Report

Akt-Mediated Phosphorylation of EZH2 Suppresses Methylation of Lysine 27 in Histone H3

See allHide authors and affiliations

Science  14 Oct 2005:
Vol. 310, Issue 5746, pp. 306-310
DOI: 10.1126/science.1118947

Abstract

Enhancer of Zeste homolog 2 (EZH2) is a methyltransferase that plays an important role in many biological processes through its ability to trimethylate lysine 27 in histone H3. Here, we show that Akt phosphorylates EZH2 at serine 21 and suppresses its methyltransferase activity by impeding EZH2 binding to histone H3, which results in a decrease of lysine 27 trimethylation and derepression of silenced genes. Our results imply that Akt regulates the methylation activity, through phosphorylation of EZH2, which may contribute to oncogenesis.

Posttranslational modification of histones plays a central role in the regulation of many biological functions (1). Histone methylation, in particular, is linked to the regulation of gene expression and chromatin conformation (24). This process, which used to be considered a permanent modification, has recently been shown to be reversible, and the enzymes responsible for demethylation have been identified (58). Perturbation of this epigenetic balance has important links to carcinogenesis (9). For instance, overexpression of EZH2, a central member of Polycomb repressive complexes (PRCs) that has intrinsic histone methyltransferase (HMTase) activity (10, 11), has been implicated in cancer progression and metastases in multiple cancer types (1214).

The phosphoinositide 3-kinase–Akt (PI3K-Akt) signaling pathway is involved in processes such as survival, proliferation, growth, and motility (15, 16). In an attempt to identify potential upstream regulators for HMTase of EZH2, we unexpectedly found that cells with high levels of activated, and therefore phosphorylated, Akt generally have lower levels of histone 3 (H3) lysine 27 (K27) trimethylation (P < 0.05; fig. S1). This inverse correlation prompted us to test whether alteration of Akt activity may change H3 K27 trimethylation. Indeed, insulin-like growth factor (IGF)–induced activation of Akt by phosphorylation reduced H3 K27 trimethylation, which could be blocked by the PI3K-Akt inhibitor, LY294002 (Fig. 1, A and B; fig. S2). Similarly, expression of dominant-negative Akt (DN-Akt) (17) (Fig. 1C) or knocking down Akt expression with small interfering RNA (siRNA) (Fig. 1D; fig. S3) also enhanced H3 K27 trimethylation. The H3 K27 trimethylation was further shown to be inversely correlated with Akt activity over time (fig. S4).

Fig. 1.

Akt negatively regulates H3 K27 trimethylation. Cell lysates from cell lines as marked on the top of each panel were analyzed by Western blotting by using antibodies indicated to the left of each panel. (A) T47D cells were pretreated with IGF (20 ng/ml) for 30 min before LY294002 (20 μM) treatment for 24 hours. (B) Cells were treated with or without 20 μM LY294002. (C) Two pairs of wild-type and stable DN-Akt transfectants (MDA-MB453/DN-Akt, HER2-neu-3T3/DN-Akt-3T3) were examined for the level of H3 K27 trimethylation. (D) Immunoblot analysis of H3 K27 trimethylation in MDA-MB453 cells transfected with control (ctrl) or Akt siRNA for 48 hours. Detailed experimental conditions are in (19).

To address whether EZH2 might be a target of Akt, we performed coimmunoprecipitation experiments and detected an association between Myc-tagged EZH2 and hemmagglutinin (HA)-tagged Akt (Fig. 2A). Both wild-type (WT)- and ΔSET-EZH2 (without the SET domain) associate more strongly with constitutively activated Akt (CA-Akt) than with DN-Akt. We also found an association between endogenous Akt and EZH2, and this association was reduced in cells treated with LY294002 (Fig. 2B; fig. S5), which indicates that these two molecules interact in vivo.

Fig. 2.

Akt interacts with and phosphorylates EZH2 on Ser21. (A) Coimmunoprecipitation assay of EZH2 and Akt. Lysates from 293T cells transfected with the indicated plasmids were immunoprecipitated with antibodies against HA or Myc. Whole-cell lysate (lysate) was also subjected to Western blot analysis to examine the expression of EZH2 and Akt. (B) Examination of the interaction between endogenous EZH2 and Akt by immunoprecipitation and Western blotting with antibodies against either EZH2 or Akt in MDA-MB453 cells (±LY294002). (C) Comparison of the amino acid sequences of the Akt-phosphorylaton motifs of EZH2 and other known Akt substrates. (D) EZH2 was immunoprecipitated from 293T cells cotransfected with Myc-tagged WT-EZH2 and either CA-Akt or DN-Akt, and then subjected to Western blot analysis with an Akt substrate antibody. (E) HA-tagged Akt was expressed and immunoprecipitated from 293T cells. The kinase reaction is described in (19). (F) Endogenous EZH2 was immunoprecipitated by an antibody against EZH2 in MDA-MB453 (±LY294002) or in DN-Akt/MDA453 cells and subjected to Western blot with an antibody against phospho-EZH2. (G) After immunoprecipitation of phosphorylated endogenous EZH2 by the antibody against phosphoEZH2, lysates were analyzed by Western blot with an antibody against EZH2. (H) EZH2 was expressed and immunoprecipitated from 293T cells transfected with wild-type (±LY294002) or mutant EZH2 and subjected to Western blot with an antibody against phospho-EZH2.

