JMJD6 Licenses ERα-Dependent Enhancer and Coding Gene Activation by Modulating the Recruitment of the CARM1/MED12 Co-activator Complex

SUMMARY

Whereas the actions of enhancers in gene transcriptional regulation are well established, roles of JmjC domain-containing proteins in mediating enhancer activation remain poorly understood. Here we report that recruitment of the JmjC domain-containing protein 6 (JMJD6) to estrogen receptor alpha (ERα)-bound active enhancers is required for RNA polymerase II recruitment and enhancer RNA production on enhancers, resulting in transcriptional pause release of cognate estrogen target genes. JMJD6 was found to interact with MED12 in the mediator complex to regulate its recruitment. Unexpectedly, JMJD6 is necessary for MED12 to interact with CARM1, which methylates MED12 at multiple arginine sites and regulates its chromatin binding. Consistent with its role in transcriptional activation, JMJD6 is required for estrogen/ERα-induced breast cancer cell growth and tumorigenesis. Our data have uncovered a critical regulator of estrogen/ERα-induced enhancer, coding gene activation and breast cancer cell potency, providing a potential therapeutic target of ER-positive breast cancers.

INTRODUCTION

The steroid hormone estrogens (17-β-estradiol, estradiol, E₂) play a vital role in various biology processes, such as normal mammary gland development, brain development, and reproduction (Couse and Korach, 1999). However, prolonged exposure to high levels of estrogen can lead to breast cancer by constitutively activating the transcription of genes predominantly implicated in metabolism and cell cycle regulation. Estrogen effects on normal mammary gland development and breast tumorigenesis are mediated by two receptors, estrogen receptor alpha (ERα) and beta (ERβ). Upon estrogen binding, ER undergoes a conformational change, translocates from cytosol to nucleus to bind to specific estrogen response elements (ERE), and regulates gene expression. Estrogen-dependent gene expression requires a highly coordinated and complex interplay between various transcription factors, epigenetic enzymes involved in post-translational modifications of histones, epigenetic readers and chromatin remodelers (Hervouet et al., 2013). For instance, bromodomain-containing protein 4 (BRD4) has been shown to play an important role in promoting estrogen-regulated transcription and proliferation of ER-positive breast cancer cells (Feng et al., 2014, Nagarajan et al., 2014, Nagarajan et al., 2015, Osmanbeyoglu et al., 2013). Despite the plethora of proteins that have been identified to play important roles in ER-positive breast cancers, a deeper understanding of the underlying molecular mechanisms is needed to uncover novel therapeutic targets and develop new drugs for treating ER-dependent breast cancers.

Enhancers, genomic regulatory elements described about forty years ago, are well established to play key roles in controlling regulated tissue-specific gene expression (Bulger and Groudine, 2011, Ong and Corces, 2011, Plank and Dean, 2014, Heintzman et al., 2009). Genomic features including histone modifications (e.g., H3K4me1/2, H3K27Ac), coactivators (e.g., CBP, p300, MED1, MED12), and an open chromatin architecture (e.g., DNase I hypersensitivity) have been identified as signatures of enhancers (Natoli and Andrau, 2012). Differential decoration of enhancers with these features in a given cell may define distinct classes of enhancers that specify distinct gene expression profiles and biological outcomes. ChIP-seq analysis of ERα under estrogen stimulation revealed that majority of its genomic binding sites localized on distal enhancers (Carroll et al., 2006, Li et al., 2013). More recently, several studies have shown that many enhancers are loaded with RNA Pol II and associated with the production of transcription, namely enhancer RNAs (eRNAs) (De Santa et al., 2010, Djebali et al., 2012, Hah et al., 2011, Kim et al., 2010, Lam et al., 2013, Melo et al., 2013, Natoli and Andrau, 2012, Wang et al., 2011, Melgar et al., 2011, Wu et al., 2014, Lai et al., 2013, Mousavi et al., 2013). eRNAs have been defined as short transcripts (50–2000 nucleotides) that are transcribed bi-directionally, or sometimes uni-directionally, from enhancer regions (Kim et al., 2010). Although whether eRNAs themselves are functional remains to be unequivocally proven, many studies have clearly demonstrated that enhancer transcription occurs before coding gene activation and may help to create an open chromatin environment, facilitate promoter and enhancer looping, and contribute to coding gene transcriptional regulation (Hsieh et al., 2014, Lai et al., 2013, Lam et al., 2013, Melo et al., 2013, Mousavi et al., 2013, Schaukowitch et al., 2014, Pnueli et al., 2015, Zhao et al., 2016, Li et al., 2013). In addition, enhancer transcription has been suggested to play an essential role in enhancer marker (H3K4me1/2) deposition at de novo enhancers (Kaikkonen et al., 2013). However, the molecular machinery that controls the appropriate transcriptional output of enhancers and how it participates in subsequent activation of coding genes remain elusive.

The large family of JmjC domain-containing proteins has been shown to be critical for gene transcription. One member of this family, the Jmjc domain-containing protein 6 (JMJD6), was originally identified as a phosphatidylserine receptor on the surface of phagocytes (Fadok et al., 2000, Wang et al., 2003, Bose et al., 2004). Subsequent studies demonstrated that it was localized in the nucleus of a cell, suggesting it might possess novel nuclear functions (Hahn et al., Cikala et al., 2004, Hahn et al., 2010, Cui et al., 2004, Tibrewal et al., 2007). JMJD6 was found to function as an iron (Fe²⁺)- and 2-oxoglutarate (2-OG)-dependent dioxygenase that demethylates methylated arginines as well as hydroxylates lysines on both histone and non-histone proteins (Chang et al., 2007, Liu et al., 2013, Poulard et al., 2014, Lawrence et al., 2014, Gao et al., 2015, Wu et al., 2015, Tikhanovich et al., 2015, Mantri et al., 2011, Unoki et al., 2013, Webby et al., 2009, Wang et al., 2014a). We recently reported a transcriptional paradigm in which JMJD6 regulates promoter-proximal Pol II pausing release in a distal manner (Liu et al., 2013). Besides its function in transcriptional control, we and others have shown that JMJD6 also interacts with multiple splicing factors, and is involved in gene splicing control (Heim et al., 2014, Webby et al., Liu et al., 2013, Rahman et al., 2011, Yi et al., 2017). JMJD6 has been implicated in a multitude of biological processes, including embryonic development (Bose et al., 2004, Li et al., 2003, Kunisaki et al., 2004), cell cycle control (Wang et al., 2014a), cellular proliferation and motility (Lee et al., 2012, Chen et al., 2014), adipocyte differentiation (Hu et al., 2015) and development of various types of cancers, such as breast (Poulard et al., 2015, Lee et al., 2012, Aprelikova et al., 2016), lung (Zhang et al., 2013) and colon cancer (Wang et al., 2014a).

The mediator is a large, multi-subunit complex that is conserved from yeast to humans (Malik and Roeder, 2010). The mammalian mediator complex comprises 30 individual subunits that are arranged in four modules called head, middle, tail and kinase modules (Malik and Roeder, Taatjes, 2010). MED12, a component in the kinase module, is located on X-chromosome and encodes a 2,177 amino acid (aa) protein. It has been shown that med12 gene is essential for early development in mouse and is involved in the transcriptional regulation of many signaling pathways (Philibert and Madan, 2007, Rocha et al., 2010). MED12 has been shown to be implicated in a number of neurological disorders (Xu et al., 2011, Philibert and Madan, 2007, Sandhu et al., 2003, Graham and Schwartz, 2013, Risheg et al., 2007, Schwartz et al., 2007, Ding et al., 2008) as well as human cancers (Schiano et al., 2014, Turunen et al., 2014, Huang et al., Shalem et al., 2014, Wang et al., 2015). Particularly, MED12 was linked to drug resistance in multiple cancer types including breast, colon, lung cancers and melanoma (Wang et al., 2015, Shalem et al., 2014, Huang et al., 2012). However, whether MED12 is involved in estrogen-induced transcriptional program and how its activity is regulated is not fully determined.

In the current study, we found that JMJD6 is specifically recruited onto ERα-bound active enhancers in response to estrogen stimulation, and is required for activation of these enhancers, including RNA Pol II recruitment and eRNA transcription, leading to transcriptional activation of cognate estrogen target genes. Using affinity purification, we revealed that JMJD6 interacts with MED12 in the mediator complex, and is required for MED12 recruitment onto ERα-bound active enhancers. Quantitative mass spectrometry (qMS) analysis revealed that, in the absence of JMJD6, MED12 interaction with CARM1 is significantly attenuated, which is found to methylate MED12 at its C-terminus at multiple arginine sites and is required for MED12 recruitment onto ERα-bound active enhancers. Consistent with its role in estrogen/ERα-induced transcriptional program, JMJD6 was found to serve as a critical regulator of breast cancer cell growth and tumorigenesis, with potential future therapeutic implications.

