|Home | About | Journals | Submit | Contact Us | Français|
The Polycomb group of proteins forms at least two distinct complexes designated the Polycomb repressive complex-1 (PRC1) and PRC2. These complexes cooperate to mediate transcriptional repression of their target genes, including the Hox gene cluster and the Cdkn2a genes. Mammalian Polycomb-like gene Pcl2/Mtf2 is expressed as four different isoforms, and the longest one contains a Tudor domain and two plant homeodomain (PHD) fingers. Pcl2 forms a complex with PRC2 and binds to Hox genes in a PRC2-dependent manner. We show that Pcl2 is a functional component of PRC2 and is required for PRC2-mediated Hox repression. Pcl2, however, exhibits a profound synergistic effect on PRC1-mediated Hox repression, which is not accompanied by major alterations in the local trimethylation of histone H3 at lysine 27 (H3K27me3) or PRC1 deposition. Pcl2 therefore functions in collaboration with both PRC2 and PRC1 to repress Hox gene expression during axial development. Paradoxically, in embryonic fibroblasts, Pcl2 is shown to activate the expression of Cdkn2a and promote cellular senescence, presumably by suppressing the catalytic activity of PRC2 locally. Taken together, we show that Pcl2 differentially regulates Polycomb-mediated repression of Hox and Cdkn2a genes. We therefore propose a novel role for Pcl2 to modify functional engagement of PRC2 and PRC1, which could be modulated by sensing cellular circumstances.
The Polycomb group (PcG) was first identified in Drosophila melanogaster as a gene cluster necessary for the maintenance of segmental identity (50, 53). The PcG proteins are required for the epigenetic repression of a number of developmental regulatory genes, including homeotic genes. They do so by forming at least two distinct multimeric protein complexes at their target loci, known as Polycomb repressive complex-1 (PRC1) and PRC2 (11, 13, 46, 57).
In Drosophila, the core PRC1 is formed by four proteins, Polycomb (Pc), Polyhomeotic (Ph), Posterior sex combs (Psc), and dRing (Sex comb extra; Sce), which inhibit the chromatin remodeling activity of SWI/SNF complexes (20, 35). Mammalian orthologs form functionally similar complexes (40). PRC1 has been shown to act as an E3 component for ubiquitination of histone H2A at lysine 119 (uH2A) via Ring1B and Ring1A catalytic subunits (9, 14, 62). In addition, regulation of higher-order chromatin structure is a recently identified function mediated by PRC1 (19). The core of the mammalian PRC2 is composed of three evolutionarily conserved proteins, Eed, an ortholog of extra sex combs [Esc]), Ezh2, an ortholog of enhancer of zeste [E(z)], and Suz12, which mediates trimethylation of histone H3 at lysine 27 by the Ezh2 component (11, 13, 46, 48, 51, 60). Although both PRC1 and PRC2 are required to maintain Hox repression and share target loci in both vertebrates and invertebrates, these complexes do not associate with each other in most cell types. A plausible model for Polycomb silencing is through the engagement of PRC1 with the target via its interaction with H3K27me3 mediated by the chromodomain of Pc or its mammalian counterparts (11, 62). However, recent genome-wide analysis of PRC1 and PRC2 occupancy revealed that PRC1 binding demarcated a portion of the PRC2 targets (37), implying that functional engagement of PRC1 and PRC2 is not constitutive but, instead, is regulated by undefined intrinsic and/or extrinsic signals.
The polycomblike (Pcl) gene was initially isolated as a PcG gene during a Pc mutation modifier screen in D. melanogaster and then identified in several studies as an enhancer for Esc mutations (17, 33). Pcl mutant phenotypes are similar to those of Pc mutants, and interactions between mutant Pcl and Pc or Esc synergistically enhance the respective phenotypes in a dose-dependent manner. The Pcl protein has a Tudor domain and two tandem plant homeodomain (PHD) fingers (43, 49, 64). Pcl has been shown to be a part of two distinct protein complexes, PRC2 and a thus far uncharacterized complex, which appear at Drosophila embryonic and larval stages, respectively (47, 56, 61). In Drosophila embryos, Pcl forms complexes with PRC2 and maximizes its catalytic activity at Polycomb target genes (47). In the larval stage, although Pcl does not form complexes with PRC2, it mediates pleiohomeotic-dependent PRC2 target binding (56). These findings imply that Pcl plays at least two distinct roles in regulating the expression of Polycomb targets by interacting with different protein complexes and suggest that these interactions depend on the developmental stage or cell type. The picture may be even more complex because functional interactions of Pcl could potentially extend to PRC1. In addition, Pcl has been shown to display extensive colocalization with Pc and Polyhomeotic (Ph) proteins on the polytene chromosomes (43), and genetic interactions between Pcl and Pc have also been demonstrated (33). These observations suggest that Pcl may have multiple regulatory functions in mediating Polycomb repression.
