O-GlcNAcylation regulates EZH2 protein stabilityand functionChi-Shuen Chua,1, Pei-Wen Loa,b,1, Yi-Hsien Yeha, Pang-Hung Hsuc,d, Shih-Huan Penga,e, Yu-Ching Tenga,Ming-Lun Kanga, Chi-Huey Wonga,2, and Li-Jung Juana,2

aGenomics Research Center, Academia Sinica, Taipei 115, Taiwan; bInstitute of Biochemistry and Molecular Biology, National Yang-Ming University, Taipei112, Taiwan; cDepartment of Life Science and dInstitute of Bioscience and Biotechnology, National Taiwan Ocean University, Keelung 202, Taiwan;and eInstitute of Molecular Medicine, National Taiwan University, Taipei 100, Taiwan

Contributed by Chi-Huey Wong, December 16, 2013 (sent for review September 25, 2013)

O-linked N-acetylglucosamine (GlcNAc) transferase (OGT) is theonly known enzyme that catalyzes the O-GlcNAcylation of pro-teins at the Ser or Thr side chain hydroxyl group. OGT participatesin transcriptional and epigenetic regulation, and dysregulation ofOGT has been implicated in diseases such as cancer. However, theunderlying mechanism is largely unknown. Here we show thatOGT is required for the trimethylation of histone 3 at K27 to formthe product H3K27me3, a process catalyzed by the histone meth-yltransferase enhancer of zeste homolog 2 (EZH2) in the polycombrepressive complex 2 (PRC2). H3K27me3 is one of the most impor-tant histone modifications to mark the transcriptionally silencedchromatin. We found that the level of H3K27me3, but not otherH3 methylation products, was greatly reduced upon OGT deple-tion. OGT knockdown specifically down-regulated the protein sta-bility of EZH2, without altering the levels of H3K27 demethylasesUTX and JMJD3, and disrupted the integrity of the PRC2 complex.Furthermore, the interaction of OGT and EZH2/PRC2 was detectedby coimmunoprecipitation and cosedimentation experiments. Im-portantly, we identified that serine 75 is the site for EZH2 O-GlcNAcylation, and the EZH2 mutant S75A exhibited reduction instability. Finally, microarray and ChIP analysis have characterizeda specific subset of potential tumor suppressor genes subject torepression via the OGT–EZH2 axis. Together these results indicatethat OGT-mediated O-GlcNAcylation at S75 stabilizes EZH2 and hencefacilitates the formation of H3K27me3. The study not only uncoversa functional posttranslational modification of EZH2 but also revealsa unique epigenetic role of OGT in regulating histone methylation.

Protein glycosylation with β-N-acetyl-D-glucosamine (O-GlcNAcylation) is a widespread and dynamic modification inboth cytosol and nucleus (1, 2). It occurs by O-linked N-acetyl-glucosamine (GlcNAc) transferase (OGT)-catalyzed glycosylationat the hydroxyl group of serine or threonine residue of the pro-tein substrate, and removal of the O-GlcNAc group is catalyzedby the glycosidase O-GlcNAcase (OGA) (3–6). How proteinO-GlcNAcylation exerts its effect is largely unknown, but previousstudies show that it could induce a conformational change toinitiate protein folding (7), compete with phosphorylation at thesame or proximal serine or threonine (8), disrupt protein–proteininteraction (9), serve as a protein recruiting signal (10), or regulateprotein stability (11).More than thousands of proteins aremodified byO-GlcNAcylation.

These proteins are involved in a variety of biological and patho-logical processes, including epigenetic regulation, transcription,translation, signal transduction, cell division, synaptic plasticity,embryonic stem cell identity, type II diabetes, Alzheimer’s disease,and tumor malignancy (8, 12–15). O-GlcNAcylation regulatestranscription and epigenetics at least through the following twomechanisms. First, OGT adds GlcNAc to DNA-binding tran-scription factors, such as p53, c-Myc, etc. (16), and to histoneproteins such as histone H2A at T101, H2B at S36 and S112, H3at S10 and T32, and H4 at S47 (17–20). O-GlcNAcylation ofH2B at S112 facilitates the ubiquitination of K120, leading to up-regulation of transcriptional elongation (18). The functions of mosthistone O-GlcNAcylations are currently not clear, nor do we

completely understand how OGT is recruited to chromatin, al-though some progress has been made. It is known that OGT canbind to corepressor mSin3A (21) and also associate with the DNAdemethylase TET family proteins TET2 and TET3 and potentiateTET family protein-mediated gene activation (22–25). Anothercritical mediator of OGT is the polycomb repressive complex2 (PRC2).PRC2 is a conserved complex that in humans contains the en-

hancer of zeste homolog 2 (EZH2), SUZ12, EED, and RbAp46/48(26, 27). As a Su(var)3-9, enhancer of zeste and trithorax domain-containing enzyme, the major function of EZH2 is to catalyze thetransfer of methyl groups to the K27 residue of histone H3 to formH3K27me3 and to induce a signal to recruit polycomb repressivecomplex 1 (PRC1) for establishing the silenced chromatin (26–29).EZH2/PRC2 has been shown to play critical roles in diverse bi-ological processes, such as development, stem cell maintenance, andX-chromosome inactivation (30). Most importantly, EZH2 over-expression in various types of cancers has been linked to onco-genesis, partly via H3K27me3 in promoters of specific tumorsuppressor genes, and thus causes gene silencing (31). TargetingEZH2 is believed to be a promising strategy for cancer therapy(32, 33).Functional association of OGT and EZH2/PRC2 was first

reported in Drosophila. Sxc/Ogt, the Drosophila homolog ofmammalian OGT, was identified as a polycomb group (PcG)protein involving in polycomb repression during larvae de-velopment (34, 35). The chromosomal location of O-GlcNAccoincides with the PcG response elements in both Drosophila(34, 35) and Caenorhabditis elegans (36). Interestingly, mutationsin PRC2 subunits decrease the level of Ogt protein and global

