Introduction

RNA interference (RNAi) is an important cellular process that uses small RNA molecules between 20 and 30 nt in length for sequence-specific recognition and regulation of complementary target sequences. RNAi was first shown to act post-transcriptionally by degrading targeted messenger RNAs (mRNAs) in the cytoplasm [1]. This process of post-transcriptional gene silencing (PTGS) has been shown well-conserved among all the model Metazoa organisms. In yeasts and plants, small RNAs were also shown to elicit transcriptional gene silencing (TGS) in the nucleus by guiding repressive epigenetic modifications such as DNA cytosine methylation and histone modifications to complementary genomic targets. Although TGS was initially thought to be absent in Metazoa, recent studies proved that similar mechanisms are active in germ cells of Caenorhabditis elegans, Drosophila, and mammals. These findings suggest that small RNA-mediated TGS is a conserved mechanism in many eukaryotes.

Small RNAs that can operate TGS includes two main types: small interfering RNAs (siRNAs) and PIWI-interacting RNAs (piRNAs). siRNAs are processed from double-stranded siRNA precursors by an RNase-III-like Dicer family protein. After being generated, siRNAs are loaded into an effector Argonaute (AGO) protein to operate TGS or PTGS functions. Epigenetic modification directed by siRNA has been most studied in plants and yeasts, but increasing evidence shows that this process might also occur in metazoan somatic cells [2]. In contrast to siRNAs, piRNA biogenesis is independent of Dicer, and piRNA precursors do not have an obvious secondary structure. The effector proteins for piRNA are PIWI proteins, a germline-specific clade of the AGO protein family. piRNAs and PIWI proteins are not present in plants or fungi, which use siRNA for TGS processes (Fig. 1).

Fig. 1
figure 1

Small RNA-directed transcriptional regulation in three model organisms. a RNA-directed DNA methylation in Arabidopsis thaliana. RNA Pol IV produces siRNA precursor transcripts which are further converted into double-stranded RNAs, processed into 24-nt siRNAs and loaded into AGO4. The AGO4-siRNA complexes then target to the Pol V-transcribed non-coding RNA through sequence complementarity and guide the de novo DNA methyltransferase, DRM2, to methylate DNA. b piRNA-mediated transcriptional silencing in Drosophila melanogaster. Long, polycistronic piRNA precursors are transcribed by Pol II, processed into 24–27-nt piRNAs and loaded into PIWI or AUB proteins. PIWI then translocate into nuclei to target the complementary nascent transcripts, leading to deposition of H3K9me2/3 repressive marks on surrounding chromatin regions. c Small RNA-mediated transcriptional regulation in Caenorhabditis elegans. piRNAs are transcribed, trimmed to 21 nt in length and loaded into PRG1 proteins. PRG1 in turn recruits RNA-dependent RNA polymerase that makes secondary 22-nt sRNAs (22G RNA) which are loaded into WAGO9 or WAGO10. WAGO9-22G RNA complex then translocates into the nucleus, scanning the whole transcriptome to find the complementary transcripts for H3K9me3 deposition on the chromatin. To protect host genes from silencing, CSR-1 binds to another group of 22G RNAs which are complementary to host genes. CSR-1 targeting of host genes increase their transcription and protect them from PRG-1-mediated silencing

Full size image

An essential and conserved function for siRNA in plants and piRNAs in Metazoa is to protect genome integrity by repressing the expression of transposable elements (TEs). Indeed, failure of piRNA repression in flies and mice leads to transposon activation, formation of double-stranded DNA breaks, and sterility. Moreover, small RNAs expressed in the germline cells of C. elegans and Drosophila are inherited by the progeny and might provide an epigenetic signal for inheritance of certain traits in the next generation.

Besides small RNA, studies of long non-coding RNAs (lncRNAs) revealed that some lncRNAs play roles in transcriptional regulation of gene expression. In this review, we outline the biogenesis and mechanisms of transcriptional regulation by different classes of non-coding RNAs.