Three homologs of the Drosophila Pcl gene have been identified in vertebrates, frogs, chickens, and mammals and are named, Pcl1 (also Phf1 or tctex3), Pcl2/Mtf2 (also M96; here, designated Pcl2) and Pcl3 (also Phf19) (12, 27, 32, 36, 64, 66). Overexpression studies using Xenopus and chicken embryos demonstrated that Pcl1 and/or Pcl2 mediates the repression of developmental regulatory genes such as En-2, Krox20, Hoxb9, and Shh (36, 64, 66). Murine and human orthologs of these genes have been shown to be the targets of PRC1 and PRC2 as revealed by genome-wide chromatin immunoprecipitation (ChIP) assays (8, 38). These findings imply that vertebrate as well as Drosophila Pcl family proteins may be part of the PcG complex. In support of this idea, hPHF1 was copurified with a subset of PRC2 and was shown to enhance the catalytic activity of PRC2 in a human cell line (10, 55). Unexpectedly, however, hPHF1 was also found to form complexes with proteins involved in the DNA damage response (25). Despite these recent advances, the molecular basis underlying the multiple functions of Pcl family proteins is not well understood. Although the biochemical properties of hPHF1 have been elucidated, a genetic approach using loss-of-function mutants of the mammalian Pcl family is still needed to clarify the genetic cascades involving the Pcl orthologs. A study using a Pcl2 gene trap allele suggested the involvement of Pcl2 in anterior-posterior (A-P) specification of the axis (63); however, the mechanistic basis of the involvement remains unclear. To address these questions, we have combined genetic and biochemical approaches to clarify the role of Pcl2 in Polycomb-mediated repression at the Hox gene cluster and Cdkn2a genes.
Data sets of RNA-Seq experiments performed by Motazavi et al. (45) were downloaded (http://woldlab.caltech.edu/~alim/RNAseq/). Results from two biological replicates were summed for three cell types (brain, liver, and muscle). Positions of the 15 Pcl2 exons (Mtf2 in the database) were identified using the BLASTN (3) program with Mtf2 as the reference sequence (GenBank accession number NM_013827.2) and mouse chromosome 5 (NCBI version 37) (see Fig. S1A posted at http://web.rcai.riken.jp/en/paper_figs/Li_Supp_figs_rev3.pdf).
A partial cDNA fragment that included the PHD fingers (amino acids [aa] 103 tp 327) of Pcl2 was isolated by yeast two-hybrid screening for Mel18-interacters and subcloned into the pGEX-4T-3 vector (Amersham Biosciences) to express a glutathione S-transferase (GST)-Pcl2 fusion protein. This fusion protein was used to generate hybridomas as described previously (29). Rabbit polyclonal antiserum was raised against a synthetic polypeptide corresponding to amino acids 20 to 36 of Pcl2 (RNQKTSASLNKLSLQDGC) conjugated to keyhole limpet hemocyanin. A purified IgG fraction of this antiserum was used for immunoprecipitation (IP) and immunofluorescence.
IP and Western blot analyses were performed as described previously (28). Antibodies used in this study are listed in Table Table11 (5, 23, 28). Nuclear proteins were extracted from nuclear pellets of K562 cells as described previously (15). Nuclear extracts were loaded onto Superose 6 columns (Amersham Biosciences), and 0.5-ml fractions were collected. Proteins from each fraction were concentrated by trichloroacetic acid-deoxycholate precipitation. Equal amounts of each sample were run on SDS-PAGE gels and subjected to Western blot analysis.
The gene-trapped embryonic stem (ES) cell line for the Pcl2 locus (M016A06) was obtained from the European Mouse Mutant Archive (24). The endogenous Pcl2 locus was disrupted by the insertion of a retroviral U3 β-galactosidase-Neor fusion gene (β-geo) gene trap vector. Its insertion into exon 4 results in disruption of the Pcl2 open reading frame (ORF) (see Fig. S2 posted at http://web.rcai.riken.jp/en/paper_figs/Li_Supp_figs_rev3.pdf).
