Research article

Functional characterization of human Polycomb-like 3 isoforms identifies them as components of distinct EZH2 protein complexes

Gaylor Boulay, Claire Rosnoblet, Cateline Guérardel, Pierre-Olivier Angrand, Dominique Leprince

Abstract

PcG (Polycomb group) proteins are conserved transcriptional repressors essential to regulate cell fate and to maintain epigenetic cellular memory. They work in concert through two main families of chromatin-modifying complexes, PRC1 (Polycomb repressive complex 1) and PRC2–4. In Drosophila, PRC2 contains the H3K27 histone methyltransferase E(Z) whose trimethylation activity towards PcG target genes is stimulated by PCL (Polycomb-like). In the present study, we have examined hPCL3, one of its three human paralogues. Through alternative splicing, hPCL3 encodes a long isoform, hPCL3L, containing an N-terminal TUDOR domain and two PHDs (plant homeodomains) and a smaller isoform, hPCL3S, lacking the second PHD finger (PHD2). By quantitative reverse transcription–PCR analyses, we showed that both isoforms are widely co-expressed at high levels in medulloblastoma. By co-immunoprecipitation analyses, we demonstrated that both isoforms interact with EZH2 through their common TUDOR domain. However, the hPCL3L-specific PHD2 domain, which is better conserved than PHD1 in the PCL family, is also involved in this interaction and implicated in the self-association of hPCL3L. Finally, we have demonstrated that both hPCL3 isoforms are physically associated with EZH2, but in different complexes. Our results provide the first evidence that the two hPCL3 isoforms belong to different complexes and raise important questions about their relative functions, particularly in tumorigenesis.

  • chromatin
  • epigenetics
  • human Polycomb-like 3 (hPCL3)
  • Polycomb (PC)
  • Polycomb-like (PCL)
  • PHD finger protein 1 (PHF1)
  • PHF19
  • Polycomb repressive complex 2 (PRC2)

INTRODUCTION

PcG (Polycomb group) proteins, first identified by genetic studies in Drosophila, are required for maintaining the transcriptional silence of homoeotic genes as well as many other genes essential for genomic programming and differentiation [13]. In Drosophila, as well as in vertebrates, PcG proteins interact with each other to form multimeric chromatin-associated protein complexes. These complexes and their underlying molecular mechanisms are well conserved, although they are recruited to their target genes through different mechanisms in flies and mammals [46]. Two types of PcG multiprotein complexes with distinct biochemical properties have been extensively studied, PRC1 (Polycomb repressive complex 1) and PRC2 [1,7]. Their co-ordinated action is required for the silencing of target genes. The PRC2 complex is the ‘writer’ of the H3K27me3 (histone H3 trimethylated Lys27) repressive epigenetic mark. H3K27me3 residues are specifically bound by the chromodomain of PC (Polycomb), a component of the PRC1 complex [1,7]. The PRC2 catalytic subunit is the SET-domain containing protein E(Z), a specific H3K27 histone methyltransferase, but three other components, SU(Z)12 (Suppressor of zeste 12), ESC (Extra sex combs) and NURF55 (nucleosome remodelling factor 55-kDa subunits) are required for the stability and full activity of the PRC2 complex [8]. In Drosophila, in addition to the prominent 600-kDa PRC2 complex, a 1-MDa PRC2 complex containing the histone deacetylase RPD3 and PCL (PC-like) has been characterized [9]. Previously, elegant genetic and biochemical studies have highlighted the functional role of PCL as a key player in PcG-mediated transcriptional repression. They demonstrated that whereas PRC2 is involved in the genome-wide mono- and di-methylation of H3K27, PCL–PRC2 is required for the tri-methylation of H3K27, a signal which is confined to PcG target genes [10]. These results have been further extended to PHF1 [PHD (plant homeodomain) finger protein 1], one of the three human orthologues of Drosophila PCL [11,12].

In mammals, three PCL paralogues have been identified: PHF1, also called hPCL1 (human PCL1) [13] or Tctex3 in mice [14]; MTF2 (metal response element-binding transcription factor 2), also called hPCL2 [13,15,16]; and hPCL3, also called PHF19 [17]. These three genes are differentially expressed, suggesting that their expression pattern could provide other potential regulatory mechanisms to PcG target genes. Indeed, PHF1 and hPCL3 are widely expressed in different normal tissues with some examples of co-expression [13,16,17]. hPCL3 is also up-regulated in many cancers [17]. By contrast, microarray analyses in mice have demonstrated that Pcl2 is highly expressed in undifferentiated embryonic stem cells and during embryonic development as well as in some adult tissues [16]. PHF1, hPCL2 and hPCL3 are highly similar and display strong sequence similarities to Drosophila PCL. In particular, they share an N-terminal module consisting of three well-defined functional domains, namely a TUDOR domain and two adjacent PHD fingers immediately followed by a domain of extended homology (EH) with Drosophila PCL (Figure 1) [17]. Finally, their C-terminal ends are highly divergent except for a common C-terminal ‘chromo-like domain’ weakly homologous to the chromo domain found in many chromatin-associated proteins (Figure 1) [17].

Figure 1 PCL paralogues in humans

The sequences for hPCL3L (also called PHF19) (Q5T6S3; 580 amino acids), hPCL3S (Q5T6S3–2; 207 amino acids), PHF1 (O43189; 567 amino acids) and hPCL2 (Q9Y483; 593 amino acids) were all derived from Expasy/SwissProt. The conserved TUDOR, PHD1 and PHD2 domains are indicated [12,17,26,43]. For hPCL3L, the PCL EH domain and a C-terminal Chromo-like domain (C) are indicated. Two predicted NLSs are shown as black ovals [17].

hPCL3 presents unique features in the Polycomb-like gene family. Owing to different polyadenylation sites and alternative splicing mechanisms in its coding exon 5, this gene encodes two distinct isoforms: a long isoform (580 amino acids), called hPCL3L, which is similar to PHF1 and hPCL2/MTF2 and a shorter isoform (207 amino acids), called hPCL3S, which contains only the N-terminal TUDOR domain and PHD1, the first of the two PHD fingers (Figure 1) [17].

