DNA-RNA hybrid immunoprecipitation (DRIP) is the primary method used to identify genome-wide RNA-DNA hybrids, the main component of an R-loop (Ginno et al., 2012) (Table 1). DRIP was developed in 2012 and relies on the recognition of RNA-DNA hybrids by the S9.6 antibody. The S9.6 antibody was first thought to recognize RNA-DNA hybrids without sequence specificity. However, a more detailed analysis has shown that it exhibits a bias toward GC-rich sequences and longer RNA-DNA hybrids (Phillips et al., 2013; König et al., 2017). S9.6 antibody also recognizes dsRNA with an affinity five times lower than RNA-DNA hybrids, forcing the removal of single- and double-stranded RNAs (ssRNA and dsRNA) with RNase A and RNase III treatments before nucleic acid fragmentation and immunoprecipitation and precluding the use of the S9.6 antibody for microscopy approaches due to the detection of RNA artifacts (Hartono, 2018; Smolka et al., 2021). DRIP relies on cell lysis and protein removal followed by total nucleic acid extraction with phenol-chloroform and ethanol precipitation. As a quality control to call peaks and identify areas in the genome rich in R-loops by DNA sequencing, a negative control sample is treated with RNase H (RNase H1 and H2) to remove RNAs from RNA-DNA hybrids. This basic DRIP protocol was first developed in Ntera2 mammalian cell lines and optimized to identify R-loop forming sequences in yeast (Wahba et al., 2016; Hartono, 2018), plant tissues (Xu et al., 2020), Drosophila embryos, and cell lines (Alecki et al., 2020). Additionally, DRIP has been improved to determine the DNA strand specificity, using either the DNA or RNA moiety of the sample, and to increase the nucleotide resolution in sequencing experiments (Sanz et al., 2016; Sanz and Chédin 2019).

In 2020, a single-molecule assay was developed to obtain a near single-nucleotide resolution of the R-loop forming sequence (Malig and Chedin 2020). Unlike the previously mentioned protocol, this new method relies on non-denaturing bisulfite conversion of cytidine (C) to uridine (U) in the displaced single-stranded DNA followed by sequencing long-read, which identifies the entire R-loop and two mutually exclusive regions forming an R-loop (Malig and Chedin 2020) (Table 1). However, bisulfite converts any C to U in a single-stranded DNA. Therefore, other non-canonical nucleic acid structures that expose an ssDNA, like G-quadruplex or I-motif, can be detected. Hence, the specificity of bisulfite sequencing to detect R-loops should be validated by DRIP-sequencing. Although C methylation make ssDNA insensitive to bisulfite (Moye et al., 2015; Miglietta et al., 2020),R-loop and DNA methylation are incompatible in mammals and plants (Ginno et al., 2012; Xu et al., 2017) and R-loops have been involved in DNA demethylation via their recognition by the protein GADD45A, which recruit the demethylase TET1 (Arab et al., 2019).

Recently, R-loops were detected in vivo by expressing a catalytically inactive RNase H1 fused to the Micrococcal nuclease (MNase) (Yan and Sarma 2020). The expression of this chimeric protein leads to the cleavage of the DNA at the RNase H1 binding site. R-loops identified by this approach might be a minor fraction because RNase H1-MNase fusion protein would only bind R-loops that are protein-free and should outcompete the endogenous RNAse H1. Hence, this technique might be biased to identify stabilized R-loops because they are likely resolved by the endogenous RNase H1. Under physiological conditions, RNase H1 is present in the nucleus, but only a small fraction is bound to the DNA as RNase H1 is essentially targeted to chromatin under stress conditions (Lockhart et al., 2019). RNase H1 has two distinct RNA-DNA hybrid binding sites separated by multiple amino acids that have to bind, meaning that this approach will not be able to detect the smallest R-loops (Nowotny et al., 2008). Additionally, the catalytically inactive RNase H1 has a lower affinity for RNA-DNA hybrid than S9.6 antibody, making this technique less sensitive for detecting R-loops than DRIP (Ginno et al., 2012). In spite of these caveats, combining this technique with bisulfite conversion has made possible to provide strand specificity of the R-loops identified with MapR (Wulfridge and Sarma 2021).

