The N Terminus of Drosophila ESC Binds Directly to Histone H3 and Is Required for E(Z)-Dependent Trimethylation of H3 Lysine 27

Abstract

Polycomb group proteins mediate heritable transcriptional silencing and function through multiprotein complexes that methylate and ubiquitinate histones. The 600-kDa E(Z)/ESC complex, also known as Polycomb repressive complex 2 (PRC2), specifically methylates histone H3 lysine 27 (H3 K27) through the intrinsic histone methyltransferase (HMTase) activity of the E(Z) SET domain. By itself, E(Z) exhibits no detectable HMTase activity and requires ESC for methylation of H3 K27. The molecular basis for this requirement is unknown. ESC binds directly, via its C-terminal WD repeats (β-propeller domain), to E(Z). Here, we show that the N-terminal region of ESC that precedes its β-propeller domain interacts directly with histone H3, thereby physically linking E(Z) to its substrate. We show that when expressed in stable S2 cell lines, an N-terminally truncated ESC (FLAG-ESC61-425), like full-length ESC, is incorporated into complexes with E(Z) and binds to a Ubx Polycomb response element in a chromatin immunoprecipitation assay. However, incorporation of this N-terminally truncated ESC into E(Z) complexes prevents trimethylation of histone H3 by E(Z). We also show that a closely related Drosophila melanogaster paralog of ESC, ESC-like (ESCL), and the mammalian homolog of ESC, EED, also interact with histone H3 via their N termini, indicating that the interaction of ESC with histone H3 is evolutionarily conserved, reflecting its functional importance. Our data suggest that one of the roles of ESC (and ESCL and EED) in PRC2 complexes is to enable E(Z) to utilize histone H3 as a substrate by physically linking enzyme and substrate.

Polycomb group (PcG) and Trithorax group (TrxG) proteins are required for maintaining stable heritable expression patterns of many developmentally important genes (2, 32). Mutations of PcG genes and TrxG genes cause abnormal development and disease in mammals (3, 14, 17, 40), which has spurred broad interest in the mechanisms underlying PcG protein-mediated gene silencing and TrxG protein-mediated gene expression. Much attention has focused on the covalent modifications of histones by PcG and TrxG protein complexes. Some noteworthy discoveries have included the identification of a histone H3 and H4 methyltransferase activity possessed by the SET domain-containing proteins SU(VAR)3-9, E(Z), TRX, and ASH1 (22).

The Drosophila melanogaster 600-kDa ESC/E(Z) complex, also known as Polycomb repressive complex 2 (PRC2), which contains the PcG proteins ESC, E(Z), and SU(Z)12 as well as the histone H4 binding protein p55 (ortholog of mammalian RbAp48 and RbAp46), is recruited to specialized Polycomb response elements (PREs) and methylates histone H3 K27 in the surrounding chromatin of PcG target genes (4, 7, 16, 25). A second PcG complex, PRC1, binds to the methylated H3 K27 and silences the promoter. The precise mechanism of silencing remains poorly understood, but silencing is lost when K27 methylation by E(Z) is perturbed.

Mono-, di-, and trimethylated forms of H3 K27 (1me-, 2me-, 3meH3K27) can be detected in vivo (30). Only trimethyl H3 K27 serves as the epigenetic mark for Polycomb silencing and binds PRC1 via the chromodomain of the PC protein. In an E(z) loss-of-function mutant, all three H3 K27 methylated species are undetectable on late larval salivary gland polytene chromosomes, strongly suggesting that all H3 K27 methylation is carried out by E(Z) (8). E(Z) histone methyltransferase (HMTase) activity in vitro requires ESC (7, 26), and global H3 K27 methylation in embryos also requires ESC (13). Similarly, global H3 K27 methylation in mammals requires the ESC homolog EED (24). What then is the molecular basis of the ESC requirement for E(Z) HMTase activity? ESC is not absolutely required for the observed in vitro binding of recombinant PRC2 to nucleosomes, which is mediated primarily by the SU(Z)12 and p55 subunits (26), although its presence in the complex does increase the affinity of PRC2 for nucleosomes (26). Therefore, ESC might either alter the conformation of E(Z) to render it enzymatically active or play a role in the recognition and binding of histone H3, thereby enabling E(Z) to efficiently utilize it as a substrate.

ESC is comprised of seven C-terminal WD repeats (residues 61 to 425) preceded by a short N terminus that contains acidic, Ser/Thr-rich, and basic regions (27, 34). We previously showed that the WD region of ESC binds directly to the N terminus of E(Z) (35), while the N terminus of ESC (residues 1 to 60) mediates ESC dimerization and phosphorylation (38). ESC appears to be required for stable association of E(Z) with the 600-kDa complex in vivo or for the stability of E(Z) itself, since we detected no E(Z) in the 600-kDa complex in esc mutant extracts that contain no ESC protein (12). Similarly, the ESC homolog EED also appears to be required for stable association of the E(Z) homolog EZH2 within the corresponding mammalian complexes (5, 24).

To determine whether ESC mediates the binding of E(Z) to H3, we tested the interaction between ESC and H3 by coimmunoprecipitation from cotransfected Drosophila S2 cells, by glutathione S-transferase (GST) pull-down assays, and by Far Western blots. With these varied approaches, we show that (i) ESC specifically interacts with histone H3 but not with other core histones, (ii) the N terminus of ESC is necessary and sufficient for histone H3 binding, (iii) the H3 N-terminal tail and C-terminal residues are not required for H3 interaction with ESC, and (iv) ESC binding does not discriminate between posttranslationally modified and unmodified forms of H3.

The Drosophila ESCL protein, a paralog of ESC, is very similar to ESC and can substitute for ESC in recombinant PRC2 complexes in an in vitro HMTase assay (43). We show that the N terminus of ESCL also specifically interacts with histone H3. Moreover, the mammalian homolog of ESC, EED, also interacts with histone H3 via its N terminus, indicating that the interaction of ESC with histone H3 is evolutionarily conserved and suggesting that the association is functionally important.

To assess whether the binding of the ESC N terminus to histone H3 is required for methylation of H3 K27, we expressed full-length ESC and an N-terminally truncated ESC (FLAG-ESC61-425) in S2 stable cell lines. Both forms of ESC were found to be incorporated into complexes with E(Z) and bound to a Ubx PRE in a chromatin immunoprecipitation (ChIP) assay. However, unlike full-length ESC, incorporation of FLAG-ESC61-425 into the E(Z) complex prevents the trimethylation of histone H3 by E(Z), indicating that binding of the ESC N terminus to histone H3 is required for E(Z)-dependent trimethylation of histone H3.

MATERIALS AND METHODS

Plasmid constructs.

Expression of FLAG-ESC proteins (full-length, 61 to 425, and 1 to 169) and GST protein in Drosophila S2 cells was driven by the inducible metallothionein promoter of pRMHA3 as previously described (38). pRMHA3 constructs for expressing FLAG-PC and FLAG-ESCL were generated similarly by PCR with NheI-NsiI sites flanking the coding sequence. pRMHA3-FLAG-ESCLΔ(109-382) was derived from pRMHA3-FLAG-ESCL by deletion of the BglII fragment. Constructs for GST fusions of Drosophila core histones H2A, H2B, H3, and H4 were generated by PCR and inserted into the SacI-SalI sites of pRMHA3. Construct pET-H3-H4 was a gift from J. Kadonaga (18). pET-H3(1-96) was derived from pET-H3-H4 by deleting the NheI-BamHI fragment, blunting the ends with Klenow, and recircularizing.

Antibodies.

