THIS ARTICLE HAS BEEN RETRACTED 
Epigenetic regulator polycomb group protein complexes control cell fate and cancer

Abstract

The chromatin‐associated Polycomb group (PcG) proteins were first identified in genetic screens for homeotic transformations in Drosophila melanogaster. Besides body patterning, members of the PcG are now known to regulate epigenetic cellular memory, stem cell self‐renewal, and cancer development. Here, we discuss the multifarious functions of the PcG family, isoforms of protein complexes, and its enzymatic activities, for example histone methylation, links to DNA methylation, its phosphorylation status, H2A mono‐ubiquitination, SUMOylation, and links to non‐coding RNA. We also discuss the function of cytosolic PcG complexes as a regulator of receptor‐induced actin polymerization and proliferation in a methylation‐dependent manner. We propose that the functional versatility of PcG protein complexes contributed significantly to the complexity of heritable gene repression mechanisms, signal transduction, and cell proliferation in cancer development. (Cancer Sci 2008; 99: 1077–1084)

Epigenetic memory of gene expression profiles is thought to be vital for the development of multicellular organisms. Polycomb group (PcG) proteins were firstly identified in Drosophila melanogaster (D. melanogaster) as factors necessary to preserve cell‐fate decisions throughout embryogenesis by silencing Hox genes in a body‐segment‐specific manner.^{( 1} , ^{2 )} Now accepted as a large family of chromatin‐associated proteins conserved from plants to humans, the PcG is involved in many cellular memory processes including body patterning, X inactivation in female mammals,^{( 3 )} and vernalization in plants.^{( 4 )}

Here, we review Polycomb group (PcG) proteins, one class of epigenetic regulators, which have connections with all three epigenetic mechanisms, for example DNA methylation, histone modification (histone code), and non‐coding RNA, and thus function as transcriptional regulators that silence specific sets of genes through chromatin modification.

PcG proteins are structurally and functionally diverse and form large multimeric complexes of two general types: Polycomb repressive complex‐1 (PRC1) and PRC2.^{( 5 )} These complexes modify histone tails and are believed to cooperate in transcriptional repression of target genes by altering higher‐order chromatin structure.^{( 6} , ^{7 )}

Although primarily known for their role in maintaining cell identity during the establishment of the body plan, several mammalian PcG members have now been implicated in the control of stemness in embryonic stem (ES) cells/hematopoietic stem cell (HSC)/neural stem cells, cellular proliferation, and neoplastic development.^{( 8} , ^{9 )}

PcG genes: Conservation and diversification of PcG homologs

Vital domains of PcG homologs are highly conserved between evolutionarily distant organisms and among paralogs in a given organism (more than 75% amino acid similarity). However, outside of key domains, PcG proteins have diverged significantly from their Drosophila counterparts and from their paralogs. Accumulating evidence suggests that PcG paralogs have specialized expression patterns and functions (see below).^{( 10 )} An important challenge is to understand the functional significance of developmentally regulated expression of paralogs and how this impacts PRC composition, genomic targeting and/or mechanism of transcriptional repression. To illustrate the functional diversification of PcG paralogs, here we focus on only homologs of Drosophila polycomb (Pc), posterior sex combs (Psc), dRing, and enhancer of zeste (E(z)). (Table 1, Fig. 1)

Table 1.

