Natural antisense transcripts

Abstract

Recent years have seen the increasing understanding of the crucial role of RNA in the functioning of the eukaryotic genome. These discoveries, fueled by the achievements of the FANTOM, and later GENCODE and ENCODE consortia, led to the recognition of the important regulatory roles of natural antisense transcripts (NATs) arising from what was previously thought to be ‘junk DNA’. Roughly defined as non-coding regulatory RNA transcribed from the opposite strand of a coding gene locus, NATs are proving to be a heterogeneous group with high potential for therapeutic application. Here, we attempt to summarize the rapidly growing knowledge about this important non-coding RNA subclass.

INTRODUCTION

Although DNA and RNA were first discovered at the same time, it took >150 years for the RNA to emerge from the shadow of its longer sibling. Only now the full structural and regulatory role of RNA is becoming apparent, leading to the emerging understanding of the eukaryotic genome as an RNA machine (1). Efforts of multiple scientists, fueled by the achievements of the FANTOM, and later GENCODE and ENCODE consortia, expanded the scope of RNA functions from a single role of carrying information from DNA to ribosomes to a wide spectrum of regulatory and structural roles in gene expression, genomic imprinting and chromatin rearrangement (2–6). The journey of the non-coding RNA (ncRNA) from ‘genomic junk’ to biological prominence has arguably been long and tortuous (1,7).

As the identity and functions of non-coding transcripts are still in the early stages of discovery, so is their classification. A commonly used grouping based mostly on the order of discovery, divides non-coding transcripts into rRNA, tRNAs, snRNAs, snoRNAs, short (miRNA, piRNA) and long non-coding RNAs (lncRNA). However, as structures and functions of multiple non-coding transcripts become better characterized, it is becoming clear that the classification based on this grouping is not very useful. In fact, it might even impede progress by artificially separating similar entities into different classes and encouraging scientists to overlook, for example, the dual lncRNA and miRNA aspects of some ncRNAs. Even the division into coding and ncRNAs is gradually becoming less obvious, as cases are discovered where a coding gene mRNA functions as a regulatory molecule in its own right or an ncRNA is found to produce a protein (8,9). The current neat picture of the RNA world could be threatened even further by the recently proposed model where transcripts are generated constantly at low levels from different points of the genome to allow transcription surveillance machinery to sample the local context and engage in appropriate action.

Despite the lack of comprehensive classification, we have to limit our review to a subgroup of RNAs due to space constraints. We will focus on natural antisense transcripts (NATs) which are relatively less studied than ribosomal/tRNAs, short ncRNAs and even another subgroup of lncRNA, long-intergenic non-coding RNAs (lincRNAs) (10,11). Furthermore, due to their highly gene locus-specific effects, NATs may provide a unique entry point for therapeutic intervention in targeted gene upregulation (12). For the purpose of this review, we will define NATs as ncRNAs transcribed from the opposite strand of a coding gene and capable of regulating the expression of their sense gene pair or of several related genes. As will be described, this group is highly heterogeneous. NATs may originate from all parts of a given protein coding locus. The effect of NATs on their partner coding genes could include suppression, activation or homeostatic adjustment, and the mechanisms may be as different as recruitment of epigenetic modifier enzymes, ncRNA/mRNA pairing or stabilization of long-range chromosomal interactions. Below we will review examples of these cases.

NATS: STRUCTURAL AND FUNCTIONAL DIVERSITY

NAT sequences

For the NAT subclass as a whole, no clear sequence motifs have yet been identified. Furthermore, NAT sequences are poorly conserved among species (13). It has been proposed that it is secondary and/or tertiary structures that are evolutionally conserved and are essential for NAT and other lncRNA functions (14). It is also possible that closer homologies will emerge when known RNA transcripts are grouped in a different way. For example, when a group of 141 intronic regions and 74 intergenic transcripts obtained by deep sequencing of intronic and intergenic chromatin-associated RNAs was analyzed, it showed significant conservation across 44 species of mammals (15).

