Targeting polycomb systems to regulate gene expression: modifications to a complex story

Abstract

Polycomb group proteins are transcriptional repressors that are essential for normal gene regulation during development. Recent studies suggest that polycomb repressive complexes recognise and are recruited to through a range of different mechanisms, which involve transcription factors, CpG island elements and non-coding RNAs. Together with the realization that the interplay between polycomb repressive complexes 1 (PRC1) and 2 (PRC2) is more intricate than previously appreciated, this has increased our understanding of the vertebrate polycomb system at the molecular level.

Introduction

A remarkable feature of multicellular organisms is the capacity to create functionally unique cell types from an essentially invariant genome sequence shared by all cells in the organism. This diversity relies on the capacity of individual cell types to initiate and then maintain specific gene expression patterns during development. To achieve this, cellular signalling events are thought to regulate the activity of cell type-specific DNA binding transcription factors that function as master regulators of gene expression networks. Interestingly, however, genetic screens for factors involved in the regulation of gene expression and cell fate specification have also identified additional genes that seem to impact these processes through chromatin structure and histone modification. This is exemplified by the polycomb group genes, which were first identified in Drosophila melanogaster as regulators of Hox gene expression and normal developmental body plan specification¹. Subsequently, orthologous genes were identified in vertebrate species where they also encode transcriptional repressors. Vertebrate polycomb group proteins are essential for normal gene regulation during embryonic development and are perturbed in a wide range of human cancers (reviewed in ²).

Since the initial identification of polycomb group genes, an immense amount of biochemical work has focussed on understanding how these chromatin-associated factors function. This has led to the discovery that polycomb group proteins usually belong to one of two multi-subunit protein complexes: Polycomb Repressive Complex 1 (PRC1) that adds a ubiquityl moiety to histone H2A at lysine 119 (H2AK119ub1) and Polycomb Repressive Complex 2 (PRC2) that catalyses the addition of one to three methyl groups to histone H3 at lysine 27, leading to H3K27me1, H3K27me2 and H3K27me3 (Box 1) (reviewed in ³).

Box 1. Core PRC complexes and their chromatin modifying activities.

Polycomb repressive complexes 1 (PRC1) and 2 (PRC2) are comprise core protein components that are necessary for their respective enzymatic activities. PRC1 is a histone H2A lysine 119 ubiquitin ligase. In PRC1 two functionally equivalent yet mutually exclusive subunits, RING1A or RING1B, function as ubiquitin E3 ligases that guide the transfer of a ubiquityl moiety onto H2AK119 in chromatin. RING1A or RING1B dimerize with a PCGF subunit (of which there are six, PCGF1-6, in vertebrates). The PCGF components are required for the core enzymatic activity of the complex, and also define how the core complex interacts with auxiliary complex components to regulate its targeting to chromatin and enzymatic activity.

PRC2 is a histone H3 lysine 27 methyltransferase. The methyltransferase activity of PRC2 resides in the SET domain of two mutually exclusive proteins, EZH1 or EZH2. Alone these proteins exhibit little enzymatic activity in vitro, but when bound to EED, SUZ12, and RBAP46 or RBAP48 they form an active methyltransferase complex that is specific towards chromatin substrates. EED, SUZ12, and RBAP46 or RBAP48 contribute to the stability and structural integrity of PRC2, but they also regulate its enzymatic activity and support targeting to chromatin, either directly or via interaction with auxiliary complex components. Despite detailed characterization of the core PRC1 and PRC2 protein components, it remains unclear whether these complexes are stable entities inside cells or if they also display some capacity to assemble in a dynamic fashion on chromatin.

PRC1 and PRC2 usually co-occupy target sites in the genome, where their combined activities create ‘polycomb chromatin domains’ consisting of the polycomb group proteins themselves, H2AK119ub1, and H3K27me3. How vertebrate polycomb protein complexes are recruited to chromatin, how different PRC1 and PRC2 complexes function together in situ, and how polycomb chromatin domains actually repress transcription still remain poorly understood. Several recent excellent review articles have described the general features and functions of polycomb systems in different phyla^3–7. In this ‘Progress Article’ we focus on exciting new advances that have begun to shed light on the molecular mechanisms that underpin recruitment of polycomb group protein complexes to target sites, the formation of polycomb chromatin domains, and the functional relevance this has for gene regulation in vertebrates.

