Sentinels of chromatin: chromodomain helicase DNA-binding proteins in development and disease

In this review, Alendar and Berns discuss emerging evidence that the chromodomain helicase DNA-binding (CHD) family of enzymes is subjected to frequent DNA copy number alterations or mutations and shows aberrant expression in malignancies and other human diseases.

Abstract

Chromatin is highly dynamic, undergoing continuous global changes in its structure and type of histone and DNA modifications governed by processes such as transcription, repair, replication, and recombination. Members of the chromodomain helicase DNA-binding (CHD) family of enzymes are ATP-dependent chromatin remodelers that are intimately involved in the regulation of chromatin dynamics, altering nucleosomal structure and DNA accessibility. Genetic studies in yeast, fruit flies, zebrafish, and mice underscore essential roles of CHD enzymes in regulating cellular fate and identity, as well as proper embryonic development. With the advent of next-generation sequencing, evidence is emerging that these enzymes are subjected to frequent DNA copy number alterations or mutations and show aberrant expression in malignancies and other human diseases. As such, they might prove to be valuable biomarkers or targets for therapeutic intervention.

All cells in our body face the challenge of accommodating ∼2 m of DNA in a 5-µm nucleus. The fundamental unit of compaction is the nucleosome that, together with linker histone H1, other structural nonhistone DNA-binding proteins, and RNA, facilitates DNA assembly into higher-order chromatin (Yadav et al. 2018). Nevertheless, the chromatin-packaged DNA must be accessible to the myriad of factors that catalyze transcription, repair, replication, and recombination (Radman-Livaja and Rando 2010). Transcriptional activity of genes is mechanistically coupled to their local chromatin environment, shaped by the actions of enzymes that either catalyze the deposition or removal of covalent modifications on histones and DNA or use ATP hydrolysis to mobilize nucleosomes (Voss and Hager 2014).

ATP-dependent chromatin remodelers have a SNF2 helicase-like ATPase domain (Ryan and Owen-Hughes 2011) and, based on the presence of other functional domains, can be broadly divided into four main protein families: SWI/SNF, ISWI, CHD, and INO80 complexes (Clapier et al. 2017).

Here, we highlight the major findings of research on the CHD family of proteins and discuss challenges that lie ahead. Other families of chromatin remodelers have been extensively reviewed elsewhere (Narlikar et al. 2013; Kadoch and Crabtree 2015; Clapier et al. 2017).

Domain organization, structural features, and regulatory principles

Almost three decades ago, the group of the late Robert Perry (Delmas et al. 1993) cloned the first member of this remarkable family, murine Chd1, and coined the term “chromodomain–helicase–DNA-binding protein.” CHD proteins are highly conserved in evolution and play pivotal roles in organismal development and tumorigenesis. Based on their constituent domains, CHD remodelers are classified into three subfamilies: subfamily I (CHD1 and CHD2), subfamily II (CHD3–5), and subfamily III (CHD6–9) (Fig. 1; Clapier et al. 2017).

Figure 1.

Structural domain organization of the CHD enzymes. Domain representation of all three CHD subfamilies as indicated in the text. All CHDs have double chromodomains (blue squares) and SNF2-like ATPase/DEXDc helicase domains (green ovals). Subfamily II members contain tandem PHD Zn finger-like domains (purple triangles). Subfamily III members have additional Brahma Kismet (BRK) domains at the C terminus (gray oval); enzymes of both subfamilies I and III contain SANT-SLIDE domains at C terminus (red hexagon). These domains facilitate engagement and stabilization of the interaction with the chromatin substrate and allow efficient ATP hydrolysis and DNA translocation. Recently, a number of novel structural domains have been identified: CHD-N-terminal, CHD-C-terminal, CHCT, DUF4208, and conserved region (CR). CHD9 associates with nuclear receptors through LxxLL recognition motifs (orange stripes).

CHDs are a family of large proteins that exist either as monomers or constituents of multimeric complexes, exerting specific functions in different cell types or developmental stages (Supplemental Table S1; Micucci et al. 2015). Moreover, alternative promoter usage or splicing can result in several protein isoforms, generating additional functional complexity (Kita et al. 2012; Nishiyama et al. 2012; Kunkel et al. 2018, 2020). They are subjected to a range of post-translational modifications (Schmidt and Schreiber 1999; Choudhary et al. 2009; Urquhart et al. 2011; Kessler et al. 2012; Piatti et al. 2015; Salomon-Kent et al. 2015; Yu et al. 2015; Luijsterburg et al. 2016; Zhao et al. 2017) and reside in the nucleus and nucleolus (Zentner et al. 2010; Kita et al. 2012; Xie et al. 2012; Salomon-Kent et al. 2015; Hoffmeister et al. 2017), as well as in the cytoplasm (Stokes and Perry 1995; Salomon-Kent et al. 2015), where they exert context-dependent functions.

The hallmark features of all CHD remodelers are the presence of double chromodomains at their N-terminal region and a central SNF2 helicase-like ATPase domain (Fig. 1). The double chromodomains mediate association with chromatin by directly binding to methylated lysines on histone H3 tails (Flanagan et al. 2005, 2007; Sims et al. 2005), nucleosomal DNA (Nodelman et al. 2017), poly(ADP-ribose) (Murawska and Brehm 2011), or RNA (Akhtar et al. 2000), serving as nucleosome recognition surface and allosteric regulators of the enzymatic activity of the ATPase motor (Hauk et al. 2010). Notably, the chromodomains of subfamily II members lack key residues required for binding to methylated histone tails but have evolved the ability to bind DNA substrates (Flanagan et al. 2007). The ATPase motor of CHD enzymes is related to the catalytic subunit of superfamily 2 (SF2) helicases (Flaus et al. 2006), and uses ATP hydrolysis to disrupt histone–DNA contacts within the nucleosome to alter chromatin packaging (Hauk and Bowman 2011). Furthermore, CHD enzymes harbor several auxiliary domains that facilitate substrate recognition and chromatin binding, ATPase activity, and coupling of ATP hydrolysis to productive DNA translocation (Fig. 1; Marfella and Imbalzano 2007; Ramírez et al. 2012; Silva et al. 2015; Luijsterburg et al. 2016; Mohanty et al. 2016).

The subfamily II members CHD3/CHD4 (Mi-2α/β) and CHD5 are core components of the multimeric enzymatic complex with both chromatin remodeling and histone deacetylase activity, termed the nucleosome remodeling and histone deacetylase complex (NuRD) (Supplemental Table S1 and references therein). They contain additional tandem N-terminal zinc finger-like, paired plant homeodomains (tPHD) (Mansfield et al. 2011), which appear to be structurally independent modules that can bind separate unmodified or K9-methylated H3 tails on a single nucleosome (Supplemental Table S1; Musselman et al. 2012).

The third subfamily (CHD6–9) includes recently identified homologs of Drosophila Kismet protein (KIS-L). They are evolutionarily conserved and harbor additional structural features such as the Brahma and Kismet domain (BRK), the conserved region (CR) domains (Schuster and Stöger 2002; Shur and Benayahu 2005), and the SANT-SLIDE-like domain (Fig. 1; Boyer et al. 2004). The BRK domains appear to mediate CHD7/8 interaction with CCCTC-binding factor (CTCF) (Ishihara et al. 2006), and between CHD7 and CHD8 themselves (Batsukh et al. 2010). The SANT-SLIDE-like DNA-binding domain (DBD) is also present at the C terminus of CHD1 and CHD2 and binds linker DNA to direct nucleosomal sliding (McKnight et al. 2011; Ryan et al. 2011; Liu et al. 2014a).

Chromatin recruitment and allosteric regulation of the ATPase activity

Since CHD proteins have limited DNA sequence specificity (Stokes and Perry 1995), their chromatin association is mediated by interactions with transcription factors, modified histones, RNA, poly(ADP-ribose), and methylated DNA (Supplemental Table S1). Current evidence supports a model in which CHD association with chromatin takes place in a stepwise fashion. Interactions with specific transcription factors (TF) or subunits of the multimeric protein complexes in which CHDs reside govern recruitment to target chromatin sites (Längst and Manelyte 2015), as exemplified by the chromatin recruitment of the NuRD complex through interactions with a panoply of developmental and tissue-specific transcription factors (Supplemental Table S1 and references therein); the NuRD complex also binds methylated DNA, acting as an important nexus between gene silencing and DNA methylation (Zhang et al. 1999).

Moreover, the presence of several distinct histone and RNA- and DNA-binding modules facilitates CHD chromatin association (Fig. 1; Clapier and Cairns 2009; Längst and Manelyte 2015; Hendrickson et al. 2016). In general, double chromodomains of CHD enzymes (CHD1/2 and CHD7–9) bind methylated histone tails associated with active chromatin, such as H3K4me1/2/3 and H3K36me3, or marks of repressive chromatin states (CHD3/4/5/9), such as H3K9me2/3 and H3K27me3 (Supplemental Table S1 and references therein). Moreover, it appears that poly(ADP-ribose)-dependent recruitment of CHD2–4 and CHD6/7 to chromatin plays an important role in DNA damage response (Luijsterburg et al. 2016; Smith et al. 2018; Moore et al. 2019; Rother et al. 2020).

