O-GlcNAcylation regulates the stability and enzymatic activity of the histone methyltransferase EZH2

Significance

Glycosylation is considered to be a major posttranslational modification, and O-GlcNAcylation is known to affect protein folding and function. In this study, we show that the methyltransferase EZH2, which catalyzes the methylation of histone 3 at lysine 27 to form H3K27m3, requires O-GlcNAcylation to enhance its stability and enzymatic activity to promote tumor progression. We further show that the O-GlcNAcylation in the N-terminal region of EZH2 stabilizes the enzyme and the O-GlcNAcylation at S729 in the catalytic domain is essential for its activity of di- and trimethylation. This study indicates that selective inhibition of EZH2 O-GlcNAcylation may suppress the methylation of H3K27 and thus inhibit tumor progression.

Abstract

Protein O-glycosylation by attachment of β-N-acetylglucosamine (GlcNAc) to the Ser or Thr residue is a major posttranslational glycosylation event and is often associated with protein folding, stability, and activity. The methylation of histone H3 at Lys-27 catalyzed by the methyltransferase EZH2 was known to suppress gene expression and cancer development, and we previously reported that the O-GlcNAcylation of EZH2 at S76 stabilized EZH2 and facilitated the formation of H3K27me3 to inhibit tumor suppression. In this study, we employed a fluorescence-based method of sugar labeling combined with mass spectrometry to investigate EZH2 glycosylation and identified five O-GlcNAcylation sites. We also find that mutation of one or more of the O-GlcNAcylation sites S73A, S76A, S84A, and T313A in the N-terminal region decreases the stability of EZH2, but does not affect its association with the PRC2 components SUZ12 and EED. Mutation of the C-terminal O-GlcNAcylation site (S729A) in the catalytic domain of EZH2 abolishes the di- and trimethylation activities, but not the monomethylation of H3K27, nor the integrity of the PRC2/EZH2 core complex. Our results show the effect of individual O-GlcNAcylation sites on the function of EZH2 and suggest an alternative approach to tumor suppression through selective inhibition of EZH2 O-GlcNAcylation.

Protein glycosylation is an important posttranslational modification, of which the addition of N-acetylglucosamine (GlcNAc) to the Ser or Thr residue (O-GlcNAcylation) without further glycosylation is commonly found in animals and plants (1). The addition and removal of O-GlcNAc by O-linked N-acetylglucosaminyltransferase (OGT) and O-linked N-acetylglucosaminidase (OGA) on nuclear or cytosolic proteins are keys to maintain the normal functions of many proteins, including nuclear pore complexes, transcription factors, dosage compensation complexes, proteasomes, kinases, neuronal proteins, and mitochondria proteins, etc. (1). Changes in the status of protein O-GlcNAcylation can influence their downstream biological processes and thus may affect the onset of chronic diseases and cancer progression (2, 3).

The polycomb-group proteins (PcGs) are a series of proteins related to embryonic development, including OGT, PRC1, and PRC2. PRC1 is the ubiquitin ligase of H2AK119, and PRC2 containing the methyltransferase EZH2 is responsible for the methylation of H3K27. PcGs are recruited to the polycomb-group response elements (PREs) to regulate the expression of homeotic genes (HOX) which encode a set of transcription factors that specify the anterior–posterior axis and segment identity in the embryonic development of Drosophila (4–6). PRC1 and PRC2 are conserved in mammalian species and involved in the progression of several types of cancer (7, 8). In Drosophila, PRC1 is composed of Polycomb (Pc), Posterior sex combs (Psc), Drosophila RING (dRING), and Polyhomeotic (Ph) (7, 8). Interestingly, Super sex combs (sxc), one of the PcG genes, encodes Drosophila OGT (9, 10) and is necessary for the repression of multiple HOX genes in Drosophila larvae (11, 12). A genome-wide profiling reveals that the PREs bound by OGT are highly associated with the regions targeted by PRC1 (9, 13). The subunits of PRC1, Ph and RING, are found to be O-GlcNAcylated to prevent Ph from aggregation and also to affect pluripotency maintenance and differentiation in embryonic stem cells (14, 15). It was suggested that O-GlcNAcylation might play an important role in the regulation of PRC1-mediated gene expression, and along this line the O-GlcNAcylation of EZH2 at S76 in the PRC2 complex was reported to stablize EZH2 in our previous study (16). The PRC2 complex is composed of Enhancer of zeste 2 (EZH2), Suppressor of Zeste 12 (Suz12), Extraembryonic endoderm (EED), AE binding protein 2 (AEBP2), and retinoblastoma binding protein 4/7 (RBBP4/7) (17, 18). Within the PRC2 complex, EZH2 catalyzes the di- and trimethylation of histone H3 at lysine 27 (K27) to form H3K27me2/3 to regulate embryonic and cancer development (19–23). In contrast to H3K27me2/3, histone H3 with monomethylation at K27 (H3K27me1) contributes to the promotion of gene transcription (24), but the mechanism of H3K27me1 formation in vivo is still ambiguous. In this study, we identified five more O-GlcNAcylation sites on EZH2, using a method of fluorescence labeling and mass spectrometry, and revealed that O-GlcNAcylation mediates EZH2 function in a glycosite-dependent manner.