DNA:RNA triplexes are composed of two DNA strands and one RNA in the major groove of the dsDNA formed by a Hoogsteen base pairing between the RNA and the purine-rich DNA strand (Li et al., 2016). The absence of antibodies or proteins that specifically recognize and bind these structures has made difficult their identification in vivo. The first protocol developed to discover DNA:RNA triplexes genome-wide (triplex capture assay) involves the incubation of biotinylated-RNA with isolated nuclei followed by streptavidin pull down to detect the DNA bound to the RNA (Postepska-Igielska et al., 2015). However, triplex capture assay creates artificial DNA:RNA triplexes, prevents the identification of endogenous triplexes, and does not distinguish DNA:RNA triplexes from hybrid G-quadruplexes or RNA-DNA hybrids. In 2019, Sentürk Cetin et al. developed other protocol to identify DNA:RNA triplexes in vivo in mammalian cells for the first time (Sentürk Cetin et al., 2019). Total nucleic acids are extracted using gentle lysis in the presence of proteinase K and RNA-DNA hybrids and ssRNA are removed by a short treatment with RNase H and RNase A, respectively. DNA- and RNA-associated DNA was immunoprecipitated with an antibody that recognizes DNA, then RNAs were eluted from the beads by digesting the DNA with DNase, and the RNA part of the triplexes were sequenced. In this protocol, samples are treated with a commercial RNase H for 30 min, which is insufficient to remove all RNA-DNA hybrids; thus, a fraction of the RNA-DNA hybrids, the main part of R-loops, might still be present and identified as DNA:RNA triplexes (Alecki and Francis 2021). As a result, they observed that 20% of the peaks of their putative DNA:RNA triplexes overlapped with previously published R-loop forming sequences, suggesting that these RNAs tend to form different non-canonical nucleic acid structures or that their protocol is not fully specific for the identification of DNA:RNA triplexes in vivo. Even so, this protocol allowed the identification of DNA:RNA triplexes in vivo for the first time, validating that the triplex capture assay using biotinylated RNA can be a tool to identify triplex-forming sequences.

RNA molecules interact with other RNA molecules directly or through proteins. RNA-RNA interactions mediated by proteins are detected using cross-linking reagents, such as UV or formaldehyde. The RNAs are fragmented, and their extremities in proximity are ligated, and the RNA hybrids or chimeras are sequenced to identify RNA-RNA interactions (Gong et al., 2018; Lin and Miles 2019). The direct interaction between two RNA molecules is detected using the compound psoralen. Psoralens were used initially to identify RNA structure because they intercalate into dsRNAs. The subsequent cross-linking of dsRNAs with UV, removal of ssRNA by RNases, and ligation of dsRNA ends afford to sequence the chimeric RNAs (Lu and Chang 2018).