Intradomain allosteric regulation

A great number of studies have addressed the biochemical and structural properties of CHD functional domains in isolation; however, their function in the context of the full-length proteins has remained ill defined. Recent studies, largely spearheaded by Bowman's group (Hauk et al. 2010; Hauk and Bowman 2011; McKnight et al. 2011; Nodelman and Bowman 2013; Nodelman et al. 2016, 2017, 2021; Tokuda et al. 2018; Winger et al. 2018), provide insight into the long-standing question of how the ATPase motor integrates information from the auxiliary domains to orchestrate engagement of the nucleosome substrates and catalyze accurate nucleosome sliding. An overarching model is emerging in which N-terminal chromodomains (yChd1) and PHD domains (CHD4) fold onto the bilobed ATPase motor, preventing nucleic acid binding and ATP hydrolysis (Fig. 2; Hauk et al. 2010; Ryan et al. 2011; Morra et al. 2012; Watson et al. 2012; Nodelman and Bowman 2013; Liu et al. 2014a; Nodelman et al. 2017, 2021). Consequently, in the absence of nucleosome substrates, key highly conserved residues across the central cleft of the ATPase motor are kept in an open conformation, thus preventing ATP hydrolysis (Hauk and Bowman 2011; Watson et al. 2012; Morra et al. 2012; Nodelman and Bowman 2013; Nodelman et al. 2017, 2021). Additionally, interactions with extranucleosomal elements such as linker DNA and protruding histone H4 tails trigger major conformational changes and dislodge the autoinhibitory domains, allowing the ATPase motor to adopt an active conformation (Fig. 2; Brehm et al. 2000; Hauk et al. 2010; Bouazoune and Kingston 2012; Farnung et al. 2017, 2020; Ludwigsen et al. 2017; Sundaramoorthy et al. 2017; Tokuda et al. 2018; Yan and Chen 2020; Nodelman et al. 2021).

Figure 2.

Autoinhibitory regulation and nucleosome sliding mechanism of SF2-type chromatin remodelers. (A) The ATPase motor of CHD enzymes is comprised of two RecA-like lobes ([red] lobe1, [blue] lobe2) that adopt a closed confirmation to form a contiguous surface that allows nucleic acid binding and efficient ATP hydrolysis. Biophysical and structural studies of yChd1, CHD2, and CHD4 revealed autoinhibitory domain organization whereby flanking domains fold onto the ATPase motor to prevent activation by improper substrates (Hauk et al. 2010; Ryan et al. 2011; Morra et al. 2012; Watson et al. 2012; Liu et al. 2014a). The N-terminal tandem chromodomains (CD; green) of yChd1 contain a linker helix with several conserved acidic residues that stacks onto the DNA-binding surface of lobe2 and negatively regulates yChd1 activity (Hauk and Bowman 2011; Narlikar et al. 2013). Full inhibition is achieved through folding of the “bridge” inhibitory domain adjacent to the ATPase motor (Bridge; orange). Moreover, autoinhibitory interaction of the ATPase motor with the C-terminal domains uncouples ATP hydrolysis from productive DNA translocation of Chd1p and CHD4, respectively (Hauk et al. 2010; Morra et al. 2012; Watson et al. 2012). Upon chromatin binding, interactions of CHD domains with linker DNA and protruding histone H4 tails (H4; gray) trigger major conformational change and dislodging of the autoinhibitory domains, allowing the ATPase motor to adopt an active conformation (Hauk and Bowman 2011; Nodelman et al. 2021). The interplay between the ATPase motor and flanking domains not only provides means for autoinhibition but also can promote substrate recognition and ATPase activity. In the case of CHD4, the N-terminal tandem chromodomains and PHD domains stimulate ATPase motor activity and DNA translocation (Morra et al. 2012; Watson et al. 2012), whereas the respective C-terminal DNA-binding regions of CHD1, CHD2, and CHD7 stimulate activity of the ATPase motor (Ryan and Owen-Hughes 2011; Bouazoune and Kingston 2012; Liu et al. 2014a). Thus, intradomain allosteric regulation keeps the enzyme in an inactive conformation to ensure substrate specificity, while on the other hand facilitating ATPase motor activity and DNA translocation. (GATED) yChd1 closed confirmation, (UNGATED) open confirmation, (light blue) nucleosome core, (gray) wrapped DNA. Features required for the efficient ATP binding and hydrolysis in lobe1 and lobe2 are indicated: ATP binding (magenta circles), ATP hydrolysis (green lines), and nucleic acid binding surfaces on each lobe (opaque areas) (Hauk and Bowman 2011). Top panel cartoon adapted with permission from Springer Nature from Mueller-Planitz et al. (2013), © 2013. Bottom panel republished with permission of Elsevier Science and Technology Journals from Hauk and Bowman (2011); permission conveyed through Copyright Clearance Center, Inc. (B) Historically, nucleosome crystal structure and earlier models of nucleosome sliding revealed that nucleosomal DNA can accommodate an extra base pair at the internal DNA location termed superhelix location 2 (SHL2) (Luger et al. 1997), which is a docking site for many ATP-dependent chromatin remodelers. Recent evidence shows that CHD, ISWI, and Snf2-type remodelers shift DNA discontinuously around the nucleosome (Farnung et al. 2017, 2020; Winger et al. 2018; Sabantsev et al. 2019; Yan and Chen 2020; Zhong et al. 2020; Nodelman et al. 2021). Remarkably, in an open conformation, the ATPase motor bound at the nucleosomal SHL2 position pulls 1–2 bp of DNA at the nucleosome entry site, creating a shift of only one DNA strand (tracking strand) with respect to the other (guide strand), resulting in a twist defect. Subsequent ATP binding induces closed confirmation of the ATPase motor, pulling the entire base pair of the DNA at SHL2 and forcing the DNA twist defect toward the nucleosomal exit site. Upon ATP hydrolysis, the motor readopts the open confirmation again, priming the nucleosome for the next translocation cycle. This sequential, ATP-driven conformational transition between open and closed states propels the DNA around the nucleosome histone core (Liu et al. 2017; Farnung et al. 2017, 2020; Armache et al. 2019; Nodelman et al. 2021; Yan and Chen 2020). Republished with permission of Elsevier Science and Technology Journals from Bowman (2019); permission conveyed through Copyright Clearance Center, Inc.

Intriguingly, the interplay between the ATPase motor and flanking domains not only permits autoinhibition, but also promotes substrate recognition and ATPase activity (Morra et al. 2012; Watson et al. 2012) (Ryan et al. 2011; Bouazoune and Kingston 2012; Liu et al. 2014a). Thus, intradomain allosteric regulation keeps the enzyme in an inactive conformation to ensure substrate specificity, while facilitating ATPase motor activity and DNA translocation once specific interactions with chromatin substrate are consolidated.

Nucleosome binding and translocase activity

Like other members of the SF2 helicase family, CHD remodelers are intrinsically active DNA translocases that harness the energy of ATP hydrolysis to propel DNA around the octamer and mobilize nucleosomes (Hauk et al. 2010; Mueller-Planitz et al. 2013; Nodelman and Bowman 2013; Nodelman et al. 2016; Farnung et al. 2017; Yan and Chen 2020). Recent evidence shows that SNF2-type remodelers shift DNA discontinuously, as DNA movement at the nucleosome entry site precedes that of the exit site (Farnung et al. 2017; Sabantsev et al. 2019; Zhong et al. 2020), inducing a “twist defect” of the nucleosomal DNA (Fig. 2; Winger et al. 2018; Farnung et al. 2020; Yan and Chen 2020; Nodelman et al. 2021).

Upon chromatin association, ATPase motor anchoring on the nucleosome and interactions of diverse CHD auxiliary domains with chromatin features (histone H4 tails, linker DNA, or other chromatin regulators) stabilize binding, prime the nucleosome for remodeling, and promote ATPase activity (Farnung et al. 2017, 2020; Liu et al. 2017; Armache et al. 2019). The concomitant cycles of ATP binding, hydrolysis, and dissociation of the hydrolysis product trigger a succession of conformational changes of the enzyme, promoting DNA translocation and nucleosome remodeling (Fig. 2; Mueller-Planitz et al. 2013; Yan and Chen 2020). The process is driven by the binding affinity of the remodeler for the nucleosome substrates, which, upon remodeling, decreases, leading to the release of the enzyme, thus marking the end point of the remodeling reaction (Längst and Manelyte 2015).

Functionally, chromatin remodelers alter histone–DNA interactions by promoting assembly, disruption, and repositioning of nucleosomes and incorporation of histone variants (Mueller-Planitz et al. 2013). The majority of CHDs (1/2/3/4/7/8/9) appear to predominantly slide nucleosomes (Lusser et al. 2005; Liu et al. 2014a; Salomon-Kent et al. 2015; Hoffmeister et al. 2017; Manning and Yusufzai 2017), whereas CHD5 and CHD6 disrupt nucleosomes without causing sliding (Quan and Yusufzai 2014; Manning and Yusufzai 2017). Notably, CHD1, CHD2, and ISWI-type remodelers are the only members of the SF2 helicases able to assemble periodic nucleosome arrays in an ATP-dependent manner (Lusser et al. 2005; Stockdale et al. 2006; Liu et al. 2014a). Importantly, CHD enzymes can also exhibit distinct remodeling activities in the absence of accessory subunits of the complexes in which they reside (CHD3/4 and NuRD) (Hoffmeister et al. 2017).

Collectively, biophysical and structural studies reveal that CHD remodelers display intricate conformational intradomain allosteric regulation and seem to share a common translocation mechanism with other SF2 ATPases, yet individual CHDs exhibit largely distinct nucleosome binding and remodeling activities (Quan and Yusufzai 2014; Liu et al. 2014a; Manning and Yusufzai 2017). This aligns with the nonredundant functions they fulfil in chromatin organization and transcription, resulting in distinct phenotypes in animal models and human disease upon their disruption.

Multimeric protein complexes

CHDs function either as monomers or constituents of multimeric enzymatic complexes (Micucci et al. 2015) that remodel chromatin and catalyze deposition or removal of histone modifications (Supplemental Table S1 and references therein). The nucleosome remodeling activity of CHD proteins facilitates interaction of histone methyltransferases (KMTs) and deacetylases (HDACs) with histone tail substrates. The combinatorial assembly of the noncatalytic subunits confers functional specificity to the CHD remodelers (Wang and Zhang 2001; Hoffmeister et al. 2017). The subunit composition is governed by the cell type and developmental stage (Micucci et al. 2015; Nitarska et al. 2016), and, in response to cellular cues, CHD enzymes are tethered to particular genomic loci where they exert specific functions (Lai and Wade 2011).

