Binding of different histone marks differentiallyregulates the activity and specificity of polycombrepressive complex 2 (PRC2)Chao Xua,1, Chuanbing Biana,1, Wei Yangb,1, Marek Galkac, Hui Ouyanga, Chen Chend, Wei Qiua, Huadong Liuc,Amanda E. Jonesb, Farrell MacKenziea, Patricia Pana,e, Shawn Shun-Cheng Lic,2, Hengbin Wangb,2, and Jinrong Mina,f,2

aStructural Genomics Consortium, University of Toronto, 101 College Street, Toronto, ON, Canada M5G 1L7; bDepartment of Biochemistry and MolecularGenetics, University of Alabama at Birmingham, Kaul Human Genetics Building Room 430, 720 South 20th Street South, Birmingham, AL 35294;cDepartment of Biochemistry, Schulich School of Medicine and Dentistry, University of Western Ontario, London, ON, Canada N6A 5C1; dSamuelLunenfeld Research Institute, Mount Sinai Hospital, Toronto, ON, Canada M5G 1X5; eDepartment of Medical Biophysics, University of Toronto,Toronto, ON, Canada M5G 2M9; and fDepartment of Physiology, University of Toronto, Toronto, ON, Canada M5S 1A8

Edited by Tony Pawson, Samuel Lunenfeld Research Institute, Toronto, Canada, and approved September 14, 2010 (received for review June 23, 2010)

Thepolycomb repressive complex2 (PRC2) is themajormethyltrans-ferase for H3K27 methylation, a modification critical for maintain-ing repressed gene expression programs throughout development.It has been previously shown that PRC2 maintains histone methyl-ation patterns during DNA replication in part through its ability tobind to H3K27me3. However, the mechanism by which PRC2 recog-nizes H3K27me3 is unclear. Herewe show that theWD40 domain ofEED, a PRC2 component, is a methyllysine histone-binding domain.The crystal structures of apo-EED and EED in complex respectivelywith five different trimethyllysine histone peptides reveal that EEDbinds these peptides via the top face of its β-propeller architecture.The ammonium group of the trimethyllysine is accommodated byan aromatic cage formed by three aromatic residues, while its ali-phatic chain is flanked by a fourth aromatic residue. Our structuraldata provide an explanation for the preferential recognition of theAla-Arg-Lys-Ser motif-containing trimethylated H3K27, H3K9, andH1K26 marks by EED over lower methylation states and other his-tone methyllysine marks. More importantly, we found that bindingof different histone marks by EED differentially regulates the activ-ity and specificity of PRC2.Whereas the H3K27me3mark stimulatesthe histone methyltransferase activity of PRC2, the H1K26me3mark inhibits PRC2 methyltransferase activity on the nucleosome.Moreover, H1K26me3 binding switches the specificity of PRC2from methylating H3K27 to EED. In addition to determining themolecular basis of EED-methyllysine recognition, ourwork providesthe biochemical characterization of how the activity of a histonemethyltransferase is oppositely regulated by two histone marks.

methyllysine-binding domain ∣ WD40 repeat-containing protein ∣X-ray crystallography

The polycomb (PcG) and trithorax groups of proteins act inconcert to maintain the transcriptional state of developmental

control genes such as the Hox genes. These patterns are estab-lished during early embryonic development and are heritablethrough many generations of cell division (1–3). Polycomb groupproteins are transcriptional repressors that maintain target genesin an inactive state, whereas trithorax group proteins are tran-scriptional activators that maintain their target genes in an activestate. This system of cellular memory is conserved in nearly allmulticellular eukaryotes (4).

PcG proteins often function as complexes, and the two best-characterized complexes are the polycomb repressive complex1 (PRC1) and 2 (PRC2). PRC2 exhibits histone methyltransferaseactivity on H3K27 (2, 5–7) and weakly on H1K26 (8). Studiesin Drosophila have shown that PRC1 contains a chromodomainprotein, polycomb, which preferentially recognizes the histoneH3K27me3 mark (9, 10). One hypothesis is that H3K27 methyl-ation by PRC2 creates a binding site recognized by the chromo-domain of polycomb of PRC1, which in turn maintains transcrip-

tional silencing through an unknown mechanism (11). In additionto silencing Hox genes, the polycomb group complexes are alsoinvolved in X-inactivation, germ-line development, stem cellpluripotency and differentiation, and cancer metastasis (2).

PRC2 complex contains four core components: EZH2, EED,SUZ12, and RbAp46/48. The catalytic activity of PRC2 is con-ferred by the suppressor of variegation [Su(var)3-9], enhancerof zeste [E(z)], and trithorax (SET) domain in EZH2, but PRC2exhibits robust methylation activity only as a complex because invitro enzymatic assays indicate that EZH2 has virtually no histonemethylation activity on its own (6). Biochemical and geneticstudies suggest that both EED and SUZ12 are required for theintegrity of PRC2 and PRC2-mediated H3K27 methylation be-cause mutations in either protein lead to EZH2 to destabilizationand deficient methylation of H3K27 (12–14).

EED contains seven WD40 repeats at its C terminus (residues81–441) preceded by a small N-terminal domain (residues 1–80).Studies of theDrosophila EED orthologue extra sex combs (ESC)revealed that the N-terminal region interacts directly with thecore domain of histone H3 and that this interaction is essentialfor E(Z)-dependent trimethylation of H3K27 (15). Furthermore,although an N-terminally truncated ESC can form a complexwith E(Z) when expressed in stable S2 cell lines, this complexcannot carry out trimethylation of histone H3 (15). A similarmechanism may regulate PRC2 function in vertebrates becausethe ESC/EED–histone H3 interaction is evolutionarily conserved(15).