To generate deletion mutant mice, Pcl2 homology arms, a 3.4-kb XbaI fragment that lies at the 5′ end of the exon 4 and a 2.4-kb KpnI/SalI fragment encompassing part of exon 5, were first introduced into the pPNT-loxP-Neo vector, in which the neomycin resistance gene (Neor) cassette was placed between two loxP sites (see Fig. S3 posted at the URL mentioned above). Homologous recombinants and germ line chimeras were isolated as described previously (1) (see Fig. S3B posted at the URL mentioned above). Heterozygous mutants were bred with CAG-Cre transgenic mice to delete the Neor cassette (54). Pcl2+/Δ mice were backcrossed six times onto a C57BL/6 background. Pcl2+/Δ mice were crossed with Phc2+/−, Mel18+/−, and Suz12+/− mice to generate compound mutants. All animal experiments were carried out according to the in-house guidelines for the care and use of laboratory animals of the RIKEN, Yokohama Institute, Japan.
We used Agilent Mouse Promoter ChIP-on-Chip Set 244K with chromatin immunoprecipitation (ChIP-chip) using antibodies against H3K27me3, Ring1B, or Pcl2. We used probes showing significant intensities above background (P < 10−7, Student's t test) and located from −4 kb to +4 kb of the transcription start site (TSS) of mouse genes. Methylated, ubiquitinated, or Ring1B- or Pcl2-binding positions were defined where the signal intensity of ChIP was 4 times (H3K27me3 and Pcl2) or 10 times (Ring1B) greater than that of input and when the statistical significance between the signals was less than 10−3. To reduce noise in these assays, we removed detected probes that did not have significantly strong signals from input around the 500-bp region flanking their respective positions. ChIP followed by high-throughput sequencing (ChIP-seq) analysis was performed as described previously (37). Since this study is aimed to specifically elucidate the role of Pcl2 at the Hox and Cdkn2a genes, whole data sets of these analyses will be published elsewhere.
Skeletal preparations were made from perinatal mice, and cleared skeletons were analyzed under a stereomicroscope as described previously (1). In situ hybridization of tissue sections was performed as described previously (1). RNA extraction from ES cells, reverse transcription, and PCRs were performed as described previously (21). The expression of Hox and Tbx3 genes was quantified relative to Hprt expression by real-time PCR analyses using an Mx3005P multiplex quantitative PCR system (Stratagene). Primer pairs used for reverse transcription-PCR (RT-PCR) analysis are listed in Table Table22.
Mouse embryonic fibroblasts (MEFs) were prepared as described previously (31). The Mel18-null, Mel18/ARF, and Mel18/p53 double-null MEFs were routinely maintained with Dulbecco's modified Eagle's medium (DMEM) supplemented with 20% fetal bovine serum (FBS), 1× nonessential amino acids (Invitrogen), and 1× penicillin-streptomycin (Invitrogen). In modified 3T9 assays, cells were passaged at 3-day intervals and counted after trypsinization, and the number of cells per dish was recorded. A total of 1 × 106 cells were replated in 60-mm diameter dishes every 3 days (31). In growth rate assays, cells diluted to 2 × 104 cells per 60-mm diameter dish were plated in duplicate, and thereafter individual cultures harvested each day were counted. The Pcl2 expression retrovirus was prepared using a retrovirus packaging system (PlatE), according to the manufacturer's instructions, and transduced into immortalized MEF Pcl2GT/GT cells.
Using database screening (GenBank accession number NM_013827.2), we identified a Pcl2 transcript that is predicted to encode a protein containing a Tudor domain, two PHD fingers in the N-terminal region, and a chromo-like domain structure in the C-terminal region (Fig. (Fig.1A,1A, upper scheme). In addition, we found a shorter transcript encoding a protein (AAI00341.1) lacking the Tudor domain (Fig. (Fig.1A,1A, lower scheme). To elucidate the origin of the Pcl2 isoforms, we examined the promoter and exon usage at the Pcl2 locus through data sets obtained from recent genome-wide analyses for H3K4me3 occupancy and RNA transcription (see Fig. S1A posted at http://web.rcai.riken.jp/en/paper_figs/Li_Supp_figs_rev3.pdf) (44, 45). Pcl2 appears to have a single predominant transcription start site, where there is extensive H3K4 trimethylation. We further found that usage of the 15 Pcl2 exons was not necessarily uniform and that exons 2, 3, 5, and 11 were differentially expressed among brain, liver, and muscle. This analysis suggests that the Pcl2 locus can give rise to several different transcripts, most likely due to alternative exon usage. We therefore inferred that the Pcl2 isoform lacking the Tudor domain could be produced from a transcript possessing exons 2 and 3 (see Fig. S1B posted at the URL mentioned above). In this case the nascent peptide initiating at the exon 1 start codon would prematurely terminate because of a frameshift; instead, productive translation would initiate at the start codon in exon 4, giving rise to a protein equivalent to AAI00341.1 (indicated by a green triangle in Fig. S1B posted at the URL mentioned above).
