RNA in unexpected places: long non-coding RNA functions in diverse cellular contexts

Abstract

The increased application of transcriptome-wide profiling approaches has led to an explosion in the number of documented long non-coding RNAs (lncRNAs). While these new and enigmatic players in the complex transcriptional milieu are encoded by a significant proportion of the genome, their functions are mostly unknown. Early discoveries support a paradigm in which lncRNAs regulate transcription via chromatin modulation, but new functions are steadily emerging. Given the biochemical versatility of RNA, lncRNAs may be used for various tasks, including post-transcriptional regulation, organization of protein complexes, cell-cell signalling and allosteric regulation of proteins.

Dedicated consortiums, such as the ENCODE (Encyclopedia of DNA Elements) project, have markedly expanded our knowledge of what lies in the dark recesses of the genome through their extensive annotation efforts¹. These findings in conjunction with previous studies looking specifically at transcriptional outputs have underscored the pervasiveness with which genomes are transcribed^2,3. An important implication of these findings is that whereas only a minuscule fraction of the human genome encodes proteins, nearly 60% is represented in processed transcripts that seem to lack protein-coding capacity⁴. Together with observations that more sophisticated organisms tend to have more non-coding DNA, this raises the possibility that the barren regions between genes are actually elysian fields’ rich with information⁵. The implications of this are undeniably intriguing, but we are still far from ascribing biological functions to the vast array of non-coding RNA (ncRNA) transcripts. With thousands of documented ncRNAs, pervasive transcription has been described in virtually all eukaryotic organisms^6,7.

For the better part of the past decade, particular attention has focused on the exploding class of transcripts referred to as long non-coding RNAs (lncRNAs), arbitrarily defined as being longer than 200 nucleotides^7,8. Given the prevalence of lncRNA expression, it has been posited that lncRNAs might constitute a significant fraction of the functional output of mammalian genomes^7–9. Such notions have been met with considerable, and quite possibly legitimate, scepticism¹⁰. Indeed, the documentation of pervasive transcription has far outpaced the molecular characterization of the transcripts produced. Although some lncRNA transcripts may represent transcriptional noise, a small but steadily growing list has authentic biological roles^6,11–13. For example, lncRNAs have been implicated in regulating imprinting, dosage compensation, cell cycle regulation, pluripotency, retro-transposon silencing, meiotic entry and telomere length, to name just a few^12,13. Despite these advances, most lncRNAs remain partially uncharacterized. Additionally, there has been a heavy focus so far on the ways that lncRNAs regulate chromatin states, and this emphasis probably underrepresents the full repertoire of lncRNA function. Nonetheless, the rapidly growing lncRNA field is already changing not just our perspective of genomic content, but also the way we think about genes.

In this Review, we focus on the functional attributes of RNA and highlight the unconventional, and perhaps underappreciated, biological contributions of lncRNAs, including the diverse mechanisms through which lncRNAs participate in transcriptional regulation. We touch briefly on the roles of lncRNAs in regulating chromatin states, as this has been explored in several recent reviews (see REFS 8,9,13–15). In addition, we highlight roles beyond transcription whereby lncRNAs function in various cellular contexts, including post-transcriptional regulation, post-translational regulation of protein activity, organization of protein complexes, cell-cell signalling, as well as recombination.

A biochemically versatile polymer

RNA is a versatile molecule making it well suited for a myriad of functions. It is this feature that inspired the ‘RNA world hypothesis’ in which it was postulated that billions of years ago, RNA provided the precursors of all life¹⁶. The multifunctionality of RNA stems from several unique physiochemical properties. First, and perhaps most obvious, is its ability to base pair with other nucleic acids (FIG. 1a). RNA is, therefore, particularly adept at recognizing both RNA and DNA targets through simple one-to-one base pairing interactions. By comparison, proteins such as transcription activator-like effectors (TALEs) and PUF proteins require 100 times more genomic sequence space than an RNA to achieve sequence-specific binding¹⁷. Moreover, because two RNA transcripts can base pair at any point during the life cycle of the target mRNA, regulatory RNAs can influence transcription, processing, editing, translation or degradation of target mRNAs. Second, RNA molecules can fold into intricate three-dimensional structures that provide complex recognition surfaces (FIG. 1b). This structure expands the large variety of molecular targets that RNA can bind with high affinity and specificity. RNA structures can even be selected for in vitro to bind to anything from small molecules to proteins¹⁸. Third, in terms of both expression and structure, RNA is dynamic. More explicitly, because RNA can be rapidly transcribed and degraded, it is well suited for dynamic, transient expression (FIG. 1c). Moreover, without the need to be translated, a regulatory RNA gene could transition faster from being transcriptionally inactive to fully functional. In addition, as conformational changes can be triggered by ligand binding, RNA structures themselves can be very dynamic¹⁹. Fourth, RNA is malleable and therefore provides an excellent platform for evolutionary innovation (FIG. 1d). Specifically, unencumbered by amino acid-coding potential, regulatory RNAs are less restricted in terms of their conservation. As such, RNAs are more tolerant of mutations, which could allow for the rapid evolution of diverse cellular activities. Last, RNA-dependent events can have the capacity to be heritable. This idea is supported by the demonstration of RNA-templated modifications to the genome (FIG. 1e). For example, retroviral genomic integrations as well as the presence of thousands of processed pseudogenes suggest that information housed within mature RNA transcripts can be integrated back into the genome^20,21. These instances of RNA-mediated events that have manifested in genomic change suggest it is possible for other RNA-dependent events to become heritable. Importantly, these defining properties of RNAs raise exciting possibilities as to what roles lncRNAs could have in the cell. Although various functional roles have now been attributed to lncRNAs, it is likely that as we dig deeper into the molecular biology of lncRNAs more functions will emerge.