Nomenclature of polycomb group proteins

Proteins			Protein domains	Functions
D.melanogaster	Human	Mouse
PRC2 complex
ESC	EED	EED	WD40 repeats	EED is required for the HMTase activity of EZH2 within the PRC2 complex
E(z)	EZH1	EZH1/ENX2	SET domain	Histone methyltransferase
	EZH2	EZH2/ENX1
SU(z)12	SUZ12	SUZ12	Zinc‐finger	Like EED, SUZ12 is a critical minimal component required for enzymatic activity of the PRC2 complexes. In Drosophila, Su(z)12 drives nucleosome binding by PRC2.
PRC1 complex
Pc	CBX2/HPC1	CBX2/M33	Chromodomain	Methyl‐lysine binding.
				Chromo domain protein, reader of the H3K27me3 mark.
				SUMO E3 ligase
	CBX4/HPC2	CBX4/MPC2
	CBX6	CBX6
	CBX7	CBX7
	CBX8/HPC3	CBX8/PC3
PH	EDR1/HPH1/PHC1	EDR1/MPH1/Rae28/PHC1	Zinc‐finger SPM domain	Polyhomeotic proteins are SAM (sterile alpha motif) and Zn‐finger proteins that are stoichiometric components of PRC1 and required for silencing.
	EDR2/HPH2/PHC2	EDR2/MPH2/PHC2
	EDR3/HPH3/PHC3	EDR3/MPH3/PHC3
RING	RING1/RNF1/RING1A	RING1/RING1A	RING‐finger domain	Ubiquitin E3 ligase.
	RING2/RING1B	RNF2/RING1B
PSC	BMI1/PCGF4	BMI1/PCGF4	RING‐finger domain	RING domain protein that is essential for the H2A ubiquitynation function of RING1B
	MEL‐18/RNF110/PCGF2	MEL‐18/RNF110/ZFP144
	MBLR/RNF134/PCGF6	MBLR/RNF134/PCGF6
Others
PHO	YY1	YY1	Zinc‐finger	DNA‐binding
PHO‐like	YY2	YY2	Zinc‐finger	DNA‐binding
SCM	SCML1	SCMH1	Zinc‐finger SPM domain
dSfmbt	L3MBTL2	L3MBTL2	Zinc finger
			Mbt repeat

Owing to the large number of putative PcG homologs in vertebrates, this table is not exhaustive.

WD repeats are a conserved domain that usually ends with Trp Asp (WD).

The SET domain is named after the first letter of the three proteins in which the domain was found – SU(VAR)3–9, E(z), and TRX.

SPM refers to the presence of this domain in SCM, PHO, and MBT proteins.

A RING finger is conserved cysteine‐rich domain named after the really interesting new gene (RING) in which it was first identified. D. melanogaster; Drosophilia melanogaster; ESC, extra sex combs; E(z), enhancer of zeste; PC, polycomb; PCL, polycomb‐like; PH, polyhomeotic; PHD, plant homeodomain; PHO, pleiohomeotic; PSC, posterior sex combs; SU(z), suppressor of zeste; SCM, sex combs on midleg; L3mbtl, lethal^{( 3 )} malignant brain tumor‐like protein.

Figure 1.

(a) Schematic representation of known core members of polycomb repressive complex‐1 (PRC1) and PRC2 complexes. Drosophila proteins are shown as colored ovals; mouse homologs of these proteins are listed adjacently. (b) Epigenetic gene silencing by polycomb protein complexes. Binding of polycomb repressive complex‐2 (PRC2) initiation complex to polycomb group (PcG) target genes induce EZH2‐mediated methylation of histone H3 lysine 27 (H3K27me3). PRC1 is able to recognize the H3K27me3 mark through the chromodomain of polycomb (Pc). This interaction might bring neighboring nucleosome into the proximity of the PRC2 to facilitate widespread methylation over extended chromosome region, and other enzymatic activity (see text). CtBP, C‐terminal binding protein; DNMT, DNA methyltransferases; ncRNA, non‐coding RNA; Su(z), suppressor of zeste.

Pc homologs.  Vertebrate have between three and five Pc homologs (known as chromobox (Cbx)), which all have highly conserved chromo domain (CD) and Pc boxes. However, paralogs differ greatly in length and in the presence of other domains and motifs; these features might contribute to differential function.

Mammalian Cbx proteins differentially effect cell‐cycle regulation. The overexpression of Cbx7 or Cbx8 (also known as HPC3/Pc3) but not Cbx4 (HPC2/mPc2) bypasses replicative senescence in human and mouse fibroblasts owing in part to the repression of the INK4a‐ARF locus.^{( 11 )} However, they do so in the context of distinct PRC1 complexes; Cbx8 depends on Bmi1 (Psc homolog, see below) to bind the INK4a‐ARF locus and to extend the lifespan, whereas the Cbx7‐mediated bypass of senescence is Bmi1‐independent.^{( 11 )} It remains to be shown if Cbx7‐containing PRC1 complexes contain another Psc homolog required for Cbx7‐mediated repression of this tumor suppressor locus.

