ArticlePDF AvailableLiterature Review

Discovering functional motifs in long noncoding RNAs

January 2022
WIREs RNA 13(1)

DOI:10.1002/wrna.1708

Authors:

Caroline Ross

University of Cape Town

Igor Ulitsky

Weizmann Institute of Science

Long noncoding RNAs (lncRNAs) are products of pervasive transcription that closely resemble messenger RNAs on the molecular level, yet function through largely unknown modes of action. The current model is that the function of lncRNAs often relies on specific, typically short, conserved elements, connected by linkers in which specific sequences and/or structures are less important. This notion has fueled the development of both computational and experimental methods focused on the discovery of functional elements within lncRNA genes, based on diverse signals such as evolutionary conservation, predicted structural elements, or the ability to rescue loss‐of‐function phenotypes. In this review, we outline the main challenges that the different methods need to overcome, describe the recently developed approaches, and discuss their respective limitations. This article is categorized under: RNA Evolution and Genomics > Computational Analyses of RNA RNA Interactions with Proteins and Other Molecules > Protein‐RNA Interactions: Functional Implications Regulatory RNAs/RNAi/Riboswitches > Regulatory RNAs

Content uploaded by Caroline Ross

Content may be subject to copyright.

ADVANCED REVIEW

Discovering functional motifs in long noncoding RNAs

Caroline Jane Ross | Igor Ulitsky

Biological Regulation and Molecular

Neuroscience, Weizmann Institute of

Science, Rehovot, Israel

Correspondence

Igor Ulitsky, Biological Regulation and

Molecular Neuroscience, Weizmann

Institute of Science, Rehovot 76100, Israel.

Email: igor.ulitsky@weizmann.ac.il

Funding information

German-Israel foundation for Scientific

Research and Development, Grant/Award

Number: I-1455-417.13/2018; Israel

Science Foundation, Grant/Award

Numbers: 2406/18, 852/19; MOST-PRC

program; The EU Joint Programme -

Neurodegenerative Disease Research

(JPND ERA-Net localMND)

Edited by: Jeff Wilusz, Editor-in-Chief

Abstract

Long noncoding RNAs (lncRNAs) are products of pervasive transcription that

closely resemble messenger RNAs on the molecular level, yet function through

largely unknown modes of action. The current model is that the function of

lncRNAs often relies on specific, typically short, conserved elements, connected by

linkers in which specific sequences and/or structures are less important. This

notion has fueled the development of both computational and experimental

methods focused on the discovery of functional elements within lncRNA genes,

based on diverse signals such as evolutionary conservation, predicted structural ele-

ments, or the ability to rescue loss-of-function phenotypes. In this review, we out-

line the main challenges that the different methods need to overcome, describe the

recently developed approaches, and discuss their respective limitations.

This article is categorized under:

RNA Evolution and Genomics > Computational Analyses of RNA

RNA Interactions with Proteins and Other Molecules > Protein-RNA Inter-

actions: Functional Implications

Regulatory RNAs/RNAi/Riboswitches > Regulatory RNAs

KEYWORDS

computational biology, long noncoding RNA, RNA-protein interactions

1|INTRODUCTION

Over the past decade, progress in high throughput sequencing and genome-wide transcriptome profiling have revealed

an unprecedented landscape of long noncoding RNAs (lncRNAs) in animal and plant genomes (J. B. Brown et al., 2014;

Cabili et al., 2011; Clark & Mattick, 2011; Guttman et al., 2009; Iyer et al., 2015; Necsulea et al., 2014; Sarropoulos

et al., 2019). Only a fraction of these lncRNAs have been studied and those have been attributed with regulatory func-

tions in various cell processes and occasionally implicated in human disease (Guo et al., 2019; Hezroni et al., 2019; Liu

et al., 2021; Statello et al., 2020). LncRNA synthesis is similar to that of mRNAs in that they are transcribed by RNA

polymerase II (Pol II), undergo splicing and are capped and polyadenylated. The classification of lncRNAs is largely

dependent on exclusion criteria and typically encompasses transcripts of at least 200 nucleotides (nt) that do not exhibit

potential to encode functional proteins and do not belong to any other well-defined group of noncoding RNAs, such as

ribosomal RNA (rRNA). Thus, the currently annotated lncRNAs comprise a hash of thousands of genes with diverse

properties, possible functions, and mechanisms for which the principles of subclassification have not yet been eluci-

dated. A striking difference between lncRNA genes and coding genes, is the rapid turnover of lncRNAs during evolu-

tion. This has underpinned earlier assumptions that the majority of lncRNAs may not be functional. However, studies

on specific lncRNAs have shown that function can be maintained across distantly related species, even in cases where

Received: 1 September 2021 Revised: 19 November 2021 Accepted: 4 December 2021

DOI: 10.1002/wrna.1708

https://doi.org/10.1002/wrna.1708

the sequences have substantially diverged (reviewed in Ulitsky, 2016). As the evidence continues to mount that a siz-

able number of lncRNAs are relevant to human disease, it is becoming more important to develop methods that can

rapidly distinguish which lncRNAs are likely to be functional. The understanding of how lncRNA functions have been

maintained across sequences that have extensively changed is fundamental to deriving principles for the large-scale

classification of these genes into functional groups. This classification will allow us to transfer functional knowledge

across lncRNAs with similar properties, and will guide the design of experimental studies to systematically probe their

numerous biological mechanisms.

The first studies that delved into mapping the homology of lncRNA sequences relied on alignment-based algo-

rithms, which were successful in comparison of closely related species. However, these algorithms assume the equiva-

lence between homology and nucleotide sequence similarity, which is problematic for the comparison of lncRNA

sequences that have largely diverged. Nevertheless, comparisons of lncRNAs between human and other mammals sug-

gest that lncRNA sequences often contain short conserved patches that evolve under purifying selection and that these

patches are disguised by their surrounding sequence context which evolves rapidly and thus seems to have limited con-

tribution to function. This model is attractive because it is sufficient for supporting interactions of lncRNAs with RNA

binding proteins (RBPs) or microRNAs (miRNAs), which has previously been shown to modulate lncRNA function and

to be regulated by them (Bitetti et al., 2018; Kleaveland et al., 2018; Ulitsky et al., 2011; Xue et al., 2016). More recent

studies have expanded on this idea, and implemented a combination of computational and experimental approaches to

uncover functional elements in an array of lncRNA genes. This concept has also underpinned the recent development

of novel alignment-free approaches which have advanced the capability to identify and compare orthologous lncRNAs

in more distal species.

In this review, we summarize the recent advancements in the computational and experimental approaches to

uncover functional motifs in lncRNA genes. We begin with a brief description of the challenges that have hindered

lncRNA annotation across diverse species and the comparative analysis of their sequences, with a particular focus on

considerations for the unbiased identification of lncRNA orthologs in distal species. We then discuss the fundamental

traits of lncRNA evolution that have been uncovered from early studies, and go on to describe how these traits have

been leveraged to develop new algorithms that present more sensitive approaches for the functional classification of

lncRNAs. Last, we give an overview of the experimental methods that have been developed to validate the functional

importance of sequence and structural motifs in lncRNA sequences.

2|THE MAIN CHALLENGES THAT HAVE STALLED COMPARATIVE

GENOMICS OF lncRNAs

Functional conservation is a well-defined evolutionary constraint that has been successfully leveraged to annotate pro-

tein coding genes (PCGs) and other noncoding RNAs (Bartel, 2009; Michel & Westhof, 1990; Woese et al., 1980). How-

ever, the application of comparative genomics to lncRNAs has been hindered by two major obstacles. The first obstacle

stems from the rapid evolution of lncRNA sequences. Conventional algorithms that were developed for PCGs are based

on multiple sequence alignments (MSAs). These can be derived from whole genome alignments (WGAs), but those

require extensive and pairwise-significant sequence similarity between the studied genomes, and much or all of the

sequence of a given lncRNA will often not have informative MSAs. Such methods thus perform poorly in identifying

and comparing orthologous lncRNAs that lack long continuous regions that are highly constrained at the sequence

level. The analysis of CHASERR lncRNA sequences from distantly related species presents an example of this problem

(Figure 1a). CHASERR is transcribed in close proximity to the transcription start site (TSS) of CHD2 and acts in cis to

regulate expression levels of CHD2 (Rom et al., 2019). This regulatory mechanism has been shown to be essential for

mouse viability (Rom et al., 2019). Analysis of RNA-seq data clearly indicates that CHASERR has syntenic counterparts

that are transcribed in species as far as zebrafish (>400 million years of evolution; Figure 1a), however BLASTN does

not detect significant alignment (E-value <0.001) between human CHASERR and species beyond amniotes (Ross

et al., 2021). An alternative to using MSAs is to directly compare sequences of lncRNAs across species by pairwise

sequence alignment. However, the number of lncRNAs that have been annotated remains very limited in most verte-

brate species. This is a consequence of both incomplete genome sequences or partial annotations of protein-coding

genes, and the limited accuracy of algorithms for reconstruction of transcripts from RNA-seq data. In some cases this

reduces the ability to identify orthologous lncRNAs, while in other cases it can lead to the false identification of a

lncRNA that appears to stand alone, but actually aligns with a region of a PCG that has not been fully annotated

2of25 ROSS AND ULITSKY

17577012, 2022, 4, Downloaded from https://wires.onlinelibrary.wiley.com/doi/10.1002/wrna.1708 by University Of Cape Town, Wiley Online Library on [09/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

FIGURE 1 Dimensions of lncRNA evolution that have been leveraged for comparative sequence analysis. (a) Genomic locus of Chaserr

across vertebrate species. RNA-seq data obtained from HPA and SRA. PhyloP scores indicate low sequence conservation, despite synteny.

Genomic gaps in respective species and high GC content of the region are shown to highlight potential challenges in identification of

lncRNA orthologs in this region. Zebrafish Chaserr reconstructed by PLAR. Short conserved motifs identified from lncLOOM analysis of

Chaserr sequences from 16 vertebrate species (ordered from Human to Zebrafish). Graded brown color indicates conservation per number of

species. (b) Computational approaches to identify functional motifs in lncRNA sequences: A composition-based approach (as used in Seekr)

vs a conservation based approach (as used in lncLOOM). (c) Schematic illustration of common secondary structures in RNA sequences

ROSS AND ULITSKY 3of25

(discussed in Chen et al., 2016). The issue can be partially addressed by using programs such as Trinity (Grabherr

et al., 2011) to reconstruct the transcriptome de novo. However, these tools also have their limitations, especially in the

reconstruction of full transcripts from short RNA-seq reads (Hölzer & Marz, 2019). These shortcomings have left sub-

stantial parts of the transcriptomes from many vertebrate species unexplored. This has been a major incentive to

develop more sensitive algorithms that can confidently predict orthologous genes by directly comparing transcripts to

annotated lncRNAs from a query species (discussed below). Another option to overcome the challenges in d§e novo

transcriptome assembly lies in long-read sequencing technologies, which have advanced substantially in recent years

(Amarasinghe et al., 2020; Oikonomopoulos et al., 2020) and have already been used in transcriptome assembly in

recent studies (Müller et al., 2021; Y. H. Sun et al., 2021). Although there are still several limitations related to the high

error rate of long-read sequencing, we expect that it will become more applicable to transcriptome assembly as the tech-

nologies continue to advance and improved analysis methods for error correction are developed.

Faced with these challenges, the first studies that began to explore the evolutionary trajectories of lncRNAs were

based on aligning (with BLASTN, BLASTZ, and LastZ) lncRNAs within clades of closely related species, particularly

across mammals (Bu et al., 2015; Carninci & Hayashizaki, 2007; Kutter et al., 2012; Washietl et al., 2014). Although

these studies were limited in that they could miss parts of the conservation because of the challenges mentioned above,

they shed light on sequence and structural characteristics that are shared between lncRNAs. These characteristics can

now be leveraged to establish more sensitive computational approaches for comparison between distantly related spe-

cies (discussed in section 3). Importantly, these studies also established standards for the analysis of lncRNA sequences

and revealed fundamental considerations that are important to avoid ascertainment bias of orthologous lncRNA identi-

fication (Hezroni et al., 2015).

2.1 |Considerations for the unbiased comparison of orthologous lncRNA genes

The majority of studies have investigated lncRNA evolution by projecting human lncRNA sequences across WGAs to

identify candidate loci where the DNA sequence is conserved and also transcribed in distal species (Bu et al., 2015;

Hezroni et al., 2015; Kutter et al., 2012; Necsulea et al., 2014; Washietl et al., 2014). These studies have largely focused

on long intergenic noncoding RNAs (lincRNAs), which are lncRNAs that do not overlap PCGs, and are therefore easier

to identify. In comparison to coding genes, lncRNA expression is notably more tissue-specific (Cabili et al., 2011) and

can also vary significantly between individuals within a population (Kornienko et al., 2016). Taking this into account,

the comprehensive catalogs of lncRNA transcripts for various species were typically reconstructed from RNA-seq data

from multiple tissues. There are two conflicting caveats in the WGA-based approach. From one side, lncRNA conserva-

tion may be underestimated if the lncRNAs in other species are not detected due to the spatiotemporal nature of

lncRNA expression (Cabili et al., 2011; Chodroff et al., 2010; Mor

an et al., 2012; X.-Q. Zhang et al., 2017) or because

reconstruction algorithms fail to properly annotate them (e.g., a lncRNA transcript is merged to a PCG next to it, and

not annotated as a lncRNA). This may be partially circumvented by carefully comparing expression levels from multiple

tissue types or combining RNA-seq data with 3P-seq data (Hezroni et al., 2015). As an alternative, Necsulea et al., 2014

increased their estimates of the evolutionary age of lncRNA families by including species for which, according to a

probabilistic likelihood, the absence of transcription could be attributed to read coverage or exonic length. Although

this improved the sensitivity, it presents potential bias in the other direction, as the assumption that if a lncRNA is tran-

scribed in some species, all sequences homologous to it in other species are also transcribed, is often too strong. Instead,

it is possible that genomic loci may be conserved due to other constraints. This potential bias is illustrated through close

inspection of the Sox21 loci in distal vertebrate species. In humans, the 20 kb region surrounding Sox21 contains three

lincRNAs, SOX21-AS1,linc-SOX21-B, and linc-SOX21-C, that are expressed and overlap DNA sequences that are

alignable to other mammals, while linc-SOX21-B even overlaps a region that is alignable to zebrafish. Despite this con-

servation between the DNA sequences, Hezroni et al. did not detect transcription of either linc-SOX21-B or linc-

SOX21-C in any of the other species that were studied, including two primates. Rather, the significant alignment of

human linc-SOX21-B to other genomes was attributed to its overlap with a highly conserved brain and neural tube

enhancer (VISTA, element hs488; Visel et al., 2007). With this scenario in mind, it is advised to first reconstruct

lncRNAs independently in each species and then subsequently compare the RNA sequences to each other.