Interestingly, common NAT sequences can be contributed by the transposable elements. Some of the active human endogenous retroviruses (HERVs) are transcribed from the coding gene loci in antisense direction and regulate these genes in a discordant fashion, thus qualifying as NATs (e.g. HERV-Ec1 and HERV-Ec6 proviruses located in PLA2G4A and RNGTT genes) (16). HERV promoters frequently originate expression of new isoforms of both coding and non-coding genes (17,18). Some NATs contain short interspersed nuclear element B2 (SINEB2), for example antisense Uchl1 which includes a part overlapping a 5′ sequence of Uchl1 and an embedded inverted SINEB2 (19). Tspo NAT was created by extension of the SINEB2 transcript to exon 3 of the Tspo gene (20).

Some of the pseudogenes, frequently generated through inverted gene duplications, may give rise to lncRNA transcripts with regulatory functions (21).

The search for common NAT motifs is further complicated by the fact that RNAs can combine coding and non-coding roles. For example, steroid receptor RNA activator thought to be non-coding was found to produce a highly conserved small protein (9).

Localization of NATs relative to coding gene sequences

Known NATs overlap introns, exons, promoters, enhancers, UTRs and flanking sequences of the partner coding genes, in all combinations. Head-to-head, tail-to-tail and fully overlapped with coding gene NAT configurations have also been observed.

Several groups have attempted associating relative position of lncRNA with its function, as deduced from correlation in temporal profiles and expression levels of their partner protein-coding genes. Batagov et al. (22) found weak positive correlation in expression for coding/non-coding gene pairs at bidirectional promoters and for sense–antisense transcript pairs during a 120-h time course of differentiation of human neuroblastoma SH-SY5Y cells into neurons after treatment with retinoic acid. In contrast, lncRNAs located in the introns and downstream of the protein-coding genes showed negative correlation. Using directional RNA-seq data from of mouse and chimpanzee tissues, Uesaka et al. (23) have noted that loci with tissue-specific expression frequently contain lncRNAs overlapping the coding gene promoter (termed promoter-associated ncRNAs or pancRNAs), whereas constitutively expressed genes usually did not have pancRNAs. Furthermore, expression of pancRNAs and their coding partners was positively correlated.

Head-to-head configuration of coding/non-coding gene pairs has been closely investigated in recent years (24,25). It has been shown that a significant fraction of the transcription start sites of protein-coding genes may be generating bidirectional transcription (23). More than 60% of the lncRNA transcripts in human and murine embryonic stem cells (ESCs) originate from bidirectional promoters (26). It has been proposed that transcription of the ncRNA from these sites may be suppressed by a gene-loop configuration formed by Ssu72 protein, which interacts with both the promoter and terminator of the active protein-coding genes and restricts divergent transcription of ncRNAs (27).

A subset of enhancers has been shown to produce lncRNAs termed eRNAs, which may form a separate NAT subgroup. Such eRNA-producing enhancer at Nanog locus exhibited decreased DNA methylation, elevated levels of the active mark H3K27Ac and DNA hydroxylase Tet1. Binding of Sall4 and Tet family proteins was necessary for eRNA transcription at this locus (28).

Mechanisms of NAT-mediated regulation

Multiple mechanisms have been proposed for NAT-dependent regulation of coding protein expression. It is likely that the mechanism of action would constitute the most meaningful basis for NAT classification. However, the currently available data are too sketchy and only a preliminary grouping by mechanism of action type can be suggested. It is possible that as more information becomes available more NATs will be found to participate simultaneously in several types of regulation described below.

Interaction with protein complexes: decoy, scaffolding and tethering mechanisms

Different authors describe lncRNAs interacting with protein complexes as decoys, scaffolds or tethers/guides. Typically, decoys interact with one partner, whereas scaffolds and tethers/guides bridge groups of targets. In practice, however, it is hard to draw a clear line between these three groups. Both tethers and scaffolds can display some aspects of decoy mechanism. The overlap is even bigger between scaffolding and tethering groups. Below, we will review some of the cases keeping the decoy, scaffold or tether/guide description given by the authors for convenience.

The decoy mechanism mostly involves different lncRNAs competing for the binding sites with other molecules (Fig. 1). A classic case of decoy mechanism has been described by Huang et al. (29). hnRNP I enhances the translation of p27 (Kip1) through interaction with its 5′-untranslated region. lncRNA UCA1 displaces p27 RNA from the hnRNP I complex which leads to a decrease in p27 protein level. An interesting variation on the decoy mechanism is described for lncRNA cardiac hypertrophy-related factor that downregulates miR-489 expression levels by directly binding to and sequestering miR-489 in a model of AngII-induced cardiac hypertrophy (30).

