|Home | About | Journals | Submit | Contact Us | Français|
The mammalian Polycomblike protein PHF1 was previously shown to interact with the Polycomb group (PcG) protein Ezh2, a histone methyltransferase whose activity is pivotal in sustaining gene repression during development and in adulthood. As Ezh2 is active only when part of the Polycomb Repressive Complexes (PRC2-PRC4), we examined the functional role of its interaction with PHF1. Chromatin immunoprecipitation experiments revealed that PHF1 resides along with Ezh2 at Ezh2-regulated genes such as the HoxA loci and the non-Hox MYT1 and WNT1 genes. Knockdown of PHF1 or of Ezh2 led to up-regulated HoxA gene expression. Interestingly, depletion of PHF1 did correlate with reduced occupancy of Bmi-1, a PRC1 component. As expected, knockdown of Ezh2 led to reduced levels of its catalytic products H3K27me2/H3K27me3. However, reduced levels of PHF1 also led to decreased global levels of H3K27me3. Notably, the levels of H3K27me3 decreased while those of H3K27me2 increased at the up-regulated HoxA loci tested. Consistent with this, the addition of PHF1 specifically stimulated the ability of Ezh2 to catalyze H3K27me3 but not H3K27me1/H3K27me2 in vitro. We conclude that PHF1 modulates the activity of Ezh2 in favor of the repressive H3K27me3 mark. Thus, we propose that PHF1 is a determinant in PcG-mediated gene repression.
Polycomb group (PcG) proteins are key regulators of homeotic genes in multicellular organisms. In addition to homeotic genes, PcG proteins regulate several other families of genes such as those controlling transcription, organogenesis, and neural development (3, 4, 22, 39, 43) by maintaining the repressed or “OFF” state of genes that were inactivated early in development. Another set of proteins, the trithorax group (trxG), antagonizes the effect of the PcG proteins by maintaining genes in the activated or “ON” state. Classical genetic studies have shown that in the absence of either the PcG or TrX proteins, the homeotic genes (Hox) are deregulated, resulting in mild to severe developmental defects (37, 40).
PcG proteins have been reported to exist in two distinct types of Polycomb repressive complexes: Polycomb (Pc)-containing complexes (PRC1 and PRC1-like) and the Enhancer of Zeste [E(z)]-containing complexes (PRC2/PRC3/PRC4). The Drosophila E(z) protein has two known human homologues, Ezh1 and Ezh2 (21). Ezh2 is a histone lysine methyltransferase (HKMT) that catalyzes the methylation of nucleosomal histone H3 at lysine 27 (7, 11, 19, 28). In addition to Ezh2, the core of the PRC2/PRC3/PRC4 complexes contains other proteins that are essential for Ezh2 activity: Suppressor of Zeste 12 (Suz12), Extra-embryonic endoderm (Eed), and the histone binding proteins Rbap46/Rbap48 (8, 29, 34). One or more of the core components also have the ability to modulate Ezh2 activity/function. For example, the core component Eed is a WD40 repeat-containing protein that exists in different isoforms due to the presence of alternative translational start sites (12). These isoforms are differentially represented in PRC2/PRC3/PRC4 and are determinant with respect to the alternate substrate specificity exhibited by Ezh2 in these complexes. In addition to containing a unique form of Eed (Eed2), the PRC4 complex is associated with the NAD-dependent histone deacetylase SirT1 and is able to methylate the linker histone H1 at lysine 26 within its N-terminal tail (18).
H3K27 methylation exists in four different forms in vivo (unmodified and mono-, di-, and trimethylated), and their relative abundances differ. The most abundant is H3K27me2, constituting approximately 50 percent of all methylated H3K27 in vivo. H3K27me1 and H3K27me3 comprise about 25% and 10%, respectively (35). The methylation of H3K27 has been associated with transcriptionally repressed chromatin. In particular, H3K27me3 is enriched on the inactive X chromosome in mammalian cells (32, 36, 42). The pericentromeric heterochromatin is characterized by the presence of the H3K27me1 modification (35).
The PRC1 complex contains the chromodomain protein Polycomb (Pc) that has been shown to have an affinity for histone H3 di- or trimethylated at K27 (14, 27). The interaction of Pc with H3K27me2/H3K27me3 led to an attractive proposition that methylation of H3K27 by Ezh2-containing complexes would recruit PRC1 via the chromodomain of Pc and result in repression of the target genes (7). That this is the pathway to repression is unclear, however, since some genes are occupied by Pc but lack both E(z) and H3K27me3 and many regions have H3K27me3 but do not bind Pc (33, 39). Moreover, plants and some fungi also have the PRC2 complex which functions in repression and yet the PRC1 complex is absent (38).