RESULTS

Estrogen induces JMJD6 binding on ERα-bound active enhancers

We mined breast cancer-linked gene expression data using the Gene Expression-Based Outcome for Breast Cancer Online (GOBO) tool, and found that high expression of JMJD6 was significantly associated with worse survival in estrogen receptor (ER)-positive breast cancer (Fig. S1A). This observation, in concert with our recent study demonstrating that JMJD6 regulates gene transcription through its actions on gene enhancers, prompted us to examine the possibility that it might exert critical roles in transcriptional programs regulated by ERα in breast cancer cells. We first examined its binding with chromatin in response to estrogen by using ChIP-seq (chromatin immunoprecipitation coupled with high throughput sequencing) in ERα-positive MCF7 breast cancer cells. MCF7 cells cultured in stripping medium for three days were treated with or without estrogen followed by ChIP-seq with anti-JMJD6 antibody. Consistent with our previous study in other cell lines in the absence of regulatory signals, the majority of JMJD6 binding sites without estrogen treatment were found to be localized at distal regions (non-promoter regions) (Fig. 1A and Fig. S1B). However, upon estrogen treatment, there were additional 629 JMJD6 binding sites that were strongly induced (fold induction (FC) > 2) (Fig. 1A–1D). The vast majority of these estrogen-induced JMJD6 binding sites (>90%) were localized at distal regions (non-promoter regions), which closely resembled that of ERα (Fig. 1D, 1E and Fig. S1C). Because estrogen effects in MCF7 cells were mainly mediated through ERα binding on distal enhancers, we first analyzed estrogen-induced JMJD6 binding sites to see their correlation with ERα binding and whether they harbor the classical estrogen response element (ERE). Motif analysis revealed that ERE was the most significant enriched motif (P = 1E-308) found in estrogen-induced JMJD6 binding sites (Fig. 1F). ERα binding was highly enriched on these sites upon estrogen treatment shown by heat map and tag density plot (Fig. 1G, the fourth column, and Fig. S1D, top panel on the right). We next assessed whether those estrogen-induced JMJD6 binding sites shared enhancer characteristics, including highly enriched H3K4me1/2, but low levels of H3K4me3. Heat map and tag density plots confirmed they were, indeed, ERα-bound enhancers (Fig. 1G, the fifth to the tenth columns, and Fig. S1D, the second to the fourth panels). Furthermore, it was reported recently that a group of enhancers, namely active enhancers, were essential for the transcriptional activation of estrogen-induced coding target genes(Li et al., 2013). These enhancers were decorated with H3K27Ac histone marker and co-activator protein p300, and associated with the highest levels of ERα binding in the presence of estrogen, but were devoid of repressive histone markers including H3K9me3 or H3K27me3. Both heat map and tag density plots revealed that H3K27Ac and p300 were indeed induced by estrogen (Fig. 1G, the eleventh and fourteenth panels, and Fig. S1D, the fifth and sixth panels), and ERα levels were much higher on these estrogen-induced JMJD6 binding sites compared with that on all ERα sites (Fig. S1D, the top two panels). Repressive histone markers, both H3K9me3 and H3K27me3, were found to be absent (Fig. 1G, the fifteenth to eighteenth panels, and Fig. S1D, the seventh and eighth panels). Binding of JMJD6, ERα, H3K4me1/2/3, H3K27Ac, p300, H3K9me3 and H3K27me3 on representative active enhancers were shown, such as the ones nearby classical estrogen-induced coding genes, FOXC1, SIAH2, GREB1 and SMAD7 (Fig. 1H, 1I and Fig. S1E, S1F). Estrogen-induced JMJD6 binding sites identified by ChIP-seq were validated by ChIP-qPCR in MCF7 cells transfected with control siRNA or two independent siRNAs targeting JMJD6 (Fig. S1G). The knockdown efficiency of JMJD6 was examined at mRNA and protein level through RT-qPCR and immunoblotting analysis, respectively (Fig. S1H and S1I). It should be noted that JMJD6 oligomerization remained intact even under denature conditions, which was consistent with previous reports (Liu et al., 2013, Yi et al., 2017, Han et al., 2012). Unlike JMJD6 binding sites being induced robustly, those 133 JMJD6 binding sites lost in response to estrogen treatment were either with low tag density at both conditions, with or without E₂ treatment, but however were falsely predicted as peaks, or the reduction of tag density were very subtle.

Figure 1. JMJD6 binding was induced by estrogen on ERα-bound active enhancers.

(A) MCF7 cells treated with or without estrogen (E₂, 10⁻⁷ M, 1 hr) were subjected to ChIP-seq with anti-JMJD6 specific antibody, and overlapping between JMJD6 ChIP-seq binding sites in the presence or absence of estrogen was shown by venn diagram (FC>2).

(B) JMJD6 ChIP-seq tag distribution, both with and without estrogen, centered on estrogen-induced JMJD6 sites (± 3,000 bp).

(C) Box plots displaying the change of tag density induced by estrogen shown in (B).

(D) Heat map representation of JMJD6 ChIP-seq tag density centered on estrogen-induced JMJD6 sites (± 3,000 bp). JMJD6 ChIP-seq binding sites were first categorized as 3’ UTR (untranslated regions) exons, 5’ UTR exons, CDS (coding sequencing) exons, intergenic, introns, non-coding, promoter (TSS, transcription start site) and TTS (transcription termination site) based on genomic location, then ChIP-seq binding sites were rank ordered from high to low tag density in each category.

(E) Genomic distribution of estrogen-induced JMJD6 sites.

(F) Motif analysis for estrogen-induced JMJD6 sites (± 100 bp from the center of ChIP-seq binding sites).

(G) Heat map representation of JMJD6, ERα, H3K4me1, H3K4me2, H3K4me3, H3K27Ac, p300, H3K9me3 and H3K27me3 ChIP-seq tag density in the presence or absence of estrogen centered on estrogen-induced JMJD6 sites (±3,000 bp).

(H, I) Genome browser views of JMJD6, ERα, H3K4me1, H3K4me2, H3K4me3, H3K27Ac, p300, MED1, MED12, H3K9me3 and H3K27me3 ChIP-seq in the presence or absence of estrogen on selected active enhancer regions, such as the ones nearby estrogen-induced coding gene FOXC1 (H) and SIAH2 (I), were shown. Boxed regions indicated active enhancers.

JMJD6 is a determinant of transcriptional activation of ERα-bound active enhancers

Several studies reported recently that enhancer RNAs (eRNAs) are involved in transcriptional regulation of nearby coding genes (Lam et al., 2013, Kim et al., 2010, Melo et al., Lai et al., 2013, Mousavi et al., 2013, Hsieh et al., 2014, Schaukowitch et al., 2014, Pnueli et al., 2015, Zhao et al., 2016, Li et al., 2013). For instance, estrogen-induced eRNAs from ERα-bound active enhancers were critical for the transcriptional activation of cognate estrogen-induced coding genes (Li et al., 2013). Our observation that binding of JMJD6 is induced on ERα-bound active enhancers prompted us to examine whether JMJD6 is required for enhancer activation (i.e. eRNA production), and therefore estrogen-induced cognate coding gene activation. To this end, Gro-seq (global run-on coupled with high throughput sequencing), which has been proven to be robust in detecting eRNAs, was performed in MCF7 cells transfected with control siRNA or siRNA specifically against JMJD6 in the presence or absence of estrogen treatment. It was found that eRNAs were dramatically induced bi-directionally on those estrogen-induced JMJD6 binding sites upon estrogen treatment (Fig. 2A and 2B), whereas there was nearly no induction of eRNA expression on those ERα-bound enhancers without JMJD6 co-occupancy (Fig. S2A and S2B) (compare siCTL (CTL) (+) to siCTL (E₂) (+) and siCTL (CTL) (−) to siCTL (E₂) (−)), suggesting JMJD6 might be a determinant for estrogen-induced eRNA production. To support this notion, eRNA production was attenuated on those estrogen-induced JMJD6 binding sites when JMJD6 was knocked down (Fig. 2A and 2B). Significance of the change of eRNA induction by estrogen and by JMJD6 knockdown was demonstrated by box plot (Fig. S2C and S2D). Serving as a control, JMJD6 knockdown had no significant impact on eRNA production on those ERα-bound enhancers without JMJD6 co-occupancy (Fig. S2A and S2B). Furthermore, when we classified estrogen-induced JMJD6 binding sites based on ChIP-seq tag intensity into three groups, high, medium and low, it was found that eRNA induction was positively correlated with JMJD6 tag intensity (Fig. 2C). Our Gro-seq experiments were highly reproducible between replicates (Fig. S2E), and the effects of JMJD6 on eRNA production were well correlated based on the siRNA described above and a second independent one targeting JMJD6 (Fig. S2F).

Figure 2. JMJD6 is required for transcriptional activation of ERα-bound active enhancers.

(A) MCF7 cells were transfected with control siRNA (siCTL) or siRNA specific against JMJD6 (siJMJD6), and treated with or without estrogen (E₂, 10⁻⁷ M, 1 hr) followed by Gro-seq. Gro-seq tag distribution, both sense (+) and anti-sense (−), centered on estrogen-induced JMJD6 sites were shown (±5,000 bp).

(B) Heat map representation of Gro-seq tag density as shown in (A).

(C) Estrogen-induced JMJD6 sites were divided into three groups, high, medium and low, based on ChIP-seq tag density, and Gro-seq tag distribution, both sense (+) and anti-sense (−), centered on estrogen-induced JMJD6 sites, were shown (±5,000 bp).

(D) MCF7 cells were transfected with control siRNA (siCTL) or two independent siRNAs specific against JMJD6 (siJMJD6), and treated with or without estrogen (E₂, 10⁻⁷ M, 6 hr), followed by RNA extraction and RT-qPCR analysis to examine the expression of selected enhancer RNAs (eRNAs) nearby estrogen-induced coding genes as indicated (± s.e.m., *P<0.05, **P<0.01, ***P<0.001).

(E) Wild type (WT) and JMJD6 knockout (KO) MCF7 cells were maintained in stripping medium for three days before treating with or without estrogen (E₂, 10⁻⁷ M, 6 hr), followed by RNA extraction and RT-qPCR analysis to examine the expression of selected enhancer RNAs (eRNAs) nearby estrogen-induced coding genes as indicated (± s.e.m., *P<0.05, **P<0.01, ***P<0.001).

(F) MCF7 cells were transfected with control siRNA (siCTL) or siRNA specific against JMJD6 (siJMJD6), and treated with or without estrogen (E₂, 10⁻⁷ M, 1 hr) followed by ChIP-seq with anti-Pol II specific antibody. Pol II ChIP-seq tag distribution centered on estrogen-induced JMJD6 sites (±6,000 bp) was shown.

(G, H) Heat map (G) and box plots (H) representation of Pol II ChIP-seq tag density centered on estrogen-induced JMJD6 sites.