Significance

The present study identifies a cross-talk of two importantposttranslational modifications, revealing that enhancer ofzeste homolog 2 (EZH2) O-GlcNAcylation (GlcNac, N-acetylglu-cosamine) at serine 75 is required for EZH2 protein stability andtherefore facilitates the histone H3 trimethylation at K27 toform H3K27me3. The finding is significant because bothO-linked GlcNAc transferase-mediated O-GlcNAcylation andEZH2-mediated H3K27me3 formation play a pivotal role indevelopment, and their up-regulation is believed to participatein tumor malignancy. The identification of O-linked GlcNActransferase association with the polycomb repressive com-plex 2 (PRC2) further provides a new approach to regulatePRC2 function.

Author contributions: C.-S.C., P.-W.L., C.-H.W., and L.-J.J. designed research; C.-S.C., P.-W.L.,Y.-H.Y., P.-H.H., S.-H.P., Y.-C.T., and M.-L.K. performed research; C.-S.C., P.-W.L., P.-H.H.,C.-H.W., and L.-J.J. analyzed data; and C.-S.C., C.-H.W., and L.-J.J. wrote the paper.

The authors declare no conflict of interest.1C.-S.C. and P.-W.L. contributed equally to this work.2To whom correspondence may be addressed. E-mail: chwong@gate.sinica.edu.tw orljjuan@gate.sinica.edu.tw.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1323226111 PNAS | January 28, 2014 | vol. 111 | no. 4 | 1355–1360

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

July

4, 2

021

http://crossmark.crossref.org/dialog/?doi=10.1073/pnas.1323226111&domain=pdf&date_stamp=2014-01-22http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplemental

O-GlcNAcylation in mouse embryonic stem cells (37). Theseobservations indicate that OGT and PRC2 may function de-pendently on each other. Nevertheless, it is obscure how OGTregulates PcG repression. Three of the PRC1 components, Ph,Pc, and Ring, are likely Sxc/Ogt substrates in Drosophila (34).However, these observations have not been confirmed by massspectrometry analysis, and no function has been reported withregard to the O-GlcNAcylation of these proteins (34). In thepresent study, using an unbiased small-scale screening, we in-dependently demonstrate that OGT depletion leads to the down-regulation of H3K27me3. We also found that OGT is able toassociate with EZH2 and the PRC2 complex and that EZH2 isO-GlcNAcylated at S75 to maintain its stability and activity.

ResultsOGT Knockdown Down-Regulates H3 Trimethylation at K27. To ex-plore whether OGT not only directly modifies histone proteinswith monosaccharides but also indirectly regulates other histonemodifications, we depleted OGT from cells and analyzed thehistone extracts by Western blot to determine whether any of thehistone modifications is altered. Here we focus on H3 methyla-tion at the sites associated with transcriptional repression, suchas methylation at K9 and K27, and the sites for transcriptionalactivation, such as methylation at K4, K36, K79, R17, and R26.Remarkably, the human breast cancer cell line MCF7 transfectedwith two different OGT siRNAs, separately or together, not onlycompletely blocked the synthesis of OGT and the global O-GlcNAcylation, but also greatly down-regulated H3K27me3 (Fig.1A). The experiments were repeated three times, and the intensityof H3K27me3 bands was quantified. As shown in Fig. S1, morethan half of H3K27me3 was lost by OGT knockdown, compared

with scramble control. In contrast, none of H3 methylation atother sites was significantly altered (Fig. 1A). Subsequently, massspectrometry was performed to further confirm the result. Histo-nes from MCF7 cells transfected with two OGT siRNAs werepurified, separated by SDS/PAGE, and analyzed by liquid chro-matography (Agilent Technologies) coupled with mass spectrom-etry (LTQ-FT; Thermo Scientific) after in-gel trypsin digestion.Consistently, the H3K27me3 level showed a more than 50% re-duction when OGT was deficient, with a concomitant increaseof H3K27me2 and H3K27me1 (Fig. 1B). This OGT-dependentH3K27me3 was unlikely cell type specific because the same re-sult was also observed in MDA-MB-231, another breast cancercell line (Fig. S2), suggesting a general control mechanism ofH3K27me3 by OGT.

OGT Knockdown Reduces EZH2 Protein Stability. The significant lossof H3K27me3 by OGT knockdown prompted us to investigatewhether OGT depletion alters the expression level of H3K27methyltransferase EZH2 or demethylases UTX and JMJD3 (38).As shown in Fig. 2A, OGT deprivation by siRNA had no sig-nificant effect on the protein levels of UTX and JMJD3; how-ever, the EZH2 protein level was greatly reduced (Fig. 2A).Importantly, the OGT knockdown-mediated down-regulation ofEZH2 and H3K27me3 could be rescued by ectopic expression ofsiRNA-resistant OGT, indicating that it was not an off-targeteffect (Fig. 2B). Moreover, we found that OGT depletion notonly reduced the protein level of EZH2 but also diminished theexpression of all other PRC2 subunits (Fig. 2C), consistent withthe previous report that disruption of an essential subunit maylead to destabilization of the whole PRC2 complex (39–41). Itshould be noted that OGT knockdown-mediated down-regula-tion of PRC2 expression was at the protein level because OGTdeprivation did not alter the mRNA levels of all PRC2 subunits(Fig. S3). These results suggested that OGT knockdown mayalter the stability of EZH2/PRC2 complex. To test this hypoth-esis, MCF7 cells with or without OGT knockdown were treatedwith cycloheximide to inhibit protein synthesis, followed by thechase of the remaining EZH2. As shown in Fig. 2D, 10 h aftercycloheximide treatment, the level of EZH2 only dropped to93% in cells with scramble control (compare lane 6 with lane 1).However, in cells with OGT siRNA, EZH2 was reduced to 54% ofits original amount (compare lane 12 with lane 7). Together theseexperiments demonstrate that OGT is required for the maintenanceof EZH2/PRC2 stability and the subsequent H3K27me3 level.