Cbx proteins also interact with non‐PcG proteins. Cbx4 is the only member of the family that binds the transcriptional corepressor C‐terminal binding protein (CtBP). CtBP interacts with transcription factors that might target Cbx4‐containing complexes to specific DNA sequences. Cbx4 is also unique among Pc homologs as an E3 SUMO ligase.^{( 12 )} The full range of Cbx4 SUMO targets is unknown, but the SUMOylation of several transcriptional regulators, including CtBP, is enhanced by Cbx4.^{( 12} , ^{13 )} Additionally, recent biochemical data suggest that the five mammalian Cbx proteins have different histone‐binding preferences: the Cbx CD bind differentially to H3K27me3 and H3K9me3, unlike Drosophila Pc CD, which prefers H3K27me3.^{( 14 )}

Psc homologs.  Mel‐18 and Bmi1 (two of six Psc homologs in mammals) are also likely to be non‐redundant paralogs, despite their 63% amino acid sequence identity. Bmi1‐ and Mel‐18‐deficient mice display similar but unique phenotypes^{( 15} , ^{16 )} and only ~30% of Bmi1‐regulated genes were found to be coregulated by Mel‐18 and vice‐versa.^{( 17 )} In some cases, Mel‐18 and Bmi1 have opposite effects on cell‐cycle regulation. Retroviral insertion of Bmi1 into Eµ‐myc transgenic mice accelerated lymphomagenesis (oncogenic),^{( 18 )} whereas cells from transgenic mice overexpressing Mel‐18 arrest in G1/S of the cell cycle (tumor‐suppressive).^{( 19 )} Moreover, stable Mel‐18‐knockdown fibroblasts induce tumor formation when injected into mice.^{( 19} , ^{20 )}

However, the relationship between these two paralogs is complicated. A recent study using stable knockdown (siRNA) of Bmi1 and Mel‐18 in cancer cell lines revealed similar, not opposing consequences on cell growth. ^{( 17 )} Furthermore, Mel‐18 can act as a tumor suppressor by repressing Bmi1 in cancer cells.^{( 21 )} Although Bmi1 and Mel‐18 form stable PRC1 complexes with similar composition, only Bmi1 has been shown to positively regulate Ring1B ubiquitination of H2AK119.

The function of individual PRC1 components in higher organisms is poorly understood. Previous studies demonstrated that while Mel‐18 and Bmi1 knockout mice have similar, though not identical, phenotypes^{( 15} , ¹⁶ , ²² , ^{23 )} the double knockout mouse embryos exhibit a more profound phenotype.^{( 24 )} This suggests that while these two proteins have independent functions, they may also exhibit some functional redundancy.

In keeping with this notion, we and others^{( 25 )} (and in unpublished data by M.K.) have found that MEL‐18 (Mel‐18) is part of a multiprotein complex with RING1/2 (Ring1A/B), HPH2a (mPh2), and CBX8 (mPc3), all of which also associate with BMI1 (Bmi1).^{( 26} , ^{27 )} However, since MEL‐18 and BMI1 components are not found together in the same complex, they probably participate in distinct polycomb repressor complexes (melPRC1 and bmi1PRC1, respectively). This may explain the non‐overlapping functions of Mel‐18 and Bmi1 defined in genetic experiments. Therefore, the presence of different Psc homologs in PRC1 might change its catalytic activity in vivo.

dRing homologs.  Mammalian homologs of dRing, Ring1A (also known as Ring1) and Ring1B (Ring2), also exhibit some functional divergence (Fig. 1). Although they share long stretches of high conservation, Ring1A‐ and Ring1B‐deficient mice have different phenotypes.^{( 28 )} Mice heterozygous for Ring1A exhibit classic homeotic transformations and skeletal defects, whereas Ring1B heterozygous mice show no skeletal phenotype.^{( 28 )} However, Ring1B is essential for normal gastrulation, and null embryos do not survive past embryo day 10.5.^{( 28 )} The differential severity of these phenotypes correlates with the extent of ubiquitinated‐H2A (H2Aub) depletion in these knockouts. Global H2Aub is drastically reduced in Ring1B‐ but not Ring1A‐null ES cells.^{( 29 )}

Biochemical study supports the differential functions revealed in these mouse studies. Full‐length recombinant Ring1B but not Ring1A has ubiquitin‐ligase activity for H2A (unlike a full‐length protein, N‐terminal Ring1A has E3 activity^{( 30 )}) and Ring1B association with Ring1A enhances this activity.^{( 27} , ^{31 )} Although global H2Aub is drastically reduced in Ring1B‐null cells, H2Aub staining is maintained on the inactive X chromosome;^{( 29 )} only in Ring1B/Ring1A double‐knockout cells is H2Aub lost from this structure, revealing functional redundancy of Ring1A and Ring1B in some contexts.^{( 29 )} These indicate more intricate mazes in vivo than the in vitro studies suggest.