Biochemically the NuRD complex is characterized best. It harbors CHD3/4/5 chromatin remodelers and HDAC1 and HDAC2 catalytic subunits, next to several auxiliary subunits (Supplemental Table S1; Tong et al. 1998; Wade et al. 1998, 1999; Xue et al. 1998; Zhang et al. 1998, 1999; Kim et al. 1999; Hoffmeister et al. 2017; Tencer et al. 2017). The concerted action of CHD3/4 (Mi2-α/β) and HDAC subunits of the NuRD complex results in histone deacetylation, chromatin compaction, and transcriptional repression (Tong et al. 1998; Xue et al. 1998). Notably, the core CHD3/4/5 remodelers can form distinct NuRD complexes with specific functions (Nitarska et al. 2016). On the other hand, activators of transcription such as CHD8 can also regulate transcription as part of the histone methyltransferase subcomplex termed WAR (WDR5, ASH2L, and RBBP5) (Supplemental Table S1; Yates et al. 2010). Curiously, experimental evidence suggests that CHD family proteins also interact with each other (CHD7–CHD8 [Batsukh et al. 2010] and CHD1–CHD2 [Luijsterburg et al. 2016]) raising an interesting possibility of direct functional cooperativity in regulating chromatin-templated processes.

Chromatin organization and transcriptional regulation

Every cell type is characterized by a distinct set of genes that can be transcriptionally on, off, poised, or oscillating, resulting from the complex interplay between genomic regulatory elements and cognate binding factors (Signolet and Hendrich 2014). Access of specific TFs and chromatin regulators to underlying DNA is determined by the nucleosome positioning and an array of histone and DNA modifications (Lai and Pugh 2017).

Association with promoters and enhancers

RNA polymerase II (RNAPII) transcription typically initiates at the transcription start site (TSS) embedded within the core promoter sequence that is usually devoid of nucleosomes, termed the nucleosome-free region (NFR), which is demarcated by the +1 nucleosome that represents a major obstacle for RNAPII productive elongation (Fig. 3). The CHDs’ activity enables RNAPII to overcome nucleosomal barriers during transcription (Li et al. 2007).

Figure 3.

Regulation of the transcription cycle. (A) CHD enzymes are involved at different stages of transcription: initiation, promoter escape, elongation, splicing, and termination, as well as higher-order chromatin organization. The promoter nucleosome-free region (NFR) is demarcated by the ±1 nucleosomes (purple). CHDs also regulate nucleosome accessibility at enhancer elements and modulate CTCF-mediated chromatin organization. (TSS) Transcription start site, (TTS) transcription termination site, (TF) transcription factors, (GTFs) general transcription factors, (U1/2/4/5/6) spliceosome, (orange rectangle) enhancer elements, (blue rectangle) CTCF-binding site. RNA polymerase II (RNAPII) phosphorylation status of the C-terminal domain is represented by the red (serine 5 phosphorylation-paused RNAPII) and green (serine 2 phosphorylation-elongating RNAPII) circles. (B) CHD enzymes can both promote and inhibit transcription of different subsets of genes in a context-dependent manner. CHD8 and linker histone H1 (blue circle) form a trimeric repressive complex with either P53 or β-catenin (green oval) to repress target gene expression.

The landmark study by the group of Matthieu Gérard (Dieuleveult et al. 2016) comprehensively charted chromatin binding of CHD proteins and their role in the gene expression regulation in mouse ES cells. It revealed a dynamic interplay between remodelers at promoter-proximal nucleosomes and a variable correlation between promoter enrichment of remodelers and transcriptional activity of the corresponding genes (Fig. 3). Moreover, it demonstrated dynamic CHD binding at gene promoters with different NFR regions, serving different functional purposes. Specifically, promoter occupancy of CHD1/2/9 directly correlates with H3K4me3 levels, CHD8 binds promoters with intermediate levels of H3K4me3, whereas CHD4/6 bind promoters irrespective of the H3K4me3 status. The observed binding profile of CHD4 aligns with reports that the NuRD complex displays a broader chromatin association in ES cells and contributes to both transcriptional activation and repression (Hu and Wade 2012; Bornelöv et al. 2018). Additional studies in primary and cancer cells show enrichment of CHD5 at the promoter-proximal regions largely devoid of H3K4me3 or H3K27me3 (Egan et al. 2013; Paul et al. 2013), and CHD3/NuRD enrichment at H3K9 acetylated (H3K9ac) target genes (Tencer et al. 2017). In contrast, CHD2 correlates with the H3K36me3 mark and is distributed throughout the gene body of transcribed genes and at transcription termination sites (TTSs) (Dieuleveult et al. 2016; Rom et al. 2019), likely modulating mRNA processing and RNAPII elongation and termination, as previously shown for yChd1 (Fig. 3; Murawska and Brehm 2011). This seminal study highlights several important general concepts. First, that CHD chromatin remodelers exhibit position-specific nucleosome binding tailored to the local chromatin architecture and DNA composition (CpG content). Second, that individual CHD remodelers can function as transcriptional activators of one subset of genes and as repressors of another.

Besides binding to promoter-proximal regions, CHDs remodel nucleosomes at enhancers to modulate enhancer–promoter communication and transcription initiation (Fig. 3). For instance, the CHD7 association with active and/or poised enhancers is important for RNAPII transcription in diverse cell lineages (Bajpai et al. 2010; Schnetz et al. 2010; Marie et al. 2018). Moreover, several lines of evidence show that CHD4/NuRD enhancer occupancy is critical for the transcriptional regulation in mouse ES cells (Bornelöv et al. 2018) and during T-cell development (Williams et al. 2004). Notably, nuclear receptors mediate CHD enhancer recruitment to regulate expression of androgen receptor (AR)-responsive genes (CHD1 and CHD8) (Menon et al. 2010; Metzger et al. 2016), progesterone (PR)-responsive genes (CHD8) (Ceballos-Chávez et al. 2015), and PPARγ- and C/EBPβ-regulated genes (CHD7) (Kita et al. 2018). These studies highlight different means of CHD recruitment to enhancer elements, eliciting distinct transcriptional outcomes governed by cellular and developmental contexts.

Transcriptional elongation and splicing

Current evidence supports a model in which chromatin structure (Schwartz et al. 2009) and RNAPII elongation rate (Batsché et al. 2006) play a critical role in splicing. The CHDs appear to act as an interface between chromatin epigenetic information and the RNAPII and pre-mRNA processing machinery (CHD1 [Sims et al. 2007; Lee et al. 2017] and CHD7 [Bélanger et al. 2018]), modulating both the RNAPII elongation (CHD8 [Rodríguez-Paredes et al. 2009] and dKisL [Srinivasan et al. 2008]) and splicing efficiency (CHD1 [Sims et al. 2007; Lee et al. 2017]). Furthermore, direct association with nascent RNA is demonstrated for dMi-2 (Murawska et al. 2011) and CHD1/2/4/7 (Hendrickson et al. 2016), providing additional means by which CHD enzymes may affect cotranscriptional processing. However, it remains unclear whether any of the CHD–RNA interactions affect biological functions of RNA molecules or CHD recruitment to specific genomic locations.

Higher-order chromatin regulation

In addition to shaping the local chromatin landscape and modulating transcription of individual genes, CHD proteins act globally to regulate higher-order chromatin structure, thereby regulating genome stability, transcriptional regulation, and DNA damage repair.

The role of CHD1 in maintaining high-order chromatin organization and genome stability has become evident in prostate cancer, where CHD1-null tumors display distinctive patterns of chromoplexy (Stephens et al. 2011; Baca et al. 2013). Moreover, several studies show that CHD1 (Boginya et al. 2019), CHD4 (Fasulo et al. 2012), and CHD6 (Sancho et al. 2015) appear to be required for the loading of cohesin, a critical mediator of transcription regulation and genome stability that is frequently mutated in cancer.

The CTCF insulator protein modulates transcription and orchestrates long-distance chromatin interactions (Ling et al. 2006), facilitating genome partitioning into functionally distinct regions called topologically associated domains (TADs) (Dixon et al. 2012). Biochemical studies reveal CHDs’ association with CTCF (CHD7 and CHD8) (Ishihara et al. 2006; Allen et al. 2007), and with the human CTCF paralog BORIS (brother of the regulator of imprinted sites) (CHD8) (Nguyen et al. 2008). Moreover, CHDs occupy many CTCF-binding sites (CHD8) and modulate insulation and topological organization of CTCF target loci (CHD8 [Ishihara et al. 2006] and CHD6 [Sancho et al. 2015]) (Fig. 3). Curiously, in mouse ES cells, the CHD4/ChAHP complex prevents CTCF binding to sequences within the transposable elements and thereby safeguards the evolutionarily conserved spatial chromatin organization (Kaaij et al. 2019). Given that CHD9 appears to play a role in regulation of chromatin organization in oocytes (Ooga et al. 2018), research into the role of the CHD family in higher-order chromatin organization is warranted.

