Fatica 2013

REVIEWS
NON-CODING RNA
Long non-coding RNAs: new players in

cell differentiation and development
Alessandro Fatica1 and Irene Bozzoni1–4
Abstract | Genomes of multicellular organisms are characterized by the pervasive expression
of different types of non-coding RNAs (ncRNAs). Long ncRNAs (lncRNAs) belong to a novel
heterogeneous class of ncRNAs that includes thousands of different species. lncRNAs have
crucial roles in gene expression control during both developmental and differentiation
processes, and the number of lncRNA species increases in genomes of developmentally
complex organisms, which highlights the importance of RNA-based levels of control in the
evolution of multicellular organisms. In this Review, we describe the function of lncRNAs in
developmental processes, such as in dosage compensation, genomic imprinting, cell
differentiation and organogenesis, with a particular emphasis on mammalian development.
microRNAs
Although the central role of RNA in cellular func- to specific RNA or DNA target sites5,7,8. This dual activ-
(miRNAs). Small non-coding tions and organismal evolution has been advocated ity is shared with small ncRNAs4, such as microRNAs
RNAs of ~22 nucleotides that periodically during the past 50 years, only recently has (miRNAs), small nucleolar RNAs and many other small
are integral components of RNA received a remarkable level of attention from the nuclear ribonucleoprotein particles (BOX 2). However,
RNA-induced silencing complex
(RISC) and that recognize
scientific community. Analyses that compare transcrip- unlike small ncRNAs, lncRNAs can fold into complex
partially complementary target tomes with genomes of mammalian species (BOX 1) have secondary and higher order structures to provide greater
mRNAs to induce translational established that approximately two-thirds of genomic potential and versatility for both protein and target rec-
repression, which is often linked DNA is pervasively transcribed, which is in sharp con- ognition5,7,8. Moreover, their flexible8,9 and modular 10,11
to degradation. Among the
RISC proteins, AGO binds to
trast to the <2% that is ultimately translated into pro- scaffold nature enables lncRNAs to tether protein fac-
miRNA and mediates the teins1,2. Moreover, the degree of organismal complexity tors that would not interact or functionally cooperate if
repressing activity. among species better correlates with the proportion of they only relied on protein–protein interactions5,8,12–14.
each genome that is transcribed into non-coding RNAs Such combinatorial RNA-mediated tethering activity has
1
Department of Biology and (ncRNAs) than with the number of protein-coding enhanced gene regulatory networks to facilitate a wide
Biotechnology “Charles
Darwin”, Sapienza University
genes, even when protein diversification by both alterna- range of gene expression programmes (FIG. 1) to provide
of Rome, Piazzale Aldo Moro 5, tive splicing and post-translational regulation are taken an important evolutionary advantage5,7,8. This complexity
00185 Rome, Italy. into account 3. This suggests that RNA-based regula- is likely to be further expanded by differential splicing
2
Institute of Molecular tory mechanisms had a relevant role in the evolution of and the use of alternative transcription initiation
Biology and Pathology of the
National Research Council,
developmental complexity in eukaryotes. sites and polyadenylation sites by lncRNAs, thus increasing
Piazzale Aldo Moro 5, 00185 The range of ncRNAs in eukaryotes is vast and the number of tethering-module combinations.
Rome, Italy. exceeds the number of protein-coding genes. Besides the The expression of lncRNAs has been quantitatively
3
Istituto Pasteur Fondazione different families of small ncRNAs4, a large proportion of analysed in several tissues and cell types by high-
Cenci Bolognetti, Piazzale Aldo
Moro 5, 00185 Rome, Italy.
transcriptomes results in RNA transcripts that are longer throughput RNA sequencing (RNA-seq) experiments,
4
Center for Life Nano Science than 200 nucleotides, which are often polyadenylated and it was generally found to be more cell type specific
@Sapienza, Istituto Italiano di and are devoid of evident open reading frames (ORFs) than the expression of protein-coding genes5,6,8,15–17.
Tecnologia, Sapienza — these are defined as long ncRNAs (lncRNAs) 5–7. Interestingly, in several cases, such tissue specificity has
University of Rome, Viale
Regina Elena 291, 00161
Many roles are emerging for lncRNAs in ribonucleo- been attributed to the presence of transposable elements
Rome, Italy. protein complexes that regulate various stages of gene that are embedded in the vicinity of lncRNA transcrip-
Correspondence to I.B. expression5,7. Their intrinsic nucleic acid nature confers tion start sites18–20. Moreover, lncRNAs have been shown
e‑mail: on lncRNAs the dual ability to function as ligands for to be differentially expressed across various stages of
irene.bozzoni@uniroma1.it
doi:10.1038/nrg3606
proteins (such as those with functional roles in gene differentiation, which indicates that they may be novel
Published online regulation processes) and to mediate base-pairing ‘fine-tuners’ of cell fate5–7. This specific spatiotemporal
3 December 2013 interactions that guide lncRNA-containing complexes expression can be linked to the establishment of both
NATURE REVIEWS | GENETICS VOLUME 15 | JANUARY 2014 | 7
© 2014 Macmillan Publishers Limited. All rights reserved

REVIEWS
Box 1 | Methodologies for lncRNA identification and analyses

The identification of long non-coding RNAs (lncRNAs) relies on the detection of transcription from genomic regions that
are not annotated as protein coding, such as regions that are devoid of open reading frames. This can be achieved by the
direct detection of the transcribed RNA. However, conventional gene expression microarrays are only designed to detect
the expression of protein-coding mRNAs, and unbiased RNA detection methods are therefore required. These include
tiling arrays, serial analysis of gene expression (SAGE), cap analysis of gene expression (CAGE) and high-throughput RNA
sequencing (RNA-seq). Alternatively, transcription can be inferred from genomic DNA analyses by detecting specific
histone marks (such as H3K4–H3K36 domains) that signify transcriptionally active chromatin.
Tiling arrays
In this technique, cDNA is hybridized to microarray slides carrying overlapping oligonucleotides that cover either
specific chromosomal regions or a complete genome. This methodology allows the analysis of global transcription
from specific genomic regions and was initially used for both identification and expression analysis of lncRNAs112.
SAGE
SAGE was the first method to use sequencing for high-throughput analyses of transcriptomes113. It is based on the
generation of short stretches of unbiased cDNA sequence (that is, SAGE tags) by restriction enzymes. SAGE tags are
concatenated before cloning and sequencing. This methodology allows both the quantification of transcripts
throughout the transcriptome and the identification of new transcripts, including lncRNAs. Several modifications to
the original SAGE strategy have been developed to improve the specificity by generating larger tags, such as
LongSAGE and SuperSAGE114,115.
CAGE
CAGE relies on the isolation and sequencing of short cDNA sequence tags that originate from the 5ʹ end of RNA
transcripts116. Similarly to SAGE, tags are concatenated before cloning and sequencing. However, in addition to
quantifying expression level, CAGE also identifies the location of each transcription start site.
RNA-seq
Sequencing of transcriptomes by RNA-seq is one of the most powerful methodologies for de novo discovery and
expression analyses of lncRNAs117. In this method, total RNA is converted to a cDNA library that is directly sequenced by
high-throughput sequencing instruments. There are several types of sequencing technologies but Illumina platforms are
currently the most commonly used for RNA-seq experiments. A single sequencing run produces billions of reads that
are subsequently aligned to a reference genome. Following alignment, the data are translated into a quantitative
measure of gene expression by specific algorithms.
Chromatin immunoprecipitation (ChIP)
ChIP allows the isolation of DNA sequences that are associated with a chromatin component of interest. When
Polyadenylation sites combined with high-throughput readouts such as microarrays (that is, ChIP–chip) and DNA sequencing (that is,
Sequences that are required ChIP–seq), these methods can infer the genomic distribution of either proteins or histone modifications. Analysis of
for the cleavage of primary loci with specific histone modifications that characterize active transcription (such as H3K4–H3K36 marks) allowed an
RNA transcripts that are indirect identification of many unknown lncRNAs77.
produced by RNA polymerase II.
As a consequence of such
cleavage, the 5′ cutoff product
becomes polyadenylated, well-defined barriers of gene expression and cell- cases, they recruit DNA methyltransferase 3 (DNMT3)
whereas the 3′ product type-specific gene regulatory programmes. Combined and histone modifiers, such as the Polycomb repressive
undergoes rapid degradation with the involvement of lncRNAs in positive or nega- complex PRC2 (REFS 12,25) and histone H3 lysine 9
that induces Pol II release
from the DNA and hence
tive feedback loops, lncRNAs can amplify and con- (H3K9) methyltransferases26,27. The resultant DNA and
transcriptional termination. solidate the molecular differences between cell types histone modifications predominantly correlate with the
that are required to control cell identity and lineage formation of repressive heterochromatin and with tran-
Polycomb repressive commitment 21–23. scriptional repression. Furthermore, the act of lncRNA
complex
In this Review, we discuss our latest understanding transcription itself can negatively affect gene expres-
(PRC). A multiprotein complex
that silences target genes by of lncRNAs according to their roles in various develop- sion28,29. Transcriptional activation has also been shown
establishing a repressive mental processes. We focus on lncRNAs for which roles through the recruitment of chromatin-modifying
chromatin state; PRC2 have been confirmed by functional studies in cellular complexes, such as the histone H3K4 methyltransferase
trimethylates histone H3 at systems, and some of these have been further charac- MLL1 complex (REFS 30,31) , and by the activation of
lysine 27, which is recognized
by PRC1 that mediates
terized through in vivo loss‑of‑function approaches specific enhancer regions through changes to three-
chromatin compaction in model organisms (TABLE 1). We describe molecular dimensional chromatin conformation 30,32,33. With
by inducing H2A mechanisms of action and roles of lncRNAs in cellular respect to the target sites, it is possible to distinguish
monoubiquitylation. processes such as genomic imprinting, maintenance of between cis- and trans-acting lncRNAs — cis-acting
pluripotency and development in various organs, and lncRNAs control the expression of genes that are posi-
MLL1 complex
A multiprotein complex that in environment-responsive developmental programmes. tioned in the vicinity of their transcription sites and can
mediates both histone H3 Finally, we summarize future challenges in the field. sometimes spread their effect to long distances on the
trimethylation at lysine 4 same chromosome, whereas trans-acting lncRNAs can
(H3K4me3) and histone Modes of action of lncRNAs either repress or activate gene expression at independ-
H4 acetylation at lysine 16
(H4K16ac), which are
Nuclear lncRNAs. Evidence so far indicates that most ent loci5,7,8. However, for both classes of lncRNAs, the
associated with nuclear lncRNAs function by guiding chromatin targeting mechanisms is still far from being understood;
transcriptionally active genes. modifiers to specific genomic loci5,7,8,24 (FIG. 1A). In most in particular, it is not known how cis-acting lncRNAs
8 | JANUARY 2014 | VOLUME 15 www.nature.com/reviews/genetics

