Polycomb Mediated Epigenetic Silencing and Replication Timing at the INK4a/ARF Locus during Senescence

Background The INK4/ARF locus encodes three tumor suppressor genes (p15Ink4b, Arf and p16Ink4a) and is frequently inactivated in a large number of human cancers. Mechanisms regulating INK4/ARF expression are not fully characterized. Principal Findings Here we show that in young proliferating embryonic fibroblasts (MEFs) the Polycomb Repressive Complex 2 (PRC2) member EZH2 together with PRC1 members BMI1 and M33 are strongly expressed and localized at the INK4/ARF regulatory domain (RD) identified as a DNA replication origin. When cells enter senescence the binding to RD of both PRC1 and PRC2 complexes is lost leading to a decreased level of histone H3K27 trimethylation (H3K27me3). This loss is accompanied with an increased expression of the histone demethylase Jmjd3 and with the recruitment of the MLL1 protein, and correlates with the expression of the Ink4a/Arf genes. Moreover, we show that the Polycomb protein BMI1 interacts with CDC6, an essential regulator of DNA replication in eukaryotic cells. Finally, we demonstrate that Polycomb proteins and associated epigenetic marks are crucial for the control of the replication timing of the INK4a/ARF locus during senescence. Conclusions We identified the replication licencing factor CDC6 as a new partner of the Polycomb group member BMI1. Our results suggest that in young cells Polycomb proteins are recruited to the INK4/ARF locus through CDC6 and the resulting silent locus is replicated during late S-phase. Upon senescence, Jmjd3 is overexpressed and the MLL1 protein is recruited to the locus provoking the dissociation of Polycomb from the INK4/ARF locus, its transcriptional activation and its replication during early S-phase. Together, these results provide a unified model that integrates replication, transcription and epigenetics at the INK4/ARF locus.


Introduction
Cellular senescence is a fundamental cellular program that is activated after a finite number of cell divisions and operates to avoid further cell proliferation. In addition cellular senescence constitutes a tumor suppressor mechanism [1,2]. The tumor suppressor pathways, ARF/MDM2/p53 and p16 INK4a /Rb, have been shown to play critical roles in the induction of cellular senescence [3]. Studies on Polycomb group genes (Pc-G) have demonstrated that beside their function in controlling the expression of Hox genes, Pc-G play a central role in cell proliferation through the repression of the INK4/ARF locus [4,5,6]. This locus encodes both p16 INK4a , which prevents inactivation of the tumor suppressor RB, and p19 ARF , which stabilizes the tumor suppressor p53 [7,8]. Evidence supporting the direct control of the cell cycle by Pc-G proteins in vertebrates came from studies on mouse Bmi1 mutants. Bmi1 was first identified as a proto-oncogene that cooperates with c-Myc to promote the generation of mouse B-and T-lymphomas [9]. Mice lacking Bmi1 exhibit strong proliferative defects during lymphocyte development. In the absence of Bmi1, M33, or Phc2, primary embryonic fibroblasts (MEFs) are unable to progress into S phase, undergo premature senescence after only a few passages in culture and show an increased accumulation of the tumor suppressors p16 INK4a , p19 ARF and p15 INK4b [4,10]. Generation of Bmi1/ Ink4a/Arf compound mutant mice have provided genetic evidence that at least part of these defects are due to activation of the INK4a/ARF locus [11].
Pc-G and Trx-G proteins function in distinct multiprotein complexes which control transcription by altering the structure of chromatin, organizing it into either a ''closed'' or an ''open'' conformation. The Trx-G protein MLL1 mediates lysine-directed histone methylation [12,13]. Methylation on lysine-4 of histone H3 (H3K4me2 and H3K4me3) is associated with a permissive and transcriptionally active state of the chromatin [14]. Pc-G proteins are transcriptional repressors that functionally can be separated into at least two different complexes: the initiation complex, Polycomb complex 2 (PRC2), which in humans consists of EZH2, EED, and SUZ12; and the maintenance complex, PRC1, with the core proteins RNF2, HPC, and BMI1. Both PRC1 and PRC2 complex members have been linked to cell cycle control. EZH2 is the active component of PRC2 through its SET domain histone methyltransferase activity specific for Lys 27 (K27) of histone H3 and K26 of histone H1 [15,16]. It has been demonstrated that PRC2 is required for PRC1 binding to chromatin [17], presumably achieved through binding of HPC protein to H3K27me3 [15]. It has recently been shown that Polycomb proteins are bound to the INK4a/ARF locus and dissociated during senescence [18]. We have now extended this study and demonstrate that both Polycomb and the trithorax (Trx-G) member MLL1 are localized at this locus and importantly associated to the Regulatory Domain (RD) of the INK4/ARF locus identified as a DNA replication origin and as a global transcriptional regulator of the entire locus [19,20]. Also, we demonstrate that BMI1 interacts with the licencing factor CDC6. Finally, we show that the late timing of replication of the INK4a/ ARF locus in young proliferating MEFs shifts to an early replication timing in senescent and notably in PRC2 mutant MEFs. Together our results demonstrate that MLL1 and Polycomb group genes directly control the INK4a/ARF locus through epigenetic chromatin modifications and that the loss of repressive epigenetic marks both in senescent and Polycomb mutant cells leads to a shift of the replication timing of the INK4a/ ARF locus.

