Introduction

Long non-coding RNAs (lncRNAs) are a heterogeneous class of long RNAs without a large open reading frame encoding proteins. To better understand the role of lncRNAs in the formation and recruitment of chromatin remodeler complexes, we will briefly discuss the organization of DNA into chromatin, and then review the chromatin remodeler subfamilies with the involved subunit proteins.

Chromatin is the state of organized DNA condensation in the nucleus. DNA is wrapped into nucleosome structures that are further packed into chromatin fibers and condensed into either “euchromatin” or highly condensed “heterochromatin”. Chromatin condensation is a dynamic process resulting in different nuclear sub‐compartments, including topologically lamina-associating domains, nucleoli, Cajal bodies, nuclear stress bodies, paraspeckles and non-chromatin bodies like nuclear speckles and PML bodies (reviewed by [1]). In the fundamental nucleosome unit, ~ 146 base-pairs (bp) of the negatively charged DNA helix are folded ~ 1.7 times around an octamer of positively charged histone proteins. A tetramer consisting of two H3 and two H4 histones binds to the DNA, where after two H2A-H2B dimers join the complex. The first ~ 20 bp of DNA that sort the fold are hold together by the histone H1 protein and a “beads on a string” structure is formed with strings of ~ 15–70 bp free DNA (Fig. 1a, b, [2]).

Fig. 1
figure 1

Organization of chromatin. Depicted is a DNA helix (in grey) with histone proteins H2A, H2B, H3, H4 and H1 (different colors). a Schematized top view of a nucleosome with tetramer (H3-H4)2 and two dimers (H2A-H2B). b Nucleosomes side view with histone H1 stabilizing the DNA linker. Methylated histone tails have stronger DNA interactions and then acetylated tails. c Presented are common histones modifications, such as methylation (squares), acetylation (triangles), phosphorylation (circles) and ubiquitination (trapeziums). Histones modifications are frequently recognized by chromatin readers. Modifications that are implicated in chromatin remodeling complexes are depicted by solid symbols and their corresponding readers are annotated (red circles). d LncRNAs (in blue) can either inhibit or stimulate chromatin remodeling via interference with remodeler complexes

Full size image

Nucleosome positioning is part of epigenetic regulation since the condensed DNA is inaccessible for transcription, replication, and repair [3]. Positioning of nucleosomes is not arbitrary; firstly, it depends on the thermodynamic bending properties of different DNA sequences. In particular, the properties of poly(deoxyAdenylic(dA): deoxyThymidylic(dT)) DNA stretches disfavoring nucleosome formation. DNA sequence motifs can also be recognized by sequence-specific factors that may lead to nucleosome remodeling or depletion. Secondly, DNA modifications, such as 5-methylation and 5-hydroxymethylation of cytosines, affect DNA flexibility and nucleosome stability. Thirdly, the DNA-interactions with histone tails may vary, as these are subjected to several posttranslational modifications, which may change electric charges and/or evoke steric hindrance (Fig. 1c). In addition, acetylation of lysine residue 16 of histone 4 (H4K16) weakens its protein–protein interaction with residues from the H2A-H2B dimer of the adjacent nucleosome and significantly reduces nucleosome stacking [4]. Moreover, modifications like methylation and acetylation of lysine residues are important for the recognition by chromatin reader proteins [5]. These chromatin readers are known to recruit complexes with writers and erasers of chromatin modifications, as well as chromatin remodelers (Fig. 1c).

Rather than modifying DNA or histones to modulate chromatin (like writers/erasers), the remodeler complexes change nucleosome positioning (Fig. 1d). They affect spacing of nucleosomes, regulate nucleosomes transfer and evoke histone variant switching. The kinetic energy needed to modulate the stable nucleosome organization comes from ATP hydrolysis. All chromatin remodeling complexes contain a catalytic unit from the superfamily 2 helicases, which categorizes them into four types of chromatin remodeling complexes: (1) the SWI/SNF, (2) the INO80/SWR1, (3) the ISWI and (4) the CHD family. The complexes of each chromatin remodeler family have distinct subunits and different functionalities that are listed in Table 1. Today, 34 lncRNAs have been described to interact with diverse subunits. Another 18 lncRNAs were reported to act as competing endogenous (ce) RNA with miRNA targets. We will summarize the current knowledge of the effect that lncRNAs have on the function of each chromatin remodeler complex family.

