JMJD6 Regulates ERα Methylation on Arginine

ERα functions are tightly controlled by numerous post-translational modifications including arginine methylation, which is required to mediate the extranuclear functions of the receptor. We report that upon oestrogenic stimulation, JMJD6, the only arginine demethylase described so far, interacts with and regulates methylated ERα (metERα) function. Moreover, by combining the silencing of JMJD6 with demethylation assays, we show that metERα is a new substrate for JMJD6. We propose that the demethylase activity of JMJD6 is a decisive regulator of the rapid physiological responses to oestrogen.


Introduction
Oestrogen (17b-oestradiol, E 2 ) a member of the steroid hormone family, plays a crucial role in many physiological processes and in disease, namely in breast cancer. The biological actions of oestrogen are mediated through ERa and ERb, which function in the nucleus as ligand-dependent transcription factors promoting gene transcription and the stimulation of cell growth in various tissues, including breast epithelial cells [1,2].
In addition to these well-documented effects, oestrogens also activate multiple signal transduction cascades outside of the nucleus via nongenomic signalling. This nongenomic pathway involves growth factor-dependent kinases and adaptor proteins leading to downstream activation of signalling molecules, such as MAPK and Akt [3][4][5][6]. Cellular responses to oestrogens are highly controlled and require the regulation of ERa function through numerous post-translational modifications that regulate both genomic and nongenomic pathways (For a review, [7]). Most nongenomic effects of oestrogen are mediated through the recruitment of the tyrosine kinase Src and PI3K [3,4]. Our team has contributed to the understanding of this pathway by demonstrating that arginine methylation of the receptor is prerequisite to oestrogen-induced formation of the ERa/Src/ PI3K complex which activates Akt [8]. Recently, we also showed that this pathway is activated in aggressive breast tumours and could constitute a new potential target for therapy [9].
Our finding of ERa methylation being a dynamic process strongly suggests the involvement of an enzyme that reverses this methylation. We therefore investigated if the only arginine demethylase identified so far, JMJD6, plays a role in this process. The enzymatic activity of this protein was described by Buick's team on histones displaying asymmetrical as well as symmetrical demethylation on arginine residues [10]. Indeed, three types of arginine methylation exist in mammalian cells: monomethylargi-nine (MMA), asymmetric dimethylarginine (ADMA) and symmetric dimethylarginine (SDMA). PRMT1 is the major Type 1 arginine methyltransferase that adds the ADMA mark and PRMT5 is the principal Type 2 enzyme which catalyzes SDMA [11].
In this paper, we demonstrate that JMJD6 demethylates the ERa methylated on R260, thereby regulating oestrogen nongenomic signalling. Moreover, global approaches suggest that JMJD6 regulates other arginine methylated-proteins and further studies should be done to validate this point, based on the data obtained by mass spectrometry.