The WD40 repeat is a structural motif of approximately40 amino acids, which forms a four-stranded antiparallel β-sheetand often contains a Gly-His dipeptide at the end of the fourthstrand and a Trp-Asp dipeptide at the end of the third strand (16).Most WD40 repeat proteins contain seven or eightWD40 repeatsand adopt a β-propeller architecture. WD40 proteins are involvedin diverse functions such as transcription regulation, signal trans-

Author contributions: C.X., C.B., W.Y., M.G., S.S.-C.L., H.W., and J.M. designed research;C.X., C.B., W.Y., M.G., H.O., W.Q., H.L., A.E.J., F.M., and P.P. performed research; C.X.,C.B., W.Y., M.G., H.O., C.C., W.Q., H.L., S.S.-C.L., H.W., and J.M. analyzed data; and J.M.wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.

Data deposition: The crystallography, atomic coordinates, and structure factors havebeen deposited in the Protein Data Bank, www.pdb.org (PDB ID codes 3JZN, 3K26,3K27, 3JZG, 3JZH, and 3JPX).1C.X., C.B., and W.Y. contributed equally to this work.2To whom correspondence may be addressed. E-mail: hbwang@uab.edu, sli@uwo.ca, or jr.min@utoronto.ca.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1008937107/-/DCSupplemental.

19266–19271 ∣ PNAS ∣ November 9, 2010 ∣ vol. 107 ∣ no. 45 www.pnas.org/cgi/doi/10.1073/pnas.1008937107

duction, vesicular trafficking, cytoskeletal assembly, cell cyclecontrol, and apoptosis (16). A common function of these WD40repeat proteins is to serve as a scaffold for protein–protein orprotein–DNA interactions and to coordinate downstream events.Since the first WD40 repeat structure Gβ was solved (17, 18), tensof WD40 structures have been reported, which, together, showthat the WD40 β-propeller is capable of recognizing a diversegroup of ligands, including proteins (17, 18), phosphopeptides(19–21), and nucleic acids (22), mainly through the small top sur-face. For example, WDR5 uses the top surface of its β-propellerto recognize histone H3K4 peptide (23–26) or a mixed lineageleukemia (MLL) fragment (27–29). WDR5 recognizes a peptidethrough an arginine residue that inserts into the central poreof the β-propeller and anchors the peptides into a shallow grooveat its top face.

WD40 proteins can also bind their ligands through the largerbottom surface. For example, the structure of the WD40 repeatdomain of EED in complex with an N-terminal fragmentof EZH2 revealed that the peptide bound to the WD40 repeatdomain of EED on the larger bottom surface (Fig. 1A) (30).Mutations in the WD40 region of the ESC perturbed its bindingto E(Z) in vitro and abolished its silencing function in vivo (31).

Recently, it was shown that the EZH2–EED–SUZ12 trimericPRC2 complex binds to and colocalizes with the H3K27me3 markat sites of active DNA replication and in the G1 phase of a cellcycle (32). However, it is unclear which component is responsiblefor recognizing the H3K27me3 mark. An examination of the topsurface of the EED structure revealed an aromatic cage consist-ing of residues Phe97, Tyr148, Trp364, and Tyr365 (Fig. 1A). Anaromatic cage is a common feature of methyllysine-bindingmotifs (33) such as the chromodomain (9, 10, 34, 35) and themalignant brain tumor repeats (36, 37). This finding led us topropose that EED may possess a methyllysine-binding activitythat is important for PRC2 function. Here, we provide evidencein support of this notion. Specifically, we show that the WD40domain of EED is a histone methyllysine-binding motif that pre-ferentially recognizes Ala-Arg-Lys-Ser (ARKS) motif-containingtrimethylated H3K27, H3K9, and H1K26 peptides and that thenucleosome methylation activity of PRC2 is enhanced by bindingto the H3K27me3 histone peptide via EED but inhibited byH1K26me3. We further investigated the molecular basis for the

selective binding of EED to trimethylated lysine histones by de-termining the crystal structure of apo-EED and its complexeswith five trimethylated histone peptides, respectively.

Results and DiscussionTheWD40 Repeat Domain of EED Selectively Recognizes TrimethylatedHistone Peptides. PRC2 methylates histone lysine residues, mainlyon H3K27 (2, 5–7), and also binds to the H3K27me3 mark (32).Efficient binding requires a ternary complex of EZH2, EED,and SUZ12 but is independent of the catalytic SET domain ofEZH2 (32). However, which component in PRC2 is responsiblefor binding to H3K27me3 is unclear. Aided by the crystal struc-ture of EED in complex with a short N-terminal fragment ofEZH2 (30), we identified an aromatic cage on the top face of theEED β-propeller structure (Fig. 1A), raising the possibility thatEEDmay mediate direct binding of PRC2 to H3K27me3 throughthis aromatic cage. To test this hypothesis, we measured thebinding of EED to a panel of peptides that represented thehistone marks H3K4, H3K9, H3K27, H3K36, H3K79, H4K20,and H1K26 in different methylation states (Fig. 1 B and C).We found that EED preferentially bound trimethylated histonepeptides, with higher affinities to the H3K9me3, H1K26me3,and H3K27me3 peptides than the other trimethylated histonepeptides. Our EED-histone peptide binding data agree withthe result of Margueron et al. that was published while this manu-script was under review (38).