Table 1 Proteins of human chromatin remodeling complexes
Full size table

Chromatin remodeling complexes of the SWI/SNF family

Subunits of the SWI/SNF complexes

The SWI/SNF proteins were discovered in yeast as important factors regulating mating-type switching (SWI = SWItch) and regulating the use of different energy sources (SNF = Sucrose Non-Fermentable) [6]. The catalytic subunit in these complexes is either the ATP-dependent helicase Brahma (SMARCA2, BRM) or Brahma-related gene-1 (SMARCA4, BRG1). The core subunits are SMARCB1 (SNF5), SMARCC1 (BAF155) and SMARCC2 (BAF170) (Table 1). SWI/SNF remodeler complexes anchors to histone proteins and translocate 1–2 bp of DNA along the surface of the nucleosome, depending on the ATPase activity, this results in nucleosome sliding, or in destabilizing and removal of H2A–H2B dimers or entire histone cores [7], thus creating open chromatin structures.

Depending on subunit composition the SWI/SNF complexes are classified as PBAF (PolyBromo-Associated Factors) and BAF (BRG1- or BRM-Associated Factors). Exclusive subunits of the PBAF complex are ARID2, PBRM1 (BAF180), PHF10, and BRD7, in combination with the helicase SMARCA4. In the embryonic stem cell, the PBAF complex is important for maintaining the stem cell transcriptome [8]. Indeed, SMARCA4 plays an important role in the NANOG signaling pathway [9]. The PBAF complex inhibits growth through down-regulation of cell cycle genes CDK2, CDK4 and CCND1 and interference with the phosphorylated RB-pathway [10]. In addition, the PBAF complex is implicated in transcriptional repression at DNA double-strand breaks [11].

The BAF complex harbors the exclusive subunits ARID1A or ARID1B, BRD9 and SS18. They contain either the helicase SMARCA4 or SMARCA2. The difference between the PBAF and BAF complexes also results in different functionality [10, 12]. The SMARCA4- and SMARCA2-containing complexes have been assigned to ‘actively marked’ and ‘repressively marked’ chromatin binding complexes, respectively [13]. In line with this, both complexes have differential affinity for histone modifications (PBAF complexes binds to H3K4-monomethylated, BAF binds to H3K4-trimethylated; [14]).

Whereas the PBAF complex is associated with an embryonic stem cell phenotype, BAF complexes seem to be tissue- and cell-type specific [15]. In particular, SMARCD3 subunits defines cardiac progenitors (cBAF complex) and SMARCD1 neuronal progenitors (npBAF complex), and the association with SS18L1 is associated with post-mitotic neurons (nBAF complex) [15].