RNA-Directed DNA Methylation in Arabidopsis

In plants, the process of RNA-directed DNA methylation (RdDM) requires two atypical DNA-dependent RNA polymerases-Pol IV and Pol V (review in [3, 4•]). Pol IV is responsible for making siRNA precursor transcripts, which are converted into double-stranded RNAs (dsRNAs) by RNA-dependent RNA polymerase 2 (RDR2). RDR2 physically interacts with Pol IV and is nonfunctional without Pol IV, suggesting that the conversion of nascent transcripts to dsRNAs is channeled [5, 6]. The dsRNAs are then cleaved into 24-nt siRNAs by Dicer-like 3 (DCL3), followed by methylation of their 3′ ends by HEN1 methyl transferase. Thereafter, one strand of 24-nt siRNAs is loaded into Argonaute 4 (AGO4) in the cytoplasm and the AGO4/siRNA complex is transported into the nucleus [7]. While Pol IV transcribes substrates for siRNA generation, another polymerase, Pol V, transcribes RNA that is targeted by the AGO4/siRNA complex. AGO4 interacts with Pol V and KTF1 protein that binds nascent transcripts generated by Pol V. The AGO4/Pol V/KTF1 complex was proposed to guide the de novo DNA methyltransferase DRM2 to chromatin at targeted loci promoting cytosine methylation at sites containing CG, CHG, and CHH motifs. Interactions between AGO4 and DRM2 were demonstrated in vivo, supporting this hypothesis (Fig. 1a) [8••].

The mechanism of DRM2-mediated DNA methylation is not yet fully understood. A recent study reported the crystal structure of DRM2 and showed that it forms a homodimer critical for catalytic activity. Furthermore, DRM2 preferentially methylates one DNA strand, likely the strand acting as the template for RNA polymerase V-mediated transcription. Taken together, tethering of the AGO4/siRNA complex to Pol V recruits DRM2 to methylate the Pol V-associated strand of DNA. DRM2 methylation seems to prefer double-stranded DNA, so how strand information for DRM2 target selection is delivered remains elusive [8••]. A multifunctional protein RDM1 may also play an important role in the recruitment of DRM2, since it is the only known protein that interacts with both AGO4 and DRM2, in addition to binding methylated single-stranded DNA [9].

How Pol IV and Pol V are specified to transcription of certain genomic regions is not fully understood. Analysis of Pol IV occupancy on DNA by ChIP-seq indicated that it is enriched at pericentromeric heterochromatin, which are regions enriched in transposons and other repetitive elements [10••]. Pol IV is recruited to a large subset of its genomic targets by a Pol IV-interacting protein Sawadee-Homeodomain Homolog1/DNA-Binding Transcription Factor 1 (SHH1/DTF1) [10••, 11••]. The SAWADEE domain of SHH1/DTF1 forms a unique tandem Tudor-like fold and functions as a dual lysine reader to probe both methylated H3K9 and unmethylated H3K4 residues of chromatin [10••, 11••]. The binding specificity of SAWADEE domains is consistent with SHH1/DTF1-dependent Pol IV ChIP-seq peaks, which are present in genomic regions enriched for H3K9me2 and depleted for H3K4me marks [10••]. The key residues in both lysine binding pockets of SHH1/DTF1 are required for Pol IV occupancy and activation of RdDM, suggesting that Pol IV is recruited to genomic regions through SHH1/DTF1 [10••]. SHH1/DTF1 may require chromatin-remodeling proteins like CLSY1 to change chromatin structure, facilitating the loading of Pol IV onto DNA [6, 10••, 11••].

