Long noncoding RNAs in DNA methylation: new players stepping into the old game
- 2.2k Downloads
Long non-coding RNAs (lncRNAs) are being discovered as a novel family of regulators of gene expression at the epigenetic level. Emerging lines of evidence demonstrate that interplays between lncRNAs and DNA methylation machinery are an important layer of epigenetic regulation. Here in this mini-review we summarize the current findings in the field and focus particularly on the interactions mediated through direct physical association between lncRNAs and DNA methyltransferases (DNMTs).
KeywordslncRNAs Dnmt Methylation Epigenetics
chromatin isolation by RNA purification
DNMT1-associated colon cancer repressed lncRNA 1
embryonic stem cell
long intergenic ncRNA
long noncoding RNA
promoter of MAT2A-antisense radiation-induced circulating lncRNA
polycomb repressive complex 2
RNA immunoprecipitation sequencing
TCF21 antisense RNA inducing demethylation
X-inactivation specific transcripts
DNA methylation is one of fundamental epigenetic mechanisms to regulate gene transcription, which involves the addition of methyl group to cytosines that are typically restricted to CpG dinucleotides . The CpG-rich regions in genome, referred to as CpG islands, often occur at promoter and first-exon regions and are normally unmethylated; when they become methylated, the transcription of the cognate genes will be blocked [2, 3]. The establishment and maintenance of methylation patterns resulting in modulation of gene expression is one of the key steps in epigenetic regulation during normal development programs. Such process is precisely orchestrated by several DNA methyltransferases (DNMTs). In mammalian cells, DNMT3a and DNMT3b are known to de novo establish DNA methylation pattern whereas DNMT1 acts to maintain methylation status during DNA replication [4, 5]. It is now clear that dynamic changes in DNA methylation accompany development and other pathophysiological processes [6, 7]. For example, low methylation level was found at CpG-rich sequences at the pluripotency state of embryonic stem cells (ESCs); when differentiation processes ensue into the three germ layers, a global gain of DNA methylation occurs at the specific regions [7, 8]. In the absence of DNMTs, ESCs are competent for self-renewal but fail to initiate cellular differentiation [8, 9, 10]. Despite a wealth of knowledge on the dynamic changes of DNA methylation in diverse settings, there is still limited understanding of the mechanisms to determine the DNA regions targeted by DNMTs since they lack any sequence specificity other than CpG dinucleotides . The interaction between histone modifiers and DNMTs provides one solution for the achieved sequence specificity. For example, EZH2 (a component of Polycomb Repressive Complex 2) was shown to interact with DNMTs and the presence of EZH2 is crucial for DNA methylation at EZH2-target promoters . The H3K9 methyltransferase SETDB1 also directly associates with DNMT3a/b and their co-binding was observed at RASSF1 promoter . Recently, genome-wide DNA methylation profiles during cell differentiation have revealed the differentially methylated regions are enriched for transcription factor binding sites, implying potential roles of transcription factors in dictating the changes of DNA methylation landscape, which is supported by several pieces of experimental evidence [3, 13, 14]. For instance, E2F6 was found to recruit DNMT3B to a subset of germline gene promoters in somatic tissue, leading to the DNA methylation and the silenced state . Notwithstanding these advances, key questions regarding how sequence-specific DNA methylation is orchestrated and whether additional players are involved remain largely unsolved. The novel family of regulators of gene expression, long non-coding RNAs (lncRNAs) seem to fill the gap.
An outstanding example comes from a paper published in Nature , Di Ruscio and colleagues reported that a class of non-polyadenylated RNAs can regulate DNA methylation at a locus-specific pattern through interacting with DNMT1. In their findings, a nuclear-enriched lncRNA termed ecCEBP (extra-coding CEBPA) is expressed from CEBPA locus and shares a concordant expression pattern with CEBPA in human tissues. Silencing of this lncRNA diminishes the expression of CEBPA, suggesting its activating role. CEBPA is a known factor subject to the regulation of DNA methylation. Intriguingly, depletion of ecCEBP results in the increased methylation intensity at CEBPA promoter region. Moreover, overexpression of a region of ecCEBP is sufficient to alleviate the methylation level at the CEBPA locus and in turn increase the CEBPA expression. It is worthy to note that regulation of DNA methylation by ecCEBP is specific to the locus, since other genomic regions show little changes in DNA methylation pattern upon ecCEBP perturbation. Additionally, the authors demonstrated that ecCEBPA is physically associated with DNMT1; a subsequent detailed analysis revealed the DNMT1 with ecCEBPA RNA domains that assumed a stem-loop structure and mapped the RNA binding interface to a region encompassing the DNMT1 catalytic domain. Finally, the authors assessed the DNMT1-interacting RNAs at a genome-wide level by performing RIP followed by high throughput sequencing. Interestingly, they found an anti-correlation between DNMT1/RNA interaction and the methylation level at corresponding locus . Taken together, this study has provided solid evidence for a model in which a class of non-polyadenylated lncRNAs regulate DNA methylation patterns through their association with DNMT1.