Transcriptional outcomes

As demonstrated in diverse model organisms, CHD enzymes regulate nucleosome density and chromatin organization at promoter and enhancer elements, which can lead to different transcriptional outcomes (Fig. 3). For instance, some, such as CHD1, are required for RNAPII transcription (Krogan et al. 2002; Simic 2003; Srinivasan et al. 2005; Warner et al. 2007; Skene et al. 2014; Dieuleveult et al. 2016; Baumgart et al. 2017) and play a role in transcription initiation (Lin et al. 2011) and in RNAPII-directed turnover of promoter-proximal nucleosomes, promoter escape, and subsequent elongation and splicing (Sims et al. 2007; Skene et al. 2014). Similarly, mounting evidence shows that mammalian CHD2 (Rom et al. 2019), CHD6 (Lutz et al. 2006), CHD8 (Rodríguez-Paredes et al. 2009; Ceballos-Chávez et al. 2015), and CHD9 (Shur et al. 2006a; Lee and Stallcup 2017; Alendar et al. 2020; Newton and Pask 2020; Yoo et al. 2020) potentiate transcription of distinct target genes in diverse cell types. CHD enzymes also contribute to transcriptional repression, best exemplified by the concerted action of chromatin remodeling and histone deacetylase catalytic subunits of the NuRD complex, which generally leads to the generation of hypoacetylated, densely packed chromatin (Lai and Wade 2011) and facilitates PRC2-mediated silencing of target genes in ESCs (Reynolds et al. 2012).

The capacity of CHD enzymes to function as either transcriptional activators or repressors is likely caused by the mode of CHD recruitment to target loci, the local chromatin landscape, and the subunit composition of the multimeric protein complexes in which they reside (Fig. 3). For instance, CHD4 alone promotes transcription in T cells (Williams et al. 2004) and ES cells (Dieuleveult et al. 2016; Bornelöv et al. 2018), whereas a great number of studies reviewed here show that in a NuRD complex it contributes to transcriptional repression. Similarly, in addition to its role in promoting transcription (Yates et al. 2010; Ceballos-Chávez et al. 2015), CHD8 inhibits β-catenin (Nishiyama et al. 2004) and P53 (Nishiyama et al. 2009) transactivation by forming a repressive trimeric complex with linker histone H1.

In some instances, different CHD remodelers are required for distinct stages of transcription—e.g., to regulate initiation (dBRM), promoter clearance (dKisL), and elongation (dCHD1) (Srinivasan et al. 2005)—or act synergistically with other chromatin regulators to modulate transcription in a given cellular context (CHD7 and PBAF [Bajpai et al. 2010], as well as CHD4 and Polycomb repressive complex 2 [PRC2] [Sparmann et al. 2013]). Moreover, different chromatin remodelers with opposing functions can engage in a “tug-of-war” action to fine-tune transcription in response to extracellular cues—SWI/SNF and NuRD/Mi-2 (Ramirez-Carrozzi et al. 2006; Gao et al. 2009), esBAF and NuRD (Yildirim et al. 2011), and Cockayne syndrome protein B (CSB) and NuRD (Xie et al. 2012).

As discussed in the following section, this mode of action may be particularly relevant in embryonic development, where cells rapidly switch transcriptional programs during lineage commitment and differentiation.

Stem cell maintenance, lineage commitment, differentiation, and development

Accurate transcriptional control is paramount for organismal development and requires cross-talk between myriad transcription factors and chromatin regulators that act in concert to control expression of the right genes at the right time. A great body of work discussed below highlights the important roles CHD enzymes play in stem cell self-renewal and pluripotency, as well as in directing cell fate decisions in development (Fig. 4).

Figure 4.

Cell fate specification and lineage commitment. CHD functions are important in diverse stem cell populations (annotated by yellow rectangles). Some CHDs are important for the emergence and self-renewal of stem cells (circular arrow), and others are vital for lineage commitment and differentiation (straight arrows). Distinct CHDs can promote (black) or inhibit (red) differentiation. For simplicity, not all intermediate precursors and terminally differentiated cells are depicted. (ESC) Embryonic stem cell, (HSC) hematopoietic stem cell, (MSC) mesenchymal stem cell, (NSC) neural stem cell, (NCSC) neural crest stem cell, (DFC) dental follicle cells.

The majority of the Chd germline knockouts in animal models display severe early developmental phenotypes, pointing to their nonredundant functions (Supplemental Table S1 and references therein). For instance, Chd7 and Chd8 germline knockout mice are embryonic-lethal due to the widespread P53-mediated apoptosis (Nishiyama et al. 2009; Nostrand et al. 2014). Notably, Chd5 and Chd9 knockouts are viable and provide an exception to the rule, with Chd9^−/− animals having no overt phenotype (Alendar et al. 2020), and Chd5^−/− displaying impaired spermatogenesis and fertility (Li et al. 2014). There are contrasting reports of phenotypes observed in Chd2 knockout mice, ranging from perinatal lethality (Marfella et al. 2006, 2008; Kulkarni et al. 2008) to brain-specific developmental defects (Meganathan et al. 2017), or absence of any overt phenotypes (Supplemental Table S1; Rom et al. 2019). Single-exon deletion Chd6 mice are viable but display motor coordination impairment (Lathrop et al. 2010). Although no germline knockout mice are reported to date, both Chd1 (Gaspar-Maia et al. 2009) and Chd4 (O'Shaughnessy-Kirwan et al. 2015) clearly play critical roles in development, as their depletion leads to impaired blastocyst formation, preimplantation defects, and embryonic lethality.

As discussed below, most of our insight into the roles of CHDs in guiding cell fate decisions and development comes from the studies of primary cell cultures and tissue-specific conditional knockouts (Supplemental Table S1). Gene dosage seems to be important, and haploinsufficiency can cause pathogenesis in both mice and humans, indicating that functions of chromatin remodelers in specific cellular processes are rate-limiting.

CHDs in stem and progenitor cells

ES cells are self-renewing, pluripotent cells characterized by high expression of chromatin remodeling enzymes (Efroni et al. 2008), pervasive global transcription, and relatively open chromatin organization (Orkin and Hochedlinger 2011; Percharde et al. 2017). The gene expression heterogeneity of ESCs creates clonal diversification and transcriptional plasticity, securing accurate response to differentiation cues while maintaining the capacity for self-renewal (Ram and Meshorer 2009). These features of pluripotent ESCs are at least in part regulated by the ATP-dependent chromatin remodeling enzymes (Landry et al. 2008; Ho et al. 2009, 2011).

Importantly, CHD enzymes form intricate regulatory networks with core pluripotency factors, forming complexes with SOX2, OCT4, and NANOG (CHD1, CHD2, CHD7, and CHD3/4/5/NuRD) (Schnetz et al. 2009; Pardo et al. 2010; Reynolds et al. 2012; Gagliardi et al. 2013; Semba et al. 2017; Bornelöv et al. 2018). They also associate with promoters and enhancers of Sox2, Pou5f1, and Nanog to modulate their expression (CHD1, CHD2, CHD4, and CHD7) (Gaspar-Maia et al. 2009; Pardo et al. 2010; Schnetz et al. 2010; Reynolds et al. 2012; Gagliardi et al. 2013; Suzuki et al. 2015; Semba et al. 2017). Moreover, expression of Chd1, Chd9, and several subunits of the NURD complex is directly regulated by the core pluripotency factors (Kim et al. 2008; Orkin et al. 2008; Orkin and Hochedlinger 2011). As such, CHD enzymes clearly play an important role in modulating the ESCs’ core pluripotency regulatory circuitry.

In ESCs, CHDs appear to occupy distinct enhancers, as well as promoters of both active and bivalent genes marked by H3K4m3 and H3K4me3/H3K27me3, respectively (Gaspar-Maia et al. 2009; Schnetz et al. 2009; Lin et al. 2011; Dieuleveult et al. 2016). Bivalent developmental genes are largely kept silent in ESCs but are poised for activation upon differentiation (Bernstein et al. 2006; Voigt et al. 2013). The CHD1-mediated (Konev et al. 2007; Gaspar-Maia et al. 2009; Baumgart et al. 2017) and CHD2-mediated incorporation of histone variants (H2A.Z and H3.3) (Harada et al. 2012; Shen et al. 2015b; Semba et al. 2017) facilitates establishment of bivalency and bookmarking of key developmental genes (Harikumar and Meshorer 2015). In turn, this promotes decommissioning of the pluripotency enhancers by CHD4/NuRD/LSD1 (Gehre et al. 2020) and concomitant shutdown of the ESC-specific transcription during differentiation (Reynolds et al. 2012; Whyte et al. 2012).

Transcriptional heterogeneity and plasticity of ESCs allow prompt cellular response to developmental cues. In this regard, CHD4 and CHD7 act as transcriptional rheostats to delimit expression of the lineage- and ES-specific genes, facilitating cell fate changes (Schnetz et al. 2010; Reynolds et al. 2012; O'Shaughnessy-Kirwan et al. 2015; Bornelöv et al. 2018). Consequently, in the absence of CHD4/NuRD, ESCs are unable to down-regulate pluripotency genes and accurately engage critical lineage-specific genes (Kaji et al. 2006; Reynolds et al. 2012).

Thus, the net effect of intricate CHD transcriptional regulatory mechanisms that operate in ESCs is twofold: curtailing expression of ESC-specific genes and core pluripotency factors while maintaining tight regulation of expression of key instructive factors required for rapid transcriptional response upon lineage commitment and differentiation.

Reprogramming

Chromatin remodelers also play critical roles during reprograming of somatic cells to induced pluripotency (iPS) by overexpression of Oct4, Sox2, Klf4, and cMyc (OSKM) factors (Orkin and Hochedlinger 2011). Notably, in both mouse and human somatic cells, the NuRD complex appears to modulate iPS reprogramming, although further studies are required to clarify its exact role, as current literature provides contrasting experimental evidence suggesting that it can both inhibit (Luo et al. 2013; Rais et al. 2013) and promote (Santos et al. 2014) reprogramming. The contribution of other CHD proteins to iPS reprograming is less understood; while CHD1 is critical for efficient reprogramming of mouse fibroblasts into iPS cells, CHD2 and CHD7 appear to be dispensable (Gaspar-Maia et al. 2009; Schnetz et al. 2009; Semba et al. 2017). Recently, iPS cells were successfully derived from patients carrying heterozygous mutations in CHD7 (CHARGE syndrome) and CHD8 (autism spectrum disorders [ASDs]), indicating that, while their heterozygosity appears to drive human neurodevelopmental pathologies, gene dosage is not critical for somatic cell reprogramming (Chai et al. 2018; Deneault et al. 2018). This holds great promise as a model system for studying underlying molecular perturbations.