Figure 1. RNA is a biochemically versatile polymer.

a | RNA is particularly well suited for sequence-specific nucleic acid targeting through base pairing interactions over a short region (for example, eight nucleotides). By contrast, proteins require repeat motifs comprising 35–39 amino acids (105–117 base pairs of genomic sequence) to recognize a single RNA base with specificity. Therefore, to recognize eight nucleotides, 280–312 amino acids (840–936 base pairs of genomic sequence) would be required. Compared to the eight base pairs required for an RNA, protein-based nucleic acid recognition requires substantially more genomic sequence¹⁷.b | RNA can fold into complex three-dimensional structures that can specifically bind various ligands, including small molecules and peptides¹⁸.c | RNA is suitable for transient expression, because a fully functional RNA can be generated immediately following transcription and processing but can also be rapidly degraded. Together, this allows RNA effectors to be produced in quick pulses. Proteins, however, require additional steps, including mRNA export and translation, to produce a functional peptide. Likewise, both the mRNA and the protein need to be degraded to turn off expression, d | RNA is malleable and, therefore, more tolerant of mutations. Although some mutations in protein-coding genes are silent, many are deleterious such as nonsense mutations that generate truncated polypeptides. RNA, however, can tolerate mutations even within the regions responsible for target recognition. e | RNA-dependent events can be heritable. For instance, processed pseudogeneswere once RNA transcripts that have beengenomically integrated. In add it ion, telomerase uses an RNA template to add telomeric repeats to the ends of chromosomes. ORF, open reading frame; Pol II, RNA polymerase II.

lncRNAs as regulators of transcription

The number of lncRNAs with described functions is steadily increasing, and many of these reports revolve around their regulatory capacity. For example, lncRNAs often function as important cis- and trans-acting modulators of protein-coding gene expression⁸. A common theme has emerged in which lncRNAs regulate transcription via chromatin modulation (for reviews, see REFS 8,13,15). lncRNAs across a broad range of eukaryotes affect chromatin context, suggesting that this is a conserved function despite the fact that the transcripts themselves are often not conserved¹². Numerous lncRNAs physically associate with, and potentially target, histone-modifying activities to specific loci^22,23 TABLE 1). lncRNAs such as HOTAIR (HOX transcript antisense RNA), ANRIL (also known as CDKN2B anti-sense RNA 1) and KCNQ1OT1 (KCNQ1 opposite strand or antisense transcript 1) have even been shown to bind more than one histone-modifying complex. As such, a paradigm in which lncRNAs can act as scaffolds that organize the concerted actions of chromatin-modifying complexes spatially and temporally is emerging^15,24–28 (FIG. 2a; TABLE 1). For example, HOTAIR physically associates not only with Polycomb repressive complex 2 (PRC2) but also with LSD1 (Lys-specific demethylase 1)²⁴. PRC2 and LSD1 are responsible for the deposition of the repressive histone mark trimethylated Lys27 of histone H3 (H3K27me3) and removal of active H3K4me2 marks, respectively. Moreover, global analyses suggest that a large number of other lncRNAs can also bind PRC2 and LSD1 (REF. 22). In addition, other lncRNAs have been shown to bind overlapping but distinct combinations of histone-modifying complexes. For example, KCNQ1OT1 binds PRC2 and the methyltransferase G9A (also known as EHMT2), whereas ANRIL binds PRC1 and PRC2 (REFS 26–28) (TABLE 1). HOTAIR and other lncRNAs have, therefore, been proposed to function as scaffolds that coordinate the targeting of distinct repressive histone-modifying complexes to target loci²⁵. However, within this framework, the detailed mechanism of how lncRNAs target specific DNA regions remains unclear.

Table 1.

lncRNA-mediated gene expression control

lncRNA	Function	Mechanism	Refs
Regulation of mRNA transcription
XIST	X inactivation	Chromatin-mediated repression	23,121
HOTA1R	Repression at the HOXD locus	Chromatin-mediated repression	24,122
HOTT1P	Activation at the HOXA locus	Chromatin-mediated activation	123
KCNQIOT1	Imprinting at the KCNQ1 cluster	Chromatin-mediated repression	27
ANR1L	Repression at the lNK4b ARF-INK4a locus	Chromatin-mediated repression	26,28
AIRN	Imprinting at the IGF2R cluster	Chromatin-mediated repression, transcription interference	30
ME4 antisense	Repression of ME4 mRNA	Transcription interference	31
IRT1	Repression of IME1 mRNA	Chromatin-mediated repression	32
GAL10 lncRNA	Repression of GAL1 and CAL10 mRNAs	Chromatin-mediated repression	35
PH084 antisense	Repression of PH084 mRNA	Chromatin-mediated repression	33
ICR1	Repression of FLO11 mRNA	Modulation of transcription factor recruitment	37,41
PWR1	Activation of FLO11 mRNA	Modulation of transcription factor recruitment	37,41
SRG1	Repression of SER3 mRNA	Nucleosome remodelling	38
fbp1 ncRNA	Activation of fbp1	Chromatin remodelling	39
UNOCR	Activation of lysozyme mRNA	Nucleosome remodelling	40
Alu repeat-containing RNA	Transcriptional repression during heat shock	Inhibition of Pol II	47
HSR1	Activation of the HSF1 transcription factor	Allosteric activation together with eEFIA	49
Non-coding DHFR	Transcriptional repression of DHFR	Inhibition of pre-initiation complex formation	48
CAS5	Repression of glucocorticoid receptor-mediated transcription	DNA mimicry	50
EVF2	Transcriptional activation of DLX2 targets, transcriptional repression of MeCP2 targets	Recruitment of DLX2 or MeCP2	51,52
CCND1 promoter RNA	Repression of CCND1 transcription	Allosteric activation of TLS	53
NRON	Repression of NFAT-mediated transcription	Inhibition of transcription factor nucleocytoplasmic shuttling	54
Regulation of mRNA processing
Neuroblastoma MYC (NAT)	Inhibition of neuroblastoma MYC intron 1 splicing	Unknown mechanism involving the inhibition of splicing via RNA-RNA duplex formation	61
Rev-ErbAalpha	Inhibition of the c-ErbAalpha 2 splice isoform	Unknown mechanism involving the inhibition of splicing via RNA-RNA duplex formation	62
ZEB2 (NAT)	Activation of ZEB2 translation	Unknown mechanism involving regulated splicing of an IRES-containing intron	59
MALAT1	Ser/Arg splicing factor regulation	Scaffolding of subnuclear domains	64
Sas10 mRNA 3′UTR	Repression of Rnp4F mRNA	Unknown mechanism involving RNA editing	66
Modulation of mRNA post-transcriptional regulatory pathways
Antisense UCHL1	Upregulation of UCHL1 protein production	SINE2B element-mediated translational upregulation	68
KCS1 antisense	Production of truncated KCS1 protein	Unknown mechanism involving base pairing	69
1/2-sbsRNA1	Down-regulation of SERPINE1 and FLJ21870 mRNAs	Staufen-mediated decaythrough Alu element base pairing	70
BACE1AS	Up-regulation of BACE1	Stabilization of BACE1 mRNA by blocking miRNA-induced repression	71,72
UNCMD1	Control of muscle differentiation through upregulation of MAML1 and MEF2C transcript ion factors	Sequestration of miRNAs	74
HULC	Downregulation of miRNA-mediated repression	Sequestration of miRNAs	75
PTENP1 pseudogene	Upregulation of PTEN	Sequestration of miRNAs	79
IPS1	Downregulation of miRNA-mediated repression	Sequestration of miRNAs	76
CDR1as	Downregulation of miRNA-mediated repression	Sequestration of miRNAs	77,78