Goat anti-histone H3 polyclonal antibody (ab 12079) against a C-terminal peptide of H3 (IQLARRIRGERA) was obtained from Abcam. Rabbit anti-3meH3K27 antibody was a generous gift from T. Jenuwein, which was previously shown to be highly specific for the trimethylated form of H3 K27 (30). Rabbit anti-1meH3K27 and -2meH3K27 were from Upstate, and their K27 specificities were confirmed by staining polytene chromosome from wild-type and E(z) null mutant flies. Rabbit anti-2meH3K4, anti-2meH3K9, anti-histone H4, and anti-H2A (against 88-IRNDEELNKL-97 of human H2A) antisera were from Upstate. These histone antibodies recognized Drosophila histones and calf thymus histones. Anti-FLAG M5 monoclonal antibody (Sigma) was used for detection of FLAG (F)-tagged proteins. Goat anti-GST antibody was from Amersham. Anti-PcG protein antibodies [rabbit anti-E(Z), rabbit anti-ESC, and guinea pig anti-SU(Z)12] were described previously (36-38). Anti-ESCL antibodies (against GST-ESCL1-95) were raised in guinea pigs and purified on an ESCL(1-95)-coupled column to specifically recognize ESCL and not ESC. Monoclonal anti-β-tubulin antibodies (E7) were from the Developmental Studies Hybridoma Bank at the University of Iowa.

Cell culture and transient transfection.

Drosophila S2 cells (derived from late embryos) obtained from the Drosophila Genomics Resource Center were cultured at 27°C in Schneider's medium (Invitrogen) supplemented with 10% (for S2) fetal bovine serum (FBS) (Invitrogen). A second Drosophila S2 cell line, D. Mel-2 (Invitrogen), adapted for growth in the absence of FBS, was grown at 27°C in Drosophila serum-free medium (SFM; Invitrogen) and used in the generation of all stable transfectants described next. For transient expression of exogenous protein in S2 cells, transfections were performed according to the manufacturer's instructions. Briefly, 5 μg of purified endotoxin-free plasmid DNA, 150 μl of incomplete medium, and 30 μl of SuperFect reagent (QIAGEN) were mixed in a 1.5-ml tube by vortexing for 10 s and incubated at room temperature for 10 min. Complete medium (1 ml) was added to the reaction tube, and the resulting mixture was added dropwise to a 60-mm plate that contained logarithmically growing attached S2 cells for a 3-h incubation. After 24 h of culture, protein expression was induced by the addition of 0.5 mM CuSO₄ (final concentration) for 1.5 to 2 days at room temperature. Cells were harvested with a cell scraper, washed once with cold phosphate-buffered saline (PBS), and stored at −80°C for future use. Protein levels of transiently expressed FLAG-ESC and GST-H3 are higher than their endogenous levels but in the range of concentrations used by others (44).

Preparation of stable S2 cell lines.

Stable S2 cell lines that express FLAG-ESC proteins, including full-length, N-terminally truncated (residues 61 to 425), and C-terminally truncated (residues 1 to 169) forms as well as FLAG-PC, were generated as described above for transient transfection, with the following minor adjustments. For the stable transfections, the expression constructs (in pRMHA3 vector) were linearized and individually combined with linearized pCoHygro (Invitrogen) in a 19:1 ratio totaling 5 μg. In addition, following the 3-h incubation with the transfection complexes, the cells were washed three times with PBS and allowed to recover under the standard growth conditions of D. Mel-2 cells for 48 h prior to the addition of both 10% FBS and 0.5 μg/ml hygromycin (Sigma). Cells were split periodically into fresh Drosophila SFM supplemented with antibiotic and FBS for a period of 3 to 4 weeks to allow for their uniform growth.

Far-Western blots.

Calf thymus free histones (1 μg/lane) or whole S2 cells (∼1 × 10⁶ cells/lane) were separated by 15% sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) and transferred to a 0.2-μm nitrocellulose membrane (Whatman). The membrane was blocked with 10% nonfat dry milk in PBS (pH 7.4)-0.2% Tween 20 for 30 min and then incubated with 0.5 μg/ml of purified GST or GST fusion proteins: GST-ESC1-60, GST-ESC1-47, GST-ESCL1-95, GST-PC1-86 for 1 h at room temperature. The membrane was subsequently incubated with goat anti-GST antibodies (1:1,000 dilution) for 1 h and then with anti-goat immunoglobulin G peroxidase conjugate (Sigma) (1:10,000 dilution) for 1 h. Each incubation was followed by multiple washings with PBS (pH 7.4)-0.2% Tween 20. SuperSignal West Dura substrate (Pierce) (1:10 dilution) was used for visualization of bound antibody.

GST pull-down assays.

S2 cells transiently coexpressing a GST-histone fusion (or GST alone as a control) and FLAG-tagged proteins (ESC, ESCL, PC) were resuspended in 0.45 ml of buffer A (50 mM PBS [pH 7.8], 0.5% Triton X-100, 10 mM EDTA, 0.5 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, plus phosphatase and protease inhibitors) and then ruptured by sonication (20 s) on ice. Supernatant was collected after centrifugation (16,000 × g, 20 min at 4°C) and mixed with glutathione-Sepharose beads (20 to 30 μl) (Amersham Biosciences) for 1 to 1.5 h at 4°C. Beads were washed four times with 0.3 ml washing buffer (40 mM HEPES, pH 7.5, 0.3 M NaCl, 10% glycerol, 0.2% NP-40). After a final wash with buffer containing 0.15 M NaCl, bound proteins were eluted with 30 μl of 2× SDS sample buffer. Western analyses were performed with anti-FLAG M5 and anti-GST antibodies. GST-ESC1-60 and GST alone purified from Escherichia coli were also used for in vitro GST pull-down of calf thymus free histones (Roche) or mononucleosomes (assembled in vitro from recombinant histones and kindly provided by Cheng-Ming Chiang) as follows. Glutathione-Sepharose beads (30 μl) bound with GST fusion proteins were incubated with 0.1 mg of calf thymus free histones in 0.2 ml of buffer (15 mM HEPES, pH 7.5, 0.4 M NaCl, 0.11 M KCl, 5% glycerol, 0.5% NP-40) for 1 h at 4°C. Beads were washed six times (0.5 ml/wash) with the same buffer and once with buffer without salt. Bound proteins were eluted in 40 μl of 2× SDS sample buffer and analyzed by Western blotting (10 μl/lane).

Preparation of extracts from S2 cell nuclei and immunoprecipitation.

Nuclei were prepared from stable S2 cell lines expressing FLAG-ESC (full-length) and FLAG-ESC61-425. Cells (grown in a 75-cm² flask without FBS and antibiotic) were collected after 20 h of induction by 0.3 to 0.4 mM CuSO₄ and washed once with cold PBS. Cell nuclei were prepared and washed once with buffer 1 as previously described for the preparation of embryo nuclei (36). To release chromatin-associated proteins and to solubilize chromatin, cell nuclei were resuspended and incubated for 6 min at 37°C in buffer 1 supplemented with 0.8 U of micrococcal nuclease (MNase) (Sigma), protease inhibitors (36), and 1 mM CaCl₂. The remaining insoluble material was removed by centrifugation (30,000 × g for 30 min at 4°C). The extent of MNase digestion was determined by isolating DNA fragments from the supernatant and running them on an agarose gel. Prominent DNA bands of ∼170, 360, and 520 bp in size were detected, corresponding to mono-, di- and trinucleosomes. The supernatant (140 μl), containing released chromatin-associated proteins and solubilized nucleosomes, was incubated with anti-FLAG M2-agarose (40 μl) for 2 h at 4°C. The agarose beads were then washed six times with 0.4 ml of 50 mM Tris-HCl (pH 7.4) plus 0.15 M NaCl to remove unbound proteins. The bound proteins were eluted in 50 μl of 2× SDS sample buffer and analyzed by Western blotting.

