lysine methylation in ER signaling
Genome-wide ChIP analysis of ER binding regions [36,37,45] and ChIP-sequence studies of RNAPII recruitment [46,47] have revealed that, in addition to ERa binding, modifications of local chromatin and epigenetic states, recruitment of specific co-factors and release of RNAPII pausing are critical for the transcriptional regulation of ER responsive genes.
Enrichment of H3K9me3 has been reported in coding regions of some active genes [48,49]; however recent genome-wide investigations support the general repressive nature of the H3K9me3 mark [13,15]. Increasing recent evidence indicates that genomic architecture is not colinear but is modular [50], and permissive and repressive chromatin regions exist within the body of gene. The intragenic permissive chromatin regions are flanked by the repressive mark, H3K9me3, and the maintenance of the intragenic chromatin boundary appear to functions as a checkpoint in elongation [51]. We observed enrichment of H3K9me3 near the distal ER binding site under steroid-depleted condition, and found that the methylation levels significantly decreased upon E2 stimulation in a JMJD2B dependent manner ( Figures 5B and 5D). In addition, JMJD2B-depletion reduced RNAPII occupancy ( Figure 5C). Thus, our data suggest that JMJD2B contributes the establishment of local epigenetic state and chromatin structure, which are required for proper induction of ER responsive genes.
We observed the binding of ERa and JMJD2B to ER binding regions in the absence of estrogen stimulation although the binding increased upon estrogen stimulation ( Figure 5A). LSD1 and JMJD2C have been shown to present at promoter and enhancer regions of AR-dependent genes prior to ligand stimulation. However, ligand-receptor interaction is needed to induce AR recruitment and subsequent demethylation [22,29]. In an analogous way, recruitment of liganded ER and other cofactors may be necessary for JMJD2B activity. Further genomewide analysis will be required to clarify these issues.

Biological significances of JMJD2B in human diseases
Recent studies have implicated H3K9 modifications in numerous biological phenomena including germ cell development, X chromosome inactivation, DNA damage repair and apoptosis [14]. Recent reports also link deregulated histone methylation to tumorigenesis [52,53].
Interestingly, focal amplification of JMJD2B locus was identified in brain tumors [54]. An H3K9 histone methyltransferase, Suv39H1, has been shown to function as a tumor suppressor by maintaining H3K9 methylation levels [55,56]. In addition, we show that JMJD2B enhances the transcriptional activation of oncogenes and anti-apoptotic genes. Dysregulation of JMJD2 thus may counteract the tumor suppressive function of Suv39H1.
In conclusion, our data indicate that JMJD2B regulates an epigenetic signature in ER binding regions distal to TSSs leading to the establishment of specific transcriptional programs and biological effects downstream of ER signaling. In addition, the recent report has identified that combination of expression levels of a hypoxic marker and JMJD2B predict patient survival [35].We thus believe that further characterization of JMJD2B and related signaling components may identify therapeutic targets and prognostic markers for human cancers.  Figure 3B. Cell lysates and a-FLAG immunoprecipitates were analyzed by western blot with antibodies against the indicated proteins. Relative intensity of the bands of the Flag-JMJD2B and mutants are shown, normalized to the bands of corresponding input. The values are presented wild type as 1. Arrowheads, ERa bands; arrows, wild type or deletion mutant JMJD2B proteins; asterisk, a-FLAG antibody. (E) Kinetics of Association between JMJD2B and ERa or SWI/SNF-B complex. Nuclear lysates were harvested at indicated time points after E2 stimulation and subjected to immunoprecipitation with control IgG or the antibodies against the indicated proteins. Input lysate and immunoprecipitated samples were then immunoblotted using antibodies against the indicated proteins. (F) JMJD2B reduces H3K9me3 levels. U2OS cells were transfected with FLAG-JMJD2B and FLAG-DFeJMJD2B and stained with anti-FLAG and anti-H3K9me3 antibodies followed by anti-mouse IgG (green) or anti-rabbit IgG (red). Nuclei were visualized using Hoechst 33258. Overlay: merge of FLAG and H3K9me3 staining. Data shown are representative of two independent preparations. (TIF) Figure S4 JMJD2B mediates induction of ER target genes and estrogen-dependent proliferation of breast cancer cells. (A) JMJD2B is required for the induction of ER target genes. MCF-7 cells were transfected with either control siRNA or JMJD2B siRNA (target sequence #1), cultured in steroid-free medium for 72 hr, and stimulated with or without E2 for 4 hr. mRNA levels of JMJD2B or the indicated ER target genes were measured by real-time RT-PCR. Results shown are mean mRNA level normalized to the amount of ACTB mRNA 6 s.d. of triplicates. **, p,0.01. (B) Microarray data for representative ER target genes. Signal intensities on microarray for representative ER target genes are shown by their mean values (6SD). (C) Heat map representation of differentially expressed genes. One thousand four hundred and thirty-two differentially expressed genes (as calculated using a false discovery rate ,0.05 and log-fold change .2) were sorted by hierarchical clustering. Each row represents a gene and each column represents a sample.