Figure 1.

Proposed mechanisms of NAT-mediated regulation. (A) Interaction with protein complexes: decoy mechanism. NAT binds to a protein complex, which either prevents RNApol binding to the coding gene's promoter and thus inhibits transcription or interferes with mRNA translation (not shown) (29,30). (B) Interaction with protein complexes: tethering mechanism. NAT is transcribed from the opposite strand of the protein-coding locus. The NAT-mediated tethering can occur by the nascent NAT at the time of NAT transcription or after NAT transcription has been completed, by pairing with DNA or nascent mRNA sequences. NAT then binds a protein complex (e.g. PRC2) thus tethering it to the coding gene locus, and/or scaffolding several proteins at the promoter site. PRC2 catalyzes the trimethylation of histone H3 at lysine 27 (H3K27me3), which is recognized by PRC1. PRC1 then catalyzes the monoubiquitylation of histone H2A, which contributes to chromatin compaction and repression of the target locus (2,31). (C) Generation of endogenous siRNAs and miRNA. NAT forms internal hairpins or duplexes with mRNA in the areas of homology. The double-stranded RNA stretches are trimmed by Dicer to form short RNA duplexes, which are then bound by the RISC complex and used as a template for recognition of mRNA. Captured mRNA is then cleaved by the RISC complex, reducing protein expression (32,33).

Tethering of epigenetic modulators has been first described for lncRNAs in the context of imprinting. These interactions for the most part involve polycomb group protein complexes PRC1 (polycomb repressive complex 1) and PRC2 tethering to defined gene loci. This group of interactions has recently been expanded to include developmentally and environmentally driven epigenetic modification of the non-imprinted loci (reviewed in 34,35) (Fig. 1). PRC2 catalyzes the di- and trimethylation of histone H3 at lysine 27 (H3K27me2 or H3K27me3), which is recognized by PRC1. PRC1 then catalyzes the monoubiquitylation of histone H2A, which contributes to chromatin compaction and repression of the target locus. The mechanism of PRC2 recruitment is currently not completely understood and is likely locus or environment dependent. Factors involved in this process may include CpG-rich domains and other sequence features of the promoters, DNA-binding proteins and IncRNA. lncRNA may be used to tether/guide/direct the complex to particular target genes through both base pairing to DNA and secondary–tertiary structure-driven protein binding. The tethering can occur by the nascent NAT at the time of NAT transcription or after NAT transcription has been completed, by pairing with DNA or mRNA sequences. There are several possible configurations for the pairing, including base pairing between the NAT and ssDNA, formation of a RNA–DNA–DNA triplex or via RNA–RNA hybrids of NATs with a nascent partner mRNA. It is possible that tethering requires participation of other protein factors because comparing human PRC2 binding to its known target lncRNAs with PRC2 binding to irrelevant transcripts, the binding constants were found to be similar in ciliates and bacteria (36). JARID2 could be one of the proteins that facilitates PRC2/lncRNA interaction, at least at imprinted loci involved in differentiation, including the Dlk1-Dio3 locus (37).

Interestingly, in mouse ESCs, PRC2 is found at both active and inactive promoters, but the H3K27me3 mark is not observed at active gene loci. Using in vivo RNA–protein cross linking, it was determined that EZH2, the catalytic subunit of PRC2, directly binds the 5′ region of nascent coding and ncRNAs at active promoters. EZH2 binding may contribute to decreased H3K27me3 deposition at these sites through the decoy mechanism (38). PRC2 also binds widely to enhancers, but H3K27me3 is only deposited at sites depleted for activating promoter motifs and enriched for motifs of developmental factors. These sequences represent blastula-stage DNA methylation-free domains that are conserved between humans, frogs and fish (39). The role of eRNAs in this process remains to be investigated.

Involvement of lncRNA in PRC2 recruitment, at least in the context of imprinting, has recently been questioned based on microarray-based epigenomic mapping and super-resolution 3D structured illumination microscopy data. Spatial separation and absence of colocalization of Xist and PRC2 was observed in the mouse ES cell line carrying an inducible Xist transgene located on chromosome 17 and in normal XX somatic cells (40).

