The Mediator complex orchestrates multiple transcription factors with the Pol II apparatus for precise transcriptional control. However, its interplay with the surrounding chromatin remains poorly understood. Here, we analyze differential histone modifications between WT and MED23−/− (KO) cells and identify H2B mono‐ubiquitination at lysine 120 (H2Bub) as a MED23‐dependent histone modification. Using tandem affinity purification and mass spectrometry, we find that MED23 associates with the RNF20/40 complex, the enzyme for H2Bub, and show that this association is critical for the recruitment of RNF20/40 to chromatin. In a cell‐free system, Mediator directly and substantially increases H2Bub on recombinant chromatin through its cooperation with RNF20/40 and the PAF complex. Integrative genome‐wide analyses show that MED23 depletion specifically reduces H2Bub on a subset of MED23‐controlled genes. Importantly, MED23‐coupled H2Bub levels are oppositely regulated during myogenesis and lung carcinogenesis. In sum, these results establish a mechanistic link between the Mediator complex and a critical chromatin modification in coordinating transcription with cell growth and differentiation.
The Mediator subunit MED23 regulates histone 2B mono‐ubiquitylation via RNF20/40 recruitment, leading to transcriptional activation of target genes and illustrating a role for MED23 in myogenesis and tumorigenesis.
The Mediator subunit MED23 is required for H2B mono‐ubiquitination.
The Mediator complex recruits E3 ligase RNF20/40 via MED23.
MED23 links H2B mono‐ubiquitination to transcriptional activation by Mediator.
Mediator and the PAF complex cooperatively stimulate RNF20/40 E3 activity to enhance H2B mono‐ubiquitination.
As an epigenetic regulator, Mediator plays a critical role in cell growth and differentiation.
Transcription in eukaryotic cells is subject to multiple layers of regulation. In addition to a host of specific transcription factors, co‐factors, general transcription factors, and RNA polymerases, the chromatin environment, such as DNA or histone modifications (Suganuma & Workman, 2011; Lee & Young, 2013), impacts gene expression. An increasing number of studies have indicated dynamic interplays between these layers of transcriptional control.
Chromosomal histone tails are subjected to multiple covalent modifications, including methylation, acetylation, phosphorylation, ADP‐ribosylation, and ubiquitination (Strahl & Allis, 2000; Turner, 2002). Studies over the past few decades have revealed that reversible chromatin modifications are closely associated with gene transcription. Some classic histone marks are associated with transcriptional activation, including H3K4 di‐ or trimethylation, H3K36 trimethylation, multiple acetylations of H3 and H4, and H2B ubiquitination, whereas some other histone markers, such as H3K9 and K27 trimethylation and H2AK119 ubiquitination, correlate with transcriptional repression (Berger, 2002; Li et al, 2007). Studies on histone modifications have led to an important concept known as the histone code, where the combination of multiple histone modifications give rise to precise transcriptional outcomes, such as activation, repression, intensity, and duration (Strahl & Allis, 2000; Jenuwein & Allis, 2001). Mechanistically, this code posits that distinct histone marks serve as a platform for binding of effectors that control transcriptional activation or silencing (Strahl & Allis, 2000; Jenuwein & Allis, 2001).
The canonical function of the Mediator complex is to bridge transcription factors and the basal transcription machinery to promote assembly of the pre‐initiation complex (PIC) (Malik & Roeder, 2010; Taatjes, 2010). In recent years, the Mediator complex has emerged as a “master coordinator” that orchestrates multiple steps involved in gene expression, including transcription initiation, elongation, termination, and mRNA processing (Donner et al, 2010; Mukundan & Ansari, 2011; Takahashi et al, 2011; Huang et al, 2012; Galbraith et al, 2013; Wang et al, 2013b; Yin & Wang, 2014). Recent studies on the regulation of Mediator complex by chromatin structure have suggested that the Mediator complex plays an important role in chromatin looping and the function of super enhancers (Kagey et al, 2010; Whyte et al, 2013). Another study revealed that MED12 links the repressive histone marker H3K9 di‐methylation with neuronal gene regulation (Ding et al, 2008). Mediator kinase sub‐module associates various chromatin remodeling complexes (Fukasawa et al, 2012). These findings raise the important question as to whether Mediator subunits can function as epigenetic regulators for specific histone modifications.
In an effort to search for epigenetic events regulated by the Mediator subunit MED23, we performed a screening assay to examine histone modifications and identified H2B mono‐ubiquitination (H2Bub) as a MED23‐dependent histone modification. Using proteomic and biochemical approaches, we found that the Mediator complex associates with the RNF20/40 complex via its MED23 subunit, and MED23 can robustly enhance H2BK120 ubiquitination through the RNF20/40 complex in vitro. ChIP‐seq experiments indicated that MED23 ablation attenuated H2Bub levels and reduced Pol II enrichment genome‐wide. Combined analyses of ChIP‐seq and microarray data revealed MED23‐dependent coupling of H2Bub and transcription levels. Furthermore, skeletal muscle tissues and myotubes expressed low levels of MED23 that also displayed low level of H2Bub, and the MED23 deletion in myoblasts facilitated myogenesis. In contrast, high levels of MED23 collaborate with high levels of H2Bub to control cancer‐related pathways in human lung cancer. Together, these findings illustrate a novel cross‐regulation between the Mediator complex and epigenetic modifications, and expand the role of the Mediator complex as an epigenetic regulator in cell growth and differentiation.