The Polycomblike (Pcl) protein was first identified in Drosophila (13). Early studies with Pcl showed that it was localized to PcG sites on Drosophila polytene chromosomes (23). Polycomblike is an essential gene, as lack of functional Pcl in Drosophila results in posterior-directed homeotic transformation and embryonic lethality. Genetic studies indicated that Pleiohomeotic (Pho), the Drosophila homolog of human YY1, interacts with both polycomb (Pc) and Pcl during embryonic development as well as in the adult. Double mutants in Pho and either Pc or Pcl enhance the relatively mild transformation seen in embryos that are mutant with respect to Pho alone (20). Biochemical analysis of the Drosophila Polycomblike protein (dPcl) showed that it is present in a 1-MDa complex in early embryos and that this complex is distinct from the 600-kDa E(z) complex PRC2 (45). The Pcl protein has a TUDOR domain and two tandem PHD finger domains that are also conserved in its three human homologues, PHF1 (hPcl1), MTF2 (hPcl2), and PHF19 (hPcl3) (10, 16, 47). In vitro translation experiments revealed that Pcl interacts with E(z) and the histone deacetylase Rpd3 via its PHD fingers. Two-hybrid interaction studies revealed that the human homolog PHF1 but not MTF2 interacts with Ezh2 via its PHD fingers (31, 45).
Given our findings that Ezh2 is active only when composing the PRC complexes and that the presence of different Eed isoforms in PRC2/PRC3/PRC4 as well as the association of SirT1 in PRC4 impacts Ezh2 substrate specificity, we examined the consequences of PHF1 association with Ezh2.
Antibodies against H3K27me2, H3K4me3, Ezh2, Suz12, Rbap46, and PHF1 were produced in the laboratory of D.R. H3K27me3 antibodies were obtained from Thomas Jenuwein's laboratory. Bmi-1 was a generous gift from Kristian Helin. Other commercial antibodies used were H3K27me1 (Upstate catalog no. 07-448), H3K79me2 (Upstate catalog no. 07-366), Gal4 (Upstate catalog no. 06-262), H3K9me2 (Upstate catalog no. 07-441), H3K9me3 (Upstate catalog no. 07-442), H4K20me3 (Upstate catalog no. 07-463), bromodeoxyuridine (BrdU) (Roche), actin (Sigma), hemagglutinin (HA) (Sigma), and YY1 (Santa Cruz 7341).
I.M.A.G.E. clones containing PHF1 in the pCMVsport6 vector were purchased from ATCC. Full-length PHF1 was amplified by PCR and cloned into the 5′ Not1 and 3′ Xba1 sites in pFLAGCMV4 vector (Sigma). Subsequent cloning was done using standard techniques.
A truncated form of PHF1 that contained the N-terminal TUDOR domain and the first PHD finger was cloned into pet102 vector (Invitrogen) by directional TOPO cloning. The protein was expressed in BL21pLysS cells and purified using nickel-nitrilotriacetic acid agarose according to the manufacturer's protocols and used as an antigen to produce rabbit polyclonal antibodies.
pFLAGCMV4-PHF1 vector was transfected into 293F cells, and stable transformants were selected with neomycin. PHF1 was cloned into modified pcDNA4TO (Invitrogen) vector containing the N-terminal Gal4 DNA binding domain and a C-terminal HA epitope tag. This construct was then transfected into a 293 Trex cell line expressing luciferase as described previously (46). Gal4-PHF1-HA was induced by addition of 1 μg/μl tetracycline for 12 h. Luciferase assays were performed as described earlier (46).
HeLa cells were grown in Dulbecco modified Eagle medium supplemented with 10% bovine serum. RNAs with a 19-bp stem-loop were expressed from pLSL-Puro vector. HeLa cells were transfected with this construct by use of Effectene and selected with 1 μg/ml puromycin to generate stable clones. Reverse transcriptase PCR (RT-PCR) was performed to analyze stable knock-down clones. For cell proliferation assays, 104 cells were plated in quadruplicate in 24-well plates and were harvested at 24-h intervals. Total DNA content was quantified by DABA (3,5-diaminobenzoic acid dihydrochloride) staining as described previously (25). The stem-loop small interfering RNAs (siRNAs) were synthesized as two complementary 59-bp oligonucleotides, annealed, and then cloned into the BamHI/EcoRI sites of pLSL-puromycin vector. The sequence (targeted sequence in uppercase) of the oligonucleotide used was as follows: 5′-gatccGCAACCGACAGCAGAGTTActtcctgtcataactctgctgtcggttgctttttg-3′.
RNA was extracted using Trizol reagent (Invitrogen) per the manufacturer's instructions. RT-PCR of isolated RNA (51) and chromatin immunoprecipitation (ChIP) (1) were performed as described previously. All primer sequences are available upon request.
All real-time PCR analyses were performed using a Stratagene MX3005p apparatus and LightCycler 480 SYBR green 1 Master Mix (Roche). All reverse transcriptase and chromatin immunoprecipitations were performed at least three times and analyzed by real-time PCR. For reverse transcriptase experiment results, RNA levels are represented as values relative to GAPDH (glyceraldehyde-3-phosphate dehydrogenase) levels. Chromatin immunoprecipitation results were quantified by correcting the background (immunoglobulin G control) from the immunoprecipitated sample and are represented as percentages of the total input. Standard deviation values for the results from three independent sets of experiments are shown.