In contrast to Pol IV, Pol V has been shown to preferentially associate with euchromatic regions, particularly at recently integrated small transposons and at promoters of genes which contain transposons or other repeats within their promoters, introns, or coding regions [12–16]. The recruitment of Pol V to target loci has recently been shown to be dependent on SUVH2 and SUVH9, both members of the Su(var)3-9 histone methyltransferase family [17••, 18••]. SUVH2 and SUVH9 showed no histone methyltransferase activity [19, 20]. Instead, they bind methylated DNA through their SRA domains, which are required for DNA methylation and transcriptional silencing functions of both SUVH2 and SUVH9 [19]. SUVH2 and SUVH9 physically interact with DMS3 and DRD1, which both belong to a chromatin-remodeling complex termed DDR (DRD1, DMS3, and RDM1) [17••, 18••]. The DDR complex was demonstrated to be essential for Pol V occupancy on chromatin and Pol V-dependent DNA methylation and siRNA accumulation [13, 14]. Therefore, SUVH2 and SUVH9 may facilitate Pol V binding to chromatin containing pre-existing DNA methylation, likely by recruiting the DDR complex and forming a positive feedback loop that further enhances DNA methylation.

An intriguing question of how new, invasive transposons are detected and silenced remained unknown because the RdDM pathway was previously investigated in steady-state systems containing long-established conditions of transcriptional silencing. However, a recent study described the sequence of events leading to de novo silencing of the Evadé (EVD) retrotransposon [21•]. At an early stage after activation of EVD (generations F8–F11), the numbers of new EVD insertions and expression level of its RNA were increased, in addition to an increase in 21–22-nt EVD siRNAs, which are processed by DCL2 and DCL4. siRNAs target EVD transcripts through PTGS; however, EVD RNA was only partly degraded, attributable to protection by retrotransposon-derived GAG protein. In generations F14, genomic insertions of EVD reached a plateau of ~40 copies, coinciding with the appearance of EVD-derived 24-nt siRNA, LTR methylation and decrease of 21–22 nt siRNA. This suggested that EVD had undergone TGS at this stage and was successfully attenuated. The PTGS-to-TGS shift seems to be initiated by the appearance of 24-nt siRNA, which may be generated by DCL3 when DCL2 and DCL4 that generate 21–22-nt siRNA are saturated. The 24-nt siRNA was loaded into AGO4 and induced de novo methylation of loci-expressing EVD transcripts, which in turn initiates antisense transcription and the spread of methylation toward the 5′-LTR [21•]. Pol IV and Pol V are believed to be responsible for antisense transcription, but a detailed mechanism is lacking.

Transcriptional Silencing by piRNA in Drosophila melanogaster

PIWI-interacting RNAs or piRNAs of length of 24 to 30 nt associate with PIWI clade Argonaute proteins and are expressed in germ cells and associated follicular cells in Drosophila ovary and testis. The population of piRNA in germ cells are highly complex as millions of distinct piRNA molecules were annotated. About 80 % of the piRNA population is mapped to annotated transposons or transposon remnants [22]. Mapping piRNAs to genomic regions revealed that they are derived from discrete genomic loci called piRNA clusters [22]. These clusters are mostly located in percentromeric and subtelomeric heterochromatic regions, which are enriched in transposable element sequences [22]. piRNA clusters are believed to be transcribed by RNA polymerase II (Pol II) as long, continuous transcripts, which have no obvious secondary structures and can span up to 200 kb in length. It is thought that piRNA biogenesis begins with the endonucleolytic cleavage in the long precursor transcript, generating shorter piRNA precursors. The cleavage is likely executed by the endonuclease (Zuc) [23–25]. After cleavage, the 5′ end of piRNA precursors is loaded into PIWI family proteins, PIWI and Aubergene (AUB). Once loaded into PIWI proteins, the piRNA precursor is trimmed at its 3′ end by an unidentified 3′-to-5′ exonuclease to mature piRNA length, followed by 2′-O-mehylation by Hen1 methyl transferase [26]. Loading of piRNA into AUB initiates biogenesis of secondary piRNA called ping-pong amplification, which simultaneously destroy transposon transcripts in cytoplasm and increases the production of piRNAs targeting active transposon sequences [22, 27]. Conversely, once loaded with piRNAs, PIWI translocates into nuclei to regulate chromatin structure, resulting in TGS (Fig. 1b).