To validate JMJD6 effects on eRNA production, total RNA extracted from MCF7 cells transfected with control siRNA or two independent siRNAs specifically against JMJD6 in the presence or absence of estrogen treatment were subjected to RT-qPCR analysis using primers specifically targeting several ERα and JMJD6 co-bound active enhancers nearby estrogen-induced coding genes, such as FOXC1, SIAH2, GREB1, NRIP1 and SMAD7. Consistent with our observation from meta-analysis (Fig. 2A), eRNA production from these active enhancers was induced by estrogen treatment, which was attenuated by JMJD6 knockdown (Fig. 2D). Furthermore, JMJD6 effects on eRNA production were confirmed in JMJD6 knockout MCF7 cells (Fig. 2E), which was generated by CRISPR (clustered, regularly interspaced, short palindromic repeats)/Cas9-mediated gene editing technology (Fig. S2G and S2H).

Transcription of eRNA has been shown to be driven by RNA Polymerase II (Pol II). We thus tested whether JMJD6 is required for Pol II recruitment onto active enhancers. Pol II ChIP-seq was performed in MCF7 cells transfected with control siRNA or siRNA specifically targeting JMJD6 in the presence or absence of estrogen. As expected, estrogen induced Pol II recruitment onto those JMJD6/ERα-co-bound active enhancers, and knockdown of JMJD6 attenuated estrogen effects, which were shown by both tag density plots and heat map analysis (Fig. 2F and 2G). Statistical test for the change of Pol II recruitment by estrogen and JMJD6 was performed (Fig. 2H). Serving as a control, Pol II binding was not affected on ERα-bound enhancers that did not exhibit JMJD6 co-occupancy (Fig. S2I and S2J). Taken together, our data suggested that JMJD6 is a determinant for transcriptional activation of ERα-bound active enhancers, involving in estrogen-induced Pol II recruitment and eRNA production.

JMJD6 is required for estrogen-induced coding gene transcriptional activation

ERα-bound active enhancers and associated eRNAs have been reported to be required for cognate coding gene transcriptional activation (Li et al., 2013). We therefore asked whether JMJD6 is required for estrogen-induced coding gene activation based on our Gro-seq described above in MCF7 cells. Of a large cohort of 1,108 genes that were induced by estrogen (FC>1.5) (Fig. 3A), expression of 731 of these genes was attenuated following knockdown of JMJD6, representing 66% of all estrogen-induced genes (Fig. 3B). These 731 genes were referred to as estrogen-induced and JMJD6-dependent genes. The expression of these genes in the presence of estrogen in control and JMJD6 knockdown conditions was shown by heat map (Fig. 3C and 3D). JMJD6 effects on representative estrogen-induced coding gene transcriptional activation were confirmed by RT-qPCR analysis in MCF7 cells with JMJD6 knockdown and knockout (Fig. 3E and 3F).

Figure 3. JMJD6 regulates estrogen-induced coding gene transcription.

(A) MCF7 cells were transfected with control siRNA (siCTL) or siRNA specific against JMJD6 (siJMJD6), and treated with or without estrogen (E₂, 10⁻⁷ M, 1 hr) followed by Gro-seq. Pie chart showed genes positively- and negatively-regulated by estrogen based on Gro-seq (FC (siCTL (E₂)/siCTL (CTL))≥1.5).

(B) Venn diagram showing genes induced by estrogen and dependent on JMJD6 for expression based on Gro-seq as described in (A) (FC (siCTL (E₂)/siJMJD6 (E₂))≥1.5).

(C, D) Heat map (C) and box plots (D) representation of the expression levels (RPKM) for genes induced by estrogen and dependent on JMJD6 in the presence of estrogen as described in (B) in both control siRNA (siCTL) and JMJD6 siRNA (siJMJD6) transfected conditions.

(E) MCF7 cells were transfected with control siRNA (siCTL) or two independent siRNAs specific against JMJD6 (siJMJD6), and treated with or without estrogen (E₂, 10⁻⁷ M, 6 hrs), followed by RNA extraction and RT-qPCR analysis to examine the expression of selected estrogen-induced coding genes as indicated (± s.e.m., **P<0.01, ***P<0.001).

(F) WT and JMJD6 (KO) MCF7 cells were maintained in stripping medium for three days before treating with or without estrogen (E₂, 10⁻⁷M, 6 hr), followed by RNA extraction and RT-qPCR analysis to examine the expression of selected estrogen-induced coding genes as indicated (± s.e.m., **P<0.01, ***P<0.001).

(G) Traveling ratio (TR) distribution calculated based on Gro-seq for genes induced by estrogen and dependent on JMJD6.

(H) Box plots displaying the change of TR as shown in (G).

(I) Gro-seq tag distribution, both sense (+) and anti-sense (−), centered on TSSs (transcription start sites) of genes induced by estrogen and dependent on JMJD6 (± 6,000 bp).

We recently reported a paradigm in gene transcriptional regulation, in which JMJD6 localized at distal enhancers to regulate promoter-proximal pausing release and transcriptional elongation of cognate coding genes (Liu et al., 2013). Therefore, we investigated whether those estrogen-induced and JMJD6-dependent genes experience promoter-proximal pausing regulation, and the role of JMJD6 in this process. Promoter-proximal pausing index (Zeitlinger et al., 2007) or traveling ratio (TR) (Reppas et al., 2006) is defined as the relative ratio of Pol II density in the promoter-proximal region and the gene body, which can be calculated based on Gro-seq as it detects transcripts generated in nuclear run-on reactions by RNA Pol II that is engaged in and competent for transcription, and it can precisely distinguish paused Pol II from backtracked and arrested Pol II (Adelman et al., 2005, Core et al., 2008). Based on our Gro-seq data, vast majority of genes induced by estrogen and dependent on JMJD6 experience promoter-proximal pausing, which was released upon estrogen treatment (i.e., TR was decreased upon estrogen treatment) (Fig. 3G). Importantly, knockdown of JMJD6 abolished estrogen-induced pausing release (Fig. 3G). The significance of the TR change caused by estrogen treatment and the impact of JMJD6 on such change was visualized by box plot analysis (Fig. 3H). Close examination of the tag density distribution surrounding transcription start sites (TSSs) (6kb upstream and downstream of TSS) revealed that there was a decrease of tag density at promoter-proximal region, but an increase along the gene body upon estrogen treatment for genes induced by estrogen and dependent on JMJD6, resembling a typical pause release (Fig. 3I). Importantly, knockdown of JMJD6 attenuated estrogen effects on tag density distribution/pausing release (Fig. 3I). Taken together, our data suggested that JMJD6 regulates promoter-proximal pausing release and transcriptional activation of a large set of estrogen-induced coding genes.

JMJD6 regulates MED12 function in estrogen-induced transcriptional activation

To further explore the molecular mechanisms underlying JMJD6 regulation of estrogen-induced transcription program, we purified JMJD6-associated proteins in an inducible stable MCF7 cell line expressing JMJD6 in the presence of estrogen. First, cells extracts were prepared in relatively low salt concentration (see materials and methods) and subjected to affinity purification followed by mass spectrometry (MS) analysis. Strikingly, 23 out of 30 subunits in the mediator complex were pulled down by JMJD6, with one of the subunits, MED12, had the most unique peptides identified (Fig. S3A). To further investigate which subunit might directly interact with JMJD6 on chromatin to regulate estrogen-induced gene transcriptional activation, chromatin-bound fraction (pellet) was extracted and subjected to affinity purification followed by MS analysis. It was found that MED12 was the only subunit in the mediator complex that still remains associated with JMJD6. MED12 has been implicated in the transcriptional regulation of a variety of signaling pathways (Philibert and Madan, 2007). We therefore focused on investigating the potential physical and functional relationship between JMJD6 and MED12 in estrogen-induced gene transcriptional activation.

JMJD6 and MED12 interaction was confirmed by affinity purification, as described above, followed by immunoblotting with anti-MED12 specific antibodies (Fig. 4A). JMJD6 interaction with MED12 was further confirmed by exogenously and endogenously expressed proteins, which was not altered by estrogen treatment (Fig. 4B and Fig. S3B). The interaction between JMJD6 and MED12 appeared to be direct because purified in vitro-expressed proteins were also found to be associated examined by surface plasmon resonance (SPR) assay (Fig. S3C). We next sought to test whether MED12 is involved in estrogen-induced gene transcriptional activation, as was the case for JMJD6. MCF7 cells were transfected with control siRNA or siRNA specifically against JMJD6 or MED12 in the presence or absence of estrogen followed by RNA-seq. It was found that JMJD6 and MED12 exerted similar effects on estrogen-induced transcriptional activation (Fig. S3D). More specifically, the expression of 61% and 69% of estrogen-induced genes were attenuated following JMJD6 and MED12 knock-down, respectively (FC>1.5), with the vast majority of affected genes overlapped (Fig. 4C). The impact of JMJD6 or MED12 on the expression of those affected genes was shown by heat map (Fig. 4D), and statistical test was performed (Fig. 4E). Our RNA-seq experiments were highly reproducible between replicates (Fig. S3E). Effects of MED12 on representative estrogen-induced target genes were confirmed by RT-qPCR analysis with two independent siRNAs targeting MED12 (Fig. 4F). The enhancer transcriptional activation/eRNA production was similarly affected by MED12 (Fig. 4G). The knockdown efficiency of MED12 was examined by RT-qPCR and immunoblotting analysis (Fig. S3F and S3G). To further explore the functional connection between JMJD6 and MED12, we tested whether MED12 is similarly recruited onto active enhancers, which might potentially confer MED12 function in estrogen-induced gene transcriptional activation, and whether its recruitment is regulated by JMJD6. To this end, MCF7 cells were transfected with control siRNA or siRNA specifically targeting JMJD6, in the presence or absence of estrogen, followed by ChIP-seq analysis with anti-MED12 specific antibody. MED12 binding was found to be significantly increased on those enhancers exhibiting estrogen-induced JMJD6 binding upon estrogen treatment, which was significantly attenuated following JMJD6 knockdown (Fig. 4H–4J). JMJD6’s impact on MED12 binding was validated by using two independent siRNAs targeting JMJD6 (Fig. S3H). Knockdown of JMJD6 did not change MED12 expression, as examined by RT-qPCR and immunoblotting analysis (Fig. S3I and S3J). Furthermore, MED12 ChIP-seq peaks identified were specific as MED12 binding was significantly impaired when MED12 was knocked down on selected enhancer regions, but not a negative control region GAPDH promoter (Fig. S3K). Our data suggested that JMJD6 exhibits an unexpected role in the recruitment of MED12 to ERα-bound active enhancers to regulate estrogen-induced transcriptional activation.