FLAG-PHF1 was cloned into the EcoRI site of the pAcHLT-a vector (BD Biosciences). Baculovirus was generated by Kinnekeet Biotechnology. Production of PRC2 complexes by use of the baculovirus system was described previously (8). Hi-5 cells were infected at a multiplicity of infection of 10 with virus expressing FLAG-PHF1. Cells were harvested after 3 days and lysed by sonication, and the lysate was incubated for 4 h with M2 agarose beads (Sigma). Washes were performed with BC500 buffer containing 50 mM Tris, 2 mM EDTA, 500 mM KCl, 10% glycerol, and protease inhibitors. Protein was eluted with FLAG peptide at a 0.2 mg/ml concentration.
Standard HMT assays were performed as described previously (30). In brief, PRC2 and PHF1 were incubated on ice for 30 min. Substrates (either recombinant octamers or nucleosomes) were added to this mixture along with 5× methylation buffer, 0.2 M dithiothreitol, 3H-labeled S-adenosylmethionine, and water. The reaction mixture was incubated at 30°C for 1 h and processed for either Western blot or activity analyses.
All transient transfections were done using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Cells were harvested 48 h after transfection for Western blot or luciferase analyses.
FLAG-PHF1 was immunoprecipitated by incubating M2 beads with nuclear extract for 4 h in BC300 buffer (50 mM Tris [pH 7.9], 300 mM KCl, 2 mM EDTA, 10% glycerol, and protease inhibitors). Washes were performed with the same buffer containing 500 mM KCl. Proteins were eluted with FLAG peptide at a 0.2 mg/ml concentration. The eluate (50 μl) was applied to a 2.4-ml S200 gel filtration column (Amersham Pharmacia), and 50-μl fractions were collected. Fractions were then analyzed by Western blotting. Immunoprecipitations of endogenous proteins were carried out overnight with protein A beads that were prepared by covalently cross-linking specific antibodies. Beads were washed with BC300 buffer and the proteins eluted using 0.1 M glycine (pH 2.5).
The direct interaction observed between the Drosophila Pcl protein and E(z) is conserved between their human counterparts PHF1 and Ezh2, respectively, as shown by the results of two-hybrid interaction studies (31). To test whether other PRC components might also interact with PHF1, we first established a stable cell line expressing full-length FLAG-tagged PHF1. Immunoprecipitation using anti-FLAG antibody revealed that PHF1 coprecipitated with components of a PRC complex, specifically Ezh2, Suz12, Eed, and RbAP46/RbAP48 (Fig. (Fig.1A).1A). The complex was identified as PRC2 based on the presence of the PRC2-specific Eed isoforms (Eed1, Eed3, and Eed4). We could not detect any association with YY1, demonstrating specificity of the interaction between PHF1 and PRC2. Additionally, cells transfected with a FLAG-empty vector failed to immunoprecipitate components of the PRC2 complex (Fig. (Fig.1A),1A), establishing that the interaction is dependent on PHF1.
To further characterize the interaction between PHF1 and Ezh2, endogenous proteins were immunoprecipitated with specific antibodies and subjected to Western blot analysis. As seen in Fig. Fig.1B,1B, the PHF1 antibodies coimmunoprecipitated Suz12 and Ezh2 as well as Eed (lane 5) from HeLa nuclear extracts whereas the Ezh2 antibodies pulled down Suz12 and Eed but not PHF1 under the same conditions (lane 3). To rule out that the lack of interaction was due to epitope masking, we used Eed antibodies and found that Ezh2 and Suz12 coimmunoprecipitated but not PHF1 (lane 4). We also repeated these experiments with antibodies against Suz12. As shown in Fig. Fig.1B1B (bottom panel), anti-Suz12 antibody is able to immunoprecipitate Ezh2 but not PHF1 (lane 3) from HeLa nuclear extracts. These results suggest that a small fraction of Ezh2 is in association with PHF1 in human cells.
Analysis of the FLAG-PHF1 immunoprecipitates by size exclusion chromatography indicated that the PHF1 protein overlapped with the PRC2 complex at approximately 600 kDa (Fig. (Fig.1C).1C). This native mass is slightly greater than that of the PRC2 complex alone. Of note, a fraction of PRC2 migrated at 500 kDa and was independent of the presence of PHF1. Previous studies have shown that the Drosophila Polycomblike (Pcl) and Enhancer of Zeste (Ez) proteins migrated as a 1-MDa complex distinct from the PRC2 complex, indicating differences in complex composition between the mammalian and fly systems. Our attempts to purify the ~600-kDa complex by conventional chromatographic columns resulted in coelution of PHF1 and PRC2 in the initial steps; however, subsequent steps resulted in PHF1 dissociation from the core PRC2 complex (data not shown). The results indicate that the majority of Ezh2 in cells is present in the PRC2/PRC3 complex and that a minor fraction of the PRC2 complex is associated with PHF1. Based on these observations and the results presented in Fig. Fig.1B,1B, we speculate that the human PHF1-PRC2 interaction is transient and perhaps requires the presence of chromatin for stability (see below).