Double-stranded RNAi in S2 stable cell lines.

To assess the function of the ESC N terminus, we used RNA interference (RNAi) to knock down endogenous esc and escl in the S2 stable cell lines described above and then induced expression of FLAG-ESC proteins (full-length, ESC61-425, and ESC1-169). RNAi was performed as previously described by Maiato et al. (20) with the following modifications. A template for double-stranded RNA (dsRNA) synthesis was generated by PCR that consisted of 490 bp of escl cloned between the 5′ and 3′ untranslated regions (UTRs) of esc flanked both 5′ and 3′ with a T7 promoter (5′-TTA ATA CGA CTC ACT ATA G GG-3′) and inserted into the pGEM-T Easy vector (Promega). The forward and reverse primers are as follows: for esc 5′ UTR, 5′-TAT TTT ACG CGA TCC GTG ACT G-3′ and 5′-CGC GCG GCA CGT TTG TAC-3′; for escl, 5′-ATG ACC GAA ACG AAT CCC A-3′ and 5′-AGT TTT TGC TGC ACT GGT GCT TA-3′; and for esc 3′ UTR, 5′-CTG GGC GCA ACC ACA AAC-3′ and 5′-AAG GCA GCA GAG TAA GCA ACA ACT ATT-3′. dsRNA was generated in vitro using the RiboMAX large-scale RNA production system-T7 (Promega) according to the manufacturer's instructions. For RNAi, Drosophila S2 stable cells were treated with 15 μg/ml of dsRNA for 8 days, with fresh dsRNA and media added every other day. During the final 4 days of RNAi, 0.2 mM CuSO₄ (final concentration) was added to half of the dsRNA-treated cells to induce expression of FLAG-ESC proteins. Cell harvests were counted to ensure that each sample had equal cell numbers. Whole-cell extracts (1 million cells/10 μl) were made by suspending the cells in an equal volumes of both 8 M urea, 4.0% CHAPS, 40 mM Tris, and 2× SDS sample buffer. Proteins were resolved by SDS-PAGE and visualized by Western blotting.

ChIP.

ChIP was conducted as described by Upstate with the following additions or alterations. After 20 h of induction with 0.3 mM CuSO₄, stable cell lines (2.0 × 10⁷), described above, were collected in 1.0 ml of Drosophila SFM and treated with 1.0% formaldehyde for 10 minutes at room temperature. The cross-linking reaction was stopped with the addition of 0.125 M glycine. Cell pellets were resuspended in 1.0 ml SDS lysis buffer, incubated on ice for 10 min, and sonicated 4 times for 10 s using a Branson Sonifier 450 (output, 2; duty cycle, 30%). The resulting sonicates were then diluted with 10 ml low-salt wash buffer without SDS (LSB-SDS) and dialyzed overnight against 2 changes of a 40× volume of LSB-SDS (Spectrum Laboratories Spectra/Por 2; molecular weight cutoff, 12,000 to 14,000) to remove excess SDS. A 3.5-ml quantity of each dialyzed extract was precleared with 70 μl protein G-Sepharose (GE Healthcare)-sonicated salmon sperm DNA (∼80 μg) for 2 h at 4°C, and 0.5 ml was set aside as input. Protein G-Sepharose (30 μl) and M2 agarose (Sigma) (30 μl) were, respectively, added to two 1.5-ml aliquots of each sample and incubated overnight at 4°C. The beads were washed once each with LSB-SDS, high-salt buffer without SDS, and LiCl wash, followed by two washes with Tris-EDTA (pH 8.0). Real-time PCR reactions were performed in a 25-μl final reaction volume using Platinum SYBR green qPCR SuperMix-UDG (Invitrogen) and a Chromo 4 real-time PCR system (MJ Research). Quantities are percentages of total input immunoprecipitated.

RESULTS

N terminus of ESC interacts with histone H3.

The PRC2 complex can methylate H3 K27, an activity that is intrinsic to the SET domain of E(Z). However, in vitro, the E(Z) protein itself has no detectable HMTase activity and does not bind to histones or nucleosomes (7, 26). The noncatalytic components in PRC2 are required for maximal HMTase activity (7, 26). SU(Z)12 is critical not only to stabilize E(Z) (24, 29) but is also required for nucleosome binding (26). Robust HMTase activity in vitro requires ESC and, to a lesser degree, p55 (7). Both ESC (27, 34) and p55 (39) contain 7 tandem WD motifs that are predicted to fold into β-propeller structures. The highly conserved mammalian ortholog of p55, RbAp48/46, binds directly to histone H4 (42) through N- and C-terminal sequences flanking its WD repeats (41). Consistent with this, a GST fusion protein containing Drosophila histone H4 (GST-H4) strongly and specifically binds to p55 in a GST pull-down assay (Fig. 1). This suggests that p55 may stimulate E(Z) activity by helping target it to histone/nucleosome substrates analogous to the stimulatory effect of RbAp48 on the histone acetyltransferase activity of HAT1 in the HAT1:RbAP48 heterodimer. Several other nuclear WD proteins have been shown to bind directly to histones, including yeast TUP1 (9) Drosophila Groucho (11), and human HIRA (19).

FIG. 1.

Drosophila histone H4 strongly and specifically binds to p55. In GST pull-down assays, in vitro-translated ³⁵S-labeled p55 binds to purified GST-H4 and GST-H4 (1-85) (lanes 3 and 4 in panel A) but not to GST or GST-H4 (1-34) (lanes 2 and 5 in panel A). ³⁵S-p55 strongly binds to GST-H4 (lane 6 in panel B) and very weakly to GST-H2A (lane 3 in panel B) but not to GST-H2B or GST-H3 (lanes 4 and 5 in panel B).

The requirement of ESC for the HMTase activity of E(Z) prompted us to test whether ESC also binds histones, particularly histone H3, a known target of E(Z) HMTase activity. Cotransfection of Drosophila S2 cells was used to test the interaction of ESC and histone H3 as previously described (38). To generate soluble H3 in vivo, we coexpressed GST-H3 (Fig. 2, lanes 4 to 6) with FLAG (F)-ESC in S2 cells and then pulled down GST-H3 and its associated proteins from cell extracts with glutathione-Sepharose. As shown in Fig. 2A, F-ESC was recovered with GST-H3 under washing conditions of 0.15 M and 0.3 M NaCl (Fig. 2A, lanes 5 and 6), but not with GST (38) or when GST-H3 is absent (Fig. 2A, lanes 2 and 3). The Polycomb (PC) protein has also been shown to bind H3 (1), specifically methylated H3 K27 (7, 10, 23). Since ESC is transiently associated with PC in early embryos (31), the association between GST-H3 and F-ESC could be indirect, due to binding of endogenous PC to GST-H3. However, this possibility is ruled out by the absence of any detectable PC in lanes 5 and 6 (Fig. 2A, bottom panel). As an additional control, we coexpressed GST-H3 with F-PC or F-ESC in S2 cells and performed GST pull-down assays. A faint band of F-PC was detected in lane 12 (Fig. 2A), indicating a relatively weak association of F-PC with GST-H3 compared to F-ESC (lane 15). Because we could detect the transiently expressed GST-H3 by Western blotting with anti-GST and anti-H3 antibodies but not with anti-trimethylated H3 K27 antibodies (data not shown), it is possible that the weaker binding of F-PC is due to the high proportion of unmethylated GST-H3. This is not surprising, since it has been estimated that no more than 10% of the total endogenous histone H3 is K27 trimethylated (8).

FIG. 2.