Figure 4. JMJD6 regulates MED12 function in estrogen-induced transcriptional activation.

(A) Inducible stable MCF7 cells expressing Flag-tagged JMJD6 were treated with doxycycline (Dox) (+: 0.5 μg/mL; ++: 1 μg/mL) for 48 hrs and subjected to immunoprecipitation (IP) with anti-Flag or anti-HSP60 antibody followed by immunoblotting (IB) with antibodies as indicated. HSP60 served as a negative control, exhibiting no interaction with JMJD6. N/A: empty lane.

(B) HEK293T cells transfected with Myc-tagged JMJD6 together with Flag-tagged MED12 or BCL2 were subjected to immunoprecipitation (IP) with anti-myc antibody followed by immunoblotting (IB) with antibodies as indicated. BCL2 served as a negative control, exhibiting no interaction with JMJD6.

(C) MCF7 cells were transfected with control siRNA (siCTL), siRNA specific against JMJD6 (siJMJD6) or MED12 (siMED12), and treated with or without estrogen (E₂, 10⁻⁷ M, 6 hr) followed by RNA-seq analysis. Estrogen-positively regulated genes (FC (siCTL (E₂)/siCTL (CTL))>1.5) which were dependent on both JMJD6 and MED12 were shown by Pie chart.

(D, E) Heat map (D) and box plots (E) representation of the expression levels (RPKM) for genes induced by estrogen and dependent on JMJD6 and MED12 in the presence of estrogen as described in (C) in control siRNA (siCTL), JMJD6 siRNA (siJMJD6) and MED12 siRNA (siMED12) transfected conditions.

(F, G) MCF7 cells were transfected with control siRNA (siCTL) or two independent siRNAs specific against MED12 (siMED12), and treated with or without estrogen (E₂, 10⁻⁷ M, 6 hr) followed by RT-qPCR analysis to examine the expression of selected estrogen-induced coding genes (F) and cognate enhancer RNAs (eRNAs) (G) as indicated (± s.e.m., ***P<0.001).

(H) MCF7 cells were transfected with control siRNA (siCTL) or siRNA specific against JMJD6 (siJMJD6), maintained in stripping medium for three days before treating with or without estrogen (E₂, 10⁻⁷ M, 1 hr) followed by ChIP-seq with anti-MED12 specific antibody. MED12 ChIP-seq tag distribution centered on estrogen-induced JMJD6 sites (± 3,000 bp) was shown.

(I, J) Heat map (I) and box plot (J) representation of MED12 ChIP-seq tag density centered on estrogen-induced JMJD6 sites (± 3,000 bp). MED12 ChIP-seq binding sites were first categorized as 3’ UTR exons, 5’ UTR exons, CDS exons, intergenic, introns, non-coding, promoter (TSS) and TTS based on genomic location, then ChIP-seq binding sites were rank ordered from high to low tag density in each category.

JMJD6 is required for MED12 association with CARM1

We next sought to explore how JMJD6 regulates MED12 association with these enhancers. We hypothesized that JMJD6 might regulate MED12 interaction with other proteins and/or MED12 post-translational modifications, and therefore its association with ERα-bound active enhancers. To this end, we first identified the JMJD6-interacting domain in MED12. Interaction assays with MED12 truncations revealed that JMJD6 specifically interacted with C-terminus of MED12, which has been shown to interact with many other known MED12 partners (Philibert and Madan, 2007) (Fig. 5A and 5B). The interaction between JMJD6 and C-terminus of MED12 was further confirmed by SPR assay (Fig. S4A). To search for the protein(s) interaction with MED12 that might be altered in the absence of JMJD6, wild type (WT) or JMJD6 knockout cells were subjected to SILAC labeling, infected with Flag-tagged C-terminus of MED12 in the presence of estrogen, pooled and followed by affinity purification and mass spectrometry (MS) analysis (Fig. 5C and 5D). In addition to known MED12 interacting partners, such as REST(Ding et al., 2008), RCOR1(Ding et al., 2008), Catenin (alpha, beta and delta) (Rocha et al., 2010, Kim et al., 2006), BRD4 (Bhagwat et al., 2016) and CARM1/PRMT4 (Chen et al., 1999), JMJD6 was also pulled down by MED12, independently confirming JMJD6 and MED12 interaction. Quantitative analysis revealed that the most affected interacting partner with MED12 was CARM1, which is a member in the protein arginine methyltransferase family (Chen et al., 1999). The ratio of the abundance of CARM1 in MED12 pull-downs from wild type (light label, 184.2) and JMJD6 knockout cells (heavy label, 15.8) was 11.658. Decreased binding affinity between CARM1 and MED12 in the absence of JMJD6 was confirmed by immunoblotting analysis (Fig. 5E, top panel). Furthermore, the interaction between CARM1 and MED12 was demonstrated by using endogenous proteins, which was not altered by estrogen treatment (Fig. S4B).

Figure 5. JMJD6 regulates MED12 interaction with CARM1 and hence MED12 methylation and chromatin binding.

(A) Schematic representation of the domain architecture of MED12 protein. Leucine-rich (L) domain (light green); Leucine-serine-rich (LS) domain (yellow); Proline-glutamine-leucine (PQL) domain (light blue); Poly-glutamine (Opa) domain (purple).

(B) HEK293T cells transfected with Myc-tagged JMJD6 and Flag-tagged MED12 (1–597), MED12 (598–964), MED12 (959–1718) or MED12 (1616–2177) were subjected to immunoprecipitation (IP) with anti-Myc antibody followed by immunoblotting (IB) with antibodies as indicated.

(C) Schematic representation of the protocol applied for detecting differential binding proteins and post translational modifications (PTMs) of C-terminus of MED12 (1616–2177) in wild type (WT) or JMJD6 (KO) MCF7 cells. Wild type (WT) and JMJD6 (KO) MCF7 cells were subjected to SILAC labeling and then infected with lenti-viral vectors expressing Flag-HA-tagged MED12 C-terminus (1616–2177). Cells were then lysed and cell lysates were pooled and subjected to affinity purification using anti-Flag M2 agarose followed by mass spectrometry (MS) analysis.

(D) Cell lysates as described in (C) were subjected to immunoblotting (IB) analysis with antibodies as indicated.

(E) Cells lysates as described in (C) were subjected to immunoprecipitation (IP) with anti-Flag antibody followed by immunoblotting (IB) with antibodies as indicated.

(F) In vitro methylation assay was performed by mixing purified bacterially-expressed CAgbRM1 with Flag-tagged MED12 truncations purified from over-expressed HEK293T cells (upper panel). The expression of MED12 truncations was examined by immunoblotting using anti-Flag antibody (bottom panel).

(G) In vitro methylation assay was performed by mixing purified bacterially-expressed MED12 C-terminus (1616–2177) with Flag-tagged PRMTs purified from over-expressed HEK293T cells (upper panel). The expression of all PRMTs was examined by immunoblotting using anti-Flag antibody and indicated by white arrows (bottom panel). Star indicates methylation of MED12.

(H) Methylated arginine residues which methylation decreased upon JMJD6 depletion were identified following the protocol as described in (C), which were shown as indicated.

(I) MCF7 cells were transfected with control siRNA (siCTL) or siRNA specific against CARM1 (siCARM1), and treated with or without estrogen (E₂, 10⁻⁷ M, 1 hr) followed by ChIP with anti-MED12 specific antibody. Fold induction of MED12 binding in response to estrogen treatment on selected active enhancer regions, such as the ones nearby estrogen-induced coding gene FOXC1, SIAH2 and GREB1, were shown (± s.e.m., **P<0.01; N.S: not significant). P: promoter; E: enhancer.

MED12 has recently been shown to be methylated by CARM1 at its C-terminus (Wang et al., 2014b), which was required for MED12 binding with chromatin and transcriptional regulation, sensitizing breast cancer cells to chemotherapy drugs (Wang et al., 2015). Based on our quantitative MS analysis above, we hypothesized that JMJD6 might regulate MED12 through modulating CARM1 binding with MED12, and therefore CARM1-mediated MED12 methylation and chromatin binding. To test this hypothesis, we first confirmed previous findings that CARM1 methylated MED12 at its C-terminus (aa 1616–2177) (Fig. 5F). Furthermore, C-terminus of MED12 was found to be exclusively methylated by CARM1 out of all eleven members of the protein arginine methyltransferase (PRMT) family tested (Fig. 5G). The activities of all PRMTs have been described in our previous report (Gao et al., 2015). Next, we attempted to identify the arginine methylation sites at C-terminus of MED12, which are presumably catalyzed by CARM1, and examine the change of such methylation in response to JMJD6 depletion using the MS data collected above in both wild type and JMJD6 knockout cells. MED12 was found to be heavily methylated at multiple arginine sites in its C-terminus, including arginine 1854 mono- and di-methylation (R1854me1/2), R1859me1/2, R1862me1/2, R1871me1/2, R1899me1/2, R1994me1/2 and R2015me1 (Fig. S4C and Supplementary Table 1). More importantly, the levels of R1854me1, R1871me1/2 and R1899me1/2 were decreased significantly in JMJD6 knockout cells (Fig. 5H and Supplementary Table 1), although events at R1854me2, R1859me1/2, R1862me1/2, R1994me1 and R2015me1 could not be accurately quantified. Decrease of arginine methylation on MED12 C-terminus was confirmed by immunoblotting using anti-H3R17me2 (a) antibody, which could largely recognize the methylated substrates for CARM1 (Fig. 5E, middle panel). MED12 C-terminus was equally pulled down in both wild type and JMJD6 knockout cells (Fig. 5E, bottom panel). We then examined whether knockdown of CARM1 would affect MED12 recruitment to ERα-bound active enhancers. A significant decrease of MED12 binding was observed when cells were transfected with siRNA specifically targeting CARM1, such as enhancer regions nearby FOXC1, SIAH2 and GREB1 (Fig. 5I). The knockdown efficiency of CARM1 was examined at mRNA and protein level through RT-qPCR and immunoblotting analysis, respectively (Fig. S4D and S4E). In consistency with its role in controlling MED12 binding, CARM1 was found to be required for selected estrogen-induced transcriptional activation of enhancer and cognate coding genes (Fig. S4F and S4G). Taken together, our data suggested that JMJD6 modulates MED12 binding with CARM1, which methylates MED12 and is required for MED12 recruitment onto JMJD6 and ERα co-occupied active enhancers.