Recently, several studies have identified the genes targeted by Ezh2 (3, 4, 22, 39, 43). Among the well-characterized non-HOX target genes is the MYT1 (myelin transcription factor 1) gene (17). In order to determine whether PHF1 colocalizes with Ezh2 at the MYT1 promoter, we used ChIP experiments. As previously described (17), we observed a strong enrichment for Ezh2 at the MYT1 distal fragment but not at the coding region (exon 5) (Fig. (Fig.2A).2A). Using an antibody generated against PHF1, we found that PHF1 is also enriched at the same location as Ezh2 (Fig. (Fig.2A2A).
We then expanded this experiment to include other Ezh2 target genes such as WNT1 and also some Hox genes, namely, HOXA9 and HOXA11, and observed that PHF1 is localized at all promoters tested but not in the coding regions of these genes (Fig. (Fig.2A).2A). The presence of PHF1 with these gene promoter sequences is specific, as PLCB4, a gene that is not regulated by Ezh2 (17), does not show any enrichment for either Ezh2 or PHF1 (Fig. (Fig.2A).2A). Previous studies have shown that Ezh2 and Suz12 bind to two distinct regions within the distal promoter of the MYT1 gene (17). To determine whether PHF1 binds in a manner similar to that of Ezh2 and of Suz12, we performed ChIP analyses using chromatin that was sheared to ~500 bp after cross-linking. In agreement with previous studies, we observed Ezh2 at two regions of the MYT1 promoter (Fig. (Fig.2B,2B, left panel), and importantly, the profile of PHF1 binding was very similar to that of Ezh2 at these areas (Fig. (Fig.2B,2B, left panel). Moreover, the levels of H3K27me3 peaked at the same regions as those of Ezh2 and PHF1 despite a lower H3 occupancy (Fig. (Fig.2B,2B, right panel). On the other hand, the levels of H3K27me2 were decreased in several regions that were enriched for PHF1, Ezh2, and H3K27me3 (Fig. (Fig.2B,2B, right panel). The cooccupancy of PHF1 and Ezh2 at several promoters, along with their ability to interact, prompted us to analyze the possible role of PHF1 in the recruitment/regulation of Ezh2.
The homologues of Polycomblike (Pcl) in chickens and Xenopus exhibit transcriptional repressive activity (48, 50). To determine whether this is the case in the mammalian system, we generated a construct containing a tetracycline-inducible gene encoding PHF1 tagged at its N terminus with the Gal4 DNA binding domain and at its C terminus with HA (Gal4-PHF1-HA). Cells were then cotransfected with this plasmid and those encoding luciferase and β-galactosidase. Protein expression levels were determined by Western blot analysis (Fig. (Fig.3A,3A, right panel) and the transfection efficiency was determined by normalizing luciferase activity to β-galactosidase levels. The results indicate that human PHF1 functions as a repressor when targeted to a promoter. A PHF1 construct without the Gal4 DNA binding domain did not repress luciferase expression (Fig. (Fig.3A,3A, left panel).
We next used this tetracycline-inducible Gal4-PHF1-HA construct in transfection experiments with 293 Trex cells that stably express the luciferase gene containing Gal4 DNA binding sites and the thymidine kinase (TK) promoter (293 Trex TK Luc)(Fig. Luc)(Fig.3A).3A). The tetracycline-dependent induction of the Gal4-PHF1-HA protein was confirmed by Western blot analysis (Fig. (Fig.3B).3B). To explore the mechanism of PHF1-mediated repression, we performed ChIP before and after its induction. After tetracycline addition, tagged PHF1 was targeted to the TK promoter, and this resulted in increased levels of the PRC2 components Ezh2 and Suz12 as well as the PRC1 component Bmi-1 (Fig. (Fig.3B).3B). In agreement with previous studies showing an overall exclusion between RNA polymerase II and Ezh2 at Ezh2-regulated genes (43), we observed a reduced occupancy of RNA polymerase II (Fig. (Fig.3B).3B). We also observed an increase in the level of the repressive mark H3K27me3 and a reduction in the level of H3K4me3, a mark associated with active transcription (Fig. (Fig.3C).3C). Of note, we observed a reduction in H3K27me2 levels at this locus. As we did not observe a change in histone H3 occupancy (H3K79me2 levels in Fig. Fig.3C3C and data not shown), this decrease in H3K27me2 might have been a consequence of active demethylation or histone exchange. However, as H3K27me2 occupancy decreased concomitantly with increased H3K27me3 levels, we hypothesize that H3K27me2 is the substrate for further methylation by Ezh2-containing complexes.