It is now generally accepted that many lncRNAs are lineage specific and do not have recognizable homologues in

distant species (Carninci & Hayashizaki, 2007; Church et al., 2009; Hezroni et al., 2015; Necsulea et al., 2014; Okazaki

et al., 2002; Paralkar et al., 2014). This is strongly substantiated by studies that examined lncRNAs expressed in specific

4of25 ROSS AND ULITSKY

tissues across closely related mammalian species (Chen et al., 2016; Kutter et al., 2012; Mor

an et al., 2012; Mustafi

et al., 2013). For example, Kutter et al. found that only 60% of the lncRNAs that are expressed in mouse liver are also

expressed in rat liver, while only 27% are also expressed in human liver (Kutter et al., 2012). As expected, the conserva-

tion of lncRNAs decreases dramatically across longer evolutionary distances. One of the first transcriptome-wide com-

parisons between lincRNA expression in zebrafish and mammals found that only 29 out of 567 (5%) lincRNAs that

were identified in zebrafish had detectable putative orthologs in mammals (Ulitsky et al., 2011). A subsequent study

explored lncRNA conservation beyond vertebrates by comparing 16 vertebrate species and sea urchin (Hezroni

et al., 2015). Although hundreds of putative orthologs were detected beyond mammals, only 99 lincRNA genes could be

traced to the last common ancestor of tetrapods and teleost fish, and no significant homology was detected between ver-

tebrates and sea urchin. Although this number of putative orthologs between human and fish was substantially less

than the 171 lncRNAs reported by Necsulea et al. (potentially due to the combination of factors previously discussed)

the overall consensus is that homology between mammals and fish is detectable for <5% of human lncRNAs. These

observations mark out two possible roadmaps for fruitful comparative genomic analysis. One possibility, if the majority

of lncRNAs did arise de novo after major speciation events, is that we can use more intermediate evolutionary distances

or a phylogenetic tree-based approach to carefully compare lncRNAs between more closely-related species. Alterna-

tively, it is also plausible that many more lncRNAs are deeply conserved and their sequences contain spasmodic simi-

larity that is not statistically detectable by alignment-based algorithms. If this is the case, it is expected that many

lncRNAs will have synthetic counterparts in distal species, but without significant sequence similarity. Indeed, such

lncRNAs have been observed (Amaral et al., 2018; Bryzghalov et al., 2020; Hezroni et al., 2015; Ponjavic et al., 2009;

Ulitsky et al., 2011). For example, Hezroni et al. found no significant sequence homology, but over 2000 human

lincRNAs had putative homologues with only positional conservation (referred to as syntologs) in the sea urchin

genome, which was 600 more than the number expected by chance, evaluated by randomly placing lncRNAs in the

sea urchin genome (Amaral et al., 2018; Bryzghalov et al., 2020; Hezroni et al., 2015; Ponjavic et al., 2009; Ulitsky

et al., 2011). Such cases are worth exploring further as they will elucidate particularly short conserved motifs that are

potentially easy to study further experimentally. Recent advances in computational methods for the comparative analy-

sis of lncRNAs have therefore focused on alignment-free methods to uncover subtle motifs that have been purified in

conserved lncRNAs from distant species or are shared between lncRNAs that perform similar functions in the same

species. In the next section we provide examples of how the following dimensions: (1) syntenic conservation across spe-

cies, (2) primary sequence conservation, and (3) the conservation of elements that adopt shared secondary structures,

have been used to increase the sensitivity of lncRNA identification and motif discovery.

3|COMPUTATIONAL APPROACHES THAT LEVERAGE SEQUENCE

FEATURES OBSERVED IN CONSERVED IN lncRNAs

3.1 |Syntenic conservation increases the sensitivity of orthologous lncRNA detection in

distal species

Although synteny alone is not informative for functional motif discovery, it is a fundamental property that can be used

to identify and construct databases of putatively orthologous lncRNAs, and these datasets can be used as input for sub-

sequent motif analysis (Table 1), which is especially useful in cases where significant sequence similarity is not detect-

able. Syntenic conservation was recently used to characterize a new subgroup of lncRNAs in mammals which are

positioned at chromatin loop anchor points and the borders of topologically associating domains (TADs; Amaral

et al., 2018). These lncRNAs were named topological anchor point RNAs (tapRNAs) and were very often associated

with developmental genes with which they are co-expressed. Motif analysis, performed by the direct alignment of

human and mouse tapRNAs, showed that they are enriched with binding sites of transcription factors (TFs) and zinc-

finger proteins such as the chromatin organizer CTCF. Synteny has also been used to construct extensive resources of

putative orthologous lincRNA sequences that are now available for future studies. For instance, thousands of full-length

orthologous lincRNAs were generated using the PLAR (lncRNA annotation from RNA-seq data) pipeline by compari-

son of lncRNAs across vertebrates and sea urchin (Hezroni et al., 2015). Another good resource for obtaining lncRNA

sequences is SyntDB a database of syntenic lncRNAs that are conserved across 11 primate species (Bryzghalov

et al., 2020). The identification of lncRNAs that are positionally conserved has also been implemented as a key step in

publicly available pipelines that have been developed for the identification of orthologous lncRNA genes. Below we

ROSS AND ULITSKY 5of25

TABLE 1 Databases and pipelines for the retrieval of lncRNA homologues across species

Resource Species analyzed Description Website References

Pipeline PLAR 17 vertebrates

and sea urchin

Pipeline for reconstruction and of lncRNA transcripts

from RNA-seq from multiple tissues and 3P-seq data

http://webhome.weizmann.

ac.il/home/igoru/PLAR

(Hezroni et al., 2015)

Slncky 29 mammals lncRNA discovery tool based on RNA-seq from

multiple species. Orthologs identified by synteny

https://slncky.umassmed.

edu/

https://slncky.github.io/

(Chen et al., 2016)

LincOFinder Vertebrates and

invertebrates

lncRNA discovery tool based on annotated orthologous

PCGs and RNA-seq from multiple species. lncRNA

orthologs identified by microsynteny cluster analysis

https://github.com/

cherrera1990/

LincOFinder

(Herrera-Úbeda et al., 2019)

lncEVO 5 mammals Pipeline for identification, reconstruction and

comparison of conserved lncRNAs based on RNA-seq

from multiple species

https://gitlab.com/

spirit678/lncrna_

conservation_nf

(Bryzghalov et al., 2021)

Database Evo-devo 6 mammals

and chicken

Catalogs of lncRNAs identified from RNA-seq in seven

major developmental organs from early

organogenesis to adulthood

https://apps.kaessmannlab.

org/lncRNA_app/

(Sarropoulos et al., 2019)

SyntDB 12 primates lncRNAs identified from RNA-seq from multiple

tissues. Implements slncky to identify syntologs

across species

http://syntdb.amu.edu.pl (Bryzghalov et al., 2020)

RNAcentral >50 species Collection of 44 ncRNA databases including

NONCODE, LncBOOK, LNCipedia, and lncRNAdb

https://rnacentral.org (RNAcentral

Consortium, 2021)

Study Kutter et al. Mouse and

rat species

lncRNA catalogs constructed from RNA-seq and

supported by H3K4me3-bound (ChIPseq) DNA data

(Kutter et al., 2012)

Washietl et al. 6 mammals lncRNA catalogs constructed from RNA-seq from

multiple tissues

(Washietl et al., 2014)

Necsulea et al. 11 vertebrates lncRNA catalogs constructed from RNA-seq from

multiple tissues

(Necsulea et al., 2014)

6of25 ROSS AND ULITSKY

describe how two such methods, slncky (Chen et al., 2016) and LincOfinder (Herrera-Úbeda et al., 2019), use syntenic

conservation to increase the sensitivity of lncRNA identification.

The slncky pipeline is based on two steps, both of which rely on syntenic conservation. First, slncky aligns putative

lncRNAs to syntenic noncoding transcripts from different species in order to evaluate coding potential. If an alignment

is detected, slncky then interrogates putative open reading frames (ORF) that are longer than 30 nt and conserved in

both species. For each ORF, slncky computes the ratio of nonsynonymous to synonymous mutations (dN/dS) and only

considers the putative lncRNA as protein-coding if the ORF has a significantly low dN/dS ratio. Second, slncky

increases the alignment score between two putative lncRNA orthologs by expanding the alignment to include the

highly conserved sequences of their flanking protein-coding genes. Specifically, once slncky has defined syntenic regions

that contain a noncoding transcript, it performs a second alignment of only the area 150,000 nt upstream and down-

stream of the syntenic region. To minimize false positive alignments (that can often arise from aligning repetitive ele-

ments), slncky aligns each lncRNA to shuffled intergenic sequences and determines an empirical 5% threshold for

classifying significant alignment scores. Using this approach, the authors were able to identify 1466 out of 1521 (>95%)

of orthologous lncRNAs previously reported between human and mouse and also identify a further 121 pairs (8%) of

the homologous human-mouse lncRNAs that were previously classified as species-specific. They also found that 18% of

lncRNAs that are expressed in mammalian pluripotent cells were likely present prior to the divergence between rodents

and primates (Chen et al., 2016). Although slncky is useful, it is largely dependent on the quality of pairwise WGAs to

project lncRNA expression to loci in any other species and may therefore have limitations in comparing more distant

species.

The use of synteny to identify orthologous lncRNAs was recently expanded to establish another pipeline, Lin-

cOFinder, that was used to infer putatively conserved lncRNAs between human and amphioxus (Herrera-Úbeda

et al., 2019). Instead of relying on WGAs to define syntenic regions, LincOFinder creates two lists of genes: one list con-

tains all the genes in the reference species sorted by genomic position and the second list contains all corresponding

orthologs of protein-coding genes in the species being interrogated. These orthologs are identified by using known sets

of orthologous families or helper programs such as Orthofinder (Emms & Kelly, 2015). Sets of candidate genes (that

may potentially define a microsyntenic region) in the interrogated species are then selected if they are orthologs of

genes that neighbor a lincRNA in the reference species. The coordinates of these candidate genes form the input to a

UPGMA hierarchical clustering algorithm which identifies microsyntenic clusters that contain the orthologous genes

that are sufficiently close together. Each of these clusters are then scanned for expression of a lincRNA in the inter-

rograted species. Using this approach, the authors were able to identify 16 lincRNAs putatively conserved between

human and amphioxus, including HOTAIRM1 which is located in the anterior part of the Hox cluster across several

vertebrate lineages (Gardner et al., 2015; H. Yu et al., 2012).

3.2 |Computational frameworks focused on short motifs

Although the primary sequences of lncRNAs are not well conserved, multiple studies have identified short conserved

motifs using alignment-based approaches (Chureau et al., 2002; Hezroni et al., 2015; Jin et al., 2021; Ulitsky et al., 2011)

and experimental techniques (Ilik et al., 2013; Quinn et al., 2014,2016) to uncover short functional domains in

lncRNAs that are evolutionary conserved. For example, in Hezroni et al., 2015 the direct comparison of RNA sequences

using BLASTN identified short patches that were alignable in lincRNAs that were conserved in human and one of

15 other vertebrates. Although short conserved patches could be detected as far as shark, the length of these patches sig-

nificantly decreased in fish species with averages less than 100 bases, and approaching the lower limit of lengths that

can be detected by BLASTN. To overcome limitations in the alignment, the study also characterized motif enrichment

across lncRNAs in each species by counting the number of occurrences of all possible 6mers in exonic sequences com-

pared to the number of random occurrences in shuffled sequences with preserved dinucleotide frequencies. This

approach is similar to the n-gram metrics that were systematically evaluated against alignment-based approaches for

the analysis of lncRNA sequences in Noviello et al. (2018), where each n-gram would comprise the six consecutive

nucleotides in each 6mer. From their analysis, Hezroni et al. identified 31 motifs that were enriched in at least 12 spe-

cies, thus capturing motifs enriched in lncRNAs in both mammals and fish. A substantial fraction of these motifs cor-

responded to exonic splicing enhancers (ESEs), were purine-rich, or composed combinations of CUG and CAG which

form binding sites for the splicing factors CUG-BP and Muscleblind. This together with additional studies (Gil &

Ulitsky, 2018; Haerty & Ponting, 2015; Schuler et al., 2014; J. Y. Tan et al., 2020) provides evidence that the

ROSS AND ULITSKY 7of25

subsequences that control the processing of lncRNA transcripts are functionally important and under purifying selec-

tion, hinting that in many cases the biological function of the lncRNA may be incidental to the process of its

maturation.

In addition to mapping motif enrichment across lncRNAs that are evolutionary conserved, motif enrichment has

also been described across more focused sets of lncRNAs that can be potentially considered as coming from the same

gene family. Gil and Ulitsky (2018) explored the differences in sequence composition and chromatin landscape between

enhancer regions that transcribed lncRNAs and enhancers that transcribed much shorter and less stable enhancer-

RNA (eRNA) molecules. They found that, in comparison to enhancers that only produce eRNAs, enhancers that harbor

lncRNA transcripts have heightened activity and are significantly enriched with motifs recognized by specific RBPs,

particularly splicing factors that are required for the maturation of the lncRNA. They inferred, from analyzing expres-

sion data in which several splicing factors were knocked down, that this heightened enhancer activity is driven by the

maturation of the lncRNAs. This was further evidenced in a similar study that reported the significant enrichment of

splicing-associated motifs in multi-exonic lincRNAs that are produced from enhancers compared to single-exonic coun-

terparts (Tan et al., 2020). Likewise, their analysis also showed that enhancers that are associated with multi-exonic

lncRNAs have a 2.5-fold increase in enhancer activity. Both studies also reported that lncRNA producing enhancers are

enriched with binding sites for proteins involved in chromatin remodeling and loop formation, specifically CTCF. To

achieve their results these studies used motif analysis tools such as AME (McLeay & Bailey, 2010), FIMO (Grant

et al., 2011), and TOMTOM (Gupta et al., 2007) that scan sequences for known motifs and are not reliant on sequence

alignments (Table 2).

There are also now several examples where the mature lncRNA product is functional, and in many of these cases the

functionality can be attributed to short conserved motifs that correspond to miRNA binding sites or RBPs that confer post-

transcriptional function. One well known example is Cyrano (OIP5-AS1 in human), which harbors an extensively paired

site to miR-7, which is required for degradation of miR-7 (Kleaveland et al., 2018; Ulitsky et al., 2011). The degradation of

miR-7 enables CDR1as (a circular RNA that has many binding sites to miR-7) to accumulate in the cytoplasm of neurons,

where it potentially regulates neuronal activity (Kleaveland et al., 2018). The miR-7 binding site in Cyrano is contained

within a deeply conserved stretch of 67 nucleotides, which is the only region that is significantly alignable between human

(8862 nt) and zebrafish (4630 nt) by BLASTN. Another well-known example is NORAD, a highly conserved cytoplasmic

lncRNA that antagonizes the repression of mRNA levels of genes involved in cell division by sequestering Pumilio proteins

(Elguindy et al., 2019;Koppetal.,2019; S. Lee et al., 2016; Tichon et al., 2016). Human NORAD contains at least 17 Pumilio

recognition elements (PREs) that bind PUM1 and PUM2 proteins (Tichon et al., 2016).

Collectively the findings in these studies support the notion that lncRNA function often requires only short con-

served patches that are under selection, and that the remainder of the sequence can tolerate extensive changes. This

concept has been applied in recent frameworks, specifically developed to compare noncoding sequences (Table 2).

Below we discuss two of these frameworks, namely SEEKR (SEquence Evaluation from Kmer Representation; Kirk

et al., 2018) and LncLOOM (Ross et al., 2021) that use alternative implementations of n-grams to home in on short

motifs (or kmers) that underlie lncRNA function (Figure 1b).

Proteins that function via similar mechanisms adopt similar folds and bear specific domains that can be described

by statistical models of amino acid sequences. This enabled the large-scale classification of protein families within and

across species. In contrast, lncRNAs that function in similar ways often lack detectable significant sequence homology.

For example, Xist and Kcnq1ot1 are both known to recruit the Polycomb Repressor Complex to repress their target

genes in cis (Lee & Bartolomei, 2013). However, the lack of sequence homology between them made it difficult to pre-

dict this shared mode of action from their linear sequences (Kirk et al., 2018). To overcome this limitation, SEEKR uses

a statistical approach to compare profiles of kmer abundance between different lncRNA sequences. Kmer profiles are

derived for each lncRNA sequence by counting the occurrences of 4-, 5-, and 6mers, and these profiles are then normal-

ized by the length of the sequence and compared with Pearson correlation. The authors used SEEKR to compare

human and mouse lncRNAs with putatively related functions. In both species, SEEKR was able to cluster well-known

cis-repressive lncRNAs (including: XIST,TSIX, KCNQ1OT1,UBE3A-ATS,ANRIL/CDKN2B-AS1, and Airn) that were

characterized by a high abundance of AU-rich kmers, separately from cis-activating lncRNAs (including: PCAT6,

HOTTIP,LINC00570,DBE-T, and HOTAIRM1) that had a significant enrichment of GC-rich kmers. Overall, their

results show that kmer-based quantitation can infer related functions in lncRNAs, irrespective of their low sequence

homology.