Lineage commitment and differentiation

The iconic concept of developmental specification as an epigenetic landscape introduced by Waddington (1957) likens differentiation of a pluripotent cell to a ball on the top of the hill going down the cascade of branching ridges and valleys, representing specific developmental trajectories leading to different cell fates. As discussed below, CHDs alter chromatin landscape and modulate barriers between distinct cell fates, either fortifying commitment toward a specified developmental trajectory (increasing the barrier) or lowering the threshold for differentiation and permitting random sampling of different cellular fates (reducing the barrier).

The association of CHDs with chromatin is dynamic during differentiation. Some family members (CHD1 and CHD7) appear important for both stemness and differentiation of ESCs, whereupon their chromatin occupancy alters, tracking the transcriptional activation and H3K4me3 redistribution to differentiation-specific genes (Gaspar-Maia et al. 2009; Schnetz et al. 2009, 2010). In contrast, CHD2 remains on a subset of bivalent genes throughout differentiation, likely preserving resolution of their poised chromatin state for later developmental stages (Semba et al. 2017).

Beyond their importance in ESC biology, the roles of CHDs extend to other embryonic and adult stem and progenitor cells (Fig. 4). Some, such as CHD1, have a general role in the expansion of stem and progenitor populations during developmental transitions (Percharde et al. 2017), or fulfil critical roles in lineage commitment and differentiation of the diverse stem cells and early progenitors (CHD4/NuRD) (Fig. 4; Williams et al. 2004; Kaji et al. 2006; Kashiwagi et al. 2007; Yoshida et al. 2008; Hung et al. 2012; Reynolds et al. 2012; Luo et al. 2013; O'Shaughnessy-Kirwan et al. 2015; Bornelöv et al. 2018; Hirota et al. 2019). Additionally, individual CHD enzymes can also contribute to the development of specific germ layers and lineages. For instance, CHD1/2/9 are important for mesoderm formation and lineage commitment of mesenchymal stem cells (MSCs) toward osteoblasts (CHD9 [Shur and Benayahu 2005; Marom et al. 2006; Shur et al. 2006a,b] and CHD1 [Baumgart et al. 2017]) and adipocytes (CHD1) (Baumgart et al. 2017). Similarly, CHD7 and CHD8 appear to be important for the osteogenic differentiation of neural crest cells (NCCs) (Liu et al. 2020; Fan et al. 2021), whereas CHD7 appears to be required for both oligodendrocyte maturation and osteoblast bone formation (He et al. 2016), indicating that CHD enzymes may act as developmental hubs of diverse lineages.

Several CHDs are also implicated in myogenic differentiation and muscle organogenesis (Fig. 4), as exemplified by the critical role of CHD2 in differentiation of myogenic progenitors in mice (Harada et al. 2012). Evidence is emerging that the CHD4/NuRD complex is an important regulator of cardiac and skeletal muscle progenitor cell identity (Mammen et al. 2009; Arco et al. 2016; Sreenivasan et al. 2020), as evident from Chd4-null cardiomyocytes that form a hybrid muscle tissue in mice (Wilczewski et al. 2018), and the notion that mutations in both CHD4 and CHD7 are implicated in congenital heart disease (Homsy et al. 2015). Similarly, CHD4/NuRD is critical for the self-renewal and multilineage differentiation potential of hematopoietic stem cells (Wade et al. 1999; Fujita et al. 2004; Williams et al. 2004; Yoshida et al. 2008) and fulfils important roles in guiding lymphocyte development (B cell [Fujita et al. 2004; Arends et al. 2019; Yoshida et al. 2019] and T-cell [Yasui et al. 2002; Williams et al. 2004; Naito et al. 2007]), governing the postnatal switch in β-globin expression (Xu et al. 2010) and establishment of immune tolerance (Tomofuji et al. 2020).

Other CHDs play important roles as well (Fig. 4). For instance, while some are vital for the emergence (CHD1 [Koh et al. 2015] and CHD8 [Shooshtarizadeh et al. 2019; Nita et al. 2021; Tu et al. 2021]) and differentiation of hematopoietic stem and progenitor cells (HSPCs) toward erythroid and myeloid precursors (CHD2) (Nagarajan et al. 2009; Varnoosfaderani et al. 2020), others appear to have an opposing role (CHD7) (Fig. 4; Hsu et al. 2013, 2020). Importantly, patient genomic data and studies in animal models point to driver roles of CHD2 (Berquam-Vrieze et al. 2011; Quesada et al. 2012; Rodríguez et al. 2015; Kim et al. 2016), CHD7 (Zhen et al. 2017), CHD8 (Shingleton and Hemann 2015; Shen et al. 2015a; Grande et al. 2019; Newell et al. 2019), and CHD9 (Mikkers et al. 2002; Shou et al. 2006) of various hematopoietic malignancies (Supplemental Table S1).

Neuronal development and pathologies

During neurodevelopment, CHD remodelers regulate chromatin accessibility and gene expression of key transcriptional networks converging on the SHH (Feng et al. 2017), Notch (Jones et al. 2015), FGF (Hurd et al. 2010; Engelen et al. 2011), TGF-β (Fueyo et al. 2018), WNT/β-catenin (Sakamoto et al. 2000; Nishiyama et al. 2004), and PI3K (Micucci et al. 2015) signaling pathways and regulating P53 transactivation (Nishiyama et al. 2009; Nostrand et al. 2014), thereby controlling many aspects of brain development and functionality. Importantly, in the last decade, a great number of whole-genome sequencing studies identified mutations of CHD enzymes in neurodevelopmental disorders, with haploinsufficiency as a mechanism driving these pathologies (Supplemental Table S1 and references therein). Phenotypic characterization of patient-derived cell lines and CHD animal models provides valuable insight into the underlying molecular mechanisms that govern development of these pathologies.

CHDs have distinct yet overlapping roles in neural stem cell niches in the cerebrum (subventricular zone [SVZ]) and hippocampus (subgranular zone of the dentate gyrus [SGZ/DG]; CHD3/4/5/NuRD) (Egan et al. 2013; Sparmann et al. 2013; Nitarska et al. 2016), where they regulate neurogenesis and neuronal differentiation (CHD2 [Shen et al. 2015b; Kim et al. 2018; Lalli et al. 2020], CHD4 [Sparmann et al. 2013], and CHD5 [Egan et al. 2013]) and neuronal migration and maturation (CHD3 and CHD5) (Fig. 5; Egan et al. 2013; Nitarska et al. 2016). Likewise, both CHD7 (Engelen et al. 2011; Feng et al. 2013, 2017; Micucci et al. 2014; Jones et al. 2015; He et al. 2016; Yao et al. 2020) and CHD8 (Sakamoto et al. 2000; Nishiyama et al. 2004; Ronan et al. 2013; Sugathan et al. 2014; Durak et al. 2016; Katayama et al. 2016; Kawamura et al. 2021) are vital for brain development and regulate functions of stem cells in diverse brain regions. Thus, CHD enzymes play vital roles both in brain development (CHD2, CHD3/4/5, CHD7, and CHD8) and adult neurogenesis (CHD2, NuRD, and CHD7). It is becoming increasingly evident that some of the CHD7/8-mediated neurodevelopmental disorders develop, at least in part, as a result of dysregulation of P53 transactivation during brain development (Nostrand et al. 2014; Marie et al. 2018; Wade et al. 2019; Hurley et al. 2021).

Figure 5.

CHDs regulate different aspects of neurodevelopment. (A, top) CHD enzymes are vital for the development and functionality of different parts of the brain: the cerebral neocortex (CHD2, CHD3/4/5/NuRD, CHD7, and CHD8), cerebellum (CHD4/NuRD, CHD6, CHD7, and CHD8), hypothalamus (CHD2), thalamus (CHD8), and hippocampus (CHD4/5/NuRD, CHD7, and CHD8). They regulate neural stem cell niches in the cerebrum (subventricular zone [SVZ]) and hippocampus (subgranular zone of the dentate gyrus [SGZ/DG]) and are essential for proper neuronal differentiation and migration. CHD7 is important for the development of the inner ear and olfactory system and regulates migration of the neuroblasts that emerge from the neural stem cells in the SVZ and proceed through the rostral migratory stream (RMS) to the olfactory bulb (OB). The cartoon represents the sagittal section of the rodent brain and highlights CHD functions in the respective brain regions (lines). (Bottom) CHDs are required at various stages of cortical neurogenesis. During the early stages of cortical neurogenesis, symmetrical division of the neuroepithelial cells (NEs) in the ventricular zone (VZ) expands their pool, whereas asymmetrical division gives rise to apical neural precursor cells (NPCs), such as apical radial glial cells (aRGs), and pioneer neurons. Apical progenitor cells give rise to neurons through basal intermediate precursors (bIP) and radial glial cells (bRG) that reside in the subventricular zone (SVZ). Neurons use aRG and bRG fibers to migrate to the specified layer in the cortical plate (CP), where they establish synaptic connections with subplate layer neurons during maturation. CHD remodelers involved at distinct stages are highlighted bellow. (MZ) Marginal zone, (IZ) intermediate zone. Cortical neurogenesis cartoon representation adapted with permission from Sokpor et al. (2018). (B, top) Synaptogenesis between neurons initiates during embryogenesis, but synaptic plasticity proceeds throughout life. Synaptic plasticity is the property of synapses to strengthen or weaken in response to changes in both the amplitude and the temporal dynamics of neuronal activity. Sensory inputs and intrinsic brain activity can effect long-term changes in synaptic efficacy and eventually increase or decrease neuronal connectivity by modulating the number of synapses (Bourgeron 2015). Some CHD enzymes regulate neuronal connectivity (CHD4, CHD5, and CHD8) and excitability (CHD2 and CHD8), whereas others regulate synaptic vesicle recycling (KisL) and expression of genes essential for synaptic homeostasis and plasticity (CHD8). (Bottom) Apoptosis and pruning are required for the maintenance and refinement of the neural circuitries, and CHD7/8 inhibition of P53-mediated apoptosis plays an active role in this process. Dysregulation of CHD enzyme function affects synaptogenesis and apoptosis and leads to aberrant neurogenesis and connectivity, which results in the emergence of diverse neurodevelopmental disorders (NDDs).