We also observed that induced, tagged PHF1 localized to the endogenous MYT1 promoter and that this correlated with the increased presence of Ezh2 and of Suz12 there (data not shown). However, since it is a natural Ezh2 target gene and thus already contains high levels of the repressive H3K27 methyl marks, we did not observe a quantitatively reliable change in either H3K27me2 or H3K27me3 levels at this locus (data not shown).
Analysis of tetracycline-inducible Gal4-tagged Ezh2 in the 293 Trex TK Luc cell line showed that recruitment of Gal4-Ezh2 to the TK promoter resulted in recruitment of Suz12, as expected (Fig. (Fig.3D).3D). However, this did not lead to an increased presence of either PHF1 or Bmi-1 (Fig. (Fig.3D).3D). This is in contrast to the observations with respect to the tagged PHF1 that led to Ezh2 and the PRC1 component Bmi-1 being recruited (see above). The absence of PHF1 recruitment by tagged Ezh2 is consistent with only a minor fraction of cellular PRC2 being associated with PHF1. This failure to recruit Bmi-1 prompted us to analyze the status of the H3K27 methyl marks upon targeting of Gal4-Ezh2. Surprisingly, the Gal4-Ezh2 that was recruited gave rise to H3K27me2 but not H3K27me3 at the TK promoter (Fig. (Fig.3D).3D). Although the Pc component of PRC1 has comparable affinity for both H3K27me2 and H3K27me3 in vitro, it appears to bind preferentially to H3K27me3 in vivo based on its localization to regions of chromatin that are trimethylated at H3K27. The absence of Bmi-1 at the TK promoter is therefore consistent with the lack of H3K27me3 in this region. To confirm the specificity of increased H3K27me2 at this site, we also used a Gal4-tagged catalytic mutant of Ezh2 (H689A) (19). Recruitment of this mutant protein to the TK promoter did not result in an increase in either H3K27me2 or H3K27me3 levels (Fig. (Fig.3E).3E). Thus, Ezh2-containing complex recruited in a manner independent of PHF1 catalyzes H3K27me2 but not H3K27me3 in vivo.
Given that PHF1-mediated recruitment of Ezh2-containing complex did give rise to increased levels of H3K27me3, we hypothesize that the presence of PHF1 is required for Ezh2-mediated catalysis of H3K27me3 in vivo, apparently at the expense of H3K27me2, either by modulating Ezh2 methyltransferase activity or prolonging its occupancy, within the context of PRC2 (see below).
We also studied the ability of artificially recruited Gal4-PHF1 to maintain repression through several cell cycles by use of the stable 293 Trex TK Luc cell line that also contains a stably integrated Gal4-PHF1 gene. Gal4-PHF1 expression was induced for 12 h followed by tetracycline removal, and cells were then harvested every 24 h for a period of 1 week. Luciferase activity and Gal4-PHF1 expression levels were measured at each time point. Upon Gal4-PHF1 induction, luciferase activity was reduced, and upon tetracycline removal, Gal4-PHF1 protein levels gradually decreased, becoming undetectable by day 4. This gradual decrease was concomitant with a gradual increase in luciferase expression that eventually reached the levels exhibited prior to Gal4-PHF1 induction (Fig. (Fig.3F).3F). Western blot analysis at each time point showed a good correlation between the level of luciferase expression and the amount of Gal4-PHF1 protein in the cells. We also performed ChIPs before and after release from tetracycline (Fig. (Fig.3G).3G). Upon addition of tetracycline, PHF1 was recruited to the TK promoter along with Ezh2. The enrichment of both PHF1 and Ezh2 continued up to day 2 after release from tetracycline. Subsequently, both PHF1 and Ezh2 were lost, suggesting that recruitment and maintenance of Ezh2 at this region are dependent on the presence of PHF1. The level of the H3K27me3 mark also increased after induction of PHF1 and was lost after day 4 of release. On the other hand, the H3K27me2 mark was reduced slightly upon recruitment of PHF1 and then increased to its original levels by day 4. These results collectively suggest that the repression observed with PHF1 in this system is not maintained.
To test whether the repression mediated by Gal4-PHF1 is dependent on Ezh2, Ezh2 was knocked down prior to Gal4-PHF1 induction. The reduced Ezh2 levels did not affect the degree of repression brought about by Gal4-PHF1 recruitment (data not shown). This suggests that PHF1 is able to repress gene expression independently of Ezh2 (see below).