Results

Additional O-GlcNAcyaltion Sites on EZH2 Other than S76.

We previously found that the O-GlcNAcyaltion of EZH2 occurred at S76 (equivalent to S75 if ignoring the first amino acid Met) and the glycosylation increased the protein stability (16). However, the S76A mutant of EZH2 still showed the O-GlcNAcyaltion signal as detected by Western blot. To enhance the signal, we labeled the O-GlcNAcylation sites of EZH2 expressed in 293T cells using a peracetylated alkyne-modified GlcNAc analog (Ac₄GlcNAc) as a substrate, followed by copper(I)-catalyzed azide-alkyne cycloaddition (CuAAC) of the pulled-down EZH2 using azido-biotin, azido-TAMRA, or azido BODIPY dye (Az2) (25) (Fig. 1 A and B and SI Appendix, Fig. S1 A and B). Using these azido probes to analyze and characterize the fluorescent triazole adduct through in-gel analysis was very convenient compared with Western blot, and of these probes, Az2 was found to be better in terms of fluorescent stability and intensity, and was therefore further exploited in the following experiments. Based on the result of metabolic labeling (Fig. 1C), we found that the S76A mutation only caused a slight reduction of O-GlcNAcylation on EZH2, and the same was observed with OGT overexpression to increase the protein level of both EZH2 wild type (WT) and the mutant EZH2 S76A (SI Appendix, Fig. S2). These data suggest the presence of other O-GlcNAcylation sites on EZH2 besides S76.

Fig. 1.

EZH2 has other O-GlcNAcylation sites in addition to S76. (A) The flowchart of GlcNAl metabolic incorporation detected by probe Az2. (B) The chemical structure of Azido-BODIPY dye (AZ2) reporter used in A. (C) There are other O-GlcNAcylation sites residing on EZH2 besides S76. The EZH2 proteins were purified from 293T cells overexpressed with EZH2 wild type or S76A. The cells were treated with Ac₄GlcNAl overnight before protein extraction. Then the O-GlcNAcylation level was examined by in-gel fluorescent assay using Az2 as shown in A. WT, wild type. Band intensities were measured by ImageJ. The quantity was determined by dividing the fluorescent signal to the signal of individual protein stain.

O-GlcNAcyaltion Distributes over the Whole Protein of EZH2.

To identify other unknown O-GlcNAcylation sites, we prepared five EZH2 truncated fragments based on the domain structure. It was found that the wild-type EZH2 and the truncated fragments including the N-terminal, the middle, and the C-terminal fragments exhibited the O-GlcNAcylation signals (SI Appendix, Fig. S3). However, the expression level of fragment 612–746 was too low to be immunoprecipitated. Likewise, the O-GlcNAcylation level of the middle and the C-terminal fragments was higher than that of the N-terminal fragment which contained the S76 residue (SI Appendix, Fig. S3). This result indicates that EZH2 has multiple O-GlcNAcylation sites.