In the present study, we have investigated in detail the functional properties of the two hPCL3 isoforms. By quantitative RT (reverse transcription)–PCR analyses, we demonstrate that both hPCL3 isoforms are widely co-expressed, albeit at various levels, in different normal or transformed cells. By transient-transfection assays, we show that both the hPCL3L and hPCL3S isoforms can co-immunoprecipitate EZH2 and EED (Embryonic Ectoderm Development) and that the small isoform interacts less efficiently than the full-length hPCL3L protein. In addition, using an antibody specific for the full-length hPCL3L protein, we demonstrate that endogenous hPCL3L proteins interact with EZH2, the catalytic component of PRC2 and EED. Through the use of various deletion constructs, we show that the shared TUDOR domain and the PHD2 domain specific for the long isoform are both involved in the interaction with EZH2. Thus the presence of two EZH2-interacting domains (TUDOR and PHD2) in hPCL3L compared with only the TUDOR domain in hPCL3S might explain the stronger interaction observed with hPCL3L. In addition, we show that hPCL3L can self-associate in vivo through the PHD2 domain which, in contrast with PHD1, is phylogenetically well conserved in PCL3 proteins and more generally among all other PCL homologues. In close agreement with all of these results, hPCL3L and hPCL3S exhibit strikingly distinct subcellular localization upon immunofluorescence analyses and cell fractionation of transfected cells. Finally, gel-filtration analyses demonstrate that the two hPCL3 isoforms are found in two multiprotein complexes with different molecular masses; both are similar in molecular mass to other complexes also containing EZH2.

Taken together, our results strongly suggest that hPCL3L and hPCL3S proteins, which harbour different combinations of functional domains, could play divergent regulatory roles in PcG-mediated transcriptional repression.

EXPERIMENTAL

RNA isolation, RT–PCR and real-time PCR analyses

Total RNA was reverse transcribed using random primers and MultiScribe™ reverse transcriptase (Applied Biosystems). Real-time PCR analysis was performed by Power SYBR Green (Applied Biosystems) in a MX3005P fluorescence temperature cycler (Stratagene). Results were normalized with respect to 18S RNAs used as an internal control [18].

Cloning of full-length hPCL3L, hPCL3S and PHF1 and of hPCL3 domains

IMAGE clones containing full-length hPCL3S (BC022374) and human PHF1 (AA465397) were purchased from RZPD. Full-length hPCL3L (AL556762) was obtained from GENOSCOPE. Full-length open reading frames and relevant domains were amplified by PCR and subcloned into pcDNA3–FLAG or pcDNA3–HA expression vectors using standard procedures. All constructs were verified by sequencing.

Other mammalian expression vectors

Mammalian expression vectors for Myc–EZH2 [19] and EED [20,21] have been previously described. For analytical gel-filtration chromatography, expression vectors were generated by site-specific recombination using the Gateway® system (Invitrogen) of PCR-amplified open reading frames into TAP-tagged Moloney murine leukaemia virus-based vectors as previously described [22]. Full-length open reading frames of all cDNAs were PCR-amplified from cDNA clones for hPCL3S, hPCL3L or EZH2 (Mammalian Gene Collection cDNA clones: EZH2; IMAGE, 3901250) and cloned into the Gateway® entry vector pDONR201 (Invitrogen) and sequence verified. Entry clones were then recombined into suitable expression vectors by Gateway® LR reactions.

Antibodies

Commercial antibodies of the following specificity were used: anti-FLAG antibody from Sigma (M2 monoclonal antibody F3165); anti-Myc-tag antibody from Santa Cruz Biotechnology [rabbit polyclonal antibody (cat. no. sc-789) used at a 1:5000 dilution]; anti-HA-tag antibody from BAbCO (rabbit polyclonal antibody for immunoprecipitation and mouse monoclonal antibody for Western blotting used at a 1:1000 dilution); anti-lamin A/C antibody from Santa Cruz Biotechnology [H-110 rabbit polyclonal antibody (cat. no. sc-20681) used at a 1:1000 dilution]; and anti-GAPDH (glyceraldehyde-3-phosphate dehydrogenase) antibody from Santa Cruz Biotechnology [6C5 monoclonal rabbit polyclonal antibody (cat. no. sc-32223) used at a 1:500 dilution]. The Protein A moiety of the TAP tag was revealed with a PAP (peroxidase anti-peroxidase) soluble complex antibody produced in rabbit (cat. no. P1291; Sigma), used at a dilution of 1:10000.

The monoclonal antibodies against EED (AA19) and EZH2 (AC22) were generously given by Kristian Helin [BRIC (Biotech Research and Innovation Centre), University of Copenhagen, Copenhagen, Denmark] [20,21]. To generate a rabbit antibody against hPCL3L, the EcoRI–SmaI fragment encoding the specific C-terminus of the hPCL3L isoform (amino acids 382–580) was cloned in-frame with GST (glutathione transferase) in the pGEX2TK vector. The GST–hPCL3L fusion protein was expressed in the Escherichia coli strain BL21pLysS, purified on a GST column and used to immunize two New Zealand rabbits (Eurogentec). To generate a rabbit antibody against hPCL3S, the full-length hPCL3S isoform was expressed as a GST-fusion protein in BL21pLysS bacteria, purified on a GST column and used to immunize two New Zealand rabbits. However, these antibodies could only detect hPCL3S in transient-transfection assays, but not the endogenous hPCL3S proteins (results not shown).