Multiple authors have reported changes in other epigenetic DNA and chromatin marks including methylation and demethylation of H3K4, H3K9, acetylation and deacetylation of histone H3 involving multiple enzymes and epigenetic complexes, such as methyltransferases G9a, GLP and HDACs (Table 1). These alternative epigenetic modifications are likely induced by changes in NAT levels and activity.

Table 1.

Localization and mechanisms of action of known NATs

Coding gene symbol	Name used in reference	NAT name	NAT localization relative to coding gene	Effect on coding gene expression	Mechanism	Therapeutic relevance	Model	Methods	Ref.
ADAM12	ADAM12	Inter genic10	Intergenic, overlaps 3′ end of ADAM12	Concordant	NAT knockdown decreases H3K4me2 in the promoter region		Human fibroblast cells	Deep sequencing, ChRIP	(15)
APOA1	APOA1	APOA1 NAT	Overlaps coding gene cluster	Discordant	NAT recruits LSD1 and SUZ12, affects epigenetic marks at the three promoters	APOA1 cluster is involved in cardiovascular disease	Human and monkey cell lines and primary hepatocytes, monkeys	ChIP, real-time PCR	(41)
APOA4	APOA4	APOA1 NAT	Overlaps coding gene cluster	Discordant	NAT recruits LSD1 and SUZ12, affects epigenetic marks at the three promoters	APOA1 cluster is involved in cardiovascular disease	Human cell lines	ChIP, real-time PCR	(41)
APOC3	APOC3	APOA1 NAT	Overlaps coding gene cluster	Discordant	NAT recruits LSD1 and SUZ12, affects epigenetic marks at the three promoters	APOA1 cluster is involved in cardiovascular disease	Human cell lines	ChIP, real-time PCR	(41)
BACE1	BACE1	BACE1-AS	Overlaps coding gene	Concordant	NAT masks binding site for miR-485-5p	BACE1-AS is upregulated in AD brains and is involved in regulation of Aβ 1–42 production	APP transgenic mice, N2a and HEK293T C3 cells, human and mouse tissues, AD brains	Knockdown, overexpression, mutation analysis, DiscoveRx	(42–44)
BDNF	BDNF	BDNF-AS	Overlaps coding gene	Discordant	NAT depletion reduced H3K27met3 and Ezh2 binding, but not H3K4met3 and H3K36met3	BDNF regulates neuronal outgrowth and differentiation	Human and mouse cell lines, mice, cultured human neurons, glia, mouse neurospheres	Knockdown in vivo and in vitro, ChIP, IHC	(31)
CDKN2B	p15 (INK4B)	ANRIL	Overlaps coding gene	Discordant?	NAT is required for the recruitment of PRC2 and SUZ12 binding to p15(INK4B)	INK4B is activated by carcinogens to stop cell propagation	Human cell lines	Knockdown	(45)
CDKN2B	p15	p15AS	Overlaps coding gene	Discordant?	NAT mediates CDKN2B repression in cis and in trans via heterochromatin formation, not DNA methylation; repression continued after NAT expression was turned off; dicer independent	p15 is implicated in leukemia	Mouse ESCs, leukemic leukocytes, human cell lines	Expression profiling, overexpression	(46)
DHRS4, DHRS4L2, DHRS4L1	DHRS4, DHRS4L2, DHRS4L1	AS1DHRS4	Antisense in intron 1 of DHRS4	Discordant for all three genes	NAT pairs with ongoing sense transcripts; mediates H3 deacetylation and H3K4 demethylation of DHRS4, interacts with G9a and EZH2 in DHRS4L2 and DHRS4L1 promoters	DHRS4 is important for metabolism of organic compounds	Normal human hepatic (HL7702) and hepatocellular carcinoma (HepG2) cells	siRNA, RNA-ChIP	(47)
DLX1	Dlx1	Dlx1as	Overlaps coding gene	Discordant?			Embryonic mouse brain	qPCR, ISH	(48)
DLX6	DLX6	DLX6-AS/Evf2	Head to head	Discordant	NAT recruits DLX and MECP2 to regulatory elements in intergenic region	Evf2 mouse mutants had reduced synaptic inhibition	Developing mouse brain, Evf2 loss-of-function mice	Evf2 electroporation into brains, qChIP-PCR	(49)
EPH1B	EPH1B	EPH1B-AS	Overlaps coding gene	Discordant		EPHB1 is important in brain development and function	Human cells	Knockdown	(31)
FOXG1	FOXG1	FOXG1-AS	Promoter	Concordant?			SH-SY5Y, normal and autistic brains	Expression profiling	(50)
GCCR	GR	Gas5	Overlaps coding gene		Gas5 bound to DNA-binding domain of GCCR through its hairpin #5 without altering the abundance of GCCR target genes	Gas5 sensitized cells to apoptosis	HeLa cells	Yeast two-hybrid screen, GR-binding defective Gas5 mutant, ChiP, overexpression, mutation analysis	(51)