O-GlcNAcylation Occurs at S73, S84, S87, T313, and S729 of EZH2.

Next, we determined the O-GlcNAcylation sites of EZH2 by liquid chromatography-electrospray ionization-mass spectrometry (LC-ESI-MS). In the beginning, we used the O-GlcNAcylated peptide enrichment method to detect the O-GlcNAcylation sites on EZH2 using the azido-biotin probe as described in SI Appendix, Fig. S4A. The MS spectrum revealed that the O-GlcNAcylation occurred at S76 (SI Appendix, Fig. S4B), which is consistent with our previous findings. Since EZH2 S76A still contained other O-GlcNAcylation sites as shown in Fig. 1C, we decided to identify the other O-GlcNAcylation sites of EZH2 from the products of Az2-CuAAC reaction to generate Az2-GlcNAl-EZH2 and two other O-GlcNAcylation–related modifications, GlcNAc- and GlcNAl-EZH2 (SI Appendix, Fig. S5A). The MS analysis of Az2-labeled EZH2 indicated an O-GlcNAcylation at T313, shown as Az2-GlcNAl signal (SI Appendix, Fig. S5B). Interestingly, either EZH2 S76A or T313A, or the double mutant had a similar O-GlcNAcylation level (SI Appendix, Fig. S6), consistent with our speculation that O-GlcNAcylation on EZH2 occurred transiently and dynamically, and therefore the MS analysis might not reveal all of the O-GlcNAcylation sites simultaneously (SI Appendix, Fig. S3).

We then overexpressed EZH2 in 293T cells for further detection of other possible O-GlcNAcylation sites because the protein level of endogenous EZH2 was too low for MS analysis. The O-GlcNAcylation of endogenous EZH2 can be detected when cells are treated with the OGA inhibitor PUGNAC (16), but we could not detect the O-GlcNAcylation on exogenous EZH2 by using Western blot (SI Appendix, Fig. S7). Perhaps the level of endogenous OGT within cells was not high enough for the O-GlcNAcylation of exogenous EZH2 to the level for MS analysis. Next, we examined whether OGT overexpression could enhance the O-GlcNAcylation level by introducing a sugar probe to the GlcNAc moiety using the Gal-T1 (Y289L) labeling method (26) (Fig. 2 A and B). The result showed that the O-GlcNAcylation level of overexpressed EZH2 was enhanced when OGT was co-overexpressed (Fig. 2C), as shown in the Western blot analysis (SI Appendix, Fig. S7). Next, we evaluated the O-GlcNAcylation sites on EZH2 co-overexpressed with OGT by MS and found three peptides showing the O-GlcNAcylation signal (SI Appendix, Figs. S8A, S9A, and S10). In addition, overexpression of OGT increased the O-GlcNAcylation on the IQPVHILTSVSSLR fragment to 25.99%, and the ECSVTSDLDFPTQVIPLK fragment and the S729 glycosite to 42.19% and 0.75%, respectively (Fig. 2D). In addition, there was a tiny proportion (about 0.061%) of the peptide ECSVTSDLDFPTQVIPLK that possessed two GlcNAc moieties. Since the O-GlcNAc moiety was labile in higher-energy collisional dissociation (HCD)-MS/MS analysis (27, 28) (SI Appendix, Figs. S8A and S9A), the O-GlcNAcylation sites at S73 and S84 were determined by electron-transfer dissociation (ETD)-MS/MS (SI Appendix, Figs. S8B and S9B), and the O-GlcNAcylation at S87 was determined by ETD-MS/MS (SI Appendix, Fig. S9C). On the basis of these results, we found six O-GlcNAcylation sites with different levels of signal in EZH2 (Fig. 2E).

Fig. 2.