1/2-sbsRNA1, half-STAU1-binding site RNA1;AIRN, antisense of IGFR2 non-coding RNA; BACE1AS, beta-site APP-cleaving enzyme 1 antisense; CCND1, cyclin D1; CDR1as, CDR1 antisense; DHFR, dihydrofolate reductase; fbp1, fructose-l,6-bisphosphatase 1; eEF1A, eukaryotic elongation factor 1A; FLO11; GAS5, growth arrest specific 5; HOTAIR, HOX transcript antisense RNA; HOTTIP, HOXA transcript at the distal tip; HOX, homeobox cluster; HSF1, heat shock factor 1; HSR1, heat shock RNA1; HULC, highly upregulated in liver cancer; IGF2R, insulin-like growth factor 2 receptor; IME, inducer of meiosis; IPS1, INDUCED BY PHOSPHATE STARVATION 1; IRES, internal ribosome entry site; IRT1, IME1 regulatory transcript 1; KCNQ1, potassium voltage-gated channel, KQT-like subfamily, member 1; KCNQ1OT1, KCNQ1 opposite strand or antisense transcript 1; LINOCR, LPS-inducible non-coding RNA; lncRNA, long non-coding RNA; MALAT1, metastasis associated lung adenocarcinoma transcript 1; MAML1, mastermind-like 1; MeCP2, methyl CpG binding-protein 2; MEF2C, myocyte enhancer factor 2C; miRNA, microRNA; NAT, natural antisense transcript; ncRNA, non-coding RNA; NFAT, nuclear factor of activated T cells; NRON, non-coding repressor of N FAT; Pol II, RNA polymerase II; PTENP1, phosphatase and tensin homologue; Rnp4F, RNA-binding protein 4F; TLS, translocated in liposarcoma; UCHL1, ubiquitin carboxyl-terminal esterase LI; UTR, untranslated region; XIST, X inactivation-specific transcript; ZEB2, zinc-finger E-box binding homeobox 2.

Figure 2. lncRNAs regulate transcription through several mechanisms.

a-c | Long non-coding RNAs (lncRNAs) can modulate chromatin through transcription-independent (part a) and transcription-dependent mechanisms (parts b and c). lncRNAs can bind one or more chromatin-modifying complexes and target their activities to specific DNA loci (part a). Depending on the nature of the enzymes bound, lncRNA-mediated chromatin modifications can activate or repress gene expression^{22,23,26,27,120}. Chromatin-modifying complexes bound to the RNA polymerase II (Pol II) carboxy-terminal domain (CTD) can modify chromatin during transcription of lncRNAs^33–35 (part b). Transcription of lncRNAs can also result in chromatin remodelling that can either favour or inhibit the binding of regulatory factors (part c). Depending on the nature of the factors that bind during remodelling, gene expression is activated or repressed ^37–40. d–g | lncRNAs can modulate both the general transcription machinery (parts d and e) as well as specific regulatory factors (parts f and g). lncRNAs can bind Pol II directly to inhibit transcription⁴⁷ (part d). Formation of lncRNA-DNA triplex structures can also inhibit the assembly of the pre-initiation complex⁴⁸ (part e). lncRNAs can fold into structures that mimic DNA-binding sites (left) orthat generally inhibit or enhance the activity of specific transcript ion factors (right)^50–53 (part f). lncRNAs can also regulate gene expression by binding specific transport factors to inhibit the nuclear localization of specific transcription factors⁵⁴ (part g).

Additionally, at least in some cases, lncRNA expression may influence epigenetic events through transcription-dependent mechanisms²⁹. The mammalian lncRNA Airn (antisense of Igf2r non-coding RNA) has been suggested to interfere with transcription during its regulation of Igf2r (insulin-like growth factor 2 receptor) because Airn transcription rather than the lncRNA product itself is required for silencing³⁰ (TABLE 1). Similarly, an antisense RNA has also been postulated to repress mRNA expression at the yeast IME4 locus through transcriptional interference³¹ (TABLE 1). In some instances (for example, the GAL10-1, IME1 and PHO84 loci in yeast), movement of the polymerase along the DNA locus can result in the deposition of histone modifications, which in turn repress expression from nearby promoters. This may be one mechanism of transcription-dependent lncRNA regulation ^32–35 (FIG. 2b; TABLE 1). Conversely, in flies non-coding transcription through Polycomb response elements is thought to counteract silencing during the switch from repressed to active states³⁶. Moreover, lncRNA transcription in various organisms can modulate the binding of regulatory factors^37–40 (FIG. 2c; TABLE 1). An interesting example is the pair of cis-acting lncRNAs, ICR1 and PWR1, which dictate the variegated expression of FLO11 mRNA in yeast. Specifically, transcription of ICR1 is thought to reset the FLO11 locus by inhibiting recruitment of the Flo8 or Sfl1 transcription factors, which promote FLO11 mRNA repression and activation, respectively. After this reset, if Fl08 binds it drives expression of PWR1, which in turn interferes with ICR1 expression in cis. ICR1 and PWR1 lncRNAs therefore represent a ‘toggle switch’, resulting in FLO11 mRNA expression when PWR1 is expressed and FLO11 mRNA repression when ICR1 is expressed^37,41. By contrast, the lncRNA SRG1 exerts chromatin regulation by directing a high level of nucleosomes to the region of the phosphoglycerate dehydrogenase SER3 mRNA promoter³⁸. In these particular cis-acting instances, it is often unclear whether the phenotype associated with the locus arises from the lncRNA itself or rather from changes in DNA-protein interactions that arise from polymerase movement.

lncRNAs have now also been implicated in transcriptional upregulation by enhancers^42,43. A specific type of lncRNA, termed enhancer RNA (eRNA), displays enhancer-like activity and upregulates expression via the Mediator complex^43,44 (TABLE 1). As studies suggest that classic enhancer elements are widely expressed, eRNAs may frequently be important for enhancer function at chromatin^42,45,46.