In an alternative approach, we recently developed LncLOOM, a new framework that identifies combinations of

short motifs that are found in the same order in putatively homologous sequences from different species. To identify

8of25 ROSS AND ULITSKY

TABLE 2 Computational tools that have been used for comparative analysis of lncRNA sequences

Program Algorithm description

Recommended dataset for lncRNA sequence

analysis Example of study where applied

BLASTN Local alignment-based discovery of segments with

statistically significant similarity

Closely related species Identification of short conserved patches in lncRNAs

across vertebrates

(Hezroni et al., 2015)

MEME Alignment-based motif discovery based on

expectation–maximization (EM) algorithm

Can be used across intermediate evolutionary

distances

Discovery of a functional conserved motif in Nron

lncRNA, that binds to E3 ubiquitin ligase CUL4B to

regulate ERαstability. (Jin et al., 2021)

AME Statistical comparison of differential motif enrichment

between sequences, based on a linear regression

model

To compare groups of lncRNAs that have similar

functions in the same or closely related species

Characterization of motifs that are enriched in

lncRNAs that are transcribed from enhancer

regions, and dictate higher enhancer activity. (Gil &

Ulitsky, 2018)

TOMTOM Statistical comparison of one or more DNA/RNA

motifs against a database of known motifs

Functional annotation of motifs discovered in

lncRNA sequences that have similar functions

Characterization of transcription factor binding sites

that are enriched in multi-exonic lincRNAs that are

produced from enhancers, compared to single-exon

counterparts. (Tan et al., 2020)

FIMO Statistical tool to locate specific motifs within a

DNA/RNA sequence

To locate motifs of interest in a set of lncRNA

sequences

Seekr Statistical comparison of kmer composition profiles

across different lncRNA sequences

To identify and compare groups lncRNAs that have

similar functions in the same species

Differential characterization of kmer profiles in cis-

activating and cis-repressive lncRNAs in human and

mouse. (Kirk et al., 2018)

LncLOOM Implementation of the longest common subsequence

problem using a graph-based approach and integer

linear programming to identify conserved

combinations of motifs that appear in the same

order in sequences from different species

To compare lncRNA sequences that are orthologous

across large evolutionary distances. Ideally, the set

of sequences should represent species that

monotonically increase in evolutionary distance

Discovery of deeply conserved, novel functional

elements in Chaserr lncRNA that were

experimentally validated to regulate CHD2

expression. (Ross et al., 2021)

TDF (Triplex

Domain

Finder)

Statistical ranking, based on motif over-

representation, of lncRNA domains that are likely to

interact with specific DNA targets, such as

protomers of differentially expressed genes upon

lncRNA perturbation

To map DNA binding domains in a single lncRNA

sequence, for which differentially expressed genes

are known from experimental knockdown of the

lncRNA

Identification of known triple helices of lncRNAs

Fenderr,HOTAIR,MEG3, and prediction of

GATA6-AS triple helices that modulate cardiac

mesoderm differentiation. (Kuo et al., 2019)

DeepLncRNA Deep learning algorithm that classifies localization of

lncRNAs based on sequence composition

To predict the localization of a specific lncRNA

sequence, based on specific sequence motifs

Large-scale prediction of the localization of human

and primate lncRNAs. (Gudenas & Wang, 2018)

iLoc-lncRNA SVM is a machine-learning algorithm based on the

statistical learning theory that classifies localization

of lncRNAs based on octamer composition lncRNA

sequences

To predict the localization of a specific lncRNA

sequence, based on specific sequence motifs

Large-scale prediction of the localization of lncRNAs

from the RNALocate database. (Su et al., 2018)

ROSS AND ULITSKY 9of25

conserved motif combinations, LncLOOM builds a directed graph from a set of homologous sequences that are ideally

ordered from species with a monotonically increasing evolutionary distance with respect to a query sequence. Each

sequence is modeled as a layer of kmers, where each kmer represents a node in the graph and identical nodes in con-

secutive layers are connected by edges. Motif discovery is then performed using integer linear programming to find long

nonintersecting paths in the graph. Essentially, LncLOOM aims to identify the longest common subsequence

(Maier, 1978) in a set of homologous sequences. The LncLOOM approach is underpinned by the assumption that the

linear order of short subsequences has been conserved across long evolutionary distances. Although the order of kmers

may not always be important for function, LncLOOM can capture combinations of kmers where the order has been

maintained because the functionality of the conserved elements is aided by sequence or structural context in the longer

RNA molecule. More importantly, it is unlikely that large combinations of short conserved kmers that appear in the

same order were independently gained during evolution. Rather, it is more plausible that these elements are remnants

of their rapidly evolving sequences and have been evolutionarily selected because they are important to the function of

the lncRNA. From a computational standpoint, the constraint of the linear order of kmers increases the power of motif

discovery because an ordered set of k-mers is much less likely to appear by chance than any one of its possible permuta-

tions, and so while the presence of each of the kmers or even their combination might not be statistically significant,

the addition of the conserved order constraint enables discovery of significant conservation. To demonstrate the power

of LncLOOM, the framework was used to analyze the sequences of Cyrano,Chaserr, and Libra (a zebrafish lncRNA that

is homologous to the Nrep mRNA in mammals) across vertebrate species. In Cyrano, LncLOOM identified a set of nine

ordered kmers that are conserved in 17 vertebrate species from human to zebrafish, seven of which were also conserved

to elephant shark. In addition to known miRNA and protein binding sites, LncLOOM identified novel functional ele-

ments in Chaserr that were conserved throughout vertebrates and were experimentally validated to regulate CHD2

expression (Ross et al., 2021).

3.3 |Comparison of lncRNA secondary structures

RNA secondary structures are known to mediate RNA-protein and RNA–RNA interactions. They have been shown to

be critical to the biogenesis and function of many ncRNAs such as miRNAs, small nucleolar RNAs (snoRNAs), and

transfer RNAs (tRNAs; Bhartiya & Scaria, 2016; Parisien et al., 2013). They also contribute to the stability, processing,

and localization of mRNA molecules as in the case of riboswitches, terminal stem loops and 30stem loops in histone

mRNAs (Mignone et al., 2002; Singh & Singh, 2019; Svoboda & Di Cara, 2006; D. Tan et al., 2013). Due to their func-

tional importance, such elements are often constrained during evolution more than their flanking linear RNA

sequences. For instance, TERC, the RNA component of the telomerase complex, is structurally very conserved across

vertebrates despite sharing only 60% sequence identity between human and mouse (Seemann et al., 2017; Wang

et al., 2016). Likewise, we expect that the low sequence conservation of lncRNAs across vertebrates does not preclude

the existence of conserved structural motifs that underpin their function (Diederichs, 2014; Guttman & Rinn, 2012;

Wutz et al., 2002). Indeed, there is some evidence from experimentally solved structures of conserved secondary struc-

ture elements, including in XIST (Pintacuda et al., 2017; Smola et al., 2016), MALAT1 (Brown et al., 2014; McCown

et al., 2019), Cyrano (Jones et al., 2020), MEG3 (Uroda et al., 2019; Zhang et al., 2010), and COOLAIR (Hawkes

et al., 2016). We describe the proposed contributions of these structures to lncRNA function in more detail in section 4,

when we consider experimental approaches to uncover functional motifs.

From a computational perspective, structural motif discovery in lncRNA sequences is a major challenge

(Figure 1C). Although many methods have been developed to predict the secondary structure of RNA sequences

(reviewed in B. Yu et al., 2020), the accuracy of these algorithms for predicting the full structures of long RNAs is lim-

ited. The most popular programs rely on minimum free energy (MFE) calculations, for which the accuracy of predicted

base pairs has been reported to be as low as 40% for RNAs longer than 500 nt (Doshi et al., 2004; Lorenz et al., 2016).

This is mainly due to an incomplete understanding of the thermodynamic parameters that govern RNA molecular

interactions and stability; such that any small change to the thermodynamic models that underpin MFE calculations,

can generate significantly different optimal structures (Rogers et al., 2017; Schroeder, 2018). This becomes particularly

relevant in the case of longer RNAs where the number of possible folds is enormous, and many possible optimal struc-

tures can be generated with no significant differences in their MFE scores. Another caveat in this approach is that the

model assumes that the optimal structure of the RNA molecule corresponds to a global free energy minimum. This

assumption does not account for the co-transcriptional folding of RNA transcripts (Watters et al., 2016; Yu et al., 2021),

10 of 25 ROSS AND ULITSKY

which may result in the RNA adopting local folds such that the kinetics of the molecule differ from the global free

energy minimum of the whole RNA. Another conceptual problem is that even random RNA sequences can adopt

highly stable structures in cells (Schultes et al., 2005), and so the presence of a particular structure cannot usually be

used as evidence for its functional importance. Similarly, random sequences have multiple predicted plausible folds,

that may comprise a similar number of base pairs and have similar stability (Rivas & Eddy, 2000). This highlights the

challenge in deciphering which RNA molecules harbor biologically relevant structures as the presence of a structural

motif in an RNA sequence from one species (determined either experimentally or predicted) does not provide evidence

that the motif is functional. To infer functional importance, it is necessary to uncover statistical evidence that the motif

has been evolutionarily conserved beyond phylogenetic expectation, that is, that the functional constraints are imposed

on the structural motif and not the primary sequence in which the motif is embedded (reviewed in Rivas, 2021). Covari-

ance models are the gold standard for such analysis (Eddy & Durbin, 1994; Griffiths-Jones et al., 2003) and have been

instrumental in predicting the structure of other ncRNAs, such as rRNA (Gutell et al., 2002). However, opinions differ

on the extent to which covariation statistically supports predicted secondary structures in lncRNA sequences

(Rivas, 2021). Here, we briefly discuss the handful of studies that have used covariance based approaches to predict con-

served structural motifs in lncRNAs.

Covariance models have been implemented in several computational tools that predict RNA secondary structure

including: R-scape (Rivas et al., 2017) and frequently used programs such as INFERNAL (Nawrocki et al., 2009),

CMFinder (Yao et al., 2005), EvoFold (Pedersen et al., 2006), and others (reviewed in Tahi et al., 2017). To circum-

vent the inaccurate prediction of the secondary structure of longer RNA molecules, studies have predicted local

structures of shorter putative functional regions within lncRNAs such as NORAD (Tichon et al., 2016,2018), Cyrano

(Jones et al., 2020), and XIST (Wutz et al., 2002). These studies used programs such as EvoFold that uses a phyloge-

netic stochastic context-free grammar (phylo-SCFG) to model coevolving base pairs within the structural motifs;

ScanFoldScan (Andrews et al., 2018) that predicts local structures across an RNA sequence and then searches the

RFam database for conservation matches; or CMFinder (Yao et al., 2005)thatusescovariancemodelstoimprove

the accuracy of the predicted structural motifs. Although these programs provide evidence of motif conservation,

they do not statistically test if the level of covariance is significant, such that the motif is conserved beyond random

expectation. Nonetheless, CMFinder was used by Seemann et al. (2017) to screen vertebrate genomes for putatively

conserved RNA structures (CRSs). The authorsreportedCRSsinseverallncRNAsincludingXIST and MALAT1 and

noted that the density of CRSs decreases from the 50to the 30end of lncRNAs. On the other hand, R-scape provides

a method that quantitatively tests whether covariance supports the presence of a conserved RNA secondary struc-

ture (Rivas et al., 2017). R-scape analysis detected no statistically significant support for the secondary structures

determined experimentally in lncRNAs HOTAIR,SRA,andXIST. A later study proposed that the covariance support

of evolutionary conserved structures in HOTAIR, and other lncRNAs, can be increased by extending the depth of

the alignment beyond mammals to increase variation or adjusting the default parameters of R-scape to scan shorter

window sizes or use different statistical metrics (Tavares et al., 2019). However, this has since been rebutted by evi-

dence that the alignments of HOTAIR and SRA do have sufficient variation to detect covariations, yet still lack sig-

nals of evolutionary conserved structures (Rivas et al., 2020). All in all, the analysis of conserved structural motifs in

lncRNA sequences remains a major challenge with very few positive controls that can be used for better method

development.

The best approach may be to consider structural elements in combination with other features as a means to

enhance the sensitivity of lncRNA classification. Along these lines, Quinn et al. (2016) used a combination of syn-

teny, microhomology, and conserved secondary structural elements to identify roX1 and roX2 lncRNAs across

diverse Drosophila species. In the past, homology between these genes has remained undetected because the similar-

ity between their sequences is comparable to that of random sequences. In an earlier study, Tycowski et al. (2012)

presented a structure-based bioinformatic screen that successfully identified conserved expression and nuclear

retention elements (ENEs) in the genome of diverse viral genomes. The ENE, a structured element that is composed

of a stem-loop structure with an asymmetric internal U-rich loop, was first identified in the genome of Kaposi's

sarcoma-associated herpesvirus (Conrad & Steitz, 2005). The element is located ~120 nt upstream of the poly-

adenylation site (PAS) of PAN lncRNA, and plays an essential role in stabilizing PAN lncRNA through triple-helix

formationwithitspoly(A)tail,therebypreventing the initiation of RNA decay (Conrad et al., 2006; Mitton-Fry

et al., 2010). The conserved ENEs identified by Tycowski et al. (2012), were indeed located in putative lncRNAs that

were transcribed from intergenic regions of the viral genomes.

ROSS AND ULITSKY 11 of 25

4|EXPERIMENTAL STUDIES OF FUNCTIONAL ELEMENTS IN lncRNA

GENES

In parallel to the computational predictions, functional elements can also be identified and studied in detail experimen-

tally, as we discuss next.

4.1 |Systematic dissection of lncRNA sequences that dictate subcellular localization

To further understand the mechanisms that drive lncRNA function, several studies have used massively parallel

reporter assays to systematically interrogate lncRNA sequences to decipher elements that dictate their function

(Figure 2a). For example, to identify elements in lncRNAs and mRNAs that can force nuclear localization, Lubelsky

and Ulitsky cloned thousands of short fragments that tiled the exons of human lncRNAs and 30UTRs of selected

mRNAs into the untranslated region of an otherwise cytoplasmic mRNA. By comparing reporter RNA from cytoplasmic

and nuclear fractions of cells that were transfected with the library, they attributed nuclear accumulation to a short

C-rich sequence (42 nt) derived from an Alu repeat, named SIRLOIN. Using RNA immunoprecipitation (RIP) with an

FIGURE 2 Experimental approaches that have been used to investigate the functional importance of conserved motifs in lncRNAs.

(a) A high throughput tilling assay to identify sequence elements that modulate RNA localization. (b) RNA-centric approaches (left) can be

used to identify proteins, RNA, or DNA bound to an RNA of interest. Cross-linking approaches allow for in situ discovery of RNA

interaction partners, through pull down with biotinylated antisense probes specific to the RNA of interest. In vitro transcription of

biotinylated RNA molecules of wild-type and mutated fragments can be used to investigate interactions with conserved regions within the

RNA molecule. Protein-centric approaches (right) identify protein:RNA interactions with a protein of interest, based on protein pulldown

with antibody affinity beads. (c) Perturb-and-rescue experiments. Left: Endogenous rescue with fragments that encode functional domains of

a lncRNA transcript, following knockdown of the endogenous express lncRNA using CRISRPi. Right: Cross-species rescue of post-

transcriptional perturbation of lncRNA transcript using ASOs that target conserved regions

12 of 25 ROSS AND ULITSKY

hnRNPK-specific antibody, they further showed that SIRLOIN drives nuclear localization by recruiting hnRNPK

(Lubelsky & Ulitsky, 2018). In a follow-up study, the authors investigated the mode of action of SIRLOIN at higher res-

olution by using a suite of massively parallel RNA assays and libraries that contained thousands of sequence variants to

pinpoint the regions within the SIRLOIN element that are essential for the nuclear retention of lncRNA transcripts

(Lubelsky et al., 2021). Although SIRLOIN contains multiple CCC elements that bind hnRNPK, their findings suggest

that nuclear retention is solely dependent on hnRNPK binding to the GCCUCCC element that is precisely positioned in

the SIRLOIN core (Lubelsky et al., 2021). In an independent study, Shukla et al. also used a tiling-reporter assay to

characterize sequence elements that dictated the localization of lncRNA transcripts. They identified a shorter C-rich

motif (15 nt), along with 108 other RNA elements that increased nuclear localization of their cytoplasmic reporter

(Shukla et al., 2018). As an alternative to the tiling strategy, a recent study introduced mutREL (RNA elements for sub-

cellular localization by sequencing), which is a high-throughput method coupled with random mutagenesis to identify

motifs that specify localization (Yin et al., 2020). In this approach, DNA fragments of candidate genes are PCR ampli-

fied from their genomic DNA and then cloned into GFP-reporters which are then stably integrated into the genome of

cells. RNA is then isolated and sequenced from the chromatin, nucleoplasm, and cytoplasm fractions. The random

mutagenesis is achieved by using an error-prone PCR (McCullum et al., 2010). The authors used mutREL to investigate

lncRNAs Malat1,Neat1,NXF1-IR, and uncovered an RNA motif that recognizes the U1 small nuclear ribonucleopro-

tein (snRNP) to promote lncRNA-chromatin retention (Yin et al., 2020).