Moreover, CHD enzymes are also critical for the maturation of glial cells and myelination of the central nervous system (CNS; CHD7/8) (Feng et al. 2013; He et al. 2016; Doi et al. 2017; Marie et al. 2018; Zhao et al. 2018a) and peripheral nervous system (CHD4) (Fig. 5; Srinivasan et al. 2006; Hung et al. 2012). Intriguingly, CHD8 can functionally compensate for the CHD7 loss at least partially during differentiation of oligodendrocyte precursor cells (OPCs) (Marie et al. 2018). These findings, together with evidence of CHD7–CHD8 association (Batsukh et al. 2010), suggest a requirement for concerted action of these remodelers.

Beyond their role in regulating cell fate decisions during neurogenesis, CHD enzymes are also important for neuronal excitability (CHD2 [Meganathan et al. 2017] and CHD8 [Platt et al. 2017; Jung et al. 2018; Ellingford et al. 2021]), and for synaptic connectivity and strength (CHD4 [Yamada et al. 2014], CHD5 [Egan et al. 2013; Pisansky et al. 2017], and CHD8 [Platt et al. 2017; Jung et al. 2018; Ellingford et al. 2021]). The CHD8 regulation of synaptic homeostasis and plasticity is vital for neocortical (Kweon et al. 2021) and cerebellar (Kawamura et al. 2021) development and is dysregulated in ASDs, as discussed below (Fig. 5).

Neurodevelopmental disorders

As highlighted above, CHD enzymes are critical for the development of the cerebrum (CHD2, CHD3/4/5/NuRD, CHD7, and CHD8), cerebellum (CHD4/NuRD, CHD6, CHD7, and CHD8), hypothalamus (CHD2), thalamus (CHD8), and hippocampus (CHD4/5/NuRD, CHD7, and CHD8) (Fig. 5). Disruptions of CHD functions can result in different numbers of mature neurons and neuroglia, mislocalization of neurons, and altered synaptic activity in various parts of the brain, thereby affecting cognitive functions, the integration of sensory information, memory, learning, and motor coordination.

Therefore, it comes as no surprise that a large number of recent whole-genome sequencing studies of patients with NDDs (ASD, CHARGE, schizophrenia, intellectual disability, developmental disorders, and epilepsy) identified de novo heterozygous mutations, SNVs, small deletions, and copy number variations in CHD1, CHD2, CHD3, CHD4, CHD6, CHD7, and CHD8 (Supplemental Table S1 and references therein; O'Roak et al. 2012a,b, 2014; Sanders et al. 2012; Bernier et al. 2014; Iossifov et al. 2014; Krumm et al. 2014; Study 2017; Coe et al. 2019).

The de novo loss-of-function mutations in CHD8 are recurrent in ASD patients, and Chd8 heterozygous mice faithfully recapitulate a variety of symptoms found in these patients (Durak et al. 2016; Katayama et al. 2016; Gompers et al. 2017; Platt et al. 2017; Jung et al. 2018; Suetterlin et al. 2018). Likewise, mounting evidence indicates that dysregulation of CHD7/SOX2 gene networks is driving human Alagille, Pallister-Hall, and Feingold syndromes (Engelen et al. 2011). Moreover, CHD7 regulates development of diverse tissues of neuronal origin that are affected in the autosomal dominant congenital CHARGE syndrome (Micucci et al. 2015), and 90% of the CHARGE patients harbor CHD7 heterozygous mutations (Supplemental Table S1 and reference therein; Ravenswaaij-Arts and Martin 2017). Importantly, Chd7 germline heterozygous and tissue-specific knockout mice display most of the features observed in CHARGE individuals (Supplemental Table S1). Biochemical studies reveal that patient-specific mutations of CHD7 chromodomains modulate ATP-dependent remodeling activity in a mutation-dependent manner (Bouazoune and Kingston 2012; Yan et al. 2020). Moreover, as reviewed earlier, dysregulation of the CHD7–P53 axis plays an important role in CHARGE etiology, and P53 mutations are reported in 10%–30% of CHARGE individuals (Nostrand et al. 2014).

Similarly, several recent studies reveal that haploinsufficiency of CHD1 (Pilarowski et al. 2018), CHD2 (Pinto et al. 2010; Capelli et al. 2012; Rauch et al. 2012; Carvill et al. 2013; Suls et al. 2013; Chénier et al. 2014; Lund et al. 2014; Thomas et al. 2015; Wilson et al. 2021), and CHD5 (Pisansky et al. 2017; Network et al. 2021) contributes to the emergence of NDDs with diverse phenotypic features, whereas recurrent de novo CHD3 and CHD4 missense mutations are drivers of novel human neurodevelopmental syndromes termed Snijders Blok-Campeu syndrome (Blok et al. 2019; Drivas et al. 2020) and Sifrim–Hitz–Weiss syndrome (SIHIWES) (Weiss et al. 2016, 2020), respectively. These CHD mutations are likely gene-disrupting mutations (LGD) that result in alteration or complete loss of protein function. Notably, gene dosage seems to play a role in the etiology of pathologies, as both inactivation and duplication of CHD3 (Drivas et al. 2020) and CHD4 (Coe et al. 2019) can result in NDDs.

Importantly, as discussed in the following section, it is increasingly appreciated that similar signaling and metabolic pathways are dysregulated in both NDDs and human cancers. This holds promise for the repurposing of anticancer drugs for the treatment of some NDD symptoms.

CHDs in cancer

The genome sequencing endeavor of human cancers has identified numerous somatic mutations, copy number alterations, and chromosomal rearrangements of chromatin proteins (Flavahan et al. 2017), including CHD family members (Supplemental Table S1 and references therein). These mutations affect mostly only one allele, likely causing protein inactivation, providing support for the notion that haploinsufficiency drives pathogenesis both in human cancers and neurodevelopmental disorders. Additionally, epigenetic silencing of CHDs by promoter hypermethylation is documented across various types of cancers (Supplemental Table S1). Notably, complete inactivation is occasionally observed in prostate cancer (CHD1; homozygous deletion) and neuroblastoma (CHD5; inactivation of both alleles) (Supplemental Table S1). Given that intradomain allosteric regulation appears to be a general mechanism controlling chromatin association and enzymatic activity of CHD enzymes, it is plausible that CHD translocations, mutations or copy number alterations modulate folding and allosteric regulation or alter assembly and subunit composition of the CHD protein complexes, resulting in aberrant enzymatic activity and pathogenicity. Functional diversification of CHD enzymes can be further expanded through association with cancer-specific chimeric TFs generated by chromosomal translocations, as illustrated by PML-RARα that repurposes the NuRD complex to drive leukemogenic transcriptional programs (Morey et al. 2008). To what extent these genetic CHD perturbations are clear drivers or mere passengers in tumorigenesis is not clear, but as discussed below, for some (CHD1, CHD4, CHD5, and CHD7), a causal role in human malignancies has been clearly established.

One of the cardinal features of cancer cells is the ability to sustain proliferation through persistent activation of mitogenic signaling pathways (Fig. 6; Hanahan and Weinberg 2011). The CHD enzymes act downstream from key developmental signaling pathways that are frequently perturbed in tumorigenesis, such as WNT/β-catenin (CHD5 and CHD8) (Nishiyama et al. 2004; Fatemi et al. 2014), LIF/STAT3 (CHD8) (Yamashina et al. 2006), SHH (CHD7) (Engelen et al. 2011), NOTCH (CHD7) (Engelen et al. 2011), SMAD/BMP (CHD7) (Jiang et al. 2012; Liu et al. 2014b), and TGF-β/SMAD (CHD4 and CHD8) (Wang et al. 2009; Fueyo et al. 2018). Thus, chromatin remodeling activity of CHD enzymes appears to be critical for translating information from ligand-mediated signaling pathways to the transcriptional machinery—e.g., ERK/CHD7 (Badodi et al. 2017), WNT/β-catenin/CHD8 (SAWADA et al. 2013), PTEN/CHDl (Zhao et al. 2017), and K-RAS^G13D/CHD1 (Yu et al. 2015)—and it is likely that the CHD modulation of developmental pathways also contributes to tumorigenesis.

Figure 6.

The role of CHDs in tumorigenesis. (A) CHD enzymes are context-dependent tumor suppressors (TS; red) and oncogenes (green) that contribute to acquisition of different hallmarks of cancer. They can drive tumorigenesis by directly promoting transcription of oncogenes and repression of TS genes (CDKN2A and TP53; blue rectangles). Some enzymes can exert both oncogenic and TS function (CHD1, CHD4/NuRD, CHD7, and CHD8). For simplicity, not all players that are mentioned in the text are indicated in the cartoon. CHD functions in human malignancies are still largely unexplored, and it is likely that all members are dysregulated and mutated in diverse cancers and contribute to different aspects of tumorigenesis. Cartoon representation adapted from Hanahan and Weinberg (2011) with permission from Elsevier. (B, top) CHD1 nucleosome remodeling activity restricts binding of AR (green ovals) to prostetic lineage enhancers (yellow rectangles) in normal prostates. (Middle) During prostate tumorigenesis, in primary tumors, CHD1 loss leads to AR cistrome redistribution to a different subset of enhancers (blue rectangles), with HOXB13 (red hexagon) binding motifs (red squares) to drive expression of pro-oncogenic gene networks. (Bottom) More advanced, castration-resistant prostate cancers (CRPCs) are characterized by amplifications or mutations in the AR ligand-binding domain (LBD; AR-V7 as an example; blue ovals) (Jeselsohn et al. 2015; Watson et al. 2015). The LBD mutations allow promiscuous activation by noncanonical ligands (adrenal androgens, estrogen, progesterone, and glucocorticoids) to secure sustained expression of a subset of critical AR target genes (Watson et al. 2015). Additionally, antiandrogen-resistant prostate cancers can fully bypass the requirement for AR signaling by activating GR-mediated (violet ovals) transcriptional regulation of the AR cistrome (Arora et al. 2013). The role of other CHDs in the regulation of alternative transcriptional pathways that sustain expression of the AR-driven oncogenic gene networks in CRPC prostate cancers is unknown.