Aside from modulating chromatin, lncRNAs can regulate transcription through additional mechanisms. For example, lncRNAs can influence the transcription machinery directly. During heat shock, lncRNAs generated from Alu SINE elements mediate transcriptional repression through direct contact with RNA polymerase II (Pol II) (FIG. 2d; TABLE 1). This interaction inhibits transcription of specific mRNAs during heat shock⁴⁷. Furthermore, several lncRNAs can regulate the binding and/or activity of transcription factors. At the DHFR (dihydrofolate reductase) locus, expression of an upstream lncRNA impairs the assembly of the transcription pre-initiation complex in trans through the formation of an RNA-DNA triplex structure⁴⁸ (FIG. 2e; TABLE 1).

Moreover, several lncRNAs act directly on specific transcription factors. For instance, during the heat shock response, heat shock factor 1 (HSF1) is activated through the combined actions of a lncRNA, HSR1 (heat shock RNA 1), and a surprising protein interaction partner and co-activator, translation elongation factor eEF1A⁴⁹ (TABLE 1). In another example, the GAS5 (growth arrest specific 5) lncRNA folds into a structure that mimics the DNA-binding site of the glucocorticoid receptor, and the resulting interaction represses GR-mediated transcription⁵⁰ (FIG. 2f; TABLE 1). By contrast, the lncRNA Evf2 (also known as Dlx6os1) can act either as a co-activator or co-repressor, depending on whether it recruits the transcriptional activator DLX2 or the transcriptional repressor MeCP2 (methyl-CpG binding-protein 2) to specific DNA regulatory elements^51,52 (FIG. 2f; TABLE 1). Furthermore, binding of lncRNAs generated from the CCND1 (cyclin D1) promoter allosterically promotes a conformational switch in the TLS (translocated in liposarcoma) protein factor from an inactive to active form. Active TLS inhibits histone acetyltransferases, ultimately leading to repression of CCND1 transcription⁵³ (FIG. 2f; TABLE 1). The lncRNA transcript thus indirectly promotes a repressive chromatin environment.

By contrast, the NRON (non-coding repressor of NFAT) lncRNA indirectly represses transcription by inhibiting nucleocytoplasmic shuttling of the transcription factor NFAT (nuclear factor of activated T cells)⁵⁴. The transport of NFAT, which is imported from the cytoplasm into the nucleus in response to calcium-dependent signals, is inhibited by NRON. NRON binds the transport receptor importin-β, and knockdown of NRON results in nuclear accumulation of NFAT, suggesting that NRON competes with NFAT for importin-β interaction⁵⁴ (FIG. 2g; TABLE 1).

Interestingly, lncRNAs have been indirectly linked to both gene activation and repression through the organization of nuclear subdomains. For instance, the lncRNAs TUG1 (taurine upregulated 1) and MALAT1 (metastasis associated lung adenocarcinoma transcript 1; also known as NEAT2) have been linked to repressive Polycomb group bodies and more active interchromatin granules, respectively (TABLE 1). Both lncRNAs bind Polycomb 2, but TUG1 binds methylated Polycomb 2 and MALAT1 binds the unmethylated protein⁵⁵. The methylation status of Polycomb 2 therefore dictates a switch in both its lncRNA-binding specificity and nuclear subcompartment localization. Importantly, this switch is accompanied by movement of Polycomb 2 target genes between active and repressive nuclear domains and ultimately influences downstream gene expression⁵⁵.

lncRNAs, therefore, can regulate transcription through several mechanisms (FIG. 2). Given the decades of research focused on transcriptional control from a transcription factor-centric point of view, it is interesting to speculate about the purpose of this additional layer of RNA-based regulation. Even at the yeast GAL locus, arguably one of the most extensively studied DNA loci during the past 50 years, a hidden layer of lncRNA-based regulation has now been described^35,56. Indeed there has been a growing interest in such RNA-based control during the past decade⁵⁷, and we and others have speculated that this extra layer of regulation reinforces the control that is imposed by protein factors at a locus. Notably, the impressive diversity of transcriptional regulatory mechanisms discussed here might just be the tip of the iceberg, with additional means of lncRNA-mediated transcriptional regulation to be uncovered in the future.

Regulators of mRNA processing

mRNA transcripts often have a complicated post-transcriptional existence⁵⁸. Immediately in the wake of transcription, nascent pre-mRNAs are spliced and processed into one of potentially many isoforms. Importantly, alternative splicing and editing contribute to increasing gene isoform diversity.

In some cases, lncRNA genes that have an anti-sense orientation to known protein-coding genes, also known as natural antisense transcripts (NATs), can influence how an mRNA arising from the sense strand is processed. For example, NATs influence splicing patterns of mRNAs at the neuroblastoma MYC, c-ErbAalpha (also known as Thra) and ZEB2 (zinc-finger E-box binding homeobox 2) loci in mammalian cells^59–62 (FIG. 3a; TABLE 1). In the case of neuroblastoma MYC and c-ErbAalpha, the NAT and pre-mRNA were suggested to form RNA-RNA duplexes, which then inhibit splicing^61,62. At the ZEB2 locus, NAT expression inhibits splicing of an internal ribosome entry site (IRES)-containing intron. Translation of ZEB2 relies on this IRES, and therefore expression of the NAT indirectly facilitates expression of ZEB2 protein⁵⁹. The mechanism by which NATs influence splicing is unclear, but it has been postulated to involve splice-site masking and a subsequent block in spliceosome recruitment⁶³.

Figure 3. lncRNAs influence mRNA processing and post-transcriptional regulation.

a,b | Long non-coding RNAs (lncRNAs) can modulate mRNA processing. Splicing patterns can be influenced by lncRNAs that associate with the pre-mRNA(part a). For example, splicing of the first intron of neuroblastoma MYC mRNA is prevented by a natural antisense transcript⁶¹. Antisense lncRNAs that associate with an mRNA could direct mRNA editing, perhaps through association of the duplex with ADAR (adenosine deaminase acting on RNA) enzymes that catalyse adenosine to inosine conversion in double-stranded RNA^63,66 (part b). c-f | lncRNAs modulate post-transcriptional regulatory events. lncRNAs containing SIN EB2 repeat elements can upregulate translation through association with the 5′ region of an mRNA⁶⁸ (part c). lncRNAs containing Alu repeat elements associate with the Alu elements in the 3′ untranslated region (UTR)of an mRNA, and this double-stranded structure can direct Staufen-mediated decaythrough a pathwaythat is molecularly similar to nonsense-mediated decay⁷⁰ (part d). lncRNAs can mask miRNA-binding sites on a target mRNA to block miRNA-induced silencing through the RNA-induced silencing complex (RISC)⁷² (part e). Linear or circular lncRNAs can function as miRNA decoys to sequester miRNAs from their target mRNAs^74,75 (part f).