OGT Interacts with EZH2 in the PRC2 Complex. Because OGT isessential to EZH2/PRC2 function, we investigated whether OGTphysically interacts with EZH2/PRC2. To test this, 293T cellsexpressing EZH2-FLAG or OGT-V5 or both were subject toimmunoprecipitation with FLAG Ab (Fig. 3A, Left) or V5 Ab(Fig. 3A, Right), followed by Western blot with the Abs indicated.Indeed, EZH2-FLAG was coprecipitated with OGT-V5 in thereciprocal experiments (Fig. 3A). Consistently, the endogenousOGT was also co-pulled down by EZH2 Ab (Fig. 3B). It is notedthat the signal appearing in the IgG immunoprecipitation (IP)control (Fig. 3B, lane 2) is not at the same position of EZH2. Webelieve that it is a nonspecific band. Glycerol-sizing gradientsedimentation was further applied to understand whether OGTnot only interacts with EZH2 but also associates with the PRC2complex. As shown in Fig. 3C, OGT coeluted with EZH2 and allother PRC2 subunits in fraction nos. 6–9. These results provideevidence that OGT physically associates with EZH2/PRC2 in cells.

EZH2 Is O-GlcNAcylated in Vivo, and the O-GlcNAcylation Site Mutantof EZH2 Shows Reduced Protein Stability. Our data so far indicatethat OGT is required for sustaining EZH2 protein stability andH3K27me3 formation and that OGT physically interacts with EZH2.Next we sought to analyze whether OGT directly modifies EZH2with O-GlcNAc, and if yes, whether EZH2 O-GlcNAcylationaffects the protein stability of EZH2. To this end, first we treatedMCF7 cells with the OGA inhibitor O-(2-acetamido-2-deoxy-d-

Fig. 1. OGT knockdown reduces H3 trimethylation at K27. (A) OGT de-pletion decreases the level of H3K27me3 but not other H3 methylationproducts. Total cell lysates or histones purified from MCF7 cells mock trans-fected (NT), transfected with scramble RNA (Scr), or two different siOGT (#1and #2) separately or together were subjected to Western blot using in-dicated Abs. (B) OGT knockdown decreases global H3K27me3 measured bymass spectrometry. Histones purified from MCF7 cells transfected with in-dicated siRNAs were subjected to SDS/PAGE, followed by LC/MS analysis.

1356 | www.pnas.org/cgi/doi/10.1073/pnas.1323226111 Chu et al.

Dow

nloa

ded

by g

uest

on

July

4, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplemental/pnas.201323226SI.pdf?targetid=nameddest=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplemental/pnas.201323226SI.pdf?targetid=nameddest=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplemental/pnas.201323226SI.pdf?targetid=nameddest=SF3

glucopyranosylidene)amino-N-phenylcarbamate (PUGNAc) andfound that the treatment not only led to increased global O-GlcNAcylation but also enriched H3K27me3 and EZH2 (Fig.4A), suggesting that the up-regulated EZH2 level might bea result of EZH2 O-GlcNAcylation. Consistent with this notion,in the presence of PUGNAc the endogenous EZH2 pulled downby EZH2 Ab could be recognized by O-GlcNAc Ab (Fig. 4B).This result indicates that EZH2 was most likely modified withthis sugar in vivo. However, the experiment cannot exclude thepossibility that the O-GlcNAcylation signal was from an EZH2-associated proteins. To rule out this possibility, we applied thefollowing cell labeling experiment using O-GlcNAz. It is knownthat proteins modified with O-GlcNAz, an O-GlcNAc derivativewith azido-acetyl group, can be conjugated with phosphine-FLAG,which then can be detected by FLAG Ab (42) (Fig. S4). As shownin Fig. 4C, in MCF7 cells metabolically labeled with O-GlcNAz,the EZH2 precipitated with EZH2 Ab in a denatured condition(34) could be recognized by FLAG Ab (arrowhead), supportingthe same conclusion that EZH2 is O-GlcNAcylated in cells. Itshould be noted that the denatured condition used in the exper-iment above in theory had disrupted protein–protein interaction.Therefore, the signal by FLAG Ab was most likely from EZH2conjugated with O-GlcNAz-FLAG but not from any EZH2-associated protein.Finally, mass spectrometry was applied to confirm EZH2

O-GlcNAcylation and to identify the modification site(s). TheFLAG-tagged EZH2 was precipitated from 293T cells over-expressing both OGT-V5 and EZH2-FLAG, followed by liquid

chromatography (Agilent Technologies) coupled with mass spec-trometry (LTQ-FT) after in-gel trypsin digestion. We found thatO-GlcNAcylation occurred on S75 of EZH2 (Fig. 4D). To un-derstand the function of EZH2 O-GlcNAcylation at S75, a mutantEZH2 with S75 substituted with alanine (S75A) was constructedand expressed in 293T cells, followed by a cycloheximide-chaseexperiment. The result showed that the S75A mutant of EZH2was more labile compared with WT EZH2 (Fig. 4E). These ex-periments strongly suggest that EZH2 O-GlcNAcylation at S75 isessential for EZH2 protein stability. Given that OGT is the soleenzyme to carry out O-GlcNAcylation in cells, it is reasonable topropose that the loss of H3K27me3 in cells depleted of OGT likelyresults from the absence of EZH2 O-GlcNAcylation at S75 andreduced stability of EZH2.