O-GlcNAcylation sites at S73, S84, S87, and S729 of EZH2 were determined by MS. (A) The flowchart of detection of the O-GlcNAcylation level of EZH2 by GalT1 Y289L labeling method using UDP-GalNAz. (B) The chemical structure of azido-biotin probe used in A. (C) OGT overexpression increases the O-GlcNAcylation level of EZH2. The overexpressed EZH2-FLAG was purified from 293T cells with or without OGT co-overexpression, followed by GalT1 Y289L labeling using UDP-GalNAz as shown in A. (D) OGT overexpression enhances the O-GlcNAcylation level of three peptides containing S73, S84, and S729 into different ratio. The signal was quantitatively determined with LC-MS by dividing the signal of indicated O-GlcNAcylated peptide to the signal of total indicated ones, n = 3. (E) The O-GlcNAcylation sites of EZH2, including S73, S76, S84, S87, T313, and S729. DNMT, DNA methyltransferase; EBD, EED binding domain; NLS, nuclear location signal.

O-GlcNAcylation in the N-Terminal Region of EZH2 Contributes to Protein Stability.

Next, we evaluated whether these newly discovered O-GlcNAcylation sites were related to EZH2 stability, since it has been known that O-GlcNAcylation contributed to the stability of EZH2 in our previous study (16). We excluded the examination on S87, since the content of O-GlcNAcylation at S87 was very low (∼0.061%). Both S76A and T313A were found to reduce the stability of EZH2 compared with the wild type (Fig. 3A and SI Appendix, Fig. S11A), but there was no statistical difference in the EZH2 half-life between S76A and T313A mutants (Fig. 3A). On the other hand, the single, double, and triple mutants of EZH2 at S73, S84, and S729 were all found to reduce the protein stability compared to the wild type (Fig. 3B and SI Appendix, Figs. S11B and S12A). Moreover, we found that the single or double mutation on S73 and S84 had more impact on the half-life of EZH2 S729A (SI Appendix, Fig. S12B), and the effect of S73A was equivalent to S84A (SI Appendix, Fig. S12 C and D). This result indicates that the stability of EZH2 is mainly regulated by the O-GlcNAcylation at S73 and S84, rather than at S729.

Fig. 3.

O-GlcNAcylation at S73 and S84 may increase EZH2 stability. (A and B) Mutations of O-GlcNAcylation sites reduce the half-life of EZH2. The half-life of EZH2-FLAG wild type (WT) or mutants is shown in A (S76A, T313A, S76A/T313A, and wild type) or (B) (S73A, S84A, S729A, and wild type). EZH2 WT and mutants were transfected to 293T cells for 2 d and subsequently treated with cycloheximide at the final concentration of 50 μg/mL. The protein lysates were harvested at the indicated time points for Western blot using proper antibodies. (C) OGT overexpression increases the protein level of isolated EZH2. EZH2 was co-overexpressed with/without OGT, EED, and SUZ12 in 293T. The protein lysates were subjected to Western blot using proper antibodies. (D) MG132 treatment rescues the decreased protein level of EZH2 S73A/S84A. EZH2 WT and S73A/S84A were transfected to 293T for 2 d and subsequently treated with MG132 at the final concentration of 25 μg/mL. Then the lysates were harvested at the indicated time points for Western blot using the indicated antibodies. Band intensities were measured by ImageJ. The protein quantity was determined by dividing the signal of EZH2-FLAG to the signal of β-actin. The results are represented as mean ± SD. *P value <0.05, **P value <0.01. n.s., no significant difference. n = 5 in A and B; n = 3 in C.

O-GlcNAcylation Stabilizes Isolated EZH2 but Not EZH2 in the PRC2 Complex.