We then used Western blot analysis using a MAb produced against a GST-Pcl2 (aa 103 to 327) fusion protein to examine the expression of Pcl2 gene products. The MAb reproducibly detected strong bands of 67, 55, and 46 kDa and a much weaker 52-kDa band in whole-cell extracts of wild-type ES cells (Fig. (Fig.1B).1B). To clarify whether these bands represent isoforms of Pcl2, we measured the effect of small interfering RNA (siRNA)-mediated knockdown of Pcl2 in ES cells (Fig. (Fig.1A)1A) on the appearance of the bands. Expression of two different siRNA constructs reduced the intensity of all four bands proportionally (Fig. (Fig.1B).1B). These data indicate that Pcl2 is expressed primarily as 67-, 55-, 52-, and 46-kDa isoforms.
To further elucidate the origin of the Pcl2 isoforms, we used two mutant alleles for Pcl2. Insertion of a gene-trap vector into exon 3 (Pcl2GT) is expected to impair the expression of Pcl2 (Fig. (Fig.1C;1C; see also Fig. S2 posted at http://web.rcai.riken.jp/en/paper_figs/Li_Supp_figs_rev3.pdf). In Pcl2GT/GT ES cells, we found this insertion diminished band intensities (Fig. (Fig.1D).1D). In another allele, we deleted exon 4 and part of exon 5 by targeted mutagenesis (Pcl2Δ) (Fig. (Fig.1C;1C; see also Fig. S3 posted at the URL mentioned above). This deletion resulted in the expression of truncated Pcl2 transcripts in which exon 4 and 5 were skipped (X. Li, unpublished observations). These truncated transcripts are predicted to encode a Pcl2 protein lacking the Tudor domain and the first PHD finger. Consistent with this prediction, the 67- and 55-kDa bands were abolished in Pcl2Δ/Δ ES cells. Therefore, the 67- and 55-kDa isoforms represent proteins with and without the Tudor domain, respectively (Fig. (Fig.1A).1A). To validate this interpretation, we generated a polyclonal antibody against a synthetic polypeptide corresponding to part of the Pcl2 Tudor domain (aa 20 to 36). This antibody exclusively immunoprecipitated the 67-kDa band, confirming the presence of the Tudor domain in this isoform (Fig. (Fig.1E1E).
Hox cluster genes are evolutionarily conserved targets for Polycomb repression during A-P specification of the axis. We thus examined axial skeletal development in Pcl2GT/GT and Pcl2Δ/Δ mice. Both mutant mice survived to birth; however, most of the Pcl2GT/GT mice died before weaning. In contrast, Pcl2Δ/Δ mice were viable and fertile. Although postnatal viabilities of the Pcl2GT/GT and Pcl2Δ/Δ mice were different, similar axial skeletal alterations that are characteristic of posterior transformations were observed in both mutants, as shown in a previous study (65) (Fig. 2A and B). Briefly, ectopic ribs that associated with the seventh cervical vertebra (C7) were frequently seen in Pcl2GT/GT and Pcl2Δ/Δ mice (indicated by arrows in Fig. Fig.2A,2A, frames b, c, e, f, n, and o). Consistent with this malformation, sternums were shifted anteriorly (indicated by brackets in Fig. Fig.2A,2A, frames b and c). The odontoid process, which is normally a characteristic of the second cervical vertebra (C2), was frequently associated with the first cervical vertebra (C1) in the mutants (indicated by red triangles in Fig. Fig.2A,2A, frames h to j). Axial skeletal changes indicating posterior transformations were also manifested at thoracolumbar and lumbosacral transitions of Pcl2GT/GT and Pcl2Δ/Δ mice (Fig. (Fig.2B).2B). Therefore, Pcl2 proteins, particularly the 67- and/or 55-kDa isoforms, are required for A-P specifications of the axis, similar to other PcG proteins.