OGT-EZH2 Axis Suppresses Specific Tumor Suppressor Gene Expression.To gain insight into the role of OGT in EZH2-mediated func-tion, microarray analysis was applied to identify genes controlledby both OGT and EZH2. Because overexpression of OGT andEZH2 has been implicated in breast tumorigenesis (43, 44) andEZH2 has been known to repress specific tumor suppressor geneexpression in breast cancer cells (45), microarray analysis wascarried out using the breast cancer cell line MCF7 treated withscramble siRNA, siOGT, or siEZH2 (Fig. S5A). It was found thata total of 63 genes were coregulated by OGT and EZH2 (Fig.S5B). Among them, we focused on 20 genes whose expressionswere increased in both OGT-KD and EZH2-KD cells (1.5-foldcutoff, P < 0.01) (Fig. S5B). The expressions of this set of geneswere further analyzed by real-time RT-PCR, and 16 genes amongthe set were found significantly up-regulated in both OGT-KD(Fig. 5A) and EZH2-KD (Fig. 5B) cells. Potentially this group ofgenes might exert tumor suppressor function under the control ofthe OGT-EZH2 axis. Indeed, IL1R1, SCUBE2, UNC5A, andSPATA17 have been shown negatively correlated with oncogenesis(46–49). In addition, CSTA can inhibit metastasis in breast cancercells and is negatively correlated with metastasis (50). Importantly,ChIP assays indicate that the promoter regions of these 16 geneswere bound by OGT (Fig. 5C) and enriched in EZH2 andH3K27me3 in an OGT-dependent manner, because depletion of

Fig. 2. OGT knockdown reduces EZH2 protein stability. (A) OGT knockdowndecreases the protein level of EZH2 but not UTX or JMJD3. (B) Reduction ofEZH2 by OGT knockdown can be rescued by adding back resistant OGT. (C)OGT knockdown decreases the protein levels of PRC2 components. Totallysates or histone extracts from MCF7 cells transfected with scramble siRNA(Scr) or siOGT were subjected to Western blot using indicated Abs. For B, cellswere treated with scramble (Scr) siRNA or siOGT for 24 h, followed bytransfection of vector alone (lanes 1 and 2) or the plasmid encoding thesiRNA-resistant OGT (lane 3) for 2 d. (D) OGT knockdown destabilizes EZH2.MCF7 cells were transfected with scramble siRNA (Scr) or siOGT for 3 d andsubsequently treated with cycloheximide at the final concentration of 50 μg/mLand harvested at the indicated time points for Western blot. Band intensitieswere measured by ImageJ. Normalization was done by dividing the EZH2signal to α-tubulin signal. P values were measured by Student’s t test. Theresults are presented as mean ± SD. *P = 0.031, n = 3.

Fig. 3. OGT stably interacts with EZH2 in the PRC2 complex. (A) Exoge-nous OGT associates with exogenous EZH2. Total lysates from 293T cellsexpressing OGT-V5 and/or EZH2-FLAG were subjected to IP with FLAG Ab(Left) or V5 Ab (Right), followed by Western blot using indicated Abs. (B)Endogenous OGT interacts with endogenous EZH2. Nuclear extracts fromMCF7 cells were subjected to IP, followed by Western blot using indicatedAbs. Asterisk indicates a nonspecific band. (C ) OGT cofractions with PRC2complex. Nuclear extracts from MCF7 cells were subjected to 10–50%glycerol sizing gradient sedimentation. Fractions were collected, pre-cipitated by trichloroacetic acid, and analyzed by Western blot usingindicated Abs.

Chu et al. PNAS | January 28, 2014 | vol. 111 | no. 4 | 1357

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

July

4, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplemental/pnas.201323226SI.pdf?targetid=nameddest=SF4http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplemental/pnas.201323226SI.pdf?targetid=nameddest=SF5http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplemental/pnas.201323226SI.pdf?targetid=nameddest=SF5http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplemental/pnas.201323226SI.pdf?targetid=nameddest=SF5http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplemental/pnas.201323226SI.pdf?targetid=nameddest=SF5

OGT abrogated EZH2 recruitment (Fig. 5D and Fig. S6) andH3K27me3 occupancy (Fig. 5E and Fig. S7). In summary, weidentified a set of tumor suppressor genes whose expressions inbreast cancer cell are down-regulated by the OGT-EZH2 axis.

DiscussionHere we provide a unique regulatory mechanism by which OGTfacilitates H3K27me3 synthesis. Our results indicate that OGTmodifies the H3K27 methyltransferase EZH2 with O-GlcNAc atS75 and by this O-GlcNAcylation maintains EZH2 protein sta-bility and function. Because O-GlcNAc is known to regulate theproteasome (51), we thus cannot exclude the possibility that thedecrease in EZH2 levels upon OGT knockdown was partlythrough proteasomal dysregulation. However, because (i) OGT isthe only enzyme to carry out protein O-GlcNAcylation, (ii), EZH2was O-GlcNAcylated at S75 in cells (Fig. 4), and (iii) the S75AEZH2 mutant showed decreased protein stability (Fig. 4E), webelieve that loss of protein stability does contribute to the decreaseof EZH2 protein level in OGT-KD cells.Although endogenous WT EZH2 in MCF7 cells treated with

siRNA or siOGT looks stable over the experiment time frame(still approximately 100% at 8 h) (Fig. 2D), less than 75% ofFLAG-tagged EZH2 content remains at 8 h when analyzed in293T cells with no treatment of any siRNA (Fig. 4E). With thesedifferences, it is not appropriate to directly compare the turnoverrate of the endogenous EZH2 in Fig. 2D with that of overex-pressed FLAG-EZH2 in Fig. 4E. Most importantly, the results inboth figures are consistent in that OGT knockdown decreasesEZH2 protein stability, and EZH2 O-GlcNAcylation mutantshows decreased protein stability.How S75 O-GlcNAcylation regulates EZH2 protein stability is