Another example of DNMT1-interacting lncRNA comes from a recent publication from Chalei and colleagues . In this study Chalei et al. revealed that a central nervous system-expressed lncRNA termed Dali is essential for neural differentiation and can regulate neural gene expression partially through interacting with DNMT1 to affect DNA methylation at distal target promoters. Notably, in this study the authors characterized Dali-associated DNA regions at a genome-wide level. They found Dali is associated with active chromatins revealed by the evidence that Dali bound regions are enriched for active histone marks like H3K4me3, H3K4me1 and H3K27ac. Additionally, the authors performed computational analysis to exclude the possibility that Dali directly interacts with DNA. On the contrary, they identified Dali interactions with DNMT1 to bring it to DNA regions; depletion of Dali increases DNA methylation at a subset of gene promoters, suggesting its inhibitory role in DNA methylation. Of note, the target promoters under Dali control are away from its site of transcription, indicating Dali modulates DNA methylation in trans. DNMT1 interacting transcription factors were thought to be required to target DNMT1 to specific loci. Interestingly, binding motifs of 9 such transcription factors are enriched in Dali-bound DNA regions, sparking a possibility that these factors determine the specificity of Dali-mediated DNA methylation change. These observations led to a model that lncRNA influences DNA methylation in trans at distal DNA regions through harnessing DNMT1 machinery .
While the above studies were focused on DNMT1, the maintenance enzyme, other DNA methyltransferases enzymes may also associate with lncRNAs, which may modulate their enzymatic activity and the patterns of de novo DNA methylation. Indeed, a very recent report  from our group identified a lncRNA, Dum which coordinates the skeletal myoblast differentiation program through interacting with several DNMTs including DNMT1, DNMT3a and DNMT3b. In this work, we found Dum is induced by MyoD with an increased expression during differentiation process. Silencing of Dum impinges on myogenic differentiation in vitro and muscle regeneration in vivo, suggesting its pro-myogenic role. Additionally, Dum knockdown influenced the expression of several nearby genes, favoring a cis-action mode; and Dppa2 was found to be a profound downstream target to recapitulate the roles of Dum in myogenesis. Dppa2 was reported to be regulated by DNA methylation and our mechanistic studies revealed it is controlled by the interaction between Dum and the DNMTs. Depletion of Dum remarkably impaired the binding of DNMTs to the CpG regions of Dppa2 promoter; while using chromatin isolation by RNA purification (ChIRP) assay, we retrieved Dppa2 promoter regions, substantiating the indispensable role of Dum in DNMTs regulation on Dppa2 promoter region. Of note, Dum/Dppa2 interaction requires the intra-chromosomal looping between these two loci. Cohesin components, NIPBL and RAD21, were shown to occupy both Dppa2 and Dum regions and knockdown of the factors attenuated the repression of Dppa2 expression, suggesting the long-range chromosomal interaction is required for the repressive role of Dum on Dppa2 locus. Collectively, our study expands the concept of DNMT-interacting lncRNAs and suggests chromatin looping may be necessary to target the lncRNAs to particular loci .
Besides these findings to dissect the connections between individual lncRNAs and DNMTs, it will be of great interest to identify the DNMT-interacting lncRNAs genome-widely. Indeed, the latest work done by Merry et al.  uncovered a total of 148 lncRNAs that are associated with DNMT1 in colon cancer cell through RNA immunoprecipitation sequencing (RIP-seq). Among them, the authors identified one lncRNA, referred to as DACOR1 (DNMT1-associated Colon Cancer Repressed lncRNA 1), that is highly and specifically expressed in normal colon tissue while repressed in colon cancer cell lines. Intriguingly, overexpression of DACOR1 in colon cancer cells resulted in a gain of DNA methylation at multiple loci without changing the DNMT1 expression level. Through ChIRP-seq (ChIRP sequencing) analysis, the authors demonstrated that DACOR1 occupies a total of 338 genomic loci, of which 161 peaks are near 150 annotated genes. Notably, 31 of these sites overlapped with differentially methylated regions (DMRs) previously found in colon cancer samples versus matched normal tissues. These findings indicate that DACOR1, via its interaction with both chromatin and DNMT1, targets DNMT1 protein complex to the right genomic loci. Moreover, DACOR1 was found to repress the expression of Cystathionine β-synthase (CBS) and in turn increase methionine, which is the substrate to produce S-adenosyl methionine (SAM). SAM is the key methyl donor for DNA methylation in mammalian cells. Thus, DACOR1 may also impinge on DNA methylation through orchestrating cellular SAM levels .