Importantly, CHD chromatin remodelers promote acquisition of different hallmarks of cancer by directly activating expression of pro-oncogenic transcriptional networks or inhibiting expression of tumor suppressor genes (Fig. 6). For instance, several examples of the pro-oncogenic function of CHD enzymes have been documented: regulating cell cycle and proliferation (CHD8 [Caldon et al. 2009; Rodríguez-Paredes et al. 2009; Subtil-Rodríguez et al. 2014; Dingar et al. 2015; Shen et al. 2015a], CHD4 [D'Alesio et al. 2016; 2019], CHD1 [Yu et al. 2015], and CHD9 [Yoo et al. 2020]), modulating angiogenesis (CHD4 [Yoo et al. 2006; Wang et al. 2020] and CHD7 [Boyd et al. 2019]), affecting the tumor microenvironment (CHD1 [Zhao et al. 2020]), impairing P53 transactivation (CHD4 [Luo et al. 2000], CHD7 [Nostrand et al. 2014], and CHD8 [Tu et al. 2021]), regulating expression of rRNA genes (CHD7 [Zentner et al. 2010], CHD4 [Zhao et al. 2018b], and CHD9 [Salomon-Kent et al. 2015]), and biosynthetic pathways (CHD1 [Bulut-Karslioglu et al. 2018] and CHD8 [Subtil-Rodríguez et al. 2014; Ceballos-Chávez et al. 2015]) (Fig. 6). Moreover, some, like the multifaceted CHD4/NuRD complex, promote tumorigenesis by epigenetic silencing of key tumor suppressor genes in colorectal cancer (Cai et al. 2014) (Xia et al. 2017).

As noted earlier, haploinsufficiency of chromatin remodelers drives tumorigenesis, and down-regulation of CHD expression is observed in several human cancers (Supplemental Table S1). CHD5 is best characterized as a tumor suppressor, and its inactivation drives progression of various human cancers (Supplemental Table S1 and references therein; Kolla et al. 2014). CHD5 is a potent regulator of cellular proliferation, senescence, and apoptosis owing to transcriptional regulation of the Cdkn2a locus, and its compromised expression in various human cancers blunts induction of key cellular gatekeeper pathways p16^INK4a/RB and p14^ARF/p53, setting the stage for cancer development (Mills 2017).

These studies, together with the roles of CHDs in the brain (CHD7) and hormone-dependent cancers (CHD1, CHD4/NuRD, and CHD8) discussed below, highlight the role of individual CHD enzymes as context-dependent oncogenes or tumor suppressors in human malignancies.

Brain cancers

CHD enzymes are vital for the proper development and functionality of the nervous system, and their reduced activity can give rise to neuronal tumors of variable origin. For instance, in neuroblastoma, loss of CHD5 (Kolla et al. 2014) and CHD9 (Lasorsa et al. 2016) correlates with poor prognosis. Notably, it appears that CHD7 inactivation drives human gliomas (Boyd et al. 2019), whereas both overexpression (Machado et al. 2019) and inactivation of CHD7 promotes medulloblastoma (MB) progression (Badodi et al. 2017, 2021). Indeed, CHD7 inactivation drives a MB subtype with increased activation of the ERK and mTOR signaling pathways (Badodi et al. 2017, 2021). Importantly, this metabolic adaptation creates a clinically relevant cancer vulnerability, as combined treatment with inositol and cisplatin greatly improves survival of MB xenografts (Badodi et al. 2021).

Hormone-driven cancers

There is ample evidence that CHD enzymes govern transcription in response to ligand-mediated activation of nuclear receptors in human cancers, either through direct association with nuclear receptors (NRs; CHD6, CHD8, and CHD9) (Surapureddi et al. 2002; Shur and Benayahu 2005; Marom et al. 2006; Surapureddi et al. 2006; Shur et al. 2006a; Menon et al. 2010), or indirectly through modulating NR-mediated transcriptional activity (CHD1, CHD4, and CHD7) (Augello et al. 2019; Zhang et al. 2020; Zhao et al. 2020; Paakinaho et al. 2021). These regulatory interactions are particularly important for prostate and breast cancers (CHD1 [Berger et al. 2011; Grasso et al. 2012; Huang et al. 2012; Liu et al. 2012; Burkhardt et al. 2013] and CHD8 (Caldon et al. 2009; Menon et al. 2010; Ceballos-Chávez et al. 2015; Metzger et al. 2016]) discussed below.

Prostate cancer

The progression of prostate cancer is mainly driven by PTEN loss, AR amplification or mutation, and ETS gene fusion (TMRPSS2-ERG) (Jamaspishvili et al. 2018). Recent genomic studies of prostate cancers also identified recurrent mutations (CHD1/3), amplifications (CHD7), or promoter methylation (CHD5/8) of CHD family members (Supplemental Table S1). Intriguingly, a CHD8 functional dichotomy is observed, whereby decreased expression through promoter methylation drives early stages of prostate cancer tumorigenesis, whereas higher levels are associated with advanced clinical features (Damaschke et al. 2014).

Next to PTEN loss, CHD1 inactivation is the most frequent event in localized prostate cancer and advanced castration-resistant prostate cancer (CRPC) samples (Supplemental Table S1 and references therein). Importantly, a large body of work in cellular and animal models established CHD1 as a bone fide tumor suppressor in prostate cancer (Berger et al. 2011; Grasso et al. 2012; Huang et al. 2012; Liu et al. 2012; Baca et al. 2013; Burkhardt et al. 2013; Rodrigues et al. 2015; Shenoy et al. 2017; Zhao et al. 2017, 2020; Augello et al. 2019; Li et al. 2020; Zhang et al. 2020). The CHD1-null tumors define a specific subset of prostate tumors with high genome instability, characterized by intrachromosomal rearrangements, gene deletions, and high frequency of recurrent small copy number amplifications (Baca et al. 2013).

However, CHD1 can also serve as an oncogenic driver of distinct PTEN⁻ and TMPRSS2-ERG⁺ subtypes of prostate cancers. Indeed, high CHD1 levels positively correlate with the Gleason grade in patients (Zhao et al. 2017). Strikingly, PTEN and CHD1 deletions are mutually exclusive, and CHD1 appears to be a synthetic-essential gene required for the growth of PTEN-deficient tumors by promoting transcriptional activation of the prosurvival TNF–NFκB pathway (Zhao et al. 2017). Additionally, CHD1 regulation of IL-6 expression fosters an immunosuppressive tumor microenvironment and promotes progression of PTEN-deficient prostate cancers (Zhao et al. 2020). This newly defined PTEN–CHD1 paradigm appears to be equally important for the growth of breast and colorectal cancers as well, and CHD1 targeting in PTEN-negative breast and prostate cancers could be a viable strategy to impair tumor growth (Zhao et al. 2017).

Importantly, it appears that CHD1 nucleosome remodeling activity commands AR chromatin distribution and prostate tumorigeneses, as it favors either occupancy of canonical prostatic lineage-specific enhancers or pro-oncogenic genes enriched for HOXB13 binding motif (Fig. 6; Burkhardt et al. 2013; Pomerantz et al. 2015; Augello et al. 2019). Specifically, upon CHD1 loss during prostate epithelial transformation, altered chromatin accessibility leads to extensive redistribution of the AR cistrome to HOBX13 target regions (Pomerantz et al. 2015), resulting in transcriptional activation of pro-oncogenic pathways (Augello et al. 2019).

Together, the data suggest that CHD1, similar to CHD8, exerts dichotomous functions, having context-dependent tumor-suppressive or oncogenic roles in prostate cancer progression. Moreover, the CHD1 pleiotropic role modulates AR cistrome transactivation and the tumor microenvironment, thereby promoting prostate cancer progression. As such, the CHD1 status in patients may serve as an indicator of chromosomal instability and a predictor of metastatic potential.

Breast cancer

Owing to its multisubunit composition and context-dependent functionality, the NuRD complex can have opposing roles in breast cancer progression (Lai and Wade 2011), both inhibiting (Fujita et al. 2003) and promoting (Fu et al. 2011) EMT and enhancing insensitivity to endocrine therapy (Fu et al. 2011). Additionally, evidence suggests that both CHD1 and CHD8 might be involved in ER⁺ breast cancer progression (Tan et al. 2014; Ceballos-Chávez et al. 2015). Notably, similar to CHD1 modulation of the AR cistrome in prostate cancer, NuRD chromatin occupancy regulates the binding potential and genomic distribution of ER (Serandour et al. 2018) and is critical for the growth of ER⁺ breast cancers (Farcas et al. 2021). These studies are indicative of complex roles of CHD family members in breast cancer progression, warranting further studies of the underlying mechanisms.