Long non-coding RNAs and the SWI/SNF family

Amongst the chromatin remodelers, the SWI/SNF complex has frequently been shown to interact with lncRNAs (Table 2, Fig. 2a). In particular, SMARCA4 physically associates with primary transcribed RNA [16], suggesting a cis-acting function as shown for the lncRNAs lncFZD6 and lncTCF7. Some lncRNAs (e.g., MANTIS, SChLAP1) do not bind the SWI/SNF complex, but compete for DNA-binding at a particular chromosomal promoter gene region (e.g., MANTIS at the ICAM-1 gene). SMARCA4-lncRNA interaction may also interfere with recruitment of SWI/SNF complexes at multiple gene regions (trans-acting) (e.g. LincRNA Cox2 and IL-7–AS). Recently, also SMARCB1 (SNF5) has been shown to interact with a pool of lncRNAs [17], including SWINGN and HOTAIR. Interestingly, lncBRM seem to inhibit SWI/SNF-SMARCA2 BAF complexes by its association with SMARCA2 (BRM), which in turn favors the assembly of SWI/SNF-SMARCA4 BAF complexes in liver cancer stem cells [18]. The BAF-specific subunit ARID1A has also been shown to bind lncRNAs (HOTAIR, MVIH, DGCR5 and LINC00163) thereby interfering with the transcription of several genes (Table 2 and Fig. 2a). Binding of lncRNA uc.291 with ACTL6A competes with BAF-ACTL6A binding and affects the expression of several differentiation genes in skin keratinocytes [19]. In addition, the MAPK pathway is affected by the interaction of SWI/SNF subunit BCL11A with CDKN2B-AS1 in lymphocytes and lncRNA uc.57 in breast cancer cells [20, 21]. Lots of lncRNAs have been shown to regulate gene expression by competing for miRNA binding [22], and for the SWI/SNF family lncRNA CASC15 and DLEU1 were reported to act as ceRNA for ARID1A and SMARCD1 transcripts, respectively [23, 24]. In addition, lncRNA DSCAM-AS1 may upregulate BCL11A expression [25].

Table 2 LncRNAs known to interact with human chromatin remodeling complexes
Full size table
Fig. 2
figure 2

Effects of long non-coding RNAs on chromatin remodeling complexes. LncRNAs (in blue) affect chromatin remodeling by changing subunit expression (transcription and/or competing for miRNA binding (ceRNAs)), by inhibiting complex binding to or by stimulating the recruitment to a genomic region. a Interaction of SWI/SNF complexes with lncRNA results in nucleosome sliding either further apart or closer together. b Histone replacement (H2A.Z) and modifications in nucleosomes are affected by lncRNAs that regulate the INO80 complexes. c LncRNAs known to regulate the different ISWI complexes are mainly recruiting these complexes. d Generally, lncRNAs that function as ceRNAs are known for the CHD family

Full size image

Chromatin remodeling complexes of the INO80/SWR1 family

Subunits of the INO80/SWR1 complexes

The INO80/SWR1 factors were first identified in budding yeast in a screen for regulators of phospholipid biosynthesis (INOsitol requiring) and for switching from glycolytic to oxidative metabolism (SWR1) [26]. This family includes three major complexes, each having a unique ATPase unit (INO80, SRCAP or p400) in combination with the two ATP-driven helicases RUVBL1 and RUVBL2. The “effector” sub-units may recruit other enzymes, such as transcription factors and histone acetyltransferases. In particular, the p400-TRRAP-TIP60 complex mediates acetylation of histone H4 and H2A-tails. This acetylation stimulates the INO80 and SRCAP complexes to replace histone H2A with the 60% homologous histone H2A.Z. The H2A.Z diverges with H2A at two loops interacting with the nucleosomal DNA and at the tail section interacting with histone H3, which may result in a reduced stability or in nucleosome sliding [27]. Nucleosomes with histone H2A.Z are enriched at flanking regions of transcription start sites. In addition, H2A.Z can also be modified by acetylation, methylation, phosphorylation, SUMOylation and ubiquitination [28]. Removal of H2A.Z from DNA by the histone chaperone ANP32E and INO80 is a primary step in DNA repair [29]. The INO80 complex may recruit additional subunits, e.g., NFKRP, implicated in double strand break DNA repair [30].

Long non-coding RNAs and the INO80/SWR1 family

Both the lncRNA LCTS5 and HAND2‐AS1 interact with the INO80 subunit, however, they have opposite effects (Fig. 2b). LCTS5 inhibits and HAND2‐AS1 stimulates INO80 complex recruitment [31,32,33]. Similarly, ANRIL and Linc-YY1 both interact with transcription factor YY1 resulting, respectively, in recruitment to promoter loci of IL6/IL8 and eviction of polycomb repressive complex (PRC2) regulated promoters [34, 35]. Interaction of lncAKHE with YEATS4 from the histone acetylation p400-TRRAP complex results in activation of NOTCH2 signaling in hepatocellular carcinomas [36].