SUZ12 has been known to contribute to the stability of EZH2 (18). We found that overexpression of SUZ12, EED, or OGT increased the protein level of EZH2 (SI Appendix, Fig. S13). Furthermore, overexpression of OGT augmented the protein level of isolated EZH2 to 7.7- to 9.5-fold (Fig. 3C). However, overexpression of OGT had less effect on EZH2 when co-overexpressed with EED (2.2-fold) (Fig. 3C and SI Appendix, Fig. S14A), SUZ12 (1.4-fold), or both EED and SUZ12 (no significant difference) (Fig. 3C and SI Appendix, Fig. S14B). This result indicated that the O-GlcNAcylation contributed mainly to the stability of isolated EZH2 but not the EZH2 in the complex. Further, we evaluated whether ubiquitin-proteasome degradation was associated with the impaired stability of EZH2 N-terminal mutants. The treatment of proteasome inhibitor MG132 was able to rescue the reduced EZH2 protein level caused by the S73A/S84A mutation (Fig. 3D). In addition, the polyubiquitylation level of EZH2 S73A/S84A was also increased compared with the wild type (SI Appendix, Fig. S15), all indicating that O-GlcNAcylation in the N-terminal region stabilized EZH2 by preventing it from proteasomal degradation.

O-GlcNAcylation on EZH2 Does Not Affect Its Association with the PRC2 Complex.

S73 is conserved in mammalian species, and S76 and S84 are conserved in vertebrates (SI Appendix, Fig. S16A). Although the ratio of O-GlcNAcylation at S87 is very low, this site is highly conserved in chordate (SI Appendix, Fig. S16A). The four O-GlcNAcylation sites are located in the β-addition motif (BAM), which is composed of three β-strands packed against the side of the β-propeller fold of the WD40 repeats of EED (29) (SI Appendix, Fig. S16C). Another N-terminal O-GlcNAcylation site related to the protein stability is T313, which is conserved in mammalian species (SI Appendix, Fig. S16B) and located in the domain named “Motif Connecting SANT1L and SANT2L” (MCSS) that bundles with the N-terminal loop of the VEFS domain of SUZ12 (SI Appendix, Fig. S16D) to hold the EED and the SET domain together (29). On the basis of the structure, we speculated that the O-GlcNAcylation in the N-terminal region of EZH2 might be related to its association with the other two components of PRC2 complex, SUZ12 or EED. Nevertheless, the EZH2 single, double, or triple mutants containing S73A, S76A, S84A, T313A, and/or S729A did not affect the formation of the PRC2 core complex composed of EZH2, SUZ12, EED, and RBBP4/7, nor did these mutated O-GlcNAcylation sites affect EZH2 interacting with SUZ12 or EED (SI Appendix, Fig. S17 A and B). In addition, co-overexpression of OGT did not influence the interaction of EZH2 wild type or S73A/S84A mutants with SUZ12 or EED (SI Appendix, Fig. S18 A and B). Overall, these results suggest that the O-GlcNAcylation on EZH2 does not affect the integrity of the PRC2 complex.

S729A Mutation Diminishes the Methyltransferase Activity of EZH2 to Form H3K27me2/3 but Has No Effect on the Formation of H3K27me1.

S729 is located at the SET domain, the methyltransferase domain, of EZH2 as shown in Fig. 2E. Therefore, we speculated that the O-GlcNAcylation at S729 might interfere with the methyltransferase activity of EZH2. In the PRC2 complex, EZH2, SUZ12, EED, and RBBP4/7 form the core complex to exhibit the minimal enzymatic activity toward the mono-, di-, or trimethylation of H3K27, while AEBP2 and RBBP4/7 are required for the optimal methyltransferase activity (17, 18). We then investigated the methyltransferase activity of the PRC2 core complex with EZH2 wild type and mutants (SI Appendix, Fig. S17 A and B) and found that the core complex containing the EZH2 mutant S729A lost the di- and trimethylation activities on H3K27 (Fig. 4 B and C and SI Appendix, Fig. S19C), but still retained a reduced monomethylation activity (Fig. 4A and SI Appendix, Fig. S19C). Furthermore, we evaluated the mutations of the other two O-GlcNAcylation sites, S76 and T313, and as predicted, the core complex containing EZH2 S76A or S76A/T313A mutants did not show changes in the enzymatic activity, while the T313A mutation enhanced the enzymatic activity slightly (SI Appendix, Fig. S19 A and B). This result suggests that only the O-GlcNAcylation occurring in the SET domain is related to the methyltransferase activity, and the O-GlcNAcylation at S729 in the EZH2 SET domain is associated with the methyltransferase activity to form H3K27me2/3.

Fig. 4.