H3K27me3 Stimulates Whereas H1K26me3 Inhibits the Methyltransfer-ase Activity of PRC2 on Nucleosomal Histone H3K27. The binding ofEED to trimethylated lysine marks on histones suggests a me-chanism for PRC2 to propagate and spread the H3K27me3 markto daughter strands during cell division. To investigate howmethylated histones may regulate the function of PRC2, we mea-sured its histone methyltransferase activity on nucleosome in thepresence of different histone peptides. Our results showed thatthe H3K27me3 peptide stimulated the enzymatic activity ofPRC2 by approximately three folds (Fig. 2A). These data are con-sistent with a recent report showing that H3K27me3 peptide sti-mulates the PRC2 activity (38). With the exception of H1K26peptide, we did not observe any significant effects of otherpeptides on the activity of PRC2. Addition of the H1K26me3

A

C

B

Fig. 1. Preferential binding of EED to trimethylatedlysine histone peptides. (A) The top surface of EED har-bors an aromatic cage consisting of Phe97, Tyr148,Trp364, and Tyr365, distinct from the EZH2-binding siteshown in the crystal structure of EED in complex withan EZH2 fragment (30). The aromatic cage residues onthe top surface of EED are displayed in a stick model.EED and the EZH2 peptide are colored in green andblue, respectively. (B) Binding of different trimethy-lated histone peptides to EED measured by fluores-cence polarization. (C) Tabulated binding affinities(Kd) of different histone peptides for EED, measuredby fluorescence polarization binding assays. The targetlysines are colored in red. NB, no detectable binding;ND, not determined.

Xu et al. PNAS ∣ November 9, 2010 ∣ vol. 107 ∣ no. 45 ∣ 19267

BIOCH

EMISTR

Y

peptide into the reaction inhibited nucleosomal histone H3K27methylation (Fig. 2A). Although histone H1 inhibition of PRC2methyltransferase activity on nucleosome was previously shown(8), the strong inhibitory effect of a methyllysine peptide onPRC2 activity has not been reported to date. To investigatewhether the methylation status of H1K26 affects the inhibitoryactivity, we carried out the in vitro methylation assay in the pre-sence of the H1K26 peptide with varying methylation states. In-terestingly, H1K26 peptides of lower methylation states exhibiteda weaker inhibitory effect compared to the H1K26me3 peptide(Fig. 2B). PRC2 has been shown to methylate H1K26 weakly(8). Presumably, H1K26 will exhibit full inhibition effect onlywhen it is trimethylated because lower methylated H1K26 pep-tides have much weaker binding affinity for EED (38).

To further investigate whether the inhibitory effect ofH1K26me3 is mediated by its direct binding to EED, we carriedout a peptide competition assay. Our results show that theH1K26me3 peptide binds to EED directly and that the bindingcan be competed off by the H3K27me3 peptide, suggesting thatH1K26me3 and H3K27me3 bind to the same site on EED(Fig. 1C and Fig. S1). We next investigated whether impairmentof H1K26me3 binding to EED affects the inhibitory effect ofPRC2 complex on nucleosome. On the basis of our EED–histonepeptide complex structures, which will be discussed later, the re-sidue at the P-2 position relative to the methyllysine at the histonepeptides is important for complex formation. Thus, an Ala to Glumutation at the P-2 position in the conserved ARKS motif ofH3K27me3, H3K9me3, and H1K26me3 would impair its binding

to EED, which is confirmed by our isothermal titration calorime-try (ITC) and peptide competition assays (Fig. S1 B and C). Con-sistent with its loss of binding, the H1K26me3A24E peptide wasunable to inhibit PRC2 activity (Fig. 2C). Taken together, thesedata clearly demonstrate that H1K26me3 inhibition of PRC2activity on nucleosome is conferred by its direct binding to EED.

It is therefore likely that, in vivo, the activity of PRC2 isregulated by both the H3K27me3 and the H1K26me3 marks thatplay opposing roles. This finding suggests a mechanism for dy-namic control of histone methylation in that the recruitmentof PRC2 by H3K27me3 enhances its activity and allows for pro-pagation of the H3K27me3 mark to sister chromatins during acell division, whereas recruitment of the PRC2 complex to theH1K26me3 mark shuts down the methylation activity toward corehistones.

Whereas the addition of the H1K26me3 peptide into the invitro methylation reaction inhibited nucleosomal H3 methyla-tion, it promoted methylation of another protein of approxi-mately 50 kDa (Fig. S2A). This protein was subsequentlyidentified as EED by liquid chromatography MS/MS. To furthercharacterize the methylation of EED, we used multiple reactionmonitoring mass spectrometry to identify the specific sites thatare methylated. We checked all 32 lysines in EED and obtainedhigh quality data on 23 lysines, of which K66, K197, K268, andK284 were methylated (Fig. S3). Moreover, all three methylationstates were detected for these four sites, and the most abundantmethylation states are shown in Fig. S3. The identities of thesemethyl peptides were confirmed by both multiple reactionmonitoring and the corresponding MS/MS spectra collected ininformation-dependent acquisition mode. Of these methylationsites identified in EED, the K66 site is located in a sequence motifof Gly-Arg-Lys-Ser that resembles the H3K27 sequence, ARKS.However, mutation of K66 to an alanine did not abolish themethylation of EED (Fig. S2B). Given the multiple sites thatcould be methylated in EED, this result is not unexpected. There-fore, our data indicate that, in contrast to H3K27me3 and otherhistone peptides, the H1K26me3 peptide changes the substratepreference of PRC2.