DNA damage repair signaling

Although not a focus of this review, it is noteworthy highlighting critical roles of CHD enzymes in modulating double-stranded break (DSB) DNA damage repair (DDR) pathway signaling (for further reading, see Stanley et al. 2013; Rother and Attikum 2017). Current evidence indicates that CHD2 (Luijsterburg et al. 2016), CHD3/4 (Rother and Attikum 2017), and CHD7 (Rother et al. 2020) recruitment to DSBs orchestrates nonhomologous end joining (NHEJ), whereas CHD1 (Kari et al. 2016) and CHD3/4 (Pan et al. 2012) are important for homology-directed repair (HDR). Moreover, displacement of CHD3 appears vital for the chromatin relaxation and repair of heterochromatic DSBs by either NHEJ or HDR (Goodarzi et al. 2011; Klement et al. 2014). In all of these instances, upon DNA damage, CHD remodeling activity is required either to promote transcription of inhibitors of oxidative damage (CHD6) (Moore et al. 2019), or to shape local chromatin structure and facilitate recruitment of the DNA repair machinery to sites of damage (CHD1 [Kari et al. 2016; Shenoy et al. 2017; Zhou et al. 2018], CHD2 [Luijsterburg et al. 2016], CHD3 [Goodarzi et al. 2011; Klement et al. 2014], CHD4 [Rother and Attikum 2017], and CHD7 [Colbert et al. 2014; Rother et al. 2020]).

Clinical relevance as potential biomarkers and actionable therapeutic targets

Current cancer treatment modalities rely on the use of standard cytotoxic chemotherapeutic agents and relatively recent approaches, such as targeted drugs, endocrine therapies, and immunotherapy. Therapeutic effects of anticancer drugs are largely determined by the landscape of genetic and epigenetic events acquired during tumor evolution. Emerging evidence indicates that CHDs may hold promise as actionable targets and serve as biomarkers and predictors of metastatic potential, patient survival, and therapeutic response, as well as chemoresistance (Wagner et al. 2018), in several cancers. As such, they could help inform patient stratification and response to therapies.

As CHDs are dysregulated across various cancers, they may serve as novel stratification biomarkers (CHD1: prostate cancer), and predictors of patient survival (CHD4: colorectal cancer and hepatocellular carcinoma [Xia et al. 2017] and nonsmall lung cancer [Xu et al. 2016], CHD8: gastric cancer [SAWADA et al. 2013], CHD5: various cancer types [Mills 2017], and CHD7: pancreatic cancer [Colbert et al. 2014] and glioblastoma [Boyd et al. 2019]) and therapeutic outcomes (CHD2: colorectal cancer [Bandrés et al. 2007]) (Supplemental Table S1).

It is increasingly apparent that some cancers rely on the tumor-promoting function of CHD enzymes, phenomena known as “oncogene addition,” as their depletion elicits a strong antitumor effect and could be exploited as vulnerabilities in at least some cancer types (CHD1: PTEN-negative prostate cancer [Zhao et al. 2017], CHD4: AML [Sperlazza et al. 2015; Heshmati et al. 2016, 2018] and glioblastoma [Chudnovsky et al. 2014; McKenzie et al. 2019], and CHD7: pancreatic cancer [Colbert et al. 2014] and glioblastoma [Machado et al. 2019]). Additionally, studies in animal models provide evidence that CHD7 (Zhen et al. 2017) and CHD8 (Shingleton and Hemann 2015; Shen et al. 2015a) are required for the growth of hematopoietic cancers and therefore might represent vulnerabilities in human hematopoietic malignancies. Alternatively, given the promoter silencing by hypermethylation in CHD nullizygous tumors (Supplemental Table S1), reactivation of the wild-type locus using demethylating agents might be a viable therapeutic strategy (CHD5) (Fatemi et al. 2014).

Importantly, from a clinical perspective, loss of CHD function renders cells sensitive to treatment with irradiation (CHD1 [Kari et al. 2016; Shenoy et al. 2017] and CHD2 [Rajagopalan et al. 2012]), PARP inhibitors (CHD1 [Kari et al. 2016; Shenoy et al. 2017] and CHD4 [Pan et al. 2012; Nio et al. 2015]), chemotherapeutic drugs (CHD1 [Kari et al. 2016; Shenoy et al. 2017] and CHD7 [Colbert et al. 2014]), genotoxic agents (CHD4 [Sperlazza et al. 2015]), monoclonal antibodies (CHD4 [D'Alesio et al. 2019]), and combined inositol hexaphosphate (IP6) and cisplatin treatment (CHD7 [Badodi et al. 2021]). Conversely, loss of CHDs can also confer resistance to therapeutic treatments, as evident from enzalutamide resistance in prostate cancer (CHD1) (Zhang et al. 2020) and the relapse of SCLC (CHD8) (Supplemental Table S1; Wagner et al. 2018).

Although showing great promise in some malignancies (Cheng et al. 2019), the field of epigenetic targeted therapy is still in its infancy. Given the CHDs’ domain conservation and structure, functional dichotomy, protein isoforms, and multimeric complexes they reside in, as well as the severity of phenotypes in homozygous animal models, there are considerable challenges regarding drug design and selectivity.

An interesting approach could be the use of synthetic histone mimic compounds, such as influenza virus nonstructural protein 1 (NS1), which, when dimethylated and trimethylated, is able to target CHD1 chromodomains faithfully mimicking the CHD1–H3K4me2/3 interaction (Marazzi et al. 2012). This might enable selective inhibition of CHD activity in various pathological conditions.

Additional concern when considering CHDs as actionable targets is the combinatorial assembly of protein subunits that governs chromatin association and functional specificity of CHD complexes. Interrogation of functional and compositional differences between physiological and tumor-specific CHD complexes could provide an opportunity for development of new therapeutic strategies. The emergence of novel biochemical and structural approaches, such as cryo-EM, will facilitate determination of the structure of tumor-specific CHD protein complexes, and may allow development of selective small molecule inhibitors aimed at disrupting key interfaces that sustain tumor-specific CHD transcriptional programs.

Concluding remarks and future prospects

Chromodomain helicase DNA-binding proteins are an evolutionarily conserved family of ATP-dependent chromatin remodeling enzymes that play pivotal roles in regulation of the DNA-templated processes. The sheer number of CHD remodelers, their functional nonredundancy, and the unique domain composition of the different classes point to more specific and programmatic functions. Curiously, many CHD genes are expressed in the nervous system, and it is conceivable that the evolutionary development of the nervous system might have prompted CHD diversification and specialization of their functions, or has been the result of the rewiring of the regulatory circuitry that adapted to the increased complexity of the non-protein-coding complement of the genome (Mattick 2007).

Given their functional complexity, it is still unclear whether the primary function of CHD proteins is to locally remodel chromatin to regulate gene expression and DNA-repair or to maintain higher-order chromatin structure, and how the enzymes have been repurposed for each function within a given cellular context. Recent efforts in mapping CHD chromatin occupancy provide insight into the mechanism of transcriptional regulation. However, the role of CHDs in higher-order chromatin organization is still poorly understood, and it is tempting to speculate that distinct CHD enzymes might, each in their own right, safeguard particular or overlapping spatially organized “chromatin territories.” Given the dynamic nature of CHD chromatin association, a systematic approach using ChIP-seq, RNA-seq, and chromosome capture conformation assays in a defined model system (differentiation: ES vs. NPC; pathology: normal vs. tumor/NDD samples) might permit integrative modeling of their function in chromatin organization and start addressing this important question.

Recently, the role of liquid–liquid phase separation (LLPS) in chromatin compartmentalization and organization has received much attention (Hnisz et al. 2017; Gibson et al. 2019). Importantly, this process is at least in part driven by nucleosome spacing, linker length, and histone H1 binding (Hnisz et al. 2017; Gibson et al. 2019), and CHD enzymatic activity is instrumental in modulating these chromatin features. Additionally, as CHDs are organized in subnuclear compartments (CHD7 and CHD9), as well as occupy enhancers (CHD1, CHD4, and CHD7) and interact with nascent RNAs (CHD4 and CHD7), both of which are shown to affect chromatin compaction, exploring the involvement of CHDs in chromatin organization by LLPS is an attractive strategy.

The CHD proteins instruct cells to maintain their identity, divide, differentiate, or die, thereby orchestrating organismal development or malignant transformation. Disruption of CHD remodelers in mice is often embryonic-lethal. Phenotypic discrepancies of individual Chd knockouts between different studies highlight the importance of careful selection of the targeting strategy (homologous recombination or CRISPR) when manipulating proteins of such domain complexity, as removal of one critical exon or functional domain can elicit phenotypically different outcomes compared with the complete loss-of-function protein.

Genetic alterations of several CHD family members are identified in human cancers and neurodevelopmental syndromes, and mouse models carrying mutated Chd7/8 are valuable to further our understanding of underlying NDD molecular mechanisms and identification of viable therapeutic strategies. So far, only patient-specific mutations of CHD3, CHD4, and CHD7 are shown to alter nucleosome remodeling activity (Bouazoune and Kingston 2012; Kovač et al. 2018; Blok et al. 2019; Yan et al. 2020). Systematic analysis of patient-specific CHD mutants in defined cell-based or animal model systems might provide insight into their mechanism of action and relevance for pathogenesis. In addition, biochemical purification and structural characterization of CHD protein complexes specific for particular developmental (Nitarska et al. 2016; Hoffmeister et al. 2017) or pathological (i.e., normal vs. tumor/NDD sample) conditions would inform about the specific functional contexts in which CHDs operate. This would potentially allow the designing of drugs specifically aimed at disrupting key interfaces that sustain tumor-specific CHD transcriptional programs.

Three decades of the intense biochemical and genetic analysis of the CHD family members in different model organisms has provided insight into the mechanistic principles that govern the function of CHD remodelers in vitro and in vivo. However, it is becoming evident that we have barely exposed the tip of the iceberg, and that their complex roles in regulating gene expression, chromatin organization, development, and disease are yet to be fully appreciated.

Competing interest statement

The authors declare no competing interests.