E(z) homologs.  The mammalian organisms that we focused on have two E(z) homologs: Ezh1 and Ezh2 (Fig. 1, Table 1). Little is known about the functional differences between these paralogs in mammals, but the ancestral E(z) gene also expanded in plant lineages. Arabidopsis has three E(z) homologs: Medea (MEA), curly leaf (CLF), and swinger (SWN) with largely non‐overlapping patterns of expression. These paralogs proteins regulate different developmental processes and non‐identical sets of genes.^{( 32} , ³³ , ^{34 )} Although similar in domain structure, ectopic expression of MEA or SWN cannot rescue CLF‐deficient plants, suggesting some functional divergence of paralogs.^{( 33 )} Additionally, phylogenetic analysis of E(z) homolog SET domains in a variety of angiosperm species revealed that Arabidopsis CLF, SWN, and MEA proteins cluster into three separate classes. This suggests that E(z) homologs have diverged and become fixed in diverse plant species, presumably for specialized, beneficial functions.^{( 33 )}

Polycomb protein complexes: Isoforms and variants

PcG proteins are functionally diverse and form large multimeric complexes of two general types: PRC1 and PRC2.^{( 5 )} These complexes modify histone tails and are believed to cooperate in transcriptional repression of target genes by altering higher‐order chromatin structure.^{( 6} , ^{7 )} Here we will describe Drosophila PRC as well as mammalian PRC to be discussed below.

PRC2 contains four core proteins: Enhancer of zeste (E(z)), extra sex combs (Esc), suppressor of zeste 12 (Su(z)12), and nucleosome remodeling factor 55‐kDa subunit (Nurf55) (Fig. 1). E(z), a histone methyltransferase, catalyzes the trimethylation of histone H3 lysine 27 (H3K27me3) via its SET domain.^{( 35 )} Interestingly, E(z) is catalytically inactive in vitro unless associated with other PRC2 complex subunits, which are responsible for either binding histones/nucleosomes or enhancing enzymatic activity.^{( 35} , ³⁶ , ^{37 )} Although H3K27me3 is associated with PcG‐mediated transcriptional repression and PRC1 binding (see below), the targeting, and inheritance of this covalent modification remains unclear.

Recently isoforms of PRC2/3/4 were described based on the subunit isoforms of Eed.^{( 38 )} The human Eed protein exists in four different isoforms that arise from alternate translation initiation sites from the same mRNA, generating different Ezh2‐containing complexes.^{( 39} , ^{40 )} The largest form (Eed1) is predominantly present in the PRC2, whereas the two shortest forms (Eed3 and Eed4) are present in the complex named as PRC3. Neither PRC2 nor PRC3 contain the Eed isoform 2 (Eed2). The PRC3 complex exclusively targets methylation of histone H3‐K27, and the activity is repressed in the presence of histone H1. PRC2 also methylates histone H3‐K27 but in the presence of histone H1, PRC2 methylates both H3‐K27 and H1‐K26.^{( 40 )}

The characterization of PRC4, a complex containing the Eed isoform 2 (Eed2), which appears to be present only in undifferentiated pluripotent cells as well as in cells that have lost their ‘normal’ regulation, has also been reported. PRC4 contains the NAD⁺‐dependent histone deacetylase SirT1 together with other subunits of the PRC2/3 complexes, to be overexpressed in breast, colon, and prostate cancers. Importantly, previous studies have indicated that EzH2 is overexpressed in late stages of prostate cancer.^{( 41 )} Overexpression of EzH2 results in the accumulation of the PRC4 complex.^{( 38 )} Taken together, the nomenclature of PRC3 and PRC4 should be considered as a variant form of PRC2.