Several lncRNAs regulate the expression of INO80 subunits. UCHL5 mediates de-ubiquitination of NFRKB (INO80G), which is prevented by its interaction with lncRNA DRAIC [37]. The subsequent NFRKB ubiquitination-mediated degradation affects resection of DNA in double strand breaks DNA-repair [30]. In addition, lncRNA PTCSC3 inhibits INO80 expression by negatively regulating STAT3 [38]. The lncRNAs CR933609, RNCR3, and CASC7 were reported to act as ceRNAs targeting INO80D, BRD8 and ING3, respectively [39,40,41]. SPAG5-AS1, HOTAIR, LINC00899 and LINC00668 were reported to act as ceRNAs targeting YY1 [42,43,44,45].

Chromatin remodeling complexes of the ISWI family

Subunits of the ISWI complexes

Initially discovered in a search for SWI/SNF genes in Drosophila [46], ISWI complexes regulate nucleosome spacing and are implicated in DNA repair. These chromatin remodelers are rather small complexes with 2–4 sub-units. The best-known ATP-helicases in these complexes are SMARCA5 and SMARCA1. Both SMARCAD1 and SMARCA6 (HELLS) are SWI/SNF genes that can also be classified in this family. SMARCAD1-TRIM28 binds to H2A-ubiquitin and stimulates acetylation [47]. HELLS-CDC7A re-modulates the nucleosome to facilitate access to DNA for DNMT3B [48, 49]. SMARCA3 and SMARCAL1 are also ATP-driven DNA-binding helicases, but they are implicated, respectively, in DNA unwinding, and in the stabilization of the replication fork and they do not interact with nucleosomes [50, 51].

The WICH (SMARCA5-BAZ1B) and CERF (SMARCA1-CECR2) complexes affect the phosphorylation of histone H2A.X (γH2A.X) [52, 53]. This histone variant is highly homologous to H2A, but has a 13 amino acid extended C-tail that includes three phosphorylation sites (T136, S139 and Y142) [54]. Phosphorylation and ubiquitination of H2A.X are key events in the detection and response to double strand breaks of DNA damage.

The ACF (SMARCA5-BAZ1A) complex can either function as a heterodimer complex playing a role in double strand break repair [55], or as the CHRAC complex (ACF interacting with CHRAC1 and POLE3). Both complexes have very similar nucleosome sliding activity [56]. The NURF (SMARCA1-BPTF-RBBP4-RBBP7) complex also mediates nucleosome sliding and enhances recruitment of transcription factors and chromatin insulator protein elements [57].

Long non-coding RNAs and the ISWI family

The LncKdm2b (KDM2B-DT) seems to have a general function in recruiting chromatin remodeler complexes, as it was reported to direct both SRCAP (INO80 family) and NURF (ISWI family) complexes to two different zinc finger protein genes (Zbtb3 and Zfp292) [58, 59] (Fig. 2b, c). The NURF complex is also recruited through binding of lncRNA NMR to BPTF [60] and of DLEU1 to SMARCA1 [61]. Interestingly, DLEU1 has a double function as it seems to inhibit other complexes via miRNA-mediated decay of SMARCD1 (subunit of SWI/SNF complex). The NEXN-AS1-mediated recruitment of BAZ1A (ACF complex) to NEXN was shown to upregulate its expression [62], which may suggest a cis-acting function similar to the recruitment of the SWI/SNF complex to primary transcribed RNA. The “lnc pRNA” has a specific action on ribosomal RNA genes as its interaction with PARP1 and BAZ2A (NoRC complex) represses these genes in particular [63].