4.2 |Affinity assays uncover lncRNA motifs that facilitate lncRNA:protein, lncRNA:

RNA, and lncRNA:genome interactions

Many of the proposed mechanisms for lncRNAs have been uncovered from affinity-based assays that can also be used

to attribute binding to a protein, DNA, or RNA to specific regions within lncRNA molecules. To date, an array of versa-

tile methods have been developed to map interactions with RNA molecules (reviewed in Ramanathan et al., 2019).

These methods can be broadly categorized into RNA-centric methods which identify interacting partners bound to an

RNA of interest; or protein-centric methods which identify RNA molecules bound to a protein of interest. In this review

we give selected examples of how some of these assays have been used to investigate the molecular mechanism of cer-

tain lncRNAs, several more examples have been described in (Constanty & Shkumatava, 2021).

RNA-centric methods include an array of hybridization methods that capture RNA interactions with protein, RNA

and DNA (Figure 2b). These methods have been widely used to explore the modes of action of many lncRNAs

(reviewed in Cao et al., 2019). In 2011 two main methods were developed to investigate the in vivo genomic binding

sites of chromatin associated lncRNA, namely: ChIRP (Chromatin isolation by RNA purification; Chu et al., 2011,

2012) and CHART (capture hybridization analysis of RNA targets; M. D. Simon, 2013; M. D. Simon et al., 2011). The

methods are similar in that they use biotin-conjugated 20–25mer DNA probes to purify a lncRNA of interest RNA and

determine its associated chromatin fraction by DNA deep sequencing. Subsequently, RAP (RNA antisense purification),

a method that increases specificity by using longer DNA probes, was developed (Engreitz et al., 2015; Engreitz

et al., 2013). All three methods have been successfully used to establish high resolution maps of genomic binding roX2

(Chu et al., 2011; Simon et al., 2011) and Xist (Engreitz et al., 2015; Engreitz et al., 2013) lncRNAs across the X chromo-

some. Using ChiRP, Chu et al. (2011) also showed that HOTAIR preferentially binds genomic regions containing a GA-

rich motif. More recently, several high throughput methods have been developed to systematically map RNA:chromatin

interactions on a genome-wide scale (reviewed in Kato & Carninci, 2020). Additionally, RNA and protein that is

enriched from purification by these methods can also be isolated and subjected to RNA and protein analysis. For exam-

ple RAP-RNA was used to show that Malat1 interacts with many nascent pre-mRNAs as a means to localize to the

chromatin at active genes (Engreitz et al., 2014). While a combination of RAP and mass spectrometry revealed that Xist

interacts directly with SHARP, which also interacts with the SMRT corepressor to activate HDAC3 deacetylation activ-

ity on chromatin (McHugh et al., 2015). In a recent study, RAP was used to characterize the functional mechanism of

ADEPTR, a lncRNA recently found to be necessary for activity-dependent changes in synaptic transmission and struc-

tural plasticity of dendritic spines (Grinman et al., 2021): ADEPTR contains a conserved sequence that binds to actin-

scaffolding regulators AnkB and Spnt1, which are then transported to distal process through ADEPTR interactions with

the motor protein Kif2A. Although cross-linking methods have proved to be useful, they can also be technically chal-

lenging due to the inefficiency of RNA-pulldowns, in particular for RNAs that are not abundant. As an alternative, the

high-throughput method incPRINT, was recently developed to screen protein interactions with an RNA of interest in

ROSS AND ULITSKY 13 of 25

cells (Graindorge et al., 2019). This is achieved by using a library of tagged proteins with a target RNA that is tethered

to a luciferase detector. The method enables the characterization of in-cell RNA-interacting proteomes, as well as the

mapping of protein bindings across different regions of long RNA transcripts.

In addition to studying the interactomes of endogenously expressed lncRNAs, affinity-based RNA pulldown

methods that are based on in vitro transcription (IVT) of biotinylated target can also be used to characterize interacting

proteomes of lncRNAs (reviewed in Cao et al., 2019). The method is advantageous in that it is fast and can be used to

enrich RBPs associated with unabundant target RNA. However, the approach also has limitations as in vitro transcribed

RNA may not have the same modifications or adopt the same structure as cellular RNA, and they may encounter in

lysates proteins that are not normally found in their proximity in cells, for example, a nuclear lncRNA may bind

strongly to cytoplasmic proteins in lysates, but not in cells. Nevertheless, IVT affinity-based RNA pulldown has been

widely used to characterize lncRNA:protein interactions (Huang et al., 2017; Noh et al., 2016; Unfried et al., 2021). We

recently used an IVT approach, combined with mutagenesis, to interrogate the functional importance of short con-

served elements in the last exon of Chaserr (Ross et al., 2021). Following incubation with cell lysate, WT and mutated

transcripts were pulled down with streptavidin beads and interacting proteins were identified by mass-spectrometry,

and comparison of the identified interactomes homed in on proteins bound specifically to the conserved sites in

Chaserr.

Protein-centric methods include well-established assays such as RIP and cross-linking and immunoprecipitation

(CLIP) to identify endogenous RNAs that interact with a protein of interest (Figure 2b). RIP is an effective assay to dis-

cover RNA molecules that interact with specific proteins, however it is limited in that it does not map the precise

regions of the RNA molecule that form the interaction. On the other hand, CLIP provides a method to map the nucleo-

tide resolution of the binding sites that interact with specific proteins (Ule et al., 2003). Many variants of CLIP have

now been developed (Hafner et al., 2021) and have been extensively used to understand the molecular mechanisms of

lncRNA function, particularly in lncRNAs that are associated with human disease (reviewed in Jonas et al., 2020 and

J.-M. Carter et al., 2021). Much of the experimental data that has been generated from CLIP experiments in recent

years, has been integrated into public repositories such as GEO (Barrett et al., 2013; Edgar et al., 2002), CLIPdb (Yang

et al., 2015), and starBASE (Li et al., 2014; Yang et al., 2010). Large-scale projects such as ENCODE have mapped the

binding sites of hundreds of RBPs in certain human cell lines (Van Nostrand et al., 2020). Publicly available CLIP data

has also been integrated in computational frameworks for lncRNA analysis: for instance, eCLIP data from ENCODE

has been incorporated into LncLOOM to annotate evolutionarily conserved motifs that correspond to binding sites of

RBPs (Ross et al., 2021).

4.3 |Discovery of functional structured RNA elements in lncRNAs

Generally, it is still unclear to what extent secondary structure is important to lncRNA function (Rivas, 2021; Ulitsky &

Bartel, 2013). As it stands, the functional importance of structural elements have been experimentally studied in rela-

tively few lncRNAs (Table 3). Indeed, these studies have provided key insights into how structural elements modulate

the function and stability of some lncRNAs. One lncRNA where structure has proved to be relevant is MEG3. Based on

computational prediction, Zhang et al. reported that the MEG3 lncRNA contains three structural motifs (termed M1,

M2, and M3), two of which were experimentally validated to be required for p53 activation by MEG3 (Zhang

et al., 2010). When the primary sequence of M2 was replaced by an entirely unrelated sequence that artificially folded

into a similar structure, the transcript retained the functions of both p53 activation and growth suppression, in compar-

ison to almost complete loss of function when the motif was deleted. More recently, chemical probing and atomic force

microscopy revealed that the p53-activating core of MEG3 comprises two evolutionary conserved distal motifs that

interact by base complementarity to form pseudoknots (Uroda et al., 2019). Notably, one of these motifs overlapped the

M2 motif identified by Zhang et al. They showed that point mutations that break the long-range interactions between

these motifs disrupt the MEG3 architecture and severely impair p53 activation, although the mechanism by which this

occurs still needs to be elucidated. In addition to mediating the function of lncRNA molecules, structural elements have

also been found to be essential to lncRNA stability. As an example, RNA crystallization of the 30-end of MALAT1 rev-

ealed a triple helix structure that protects it from RNase degradation (Brown et al., 2014).

Multiple methods have been developed to determine the secondary structure of RNA molecules in vitro and in cells

(reviewed in Zampetaki et al., 2018). So far, the majority of structural elements found in lncRNA sequences have been

solved using a combination of chemical and enzymatic probing such as: SHAPE-seq, DMS-seq, and PARS. To overcome

14 of 25 ROSS AND ULITSKY

the challenges imposed from the length of lncRNA sequences Novikova et al. (2013) introduced the (3S) shotgun

approach, which consists of two steps to determine subdomains that adopt modular folds in the context of full-length

RNA structure. In the first step, the entire lncRNA sequence is chemically probed using SHAPE and DMS. Next, the

sequence is divided into overlapping segments which are then probed individually to determine local structural

domains that are subsequently compared to their folds obtained from the full-length sequence. This approach has been

successfully applied to several lncRNAs including Cyrano (Jones et al., 2020), COOLAIR (Hawkes et al., 2016), and

Braveheart (Xue et al., 2016). In particular, the structural study of Braveheart revealed key insights into its mode of

action. It was found that Braveheart adopts a modular secondary structure and contains a 50asymmetric G-rich internal

loop (AGIL) that recruits and antagonizes the zinc-finger protein CNBP, a TF that represses cardiac differentiation, to

promote cardiac fate (Xue et al., 2016). More recent technologies have now advanced our capability to determine RNA

structures in living cells and thereby map long-range RNA interactions across the transcriptome (Lu et al., 2016; Ziv

et al., 2018). One such approach, PARIS (psoralen analysis of RNA interactions and structures) showed that XIST A-

repeats form complex inter-repeat duplexes that span many kilobases across the X chromosome. The XIST A-repeats

coordinate the assembly of SPEN, an Xist-binding transcriptional repressor that is required for X chromosome inactiva-

tion (XCI), across the X chromosome (Lu et al., 2016). The functional importance of structural motifs in RNA has also

been demonstrated in viral genomic RNA (vRNA) and viral mRNA (reviewed in Ferhadian et al., 2018; Fern

andez-

Sanlés et al., 2017). For example, in vivo enzymatic probing revealed that the efficient replication and infectivity of

Influenza A virus (IAV) is dependent on the formation of stable structural motifs in the viral mRNAs (Simon

et al., 2019), while conserved G-quadruplex structures have been reported in IAV vRNA (Tomaszewska et al., 2021).

TABLE 3 Examples of structural elements that have been supported by experimental data

LncRNA Description Method References

MEG3 The p53-activating core comprises two evolutionary conserved

distal motifs that interact to form pseudoknots

SHAPE, probing with 1M7,

NMIA, 1M6, and DMS

Hydroxyl radical footprinting

Atomic force microscopy

(Uroda

et al., 2019)

XIST XIST A-region contains repeated stem-loop structures that

recruit the PRC2 complex

Probing with RNAse V1 and DMS

Targeted Structure-Seq

(Fang et al., 2015;

Maenner

et al., 2010)

XIST A-repeat inter-repeat duplexes that span across the X-

chromosome to facilitate the assembly SPEN

PARIS (Lu et al., 2016)

CYRANO Cloverleaf structure (100 nt) that contains the highly

complementary site to miR-7

3S shotgun approach, SHAPE,

probing with 1M7, NMIA, and

S1 nuclease

(Jones

et al., 2020)

SRA Complex structural organization consisting of four domains

that contain multiple loops and helical structures

SHAPE, probing with DMS and

RNase V1

(Novikova

et al., 2012)

HOTAIR Four independently-folded highly structured domains SHAPE, probing with DMS and

terbium

(Somarowthu

et al., 2015)

COOLAIR Evolutionarily conserved complex structure with multi-helix

junctions

3S shotgun approach, SHAPE,

probing with 1M7

(Hawkes

et al., 2016)

Braveheart Adopts a modular fold that contains a short asymmetric G-rich

internal loop that binds to CNBP

SHAPE, probing with 1M7 and

DMS

(Xue et al., 2016)

ROX1 and

ROX2

Evolutionarily conserved domains that contain tandem stem-

loops, in which the roXbox motif is embedded

SHAPE and parallel analysis of

RNA structure (PARS) analysis

(Ilik et al., 2013)

MALAT1 30-end forms triple helix structures that stabilize the lncRNA RNA crystallization (Brown

et al., 2014)

30-end contains evolutionary conserved tRNA-like structures SHAPE, probing with DMS (Zhang

et al., 2017)

PAN Cis-acting elements MRE and ENE are organized within a

branched secondary structure comprised of three domains

that contain multiple hairpin loops

SHAPE-mutational profiling

(SHAPE-MaP)

(Sztuba-Solinska

et al., 2017)

ROSS AND ULITSKY 15 of 25

Similarly, SHAPE analysis revealed conserved structural motifs in the RNA genome and ORF of hepatitis C virus that

contribute to viral fitness and regulate specific stages in the life cycle of the virus (Mauger et al., 2015; Pirakitikulr

et al., 2016).

Overall experimental studies corroborate the role of structural elements in lncRNA function, however some features

may be worth exploring in future studies. First, previous analysis of the sequence and context preferences of RBPs has

shown that RBPs often bind low-complexity RNA motifs and their specificity is influenced by the surrounding sequence

context including secondary structures and their flanking nucleotides (Dominguez et al., 2018). Therefore, to fully

understand how the functions of lncRNAs are encoded in their sequences, we also need to determine the factors that

control the functionality of structured domains that are positioned in precise locations within their lncRNA molecules.

Second, there is evidence to suggest that the structures of lncRNA molecules are dynamic and adopt different conforma-

tions across different subcellular compartments and/or cell states, subject to modifications or interactions with chroma-

tin, proteins, or other RNA molecules (Sun et al., 2019). This kind of structural dynamics could dictate temporally

variable roles of the same lncRNA, or nonmutually exclusive functions that are performed in different compartments of

the cell. What may also be of interest is the relationship between structural changes that occur due to species-specific

factors and the diversification of lncRNA function. Substantial levels of species-specific structural divergence has been

previously reported. For example, the structures of small RNAs that are highly conserved between human and mouse

undergo both shared and unique conformational changes (Sun et al., 2019).

4.4 |Perturb-and-rescue validates motifs that mediate regulatory activity of specific

lncRNAs

Perturbation and rescue experiments provide a powerful approach to validate the function of conserved motifs and have

been included in multiple studies to further understand how selected lncRNAs perform their function (Figure 2c). The

Mendell Lab recently used an array of comparative genetic rescue experiments to interrogate two alternative pathways

that have been proposed for NORAD function (Elguindy et al., 2019), a lncRNA that is required for genome stability in

mammals (Lee et al., 2016). As mentioned earlier in this review, several studies have now found that NORAD acts in

the cytoplasm as a binding decoy for PUM1 and PUM2 proteins, which in turn reduces the repression of PUM targets

including key regulators of mitosis, DNA repair, and DNA replication genes (Elguindy et al., 2019; Kopp et al., 2019;

Lee et al., 2016; Tichon et al., 2016,2018). This mechanism is strongly supported by the high number of PREs located

in the NORAD lncRNA. Alternatively, Munschauer et al. proposed that the regulation of genomic stability by NORAD

is mediated through NORAD:RBMX interaction that facilitates the assembly of a ribonucleoprotein (RNP) complex in

the nucleus (Munschauer et al., 2018), which includes proteins such as Topoisomerase I (TOP1) that are critical for

genome maintenance. Similarly, this mechanism is supported by an RBMX binding site that spans the first 898 nt (15%)

of the NORAD sequence. To directly interrogate the importance of PUM1/PUM2 and RBMX binding for NORAD func-

tion, Elguindy et al. generated a series of mutant NORAD constructs lacking either PUM or RBMX binding sites; as well

as truncated constructs of domains that contain multiple PREs or the RBMX binding site. These constructs were inte-

grated as potential endogenous rescues into the AAVS1/PPP1R12C locus of HCT116 cells. Following the depletion of

endogenous NORAD transcripts using CRISPR interference (CRISPRi), it was found that the PRE-mutant construct was

not able to rescue the increase in aneuploid cells, while the construct lacking only the RBMX binding site fully restored

genome stability. Furthermore, no increase in aneuploid cells was observed after knockdown of NORAD in cells that

expressed the truncated fragment containing one PRE-rich domain (referred to as the ND4 domain). This demonstrated

the ability of the ND4 domain to function independently from the full NORAD transcript. In contrast to ND4, expres-

sion of the 50fragment comprising only the RBMX binding site had no rescue activity. These results by Elguindy

et al. (2019) strongly suggest that NORAD interaction with PUM is critical to maintaining genome stability, while its

interaction with RBMX may be dispensable for this function.