MED23 deficiency selectively reduces H2B mono‐ubiquitination
We recently found that the Mediator complex interacts with the splicing machinery through an interaction between MED23 and the splicing regulator HnRNP L (Huang et al, 2012). HnRNP L is also known as a subunit of the human KMT3a complex that controls H3K36 tri‐methylation (H3K36me3) (Yuan et al, 2009). This result led us to ask whether MED23 regulates H3K36me3 levels and/or other histone modifications. To further examine epigenetic regulation by the Mediator subunit MED23, we screened histone modifications in MED23+/+ (WT), MED23−/− (KO), and MED23−/− with re‐expressed MED23 (KO+23) MEFs. We examined both global methylation and acetylation of H3 and H4 and found no significant difference in the cells (Fig 1A). Despite the observation that MED23 depletion decreased H3K36me3 levels in a set of genes whose alternative splicing (AS) events are co‐regulated by MED23 and HnRNP L (Huang et al, 2012), H3K36me3 did not globally change in KO cells when compared with WT or KO+23 cells (Fig 1A). No significant difference was observed in any of the histone modifications examined (Fig 1A). However, global H2B lysine 120 mono‐ubiquitination (H2Bub) was greatly reduced upon MED23 depletion, and these levels could be recovered in KO cells re‐expressing MED23 (Fig 1B and C). These results suggested that at the global level, H2Bub appears to be regulated by MED23. In contrast, H2A lysine 119 mono‐ubiquitination (H2Aub) was not affected by MED23 (Fig 1A). Previous studies have shown that H2B mono‐ubiquitination is required for H3K4 and H3K79 methylation by Set1 and Dot1, respectively (Dover et al, 2002; Kim et al, 2009), but we found that methylation on H3K4 and H3K79 was hardly changed in KO MEFs. Therefore, MED23‐dependent H2Bub regulation does not seem to be coupled with global H3 methylation in our system (Fig 1A). However, further details on the coupling between H2Bub, H3K4, and H3K79 methylations will be revealed later in Fig 5.
To test whether immortalization of the MEFs may have affected H2Bub regulation and induced an indirect effect of MED23 on H2Bub, primary MEFs were isolated from embryonic day 12.5 MED23F/F mice in the presence or absence of Cre‐expressing adenovirus. H2Bub modification was significantly and consistently reduced when MED23 was acutely deleted by Ad‐cre, a Cre‐expressing adenovirus (Fig 1D). To examine the specificity of Mediator subunits for H2Bub modification, multiple Mediator components were depleted in WT MEFs using a retrovirus‐mediated knockdown method, and H2Bub levels were significantly reduced in MED23 knockdown cells (Fig EV1A and B). Interestingly, for reasons that are presently unknown, depletion of Mediator subunits, such as MED6, MED12, MED14, and MED24, increased H2Bub levels compared with control cells. Moreover, MED23 regulation of H2Bub was also observed in other cell types. For example, like knockdown of H2B ubiquitination E3 ligase RNF20, depletion of MED23 in HeLa cells also exhibited decreased levels of H2Bub when two different shRNAs targeting MED23 were used (Figs 1E and EV1C). Taken together, these results suggest that MED23, unlike several other Mediator subunits, specifically regulates H2Bub but not other histone modifications such as multiple histone acetylation, methylation, or H2Aub.
H2B mono‐ubiquitination occurs when the 76‐residue ubiquitin protein is covalently attached to Lys120 of histone H2B in metazoans (Lys123 in yeast). This reaction is catalyzed by the specific heterodimeric E3 ligase RNF20/40 (Bre1A/B in yeast) (Hwang et al, 2003; Kim et al, 2005; Zhu et al, 2005). To determine whether MED23‐dependent H2Bub occurred via RNF20/40, we analyzed the expression levels of RNF20/40 and found that loss of MED23 did not change the expression of RNF20 and RNF40 (Fig 1B). We have also found that there is no difference in the expression of USP22, the known deubiquitinase enzyme for H2Bub (Zhang et al, 2008a). Therefore, the regulation of H2Bub by MED23 likely occurs through mechanisms other than transcriptional regulation of the related enzymes.
The Mediator complex associates with the RNF20/40 E3 ligase via MED23
To understand the mechanistic relationship between the Mediator subunit MED23 and H2Bub, we employed a His‐Flag double tag tandem affinity purification procedure to search for MED23‐interacting proteins. Using a baculovirus expression system, double‐tagged MED23 was purified as a soluble protein (Fig 2A), and its specific interactions with E1A and ELK1 were verified (Wang et al, 2009; Huang et al, 2012). The MED23‐bound Ni‐NTA column was then loaded with HeLa cell nuclear extract, washed, and eluted with imidazole buffer. MED23 subsequently bound to anti‐Flag beads and was eluted with a FLAG peptide. The eluted protein mixture was analyzed by MS/MS spectrometry (Fig 2B and C). In addition to a group of RNA‐processing proteins, including HnRNP L (Huang et al, 2012), peptides of the RNF20/40 enzyme were identified among the MED23‐binding proteins (Table EV1), suggesting a potential mechanistic link between the Mediator complex and H2Bub ligases (Fig 2D). By contrast, neither RNF20 nor RNF40 was detected by MS/MS spectrometry when an identical procedure was used for five other Mediator components, including CDK8, MED15, MED16, MED24, and MED29 (Huang et al, 2012; Wang et al, 2013b). Consistently, RNF20 and RNF40 were eluted from the His‐Flag‐tagged MED23 affinity purification column but not from the MED29 column (Fig 2E).