REVIEWS
Box 2 | Examples of ncRNAs with guiding activity for proteins

are retained to the sites of their transcription and how
trans-acting lncRNAs find distantly located targets.
a Pre-mRNA splicing Different recognition mechanisms have been proposed,
U1 snRNP including recruitment by bridging proteins34, formation
of an RNA–DNA triplex 35 and DNA recognition by RNA
U1-70K
SR structures36. Nuclear lncRNAs can also have indirect
regulatory effects on gene loci (FIG. 1B); for example, by
acting as decoys that sequester transcription factors37,38,
U2AF by allosterically modulating regulatory proteins 39,
Exon A - YRYRYRYY Exon and by altering nuclear domains 40 and long-range
Pre-mRNA three-dimensional chromosomal structures41.
Cytoplasmic lncRNAs. Many lncRNA-mediated mech-

b Site-specific RNA modification
anisms of gene regulation have been identified in the
H/ACA-box snoRNP C/D-box snoRNP
(pseudouridylation) (ribose 2′-O- cytoplasm7. These lncRNAs often show sequence com-
D′ plementarity with transcripts that originate from either
methylation)
CH3 the same chromosomal locus or independent loci. Upon
C′
recognition of the target by base pairing, they can mod-
Fibrillarin ulate translational control, examples of which include
ψ Dyskerin ψ Dyskerin
CH3
positive regulation by the ubiquitin carboxy-terminal
RNA hydrolase L1 antisense RNA 1 (Uchl1‑as1)42 and nega-
A
UG
tive regulation by tumour protein p53 pathway core-

A CU
RNA 5′ RUG GA
C Box pressor 1 (Trp53cor1; also known as lincRNA‑p21)43.
5′
5′ ANANNA ACA NN 3′ D Box
Similarly, lncRNAs can modulate mRNA stability; for
H Box ACA Box 5′ 3′
example, both β-site APP-cleaving enzyme 1‑antisense
(BACE1‑AS)44 and tissue differentiation-inducing non-
AGO2 protein-coding RNA (TINCR)45 increase the stability
c RNA interference mRNA
of their target mRNAs, whereas half‑STAU1 (staufen
siRNAs AAAAA double-stranded RNA-binding protein 1)‑binding
site RNAs (1/2sbsRNAs)46,47 decrease target mRNA
GW182
stability (FIG. 2A).
miRNP-mediated
A peculiar mode of action is that of lncRNAs that
mRNA AGO
silencing function as competing endogenous RNAs (ceRNAs)48
AAAAA — by binding to and sequestering specific miRNAs,
Ribosome miRNAs
ceRNAs function as ‘miRNA sponges’ to protect the
target mRNAs from repression. This represents a new
d Telomere formation DNA Telomerase type of regulatory circuitry in which different types of
RNAs (both coding and non-coding) can crosstalk to
TERC RNA each other by competing for shared miRNAs. This activ-
ity was initially described in plants49 and, subsequently,
in mammals50, in which it was shown to be relevant in
many processes, including tumorigenesis51,52, cell differ-
entiation21 and pluripotency 23. Recently, an additional
Guiding activity of RNAs has been described over many years inNaturedifferent classes| Genetics
Reviews of well- example of ceRNA was found in a newly identified
characterized non-coding RNAs (ncRNAs). These ncRNAs function in trans by bringing class of circular RNAs (circRNAs)53–55, which function
specific interactors to specific targets (shown in red) through base pairing with other
as sponges for miRNAs in neuronal cells (FIG. 2B). It is
RNA or DNA molecules. For example, the spliceosomal U1 small nuclear ribonucleoprotein
particles (snRNPs) work in the splicing process by recognizing 5′ splice sites in pre-mRNA
interesting that, whereas the linear ceRNAs have a short
molecules through the U1 small nuclear RNA component and by contacting splicing half-life that allows a rapid control of sponge activity,
factors (such as SR proteins and U2 auxiliary factor (U2AF)), which are bound to the 3′ circRNAs have much greater stability and their turnover
portion of the intron, through the U1‑70K protein component of the U1 snRNP118 (see the can be controlled by the presence of a perfectly matched
figure, part a). Similarly, small nucleolar RNAs (snoRNAs) guide modifying enzymes to miRNA target site53–55.
produce site-specific pseudouridylation (Ψ; for H/ACA-box snoRNAs, catalysed by
dyskerin) and ribose 2ʹ-O-methylation (for C/D-box snoRNAs, catalysed by fibrillarin) on Dosage compensation and genomic imprinting
target RNAs119 (see the figure, part b). In the cytoplasm, small interfering RNAs (siRNAs) Among the first and best-characterized examples of
and microRNAs (miRNAs) direct Argonaute (AGO) proteins to target RNAs. The lncRNAs that have specific developmental roles and
AGO2–siRNA complex induces RNA cleavage, whereas the AGO–miRNA complex, in
robust loss‑of‑function phenotypes in vivo are those
combination with GW182, triggers translational repression and mRNA degradation22
(see the figure, part c). Finally, telomerase RNA component (TERC) recognizes single-
involved in dosage compensation and genomic imprinting
stranded DNA at telomeres and serves as a template for the associated telomerase9 (FIG. 1Aa). These two processes are required for normal
(see the figure, part d). In this case, the ncRNA molecules provide DNA-binding development and rely on the formation of silenced
specificity to a protein complex. In all these examples, target specificity is provided by chromatin to produce monoallelic expression of specific
short complementary sequences. miRNP, microRNA ribonucleoprotein complex. genes in mammals56.

REVIEWS
A Cis-acting IncRNAs Repressive histone modification

Chromatin- Activating histone modification
modifying
complex
IncRNA gene
Aa Xist, Kcnq1ot1 and Airn Ab HOTTIP

H3K9me2 or H3K9me3
5mC H3K4me3
MLL1
H3K27me3 complex
DNMT3
EHMT2
PRC2
IncRNA gene
Chromatin-
B Trans-acting IncRNAs modifying
complex
IncRNA gene IncRNA
Transcriptional regulator
IncRNA
Ba HOTAIR
KDM1A–coREST–
PRC2 REST complex
complex
Bb IncRNA-ES1 and IncRNA-ES2
OCT4 SOX2
H3K27me3 H3K4me3
Bc Jpx
PRC2
complex
CTCF
Xist promoter
H3K27me3
10 | JANUARY 2014 | VOLUME 15 Nature Reviews | Genetics

www.nature.com/reviews/genetics

REVIEWS
◀ Figure 1 | Models of nuclear lncRNA function. Examples of long non-coding RNAs for a naturally occurring antisense transcript in gene
(lncRNAs) that regulate transcription in cis (part A) and in trans (part B), by recruiting expression regulation. Furthermore, Xist activation also
specific transcriptional regulators onto specific chromosomal loci, are shown. requires the lncRNA Jpx 62, which induces Xist tran-
Aa | lncRNAs that are involved in dosage compensation and genomic imprinting scription through the sequestration of transcriptional
include X‑inactive specific transcript (Xist), Kcnq1 overlapping transcript 1 (Kcnq1ot1)
repressor CTCF38 (FIG. 1Bc).
and Airn (antisense Igf2r (insulin-like growth factor 2 receptor) RNA). These lncRNAs
induce the formation of repressive chromatin through the recruitment of DNA
Xist, which is transcriptionally regulated by a
methyltransferase 3 (DNMT3), which induces DNA methylation; Polycomb repressive network of pluripotency factors, may also have an
complex 2 (PRC2), which produces histone H3 lysine 27 trimethylation (H3K27me3); important role in differentiation. Indeed, both the
and histone lysine N-methyltransferase EHMT2, which is responsible for producing homozygous and heterozygous conditional deletion of
H3K9me2 and H3K9me3 (REF. 56). Ab | HOXA distal transcript antisense RNA (HOTTIP) Xist in mouse haematopoietic stem cells produced an
functions through the recruitment of the MLL1 complex, which drives the formation of aberrant maturation of haematopoietic progenitors in
the activating H3K4me3 mark30. Ba | HOXA transcript antisense RNA (HOTAIR) is a females63, which resulted in the development of blood
trans-acting regulator of the HOXD genes12. It is characterized by a modular scaffold cell cancers and in accelerated death. Aberrant XIST
structure that allows the recruitment of two distinct repressive complexes, PRC2 and expression has been observed in human cancers, which
the H3K4 demethylating complex KDM1A–coREST–REST (lysine-specific histone
further suggests that alteration in the X inactivation
demethylase 1A–REST corepressor 1–RE1-silencing transcription factor) on the same
genomic region11. Bb | The pluripotency RNAs lncRNA-ES1 and lncRNA‑ES2 associate
process contributes to tumorigenesis.
with both PRC2 and the transcription factor sex-determining region Y-box 2 (SOX2),
which suggests that these lncRNAs control embryonic stem cell pluripotency by Genomic imprinting. Imprinted genes generally asso-
silencing SOX2‑bound developmental genes14; this function is alternative to OCT4- ciate in clusters and are epigenetically marked in
and SOX2‑dependent activation of pluripotency genes. Bc | The lncRNA Jpx (Jpx sex-dependent ways during male and female game-
transcript, Xist activator) that binds to the transcriptional repressor CTCF inhibits its togenesis; they are subsequently silenced on only
binding to the Xist promoter, thus activating Xist transcription38. one parental chromosome in the embryo. Imprinted
regions encode different species of ncRNAs, includ-
ing lncRNAs that, in many cases, bind to imprinted
X chromosome inactivation. The identification of regions and are directly involved in silencing 56. These
X‑inactive specific transcript (XIST) as a regulator lncRNAs are generally long (more than 100 kb) and
of X chromosome inactivation in mammals provided function in cis. The best-characterized example at both
one of the first examples of a lncRNA that is directly the genetic and the molecular levels are the lncRNAs
involved in the formation of repressive chromatin 56. Kcnq1 overlapping transcript 1 (Kcnq1ot1) and Airn
Xist deletion in mice causes a loss of X chromosome (antisense Igf2r (insulin-like growth factor 2 recep-
inactivation and female-specific lethality 57. Various tor) RNA). These lncRNAs are paternally expressed;
studies both in mice and in mouse embryonic stem they function by repressing flanking protein-coding
cells (ESCs) — a major model system for X chro- genes in cis and are involved in early development
mosome inactivation — have demonstrated that, in in mice56. The loss of function of these lncRNAs in
female cells, Xist acts in cis by inducing the forma- the embryo is not lethal — paternal inheritance of a
tion of transcriptionally inactive heterochromatin loss‑of‑function allele results in a loss of imprinting
on the X chromosome from which it is transcribed56. and in growth defects, whereas maternal inheritance
Xist is required only for the initiation and not for the of this allele does not affect imprinting or growth64–66.
maintenance of X inactivation, and its spatiotemporal These studies showed that multiple repressive pathways
expression must be properly controlled56. Xist induces regulate imprinted gene silencing by lncRNAs during
the formation of repressive heterochromatin, at least development, and that the extent of silencing along the
in part, by tethering PRC2 to the inactive X chromo- chromosome varies in different tissues26,27,66. For exam-
some25. However, parallel PRC2‑independent pathways ple, during embryonic development, Kcnq1ot1 func-
have been recently demonstrated in both mouse and tions by establishing and maintaining repressive DNA
Dosage compensation
human ESCs58,59. methylation on surrounding genes, whereas, in the
The process that ensures
equal levels of X‑linked gene The interaction between Xist and chromatin may placenta, it functions by recruiting the repressive his-
expression in males (XY) and involve, among others, transcriptional repressor tone modifiers PRC2 and the H3K9 methyltransferase
females (XX). protein YY1 that is thought to function as a recruit- EHMT2 (also known as G9a) on genes that are located
ment platform for Xist by binding to its first exon34. further away from the imprinted region66. It is worth
Genomic imprinting
Epigenetic silencing of genes
Moreover, it has been recently shown that Xist itself noting that, in the establishment of transcriptional gene
on the basis of their parental is able to recognize the three-dimensional conforma- silencing by cis-acting lncRNAs, continuous transcrip-
origin, which results in tion of the X chromosome41. Notably, Xist expression tion might be more important than the production of
monoallelic expression. is itself controlled by other lncRNAs in both a positive mature RNA. This has been elegantly shown for Airn,
and a negative manner 56. One of the best-characterized which is expressed from the paternal chromosome and
EHMT2
A histone lysine Xist regulators is its natural antisense non-coding is antisense to the Igf2r gene. Airn functions in cis to
methyltransferase that is transcript Tsix. Tsix counteracts Xist expression by silence the paternal Igf2r allele, whereas the maternal
responsible for dimethylation inducing repressive epigenetic modifications at the Igf2r allele remains expressed. In embryonic tissues,
and trimethylation at histone Xist promoter 56. The loss of Tsix function in vivo Airn silences paternal Igf2r through a mechanism that
H3 lysine 9, which creates
epigenetic marks that
resulted in ectopic Xist expression, aberrant X inacti- does not require a stable RNA product but that is based
predominantly correlate with vation and early embryonic lethality 60,61. These mouse on continuous Airn transcription, which interferes with
transcriptional repression. models showed, for the first time, an important role the recruitment of RNA polymerase II29. By contrast,