In the study of lncRNAs that have low sequence homology between species, the demonstration that loss of function

of a lncRNA in one species can be rescued by the exogenous expression of the homologue from different species is a

compelling approach to ascertain that the function of the lncRNA may indeed be encoded in more subtle sequence fea-

tures that are common to both orthologs. This concept was illustrated by a comprehensive analysis of roX1 and roX2

lncRNAs that correlated changes in their binding potency and their functionality in diverse Drosophilid species across

40 million years of evolution (Quinn et al., 2016). Although roX1 and roX2 functionality is conserved across species, the

sequence identity between orthologs is similar to that of random sequences. Likewise, roX1 and roX2 differ greatly in

16 of 25 ROSS AND ULITSKY

sequence and size, yet they initially appeared to be functionally redundant in D. melanogaster and disruption of either

lncRNA results in failed dosage compensation and male-specific lethality (Meller & Rattner, 2002), though more recent

studies found some hierarchy in their functional importance (Valsecchi et al., 2021). The functionality of roX lncRNAs

has been attributed to repeats of the short roXbox motif (8 nt), that is embedded in tandem stem-loop structures across

their sequences (Ilik et al., 2013; Seung-Won et al., 2008). Quinn et al. used genomic occupancy maps in four species to

show that the targeting of the roX lncRNAs to the X chromosome is conserved, however their precise binding sites dif-

fer considerably across the species. Nonetheless, when roX lncRNAs from other species were introduced into roX-null

D. melanogaster, they were able to modestly rescue ~20% of the roX-null D. melanogaster male mutants by binding to

D. melanogaster high-affinity sites (HASs; Quinn et al., 2016). Importantly, the modest cross-species rescue was consis-

tent with prior reports that rescue efficiency decreases with increasing evolutionary distance (Seung-Won et al., 2008),

suggesting that functionality decreases as the sequence similarity degrades but is still maintained at a lower level across

evolution. Notably, rescue efficiency was substantially improved by chimeric constructs with an increased number of

stem-loops. Furthermore, when the genomic occupancy maps of roX1 and roX2 were compared within the same spe-

cies, it was found that in certain species, including D. virilis and D. busckii, the redundancy between the two lncRNAs

had degenerated and roX1 binding was less potent. Interestingly, genetic rescue with engineered constructs that con-

tained varying numbers of the repetitive stem-loop structure from different ortholog species, confirmed that the

decreasing potency of roX1 in these species correlated with loss of stem-loops and roXbox motifs. These results suggest

that D. virilis and D. busckii roX1 lncRNAs have vestigial function due to loss of key sequence elements, while the pres-

ervation of these elements in in D. melanogaster has maintained the overall roX1-roX2 functional redundancy (Quinn

et al., 2016). As another example, a cross-species rescue was also used to characterize the function of TERMINATOR

and PUNISHER, two lncRNAs that play a critical role in cardiovascular development across vertebrates (Kurian

et al., 2015). Both lncRNAs are conserved across vertebrates and PhyloP analysis revealed relatively short regions (250–

500 bp) conserved across their sequences. Loss-of-function experiments in zebrafish embryos using morpholino anti-

sense oligonucleotides (MOs) against the conserved regions and splice sites in terminator compromised development at

the gastrulation stage and resulted in >70% lethality, while MO injections targeting punisher resulted in severe defects

in branching and vessel formation. Both morphant phenotypes observed in zebrafish embryos were sufficiently rescued

upon coinjection with the respective human lncRNA sequences of TERMINATOR and PUNISHER. A similar approach

was also previously used to investigate if Cyrano lncRNA has conserved function in human and zebrafish embryonic

development, given that short patches of conservation were interspersed across their sequences (Ulitsky et al., 2011).

Another powerful way to ascertain the essential function of a specific sequence domain, is to rescue the perturbed phe-

notype by introducing a similar domain encoded from an alternative gene that performs a similar function. This approach

wasrecentlyusedtoestablishanevolutionarylinkbetweenXCIbyXist lncRNA and RNA-mediated endogenous retrovirus

(ERV) silencing (Carter et al., 2020). The authors show that SPEN also interacts with structural motifs in ERV RNA to

recruit chromatin silencing machinery to loci from which ERVs are transcribed. These structural motifs are similar to the

A-repeats in Xist lncRNA, which are derived from TEs and required for the recruitment of Spen and subsequent gene

silencing in cis (Giorgetti et al., 2016;Wutzetal.,2002). Using CRISPR-Cas9, they show that knock-in of an ERV-derived

Spen binding site into an A-repeat deficient Xist is sufficient to rescue interaction with Spen and restores strict local gene

silencing in cis. Overall their findings suggest that Xist mediates XCI by repurposing antiviral chromatin silencing machin-

ery and that Xist acquired its ability to silence genes in cis from TEs derived from ancient ERVs. This confirms previous

suggestions that the A-repeat sequence is derived from the insertion and duplication of an ERV (Elisaphenko et al., 2008),

and sheds light on how ERVs have impacted the evolution of functional lncRNAs.

5|CONCLUSION

Progress in understanding the biology of individual lncRNA genes has given rise to specific concepts of what types of

functional elements are expected to be found in lncRNA genes. These have been the basis of computational tools

looking for short conserved elements, for evidence of selection on secondary structures, and/or for repeated occurrences

of short elements. Still the number of lncRNAs for which there is convincing evidence of a specific mode of action and

the sequences underlying it remains very low, and they correspond to a small minority of lncRNAs with reported func-

tions. Since this is now the main frontier in lncRNA research, we expect that many additional functional domains

within lncRNAs will be characterized in the near future, leading to both the refinement of the computational and

experimental tools developed so far, and catalyzing development of new methodologies. These will likely also be able to

ROSS AND ULITSKY 17 of 25

go beyond discovery of individual domains and decode combinations of functional elements that co-exist within the

same lncRNAs, as well as hierarchical grouping of functional domains found in different lncRNAs into families with

different granularity levels, similar to the situation in the research of domains in proteins. Once such tools become

available, they will enable large-scale prediction of lncRNA mode of action and possibly even functions. Since valida-

tion of predicted functional aspects is typically easier than their de novo experimental discovery, these tools are likely

to play instrumental roles in a leap forward of research programs focused on lncRNA biology.

ACKNOWLEDGMENT

The authors thank Yoav Lubelsky and members of the Ulitsky laboratory for helpful discussions and comments on the

manuscript.

CONFLICT OF INTEREST

The authors have declared no conflicts of interest for this article.

AUTHOR CONTRIBUTIONS

Caroline Jane Ross: Conceptualization (equal); visualization (lead); writing –original draft (lead); writing –review

and editing (equal). Igor Ulitsky: Conceptualization (equal); funding acquisition (lead); supervision (lead); writing –

review and editing (lead).

DATA AVAILABILITY STATEMENT

Data sharing is not applicable to this article as no new data were created or analyzed in this study.

ORCID

Caroline Jane Ross https://orcid.org/0000-0003-4354-6372

Igor Ulitsky https://orcid.org/0000-0003-0555-6561

RELATED WIRESARTICLES

Evolutionary clues in lncRNAs

Cytoplasmic functions of long noncoding RNAs

Long noncoding RNAs in mammalian cells: what, where, and why?

REFERENCES

Amaral, P. P., Leonardi, T., Han, N., Viré, E., Gascoigne, D. K., Arias-Carrasco, R., Büscher, M., Pandolfini, L., Zhang, A., Pluchino, S.,

Maracaja-Coutinho, V., Nakaya, H. I., Hemberg, M., Shiekhattar, R., Enright, A. J., & Kouzarides, T. (2018). Genomic positional conser-

vation identifies topological anchor point RNAs linked to developmental loci. Genome Biology,19(1), 32.

Amarasinghe, S. L., Su, S., Dong, X., Zappia, L., Ritchie, M. E., & Gouil, Q. (2020). Opportunities and challenges in long-read sequencing data

analysis. Genome Biology,21(1), 30.

Andrews, R. J., Roche, J., & Moss, W. N. (2018). ScanFold: An approach for genome-wide discovery of local RNA structural elements—

Applications to Zika virus and HIV. PeerJ,6, e6136.

Barrett, T., Wilhite, S. E., Ledoux, P., Evangelista, C., Kim, I. F., Tomashevsky, M., Marshall, K. A., Phillippy, K. H., Sherman, P. M.,

Holko, M., Yefanov, A., Lee, H., Zhang, N., Robertson, C. L., Serova, N., Davis, S., & Soboleva, A. (2013). NCBI GEO: Archive for func-

tional genomics data sets—Update. Nucleic Acids Research,41(Database issue), D991–D995.

Bartel, D. P. (2009). MicroRNAs: Target recognition and regulatory functions. Cell,136(2), 215–233.

Bhartiya, D., & Scaria, V. (2016). Genomic variations in non-coding RNAs: Structure, function and regulation. Genomics,107(2–3), 59–68.

Bitetti, A., Mallory, A. C., Golini, E., Carrieri, C., Carreño Gutiérrez, H., Perlas, E., Pérez-Rico, Y. A., Tocchini-Valentini, G. P., Enright, A. J.,

Norton, W. H. J., Mandillo, S., O'Carroll, D., & Shkumatava, A. (2018). MicroRNA degradation by a conserved target RNA regulates ani-

mal behavior. Nature Structural & Molecular Biology,25(3), 244–251.

Brown, J. A., Bulkley, D., Wang, J., Valenstein, M. L., Yario, T. A., Steitz, T. A., & Steitz, J. A. (2014). Structural insights into the stabilization

of MALAT1 noncoding RNA by a bipartite triple helix. Nature Structural & Molecular Biology,21(7), 633–640.

Brown, J. B., Boley, N., Eisman, R., May, G. E., Stoiber, M. H., Duff, M. O., Booth, B. W., Wen, J., Park, S., Suzuki, A. M., Wan, K. H., Yu, C.,

Zhang, D., Carlson, J. W., Cherbas, L., Eads, B. D., Miller, D., Mockaitis, K., Roberts, J., …Celniker, S. E. (2014). Diversity and dynamics

of the Drosophila transcriptome. Nature,512(7515), 393–399.

Bryzghalov, O., Makałowska, I., & Szcze

sniak, M. W. (2021). lncEvo: Automated identification and conservation study of long noncoding

RNAs. BMC Bioinformatics,22(1), 59.

18 of 25 ROSS AND ULITSKY

Bryzghalov, O., Szcze

sniak, M. W., & Makałowska, I. (2020). SyntDB: Defining orthologues of human long noncoding RNAs across primates.

Nucleic Acids Research,48(D1), D238–D245.

Bu, D., Luo, H., Jiao, F., Fang, S., Tan, C., Liu, Z., & Zhao, Y. (2015). Evolutionary annotation of conserved long non-coding RNAs in major

mammalian species. Science China. Life Sciences,58(8), 787–798.

Cabili, M. N., Trapnell, C., Goff, L., Koziol, M., Tazon-Vega, B., Regev, A., & Rinn, J. L. (2011). Integrative annotation of human large inter-

genic noncoding RNAs reveals global properties and specific subclasses. Genes & Development,25(18), 1915–1927.

Cao, M., Zhao, J., & Hu, G. (2019). Genome-wide methods for investigating long noncoding RNAs. Biomedicine & Pharmacotherapy =Biome-

decine & Pharmacotherapie,111, 395–401.

Carninci, P., & Hayashizaki, Y. (2007). Noncoding RNA transcription beyond annotated genes. Current Opinion in Genetics & Development,

17(2), 139–144.

Carter, A. C., Xu, J., Nakamoto, M. Y., Wei, Y., Zarnegar, B. J., Shi, Q., Broughton, J. P., Ransom, R. C., Salhotra, A., Nagaraja, S. D., Li, R.,

Dou, D. R., Yost, K. E., Cho, S.-W., Mistry, A., Longaker, M. T., Khavari, P. A., Batey, R. T., Wuttke, D. S., & Chang, H. Y. (2020). Spen

links RNA-mediated endogenous retrovirus silencing and X chromosome inactivation. eLife,9, e54508. https://doi.org/10.7554/eLife.

54508

Carter, J.-M., Ang, D. A., Sim, N., Budiman, A., & Li, Y. (2021). Approaches to identify and characterise the post-transcriptional roles of

lncRNAs in cancer. Non-Coding RNA,7(1), 19. https://doi.org/10.3390/ncrna7010019

Chen, J., Shishkin, A. A., Zhu, X., Kadri, S., Maza, I., Guttman, M., Hanna, J. H., Regev, A., & Garber, M. (2016). Evolutionary analysis

across mammals reveals distinct classes of long non-coding RNAs. Genome Biology,17, 19.

Chodroff, R. A., Goodstadt, L., Sirey, T. M., Oliver, P. L., Davies, K. E., Green, E. D., Moln

ar, Z., & Ponting, C. P. (2010). Long noncoding

RNA genes: Conservation of sequence and brain expression among diverse amniotes. Genome Biology,11(7), R72.

Chu, C., Qu, K., Zhong, F. L., Artandi, S. E., & Chang, H. Y. (2011). Genomic maps of long noncoding RNA occupancy reveal principles of

RNA-chromatin interactions. Molecular Cell,44(4), 667–678.

Chu, C., Quinn, J., & Chang, H. Y. (2012). Chromatin isolation by RNA purification (ChIRP). Journal of Visualized Experiments: JoVE,61,

3912. https://doi.org/10.3791/3912

Church, D. M., Goodstadt, L., Hillier, L. W., Zody, M. C., Goldstein, S., She, X., Bult, C. J., Agarwala, R., Cherry, J. L., DiCuccio, M.,

Hlavina, W., Kapustin, Y., Meric, P., Maglott, D., Birtle, Z., Marques, A. C., Graves, T., Zhou, S., Teague, B., …Mouse Genome Sequenc-

ing Consortium. (2009). Lineage-specific biology revealed by a finished genome assembly of the mouse. PLoS Biology,7(5), e1000112.

Chureau, C., Prissette, M., Bourdet, A., Barbe, V., Cattolico, L., Jones, L., Eggen, A., Avner, P., & Duret, L. (2002). Comparative sequence

analysis of the X-inactivation center region in mouse, human, and bovine. Genome Research,12(6), 894–908.

Clark, M. B., & Mattick, J. S. (2011). Long noncoding RNAs in cell biology. Seminars in Cell & Developmental Biology,22(4), 366–376.

Conrad, N. K., Mili, S., Marshall, E. L., Shu, M.-D., & Steitz, J. A. (2006). Identification of a rapid mammalian deadenylation-dependent

decay pathway and its inhibition by a viral RNA element. Molecular Cell,24(6), 943–953.

Conrad, N. K., & Steitz, J. A. (2005). A Kaposi's sarcoma virus RNA element that increases the nuclear abundance of intronless transcripts.

The EMBO Journal,24(10), 1831–1841.

Constanty, F., & Shkumatava, A. (2021). lncRNAs in development and differentiation: From sequence motifs to functional characterization.

Development,148(1), dev182741. https://doi.org/10.1242/dev.182741

Diederichs, S. (2014). The four dimensions of noncoding RNA conservation. Trends in Genetics: TIG,30(4), 121–123.

Dominguez, D., Freese, P., Alexis, M. S., Su, A., Hochman, M., Palden, T., Bazile, C., Lambert, N. J., Van Nostrand, E. L., Pratt, G. A.,

Yeo, G. W., Graveley, B. R., & Burge, C. B. (2018). Sequence, structure, and context preferences of human RNA binding proteins. Molecu-

lar Cell,70(5), 854–867.e9.

Doshi, K. J., Cannone, J. J., Cobaugh, C. W., & Gutell, R. R. (2004). Evaluation of the suitability of free-energy minimization using nearest-

neighbor energy parameters for RNA secondary structure prediction. BMC Bioinformatics,5, 105.

Eddy, S. R., & Durbin, R. (1994). RNA sequence analysis using covariance models. Nucleic Acids Research,22(11), 2079–2088.

Edgar, R., Domrachev, M., & Lash, A. E. (2002). Gene expression omnibus: NCBI gene expression and hybridization array data repository.

Nucleic Acids Research,30(1), 207–210.