Transfection and co-immunoprecipitation assays

HEK-293T cells (human embryonic kidney cells expressing the large T-antigen of simian virus 40) were maintained in DMEM (Dulbecco's modified Eagle's medium) medium (Gibco) supplemented with 10% (v/v) FBS (fetal bovine serum) and non-essential amino acids. Cells were transfected in OptiMEM (Gibco) by the PEI (polyethyleneimine) method using ExGen 500 (Euromedex), as previously described [23]. Cells were transfected for 6 h and then incubated in fresh complete medium. For co-immunoprecipitation assays, 48 h after transfection, cells were rinsed twice in ice-cold PBS and lysed in ice-cold IPH buffer [50 mM Tris/HCl, pH 8, 150 mM NaCl, 5 mM EDTA, 0.5% NP-40 (Nonidet P40) and 1× protease inhibitor cocktail (Roche)]. Cell lysates were cleared by centrifugation (20000 g, 4 °C, 30 min). The supernatants were incubated overnight at 4 °C with 2 μg of antibody. Then, protein A/G–Sepharose beads (Amersham Biosciences) were added for 30 min. The beads were washed three times with IPH buffer. Proteins were eluted by boiling in Laemmli loading buffer [0.12 M Tris/HCl, pH 6.8, 4% (w/v) SDS, 4% (v/v) 2-mercaptoethanol and 20% (v/v) glycerol] and separated by SDS/PAGE (8%, 12% or 15% gels) before Western blotting. Western blot analysis was performed as previously described [23]. The secondary antibodies were horseradish peroxidase-linked antibodies raised against rabbit, goat or mouse immunoglobulins (Amersham).

Similarly, the erythromyeloblastoid leukaemia cell line K562, which expresses high amounts of hPCL3 mRNAs [17], was used to perform co-immunoprecipitation analyses of endogenous proteins.

Co-immunoprecipitation of chromatin-associated proteins

Chromatin-associated proteins were extracted as previously described with minor modifications [24]. Transiently transfected HEK-293T cells were harvested and lysed for 45 min on a rotator at 4 °C in buffer A [50 mM Tris/HCl, pH 7.5, 1 mM DTT (dithiothreitol), 20 mM NaF, 0.5% Triton X-100 supplemented with 1× protease inhibitor cocktail containing no EDTA (Roche)]. After centrifugation (1800 g, 4 °C, 10 min), pellets were washed twice with buffer A, resuspended in buffer B (50 mM Tris/HCl, pH 8.0, 1.5 mM CaCl2 and 20 mM NaF) and finally treated with 30 units of micrococcal nuclease (Fermentas) for 35 min at 37 °C under mild agitation. Solubilized proteins were recovered by two rounds of centrifugation (5000 g, 4 °C, 2 min). Before immunoprecipitation, the chromatin fraction was adjusted for a final concentration of 150 mM NaCl and 0.5% Triton X-100. FLAG-tagged hPCL3S proteins were immunoprecipitated with anti-FLAG antibodies coupled to Sepharose (Sigma) for 2 h on a rotator at 4 °C. Resins were washed four times with TBS (Tris-buffered saline) containing 0.1% Triton X-100 and 0.25% NP-40. Finally, purified proteins were eluted in Laemmli buffer and subjected to immunoblotting.

Immunofluorescence analyses

U20S (human osteosarcoma) cells (European Collection of Animal Cell Cultures) were maintained in DMEM medium (Gibco) supplemented with 10% (v/v) FBS and non-essential amino acids. Cells were cultured on coverslips, transfected by the PEI method as described above and processed at room temperature (20 °C) as previously described [25]. Briefly, cells were washed, fixed for 20 min in cold 3% (v/v) PFA (paraformaldehyde), permeabilized in 0.1% Triton X-100 for 5 min, saturated for 30 min in 300 μl of PBS containing 10% (v/v) goat serum, incubated for 30 min with the anti-FLAG antibody diluted 1:500 in PBS containing 10% (v/v) goat serum, and incubated in the dark for 30 min with the secondary antibody diluted 1:200 in PBS containing 10% (v/v) goat serum. Between each stage, the cells were washed three times for 5 min in PBS. Nuclei were stained with 10 μg/ml Hoechst 33342 for 1 min. The cells were then placed inverted on a drop of Immuno Floure Mounting Medium (ICN) on a slide. The slides were stored in the dark at 4 °C and visualized under fluorescence.

Cell-fractionation experiments

HEK-293T cells were transiently transfected with the pcDNA3–FLAG–hPCL3S and hPCL3L expression vectors. Cytosolic and nuclear fractions were obtained using the Nuclear Extraction kit (Millipore) according to manufacturer's instructions. For Western blot analyses, 50 μg of protein from each fraction were loaded and analysed with anti-FLAG antibodies. GAPDH and lamin A/C were used as cytoplasmic and nuclear markers.