To further verify the association between RNF20/40 and MED23, a Co‐IP assay was performed by co‐transfecting plasmids of tagged MED23, RNF20, and RNF40 into HeLa cells. While individual RNF20 or RNF40 proteins could only weakly interact with MED23, co‐expression of both RNF20/40 resulted in strong Co‐IP of MED23 (Figs 2F and EV2A). UBE2A, an E2‐conjugating enzyme which had previously been shown to directly interact with the RNF20/40 complex (Kim et al, 2009), was also pulled down with MED23 (Fig 2F). In addition, antibodies against RNF20 or RNF40 pulled down endogeneous MED23 in HeLa cell extract (Fig 2G). Notably, the mixed antibodies to RNF20 and RNF40 can be more effective than antibody to either RNF20 or RNF40 in pulling down MED23 as well as other Mediator components, such as MED1 (Fig 2G). Conversely, antibodies against MED23 pulled down RNF20/40 as well as MED1 in HeLa cells, but not GAPDH and STAT3 (as a negative control) (Fig 2H). Furthermore, to verify that MED23 interacts with RNF20/40 as part of the Mediator complex, we perform the Co‐IP assay in HeLa nuclear extract using the antibody against another Mediator subunit CDK8. CDK8 antibody can efficiently pull down the Mediator complex (as indicated by multiple components), as well as RNF20 and RNF40 (Fig EV3A). Consistently, when HeLa nuclear extract was fractionated by Superose 6 gel filtration column, RNF20, RNF40, and PAF1 were detected in fractions that overlap with those for the Mediator complex, as indicated by MED1, MED12, MED6, MED16, MED23, and CDK8 (Fig EV2B, fractions 10–13). Further, the RNF20/40 did not co‐elute with the monomeric MED23 fractions (Fig EV2B, fractions 19–22), suggesting that RNF20/40 acts together with the Mediator complex, but not the monomeric MED23, to regulate H2B mono‐ubiquitination. Domain‐mapping with multiple deletions of RNF40 revealed that the second coiled‐coiled domain of RNF40 was important for the RNF20/40‐MED23 association (Fig EV2C), which is consistent with previous observations that the second coiled‐coiled domain is critical for both RNF20–RNF40 and RNF20/40–UBE2A/B interactions (Kim et al, 2009). For MED23, the N‐terminal region (N1‐327) and the C‐terminal region (N878‐1368) seem to be important for the MED23–RNF20/40 interaction, whereas other regions appear to play a negative role in this interaction (Fig EV2D).
Because MED23 regulates H2Bub and interacts with RNF20/40, we reasoned that MED23 might be required for RNF20/40 recruitment to MED23‐targeted genes in vivo. We therefore developed an RNF20/40 ChIP assay to analyze binding of RNF20/40 to the MED23‐targeted Egr1 gene locus using a mixture of antibodies specific to RNF20 and RNF40. As indicated in Fig 2I, MED23 deficiency reduced the recruitment of RNF20/40 by approximately threefold at the Egr1 promoter region, which coincided with our previous finding that MED23 deficiency reduces Mediator recruitment to a similar degree at the Egr1 gene promoter (Wang et al, 2005).
We next tested if the differential recruitment of RNF20/40 in WT and KO MEFs affects association of RNF20/40 with chromatin. Cell extracts derived from either WT or KO MEFs were respectively fractionated into chromatin and soluble fractions, followed by immunoblotting using various antibodies. Deficiency of MED23 decreased the chromatin binding of RNF20 but not UBE2A, indicating that MED23 is important for recruiting the RNF20/40 E3 ligase to chromatin for H2B mono‐ubiquitination (Fig 2J). Moreover, we also observed that the levels of MED24 and MED6 were moderately reduced in chromatin of KO cells compared with that of WT cells, suggesting that MED23 deletion also impairs Mediator association with chromatin (Fig 2J). Taken together, these data indicate that Mediator associates with the RNF20/40 E3 ligase complex via MED23 and that this association may be responsible for targeting RNF20/40 to chromatin for H2B mono‐ubiquitination.
Mediator and PAF complexes collaboratively promote H2Bub in vitro
To demonstrate direct functional consequence of the interaction between MED23 and RNF20/40 in H2B ubiquitination, we employed an in vitro assay to check if Mediator MED23 affects H2B ubiquitination. Lysine 120 ubiquitination of H2B took place in a complete in vitro reaction containing purified E1, UBE2A (E2), RNF20/40 complex (E3) (Fig 3A), ubiquitin, ATP, and histone octamer, but not in the reactions missing any of the aforementioned components (Fig EV3B, lanes 1–6). To examine whether MED23 directly stimulates H2B ubiquitination in vitro, purified recombinant MED23 (Fig 3A) was introduced to the H2Bub assay. Indeed, it increased the UBE2A–RNF20/40 complex‐mediated H2B ubiquitination on the histone octamer substrate (Fig EV3B, lane 7), suggesting that MED23 can enhance H2BK120 ubiquitination via the UBE2A–RNF20/40 complex.
Because nucleosomes constitute a more relevant substrate in vivo, we further analyze the role of MED23 in promoting H2Bub using assembled recombinant chromatin as a substrate. However, the recombinant MED23 failed to enhance H2Bub on an oligonucleosome substrate (Fig 3B, compare lane 2 to lane 1). Purified Mediator complex could barely enhance H2Bub levels further (Fig 3B, lane 3). We thus speculated that a critical component for Mediator‐dependent regulation of H2Bub might be lacking in these reactions. We noticed that during gel filtration analysis of the Mediator, PAF complex partially co‐eluted with RNF20/40 and the Mediator complex (Fig EV2B). Since it was also previously reported that the PAF complex interacts with E2 and E3 (RNF20/40) to facilitate mono‐ubiquitination of H2B on chromatin templates (Zhu et al, 2005; Kim et al, 2009), we went on to test if the purified PAF complex may help Mediator in the in vitro H2Bub reaction (Fig 3A). While PAF complex alone weakly stimulated H2Bub, recombinant MED23 plus PAF complex significantly increased H2Bub levels (Fig 3B, compare lane 5 to lane 4). Most noticeably, the reaction containing the PAF complex and purified endogenous Mediator complex dramatically increased the level of H2Bub on chromatin substrate (Fig 3B, lane 6). In addition, the Mediator complex acted more effectively than MED23 alone on H2B ubiquitination (Fig 3B, lane 6 compared to lane 5). Consistent with in vitro results, we observed that PAF complex recruitment at the MED23‐target gene Egr1 was reduced by threefold with MED23 depletion in HeLa cells (Fig 3C). Taken together, these results strongly suggest interplays between Mediator, the PAF complex, and the H2B mono‐ubiquitination machinery, and Mediator and PAF complexes may collaboratively promote H2B lysine 120 ubiquitination through RNF20/40 (Fig 8).