Mouse embryonic fibroblasts (MEFs)
Day-12p.c. wild type, M33 mutant and Bmi1 mutant embryos were mechanically dissociated into single cells and cultured in DMEM with 10% FCS and penicillin and streptomycin (Invitrogen, Breda, The Netherlands). MEFs were passaged every two to three days and viable cells were counted by Trypan Blue (Invitrogen) exclusion. Senescent cells were detected using the beta-Galactosidase Staining Kit (BioVision) following the manufacturer's recommendations.

Quantification of mRNA levels by Q-PCR
For reverse transcriptase Q-PCR (Q-RT-PCR) analysis, total RNA was extracted from MEFs with the RNeasy Mini Kit (QIAGEN) according to the manufacturer's protocol. cDNA was synthesized from 1 mg of total RNA using the QuantiTect Reverse Transcription Kit (QIAGEN). Q-PCR amplification of either mouse Ink4a transcripts or Arf was performed with the SYBR Green PCR Master Mix Kit (Applied Biosystems) in a 7500 Real-Time PCR System From Applied Biosystems. Primers used are given in Table 1. Q-RT-PCR analysis of Gapdh mRNA expression was performed as a positive control and for normalization.

Coimmunoprecipitations
NIH3T3 cells were transiently transfected with pCG-HAcdc6 using Lipofectamine TM 2000 Reagent (Invitrogen) following the manufacturer's instructions. 72 h after transfection CoIP were performed using Nuclear Complex Co-IP Kit (active motif) with mouse anti-HA (2367; Cell Signaling). Thymocytes were isolated from 5 to 7 weeks old mice and protein extraction and CoIP were performed using Nuclear Complex Co-IP Kit (active motif) with mouse anti-BMI1 (05-637; Upstate Biotechnology). Western blotting procedure and reagents for western blot analysis were previously described [4]. Antibodies used were mouse anti-CDC6 (sc-9964; Santa Cruz Biotechnology), goat anti-BMI1 (sc-8906; Santa Cruz Biotechnology). Replication timing analysis BrdU labeling, fixation in cold 70% ethanol, cell cycle fractionation by flow cytometry and isolation of BrdU-labeled DNA by immunoprecipitation were carried out as previously described [22].

Results and Discussion
Polycomb expression and cellular senescence in mouse embryonic fibroblasts To study cellular senescence we have used serial passaging of mouse embryonic fibroblasts (MEFs) cultured from 12-day-old C57BL/6 embryos. As already described, MEFs were able to undergo a limited number of population doublings (P12) before they senesced. Senescence was assessed by monitoring the endogenous B-galactosidase (B-gal) activity at pH 6 ( Fig. 1B). At this stage (P10), cells overexpress both p16 Ink4a and p19 Arf . As expected knock out cells derived from PRC1 members Bmi1 and M33 overexpress both p16 Ink4a and p19 Arf as soon as passage P3 [4] (Fig. 1A). Since EZH2 is required for Polycomb silencing we measured differences in expression levels of Ezh2 in young and senescent MEFs by Q-RT-PCR (Fig. 1C). Protein levels of EZH2 (not shown) in MEFs correlated with RNA levels; Ezh2 is more abundant in early passage (P3) than in senescent MEFs (P10) (Fig. 1C). In PRC1 mutant cells (M33, Bmi1) the expression level of Ezh2 is strongly diminished. Interestingly, expression level of the PRC1 member Bmi1 is not modified in senescent cells as compared to young proliferating cells (Fig. 1C) [18].

