Dangerous liaisons: interplay between SWI/SNF, NuRD, and Polycomb in chromatin regulation and cancer

Abstract

Changes in chromatin structure mediated by ATP-dependent nucleosome remodelers and histone modifying enzymes are integral to the process of gene regulation. Here, we review the roles of the SWI/SNF (switch/sucrose nonfermenting) and NuRD (nucleosome remodeling and deacetylase) and the Polycomb system in chromatin regulation and cancer. First, we discuss the basic molecular mechanism of nucleosome remodeling, and how this controls gene transcription. Next, we provide an overview of the functional organization and biochemical activities of SWI/SNF, NuRD, and Polycomb complexes. We describe how, in metazoans, the balance of these activities is central to the proper regulation of gene expression and cellular identity during development. Whereas SWI/SNF counteracts Polycomb, NuRD facilitates Polycomb repression on chromatin. Finally, we discuss how disruptions of this regulatory equilibrium contribute to oncogenesis, and how new insights into the biological functions of remodelers and Polycombs are opening avenues for therapeutic interventions on a broad range of cancer types.

Keywords: NuRD; Polycomb; SWI/SNF; cancer; chromatin.

Figure 1.

ATP-dependent chromatin remodeling. (A) Different outcomes of ATP-dependent remodeling of nucleosomes. Remodeler action can drive the sliding of a nucleosome to another position on the DNA, thus exposing a previously bound sequence. Alternatively, remodelers can make the nucleosomal DNA more accessible, while the histone octamer remains associated. Remodeling can also disrupt the octamer structure causing a partial disassembly, typically through eviction of histone H2A/H2B dimers. Specialized remodelers can mediate the exchange between histone variants. Finally, remodeling can result in the complete eviction of the histone octamer. (B) Structural domains of the four major Snf2 ATPase subfamilies, SWI/SNF, CHD, ISWI, and INO80. The translocase/ATPase domain of all remodelers comprises two RecA-like lobes separated by an insertion (highlighted in gray). Members of the INO80 family have a longer insertion than other remodelers. Each subfamily is characterized by a unique set of additional domains, including the HSA (helicase SANT-associated) and post-HSA domains, SnAC (Snf2 ATP coupling), AT hooks (A/T-rich DNA-binding domains), Bromo (bromodomains), Chromo (chromodomains), SANT-SLIDE domain, PHD finger (plant homeodomain), HAND-SANT-SLIDE domain, AutoN (autoinhibitory N-terminal), and NegC (negative regulator of coupling). See the text for details and references.

Figure 2.

Model of nucleosome remodeling. (A) Cartoon of a generic ATP-dependent translocase RecA lobe 1 and lobe 2 moving along the tracking strand in a 3′ to 5′ direction. A cycle of ATP-binding and hydrolysis drives conformational changes through which the translocase “inchworms” along the tracking strand with 1-bp steps per every ATP hydrolysis. (B) Top and side view of remodeler ATPase binding to a nucleosome. The ATPase subunits of SWI/SNF, ISWI, CHD1, and SWR1 bind the nucleosomal DNA at superhelical position 2 (SHL + 2). (C) Remodeler ATPases make additional contacts through (1) binding to the opposite DNA gyre, ∼90 bp away; (2) contacting the histone core, typically at an acidic patch formed by H2A and H2B; (3) binding the linker DNA; and (4) interacting with the N-terminal tail of (usually) histone H4. Due to these additional contacts, the ATPase does not move along the nucleosomal DNA but rather pulls the DNA toward the octamer dyad. Multiple cycles of ATP-binding and hydrolysis generates a ratcheting motion that locally distorts the DNA and peels it off the histone core. See the text for details and references.

Figure 3.

Remodeler functions in organizing the chromatin template. (A) Remodelers such as the ISWI class ACF mediate the formation of regularly spaced nucleosomal arrays; e.g., following DNA replication or other disruptions of chromatin organization. Well-organized arrays help prevent spurious initiation of transcription. (B) SWI/SNF remodelers promote transcription activation by generating an open chromatin conformation at promoters and enhancers, which may involve the sliding, displacement, or restructuring of nucleosomes. The relative importance of each of these mechanisms in vivo remains to be determined. Remodeler targeting involves recruitment by sequence-specific transcription factors and the local chromatin state; e.g., through recognition of acetylated histones by one of the bromodomains of SWI/SNF. In addition, SWI/SNF remodelers counteract Polycomb-repressive complexes (PRCs). (C) NuRD remodelers antagonize SWI/SNF function. NuRD mediates nucleosome invasion of regulatory DNA, and removal of acetylation marks. NuRD activity is then thought to promote the subsequent recruitment of the Polycomb system via its deacetylation of H3K27 and/or nucleosome remodelling, to further the formation of repressive chromatin. Green flags represent histone acetylation, and red flags represent H3K27me3.

Figure 4.