Recruitment of polycomb complexes

In the Drosophila melanogaster genome, polycomb responsive elements (PREs) act as recruitment sites for polycomb repressive complexes. DNA binding transcription factors are thought to play an important role in bringing polycomb protein complexes to these sites (reviewed in⁴). However, attempts to define vertebrate PREs have proven largely unsuccessful and emerging evidence supports the idea that vertebrate polycomb complexes are directed to DNA by both locus-specific and more generalized targeting mechanisms.

Locus-specific targeting

Inspired by observations in D. melanogaster, it has been proposed that vertebrate transcription factors might recruit polycomb repressive complexes to target sites in chromatin (Figure 1A). However, there are only a limited number of cases in which site-specific DNA binding transcription factors have been identified in unbiased biochemical isolations of polycomb complexes. Examples include E2F, MGA and MAX, which were found to interact with PRC1^{8, 9}. Thus, it has been brought into question whether interactions with transcription factors broadly underpin targeting. Candidate-based approaches have identified additional DNA binding factors, including REST, SNAIL and RUNX1 that interact with PRC1 or PRC2, but these factors seem to contribute to polycomb protein recruitment in only specific instances^10–14. More recently, detailed proteomic profiling of polycomb protein complexes has identified novel interactors, including proteins that contain zinc finger domains (ZNF518A and ZNF518B¹⁵) that are often associated with DNA binding activity. Whether these newly identified proteins contribute to polycomb complex targeting remains to be determined.

Figure 1. Getting polycomb repressive complexes to chromatin.

a | Locus-specific targeting of polycomb repressive complexes. PRC1 and PRC2 can associate with DNA binding transcription factors (top-left), such as RUNX1 and REST, which guide these complexes to chromatin. Similarly, interactions with long non-coding RNAs such as Xist function in chromosome- and locus-specific targeting of polycomb complexes (bottom-left). The chromatin remodelling protein ATRX may remodel the structure of Xist to achieve interaction with PRC2. Following recruitment to chromatin, PRC1 catalyses ubiquitylation of H2AK119 and PRC2 catalyses trimethylation of H3K27, as indicated by rounded arrows. The square arrows indicate transcription start sites.

b | Generic targeting of polycomb repressive complexes to gene regulatory elements. A variant PRC1 complex containsthe KDM2B protein. KDM2B has a zinc-finger CxxC DNA binding domain (red area) that specifically recognizes non-methylated CpG dinucleotides. This allows KDM2B to bind at CpG islands genome-wide, and contributes to PRC1 occupancy at these elements (top-right). PRC2 interacts with nascent RNA polymerase II (Pol II) transcripts at 5’-ends of genes, which may provide a mechanism to maintain repression at silent genes following stochastic transcription initiation events. Alternatively, at active genes, interaction of PRC2 with nascent RNA may constrain the catalytic activity of PRC2 and protect against polycomb-mediated repression (middle-right). A subset of PRC2 complexes contain PCL proteins which bind to H3K36me3, a modification associated with active transcription. This may enable PRC2 to bind at, or spread into, previously transcribed regions, catalysing H3K27me3 at these regions.

In addition to transcription factors, it has been suggested that long noncoding RNAs (lncRNAs) could recruit polycomb complexes to specific loci, however the generality of this as a targeting mechanism remains a topic of debate (reviewed in ¹⁶). This idea originated from the observation that the Xist lncRNA is required for localization of PRC2^17–20 to the inactivated X-chromosome during mammalian dosage compensation in a manner that relies on the sub-stoichiometric PRC2 component JARID2²¹. Recent studies, aimed at understanding the relationship between PRC2 and Xist, have suggested that the interaction between Xist and the chromatin remodelling protein ATRX induces conformational changes within the Xist RNA which favours specific and direct interaction with PRC2²² (Figure 1A). However, it has also been proposed that the relationship between Xist and PRC2 may be indirect. In support of an indirect interaction, super-resolution imaging studies have indicated that PRC2 is not intimately associated with Xist RNA on a Xist-inactivated chromosome²³ and the Xist-binding protein SHARP was required to recruit PRC2 to Xist-coated chromosomes²⁴. Nevertheless, there is additional evidence in favour of direct roles for lncRNAs in the recruitment of PRCs to chromatin. For example, PRCs are recruited to the Kcnq1ot1 locus on a paternally imprinted region of mouse chromosome 7 that transcribes lncRNAs²⁵. Moreover,the lncRNA HOTAIR, which is expressed from the HOXC locus on human chromosome 12, seems to function in trans to target PRC2 to the HOXD locus on chromosome 2²⁶.