MED23‐dependent and MED23‐independent H2Bub regulation and transcriptional activities
Mono‐ubiquitination of H2B enhances the accessibility of chromatin to transcriptional activators (Fierz et al, 2011) and facilitates FACT‐dependent Pol II stimulation (Pavri et al, 2006; Fleming et al, 2008). We therefore examined whether MED23 could control transcriptional activation through H2B mono‐ubiquitination. First, we performed ChIP‐seq assays to compare Pol II and H2Bub binding profiles in WT and KO MEF cells. Consistent with the Western blot results shown in Fig 1, MED23 deficiency reduced enrichment of H2Bub at the gene body genome‐wide and impaired Pol II occupancy at the gene promoters and in coding regions (Figs 4A and EV4A). And the heatmaps of ChIP‐seq density for H2Bub and Pol II also indicated that MED23 deletion significantly reduced H2Bub level at the coding regions of about one‐third of total genes, and Pol II enrichment at the same set of genes' bodies also attenuated accordingly. As a comparison, there is no difference in the H3K4me3 level at TSS (transcription start sites) between WT and KO cells (Fig 4B). Cumulative distribution function analysis showed that the differences in H2Bub and Pol II binding between WT and KO MEF cells were highly significant (Fig EV4B). These results suggest a MED23‐dependent correlation between H2Bub and Pol II occupancy.
To further explore the detail relationship among MED23, H2Bub, and transcriptional activity, we performed a combinatorial analysis using the ChIP‐seq data together with microarray data. Based on the microarray analysis of WT and KO MEFs (Yin et al, 2012), three sets of genes were stratified: (i) down‐regulated more than twofold in MED23 KO cells (n = 903), (ii) up‐regulated more than twofold in MED23 KO cells (n = 1,009), and (iii) minimal fold changes (less than 0.01‐fold), also called unchanged genes by MED23 KO (n = 1,000). Interestingly, only for the set of genes down‐regulated by MED23 KO, loss of MED23 largely reduced the occupancy of H2Bub at gene bodies (Fig 4C). Cumulative distribution function analysis confirmed the significant difference (Fig EV4C). Consistently, the heatmaps for this set genes showed that the enrichment levels of H2Bub and Pol II, but not H3K4me3, were dramatically decreased by MED23 deletion (Fig 4D). By contrast, for the other two sets of genes that were unchanged or up‐regulated by MED23 KO, both the average profiles and heatmaps demonstrated that there were no significant differences in H2Bub levels between the WT and KO MEF cells (Figs 4C and D, and EV4C), suggesting that both MED23‐dependent and MED23‐independent H2Bub regulation existed. Moreover, when an integrated analysis was applied to the microarray and H2Bub ChIP‐seq data of WT and KO cells, the number of genes whose expression (> 2 folds up or down) and H2Bub levels (> 2 folds up or down) were consistently regulated by MED23 significantly outnumbered genes that were inversely regulated (P = 8.9e‐26) (Fig 4E), indicating the MED23‐dependent coupling of the transcriptional activities and H2Bub levels.
To analyze the global distribution relationship between Mediator and H2Bub, we also compared the average binding profiles of H2Bub and Mediator using our H2Bub ChIP‐seq and MED1 ChIP‐seq data (Kagey et al, 2010), which revealed that Mediator complex (in green) is mainly enriched at TSS, and H2Bub (in red) is mainly enriched at coding region with its peak very close to the TSS (Fig 4F), thus further supporting a close relationship between Mediator complex and H2Bub modification.
To examine the gene‐specific regulation of H2Bub by MED23, we checked the proto‐typical MED23 target gene Egr1, Krox20, and Egr3, and found that occupancy of both Pol II and H2Bub were decreased upon MED23 deletion (Figs 5A–E and EV5A and B), consistent with the Western blotting screen and genome‐wide analyses (Figs 1 and 4). Noticeably, results in Fig 1 showed that MED23‐dependent H2Bub regulation does not seem to be coupled with global H3 methylation. However, at the Egr1 gene locus, we observed that H3K4me3 and H3K79me3 modification levels were decreased in the coding region but not at the promoter region of Egr1 in MED23‐depleted cells (Fig 5F and G). ChIP‐seq also revealed that the enrichment of H3K4 and K79 tri‐methylation at gene coding regions were more or less reduced by MED23 deletion (Fig 5H and I). Therefore, the coupling of H2Bub with H3 methylation appears to occur mainly in coding regions, consistent with that H2Bub modification occurs in these regions.
Collectively, these data describe a novel function of Mediator involved in epigenetic regulation. Mediator MED23 subunit can act as an epigenetic regulator for H2Bub modification in coupling with the transcriptional activation.
MED23‐dependent H2Bub exerts opposing effects during myogenesis and carcinogenesis
Among multiple tissues, Mediator subunit MED23 shows the most reduced expression in muscle when compared with many other tissues (Fig 6A), we ask if the levels of MED23 and H2Bub may couple to regulate myogenesis. First, we checked the H2B ubiquitination in multiple mouse tissues, and found that H2Bub was almost lost in the skeletal muscle, which well correlates with the greatly reduced MED23 expression relative to other tissues (Fig 6A). Then, we compared myoblasts with myotubes by examining Mediator components and H2Bub levels using Western blotting and found that MED1 and MED23 were both detected in myoblasts but not much in myotubes (Fig 6B), suggesting both Mediator subunits and H2Bub levels were reduced in myotubes (Fig 6B). When we knocked down MED23 in myoblasts, the attenuated H2Bub levels was also attenuated, suggesting that loss of MED23 in myoblasts is coupled with H2Bub levels (Fig 6C). We then explored the effects of MED23 depletion on myoblast differentiation. The standard horse serum‐induced myogenesis assay was performed using myoblasts with or without MED23 depletion. Three days post‐differentiation, MED23‐depleted C2C12 cells contained more myotubes with enhanced stress fiber and focal adhesion formation compared with control cells (Fig 6D). Immunostaining with myotube markers including MHC and myogenin revealed further confirmed that MED23 deficiency facilitated the process of myogenesis (Fig 6D). Consistent with this result, Western blot analysis of MED23‐deficient cells revealed increased expression of both MHC and myogenin during different stages of myogenesis (Fig 6E). Therefore, the reduced MED23 expression can result in attenuated H2Bub levels and is important for promoted myoblast differentiation.