GDNF	GDNF	GDNF-AS	Overlaps coding gene	Discordant		EPHB1 is important in brain development and function	Human cells	Knockdown	(31)
HIF1A	HIF1 alpha	aHIF	Complementary to the 3′ UTR	Discordant?		NAT is overexpressed in non-papillary kidney cancer	Renal carcinomas, normal lymphocyte cell line	Ribonuclease protection, expression profiling	(52)
HRAS	p21	p21 antisense	Bidirectional transcription	Discordant	Ago-2 is required for NAT effect; loss of NAT correlates with loss of H3K27me3 at the p21 promoter		MCF-7 cells	Nuclear run on analysis, ChIP	(53)
IL1A	IL-1a	Anti-IL-1a/AK042010	Overlaps coding gene	Concordant?		NAT is involved in regulation of cytokine production	Mouse macrophage cell line and primary macrophages	Overexpression, ChIP, expression profiling	(54)
IL1B	IL-1β	Anti-IL-1β/ AK076405	Head to head, overlaps 5′ and promoter	Discordant	NAT inhibited IL1B at transcriptional level	NAT is involved in regulation of cytokine production	Mouse macrophage cell line and primary macrophages	Overexpression, ChIP, expression profiling	(54)
IL4I1	IL-4i1	Anti-IL-4i1/ AK087754	Overlaps coding gene	Concordant?		NAT is involved in regulation of cytokine production	Mouse macrophage cell line and primary macrophages	Overexpression, ChIP, expression profiling	(54)
KCNMB4	Kcnmb4	panc Kcnmb4	Bidirectional head to head	Concordant			Mouse and chimpanzee tissue samples	Directional RNA-seq, RT-PCR, knockdown	(23)
MSTN	Myostatin	Mstn promoter-associated RNA	Promoter	Concordant	NAT knockdown increases H3K9me2, but not H3K27me3 at the Mstn promoter; histone deacetylase dependent	Mstn negatively regulates muscle mass	Differentiated mouse muscle cell lines and myoblasts with dystrophin mutation	Knockdown, expression profiling	(55)
MYH	Myosin heavy chain (MHC)	bII NAT	NAT starts in IIb-Neo intergenic region and overlaps IIb MHC	Discordant		bII NAT is regulated during postnatal development and in response to hypothyroidism	Developing rat muscle	Expression profiling	(56)
MYH1	IIx MHC	aII NAT	NAT begins in the IIa–IIx intergenic region, overlaps IIa MHC	Concordant?		NAT is involved in slow to fast MHC transformation	Rat muscle after spinal cord isolation	Expression profiling	(57)
MYH1	IIx MHC	xII NAT	Shares bidirectional promoter with IIb MHC, overlaps IIx	Discordant?		Shift from IIb to IIx expression occurs after exercise	Rat tissues	Expression profiling	(58)
MYH2	IIa MHC	aII NAT	NAT begins in the IIa-IIx intergenic region, overlaps IIa	Discordant?		NAT is involved in slow to fast MHC transformation	Rat muscle after spinal cord isolation	Expression profiling	(57)
MYH3	Neo MHC	bII NAT	Begins in IIb-Neo intergenic region and overlaps IIb	Concordant?		NAT regulates transition between neo and IIb MHC	Normal and thyroid-deficient rat neonates treated with PTU	Expression profiling	(56)
MYH4	IIb MHC	bII NAT	Starts in IIb-Neo intergenic region and overlaps IIb	Discordant?		NAT regulates transition between neo and IIb MHC	Normal and PTU-treated rat neonates	Expression profiling	(56)
NFKB1	p105	Anti-p105/AK090099	Overlaps coding gene	Concordant?		NAT is involved in regulation of cytokine production	Mouse macrophage cell line and LPS-activated primary macrophages	Expression profiling	(54)
NFKB2	p100	Anti-p100/ AK029443	Head to head, overlaps the 5′ region and promoter	Concordant?		NAT is involved in regulation of cytokine production	Mouse macrophage cell line and LPS-activated primary macrophages	Overexpression, expression profiling	(54)
NIPBL	NIPBL	NIPBL-AS	Promoter		NAT localized to nucleoplasm or chromatin		SH-SY5Y, normal and autistic brains	Expression profiling	(50)
NOS2A	iNOS	iNOS asRNA	3′UTR of rat and mouse iNOS	Concordant	NAT stabilizes iNOS mRNA	Reducing iNOS mRNA levels is needed in inflammatory diseases and cancers	Primary hepatocytes from LPS-treated rats; IL-1β-treated hepatocytes; human tumors	Knockdown	(59)
PACSIN1	Pacsin1	panc Pacsin1	Bidirectional transcription	Concordant			Mouse and chimpanzee tissue samples	RNA-seq, knockdown	(23)