Although it is clear that in some specific cases transcription factors and lncRNAs contribute to locus-specific recruitment of polycomb repressive complexes, it seems unlikely that these targeting mechanisms are sufficient to create the widespread, and often tissue-specific, polycomb complex binding patterns observed in vivo.

Generic targeting

Important recent discoveries have shown that polycomb repressive complexes bind to regulatory sites across the genome through generic chromatin binding activities, with local chromatin features dictating residency and function at these sites.

In vertebrate genomes, the most universal and striking feature of polycomb-occupied sites is the presence of a CpG island (CGI). CGIs are approximately 1-2 kilobase regions of CpG-rich DNA that lack DNA methylation and typically encompass gene promoters^{27, 28}. The striking correlation between polycomb protein occupancy and CGIs, coupled with the observation that artificial DNA sequences with a high CpG content are sufficient to nucleate polycomb group proteins in vivo²⁹, has led to suggestions that CGIs may play a direct role in polycomb protein recruitment. Attempts to define a molecular link between polycomb complexes and CGIs have largely focussed on the KDM2B protein, which stably associates with PRC1. KDM2B encodes a ZF-CxxC DNA binding domain that specifically recognizes non-methylated CpG dinucleotides allowing KDM2B to bind to CGIs genome-wide^30–32. Recent studies have demonstrated that KDM2B contributes to PRC1 occupancy at CGIs (Figure 1B). However, somewhat paradoxically, high-level PRC1 enrichment is only achieved at the most repressed sites, despite KDM2B residing at all CGIs. This suggests that the recruitment of PRC1 to CGIs by KDM2B, or the repressive activity of PRC1 at CGIs is regulated by additional mechanisms that permit the establishment of polycomb chromatin domains. Furthermore, KDM2B-independent mechanisms must also function at some CGIs to recruit or stabilize PRC1 binding, as removal of KDM2B does not lead to a complete loss of PRC1 at all CGI target sites^30–33. Interestingly, it has been reported that the PRC2 component JARID2 preferentially recognises GC-rich DNA³⁴. JARID2 could thus provide a complementary mechanism for targeting PRC2 to CGIs independently of KDM2B. Together these observations suggest that polycomb complexes can be recruited to all CGIs, though other mechanisms define whether this results in the formation of stable polycomb chromatin domains.

Although lncRNA-mediated targeting has been proposed as a locus-specific recruitment mechanism for polycomb complexes, it has been recently reported that PRC2 binds RNA promiscuously with little sequence specificity^{35, 36}. This has led to several new suggestions as to how PRC2-RNA interactions may contribute to the generic recruitment or functionality of polycomb complexes at gene regulatory elements and genes. At repressed genes, binding of PRC2 to short abortive RNA transcripts may help to retain polycomb complexes and stabilize transcriptional repression during aberrant transcriptional initiation events^{37, 38}. Alternatively, nascent transcripts produced from active genes may interact with PRC2 and provide a ‘decoy’ to block stable interactions between PRC2 and chromatin, thereby protecting the transcribed gene from repressive polycomb activity. This latter idea has received some support from the observation that interactions with nascent transcripts can constrain the histone methyltransferase activity of PRC2^{39, 40}. An alternative possibility is that PRC2 utilises interactions with nascent transcripts as a generalised targeting mechanism to increase residency around gene promoters (Figure 1B) and that other counteracting signals, including H3K4 and H3K36 methylation, prevent stable polycomb domain formation at active genes⁴¹. These emerging links between ncRNAs and PRC2 enzymatic activity seem to be evolutionarily conserved, as regulated bidirectional ncRNA expression from a D. melanogaster PRE acts as a switch to alter PRC2 activity and regulate gene expression⁴². Interestingly, vertebrate polycomb target sites often exhibit bidirectional transcription, so it will be important to understand if similar switch-like mechanisms are involved in regulating vertebrate PRC2 activity.