The genome of Drosophila lacks DNA methylation, and the properties of chromatin are mainly defined by histone modifications and histone variants. It is believed that PIWI/piRNA mediates TGS by directing the deposition of repressive histone marks on targets recognized by piRNAs. Indeed, abolishment of the ability for PIWI to localize to the nucleus leads to transposon de-repression, coupled with an increase of active histone marks (H3K79me2 and H3K4me2) and a decrease of repressive marks (H3K9me2/3) over several transposable elements [28, 29]. These observations were further expanded by three recent genome-wide studies. Knockdown of PIWI leads to transcriptional de-repression of a significant fraction of transposons accompanied by loss of H3K9me2/3 marks and increase in RNA pol II occupancy on their sequences in both cell culture and in fly ovaries [30••, 31••, 32••].

PIWI is guided to genomic targets by piRNA. This is supported by the fact that the piRNA pathway does not target most host genes and that PIWI is required to be associated with piRNAs to repress TEs [31••]. Moreover, PIWI loaded with artificial piRNAs targeting a LacZ transgene leads to efficient lacZ silencing accompanied by accumulation of repressive chromatin marks and a decrease in both active chromatin marks and RNA pol II occupancy [31••]. Although the mechanism of target recognition by the PIWI/piRNA complex is unclear, it is believed that piRNAs recognize nascent transcripts through sequence complementarity. Indeed, PIWI knockdown causes a decrease in H3K9me3 signal only on actively transcribed TEs but not on untranscribed transposon remnants, which are transcriptionally inactive [30••].

Recently, two factors were proposed to be involved in PIWI-mediated TGS. Knockdown of Maelstrom (Mael) resulted in increased Pol II occupancy over transposons similar to changes seen in PIWI knockdown. Interestingly, in contrast to PIWI knockdown that resulted in strong H3K9me3 depletion on piRNA target loci, Mael knockdown does not significantly change H3K9me3 mark [30••]. Another factor Asterix (CG3893) is required for transposon silencing and PIWI-mediated establishment of H3K9me3 mark over at least of a subset of TEs [33]. Overall, while the details of piRNA-mediated transcriptional silencing remain to be elucidated, it is clear that PIWI plays an essential role in transcriptional silencing of TEs and establishment of a repressive genomic environment.

Chromatin structure also plays a central role in piRNA biogenesis, in addition to defining piRNA-generating genomic loci, such as piRNA clusters. Increasing evidence suggests that piRNA clusters have a unique chromatin environment. piRNA clusters show features of heterochromatin, which is enriched with repressive H3K9me3 marks and heterochromatin protein 1 (HP1) that interacts with H3K9me3 marks. Moreover, the heterochromatic environment seems to be required for expression and proper processing of piRNA precursors. Mutation or knockdown of the histone methyltransferase SetDB1/eggless that methylates H3K9 disrupts transcription of piRNA clusters [34]. Nevertheless, deposition of H3K9me3 cannot be sufficient for distinguishing piRNA clusters, because these features are present in many genomic loci that do not produce piRNAs. Interestingly, a germline-specific HP1 homolog protein, Rhino, interacts with H3K9me3 mark through its chromodomain, associates with a majority of piRNA clusters expressed in germ cells and seems to be absent from other genomic regions [35, 36•, 37•, 38•]. Furthermore, mutation of rhino results in piRNA loss and transposon de-repression, indicating its crucial role in piRNA pathway. Several recent studies further suggest that Rhino binding to H3K9me3 may serve as a platform to initiate assembly of a complex required for piRNA precursor transcription and/or processing. Indeed, Rhino forms a complex with two previously identified piRNA pathway proteins, Deadlock and Cutoff, and is required for efficient transcription of piRNA clusters [36•, 37•]. In another study, Rhino was shown to work together with Cutoff and a nuclear DEAD-box RNA helicase protein, UAP56, to inhibit splicing of piRNA precursors. It was proposed that a stalled splicing complex differentiates piRNA precursors from mRNAs and directs their processing to small RNAs [38•, 39]. In both models, Cutoff seems to play a crucial role although the molecular function of Cutoff remains unknown. Further characterization of Cutoff function should help understand more about piRNA biogenesis.