JMJD6 Specifically Interacts with the Methylated Form of ERa
To investigate a functional link between JMJD6 and ERa, we firstly analysed whether the two proteins could interact. GST pulldown assays indicated that JMJD6 directly associates with ERa ( Figure 1A) specifically within the hinge domain ( Figure S1 A and B). To assess if JMJD6 interacts in vitro with metERa, we performed competition experiments adding the peptide containing R260 methylated or not within the reaction, however, it does not modify the interaction ( Figure S2). Next, we wished to validate the binding between endogenous JMJD6 and ERa. Because ERa is rapidly methylated after E 2 treatment, we performed reciprocal immunoprecipitation on MCF-7 cells treated with E 2 for the indicated times ( Figure 1B and Figure S3). Oestrogen treatment induced a rapid and transient interaction between ERa and JMJD6. Interestingly, the kinetics of interaction were reminiscent of those previously described for ERa methylation on R260 [8]. However, while the ERa methylation can vary between five and fifteen minutes after estrogen treatment along the experiments, the kinetics of interaction of ERa with JMJD6 were concomitant with the kinetic of ERa methylation, suggesting that JMJD6 could specifically interact with methylated ERa (metERa). Several additional findings supported this hypothesis: i) siRNA-mediated reduction of the arginine methyltransferase PRMT1 responsible for ERa methylation, impaired both ERa methylation and the ERa/JMJD6 interaction ( Figure 1C); ii) depletion of metERa by In vitro synthesized and radiolabelled ERa protein was incubated with GST-JMJD6 and binding was analysed by autoradiography. Luciferase was used as a negative control. The lower panel shows the corresponding coomassie staining. (B) MCF-7 were treated with E 2 for the indicated times and the interaction of JMJD6 with ERa was analysed by immunoprecipitation. On the same extract methylated ERa (metERa) was studied by immunoprecipitation with the anti-metERa revealed with an anti-ERa. (C) PRMT1 is essential to JMJD6/ERa interaction. Lysates of MCF-7 cells transfected with control siRNAs or with specific PRMT1 siRNAs were treated with vehicle or E 2 (10 28 M), analysed for ERa methylation and immunoprecipitated with the anti-JMJD6 antibody before analysis with the indicated antibodies. The amount of ERa, JMJD6 and PRMT1 in the different samples was determined by western blot in the input. (D) JMJD6 interacts specifically with metERa. MCF-7 treated with E 2 as above were lysed and analysed for JMJD6/ERa interaction. In parallel, cell extracts were immunoprecipitated with metERa antibody. Unbound fractions were then used for a second immunoprecipitation using anti-JMJD6 antibody followed by western blotting with anti-ERa and anti-JMJD6 antibodies. The amount of ERa and JMJD6 in the different samples was determined by western blot in the input. doi:10.1371/journal.pone.0087982.g001 Lysates of MCF-7 cells, control cells (shCtrl), and JMJD6 knockdown cells (shJMJD6) were immunoprecipitated with JMJD6 antibody. The immunoprecipitates were then blotted with antibodies against PI3K (p85), Src, ERa and JMJD6. The amount of ERa, Src, p85 (PI3K) and GAPDH in the different samples was determined by western blot. (B) Detection of endogenous interaction between JMJD6 and ERa by PLA. MCF-7 cells were incubated with E 2 10 28 M for 5 min. After fixation, in situ PLA for JMJD6/ERa was performed using antibodies raised against JMJD6 and ERa (panels a,b). The detected dimers are represented by red dots. The nuclei were counterstained with DAPI (blue) (Obj: X63). The lower panel shows immunoprecipitation using an antibody specifically recognizing the ERa methylated on R260, also disrupted the ERa/JMJD6 interaction ( Figure 1D); and finally, iii) analysis in other breast cancer cell lines showed that the interaction of ERa with JMJD6 was restricted to cells expressing metERa, reinforcing the idea that JMJD6 interacts specifically with metERa ( Figure S4).
Altogether these data clearly establish that contrarily to in vitro results showing the interaction between JMJD6 and ERa was estrogen independent, in cell lines JMJD6 interacts specifically with metERa upon E 2 stimulation.

JMJD6 Interacts with metERa within the Complex metERa/Src/PI3K
The methylation of ERa is a critical upstream signal, required to mediate the interaction of the receptor with Src and the p85 subunit of PI3K, which propagates the signal to downstream transduction cascades responsible for orchestrating cell proliferation and survival. Disappearance of the methylated ERa occurs concomitantly with the dissociation of the complex, leading to the extinction of downstream kinase activation [8].
In view of these earlier findings, we wondered whether JMJD6 might participate towards regulating the E 2 nongenomic signalling. To test this hypothesis, we performed co-immunoprecipitation experiments on E 2 -treated cells. We found that JMJD6 coprecipitated ERa with Src and PI3K. In addition, these interactions were clearly reduced in cells in which JMJD6 had been invalidated (shJMJD6) (Figure 2A). Interestingly, the kinetics of interactions were concomitant with the kinetics of ERa methylation suggesting that the JMJD6/metERa interaction occurs when the receptor is in complex with Src and PI3K ( Figure 2A). We then examined in which compartment this interaction occurred by Proximity Ligation Assay (PLA). Figure 2B shows that JMJD6 interacted with ERa (panel a) in the cytoplasm of MCF-7 cells, as indicated by the presence of red dots. The signals strongly decreased in MCF-7 cells in which expression of JMJD6 was downregulated ( Figure 2B, panel b). The interactions were quantified by counting the number of dots per cell ( Figure 2B, lower panel). Moreover, knockdown of PRMT1 strongly decreased JMJD6/ERa interaction analyzed by PLA ( Figure 2C), supporting that JMJD6 specifically interacts with methylated form of ERa which is localized in the cytoplasm of the cells [8]. The recruitment of JMJD6 seemed to be mainly mediated by ERa since JMJD6 could only weakly interact or not at all with Src or p85 of PI3K ( Figure S5A and B). However, we cannot exclude that other cofactors are important to recruit JMJD6 within the complex.
To assess whether JMJD6 affects the methylation of ERa on R260, we used JMJD6-knocked down cells (shJMJD6). After E 2 treatment, ERa methylation was strongly increased in shJMJD6 cells compared with in control cells ( Figure 2D). However, ERa methylation still decreased after 30 min of E 2 treatment probably due to the residual presence of JMJD6. Of note, this phenomenon occurred concomitantly with an increase in metERa/Src/PI3K complex ( Figure 2E).
Taken together, these data strongly suggest that JMJD6, by regulating the methylation of ERa, could play a role in the oestrogen rapid signalling pathway.