EZH2 S729A diminishes the methyltransferase activity to form H3K27me2/3. (A–C) EZH2 S729A showed a reduced methyltransferase activity for the formation of H3K27me2/3 but had no effect on the formation of H3K27me1. In vitro histone methyltransferase (HMT) assays of the PRC2 core complexes containing EZH2 wild type (WT) or mutants were performed for quantification as indicated. The results of HMT assay were evaluated by Western blot using antibodies against H3K27me1 (A), H3K27me2 (B), or H3K27me3 (C). Band intensities were measured by ImageJ. The quantity was determined by dividing the signal of H3K27me1, -2, or -3 to the signal of H3. n = 3. *P value <0.05, **P value <0.01.

Discussion

The O-GlcNAcylation of EZH2 is a dynamic and transient process and it is difficult to detect all glycosites constantly. We have used the sensitive glycosylation probes together with the overexpression and mass spectrometry techniques to identify five O-GlcNAcylation sites in EZH2 and elucidated the role of individual glycosites. We have found four O-GlcNAcylation sites, S73, S76, S84, and S87 (Fig. 2E and SI Appendix, Fig. S16C) in the N-terminal region of EZH2 and the O-GlcNAcylation in this region seems to stabilize EZH2 (Fig. 3B and SI Appendix, Fig. S11B) from ubiquitin-proteasome degradation (Fig. 3 D and E). In addition, our results show that O-GlcNAcylation in the BAM region do not affect EZH2’s association with EED or SUZ12 (SI Appendix, Fig. S18 A and B), and OGT-mediated protein stability merely contributes to the isolated EZH2 but not the EZH2 within the PRC2 complex (Fig. 3C). These results suggest that O-GlcNAcylation in the BAM domain of EZH2 is important for the stabilization of isolated EZH2 before the formation of PRC2 complex.

Although PRC2/EZH2 is known to catalyze the mono-, di-, and trimethylation of H3K27 in vitro (19), the role of PRC2/EZH2 in the formation of H3K27me1 in vivo is still ambiguous. Previous study indicated that the formation of H3K27me1 was catalyzed by G9a, a well-known histone methyltransferase of H3K9 (30, 31). However, no difference in di- and trimethylation of H3K27 was observed between wild-type and G9a knockout cells even if the extent of monomethylation was significantly decreased in G9a knockout cells. Thus, G9a is thought to compensate the loss of EZH2 for the formation of H3K27me1 in vivo (32). In this study, we found that the methyltransferase activity of the S729A mutant for the formation of H3K27me2/3 was significantly reduced (Fig. 4), suggesting that the S729 of EZH2 or its O-GlcNAcylation would assist PRC2/EZH2 in the formation of H3K27me2/3 from H3K27me1. The S729 residue of EZH2 is highly conserved in chordate (Fig. 5A) and is located in the post-SET region (residue 726–729) (Fig. 5B). EZH2-catalyzed formation of H3K27me3 is a process critical to many types of cancer (19, 33–36). Therefore, some inhibitors have been developed to target the post-SET region of EZH2, including Y726, R727, and Y728 (37), but the significance of the S729 residue has not been addressed. To further investigate whether the O-GlcNAcylation at S729 affects the formation of di- and trimethylation, we aligned the structures of the post-SET region of the isolated EZH2 [EZH2 520–729, Protein Data Bank (PDB) ID code 4MI0] and the complex form of EZH2 (comprising EZH2, EED, VEFS domain of SUZ12, H3 peptide 22–30, and S-adenosylhomocysteine, PDB ID code 5HYN) (SI Appendix, Fig. S20). The chain from residue 726 in the complex forms a canonical post-SET structure, leading to a translocation of residues Y726 and Y728 to the lysine-accessible channel (29). S729 is also translocated to the opposite position when the isolated form transitions to the complex form (SI Appendix, Fig. S20). Next, we aligned the post-SET domain of the complex form with the inhibitor [comprising EZH2, EED, VEFS domain of SUZ12, and EZH2 inhibitor CPI-1205 (38), PDB ID code 5LS6] and without the inhibitor (PDB ID code 5HYN) (Fig. 5B), and found that Y726, R727, and Y728 were translocated (Fig. 5B). Although S729 was not represented in the complex form with inhibitor, we speculated that it would be in an alternate conformation (Fig. 5B). These observations suggest that the O-GlcNAcylation may regulate the enzymatic activity of EZH2 by altering the subconformation of the EZH2 SET domain.