REVIEWS
Table 1 | lncRNA manipulation and resulting phenotypes in model animal systems

lncRNA Process Site of action Loss‑of‑function methods Phenotype Refs
Mouse
Xist Dosage compensation Nucleus Gene disruption in embryo Embryonic lethality 57
Conditional disruption in Aberrant haematopoiesis and 63
haematopoietic stem cells blood cell cancer
Tsix Dosage compensation Nucleus Embryonic gene inactivation Embryonic lethality 60,61
by either promoter deletions or
premature termination
Kcnq1ot1 Genomic imprinting Nucleus Embryonic gene inactivation Growth defects 64,66
Airn Genomic imprinting Nucleus Embryonic gene inactivation Growth defects 29,65
Fendrr Gene expression regulation Nucleus Gene disruption in embryo Embryonic lethality 95
in mesoderm
60% reduction in embryo by RNA Normal development 95
interference
Hotair Hox gene regulation Nucleus Gene disruption in embryo Defects in skeletal system 72
development
Dlx1os Homeodomain transcription Nucleus Embryonic gene inactivation by Morphologically normal with 92
factor regulation in premature termination mild skull and neurological
developing forebrain defects
Dlx6os1 Homeodomain transcription Nucleus Embryonic gene inactivation by Morphologically normal with 90
factor regulation in premature termination altered GABAergic interneuron
developing forebrain development
Malat1 Tumorigenesis Nucleus Gene disruption in embryo Normal development 82
Miat Retina development Nucleus Knockdown and overexpression in Defects in specification of retina 126
neonatal retina cell types
Six3os1 Retina development Nucleus Knockdown and overexpression in Defects in specification of retina 127
neonatal retina cell types
Tug1 Retina development Nucleus Knockdown in neonatal retina Defects in differentiation of 128
photoreceptor progenitor cells
Vax2os Retina development Nucleus Overexpression in neonatal retina Defects in differentiation of 129
photoreceptor progenitor cells
Zebrafish
Cyrano Embryogenesis Not analysed Knockdown and functional Developmental defects 110
inactivation in embryo by
morpholino oligonucleotides
Megamind Embryogenesis Not analysed Knockdown and functional Defects in brain morphogenesis 110
inactivation in embryo by and in eye development
morpholino oligonucleotides
Chicken
HOTTIP HOXA regulation Nucleus Knockdown in chick embryos by RNA Altered limb morphology 30
interference
Airn, antisense Igf2r (insulin-like growth factor 2 receptor) RNA; Dlx1os, distal-less homeobox 1, opposite strand; Dlx6os1, Dlx6 opposite strand transcript 1; Fendrr,
Foxf1 adjacent non-coding developmental regulatory RNA; GABA, γ-aminobutyric acid; Hotair, HoxA transcript antisense RNA; HOTTIP, HOXA distal transcript
antisense RNA; Kcnq1ot1, Kcnq1 overlapping transcript 1; lncRNA, long non-coding RNA; Malat1, metastasis-associated lung adenocarcinoma transcript 1; Miat,
myocardial infarction-associated transcript (also known as Rncr2); Six3os1, Six3 opposite strand transcript 1; Tsix, X (inactive)-specific transcript, opposite strand; Tug1,
taurine upregulated gene 1; Vax2os, Vax2 opposite strand transcript; Xist, X‑inactive specific transcript.
in the placenta, mature Airn recruits EHMT2 to induce lncRNAs to regulate a gene cluster. The in situ produc-
the formation of repressive chromatin65. Altogether, tion of regulators at their site of function is intrinsically
these studies showed that a single lncRNA could work more robust than dedicated trans-acting proteins. Thus,
by different mechanisms depending on the cell type, it is not surprising that the use of cis-acting lncRNAs
which might reflect the presence of either different to silence gene transcription is an evolutionarily con-
interactors or chromatin modifications that influence served mechanism and is not restricted to complex and
lncRNA functions in diverse cellular contexts. These multicellular organisms, as in the case of yeast cryptic
examples also show the advantages of using cis-acting unstable transcripts67.

REVIEWS
Regulation of HOX genes Trans-acting lncRNAs. HOX transcript antisense RNA

HOX genes encode an evolutionary conserved family (HOTAIR) was one of the first trans-acting lncRNAs to
of transcription factors that regulate the embryo body be identified12. HOTAIR is transcribed from the HOXC
plan and that contribute to cell specification in several gene cluster but acts as a repressor of the HOXD cluster,
adult differentiation processes68. In mammals, there which is located on a different chromosome. HOTAIR
are 39 HOX genes that are grouped in four clusters interacts with the PRC2 and KDM1A–coREST–REST
(HOXA, HOXB, HOXC and HOXD), which allow (lysine-specific histone demethylase 1A–REST core-
precise spatiotemporal coordination of expression. pressor 1–RE1‑silencing transcription factor) histone-
In addition to protein-coding genes, these clusters modifying complexes11,12, and it is proposed to function
produce hundreds of lncRNAs that show similar in trans through the recruitment of these two repres-
spatiotemporal windows of expression to their neigh- sive complexes to specific target genes (FIG. 1Ba). Indeed,
bouring protein-coding genes12. Some of these lncRNAs the knockdown of HOTAIR in human fibroblasts led to
have been shown to be directly involved in the regulation decreased activity of these repressor complexes and
of HOX genes. to an increase in the expression of HOXD genes. Mouse
models that carry targeted knockout of Hotair have been
Cis-acting lncRNAs. The cis-acting lncRNA HOXA recently produced71. Hotair deletion did not affect via-
distal transcript antisense RNA (HOTTIP), which bility but led to developmental defects and to homeotic
is produced from the 5ʹ end of the human HOXA transformations in the skeletal system71. Consistent with
locus upstream of HOXA13, was identified in human its association with repressive histone-modifying com-
primary fibroblasts. The downregulation of HOTTIP plexes, derepression of several genes, including HoxD
levels in primary fibroblasts induced the transcription components, was reported upon Hotair knockout 72.
of several downstream 5ʹ-HOXA genes. HOTTIP is con- Interestingly, the deletion of the entire mouse HoxC
served in vertebrates, and its knockdown by short hair- locus is perinatally lethal and does not show skeletal
pin RNAs in chick embryos altered limb morphology 30. transformations, whereas individual knockouts of its
The mechanism by which HOTTIP regulates HOXA components, including Hotair, are viable but show devel-
expression relies on its interaction with the activating opmental defects73,74. This may reflect the presence either
histone-modifying MLL1 complex and on the forma- of compensatory mechanisms among members of the
tion of chromatin loops that connect distally expressed HoxD cluster or of genes with functions that are antag-
HOTTIP transcripts with various HOXA gene promot- onistic to Hotair 72, thus emphasizing the importance
ers30 (FIG. 1Ab). Notably, many DNA enhancer elements of using appropriate in vivo models to define lncRNA
produce enhancer RNAs, which might work by a function. Finally, HOTAIR was found to be upregulated
similar mechanism32,33. in different cancers; in breast cancer metastasis, such
An identical mode of action has been described upregulation was shown to result in the re‑targeting of
for mistral lncRNA (Mira), which is a mouse-specific PRC2 to silence tumour suppressor genes75.
lncRNA that is transcribed from the HOXA locus 31.
Mira was identified in retinoic acid-induced dif- Pluripotency versus differentiation commitment
ferentiation of mouse ESCs, in which it positively Several lncRNAs that are associated with pluripotency
controls the transcription of two adjacent genes, have been identified either as species that are induced
HOXA6 and HOXA7. The knockdown of Mira in upon the reprogramming of fibroblasts to induced pluri-
mouse ESCs inhibited the activation of germ layer potent stem cells (iPSCs)76 or as species that are expressed
specification genes, which suggests a role for Mira in in mouse 13,77,78 and human ESCs 14. Notably, these
early mouse ESC differentiation31. However, it is not lncRNA species show expression profiles that corre-
clear how this is linked to the regulation of HOXA6 late well with those of OCT4 (also known as POU5F1),
and HOXA7, as the deletion of these genes in mouse homeobox protein NANOG and sex-determining
Germ layer
embryos indicated that they are involved in later region Y-box 2 (SOX2), which are core components
Primary germ layers (that is,
ectoderm, endoderm and developmental stages69. of the transcriptional network that controls pluripo-
mesoderm) are specified HOX genes are also involved in cell differentiation, tency 78, and the promoters of these lncRNA species
during vertebrate and their deregulation is associated with different types are bound by at least one of these core pluripotency
embryogenesis and, through of human disease, including cancer 67. HOXA transcript transcription factors13. Loss‑of‑function experiments
further differentiation, give
rise to the organs and tissues
antisense RNA myeloid-specific 1 (HOTAIRM1) was resulted in either exit from the pluripotent state or the
of the body. identified as a lncRNA that is produced from the 3ʹ upregulation of lineage commitment gene expression
end of the HOXA locus specifically in myeloid line- programmes, which was comparable to the knockdown
Pluripotency ages70. The knockdown of HOTAIRM1 in myeloid leu- of well-known ESC regulators13,79. Consistent with the
The ability of a cell to
kaemia cell lines inhibited the expression of 3ʹ‑HOXA ‘modular scaffold’ hypothesis13, both the lncRNA‑ES1
differentiate into one of
many cell types. genes by a currently uncharacterized mechanism, (also known as LINC01108) and lncRNA‑ES2 pluripo-
which indicates a positive role in gene expression, tency-associated lncRNAs were found to interact with
Induced pluripotent stem but it did not produce substantial effects on granulo- Polycomb protein SUZ12 and SOX2, which suggests
cells cytic differentiation. By contrast, linc‑Hoxa1 RNA was a model whereby pluripotency-associated lncRNAs
(iPSCs). In vitro-derived
pluripotent cells that originate
recently identified in mouse ESCs, in which it func- function as scaffolds to recruit SUZ12 — part of the
from non-pluripotent cells in a tions in cis to repress Hoxa1 transcription by recruiting repressive PRC2 — to silence neural targets of SOX2 in
process called reprogramming. transcriptional activator protein Pur-β (PURB)71. pluripotent human ESCs14 (FIG. 1Bb).