Composition of mammalian SWI/SNF and NuRD complexes. (A) Schematic representation of mammalian SWI/SNF complexes BAF, PBAF, and GBAF. Due to gene duplication events, several components of each complex are encoded by up to three paralogous genes in mammals. Alternative names and orthologous subunits in yeast and Drosophila are in Table 1. (B) Mammalian NuRD complex. A NuRD-related HDAC module lacking CHD3-5 and MBD2/3, associated with PWWP2A/B, is also illustrated. For details and references, see the text.

Figure 5.

Polycomb group protein complexes and chromatin repression. (A) Schematic representation of PRC2.1 and PRC2.2 complexes. Drosophila melanogaster PRC2 components are shown in the colored ovals, and their mammalian homologs are also indicated. In mammals, PRC2 has a trimeric enzymatic core composed of EZH1/2–SUZ12–EED. PRC2.1 contains one PCL protein and the vertebrate- and eutherian-specific proteins PALI1/2 and EPOP, respectively. In PRC2.2, these proteins are replaced by AEBP2 and JARID2. (B) Schematic representation of cPRC1 and ncRC1. D. melanogaster PRC1 are indicated in the colored ovals. The names of their sometimes multiple mammalian homologs are also indicated. The enzymatic core of PRC1 is a RING-PCGF heterodimer (which is present in cPRC1) and ncPRC1. In cPRC1, RING-PCGF associates with one of each of the CBX, PHC, and SCM proteins. In ncPRC1, RING-PCGF associates with either a RYBP or YAF2 subunit. (Right panels) The different ncPRC1 complexes are defined by their specific PCGF subunit, which in turn associate with divergent subsets of interacting proteins. (C) Multiple ways of PRC recruitment to chromatin. As detailed in the text, KDM2B binds CpG islands (CGIs), thus targeting ncPRC1 (1) and H2A ubiquitylation (2). Polycomb-like proteins also bind CpG islands (3) and mediate H3K27 methylation by PRC2.1 (4). (5) H3K27me3 is recognized by EED, which then allosterically activates PRC2, thus facilitating the establishment of H3K27me3 domains. (6) The ncPRC1 mark H2Aub is recognized by JARID2, promoting local H3K27me3 by PRC2.2. (7) H3K27me3, in turn, is bound by CBX proteins in cPRC1 complexes that mediate chromatin compaction. (8) PCGF3/5/6 ncPRC1 complexes harbor sequence-specific DNA binding proteins that target H2Aub to chromatin. While, they also target nonclassical PcG sites, they may also contribute to promoting deposition of H3K27me3 via PRC2.2 binding to H2Aub. (D) Recruitment of the various PRC complexes generates a repressive chromatin environment characterized by H3K27me3, H2Aub, and chromatin compaction mediated by long-range interactions involving SAM domain-mediated polymerization of the CBX2 and PHC1-3 subunits. These PcG silenced domains are sometimes referred to as Polycomb bodies (red), that are separate in nuclear space from open transcribed chromatin (green).

Figure 6.

Chromatin remodelling complexes are frequently mutated in cancer. Mutations in different SWI/SNF subunits associate with different types of cancer. The percentage of mutations in specific types of cancer are indicated. Adult cancers are shown in blue, pediatric cancers in red. Cancer-associated mutations in NuRD do occur but are less frequent than in SWI/SNF. The enzymatic core of PRC2 is subject to both activating (in diffuse large B-cell and follicular lymphoma) and inactivating (in T-cell acute lymphoblastic leukemia and malignant peripheral nerve sheath tumors) mutations. The primary substrate of the PRC2 complex, H3K27, is also the target of an oncogenic mutation in the majority of pediatric diffuse intrinsic pontine glioma tumors. Mutation rates, which are indicated for each disease, are taken from the Cancer Genome Atlas (TCGA) pan-pediatric atlas (Gao et al. 2018) for adult cancers and for pediatric diseases from recent pan-pediatric cancer genmics studies (Gröbner et al. 2018; Ma et al. 2018). Cancer studies are indicted by their TCGA abbreviations. For details and references, see the text.

Figure 7.

Therapeutic opportunities. (A) Maintaining chromatin equilibrium. Physiological gene expression depends on the balanced interplay between SWI/SNF, NuRD, and Polycomb. A disturbance in this chromatin equilibrium—for example, due to loss of one of the SWI/SNF subunits—can promote oncogenesis due to misexpression of genes that regulate cell proliferation, cell differentiation, EMT, cellular senescence, or apoptosis. A therapeutic strategy might involve restoring the chromatin balance by compensating for loss of SWI/SNF function by inhibition of PRC2 or NuRD. Conversely, loss of Polycomb function might be compensated for by inhibition of SWI/SNF. (B) Tipping the chromatin balance. Synthetic lethality provides another potential therapeutic strategy for cancers with loss-of-function mutations in SWI/SNF. The residual SWI/SNF complexes are often essential for the viability of these tumor cells, and therefore present attractive therapeutic targets. For example, SMARCB1 mutant RT cells depend on BRD9 function. Alternatively, loss of ARID1A might create a crucial requirement for PBAF or GBAF. The loss of one subunit (e.g., SMARCA4 in lung cancer cells) can also create a crucial dependency on its paralog, SMARCA2. Consequently, targeting paralog function in these settings may provide an effective therapeutic option. For details and references, see the text.