The MALAT1 lncRNA also affects splicing, but through a more indirect mechanism. This lncRNA, which is retained in the nucleus and associates with interchromatin granules, has been implicated in alternative splicing through the modulation of active Ser/Arg splicing factors, named after characteristic Ser- and Arg- rich domains. Ser/Arg proteins are important regulators of alternative splicing, and MALAT1 interacts with, and influences the nuclear distribution and levels of, phosphorylated Ser/Arg proteins. Importantly, depletion of MALAT1 changes the alternative splicing patterns of the pre-mRNAs that they target⁶⁴.

In addition to modulating splicing, overlapping antisense lncRNAs have in principle the potential to direct mRNA editing (FIG. 3b). During editing, ADAR (adenosine deaminase acting on RNA) enzymes catalyse adenosine to inosine conversion in double-stranded RNA, and this conversion can influence RNA structure, splicing patterns, coding potential and targeting by microRNAs (miRNAs)⁶⁵. In Drosophila melanogaster, editing of Rnp4F (RNA-binding protein 4F) mRNA depends on developmentally restricted expression of a long isoform of the partially overlapping Sas10 transcript (TABLE 1). Although, in this case, an mRNA isoform with an extended 3′ untranslated region (UTR) provides the source of an antisense RNA, lncRNAs could act in a similar manner to direct editing⁶⁶. Given that many, if not most, mammalian genomic loci produce multiple RNA transcripts from both strands with at least partial overlap, the potential for double-stranded RNA editing substrates is extensive⁶⁷. With many of these pervasive transcripts anticipated to be lncRNAs, lncRNAs are likely to help diversify the transcriptome and proteome through control of RNA editing.

Modulators of post-transcriptional control

Following processing and nuclear export, mRNAs are subjected to various post-transcriptional regulatory pathways that modulate gene expression levels. For example, the overall level of protein produced from an mRNA depends on translation efficiency, mRNA turnover kinetics and small RNA-mediated translational repression. A growing number of reports implicate lncRNAs in control of these post-transcriptional events.

Translation control

The mouse Uchl1AS lncRNA produced from the Uchl1 (ubiquitin carboxyl-terminal esterase L1) locus was shown to upregulate translation of Uchl1 mRNA through a repeat element (FIG. 3c; TABLE 1). In this instance, sense and antisense transcripts are oriented in a 5′ head-to-head fashion such that the mature lncRNA contains a 73-nucleotide motif complementary to the 5′ end of the Uchl1 mRNA. This sequence-specific interaction serves to position the effector domain, which is contained in the non-overlapping 3′ region of Uchl1AS and consists of a SINEB2 repeat element that upregulates protein expression without changing Uchl1 mRNA levels. Bioinformatic analysis has identified 59 other cDNAs with similar antisense orientations and SINEB2 elements, suggesting that this regulatory mechanism might be used at other loci⁶⁸. lncRNA-mediated translational regulation has also been documented in yeast, in which an antisense KCS1 lncRNA was suggested to regulate translation of the inositol pyrophosphate synthase KCS1 mRNA expressed from the same locus. Through an unknown mechanism, which is thought to involve base pairing interactions between the antisense and sense RNAs, expression of KCS1 antisense RNA results in the production of truncated KCS1 protein⁶⁹ (TABLE 1).

mRNA stability control

lncRNAs have also been implicated in both positive and negative regulation of mRNA stability. For instance, Alu repeat-containing lncRNAs are involved in targeting mRNA transcripts for Staufen-mediated decay (SMD)⁷⁰. SMD is induced by Staufen 1 (STAU1) binding to a double-stranded structure in mRNA 3′ UTRs. Through imperfect base pairing interactions with Alu elements in the 3′ UTR, Alu repeat-containing lncRNAs create STAU1-binding sites that trans-activate SMD and destabilize the target mRNA (FIG. 3d; TABLE 1).

By contrast, BACE1AS, an antisense lncRNA that arises from the BACE1 (beta-site APP-cleaving enzyme 1) locus, increases stability of BACE1 mRNA⁷¹ (TABLE 1). BACE1AS and BACE1 mRNA form an RNA-RNA duplex, which has been suggested to stabilize the mRNA by abrogation of miRNA-induced repression. More specifically, the antisense transcript and miR-485-5p compete for binding to the same region in the BACE1 mRNA^71,72 (FIG. 3e; TABLE 1). BACE1 mRNA encodes β-secretase, the rate-limiting enzyme in amyloid-β synthesis. Regulation of BACE1 expression, therefore, has important implications in Alzheimer’s disease. Intriguingly, BACE1AS levels are increased in the brains of patients with Alzheimer’s disease, which perhaps suggests that the regulation of this lncRNA might be relevant in this condition⁷¹.

miRNA sponges

Aside from competing with small RNAs for binding sites on target mRNAs, lncRNAs also can act as decoys to attenuate small RNA regulation, for example through sequestration of proteins or RNA-dependent effectors. The competing endogeneous RNA (ceRNA) hypothesis is based on this idea. It postulates that a widespread network of crosstalk exists between coding and non-coding RNAs that manifests through competition for miRNA binding⁷³. Examples of potential ceRNAs include LINCMD1, HULC (highly upregulated in liver cancer), PTENP1 (PTEN pseudogene 1), IPS1 (INDUCED BY PHOSPHATE STARVATION 1) and CDR1as (CDR1 antisense; also known as ciRS-7)^74–78 (TABLE 1). Specifically, the muscle-specific lncRNA LINCMD1 regulates muscle differentiation by binding and sequestering miR-133 and miR-135 (REF. 74). Normally, these miRNAs negatively regulate expression of the MAML1 (mastermind-like 1) and MEF2C (myocyte enhancer factor 2C) transcription factors, which drive muscle-specific gene expression. So, by sequestering these miRNAs, LINCMD1 indirectly activates MAML1 and MEF2C⁷⁴. Similarly, the HULC lncRNA has been suggested to act as a ‘sponge’ that inhibits miR-372 by sequestering it away from potential mRNA targets⁷⁵. This regulatory principle is shared with pseudogenes, which can also act as miRNA decoys to upregulate expression of their cognate genes. This has been shown, for example, in the case of the pseudogene PTENP1 (REF. 79).