Next, we used in situ hybridization analysis to determine whether homeotic transformations in Pcl2Δ/Δ mice were accompanied by Hox gene derepression. The anterior boundaries of Hoxb4, Hoxb6, Hoxb8, and Hoxd4 expression in the developing sclerotome were shifted cranially in Pcl2Δ/Δ embryos at 11.5 days postcoitus (dpc) (Fig. (Fig.2C).2C). These results are consistent with homeotic transformations in the cervical and cervicothoracic boundary regions. We also found subtle but significant derepression of several Hox genes in Pcl2Δ/Δ ES cells (Fig. (Fig.2D).2D). These results show that Pcl2 regulates Hox gene expression.
Recent chromatin immunoprecipitation (ChIP) analyses revealed deposition of PRC2, PRC1, and H3K27me3 at the Hox cluster in developing embryos and ES cells. In parallel, Drosophila Pcl and human PHF1 were shown to be included in PRC2 in Drosophila and human cells, respectively (10, 47, 55, 56). Moreover, ChIP-seq studies reported significant overlap of genes bound by Pcl2 and Ezh2 in ES cells (41). These observations prompted us to hypothesize that Pcl2 mediates Hox repression via direct association with the Hox genes. By conventional ChIP assay using an anti-Pcl2 polyclonal antibody, we immunoprecipitated promoter regions of Hoxb4, Hoxb6, and Hoxb8 from the wild-type ES cells but not from Pcl2Δ/Δ ES cells, which were used as a negative control (Fig. (Fig.3A).3A). This result was further validated by ChIP-chip analysis using Pcl2GT/GT ES cells, which are reconstituted with the 67-kDa isoform of myc-tagged Pcl2. Here, an anti-myc antibody was used instead of the polyclonal antibody against the N-terminal peptide of Pcl2 (anti-N). In this experiment, too, we found considerable Pcl2 deposition around the transcription start sites (TSSs) of the Hox gene cluster, all of which were cooccupied by H3K27me3 and Ring1B (Fig. (Fig.3B;3B; also D. Yamada and H. Koseki, unpublished data). Based on these results, we extended our analyses to the entire Hoxd cluster by ChIP-seq analysis using the anti-Pcl2 polyclonal antibody (Fig. (Fig.3C)3C) (37). Indeed, the distribution profile of Pcl2 at the Hoxd cluster was similar to that of both Ezh2 and Ring1B. In particular, clear peaks that represented extensive binding of Pcl2, Ezh2, and Ring1B appeared concurrently around Hoxd9, Hoxd10, Hoxd11, Hoxd12, Hoxd13, and Evx2. Therefore, Pcl2 binds to Hox cluster genes, and its binding is correlated with local depositions of PRC2 and PRC1.
We went on to test whether Pcl2 expression and target binding depend on PRC2 because the expression of PRC2 components is known to be mutually dependent, and Pcl2 has already been reported as a PRC2-associated protein in mouse ES cells (21, 41, 63) (see Fig. S4 posted at http://web.rcai.riken.jp/en/paper_figs/Li_Supp_figs_rev3.pdf). In agreement with this view, we observed that the expression of Pcl2, as well as of Eed and Ezh2, was considerably decreased in Suz12−/− ES cells (Fig. (Fig.3D)3D) (21). Particularly, the 67- and 55-kDa isoforms of Pcl2 were almost absent. We further used ChIP assays to confirm this result. Consistent with the reduction of Pcl2 expression in Suz12−/− ES cells, Pcl2 binding was almost nonexistent (Fig. (Fig.3E).3E). It is worth noting that level of Pcl2 transcripts was not proportionally altered in comparison to that of Pcl2 proteins in Suz12−/− ES cells (Fig. (Fig.3F).3F). This suggests that Pcl2 expression is posttranscriptionally regulated in a PRC2-dependent manner. In contrast, the expression of Pcl2 and its binding to Hox genes were unaffected in Ring1B−/− ES cells (Fig. 4A and B). Moreover, the expression of PRC2 and PRC1 and their binding to Hox genes were not affected in Pcl2Δ/Δ ES cells (Fig. 4C and D). In summary, we conclude that Pcl2 binds to Hox genes in a PRC2-dependent manner.