JMJD6 is required for estrogen-induced breast cancer cell growth and tumorigenesis.

Due to its critical role in estrogen/ERα-induced gene transcriptional activation, we then tested whether JMJD6 regulates estrogen/ERα-induced breast cancer cell growth and tumorigenesis. Using MCF7 breast cancer cell line as a model system, we demonstrated that knockdown of JMJD6 decreased the proliferation rate of cells cultured in normal medium as well as in stripping medium followed by estrogen treatment (Fig. 6A and 6B). The effects of JMJD6 on MCF7 cell proliferation in normal medium as well as stripping medium followed by estrogen treatment were confirmed in JMJD6 knockout cells (Fig. 6C and 6D), which was further demonstrated by colony formation assay (Fig. 6E and 6F). We also found that JMJD6 knockout decreased cell migration significantly compared with control cells analyzed by wound healing assay (Fig. 6G–6J).

Figure 6. JMJD6 is required for estrogen-induced breast cancer cell growth and tumorigenesis.

(A, B) MCF7 cells were transfected with control siRNA (siCTL) or siRNA specifically targeting JMJD6 (siJMJD6) and maintained in normal growth medium (A) or stripping medium (phenol red free) for two days before treating with or without estrogen (E₂, 10⁻⁷ M) (B) followed by cell proliferation assay (± s.e.m., *P<0.05, **P<0.01, ***P<0.001).

(C, D) Wild type (WT) and JMJD6 knockout (KO) MCF7 cells were maintained in normal growth medium (C) or stripping medium for two days before treating with estrogen (E₂, 10⁻⁷M) (D) followed by cell proliferation assay (± s.e.m., *P<0.05, **P<0.01, ***P<0.001).

(E, F) Wild type (WT) and JMJD6 knockout (KO) MCF7 cells were maintained in normal growth medium (E) or stripping medium in the presence of estrogen (E₂, 10⁻⁷ M) (F) for 10 days for colony formation, and cell colonies were fixed and stained with crystal violet.

(G, I) Wild type (WT) and JMJD6 knockout (KO) MCF7 cells were grown to confluence in normal growth medium (G) or stripping medium before treating with estrogen (E₂, 10⁻⁷ M) (I) followed by wound-healing assay.

(H, J) Quantification of wound closure shown in panel (G) (H) and (I) (J) (± s.e.m., **P<0.01, ***P<0.001).

(K) Wild type (WT) and JMJD6 knockout (KO) MCF7 cells were injected subcutaneously into female BALB/C nude mice, and brushed with or without estrogen (E₂, 10⁻² M) on the neck every three days for six weeks. Mice were then euthanized and tumors were excised, photographed and weighted. (L) Significance test for tumor weight shown in (K) was performed (± s.e.m., ***P<0.001).

To test JMJD6 effects on tumor growth in vivo, we injected nude mice subcutaneously with control or JMJD6 knockout MCF7 cells, with or without estrogen administration. After six weeks, it was found that tumor volume was dramatically induced when mice were estrogen-treated compared to the control group. Importantly, JMJD6 knockout diminished the effects of estrogen-induced tumorigenesis (Fig. 6K–6L). We noted that body weight of mice was decreased slightly with estrogen treatment (Fig. S5A and S5B).

We next sought to examine whether JMJD6 regulation of estrogen/ERα-induced breast cancer cell growth is dependent on its enzymatic activity. To this end, control and JMJD6 knock out MCF7 cells were infected with or without control lentiviral vector and vectors expressing wild-type JMJD6 (WT) or its enzymatically deficient mutant containing a substitution of histidine 187 to alanine (H187A), followed by cell proliferation, colony formation and wound healing assays. It was found that JMJD6 (WT), but not JMJD6 (H187A), fully rescued the effects of JMJD6 depletion (Fig. S5C-S5F). JMJD6 (H187A) even exhibited a dominant-negative effect (Fig. S5C-S5F). Both JMJD6 (WT) and JMJD6 (H187A) expressed equally well as examined by immunoblotting (Fig. S5G). It should be noted that JMJD6 oligomerization was lost when histidine 187 was mutated to alanine, which was consistent with previous reports (Yi et al., 2017, Liu et al., 2013, Han et al., 2012). Taken together, our data suggested that JMJD6 is required for estrogen-induced MCF7 breast cancer cell and tumor growth, which is dependent on its enzymatic activity.

DISCUSSION

Enhancers critically regulate both the development of specific cell type, and the subsequent responses of their transcriptional programs by diverse signals, including ligands of nuclear receptors. One of the characteristics of ligand-activated enhancers is the production of eRNAs, which, directly or indirectly, exert functional roles in regulating their cognate coding gene transcription. However, the protein factors which govern eRNA production remain poorly characterized. Here we reported that JMJD6 is a critical regulator for estrogen/estrogen receptor (ER)-induced enhancer activation and coding target gene transcription based on its regulated recruitment to ERα-bound active enhancers, affecting both breast cancer cell growth and tumorigenesis. JMJD6 was found to be required for RNA Pol II recruitment and eRNA production on these enhancers, leading to transcriptional activation of cognate estrogen target genes. Mechanistically, JMJD6 was found to be required for MED12 recruitment, a component of the mediator complex, impacting estrogen-induced transcriptional activation.

A number of JmjC domain-containing proteins have been shown to play key roles in breast cancer growth, including LSD1/KDM1A, KDM2 subfamily, KDM3A, KDM4 subfamily, KDM5 subfamily, KDM6A, KDM6B and JMJD6 (Berry and Janknecht, 2013, Wang et al., 2009, Ramadoss et al., 2017, Taylor-Papadimitriou and Burchell, 2017, Kwok et al., 2017). Here we focused on investigating JMJD6 function in breast cancer enhancer function based on our previous observations that JMJD6 binding at enhancers plays a critical role in transcriptional pause release. Consistent with our previous finding that it is mainly localized on distal enhancers, JMJD6 was found to be specifically recruited onto ERα-bound active enhancers upon estrogen stimulation. Importantly, JMJD6 appeared to be a critical determinant for enhancer activation, including RNA Pol II recruitment and eRNA transcription, and knockdown of JMJD6 attenuated the transcriptional activation of vast majority of estrogen-induced coding genes. Specifically, JMJD6 was required for transcriptional pausing release of estrogen-induced coding genes, presumably through its interaction with P-TEFb complex (Liu et al., 2013). To further explore the molecular mechanisms underlying JMJD6 regulation of enhancer and cognate coding gene transcriptional activation, we purified JMJD6-associated proteins in the presence of estrogen. To our surprise, besides BRD4, which is required for JMJD6 binding with chromatin to regulate transcriptional pause release (Liu et al., 2013), 23 out of 30 subunits in the mediator complex were recovered in our purification. Further extraction with high salt from chromatin fraction revealed that MED12 was likely the subunit directly interacted with JMJD6. The role of BRD4 in estrogen-induced enhancer and coding gene transcriptional activation has been recently described (Nagarajan et al., 2014). As might be predicted from their physical interactions, JMJD6 and MED12 co-regulated a large program of estrogen-induced genes, and JMJD6 was required for MED12 to bind to ERα-bound active enhancers. Requirement of JMJD6 for MED12 to bind with chromatin led us to explore whether JMJD6 might regulate MED12 affinity with its associated proteins and/or MED12 post-translational modifications. Quantitative MS analysis revealed that one of the most dramatically altered MED12 binding protein upon JMJD6 knockdown was CARM1, which has been shown to be involved in estrogen-induced transcriptional activation (Al-Dhaheri et al., 2011, Lupien et al., 2009). Intriguingly, MED12 was recently shown to be exclusively methylated by CARM1 and such methylation was required for MED12 binding with chromatin, sensitizing breast cancer cells to chemotherapy drugs (Wang et al., 2015). We investigated the JMJD6-dependent alterations in MED12 methylation by quantitative MS analysis. Seven arginine residues (R1854, R1859, R1862, R1871, R1899, R1994 and R2015) were found to be methylated with high confidence at the C-terminus of MED12, which nearly covered all the methylation sites, except R1910 and R1912, currently reported in a comprehensive protein post-translational modification database (phosphosite.org). Further analysis revealed that the levels of methylation on several sites (R1854, R1871 and R1899) were significantly reduced in the absence of JMJD6, consistent with the notion that MED12 interaction with CARM1 was attenuated in the absence of JMJD6. It should be noted that, despite the observation that arginine methylation on only three sites was found to be decreased, we can not exclude the possibility that methylation on other sites, which could not be confidently quantified, indeed also decrease in the absence of JMJD6. It should also be noted that not all arginine residues at the C-terminus of MED12 were recovered in our MS analysis. Therefore, the exact methylation site that is critical for MED12 binding with chromatin and its regulated estrogen-induced transcription remains to be determined. The most likely scenario is that all the methylation sites combinatorially confer MED12 function herein. Thus, a BRD4, JMJD6, CARM1 and MED12 molecular axis/network was revealed to regulate estrogen-induced transcriptional activation. Intriguingly, based on our previous and current findings along with others, the components in this axis seem to interact highly with each other (Fig. 7).

Figure 7. A proposed model of JMJD6 function in estrogen/ERα-regulated enhancer and coding gene activation.