The Arabidopsis thaliana lncRNA IPS1 also sequesters miR-399 away from its target mRNAs⁷⁶. Whereas most miRNAs in plants have perfect complementarity to their targets, which results in mRNA cleavage, IPS1 contains an imperfect binding site for miR-399. Thus, miR-399 binding to IPS1 does not result in its cleavage but instead limits the levels of miR-399 available for other targets. This ability to evade cleavage is an important aspect of IPS1 regulation, because mutant IPS1 with perfect complementarity to miR-399 no longer regulates miR-399 (REF. 76).

More recently, another example of lncRNA-based miRNA sponges has been described, but these RNAs are unique in that they have a circular structure^77,78. In humans, the highly stable circular RNA (circRNA) CDR1as has numerous miR-7-binding sites^77,78 (FIG. 3f; TABLE 1). Importantly, a similar CDR1as genomic locus can be found across eutherian mammals, suggesting that, unlike many other lncRNAs, this RNA might be conserved⁷⁷. Moreover, bioinformatic analyses indicate that there may be thousands of expressed circRNAs across a broad range of multicellular eukaryotes⁷⁸.

lncRNAs can, therefore, modulate gene expression by diverse post-transcriptional regulatory pathways (FIG. 3c–f; TABLE 1). Whereas some lncRNAs seem to influence translation, others operate at the RNA level. As more and more lncRNAs are functionally characterized, we will probably see additional examples of post-transcriptional regulation by lncRNAs.

Regulators of protein activity

In addition to lncRNA-mediated modulation of gene expression events through effects on mRNAs, lncRNAs can also act at the protein level. Indeed, some of the same lncRNAs that affect mRNAs, such as GAS5, EVF2 and CCND1, alter the activity of transcription factors (TABLE 2). However, the ability of lncRNAs to bind and modulate protein activity extends beyond factors involved in transcription.

Table 2.

lncRNA-mediated regulation of proteins

lncRNA	Function	Mechanism	Refs
Regulation of protein activity
CAS5	Repression of glucocorticoid receptor-mediated transcription	DNA mimicry	50
EVF2	Transcriptional activation of DLX2 targets	Activation of DLX2	51,52
CCND1 promoter RNA	Repression of CCND1 transcription	Allosteric activation of TLS	53
NRON	Repression of NFAT-med iated transcription	Inhibition of transcription factor nucleocytoplasmic shuttling	54
15q11-q13sno-lncRNA	Regulation of alternative splicing	Inhibition of FOX2 function	80
rncs-1	Inhibition of Dicer-mediated repression	Sequestration of Dicer or accessory double-stranded RNA-binding proteins	81
sfRNA	Stabilization of viral and host mRNAs	Inhibition of XRNl-mediated mRNA degradation	82,83
gadd7	Inhibition of TDP43-mediated regulatory events	Sequestration of TDP43	84
Organization of protein complexes
HOTA1R	Repression at the HOXD locus	Recruitment of PRC2 and LSD1	24
KCNQ1OT1	Imprinting at the KCNQ1 cluster	Recruitment of PRC2 and G9A	27
ANR1L	Repression at the INK4b–ARF–INK4a locus	Recruitment of PRC1 and PRC2	26,28
TERC	Add it ion of telomeric repeats to the ends of chromosomes	Organizational scaffold fortelomerase components and template for repeat addition	90
SRP RNA	Directing of proteinstothe ER	Organizational scaffold for SRP components	91
NEAT1	Assembly of paraspeckles	Nucleation of subnucleardomains	95–97

CCND1, cyclin Dl; ER, endoplasmic reticulum; GAS5, growth arrest specific 5; HOTAIR, HOX transcript antisense RNA; HOXD, homeobox D cluster; KCNQ1, potassium voltage-gated channel, KQT-like subfamily, member 1; KCNQ1OT1, KCNQ1 opposite strand or antisense transcript 1; LSD1, Lys-specific demethylase 1; NFAT, nuclear factor of activated T cells; NRON, non-coding repressor of NFAT; PRC, Polycomb repressive complex; sfRNA, subgenomic flavivirus RNA; sno-lncRNA, small nucleolar long non-coding RNA; SRP, signal recognition particle; TDP43, TAR DNA-binding protein 43; TERC, telomerase RNA component; TLS, translesion DNA synthesis; XRN1, 5′ to 3′exoribonuclease 1.

For example, a new class of lncRNAs flanked by small nucleolar RNA (snoRNA) sequences, termed sno-lncRNAs, influence splicing patterns via physical interactions with an alternative splicing regulator in human cell lines⁸⁰. These sno-lncRNAs are derived from introns and are nuclear-enriched. A particularly abundant member of the sno-lncRNA family, generated from the 15q11-q13 chromosomal region, directly associates with the FOX2 alternative splicing factor (FIG. 4a; TABLE 2). Importantly, sno-lncRNA knockdown results in changes in FOX2-regulated splicing, and it has been speculated that the sno-lncRNA might inhibit FOX2 function via a sequestration mechanism⁸⁰. Similarly, the Caenorhabditis elegans lncRNA rncs-1 has been suggested to influence the processing of small RNAs via Dicer inhibition⁸¹. The rncs-1 lncRNA forms an extensive double-stranded helix, but is not cleaved by Dicer due to inhibitory secondary structures flanking this helix (FIG. 4b; TABLE 2). It has been suggested that rncs-1 competitively binds either Dicer or accessory double-stranded RNA-binding proteins to preclude processing of small RNAs from double-stranded RNA precursors⁸¹.

Figure 4. lncRNAsare involved in various cellular contexts.