Core PRC1 is composed of Pc, dRing, Psc, and Polyhomeotic (Ph).^{( 5 )} Pc has an N‐terminal CD and a C‐terminal Pc box. CD are found in many chromatin‐associated proteins and are well‐characterized methyl lysine‐binding modules.^{( 42 )} Specifically, the CD of Drosophila Pc binds most strongly to H3K27me3, the modification generated by PRC2.^{( 43} , ^{44 )} The Pc box is a ~15‐amino acid motif necessary for transcriptional repression of target genes and for interaction with dRing, the catalytically active subunit of PRC1.^{( 45 )} dRing, named for its RING‐type zinc finger, is an E3 ubiquitin ligase that monoubiquitylates histone H2A at lysine 119 (H2AK119ub).^{( 27 )} This modification, along with H3K27me3, is important for PcG‐mediated gene repression.^{( 27} , ^{29 )} Psc is also a RING finger protein, and a murine Psc homolog enhances the ubiquitination activity of dRing homologs in vitro and in vivo. ^{( 30} , ^{31 )} The precise function of Ph in PRC1 complexes remains to be characterized, but it has been speculated that Ph might promote the spreading of PcG complexes.^{( 46 )}

Purified PRC also include other PcG proteins, different PcG isoforms, DNA‐binding proteins, and chromatin‐modifying enzymes, such as histone deacetylases.^{( 38} , ⁴⁰ , ⁴⁷ , ^{48 )} Such factors probably contribute to transcriptional repression by modifying chromatin structure or by stabilizing PRC complexes at their target genes. In Drosophila, sequence‐specific DNA‐binding proteins, such as Pleiohomeotic (Pho), interact with PRC subunits and can induce PRC binding to DNA regulatory elements known as PRE (PcG response elements).^{( 5} , ^{49 )} Although PRE are necessary for PcG recruitment in Drosophila, such elements have yet to be identified in vertebrates or plants.

In mammalians, as mentioned above, we and others^{( 25 )} (and in unpublished data by M.K.) have found that MEL‐18 and BMI1 components are not found together in the same PRC1 complex; they probably participate in distinct polycomb repressor complexes (melPRC1 and bmi1PRC1, respectively). This mutually exclusive relationship of these two complexes may explain the non‐overlapping functions of Mel‐18 and Bmi1 defined in genetic experiments.

PRC and their versatility

Histone methylation and demethylation.  An enzymatic subunit of the polycomb PRC2 complex, EZH2, methylates H3K27 to mediate gene silencing. Although originally identified as Hox repressors, PRC2‐complex proteins occupy many developmentally important genes and are involved in mammalian X chromosome inactivation, germline development, stem cell identity, cell‐cycle regulation, and cancer. H3K27me3 contributes to the recruitment and/or stabilization of the polycomb PRC1 complex on chromatin, which in turn mediates gene silencing.^{( 49 )}

The recently discovered H3K27 demethylases, UTX and JMJD3, contain highly homologous JmjC domains; another member of the same evolutionarily conserved JmjC subfamily is UTY. In addition to the JmjC domain, both UTX and UTY, but not JMJD3, contain tetratricopeptide motifs, predicted to mediate protein–protein interactions. In vitro, UTX and JMJD3 catalyze transition of H3K27me3 and H3K27me2 to H3K27me1 on bulk histone substrates, with H3K27me3 being a preferred substrate.^{( 50} , ⁵¹ , ⁵² , ^{53 )} In contrast to bulk histones, activity of the recombinant UTX and JMJD3 on the nucleosomal substrates is very weak, indicating that they require protein cofactors for chromatin recognition.^{( 53 )}

Interestingly, recent genomic analysis revealed an unexpected result: in contrast to H3K27me3, which marks silent loci, H3K27me1 is enriched downstream from transcription start sites of active genes.^{( 54 )} Could this be a signature of UTX/JMJD3 present at these genes counteracting methylation by the PRC2 complex? Control of the equilibrium between tri‐methylation and demethylation of H3K27 by PRC2 and UTX/JMJD3 will tell us more dynamic nature of epigenetics in our body.^{( 55 )}

Link to DNA methylation.  Recent studies have shown a direct interaction between DNA methyltransferases (DNMT) and the polycomb protein EZH2 that hints at how these complex systems might collaborate to set up cellular memory.^{( 56 )}

These results reveal that EZH2, as part of the PRC2 complexes, can physically recruit DNMT to certain target‐genes and that this process is essential for silencing the genes. Surprisingly, reducing the levels of any of the DNMT separately results in similar de‐repression of the EZH2 target‐genes. This is unexpected because the different DNMT by and large seem to fulfill distinct roles in development and cellular viability.^{( 57 )} It is possible that various EZH2‐containing complexes might exist that can interact with the “de novo” DNMT (DNMT3a and 3b) or the “maintenance” DNMT (DNMT1) under different circumstances. This would allow PcG complexes to take part in initiating and maintaining transcriptional programs depending on the chromatin environment and stimuli.