REVIEWS
Aa An BACE1-AS Among pluripotency-associated lncRNAs, LINC-

Increase in
ORF stability ROR (long intergenic non-protein coding RNA, regu-
AAAAA
lator of reprogramming) is consistently enriched in
human iPSCs, regardless of the cell of origin of these
iPSCs76. LINC-ROR functions as a ceRNA to regulate
Ab An 1/2sbsRNA UPF1 the expression of the core pluripotency transcription
STAU1-mediated
STAU1
mRNA decay
factors by competing for miR‑145 binding 23 (FIG. 3). Thus,
ORF it is a powerful example of how regulatory networks
SBS AAAAA
among transcriptional factors, miRNAs and lncRNAs
are relevant for controlling the alternative fate between
TINCR proliferation and differentiation commitments.
Ac An STAU1 STAU1-mediated
ORF AAAAA mRNA stabilization
Brain and CNS development
Roles have been identified for lncRNAs in establish-
ing and maintaining cell-type-specific gene expression
Ad patterns during organ development. In particular, the
An central nervous system (CNS), which is characterized
Trp53cor1 Inhibition of
RCK by a vast range of neuronal and glial subtypes, is by far
ORF translation
Ribosome AAAAA the most complex and diversified organ in terms of
ncRNAs80. Cells in the CNS show intense transcription
of lncRNAs, and the increase in number of lncRNAs
Ae has been linked to evolutionary complexity. Below, we
Uchl1-as1 Cap-independent
An describe several examples of lncRNAs that have roles in
AAAAA translation
SINE
neurogenesis.
Abundant lncRNAs. Among the most abundant

lncRNAs of the nervous system, metastasis-associated
B circRNAs lung adenocarcinoma transcript 1 (Malat1) is by far the
most extensively studied. This lncRNA was described
AAAAA as a highly expressed species in different types of mouse
IncRNA
neurons81. Malat1 is localized in nuclear speckles and has
AAAAA
been shown to regulate synapse formation by modu-
miRNAs lating a subset of genes that have roles in nuclear and
Pseudogenes
synapse function81. Indeed, the knockdown of Malat1
AAAAA in cultured mouse hippocampal neurons produced
ORF AAAAA decreased synapse density and decreased dendrite
mRNA growth80. However, a loss‑of‑function genetic model of
ORF AAAAA Malat1 indicated that it is not essential for mouse pre-
natal and postnatal development82. These mice showed
Figure 2 | Models of cytoplasmic lncRNA function. A | A common recognition only a minor effect of deregulation of several genes, such
mechanism is through base pairing of complementary regionsNature betweenReviews | Genetics
the long as Malat1‑neighbouring genes, thus indicating a poten-
non-coding RNA (lncRNA) and their target RNA sequence. Aa | Base pairing between tial cis-regulatory role for Malat1 in gene transcription.
specific regions of the human β-site APP-cleaving enzyme 1 (BACE1) mRNA and its Consistent with its original identification in cancer,
antisense transcript BACE1‑AS induces stabilization of the target mRNA and increases the downregulation of MALAT1 by 1,000‑fold in human
BACE1 protein expression44. Ab | Staufen double-stranded RNA-binding protein 1
lung tumour cells, which was achieved by integrating
(STAU1)-mediated mRNA decay is induced when intermolecular base pairing is formed
RNA-destabilizing elements into MALAT1 using zinc-
between an Alu element (or short interspersed element (SINE) in mice) in the 3′
untranslated region of the mRNA and an Alu element within a long half‑STAU1‑binding finger nucleases, revealed a MALAT1‑controlled meta-
site RNA (1/2sbsRNA)46,47. This mRNA decay mechanism also involves the RNA helicase static gene expression programme83. In this setting,
up-frameshift 1 (UPF1). Ac | By contrast, STAU1‑mediated mRNA stabilization has been MALAT1 was found in nuclear speckles and interacted
described in the case of tissue differentiation-inducing non-protein coding RNA with E3 SUMO-protein ligase CBX4 (also known as
(TINCR), which recognizes its target mRNAs through a 25 nucleotide‑long motif45. PC2), which is a component of PRC1. Overall, we still
Antisense recognition has been shown to also control translation. Ad | A repressive await clarification of the exact mechanisms of MALAT1
effect on translation was shown for the targets of tumour protein p53 pathway activity in the distinct processes of tumour progression
corepressor 1 (Trp53cor1) lncRNA43, which functions with the RNA helicase RCK. and neuronal differentiation.
Ae | Translation is induced upon stress induction of ubiquitin carboxy-terminal
hydrolase L1 antisense RNA 1 (Uchl1‑as1)42. B | Base pairing is also the mode of action of
lncRNAs with connections to key neural developmental
competing endogenous RNAs. In this case, however, the complementarity is between
microRNAs (miRNAs) and different targets (including circular RNAs (circRNAs)54,55, protein-coding genes. lncRNAs have been profiled by
lncRNAs21, pseudogene transcripts50 and mRNAs48). The effect of these interactions is complementary genome-wide techniques and in situ
that protein-coding RNAs and non-coding RNAs can crosstalk to each other by hybridizations on different adult mouse brain regions,
competing for miRNA binding through their miRNA recognition motifs. ORF, open which indicated that lncRNAs are associated with
reading frame; SBS, STAU1‑binding site. distinct brain cell types and are expressed in a more

REVIEWS
in mice. These Dlx genes are related to the Drosophila

LIN28 let-7
CPTFs melanogaster Distal-less (Dll) gene; they encode homeo-
miR-134 domain transcription factors that are expressed in the
NANOG Maintenance of pluripotency
developing ventral forebrain and have been postulated
miR-296 to have a role in both forebrain and craniofacial develop-
Reprogramming of iPSCs
miR-470 SOX2 ment. Dlx6os1 controls the expression of Dlx5, Dlx6 and
Differentiation the glutamate decarboxylase 1 gene (Gad1; also known
miR-145
OCT4 as Gad67) through both cis- and trans-acting mecha-
lncRNA-ES1 nisms90. In cis, the transcription of Dlx6os1 negatively
lncRNA-ES2 regulates Dlx6 expression. By contrast, in trans, Dlx6os1
LINC-ROR recruits the transcription factors homeobox protein
Positive feedback
DLX2 (which is an activator) and methyl-CpG-binding
protein 2 (MECP2, which is a repressor) to regulate the
Figure 3 | Pluripotency control. A schematic representation Nature
of the regulatory
Reviews | Genetics expression of Dlx5 and Gad1 (which encodes an enzyme
circuitries that involve the homeobox protein NANOG, sex-determining region Y-box 2 that is responsible for γ-aminobutyric acid (GABA) syn-
(SOX2) and OCT4 core pluripotency transcription factors (CPTFs), microRNAs (miRNAs) thesis)90,91. The loss of Dlx6os1 function in mice produced
and long non-coding RNAs (lncRNAs) in pluripotency control is shown. Several a specific neural phenotype that had reduced numbers of
miRNAs were described as being necessary and sufficient to control self-renewal and
GABAergic interneurons in the early postnatal hippocampus.
pluripotency in human embryonic stem cells or to trigger differentiation120,121 by a direct
link with CPTFs. LINC-ROR (long intergenic non-protein coding RNA, regulator of
Although the number of GABAergic interneurons and
reprogramming) contributes to this circuitry by maintaining high levels of CPTFs — it Gad1 RNA levels returned to normal in the adult hip-
competes for miR‑145 binding through its miR‑145‑recognition motif. When present, pocampus of Dlx6os1 mutants, defects in synaptic inhi-
LINC-ROR prevents miR‑145 from repressing the translation of CPTFs and therefore bition were observed, which indicates a crucial role for
ensures the stem cell fate; when LINC-ROR is downregulated, the synthesis of CPTFs is Dlx6os1 in neuronal activity in vivo90,91. The characteriza-
repressed. Notably, within this circuitry, CPTFs activate their own synthesis through a tion of Dlx6os1 has opened the way to the identification
positive feedback loop, thus reinforcing the regulatory circuit122. Additionally, of a large number of brain lncRNAs that are transcribed
lncRNA‑ES1 and lncRNA‑ES2 contribute to maintaining pluripotency of embryonic stem from UCRs, which can constitute a new class of
cells by repressing SOX2 neural targets14. iPSCs, induced pluripotent stem cells. transcriptional regulators90.
Dlx1os shares some functional properties with
Dlx6os1 in that it regulates, in cis, the antisense Dlx1
tissue-specific manner than mRNAs15,84. Additionally, gene by modulating the level and stability of its tran-
they show specific temporal expression patterns during script 92. The loss of Dlx1os function produced viable
brain development 15,84. and fertile mice that had mild skeletal and neurologi-
Interestingly, transcriptome analysis indicated that cal phenotypes, which essentially replicated a Dlx1
Nuclear speckles many brain-expressed lncRNAs are primate- or human- gain‑of‑function phenotype92.
A class of nuclear body that is specific lncRNAs85; along with the identification of
located in interchromatin
signatures of positive selection and the accelerated lncRNAs as inducers of neurogenesis. A large screen for
regions of the nucleoplasm of
mammalian cells, which are evolution in lncRNA regions, these findings indicate that lncRNAs that are involved in neurogenesis identified
enriched in pre-mRNA splicing lncRNAs may be crucial effectors in human brain evolu- various lncRNAs for which knockdown blocked the
factors. tion and, possibly, in cognitive and behavioural reper- differentiation of human ESCs into mature neurons14.
toires86,87. A lncRNA that has rapidly evolved since the Interestingly, the nuclear localized lncRNA‑N1 (also
Zinc-finger nucleases
Artificial proteins that contain
divergence of humans from the other great apes is highly known as LINC01109) and lncRNA‑N3 were shown to
a zinc-finger DNA-binding accelerated region 1A (HAR1A). Its expression level bind to SUZ12 and REST, which suggests a model in
element fused to an correlates with that of reelin, a protein that is crucial for which neuronal lncRNAs reinforce REST-repressing
endonuclease domain. brain development, which suggests that it could coordi- activity by recruiting PRC2 to specific glial lineage
Double-stranded breaks are
nate the establishment of regional forebrain organization genes, thereby promoting neurogenesis. By contrast,
produced at specific DNA
sequences to induce natural in a similar manner. the cytoplasmic lncRNA‑N2 (also known as MIR100HG)
DNA repair. This strategy By contrast, other classes of brain-expressed lncRNAs seemed to function as precursor molecules for let‑7 and
allows targeted gene deletions, seem to be highly conserved from birds to mammals and for the neurogenic miR‑125b, which are miRNAs that
integrations or modifications. have similar spatiotemporal expression profiles, which promote proliferation arrest and neuronal differentiation,
GABAergic interneurons
indicate ancient roles for these lncRNAs in brain develop- respectively 22,93.
Neurons of the central nervous ment 88. Moreover, brain-expressed lncRNAs that origi-
system that form a connection nate from ultraconserved regions (UCRs) of DNA have lncRNAs in the retina. Interestingly, several lncRNAs
between other types of neurons been shown to be transcribed from complex genetic loci, were found to be specifically expressed in the retina,
and use the neurotransmitter
where they often overlap or are antisense to genes that which is a specialized part of the CNS. The retina is
γ-aminobutyric acid (GABA),
which inhibits excitatory encode key developmental regulator proteins89,90. Such a tractable tissue type for in vivo studies because loss
responses. lncRNAs modulate the activity of their nearby genes by of gene functions can be achieved by locally adminis-
acting as molecular scaffolds to recruit specific factors90,91. tered RNA interference (RNAi) reagents, in contrast to
Hippocampus For example, the co-activator lncRNA Dlx6 opposite germline genetic modifications that are required for
A part of the brain that is
specifically responsible for
strand transcript 1 (Dlx6os1; also known as Evf2)90 is many in vivo studies of the CNS. TABLE 1 shows several
storing and retrieving located in a UCR; it is an antisense RNA to distal-less examples for which a clear function of lncRNAs in retinal
memories. homeobox 6 (Dlx6) and is located downstream of Dlx5 patterning and specification has been established.