Overall Structure of Apo-EED and EED-H3K27me3. To investigate themolecular basis for the selective binding of EED to trimethylatedhistone peptides, we determined the crystal structure of apo-EEDand its complex with the H3K27me3 peptide (Fig. 3). As shownpreviously (30), the WD40 repeat domain of EED folds into aseven-blade β-propeller structure (Fig. 3A). The seven blades aresymmetrically arranged around a pseudosymmetry axis, and anarrow channel runs through the middle of the β-propeller struc-ture. The crystal structures of apo-EED and the EED–

H3K27me3 peptide complex were almost identical, with an rmsdof ∼0.2 Å for all aligned Cα atoms. As predicted from the struc-tural analysis, EED utilizes the smaller top surface of the β-pro-peller to recognize H3K27me3. Since the first WD40 repeatstructure Gβ was solved (17, 18), many WD40–ligand complexstructures have been determined, most of which use the smallertop surface to accommodate a ligand with a small interface area(Figs. S4 and S5).

The histone-binding mode of EED is reminiscent of WDR5,which also binds histone and other arginine-containing peptidesthrough its top surface (23–29). WDR5 binds histone and MLLpeptides mainly via an arginine residue of the peptide that actslike a pin and inserts into the central pore of the β-propeller toform a large network of hydrogen bonds and cation-π interactionswithWDR5 (23) (Fig. 3E). In the EED–H3K27me3 peptide com-plex structure, the histone peptide is anchored to EED throughthe trimethyllysine, which is accommodated by a binding pocketformed by four aromatic residues (Fig. 3 B and C). The side chainof the trimethylated lysine is inserted into an aromatic cageformed by F97, Y148, and Y365. The aromatic rings of these

293T No peptide me0 me1 me2 me3

EED WT + H1K26

EED Radiograph

A

B

H3 Radiograph

H3 Radiograph

293T no peptide H1K26me3 H1K26me3A24E

EED WT

H3 Radiograph

Coomassie

C

Coomassie

EED WT

1 2 3 4 5 6

Fig. 2. The H3K27me3 peptide specifically stimulates, but the H1K26me3peptide inhibits, the activity of PRC2. (A) Effects of different histone peptideson the activity of PRC2 on histone H3 methylation. Top is an autoradiographof the Middle (Coomassie blue staining). Quantification of two independentexperiments is shown in Bottom. Error bars are standard deviations of thetwo experiments. (B) Effects of different methylation states of the H1K26peptide on the methylation activity of PRC2. (C) The H1K26me3 peptidecontaining an A24E mutation lost its ability to inhibit PRC2 HMT activityon nucleosome.

19268 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1008937107 Xu et al.

three residues make close contact and are oriented roughly per-pendicular to each other, forming the three sides of the methyl-lysine-binding pocket and establishing van der Waals and cation-πinteractions with the trimethylated ammonium group. The ex-tended aliphatic chain of the methyllysine lies against the planarring of W364, further enhancing the hydrophobic interactionbetween the protein and the methylated lysine. This interactionmode is a unique methyllysine-binding feature for EED. The fouraromatic residues identified in the EED–H3K27me3 complex arehighly conserved in EED orthologues (Fig. S6). The trimethylly-sine-binding mode of EED is similar to that of other classical tri-methylated histone-binding proteins, such as HP1 and polycomb(9, 10, 34, 35).

The importance of the methyllysine-binding residues in EEDwas verified by subsequent mutagenesis experiments. Fluores-cence polarization measurements showed that mutating F97,Y148, W364, or Y365 to alanine, or W364 to leucine, completelyabolished binding of the resulting EEDmutant to the H3K27me3peptide (Fig. S7A). We next reconstituted the PRC2 complexwith the EEDmutant W364A or Y365A and examined its histonelysine methyltransferase (HKMT) activity. The activity of the re-

constituted PRC2 was severely diminished when these methylly-sine-binding residues in EED were mutated (Fig. S7B). Besidesproviding biochemical evidence to support the structure, thesedata suggest that the function of PRC2 is dependent on EED,in particular its WD40 repeats. Our data agree with that obtainedfor the Drosophila EED orthologue, ESC (31), in which certainpoint mutations in the WD40 repeats of ESC perturbs its bindingto E(Z) (31).