Long non-coding RNAs(lncRNAs) modulate protein activity by post-translational mechanisms (parts a-c). a | Small nucleolar lncRNAs(sno-lncRNAs) generated from the 15qll-ql3 locus bind and modulate the activity of the FOX2 alternative splicing factor, and this can inhibit FOX2-mediated splicing⁸⁰.b | The highly structured rncs-1 lncRNA binds Dicer to inhibit the processing of small RNAs⁸¹c | The gadd 7 IncRN A binds and modulates the ability of TDP43 (TAR DNA-binding protein 43) to target and process specific mRNAs⁸⁴.d | lncRNAscan act as scaffolds to organize several complexes²⁴.e | As the cargo of exosomesthat mediate transfer of material between cells, exosomal shuttle RNAs(exRNAs)may act as signalling molecules during cell-cell communication; exosomal cargo includes mRN As, microRNAs (miRNAs) and lncRNAs¹⁰².f | lncRNAs expressed from the switch region of genes encoding antibodies form R-loops to direct class switch recombinationvia activation-induced deaminase (AID) recruitment¹¹¹

Flaviviruses, such as West Nile virus, also produce a highly structured lncRNA termed subgenomic flavivirus RNA (sfRNA), which is resistant to destruction by host nucleases. sfRNA is essential for pathogenicity and is thought to stall the host 5′ to 3′ exoribonuclease, XRN1, during viral RNA genome degradation⁸². The inhibition of XRN1 induced by sfRNA is even strong enough to stabilize host cellular mRNAs⁸³ (TABLE 2). Although this is an example of a viral lncRNA that inhibits a host cellular enzyme, it illustrates that structured lncRNAs have the capacity to inhibit wide-ranging enzymatic activities.

A ultraviolet (UV) light-induced lncRNA, gadd7, has also been shown to influence cellular mRNA stability⁸⁴. This lncRNA, however, does so by modulating the activity of the RNA-binding protein TDP43 (TAR DNA-binding protein 43). TDP43 has been implicated in pre-mRNA splicing as well as mRNA transport, translation and stability^85–88. It binds 3′ UTR elements in a large number of genes, and this binding can result in either the stabilization or destabilization of mRNA targets^84,86–88. The association of gadd7 with TDP43 impairs TDP43 binding to several of its targets (FIG. 4c; TABLE 2). For example, by preventing TDP43 association with cyclin-dependent kinase 6 (CDK6) mRNA, gadd7 alters the role of TDP43 in modulating mRNA stability⁸⁴. Interestingly gadd7 is not the only lncRNA that TDP43 binds. TDP43 also associates the MALAT1 and NEAT1 (also known as Men ε/β) lncRNAs⁸⁹. As both MALAT1 and TDP43 are implicated in control of alternative splicing, it will be interesting to further explore this interaction in future studies.

Scaffolds for higher-order complexes

RNA transcripts associate with proteins to form ribonucleoprotein particles (RNPs). Compared with other RNAs such as snRNAs and rRNAs, we know very little about the composition of RNPs formed by lncRNAs. Some specific lncRNA-protein interactions have been characterized, but the lncRNA interaction network in cells is likely to be more complicated than single lncRNAs interacting with single proteins. Indeed there are indications that lncRNAs can act as scaffolds to organize higher-order complexes.

Some of the lncRNAs involved in chromatin-dependent events (such as HOTAIR, KCNQ1OT1 and ANRIL) have been suggested to act as scaffolds that coordinate the activities of histone-modifying complexes^15,25 (FIG. 2; TABLE 2). There are also notable examples of classic ncR-NAs such as the RNA component of telomerase (TERC) and signal recognition particle (SRP) RNA that can act as scaffolds at telomeres and on translating ribosomes during protein targeting to the endoplasmic reticulum (ER), respectively^90,91 (TABLE 2). Although the SRP and TERC ncRNAs are not generally considered to be lncRNAs, they demonstrate that RNA is particularly adept as a scaffold and that many lncRNAs could function as scaffolds in diverse contexts.

The telomerase RNP complex is responsible for adding telomeric repeats to chromosomal ends and thereby maintains their length in replicating cells. The RNA component of telomerase is not only responsible for templating the addition of telomeric repeats but also provides a scaffold that organizes telomeric regulatory proteins⁹⁰ (FIG. 4d; TABLE 2). Interestingly, other lncRNAs generated from telomeric repeats, termed TERRA, have a distinct role in telomere biology. Rather than extending telomere ends, these lncRNAs promote telomere shortening via exonuclease 1-dependent resection of chromosome ends⁹². lncRNA-mediated events thus serve critical functions in telomere homeostasis.

The SRP is a highly conserved RNP complex, consisting of the SRP RNA transcript and six proteins, which directs proteins to the ER. SRP co-translationally binds the signal sequence in nascent peptides, stalls translational elongation and then targets the ribosome-nascent chain complex to the SRP receptor on the ER. Whereas specific protein domains in SRP mediate peptide recognition and arrest of translational elongation, SRP RNA provides a scaffold to organize and coordinate distally occurring events at the sites of peptide exit and elongation factor binding on the ribosome⁹¹ (TABLE 2).

In addition to serving as scaffolds for specific multi-protein complexes, lncRNAs have been implicated in nuclear organization through the scaffolding of sub-nuclear domains⁹³. Indeed, RNA, both coding and non-coding, has been implicated in the nucleation of histone locus bodies, interchromatin granules, paraspeckles and nuclear stress bodies⁹⁴. Perhaps the best-studied lncRNA of this type is NEAT1, which is important for the de novo assembly of paraspeckles (subnuclear domains that may mediate retention of hyperedited mRNAs in the nucleus)^95,96. Interestingly, the nascent lncRNA is important for this because ongoing NEAT1 lncRNA transcription is required for paraspeckle maintenance⁹⁷.

It is enticing to speculate that other uncharacterized lncRNAs may serve as scaffolds to organize and hold together other higher-order complexes. Imagine what might have been missed through the routine treatment of protein preparations with nuclease to remove RNA contaminates before purification and identification of interacting partners. Perhaps lncRNAs could even hold together enzymes involved in fundamental metabolic processes such as glycolysis or the Krebs cycle. Indeed, the orchestration of electron transport factors on the inner lumen of the mitochondria illustrates that spatial arrangements of enzymes can partly facilitate the catalysis of reactions by overcoming the limits imposed by diffusion. Similarly to cell membranes, lncRNA might also help facilitate this purpose by bringing enzymes closer together. Perhaps this is not such a far stretch, as metabolic enzymes such as aconitase and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) are known to have RNA-binding activity^98–100.

Signalling molecules

RNA can be transferred between cells in small vesicles known as exosomes^101–103. Not to be confused with the molecular machine with the same name that mediates RNA degradation, these exosomes are membrane-bound vesicles of endosomal origin that are released from various cell types in mammals. Upon fusion with another cell, both their RNA and protein cargo can be transferred¹⁰². The RNAs that have been found in exosomes, termed exosomal shuttle RNAs (exRNAs), do not simply reflect the RNA composition of the cell of origin, suggesting that there may be selective loading of RNAs into exosomes¹⁰⁴. Because transmitted RNAs can function in the recipient cell, it has been suggested that exRNAs might be used as a signal to change gene expression patterns in the recipient cell^101–104. Although exosomes contain large amounts of exRNAs, so far miRNAs and mRNAs have been a primary focus of study^103,105,106 (FIG. 4e). However, a recent report characterizing the full complement of human plasma-derived exRNA indicates that lncRNAs are indeed present in exosomes¹⁰⁶. The presence of lncRNAs certainly raises the exciting possibility that they might provide signals that impinge on various gene expression events.