Polycomb proteins are localized at the RD INK4a/ARF
The RD INK4a/ARF has been identified as a putative DNA replication origin that assembles a multiprotein complex containing CDC6 and that coincides with a conserved non coding DNA element identified as a transcriptional regulatory element [20]. Whether PRC1 and PRC2 members bind directly to RD INK4/ARF had not been tested yet. In order to assess this question, we performed ChIP assays to examine whether EZH2 directly binds this transcriptional regulatory element. Oligonucleotide primers were designed in order to examine in young proliferating MEFs the distribution of EZH2 at the RD element and along the INK4a/ ARF locus. We found that EZH2 is bound to the first exon of ARF, exon 1b, and with a maximum peak to the shared exon of INK4a and Arf, exon 2, in early-passage MEFs (Fig. 1D). Interestingly, we found that both EZH2 and PRC1 members BMI1 and M33 are also localized at the RD origin of replication in young proliferating MEFs. In contrast, in senescent cells most of the bound EZH2 and M33 protein was lost at all the examined sites along the INK4a/ ARF locus ( Fig. 1D and E). A significant part of BMI1 is still retained at both exon 1b (p19 ARF ) and at the shared exon 2 upon senescence (Fig. 1F) however interestingly, BMI1 completely disappeared from RD (Fig. 1F). Since EZH2 and M33 are lost from the entire locus in senescent cells, these results suggest that BMI1 may bind to some parts of the locus (exon 1b and exon 2) in a manner that is independent of EZH2 and M33, but its binding to RD appears dependent on those two proteins. It has recently been shown that the BMI1 protein was dissociated from the locus at senescence [18]. While we do not explain this difference, the detected BMI1 protein bound at the locus correlates well with the fact that the expression level of Bmi1 is not modified in senescent cells. However, it was demonstrated using genome wide analysis that Polycomb domains can be segregated in two classes: the first occupied by both PRC2 and PRC1 (PRC1-positive) and the second specifically bound by PRC2 (PRC2-only) [23,24].
Several experiments indicate that H3K27 methylation by E(z) has a critical function in the establishment of transcriptional repression of PRC2 target genes. It has been demonstrated that the PRC2 complex containing E(z)/EZH2 is an active enzyme capable of methylating the histone H3 tail at lysine-9 (K9) and more importantly at K27 [15,25]. In E(z) Drosophila mutants the loss of functional E(z) induces a loss of Pc-G protein binding to polytene chromosomes [26]. Methylation of histone H3K27 by E(z) protein is strictly required for maintenance of HOX gene silencing in Drosophila [27]. Drosophila Polycomb (Pc or M33 in mouse), a core subunit of PRC1, selectively binds to histone H3 tail peptide trimethylated at K27, suggesting that H3K27me3 may contribute to targeting of PRC1 to HOX genes [28,29]. We therefore examined the status of methylation of histone H3 at the RD element and at the INK4a/ARF locus in young, senescent and Polycomb mutant cells. As shown in figure 2, H3K27me3 marks are lost from the RD element in senescent cells and from the shared exon 2 in senescent, Bmi1 2/2 and M33 mutant MEFs. The loss of the H3K27 repressive mark correlates with the higher transcription of Arf and Ink4a in senescent and Polycomb mutant cells (Fig. 1A). Histone acetylation is generally viewed as a central switch that allows exchange between permissive and repressive chromatin domains in terms of transcriptional competence. It was shown that the levels of p19 Arf are strongly upregulated in murine cells treated with histone deacetylase inhibitors (HDACis) [30]. Yet, examination of acetylation of histone H3(K9,K14) in actively cycling and senescent cells shows low level of H3 acetylation at the INK4a/ARF locus (Fig. 2). In contrast, we observe a strong enrichment of acetylated H3 at the Arf promoter in M33 and Bmi1 knockout MEFs (Fig. 2) similar to the effects observed after treatment with HDACis [30].