In addition to the interactions of methyllysine with the aro-matic cage, other interactions may also help stabilize the complex.These include a backbone hydrogen bond between Trp364 inEED and H3R26 in the peptide, a hydrogen bond between theside chain of EED Trp364 and the backbone of H3A25, andthe side chain of EED Arg414 and the backbone of methyllysine(Fig. 3C). Interestingly, Arg414 is absolutely conserved in allorthologues of EED and forms a salt bridge with anotherconserved residue Asp430 (Fig. 3C). The Arg414–Asp430 pairmay therefore play an important role in binding methyllysine. Re-sidues C-terminal to the methyllysine in the peptide form virtuallyno interactions with the protein, which explains why EED bindsits ligands more weakly (Kd: 108 μMwith H3K27me3) than do the

Fig. 3. Structure of EED in complex with a trimethylatedH3K27 peptide. (A) Top view of the EED WD40 domain incomplex with a trimethylated H3K27 peptide. EED is shownin a cartoon representation and colored green, and theH3K27me3 peptide is shown in a stick model. (B) The com-plex structure of EED and H3K27me3. EED is shown in a sur-face representation with positive electrostatic potentialdenoted in blue and negative potential by red. TheH3K27me3 peptide is depicted in a stick model. (C) Detailedview of intermolecular interactions between EED andH3K27me3. EED residues in contact with the H3K27me3peptide are shown in sticks. Hydrogen bonds are markedwith black dashed lines. (D) Superposition of the crystalstructures of EED in complex with different trimethylatedhistone peptides. Methyllysines are shown in a stick model.(E) Top view of WDR5 in complex with an H3K4me2 peptide(23). The H3K4me2 peptide and the H3R2-binding residuesin WDR5 are shown in stick models. Hydrogen bonds areshown in red dashed lines.

Xu et al. PNAS ∣ November 9, 2010 ∣ vol. 107 ∣ no. 45 ∣ 19269

BIOCH

EMISTR

Y

HP1 chromodomain (Kd: 4 μMwith H3K9me3) and the JMJD2Atudor domains (Kd: 0.4 μM with H3K4me3) to their respectivepeptide ligands (10, 39).

Preferential Recognition of the ARKme3S Motif by EED. Besides thecomplex structure of EED with the H3K27me3 peptide, we alsodetermined the crystal structures of EED complexed with theH3K4me3, H3K9me3, H3K79me3, and H4K20me3 peptides,respectively. In all five complex structures, the backbone orienta-tions of the peptides are almost identical, suggesting that theinteraction of methyllysine with the aromatic cage drives the com-plex formation (Fig. 3D). However, the five histone peptides bindto EED with varying affinities, and EED preferentially bindsH3K9me3, H1K26me3, and H3K27me3, which share a conservedARKS sequence motif.

We compared the five EED complex structures to understandthe structural basis for the preferential binding of the H3K27me3,H1K26me3, and H3K9me3 marks by EED. This comparativeanalysis identified two residues preceding the methyllysine in apeptide (referred to as the P-1 and P-2 positions) as key determi-nants of such specificity. In the EED-H3K27me3 complex, theP-2 residue alanine (H3A25) is accommodated in a relativelyhydrophobic and shallow pocket consisting of Y308, C324, andW364 (Fig. 3C). H3K9me3 also contains an alanine residue(H3A7) at the same position, so this binding mode is conservedin the EED–H3K9me3 complex (Fig. S8A). In contrast, a histi-dine residue is found at the corresponding position in theH4K20me3 peptide. The histidine side chain would be too bulkyto be inserted into the P-2 pocket without causing significant ster-ic hindrance. Indeed, in the corresponding crystal structure, thishistidine residue (H3H18) points to the solvent and makes a hy-drogen bond with the backbone carbonyl oxygen of D362 in EED(Fig. S8B). This alternative mode of P-2 recognition could explainwhy the H4K20me3 peptide has comparable binding affinity toEED. In the EED–H3K79me3 complex structure, the corre-sponding residue Asp (H3D77) is disordered, suggesting that itdid not contribute significantly to binding (Fig. S8C). In the caseof the H3K4me3 peptide, the long side chain of the H3R2 couldnot be accommodated by the shallow binding pocket and is alsodisordered (Fig. S8D).

In all the EED–peptide complexes, the peptide residue imme-diately preceding the methyllysine (P-1 position) protrudes outof the binding interface and points to the solvent. Therefore, ahydrophobic residue at this position would not be favored ener-getically, which is the case for H3K36me3 (V35 at P-1) andH3K79me3 (F78 at P-1). Together, these structures explainwhy EED preferentially binds H3K9me3, H3K27me3, andH1K26me3 marks.

Implication of EED-H3K27me3 Binding in the Transmission of theH3K27me3 Epigenetic Mark. On the basis of our binding assaysand structural studies, we conclude that (i) EED binds to multipletrimethylated histone peptides, but preferably to repressivemarks, (ii) the recognition of the methyllysine residue by the aro-matic cage plays a key role in the formation of an EED–methyl-lysine peptide complex, (iii) residues at the P-1 and P-2 positionsrelative to the methyllysine in a peptide are responsible for thedifferential binding affinities, and (iv) the H3K27me3, H3K9me3,and H1K26me3 peptides all bind EED with comparable affinitiesbut exhibit different outcomes: The H3K27me3 peptide stimu-lates PRC2 HKMT activity, H1K26me3 inhibits PRC2 HKMTactivity, and H3K9me3 has no effect.

It has been proposed that H3K27me3 binding to PRC2 intro-duces a conformational change in PRC2 and activates the PRC2HKMT activity (38). It is likely that H1K26me3 changes theconformation of PRC2 in a different way that makes EED a pre-ferred substrate for PRC2, whereas H3K9me3 binding to PRC2does not change its conformation significantly. Furthermore,

H3K9 is not a substrate of PRC2, and, even though H3K9me3can localize with PRC2 in cells, it does not form a feedback loopthat helps spread the H3K9me3 mark. The structures of theholoenzyme in complex with these peptides would help clarifythe distinct effects of different histone peptides on PRC2.