In addition to the above examples of lncRNAs which regulate DNA methylation through physical association with DNMTs, many other lncRNAs may participate in modulating DNA methylation indirectly through interacting with other RNA binding proteins. A prominent example comes from a recent report from Bao and colleagues . In their work, Bao et al. unraveled the mechanisms underlying lincRNA-p21 functions in somatic cell reprogramming. They found lincRNA-p21 is induced by p53 but not affect cell apoptosis or cell senescence. Instead, lincRNA-p21 impedes cell reprogramming by maintaining the repressive histone mark H3K9me3 or DNA methylation at the promoters of different sets of pluripotent genes. Mechanistically, such actions are mediated via the interactions of lincRNA-p21 with H3K9 methyltransferase SETDB1 and DNMT1 respectively, both of which are dependent on the RNA binding protein hnRNPk. Knockdown of hnRNPk attenuated the association between lincRNA-p21 with SETDB1 or DNMT1 and in turn enhanced reprogramming efficiency. Hence, hnRNPk serves as a platform facilitating the connection between lincRNA-p21 and DNMT1 to promote DNA methylation at specific loci .
Other lncRNAs involved in DNA methylation
Another well-known lncRNA that affects DNA methylation is Kcnq1ot1, transcribed from the paternal allele of the imprinted Kcnq1 locus. Besides the Kcnq1 gene, this locus encodes eight other genes that are either ubiquitously or placental-specifically imprinted in a Kcnq1ot1-expression-dependent manner . An 890 nt region close to the 5′ end of Kcnq1ot1 is important for silencing and its genetic deletion in mice leads to abnormal silencing of ubiquitously imprinted genes, where a decrease in DNMT1 recruitment and subsequent reduction of CpG methylation occur . However, a direct DNMT1–Kcnq1ot1 interaction has not been demonstrated. Additionally, a recently characterized lncRNA PARTICLE (promoter of MAT2A-antisense radiation-induced circulating lncRNA)  forms a DNA-lncRNA triplex upstream of a MAT2A promoter CpG island and represses MAT2A via methylation. Interestingly, MAT2A gene encodes the catalytic subunit of methionine adenosyltransferase which is an essential enzyme in methylation cycle. It means that lncRNA could interact with DNA methylation to regulate the expression of the methylation-modifying player, reflecting the close relationship between lncRNA and DNA methylation. Instead of showing the direct interaction between PARTICLE and DNMTs, the authors found that PARTICLE could interact with several chromatin suppressors such as G9a and SUZ12 (subunit of PRC2) thus serving as a recruiting platform for gene silencing complex. However, whether the methylation of MAT2A by PARTICLE was mediated by these chromatin suppressive proteins remains to be confirmed . In addition to causing DNA methylation to silence gene expression, lncRNAs could also participate in demethylation leading to gene activation. For example, an antisense lncRNA, TARID (for TCF21 antisense RNA inducing demethylation), could activates TCF21 transcription by inducing demethylation of the TCF21 promoter . Mechanistically, it occurs through TARID binding to the TCF21 promoter and recruiting GADD45A, which then recruits TDG together with TETs to direct base excision repair for demethylation.
YZ drafted the manuscript. HS and HW edited and revised the manuscript. All authors read and approved the final manuscript.
Availability of data and material
The authors declare that they have no competing interests.
The work described in this paper was substantially supported by four General Research Funds (GRF) to HW and HS from the Research Grants Council (RGC) of the Hong Kong Special Administrative Region, China (Project number: 2140874 to HW, 2140806 and 2140861 to HS); One Collaborative Research Fund (CRF) to HW and HS from RGC (Project number: C6015-14G); and one Grant from the Ministry of Science and Technology of China 973 Program (2014CB964700) to HW.