In contrast with the reduced expression of MED23 observed during myogenesis, we learned from previous study that MED23 is highly expressed in lung cancer (Yang et al, 2012; Shi et al, 2014). We then asked whether the levels of H2Bub and MED23 were elevated in cancer tissues. Human clinical samples of lung adenocarcinoma were immunostained in parallel with antibodies against either H2Bub or MED23. When compared with corresponding adjacent normal tissues, both H2Bub and MED23 exhibited positive staining patterns in tumor sections (Fig 7A). To quantitatively assess the differences between cancer and normal tissues, we employed a scoring assay that was performed by a pathologist to quantify the staining intensity of with each tissue. Statistical analysis showed that the levels of both H2Bub and MED23 in tumors were significantly higher than those found in normal tissues (P = 0.0006 and P < 0.0001, respectively) (Fig 7B). When using H2Bub staining intensity as an index, high H2Bub staining intensity correlated well with high MED23 staining signals (Fig 7C), suggesting that the elevated expression levels of MED23 and H2Bub are coupled during tumorigenesis.
We then asked how the coupled MED23:H2Bub is related to tumorigenesis. We first used pathway enrichment analysis to examine the gene set which showed > twofold decreased expression in Med23 KO MEFs compared to WT MEFs and exhibited significantly reduced H2Bub enrichment (Fig 4C). Interestingly, multiple top‐ranked categories were related to cancer (Fig 7D, column 1). Secondly, analyzing the gene set with H2Bub decreased by more than twofold in KO compared to WT MEFs, and we found that there were more cancer‐related pathways among the top‐ranked groups (Fig 7D, column 2). Lastly, more striking enrichment of cancer‐related pathways were found for the gene set whose expression and H2Bub modification levels were both down‐regulated by MED23 KO (Fig 7D, column 3). These analyses indicated that MED23–H2Bub coupling is indeed a critical mechanism in involved in tumorigenesis.
The Mediator complex has recently emerged as a versatile coordinator at multiple post‐recruitment steps of PIC formation, including transcription elongation, mRNA processing, and termination in addition to its canonical role in establishing pre‐initiation complex (PIC) formation (Donner et al, 2010; Mukundan & Ansari, 2011; Takahashi et al, 2011; Huang et al, 2012; Wang et al, 2013b; Yin & Wang, 2014). The Mediator also associates with ncRNAs to activate transcription (Lai et al, 2013). In addition, the Mediator was found to cooperate with cell identity‐related activators and cohesin complexes to establish “super enhancers” by looping enhancer‐promoter DNA during gene activation (Kagey et al, 2010; Lovén et al, 2013; Whyte et al, 2013). These findings suggest that the Mediator may play a vital role in shaping the chromatin architecture. Collectively, other studies (Dhawan et al, 2009) as well as ours have delineated the Mediator as an “epigenetic regulator” for the chromatin environment. Specifically in our study, the Mediator can mediate transcription activation through a MED23‐dependent H2B mono‐ubiquitination mechanism.
The regulation of H2Bub by MED23 appears to be highly specific. We screened many major histone modifications related to transcriptional regulation using a large panel of available antibodies, and H2Bub appeared to be a MED23‐dependent histone modification. MED23‐dependent H2Bub regulation was also observed in different types of cells, including primary MEF, HeLa, and C2C12 cells (Figs 1, 6 and EV1). Importantly, individual depletion of multiple Mediator subunits in MEFs also showed that MED23, but not other Mediator components, was specifically coupled with H2Bub (Fig EV1A and B).
The association between MED23 and RNF20/40 establishes a mechanistic link between the Mediator complex and epigenetic regulation during transcription, which may provide further insight into the roles that the Mediator complex plays in (i) crosstalk between multiple histone modifications, (ii) transcription elongation, (iii) histone modifications and cancer, and (iv) mRNA processing, as we discuss here.
Firstly, previous studies regarding trans‐histone modification crosstalk indicate that a strict linkage exists between ubiquitinated H2B and histone H3 methylation (Dover et al, 2002; Shilatifard, 2006; Kim et al, 2013; Wu et al, 2013). However, other studies suggest that H2Bub is independent of H3 methylation (Tanny et al, 2007; Lee et al, 2012). In our study, MED23 deletion reduced H2Bub levels but did not change global levels of H3K4me3 or H3K79me3 (Fig 1), which appears to favor the view that methylation and ubiquitination are uncoupled. However, for the MED23‐targeted gene Egr1, we found that modification levels of H2Bub, H3K4me3, and H3K79me3 were decreased by MED23 depletion in coding region. Genome‐wide ChIP‐seq assays revealed that the three modifications could be coupled in gene coding regions (Figs 4A and B, and 5H and I). Thus, MED23‐dependent H2Bub appears to exhibit epigenetic crosstalk with H3 methylation only in the coding regions of the genes, and crosstalk between H2Bub and H3K4me3 or H3K79me3 may only occur in a gene region‐specific manner in vivo.
Recent studies have suggested that the Mediator complex can facilitate transcriptional elongation (Donner et al, 2010; Takahashi et al, 2011; Galbraith et al, 2013; Wang et al, 2013b), and H2Bub plays a positive role in transcription elongation through an association with the PAF complex (Fig 3C) (Zhu et al, 2005; Tanny et al, 2007; Weake & Workman, 2008). Surprisingly, our study found that corporation of Mediator and PAF complexes can enhance RNF20/40‐mediated H2B ubiquitination in vitro (Figs 3B and EV3), revealing interplays between Mediator, PAF complex, and RNF20/40 in regulating the ubiquitination level of H2B lysine 120 in transcription (Fig 8). These findings suggest a novel mechanism that Mediator and PAF1 complexes may cooperate to enhance the recruitment and affinity of RNF20/40 complex to the chromatin for increased H2Bub and efficient transcription elongation (Fig 8).