The ability of EED to bind to the product of PRC2(H3K27me3) provides an effective means for the propagationand/or spreading of the H3K27me3 repressive mark. In quiescentcells, recruitment of PRC2 via EED to an H3K27me3 mark mayhelp propagate the mark to neighboring nucleosomes. In postmo-totic cells, targeting of PRC2 to the H3K27me3 mark associatedwith the parent DNA strand may help spread this mark to thenascent sister strand. This mode of dual function for a methyl-transferase or coexistence of catalytic and effector functions inthe same protein complex is not without precedent. For instance,Clr4 is a histone H3K9 methyltransferase (40, 41) that also con-tains a chromodomain at its N terminus, which binds specificallyto the H3K9me3 mark it creates. The ability of Clr4 to generateand bind H3K9me3 is essential for the spreading of heterochro-matin silencing (42). Similar modes of regulation may underliethe function of other histone-modifying enzymes.

In summary, our work identifies EED as a central playerin PRC2 function. The unique ability of EED to bind multiplehistone marks, which leads to distinct outcomes, may play apivotal role in controlling the activity of PRC2 and its functionin propagating and spreading repressive histone marks in the nu-cleus. By mediating a series of interactions—binding the histoneH3 core domain via the N-terminal fragment, the N-terminalmotif of EZH2 via the bottom surface, and the H3K27me3/H1K26me3 mark via the top surface of the EED WD40 do-main—EED may function as an adaptor and/or conformationalswitch that controls the integrity, targeting, activity, and specifi-city of PRC2. It should be noted, however, that the affinity ofEED for H3K27me3 is significantly weaker than that of HP1and Pc chromodomains for the H3K9me3 or H3K27me3 peptide.The recruitment of the PRC2 complex to the correct chromatinlocation must therefore involve multiple sites of EED and, likely,multiple components of PRC2. Indeed, it has been shown that allcomponents of PRC2 are required for optimal binding (32). Inaddition, it has been shown that RbAp46/46 is capable of bindingthe first helix of histone H4 (43–45).

MethodsProtein Expression and Purification. Two fragments of human EED (aminoacids 40–441 and 76–441) covering the WD40 repeats were subcloned intoa modified pET28GST-LIC vector. The recombinant proteins were overex-pressed at 16 °C in Escherichia coli BL21 (DE3) Codon plus RIL (Stratagene)as N-terminal GST-tagged fusion proteins and were purified by affinity chro-matography on glutathione-sepharose (GE Healthcare). After cleavage usingthrombin (Sigma Aldrich) on the column at room temperature for 4–5 h, theflow-through was collected and purified further by size exclusion chromato-graphy (Superdex 200; GE Healthcare). The proteins were concentrated to15 mg∕mL in a buffer containing 20 mM Tris-HCl, pH 7.5, 0.2 M NaCl, and1 mM DTT. Mutated EED cDNAs were made by using the QuikChange IIXL Site-Directed Mutagenesis Kit (Stratagene); mutations were confirmedby sequencing complete cDNAs. Mutated proteins were expressed in bacter-ial cells and purified as above.

Histone Methyltransferase Assay. Lentiviral vectors encoding wild-type andmutant FLAG-EED were transfected into a 293T cell by using Effectenereagents (Qiagen). Then 48 h after transfect ion, cells were harvested andlysed in nondenaturing lysis buffer (10 mM Tris-HCl, pH 7.5, 150 mM NaCl,1% NP-40, 1 mM EDTA) at 4 °C for 1 h. After brief centrifugation(16;100 × g for 10min), cell lysates were subjected to immunoprecipitate withanti-FLAG M2 agarose beads (Sigma) at 4 °C for 3 h. PRC2 complexes wereeluted from the beads with 0.4 mg∕mL FLAG peptide (0.4 mg∕mL) anddialysed against BC50 before used for the histone methyltransferase assay.The histone methyltransferase assay was performed as described previously(46). Briefly, affinity purified PRC2 samples were incubated with 10 μg ofoligonucleosome or mononuclesomes prepared from a HeLa-S3 cell (47),1 μL 3H-S-adenosylmethionine (NEN Life Science Products), and histone H3

19270 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1008937107 Xu et al.

peptides (40 mM) in histone methyltransferase buffer (20 mM Tris-HCl, pH8.0, 4 mM EDTA, 1 mM PMSF, 0.5 mM DTT) at 30 °C for 1 h. The reaction pro-ducts were resolved on 15% SDS-PAGE. After staining and destaining, gelswere treated with 3H enhancer (En3hance; Perkin Elmer), dried, and exposedto X-film for 1–3 d.

Crystallization and Structure Determination. Purified protein EED (40–441) wasmixed with histone trimethylated peptides by directly adding a twofoldmolar excess of peptide to the protein solution and crystallized by usingthe sitting drop vapor diffusion method at 18 °C. The complex of GST-tagcleaved EED and H4K20me3 peptide was crystallized in a buffer containing3.5 M sodium formate, 10 mM Tris(2-carboxyethyl)phosphine (TCEP) chloride.For the other complexes, EED (76–441) was mixed with the histone peptidesby directly adding a twofold molar excess of peptides to the protein solution.All those complexes were crystallized by using the sitting drop vapor diffu-sion method at 18 °C and in a buffer containing 0.1 M 1,3-bis[tris(hydroxy-methyl)methylamino]propane, 3.5 M sodium formate, 10 mM TCEP, and15–20% glycerol. Prior to flash-freezing the crystals in liquid nitrogen, thecrystals were soaked in a cryoprotectant consisting of 100% reservoir solutionand 20% glycerol.