The N terminus of ESC interacts with histone H3. (A) FLAG-ESC (F-ESC) associates with GST-H3 in S2 cells. Affinity pull-down assays with glutathione-Sepharose were carried out from extracts of Drosophila S2 cells expressing F-ESC alone (lanes 1 to 3), F-ESC with GST-H3 (lanes 4 to 6, 13 to 15), F-PC with GST (lanes 7 to 9), and F-PC with GST-H3 (lanes 10 to 12). Western analysis was done with anti-FLAG M5 monoclonal antibody, rabbit anti-PC, and goat anti-GST antibodies. All input (in) lanes contain 20% of total sample used in pull-down assays. Lanes 1 to 3 and 7 to 9 are negative controls for affinity pull-downs. Another negative control involving coexpression of F-ESC with GST alone was shown previously (38) (also see lane 2′ in Fig. 3A). Lanes 2 and 5 show bound (b) proteins on beads under 0.15 M NaCl washing conditions, and lanes 3, 6, 9, 12, and 15 under 0.3 M NaCl washing conditions. Unbound (u, or flowthrough) proteins are shown in lanes 8, 11, and 14. (B) N-terminal deletions abolish binding of ESC to coexpressed GST-H3 in S2 cells. Similar affinity pull-down assays were performed under 0.3 M NaCl washing conditions for different deletion forms of F-ESC. (C) The N terminus of ESC retains its binding to GST-H3. S2 cells coexpressing F-ESC1-169 and GST-H3 (lanes 1 to 3) or FLAG-ESC1-169 and GST (lanes 4 to 6) were pulled down by glutathione-Sepharose as described above for panels A and B. Endogenous protein RPD3 (at bottom of panel B) and E(Z) (at bottom of panels B and C) show no binding to transiently expressed GST fusion protein, serving both as a negative control and as a sample loading control. (D) Summary of GST-H3 binding by all tested ESC proteins in S2 cells. +, binding; −, no binding; ±, weak binding.

To map the region of ESC that interacts with H3, we coexpressed GST-H3 with F-ESC constructs that contained deletions of various regions in S2 cells and conducted pull-down assays as above (Fig. 2B and C). Deletion of the ESC residues preceding the WD region (residues 1 to 60) (F-ESC61-425) abolished ESC binding to H3 (Fig. 2B, compare lane 6 to lane 3 in top panel), suggesting that the N terminus of ESC is responsible for H3 binding. Constructs containing deletions of the acidic region (residues 1 to 22) (Fig. 2B, lane 9) or Ser/Thr-rich region (residues 23 to 44) (Fig. 2B, lane 12), also disrupted ESC binding to histone H3. Deletion of ESC residues 48 to 149, which includes part of the basic region and part of the WD region, severely impaired its binding to H3 (Fig. 2B, lane 15). These results indicate that the acidic, Ser/Thr-rich, and basic regions in the ESC N terminus are required for H3 binding. Association of F-ESC1-169 with GST-H3 (Fig. 2C, lane 3) further indicates that the ESC N terminus mediates the interaction and that the C-terminal region of ESC containing the WD motifs is not required for H3 binding. The association of GST-H3 with F-ESC in S2 cell extracts survived high-salt wash conditions (400 mM NaCl, 110 mM KCl, and 0.5% NP-40). Thus, our results indicate a specific interaction of ESC with H3. As previously noted (38), endogenous E(Z) complexes were not pulled down with transiently expressed F-ESC (Fig. 2B and C, bottom panels), suggesting a direct interaction between F-ESC and GST-H3. The ESC constructs tested for H3 binding are summarized in Fig. 2D.

ESC interacts selectively with histone H3 but not H2A, H2B, or H4.

To further confirm a specific interaction of the ESC N terminus with H3, we coexpressed F-ESC in S2 cells with similar constructs that express the other core histones: GST-H2A, GST-H2B, or GST-H4. GST or GST-histone fusion proteins were pulled down with glutathione-Sepharose from cell extracts (Fig. 3A, lanes 2′ to 6′, bottom panel). While F-ESC was recovered with GST-H3 (lane 5′), no F-ESC was recovered with GST-H2A, GST-H2B, or GST-H4 (Fig. 3A, lanes 3′, 4′, and 6′, top panel). The levels of GST-H2A and GST-H2B were lower than GST-H3 and GST-H4 in both input and pull-down lanes (Fig. 3A, bottom panel) probably due to lower solubility and/or stability.

FIG. 3.

ESC specifically interacts with histone H3 but not other core histones. (A) F-ESC was coexpressed with GST (lane 2), GST-H2A (lane 3), GST-H2B (lane 4), GST-H3 (lane 5), and GST-H4 (lane 6) in S2 cells. Affinity pull-down assays (lanes 1′ to 6′) and Western blots were carried out as described for Fig. 2. A nonspecific band recognized by anti-GST in lanes 1 to 6 is indicated by an asterisk. Protein levels of GST-H2A and GST-H2B (lanes 3 and 4 in bottom panel) in cell extracts are lower than GST-H3 and GST-H4 (lanes 5 and 6), probably due to solubility. (B) Free histones from calf thymus and purified GST fusion proteins from E. coli were used for an in vitro pull-down assay with washing buffer containing 0.4 M NaCl, 0.11 M KCl, and 0.5% NP-40. Histones are detected by Western blotting. Five micrograms of free histones is loaded in lane 1 (Input) showing 20% of total sample. GST and GST-PCL-PHD (423 to 567 amino acids) in lanes 2 and 4 serve as negative controls.

To verify the S2 cell results and test whether the interaction between the N terminus of ESC and histone H3 is direct, we used GST-ESC1-60, purified from E. coli, to pull down free histones using 0.5 M salt washing conditions (400 mM NaCl, 110 mM KCl, and 0.5% NP-40) and calf thymus histones. Consistent with the S2 cell transfection data above, only H3 (Fig. 3B, panel a), but not H2A or H4 (panels b and c), was pulled down by GST-ESC1-60 (Fig. 3B, lane 3). Thus, these data confirm a specific and direct interaction of ESC with H3. They also indicate that this interaction does not simply involve a nonspecific charge interaction between the highly basic histones and the acidic region of ESC, since this would not be expected to discriminate between the different histones. Calf thymus free histones contain a variety of posttranslational modifications. Both dimethylated H3 K27 (2meH3K27) and H3 K4 (2meH3K4) were pulled down by GST-ESC1-60 (Fig. 3B, panels d and e), but acetylated H4 K16 (AcH4K16) was not (panel f), suggesting that ESC binds to histone H3 with no preference for H3 molecules containing specific modifications (see below). When we used in vitro-assembled mononucleosomes instead of free histones in pull-down assays with GST-ESC1-60, we failed to detect any histones in the bound fractions (data not shown), suggesting that ESC1-60 cannot interact with the histone H3 in a mononucleosome, consistent with a recent report that ESC alone does not bind to nucleosomes in vitro (26).

The N-terminal tail of H3 is not required for binding to ESC.