PINK1	PINK1	naPINK1	Overlaps coding gene	Concordant?		PINK1 mutations are implicated in early-onset Parkinson's	Neuronal cell lines	Knockdown	(60)
PLA2G4A	PLA2G4A	HERV-Ec1	Between exons 7 and 8, in antisense orientation	Discordant, mutual		NAT is downregulated in urothelial carcinoma	Human tumors and urothelial cell line (UROtsa)	Expression profiling, knockdown	(16)
PQBP1	PQBP1	PQBP1-AS	Overlaps exons and promoter	Discordant?			SH-SY5Y, normal and autistic brains	Expression profiling	(50)
PTCHD1	PTCHD1	PTCHD1AS1, PTCHD1AS2, PTCHD1AS3	Overlaps coding gene			Mutations in PTCHD1 locus are associated with autism spectrum disorder	Mouse and human tissues, 10T1/2 cells	ISH, overexpression	(61)
RASSF1A	RASSF1A	ANRASSF1	Overlaps coding gene	Discordant	ANRASSF1 forms an RNA/DNA hybrid, recruits PRC2 and increases H3K27me3 at RASSF1A promoter	RASSF1A is a tumor suppressor	HeLa, MDA-MB-231 and MCF-7 cells	Overexpression, knockdown, RNA-IP, RNase-ChIP	(62)
RNGTT	RNGTT	HERV-Ec6	Antisense between exons 14 and 15	Discordant?		NAT expression is reduced in urothelial carcinoma	Human tumors and urothelial cell line	Expression profiling	(16)
SH3RF3	Sh3rf3	pancSh3rf3	Overlaps coding gene	Concordant			Mouse and chimpanzee tissue samples	Directional RNA-seq, strand-specific RT-PCR, knockdown	(23)
SPI1	PU.1	PU.1	Overlaps coding gene	Discordant	NAT inhibits PU.1 by forming RNA duplex and decreasing PU.1 mRNA association with eEF1A	PU.1 expression is needed for suppression of leukemia	HL-60 and RAW 264.7 cells	Knockdown, expression profiling, RNA IP	(63)
STAR	Star	Star NAT	Overlaps coding gene	Discordant	HCG stimulates Star NAT expression via cAMP	STAR is important in steroidogenesis	MA-10 Leydig cells, murine tissues	Overexpression	(64)
SYNGAP1	SYNGAP1	SYNGAP1-AS	Overlaps exons and promoter	Discordant?	NAT localized to nucleoplasm or chromatin	SYNGAP1 AS is upregulated in autistic brain	SH-SY5Y, normal and autistic brains	Expression profiling	(50)
TFPI2	TFPI-2	TFPI-2as, LCT13	Overlaps coding gene	Discordant	Decreased TFPI-2 expression corresponds to increase in H3K9me3 and H4K20me3	TFPI-2 is a metastasis-suppressor gene	Transgenic mouse ES, breast and colon cancer cell lines	Overexpression, knockdown	(65)
TGFB	TGFbeta1,3	TGFbeta2	Overlaps coding gene				Human and rodent tissues	Northern, ISH with sense probe	(66)
TNFA	TNF-α	TNF-α asRNA	Overlaps coding gene	Discordant	Reduces TNF-α mRNA stability?		Primary rat hepatocytes; human tumor and blood cells	Knockdown	(67)
TNFSF9	4-1BBL	Anti-4-1BBL/AK154177	Overlaps 5′ region and promoter	Discordant?		TNFSF9 is involved in regulation of cytokine production	Mouse macrophage cell line and primary macrophages	Overexpression, expression profiling	(54)
TSPO	TSPO	Tspo-NAT	Overlaps SINEB2 and exon 3 of Tspo	Discordant	cAMP increases Tspo-NAT expression	Tspo is involved in the rate-limiting step in steroidogenesis	MA-10 mouse tumor Leydig cells	Overexpression, real-time PCR, western blot	(20)
TYMS	thymidylate synthase	rTSα	Coding NAT, overlaps partner gene		Induces transformation of the adenosine to inosine nucleotide in sense pre-mRNA leading to downregulation		HeLa cells	ISH, single cell real-time PCR, ribonuclease protection assay, knockdown, Northern	(68)
UCHL1	Uchl1	Antisense Uchl1	Overlaps coding gene		NAT is required for association of UCHL1 with polysomes for translation; NAT activity needs embedded SINEB2	Uchl1 is involved in brain function and neurodegenerative diseases	Mouse cells	Mutation analysis	(19)
VPS13B	VPS13B	VPS13B-AS	Promoter		Localized to nucleoplasm or chromatin		SH-SY5Y, normal and autistic brains	Expression profiling	(50)
VWA5B2	Vwa5b2	panc Vwa5b2	Overlaps coding gene	Concordant			Mouse and chimpanzee tissue samples	Directional RNA-seq, RT-PCR, knockdown	(23)
WDR83	WDR83	DHPS (protein coding)	Overlaps coding gene, head to head	Concordant, mutual	WDR83 and DHPS form RNA duplex at overlapping 3′ UTRs which increases their stability	WDR83 or DHPS promote cell proliferation in gastric cancer	MGC803 cells (gastric cancer)	Overexpression	(25)