In addition to generic DNA and RNA binding activities, recent evidence suggests that certain polycomb repressive complexes recognise chromatin modifications placed by other histone modifying systems that are broadly associated with genes and regulatory elements. For example, the sub-stoichiometric PRC2 subunits Polycomblike 1-3 (PCL1, PCL2, PCL3) encode TUDOR domains that recognise the H3K36me3 modification^43–46. H3K36me3 is typically associated with actively transcribed gene bodies, and during cellular differentiation, PCL proteins were proposed to facilitate spreading of PRC2 into these regions (Figure 1B). In addition, links have been identified between PRC2 and the H3K9 methylation systems. Most notably, biochemical interactions were identified between PRC2 and the H3K9 methyltransferases G9A and GLP, and loss of G9A resulted in impaired PRC2 recruitment at a subset of target sites^{15, 47}.

Interplay between PRCs

Outlined above are simple examples of how PRC1 or PRC2 are individually recruited to target sites by locus-specific or generic targeting mechanisms. However, after their recruitment to chromatin, the functions of PRC1 and PRC2 are intimately related, as the enzymatic activity of each complex influences the other’s occupancy on chromatin and the full establishment of polycomb chromatin domains.

The prevailing hierarchical model

PRC1 and PRC2 typically co-localise at target sites throughout the genome. This has largely been attributed to a mechanism discovered over a decade ago in D. melanogaster which posits that de novo recruitment of PRC2 to target site catalyses H3K27me3, which is subsequently recognized by a chromobox (CBX)-containing protein in PRC1, leading to H2AK119ub1 placement and polycomb chromatin domain formation (Figure 2A)^{48, 49}. This pathway is generally referred to as the ‘hierarchical’ recruitment mechanism and places PRC1 recruitment and activity downstream of PRC2 function. Based on the conservation of CBX proteins, and on the evidence that PRC1 binding to chromatin is sensitive to loss of PRC2⁵⁰, the hierarchical recruitment mechanism was largely adopted to explain polycomb chromatin domain formation in vertebrates. However, detailed studies of the relationship between PRC1 and PRC2 in vertebrates indicate that other mechanisms are involved in the formation of polycomb domains. For example, deletion of PRC2 in mouse embryonic stem cells led to a reduction, but not loss, of PRC1 proteins at target sites and had little effect on global levels of H2AK119ub1⁵¹. These observations suggested that the relationship between PRC1 and PRC2 is more complex than originally envisaged.

Figure 2. Beyond simple recruitment to more complex interactions.

a | In the ‘hierarchical’ model of polycomb complex function, PRC2 first binds to chromatin and places H3K27me3. It is then proposed that H3K27me3 is recognized by chromobox (CBX) proteins, which are a subunit of canonical PRC1 complexes, thereby allowing PRC1 to bind and mono-ubiquitylate H2AK119.

b | In the ‘Alternative’ model for polycomb recruitment, the initial event is the binding of variant PRC1 complexes (which contain RYBP instead of CBX) to chromatin. PRC1 then catalyses the mono-ubiquitylation of H2AK119, independently of PRC2 activity and H3K27me3. The H2AK119ub1 modification then promotes the recruitment of PRC2, possibly via direct recognition of H2AK119ub1 by the AEBP2-JARID2-PRC2 complex, and placement of the H3K27me3 mark.

c | PRC1 complexes functionally segregate into ‘canonical’ complexes that contain chromobox (CBX2/4/6/7/8) and polyhomeotic (PHC1/2/3) proteins, and a series of ‘variant’ complexes that contain RYBP (or the closely related YAF2 protein) and interact with proteins that are unique to individual variant PRC1 complexes (note that only selected proteins are shown here). Under some conditions the PRC2 subunit EED may interact with a CBX-containing canonical PRC1 complex (dashed arrow).