Mammalian RbAp48 and RbAp46 (homologs of p55) bind to the first α helix of the H4 globular domain and not to the H4 N-terminal tail (41, 42). Similarly, Drosophila p55 binds to GST-H4(1-85) but not GST-H4(1-34) (Fig. 1A). In GST pull-down assays, ESC binds to free histone H3 but not to nucleosomal H3, suggesting that the ESC-binding region of H3 may not be exposed in nucleosomes and that the N-terminal tail of H3 is probably not responsible for the interaction with ESC. To test this possibility, we coexpressed F-ESC1-169 with GST-H3(1-47), which contains only the N-terminal tail of H3, and GST-H3Δ(1-47), which is missing the N-terminal tail of H3, in S2 cells and pulled down GST fusion proteins from cell extracts. Both GST-H3 and GST-H3(1-47) were strongly recovered by glutathione-Sepharose beads (Fig. 4, middle panel, lanes 3 and 9, respectively). As previously shown in Fig. 2C (lane 3), F-ESC1-169 was strongly pulled down by GST-H3 (Fig. 4, top panel, lane 3), but none was pulled down by GST-H3(1-47) (lane 9), which indicates that the N-terminal tail of H3 does not interact with ESC. Truncation of the N-terminal tail of H3 seems to decrease the solubility/stability of GST-H3Δ(1-47), which resulted in a weak band in the bound fraction of GST-H3Δ(1-47) (Fig. 4, middle panel, lane 6). A weak band of F-ESC1-169 was also pulled down by GST-H3Δ(1-47) (Fig. 4, top panel, lane 6). Thus, these data indicate that the C-terminal globular domain of H3 is responsible for ESC binding (although it is possible that, in our assay, the N-terminal tail of H3 might enhance ESC binding to H3 due to the greater solubility of full-length H3).

FIG. 4.

The N-terminal tail of histone H3 does not interact with ESC. F-ESC1-169 was coexpressed with GST-H3 (lanes 1 to 3), GST-H3Δ(1-47) (lanes 4 to 6), and GST-H3(1-47) (lanes 7 to 9) in S2 cells. Affinity pull-down assays were done as described for Fig. 2B. For Western blots, the bottom halves of membranes were probed with anti-FLAG M5 or anti-GST antibodies; the top halves of membranes were probed with anti-RPD3 or anti-E(Z) antibodies. Endogenous RPD3 and E(Z) (two panels at bottom) in input lanes 1, 4, and 7 show equal sample loading, and their absences in bound fractions (lanes 3, 6, and 9) serve as negative controls.

N terminus of ESCL also interacts with H3.

Drosophila, unlike vertebrates, contains an esc paralog, called esc-like (escl), that is very similar to esc (43). We previously showed that the conserved WD region of ESC binds directly to the N terminus of E(Z) (35). Similarly, GST-ESCLΔ(1-95), which contains just the 7 WD motifs of ESCL, pulled down in vitro-translated E(Z) (R. L. Kurzhals, F. Tie, C. Stratton, and P. J. Harte, unpublished data). Aside from the additional 33 residues at the ESCL N terminus, the ESC and ESCL N-terminal sequences immediately preceding the WD motifs share a similar sequence composition and organization, with acidic, Ser/Thr-rich, and basic regions.

We tested whether ESCL also binds histone H3 by coexpressing GST-H3 or GST-H4 with F-ESCL in S2 cells (Fig. 5A, lanes 3 and 4) as above. As expected, F-ESCL was detected in the GST-H3 pull-down assay (lane 3′, second panel) but not in the GST-H4 pull-down (lane 4′ in second panel). To test whether the N terminus of ESCL is also required for its interaction with histone H3, we cotransfected GST-H3 with F-ESCL100-462, F-ESCL34-462 or F-ESCLΔ(109-382) and pulled it down as in Fig. 2B. Deletion of the 99 N-terminal residues of ESCL abolished the interaction between F-ESCL and GST-H3 (Fig. 5B, compare lane 3′ to lane 2′ in top panel), suggesting that the N terminus of ESCL is also required for H3 binding. The first 33 residues in ESCL, which are not present in ESC, are not required for H3 binding (Fig. 5B, lane 5′). Similar to F-ESC1-169 above (Fig. 2C), F-ESCLΔ(109-382), deletion of the ESCL WD repeats 1 to 5, retained its binding to GST-H3 (Fig. 5B, lane 4′). The ESCL constructs tested for H3 binding are summarized in Fig. 5C.

FIG. 5.

The N terminus of ESCL also binds to histone H3. Transiently cotransfected proteins in S2 cells are indicated at the top of each lane. Cell extracts were used for affinity pull-down assays as described for Fig. 2B. Western analyses of cell extracts (lanes 1 to 4 in panel A) or whole-cell proteins (lanes 1 to 5 in panel B) and affinity pull-down assays (lanes 1′ to 4′ in panel A and lanes 1′ to 5′ in panel B) were carried out as described above. β-Tubulin at the bottom in panels A and B serves both as sample loading and negative controls. Nonspecific bands recognized by anti-GST antibodies in the top panel and by anti-FLAG M5 antibodies in the second panel are indicated by asterisks. GST fusion and FLAG-tagged proteins are indicated by arrows. The results of mapping the region of ESCL for binding to GST-H3 are summarized in panel C.

N terminus of ESC is the minimal H3-binding region and binds both modified and unmodified histone H3.

Data from Fig. 2 and 3 show that ESC1-60 specifically binds to histone H3, and both the acidic region and Ser/Thr-rich region at the N terminus of ESC are required for the interaction. We also used far-Western blots to test the N-terminal region of ESC for histone H3 interaction. Free histones were separated by 15% SDS-PAGE and visualized by Coomassie blue staining (Fig. 6A, lane 1) or transferred to the membrane and overlaid with GST (lane 2, negative control) and GST fusion proteins (lanes 3 to 6) for far-Western blots. PC1-86 contains the chromodomain that specifically recognizes 3meH3K27. Consistent with this, overlay of GST-PC1-86 showed a strong signal at the H3 position (lane 6) and served as a positive control in far-Western blots. As expected, GST-ESC1-60 recognized histone H3 (lane 3). GST-ESCL1-95 also recognized histone H3 (lane 5). Partial deletion of the basic domain from ESC1-60 disrupted its binding to H3 (Fig. 6A, lane 4). Similarly, GST-ESC1-60, but not GST-ESC1-47, bound to histone H3 in far-Western blots using whole S2 cell extracts (Fig. 6B, lanes 6 and 5). Taken together, the results of these experiments and the transfection experiments described above (Fig. 2B) indicate that the N-terminal 60 residues of ESC comprise the minimal region for histone H3 binding.

FIG. 6.

Far Western assays for histone H3 binding. (A) Calf thymus free histones (∼1 μg/lane) separated by SDS-PAGE were stained by Coomassie blue (lane 1) or used for far-Western blots (lanes 2 to 6). Overlaid GST and GST fusion proteins are indicated on the top of lanes 2 to 6 and are detected by anti-GST antibodies. Arrows indicate bands showing that histone H3 was bound by GST-ESC1-60 (lane 3) or by GST-PC1-86 (lane 6). (B) Whole S2 cell lysates were separated as above and were visualized by Coomassie blue staining (lane 1), by Western blotting with anti-H3, -H4, and -H2A antibodies (lanes 2, 3, and 4, respectively), or by far-Western blots with overlay of GST-ESC1-47 (lane 5) or GST-ESC1-60 (lane 6). (C) E. coli cells (lanes 1 and 2) with (+) or without (−) IPTG induction for histone H3 expression and nuclei (lane 3) from overnight embryos were analyzed by Western blotting with anti-H3 antibodies (top panel) or by far-Western blotting with overlay of GST-ESC1-60 (bottom panel). (D) E. coli cells (lanes 1 and 2) with (+) or without (−) IPTG induction for histone H3 (1 to 96 amino acids) (lane 2) expression were Coomassie blue stained (top) or analyzed by far-Western blotting (bottom) as for panel C. (E) Nuclear extracts (NE, lane 1) and nuclear pellets (lane 2) from 293 cells (human embryonic kidney cell line) were analyzed by Western blotting using anti-CK2α, anti-H3 (second panel), and then anti-H2A and -H4 antibodies (third panel), or by far-Western blotting with overlay of GST-EED1-81 (right panel). (F) 293 cell nuclei partially digested with trypsin and were analyzed by anti-H3 antibodies (which only recognize the C-terminal region containing histone H3, left panel) or by far-Western blotting (right) with overlay of GST-EED1-81, which contains 81 residues of EED starting from the first methionine to the residue before the WD repeats.