We noticed that there was one potential Akt phosphorylation site, serine 21, in the EZH2 protein and that this site was highly conserved across species (Fig. 2C). To examine whether Akt can phosphorylate EZH2, we immunoprecipitated EZH2 and probed it with an antibody that recognizes the phosphorylated form of Akt substrates. EZH2 was phosphorylated in CA-Akt–expressing cells three times as much as in DN-Akt (Fig. 2D). An in vitro kinase assay also demonstrated that Akt phoshorylated wild-type EZH2 but not variants of EZH2 in which Ser21 was replaced by Ala or Asp (S21A and S21D, respectively) and glutathione S-transferase (GST) (Fig. 2E). To further confirm that Akt can indeed phosphorylate EZH2 at Ser21 in vivo, we generated a polyclonal antibody that specifically recognized an EZH2 peptide harboring phosphorylated Ser21 but failed to detect an unphosphorylated EZH2 peptide or an unrelated peptide (fig. S6). This antibody was used to show that Ser21 phosphorylation of EZH2 is correlated with the presence of activated, phosphorylated Akt in multiple cell lines (fig. S7). To further test whether EZH2 phosphorylation is regulated by Akt, endogenous EZH2 from MDA-MB453 cells with or without LY294002 treatment or from DN-Akt/MDA453 cells was immunoprecipitated and probed with the phospho-EZH2 antibody. The phosphorylation of EZH2 was detected in wild-type MDA-MB453 cells but not in cells treated with LY294002 or in DN-Akt/MDA453 cells (Fig. 2F). When the same lysates were subjected to immunoprecipitation with phospho-EZH2 antibody followed by Western blotting with EZH2-specific antibody, blocking Akt activity decreased the intracellular level of phosphorylated EZH2 (Fig. 2G). A similar result was obtained in human embryonic kidney (HEK) 293T cells transiently transfected with wild-type or mutant EZH2 (Fig. 2H). Together, these results indicate that Akt physically interacts with EZH2 and phosphorylates it at Ser21 in vivo.

Next, we examined whether Akt-mediated phosphorylation of EZH2 inhibits K27 trimethylation of histone H3. We observed that DN-Akt enhanced H3 K27 trimethylation and CA-Akt suppressed it (Fig. 3A). Also, H3 K27 trimethylation level was indeed 3 to 5 times that in cells transfected with S21A-EZH2 (mimicking the unphosphorylated state), than cells transfected with WT-EZH2 and S21D-EZH2 (mimicking the phosphorylated state) (Fig. 3B). Moreover, knock-down of EZH2 by siRNA rendered DN-Akt unable to enhance H3 K27 trimethylation (Fig. 3C). Similar results were also obtained by immunofluorescence analysis of K27 trimethylation on transfected 293T cell lines (fig. S8). Thus, phosphorylation of EZH2 at Ser21 by Akt decreases H3 K27 trimethylation.

Fig. 3.

Akt suppresses EZH2 HMTase activity. Detection of various proteins was measured by Western blotting with antibodies shown to the left of each panel. (A) 293T cells were cotransfected with EZH2 and CA-Akt or DN-Akt. (B) Wild-type or mutant EZH2 (S21A, S21D) was transfected into 293T cells. (C) DN-Akt/MDA453 cells were transfected with control (ctrl) or EZH2 siRNA. (D) In vitro HMTase assays were performed with immunoprecipitated wild-type or mutant Myc-tagged EZH2 incubated with recombinant histone H3 and 3H-labeled S-adenosylmethionine (SAM). The quantification of enzyme activity with error bars from three independent experiments is shown. (E) Immunoprecipitated wild-type or mutant Myc-tagged EZH2 or vector was incubated with histone H3 in the presence of cold SAM; H3 K27 trimethylation was detected by Western blotting analysis. (F) Wild-type or mutant EZH2 immunoprecipitates isolated from 293T cells (±LY294002) were subjected to an in vitro methyltransferase assay by using native nucleosome as substrate (upper panel). The amount of protein in the methyltransferase reactions was controlled by Coomassie blue staining (lower panel). Expression of transfected Myc-tagged EZH2 was measured by Western blotting using an antibody against Myc.