currently under investigation. As mentioned in the Introduction,O-GlcNAcylation is known to compete with phosphorylation(52). It is likely that S75 O-GlcNAcylation prevents phosphory-lation at the same site that serves as a signal for EZH2 degra-dation. Alternatively, it is likely that O-GlcNAcylation at S75disrupts EZH2 modification(s) at other sites required for deg-radation. A previous study indicates that EZH2 phosphorylatedat T345 by cyclin-dependent kinase 1 (CDK1) is unstable duringmitosis and prone to be degraded via proteasome-mediatedpathway (53). Whether the monosaccharide at S75 disturbs EZH2phosphorylation at T345 is currently not known. Interestingly, theCDK1 activity can be inhibited by OGT overexpression (54).Together, these studies imply a potential regulatory networkamong OGT, CDK1, and EZH2.Unlike other hit-and-run enzyme–substrate interactions, OGT

is known to bind to some of its substrates tightly (55, 56).Therefore, it is not surprising to find that the binding of OGT toEZH2 is stable enough to be detected by co-IP and cosedi-mentation (Fig. 3). This and other results of the present studysupport previous observations that OGT and EZH2/PRC2 arefunctionally associated with each other (34, 35, 37). Neverthe-less, discrepancy still exists. In contrast to our study, Myers et al.(37) did not observe reduction in EZH2 or H3K27me3 in Ogt-knocked-down mouse ES cells. Gambetta et al. (34) failed todetect O-GlcNAcylation of E(z), the Drosophila homolog ofEZH2, nor did they observe down-regulation of global H3K27me3in Ogt mutants, although their data showed reduction of H3K27me3in specific genes (34). Currently we do not know what causesthese differences. One possibility is that different species and celltypes were analyzed. We used human breast and kidney epi-thelium cells, whereas Drosophila and mouse were analyzed inprevious studies.OGT knockdown only reduced the H3K27me3 to half of its

original level (Fig. 1). Therefore, it is unlikely that OGT knock-down has an overwhelming effect on global gene expression.

Fig. 4. S75 O-GlcNAcylation of EZH2 is required for EZH2 stability. (A) Ac-cumulation of global O-GlcNAcylation level increases EZH2 and globalH3K27me3. Total lysates or histones from MCF7 treated with DMSO orPUGNAc for 16 h were subjected to Western blot using indicated Abs. (B)Endogenous EZH2 is O-GlcNAcylated. Nuclear extracts from MCF7 cells withor without PUGNAc were subjected to Western blot with O-GlcNAc Ab or IPwith IgG or EZH2, followed by Western blot using O-GlcNAc or EZH2 Ab. (C)Detection of EZH2-O-GlcNAz-FLAG. Nuclear extracts from MCF7 cells treatedwith O-GlcNAz for 16 h were incubated with phosphine-FLAG overnight at4 °C. Subsequently lysates were subjected to IP in denatured condition, fol-lowed by Western blot using indicated Abs. Arrowheads indicate the band ofEZH2. Asterisk indicates a nonspecific band. (D) EZH2 is O-GlcNAcylated atserine 75. An MS/MS spectrum was generated from microliquid chroma-tography/tandem MS. (E) Serine 75 mutation reduces EZH2 protein stability.293T cells expressing EZH2-WT-FLAG or EZH2-S75A-FLAG were treated withcycloheximide at the final concentration of 50 μg/mL and harvested at in-dicated time points for Western blot using indicated Abs. Band intensities

were measured by ImageJ. Normalization was done by dividing the FLAGsignal to β-tubulin signal. P values were measured by Student’s t test. Theresults in E are presented as mean ± SD. *P < 0.05, n = 3.

1358 | www.pnas.org/cgi/doi/10.1073/pnas.1323226111 Chu et al.

Dow

nloa

ded

by g

uest

on

July

4, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplemental/pnas.201323226SI.pdf?targetid=nameddest=SF6http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplemental/pnas.201323226SI.pdf?targetid=nameddest=SF7

Indeed, OGT and EZH2 only coregulated 63 genes (Fig. S5B).In addition, similar to DNA methylation and other histone

posttranslational modifications, not all changes in H3K27me3result in an alteration of DNA transcription level. For example,Hosogane et al. (57) show that, in NIH 3T3 cells, H3K27me3change caused by depleting SUZ12, Ras, or Raf does not cor-relate with changes in target gene expressions. It is unlikely ei-ther that the small number gene effect is because OGT onlyaffects H3K27me3 level at the repeated sequences. Vella et al.(23) has shown that, in mouse ES cells, 82% of OGT ChIP sig-nals are within promoter-transcription start site and gene body,and only 16% of signals are present in intergenic regions. An-other ChIP-seq analysis indicates that, in human 293T cells, OGTbinds to promoter region but not gene body or intergenic region(24). Thus, the majority of OGT does not associate with repeatedsequences, excluding the possibility that OGT regulates H3K27me3at these sequences.Although OGT associated with EZH2 in the PRC2 complex,

these two proteins also exist in free form independently of eachother (Fig. 3). This is consistent with the microarray analysis inFig. S5, which indicates that OGT and EZH2 only coregulatea subset of genes. Among the genes most significantly controlledby the OGT–EZH2 axis, two (IL1R1 and UNC5A) have beenshown to be negatively associated with tumor malignancy. One isIL1R1 with low expression in colon cancer cell line HCT116(47), and the other is UNC5A as a downstream target of p53 toinduce apoptosis (46). Together these results uncovered a pre-viously unknown OGT–EZH2 regulatory axis that may play acritical role in tumor malignancy.