In Drosophila, transgenerational inheritance of piRNA seems to provide a critical trigger that defines genomic regions for piRNA production in progeny. A study by de Vanssay et al. showed that a transgenic locus that generates piRNAs can induce piRNA production from a homologous locus that was originally incompetent for piRNA generation [40]. Significantly, the recipient locus is activated in the progeny of the females that express active locus even if the active locus itself is not inherited, indicating that activation is caused by epigenetic signal transmitted to the progeny independently of the genomic locus. It was proposed that piRNAs in the maternal cytoplasm are the epigenetic signal that activates piRNA biogenesis in the next generation. This hypothesis is consistent with the fact that PIWI/piRNA complexes that are expressed during oogenesis are deposited into the developing eggs and are present in the early embryo [22, 41, 42]. Transgenerationally inherited piRNAs may act as an epigenetic signal for identification of substrates for piRNA biogenesis using two different mechanisms. First, piRNAs associated with nuclear PIWI protein install the H3K9me3 mark on homologous genomic sequences, leading to recruitment of Rhino and other biogenesis components. Second, inherited piRNAs enhance processing of homologous transcripts into mature piRNAs by ping-pong amplification in the cytoplasm [36•, 43•].

piRNA-Mediated DNA and H3K9 Methylation in Mouse

In mouse, piRNA and PIWI proteins are expressed in male germ cells and their expression is tightly regulated during spermatogenesis. Abundant populations of piRNA are expressed during the pachytene stage of meiosis and are called pachytene piRNA. The function and targets of pachytene piRNAs remain unknown; however, it is unlikely that they are involved in transcriptional repression since pachytene piRNAs associate with two PIWI proteins, MIWI and MILI, which are exclusively localized in the cytoplasm. An interesting feature of the genomic regions that encode pachytene piRNAs is that they are transcribed divergently from a central promoter in both directions [44, 45]. A recent study has identified a transcription factor, A-myb, which regulates pachytene piRNA cluster transcription through the binding of a conserved sequence motif in the promoter. A-myb also controls its own transcription and that of several key proteins in the piRNA pathway including MIWI. This suggests that A-myb orchestrates the transcription of piRNA clusters along with expression of the proteins necessary to produce piRNA from those cluster transcripts [46•]. Precursors of pachytene piRNAs are transcribed by Pol II and have features resembling canonical mRNA transcription, such as 5′-cap and 3′-poly(A) tail.

In contrast to pachytene piRNAs, piRNAs expressed in developing germ cells during embryogenesis are involved in transcriptional repression of transposable elements in the nucleus. These embryonic piRNAs associate with two PIWI proteins, MIWI2 and MILI and the loss of either of these proteins leads to transposon activation and sterility [47, 48]. In mouse male germ cells, transposon silencing is established in a narrow window during embryonic germ cell development after global genome de-methylation. This is a critical stage of development, since the failure to establish CpG DNA methylation on TEs leads to their activation and subsequent meiotic failure and sterility [49]. As MIWI2 is located in the nucleus and expressed only in a narrow developmental window exactly overlapping the time of de novo DNA methylation, it is considered as the effector of piRNA-directed DNA methylation in mice. Indeed, CpG methylation patterns were not reestablished on retrotransposon sequences in MIWI2-deficient mice [50]. As MILI is exclusively cytoplasmic, it is unlikely to be the direct effector of DNA methylation. However, MILI is necessary for piRNA loading and nuclear localization of MIWI2, explaining its role in DNA methylation.