Recently, it was shown that Pauper interacts with KAP1 to promote H3K9me3 deposition at a subset of distal targets, through formation of a Paupar-KAP1-PAX6 complex [64]. If the ATP-dependent helicase SMARCAD1, also interacting with KAP1, is part of this complex was not assessed in this study. The CHIRRC complex was reported to be affected by lncRNA BlackMamba via interaction with HELLS and by FGD5-AS1 as ceRNA inhibiting CDCA7 degradation [65, 66].

Chromatin remodeling complexes of the CHD family

Subunits of the CHD complexes

The chromodomain helicase DNA-binding (CHD) proteins that were initially identified as mammalian DNA-binding factors with a SWI-like helicase domain [67], are implicated in DNA repair and in recruitment of histone deacetylase/demethylase enzymes. CHD1 and CHD2 belong to the same subclass I. CHD1 has been shown to promote stabilization of H2A.X and the efficient repair of double strand breaks through homologous recombination [68]. CHD1 has no specific subunit, but has been described to interact with several factors, such as SSRP1 and the NCoR complex [69, 70]. CHD1 is a critical regulator of transcription initiation and elongation stimulating androgen receptor (AR)-mediated regulation and pluripotency gene expression (Oct 4 and Nanog) [67, 71].

The CHD2 protein has been shown to interact with PARP1 complex stimulating histone variant H3.3 deposition in non-homologous end joining (NHEJ) DNA-repair regions [72]. The histone H3.3 has only four amino acid differences compared to H3, introducing a posttranslational phosphorylation site (S31). Phosphorylated H3.3S31p stimulates p300-mediated histone H3K27 acetylation of neighboring nucleosomes [73]. In addition, the combination of histone H3.3 with H2A.Z in nucleosomes enriched near promoters and enhancers also evokes a loose nucleosomal packaging [74].

In the NuRD complex, one of the helicase from sub-family II, CHD3, CHD4 or CHD5 is present where they interact with GATAD2A/B. Interaction of the GATAD2 proteins with MDB2/3 results in recruitment of the NuRD histone deacetylase sub-complex (HDAC1/2, RBBP4/7, CTBP2, MTA1/2/3) [75]. The NuRD complex may also interact with several additional proteins, such as CDK2AP1, SALL1/4 and ZMYND8 that may direct the NuRD complex to specific regions [75]. CHD4 is also binding to RNF8, which is implicated in histone H1-ubiquitinylation and loading of the BRCA1 complex [76].

Subfamily III of CHD remodelers includes CHD6–CHD9. They interact with diverse transcription factors and posttranslationally modified histone H3. For CHD6, CHD7 and CHD8 it has been shown that they bind to the DNA strings in between nucleosomes. CHD6 and CHD7 both bind to short linker DNA, whereas CHD8 requires longer DNA sequences for binding and thus slides nucleosomes further apart [77]. CHD6 rather disrupts nucleosomes and was recently shown to relocalize to sites of DNA damage, suggesting a role in DNA repair [77, 78]. CHD7 and CHD9 have been shown to interact with chromatin remodelers from the SWI/SNF family (PBAF/BAF) [79]. CHD8 can directly bind β-catenin mediating interaction with histone H1 or other factors like the Gfi1b-complex regulating Wnt/β-catenin-dependent gene expression [80, 81]. CHD8 was further shown to interact with diverse methyltransferase complexes, such as BRD4-NSD3 [82], BACH1-MAFG-DNMT3 [83], SET1 (KMT2) [84] and WDR5-ASH2L-RbBP5 [85]. In addition, CHD7 was shown to interact with CHD8/FAM124B [86]. Both CHD7 and CHD8 are implicated in neuronal and developmental disorders (e.g., CHARGE, autism [77]).