Fig. 5.

O-GlcNAcylation may regulate EZH2 in a glycosite-dependent manner. (A) Partial sequence alignment of the EZH2 SET domain. (B) S729 in the post-SET region is translocated in the complex (PDB ID code 5HYN) (purple). Orange shows the complex form of EZH2 with inhibitor (PDB ID code 5LS6); cyan, H3 peptide (26–28); and M27 are highlighted by side chain. The image was captured by PyMOL Molecular Graphics System (version 2.1.0, Schrödinger, LLC). (C) O-GlcNAcylation at S73 and S84 may contribute to the protein stability of EZH2, and O-GlcNAcylation at S729 may promote the methyltransferase activity for the formation H3K27me2/3.

The ratio of O-GlcNAcylation at S729 was relatively low (0.75%) even if OGT was overexpressed, thus S729A is difficult to represent S729 without O-GlcNAcylation. Further study will be needed to understand the differential role of O-GlcNAcylated EZH2-S729, EZH2-S729, and EZH2-S729A. Although we did not find any phosphorylation signal at S729, the phosphorylation at S734 of EZH2 isoform a (the same site as S729 of EZH2 isoform c used in this study), catalyzed by ataxia-telangiectasia mutated (ATM) kinase, was reported to reduce the half-life of EZH2 (39). Thus, we cannot rule out that O-GlcNAcylation may coregulate EZH2 with phosphorylation at S729.

There are 16 genes coregulated by the OGT–EZH2 axis (16), suggesting that O-GlcNAcylation may regulate other functions in addition to gene expression, and our results show that O-GlcNAcylation influences the stability and the methyltransferase activity of EZH2 in a glycosite-dependent manner (Fig. 5C), suggesting that the O-GlcNAcylation of EZH2 may be used as a target for anticancer drug discovery.

Materials and Methods

SI Appendix, Materials and Methods provides information on cell culture, transfection, drug treatment, antibodies and reagents, plasmids, Western blotting, in vivo ubiquitylation assay, and in vitro histone methyltransferase (HMT) assay.

Galactosyltransferase-Catalyzed Incorporation of GalNAz to GlcNAc on EZH2 for Click Reaction with the Alkynyl-Biotin Reporter.

The N-acetylglucosamine (GlcNAc) moieties were detected by using the Click-iT kit (Invitrogen) according to the manufacturer’s instructions. EZH2-FLAG was purified by anti-FLAG beads and washed by TBS once and TBST three times. EZH2 on beads was incubated with galactosyltransferase (Gal-T1 Y289L) and 25 μM UDP-GalNAz in a mixture containing 20 mM Hepes (pH 7.9), 50 mM NaCl, 2% Nonidet P-40, and 7.5 mM MnCl₂ at 4 °C overnight. The reactions were terminated by adding an appropriate volume of 4× SDS loading dye containing 10% β-mercaptoethanol and subjected to SDS/PAGE. The gel was transferred onto a PVDF membrane, followed by on-membrane CuAAC reaction using the alkynyl-biotin reporter after blocking with 5% BSA in PBST for 1 h. The membrane was then incubated with HRP-streptavidin in 5% BSA/PBST for 1 h. After washing three times with PBST, the membrane was exposed with ECL (Millipore) and detected by LAS 4000 (Fujifilm).

Protein Labeling with Azido Probes.