X-ray diffraction data were collected at 100 K by using the Rigaku FREHigh Brilliance X-Ray Generator with R-AXIS IV detector and Canadian LightSource. Data were processed by using the HKL software package (48). Allstructures were solved by molecular replacement by using the program

MOLREP (49). The crystal structure of EED and EZH2 peptide was used asthe search model (Protein Data Bank ID code 2QXV). ARP/wARP was usedfor automatic model building (50). Graphics program COOT (51) was usedfor model building and visualization. Crystal diffraction data and refinementstatistics for these structures are displayed in Table S1.

Note. While this manuscript was under consideration, R. Margueron andcolleagues published similar work (38).

ACKNOWLEDGMENTS. We thank Alexey Bochkarev and Guillermo Senisterrafor advice and technical assistance and Yanming Wang, Aled Edwards,and Cheryl Arrowsmith for critically discussing and reading the manuscript.This work was supported by the Structural Genomics Consortium, a regis-tered charity (1097737) that receives funds from the Canadian Institutesfor Health Research, the Canadian Foundation for Innovation, GenomeCanada through the Ontario Genomics Institute, GlaxoSmithKline, Karolins-ka Institutet, The Knut and Alice Wallenberg Foundation, the Ontario Inno-vation Trust, the Ontario Ministry for Research and Innovation, Merck & Co.,Inc., the Novartis Research Foundation, the Swedish Agency for InnovationSystems, the Swedish Foundation for Strategic Research, and the WellcomeTrust. H.W. is a Sidney Kimmel Scholar and is supported by National Institutesof Health Grant GM081489. S.S.-C.L. holds a Canada Research Chair in Func-tional Genomics and Cellular Proteomics. Open access publication costs weredefrayed by the Ontario Genomics Institute Genomics Publication Fund.

1. Schwartz YB, Pirrotta V (2007) Polycomb silencing mechanisms and the managementof genomic programmes. Nat Rev Genet 8:9–22.

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

3. Martin C, Zhang Y (2007) Mechanisms of epigenetic inheritance. Curr Opin Cell Biol19:266–272.

4. Pirrotta V, Gross DS (2005) Epigenetic silencing mechanisms in budding yeast and fruitfly: Different paths, same destinations. Mol Cell 18:395–398.

5. Czermin B, et al. (2002) Drosophila enhancer of Zeste/ESC complexes have a histone H3methyltransferase activity that marks chromosomal Polycomb sites. Cell 111:185–196.

6. Muller J, et al. (2002) Histone methyltransferase activity of a Drosophila polycombgroup repressor complex. Cell 111:197–208.

7. Cao R, et al. (2002) Role of histone H3 lysine 27 methylation in polycomb-groupsilencing. Science 298:1039–1043.

8. Kuzmichev A, Jenuwein T, Tempst P, Reinberg D (2004) Different EZH2-containingcomplexes target methylation of histone H1 or nucleosomal histone H3. Mol Cell14:183–193.

9. Min J, Zhang Y, Xu RM (2003) Structural basis for specific binding of polycombchromodomain to histone H3 methylated at Lys 27. Genes Dev 17:1823–1828.

10. Fischle W, et al. (2003) Molecular basis for the discrimination of repressivemethyl-lysine marks in histone H3 by polycomb and HP1 chromodomains. GenesDev 17:1870–1881.

11. Wang L, et al. (2004) Hierarchical recruitment of polycomb group silencing complexes.Mol Cell 14:637–646.

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

13. Pasini D, Bracken AP, Jensen MR, Lazzerini Denchi E, Helin K (2004) Suz12 is essentialfor mouse development and for EZH2 histone methyltransferase activity. EMBO J23:4061–4071.

14. Cao R, Zhang Y (2004) SUZ12 is required for both the histone methyltransferaseactivity and the silencing function of the EED-EZH2 complex. Mol Cell 15:57–67.

15. Tie F, Stratton CA, Kurzhals RL, Harte PJ (2007) The N terminusof Drosophila ESC bindsdirectly to histone H3 and is required for E(Z)-dependent trimethylation of H3 lysine27. Mol Cell Biol 27:2014–2026.

16. Smith TF, Gaitatzes C, Saxena K, Neer EJ (1999) TheWD repeat: A common architecturefor diverse functions. Trends Biochem Sci 24:181–185.

17. Lambright DG, et al. (1996) The 2.0 A crystal structure of a heterotrimeric G protein.Nature 379:311–319.

18. Wall MA, et al. (1995) The structure of the G protein heterotrimer Gi alpha 1 beta 1gamma 2. Cell 83:1047–1058.

19. Nash P, et al. (2001) Multisite phosphorylation of a CDK inhibitor sets a threshold forthe onset of DNA replication. Nature 414:514–521.

20. Wu G, et al. (2003) Structure of a beta-TrCP1-Skp1-beta-catenin complex: Destructionmotif binding and lysine specificity of the SCF(beta-TrCP1) ubiquitin ligase. Mol Cell11:1445–1456.

21. Orlicky S, Tang X, Willems A, Tyers M, Sicheri F (2003) Structural basis for phosphode-pendent substrate selection and orientation by the SCFCdc4 ubiquitin ligase. Cell112:243–256.