Elguindy, M. M., Kopp, F., Goodarzi, M., Rehfeld, F., Thomas, A., Chang, T.-C., & Mendell, J. T. (2019). PUMILIO, but not RBMX, binding

is required for regulation of genomic stability by noncoding RNA NORAD. eLife,8, e48625. https://doi.org/10.7554/eLife.48625

Elisaphenko, E. A., Kolesnikov, N. N., Shevchenko, A. I., Rogozin, I. B., Nesterova, T. B., Brockdorff, N., & Zakian, S. M. (2008). A dual ori-

gin of the Xist gene from a protein-coding gene and a set of transposable elements. PLoS One,3(6), e2521.

Emms, D. M., & Kelly, S. (2015). OrthoFinder: Solving fundamental biases in whole genome comparisons dramatically improves orthogroup

inference accuracy. Genome Biology,16, 157.

Engreitz, J., Lander, E. S., & Guttman, M. (2015). RNA antisense purification (RAP) for mapping RNA interactions with chromatin. Methods

in Molecular Biology,1262, 183–197.

Engreitz, J. M., Pandya-Jones, A., McDonel, P., Shishkin, A., Sirokman, K., Surka, C., Kadri, S., Xing, J., Goren, A., Lander, E. S., Plath, K., &

Guttman, M. (2013). The Xist lncRNA exploits three-dimensional genome architecture to spread across the X chromosome. Science,

341(6147), 1237973.

Engreitz, J. M., Sirokman, K., McDonel, P., Shishkin, A. A., Surka, C., Russell, P., Grossman, S. R., Chow, A. Y., Guttman, M., &

Lander, E. S. (2014). RNA-RNA interactions enable specific targeting of noncoding RNAs to nascent pre-mRNAs and chromatin sites.

Cell,159(1), 188–199.

ROSS AND ULITSKY 19 of 25

Fang, R., Moss, W. N., Rutenberg-Schoenberg, M., & Simon, M. D. (2015). Probing Xist RNA structure in cells using targeted structure-Seq.

PLoS Genetics,11(12), e1005668.

Ferhadian, D., Contrant, M., Printz-Schweigert, A., Smyth, R. P., Paillart, J.-C., & Marquet, R. (2018). Structural and functional motifs in

influenza virus RNAs. Frontiers in Microbiology,9, 559.

Fern

andez-Sanlés, A., Ríos-Marco, P., Romero-L

opez, C., & Berzal-Herranz, A. (2017). Functional information stored in the conserved struc-

tural RNA domains of Flavivirus genomes. Frontiers in Microbiology,8, 546.

Gardner, P. P., Fasold, M., Burge, S. W., Ninova, M., Hertel, J., Kehr, S., Steeves, T. E., Griffiths-Jones, S., & Stadler, P. F. (2015). Conserva-

tion and losses of non-coding RNAs in avian genomes. PLoS One,10(3), e0121797.

Gil, N., & Ulitsky, I. (2018). Production of spliced long noncoding RNAs specifies regions with increased enhancer activity. Cell Systems,7(5),

537–547.e3.

Giorgetti, L., Lajoie, B. R., Carter, A. C., Attia, M., Zhan, Y., Xu, J., Chen, C. J., Kaplan, N., Chang, H. Y., Heard, E., & Dekker, J. (2016).

Structural organization of the inactive X chromosome in the mouse. Nature,535(7613), 575–579.

Grabherr, M. G., Haas, B. J., Yassour, M., Levin, J. Z., Thompson, D. A., Amit, I., Adiconis, X., Fan, L., Raychowdhury, R., Zeng, Q.,

Chen, Z., Mauceli, E., Hacohen, N., Gnirke, A., Rhind, N., di Palma, F., Birren, B. W., Nusbaum, C., Lindblad-Toh, K., …Regev, A.

(2011). Full-length transcriptome assembly from RNA-Seq data without a reference genome. Nature Biotechnology,29(7), 644–652.

Graindorge, A., Pinheiro, I., Nawrocka, A., Mallory, A. C., Tsvetkov, P., Gil, N., Carolis, C., Buchholz, F., Ulitsky, I., Heard, E.,

Taipale, M., & Shkumatava, A. (2019). In-cell identification and measurement of RNA-protein interactions. Nature Communications,

10(1), 5317.

Grant, C. E., Bailey, T. L., & Noble, W. S. (2011). FIMO: Scanning for occurrences of a given motif. Bioinformatics,27(7), 1017–1018.

Griffiths-Jones, S., Bateman, A., Marshall, M., Khanna, A., & Eddy, S. R. (2003). Rfam: An RNA family database. Nucleic Acids Research,

31(1), 439–441.

Grinman, E., Nakahata, Y., Avchalumov, Y., Espadas, I., Swarnkar, S., Yasuda, R., & Puthanveettil, S. V. (2021). Activity-regulated synaptic

targeting of lncRNA ADEPTR mediates structural plasticity by localizing Sptn1 and AnkB in dendrites. Science Advances,7(16),

eabf0605. https://doi.org/10.1126/sciadv.abf0605

Gudenas, B. L., & Wang, L. (2018). Prediction of LncRNA subcellular localization with deep learning from sequence features. Scientific

Reports,8(1), 16385.

Guo, J., Liu, Z., & Gong, R. (2019). Long noncoding RNA: An emerging player in diabetes and diabetic kidney disease. Clinical Science,

133(12), 1321–1339.

Gupta, S., Stamatoyannopoulos, J. A., Bailey, T. L., & Noble, W. S. (2007). Quantifying similarity between motifs. Genome Biology,8(2), R24.

Gutell, R. R., Lee, J. C., & Cannone, J. J. (2002). The accuracy of ribosomal RNA comparative structure models. Current Opinion in Structural

Biology,12(3), 301–310.

Guttman, M., Amit, I., Garber, M., French, C., Lin, M. F., Feldser, D., Huarte, M., Zuk, O., Carey, B. W., Cassady, J. P., Cabili, M. N.,

Jaenisch, R., Mikkelsen, T. S., Jacks, T., Hacohen, N., Bernstein, B. E., Kellis, M., Regev, A., Rinn, J. L., & Lander, E. S. (2009). Chroma-

tin signature reveals over a thousand highly conserved large non-coding RNAs in mammals. Nature,458(7235), 223–227.

Guttman, M., & Rinn, J. L. (2012). Modular regulatory principles of large non-coding RNAs. Nature,482(7385), 339–346.

Haerty, W., & Ponting, C. P. (2015). Unexpected selection to retain high GC content and splicing enhancers within exons of multiexonic

lncRNA loci. RNA,21, 320–332. https://doi.org/10.1261/rna.047324.114

Hafner, M., Katsantoni, M., Köster, T., Marks, J., Mukherjee, J., Staiger, D., Ule, J., & Zavolan, M. (2021). CLIP and complementary methods.

Nature Reviews Methods Primers,1(1), 1–23.

Hawkes, E. J., Hennelly, S. P., Novikova, I. V., Irwin, J. A., Dean, C., & Sanbonmatsu, K. Y. (2016). COOLAIR antisense RNAs form evolu-

tionarily conserved elaborate secondary structures. Cell Reports,16(12), 3087–3096.

Herrera-Úbeda, C., Marín-Barba, M., Navas-Pérez, E., Gravemeyer, J., Albuixech-Crespo, B., Wheeler, G. N., & Garcia-Fernàndez, J. (2019).

Microsyntenic clusters reveal conservation of lncRNAs in chordates despite absence of sequence conservation. Biology,8(3), 61. https://

doi.org/10.3390/biology8030061

Hezroni, H., Koppstein, D., Schwartz, M. G., Avrutin, A., Bartel, D. P., & Ulitsky, I. (2015). Principles of long noncoding RNA evolution

derived from direct comparison of Transcriptomes in 17 species. Cell Reports,11, 1110–1122. https://doi.org/10.1016/j.celrep.2015.04.023

Hezroni, H., Tov Perry, R. B., & Ulitsky, I. (2019). Long noncoding RNAs in development and regeneration of the neural lineage. Cold Spring

Harbor Symposia on Quantitative Biology,84, 165–177.

Hölzer, M., & Marz, M. (2019). De novo transcriptome assembly: A comprehensive cross-species comparison of short-read RNA-Seq assem-

blers. GigaScience,8(5), giz039. https://doi.org/10.1093/gigascience/giz039

Huang, M., Hou, J., Wang, Y., Xie, M., Wei, C., Nie, F., Wang, Z., & Sun, M. (2017). Long noncoding RNA LINC00673 is activated by SP1

and exerts oncogenic properties by interacting with LSD1 and EZH2 in gastric cancer. Molecular Therapy: The Journal of the American

Society of Gene Therapy,25(4), 1014–1026.

Ilik, I. A., Quinn, J. J., Georgiev, P., Tavares-Cadete, F., Maticzka, D., Toscano, S., Wan, Y., Spitale, R. C., Luscombe, N., Backofen, R.,

Chang, H. Y., & Akhtar, A. (2013). Tandem stem-loops in roX RNAs act together to mediate X chromosome dosage compensation in dro-

sophila. Molecular Cell,51(2), 156–173.

Iyer, M. K., Niknafs, Y. S., Malik, R., Singhal, U., Sahu, A., Hosono, Y., Barrette, T. R., Prensner, J. R., Evans, J. R., Zhao, S., Poliakov, A.,

Cao, X., Dhanasekaran, S. M., Wu, Y.-M., Robinson, D. R., Beer, D. G., Feng, F. Y., Iyer, H. K., & Chinnaiyan, A. M. (2015). The land-

scape of long noncoding RNAs in the human transcriptome. Nature Genetics,47(3), 199–208.

20 of 25 ROSS AND ULITSKY

Jin, F., Li, J., Zhang, Y.-B., Liu, X., Cai, M., Liu, M., Li, M., Ma, C., Yue, R., Zhu, Y., Lai, R., Wang, Z., Ji, X., Wei, H., Dong, J., Liu, Z.,

Wang, Y., Sun, Y., & Wang, X. (2021). A functional motif of long noncoding RNA Nron against osteoporosis. Nature Communications,

12(1), 1–15.

Jonas, K., Calin, G. A., & Pichler, M. (2020). RNA-binding proteins as important regulators of long non-coding RNAs in cancer. International

Journal of Molecular Sciences,21(8), 2969. https://doi.org/10.3390/ijms21082969

Jones, A. N., Pisignano, G., Pavelitz, T., White, J., Kinisu, M., Forino, N., Albin, D., & Varani, G. (2020). An evolutionarily conserved RNA

structure in the functional core of the lincRNA Cyrano. RNA,26(9), 1234–1246.

Kato, M., & Carninci, P. (2020). Genome-wide technologies to study RNA–chromatin interactions. Non-Coding RNA,6(2), 20.

Kirk, J. M., Kim, S. O., Inoue, K., Smola, M. J., Lee, D. M., Schertzer, M. D., Wooten, J. S., Baker, A. R., Sprague, D., Collins, D. W.,

Horning, C. R., Wang, S., Chen, Q., Weeks, K. M., Mucha, P. J., & Calabrese, J. M. (2018). Functional classification of long non-coding

RNAs by k-mer content. Nature Genetics,50(10), 1474–1482.

Kleaveland, B., Shi, C. Y., Stefano, J., & Bartel, D. P. (2018). A network of noncoding regulatory RNAs acts in the mammalian brain. Cell,

174(2), 350–362.

Kopp, F., Elguindy, M. M., Yalvac, M. E., Zhang, H., Chen, B., Gillett, F. A., Lee, S., Sivakumar, S., Yu, H., Xie, Y., Mishra, P., Sahenk, Z., &

Mendell, J. T. (2019). PUMILIO hyperactivity drives premature aging of Norad-deficient mice. eLife,8,8.https://doi.org/10.7554/eLife.

42650

Kornienko, A. E., Dotter, C. P., Guenzl, P. M., Gisslinger, H., Gisslinger, B., Cleary, C., Kralovics, R., Pauler, F. M., & Barlow, D. P. (2016).

Long non-coding RNAs display higher natural expression variation than protein-coding genes in healthy humans. Genome Biology,

17, 14.

Kuo, C.-C., Hänzelmann, S., Sentürk Cetin, N., Frank, S., Zajzon, B., Derks, J.-P., Akhade, V. S., Ahuja, G., Kanduri, C., Grummt, I.,

Kurian, L., & Costa, I. G. (2019). Detection of RNA–DNA binding sites in long noncoding RNAs. Nucleic Acids Research,47(6), e32–e32.

Kurian, L., Aguirre, A., Sancho-Martinez, I., Benner, C., Hishida, T., Nguyen, T. B., Reddy, P., Nivet, E., Krause, M. N., Nelles, D. A.,

Esteban, C. R., Campistol, J. M., Yeo, G. W., & Belmonte, J. C. I. (2015). Identification of novel long noncoding RNAs underlying verte-

brate cardiovascular development. Circulation,131(14), 1278–1290.

Kutter, C., Watt, S., Stefflova, K., Wilson, M. D., Goncalves, A., Ponting, C. P., Odom, D. T., & Marques, A. C. (2012). Rapid turnover of long

noncoding RNAs and the evolution of gene expression. PLoS Genetics,8(7), e1002841.

Lee, J. T., & Bartolomei, M. S. (2013). X-inactivation, imprinting, and long noncoding RNAs in health and disease. Cell,152(6), 1308–1323.

Lee, S., Kopp, F., Chang, T.-C., Sataluri, A., Chen, B., Sivakumar, S., Yu, H., Xie, Y., & Mendell, J. T. (2016). Noncoding RNA NORAD regu-

lates genomic stability by sequestering PUMILIO proteins. Cell,164(1–2), 69–80.

Li, J.-H., Liu, S., Zhou, H., Qu, L.-H., & Yang, J.-H. (2014). starBase v2.0: Decoding miRNA-ceRNA, miRNA-ncRNA and protein–RNA inter-

action networks from large-scale CLIP-Seq data. In. Nucleic Acids Research,42(D1), D92–D97. https://doi.org/10.1093/nar/gkt1248

Liu, S. J., Dang, H. X., Lim, D. A., Feng, F. Y., & Maher, C. A. (2021). Long noncoding RNAs in cancer metastasis. Nature Reviews. Cancer,

21(7), 446–460.

Lorenz, R., Wolfinger, M. T., Tanzer, A., & Hofacker, I. L. (2016). Predicting RNA secondary structures from sequence and probing data.

Methods,103,86–98.

Lu, Z., Zhang, Q. C., Lee, B., Flynn, R. A., Smith, M. A., Robinson, J. T., Davidovich, C., Gooding, A. R., Goodrich, K. J., Mattick, J. S.,

Mesirov, J. P., Cech, T. R., & Chang, H. Y. (2016). RNA duplex map in living cells reveals higher-order Transcriptome structure. Cell,

165(5), 1267–1279.

Lubelsky, Y., & Ulitsky, I. (2018). Sequences enriched in Alu repeats drive nuclear localization of long RNAs in human cells. Nature,

555(7694), 107–111.

Lubelsky, Y., Zuckerman, B., & Ulitsky, I. (2021). High-resolution mapping of function and protein binding in an RNA nuclear enrichment

sequence. The EMBO Journal,40, e106357. https://doi.org/10.15252/embj.2020106357

Maenner, S., Blaud, M., Fouillen, L., Savoye, A., Marchand, V., Dubois, A., Sanglier-Cianférani, S., Van Dorsselaer, A., Clerc, P., Avner, P.,

Visvikis, A., & Branlant, C. (2010). 2-D structure of the a region of Xist RNA and its implication for PRC2 association. PLoS Biology,8(1),

e1000276.

Maier, D. (1978). The complexity of some problems on subsequences and supersequences. Journal of the ACM,25(2), 322–336.

Mauger, D. M., Golden, M., Yamane, D., Williford, S., Lemon, S. M., Martin, D. P., & Weeks, K. M. (2015). Functionally conserved architec-

ture of hepatitis C virus RNA genomes. Proceedings of the National Academy of Sciences of the United States of America,112(12), 3692–

3697.

McCown, P. J., Wang, M. C., Jaeger, L., & Brown, J. A. (2019). Secondary structural model of human MALAT1 reveals multiple structure–

function relationships. International Journal of Molecular Sciences,20(22), 5610.

McCullum, E. O., Williams, B. A. R., Zhang, J., & Chaput, J. C. (2010). Random mutagenesis by error-prone PCR. Methods in Molecular Biol-

ogy,634, 103–109.