RNA-protein interactions are commonly studied by RNA immunoprecipitation (RIP) and cross-linking immunoprecipitation (CLIP). CLIP relies on UV cross-linking before pulling down the protein of interest, followed by RNA sequencing in order to identify bound RNAs (Ramanathan et al., 2019). CLIP is also used to identify the specific nucleotide sequence bound by a protein. For this purpose, the unprotected RNA is digested with RNases, and protein-protected RNA analyze by sequencing (Hafner et al., 2021). However, the presence of post-translational modification can decrease the efficiency of UV cross-linking (Vieira-Vieira et al., 2021). Complementary techniques to CLIP have been developed to pull down the RNA and identify its interacting proteome by mass spectrometry. Most of these methods require the insertion of exogenous sequences, like the MS2 stem-loops, into the RNA sequence to facilitate its pull down with the cognate binding protein (Slobodin and Gerst 2010; Hafner et al., 2021). Alternatively, the exogenous expression of a modified CRISPR-Cas9 (clustered regularly interspaced short palindromic repeats-Cas) system where the inactive Cas9 fused to a BirA protein is directed to a specific mRNA and interacting proteins get labeled and recognized by BioID (Yi et al., 2020). A recently developed method identifies proteins at the proximity of a particular endogenous RNA without exogenous expression of factors or modification of RNAs (Yap et al., 2021). The RNA of interest is hybridized with digoxigenin-labeled antisense probes, and the digoxigenin is then bound by the so-called hybridization proximity (HyPro) enzyme. The RNAs and proteins in the proximity of the target mRNA are biotinylated by HyPro enzymes, pulled-down using streptavidin beads, and identified by RNA-seq and mass spectrometry, respectively. Even if this technique is limited to the RNA and protein interactome of a specific RNA, it may provide information on low abundant RNAs because it doesn’t require the immunoprecipitation step. A limitation of this method is the difficulty to discern between proximal or interacting proteins and RNAs. All techniques aiming to identify RNA-protein interactions lead to the detection of direct and indirect associations. In addition, these techniques are more qualitative than quantitative, and quantification only refers to a relative comparison between experimental conditions (Vieira-Vieira et al., 2021). Their main caveat is the limitation to a single RNA or protein and miss the complete picture of all cellular RNA-protein interactions.

Single-molecule fluorescence in situ hybridization (smFISH) is a robust technique widely used to quantify and localize single RNA molecules inside cells. smFISH was designed initially to detect mRNAs through the multimerization of specific antisense DNA oligonucleotides that are fluorescently labeled (Femino 1998). The localization of several fluorophores on an RNA molecule provides the sufficient signal-to-noise ratio to specifically detect the fluorescence signal from an mRNA, whose localization is later analyzed by specific computational frameworks, like Fishquant (Mueller et al., 2013). Given that an RNA molecule must accommodate several probes to be detected, smFISH has worked to detect, quantify the abundance, and analyze the distribution of long ncRNAs (Cabili et al., 2015). The distribution of long ncRNAs in the cell is diverse. Still, it shows a predominantly nuclear localization that appears as discrete nuclear aggregates in 20% of the cases, supporting their role in gene expression and chromatin structure. A simultaneous two-color smFISH has distinguished the long ncRNA accumulation at different nuclear locations and transcription sites. In this protocol, the second set of probes, labeled with a different spectral fluorophore, detect the transcription of a proximal divergent gene to the ncRNA or a gene known to be regulated by the ncRNA (Cabili et al., 2015).

Given the multimerization nature of smFISH, the original protocol had to be adapted for the specific detection of small ncRNAs, whose sequence is not long enough to accommodate several antisense probes. The single-molecule resolution for small ncRNAs has been accomplished by increasing the signal provided by fewer antisense oligonucleotides and branching DNA probes. Amplification of the signal provided by a single antisense oligonucleotide with a hapten by tyramide signal amplification (TSA), TSA-FISH. TSA signal amplification utilizes the enzymatic reaction of the horseradish peroxidase and fluorescently labels tyramide substrates that are deposited at the reaction site (Kwon 2013). In branched FISH (bDNA FISH), two antisense oligonucleotides hybridized the RNA and provided a linker sequence. The linker sequence is hybridized by an amplifier that provides a sequence to multimerized oligonucleotides that are recognized by fluorescently labeled probes (Battich et al., 2013; Sinnamon and Kevin 2014).