As shown in Fig. 3B above, GST-ESC1-60 pulled down free histone H3 with no preference for a particular modification state. By far-Western blotting, we further showed that GST-ESC1-60 binds similarly to recombinant histone H3 expressed in E. coli (no modification) (Fig. 6C, lane 2) and to H3 isolated from embryo nuclei (with all modifications) (Fig. 6C, lane 3). While the N-terminal tail of histone H3 does not bind to ESC (Fig. 4, lane 9), E. coli-expressed H3(1-96), a C-terminally truncated form of H3, does bind to ESC (Fig. 6D, lane 2). Thus, these data suggest that H3(47-96) contains the primary determinants of ESC binding.

Interaction between ESC and histone H3 is conserved in mammals.

If the binding of ESC (and ESCL) to histone H3 is functionally important, we would expect it to be evolutionarily conserved in EED, a mammalian ortholog of ESC and ESCL. We tested whether EED1-81, the N-terminal portion of EED corresponding to ESC1-60 that is sufficient for histone H3 binding (Fig. 2 to 5), also binds to human histone H3. Like ESC1-60, this portion of EED contains acidic, Ser/Thr-rich, and basic regions (38). GST-EED1-81 was purified from E. coli and used as a probe on a far-Western blot containing human 293 cell nuclear extracts (Fig. 6E, lane 1) and nuclear pellets (lane 2). Although all core histones are present in nuclear pellets (Fig. 6E, middle panels), GST-EED1-81 bound to histone H3 only (Fig. 6E). Partial trypsin digestion of cell nuclei removes the N-terminal tail of histone H3 (Fig. 6F, left panel). Like ESC1-60, EED1-81 bound both histone H3 and its N-terminally truncated form (Fig. 6F, right panel). Thus, these data indicate that the interaction between ESC and histone H3 is conserved in mammals.

ESC links E(Z) to histone H3 in chromatin in vivo.

Binding of ESC to both histone H3 and E(Z) suggests that the observed requirement for ESC for E(Z) enzyme activity may result from the physical linkage that ESC creates between E(Z) and its substrate. To determine the requirement of the ESC N terminus in the histone H3/ESC/E(Z) complex in vivo, we generated stable cell lines that inducibly express FLAG-ESC and FLAG-ESC61-425, respectively, since FLAG-ESC expressed by transient transfection did not associate with endogenous E(Z) (38). FLAG-ESC and FLAG-ESC61-425 were immunoprecipitated from extracts of these stable cells with anti-FLAG M2 agarose, and endogenous E(Z) was found to coimmunoprecipitate with both (Fig. 7A), indicating that both FLAG-ESC and FLAG-ESC61-425 form complexes with endogenous E(Z) in vivo. This is consistent with our previous evidence (35) that the N terminus of ESC is not required for E(Z) binding.

FIG. 7.

Truncation of the ESC N terminus abolishes binding to H3 but does not affect its ability to form complexes with E(Z) or bind to the Ubx PRE. (A) ESC links E(Z) to histone H3 in vivo via the N terminus of ESC. Nuclear extracts after MNase treatment from stable cell lines expressing FLAG-ESC (lanes 1 to 3) and FLAG-ESC61-425 (lanes 4 to 6) were immunoprecipitated by anti-FLAG M2 agarose. Bound proteins were analyzed by Western blots with anti-FLAG M5 (top panel), anti-E(Z), and anti-H3 antibodies (middle panels). FLAG-ESC full-length and 61 to 425 are indicated by arrows in the top panel. Immunoglobulin G heavy chains, IgG(H), are indicated by lines. Input (lanes 1 and 3) shows 5% of the total sample. A component of CAF-1 complex, p105, serves as both a loading control and a negative control. Note at the bottom panel that the same amount of p105 is present in lanes 1 and 2 and lanes 4 and 5 (input and unbound) but is absent in lanes 3 and 6 (bound). (B) Chromatin immunoprecipitation from stable cell lines expressing FLAG-ESC, FLAG-ESC61-425, and FLAG-PC. Two independent ChIPs were carried out (a and b) with anti-FLAG M2 agarose (columns 1 to 3, M2 IP) or protein G beads (column 4 to 6, mock). The Ubx PRE fragments were detected by real-time PCR (top). The expression of FLAG-tagged proteins (as arrows indicate) in stable cell lines were analyzed by Western blots (bottom). A nonspecific band is marked by an asterisk.

To examine the association of histone H3/ESC/E(Z) in vivo, we prepared extracts from cell nuclei treated with MNase to release chromatin-associated proteins and to partially solubilize nucleosomes (see Materials and Methods) and then performed coimmunoprecipitation assays with anti-FLAG M2. As shown in Fig. 7A, both FLAG-ESC and FLAG-ESC61-425 were pulled down by anti-FLAG M2 beads from the soluble nuclear fraction after MNase treatment (lanes 3 and 6, top panel). Endogenous E(Z) coimmunoprecipitated with both FLAG-ESC and FLAG-ESC61-425, consistent with our previous evidence that the WD region of ESC is sufficient for E(Z) binding (35). In addition, a small fraction of total histone H3 coimmunoprecipitated with FLAG-ESC but not FLAG-ESC61-425. However, we were unable to pull down any intact nucleosomes (see Discussion). Since there are no detectable (free) histones present in soluble nuclear extracts obtained without MNase treatment (Fig. 6E, lane 1), we presume that this ESC-associated histone H3 is likely to be of nucleosomal origin. In any case, these data support the cotransfection results in Fig. 2 and indicate that ESC physically links E(Z) to histone H3 in vivo via its N terminus.

Trimethylation of H3 K27 in vivo requires the N terminus of ESC.

E(Z), ESC, and SU(Z)12 proteins are chromatin associated and act through PREs. Since both full-length FLAG-ESC and FLAG-ESC61-425 can form complexes with E(Z), we tested whether truncation of the N terminus of ESC will affect the chromatin binding activity of PRC2 and the E(Z)-dependent methylation of H3 K27. We performed ChIP on the S2 stable cell lines with anti-FLAG M2 agarose and real-time PCR for detection and found that both FLAG-ESC and FLAG-ESC61-425 were bound to the bxd PRE of the Ubx gene (Fig. 7B), indicating that truncation of the ESC N terminus does not abolish the binding of complexes containing it to their specific target sites in chromatin. This is consistent with evidence that SU(Z)12 and p55 are the principal determinants of PRC2 chromatin binding activity (26).

In Drosophila S2 cells, either ESC or ESCL forms complexes with E(Z)and RNAi knockdown of both ESC and ESCL or E(Z) alone abolished 2me- and 3meH3K27 (data not shown). To test whether the N terminus of ESC is required for the methylation of H3 K27 by E(Z) in vivo, we knocked down endogenous ESCL and ESC and then examined H3 K27 methylation in the stable S2 cell lines after inducing expression of the FLAG-ESC or FLAG-ESC61-425, respectively. To avoid affecting the expression of the inducible FLAG-ESC constructs, we used a composite dsRNA that contained the esc 5′ and 3′ UTRs (but not the esc coding sequence) flanking the initial 490 bp of the escl coding region. With this dsRNA, knockdown of ESCL was efficient (Fig. 8A, top panel), but knockdown of ESC was less efficient than when using dsRNAs covering the esc coding sequence (data not shown). However, as shown below, it worked well enough to strongly reduce the level of trimethylated H3 K27. In normal S2 cells (Fig. 8A, left lane) and untreated S2 stable cell lines (Fig. 8A, lanes 1 to 3), the levels of ESCL, E(Z), and SU(Z)12 proteins were comparable, but in untreated S2 stable cell lines (Fig. 8A, lanes 1 to 3) the endogenous ESC level was markedly reduced for an unknown reason. (This was likely compensated for by ESCL since, as shown below, it did not affect methylated H3 K27 levels in untreated cells.) In dsRNA-treated cells, a low level of ESC protein was still detectable (Fig. 8A, second panel). Nonetheless, a great reduction in the level of trimethylated H3 K27 was observed. No significant effect was observed on the levels of mono- and dimethylated H3 K27 (see below). RNAi appeared to have no effect on expression of the three stably transfected FLAG-ESC constructs under copper ion induction (Fig. 8A, lanes 7 to 9 in bottom panel). Even though small amounts of endogenous ESC and ESCL remained after RNAi treatment, we expected that induction of FLAG-ESC61-425 might exert a dominant-negative effect if the ESC N terminus is required for H3 K27 methylation, since it appears to be efficiently incorporated into E(Z) complexes that are bound to their normal target sites in chromatin.