JMJD6 Preferentially Demethylates Asymmetric Dimethylation on Arginine Residues
In light of our finding that JMJD6 binds to metERa and regulates its methylation, we wondered whether JMJD6 could demethylate ERa in vitro. We performed a demethylation assay using recombinant JMJD6 on ERa methylated in vitro by PRMT1 as already described [8]. Using a specific antibody directed against ERa asymmetrically methylated on R260 ( Figure 3A left upper panel), we found that JMJD6 reduced the methylation of ERa R260 ( Figure 3A, right panel). We obtained the same demethylation results when we used [ 3 H]. AdoMet for the methylation assay and the methylation was visualized by autoradiography excluding possible artefactual antibody detection ( Figure 3B). However, as described by Bruick, its enzymatic activity remained faint, probably due to the need of a coactivator to potentiate its activity, as has been described for PRMT5 [15,16] and for the lysine demethylase LSD1 [17]. In the same type of assay, JMJD6 was not able to demethylate the peptide of ERa containing monomethylated K302 (K302me1) [18] ( Figure 3A, lower panel), thus highlighting the specificity of the catalytic activity of the recombinant protein.
Considering that histone H4 at arginine 3 (H4R3) is a substrate of JMJD6 [10] and that JMJD6 is able to demethylate ERa in vitro as shown above, we tested whether JMJD6 can remove dimethylation from histone H4 methylated in vitro by recombinant PRMT1. As shown in Figure 3C, JMJD6 was indeed able to decrease H4R3 ADMA in vitro. These results support the role of JMJD6 in the demethylation of the ADMA R3 residue on histone H4.
After validation of H4R3me2a antibody ( Figure S6), concordant results were obtained in our in vivo test in which, strikingly, knockdown of JMJD6 significantly increased H4R3me2a, detected by a specific antibody, when compared with the control cells ( Figure. 3D, compare lane 2 and 3 to lane 1), while the histone H4 level remained unchanged. Again, immunofluorescence analysis showed an enhanced staining of H4R3me2a in the nuclei of knocked down cells compared to control ( Figure 3E).
To provide further evidence for the demethylating function of JMJD6, we used two complementary approaches. Firstly, we analysed the global ADMA of total proteins in MCF-7 cells depleted for JMJD6. Cell lysates of cells invalidated for JMJD6 with three different shRNAs were immunoblotted with an antibody recognizing specifically ADMA (meRG). As shown in Figure 3F, knockdown of JMJD6 had a strong effect on the amount of dimethylated proteins (compare lanes 2-4 to lane 1). We confirmed these results with another antibody recognizing ADMA (ASYM24). However, the same cell extracts analyzed with an antibody directed against SDMA did not show any difference in the amount of methylated proteins, suggesting that JMJD6 may have more affinity for target asymmetric dimethylation ( Figure 3E). quantification of the number of signals per cell was performed by computer-assisted analysis as reported in the Materials and Methods. The mean +/2 s.e.m. of four experiments is shown. P-value was determined by Student's t-test. (C) JMJD6/ERa was analyzed by PLA as in B, in MCF-7 cells knocked down for PRMT1 (panel b) or control cells (panel a). The lower panel shows quantification of the number of signals per cell (D) ERa methylation was analysed after E 2 stimulation on control (shCtrl) and shJMJD6 cells (as described in Figure 1B). The amount of ERa, JMJD6 and GAPDH in the different samples was determined by western blot. (E) ERa methylation was analysed after E 2 stimulation on control and ShJMJD6 cells as in D and from the same extracts, immunoprecipitation with anti-src antibody was performed. The presence of ERa and p85 was then analysed. The amount of ERa, p85 and PRMT1 were determined in the inputs. The lower panel shows the relative quantification of metERa/ERa. doi:10.1371/journal.pone.0087982.g002 Of note, the overexpression of JMJD6 reduced global asymmetrical dimethylation assessed with ASYM24 antibody ( Figure S7).
Next, we identified putative JMJD6 partners by immunoprecipitation of MCF-7 transfected with JMJD6, coupled to mass spectrometry analysis. In agreement with data in the literature [14,19,20] most of the identified proteins were RNA binding proteins implicated in RNA processing. Interestingly, 23/25 proteins identified have been shown to be methylated on arginine residues (Table 1). Of interest, one of the identified putative partners of JMJD6 was the arginine methyltransferase CARM1, adding further evidence for its role in ADMA regulation. This interaction has been validated by immunoprecipitation ( Figure  S8).
Altogether, these data support the notion that JMJD6 has a dual enzymatic role, as arginine demethylase and hydroxylase, and participate in various regulation pathways. Significantly, our results have shown that JMJD6 may function as an arginine demethylase for histone and, for the first time, for a non-histone target. Specifically JMJD6 negatively regulates ERa nongenomic signalling. After E 2 treatment the methylation of ERa at R260 is crucial to propagate the signal to downstream transduction cascades. We therefore propose that the demethylase activity of JMJD6 is a decisive regulator of the rapid physiological responses to oestrogen, inducing the extinction of downstream kinases activation. Knowing that JMJD6 has been demonstrated to be a factor of poor prognosis in breast cancer [21], we could hypothesize that its demethylase activity could be involved in this effect maybe in part by regulating metERa/Src/PI3K/Akt signalling.
JMJD6 knockdown cells (shJMJD6A and B) were analysed by western blot with an antibody recognizing specifically the asymmetric methylation of arginine 3 of Histone H4 anti-H4R3me2a. Control western blotting was performed with anti-histone H4. The lower panel shows the relative quantification of H4R3me2a/H4. (E) H4R3 dimethylation was also analysed by immunofluorescence in the same cells as described above (panel a,b). DAPI staining is shown in panels c and d. (F) Lysates of control and shJMJD6 cells were analysed by western blotting with antibodies recognizing asymmetric dimethylation (MeRG and ASYM24) and symmetric dimethylation (SYM10 Cell Culture and Transfections MCF-7, ZR75-1 and Cama-1 mammary cells were maintained at 37uC in the appropriate medium supplemented with 10% foetal calf serum 1% non-essential amino acids. Prior to experiments, cells were grown for 48 hr in phenol red-free medium and 10% charcoal-treated serum (Biowest). MCF-7, ZR75-1 and Cama-1 cells come from ATCC. MCF-7 and ZR75-1 have been certified by CelluloNet, Lyon, France.