Like Mediator complex, SAGA is another large multi‐subunit transcription coactivator complex (Koutelou et al, 2010; Weake & Workman, 2012), which can also regulate both H2B ubiquitination and transcription elongation (Samara & Wolberger, 2011). However, SAGA contains a subunit Ubp8, which is an H2B deubiquitinase and can facilitate Ctk1 kinase recruitment to activate RNA polymerase II elongation (Wyce et al, 2007). The opposite regulation of H2B ubiquitination by Mediator and SAGA in transcription activation seems to be paradoxical, which may be reconciled by a possible model in which multiple rounds of H2B ubiquitination and deubiquitination provide the molecular basis for transcriptional elongation (Weake & Workman, 2008). In this model, while the SAGA complex acts as H2B deubiquitinase, our study provides the missing link by which the Mediator complex recruits the H2B ubiquitinase RNF20/40 to increase H2Bub levels for transcription elongation. However, the detailed mechanisms and dynamics of H2B mono‐ubiquitination involving multiple transcriptional cofactors remain to be fully elucidated in the future.
Aberrant chromatin state of cancer cells tightly correlated with abnormal transcriptional program. In our study, the levels of both H2Bub and MED23 are increased in cancer tissues than in normal tissues (Fig 7A), and the genes regulated by MED23‐H2Bub are enriched in cancer‐related pathways, suggesting that MED23‐dependent H2Bub could be an important novel mechanism in tumorigenesis. Consistently, the RNF20, specific H2B E3 ubiquitin ligase, is required for proliferation of leukemia cells and prostate cancer cells (Jaaskelainen et al, 2012; Wang et al, 2013a). On the other hand, we also noticed that previous studies also revealed RNF20 as a putative tumor suppressor in HeLa cells and breast cancer cells, implying that H2Bub could be in low level for cancer cells (Shema et al, 2008, 2011). Also the loss of H2Bub is reported in malignant breast cancer, colon cancer, lung cancer, and parathyroid carcinoma (Prenzel et al, 2011; Hahn et al, 2012; Urasaki et al, 2012). These seemingly contradicted literatures suggest a possibility that in different types and stages of cancer, the dynamic and distinctive mechanisms of H2Bub regulation could be involved and deserved to be further studied.
Finally, recent studies regarding H2Bub have highlighted its critical role in alternative splicing of mRNA (Jung et al, 2012; Zhang et al, 2013). Considering our previous work regarding the involvement of the Mediator subunit MED23 in alternative splicing (Huang et al, 2012), it is tempting to speculate that MED23‐dependent H2Bub may also provide a mechanism by which MED23 regulates mRNA processing.
Overall, as the core integrator of transcription, the Mediator complex coordinates diverse signaling pathways to regulate specific gene transcription temporally and spatially with high accuracy. The present study described a novel function of Mediator MED23 in coordinating the histone modification, cell lineage development, and diseases. Considering the complexity of the composition and structure of the Mediator complex, the mechanisms by which multiple subunits of the Mediator complex mediate the dynamic connection between general transcription machinery and chromatin context awaits further investigation.
Materials and Methods
Cell culture and skeleton muscle differentiation
MED23+/+ (WT), MED23−/− (KO) MEFs were isolated from day 9.5 embryos and self‐immortalized following standard procedures (Balamotis et al, 2009; Wang et al, 2009). Primary MED23F/F MEFs were isolated from day 12.5 embryos of MED23F/F mice (Yin et al, 2012). MEFs, HeLa, C2C12, and 293T cells were all maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% (v/v) FBS (Hyclone). Confluent C2C12 cells were induced to differentiate into myotubes by 2% horse serum. All cells were maintained in the incubator with 5% CO2.
Retrovirus and adenovirus infection
Retroviral‐mediated knockdown or overexpression stable cell lines were established following the manufacturer's recommendation (Clontech) and have been described previously (Huang et al, 2012). The specific target sequences of siRNA were designed using siDESIGN Center tools (Thermo Scientific). Retroviruses were generated by the co‐transfection of pSiren‐RetroQ (for knockdown) or pMSCV‐puro (for overexpression) vectors with pCL10A1 helper plasmid into 293T cells using Lipofectamine 2000 (Invitrogen). After 48 h, cell culture medium containing retroviruses were harvested, supplemented 20 μg/ml polybrene (Sigma‐Aldrich), and passed through a 0.45‐μm filter. Then, retroviruses were added to the intended cells for a centrifugation at 2,500 rpm for 1.5 h at 30°C. Twenty‐four hours after spin infection, MEFs were selected with 50 μg/ml puromycin (Sigma‐Aldrich), HeLa were selected with 2 μg/ml, and C2C12 were selected with 10 μg/ml.
Cre‐expressing adenoviruses (Ad‐cre) were generated based on commercial manual (Qbiogene). For adenoviral infection of primary MEFs, 50% confluent cell cultures were infected with Ad‐cre (multiplicity of infection 50) for 48 h in complete growth medium, and then, cells were harvested for Western blot analysis.
Plasmids, oligonucleotides, and primers
Myc‐MED23, Flag‐MED23 full‐length and deletion mutants were generated previously (Huang et al, 2012). The human RNF20 and RNF40 cDNAs were obtained from Thermo Scientific. Then, full‐length cDNA and RNF40 deletion mutants were cloned into 3xFlag‐CMV‐10 (Sigma‐Aldrich) or 5xMyc‐pcDNA3 plasmids (Invitrogen). si‐MED23, si‐RNF20, and Ctrl shRNA oligonucleotides were cloned into pSiren‐RetroQ (Clonetech) (see Table EV2 for oligonucleotide sequences).