Materials and MethodsSI Materials and Methods provides information on cell culture, knockdownby siRNA, antibodies and reagents, plasmids, quantitative real-time RT-PCR(primer sets are listed in Table S1), Western blotting and histone extraction,IP and co-IP, and microarray.

Metabolic Labeling and Denatured IP. Metabolic labeling was performed aspreviously described (42). Briefly, MCF7 cells were treated with 40 μMO-GlcNAz overnight at 37 °C with 5% (vol/vol) CO2. Nuclei were prepared andincubated with equal volumes of 500 μM phosphoine-FLAG overnight at 4 °Cwith rotation. Denatured IP was performed as previously described (34)with modifications. In brief, the nuclear lysates were adjusted to denaturedcondition by the addition of 10% (wt/vol) SDS (final concentration 0.5%)and 1 M DTT (final concentration 5 mM). The lysates were boiled for5 min, and SDS was diluted to 0.1% and DTT to 1 mM, followed by proce-dures described in SI Materials and Methods, Immunoprecipitation andCoimmunoprecipitation.

Glycerol Gradient Sedimentation Analysis. The glycerol gradient sedimenta-tion analysis was performed as previously described (58), with modifications.Briefly, nuclear extracts from MCF7 cells were prepared in BC100 buffer[5 mMHepes/KOH (pH 7.3), 100 mMNaCl, 1 mMMgCl2, 0.5 mM EGTA, 0.1 mMEDTA, 10% (vol/vol) glycerol, 1 mMDTT, and 0.2 mM PMSF]. Glycerol gradients[10–50% (wt/vol)] were prepared in BC100 buffer using a gradient maker(Hoefer). A 100-μL sample containing 1 mg of MCF7 nuclear extract or 30 mgof individual standards (Sigma, MW-GF-1000) was loaded on top of the gra-dient. Gradients were kept at 4 °C for 30 min, then spun in a SW41 rotor(Beckman) at 41,000 rpm for 28 h at 4 °C. Fractions (0.5 mL) were collected andanalyzed by 8% SDS/PAGE, followed by Western blot.

Identification of O-GlcNAcylation Sites on EZH2 by Mass Spectrometry. EZH2-FLAG and OGT-V5 were coexpressed in 293T cells for 2 d. EZH2-FLAG wereimmunoprecipitated by anti-FLAG antibody and subjected to SDS/PAGE. Afterprotein gel electrophoresis, immunoprecipitated EZH2 were excised from gelsand subjected to in-gel trypsin digestion. The digested peptides were extractedand analyzed by microliquid chromatography/tandem MS.

ChIP. MCF7 transfected with scramble RNA, siOGT, or siEZH2 for 3 d weresubjected to ChIP as previously described (59). The precipitated chromatinwas washed, and DNAs were purified for measurement by quantitative PCRusing LightCycler 480 SYBR Green I Master (04 887 352 001, Roche) withprimers against specific promoter region. The primers used for ChIP real-time PCR are listed in Table S2.

Fig. 5. The OGT-EZH2 axis suppresses specific tumor suppressor gene ex-pression. mRNA levels of 16 genes corepressed by OGT and EZH2 wereconfirmed with RT-PCR in MCF7 depleted of OGT (A) or EZH2 (B). (C) OGToccupancy at the promoter regions of the 16 genes in A and B. MCF7 cellswere subjected to ChIP with control IgG (white bar) or OGT Ab (black bar). (Dand E) EZH2 occupancy (D) and H3K27me3 (E) associated with the promoterregions of the 16 genes in A and B depend on OGT. MCF7 cells with scramblesiRNA (white bar) or siOGT (black bar) were subjected to ChIP with controlIgG, EZH2, or H3K27me3 Ab. In D and E the data are shown as fold en-richment relative to IgG. GAPDH gene was used as a negative control. Allresults from A to E are presented as mean ± SD. P values were measured byStudent’s t test. *P < 0.05, **P < 0.01, ***P < 0.001, n = 3.

Chu et al. PNAS | January 28, 2014 | vol. 111 | no. 4 | 1359

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

July

4, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplemental/pnas.201323226SI.pdf?targetid=nameddest=SF5http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplemental/pnas.201323226SI.pdf?targetid=nameddest=SF5http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplemental/pnas.201323226SI.pdf?targetid=nameddest=STXThttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplemental/pnas.201323226SI.pdf?targetid=nameddest=ST1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplemental/pnas.201323226SI.pdf?targetid=nameddest=STXThttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplemental/pnas.201323226SI.pdf?targetid=nameddest=STXThttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1323226111/-/DCSupplemental/pnas.201323226SI.pdf?targetid=nameddest=ST2

ACKNOWLEDGMENTS. We thank Drs. W. H. Lee at University of Califor-nia, Irvine, and Y. Zhang at Harvard Medical School for critical sugges-tions, and Affymetrix Gene Expression Service Laboratory at Academia

Sinica for performing the microarray experiments. This study was sup-ported by grants from Academia Sinica [to L.-J.J. (career developmentgrant) and C.-H.W.].

1. Torres CR, Hart GW (1984) Topography and polypeptide distribution of terminalN-acetylglucosamine residues on the surfaces of intact lymphocytes. Evidence forO-linked GlcNAc. J Biol Chem 259(5):3308–3317.

2. Holt GD, Hart GW (1986) The subcellular distribution of terminal N-acetylglucosaminemoieties. Localization of a novel protein-saccharide linkage, O-linked GlcNAc. J BiolChem 261(17):8049–8057.

3. Haltiwanger RS, Holt GD, Hart GW (1990) Enzymatic addition of O-GlcNAc to nuclearand cytoplasmic proteins. Identification of a uridine diphospho-N-acetylglucosamine:peptide beta-N-acetylglucosaminyltransferase. J Biol Chem 265(5):2563–2568.