McHugh, C. A., Chen, C.-K., Chow, A., Surka, C. F., Tran, C., McDonel, P., Pandya-Jones, A., Blanco, M., Burghard, C., Moradian, A.,

Sweredoski, M. J., Shishkin, A. A., Su, J., Lander, E. S., Hess, S., Plath, K., & Guttman, M. (2015). The Xist lncRNA interacts directly with

SHARP to silence transcription through HDAC3. Nature,521(7551), 232–236.

McLeay, R. C., & Bailey, T. L. (2010). Motif enrichment analysis: A unified framework and an evaluation on ChIP data. BMC Bioinformatics,

11, 165.

ROSS AND ULITSKY 21 of 25

Meller, V. H., & Rattner, B. P. (2002). The roX genes encode redundant male-specific lethal transcripts required for targeting of the MSL com-

plex. The EMBO Journal,21(5), 1084–1091.

Michel, F., & Westhof, E. (1990). Modelling of the three-dimensional architecture of group I catalytic introns based on comparative sequence

analysis. Journal of Molecular Biology,216(3), 585–610.

Mignone, F., Gissi, C., Liuni, S., & Pesole, G. (2002). Untranslated regions of mRNAs. Genome Biology,3(3), REVIEWS0004.

Mitton-Fry, R. M., DeGregorio, S. J., Wang, J., Steitz, T. A., & Steitz, J. A. (2010). Poly(A) tail recognition by a viral RNA element through

assembly of a triple helix. Science,330(6008), 1244–1247.

Mor

an, I., Akerman, I., van de Bunt, M., Xie, R., Benazra, M., Nammo, T., Arnes, L., Naki

c, N., García-Hurtado, J., Rodríguez-Seguí, S.,

Pasquali, L., Sauty-Colace, C., Beucher, A., Scharfmann, R., van Arensbergen, J., Johnson, P. R., Berry, A., Lee, C., Harkins, T., …

Ferrer, J. (2012). Human βcell transcriptome analysis uncovers lncRNAs that are tissue-specific, dynamically regulated, and abnormally

expressed in type 2 diabetes. Cell Metabolism,16(4), 435–448.

Müller, T., Boileau, E., Talyan, S., Kehr, D., Varadi, K., Busch, M., Most, P., Krijgsveld, J., & Dieterich, C. (2021). Updated and enhanced pig

cardiac transcriptome based on long-read RNA sequencing and proteomics. Journal of Molecular and Cellular Cardiology,150,23–31.

Munschauer, M., Nguyen, C. T., Sirokman, K., Hartigan, C. R., Hogstrom, L., Engreitz, J. M., Ulirsch, J. C., Fulco, C. P., Subramanian, V.,

Chen, J., Schenone, M., Guttman, M., Carr, S. A., & Lander, E. S. (2018). Publisher correction: The NORAD lncRNA assembles a topo-

isomerase complex critical for genome stability. Nature,563(7733), E32.

Mustafi, D., Kevany, B. M., Bai, X., Maeda, T., Sears, J. E., Khalil, A. M., & Palczewski, K. (2013). Evolutionarily conserved long intergenic

non-coding RNAs in the eye. Human Molecular Genetics,22(15), 2992–3002.

Nawrocki, E. P., Kolbe, D. L., & Eddy, S. R. (2009). Infernal 1.0: Inference of RNA alignments. Bioinformatics,25(10), 1335–1337.

Necsulea, A., Soumillon, M., Warnefors, M., Liechti, A., Daish, T., Zeller, U., Baker, J. C., Grützner, F., & Kaessmann, H. (2014). The evolu-

tion of lncRNA repertoires and expression patterns in tetrapods. Nature,505(7485), 635–640.

Noh, J. H., Kim, K. M., Abdelmohsen, K., Yoon, J.-H., Panda, A. C., Munk, R., Kim, J., Curtis, J., Moad, C. A., Wohler, C. M., Indig, F. E., de

Paula, W., Dudekula, D. B., De, S., Piao, Y., Yang, X., Martindale, J. L., de Cabo, R., & Gorospe, M. (2016). HuR and GRSF1 modulate

the nuclear export and mitochondrial localization of the lncRNA RMRP. Genes & Development,30(10), 1224–1239.

Noviello, T. M. R., Di Liddo, A., Ventola, G. M., Spagnuolo, A., D'Aniello, S., Ceccarelli, M., & Cerulo, L. (2018). Detection of long non–

coding RNA homology, a comparative study on alignment and alignment–free metrics. BMC Bioinformatics,19(1), 407.

Novikova, I. V., Dharap, A., Hennelly, S. P., & Sanbonmatsu, K. Y. (2013). 3S: Shotgun secondary structure determination of long non-coding

RNAs. Methods,63(2), 170–177.

Novikova, I. V., Hennelly, S. P., & Sanbonmatsu, K. Y. (2012). Structural architecture of the human long non-coding RNA, steroid receptor

RNA activator. Nucleic Acids Research,40(11), 5034–5051.

Oikonomopoulos, S., Bayega, A., Fahiminiya, S., Djambazian, H., Berube, P., & Ragoussis, J. (2020). Methodologies for transcript profiling

using long-read technologies. Frontiers in Genetics,11, 606.

Okazaki, Y., Furuno, M., Kasukawa, T., Adachi, J., Bono, H., Kondo, S., Nikaido, I., Osato, N., Saito, R., Suzuki, H., Yamanaka, I.,

Kiyosawa, H., Yagi, K., Tomaru, Y., Hasegawa, Y., Nogami, A., Schönbach, C., Gojobori, T., Baldarelli, R., …RIKEN Genome Explora-

tion Research Group Phase I & II Team. (2002). Analysis of the mouse transcriptome based on functional annotation of 60,770 full-

length cDNAs. Nature,420(6915), 563–573.

Paralkar, V. R., Mishra, T., Luan, J., Yao, Y., Kossenkov, A. V., Anderson, S. M., Dunagin, M., Pimkin, M., Gore, M., Sun, D., Konuthula, N.,

Raj, A., An, X., Mohandas, N., Bodine, D. M., Hardison, R. C., & Weiss, M. J. (2014). Lineage and species-specific long noncoding RNAs

during erythro-megakaryocytic development. Blood,123(12), 1927–1937.

Parisien, M., Wang, X., & Pan, T. (2013). Diversity of human tRNA genes from the 1000-genomes project. RNA Biology,10(12), 1853–1867.

Pedersen, J. S., Bejerano, G., Siepel, A., Rosenbloom, K., Lindblad-Toh, K., Lander, E. S., Kent, J., Miller, W., & Haussler, D. (2006). Identifi-

cation and classification of conserved RNA secondary structures in the human genome. PLoS Computational Biology,2(4), e33.

Pintacuda, G., Young, A. N., & Cerase, A. (2017). Function by structure: Spotlights on Xist long non-coding RNA. Frontiers in Molecular Bio-

sciences,4, 90.

Pirakitikulr, N., Kohlway, A., Lindenbach, B. D., & Pyle, A. M. (2016). The coding region of the HCV genome contains a network of regula-

tory RNA structures. Molecular Cell,62(1), 111–120.

Ponjavic, J., Oliver, P. L., Lunter, G., & Ponting, C. P. (2009). Genomic and transcriptional co-localization of protein-coding and long non-

coding RNA pairs in the developing brain. PLoS Genetics,5(8), e1000617.

Quinn, J. J., Ilik, I. A., Qu, K., Georgiev, P., Chu, C., Akhtar, A., & Chang, H. Y. (2014). Revealing long noncoding RNA architecture and

functions using domain-specific chromatin isolation by RNA purification. Nature Biotechnology,32(9), 933–940.

Quinn, J. J., Zhang, Q. C., Georgiev, P., & Ilik, I. A. (2016). Rapid evolutionary turnover underlies conserved lncRNA–genome interactions.

Genes Dev 30(2), 191–207. http://genesdev.cshlp.org/content/30/2/191.short

Ramanathan, M., Porter, D. F., & Khavari, P. A. (2019). Methods to study RNA–protein interactions. Nature Methods,16(3), 225–234.

Rivas, E. (2021). Evolutionary conservation of RNA sequence and structure. WIREs RNA,12(5), e1649. https://doi.org/10.1002/wrna.1649

Rivas, E., Clements, J., & Eddy, S. R. (2017). A statistical test for conserved RNA structure shows lack of evidence for structure in lncRNAs.

Nature Methods,14(1), 45–48.

Rivas, E., Clements, J., & Eddy, S. R. (2020). Estimating the power of sequence covariation for detecting conserved RNA structure. Bioinfor-

matics,36(10), 3072–3076.

22 of 25 ROSS AND ULITSKY

Rivas, E., & Eddy, S. R. (2000). Secondary structure alone is generally not statistically significant for the detection of noncoding RNAs. Bioin-

formatics,16(7), 583–605.

RNAcentral Consortium. (2021). RNAcentral 2021: Secondary structure integration, improved sequence search and new member databases.

Nucleic Acids Research,49(D1), D212–D220.

Rogers, E., Murrugarra, D., & Heitsch, C. (2017). Conditioning and robustness of RNA Boltzmann sampling under thermodynamic parame-

ter perturbations. Biophysical Journal,113(2), 321–329.

Rom, A., Melamed, L., Gil, N., Goldrich, M. J., Kadir, R., Golan, M., Biton, I., Perry, R. B.-T., & Ulitsky, I. (2019). Regulation of CHD2 expres-

sion by the Chaserr long noncoding RNA gene is essential for viability. Nature Communications,10(1), 5092.

Ross, C. J., Rom, A., Spinrad, A., Gelbard-Solodkin, D., Degani, N., & Ulitsky, I. (2021). Uncovering deeply conserved motif combinations in

rapidly evolving noncoding sequences. Genome Biology,22(1), 29.

Sarropoulos, I., Marin, R., Cardoso-Moreira, M., & Kaessmann, H. (2019). Developmental dynamics of lncRNAs across mammalian organs

and species. Nature,571(7766), 510–514.

Schroeder, S. J. (2018). Challenges and approaches to predicting RNA with multiple functional structures. RNA,24(12), 1615–1624.

Schuler, A., Ghanbarian, A. T., & Hurst, L. D. (2014). Purifying selection on splice-related motifs, not expression level nor RNA folding,

explains nearly all constraint on human lincRNAs. Molecular Biology and Evolution,31(12), 3164–3183.

Schultes, E. A., Spasic, A., Mohanty, U., & Bartel, D. P. (2005). Compact and ordered collapse of randomly generated RNA sequences. Nature

Structural & Molecular Biology,12(12), 1130–1136.

Seemann, S. E., Mirza, A. H., Hansen, C., Bang-Berthelsen, C. H., Garde, C., Christensen-Dalsgaard, M., Torarinsson, E., Yao, Z.,

Workman, C. T., Pociot, F., Nielsen, H., Tommerup, N., Ruzzo, W. L., & Gorodkin, J. (2017). The identification and functional annota-

tion of RNA structures conserved in vertebrates. Genome Research,27(8), 1371–1383.

Seung-Won, P., Kuroda, M. I., & Yongkyu, P. (2008). Regulation of histone H4 Lys16 acetylation by predicted alternative secondary struc-

tures in roX noncoding RNAs. Molecular and Cellular Biology,28(16), 4952–4962.

Shukla, C. J., McCorkindale, A. L., Gerhardinger, C., Korthauer, K. D., Cabili, M. N., Shechner, D. M., Irizarry, R. A., Maass, P. G., &

Rinn, J. L. (2018). High-throughput identification of RNA nuclear enrichment sequences. The EMBO Journal,37(6), e98452. https://doi.

org/10.15252/embj.201798452

Simon, L. M., Morandi, E., Luganini, A., Gribaudo, G., Martinez-Sobrido, L., Turner, D. H., Oliviero, S., & Incarnato, D. (2019). In vivo analy-

sis of influenza A mRNA secondary structures identifies critical regulatory motifs. Nucleic Acids Research,47(13), 7003–7017.

Simon, M. D. (2013). Capture hybridization analysis of RNA targets (CHART). Current Protocols in Molecular Biology. 2013, Chapter 21, Unit

21 21.25.

Simon, M. D., Wang, C. I., Kharchenko, P. V., West, J. A., Chapman, B. A., Alekseyenko, A. A., Borowsky, M. L., Kuroda, M. I., &

Kingston, R. E. (2011). The genomic binding sites of a noncoding RNA. Proceedings of the National Academy of Sciences of the

United States of America,108(51), 20497–20502.

Singh, N. N., & Singh, R. N. (2019). How RNA structure dictates the usage of a critical exon of spinal muscular atrophy gene. Biochimica et

Biophysica Acta, Gene Regulatory Mechanisms,1862(11–12), 194403.

Smola, M. J., Christy, T. W., Inoue, K., Nicholson, C. O., Friedersdorf, M., Keene, J. D., Lee, D. M., Calabrese, J. M., & Weeks, K. M. (2016).

SHAPE reveals transcript-wide interactions, complex structural domains, and protein interactions across the Xist lncRNA in living cells.

Proceedings of the National Academy of Sciences of the United States of America,113(37), 10322–10327.

Somarowthu, S., Legiewicz, M., Chill

on, I., Marcia, M., Liu, F., & Pyle, A. M. (2015). HOTAIR forms an intricate and modular secondary

structure. Molecular Cell,58(2), 353–361.

Statello, L., Guo, C.-J., Chen, L.-L., & Huarte, M. (2020). Gene regulation by long non-coding RNAs and its biological functions. Nature

Reviews. Molecular Cell Biology,22(2), 96–118.

Su, Z.-D., Huang, Y., Zhang, Z.-Y., Zhao, Y.-W., Wang, D., Chen, W., Chou, K.-C., & Lin, H. (2018). iLoc-lncRNA: Predict the subcellular

location of lncRNAs by incorporating octamer composition into general PseKNC. Bioinformatics,34(24), 4196–4204.

Sun, L., Fazal, F. M., Li, P., Broughton, J. P., Lee, B., Tang, L., Huang, W., Kool, E. T., Chang, H. Y., & Zhang, Q. C. (2019). RNA structure

maps across mammalian cellular compartments. Nature Structural & Molecular Biology,26(4), 322–330.

Sun, Y. H., Wang, A., Song, C., Shankar, G., Srivastava, R. K., Au, K. F., & Li, X. Z. (2021). Single-molecule long-read sequencing reveals a

conserved intact long RNA profile in sperm. Nature Communications,12(1), 1361.

Svoboda, P., & Di Cara, A. (2006). Hairpin RNA: A secondary structure of primary importance. Cellular and Molecular Life Sciences: CMLS,

63(7–8), 901–908.

Sztuba-Solinska, J., Rausch, J. W., Smith, R., Miller, J. T., Whitby, D., & Le Grice, S. F. J. (2017). Kaposi's sarcoma-associated herpesvirus

polyadenylated nuclear RNA: A structural scaffold for nuclear, cytoplasmic and viral proteins. Nucleic Acids Research,45(11), 6805–6821.

Tahi, F., Du T Tran, V., & Boucheham, A. (2017). In silico prediction of RNA secondary structure. Methods in Molecular Biology,1543,

145–168.

Tan, D., Marzluff, W. F., Dominski, Z., & Tong, L. (2013). Structure of histone mRNA stem-loop, human stem-loop binding protein, and

3'hExo ternary complex. Science,339(6117), 318–321.

Tan, J. Y., Biasini, A., Young, R. S., & Marques, A. C. (2020). Splicing of enhancer-associated lincRNAs contributes to enhancer activity. Life

science Alliance,3(4), e202000663. https://doi.org/10.26508/lsa.202000663

Tavares, R. C. A., Pyle, A. M., & Somarowthu, S. (2019). Phylogenetic analysis with improved parameters reveals conservation in lncRNA

structures. Journal of Molecular Biology,431(8), 1592–1603.

ROSS AND ULITSKY 23 of 25

Tichon, A., Gil, N., Lubelsky, Y., Havkin Solomon, T., Lemze, D., Itzkovitz, S., Stern-Ginossar, N., & Ulitsky, I. (2016). A conserved abundant

cytoplasmic long noncoding RNA modulates repression by Pumilio proteins in human cells. Nature Communications,7, 12209.

Tichon, A., Perry, R. B.-T., Stojic, L., & Ulitsky, I. (2018). SAM68 is required for regulation of Pumilio by the NORAD long noncoding RNA.