Western blot and real‐time PCR assays
Methods for Western blot and real‐time PCR assays have been described previously (Wang et al, 2009) (see Table EV2 for primers used for real‐time PCR). Antibodies for Western blot include the following: MED23 (BD Biosciences, 550492), MED1 (Bethyl Lab, A300‐793A), MED6 (Santa Cruz, sc‐9434), MED12 (Bethyl Lab, A300‐774A), MED16 (Bethyl Lab, A303‐668A), CDK8 (BD Biosciences, 552053), RNF20 (Abcam, ab32629), RNF40 (Bethyl Lab, A300‐719A), RNF40 (Santa Cruz, sc‐102097), PAF1 (Abcam, ab20662), CTR9 (Bethyl Lab, A301‐395A), HnRNP L (Sigma‐Aldrich, R4903), CPSF5 (Proteintech, 10322‐1‐AP), STAT3 (Cell Signaling, 9139), TBP (Santa Cruz, sc‐273), Flag (Sigma‐Aldrich, F3165), Myc (Sigma‐Aldrich, C3956), γ‐tubulin (Sigma‐Aldrich, T6557), GAPDH (Proteintech, 60004‐1‐Ig), β‐actin (Sigma‐Aldrich, A5441), UBE2A (Bethyl Lab, A300‐281A), myogenin (BD Pharmingen, 556358), MHC (Millipore, 05‐716), H2B (Abcam, ab1790), H2Bub (Millipore, 05‐1312), H2Bub (Cell Signaling, 5546), H3K4me3 (Abcam, ab8580), H3K4me1 (Abcam, ab8895), H3K79me3 (Abcam, ab2621), H3K79me2 (Abcam, ab3594), H3K79me1 (Abcam, ab2886), H3 (Abcam, ab1791), H3K27me3 (Abcam, ab6002), H3K36me3 (Abcam, ab9050), H3K36me2 (Abcam, ab9049), H3K36me1 (Abcam, ab9048), H3K9me3 (Abcam, ab8898), and H2Aub (Millipore, 05‐678). Antibodies to H3K9me2, H3K9me1, H3K27me2, H3K27me1, H3K4me2, H3K9ac, H3K27ac, and H4K8ac were generous gifts from Dr. Charlie Chen (Institute of Biochemistry and Cell Biology, Shanghai). Antibodies of H3R2me2 and H4ac were generous gifts from Dr. Jinqiu Zhou (Institute of Biochemistry and Cell Biology, Shanghai).
Protein purification of MED23, two‐step affinity column purification, and MS/MS
Soluble His‐Flag‐MED23 expression, two‐step affinity column purification, and MS/MS analysis have been described previously (Huang et al, 2012).
Co‐immunoprecipitation (Co‐IP) assay
For transient co‐transfection, HeLa cells plated in 10‐cm dishes (90% confluency) were co‐transfected with 6 μg of each plasmid. After 48 h, the cells were washed with PBS buffer and lysed in 1 ml lysis buffer (20 mM Tris–Cl (pH 8.0), 135 mM NaCl, 1% NP‐40, and 10% glycerol, and freshly supplemented with 10 mM NaF, 2 mM NaVO4, and proteinase inhibitors (Roche)). Then, after a rotation of 30 min at 4°C, the lysates were spun down at 12,000 g for 5 min. The supernatant was added to 20 μl of equilibrated anti‐Flag M2 beads (Sigma‐Aldrich, A2220) and incubated at 4°C. After overnight incubation, the beads were washed three times with lysis buffer. Then, 60 μl 1× SDS loading buffer was added to the beads and boiled at 99°C for 5 min for Western blot analysis with the indicated antibodies.
As for endogenous Co‐IP, HeLa cells in 10‐cm dishes were lysed in a buffer containing 50 mM HEPES (pH 7.5), 150 mM NaCl, 1 mM EDTA, 2.5 mM EGTA, and 0.1% Tween‐20 and freshly supplemented with 1 mM PMSF and a proteinase inhibitor cocktail tablet (Roche) as previously described (Kim et al, 2005). Then, the lysates were centrifuged and the supernatant was incubated with 20 μl antibodies to RNF20 (Cell Signaling, 11974), or RNF40 (Cell Signaling, 12187), or a mixture of RNF20 and RNF40 antibodies, or MED23 (Novus, NB200‐338) or an equal amount of rabbit IgG (Cell Signaling, 2729) overnight. Then, 20 μl Dynabeads Protein G (Invitrogen) was added and incubated for 2 h at 4°C. The beads were then washed with lysis buffer three times and boiled with SDS loading buffer for Western blot assay with the indicated antibodies. For the Co‐IP experiment using CDK8 antibody (Santa Cruz, sc‐1521), 4 ml crude HeLa nuclear extract was diluted with equal volume of D100 buffer containing 20 mM HEPES pH 7.9, 100 mM KCl, 10% glycerol, 0.2 mM EDTA, 10 mM β‐mercaptoethanol, and 1 mM PMSF and incubated with CDK8 antibody or goat IgG overnight. At day 2, the mixtures were supplemented with 20 μl Dynabeads Protein G for 2 h at 4°C. The beads were then washed with D200 buffer (20 mM HEPES pH 7.9, 200 mM KCl, 10% glycerol, 0.2 mM EDTA, 10 mM β‐mercaptoethanol, 1 mM PMSF) 3 times and boiled with SDS loading buffer for Western blot using the indicated antibodies.
Nuclear extract preparation and Superose 6 gel filtration analysis
Preparation of un‐dialyzed HeLa cell nuclear extract was described previously (Wang et al, 2001). HeLa nuclear extract was centrifuged at top speed for 10 min, and supernatants were loaded onto a Superose 6 column. Then, all fractions were collected and subjected to Western blot using the indicated antibodies.
Chromatin fraction isolation
Chromatin fractionated as previously described (Wu et al, 2014), and WT and KO MEFs were harvested and suspended in buffer A (10 mM HEPES pH 7.9, 10 mM KCl, 1.5 mM MgCl2, 0.34 M sucrose, 10% glycerol, 0.1% Triton X‐100) for 5 min and then buffer B (10 mM HEPES pH 7.9, 10 mM EDTA) for 30 min. The supernatant contains both cytoplasmic and nucleoplasmic fractions. The loading buffer was then added directly to the pellet to make the chromatin fraction.
Cells plated on coverslips were washed with PBS twice and fixed using 4% paraformaldehyde in PBS for 10 min at room temperature. Then, cells were permeabilized with 0.1% Triton X‐100 for 4 min and incubated for 30 min in blocking buffer (2% BSA in PBS). Primary antibodies were diluted with blocking buffer and applied for 1 h at room temperature. Nuclei were stained with DAPI (1:1,000) for 4 min. MHC (Millipore, 05‐716) antibody was used for immunostaining.