MLL1 is recruited to the INK4a/ARF locus during senescence
In Drosophila and in mouse the activity of Trx-G/MLL complexes is required to prevent Pc-G-mediated silencing of transcribed Hox genes [31,32,33,34]. MLL1 protein complexes catalyze the trimethylation of H3K4, which is generally associated with active transcription [35]. Accordingly, H3K4me3-modified nucleosomes are specifically enriched at the promoters of active genes [36]. We therefore monitored the binding of the MLL1 protein and the associated H3K4me3 positive transcriptional mark at the RD element and at the INK4a/ARF locus in MEFs during senescence and in Polycomb mutant cells. MLL1 was bound to the RD element and to both exon 1b and p16 INK4a /p19 ARF shared exon 2 in young cells (Fig. 3A). However, in both senescent and Polycomb mutant cells we observe a strong enrichment of MLL1 binding at the locus demonstrating that MLL1 participates to the transcription of Arf and Ink4a. Surprisingly, we did not observe a similar increase of the H3K4 methyl mark during senescence and in mutant cells (Fig. 3A). This positive mark is equally present in young, senescent or Polycomb mutant cells at the INK4a/ARF locus. Patterns of methylation at lysine 4 and 27 of histone H3 have been associated with gene activation and repression that are developmentally regulated and are thought to elicit the coordination of lineage specific gene expression programs [37]. Interestingly, in ES cells, the Polycomb Hox target gene promoters often display both H3K4me3 and H3K27me3 marks, and such regions, containing both repressing and activating chromatin modifications, were referred to as ''bivalent domains'' [38]. In stem cells, these bivalent domains may keep selected genes ''poised'' for activation. H3K27me3 is a rather stable modification, which could be progressively lost in the absence of PRC2, along with cell divisions [39]. However, studies in ES cells indicated that changes  Table 1). Wild type P3 (young), P10-12 (senescent), M332/2 (P4) and Bmi2/2 (P4) MEFs were subjected to ChIP assays using anti EZH2, BMI1 and M33 antibodies. DNA enrichment was calculated as described in Materials and Methods. Bars represent the mean+/2s.d. of quantifications from two to four separate immunoprecipitations analyzed in triplicate. doi:10.1371/journal.pone.0005622.g001 in chromatin associated with Hox gene activation are likely to occur promptly, and involving an appropriate demethylase activity. We therefore monitored the expression of both Jmjd3 and Utx H3K27 histone demethylases. As shown in figure 3B expression of Utx is not modified in young, senescent or M33 mutant cells. However, transcription of Jmjd3 is significantly induced in senescent MEFs. These results strongly suggest that the upregulation of Jmjd3 (Fig. 3B) and the downregulation of Ezh2 (Fig. 1C) are critical determinants of the transcriptional activation of the INK4/ARF locus during senescence. It has been shown that UTX can interact with components of the MLL2 complex [40,41]. This physical association between enzymes removing the H3K27me3 repressive mark, on the one hand, with protein complexes promoting the deposition of the active H3K4me3 mark, on the other hand, suggests that both activities are required for a rapid and stringent response of target genes. MLL1 is cleaved by Taspase1, generating an N-terminal and a C-terminal fragment, which can heterodimerize in vitro [42,43]. In Drosophila it was demonstrated that TRX-N is present at thousand genomic sites, where no Pc-G binding can be observed. However, it was shown that TRX-C is strongly bound at Pc-G binding sites [44]. These results suggest the C-terminal part of TRX is specifically linked to Pc-G function. It was suggested that Pc-G proteins might repress transcription by anchoring the C-terminal portion of TRX at Polycomb response elements (PREs/TREs) or that constitutive TRX-C binding at PREs/TREs might allow Pc-G target genes to switch their state upon transcriptional induction [44].