We went on to test whether Hox repression by Pcl2 involves its interactions with PRC2. It is well established that Polycomb components cooperate to mediate Hox repression via physical interactions. Indeed, Drosophila Pcl was isolated as an enhancer for the extra sex combs (esc) mutation (33). Therefore, we examined the interactions between Pcl2 and Suz12 (an essential component of PRC2) mutant alleles. Since Suz12 homozygous mutants die around 8.5 dpc, heterozygous embryos were used. Notably, even the heterozygous mutants exhibited homeotic transformations of the axis, albeit to a lesser extent than Pcl2Δ/Δ mutants (Fig. 5A and B) (21). The skeletal phenotypes were compared between Pcl2Δ/Δ and Pcl2Δ/Δ; Suz12+/− mice at the newborn stage, and significant differences were identified around the cervicothoracic boundary. Anterior tuberculums, which are characteristic of the C6 vertebra, were associated instead with the C5 in Pcl2Δ/Δ; Suz12+/−. The C7 vertebrae were associated with perfect ribs, which were attached to the anteriorly shifted sternum. A prominent spinous process, which appeared on thoracic vertebra 2 (T2) in the wild-type, Suz12+/−, and Pcl2Δ/Δ pups, was frequently associated with the T1 in Pcl2Δ/Δ; Suz12+/− mutants. In summary, heterozygous loss of Suz12 synergistically enhanced the homeotic transformations of Pcl2 single mutants. Taken together, we conclude that Hox repression by Pcl2 involves its interaction with PRC2. Since Pcl and PHF1 are known to enhance the catalytic activity of PRC2, we inferred that Pcl2 contributes to Hox repression by regulating PRC2 activity. We did not, however, observe extensive changes in global and local levels of H3K27me3 in Pcl2Δ/Δ ES cells (Fig. 4C and D). Supporting this observation, binding of Ring1B to Hox genes was not changed despite Hox derepression in ES cells. As the molecular mechanisms that underlie Pcl2-dependent regulation of PRC2 are still unclear, we speculated that Pcl2-mediated Hox repression might involve PRC1.
We thus examined whether the Pcl2 mutation functionally interacts with mutations in genes encoding PRC1 components since previous reports have shown that compound mutants of PRC1 genes have a more severe phenotype than single mutants, suggesting the importance of such cooperativity (2, 6). We therefore compared the axial skeletal abnormalities in Pcl2Δ/Δ; Mel18−/− and Pcl2Δ/Δ; Mel18+/+ mice. Notably, no viable Pcl2Δ/Δ; Mel18−/− mice were seen at the perinatal stage, whereas the respective single homozygotes (Pcl2Δ/Δ; Mel18+/+ and Pcl2+/+; Mel18−/−) survived. Because of this, the skeletal phenotypes were examined at 16.5 dpc. Mel18−/− fetuses had phenotypes that were nearly identical to those in the Pcl2Δ/Δ fetuses, as reported previously (1). The Pcl2Δ/Δ; Mel18−/− double mutants showed more stringent transformations than either of the single mutants (Fig. 5C and D). Occipital bones were segmented, resulting in the formation of an ectopic arch. The C1 vertebrae were identical to the wild-type C2, and perfect ribs bridged an anteriorly shifted sternum and the C7. Interestingly, a hole reproducibly appeared at the central region of the scapula of Pcl2Δ/Δ; Mel18−/− mice, which was not seen in either of the single mutants. A similar synthetic defect in the scapula has been reported in Mel18−/−; Phc2−/− and Mel18−/−; Bmi1+/− mice due to the combined effects of the respective mutations (2, 28). Taken together, the genetic interactions between Pcl2 and Mel18 mutations synergistically enhanced skeletal abnormalities of the respective mutants and influenced the survival of late-gestation fetuses. To further clarify whether interactions between Pcl2 and Mel18 involve PRC1, we extended our genetic studies to Phc2, which is another PRC1 component and is known to physically and genetically interact with Mel18 (28). The Pcl2Δ/Δ; Phc2−/− fetuses exhibited skeletal phenotypes and late-gestational lethality very similar to the patterns seen in not only Pcl2Δ/Δ; Mel18−/− but also Mel18−/−; Phc2−/− and Phc1+/−; Phc2−/− mutants (Fig. 5C and D). These observations indicate a role for Pcl2 in reinforcing Hox repression that is mediated by PRC1. The cooccupancy of Hox genes by Pcl2 together with PRC2 and PRC1 suggests that Pcl2 positively regulates PRC2 and PRC1 to repress Hox genes during A-P specification of the axis.