Analytical gel-filtration chromatography

Cells were lysed in low-salt buffer [10 mM Tris/HCl, pH 7.4, 25 mM NaCl, 2 mM mgOAc, 1 mM DTT, 1mM EDTA, 0.05% NP-40 and complete protease inhibitors (Roche)] for 15 min at 4 °C. Nuclei were pelleted by centrifugation (800 g, 4 °C, 10 min) and lysed in E1A buffer [50 mM Hepes, pH 7.9, 1 mM EGTA, 250 mM NaCl, 1 mM DTT, 1 mM EDTA and complete protease inhibitors (Roche)] by gently shaking for 2 h at 4 °C. Nuclear extracts were recovered by centrifugation (2500 g, 4 °C, 10 min) and clarified by two centrifugations at 10000 g for 1 min. Fresh nuclear extract (1 mg) was fractionated on a Superose 6 HR 10/300 GL column (GE Healthcare) equilibrated in 50 mM Tris/HCl, pH 7.4, and 100 mM KCl. Size-exclusion chromatography was performed on an FPLC system and an ÄKTA™ purifier (GE Healthcare). Elution profiles of Blue Dextran (V0), thyroglobulin (669 kDa), apoferritin (443 kDa), β-amylase (200 kDa), alcohol dehydrogenase (150 kDa), BSA (66 kDa) and carbonic anhydrase (29 kDa) (all from Sigma) were used for calibration. All experiments were run at room temperature. Fractions of 0.5 ml were collected and 25-μl aliquots from even-numbered samples were electrophoresed on 4–12% Bis-Tris gels (Invitrogen) and transferred to nitrocellulose membranes (Schleicher & Schuell) for Western blotting.

RESULTS

The two hPCL3 isoforms are expressed at comparable levels in various transformed and normal cell types

To accurately measure the expression levels of each hPCL3 isoform, we performed quantitative RT–PCR analyses on total RNAs prepared from various transformed cell lines or from normal human cells including WI38 fibroblasts, hTERT-HME-1 and MCF10A mammary epithelial cells and HUVECs (human umbilical vein endothelial cells). We used sets of primers able to specifically amplify hPCL3L and hPCL3S isoforms as well as full-length PHF1 (Table 1).

View this table:
Table 1 Primers used for quantitative PCR amplification

for, forward; rev, reverse.

As shown in Figure 2, hPCL3 and PHF1 are differentially expressed in most cell types analysed. The highest expression levels of both hPCL3 isoforms was observed in the erythromyeloblastoid leukaemia cell line K562, as previously shown by RNA dot-blot experiments [17], and in the medulloblastoma cell lines D283 and DAOY. In D283 cells, the variant encoding the small isoform, hPCL3S, is expressed at a higher level than the transcript encoding the full-length protein, hPCL3L. hPCL3 is more highly expressed than PHF1 in normal endothelial cells (HUVECs) and in mammary epithelial cells (MCF10A and hTERT-HME1). However, the converse situation is observed in normal WI38 fibroblasts, which expressed conspicuous levels of PHF1 compared with a much weaker expression of hPCL3.

Figure 2 Expression of the hPCL3 and PHF1 transcript variants in various cell lines

Quantitative expression of PHF1 and of the two hPCL3 isoforms in various normal or transformed cells. Transcript levels were measured by quantitative RT–PCR analyses. Results are normalized to 18S mRNA and are presented as means±S.D. from two independent experiments in triplicate.

Although different total levels of hPCL3 are observed, both hPCL3 isoforms are expressed at similar levels in all cell types examined (Figure 2). By contrast, for PHF1, semi-quantitative RT–PCR analyses demonstrated that the PHF1 smaller isoform (O43189–2; 457 amino acids) is virtually undetectable in most cell types tested except in the colon carcinoma cell line SW-480, where it is weakly expressed compared with full-length PHF1 (results not shown).

Thus both hPCL3 isoforms are expressed at comparable levels in various cell types despite major differences in their coding sequences and probably their biochemical properties.

Both hPCL3 isoforms interact with two components of PRC2, EZH2 and EED

Drosophila PCL, human PHF1 and human PCL2 interact with PRC2 [1012,16,26]. To demonstrate that hPCL3 also interacts with PRC2 components, we performed co-immunoprecipitation experiments in HEK-293T cells using expression vectors encoding each hPCL3 isoform and the enzymatic component of PRC2, the specific H3K27 histone methyltransferase EZH2. Myc–EZH2 was co-precipitated by FLAG–hPCL3L, demonstrating that full-length hPCL3 and EZH2 proteins associate in vivo (Figure 3A, right-hand panel, lane 6). Furthermore, Myc–EZH2 was also co-precipitated by FLAG–hPCL3S, demonstrating that this smaller isoform is also able to interact with EZH2, albeit less efficiently than the full-length hPCL3L protein (Figure 3A, right-hand panel, compare lanes 5 and 6).

Figure 3 The two hPCL3 isoforms interact with the PRC2 components EZH2 and EED

(A) Association of hPCL3L and hPCL3S with EZH2. HEK-293T cells were transiently transfected for 48 h with the indicated expression plasmids as well as with the empty pcDNA3–FLAG vector as a control. Whole-cell extracts were prepared in IPH buffer and incubated with anti-FLAG antibodies [right-hand panel: immunoprecipitated (IP) FLAG (hPCL3), lanes 16] and immunoblotted (WB) with polyclonal antibodies against Myc to detect EZH2 co-immunoprecipitation. The membrane was then stripped and re-probed with anti-FLAG monoclonal antibodies to ascertain the presence of hPCL3L or hPCL3S. A portion (2%) of each lysate was directly resolved by SDS/PAGE and immunoblotted with the indicated antibodies (left-hand panel: input, lanes 1–6). (B) Association of hPCL3L and hPCL3S with EED3. A similar experiment was performed with an expression vector for EED3. To detect EED, we used the AA19 monoclonal antibody (generously given by K. Helin). The arrowhead indicates the heavy chains of the anti-FLAG monoclonal antibody used in each lane to immunoprecipitate the FLAG-tagged constructs. The presence of a faint amount of EED in lane 4 presumably corresponds to non-specific binding of EED to the beads during the co-immunoprecipitation step. (C) In vivo, endogenous hPCL3L interacts with EZH2. K562 cells were lysed with IPH buffer and 2% of the lysates were kept as input (left-hand panel). Equal amounts of lysates were then immunoprecipitated with normal rabbit immunoglobulins (IgG) or with the rabbit anti-hPCL3L antibodies (hPCL3L). The immunoprecipitates were analysed by immunoblotting with the mouse anti-EZH2 monoclonal antibody (upper panel). The asterisk refers to a non-specific band. The membrane was then incubated with the anti-hPCL3 polyclonal antibodies as a control (lower panel). (D) In vivo, endogenous hPCL3L interacts with another component of the PCR2 complex EED. A similar experiment was conducted in K562 cells, except that the immunonoprecipitates were first analysed by immunoblotting with the mouse anti-EED monoclonal antibody. The EED isoforms detected in the input or co-immunoprecipitated with hPCL3L are numbered as previously described by Kuzmichev et al. [35]. The arrowhead corresponds to the heavy chains of the rabbit immunoglobulins.