Generation of endogenous siRNAs and miRNA/modulation of mRNA stability

The first proposed mechanism of NAT-mediated gene expression regulation was based on the presence of overlapping exons between NATs and their sense gene partners as well as known information about how siRNA/miRNA acts upon messenger RNA. This evidence was used to formulate the theory that sense/antisense RNA duplex formation during or after transcription leads to either sense transcript degradation or stabilization. This mechanism overlaps with another suggested function of ncRNAs, namely serving as precursors for endogenous siRNAs and miRNA production. Although multiple examples of gene expression regulation through sense–antisense RNA interaction are known (Table 1), it is likely that these mechanisms are not very widespread, because only 1% of the genome is transcribed from both plus and minus strands (23).

The biological importance of the sense–antisense RNA interaction is supported by recent findings showing that a subset of lncRNA is enriched in the cytosol and in ribosomal fractions rather than in the nucleus (69). In mouse CD4+ T cells multiple endogenous siRNA and miRNA transcripts were identified. These transcripts interacted with argonaute (AGO) proteins that mediated RNA interference and posttranscriptional gene silencing (32). Short RNA-seq and cross linking, ligation and sequencing of hybrids of human embryonic kidney cell (HEK293) RNA yielded antisense transcripts from 378 genes with a characteristic endo-siRNA footprint (co-occurrence of RNAPII and AGO1); (33).

lncRNA in stabilization of long-range chromosomal interactions

Advanced methods, including chromatin conformation capture and its modifications, 4C, 5C and Hi-C (1–3), have revealed long-range chromosomal interactions in B- and T-cell receptor loci (reviewed in 70). Multiple chromatin loops in these loci may form rosette-like structures that bring distant chromosomal regions into the same transcription factory. These structures may be formed with participation of CTCF and cohesin proteins and lncRNAs.

THERAPEUTIC APPLICATIONS OF NATS

The high target specificity of NATs that normally regulate one gene or a small group of related genes makes them a welcome addition to the list of targets available for therapeutic intervention. A growing number of NATs in disease-relevant loci is being characterized (Tables 1 and 2). Although functional NATs were described as early as mid-1990, the therapeutic application for them was not proposed until mid-2000s (81). Whereas some authors suggest using NATs and other lncRNAs for gene downregulation (55), a more unique therapeutic aspect of NATs is their ability to increase the expression of specific genes (12,82–85). This approach is currently being brought into practice by at least two biotechnology companies utilizing related NAT-targeting technologies: (i) oligonucleotides interfering with NAT function through steric hindrance or RNAse H-mediated degradation (OPKO CURNA, Miramar, FL, USA; founded 2008) and (ii) oligonucleotides targeting NAT interaction with PRC2 (RaNA, Cambridge, MA, USA; founded 2011). Additionally, manipulation of regulatory pseudogene transcripts through the use of synthetic antisense oligonucleotides, siRNAs, aptamers or gene therapy has been proposed as a novel pharmacological strategy (21). An algorithm to design short ncRNAs for the epigenetic transcriptional silencing or activation of specific genes has been published (86).