Upon estrogen stimulation, a coactivator complex constituting BRD4, JMJD6, MED12, CARM1 and others, accompanied by the exchange of co-repressors, was co–recruited with ERα onto active enhancers, leading to the transcriptional activation of these enhancers (Pol II recruitment and eRNA production) and cognate nearby coding genes (promoter-proximal Pol II pausing release and mRNA production). Mechanistically, BRD4/JMJD6 is required for MED12 to interact with CARM1, which methylates MED12 at multiple arginine residues at its C-terminus and mediates its chromatin binding.

Our data thereby reveal that JMJD6 is required for estrogen/ER-induced transcriptional activation, breast cancer cell growth and tumorigenesis, suggesting JMJD6 might serve as a potential drug target in ER-positive and endocrine therapy-resistant breast cancer. Indeed, JMJD6 has been found to be highly expressed in clinical ERα-positive breast tumor samples (Aprelikova et al., 2016, Lee et al., 2012, Poulard et al., 2015). JMJD6 has been shown to possess, at least two types of enzymatic activities, demethylation (Liu et al., 2013, Chang et al., 2007) and hydroxylation (Webby et al., 2009), which are important for its function in gene transcription and splicing. As JMJD6 function in breast cancer is apparently dependent on its enzymatic activity, developing small molecule inhibitors targeting JMJD6 will provide an additional therapeutic adjunct for ERα-positive and endocrine therapy-resistance breast cancers. Further, small molecule inhibitor capable of interfering with the molecular axis including BRD4, JMJD6, CARM1 and MED12, and antagonizing the enhancer activation program would also be efficacious in battling ERα-positive and endocrine therapy-resistance breast cancer.

STAR METHODS

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Wen Liu (w2liu@xmu.edu.cn).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Cell Culture

MCF7 and HEK293T cells obtained from ATCC were cultured in DMEM (GIBCO) media supplemented with 10% fetal bovine serum (FBS) in a 5% CO₂ humidified incubator at 37 °C. If estrogen (E₂) was added, cells were maintained in stripping medium (phenol red free) plus 5% charcoal-depleted FBS for 72 hrs before treating with or without E₂ (Sigma).

Animal Studies

Female nude mice (BALB/C, 15–20 g, 4–6 weeks old) were purchased from Beijing Vital River Laboratory Animal Technology, China, and maintained in animal room with 12 hrs light/12 hrs dark cycles at Animal Facility in Xiamen University. They were cared with free access to standard rodent chow and water in accordance with institutional guidelines. All of the animal experiments were approved by the Xiamen Animal Care and Use Committee.

METHOD DETAILS

Cloning Procedures

JMJD6 was PCR-amplified from p3XFLAG-CMV-10-JMJD6 we reported previously (Liu et al., 2013), and then cloned into pBobi, pRevTRE (Clontech) or pCMV-myc (Clontech) expression vector. Flag- and HA-tag were added to the amino- and carboxy-terminus of JMJD6 when cloned into pBobi and pRevTRE vectors, respectively. MED12 full length (FL) or truncations were PCR-amplified from cDNAs by using Transstart fastpfu fly polymerase (TransGen Biotech) and then cloned into p3XFLAG-CMV-10 (Sigma), pBobi (Flag- and HA-tag were added to the amino- and carboxy-terminus, respectively), pET-28a (+) (Novagen) or pGEX-6P-1 (GE Healthcare) expression vectors.

SiRNA Transfection, RNA Isolation, and RT-qPCR

SiRNA transfections were performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s protocol. Total RNA was isolated using RNeasy Mini Kit (Qiagen) following the manufacturer’s protocol. First-strand cDNA synthesis from total RNA was carried out using GoScript Reverse Transcription System (Promega), followed by quantitative PCR (qPCR) using AriaMx Real-Time PCR machine (Agilent Technologies). RNA samples from three biological repeats were pooled together for RT-qPCR analysis, and at least three technical repeats have been done for each pooled sample. Standard error of the mean is depicted. Sequence information for all primers used to check gene expression was presented in Table S2.

Plasmids Transfection, Lentivirus Packaging and Infection, Immunoblotting and Immunoprecipitation

Plasmid transfections were performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s protocol.

Lentivirus Packaging and Infection: HEK293T cells were seeded in culture plates coated with poly-D-lysine (0.1% (w/v), Sigma, P7280) and transfected with lentiviral vectors together with packaging vectors, pMDL, VSVG and REV, at a ratio of 10:5:3:2 using Lipofectamine 2000 for 48 hrs. Virus was collected, filtered and added to MCF7 cells in the presence of 10 μg/mL polybrene (Sigma, H9268), followed by centrifugation for 30 mins at 1,500 g at 37 °C. Medium was replaced 24 hrs later.

Protein immunoprecipitation and immunoblotting were performed as following. Briefly, cells were lysated in lysis buffer (50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1 mM EDTA and 1% Triton X-100) containing protease inhibitor cocktail (sigma, P2714–1BTL) on ice for 30 mins followed by centrifugation. For immunoblotting, the resultant supernatant was directly boiled in SDS sample buffer (1% SDS, 5% glycerol, 50 mM DTT, 30 mM Tris-HCl, pH 6.8, 0.25% bromophenol blue) for 5 mins, resolved by 10% SDS-PAGE gel in SDS running buffer (25 mM Tris, 250 mM glycine, 0.1% SDS) and transferred to nitrocellulose membrane (Bio-Rad). The membrane was then blocked at room temperature (RT) in blocking buffer (5% skim milk in TBST (10 mM Tris, pH 8.0, 150 mM Nacl, 0.05% Tween-20)) for 1 hr, incubated with primary antibody diluted in blocking buffer, and washed five times with TBST, followed by incubation with HRP-conjugated secondary antibody. Membranes were then rinsed with TBST extensively before imaging. For immunoprecipitation, the resultant supernatant was incubated with antibodies (2 to 5 μg) at 4°C overnight. Fifty μl of SureBeads Protein A or G magnetic beads (Bio-Rad, 161–4013 or 161–4023) were then added and incubated for an additional 4 hrs before washing five times with washing buffer (the same as lysis buffer). Beads were then re-suspended and boiled in SDS sample buffer, and the associated proteins were resolved by SDS-PAGE gel, transferred to nitrocellulose membrane. Immunoblotting was then performed as described above.

Cell Proliferation Assay, Colony Formation Assay, Wound Healing Assay, and Tumor Xenograft Assay

Cell viability was measured by using a CellTiter 96 AQueous one solution cell proliferation assay kit (Promega) following the manufacturer’s protocol. Briefly, MCF7 cells were transfected with siRNA and maintained in normal growth medium for different time points followed by cell proliferation assay. If estrogen (E₂) was added, cells were maintained in stripping medium (phenol red free) for two days before treating with or without estrogen (E₂, 10⁻⁷ M) for different time points followed by cell proliferation assay. When JMJD6 (WT) and JMJD6 (KO) were subjected to cell proliferation assay, cells were seeded at the same density and maintained in normal growth medium for different time points followed by cell proliferation assay. If estrogen (E₂) was added, cells were seeded at the same density and maintained in stripping medium (phenol red free) for two days before treating with or without estrogen (E₂, 10⁻⁷ M) for different time points followed by cell proliferation assay. To measure cell viability, 20 μL of CellTiter 96 AQueous one solution reagent was added per 100 μL of culture medium, and the culture plates were incubated for 1 hr at 37 °C in a humidified, 5% CO₂ atmosphere incubator. The reaction was stopped by adding 25 μL of 10% SDS. Data was recorded at wavelength 490 nm using a Thermo Multiskan MK3 Microplate Reader.

For colony formation assays, 2,000 cells, either JMJD6 (WT) or JMJD6 (KO), were seeded in one well in a 6-well plate, and colonies were examined 10 days after. Briefly, colonies were fixed with methanol/acid solution (3:1) for 5 mins and stained with 0.1% crystal violet for 15 mins.

For wound-healing assay, cells were grown to confluence in 6-well plates in normal growth medium or stripping medium, and wounds were performed with a P200 pipette tip. After removing cellular debris by washing cells with PBS, three images of each well were taken. The wounded area was measured by using image J and recorded as A0. For cells maintained in stripping medium, estrogen (E₂, 10⁻⁷ M) was then added. The cells were then allowed to migrate back into the wounded area, and three images were taken and the wounded area was measured again 24 hrs later and recorded as A1. Cell migration was presented as wound closure (%) = (wounded area (A0-A1)/wounded area A0)×100%.

For Tumor xenograft assay, four groups (4 mice/group) of female BALB/C nude mice (age 4–6 weeks) were subcutaneously implanted with 5×10⁶ of JMJD6 (WT) or JMJD6 (KO) cells suspended in DMEM medium without FBS. To supplement the estrogen for MCF7 cell proliferation, each nude mouse was brushed with estrogen (E₂, 10⁻² M) every 3 days for the duration of the experiments. All mice were euthanized 6 weeks after subcutaneous injection. Tumors were then excised, photographed and weighted. Animals were housed in the Animal Facility at Xiamen University under pathogen-free conditions, following the protocol approved by the Xiamen Animal Care and Use Committee.

Generation of JMJD6 Knockout Cell Lines Using CRISPR/Cas9 Gene Editing Technology

JMJD6 knock out (KO) MCF7 cells were generated by using CRISPR/Cas9 system. Specifically, gRNA sequence (5’ -GCTCTCGTAGTAGTTGTGCCGGG-3’) targeting JMJD6 was first cloned into gRNA cloning vector (Addgene, 41824) and confirmed by sequencing. MCF7 cells were then transfected with pcDNA3.3-hCas9 (Addgene, 41815) and gRNA expression vectors, followed by G418 (1 mg/mL) selection. Single colonies were subjected to immunoblotting to select knock-out ones, which were further validated by PCR using genomic DNA as template followed by Sanger sequencing. The sequencing information for primer set used was as follows: Forward (F) 5’- GTGCGTTAGTGTCAGGAAGC-3’ and Reverse (R) 5’ - GCCCAGAGAAAGGTGCGTA-3’.