Visualizing RNAs in live cells to investigate their expression dynamics relies upon the same fluorescent multimerization principle that smFISH. The MS2 and PP7 are the two most successfully used systems for ncRNAs, like telomeric RNA TERRA (Cusanelli et al., 2013), the X-inactive specific transcript Xist , and the antisense GAL10 ncRNA (Lenstra et al., 2015). The MS2 and PP7 systems are RNA aptamers derived from bacteriophages that are inserted as an array of 12 to 24 stem-loops in the gene encoding for the ncRNA. Each of the stem-loops is tightly bound by the cognate protein, MCP and PCP, respectively, fused to a fluorescent protein, like the green fluorescent protein (GFP) (Vera et al., 2016). The MS2 and PP7 have been combined to tag in the same cell an ncRNA and its regulated mRNAs to link ncRNA expression dynamics to their regulatory function on mRNA transcription (Lenstra et al., 2015). The caveats of the MS2 and PP7 systems to follow the expression of small ncRNAs is that they preclude their function by adding a long RNA sequence and a bulky protein load.

Genome-wide studies have identified R-loops during the development of several organisms. R-loops have been observed in Drosophila melanogaster embryos at two different stages of development: at two to 6 h (h) that corresponds to the steps of cellularization and gastrulation, and at 10–14 h during the embryo segmentation (Alecki et al., 2020). The study was restricted to detect R-loops in DNA elements called Polycomb response elements (PREs). PREs can be transcribed into ncRNAs and form R-loops. A subset of R-loops formed in PREs in 2–6 h old embryos was absent in 10–14 h embryos. PREs covered by R-loops had a higher level of the repressive transcription mark H3K27me3 than PREs lacking R-loops, and PREs which have formed an R-loop only in 2–6 h old embryos are more likely to be repressed by Polycomb group proteins in 10–14 h old embryos. Polycomb group trimethylated the lysine 27 of histone 3 (H3K27me3) and bound to it to compact the chromatin and inhibit its remodeling. Hence, this result suggested that forming an R-loop at PRE sequences during an early developmental stage may lead to Polycomb-dependent gene repression at a later stage. Their data shows that only a small number of R-loops are exclusively formed in 2–6 h embryos, but a higher number of them formed in early embryos and showed a decrease in the R-loop level in 10–14 h embryos, suggesting that upon development and cell differentiation some cells form R-loop and some other don’t.

R-loops forming at PREs are not the only developmentally regulated R-loops in Drosophila. Drosophila embryos also form R-loops in retrotransposon sequences known as Long interspersed nuclear elements (LINE). R-loops are enriched in LINEs and retrotransposons in embryos but S2 cells (derived from 20 to 24 h old embryos) have them enriched at low complexity sequences and simple repeats (Zeng et al., 2021). The formation of R-loops over LINE sequences was previously reported in the human LINE-1 sequence, and it is related to the Aicardi-Goutières syndrome, an inherited encephalopathy that affects newborn infants. The Aicardi-Goutières syndrome is associated with mutations in the exonucleases TREX1 or SAMHD1, the RNA deaminase ADAR1R, and the RNase H2 (Benitez-Guijarro et al., 2018). All these mutations lead to an increase of R-loop levels and DNA hypomethylation, a decrease of LINE-1 retrotransposition, which result in an induction of the innate immune response and neurological dysfunction (Lim et al., 2015). RNase H2 activity is also required to degrade the RNA part of the RNA-DNA hybrid intermediate and complete the retrotransposition, suggesting that R-loop formation and resolution are two essential steps of transposon insertion (Benitez-Guijarro et al., 2018). R-loops misregulation has also been linked to myelodysplastic syndrome, a blood cell cancer in which blood cells don’t mature (Sperling et al., 2017). The myelodysplastic syndrome can be caused by a decrease of the DEAD-box protein DDX41, a putative RNA helicase that was shown to interact with RNA-DNA hybrids (Wang et al., 2018). The reduction of DDX41 leads to an increase of R-loops in hematopoietic stem and progenitor cells and increases inflammatory signaling, leading to a rise in the number of hematopoietic stem and progenitor cells (Weinreb et al., 2021).