4. Kreppel LK, Blomberg MA, Hart GW (1997) Dynamic glycosylation of nuclear andcytosolic proteins. Cloning and characterization of a unique O-GlcNAc transferasewith multiple tetratricopeptide repeats. J Biol Chem 272(14):9308–9315.

5. Dong DL, Hart GW (1994) Purification and characterization of an O-GlcNAc selectiveN-acetyl-beta-D-glucosaminidase from rat spleen cytosol. J Biol Chem 269(30):19321–19330.

6. Vocadlo DJ (2012) O-GlcNAc processing enzymes: catalytic mechanisms, substratespecificity, and enzyme regulation. Curr Opin Chem Biol 16(5-6):488–497.

7. Simanek EE, et al. (1998) Glycosylation of threonine of the repeating unit of RNApolymerase II with beta-linked N-acetylglucosame leads to a turnlike structure. J AmChem Soc 120(45):11567–11575.

8. Hart GW, Slawson C, Ramirez-Correa G, Lagerlof O (2011) Cross talk betweenO-GlcNAcylation and phosphorylation: Roles in signaling, transcription, and chronicdisease. Annu Rev Biochem 80:825–858.

9. Yang X, et al. (2001) O-linkage of N-acetylglucosamine to Sp1 activation domain in-hibits its transcriptional capability. Proc Natl Acad Sci USA 98(12):6611–6616.

10. Gewinner C, et al. (2004) The coactivator of transcription CREB-binding protein inter-acts preferentially with the glycosylated form of Stat5. J Biol Chem 279(5):3563–3572.

11. Yang WH, et al. (2006) Modification of p53 with O-linked N-acetylglucosamine reg-ulates p53 activity and stability. Nat Cell Biol 8(10):1074–1083.

12. Jang H, et al. (2012) O-GlcNAc regulates pluripotency and reprogramming by directlyacting on core components of the pluripotency network. Cell Stem Cell 11(1):62–74.

13. Bond MR, Hanover JA (2013) O-GlcNAc cycling: A link between metabolism andchronic disease. Annu Rev Nutr 33:205–229.

14. Hanover JA, Krause MW, Love DC (2012) Bittersweet memories: Linking metabolismto epigenetics through O-GlcNAcylation. Nat Rev Mol Cell Biol 13(5):312–321.

15. Slawson C, Hart GW (2011) O-GlcNAc signalling: Implications for cancer cell biology.Nat Rev Cancer 11(9):678–684.

16. Ozcan S, Andrali SS, Cantrell JE (2010) Modulation of transcription factor function byO-GlcNAc modification. Biochim Biophys Acta 1799(5-6):353–364.

17. Sakabe K, Wang Z, Hart GW (2010) Beta-N-acetylglucosamine (O-GlcNAc) is part ofthe histone code. Proc Natl Acad Sci USA 107(46):19915–19920.

18. Fujiki R, et al. (2011) GlcNAcylation of histone H2B facilitates its monoubiquitination.Nature 480(7378):557–560.

19. Zhang S, Roche K, Nasheuer HP, Lowndes NF (2011) Modification of histones by sugarβ-N-acetylglucosamine (GlcNAc) occurs on multiple residues, including histone H3serine 10, and is cell cycle-regulated. J Biol Chem 286(43):37483–37495.

20. Fong JJ, et al. (2012) β-N-Acetylglucosamine (O-GlcNAc) is a novel regulator of mitosis-specific phosphorylations on histone H3. J Biol Chem 287(15):12195–12203.

21. Yang X, Zhang F, Kudlow JE (2002) Recruitment of O-GlcNAc transferase to promotersby corepressor mSin3A: Coupling protein O-GlcNAcylation to transcriptional re-pression. Cell 110(1):69–80.

22. Chen Q, Chen Y, Bian C, Fujiki R, Yu X (2013) TET2 promotes histone O-GlcNAcylationduring gene transcription. Nature 493(7433):561–564.

23. Vella P, et al. (2013) Tet proteins connect the O-linked N-acetylglucosamine trans-ferase Ogt to chromatin in embryonic stem cells. Mol Cell 49(4):645–656.

24. Deplus R, et al. (2013) TET2 and TET3 regulate GlcNAcylation and H3K4 methylationthrough OGT and SET1/COMPASS. EMBO J 32(5):645–655.

25. Shi FT, et al. (2013) Ten-eleven translocation 1 (Tet1) is regulated by O-linked N-acetylglucosamine transferase (Ogt) for target gene repression in mouse embryonicstem cells. J Biol Chem 288(29):20776–20784.

26. Cao R, et al. (2002) Role of histone H3 lysine 27 methylation in Polycomb-group si-lencing. Science 298(5595):1039–1043.

27. Kuzmichev A, Nishioka K, Erdjument-Bromage H, Tempst P, Reinberg D (2002) His-tone methyltransferase activity associated with a human multiprotein complex con-taining the Enhancer of Zeste protein. Genes Dev 16(22):2893–2905.

28. Czermin B, et al. (2002) Drosophila enhancer of Zeste/ESC complexes have a histoneH3 methyltransferase activity that marks chromosomal Polycomb sites. Cell 111(2):185–196.

29. Müller J, et al. (2002) Histone methyltransferase activity of a Drosophila Polycombgroup repressor complex. Cell 111(2):197–208.

30. Cao R, Zhang Y (2004) The functions of E(Z)/EZH2-mediated methylation of lysine 27in histone H3. Curr Opin Genet Dev 14(2):155–164.