Antibodies
See Table 2.

JMJD6 Stable Knockdown Using Lentiviral Short Hairpin RNA
Three pre-made lentiviral short hairpin RNA (shRNA) constructs targeting human JMJD6 and one negative control construct created in the same vector backbone (pLKO.1-Puro) were purchased from SIGMA. Puromycin selection (1 mg/mL) was started 48 h after lentiviral infection.

GST-pull-down Assay
Psg5-ERa and pCDNA3-JMJD6 plasmid were transcribed and translated in vitro using T7-coupled reticulocyte lysate (Promega), in the presence of [ 35 S]methionine. Labelled proteins were incubated with 10 mg of purified recombinant GST-fusion proteins in 200 ml of binding buffer (Tris 20 mM pH 7.4, NaCl 0.1 M, EDTA 1 mM, glycerol 10%, Igepal 0.25%) with 1 mM DTT and 1% milk) for 2 h at room temperature. In competition experiments, peptide containing R260 of 1 mM of ERa methyl-ated or not was added to the reactions [8]. Beads were washed three times in binding buffer, and then bound proteins were resolved on SDS-polyacrylamide gel electrophoresis (PAGE), and visualized by autoradiography.

Affinity Purification and Mass Spectrometry
MCF-7 cells were transiently transfected with V5-tagged JMJD6. Cell extracts were purified with agarose beads coupled to V5 antibody and loaded on SDS-PAGE. After elution, proteins were digested by trypsin and peptides were analysed by LC/MS/ MS at IBCP, Lyon.

Immunoprecipitation and Western Blotting
To study the effect of oestrogen, the cells were treated for different times with E 2 (Sigma). After treatment, cells were lysed in RIPA buffer. Protein extracts were incubated with primary antibodies overnight at 4uC with shaking. Protein A-agarose or protein L-agarose beads were added and incubated 1 h at 4uC. The immunoprecipitates were separated on SDS-PAGE. The proteins were visualized by an enhanced chemiluminescence kit (Roche Molecular Biochemicals) following the manufacturer's instructions.
For histone extraction, cells were lysed in a solution containing 3% of SDS and 10% of b-mercaptoethanol for 2 min at 95uC, then cooled in ice. After addition of DNAase, the samples were loaded on SDS-PAGE.
Immunoblot images were digitized and quantified using the Image J Software.

Methylation and Demethylation Assays
The in vitro methylation protocol has been described previously [8]. The demethylation assay was performed as described for methylated histones [10]. We also used GST-hinge-ERa or histone H4 methylated by PRMT1 and purified on microspin columns to remove S adenosylmethionine (SAM) or a peptide corresponding to amino acids 297-308 non methylated or monomethylated on K302 [18]. Some experiments were per-

Proximity Ligation Assay
This technology developed by Olink Bioscience (Sweden) allows the visualisation of protein/protein interactions in situ and was first published in 2006 [23]. The experiments were performed following the manufacturer's instructions and have been previously described [9].

Image Acquisition and Analysis
The hybridized fluorescent slides were viewed under a Leica DM6000B microscope. Images were acquired under identical conditions at objective X63. For each sample, at least one hundred cells were counted. Analyses and quantifications of these samples were performed using Image J software.