These studies, together with the fact that R-loops are detected at repetitive Satellite sequences in mammalian cells (Zeng et al., 2021), suggest that R-loop dynamics during mammalian development might support gene expression changes during this process. Mouse embryonic stem cells (mESC) have been used to identify R-loop interactors using immunoprecipitation with the S9.6 antibody followed by mass spectrometry (Wu et al., 2021). RBP1, a subunit of RNA polymerase II, and DHX9, RNA helicase, were among the already known R-loop interactors and regulators identified in this study. Additionally, nucleolar proteins and proteins involved in rRNA transcription and processing were particularly enriched. This result was expected because the nucleolus is one of the hotspots for R-loop formation (Sollier et al., 2014). Some proteins of the DEAD-box family, which are known to be involved in R-loop regulation and bind RNAs were also enriched (Cargill et al., 2021). Interestingly, the knockdown of several members of the DEAD-box family leads to the overexpression of genes involved in cell fate commitment and neuronal differentiation (Wu et al., 2021), suggesting that the R-loops regulated by these proteins may be important for the pluripotency of mESC.

The idea that R-loop regulation is essential for the pluripotency of stem cells is further supported by experiments done to reprogram Mouse Embryonic Fibroblast (MEFs) into pluripotent stem cells (PSC) (Li et al., 2020). During the dedifferentiation of MEFs using the Yamanaka cocktail, R-loop forming sequence changes preceded gene expression changes (Li et al., 2020). Interestingly, the stabilization of R-loops by knocking down RNase H1 using shRNA or expressing a catalytically inactive RNase H1 decreased or prevented MEF reprogramming, respectively. Genes involved in somatic cell state had increased R-loop levels upon the expression of the catalytically inactive RNase H1 in MEFs, which required a higher amount of the Yamanaka factor Sox2 to restore pluripotency. Surprisingly, Sox2 did not reduce R-loops’ total level but induced R-loops formation at other genes, including pluripotency genes. Sox2 interacts with Ddx5, a member of the DEAD-box family, and Dhx9, an RNA helicase, and inhibits their R-loop resolution activity. Overall, (Yaoyi et al., 2020) have highlighted the relevance of the tight regulation of R-loops to maintain the cell somatic state or reprogram pluripotency.

R-loop formation also coordinates different aspects of plant development. One study in Arabidopsis thaliana and another in rice have looked at R-loop formation genome-wide in various tissues and under different growth conditions (Fang et al., 2019; Xu et al., 2020). Differences in R-loop formation were found between seeds and flowers in Arabidopsis thaliana and seed and leaf callus in rice, where 30% of the R-loop forming sequences were different. Although the expression of ncRNA was not investigated, differences in R-loop formation between seeds and developed plants suggested that R-loops are developmentally regulated and, therefore, differentially expressed ncRNAs are probably involved in the formation of these non-canonical nucleic acid structures. Hence, efforts done over the last decade have identified R-loop forming sequences genome-wide in multiple organisms.