A similar experiment was conducted with an expression vector for EED (isoform 3) [20]. As shown for EZH2, both hPCL3 isoforms interacted with the core PRC2 component EED. Again the interaction appeared to be stronger with the full-length hPCL3L protein (Figure 3B, right-hand panel, compare lanes 5 and 6).

To confirm these interactions between endogenous proteins, we generated rabbit polyclonal antibodies directed against hPCL3L, but were unable to obtain antibodies specific for the hPCL3S isoform (see the Experimental section; results not shown). In total cell extracts from K562 erythromyeloblastoid cells expressing high amounts of hPCL3 (Figure 2) [17], EZH2 as well as EED isoforms 1 and 4, were immunoprecipitated by these anti-hPCL3L antibodies, but not by normal rabbit immunoglobulins (Figures 3C and 3D).

Taken together, our results demonstrate that both hPCL3 isoforms interact with EZH2 and EED, two components of the PRC2 complex. In addition, the weaker interaction observed in both cases with hPCL3S strongly suggested that domain(s) not present in the smaller isoform may be involved in the interaction with PRC2 components.

The TUDOR and PHD2 domains, but not the PHD1 domain of hPCL3, are involved in the interaction with EZH2

hPCL3L, like PHF1 and PCL2, contains an N-terminal TUDOR domain followed by two zinc-finger-like PHD domains, whereas hPCL3S contains a TUDOR domain and only the first PHD1 finger (Figure 1). To identify the hPCL3 domains implicated in the interaction with EZH2, we constructed several deletion mutants. The TUDOR and PHD1 domains were amplified by PCR from the N-terminal and C-terminal portions of hPCL3S and cloned into a pcDNA3–FLAG vector. In addition, the TUDOR domain was deleted in the context of the full-length hPCL3L protein (Figure 4A). These constructs were transiently expressed in the absence or in presence of Myc-tagged EZH2 and co-immunoprecipitation experiments were performed. As shown in Figure 4(B), the TUDOR domain interacts with EZH2, whereas the C-terminal moiety of hPCL3S containing the PHD1 domain is unable to co-precipitate Myc–EZH2 (third panel, lanes 7 and 8). Deletion of the TUDOR domain in the hPCL3L isoform did not abolish the interaction with Myc–EZH2, demonstrating that the TUDOR domain is not the sole domain implicated in this interaction (Figure 4B, lane 10). Indeed, the isolated PHD2 domain that is specific to the hPCL3L isoform is also able to interact with EZH2 (Figure 4B, lane 9). Interestingly, CLUSTAL alignments of hPCL3 proteins from various species demonstrate that their TUDOR and PHD2 domains are phylogenetically well conserved, whereas the PHD1 domain is not (results not shown).

Figure 4 The TUDOR and PHD2 domains of hPCL3L, but not its PHD1 domain, interact with EZH2

(A) Schematic representation of the hPCL3L and hPCL3S isoforms and deletion mutants used. (B) Association of EZH2 with hPCL3 is dependent on the TUDOR and PHD2 domains. HEK-293T cells were transiently transfected with the indicated expression plasmids as well as with the empty pcDNA3–FLAG vector as a control for 48 h. Whole-cell extracts were prepared in IPH buffer and incubated with anti-FLAG antibodies (upper three panels: IP FLAG (hPCL3), lanes 1–10) and immunoblotted (WB) with anti-FLAG monoclonal antibodies to verify the presence of hPCL3L or its deletion mutants in the immunoprecipitates and with antibodies against the Myc epitope to detect EZH2 co-immunoprecipitation. A portion (2%) of each lysate was directly resolved by SDS/PAGE and immunoblotted with antibodies against Myc to ascertain the presence of EZH2 (bottom panel: input, lanes 1–10). In the top panel, the arrowhead indicates the heavy chains of the anti-FLAG monoclonal antibody used in each lane to immunoprecipitate the FLAG-tagged constructs.

The presence of two EZH2-interacting domains (TUDOR and PHD2) in hPCL3L compared with only one TUDOR domain in hPCL3S might thus explain the stronger interaction observed with hPCL3L (Figures 3A and 3B, and Figure 4B). Therefore, the presence of different functional domains in the two hPCL3 isoforms is well correlated with their different biochemical properties, particularly in terms of their interaction with core PRC2 components

The hPCL3L isoform is able to self-associate through the conserved PHD2 domain

The N-terminus of ESC, the Drosophila orthologue of EED, mediates its dimerization [27]. To investigate whether hPCL3 can also self-associate, we generated HA-tagged hPCL3L and hPCL3S constructs for use in co-immunoprecipitation assays. In transiently transfected HEK-293T cells, HA–hPCL3L was co-precipitated by FLAG–hPCL3L, demonstrating that the full-length hPCL3L proteins can self-associate in vivo (Figure 5A, upper panel, lane 8). Strikingly, no significant interaction was detected between the HA-tagged small isoform (HA–hPCL3S) and the FLAG-tagged full-length protein (hPCL3L) (Figure 5A, upper panel, lane 7). Thus, despite their co-expression in many cell types (Figure 2) [17], the two hPCL3 isofoms are unlikely to stably hetero-oligomerize in vivo.