REVIEWS
lncRNAs with non-canonical structures. A circRNA importance of null-mutant models for uncovering roles for
that is derived from non-canonical splicing of an anti- lncRNAs95. Although Fendrr was suggested to inter-
sense transcript (CDR1AS; also known as ciRS‑7) to act with components of both repressive chromatin-
the cerebellar degeneration-related protein 1 (CDR1) associated complexes (such as PRC2) and activating
mRNA was recently identified in the human brain53, chromatin-associated complexes (such as MLL1) in
as well as in mouse cortical pyramidal neurons and mouse embryos, chromatin immunoprecipitation (ChIP)
interneurons53. Interestingly, this circRNA functions as analysis following Fendrr deletion showed a change in
a sponge for miR‑7 through 70 selectively conserved occupancy at Fendrr-target genes only for the repressive
miR‑7 target sites, thus regulating endogenous miR‑7 PRC2 (REF. 95). Unlike Bvht, Fendrr has a human orthol-
targets54,55. Zebrafish was used to study the in vivo func- ogous transcript FENDRR that, similarly to murine
tion of this circRNA because it has lost the cdr1 locus Fendrr, is also associated with PRC2 (REF. 24).
while maintaining miR‑7 expression in the embryonic
brain during evolution. Embryos that expressed ectopic Skeletal muscle. One of the first lncRNAs that was
CDR1AS developed brain defects and had a smaller identified with a role in myogenesis was Linc‑MD1
midbrain region, which is similar to the phenotype of (long non-coding RNA, muscle differentiation 1).
the loss of miR‑7 function obtained by treatment with This lncRNA is expressed in a specific temporal win-
morpholino oligonucleotides55. Therefore, circRNAs may dow during in vitro muscle differentiation of mouse
also have roles in neuronal function and in neurological myoblasts and was shown to control the progression
disorders54,55. from early to late phases of muscle differentiation by
In conclusion, the multifaceted functions of lncRNAs functioning as a ceRNA. Through competition for
seem appropriate for the complex regulatory demands the binding of miR‑133 and miR‑135, it regulates the
of the CNS, and further studies of lncRNAs may expression of mastermind-like protein 1 (MAML1)
uncover details of even more complex brain function and myocyte-specific enhancer factor 2C (MEF2C),
and of the pathogenetic events that underlie neuro- which are transcription factors that activate late-dif-
degenerative disorders. However, deeper analyses of ferentiation muscle genes21. LINCMD1 is conserved
the differences that are often found between in vivo between mice and humans 96 , and its expression
and in vitro systems, as well as those between different is strongly reduced in myoblasts of patients with
knockdown strategies, are required for a more reliable Duchenne muscular dystrophy21. Interestingly, in these
understanding of lncRNA functions in the development cells, the recovery of LINCMD1 levels rescued the
of the brain and the CNS. correct timing of in vitro differentiation, which sug-
gests a relevant conserved role in the control of muscle
Development of other organs differentiation21 (FIG. 4).
In addition to extensive roles in brain development, More recently, the imprinted H19 lncRNA, which is
lncRNAs are known to function in the develop- highly expressed in the developing embryo and in adult
ment of diverse organs and tissue types, which are muscle, was shown to work as a ceRNA for let‑7 and to
described below. control muscle differentiation. Indeed, the depletion of
H19 caused precocious muscle differentiation — a phe-
Heart. One of the best examples of the importance notype that is recapitulated by let‑7 overexpression97. As
Morpholino of lncRNAs in organ development is provided by two high let‑7 levels are generally associated with increased
oligonucleotides lncRNAs that are involved in mouse cardiac develop- cellular differentiation, it was hypothesized that H19
Oligonucleotides that are
modified to be highly stable in
ment — braveheart (Bvht; also known as Gm20748)94 inhibits let‑7 activity, thereby preventing precocious
the cell; they are used as and Foxf1 adjacent non-coding developmental regula- differentiation97.
antisense RNA to block cell tory RNA (Fendrr)95. These lncRNAs were identified Another lncRNA that is linked to neuromuscular
components from accessing from the mesoderm, from which the heart originates94,95. disease is D4Z4‑binding element transcript (DBE‑T),
the target site for which they
The knockdown of Bvht by RNAi in mouse ESCs which is selectively expressed in patients with faci-
are designed.
and neonatal cardiomyocyte cultures affected cardiac- oscapulohumeral muscular dystrophy (FSHD). DBE‑T
Chromatin specific gene expression and altered development into recruits histone-lysine N-methyltransferase ASH1L — a
immunoprecipitation mature cardiomyocytes94, thus suggesting a possible component of the MLL1 complex — which results in
(ChIP). A method used to role for Bvht in cardiac tissue regeneration after inju- H3K36 dimethylation and in aberrant transcriptional
determine whether a given
protein binds to, or is localized
ries. Bvht was shown to interact with PRC2, which activation of the FSHMD1A (also known as FSHD)
to, specific chromatin loci suggests that it functions by mediating epigenetic locus in patients with FSHD98.
in vivo. regulation of cardiac commitment 94. Notably, Bvht is Moreover, lncRNAs that regulate gene expres-
specific to mice and is not expressed in rats or humans; sion by driving STAU1-mediated mRNA decay have
Duchenne muscular
whether alternative molecular components carry out also been recently linked to myogenesis — sbsRNAs
dystrophy
A severe genetic disorder that roles that are equivalent to Bvht in other mammals is induce mRNA degradation by recruiting STAU1 to
is characterized by the rapid currently unclear. target mRNAs through base pairing with short inter-
progression of muscle In the case of Fendrr, a 60% reduction of expres- spersed elements (SINEs) in the 3ʹ untranslated region
degeneration, which leads to a sion by RNAi in vivo did not show any apparent phe- of target mRNAs (FIG. 2Ab). Remarkably, downregulating
loss of ambulation and death.
It is due to mutations in the
notypes95. By contrast, the knockout of Fendrr resulted the abundance of three of the four sbsRNAs that were
dystrophin gene that prevent in embryonic lethality owing to impaired heart func- tested altered the rate of mouse myoblast differentiation
its production. tion and to deficits in the body wall, thus indicating the in vitro47.

REVIEWS
a
miR-206, miR-31 Linc-MD1 miR-133, miR-1
Expression level
Differentiation stage
•PAX3 •MEF2C •Dystrophin
•PAX3 •PAX7 •MYF5 •MYOD •Utrophin
•PAX7 •MYF5 •MYOD •Myogenin •Myosin
Progenitors Determination Myoblasts Myocytes Myotubes
b c
PAX7 Self-renewal
miR-206 Mef2c AAAAA
Utrophin
Late myogenesis
Dystrophin miR-135
miR-31
MYF5 Early myogenesis
Linc-MD1 AAAAA
MEF2
miR-1 HDAC4 Differentiation
Myogenin
miR-133
MyoD
MAML1 Differentiation
miR-133
SRF Proliferation Maml1 AAAAA
Figure 4 | ncRNAs and muscle differentiation. a | A schematic representation of the differentiation stages from
Nature
progenitor muscle cells to terminally differentiated fibres is shown. The cells are labelled with the Reviews | Genetics
characteristic proteins
that are expressed at each stage. These include master transcription factors that regulate the switch from one stage to
the following one — such as paired box protein Pax‑3 (PAX3), PAX7, myogenic factor 5 (MYF5), myoblast determination
protein (MYOD), myocyte enhancer factor 2C (MEF2C) and myogenin — as well as the late myogenic proteins dystrophin,
utrophin and myosin123. The graph shows the corresponding temporal expression patterns of selected non-coding RNAs
(ncRNAs). b | MicroRNAs (miRNAs) cooperate with transcription factors to sharpen their temporal expression pattern124;
for example, miR‑206 and miR‑31 repress expression of the self-renewal factor PAX7 and the early myogenic factor
MYF5, respectively. The same miRNAs prevent the early activation of late myogenic proteins, such as utrophin and
dystrophin125. By contrast, late myogenic miRNAs reinforce late differentiation stages; for example, miR‑1 controls the
expression of later myogenic transcription factors MEF2C and myogenin through the repression of histone deacetylase 4
(HDAC4). c | In these circuitries, the role of Linc‑MD1 (long non-coding RNA, muscle differentiation 1) is crucial. It further
reinforces the switch from early to late differentiation gene expression by acting as a ‘sponge’ to limit the repressive
effect of miR‑133 on mastermind-like 1 (Maml1) and of miR‑135 on Mef2c. SRF, serum response factor.
Skin, haematopoietic and adipose development. Roles such as keratin 80 (KRT80), to ensure their expression
for lncRNAs have been identified in the epidermis. and cellular differentiation45 (FIG. 2Ac).
Transcriptome sequencing of progenitor and differenti- Relevant lncRNAs have also been identified in
ating human keratinocytes identified TINCR as the most haematopoiesis and adipogenesis99,100. The analysis of
highly induced lncRNA during keratinocyte differentia- lncRNAs during erythroid differentiation of mouse
tion45. TINCR-deficient epidermis lacked terminal differ- fetal liver progenitors allowed the identification of
entiation ultrastructure, including keratohyalin granules lincRNA-EPS (erythroid prosurvival). The knockdown
and intact lamellar bodies. Interestingly, TINCR also of lincRNA-EPS in mouse erythroid progenitors blocked
binds to STAU1; however, unlike the sbsRNAs described differentiation and promoted apoptosis by inhibiting
above, the TINCR–STAU1 complex targets mRNAs that the expression of the pro-apoptotic PYD and CARD
have a 25‑nucleotide ‘TINCR box’ motif, which results domain-containing gene (Pycard) through a mecha-
in the stabilization of differentiation-associated mRNAs, nism that is still undefined99. More recently, lncRNAs