Recent studies have started to highlight the role of R-loops in regulating the expression of ncRNAs and transcription factors with a role in development. There are two examples, one in plants and another in mammalian cell lines, that have been well documented. Among R-loops whose expression changed between the vegetative and reproductive state in plants, one of the best studied examples is in Flowering locus C (FLC) gene. In order to flower only after winter, plants like Arabidopsis thaliana need to repress the FLC gene (Costa and Dean 2019). Long cold exposure during vernalization leads to the progressive epigenetic silencing of the FLC gene. Three ncRNAs are transcribed from the FLC locus: COOLAIR, COLDAIR, and COLDWRAP. COOLAIR is a set of ncRNA antisense transcripts that originate close to the termination region of FLC. COOLAIR forms an extensive R-loop covering the FLC 3′ region (Sun et al., 2013), which in warm conditions triggers chromatin modifications that decrease transcriptional output from the whole locus. The R-loop is resolved by 3′ processing factors terminating COOLAIR transcription, in a process involving m⁶A methylation (Xu et al., 2021) (Figure 2). The R-loop is also important after 1–2 weeks of cold exposure where COOLAIR expression increases and clouds of nascent transcript form around the locus. This leads to a decrease of FLC expression because of their mutually exclusive production (Rosa et al., 2016). After 20 days of vernalization, COLDAIR, sense ncRNA transcribed from the first intron of FLC, is expressed. It is reported to interact with Polycomb complex PRC2 at the FLC locus (Kim et al., 2017). There is an initial nucleation site of PRC2 which tri-methylates the lysine 27 of histone 3, spread along the chromatin leading to stable repression of FLC (Yang et al., 2017). Finally, COLDWRAP, sense ncRNA transcribed in the promoter region of FLC, is expressed after COLDAIR and can serve as a backup mechanism in the absence of COLDAIR, allowing the Polycomb-dependent repression of FLC following transfer back to warm (Kim and Sung 2017). The promoter of FLC and its first intron form a gene loop identified by 3C depending on ncRNA COLDAIR and COLDWRAP (Kim and Sung 2017). The loop formation is thought to reinforce recruitment of PRC2 and its epigenetic mark H3K27me3. A second loop between the FLC promoter and termination site associates with FLC gene activity and this is lost as plants are transferred to cold (Crevillén et al., 2013).

The transcription factor TCF21 plays a critical role in developing several cell types during embryogenesis (Lotfi et al., 2021). The expression of the TCF21 gene in human embryonic cells is controlled by an R-loop formed by the lncRNA TARID in the promoter of TCF21 gene (Figure 3). The RNA-DNA hybrid part of this R-loop is recognized by the growth arrest and DNA-damage-inducible protein GADD45 alpha (GADD45A) protein, which interacts with the enzyme Ten-eleven translocation methylcytosine dioxygenase 1 (TET1). TET1 mediates the DNA demethylation of TCF21 promoter, leading to its expression (Arab et al., 2019). Interestingly, all the previously mentioned events occur following DNA replication in a specific order during the cell cycle. DNA replication leads to the eviction of nucleosomes and their distribution between the two new DNA molecules. Likewise, the level of DNA methylation is divided by two and must be restored to maintain gene expression. Transcription of TARID starts at the beginning of the replication phase, and it is followed by R-loop formation, which recruits GADD45A and DNA demethylation to the TCF21 promoter in the middle of the replication phase and leads to the transcription of TCF21 during the S phase. This R-loop dependent targeting of TET1 is not restricted to the TCF21 gene because the overexpression of RNase H1, enzyme degrading the RNA moiety of an RNA-DNA hybrid, leads to a decreased binding of GADD45A and TET1 at other binding sites controlling the expression of transcriptional coactivators and repressors, and chromatin binding proteins. On the contrary, knockdown of RNase H1 induces the stabilization of the R-loop and an increase in the general DNA methylation level (Arab et al., 2019).

Both RNA and DNA involved in DNA:RNA triplexes have been identified in human cells independently (Sentürk Cetin et al., 2019). As expected, more than 50% of the RNAs involved in DNA:RNA triplex formation come from genes, and 8% of these structures are composed of ncRNAs. Likewise, 50% of the DNA sequences involved in DNA:RNA triplexes correspond to coding genes and 8% to ncRNAs. DNA and RNA involved in triplex are present in transcription start sites and enriched with active epigenetic marks H3K4 methyl. Surprisingly, the RNA part but not the DNA of DNA:RNA triplexes are enriched at super-enhancer, which are transcribed bidirectionally into ncRNAs (Xiao et al., 2021), suggesting that this RNA transcribed from enhancer can form triplexes in trans. Hence, it was proposed that these ncRNAs produced at enhancers might contact distal regions via the formation of DNA:RNA triplexes and induce transcription (Sentürk Cetin et al., 2019). However, the examples for which mechanistic details are available involved gene repression by DNA:RNA triplex.