FIG. 8.

The N terminus of ESC is required for trimethylation of H3 K27 in vivo. Whole S2 stable cells (6 × 10⁵ cells/lane), which were stably transfected to express FLAG-ESC (lanes 1, 4, and 7), FLAG-ESC61-425 (lanes 2, 5, and 8), and FLAG-ESC1-169 (lanes 3, 6, and 9) under copper ion induction, were analyzed by Western blotting for proteins (A) and histones (B). Signals of each stable cell line in three groups were compared. Lanes 1 to 3 (group 1) are untreated S2 stable cells. Lanes 4 to 6 and 7 to 9 (groups 2 and 3) are cells treated with dsRNA for 8 days to knock down endogenous ESCL and ESC. Group 3 (lanes 7 to 9) was also treated with 0.2 mM CuSO₄ to induce FLAG-ESC expression for 4 days. Anti-FLAG M5 antibodies were used to detect FLAG-ESC (the bottom panel of A), and a nonspecific band is marked by an asterisk. Anti-ESC antibodies (against ESC1-60) recognize full-length ESC (the second panel of panel A) and FLAG-ESC1-169 (arrow indicating band in lane 9) but not FLAG-ESC61-425 (lane 8). Whole histone H3 (the top panel of B) was detected with goat anti-H3 antibodies after stripping signals of the previous Western blot (using rabbit anti-3meH3K27 antibodies) by incubation in the solution of 0.1 M NaOH. Tubulin and histone H3 serve as loading controls for proteins and histones, respectively. Arrows indicate 3meH3K27 in panel B. Note that the 3meK27 level increases in lane 7 of panel B (compare to lane 4) and decreases to undetectable in lane 8 (compare to lane 5) with similar E(Z) levels in lanes 7 and 8 in panel A.

We examined the levels of mono-, di-, and trimethylated H3 K27 as well as total H3 levels in whole-cell lysates derived from the stable transfectants described above. All forms of methylated H3 K27 and histone H3 were readily detected in untreated stable cells (Fig. 8B, lanes 1 to 3). No significant differences were observed in 1me- and 2meK27 levels before induction (Fig. 8B, lanes 4 to 6, two panels at the bottom) or after induction of the three different FLAG-ESC proteins (lanes 7 to 9) in the RNAi-treated cells (Fig. 8B, lanes 4 to 9). In contrast, with histone H3 as a loading control (Fig. 8B, top panel), 3meK27 was greatly decreased in RNAi-treated samples (Fig. 8B, lanes 4 to 6, second panel). In RNAi-treated cells in which expression of full-length FLAG-ESC was induced, 3meK27 was restored to a higher level than that of uninduced cells (Fig. 8B, compare lane 7 to lane 4, second and third panels). In contrast, in RNAi-treated cells where expression of FLAG-ESC61-425 was induced, the residual 3meK27 level was further reduced below detection (compare lane 8 to lane 5), suggesting that ESC61-425 exerts a dominant-negative effect on trimethylation. In RNAi-treated cells expressing FLAG-ESC1-169, which does not interact with E(Z) (data not shown), as expected, there was no significant difference in 3meK27 levels between induced and uninduced samples (lanes 6 and 9). These data strongly suggest that the N terminus of ESC is required for trimethylation of H3 K27 and serves to link E(Z) to its H3 substrate.

DISCUSSION

Interaction of ESC with histone H3.

Methylation of nucleosomal histone H3 by E(Z) complexes in vitro and in vivo requires ESC (7, 26, 43). Our results suggest that binding of ESC, via its N terminus, to histone H3, and to E(Z), via its β-propeller domain (35), constitutes the molecular basis of the stimulatory effect of ESC on E(Z) HMTase activity by linking E(Z) to its H3 substrate. The evolutionary conservation of the H3 binding activity of ESC, reflected in the binding of the conserved N termini of ESCL and mammalian EED to H3, underscores its importance for PRC2 function.

p55, the other histone-binding WD protein associated with ESC/E(Z) complexes, is required for stable chromatin binding but is not absolutely required for the HMTase activity of PRC2 complexes. However, it does appear to play a similar role in linking enzyme and substrate in other complexes that target H4. The two closely related mammalian p55 homologs, RpAp48 and RbAp46, appear to play such a substrate-targeting stimulatory role in the HAT1 histone acetyltransferase complex (a heterodimer of RbAp48 and HAT1) (28, 42) and in CBP complexes (45). In the latter case, the stimulatory effect of RbAp48 on CBP histone acetyltransferase activity was shown to be due to its lowering of the K_m of CBP for histones by three- to fourfold (45), as would be expected if it serves a substrate-binding/targeting function. RbAp48 has been shown to bind directly to the first α-helix of the histone fold of H4, a region that is not exposed when H4 is incorporated into nucleosomes (41). As expected from its high degree of similarity to RbAp48, p55 also binds to H4 and H4(1-85) in GST pull-down assays and does not bind to the N-terminal tail of H4 (residues 1 to 34) (Fig. 1A, lanes 3 to 5). Similarly, our data indicate that ESC binds to full-length H3 and H3(1-96) but does not interact with the N-terminal tail of H3, suggesting that, like RbAp48 and p55, ESC also binds to residues within the histone fold of H3. We have not further defined the region within the histone fold that binds ESC, so it remains possible that it also binds to a region of H3 that is not exposed in nucleosomal H3. If so, this raises an interesting question with respect to how the ESC present in PRC2 recognizes H3 in a native chromatin context. It is possible that a nucleosome remodeling event is required to enable ESC binding to H3 in vivo. However, the ability of recombinant PRC2 to methylate nucleosomal H3 in vitro raises the alternative possibility that the nucleosome binding activity of the SU(Z)12/p55 subcomplex of PRC2 may itself promote an alteration in local DNA histone contacts that facilitates ESC binding to H3.

ESC is not crucial for the in vitro chromatin binding activity of the PRC2 complex, which predominantly depends on its SU(Z)12 and p55 subunits. This, together with evidence presented here, suggests that H3 binding by ESC plays a different role, perhaps orienting/positioning E(Z) favorably for optimal interaction with H3 K27 during the docking of the complex onto chromatin. The Far Western data in Fig. 6A suggest that the interaction between ESC and histone H3 (lane 3) may be weak compared to the positive control (lane 6) (binding of the PC chromodomain to methylated H3), although this could be overstated due to altered conformation of H3 on the Far Western blot and the ability of PC to bind a short linear peptide encompassing H3 K27. In any case, weak binding of ESC to H3 is not incompatible with ESC functioning to coordinate the positioning of E(Z) and H3 while SU(Z)12 and p55 provide the predominant chromatin binding forces. It should also be kept in mind that the assays used here to isolate the interaction between ESC and H3 have limitations, and the situation in vivo is likely to be more complex. It was recently reported that mammalian PRC2 complexes, which contain two or three EED isoforms (15) together with RpAp48 and RpAp46 (21), exhibit stronger HMTase activity on dinucleosomes than on mononucleosomes (21). This suggests that in vivo a particular chromatin structure may be required for strong PRC2 binding and/or methylation, which may involve multiple interactions with several adjacent nucleosomes, perhaps by a dimer of PRC2 complexes. In addition, we have previously shown that the ESC N terminus mediates its own dimerization and phosphorylation (38). It would be interesting to determine whether or not histone H3 interacts preferentially with phosphorylated ESC.