Figure 5 The PHD2 domain of hPCL3L is implicated in its self-association

(A) hPCL3L can self-associate, but cannot heterodimerize with hPCL3S. HEK-293T cells were transiently transfected with the indicated combination of FLAG- and HA-tagged versions of hPCL3L or hPCL3S isoforms as well as with the empty pcDNA3–FLAG vector as a control. Whole-cell extracts were prepared in IPH buffer and incubated with anti-HA antibodies (IP: HA, lanes 5–8) and immunoblotted (WB) with anti-FLAG antibodies to detect co-immunoprecipitation of hPCL3L (upper right-hand panel). The blot was then stripped and probed with an anti-HA antibody to verify the presence of hPCL3L and hPCL3S in the relevant immunoprecipitates (lower right-hand panel). A portion (2%) of each lysate was directly resolved by SDS/PAGE and successively immunoblotted with anti-FLAG and anti-HA antibodies (left-hand panels: input, lanes 1–4). (B) The PHD2 domain of hPCL3L is implicated in its self-association. A similar experiment was performed with the HA-tagged hPCL3L isoform and the FLAG-tagged isolated domain as well as the empty pcDNA3–FLAG vector as a control. The asterisk refers to a non-specific band detected in the input by the anti-HA antibodies.

To define the hPCL3L functional domain(s) responsible for its self-association, we performed co-immunoprecipitation experiments using HA-tagged hPCL3L and the FLAG-tagged isolated TUDOR, PHD1 or PHD2 domains (Figure 4). The results of these experiments clearly demonstrated that only the PHD2 finger is able to mediate stable interaction with hPCL3L in vivo (Figure 5B, lower panel, lane 8).

Thus our results clearly establish different functional roles for the two PHD fingers of hPCL3L and highlight the importance of the well-conserved PHD2 domain.

The two hPCL3 isoforms display different subcellular localizations

In transient-expression assays in HEK-293T cells, the two hPCL3 isoforms exhibited strikingly different properties as exemplified by their interactions with PRC2 components (Figures 3 and 4) and the self-association of hPCL3L mediated by its unique PHD2 finger (Figure 5). To gain insights into the properties of these hPCL3 proteins, we decided to analyse their subcellular localization. To this end, FLAG-tagged hPCL3L and hPCL3S expression vectors were transiently transfected into U2OS cells. By immunofluorescence analyses, the two hPCL3 isoforms exhibited distinct subcellular localization. Indeed, FLAG–hPCL3L proteins were shown to be exclusively nuclear (Figure 6A, upper panels). By contrast, FLAG–hPCL3S proteins were detected in the cytoplasm as well as in the nucleus (Figure 6A, lower panels). To confirm these results using a different technique, we performed cell-fractionation experiments of transiently transfected HEK-293T cells. The cytosolic and nuclear fractions were prepared and aliquots were analysed by Western blotting with the monoclonal anti-FLAG antibodies or with anti-GAPDH and anti-lamin antibodies as controls for cytosolic and nuclear proteins respectively. The results clearly demonstrated that hPCL3L is almost exclusively found in the nucleus with a faint presence in the cytosolic fraction (Figure 6B, upper panel, lanes 3 and 6). By contrast, hPCL3S is found in both fractions with a majority in the cytoplasm (Figure 6B, upper panel, lanes 2 and 5)

Figure 6 Distinct subnuclear localization of FLAG-tagged hPCL3L and hPCL3S proteins

(A) Immunofluorescence analyses. U2OS cells transiently transfected with expression vectors for the FLAG-tagged hPCL3L and hPCL3S isoforms, used in the co-immunoprecipitation experiments, were fixed with paraformaldehyde and analysed by conventional immunofluorescence microscopy using the anti-FLAG monoclonal antibody. Nuclei are seen as Hoechst-positive staining. The merging of the two images is shown in the right-hand panels. (B) Cell-fractionation experiments. HEK-293T cells transiently transfected with expression vectors for the FLAG-tagged hPCL3L and hPCL3S isoforms or with the empty pcDNA3–FLAG vector as a control were lysed. Cytoplasmic and nuclear fractions prepared with the Nuclear Extraction kit (Millipore) were resolved by SDS/PAGE and immunoblotted (WB) with anti-FLAG antibodies (upper panel) to detect the hPCL3 isoforms. Anti-GAPDH and anti-lamin A/C antibodies were used as controls respectively for detection of cytoplasmic and nuclear proteins. WCE, whole cell extract.

Thus the hPCL3L and hPCL3S proteins have overlapping but distinct subcellular localization patterns. Together with their distinct biochemical properties, our results taken as a whole strongly suggest that these two proteins probably participate in different complexes with specific properties.