REVIEWS
were profiled in mice during differentiation to white and comprehension of the structure, function and evolution
brown adipose tissue. Loss‑of‑function studies identified of our genome. Moreover, despite the burst of interest in
ten lncRNAs that have specific roles in adipogenesis100. identifying new lncRNAs and in setting up new meth-
odologies to characterize their function, a future topic
lncRNAs in environmental and stress responses of interest will be the origin and evolution of lncRNAs.
An emerging function for lncRNAs is their contribution One interesting feature relates to the contribution of
to various genetic programmes that enable response to transposable elements to the genesis and regulation of
different environmental conditions. One of the first and lncRNAs18,20. Their relevance is supported by the discov-
best-studied examples is the regulation of flowering ery that, in vertebrates, transposable elements occur in
in plants. In Arabidopsis thaliana, the transcriptional more than two-thirds of mature lncRNAs, whereas they
repressor gene FLOWERING LOCUS C (FLC) has an seldom occur in protein-coding transcripts. Moreover,
important role in this process by blocking the expression transposable elements were found in biased positions
of genes that are required for the switch to flowering. and orientations within lncRNAs, particularly at their
lncRNAs have been shown to function in FLC regula- transcription start sites, which suggests a role in the
tion in various ways101. The long exposure to cold dur- regulation of lncRNA transcription18,20. Therefore, it
ing winter — a process known as vernalization — seems has been proposed that transposable elements may con-
to induce the expression of a sense transcript from FLC tribute to lncRNA evolution and that they function by
called COLD-ASSISTED INTRONIC NON-CODING conferring on lncRNAs tissue-specific expression from
RNA (COLDAIR). COLDAIR is thought to function sim- existing transcriptional regulatory signals18,20.
ilarly to animal lncRNAs in the formation of repressive Phylogenetic analysis is generally one of the first
heterochromatin through a physical association with approaches to be considered when searching for lncRNA
PRC2 (REF. 102). FLC is also regulated by a set of antisense function. However, bioinformatic analysis tools should
lncRNAs called COLD-INDUCED LONG ANTISENSE be implemented to account for the differential evolu-
INTRAGENIC RNA (COOLAIR) that encompass the tionary pressure that operates on the various lncRNA
whole FLC sense transcription unit 101. These antisense subdomains; such pressure acts either on the primary
RNAs are upregulated in response to cold temperatures, sequence of lncRNAs (for antisense effectors against
whereas they are alternatively polyadenylated in warm RNA or DNA targets) or through their secondary struc-
temperatures103. The use of the proximal polyadenylation ture (for protein-binding domains). In this respect, the
site in warm temperatures is linked to histone demeth- modular scaffold hypothesis suggests that lncRNAs have
ylation in the gene body and leads to reduced FLC undergone extensive molecular bricolage by the gain or
transcription104. COOLAIR transcription is repressed loss of different modules, which provides alternative
in warm temperatures by a mechanism that involves and more complex functions that might be subjected
the stabilization of an R‑loop (that is, an RNA–DNA to evolutionary selection8,9,13,14. Moreover, the degree of
hybrid structure) in its promoter region by the NDX1 lncRNA conservation often does not indicate functional
homeobox protein homologue105. relevance; for example, non-coding genes such as XIST
More recently, a novel lncRNA has been identified and nuclear paraspeckle assembly transcript 1 (NEAT1)
in mice as being activated by a stress signalling path- have undergone rapid sequence evolution while preserv-
way that controls the activity of the mammalian target ing their functional roles106,107, and highly accelerated
of rapamycin (mTOR) kinase, which is an important evolution in ncRNA regions has been suggested to con-
regulator of translation42. The lncRNA Uchl1‑as1 is an tribute to the development of complex structures, such
antisense transcript to the neuron-specific Uchl1 gene, as the brain86,87.
which functions in protein ubiquitylation and has roles Another relevant question concerns the non-coding
in brain function and various neurodegenerative dis- definition of a transcript. In fact, it is possible that spe-
eases. Uchl1‑as1 contains an embedded SINEB2 element cific lncRNAs have previously uncharacterized coding
that stimulates Uchl1 translation and thus UCHL1 pro- potential for small peptides (<50 amino acids) with
tein expression under stress conditions42. In particular, biological function. Even if lncRNAs are bound by ribo-
upon stress-induced inhibition of mTOR activity and somes108, it has been recently observed that they show
the resulting repression of cap-dependent translation, patterns of ribosome occupancy that are similar to
Uchl1‑as1 is exported from the nucleus to the cytoplasm, those typical of non-coding sequences, which indicates
where it can base pair with the Uchl1 mRNA and stimu- that this assay is not sufficient to classify transcripts as
late its cap-independent translation. As this activation coding or non-coding 109. Therefore, additional efforts
of UCHL1 expression does not require de novo RNA are required to define the functional implications of the
Phylogenetic analysis synthesis, it provides a rapid response to environmental association between lncRNAs and ribosomes, and to
Comparison of DNA, RNA or
changes. establish whether specific subclasses of lncRNAs with
protein sequences in different
organisms that enables one to coding potential do indeed exist.
establish their evolutionary Conclusions and perspectives Although mechanistic models are starting to emerge,
relationships. The discoveries linked to lncRNA function go far beyond at the core of lncRNA functional studies is the need for
the identification of new mechanisms that regulate gene appropriate model systems for in vivo studies, which
Bricolage
Construction or creation from
expression. The organization of lncRNA-coding loci, should allow a better understanding of the evolution
a diverse range of available which are often finely intertwined with protein-coding and functions of lncRNAs, and their roles in both devel-
things. ones, has added a high degree of complexity in the opment and differentiation. However, owing to the great

REVIEWS
variability in the evolutionary conservation or diversifi- the surrounding locus unaffected. Recent designs for
cation of such RNAs, appropriate animal model systems lncRNA inactivation have successfully used the targeted
are not always available. Notably, in a large screen car- insertion of multiple polyadenylation sites, which pre-
ried out in zebrafish, although many lncRNAs shared vents the transcription of full-length lncRNAs29,91,92,111.
characteristics with their mammalian orthologues, Additional novel strategies need to be developed for
only a few of them had detectable sequence similar- generating suitable conditional and loss‑of‑function
ity 110. Even among mammals, conservation might be model systems for lncRNA studies. An important issue
weak; hence, mouse models might not always reflect to consider when analysing loss‑of‑function phenotypes
functions in humans. Moreover, given the highly cell- of lncRNAs in vivo is the possibility of functional redun-
type-specific expression pattern of many lncRNAs15–17, dancy or of compensatory circuitries that would hide
they are likely to elicit differential developmental or their direct activity, similar to what has been observed
differentiation programmes in different organs, as is for in vivo miRNA depletions22.
the case for MALAT1 (REFS 81–83). Therefore, a more The regulation of lncRNA expression is also a relevant
exhaustive knowledge of their activity in different cells topic that has so far been poorly addressed. Besides tran-
and tissues of the body is required to elucidate possible scriptional control, post-transcriptional regulation will
tissue-specific functions. also be a relevant aspect to investigate. Major issues are
One of the most powerful techniques to study the related to understanding how polyadenylated lncRNAs
function of a gene in vivo is to disrupt its expression are retained in the nucleus and to dissecting which pro-
through targeted recombination. However, this meth- tein interactions control the maturation and subcellular
odology requires special consideration when it is applied localization of lncRNAs. For example, it remains to be
to lncRNA loci — their complex structure and frequent determined how some lncRNAs — such as circRNAs
overlap with other transcripts mean that the disrup- or polyadenylated lncRNAs that overlap with primary
tion of lncRNA loci might interfere with the function of miRNA sequences — accumulate in the cytoplasm21,53,54.
nearby genes, thus confounding the interpretation of the Such lncRNAs are much more abundant than previously
molecular causes of any resultant phenotype. Therefore, thought, and the nature of the cis- and trans-acting fac-
gene targeting should be carefully conceived to ensure tors that regulate their biogenesis and cellular localization
a truncation of the lncRNA of interest while leaving are interesting new issues to be studied.
1. Maeda, N. et al. Transcript annotation in FANTOM3: differentiation; examples of lncRNAs that interact 25. Zhao, J., Sun, B. K., Erwin, J. A., Song, J. J. & Lee, J. T.
mouse gene catalog based on physical cDNAs. PLoS with multiple transcriptional modulators were Polycomb proteins targeted by a short repeat RNA to
Genet. 2, e62 (2006). found, which supports the modular scaffold the mouse X chromosome. Science 322, 750–756
2. Djebali, S. et al. Landscape of transcription in human hypothesis. (2008).
cells. Nature 489, 101–108 (2012). 15. Mercer, T. R., Dinger, M. E., Sunkin, S. M., 26. Nagano, T. et al. The Air noncoding RNA epigenetically
3. Taft, R. J., Pheasant, M. & Mattick, J. S. Mehler, M. F. & Mattick, J. S. Specific expression of silences transcription by targeting G9a to chromatin.
The relationship between nonprotein-coding DNA and long noncoding RNAs in the mouse brain. Proc. Natl Science 322, 1717–1720 (2008).
eukaryotic complexity. Bioessays 29, 288–299 (2007). Acad. Sci. USA 105, 716–721 (2008). 27. Pandey, R. et al. Kcnq1ot1antisense noncoding RNA
4. Mattick, J. S. & Makunin, I. V. Non-coding RNA. 16. Cabili, M. N. et al. Integrative annotation of human mediates lineage-specific transcriptional silencing
Hum. Mol. Genet. 1, 17–29 (2006). large intergenic noncoding RNAs reveals global through chromatin-level regulation. Mol. Cell 32,
5. Rinn, J. L. & Chang, H. Y. Genome regulation by long properties and specific subclasses. Genes Dev. 25, 232–246 (2008).
noncoding RNAs. Annu. Rev. Biochem. 8, 145–166 1915–1927 (2011). 28. Martianov, I., Ramadass, A., Serra Barros, A.,
(2012). 17. Pauli, A. et al. Systematic identification of long Chow, N. & Akoulitchev, A. Repression of the human
6. Derrien, T. et al. The GENCODE v7 catalog of human noncoding RNAs expressed during zebrafish dihydrofolate reductase gene by a non-coding
long noncoding RNAs: analysis of their gene structure, embryogenesis. Genome Res. 22, 577–591 (2012). interfering transcript. Nature 445, 666–670 (2007).
evolution, and expression. Genome Res. 22, 18. Kelley, D. & Rinn, J. Transposable elements reveal a 29. Latos, P. A. et al. Airn transcriptional overlap, but not
1775–1789 (2012). stem cell-specific class of long noncoding RNAs. its lncRNA products, induces imprinted Igf2r silencing.
7. Batista, P. J. & Chang, H. Y. Long noncoding RNAs: Genome Biol. 13, R107 (2012). Science 338, 1469–1472 (2012).
cellular address codes in development and disease. 19. Liang, Q., Xu, Z., Xu, R., Wu, L. & Zheng, S. 30. Wang, K. C. et al. A long noncoding RNA maintains
Cell 152, 1298–1307 (2013). Expression patterns of non-coding spliced active chromatin to coordinate homeotic gene
8. Guttman, M. & Rinn, J. L. Modular regulatory transcripts from human endogenous retrovirus expression. Nature 472, 120–124 (2011).
principles of large non-coding RNAs. Nature 482, HERV‑H elements in colon cancer. PLoS ONE 7, 31. Bertani, S., Sauer, S., Bolotin, E. & Sauer, F.
339–346 (2012). e29950 (2012). The noncoding RNA Mistral activates Hoxa6 and
9. Zappulla, D. C. & Cech, T. R. Yeast telomerase RNA: 20. Kapusta, A. et al. Transposable elements are major Hoxa7 expression and stem cell differentiation
a flexible scaffold for protein subunits. Proc. Natl contributors to the origin, diversification, and by recruiting MLL1 to chromatin. Mol. Cell 43,
Acad. Sci. USA 101, 10024–10029 (2004). regulation of vertebrate long noncoding RNAs. 1040–1046 (2011).
10. Wutz, A., Rasmussen, T. P. & Jaenisch, R. PLoS Genet. 9, e1003470 (2013). 32. Ørom, U. A. et al. Long noncoding RNAs with
Chromosomal silencing and localization are mediated 21. Cesana, M. et al. A long noncoding RNA controls enhancer-like function in human cells. Cell 143,
by different domains of Xist RNA. Nature Genet. 30, muscle differentiation by functioning as a competing 46–58 (2010).
167–174 (2002). endogenous RNA. Cell 147, 358–369 (2011). 33. Li, W. et al. Functional roles of enhancer RNAs for
11. Tsai, M. C. et al. Long noncoding RNA as modular This paper describes a lncRNA that controls the oestrogen-dependent transcriptional activation.
scaffold of histone modification complexes. Science translation of late myogenic factors and the Nature 498, 516–520 (2013).
329, 689–693 (2010). progression to late differentiation stages through 34. Jeon, Y. & Lee, J. T. YY1 tethers Xist RNA to the
12. Rinn, J. L. et al. Functional demarcation of active and competition for common miRNAs. inactive X nucleation center. Cell 146, 119–133
silent chromatin domains in human HOX loci by 22. Ebert, M. S. & Sharp, P. A. Roles for microRNAs in (2011).
noncoding RNAs. Cell 129, 1311–1323 (2007). conferring robustness to biological processes. Cell 35. Schmitz, K. M., Mayer, C., Postepska, A. & Grummt, I.
13. Guttman, M. et al. lincRNAs act in the circuitry 149, 515–524 (2012). Interaction of noncoding RNA with the rDNA promoter
controlling pluripotency and differentiation. Nature 23. Wang, Y. et al. Endogenous miRNA sponge lincRNA- mediates recruitment of DNMT3b and silencing of
477, 295–300 (2011). RoR regulates Oct4, Nanog, and Sox2 in human rRNA genes. Genes Dev. 24, 2264–2269 (2010).
This study reports an exhaustive identification and embryonic stem cell self-renewal. Dev. Cell 25, 69–80 36. Chu, C., Qu, K., Zhong, F. L., Artandi, S. E. &
characterization of lncRNAs that are involved in the (2013). Chang, H. Y. Genomic maps of long noncoding RNA
control of pluripotency and differentiation. This paper describes a lncRNA that controls the occupancy reveal principles of RNA–chromatin
14. Ng, S. Y., Johnson, R. & Stanton, L. W. Human long expression of core pluripotency transcriptional interactions. Mol. Cell 44, 667–678 (2011).
non-coding RNAs promote pluripotency and neuronal factors by competing for common miRNAs. 37. Hung, T. et al. Extensive and coordinated transcription
differentiation by association with chromatin modifiers 24. Khalil, A. M. et al. Many human large intergenic of noncoding RNAs within cell-cycle promoters. Nature
and transcription factors. EMBO J. 31, 522–533 (2012). noncoding RNAs associate with chromatin-modifying Genet. 43, 621–629 (2011).
This paper characterizes several lncRNAs that are complexes and affect gene expression. Proc. Natl 38. Sun, S. et al. Jpx RNA activates Xist by evicting CTCF.
involved in the control of pluripotency and neural Acad. Sci. USA 106, 11667–11672 (2009). Cell 153, 1537–1551(2013).