Table 2.

NATs involved in rare genetic disorders

Coding gene symbol	Name used in reference	NAT name	NAT localization relative to coding gene	Effect on coding gene expression	Mechanism	Therapeutic relevance	Model	Methods	Ref.
ARHGEF26	ARHGEF26	AK087060	Starts 225 bps downstream of the coding gene	Concordant?		Mecp2 is mutated in Rett syndrome	Brain of Rett syndrome mice	Overexpression, microarray, ChIP, bisulfite genomic sequencing	(71)
ATXN7	Ataxin-7	SCAANT1	Overlaps coding gene	Discordant	NAT induces repression in cis-, accompanied by increase in H3K27me3 and decrease in H3 acetylation at ATXN7 promoter; overexpression of NAT did not affect ATXN7	In SCA7, SCAANT1 is decreased and ATXN7 is increased	Mice with ataxin-7 mini-genes, human retinoblastoma cell lines, primary cerebellar astrocytes	ChIP, deletion analysis, overexpression	(72)
ATXN8	ATXN8	ATXN8OS, protein coding	Overlaps coding gene			ATXN8 is mutated in SCA8	Human cell lines	Expression profiling	(73)
FMR1	FMR1	FMR4 or FMR1-AS1	Bidirectional transcription	None		FMR4 and FMR1 are silenced in fragile X patients and upregulated in premutation carriers	Human and monkey tissues, HEK293T, HeLa cells	Knockdown, overexpression, expression profiling	(74)
FMR1	FMR1	FMR6	Overlaps the 3′ of FMR1, antisense			FMR6 is silenced in full Fragile X Syndrome mutation and permutation carriers	Human tissues, leukocytes	Expression profiling, deep RACE, next generation sequencing	(75)
GABRR2	Gabrr2	AK081227 (same direct as coding)	Intron of Gabrr2	Discordant?		Involved in Rett syndrome	Brain of Rett syndrome mice	Overexpression, microarray, ChIP, bisulfite genomic sequencing	(71)
HTT	HTT	HTTAS	Overlaps coding gene	Discordant	NAT effect is Dicer-dependent, repeat expansion reduces expression	HTTAS is reduced in human HD	Human cells	Overexpression, siRNA	(76)
KLHL1	KLHL1	SCA8, KLHL1AS	Overlaps coding gene			NAT contributes to neurotoxicity in SCA9	Human and mouse tissues	Expression profiling	(77)
MKRN3	ZNF127	ZNF127AS, MKRN3-AS1	Overlaps coding gene, transcribed only from paternal allele			Prader–Willi syndrome is caused by deletion on the paternal chromosome including MKRN3/ZNF127	Human and mouse tissues	Expression profiling	(78)
SCA8	SCA8		Overlaps coding gene			NAT contributes to neurotoxicity in SCA8	HEK293 and SH-SY5Y cells	Expression profiling	(79)
UBE3A	Ube3a	Ube3a-as	Overlaps coding gene	Discordant		Involved in Angelman syndrome	Transgenic mice	Mutational mapping, expression profiling	(80)

The two main areas for which ncRNA-based therapies are now being developed are cancers and rare genetic disorders. Cancer applications are mostly based on lincRNAs (reviewed in 87–89). Examples of ncRNAs likely associated with rare genetic disorders are given in Table 2 and have been recently reviewed (90).

CONCLUSIONS

After a long struggle, studies of lncRNAs including NATs are now expanding at a very high rate, leading to the understanding of the major role of RNA in the functioning of the genome as well as the discovery of novel targets for therapeutic intervention.

Conflict of Interest statement. None declared.

FUNDING

NIH funded, grant numbers 1R01MH084880, R01MH083733 and R01NS063974.