To further investigate whether this phenomenon is due to alteration of HMTase activity after EZH2 is phosphorylated by Akt, we performed in vitro HMTase activity assays. S21A-EZH2 exhibited 3 to 5 times the HMTase activity of WT-EZH2 and S21D-EZH2 when we used either recombinant histone H3 (Fig. 3, D and E) or oligonucleosomes (Fig. 3F) as substrates. In addition, inhibition of Akt activity by LY294002 also enhanced EZH2 HMTase activity 5.7 times in vitro (Fig. 3F).

Akt-mediated EZH2 phosphorylation did not change its subcellular localization or its interaction with Polycomb group (PcG) proteins Suz12 and Eed (fig. S9, supporting online text), rather, phosphorylation altered the affinity of EZH2 for its substrate, histone H3. Immunoprecipitation experiments showed that S21A-EZH2 has greater association with H3 than WT-EZH2 or S21D-EZH2 had by a factor of 2 to 4; and treatment with LY294002 enhanced the association between H3 and WT-EZH2 (Fig. 4A). In addition, we isolated soluble and insoluble (chromatin-bound) nuclear fractions from 293T cells that were transfected with wild-type or mutant EZH2. S21A-EZH2 predominantly existed in an insoluble nuclear fraction, which suggested that it strongly associated with chromatin (Fig. 4B). We further investigated the chromatin association of endogenous EZH2 in cells in which Akt activity was either induced or blocked. Blockage of Akt consistently increased the chromatin-bound EZH2, whereas activation of Akt had the opposite effect (Fig. 4, D and E; fig. S10).

Fig. 4.

Akt-mediated phosphorylation of EZH2 changes substrate affinity. (A) Coimmunoprecipitation between endogenous histone H3 and transfected wild-type or mutant EZH2 in 293T cells. (B) Western blot analysis of soluble (C) and chromatin-bound (B) fractions from cells in response to exogenous expression of wild-type or mutant EZH2. (C) Expression of HOXA9 in response to exogenous expression of wild-type or mutant EZH2 in 293T cells. (D and E) Western blot analysis of endogenous EZH2 in soluble (C) and chromatin-bound (B) fractions from MDA-MB453 (±LY294002) and DN-Akt/MDA453 cells (D) or from T47D cells (E) treated with IGF (20 ng/ml) or the combination of IGF and LY294002. The loading protein amounts were 150 μg (total ∼2000 μg) for soluble and 40 μg (total ∼150 μg) for chromatin-bound fraction, respectively. (F and G) Chromatin immunoprecipitation analysis on HOXA9 promoter by using antibodies as indicated to the left of the panel were performed in HeLa cells transfected with wild-type or mutant EZH2, (F) in MDA453 cells (±LY294002) and (G) in DN-Akt stable transfectants. The primers used for PCR are described in (19).

To determine whether EZH2 phosphorylation mediated by Akt is functionally important, we tested whether it affects gene expression. Reverse transcription polymerase chain reaction (RT-PCR) showed that expression of HOXA9, a known EZH2-targeted gene (18), was suppressed by S21A-EZH2 (Fig. 4C), and this result was consistent with the high H3 K27 trimethylation on the HOXA9 promoter (Fig. 4F). This phenomenon was also observed in MDA453 cells treated with LY294002 or in DN-Akt stable transfectants (Fig. 4G). Similar results were observed when other PcG proteins such as Suz12 and Eed were examined by chromatin immunoprecipitation (ChIP) assay (Fig. 4G). Thus, phosphorylation of EZH2 on Ser21 by Akt does not seem to change the composition of the PRC, but it does reduce the affinity of EZH2 toward histone H3, which results in a decrease in H3 K27 trimethylation and derepression of genes normally silenced by EZH2.

Our study demonstrates that the HMTase activity of EZH2 responsible for H3 K27 trimethylation is regulated by the Akt signaling pathway through phosphorylation of EZH2. Phosphorylation of EZH2 suppresses H3 K27 trimethylation and disrupts gene silencing and thus may contribute to oncogenesis. Furthermore, S21D-EZH2 enhanced cell growth in culture and tumor development in animals, and phosphorylated EZH2 also correlated with Ki67, a proliferative marker in primary tumor tissues (figs. S11 and S12, tables S1 and S2, supporting online text) (19). Although Akt-mediated phosphorylation of EZH2 reduces its affinity toward histone H3, it did not compromise PRC composition, which is known to be responsible for its methyltransferase activity: The phosphorylated-EZH2 complex, instead, may target nonhistone substrates that are important for tumorigenicity. In this scenario, the dynamic changes in EZH2 substrate affinity induced by Akt may provide an explanation for why EZH2, which is overexpressed in cancers, has less HMTase activity toward histone H3.

Supporting Online Material

www.sciencemag.org/cgi/content/full/310/5746/306/DC1

Materials and Methods

SOM Text

Figs. S1 to S12

Tables S1 and S2

References and Notes

References and Notes

View Abstract

Navigate This Article