d | A model for the propagation of polycomb domains.Within the core PRC2 complex, the EED subunit is able to recognise H3K27me3 via its WD40 repeat domain. This interaction potentially recruits PRC2 to sites of pre-existing H3K27me3, as well as stimulating the enzymatic activity of PRC2. The EED-H3K27me3 interaction may facilitate spreading of H3K27me3 domains (left panel, dashed arrows) or copying of H3K27me3 onto newly incorporated histones (right panel, blue nucleosomes) during DNA replication, thereby stably propagating H3K27me3 domains in actively dividing cells.

A new twist in hierarchy: PRC1 recruits PRC2

Our understanding of polycomb systems has recently evolved with the surprising finding that PRC1 can recruit PRC2 to chromatin via a mechanism that involves recognition of H2AK119ub1 (Figure 2B). This alternative pathway was discovered using a cell-based system in which PRC1 was artificially recruited to a region of the genome devoid of genes and not normally occupied by polycomb proteins. Strikingly, de novo PRC1 binding to this site and monoubiquitylation of H2AK119 resulted in the subsequent recruitment of PRC2 and H3K27me3 deposition, creating a new polycomb chromatin domain⁵². A similar outcome was observed when PRC1 was artificially recruited to pericentric regions of the genome⁵³.

This newly discovered recruitment mechanism seems to play a part in polycomb chromatin domain formation at natural target sites, as perturbation of PRC1 and loss of H2AK119ub1 caused a significant reduction in PRC2 binding and H3K27me3 across the genome. The precise mechanism by which PRC1-dependent H2AK119ub1 underpins de novo polycomb chromatin domain formation remains to be elucidated. However, in vitro studies have identified a PRC2 complex containing the auxiliary proteins AEBP2 and JARID2 that preferentially binds to and stimulates catalysis of H3K27me3 on chromatin containing H2AK119ub1⁵⁴ suggesting this may provide the molecular link between PRC1 activity on chromatin and recruitment of PRC2 and H3K27me3. Additional unexpected connections between the two polycomb complexes include the finding that PRC2 can interact directly with an alternative H2AK119ub1 E3 ligase, TRIM37, which is highly expressed in breast cancer cells carrying the 17q23 amplification⁵⁵. Furthermore, EED, a core PRC2 subunit, has been reported to directly interact with PRC1⁵⁶ (Figure 2C). Together these observations indicate the potential for more functional overlap between PRC1 and PRC2 than previously appreciated.

Alternative roles for canonical PRC1

The observation that PRC1 can promote the recruitment of PRC2 to target sites on chromatin has placed a new focus on understanding how PRC1 complexes contribute to polycomb chromatin domain formation and function. Systematic biochemical purifications have revealed that PRC1 in vertebrates can be separated into ‘canonical’ PRC1 complexes, which have CBX proteins and presumably function as part of the hierarchical recruitment pathway, and the less well studied ‘variant’ PRC1 complexes, which lack CBX proteins and contain either RYBP or YAF2 (Figure 2C)⁸. Importantly, RYBP and YAF2 lack the capacity to bind H3K27me3, and can be recruited to chromatin in cells lacking functional PRC2, suggesting that variant PRC1 complexes must be recruited to chromatin via PRC2-independent mechanisms⁵¹. Furthermore, individual canonical and variant PRC1 complexes contain distinct protein subunits that likely contribute to target site recognition or catalysis⁸.

Examining the function of individual PRC1 complexes in vivo has revealed that variant PRC1 complexes are proficient at catalysing H2AK119ub1 on chromatin, whereas canonical complexes catalyse little of this modification⁵². This is consistent with observations that RYBP-containing PRC1 variant complexes have enhanced H2AK119ub1 catalysis in vitro⁸. As a consequence of their restricted catalytic activity in vivo, canonical PRC1 complexes seem to have limited capacity to recruit PRC2 and form polycomb chromatin domains. One exception to this generality was recently reported: a CBX2-containing canonical PRC1 complex seems to have a very specific role in atypical polycomb chromatin domain formation by directly recognizing pericentromeric heterochromatin during early mouse development⁵⁷.