22. Scrima A, et al. (2008) Structural basis of UV DNA-damage recognition by theDDB1-DDB2 complex. Cell 135:1213–1223.

23. Schuetz A, et al. (2006) Structural basis for molecular recognition and presentation ofhistone H3 by WDR5. EMBO J 25:4245–4252.

24. Han Z, et al. (2006) Structural basis for the specific recognition of methylated histoneH3 lysine 4 by the WD-40 protein WDR5. Mol Cell 22:137–144.

25. Ruthenburg AJ, et al. (2006) Histone H3 recognition and presentation by the WDR5module of the MLL1 complex. Nat Struct Mol Biol 13:704–712.

26. Couture JF, Collazo E, Trievel RC (2006) Molecular recognition of histone H3 by theWD40 protein WDR5. Nat Struct Mol Biol 13:698–703.

27. Song JJ, Kingston RE (2008) WDR5 interacts with mixed lineage leukemia (MLL)protein via the histone H3-binding pocket. J Biol Chem 283:35258–35264.

28. Patel A, Dharmarajan V, Cosgrove MS (2008) Structure of WDR5 bound to mixedlineage leukemia protein-1 peptide. J Biol Chem 283:32158–32161.

29. Patel A, Vought VE, Dharmarajan V, Cosgrove MS (2008) A conserved arginine-containing motif crucial for the assembly and enzymatic activity of the mixed lineageleukemia protein-1 core complex. J Biol Chem 283:32162–32175.

30. Han Z, et al. (2007) Structural basis of EZH2 recognition by EED. Structure15:1306–1315.

31. Tie F, Furuyama T, Harte PJ (1998) The Drosophila polycomb group proteins ESCand E(Z) bind directly to each other and co-localize at multiple chromosomal sites.Development 125:3483–3496.

32. Hansen KH, et al. (2008) A model for transmission of the H3K27me3 epigenetic mark.Nat Cell Biol 10:1291–1300.

33. Adams-Cioaba MA, Min J (2009) Structure and function of histone methylationbinding proteins. Biochem Cell Biol 87:93–105.

34. Jacobs SA, Khorasanizadeh S (2002) Structure of HP1 chromodomain bound to a lysine9-methylated histone H3 tail. Science 295:2080–2083.

35. Nielsen PR, et al. (2002) Structure of the HP1 chromodomain bound to histone H3methylated at lysine 9. Nature 416:103–107.

36. Min J, et al. (2007) L3MBTL1 recognition of mono- and dimethylated histones. NatStruct Mol Biol 14:1229–1230.

37. Li H, et al. (2007) Structural basis for lower lysine methylation state-specific readout byMBT repeats of L3MBTL1 and an engineered PHD finger. Mol Cell 28:677–691.

38. Margueron R, et al. (2009) Role of the polycomb protein EED in the propagation ofrepressive histone marks. Nature 461:762–767.

39. Botuyan MV, et al. (2006) Structural basis for the methylation state-specific recogni-tion of histone H4-K20 by 53BP1 and Crb2 in DNA repair. Cell 127:1361–1373.

40. Nakayama J, Rice JC, Strahl BD, Allis CD, Grewal SI (2001) Role of histone H3 lysine 9methylation in epigenetic control of heterochromatin assembly. Science 292:110–113.

41. Min J, Zhang X, Cheng X, Grewal SI, Xu RM (2002) Structure of the SET domain histonelysine methyltransferase Clr4. Nat Struct Biol 9:828–832.

42. Zhang K, Mosch K, Fischle W, Grewal SI (2008) Roles of the Clr4 methyltransferasecomplex in nucleation, spreading and maintenance of heterochromatin. Nat StructMol Biol 15:381–388.

43. Verreault A, Kaufman PD, Kobayashi R, Stillman B (1998) Nucleosomal DNA regulatesthe core-histone-binding subunit of the human Hat1 acetyltransferase. Curr Biol8:96–108.

44. Song JJ, Garlick JD, Kingston RE (2008) Structural basis of histone H4 recognition byp55. Genes Dev 22:1313–1318.

45. Murzina NV, et al. (2008) Structural basis for the recognition of histone H4 by thehistone-chaperone RbAp46. Structure 16:1077–1085.

46. Wang H, et al. (2003) mAM facilitates conversion by ESET of dimethyl to trimethyllysine 9 of histone H3 to cause transcriptional repression. Mol Cell 12:475–487.

47. Fang J, Wang H, Zhang Y (2004) Purification of histone methyltransferases from HeLacells. Methods Enzymol 377:213–226.

48. Minor ZOaW (1997) Processing of X-ray diffraction data collected in oscillation mode.Methods Enzymol 276:307–326.

49. Vagin A, Teplyakov A (2000) An approach to multi-copy search in molecular replace-ment. Acta Crystallogr D 56:1622–1624.

50. Perrakis A, Harkiolaki M, Wilson KS, Lamzin VS (2001) ARP/wARP and molecular re-placement. Acta Crystallogr D 57:1445–1450.

51. Emsley P, Cowtan K (2004) Coot: Model-building tools for molecular graphics. ActaCrystallogr D 60:2126–2132.

Xu et al. PNAS ∣ November 9, 2010 ∣ vol. 107 ∣ no. 45 ∣ 19271

BIOCH

EMISTR

Y