As our results thus far demonstrated that artificial recruitment of PHF1 to the integrated TK promoter correlates with Ezh2 localization at this site, we next tested whether PHF1 directly recruits Ezh2 to its target genes in vivo. To this end, we adopted a vector-based RNA interference approach. We first knocked down Ezh2 transiently in HeLa cells (Ezh2-KD) and analyzed its effect on the levels of PHF1 (Fig. (Fig.4A).4A). ChIP experiments showed that there was no effect on the levels of PHF1 at the MYT1 and HoxA9 promoters, suggesting that the presence of PHF1 at these loci is independent of Ezh2 (Fig. (Fig.4A).4A). We next examined the effects of PHF1 knockdown (PHF1-KD) (Fig. (Fig.4B,4B, left panel). We confirmed the specificity of short hairpin RNAs (shRNAs) against PHF1 by use of quantitative RT-PCR (qRT-PCR) to gauge the RNA levels of PHF1 as well as those of MTF2, the closest Polycomblike homolog to PHF1. While the RNA levels of PHF1 were reduced twofold, MTF2 gene expression was not affected (Fig. (Fig.4B,4B, left panel). The localization and the levels of both Ezh2 and Suz12 were not affected in the PHF1-KD cells in the case of the MYT1 and HoxA9 loci (Fig. (Fig.4B,4B, middle and right panels, and data not shown). These results show that while PHF1 and Ezh2 have overlapping natural target promoters, their associations with target genes occur independently.
The loss of the PRC2 component Suz12 has been shown to result in the loss of the PRC1 component Bmi-1 from the HoxA9 locus (6). Interestingly, association of Bmi-1 at both the MYT1 and HoxA9 promoters was decreased upon knockdown of PHF1. Given that the preferred target of PRC1 in vivo is H3K27me3, this result is consistent with H3K27me3 being undetectable at the artificial promoter tested when Ezh2 was recruited without PHF1 (see above).
Having shown that the recruitment of PHF1 results in a specific increase in H3K27me3 levels (Fig. (Fig.3C),3C), we next investigated whether the absence of PHF1 affects histone H3K27 methylation at the endogenous Ezh2 target loci. We first confirmed that the Ezh2-KD results showed a decrease in H3K27me2 and H3K27me3 levels, as would be expected (Fig. (Fig.5A).5A). Surprisingly, a different result was obtained upon investigation of PHF1-KD. While the levels of H3K27me3 were decreased (as seen with HoxA6, A9, and A11), as in the case of Ezh2-KD (Fig. (Fig.5B,5B, right panel), the levels of H3K27me2 were not and instead were slightly increased at these same loci (Fig. (Fig.5B,5B, left panel). H3K27 methylation levels remained unchanged at the HoxA4 locus coincident with this gene not being affected by PHF1-KD (see below). These results are consistent with the possibility that when associated with an Ezh2-containing complex such as PRC2, PHF1 stimulates the formation of H3K27me3 at the expense of H3K27me2. These results suggest that rather than functioning to recruit Ezh2, PHF1 is able to modulate its enzymatic properties in vivo at defined target genes.
Previous studies have shown that the loss of the Suz12 component of PRC2/PRC3/PRC4 at the Hox A9 promoter results in its up-regulation in HeLa cells (8). We compared the expression levels of the HoxA gene cluster in HeLa versus PHF1-KD cells. The levels of several Hox genes, HoxA3, HoxA6, HoxA9, HoxA10, and HoxA11, were up-regulated upon the loss of PHF1 (Fig. (Fig.5C5C and data not shown). These Hox genes were also up-regulated in response to Ezh2 depletion (Fig. (Fig.5C5C and data not shown). Moreover, the combined knockdown of both PHF1 and Ezh2 did not result in a further up-regulation of these genes (Fig. (Fig.5C).5C). Although H3K27me3 levels were reduced, H3K27me2 levels were increased at these promoters when PHF1 was knocked down (Fig. (Fig.5B5B and and5C).5C). This profile was not observed in the case of the HoxA4 gene, whose expression levels were unaffected by either PHF1-KD or Ezh2-KD (Fig. 5B and C). This suggests that increased H3K27me2 levels are not sufficient for maintaining the repressed states of HoxA6, HoxA9, and HoxA11 and that H3K27me3 is required for gene silencing at these loci.
Since our results showed that PHF1 modulates the ability of Ezh2 to dimethylate or trimethylate its substrate locally, we next sought to determine whether PHF1 has a more general effect on Ezh2 function. We compared the global histone modification profile from HeLa cells that were knocked down for PHF1 versus that from untreated cells. Histones were purified by acid extraction and analyzed by Western blotting for changes in levels of methylation status by use of specific antibodies (as indicated in Fig. Fig.6A).6A). The levels of H3K27me3 were severely reduced but not those of H3K27me1 or of H3K27me2. There were no detectable changes in the global levels of H3K9me3, H3K4me3, or H4K20me3 (Fig. (Fig.6A6A and data not shown). We observed a similar reduction in H3K27me3 levels by use of another approach to reduce PHF1, that is, by transfection of shRNA against PHF1 (Fig. (Fig.6C).6C). The transfected cells exhibited similar levels of Ezh2 and Suz12 relative to those of untreated cells in Western analysis, ruling out the possibility that the reduced H3K27me3 levels stemmed from decreased amounts of Ezh2 (or Suz12) in these cells (Fig. 6B and C). Thus, PHF1 is essential for global H3K27me3 in human cells.