Generation of Inducible MCF7 Cells Stably Expressing pRevTRE-Flag-JMJD6-HA and Purification of JMJD6-associated Proteins

MCF7 cells stably expressing pTet-On-Advanced (Clontech) were transfected with pRevTRE-Flag-JMJD6-HA, and then selected with hygromycin (200 μg/mL). To induce the expression of JMJD6, doxycycline (Dox) was added at a final concentration of 1 μg/mL for 48 hrs before adding estrogen (E₂, 10⁻⁷ M, 1 hr). To purify proteins associated with JMJD6, cell extracts were prepared in a buffer containing 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1 mM EDTA and 1% Triton X-100, and then subjected to affinity purification by using anti-Flag M2-agarose, washed extensively and eluted with 3XFlag peptides. Elutes were then subjected to in solution digestion and LC-MS/MS analysis following the protocol described below. To further purify proteins associated with JMJD6 on chromatin, resultant pellets as described above were further extracted with a buffer containing 50 mM Tris-HCl (pH 7.4), 420 mM NaCl, 1 mM EDTA and 1% Triton X-100, and subjected to affinity purification, elution, in solution digestion and LC-MS/MS analysis.

Stable Isotope Labeling by Amino Acids in Cell Culture (SILAC), Affinity Purification, In Solution Digestion and LC-MS/MS Analysis

Wild type and JMJD6 KO MCF7 cells were grown in SILAC DMEM (Invitrogen) supplemented with L-lysine/arginine and L-lysine/arginine-U-13C6 (Cambridge Isotope Laboratories), respectively, together with 10% dialyzed FBS, L-glutamine and penicillin/streptomycin for 2 weeks followed by infection with Lenti-viral vectors expressing pBobi-Flag-MED12 (1616–2177)-HA for 48 hrs before adding estrogen (10⁻⁷ M) for 1 hr. Cells were then lysed in a buffer containing 50 mM Tris-HCl (pH 7.4), 420 mM NaCl, 1 mM EDTA and 1% Triton X-100, pooled and subjected to affinity purification by using anti-Flag M2-agarose, washed extensively with a buffer containing 50 mM Tris-HCl (pH 7.4), 420 mM NaCl, 1 mM EDTA and 1% Triton X-100, and eluted with 3X Flag peptides (Sigma). Elutes were then subjected to in solution digestion and LC-MS/MS analysis following the protocol described below. The elutes after IP were firstly reduced in 20 mM dithiothreitol (DTT) (Sigma) at 95 °C for 5 mins, and subsequently alkylated in 50 mM iodoacetamide (IAA) (Sigma) for 30 mins in the dark at room temperature (RT). After alkylation, the samples were transferred to a 10 kD centrifugal spin filter (Millipore) and sequentially washed with 200 μL of 8 M urea for three times and 200 μL of 50 mM ammonium bicarbonate for two times by centrifugation at 14,000 g. Next, tryptic digestion was performed by adding trypsin (Promega) at 1:50 (enzyme/substrate, m/m) in 200 μL of 50 mM ammonium bicarbonate at 37 °C for 16 hrs. Peptides were recovered by transferring the filter to a new collection tube and spinning at 14,000 g. To increase the yield of peptides, the filter was washed twice with 100 μL of 50 mM ammonium bicarbonate. Peptides were desalted by StageTips. MS experiments were performed on a nanoscale UHPLC system (EASY-nLC1000, Proxeon Biosystems) connected to an Orbitrap Q-Exactive equipped with a nanoelectrospray source (Thermo Fisher Scientific). The peptides were dissolved in 0.1% formic acid (FA) with 2% acetonitrile (ACN) and separated on a RP-HPLC analytical column (75 μm×15 cm) packed with 2 μm C18 beads (Thermo Fisher Scientific) using a 4 hrs gradient ranging from 5% to 35% ACN in 0.1% FA at a flow rate of 300 nL/min. The spray voltage was set at 2.5 kV and the temperature of ion transfer capillary was 275 °C. A full MS/MS cycle consisted of one full MS scan (resolution, 70,000; automatic gain control (AGC) value, 1e6; maximum injection time, 100 ms) in profile mode over a mass range between m/z 350 and 1800, followed by fragmentation of the top twenty most intense ions by high energy collisional dissociation (HCD) with normalized collision energy at 28% in centroid mode (resolution, 17,500; AGC value: 1e5, maximum injection time: 100 ms). The dynamic exclusion window was set at 30 s. One microscan was acquired for each MS and MS/MS scan. Unassigned ions or those with a charge of 1+ and >7+ were rejected for MS/MS. Raw data was processed using Proteome Discoverer (PD, version 2.1), and MS/MS spectra were searched against the reviewed Swiss-Prot human proteome database. All searches were carried out with precursor mass tolerance of 10 ppm, fragment mass tolerance of 0.02 Da, oxidation (Met) (+15.9949 Da), methylation (Arg, Lys) (+14.0266 Da), dimethylation (Arg, Lys) (+28.0532 Da) and acetylation (protein N-terminus) (+42.0106 Da) as variable modifications, carbamidomethylation (Cys) (+57.0215 Da) as fixed modification and three trypsin missed cleavages allowed. Only peptides with at least six amino acids in length were considered. The peptide and protein identifications were filtered by PD to control the false discovery rate (FDR) <1%. At least one unique peptide was required for protein identification.

Protein Purification from Bacterial Cells or HEK293T Cells

GST- and His-tagged proteins were expressed in BL21 (DE3) bacterial cells (Agilent Technologies) and purified by using Glutathione agarose (Sigma) and Ni-NTA agarose (Qiagen), respectively. Flag-tagged proteins were expressed in HEK293T cells and cells were lysed in lysis buffer containing 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1 mM EDTA, 1% Triton X-100. Flag-tagged proteins were then affinity-purified by using anti-Flag M2 agarose and washed extensively with washing buffer containing 50 mM Tris-HCl (pH 7.4), 420 mM NaCl, 1 mM EDTA, 1% Triton X-100 before elution with 3X Flag peptides (Sigma).

Surface Plasmon Resonance

The binding kinetics between JMJD6 and MED12 was analyzed at room temperature (RT) on a BIAcore T200 machine. A CM5 sensorchip (GE Healthcare) was chemically activated by injecting 100 μL of N-ethyl-N′−3-(diethylaminopropyl) carbodiimide (EDC) (200mM) and N-hydroxysuccinimide (NHS) (50 mM) (v/v 1:1). In vitro purified bacterially-expressed JMJD6 in 10 mM sodium acetate (pH 5.0) was immobilized on the pre-activated CM5 chip using amine coupling according to the manufacturer’s instructions. The remaining ester groups were blocked by injecting 100 μL of 1 M ethanolamine-HCl (pH 9.5). The amount of immobilized JMJD6 was detected by mass concentration-dependent changes in the refractive index on the sensorchip surface, and corresponded to about 10,000 resonance units (RU). A serial concentration of MED12 ranging from 1.25 to 20 nM was added at a flow rate of 20 μL/min. When the data collection was finished in each cycle, the sensor surface was regenerated with glycine-HCl (10 mM, pH 2.5). Sensorgrams were fit globally with BIAcore T200 analysis using 1:1 Langmuir binding mode.

In vitro Methylation Assay

In vitro methylation assay was performed in methylation buffer (50 mM Tris-HCl (pH 8.0), 20 mM KCl, 5 mM DTT, 4 mM EDTA) in the presence of 2μCi L-[methy/−³H]-methionine at 37 °C for 1 hr. The reaction was stopped by adding SDS sample buffer followed by SDS-PAGE gel and autoradiogram.

Chromatin Immunoprecipitation Coupled with High Throughput Sequencing (ChIP-Seq)

For ChIP assays, cells were maintained in stripping medium (phenol red free) for three days before treating with or without estrogen (E₂, 10⁻⁷ M) for 1 hr. Cells were then fixed with 1% formaldehyde (Sigma) for 10 mins at room temperature (RT) (for MED12 and Pol II ChIP), or fixed with disuccinimidyl glutarate (DSG) (2 mM) (Proteochem) for 45 mins at RT, washed twice with PBS and then double-fixed with 1% formaldehyde for another 10 mins at RT (for JMJD6 ChIP). Fixation was stopped by adding glycine (0.125 M) and incubated for 5 mins at RT, followed by washing with PBS twice. Chromatin DNA was sheared to 300~500 bp average in size through sonication. Resultant was immunoprecipitated with anti-MED12, anti-Pol II or anti-JMJD6 antibody overnight at 4 C, followed by incubation with protein G magnetic beads (Invitrogen) for an additional 2 hrs. After washing and elution, the protein-DNA complex was reversed by heating at 65 C overnight. Immunoprecipitated DNA was purified by using QIAquick spin columns (Qiagen) and subjected to high throughput sequencing.