Nevertheless, if canonical PRC1 complexes are usually limited in their capacity to deposit H2AK119ub1 in vivo, why are they recruited to polycomb chromatin domains at all? A hint as to possible functions came with the recent demonstration that a stable component of canonical PRC1 complexes, polyhomeotic 2 (PHC2), can auto-polymerise via its sterile-alpha motif (SAM) domain leading to chromatin compaction and gene silencing^58–60. This is consistent with previous reports detailing a ubiquitin ligase-independent role for PRC1 in chromatin compaction⁶¹. Interestingly, inhibition of PHC2 polymerisation resulted in loss of canonical PRC1 binding to chromatin only at sites marked with H3K27me3, suggesting that both H3K27me3-CBX interactions and SAM domain polymerisation may have important structural roles in creating and translating repressive chromatin structures on chromatin⁵⁹.

Propagation of polycomb domains

Biochemical studies have demonstrated that polycomb repressive complexes can bind the histone modifications that they themselves place. For example, the EED subunit of PRC2 core complex binds H3K27me3 via its WD40 repeat and this interaction seems to stimulate the catalytic activity of PRC2, thus forming an activity-based feedback loop^{62, 63}. It has been proposed that this could promote the spreading of H3K27me3 along chromatin as well as ensure the propagation of H3K27me3 on newly replicated chromatin⁶² (Figure 2D). Based on the finding that PRC1, PRC2 and their chromatin modifying activities are more intimately linked than previously appreciated ^52–54, it seems plausible that this robust series of PRC-dependent feedback mechanisms could underpin both the spreading and maintenance of polycomb chromatin domains once they are initially established. It is tempting to speculate that this would not only contribute to epigenetic maintenance of polycomb chromatin domains, but also to rigidly maintain gene expression states during cell division and development. A detailed understanding of the epigenetic nature of polycomb chromatin domains is an interesting and evolving area of polycomb biology.

Polycomb systems and gene regulation

It is often suggested that vertebrate polycomb group proteins are recruited to target sites to actively drive transcriptional repression (Figure 3A). However, recent evidence suggests that this may not be the central modality connecting polycomb chromatin domains and gene repression. Firstly, kinetic analysis of gene expression in a cell culture model of Ras-induced transformation revealed that cessation of transcription preceded H3K27me3 acquisition⁶⁴. Secondly, experiments in mouse embryonic stem cells demonstrated that small-molecule inhibitors that block transcription caused recruitment of PRC2 and H3K27me3 to previously active genes⁶⁵. In light of these conceptually important findings, one interesting possibility is that polycomb systems may exploit generic targeting activities to constantly interface with or ‘sample’ gene regulatory elements and respond to the transcriptional state of individual genes. Within the context of this model, features of active transcription, for example the presence of RNA PolII or transcription-associated histone modifications, would decrease the residency time or catalytic activity of polycomb complexes at transcribed genes, meaning that full polycomb chromatin domain establishment would only occur at sites where transcriptional silencing has already been achieved (Figure 3B) (reviewed in ⁶⁶).

Figure 3. Polycomb systems and gene regulation.

a | An instructive model for polycomb complex-mediated silencing would posit that newly recruited polycomb complexes lead to polycomb chromatin domain formation, which then directs repression of transcription by RNA Polymerase II (Pol II) at the associated gene.

b | A responsive model would posit that polycomb complexes constantly ‘sample’ chromatin at regulatory elements via generic targeting modalities in order to respond to the transcriptional state of the associated gene. Transcriptional cessation would lead to the subsequent establishment of polycomb chromatin domains that protect against low-level or stochastic reactivation signals. However, in response to active transcription, the presence of Pol II or other features of transcriptional initiation would block polycomb domain establishment.