Two recent studies expand our knowledge of piRNA-mediated transposon silencing in mice [51•, 52•]. According to these studies, there are two waves of de novo DNA methylation during epigenetic reprogramming in male germ cells. The first wave of de novo methylation acts on most genomic DNA by default; however, a fraction of retrotransposon insertions evade this first wave and remain active. The active retrotransposon transcripts then enter the ping-pong amplification pathway to produce antisense piRNAs that are loaded into MIWI2, adaptively triggering the second wave of de novo DNA methylation in a sequence-specific manner [52•]. Besides DNA methylation, MIWI2 is also required to establish repressive H3K9me3 histone marks on transposon loci indicating that the nuclear functions of piRNAs are well conserved between flies and mammals. It is noteworthy that MIWI2 only targets full-length but not truncated copies of retrotransposons for deposition of repressive chromatin marks, supporting the model that piRNA require nascent transcription to change the chromatin state of transposons [51•].

Small RNA-Mediated Transcriptional Regulation in Caenorhabditis elegans

Small RNA pathways in C. elegans are very complex due to the fact that the worm genome contains more than 20 different Argonaute proteins compared with five in flies and eight in mouse [53]. Each of these Argonaute proteins associate with a distinct set of small RNAs, but there is substantial cross-talk between distinct Argonaute/small RNA pathways. One example of such cross-talk is a complex interaction between four Argonaute proteins, PRG1, WAGO9/10 and CSR1, and distinct small RNA populations associated with each of these proteins that together lead to stable transcriptional silencing of foreign genomic elements alongside protection of host genes.

PRG1 is the homolog of fly and mammalian PIWI proteins in worms. PRG1-bound piRNAs were also termed 21U-RNA for its precise 21-nt length and enrichment for uridine residue at the 5′ terminus [54–56]. In contrast to piRNA in fly and mouse that are processed from long precursor RNAs, each 21U-RNA represents a tiny, autonomous unit. The majority of 21U-RNAs are derived from two 21U-generating clusters on chromosome IV [57]. The C. elegans, the 21U-RNA cassette requires only the 21U-RNA sequence itself and two upstream motifs. One of the motifs is recognized by Forkhead transcription factors to enhance transcription of 21U-RNA precursors by Pol II [58•, 59•]. The transcription of 21U-RNA precursors begins at the second motif, making a ~26-nt capped-21U-RNA precursor followed by trimming extra nucleotides at 5′ and 3′ ends to produce mature 21U-RNAs. Recent studies have identified several 21U-RNA biogenesis factors in C. elegans, some of which are involved in precursor synthesis, while others are involved in precursor processing [58•, 60–62].

In contrast to fly and mammalian PIWI, PRG1 does not relocate into nucleus and is not directly involved in transcriptional silencing. Instead, it recruits an RNA-dependent RNA polymerase that makes a massive number of secondary 22-nt small RNAs (22G RNAs) that are bound by the worm-specific Argonaute proteins WAGO-9 (HRDE-1) and WAGO-10. WAGO-9 enters the nucleus, where it targets RNA transcripts with homology to 22G RNA, mediating TGS along with deposition of the H3K9me3 marks (Fig. 1c) [63•, 64•, 65•]. Once TGS is established, silencing of targets can be maintained for many generations independently of 21U-RNA or PRG-1 expression [63•, 65•, 66•].

Considering that there are more than 30,000 distinct 21U-RNA sequences, in addition to the fact that only partial complementarity is needed to identify targets, almost any foreign sequence can be targeted by PRG1/21U-RNA [67]. On the other hand, this also means that host genes can be potentially repressed. C. elegans has evolved a unique mechanism to protect its own DNA from WAGO-9 targeting: CSR-1 associates with RdRP-derived small RNAs (22G RNA) that are antisense to most (if not all) germline-expressed mRNAs [68, 69]. Strikingly, targeting of CSR-1 to host genes leads to their enhanced transcription and protection from PRG-1-mediated silencing [64•, 67, 70•, 71•, 72•]. Therefore, while fly and mouse use small RNAs exclusively for repression of foreign elements, the worm small RNAs scans throughout the genome, tags self-genes as “good” and represses the unlabeled “bad” invader genes (Fig. 1c). Since CSR-1-associated 22G-RNAs are transmitted into the progeny [69], the small RNA system in worm appears to be a molecular memory for both foreign and host genes.