We went on to examine the functional implication of Pcl2 to mediate Polycomb repression in other cellular contexts. Senescence is a fundamental cellular program that is activated under various forms of stress and acts to prevent further cell proliferation. When a population of cells is propagated in culture, they are exposed to various extrinsic and intrinsic stresses, and the cell population gradually stops dividing. Two definitive tumor suppressor pathways, ARF/MDM2/p53 and p16INK4a/Rb, have been shown to play critical roles in the induction of cellular senescence (59). Central mediators for cellular senescence, p19arf and p16Ink4a, are encoded by the Cdkn2a locus, which is known as an essential target of PRC1 (30). PRC1 represses p19arf and p16Ink4a in embryonic fibroblasts (MEFs) and thus protects MEFs to some extent from stress-induced premature senescence. This PRC1-dependent regulation of Cdkn2a has also been shown to be active in various stem and progenitor cells.
We first examined the binding of Pcl2 to the Cdkn2a locus by ChIP-seq analysis and found considerable Pcl2 deposition together with Ezh2, H3K27me3, and Ring1B in ES cells (Fig. (Fig.6A).6A). To test the role of Pcl2 in the regulation of p19arf and p16Ink4a expression, we generated MEFs from Pcl2Δ/Δ and wild-type littermates and compared progression of cellular senescence by a 3T9 assay (Fig. (Fig.6B).6B). To our surprise, growth of Pcl2Δ/Δ MEFs was similar to that of wild-type MEFs until passage 10, but then the cells failed to stop dividing and continued exponential growth over 50 passages while Phc2−/− MEFs prematurely senesced. We further found that Pcl2GT/GT MEFs similarly bypassed senescence (Fig. (Fig.6C).6C). To confirm that this effect is directly due to Pcl2 loss, we expressed Pcl2 in immortalized Pcl2GT/GT MEFs by retroviral transduction. Ectopic expression of Pcl2 indeed induced cellular senescence in immortalized Pcl2GT/GT MEFs (Fig. (Fig.6D).6D). Consistent with these phenotypes, Pcl2Δ/Δ MEFs did not exhibit morphological features characteristic of senescent MEFs at passage 12 (Fig. (Fig.6E).6E). We went on to investigate the expression of p19arf and p16Ink4a in Pcl2Δ/Δ MEFs and observed their strong repression at passage 10 (Fig. (Fig.6F).6F). This was accompanied by a reduction of p19arf and p16Ink4a transcripts (Fig. (Fig.6G).6G). Therefore, these results suggest that Pcl2Δ/Δ MEFs fail to undergo cellular senescence.
We next tested the functional involvement of Cdkn2a repression in terminating cellular senescence. To this end, we expressed p19arf or p16Ink4a in Pcl2Δ/Δ MEFs by retroviral transduction. Proliferation of Pcl2Δ/Δ MEFs was clearly inhibited by either p19arf or p16Ink4a (Fig. (Fig.6H),6H), thus suggesting that Pcl2 promotes cellular senescence by activating the expression of Cdkn2a genes in MEFs. To further test this hypothesis, we overexpressed Pcl2 in MEFs. We generated a transgenic mouse, in which expression of the 67-kDa isoform of Pcl2 could be induced by tamoxifen-dependent activation of ERT2-Cre (Fig. (Fig.7A).7A). Proliferation of these MEFs was impaired by tamoxifen treatment (Fig. (Fig.7B),7B), and the mitotic arrest was accompanied by activation of p19arf (Fig. (Fig.7C).7C). Therefore, we showed that Pcl2 plays a role as an activator of Cdkn2a expression, whereas PRC1 functions as a repressor in MEFs.
We went on to test whether this positive effect of Pcl2 on Cdkn2a transcription involved regulation of the catalytic activity of PRC2. We performed ChIP analysis to test H3K27me3 occupancy at the p16Ink4a promoter region in Pcl2GT/GT MEFs (Fig. 8A and B). Local H3K27me3 levels at this promoter were significantly increased in Pcl2GT/GT MEFs compared to levels the wild-type cells. Consistent with the transcriptional status of p16Ink4a, we also found decrease H3K4me3 in Pcl2GT/GT MEFs. Pcl2 therefore activates Cdkn2a expression, presumably by suppressing local catalytic activity of PRC2.