REVIEWS
39. Wang, X. et al. Induced ncRNAs allosterically modify 63. Yildirim, E. et al. Xist RNA is a potent suppressor 89. Amaral, P. P. et al. Complex architecture and regulated
RNA-binding proteins in cis to inhibit transcription. of hematologic cancer in mice. Cell 152, 727–742 expression of the Sox2ot locus during vertebrate
Nature 454, 126–130 (2008). (2013). development. RNA 11, 2013–2027 (2009).
40. Yang, L. et al. ncRNA- and Pc2 methylation-dependent 64. Fitzpatrick, G. V., Soloway, P. D. & Higgins, M. J. 90. Feng, J. et al. The Evf‑2 noncoding RNA is transcribed
gene relocation between nuclear structures mediates Regional loss of imprinting and growth deficiency in from the Dlx‑5/6 ultraconserved region and functions
gene activation programs. Cell 147, 773–788 (2011). mice with a targeted deletion of KvDMR1. Nature as a Dlx‑2 transcriptional coactivator. Genes Dev. 20,
41. Engreitz, J. M. et al. The Xist lncRNA exploits three- Genet. 32, 426–431 (2002). 1470–1484 (2006).
dimensional genome architecture to spread across the 65. Sleutels, F., Zwart, R. & Barlow, D. P. The non-coding 91. Bond, A. M. et al. Balanced gene regulation by an
X chromosome. Science 341, 1237973 (2013). Air RNA is required for silencing autosomal imprinted embryonic brain ncRNA is critical for adult
42. Carrieri, C. et al. Long non-coding antisense RNA genes. Nature 415, 810–813 (2002). hippocampal GABA circuitry. Nature Neurosci. 12,
controls Uchl1 translation through an embedded 66. Mancini-Dinardo, D., Steele, S. J., Levorse, J. M., 1020–1027 (2009).
SINEB2 repeat. Nature 491, 454–457 (2012). Ingram, R. S. & Tilghman, S. M. Elongation of the 92. Kraus, P. et al. Making sense of Dlx1 antisense RNA.
This study shows that a brain-specific lncRNA Kcnq1ot1 transcript is required for genomic Dev. Biol. 376, 224–235 (2013).
activates translation of an antisense mRNA imprinting of neighboring genes. Genes Dev. 20, 93. Le, M. T. et al. MicroRNA‑125b promotes neuronal
through an embedded repetitive element. 1268–1282 (2006). differentiation in human cells by repressing multiple
43. Yoon, J. H. et al. LincRNA‑p21 suppresses target 67. Colin, J., Libri, D. & Porrua, O. Cryptic transcription targets. Mol. Cell. Biol. 29, 5290–5305 (2009).
mRNA translation. Mol. Cell 47, 648–655 (2012). and early termination in the control of gene 94. Klattenhoff, C. A. et al. Braveheart, a long noncoding
44. Faghihi, M. A. et al. Expression of a noncoding RNA is expression. Genet. Res. Int. 2011, 653494 (2011). RNA required for cardiovascular lineage commitment.
elevated in Alzheimer’s disease and drives rapid feed- 68. Barber, B. A. & Rastegar, M. Epigenetic control of Hox Cell 152, 1–14 (2013).
forward regulation of β-secretase. Nature Med. 14, genes during neurogenesis, development, and disease. 95. Grote, P. et al. The tissue-specific lncRNA Fendrr is an
723–730 (2008). Ann. Anat. 192, 261–274 (2010). essential regulator of heart and body wall development
45. Kretz, M. et al. Control of somatic tissue 69. Kostic, D. & Capecchi, M. R. Targeted disruptions of in the mouse. Dev. Cell 24, 206–214 (2013).
differentiation by the long non-coding RNA TINCR. the murine Hoxa‑4 and Hoxa‑6 genes result in References 94 and 95 highlight the importance of
Nature 493, 231–235 (2013). homeotic transformations of components of the lncRNAs during lineage commitment and in
This paper shows that STAU1 can be tethered by a vertebral column. Mech. Dev. 46, 231–247 (1994). providing a new layer of regulation that is involved
lncRNA on specific target mRNAs that share a 70. Zhang, X. et al. A myelopoiesis-associated regulatory in determining cardiac cell fate.
25‑nucleotide conserved box. However, different intergenic noncoding RNA transcript within the human 96. Twayana, S. et al. Biogenesis and function of
from STAU1‑mediated decay, such tethering HOXA cluster. Blood 113, 2526–2534 (2009). non-coding RNAs in muscle differentiation and in
confers stability on the target mRNAs. 71. Maamar, H., Cabili, M.N., Rinn, J. & Raj, A. Duchenne muscular dystrophy. Bioch. Soc. Trans. 41,
46. Gong, C. & Maquat, L. E. lncRNAs transactivate linc‑HOXA1 is a noncoding RNA that represses Hoxa1 844–849 (2013).
STAU1‑mediated mRNA decay by duplexing with 3′ transcription in cis. Genes Dev. 27, 1260–1271 97. Kallen, A. N. et al. The imprinted H19 lncRNA
UTRs via Alu elements. Nature 470, 284–288 (2011). (2013). antagonizes let‑7 microRNAs. Mol. Cell 52, 101–112
47. Wang, J., Gong, C. & Maquat, L. E. Control of 72. Li, L. et al. Targeted disruption of Hotair leads to (2013).
myogenesis by rodent SINE-containing lncRNAs. homeotic transformation and gene derepression. Cell 98. Cabianca, D. S. et al. A long ncRNA links copy number
Genes Dev. 27, 793–804 (2013). Rep. 5, 3–12 (2013). variation to a Polycomb/Trithorax epigenetic switch in
48. Salmena, L., Poliseno, L., Tay, Y., Kats, L. & 73. Suemori, H. & Noguchi, S. Hox C cluster genes are FSHD muscular dystrophy. Cell 149, 819–831
Pandolfi, P. P. The ceRNA hypothesis: the Rosetta dispensable for overall body plan of mouse embryonic (2012).
stone of a hidden RNA language. Cell 146, 353–358 development. Dev. Biol. 220, 333–342 (2000). 99. Hu, W., Yuan, B., Flygare, J. & Lodish, H. F.
(2011). 74. Schorderet, P. & Duboule, D. Structural and Long noncoding RNA-mediated anti-apoptotic activity
49. Franco-Zorrilla, J. M. et al. Target mimicry provides a functional differences in the long non-coding in murine erythroid terminal differentiation. Genes
new mechanism for regulation of microRNA activity. RNA Hotair in mouse and human. PLoS Genet. 7, Dev. 25, 2573–2578 (2011).
Nature Genet. 39, 1033–1037 (2007). e1002071 (2011). 100. Sun, L. et al. Long noncoding RNAs regulate
50. Poliseno, L. et al. A coding-independent function of 75. Gupta, R. A. et al. Long non-coding RNA HOTAIR adipogenesis. Proc. Natl Acad. Sci. USA 110,
gene and pseudogene mRNAs regulates tumour reprograms chromatin state to promote cancer 3387–3392 (2013).
biology. Nature 465, 1033–1038 (2010). metastasis. Nature 464, 1071–1076 (2010). 101. Ietswaart, R., Wu, Z. & Dean, C. Flowering time
This paper proposes the ceRNA hypothesis, 76. Loewer, S. et al. Large intergenic non-coding RNA-RoR control: another window to the connection between
in which coding and non-coding RNAs can modulates reprogramming of human induced antisense RNA and chromatin. Trends Genet. 28,
crosstalk through competition for shared pluripotent stem cells. Nature Genet. 42, 1113–1117 445–453 (2012).
miRNA-binding motifs. (2010). 102. Heo, J. B. & Sung, S. Vernalization-mediated
51. Karreth, F. A. et al. In vivo identification of 77. Guttman, M. et al. Chromatin signature reveals over a epigenetic silencing by a long intronic noncoding RNA.
tumor-suppressive PTEN ceRNAs in an oncogenic thousand highly conserved large non-coding RNAs in Science 331, 76–79 (2011).
BRAF-induced mouse model of melanoma. Cell 147, mammals. Nature 458, 223–227 (2009). 103. Liu, F. et al. The Arabidopsis RNA-binding protein
382–395 (2011). 78. Dinger, M. E. et al. Long noncoding RNAs in mouse FCA requires a lysine-specific demethylase 1
52. Sumazin, P. et al. An extensive microRNA-mediated embryonic stem cell pluripotency and differentiation. homolog to downregulate FLC. Mol. Cell 28,
network of RNA–RNA interactions regulates Genome Res. 9, 1433–1445 (2008). 398–407 (2007).
established oncogenic pathways in glioblastoma. 79. Sheik Mohamed, J., Gaughwin, P. M., Lim, B., 104. Liu, F., Marquardt, S., Lister, C., Swiezewski, S. &
Cell 147, 370–381 (2011). Robson, P. & Lipovich, L. Conserved long noncoding Dean, C. Targeted 3ʹ processing of antisense
53. Hansen, T. B. et al. miRNA-dependent gene silencing RNAs transcriptionally regulated by Oct4 and Nanog transcripts triggers Arabidopsis FLC chromatin
involving Ago2‑mediated cleavage of a circular modulate pluripotency in mouse embryonic stem cells. silencing. Science 327, 94–97 (2010).
antisense RNA. EMBO J. 30, 4414–4422 (2011). RNA 16, 324–337 (2010). 105. Sun, Q., Csorba, T., Skourti-Stathaki, K.,
54. Hansen, T. B. et al. Natural RNA circles function as 80. Qureshi, I. A. & Mehler, M. F. Emerging roles of non- Proudfoot, N. J. & Dean, C. R‑loop stabilization
efficient microRNA sponges. Nature 495, 384–388 coding RNAs in brain evolution, development, represses antisense transcription at the Arabidopsis
(2013). plasticity and disease. Nature Rev. Neurosci. 13, FLC locus. Science 340, 619–621 (2013).
55. Memczak, S. et al. Circular RNAs are a large class of 528–541 (2012). 106. Clemson, C. M. et al. An architectural role for a
animal RNAs with regulatory potency. Nature 495, 81. Bernard, D. et al. A long nuclear-retained non-coding nuclear noncoding RNA: NEAT1 RNA is essential for
333–338 (2013). RNA regulates synaptogenesis by modulating gene the structure of paraspeckles. Mol. Cell 33, 717–726
This paper identifies a large class of cellular expression. EMBO J. 29, 3082–3093 (2010). (2009).
circRNAs; for one of these, a ceRNA function is 82. Zhang, B. et al. The lncRNA Malat1 is dispensable 107. Nesterova, T. B. et al. Characterization of the genomic
shown in the control of neuronal functions. for mouse development but its transcription plays a Xist locus in rodents reveals conservation of overall
56. Lee, J. T. & Bartolomei, M. S. X‑inactivation, cis-regulatory role in the adult. Cell Rep. 2, 111–123 gene structure and tandem repeats but rapid
imprinting, and long noncoding RNAs in health and (2012). evolution of unique sequence. Genome Res. 11,
disease. Cell 152, 1308–1323 (2013). 83. Gutschner, T. et al. The noncoding RNA MALAT1 is a 833–849 (2001).
57. Penny, G. D., Kay, G. F., Sheardown, S. A., Rastan, S. & critical regulator of the metastasis phenotype of lung 108. Ingolia, N. T., Brar, G. A., Rouskin, S., McGeachy, A. M.
Brockdorff, N. Requirement for Xist in X chromosome cancer cells. Cancer Res. 73, 1180–1189 (2013). & Weissman, J. S. The ribosome profiling strategy for
inactivation. Nature 379, 131–137 (1996). 84. Ramos, A. D. et al. Integration of genome-wide monitoring translation in vivo by deep sequencing of
58. Schoeftner, S. et al. Recruitment of PRC1 function at approaches identifies lncRNAs of adult neural stem ribosome-protected mRNA fragments. Nature Protoc.
the initiation of X inactivation independent of PRC2 cells and their progeny in vivo. Cell Stem Cell. 12, 7, 1534–1550 (2012).
and silencing. EMBO J. 25, 3110–3122 (2006). 616–628 (2013). 109. Guttman, M., Russell, P., Ingolia, N. T., Weissman, J. S.
59. Nozawa, R. S. et al. Human inactive X chromosome is 85. Xu, A. G. et al. Intergenic and repeat transcription & Lander, E. S. Ribosome profiling provides evidence
compacted through a PRC2‑independent in human, chimpanzee and macaque brains that large noncoding RNAs do not encode proteins.
SMCHD1‑HBiX1 pathway. Nature Struct. Mol. Biol. measured by RNA-seq. PLoS Comput. Biol. 6, Cell 154, 240–251 (2013).
20, 566–573 (2013). e1000843 (2010). 110. Ulitsky, I., Shkumatava, A., Jan, C. H., Sive, H. &
60. Lee, J. T. Disruption of imprinted X inactivation by 86. Pollard, K. S. et al. An RNA gene expressed during Bartel, D. P. Conserved function of lincRNAs in
parent-of‑origin effects at Tsix. Cell 103, 17–27 cortical development evolved rapidly in humans. vertebrate embryonic development despite rapid
(2000). Nature 443, 167–172 (2006). sequence evolution. Cell 147, 1537–1550 (2011).
61. Sado, T., Wang, Z., Sasaki, H. & Li, E. Regulation of 87. Lindblad-Toh, K. et al. A high-resolution map of human 111. Ohhata, T., Hoki, Y., Sasaki, H. & Sado, T. Crucial role
imprinted X‑chromosome inactivation in mice by Tsix. evolutionary constraint using 29 mammals. Nature of antisense transcription across the Xist promoter in
Development 128, 1275–1286 (2001). 478, 476–482 (2011). Tsix-mediated Xist chromatin modification.
62. Tian, D., Sun, S. & Lee, J. T. The long noncoding RNA, 88. Dorus, S. et al. Accelerated evolution of nervous Development 135, 227–235 (2008).
Jpx, is a molecular switch for X chromosome system genes in the origin of Homo sapiens. Cell 119, 112. Rinn, J. L. et al. The transcriptional activity of human
inactivation. Cell 143, 390–403 (2010). 1027–1040 (2004). chromosome 22. Genes Dev. 17, 529–540 (2003).