The two hPCL3 isoforms are present in two different EZH2-containing complexes

As a first step towards the identification of complexes containing hPCL3L and hPCL3S, we fractionated nuclear extracts of HEK-293T cells transiently expressing TAP-tagged versions of hPCL3L or hPCL3S by gel-filtration chromatography on a Superose 6 HR 10/300 GL column to separate protein complexes of different sizes. The separated fractions were analysed by Western blot analyses with a PAP soluble complex antibody that recognizes the Protein A moiety of the TAP tag in order to identify hPCL3 isoform-containing fractions. The results of these fractionation experiments shown in Figure 7(A) demonstrate that the two hPCL3 isoforms are found in two complexes of different molecular masses. hPCL3L is found in a high-molecular-mass complex with an estimated size of roughly 2000 kDa, which could be the PCL–PRC2 complex identified in previous studies [12]. By contrast, hPCL3S is also found in a multiprotein complex, albeit of much lower molecular mass. Size-exclusion chromatography experiments performed with nuclear extracts derived from TAP-tagged EZH2 expressing cells revealed that EZH2 elutes in the same molecular-mass fractions as hPCL3L and hPCL3S. The same results were obtained when these fractionation experiments were performed in HeLa or U2OS cells (results not shown). Although these experiments might not be viewed as definitive, we have demonstrated that endogenous EZH2 and EED can be co-immunoprecipitated with hPCL3L (Figures 3C and 3D). Unfortunately, owing to the unavailability of highly sensitive anti-hPCL3S antibodies, the co-immunoprecipitation of endogenous hPCL3S proteins with EZH2 and EED could not be addressed. Nevertheless, in transient-transfection assays, we have shown that hPCL3S interacts with EZH2 and EED, albeit less efficiently than hPCL3L (Figures 3C and 3D).

Figure 7 The two hPCL3 isoforms are present in two different complexes

(A) Gel-filtration analyses. Nuclear extracts from HEK-293T cells transiently expressing TAP-tagged hPCL3S, hPCL3L or EZH2 as indicated were subjected to Superose 6 gel filtration. Fractions were analysed by Western blotting using a PAP soluble complex antibody to reveal the Protein A moiety of the TAP tag. (B) Co-immunoprecipitation analyses of chromatin-associated proteins. HEK-293T cells transiently transfected with the expression vector for FLAG-tagged hPCL3S or with the empty pcDNA3–FLAG vector as a control were lysed. The chromatin-associated fractions were incubated with anti-FLAG antibodies and immunoblotted (WB) with anti-EZH2 antibodies to detect co-immunoprecipitation with exogenous FLAG–hPCL3S (upper right-hand panel: IP FLAG, lanes 3 and 4). The blot was then stripped and probed with anti-FLAG to verify the presence of hPCL3S in the relevant immunoprecipitates (lower right-hand panel: IP FLAG, lanes 3 and 4). A portion (2%) of each nuclear fraction was directly resolved by SDS/PAGE and immunoblotted with the same antibodies (input: lanes 1 and 2). A shorter exposure of the input lanes with the EZH2 antibodies is shown on the left to see the lower non-specific band (shown by an asterisk) as well as the upper EZH2 band.

To build upon these results, we next analysed the co-immunoprecipitation of endogenous EZH2 proteins after transient transfection of pcDNA3–FLAG–hPCL3S, an experimental condition mimicking our TAP-tag assay. However, since most of the overexpressed FLAG–hPCL3S protein is found in the cytoplasm (Figure 6), we used a recently published fractionation protocol giving access to the chromatin-associated protein fraction [24]. As shown in Figure 7(B), chromatin-associated FLAG–hPCL3S proteins can efficiently co-immunoprecipitate endogenous chromatin-bound EZH2 proteins.

Therefore, these gel-fractionation experiments considered together with all of the co-immunoprecipitation assays (Figures 3, 4, 5 and 7B) raise the possibility that EZH2 could be part of a least two different protein complexes, one containing the short isoform of hPLC3, and the other containing the long isoform.

DISCUSSION

In the present report, we studied hPCL3, one of the three human paralogues of the Drosophila PCL gene. Owing to alternative splicing mechanisms in its coding exon 5, hPCL3 encodes two different isoforms, hPCL3L and hPCL3S, containing various combinations of functional domains (Figure 1). The PHD finger originally characterized as a protein–protein-interaction domain is involved in many biochemical functions, notably methyl–lysine binding [28,29]. The TUDOR domain initially characterized in RNA-binding proteins is now also considered to be a domain interacting with methylated proteins, including histones [29,30] We have demonstrated that hPCL3L interacts with EZH2, as shown for PHF1 and PCL2 in previous studies [12,16], through its TUDOR and PHD2 finger domains, but not its PHD1 finger.

Co-immunoprecipitation analyses of Drosophila embryonic extracts demonstrated that E(Z) interacts with PCL in vivo, but the domains involved are strikingly different [26]. Indeed, both PHD fingers appeared to be important for this interaction, as shown by directed two-hybrid experiments and by in vitro point mutagenesis, which have highlighted conserved residues in the PHD2 Loop1 [26,31,32]. The NMR structure of the TUDOR domain from Drosophila PCL demonstrates that its TUDOR domain contains an atypical incomplete aromatic cage, unlikely to interact with methylated arginine or lysine residues [33]. By contrast, the three human PCL orthologues exhibit a complete aromatic cage, suggesting that they might bind methylated lysines [33]. Thus, even though the interaction between PCL and E(Z) that is essential for PcG-mediated transcriptional repression is conserved in their human homologues, the precise molecular mechanisms involved might be slightly different.

We have also shown that hPCL3L is able to self-associate in vivo. ESC and its human homologue EED can also dimerize [27], but to the best of our knowledge this property has never been ascribed to any PCL protein. The interaction of hPCL3L with EZH2 as well as its self-association in vivo both rely on the PHD2 finger. Given the strong conservation of the PHD2 domains between hPCL3 proteins (results not shown) as well as between the various PCL paralogues, it is tempting to speculate that self-association might be a general property shared by all PCL family members. Concerning the hPCL3 PHD1 domain, which is not evolutionarily well conserved by contrast with the PHD2 finger (results not shown), its function still remains elusive, but its role in the interaction between PCL and E(Z) in Drosophila [26] seems to have been lost in mammals.