DNMT recruitment to EZH2 target‐genes not only adds another layer of repression to ‘lock in’ the silent state by methylating the surrounding DNA, but it might also help to recruit PRC1 proteins.^{( 8 )} It is also possible that EZH2 acts independently of PRC1 by recruiting corepressors that associate with DNMTs. In both scenarios, DNA methylation of EZH2 target‐genes can lead to the recruitment of proteins that bind to the methyl‐DNA marks, and the formation of more repressive complexes. So are all EZH2 target‐genes also silenced by DNA methylation, and are they overlapping with or separate from genes repressed by the EZH2–PRC1 interaction? Notably, the mechanism by which PcG‐induced silencing is inherited is still not known. Perhaps, for a select group of target genes, EZH2‐coupled DNA methylation can participate in preserving PcG‐related cellular memory during DNA replication, through the maintenance DNMT.^{( 56 )}

Cytosolic function of PcG complexes connected to cytoskeletal actin dynamics in receptor‐mediated signal transduction.  Intriguingly, Su et al. have recently added to our knowledge of PcG by characterizing a role of a cytosolic complex containing Ezh2, a protein with known histone‐methyltransferase (HMT) activity, demonstrating that Ezh2 methyltransferase activity is necessary for receptor‐mediated signals leading to actin reorganization and proliferation in both T lymphocytes and fibroblasts.^{( 58 )} Su et al. provide a novel means by which receptor‐proximal protein (lysine) methylation could occur during ligand‐induced cellular responses. Although the exact significance of the Vav1–Ezh2 interaction is not entirely clear, and potential substrates of the cytosolic Ezh2 complex have yet to be identified, the idea that protein methylation plays a role in signal transduction is attractive. Therefore, identification of the methylated targets of the Ezh2 complex following receptor stimulation will enhance our understanding of such novel modes of protein regulation and will further illuminate the complexities of protein structure and function.

Phosphorylation status of PcG.  Recent studies have begun to identify how cell‐signaling pathways regulate gene repression by PcG proteins. Phosphorylation has been shown to reduce the histone methyltransferase (HMTase) activity of the PRC2 complex (see above and reference^{( 59 )}) Similarly, it has been suggested that phosphorylation of Bmi1 by MAPKAP kinase 3pK results in dissociation of bmiPRC1 from chromatin.^{( 60 )} Also Jak‐Stat signaling has been shown to down‐regulate genes encoding PRC1 components, and this is important in transdifferentiation of imaginal disc cells in D. melanogaster.^{( 61 )}

Recently it was identified that nine phosphoserine and a single phosphothreonine residue existed in Mel‐18. Sequence analysis reveals that these are consensus sites for multiple serine/threonine kinases, including the casein kinase, cyclin‐dependent kinase, and MAPK families. We also suggested previously that phosphorylation of Mel‐18 correlates its dimerization status.^{( 62 )} The unambiguous identification of the regulatory kinase for melPRC1 awaits further study. BMI1 can also be phosphorylated, but only at mitosis, correlating with its dissociation from chromatin.^{( 60 )} Although this is functionally distinct from phosphorylation of Mel‐18, it is possible that some of the sites of phosphorylation overlap. How, then, does phosphorylation of Mel‐18 impact on the H2A ubiquitination activity of melPRC1 (see below)? One possible mechanism is that phosphorylation of Mel‐18 is required for the complex to bind to the surface of the nucleosome. Alternatively, phosphorylation may induce a conformational switch in Mel‐18 already bound to a nucleosome, positioning Ring1B:E2 for transfer of ubiquitin onto H2A lysine 119. The fact that both phosphorylated and unphosphorylated Mel‐18 is found in the chromatin fraction of HeLa cell nuclei perhaps favors the latter hypothesis.^{( 25 )}