31. Chang CJ, HungMC (2012) The role of EZH2 in tumour progression. Br J Cancer 106(2):243–247.

32. Knutson SK, et al. (2013) Durable tumor regression in genetically altered malignantrhabdoid tumors by inhibition of methyltransferase EZH2. Proc Natl Acad Sci USA110(19):7922–7927.

33. Knutson SK, et al. (2012) A selective inhibitor of EZH2 blocks H3K27 methylation andkills mutant lymphoma cells. Nat Chem Biol 8(11):890–896.

34. Gambetta MC, Oktaba K, Müller J (2009) Essential role of the glycosyltransferase sxc/Ogt in polycomb repression. Science 325(5936):93–96.

35. Sinclair DA, et al. (2009) Drosophila O-GlcNAc transferase (OGT) is encoded by thePolycomb group (PcG) gene, super sex combs (sxc). Proc Natl Acad Sci USA 106(32):13427–13432.

36. Love DC, et al. (2010) Dynamic O-GlcNAc cycling at promoters of Caenorhabditis el-egans genes regulating longevity, stress, and immunity. Proc Natl Acad Sci USA107(16):7413–7418.

37. Myers SA, Panning B, Burlingame AL (2011) Polycomb repressive complex 2 is nec-essary for the normal site-specific O-GlcNAc distribution in mouse embryonic stemcells. Proc Natl Acad Sci USA 108(23):9490–9495.

38. Hübner MR, Spector DL (2010) Role of H3K27 demethylases Jmjd3 and UTX in tran-scriptional regulation. Cold Spring Harb Symp Quant Biol 75:43–49.

39. Bracken AP, Dietrich N, Pasini D, Hansen KH, Helin K (2006) Genome-wide mapping ofPolycomb target genes unravels their roles in cell fate transitions. Genes Dev 20(9):1123–1136.

40. Bracken AP, et al. (2003) EZH2 is downstream of the pRB-E2F pathway, essential forproliferation and amplified in cancer. EMBO J 22(20):5323–5335.

41. Montgomery ND, et al. (2005) The murine polycomb group protein Eed is required forglobal histone H3 lysine-27 methylation. Curr Biol 15(10):942–947.

42. Laughlin ST, Bertozzi CR (2007) Metabolic labeling of glycans with azido sugars andsubsequent glycan-profiling and visualization via Staudinger ligation. Nat Protoc2(11):2930–2944.

43. Fardini Y, Dehennaut V, Lefebvre T, Issad T (2013) O-GlcNAcylation: A new cancerhallmark? Front Endocrinol (Lausanne) 4:99.

44. Yoo KH, Hennighausen L (2012) EZH2 methyltransferase and H3K27 methylation inbreast cancer. Int J Biol Sci 8(1):59–65.

45. Du J, et al. (2012) FOXC1, a target of polycomb, inhibits metastasis of breast cancercells. Breast Cancer Res Treat 131(1):65–73.

46. Miyamoto Y, et al. (2010) Identification of UNC5A as a novel transcriptional target oftumor suppressor p53 and a regulator of apoptosis. Int J Oncol 36(5):1253–1260.

47. Mustafi R, et al. (2012) Both stromal cell and colonocyte epidermal growth factorreceptors control HCT116 colon cancer cell growth in tumor xenografts. Carcino-genesis 33(10):1930–1939.

48. Cheng CJ, et al. (2009) SCUBE2 suppresses breast tumor cell proliferation and confersa favorable prognosis in invasive breast cancer. Cancer Res 69(8):3634–3641.

49. Nie DS, Liu Y, Juan H, Yang X (2013) Overexpression of human SPATA17 protein in-duces germ cell apoptosis in transgenic male mice. Mol Biol Rep 40(2):1905–1910.

50. Lee S, et al. (2012) Differentially expressed genes regulating the progression of ductalcarcinoma in situ to invasive breast cancer. Cancer Res 72(17):4574–4586.

51. Zhang F, et al. (2003) O-GlcNAc modification is an endogenous inhibitor of the pro-teasome. Cell 115(6):715–725.

52. Kamemura K, Hart GW (2003) Dynamic interplay between O-glycosylation andO-phosphorylation of nucleocytoplasmic proteins: a new paradigm for metabolic control ofsignal transduction and transcription. Prog Nucleic Acid Res Mol Biol 73:107–136.

53. Wu SC, Zhang Y (2011) Cyclin-dependent kinase 1 (CDK1)-mediated phosphorylationof enhancer of zeste 2 (Ezh2) regulates its stability. J Biol Chem 286(32):28511–28519.

54. Wang Z, et al. (2010) Extensive crosstalk between O-GlcNAcylation and phosphory-lation regulates cytokinesis. Sci Signal 3(104):ra2.

55. Iyer SP, Hart GW (2003) Roles of the tetratricopeptide repeat domain in O-GlcNActransferase targeting and protein substrate specificity. J Biol Chem 278(27):24608–24616.

56. Iyer SP, Akimoto Y, Hart GW (2003) Identification and cloning of a novel family ofcoiled-coil domain proteins that interact with O-GlcNAc transferase. J Biol Chem278(7):5399–5409.

57. Hosogane M, Funayama R, Nishida Y, Nagashima T, Nakayama K (2013) Ras-inducedchanges in H3K27me3 occur after those in transcriptional activity. PLoS Genet 9(8):e1003698.

58. Klymenko T, et al. (2006) A Polycomb group protein complex with sequence-specificDNA-binding and selective methyl-lysine-binding activities. Genes Dev 20(9):1110–1122.

59. Shi-Chen Ou D, et al. (2011) Transcriptional activation of endoplasmic reticulumchaperone GRP78 by HCMV IE1-72 protein. Cell Res 21(4):642–653.

1360 | www.pnas.org/cgi/doi/10.1073/pnas.1323226111 Chu et al.

Dow

nloa

ded

by g

uest

on

July

4, 2

021