Global Run-On Sequencing (Gro-seq)

For Gro-seq experiments, cells were washed 3 times with cold 1X phosphate buffered saline (PBS) buffer and then incubated in swelling buffer (10 mM Tris-Cl (pH7.5), 2 mM MgCl₂, 3 mM CaCl₂) for 5mins on ice and harvested. Cells were first re-suspended and lysed in lysis buffer (swelling buffer with 0.5% IGEPAL and 10% glycerol). The resultant nuclei were washed once again with 10 mL lysis buffer and then re-suspended in 100 μL of freezing buffer (50 mM Tris-Cl (pH 8.3), 40% glycerol, 5 mM MgCl₂, 0.1 mM EDTA). For run-on assay, re-suspended nuclei were mixed with an equal volume of reaction buffer (10 mM Tris-Cl (pH 8.0), 5 mM MgCl₂, 1 mM DTT, 300 mM KCl, 20 units of SUPERase In, 1% sarkosyl, 500 μM ATP, GTP, and Br-UTP, 2 μM CTP) and incubated for 5 mins at 30 °C. The nuclear-run-on RNA (NRO-RNA) was then extracted with TRIzol LS reagent (Invitogen) following manufacturer’s instructions. NRO-RNA was then subjected to base hydrolysis on ice for 40 mins and followed by treatment with DNase I and antarctic phosphatase. To purify the Br-UTP labeled nascent RNA, the NRO-RNA was immunoprecipitated with an anti-BrdU agarose beads (Santa Cruz Biotech) in binding buffer (0.5 X SSPE, 1 mM EDTA, 0.05% tween-20) for 1hr at 4°C while rotating. To repair the end, the immunoprecipitated BrU-RNA was re-suspended in 50 μL reaction (43 μL DEPC water, 5.2 μL T4 PNK buffer, 1 μL SUPERase In and 1 μL T4 PNK (New England BioLabs)) and incubated at 37°C for 1 hr. RNA was then extracted, precipitated using acidic phenol-chloroform (Ambion) and subjected to poly-A tailing reaction by using poly-A polymerase (New England BioLabs) for 30 mins at 37 °C. Subsequently, reverse transcription was performed using oNTI223-primers (5’-pGATCGTCGGACTGTAGAACTCT; CAAGCAGAAGACGGCATACGATTTTTTTTTTTTTTTTTTTTVN-3’) where the p indicates 5’ phosphorylation, ‘;’ indicates the abasic dSpacer furan and VN indicates degenerate nucleotides. Specifically, tailed RNA (8.0 μL) was subjected to reverse transcription using superscript III (Invitrogen), after which the cDNA products were separated on a 10% polyacrylamide TBE-urea gel. The extended first-strand product (100–500bp) was excised and recovered by gel extraction. After that, the first-strand cDNA was circularized by CircLigase (Epicenter) and relinearized by Ape1 (New England BioLabs). Relinearized single strand cDNA (sscDNA) was separated in a 10% polyacrylamide TBE gel as described above and the product of needed size was excised (~170–400bp) for gel extraction. Finally, sscDNA template was amplified by PCR using the Phusion High-Fidelity enzyme (NEB) according to the manufacturer’s instructions. Primers oNTI200 (5’- CAAGCAGAAGACGGCATA-3 ‘) and oNTI201 (5’- AATGATACGGCGACCACCGACAGGTTCAGAGTTCTACAGTCCGACG-3 ‘) were used to generate DNAs for deep sequencing. Sequencing was performed on the Illumina Genome Analyzer II (GA II) according to the manufacturer’s instructions with small RNA sequencing primer 5’-CGACAGGTTCAGAGTTCTACAGTCCGACGATC-3’.

RNA Sequencing (RNA-seq)

Total RNA was isolated using RNeasy Mini Kit (Qiagen) following the manufacturer’s protocol. DNase I in column digestion was included to ensure the RNA quality. RNA library preparation was performed by using NEBNext® Ultra™ Directional RNA Library Prep Kit for Illumina (E7420L). Paired-end sequencing was performed with Illumina HiSeq 3000 at RiboBio Co., Ltd.

QUANTIFICATION AND STATISTICAL ANALYSIS

Computational Analysis of ChIP-Seq Data

For all ChIP-seq done in this manuscript, three biological repeats were performed and then pooled together. For each pooled sample, two technical repeats were repeated during library construction and sequencing to reduce errors from these steps. ChIP-seq sample preparation and computational analysis of ChIP-seq data were performed as following. Library construction: the libraries were constructed following Illumina ChIP-seq Sample prep kit. Briefly, ChIP DNA was end-blunted and added with an ‘A’ base so the adaptors from Illumina with a ‘T’ can ligate on the ends. Then 200–400 bp fragments are gel-isolated and purified. The library was amplified by 18 cycles of PCR. Primary analysis of ChIP-Seq datasets: the image analysis and base calling were performed by using Illumina’s Genome Analysis pipeline. The sequencing reads were aligned to hg19 Refseq database by using Bowtie2 (Langmead and Salzberg, 2012) (http://bowtie-bio.sourceforge.net/bowtie2/index.shtml) with default parameters. Both uniquely aligned reads and reads that align to repetitive regions were kept for downstream analysis (if a read aligned to multiple genomic locations, only one location with the best score was chosen). Clonal amplification was circumvented by allowing maximal one tag for each unique genomic position. The identification of ChIP-seq peaks was performed using HOMER ((http://homer.ucsd.edu/homer/). The threshold for the number of tags that determined a valid peak was selected at a false discovery rate (FDR) of 0.001. Fourfold more tags relative to the local background region (10 kb) were also required to avoid identifying regions with genomic duplications or non-localized binding. Genomic distribution was done by using the default parameters from HOMER with minor modifications, in which promoter peaks were defined as those with peak center falling between 1,00 bp downstream and 5,000 bp upstream of transcript start sites (TSS). To define estrogen-induced JMJD6 binding sites, only when fold change (FC) of ChIP-seq tag density of a peak in estrogen treatment versus control was larger than 2, that peak was considered as estrogen specific (P < 0.001). Motif analysis was performed using HOMER. Tag density for histograms (50 bp/bin), box plots and heat maps were generated by using HOMER. Box plots were then generated by R software (https://www.r-project.org/) and significance was determined using Student’s t test. Heat maps were visualized using Java TreeView (Saldanha, 2004) (http://jtreeview.sourceforge.net).

Computational Analysis of Gro-seq Data

Gro-seq sequencing reads were aligned to hg19 Refseq database by using Bowtie2. Both uniquely aligned reads and the sequencing reads that align to repetitive regions were kept for downstream analysis (if a read aligned to multiple genomic locations, only one location with the best score was chosen). Clonal amplification was circumvented by allowing maximal three tags for each unique genomic position. When analyzing estrogen effects on gene transcription, mapped reads from the first 30 kb of gene body were counted by HOMER, excluding promoter-proximal region (transcription start site (TSS) to 500 bp downstream of TSS). If the length of a gene is shorter than 30kb, then mapped reads from the whole gene were counted, excluding promoter-proximal region and gene end (500 bp upstream of transcription termination site (TTS) to TTS). The mapped reads were computed by edgeR (Oshlack et al., 2010) (http://www.bioconductor.org/packages/release/bioc/html/edgeR.html) to determine differentially expressed genes. Estrogen-regulated gene program was determined by exact test (P < 0.001) as well as fold change (FC) of gene RPKM (reads per kilobase per million mapped reads) in control and estrogen-treated samples (FC > 1.5). Coding genes with RPKM larger than 0.5, either in control or estrogen-treated sample, were included in our analysis. RPKM of a gene was calculated as mapped reads divided by gene length and the total number of mapped reads. Tag density for histograms (50 bp/bin), box plots and heat maps were generated by using HOMER. Box plots were then generated by R software and significance was determined using Student’s t test. Heat maps were visualized using Java TreeView.

Three biological repeats were performed and then pooled together. For each pooled sample, two technical repeats were repeated during library construction and sequencing to reduce errors from these steps.

Computational Analysis of RNA-seq Data

Three biological repeats were performed and then pooled together. For each pooled sample, two technical repeats were repeated during library construction and sequencing to reduce errors from these steps. Sequencing reads were aligned to hg19 Refseq database by using Tophat (Trapnell et al., 2012) (http://ccb.jhu.edu/software/tophat/index.shtml). Both uniquely aligned reads and the sequencing reads that align to repetitive regions were kept for downstream analysis (if a read aligned to multiple genomic locations, only one location with the best score was chosen). Only reads on exons were counted for quantifying gene expression by HOMER. The mapped reads were computed by edgeR (http://www.bioconductor.org/packages/release/bioc/html/edgeR.html) to determine differentially expressed genes. Estrogen-regulated gene program was determined by exact test (P < 0.001) as well as fold change (FC) of gene FPKM (fragments per kilobase per million mapped reads) in control and estrogen-treated samples (FC > 1.5). Coding genes with FPKM larger than 0.5, either in control or estrogen-treated sample, were included in our analysis. FPKM of a gene was calculated as mapped reads on exons divided by exonic length and the total number of mapped reads. Tag density for histograms (50 bp/bin), box plots and heat maps were generated by using HOMER. Box plots were then generated by R software and significance was determined using Student’s t test. Heat maps were visualized using Java TreeView.

Traveling Ratio Calculation

Traveling ratio (TR) calculated based on Gro-seq tag density was defined as ratio of Gro-seq reads density at the promoter-proximal bin (from 50 bp upstream to 300 bp downstream of TSS) to that at the gene body bin (from 300 bp downstream of TSS to 30K bp of gene body). The significance of the change of TR was displayed using box plot and determined using Student’s t test.

DATA AND SOFTWARE AVAILABILITY

The accession number for the ChIP-seq, Gro-seq and RNA-seq data reported in this paper is GSE101562. ERα ChIP-seq was from GSE45822; H3K27Ac and p300 ChIP-seq were from GSE62229; MED1 ChIP-seq was from GSE60272; H3K4me3, H3K9me3 and H3K27me3 ChIP-seq were from GSE23701. ChIP-seq (JMJD6, Pol II and MED12), Gro-seq and RNA-seq data were deposited in the Gene Expression Omnibus database under accession GSE101562.

Supplementary Material

Table S1

Table S1. MS2 spectrum of arginine-methylated peptides in MED12 which methylation decreased upon JMJD6 depletion. Related to Figure 5. MS2 spectrum of methylated arginine residues in MED12 C-terminus which methylation decreased upon JMJD6 depletion as described in Fig. 5H were shown as indicated.

Table S2

Table S2. Sequence information for all qPCR primers used in the current study. Related to STAR Methods. Sequence information of qPCR primers designed to detect the expression of FOXC1, GREB1, NRIP1, SIAH2 and SMAD7, and their cognate enhancer RNAs (eRNAs) as well as ChIP primers specifically targeting GAPDH promoter, FOXC1, SIAH2 and GREB1 enhancer regions were shown. F: forward; R: reverse.

HIGHLIGHTS.

• JMJD6 binds to ERα-bound active enhancers in the presence of E₂

• JMJD6 is a determinant of E₂-induced enhancer and cognate coding gene activation

• JMJD6 is required for MED12’s binding and function on enhancers

• JMJD6 is required for E₂-induced breast cancer cell growth and tumorigenesis