Long Non-coding RNA in Transcriptional Regulation

To date, more than 10,000 intergenic long non-coding RNAs (lncRNAs) have been annotated in mammals, and a rapidly growing number of lncRNAs have been implicated in a variety of biological processes [73]. Many lncRNAs localize exclusively or partially within the nucleus. There are various genomic origins of lncRNAs, which can be roughly classified into five categories: (1) stand-alone lncRNAs that are intergenic, (2) antisense transcripts that occurs at the opposite DNA strand orientation to a protein coding gene, (3) pseudogenes which have lost their coding ability due to mutations, (4) long intronic ncRNAs, and (5) divergent transcripts transcribed from promoters or enhancers [74].

A group of enhancer RNAs (eRNAs) were demonstrated to modulate expression of neighboring protein coding genes [75, 76•, 77–79]. Two studies have reported that eRNAs associated with their site of synthesis and induced looping of the local chromosomes to bring the enhancers near to the promoters of nearby target genes [76•, 80•]. Interestingly, eRNAs are not limited to the adjacent genes but also a long-range DNA looping interaction up to 27 Mb in length has also been documented [76•]. Bridging of enhancer and promoter loci by eRNAs were thought to act in-cis on the same chromosome, however lncRNA can also modulate gene expression in trans [81–83]. In one of these cases, lncRNA was shown to inhibit DNA methylation of an enhancer to control expression of adjacent genes on the sister chromosome [82]. Moreover, determination of the genomic binding profiles of several lncRNAs has revealed that a single type of lncRNA can associate with up to thousands of genomic loci distributed on different chromosomes to control large-scale gene expression [84–89]. The mechanisms of genome targeting by lncRNAs are currently not well understood, but the spatial organization of the chromosomes in the nucleus may play a role. A process called “proximity transfer” is easily speculated in the cases of local gene expression regulation by nearby eRNAs. Recent reports further suggest that the three-dimensional conformation of genomic DNA may guide lncRNAs to distal but spatially localized target sites [88, 90, 91]. However, spatial organization should not be the only factor to identify correct targets, since several nuclear lncRNAs can regulate transcription when expressed from ectopic loci [81, 83, 87, 92]. Thus, some lncRNAs may form complexes with proteins to bind specific regions for gene regulation. For example, NeST lncRNA interacts with WD repeat domain 5 (WDR5) and activate gene transcription of interferon gamma by altering the active histone mark H3K4me3 [92].

The molecular mechanism that allows targeting of lncRNAs to specific genomic loci are largely unknown, but at least three possible mechanisms were proposed: (1) forming RNA-DNA-DNA triplex, (2) RNA-RNA base-pairing, and (3) indirect recruitment through forming an RNA-protein-DNA complex [93•]. The results of lncRNA-targeting to genomic loci are only now beginning to be investigated. One of the possible outcomes of lncRNA function is the recruitment of protein factors that regulate the chromatin states to targeted loci. In addition, lncRNAs can directly interact with transcriptional machinery. It may act as a decoy or co-regulator for transcription factors or as an inhibitor of Pol II [74].

Conclusion

In the 1960s, the central dogma of molecular biology stated that RNA is a simple messenger between DNA and protein. Studies over the past several decades have revealed the presence of large numbers of non-coding RNAs that regulate diverse biological processes. The two major classes of regulatory RNAs described here, small RNAs and lncRNAs, show both similar and distinct features. They both need to collaborate with effector proteins to regulate gene expression. Similar to small RNAs, it was proposed that some of lncRNAs target genomic regions through sequence homology and regulate gene expression by altering histone marks. The functions of the vast majority of lncRNAs are currently unknown, and future studies should reveal the molecular mechanisms of their action.