What would be the purpose of establishing polycomb chromatin domains at already silenced genes? One possibility is that following transcriptional silencing, polycomb systems form repressive chromatin domains to limit the potential for stochastic reactivation events in inappropriate tissues. This general concept is consistent with observations in D. melanogaster, where polycomb group genes are important for the maintenance, but not the establishment, of gene expression programmes⁶⁷. Interestingly, in pluripotent cells, polycomb-occupied gene promoters also typically exhibit low levels of the transcriptionally permissive histone modification H3K4me3, which is placed by trithorax group proteins^{68, 69}. This modification is thought to result from the capacity of the trithorax system to generically ‘sample’ regulatory elements in a manner analogous to polycomb proteins. Given that polycomb and trithorax are two opposing chromatin-based gene regulatory systems, these constant sampling activities at gene regulatory elements might form the basis for a chromatin-encoded bistable switch that helps to regulate the transition between, and then stable maintenance of, chromatin states that are either permissive or repressive to transcription (discussed in detail in ^{4, 27, 66, 70}).

Although polycomb chromatin domains are predominantly found at silenced genes, recent evidence has implicated polycomb proteins and their chromatin modifications in active transcription and other gene regulatory functions. This is the case for a variant PRC1 complex containing the AUTS2 protein, which recruits Casein Kinase 2 (CSNK2A) to counteract PRC1 ubiquitin ligase activity and repression⁷¹. Similarly, in erythroid lineages, a PRC2 complex containing an alternative catalytic core associated with regions of the genome characterized by chromatin modifications associated with active transcription (H3K4me3 and H3K27ac) and promoted gene expression⁷². Furthermore, it was recently demonstrated that PRC2-dependent H3K27 mono-methylation is enriched in the bodies of transcribed genes, while the H3K27 dimethyl state covers much of the genome, including non-genic locations, possibly as a mechanism to block aberrant function of enhancer elements⁷³. These interesting recent observations highlight an incomplete understanding of the precise mechanisms by which polycomb systems regulate gene expression, and elucidating such mechanisms remains a key challenge.

Conclusion and future directions

A complex series of interactions between PRC1 and PRC2, chromatin, and transcription are essential for normal polycomb chromatin domain formation and function. Although transcription factor and lncRNA-based mechanisms contribute to polycomb complex targeting to specific chromatin regions, recent evidence suggests that polycomb complexes also constantly engage regulatory elements throughout the genome and establish repressive polycomb chromatin domains in a manner that is often responsive to the transcriptional state of associated genes. These emerging principles prompt alternative ways to think about the vertebrate polycomb system, and further experiments will be required to fully understand the molecular mechanisms underlying PRC function. Importantly, there is little evidence that the chromatin modifications placed by the polycomb systems directly repress transcription, suggesting that components of the polycomb protein complexes, such as PHC, may induce transcriptional inhibition by directly modifying chromatin structure.

Understanding the precise molecular mechanisms by which polycomb protein complexes repress gene expression remains a central and outstanding question in the field. A cryo-EM structure of the PRC2 holocomplex⁷⁴ and the recently solved atomic structure of the PRC1 core complex bound to a nucleosome⁷⁵ have provided exciting new insights into the interactions that occur within polycomb complexes and also with their chromatin substrates at high-resolution. These and other new discoveries will contribute to the understanding of how polycomb systems regulate gene expression programs during normal multicellular development. They will also pave the way for new approaches to therapeutic interventions in human diseases, including cancer, where polycomb systems are frequently perturbed.

Author biographies

Rob Klose is Wellcome Trust Senior Research Fellow and Professor of Cell and Molecular Biology in the Department of Biochemistry at the University of Oxford. He carried out his doctoral work with Adrian Bird at the University of Edinburgh, UK, and his postdoctoral studies with Yi Zhang at the University of North Carolina, Chapel Hill, USA. His laboratory studies how chromatin modification and architecture contribute to gene regulation.

Neil Blackledge is a senior post-doctoral researcher in the group of Rob Klose. His doctoral studies were carried out with Ann Harris at Northwestern University, Chicago, USA. His post-doctoral work is aimed at understanding the role of CpG islands in gene regulation, with a particular focus on the polycomb repressive system.

Nathan Rose is a Junior Research Fellow in the Department of Biochemistry and St John’s College, University of Oxford. His doctoral studies were carried out with Christopher Schofield at the University of Oxford, and his postdoctoral work with Rob Klose focuses on the biochemistry of polycomb repressive complexes.