As a first step towards identifying signature differences between ER-positive and ER-negative breast cancers, we searched the ONCOMINE database 
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 (; 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 (). 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 (). Real-time RT-PCR analysis confirmed that JMJD2B
mRNA, but not JMJD2A
mRNA, was induced upon E2 treatment ().
Elevated expression of JMJD2B in ER-positive breast cancers.
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 (). 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 (). Reduced colony formation was also observed for MCF-7 cells expressing JMJD2B-targeted shRNA (Figure S2
JMJD2B positively regulates the proliferation of ER-positive breast cancer cells.
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 (). 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 
. Furthermore, knockdown of JMJD2C in prostate cancer cell lines impairs the response to androgen receptor ligand 
. We, therefore, speculated that JMJD2B may be required for ER transcriptional activity. We co-transfected 293T cells with MYC-tagged ERα and FLAG-tagged JMJD2B, and observed that JMJD2B co-immunoprecipitated with ERα (). We then generated a series of FLAG-tagged JMJD2B mutant proteins bearing carboxy-terminal deletions () and tested them for co-immunoprecipitation with ERα. The JMJD2B protein contains N-terminal JmjN and JmjC domains, a central Pro-rich domain, and C-terminal double PHD and double Tudor domains (). We found that only the JmjC domain was required for the interaction with ERα ( and S3D
). We then tested whether ERα interacts with a catalytically inactive JMJD2B mutant (ΔFeJMJD2B; Figure S3F
) which contains point mutations (H189Y and E191A) in the iron-binding region 
. The data suggest that catalytic activity of JMJD2B does not affect the interaction with ERα (Figure S3A
). We found that the benzonase treatment did not influence the interaction between ERα and JMJD2B (Figures S3B and S3C
). These data demonstrate that the ERα-JMJD2B interaction is DNA independent.
JMJD2B interacts with ERα and SWI/SNF-B complex.
Given that the JmjC domain of JMJD2B mediates the interaction with ERα and that the JmjC domain is conserved among JMJD2 proteins, it is possible that JMJD2A and JMJD2C also interact with ERα. 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 ERα. In contrast, complexes between JMJD2C and ERα appeared to be less abundant (). We also performed co-immunoprecipitation assay for the extracts treated with benzoase.
ERα 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 (). JMJD2A and JMJD2B also interacted with Polybromo1 but not ARID1B, whereas JMJD2C interacted with ARID1B and, to a lesser extent, Polybromo1 (). 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 ERα and an increased association between ERα and JMJD2B ( and S3E
). In contrast, estrogen stimulation did not affect the interaction between JMJD2B and subunits of the SWI/SNF complex ( 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 () which is supposedly recruited to ERα with kinetics distinct to those of the SWI/SNF complex 
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 (). MYB
oncogene is an ER target gene and required for ER-induced cell proliferation 
. 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 (). These results are consistent with the observation that knockdown of JMJD2A had little effect on the proliferation of T-47D cells (). Collectively, these results indicate that JMJD2B is required for the full extent of ERα transcriptional activity and that the JMJD2A has only a limited role in the transcriptional activity of ERα 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
, and BCL-2
was also reduced in JMJD2B-depleted T-47D cells () and JMJD2B-depeleted MCF-7 cells (Figure S4A
JMJD2B mediates induction of ER target genes and estrogen-dependent proliferation of breast cancer cells.
To further assess the effect of JMJD2B depletion, we performed genome-wide gene expression analysis using the Affymetrix Human Gene 1.0 ST array. RNA was extracted from steroid-depleted control MCF-7 cells (control E2(−)), control MCF-7 cells treated with E2 (control E2(+)), steroid-depleted JMJD2B-depleted MCF-7 cells (siJ2B E2(−)), and JMJD2B-depleted MCF-7 cells treated with E2 (siJ2B E2(+)). After normalizing the data, we identified genes that exhibited a log-fold change of greater than 2 between control E2(−) and control E2(+) cells, or between control E2(+) and siJ2B E2(+) cells. (The false discovery rate (FDR) was set at <0.05.) In total, we identified 1432 differentially expressed genes ( and S4C
; Table S2
). 878 genes were up-regulated or down-regulated by JMJD2B knockdown, but not affected by E2 stimulation. These genes represent candidate JMJD2B target genes that are independent of ER signaling. 360 genes were induced by E2, whereas 194 genes were down-regulated by E2. Among these 360 genes, 53 genes (14.7%) were down-regulated in siJ2B E2(+) relative to control E2(+) cells (), indicating that a subset of E2-induced genes were dependent on JMJD2B for full induction. Again we observed impaired induction of MYB
, and BCL2
in JMJD2B-depleted cells (Figure S4B
), although the changes in transcript levels did not pass the criteria of <0.05 FDR and >2 log-fold difference.
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 JMJD2B-depleted cells ( and S4D
), indicating that JMJD2B is required for the proliferative response to estrogen. Since the induction of MYB
, 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 
We next analyzed the interaction of ERα 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 
and confirmed the binding of ERα to ER target genes including MYB
, and GREB1
(). Recently, it has been reported that most estrogen receptor binding sites are distal to transcription start sites (TSSs). Indeed, ER binds to MYB
loci more than 20 kb downstream of their respective TSSs (Figure S5A
). The interaction of ERα with these sites increased 45 min after E2 stimulation and then declined 4 hr later (). We also observed marked increase of JMJD2B binding to these sites following E2 stimulation (). Notably, ERα and JMJD2B were both detected at the ERα binding site in the JMJD2B
locus, suggesting that JMJD2B regulates itself at the transcriptional level in concert with ERα, which is consistent with the previous genome wide analysis of ERα binding site 
JMJD2B is recruited to the ER binding site; H3K9me3 is demethylated at the ER binding site.
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 JMJD2B-depleted T-47D cells (), 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 (). 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 
, we performed ChIP on 12 regions within MYB
locus. Loading of ERα occurred almost exclusively at the region identified above (). 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 (). We also performed ChIP for ERα. We found that JMJD2B depletion reduced ERα enrichment (), suggesting that JMJD2B depletion reduces recruitment of ERα or stability of ERα 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 (). Histone methyltransferases has been proposed to associate with ERα to establish a basal repressive state at ER targets 
. 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 (Jmjd2bflox/flox
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 
. We analyzed whole-mount specimens of mammary fat pads and noticed that mammary gland development was delayed in Jmjd2bflox/flox
mice (). Relative to control mice, Jmjd2bflox/flox
mice exhibited reduced branching in the mammary gland (), which is consistent with a previous report that ERα is required for ductal morphogenesis 
Conditional deletion of Jmjd2b in mammary epithelium results in defective mammary gland development.
We next examined gene expression in isolated MECs. We isolated CD45-Ter119-CD31- (Lin−) MECs 
, and found that mRNA levels of Myb
were significantly reduced in Jmjd2bflox/flox
mice compared to Jmjd2bflox/flox
mice (). 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 Jmjd2bflox/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 (). 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 ().
It is noteworthy that we have obtained viable Jmjd2b−/− 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.