In addition, we examined whether Pcl2-dependent regulation of Cdkn2a involves PRC1. We generated Pcl2 and Phc2 double mutant MEFs and compared the level of cellular senescence to that observed in the single mutants (Fig. (Fig.6B).6B). Notably, Pcl2Δ/Δ; Phc2−/− MEFs, as well as Phc2−/− MEFs, stopped replicating at passage 5. Therefore, the Pcl2-deficient phenotype was strongly suppressed by loss of Phc2, which implies that Pcl2 functions require PRC1 in regulation of cellular senescence. Consistently, we found a considerable quantity of Ring1B retained at the p16Ink4a promoter in Pcl2GT/GT MEFs although its local level was significantly decreased compared to that of the wild type (Fig. 8A and B). Pcl2 thus profoundly modulates functional engagement of PRC2 and PRC1 to maintain proper expression of Cdkn2a under replicative stress.
In this study, we have investigated the role of Pcl2 at two canonical Polycomb targets, the Hox cluster and Cdkn2a genes, by combining genetic and biochemical approaches. We have first identified Pcl2 gene products and their functions in Hox repression during A-P specification by using two mutant alleles. The expression of Pcl2 and its binding to Hox genes were shown to depend on its physical association with PRC2. Interactions of mutant Pcl2 with Suz12 indicated the requirement of Pcl2 for PRC2 to exert its functions in Hox repression. Therefore, Pcl2 is a functional component of PRC2. It is also noteworthy, however, that the Pcl2 mutation affects PRC1-mediated repression of Hox genes as manifested by genetic interactions between Pcl2 and Mel18 or Phc2 although local depositions of H3K27me3 or Ring1B were not significantly altered in the Pcl2 mutants. Although the detailed underlying mechanisms are as yet unclear, our results strongly indicate that Pcl2 cooperates with both PRC2 and PRC1 to regulate the expression of Hox cluster genes during axial development (Fig. (Fig.8C).8C). In contrast, Pcl2 activates the expression of Cdkn2a genes once primary fibroblasts become predisposed to stress-induced senescence (Fig. (Fig.8C).8C). In this case, Pcl2 could primarily act by suppressing the local catalytic activity of PRC2, which may in turn enable bypassing of PRC1-mediated repression. Taken together, these results show that Pcl2 is required to modify the functional engagement of PRC2 and PRC1 complexes, presumably by sensing cellular circumstances. However, we could not unequivocally exclude the possibility that Pcl2 might have functions independent of PRC2 in regulation of cellular senescence.
Molecular mechanisms that link Pcl2 to such cellular circumstances remain as yet unknown. One intriguing possibility is that Pcl2 might recognize local chromatin cues by its Tudor domain and/or PHD fingers. By using the Pcl2Δ allele, we have shown a requirement for the 67- and 55-kDa isoforms of Pcl2 to mediate Hox repression in developing embryos and Cdkn2a activation in MEFs. Based on other recent studies, both Tudor and PHD finger domains have emerged as binding modules for methylated histone tails (7, 26, 34). Particularly, the PHD fingers of Pcl2 are predicted to recognize unmethylated or trimethylated histone H3K4 based on their respective primary structures (42, 52, 60, 65). Polycomb target genes, including Hox genes, are known to be bivalently marked by H3K27me3 and H3K4me3 in several cell types. Pcl2 may contribute to the Hox regulation via its recognition of the methylation status of H3K4 in developing embryos. In addition to detection of the methylation status of histone tails, Tudor and PHD finger domains have also been shown to contribute to recognition of RNA and phosphatidylinositols, respectively (7, 22, 26). We thus postulate that Pcl2 might use a combination of the Tudor domain and PHD fingers to discern chromatin circumstances and affect local activity of PRC2 via physical interaction. Identification of ligands for the Pcl2 Tudor domain and PHD fingers will be necessary to fully clarify this issue.
It is well known that the expression of Hox genes continues to be dynamically regulated in developing tissues after the initial establishment of their early expression domains (16). This implies that Polycomb-dependent Hox repression is plastic with the potential to be reactivated, presumably by various differentiation cues. In line with this, PRC1 has been shown to be linked to inductive signals mediating cellular differentiation, survival, and/or proliferation in cerebellar progenitors and ES cells (18, 39). Moreover, recent studies demonstrate that dynamic regulation of Polycomb activity orchestrated by the Jumonji (or Jarid2) protein included in PRC2 balances self-renewal and differentiation of ES cells (58). In this regard, it is worthy of note that Pcl2 is copurified with Jumonji (41). Pcl2, in collaboration with proteins such as Jumonji, could act as a module that mediates such plasticity of Polycomb repression by discerning chromatin circumstances induced by differentiation cues.
This project was in part supported by Genome Network Project from the Ministry of Education, Culture, Sports, Science and Technology of the Japanese Government (H.K.).
Published ahead of print on 8 November 2010.