Long non-coding RNAs and the CHD family

So far, only a few lncRNAs have been described in the context of CHD complexes (Fig. 2d). Recently, a negative regulation loop beween lncRNA CHASERR and CHD2 has been described. The CHASERR gene is located on the same strand (1.7 kb apart) as the CHD2 gene at chr15q26. The CHASERR transcript can be bound by CHD2 protein and regulate transcription of the CHD2 gene [87]. Some lncRNAs have been shown to affect transcriptional regulation of CHD subunits. The lncRNA MTA2TR transcriptionally upregulates MTA2 expression [88]. LUCAT1 upregulates MTA1 in cervical cancer and HDAC1 in papillary thyroid cancer [89, 90]. Six lncRNAs have been described as ceRNAs for CHD subunits (MATN1-AS1 ceRNA CHD1 [91], LINC01410 ceRNA CHD7 [92], ANRIL ceRNA HDAC1 [93], SNHG15 and ARAP1-AS1 ceRNA HDAC2 [94, 95] and HOTAIR ceRNA MTA2 [96]).

Concluding remarks

The lncRNAs described in this review often recruit chromatin remodeler complexes at promoter and enhancer regions by binding a particular subunit. Some other lncRNAs, not discussed herein, are related to the chromatin remodeler complex through their position, but they seem to have unrelated functions (lnc-Arid2-IR, lnc-SMARCC2, lnc-MDB2, lnc-Dpf3) [97,98,99,100].

A recurrent chromatin regulatory mechanism is recruitment of remodeler complexes by interaction with primary transcribed RNAs, either by antisense lncRNA encoded at genomic proximity, or by trans-acting lncRNAs (e.g., LncKdm2b-mediated recruitment of SRCAP and NURF complexes [58, 59]). The interaction of chromatin remodeler subunits with some lncRNAs results in a distinct cellular function. For example, interaction of lncBRM with BRM inhibits the SWI/SNF complex associated with a differentiated phenotype and favors assembly of a stem cell-related SWI/SNF complex [18]. Also lnc-pRNA regulates primarily ribosomal RNA gene expression via the NoRC complex [63]. The interaction between chromatin remodeler subunits, in particular, the flexible and/or successive nature of complex formation in the different processes like NHEJ or HR-mediated DNA repair will be an important asset to further define the role of lncRNAs in these processes. The role of the chromatin remodelers in the different stages of DNA repair is closely related to cancer (e.g., [67, 101,102,103]). Indeed, from the lncRNAs that affect the chromatin remodeling complexes (Table 2) at least 35 are associated with cancer. Cellular capture-based approaches can be employed to further analyze the interactions of lncRNAs with the chromatin remodeler complexes, using either RNA-specific, DNA-locus specific or genome wide approaches (reviewed by [104]), although non-specific background interaction signals and sequential complex formation remains a major issue. The CRISPR-Cas9 engineering revolution now enables in situ capture of RNA-chromatin interactions by biotinylated dCas9 [104], which may also reveal implication of lncRNAs in the cancer-related nucleosome remodeling evidenced by Druliner et al. [105]. In addition, analyzing CRISPR/Cas9-based gene perturbations (e.g., lncRNAs invalidation/activation) combined with single cell sequence analysis (recently reported by [106,107,108]), may unravel the specific actions of lncRNAs, that are measured as low expressed in bulk-analyzing experiments but can be abundantly expressed in individual, and phenotype-specific cells (e.g., shown in neuronal cells and cardiomyocytes [109, 110]). Although today still presenting a huge technical challenge, these analyses could be combined with studies of chromatin organization and protein interactions in single cells [111]. Identifying the lncRNAs that contribute to epigenetic regulation by controlling the specific chromatin modifications associated with disease, may result in interesting novel targets for e.g. oligonucleotide-based therapy [112]. In particular, DNA-Gapmer antisense oligonucleotides, of which several have received therapeutic FDA-approval, are suitable for targeting lncRNAs as they can evoke RNASEH1-mediated transcript degradation both within the nucleus and cytosol.

In conclusion, the description of lncRNAs interacting with chromatin remodeler complexes in this review, intends to highlight the importance of lncRNA-mediated chromatin remodeling via remodeler complexes in physiological and carcinogenic conditions.