Knockdown of one of the PRC2/PRC3/PRC4 components, Suz12, was shown previously to lead to reduced cell proliferation and detectable morphological defects (8). To test whether PHF1 is essential for cell proliferation, we performed a cell proliferation assay and compared untreated HeLa cells to those that stably express PHF1 shRNA. We found that PHF1-KD cell lines were severely compromised in their ability to proliferate (Fig. (Fig.6D)6D) and also displayed morphology distinct from that of untreated cells (Fig. (Fig.6E).6E). We did not detect activation of the caspase pathway in PHF1-KD cells, suggesting that the decreased proliferation observed is a result of an alteration in cell cycle progression and not apoptosis (data not shown). It was shown previously that knockdown of Ezh2 resulted in a decreased amount of cells in the S phase of the cell cycle (5). We also observed a decrease in the S-phase population in the case of PHF1-KD cells by use of BrdU staining (Fig. (Fig.6F6F).
To further elucidate the mechanism of PHF1 action with respect to the methyltransferase activity of PRC2, we purified PHF1 and PRC2 independently from Hi-5 cells (Fig. (Fig.7A).7A). PHF1 was added to PRC2, and histone methyltransferase assays were performed using wild-type as well as mutant recombinant octamers or nucleosomes. PRC2 was able to methylate mainly H3K27 in vitro (Fig. (Fig.7B7B and data not shown), and the addition of PHF1 did not alter this specificity but instead stimulated the PRC2-associated histone methyltransferase activity to increase by two- to threefold (Fig. (Fig.7C,7C, middle panel). The highest point of PHF1 titration represents equal amounts of FLAG-PHF1 and FLAG-Eed (data not shown). The stimulation observed was not a result of a contaminant associated with the PHF1 protein (Fig. (Fig.7C,7C, left panel) and appears to be specific, as the addition of bovine serum albumin (BSA) did not affect the methylation activity (Fig. (Fig.7C,7C, right panel). We next performed Western blot analysis on the PRC2 and PRC2-PHF1 reaction products by use of antibodies specific to the mono-, di-, or tri-methyl forms of H3K27. The primary product of the PRC2 reaction was H3K27me1, and there were also small but detectable amounts of H3K27me2 and H3K27me3 consistent with a stepwise process. While the levels of H3K27me1 and of H3K27me2 remained unchanged upon PHF1 addition, the H3K27me3 levels were moderately but consistently increased (Fig. (Fig.7D;7D; compare lanes 4 and 5, lanes 7 and 8, and lanes 9 and 10). This increase was not a result of the presence of endogenous histones that might have associated with the PRC2 complex or the PHF1 protein (Fig. (Fig.7D,7D, lanes 1 and 2). Instead, these results strongly suggest that PHF1 functions to stimulate Ezh2 catalysis of H3K27me3 in vitro and are consistent with the results we observed in vivo.
Our results demonstrate that efficient H3K27me3 production by PRC2 requires its association with PHF1. The interaction between PHF1 and the PRC2 complex we observed in vitro also occurs in vivo at the promoters of the Ezh2 target genes tested. In similarity to the results seen with Ezh2 and its associated proteins, PHF1 is absent from the coding region of these genes. It is interesting that the regions occupied by PHF1 and Ezh2, and therefore by H3K27me3, across the MYT1 locus correspond to highly conserved regions in the genome (as shown earlier; see reference 44). PHF1 functions as a transcriptional repressor, and its targeting to an artificial promoter results in elevated levels of the Ezh2 complex components and of H3K27me3 at the promoter. Yet the presence of Ezh2 and the presence of PHF1 at their common target gene promoters occur independently of each other. The loss of PHF1 at the HOXA6, HOXA9, and HOXA11 promoters results in a decrease in the level of H3K27me3, an increase in the level of H3K27me2, and a deregulation of these genes, even though the levels of Ezh2 are unchanged. Further analysis also showed that this decrease in the level of H3K27me3 is a global phenomenon.
To our knowledge, this is the first time that PHF1 has been defined as a factor affecting Ezh2 activity. On a gel filtration column, PHF1 and PRC2 overlapped but did not perfectly coelute as would be expected of components of a stable complex. Therefore, unlike the core components Eed and SirT1 that affect PRC2/PRC3/PRC4 activity through changes in Ezh2 substrate specificity, PHF1 is not an integral component of these complexes but is rather an associated protein that does not alter Ezh2 substrate specificity but instead stimulates its trimethyltransferase activity.
Upon knockdown of PHF1, there is a loss of H3K27me3 and a concomitant increase in H3K27me2 levels at some of the Hox loci. These same genes were deregulated despite the increased presence of H3K27me2. Ezh2-KD also deregulated these genes to the same extent as PHF1-KD, and there was no additional derepression observed in a double knockdown. Thus, both PHF1 and Ezh2 are required to maintain the repressed state of these Hox genes, as loss of either protein causes complete derepression. Our studies are in agreement with those of Nekrasov et al., who reported that the Drosophila Pcl protein is required for H3K27me3 at polycomb target genes (28a). Their studies examining the Ubx and Abd-B loci showed that loss of Pcl results in a decrease in H3K27me3 levels and a concomitant increase in H3K27me1 and H3K27me2 levels. Another recent study by Cao et al. suggests that mPHF1 is required for global H3K27me2 and H3K27me3 formation (6a). While we clearly detected a decrease in global H3K27me3, we did not observe any changes in global H3K27me2 levels but instead observed an increase in H3K27me2 levels at polycomb target genes that were depleted of PHF1.