H2A mono‐ubiquitination activity.  Covalent modification of histones is important in regulating chromatin dynamics and transcription.^{( 63} , ^{64 )} One example of such modification is ubiquitination, which mainly occurs on histones H2A and H2B. Although several studies have uncovered the enzymes involved in histone H2B ubiquitination and a ‘cross‐talk’ between H2B ubiquitination and histone methylation,^{( 65 )} the responsible enzymes and the functions of H2A ubiquitination have been unknown until recently. Wang et al. have reported the purification and functional characterization of an E3 ubiquitin ligase complex that is specific for histone H2A.^{( 27 )} The complex, termed hPRC1L (human PCR1–like), is composed of several PcG proteins including Ring1A, Ring1B, Bmi1, and HPH2. hPRC1L monoubiquitinates nucleosomal histone H2A at lysine 119. Reducing the expression of Ring1B results in a dramatic decrease in the level of ubiquitinated H2A in HeLa cells. Chromatin immunoprecipitation analysis demonstrated colocalization of dRing with ubiquitinated H2A at the PRE and promoter regions of the Drosophila Ubx gene in wing imaginal discs.

It is intriguing that ubiquitination of H2A and H2B has opposite effects on transcription. In the case of H2B, ubiquitination is associated with gene activation^{( 23} , ^{24 )} and is required for H3‐K4, H3‐K79 methylation, a phenomenon referred to as ‘trans‐tail’ regulation.^{( 7} , ²⁵ , ^{26 )} It is yet to be determined whether H2A ubiquitination also participates in trans‐tail regulation since H3‐K27 methylation, a modification that is central to PcG transcriptional silencing, appears to be independent of H2A ubiquitination.^{( 27 )}

In mammals, X inactivation is initiated by expression of Xist RNA and involves the recruitment of PRC1, which mediate chromosome‐wide ubiquitination of histone H2A. Recent study shows that PRC1 recruitment by Xist RNA is independent of gene silencing. Moreover Eed is required for the recruitment of the canonical PRC1 proteins Mph1 and Mph2 by Xist. Xist expression early in ES cell differentiation establishes a chromosomal memory, which allows efficient H2A ubiquitination in differentiated cells and is independent of silencing and PRC2. These data show that Xist recruits PRC1 components by both PRC2 dependent and independent modes and in the absence of PRC2 function is sufficient for the establishment of polycomb‐based memory systems in X inactivation.^{( 14} , ⁶⁶ , ^{67 )}

SUMOylation activity.  Pc2 is a Polycomb protein, which has SUMO E3 activity for the corepressors CtBP and CtBP2. Recent study demonstrates that, in vivo, Pc2 adapter function contributes to enhancement of CtBP SUMOylation. Mutation of the CtBP binding site on Pc2 abolishes E3 activity toward CtBP. Kagey et al. have demonstrated the presence of two domains in Pc2 that contribute to full in vivo E3 activity, and suggested that SUMO E3s are more than simple platforms to which E2 and substrate bind.^{( 68 )}

It is possible that Pc2 is recruited to a specific DNA response element via interaction with a DNA binding protein and enhances SUMOylation of DNA‐bound transcriptional regulatory complexes. Additionally, SUMO modification of histone H4 has been demonstrated, and based on the likely interaction of the Pc2 chromodomain with methylated lysine 27 of histone H3, it is tempting to speculate that Pc2 could be recruited to methylated lysine 27 of H3 and then stimulate SUMOylation of H4.^{( 12} , ^{68 )}

Link to non‐coding RNA.  A distinguishing feature of metazoan genomes is the abundance of non‐coding RNA (ncRNA), which function by means other than directing the production of proteins. In addition to small regulatory RNAs such as miRNAs, recent studies have predicted the existence of long ncRNAs that are spliced, polyadenylated, and roughly as diverse in a given cell type as protein‐coding mRNAs.^{( 69 )} Long ncRNAs may have diverse roles in gene regulation, especially in epigenetic control of chromatin.^{( 70 )} Perhaps the most prominent example is silencing of the inactive X chromosome by the ncRNA XIST. To normalize the copy number of X chromosomes between male and female cells, transcription of XIST RNA from one of the two female X chromosome is involved in recruiting PcG proteins to trimethylate histone H3 on lysine 27 (H3K27me3), rendering the chromosome transcriptionally silent.^{( 71 )} Presently, the mechanism by which XIST ncRNA guides Polycomb activity is unclear. Several PcG proteins possess RNA‐binding activity, and RNA is required for PcG binding to DNA, suggesting that specific ncRNAs may be critical interfaces between chromatin‐remodeling complexes and the genome.^{( 14} , ^{72 )}