During evolution, the repertoire of PcG proteins has been greatly extended and diversified to accommodate the increasing complexity of heritable gene-silencing mechanisms in multicellular organisms [3,7,34]. For each PcG protein in Drosophila, several paralogues with specialized functions and different expression patterns are present in mammals. This holds true both for PRC1 and for different subsets of PRCs (PRC2, PRC3 and PRC4), which exhibit differential histone substrate specificities [35] and fulfil non-redundant and complementary roles [36,37]. Similarly, the single PCL gene in Drosophila has three paralogues in humans, with distinct patterns of expression [11,12,16,17] (Figure 2). Their coding sequence harbours an N-terminal functional module consisting of a TUDOR domain, a PHD1 finger and a PHD2 finger followed by an extended domain of homology with PCL. The strong conservation of these domains, with the notable exception of PHD1, suggests that they could fulfil similar functions at least in PcG-mediated transcriptional repression. Their C-terminal ends are, in contrast, highly divergent except for a C-terminal Chromo-like domain. It would be interesting to determine whether these specific C-terminal ends can be engaged in different interactions with other PcG components, thus specifying subsets of complexes with other chromatin components or even with other nuclear proteins. For example, PHF1, but neither PCL2 nor PCL3, is involved in the response to DNA DSBs (double-strand breaks) and their repair [38]. Their recruitment to DSBs is Ku70/Ku80-dependent and implies a requirement for both the TUDOR domain and a small fragment of the specific C-terminal end (amino acids 340–431), overlapping the NLSs (nuclear localization signals) [38]. Thus this expansion and diversification of the PcG gene together with distinct expression patterns should permit the formation of many different complexes able to exquisitely regulate the myriad of PcG-target genes in all tissues of a multicellular organism and perhaps also perform other functions, as recently demonstrated for PHF1 in the DNA damage response [38].

Most of the proteins or complexes involved in chromatin transactions contain several so-called effector domains able to engage multiple interactions with specific histone modifications or binding partners [39,40]. Therefore the hPCL3 isoforms might have significantly different functions in view of the partially functional module in the small isoform, hPCL3S. As a first step to define the specific functions of each hPCL3 isoform, we demonstrated that they exhibit different subcellular localizations and belong to different high-molecular-mass complexes, containing at least EZH2 (Figures 6 and 7). To gain more insight into the precise physiological functions of these complexes, a future challenge would be to define by proteomic analyses which proteins besides EZH2 and presumably other members of PRC2 complexes are associated with hPCL3L and, more importantly, with hPCL3S. In addition, ChIP-seq (chromatin immunoprecipitation followed by sequencing) [41] would define the subsets of PRC2 targets bound respectively by hPCL3L and/or hPCL3S and would allow their comparison with the target genes bound by PHF1 and hPCL2 [42]. All of these studies would help to better define the functional roles of both hPCL3 isoforms, most notably during tumorigenesis.

AUTHOR CONTRIBUTION

Gaylor Boulay and Claire Rosnoblet performed the experiments and analysed the results. Cateline Guérardel contributed several new reagents. Pierre-Olivier Angrand and Dominique Leprince conceived the project and analysed the results. Dominique Leprince wrote the paper with assistance from all authors.

FUNDING

This work was supported by the Centre National de la Recherche Scientifique (CNRS), the Pasteur Institute, the Ligue Nationale contre le Cancer (Comité Interrégional du Septentrion), the Fondation pour la Recherche Médicale (Comité du Nord) and the Association pour la Recherche contre le Cancer (ARC) [grant numbers ARC 3983 and ARC 1081] (to D.L.). G.B. was supported by fellowships from the CNRS/Région Nord-Pas de Calais and from the ARC. The work in P.-O.A.'s laboratory was supported by the CNRS, l'Université Lille Nord de France, and by the Ministère de la Recherche et de l'Enseignement Supérieur, the Région Nord-Pas de Calais and the European Regional Developmental Funds [through the ‘Contrat de Projets Etat-Région (CPER) 2007–2013’].

Acknowledgments

We thank Dr Stéphane Ansieau (Center Léon Bérard, Lyon, France), Dr Thomas Jenuwein (Max Planck Institute of Immunobiology, Freiberg, Germany), Professor Achim Leutz (Max Delbrück Center for Molecular Medicine, Berlin, Germany) and Dr Kristian Helin for monoclonal antibodies and expression vectors. We thank Dr Etienne Lelievre (CNRS UMR 8161, Institut de Biologie de Lille, Lille, France) for the HUVEC cells. We are indebted to Dr Brian Rood for critical reading of the manuscript prior to submission.

Abbreviations: DMEM, Dulbecco's modified Eagle's medium; DSB, double-strand break; DTT, dithiothreitol; EED, Embryonic Ectoderm Development; EH, extended homology; ESC, Extra sex combs; FBS, fetal bovine serum; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GST, glutathione transferase; H3K27me3, histone H3 trimethylated Lys27; HEK-293T, cells, human embryonic kidney cells expressing the large T-antigen of simian virus 40; HUVEC, human umbilical vein endothelial cell; MTF2, metal response element-binding transcription factor 2; NLS, nuclear localization signal; NP-40, Nonidet P40; PAP, peroxidase anti-peroxidase; PcG, Polycomb group; PC, Polycomb; PCL, PC-like; hPCL, human PCL; hPCL3L, long isoform of hPCL3; hPCL3S, short isoform of hPCL3; PEI, polyethyleneimine; PHD, plant homeodomain; PHF1, PHD finger protein 1; PRC, Polycomb repressive complex; RT, reverse transcription

References

View Abstract