Our results underscore the role of H3K27me3 in repression while raising questions regarding the role of H3K27me2. The results with respect to the abundance of H3K27me2 (from 40 to 60% of total histone H3, depending on the studies and on the organism) do not support the idea of its being a functional repressive mark. In the case of Drosophila, a large fraction of the genome bears the H3K27me2 mark without E(z) association. While the genomic distribution of H3K27me2 is not yet clear in the mammalian case, we suggest that H3K27me2 is an intermediate H3K27 methylation state that marks genes as being potentially repressible and that the concerted action of Ezh2 (in the context of its associated proteins) and PHF1 is required to achieve complete repression (Fig. (Fig.7E).7E). While the association of PRC2 with its target genes may be required for H3K27me3 catalysis, the presence of PHF1 at these loci allows PRC2 to efficiently catalyze H3K27me3.
More insights into the roles of H3K27me2 and H3K27me3 will likely come about upon the identification of the downstream effectors and their specificities. Of particular interest is the chromodomain protein, Polycomb (Pc), which has been shown to bind both H3K27me2 and H3K27me3 marks without apparent discrimination in vitro, although our studies suggest that Pc binds preferentially to H3K27me3 in vivo, as measured indirectly by the presence of Bmi1. Another interesting possibility is that in the absence of PHF1, Ezh2 is able to catalyze H3K27me3 only inefficiently; perhaps the establishment of these low levels of H3K27me3 stabilizes PHF1 at the site (through the TUDOR or PHD domains as discussed below), with resultant stimulation of Ezh2 activity (Fig. (Fig.77).
In vitro reconstitution experiments have indicated that the addition of PHF1 stimulates the activity of PRC2 to increase twofold. While the major product formed in an in vitro reaction with PRC2 is H3K27me1, minor amounts of H3K27me2 and H3K27me3 were also formed. We observed that addition of PHF1 resulted in a specific increase in H3K27me3 levels without changes in H3K27me1 or H3K27me2 levels. Interestingly, PHF1 contains a TUDOR domain as well as two PHD fingers. The TUDOR domain of 53BP1 was shown to bind histone H3K79me2 as well as H4K20me2 (2, 15). Similarly, the PHD finger of ING2 was shown to bind H3K4me3 (41). The presence of both the TUDOR and PHD fingers raises the possibility that PHF1 could bind to methylated H3K27 and promote Ezh2-dependent trimethylation at this residue. Whether PHF1 exhibits preferential binding to methylated histone residues and whether PHF1 can discriminate between the levels of methylation remains to be seen. The PHD fingers of some proteins were also shown to contain ubiquitin E3 ligase activity (9, 24, 49). Our biochemical studies failed to demonstrate PHF1-mediated ubiquitination of histone proteins, but whether the PHD fingers of PHF1 are able to function in this manner in the presence of additional factors remains an open question.
Of note, overexpression of Ezh2 as well as other PRC components has been correlated with cellular transformation. In fact, one of the Eed isoforms (Eed2) is present specifically in cells that are either undifferentiated or tumorigenic and Eed1 regulates a specific set of genes compared to other Eed isoforms (18, 43). The expression level of hPCL3, the human homologue of Polycomblike, is also increased in several types of cancers (47). More recently, the PHF1 gene was reported to be rearranged in three cases of endometrial stromal sarcoma, resulting in the production of two types of chimeric proteins (JAZF1/PHF1 and EPC1/PHF1) that retained the complete open reading frame of PHF1 but under the control of the JAZF1 (Juxtaposed with another zinc finger 1) and the EPC1 (Enhancer of polycomb 1) promoters, respectively (26). This suggests that in addition to the production of an abnormal fusion protein, another cause of pathogenicity might be the altered regulation of the PHF1 gene. Given that PHF1 interacts with the PRC2 complex in vitro and modulates the ability of Ezh2 to catalyze the repressive H3K27me3 mark in vitro and in vivo, PHF1 becomes another likely candidate for investigation with regard to cancer progression.
We thank Lynne Vales for comments on the manuscript and members of the Reinberg laboratory for helpful discussions. We are very grateful to Thomas Jenuwein for the gift of antibodies that are specific for H3K27me2 and H3K27me3 and to Kristian Helin for the Bmi-1 antibody. We are also grateful to Deborah Hernandez for excellent technical assistance. A.I. thanks Frank Rauscher for his support.
This work was supported by grants from DOD (PC050535 to R.M.), NIH (GM64844 to D.R.), and the HHMI (D.R.).
Published ahead of print on 19 February 2008.