The discovery of a long ncRNA that can mediate epigenetic silencing of a chromosomal domain in trans has several important implications. First, ncRNA guidance of PRC2‐mediated epigenetic silencing may operate more globally than just in the HOX loci, and it is possible that other ncRNAs may interact with chromatin‐modification enzymes to regulate gene expression in trans. Second, PcG proteins are important for stem cell pluripotency and cancer development;^{( 6 )} these PcG activities may also be guided by stem cell‐ or cancer‐specific ncRNAs. Third, Suz12 contains a zinc finger domain, a structural motif that can bind RNA^{( 73 )} and EZH2 and EED both have in vitro RNA‐binding activity. The interaction between HOTAIR (a 2.2 kb ncRNA residing in the HOXC locus) and PRC2 may also be indirect and mediated by additional factors.^{( 74 )}

Connection between PcG and cancer prognosis and therapy

Transcriptional profiling of human tumor samples holds significant promise for the advancement of cancer therapy, both in terms of improving diagnosis as well as predicting patient response to treatment. Recently, an RNA expression signature associated with ‘stem‐cell‐ness’, partly based on PcG‐driven transcriptional changes, was postulated to predict poor therapeutic outcome for patients with various types of cancer.^{( 75 )} Although these claims await further validation, it indicates that PcG expression levels might prove valuable as prognostic markers, especially as EZH2 overexpression seems to be tightly correlated with poor prognosis in both breast and prostate cancer^{( 76} , ^{77 )} (Table 2).

Table 2.

Human cancers with altered expression of polycomb group (PcG) proteins

Protein	Cancer type	References
EZH2	B‐cell non‐Hodgkin lymphoma	( 81 )
	Bladder	( 82 , 83 , 84 )
	Breast	( 85 , 86 , 87 , 88 )
	Colon	( 89 )
	Gastric	( 90 )
	Hodgkin lymphoma	( 91 )
	Liver	( 92 )
	Mantle cell lymphoma	( 93 )
	Melanoma	( 86 )
	Prostate	( 41, 86 )
SUZ12	Breast	( 94 )
	Colon	( 94 )
	Liver	( 94 )
BMI1	B‐cell non‐Hodgkin lymphoma	( 81 )
	Leukemia	( 95 )
	Mantle cell lymphoma	( 96 )
	Non‐small cell lung cancer	( 97 )
	Neuroblastoma	( 98 )
	Medulloblastoma	( 99 )

Listings are limited to publications that show overexpression of PcG proteins on the protein level, although RNA expression array analyses also indicate up‐regulation of PcG in human tumor samples. Due to the limitation of references, we have listed only EZH2, Suz12, and Bmi1. Other PcG (PCL3 and Rae28, etc.) have also been reported as showing the overexpression on the protein level, but these references have not been cited here.

The existence of cancer stem cells also has implications for treatment strategies. Cancer stem cells are postulated to have increased resistance to conventional therapy, and their persistence might therefore be one of the causes of rapid tumor relapse. Regardless of their origin, these tumor re‐initiating cells would be prime targets for therapy. Theoretically, treatment that interferes with the maintenance of the stem cell state could lead to the exhaustion of the pool of cancer stem cells and induce the conversion of malignant cancers into benign tumors. In this regard, PcG proteins and their associated enzymatic functions might serve as attractive targets for therapeutic intervention. Although an interesting concept, several challenges remain, such as the unambiguous identification of cancer stem cells, the assessment of their sensitivity, and a careful evaluation of the risk of such a treatment for normal tissue stem cells.^{( 6} , ⁷⁸ , ⁷⁹ , ^{80 )}

Conclusion

Carcinogenesis is multistage and may involve not only gene mutation but also abnormal epigenetic regulation, possibly caused by non‐genotoxic factors/environments. For a better understanding of the entirety of neoplasia, more needs to be learned from basic epigenetic regulators as well as clinical features. The authors have summarized the function of PRC connecting to many biological aspects involving tumorigenesis. In the research field of epigenetics, elucidation of molecular mechanisms by which physicochemical stress are converted to dynamic control of PRC as a master regulator of epigenetics has just begun. In addition, elucidation of epigenetic networks will provide clues to attractive therapeutic targets for epigenetic disease such as cancer.