Tissue microarray and immunohistochemistry
The tissue microarray containing a total of 55 pairs of human lung tumor and matched adjacent tissues was purchased from Shanghai Biochip Company Ltd. The primary antibodies used for immunostaining of tissue microarray were as follows: MED23 (BD Biosciences, 550492) and H2Bub (Millipore, 05‐1312). Immunohistochemistry was performed as previously described (Yang et al, 2012). Briefly, slides were deparaffinized, hydrated, and microwave retrieved. After quenching of endogenous peroxidases, slides were blocked in blocking serum and incubated with primary antibodies overnight at 4°C. Slides then were washed and incubated with secondary antibodies before being treated with HRP‐conjugated streptavidin. Finally, slides were developed using 3,3′‐diaminobenzidine tetrahydrochloride and counterstained with hematoxylin. The scoring system for immunostaining of tissue microarray has been described previously (Yang et al, 2012).
ChIP and ChIP‐seq
ChIP and ChIP‐seq analyses were performed as previously described (Wang et al, 2009; Huang et al, 2012). Antibodies for ChIP were as follows: PAF1 (Abcam, ab20662), RNF20 (Cell Signaling, 11974), RNF40 (Cell Signaling, 12187). H2Bub (Millipore, 05‐1312), Pol II (Santa Cruz, sc‐899), H3K4me3 (Abcam, ab8580), and H3K79me3 (Abcam, ab2621). For RNF20/40 ChIP assay, ChIP‐IT® High Sensitivity kit (Acivemotif) was used. ChIPed DNA was quantified by real‐time PCR, and primers were summarized in Table EV2. ChIP‐seq of H2Bub was performed by BGI (Beijing Genomics Institute, Shenzhen). For ChIP‐seq of Pol II, about 10 ng immunoprecipitated DNA was prepared for DNA library using ChIP‐seq Sample Prep Kit (Illumina). After adapter ligation and PCR amplification, 7 pmol library DNA was applied to high‐throughput sequencing using Genome Analyzer IIx (Illumina) at the Core Facility of Molecular Biology, Institute of Biochemistry and Cell Biology, Shanghai.
Global gene expression analysis was performed using Affymetrix Mouse Gene 430 2.0 array and has been described previously (Yin et al, 2012). Based on the gene expression levels in WT and KO MEFs, three groups of genes were classified (Fig 3C and D): (i) genes down‐regulated more than twofold by MED23 KO (n = 903), (ii) genes up‐regulated more than twofold by MED23 KO (n = 1,009), (iii) and genes with unchanged expression by KO (n = 1,000, selected by minimal fold change (less than 0.01‐fold) in gene expression values between WT and KO).
For ChIP‐seq analysis, sequenced tags were aligned to mouse genome using Bowtie (Langmead & Salzberg, 2012) with following parameters: ‐v1, –best, –strata, ‐k2, and ‐m1. Peak calling was performed independently for each sample using MACS algorithm (Zhang et al, 2008b) by P‐value threshold of 10−5. ChIP‐seq datasets of H2Bub, H3K4me3, and H3K79me3 was normalized to mapped reads. As the mapping ratio of Pol II ChIP‐seq of WT and KO cells was uneven, the sequencing reads of Pol II ChIP‐seq was subjected to relative normalization (i.e., using average RPM of 500 minimal fold‐change genes as background). To analyze the genome‐wide distribution of H2Bub and Pol II, the average ChIP‐seq intensity between 2 kb upstream of TSS and 2 kb downstream of TTS was calculated. For ChIP‐seq density of heatmaps, each dataset signals from −10 kb of TSS to +5 kb of TTS were rescaled by ratio relative to the corresponding Poissonian background, respectively. All ChIP‐seq analysis was performed based on the mouse genome (NCBI Build 37, mm9).
Protein purification, recombinant chromatin assembly, and in vitro H2B ubiquitination assays
All proteins except endogenous Mediator complex in H2B ubiquitylation assay were purified as described (Kim et al, 2009). Endogenous Mediator complex was purified as described (Malik & Roeder, 2003). Recombinant chromatin was assembled as described (Tang et al, 2013). In vitro H2B ubiquitylation was established as described with minor modifications (Kim et al, 2009). Briefly, complete reaction containing 100 ng E1, 100 ng E2, 200 ng RNF20/40 complex, 2.5 μg ubiquitin, and 1.0 μg histone octamer or 100 ng recombinant chromatin in 20 μl reaction buffer (50 mM Tris–Cl pH 7.9, 5 mM MgCl2, 2 mM NaF, 0.4 mM DTT, and 4 mM ATP) was incubated at 37°C for 1 h and then subjected to SDS–PAGE and Western blot using the indicated antibodies.
Microarray data in this study have been deposited previously at Gene Expression Omnibus (GEO) (http://www.ncbi.nlm.nih.gov/geo/) under accession ID GSE40286 (Yin et al, 2012). MED1 ChIP‐seq was deposited by Richard Young under accession ID GSE22557 (Kagey et al, 2010). ChIP‐seq of H2Bub, Pol II, H3K4me3, and H3K79me3 are available for download from GSE59518.
XY and GW designed the study; XY performed the most experiments; ZT performed the in vitro H2Bub assay; JY and YL performed the myogenesis‐related experiments; XF analyzed the NGS data; CL and HL carried out Co‐IP assays; QT contributed to cell culture; and XY, RGR, and GW wrote the manuscript.
Conflict of interest
The authors declare that they have no conflict of interest.
Expanded View Figures PDF
Source Data for Expanded View
We thank Drs. Sohail Malik (Rockefeller), Bin Tian (Rutgers), Zhen Shao (PICB), Luonan Chen (SIBCB), Dangsheng Li (Cell Research), Ronggui Hu (SIBCB), Ming Lei (SIBCB), and Ping Hu (SIBCB) for comments and helpful suggestion; Drs. Charlie Chen and Jinqiu Zhou at SIBCB for antibodies of histone modifications; and Dongming Hou for technical assistance. This study was supported in part by grants from the Chinese Academy of Sciences (XDA01010401), the Ministry of Science and Technology of China (2011CB510104, 2014CB964702), and the National Natural Science Foundation of China (81030047). GW is a scholar of the Hundred Talent Program.
FundingChinese Academy of SciencesXDA01010401
- © 2015 The Authors