REVIEWS
113. Velculescu, V. E., Zhang, L., Vogelstein, B. & 120. Tay, Y., Zhang, J., Thomson, A. M., Lim, B. & 127. Rapicavoli, N. A., Poth, E. M., Zhu, H. & Blackshaw, S.
Kinzler, K. W. Serial analysis of gene expression. Rigoutsos, I. MicroRNAs to Nanog, Oct4 and The long noncoding RNA Six3OS acts in trans to
Science 270, 484–487 (1995). Sox2 coding regions modulate embryonic stem regulate retinal development by modulating Six3
114. Wei, C. L. et al. 5ʹ long serial analysis of gene cell differentiation. Nature 455, 1124–1128 activity. Neural Dev. 6, 32 (2011).
expression (LongSAGE) and 3ʹ LongSAGE for (2008). 128. Young, T. L., Matsuda, T. & Cepko, C. L. The noncoding
transcriptome characterization and genome 121. Melton, C., Judson, R. L. & Blelloch, R. RNA taurine upregulated gene 1 is required for
annotation. Proc. Natl Acad. Sci. USA 101, Opposing microRNA families regulate self-renewal differentiation of the murine retina. Curr. Biol. 15,
11701–11706 (2004). in mouse embryonic stem cells. Nature 463, 501–512 (2005).
115. Matsumura, H., Krüger, D. H., Kahl, G. & Terauchi, R. 621–626 (2010). 129. Meola, N., Pizzo, M., Alfano, G., Surace, E. M. &
SuperSAGE: a modern platform for genome-wide 122. Boyer, L. A. et al. Core transcriptional regulatory Banfi, S. The long noncoding RNA Vax2os1 controls
quantitative transcript profiling. Curr. Pharm. circuitry in human embryonic stem cells. Cell 122, the cell cycle progression of photoreceptor progenitors
Biotechnol. 9, 368–374 (2008). 947–956 (2005). in the mouse retina. RNA 18, 111–123 (2012).
116. Shiraki, T. et al. Cap analysis gene expression 123. Buckingham, M. & Vincent, S. D. Distinct and
for high-throughput analysis of transcriptional dynamic myogenic populations in the vertebrate Acknowledgements
starting point and identification of promoter usage. embryo. Curr. Opin. Genet. Dev. 19, 444–453 The authors thank members of the Bozzoni and Fatica labo-
Proc. Natl Acad. Sci. USA 100, 15776–15781 (2009). ratories for discussion, and J. Rinn, C. Dean and P. Avner for
(2003). 124. Marrone, A. K. & Shcherbata, H. R. Dystrophin their suggestions. They apologize for papers that are not dis-
117. Mortazavi, A., Williams, B. A., McCue, K., Schaeffer, L. orchestrates the epigenetic profile of muscle cells via cussed owing to space limitations. They acknowledge support
& Wold, B. Mapping and quantifying mammalian miRNAs. Front. Genet. 2, 64–72 (2011). from FP7‑PEOPLE‑2011‑ITN Project HemID (289611),
transcriptomes by RNA-seq. Nature Methods 5, 125. Cacchiarelli, D. et al. miR‑31 modulates dystrophin Telethon (GPP11149), Parent Project Italia, Associazione
621–628 (2008). expression: new implications for Duchenne Italiana per la Ricerca sul Cancro, Italian Institute of
118. Wahl, M. C., Will, C. L. & Lührmann, R. muscular dystrophy therapy. EMBO Rep. 12, Technology “SEED”, Fondo per gli Investimenti per la Ricerca
The spliceosome: design principles of a dynamic 136–141 (2011). di Base, Programmi di Ricerca di Interesse Nazionale and the
RNP machine. Cell 136, 701–718 (2009). 126. Rapicavoli, N. A., Poth, E. M. & Blackshaw, S. EPIGEN Project.
119. Kiss, T. Small nucleolar RNAs: an abundant group of The long noncoding RNA RNCR2 directs mouse
noncoding RNAs with diverse cellular functions. Cell retinal cell specification. BMC Dev. Biol. 10, 49 Competing interests statement
109, 145–148 (2002). (2010). The authors declare no competing interests.

Fatica 2013

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

Fatica 2013

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fatica 2013

Uploaded by

Copyright:

Available Formats

REVIEWS

Long non-coding RNAs: new players in

NATURE REVIEWS | GENETICS VOLUME 15 | JANUARY 2014 | 7

© 2014 Macmillan Publishers Limited. All rights reserved

Box 1 | Methodologies for lncRNA identification and analyses

8 | JANUARY 2014 | VOLUME 15 www.nature.com/reviews/genetics

© 2014 Macmillan Publishers Limited. All rights reserved

Box 2 | Examples of ncRNAs with guiding activity for proteins

Cytoplasmic lncRNAs. Many lncRNA-mediated mech-

tive regulation by tumour protein p53 pathway core-

NATURE REVIEWS | GENETICS VOLUME 15 | JANUARY 2014 | 9

© 2014 Macmillan Publishers Limited. All rights reserved

A Cis-acting IncRNAs Repressive histone modiﬁcation

Aa Xist, Kcnq1ot1 and Airn Ab HOTTIP

Bb IncRNA-ES1 and IncRNA-ES2

10 | JANUARY 2014 | VOLUME 15 Nature Reviews | Genetics

© 2014 Macmillan Publishers Limited. All rights reserved

NATURE REVIEWS | GENETICS VOLUME 15 | JANUARY 2014 | 11

© 2014 Macmillan Publishers Limited. All rights reserved

Table 1 | lncRNA manipulation and resulting phenotypes in model animal systems

12 | JANUARY 2014 | VOLUME 15 www.nature.com/reviews/genetics

© 2014 Macmillan Publishers Limited. All rights reserved

Regulation of HOX genes Trans-acting lncRNAs. HOX transcript antisense RNA

NATURE REVIEWS | GENETICS VOLUME 15 | JANUARY 2014 | 13

© 2014 Macmillan Publishers Limited. All rights reserved

Aa An BACE1-AS Among pluripotency-associated lncRNAs, LINC-

Abundant lncRNAs. Among the most abundant

14 | JANUARY 2014 | VOLUME 15 www.nature.com/reviews/genetics

© 2014 Macmillan Publishers Limited. All rights reserved

in mice. These Dlx genes are related to the Drosophila

NATURE REVIEWS | GENETICS VOLUME 15 | JANUARY 2014 | 15

© 2014 Macmillan Publishers Limited. All rights reserved

16 | JANUARY 2014 | VOLUME 15 www.nature.com/reviews/genetics

© 2014 Macmillan Publishers Limited. All rights reserved

Progenitors Determination Myoblasts Myocytes Myotubes

NATURE REVIEWS | GENETICS VOLUME 15 | JANUARY 2014 | 17

© 2014 Macmillan Publishers Limited. All rights reserved

18 | JANUARY 2014 | VOLUME 15 www.nature.com/reviews/genetics

© 2014 Macmillan Publishers Limited. All rights reserved

NATURE REVIEWS | GENETICS VOLUME 15 | JANUARY 2014 | 19

© 2014 Macmillan Publishers Limited. All rights reserved

20 | JANUARY 2014 | VOLUME 15 www.nature.com/reviews/genetics

© 2014 Macmillan Publishers Limited. All rights reserved

NATURE REVIEWS | GENETICS VOLUME 15 | JANUARY 2014 | 21

© 2014 Macmillan Publishers Limited. All rights reserved

You might also like