An increasing number of WD proteins are being found to serve as histone binding subunits in chromatin-modifying complexes. Perhaps surprisingly, their WD motifs, often the only thing they have in common, are most often not responsible for their histone-binding activity. The interactions between histone H3 and ESC reported here and between histone H4 and RbAp48 (p55) reported by Vermaak et al. (41) are not mediated by their WD regions but by flanking sequences that have little in common. It is worth noting that the WD-repeat protein WDR5, a component of the MLL complex, also binds directly to histone H3 and is required for the methylation of H3 K4 by MLL (44). Like ESC, WDR5 has no selectivity for either unmethylated or methylated isoforms of H3 K4 (6, 33), as originally proposed (44). However, unlike ESC, WDR5 binds to the H3 N-terminal tail, and it does so via its β-propeller domain (WD repeats) (6, 33).

Chromatin and nucleosome binding of ESC and trimethylation of H3 K27.

Recent in vitro studies of Drosophila PRC2 HMTase activity on nucleosome substrates (26) indicate that PRC2 binding to nucleosomes requires cooperation between the SU(Z)12, p55, and ESC subunits. Although E(Z), the catalytic subunit, appears to make no contribution to nucleosome binding, it links ESC with SU(Z)12 and p55 within the complex (13, 26). While SU(Z)12 and p55 form the minimal nucleosome-binding complex, adding ESC to a E(Z)/SU(Z)12/p55 ternary complex enhances binding of the complex to nucleosomes (26). Moreover, at high concentrations, the ESC/E(Z)/SU(Z)12 subcomplex (i.e., lacking p55) is capable of binding nucleosomes (26). The binding of p55 and ESC to histones H4 and H3, respectively, could explain the effect of both proteins on the observed enhancement of nucleosome binding in vitro, provided that they are able to productively interact with the regions of H3 and H4 to which they bind. Since the region of H4 that binds RbAP48/p55 between residues is not exposed in the nucleosome (41) and the same is likely for the region of H3 that binds ESC, we therefore favor a model in which the binding of PRC2 changes the conformation of the nucleosome to allow full contact between ESC and H3 and p55 and H4, respectively. Of course, it is possible in vivo, where PRC2 function depends on local recruitment to PREs by DNA binding proteins and where a lower effective concentration of PRC2 complexes may bind productively to their nucleosome substrates, that a local nucleosome remodeling event, or activities of other endogenous proteins, may be a necessary accompaniment to productive PRC2 nucleosome binding. In any case, data presented here suggest that while ESC may measurably enhance nucleosome binding, its primary role is to link E(Z) to its K27 substrate on H3, perhaps positioning E(Z) favorably for optimal utilization of K27 as a substrate.

In our stable cell lines, induced expression of FLAG-ESC and FLAG-ESC61-425 resulted in the formation of FLAG-ESC/E(Z) and FLAG-ESC61-425/E(Z) complexes that are bound to the bxd PRE in ChIP assays. Similarly, induced expression of FLAG-PC, which binds strongly to 3meH3K27, also binds to the bxd PRE (Fig. 7B, top panel, lane 3), although it does not interact with E(Z). Our data indicate that truncation of the ESC N terminus does not abolish chromatin binding. Our current assays did not allow us to distinguish differences in the chromatin-binding affinities of complexes containing full-length FLAG-ESC and FLAG-ESC61-425 because their levels of expression differed and their relative solubilities and propensities to form complexes with E(Z) remain unknown.

The MNase treatment of isolated cell nuclei released many chromatin-associated proteins including ESC and E(Z), and mono-, di- and trinucleosomes. However, intact nucleosomes did not coimmunoprecipitate with FLAG-ESC. Given that induced FLAG-ESC forms complexes with E(Z) that bind to chromatin, there are several possible explanations for the failure of FLAG-ESC to pull down nucleosomes. First, the 2-h incubation of extracts with M2 agarose may be insufficient, or the washing buffer (with 0.15 M NaCl) may be too stringent for a stable FLAG-ESC-nucleosome association. However, increasing incubation time to overnight or decreasing the salt concentration in the wash buffer caused a significant rise in the nonspecific background. Second, the soluble nucleosomes released by MNase treatment may have altered structure and/or associated factors that preclude the FLAG-ESC-nucleosome association. Another possibility is that although the partial digestion of nuclei by MNase solubilized some nucleosomes and released some FLAG-ESC, chromatin regions enriched for FLAG-ESC-nucleosome interactions may for the most part remain insoluble. The high levels of FLAG-ESC and enrichment of 3meH3K27 in nuclear pellets after MNase treatment (data not shown) argues for this possibility.

The robust methylation of histone H3 K27 requires the 600-kDa E(Z)/ESC complex (PRC2). In our RNAi-treated cells, knockdown of ESCL and ESC caused a concomitant reduction of E(Z) (Fig. 8A, lane 4 to 6), while induced expression of FLAG-ESC increased the E(Z) level (Fig. 8A, lane 7 to 9). The 1me- and 2meH3K27 levels were barely changed in RNAi-treated cells (Fig. 8B, compare lanes 4 to 6 to lanes 1 to 3, respectively), probably due to incomplete depletion of ESC and ESCL. In contrast, the 3meH3K27 level dramatically decreased in RNAi-treated cells (Fig. 8B, lane 4 to 6), suggesting that K27 trimethylation is more sensitive to a partial reduction of PRC2 activity. The key evidence for the critical function of the N terminus of ESC comes from the induced expression of FLAG-ESC in RNAi-treated cells (Fig. 8, lane 7 to 9), with similar levels of E(Z) and SU(Z)12 in all three stable cell lines, FLAG-ESC-expressing cells restore 3meK27 to a higher level, but FLAG-ESC61-425-expressing cells exhibit further reduction of 3meK27 (Fig. 8B, lanes 7 and 8), presumably by a dominant-negative effect.

PRC2 is recruited to the PRE and methylates H3 K27 of neighboring nucleosomes, including those at the promoter. The trimethylated H3 K27 serves as an epigenetic mark for Polycomb silencing. Our data not only demonstrate that the N terminus of ESC binds directly to histone H3 but also provide evidence that the N terminus of ESC is required for the trimethylation of H3 K27 (Fig. 8B). Taken together, these and previous data strongly suggest that ESC stimulates the HMTase activity of E(Z) by linking E(Z) directly to its histone H3 substrate by its N terminus contacting histone H3 and its WD repeats interacting with the E(Z) N terminus.

Finally, we should add that while we have limited our conclusion to saying that the ESC N terminus is required for trimethylation of H3 K27, our results do not rule out the possibility that it is also required for mono- and dimethylation. Our RNAi experiment was designed to knock down endogenous ESC and ESCL without interfering with the expression of the stably transfected esc transgenes. For this reason, we were not able to use dsRNAs spanning the coding region of esc, which more efficiently deplete ESC (data not shown). In fact, more efficient knockdown of both ESC and ESCL substantially reduces both di- and trimethylation (data not shown).