To probe EZH2 O-GlcNAcylation, EZH2-FLAG was overexpressed in 293T cells treated with Ac₄GlcNAc or Ac₄GlcNAl overnight. EZH2-FLAG was then pulled down by anti-FLAG beads and then incubated in a mixture for Cu(I)-catalyzed azide-alkyne cycloaddition (CuAAC) reaction in the presence of 0.1 μM of Az2, 100 μM of Tris-triazole ligand, 1 mM of CuSO₄, and 2 mM of sodium ascorbate at room temperature for 1 h in the dark. The azido probes used include Az2, azido-biotin, or TAMRA (Invitrogen). Each sample was mixed with an appropriate volume of 4× SDS loading dye containing 10% β-mercaptoethanol, and gradually loaded onto 4–12% Bis-Tris gel. The gel was imaged by a Typhoon 9400 Variable Mode Imager (Amersham Biosciences) (λ_ex = 532 nm; λ_em = 555 nm) and stained with Imperial stain (Invitrogen). The result was detected by HRP-streptavidin when azido-biotin was used. The immunoprecipitation of proteins was performed with Imperial stain.

In-Gel Digest, Chemoenzymatic Tagging, and Chemical Derivatization.

The protein EZH2-FLAG obtained from 293T cells after treating with Ac₄GlcNAc overnight was resolved by SDS/PAGE and stained with the Imperial stain (Invitrogen). The protein bands of EZH2-FLAG were excised and digested based on a standard in-gel digestion protocol (40). The azido-containing UDP-N-azidoacetylgalactosamine (UDP-GalNAz) (Invitrogen) was added (2× in excess) to EZH2-FLAG and the mixture was incubated overnight with Gal-T1 (Y289L) as described. After the reaction, the excess of UDP-GalNAz was removed by passing the mixture through a C18 spin column (Thermo Fisher). Peptides were eluted in 70% acetonitrile and dried up by centrifugal evaporator. The peptides were resuspended in PBS by sonication for 10 min. The CuAAC reaction was performed in a solution of 20 μL containing 0.1 μM of alkynyl-biotin, 100 μM of azidopeptide, 1 mM of CuSO₄, and 2 mM of sodium ascorbate in PBS buffer at room temperature for 1 h. After the reaction was completed, the solution was allowed to bind to streptavidin-agarose beads (Pierce) in an IP buffer containing 1% BSA for 2 h at room temperature, followed by extensive washing. β-Elimination and Michael addition with DTT (BEMAD) directly on the bead was performed using the protocol previously described (40). The biotin-binding streptavidin beads were incubated in a 500 μL of BEMAD solution composed of 0.1% (vol/vol) NaOH, 1% (vol/vol) triethylamine, and 10 mM of DTT (made fresh) (pH adjusted to 12.0–12.5 with triethylamine) for 2.5 h at 50 °C. The reaction was stopped by addition of TFA to a final concentration of 1% (vol/vol). Peptides in the supernatant were cleaned up by C18 spin column (Thermo Fisher). The peptides containing DTT were detected by MS.

Mass Spectrometry and Data Analysis.

Samples were detected by LC-ESI-MS on an Orbitrap Fusion mass spectrometer (Thermo Fisher Scientific) equipped with the Ultimate 3000 RSLC system from Dionex (Dionex Corporation) and nanoelectrospray ion source (New Objective, Inc.). The digestion solution was injected (6 nL) at a flow rate of 10 μL/min to a self-packed precolumn (150 µm i.d. × 30 mm, 5 µm, 200 Å). Chromatographic separation was performed on a self-packed reversed phase C18 nanocolumn (75 µm i.d. × 200 mm, 2.5 µm, 100 Å) using 0.1% formic acid in water as mobile phase A and 0.1% formic acid in 80% acetonitrile as mobile phase B, operated at 300 nL/min flow rate. The full-scan MS condition was: mass range m/z 200–2,000 (AGC target 4E5) with easy ion chromatography, resolution 120,000 at m/z 200, and maximum injection time of 50 ms. The 20 most intense ions were sequentially isolated for HCD and detected (AGC target 1E4) with maximum injection time of 200 ms. The inclusion list m/z was isolated for ETD (reaction time based on charge) with maximum injection time of 250 ms. Both HCD and ETD were performed together with tandem mass (MS²) analysis to elucidate the glycosylation site and peptide sequence.