Physical and Functional Association of BMI1 and CDC6 in replication control
The identification of a DNA replication origin (RD) adjacent to INK4b [19,20] and its high degree of sequence conservation led to ask whether this domain might contribute to the regulation of transcription. Interestingly, overexpressing and loading CDC6 to the RD element, results in the transcriptional repression of all three genes in the INK4b-ARF-INK4a locus [45] leading to increased foci formation and enhanced transformation by oncogenic RAS. Importantly, silencing is accompanied by the recruitment of histone deacetylases and increased methylation of histone H3 on lysine 9 (H3K9), which are hallmarks of heterochromatin. We have demonstrated that both components of the PRC2 and PRC1 complex are localized at the RD element. This prompted us to search for possible physical interactions between the replication complex containing CDC6 and the components of Polycomb complexes. Co-Immunoprecipitation experiments using an antibody against BMI1 demonstrated that CDC6 is associated in a complex with BMI1 (Fig. 4B). Moreover using thymocytes we show that the CDC6-BMI1 interaction occurs in wild type non-transfected cells (Fig. 4C) indicating that the CDC6-BMI1 interaction is not due to the forced expression of CDC6 in transfected MEFs. In order to test if BMI1 is required for Ink4a/Arf repression mediated by CDC6, we transfected CDC6 in Bmi1 knock out and wild type MEFs. As already described [20], the forced expression of Cdc6 in wild type MEFs decreased the protein levels of ARF and INK4a (Fig. 4A,D). However, overexpression of Cdc6 in mutant Bmi1 cells failed to mediate Arf or Ink4a repression (Fig. 4A,D) demonstrating that CDC6 induced repression of the INK4a/ARF locus is dependent on Polycomb function.
Next we asked whether the localization of Polycomb proteins at an origin of replication together with the replication machinery [19] could affect replication of the locus. To investigate whether the induction of senescence could result in a modification of the replication timing of the INK4a/ARF locus we examined young proliferating (P3), pre-senescent (P7) and Polycomb M33 mutant (P3) MEFs. The replication timing was assessed using a PCR based approach [22,46]. Non-synchronized cells were pulse labeled with 59Bromodeoxyuridine (BrdU), stained with propidium iodide (PI) and sorted according to DNA content by flow cytometry (Fig. 5). Newly synthesized DNA was isolated by immuno-precipitation with anti-BrdU antibody. As shown in figure 5 the exon 1b (p19 ARF ) is late replicating in young cells which do not express the Arf or Ink4a genes; whereas this region becomes early replicating in pre-senescent and in M33 mutant cells when the Arf or Ink4a genes are expressed. In higher eukaryotes, it has been observed that the time of replication and transcriptional activity are often correlated; genes which are late replicating are not expressed while transcriptionally active regions are early replicating [47]. In addition, when the transcriptional stage of a gene switches from an active to inactive state, replication timing shifts from early-to late replicating. Importantly, it has recently been demonstrated that histone modifications at an origin of replication serve as a binary switch for controlling the timing of replication of the Beta-globin locus in human [48]. This replication switch is also observed in human diseases such as the fragile X syndrome. FMR1 silencing by the CGG expansion was shown to be mainly attributed to epigenetic regulated transcriptional silencing [49]. The Fmr1 gene normally transcribed is replicated early whereas it becomes silent and late replicating in patients [50,51]. In yeast it was recently shown that Swi6, an S. pombe counterpart of heterochromatin protein 1 (HP1), is required for early replication of the pericentromeric region and the mat locus [52]. In our study we show that in proliferating MEFs the INK4a/ARF locus is silent and late replicating whereas in Polycomb mutant the locus tends to be early replicating and expressed. It has recently been demonstrated in the Encode project that the H3K27me3 mark shows a positive correlation with late replication of large DNA segments [53]. We have demonstrated in senescent and Polycomb mutant cells that the ''bivalent'' domain at the INK4a/ARF locus (H3K27me3 and H3K4me3) is resolved and the locus remains only enriched in H3K4me3 positive marks correlating with the recruitment of MLL1 protein. Jmjd3 overexpression in senescent cells could indicate that this histone demethylase participates in removing the H3K27 marks at the INK4A/ARF locus. The epigenetic modifications could be responsible for the observed replication-timing shift at senescence (Fig. 6). Together, our results demonstrate that MLL1 and Polycomb group genes directly control the INK4a/ARF locus through chromatin epigenetic modifications and that the loss of the repressive epigenetic marks both in senescent and Polycomb mutant cells at an origin of replication leads to a shift of the replication timing of the locus.