Vehicles for increasing genetic diversity

Genetic diversity is crucial for the survival of a species and, within individuals, genetic innovation is of paramount importance to the adaptive immune system. Diversity in developing lymphocytes is achieved through genomic rearrangements in the form of class switch recombination (CSR; also known as isotype switching) and V(D)J recombination events. Interestingly, non-coding transcription has been implicated in both forms of recombination^107,108. Through CSR, the constant regions of antibodies are exchanged. As such, this process increases the range of effectors that a particular antibody can interface with and, therefore, increases its versatility^109,110. During CSR, the non-coding switch region (S region) is transcribed, and the lncRNAs generated from this S region are likely to be important guides in dictating the locations of recombination. The nascent lncRNA forms an RNA-DNA hybrid or R-loop structure, which displaces one strand of DNA and this, in turn, is thought to facilitate targeting of activation-induced deaminase (AID), the enzyme that initiates CSR^107,111,112 (FIG. 4f). Transcription through non-coding regions also has a role in V(D)J recombination, the genomic rearrangement that generates diversity in antigen receptor-binding pockets in antibodies^113,114. During V(D)J recombination, chromatin accessibility has been suggested to affect recombinase targeting¹¹⁴, and production of non-coding transcripts from the mouse Tcra (T cell receptor alpha chain) locus can trigger changes in chromatin structure that then influence recombination¹⁰⁸.

It is tempting to postulate that non-coding transcription might also increase genetic diversity outside the immune system. During meiosis, sites of recombination are not distributed randomly but tend to occur in discrete locations¹¹⁵. Intriguingly, in fission yeast these hot spots correlate with lncRNA-expressing loci¹¹⁶. How exactly lncRNAs contribute to recombination-site selection is currently unclear, but one possibility is that this could involve similar mechanisms to those used during recombination in lymphocytes.

Conclusions and perspectives

Amidst the exciting discoveries being made during this time of genome exploration, RNA is taking centre stage. The burgeoning lncRNA field has a strong part in this, and lncRNAs have now been demonstrated to regulate all aspects of gene expression, including transcription (FIG. 2), processing and post-transcriptional control pathways (FIG. 3). Likewise, lncRNAs have also been shown to regulate protein function and organize multiprotein complex assembly. Now with hints that lncRNAs might participate in cell-cell communication and recombination, the possible reach of lncRNA functions seems endless (FIG. 4). With most biologists trained to dissect function based on a protein-centric view of the cell, the task of functionally characterizing this new RNA world seems daunting. It is important, therefore, as we move forward, to utilize and develop more functional characterization methods that play to the strengths of RNA. Indeed technical advances are already underway that have the promise of greatly improving the in vivo functional characterization of lncRNAs. For instance, techniques to probe RNA chemical structure have often been limited to in vitro studies, but recently developed chemical probes that can be used in living cells have the promise of greatly improving our ability to determine in vivo RNA structures¹¹⁷. Additionally, the application of high-throughput microfluidics-based screening technologies towards the functional analysis of pre-programmed RNA libraries has the potential to streamline the process of discovering functional motifs within lncRNAs¹¹⁸. Last, recently developed RNA aptamers such as Spinach have adapted GFP tagging for RNA transcripts to allow RNA fusions to be imaged in living cells¹¹⁹.

Much like the multifunctional nature of a Swiss army knife, RNA has the biochemical diversity to function in diverse contexts. It may, however, take some time to determine in which contexts the cell uses some of the more exotic RNA tools. With eyes open to new possibilities, undoubtedly we will be surprised by what we find.

Glossary

Chromatin

Condensed DNA structure that is associated with histone proteins and other l)NA-binding proteins

Transcription activator-like effectors

TALEs). Naturally found in some bacteria, TALEs are proteins that bind DNA through repeat domains, and their code for sequence specificity has been elucidated allowing sequence specific TALEs to be engineered

PUF proteins

A family of sequence-specific RNA-binding proteins, which bind 3′ untranslated regions within mRNAs to repress target mRNA translation

Pseudogenes

Dysfunctional relatives of normal genes thought to arise from duplication or retrotransposition

Chromatin-modifying complexes

Protein complexes that catalyse the covalent chemical modification of chromatin

Enhancers

Short regions of DNA that enhance the expression of genes at varying distances. Effects can be mediated by transcription factor binding to these sites

Alu SINE elements

Highly abundant retrotransposons of the short interspersed nuclear elements (SINE) family

Nuclear subdomains

Non-membrane bound subcompartmentsof eukaryotic nuclei where factors with similar functions colocalize

GAL locus

An inducible locus in yeast comprising the GAL1 and GAL10 genes, which are required for galactose metabolism

Alternative splicing

An mRNA processing step whereby exons can be alternatively used to generate different isoforms of the same gene

Internal ribosome entry sites

IRESs). Nucleotide sequence that allows cap-independent translation initiation within the middle of an mRNA transcript

Spliceosome

The macromolecular machinery (composed of both RNA and protein) responsible for pre-mRNA splicing

miRNAs

(miRNAs). A class of short (~ 23 nucleotides) endogenous non-coding RNAs that control gene expression post-transcriptionally through either translational repression or mRNA degradation

Competing endogeneous RNA

(ceRNA). RNA transcripts (both coding and non-coding), which share microRNA-targeting sites and thus regulate each other via direct competition for microRNA binding

Circular RNA

(circRNA). As opposed to conventional linear RNA transcripts, the 5′ and 3′ ends of circular RNAs are covalently linked together.

Small nucleolar RNA

(snoRNA). A class of small RNA molecules that guide the chemical modification of other RNA transcripts

sno-lncRNAs

(small nucleolar long non-coding RNAs). Class of intron-derived long non-coding RNA flanked by snoRNA ends

Dicer

An RNase III family endoribonuclease responsible for the processing of pre-miRNAs into short double-stranded RNAs to be loaded into the RNA-induced silencing (RISC) complex

Adaptive immune system

A system of specialized cells that create immunological memory via specific antibodies after an initial response to a pathogen