Two examples highlight the relevance of ncRNAs forming DNA:RNA triplexes during development. Fendrr is an essential ncRNA expressed in the mesoderm, whose depletion leads to embryonic death (Grote et al., 2013). The depletion of Fendrr leads to a decrease of the repressive mark H3K27me3 and an increase of the active mark H3K4me3 in the promoter region of transcription factor like Foxf1 (Grote et al., 2013). Fendrr interacts with the Polycomb complex PRC2, which trimethylates the lysine 27 of histone 3 promoting stable gene repression (Khalil et al., 2009). Using Chromatin Isolation by RNA purification (ChIRP)-seq analysis, it was shown the interaction of Fendrr with the promoter region of Foxf1 and Gata6, in a manner non-sensitive to RNase H, suggesting that Fendrr form an DNA:RNA triplex (Grote and Herrmann 2013) (Figure 4). The authors proposed a model in which Fendrr ncRNA bound by PRC2 includes an DNA:RNA triplex in the promoter region of its target genes, allowing the sequence-specific recruitment of PRC2, the deposition of H3K27me3, and the stable gene repression necessary for the proper embryonic development.

Promoter-associated RNAs, pRNAs, of ribosomal genes (pRNAs) are ncRNAs transcribed from the PolI promoter located in the intergenic region of ribosomal DNA (rDNA) (Vydzhak et al., 2020; Wehner et al., 2014; Santoro et al., 2010). These ncRNAs are degraded to ∼250 nucleotides long fragments that are bound by the chromatin remodeling complex NoRC (Vydzhak et al., 2020). pRNAs can bind the promoter sequence of rDNA by sequence complementation. The structure formed by pRNA and its target DNA is not sensitive to RNase H, nor RNase VI indicating that it is not composed of RNA-DNA hybrid nor dsRNA, respectively, and is not mediated by protein. Thus it is likely an DNA:RNA triplex as the EMSA (electrophoretic mobility shift assay) with dsDNA and pRNAs suggested (Schmitz et al., 2010). The formation of the DNA:RNA triplex between pRNAs and the promoter of rDNA leads to the targeting of the methyltransferase DNMT3b, which methylates the DNA (Schmitz et al., 2010) (Figure 5). NoRC binds pRNAs by recognizing a stem-loop structure, and this binding is essential for NoRC targeting to the nucleolus (Mayer et al., 2008). NoRC complex is then implicated in the deacetylation of histone H4 and the trimethylation of H3K20 an epigenetic repressive mark, allowing the stable repression of rDNA genes (Guetg et al., 2010).

More examples for DNA:RNA triplexes in development are expected, given recent findings in cell lines in culture. Taking advantage of the technique developed by Sentürk Cetin that identifies both the DNA and RNA part of DNA:RNA triplexes, it has been found that the ncRNA NEAT1 has two regions able to form DNA:RNA triplex in vivo. One of the triplex-forming regions was identified in HeLa S3 cells and the other in U2OS, suggesting that multiple domains from a single RNA can form triplex and their formation can be developmentally regulated or, at least, cell type-specific (Sentürk Cetin et al., 2019). ncRNA NEAT1 was shown to bind DNA and regulate gene expression by binding transcription factors and preventing them from being targeted to their binding sites. It can also bind and recruit PRC2 to repress specific genes. Finally, NEAT1 is also involved in RNA splicing and RNA stability (Wang et al., 2020, 1). The pull-down of NEAT1 RNA after incubation with protein-free DNA reveals that this ncRNA can bind multiple DNA sequences suggesting that NEAT1 can act in trans to target proteins by forming DNA:RNA triplexes (Sentürk Cetin et al., 2019). The research done so far to investigate the mechanism by which DNA:RNA triplexes operate during development shows that they mediate the modification of the DNA and chromatin that lead to the repression of protein-encoding and ribosomal genes.