Genes & Development,32(1), 70–78.

Tomaszewska, M., Szabat, M., Zieli

nska, K., & Kierzek, R. (2021). Identification and structural aspects of G-Quadruplex-forming sequences

from the influenza a virus genome. International Journal of Molecular Sciences,22(11), 6031. https://doi.org/10.3390/ijms22116031

Tycowski, K. T., Shu, M.-D., Borah, S., Shi, M., & Steitz, J. A. (2012). Conservation of a triple-helix-forming RNA stability element in noncod-

ing and genomic RNAs of diverse viruses. Cell Reports,2(1), 26–32.

Ule, J., Jensen, K. B., Ruggiu, M., Mele, A., Ule, A., & Darnell, R. B. (2003). CLIP identifies Nova-regulated RNA networks in the brain. Sci-

ence,302(5648), 1212–1215.

Ulitsky, I. (2016). Evolution to the rescue: Using comparative genomics to understand long non-coding RNAs. Nature Reviews. Genetics,

17(10), 601–614.

Ulitsky, I., & Bartel, D. P. (2013). lincRNAs: Genomics, evolution, and mechanisms. Cell,154(1), 26–46.

Ulitsky, I., Shkumatava, A., Jan, C. H., Sive, H., & Bartel, D. P. (2011). Conserved function of lincRNAs in vertebrate embryonic development

despite rapid sequence evolution. Cell,147(7), 1537–1550.

Unfried, J. P., Marín-Baquero, M., Rivera-Calzada, 

A., Razquin, N., Martín-Cuevas, E. M., de Bragança, S., Aicart-Ramos, C., McCoy, C.,

Prats-Mari, L., Arribas-Bosacoma, R., Lee, L., Caruso, S., Zucman-Rossi, J., Sangro, B., Williams, G., Moreno-Herrero, F., Llorca, O.,

Lees-Miller, S. P., & Fortes, P. (2021). Long noncoding RNA NIHCOLE promotes ligation efficiency of DNA double-strand breaks in

hepatocellular carcinoma. Cancer Research,81, 4910–4925. https://doi.org/10.1158/0008-5472.CAN-21-0463

Uroda, T., Anastasakou, E., Rossi, A., Teulon, J.-M., Pellequer, J.-L., Annibale, P., Pessey, O., Inga, A., Chill

on, I., & Marcia, M. (2019). Con-

served Pseudoknots in lncRNA MEG3 are essential for stimulation of the p53 pathway. Molecular Cell,75(5), 982–995.e9.

Valsecchi, C. I. K., Basilicata, M. F., Georgiev, P., Gaub, A., Seyfferth, J., Kulkarni, T., Panhale, A., Semplicio, G., Manjunath, V., Holz, H.,

Dasmeh, P., & Akhtar, A. (2021). RNA nucleation by MSL2 induces selective X chromosome compartmentalization. Nature,589(7840),

137–142.

Van Nostrand, E. L., Freese, P., Pratt, G. A., Wang, X., Wei, X., Xiao, R., Blue, S. M., Chen, J.-Y., Cody, N. A. L., Dominguez, D., Olson, S.,

Sundararaman, B., Zhan, L., Bazile, C., Bouvrette, L. P. B., Bergalet, J., Duff, M. O., Garcia, K. E., Gelboin-Burkhart, C., …Yeo, G. W.

(2020). A large-scale binding and functional map of human RNA-binding proteins. Nature,583(7818), 711–719.

Visel, A., Minovitsky, S., Dubchak, I., & Pennacchio, L. A. (2007). VISTA Enhancer Browser—A database of tissue-specific human

enhancers. Nucleic Acids Research,35(Database issue, D88–D92.

Wang, Y., Yesselman, J. D., Zhang, Q., Kang, M., & Feigon, J. (2016). Structural conservation in the template/pseudoknot domain of verte-

brate telomerase RNA from teleost fish to human. Proceedings of the National Academy of Sciences of the United States of America,

113(35), E5125–E5134.

Washietl, S., Kellis, M., & Garber, M. (2014). Evolutionary dynamics and tissue specificity of human long noncoding RNAs in six mammals.

Genome Research,24(4), 616–628.

Watters, K. E., Strobel, E. J., Yu, A. M., Lis, J. T., & Lucks, J. B. (2016). Cotranscriptional folding of a riboswitch at nucleotide resolution.

Nature Structural & Molecular Biology,23(12), 1124–1131.

Woese, C. R., Magrum, L. J., Gupta, R., Siegel, R. B., Stahl, D. A., Kop, J., Crawford, N., Brosius, J., Gutell, R., Hogan, J. J., & Noller, H. F.

(1980). Secondary structure model for bacterial 16S ribosomal RNA: Phylogenetic, enzymatic and chemical evidence. Nucleic Acids

Research,8(10), 2275–2293.

Wutz, A., Rasmussen, T. P., & Jaenisch, R. (2002). Chromosomal silencing and localization are mediated by different domains of Xist RNA.

Nature Genetics,30(2), 167–174.

Xue, Z., Hennelly, S., Doyle, B., Gulati, A. A., Novikova, I. V., Sanbonmatsu, K. Y., & Boyer, L. A. (2016). A G-rich motif in the lncRNA

Braveheart interacts with a zinc-finger transcription factor to specify the cardiovascular lineage. Molecular Cell,64(1), 37–50.

Yang, J.-H., Li, J.-H., Shao, P., Zhou, H., Chen, Y.-Q., & Qu, L.-H. (2010). starBase: A database for exploring microRNA–mRNA interaction

maps from Argonaute CLIP-Seq and Degradome-Seq data. Nucleic Acids Research,39(suppl_1), D202–D209.

Yang, Y.-C. T., Di, C., Hu, B., Zhou, M., Liu, Y., Song, N., Li, Y., Umetsu, J., & Lu, Z. J. (2015). CLIPdb: A CLIP-seq database for protein-

RNA interactions. BMC Genomics,16, 51.

Yao, Z., Weinberg, Z., & Ruzzo, W. L. (2005). CMfinder—A covariance model based RNA motif finding algorithm. Bioinformatics,22(4),

445–452.

Yin, Y., Lu, J. Y., Zhang, X., Shao, W., Xu, Y., Li, P., Hong, Y., Cui, L., Shan, G., Tian, B., Zhang, Q. C., & Shen, X. (2020). U1 snRNP regu-

lates chromatin retention of noncoding RNAs. Nature,580(7801), 147–150.

Yu, A. M., Gasper, P. M., Cheng, L., Lai, L. B., Kaur, S., Gopalan, V., Chen, A. A., & Lucks, J. B. (2021). Computationally reconstructing

cotranscriptional RNA folding from experimental data reveals rearrangement of non-native folding intermediates. Molecular Cell,81(4),

870–883.e10.

Yu, B., Lu, Y., Zhang, Q. C., & Hou, L. (2020). Prediction and differential analysis of RNA secondary structure. Quantitative Biology (Beijing,

China),8(2), 109–118.

Yu, H., Lindsay, J., Feng, Z.-P., Frankenberg, S., Hu, Y., Carone, D., Shaw, G., Pask, A. J., O'Neill, R., Papenfuss, A. T., & Renfree, M. B.

(2012). Evolution of coding and non-coding genes in HOX clusters of a marsupial. BMC Genomics,13, 251.

24 of 25 ROSS AND ULITSKY

Zampetaki, A., Albrecht, A., & Steinhofel, K. (2018). Long non-coding RNA structure and function: Is there a link? Frontiers in Physiology,9,

1201. https://doi.org/10.3389/fphys.2018.01201

Zhang, B., Mao, Y. S., Diermeier, S. D., Novikova, I. V., Nawrocki, E. P., Jones, T. A., Lazar, Z., Tung, C.-S., Luo, W., Eddy, S. R.,

Sanbonmatsu, K. Y., & Spector, D. L. (2017). Identification and characterization of a class of MALAT1-like genomic loci. Cell Reports,

19(8), 1723–1738.

Zhang, X., Rice, K., Wang, Y., Chen, W., Zhong, Y., Nakayama, Y., Zhou, Y., & Klibanski, A. (2010). Maternally expressed gene 3 (MEG3)

noncoding ribonucleic acid: Isoform structure, expression, and functions. Endocrinology,151(3), 939–947.

Zhang, X.-Q., Wang, Z.-L., Poon, M.-W., & Yang, J.-H. (2017). Spatial-temporal transcriptional dynamics of long non-coding RNAs in human

brain. Human Molecular Genetics,26(16), 3202–3211.

Ziv, O., Gabryelska, M. M., Lun, A. T. L., Gebert, L. F. R., Sheu-Gruttadauria, J., Meredith, L. W., Liu, Z.-Y., Kwok, C. K., Qin, C.-F.,

MacRae, I. J., Goodfellow, I., Marioni, J. C., Kudla, G., & Miska, E. A. (2018). COMRADES determines in vivo RNA structures and inter-

actions. Nature Methods,15(10), 785–788.

How to cite this article: Ross, C. J., & Ulitsky, I. (2022). Discovering functional motifs in long noncoding

RNAs. Wiley Interdisciplinary Reviews: RNA,13(4), e1708. https://doi.org/10.1002/wrna.1708

ROSS AND ULITSKY 25 of 25

Exosomes to exosome-functionalized scaffolds: a novel approach to stimulate bone regeneration

Article

Full-text available

Nov 2024

Bone regeneration is a complex biological process that relies on the orchestrated interplay of various cellular and molecular events. Bone tissue engineering is currently the most promising method for treating bone regeneration. However, the immunogenicity, stable and cell quantity of seed cells limited their application. Recently, exosomes, which are small extracellular vesicles released by cells, have been found to effectively address these problems and better induce bone regeneration. Meanwhile, a growing line of research has shown the cargos of exosomes may provide effective therapeutic and biomarker tools for bone repair, including miRNA, lncRNA, and proteins. Moreover, engineered scaffolds loaded with exosomes can offer a cell-free bone repair strategy, addressing immunogenicity concerns and providing a more stable functional performance. Herein, we provide a comprehensive summary of the role played by scaffolds loaded with exosomes in bone regeneration, drawing on a systematic analysis of relevant literature available on PubMed, Scopus, and Google Scholar database.

Conserved RNA-binding protein interactions mediate syntologous lncRNA functions

Preprint

Full-text available

Aug 2024

Syntologous long noncoding RNAs (lncRNAs) are loci with conserved genomic positions that often show little or no sequence similarity. Despite diverging primary sequences, lncRNA syntologs from distant species can carry out similar functions. However, determinants underlying conserved functions of syntologous lncRNA transcripts with no sequence similarity remain unknown. Using CASC15 and melanoma formation as a paradigm for fast evolving lncRNAs and their functions, we found that human and zebrafish CASC15 syntologs with no detectable sequence similarity retained their function across 450 million years of evolution. Similar to the casc15 -deficient zebrafish, CASC15- mutant human melanoma cells show increased cell migration. Expression of human CASC15 in zebrafish rescues loss of casc15 function by attenuating melanoma formation. This conserved function is supported by a set of RNA-binding proteins, interacting with both zebrafish and human CASC15 transcripts. Together, our findings demonstrate that conserved RNA-protein interactions can define functions of rapidly evolving lncRNA transcripts.

Deep dive into RNA: a systematic literature review on RNA structure prediction using machine learning methods

Article

Full-text available

Aug 2024
ARTIF INTELL REV

The discovery of non-coding RNAs (ncRNAs) has expanded our comprehension of RNAs’ inherent nature and capabilities. The intricate three-dimensional structures assumed by RNAs dictate their specific functions and molecular interactions. However, the limited number of mapped structures, partly due to experimental constraints of methods such as nuclear magnetic resonance (NMR), highlights the importance of in silico prediction solutions. This is particularly crucial in potential applications in therapeutic drug discovery. In this context, machine learning (ML) methods have emerged as prominent candidates, having previously demonstrated prowess in solving complex challenges across various domains. This review focuses on analyzing the development of ML-based solutions for RNA structure prediction, specifically oriented toward recent advancements in the deep learning (DL) domain. A systematic analysis of 33 works reveals insights into the representation of RNA structures, secondary structure motifs, and tertiary interactions. The review highlights current trends in ML methods used for RNA structure prediction, demonstrates the growing research involvement in this field, and summarizes the most valuable findings.

sincFold: end-to-end learning of short- and long-range interactions in RNA secondary structure

Article

Full-text available

Jun 2024

Motivation Coding and noncoding RNA molecules participate in many important biological processes. Noncoding RNAs fold into well-defined secondary structures to exert their functions. However, the computational prediction of the secondary structure from a raw RNA sequence is a long-standing unsolved problem, which after decades of almost unchanged performance has now re-emerged due to deep learning. Traditional RNA secondary structure prediction algorithms have been mostly based on thermodynamic models and dynamic programming for free energy minimization. More recently deep learning methods have shown competitive performance compared with the classical ones, but there is still a wide margin for improvement. Results In this work we present sincFold, an end-to-end deep learning approach, that predicts the nucleotides contact matrix using only the RNA sequence as input. The model is based on 1D and 2D residual neural networks that can learn short- and long-range interaction patterns. We show that structures can be accurately predicted with minimal physical assumptions. Extensive experiments were conducted on several benchmark datasets, considering sequence homology and cross-family validation. sincFold was compared with classical methods and recent deep learning models, showing that it can outperform the state-of-the-art methods.

Long Noncoding RNAs in Response to Hyperosmolarity Stress, but Not Salt Stress, Were Mainly Enriched in the Rice Roots

Article

Full-text available

Jun 2024
INT J MOL SCI

Due to their immobility and possession of underground parts, plants have evolved various mechanisms to endure and adapt to abiotic stresses such as extreme temperatures, drought, and salinity. However, the contribution of long noncoding RNAs (lncRNAs) to different abiotic stresses and distinct rice seedling parts remains largely uncharacterized beyond the protein-coding gene (PCG) layer. Using transcriptomics and bioinformatics methods, we systematically identified lncRNAs and characterized their expression patterns in the roots and shoots of wild type (WT) and ososca1.1 (reduced hyperosmolality-induced [Ca2+]i increase in rice) seedlings under hyperosmolarity and salt stresses. Here, 2937 candidate lncRNAs were identified in rice seedlings, with intergenic lncRNAs representing the largest category. Although the detectable sequence conservation of lncRNAs was low, we observed that lncRNAs had more orthologs within the Oryza. By comparing WT and ososca1.1, the transcription level of OsOSCA1.1-related lncRNAs in roots was greatly enhanced in the face of hyperosmolality stress. Regarding regulation mode, the co-expression network revealed connections between trans-regulated lncRNAs and their target PCGs related to OsOSCA1.1 and its mediation of hyperosmolality stress sensing. Interestingly, compared to PCGs, the expression of lncRNAs in roots was more sensitive to hyperosmolarity stress than to salt stress. Furthermore, OsOSCA1.1-related hyperosmolarity stress-responsive lncRNAs were enriched in roots, and their potential cis-regulated genes were associated with transcriptional regulation and signaling transduction. Not to be ignored, we identified a motif-conserved and hyperosmolarity stress-activated lncRNA gene (OSlncRNA), speculating on its origin and evolutionary history in Oryza. In summary, we provide a global perspective and a lncRNA resource to understand hyperosmolality stress sensing in rice roots, which helps to decode the complex molecular networks involved in plant sensing and adaptation to stressful environments.

Transcriptomic Profiles of Long Noncoding RNAs and Their Target Protein-Coding Genes Reveals Speciation Adaptation on the Qinghai-Xizang (Tibet) Plateau in Orinus

Article

Full-text available

May 2024

Long noncoding RNAs (lncRNAs) are RNA molecules longer than 200 nt, which lack the ability to encode proteins and are involved in multifarious growth, development, and regulatory processes in plants and mammals. However, the environmental-regulated expression profiles of lncRNAs in Orinus that may associated with their adaptation on the Qinghai-Xizang (Tibet) Plateau (QTP) have never been characterized. Here, we utilized transcriptomic sequencing data of two Orinus species (O. thoroldii and O. kokonoricus) to identify 1624 lncRNAs, including 1119 intergenic lncRNAs, 200 antisense lncRNAs, five intronic lncRNAs, and 300 sense lncRNAs. In addition, the evolutionary relationships of Orinus lncRNAs showed limited sequence conservation among 39 species, which implied that Orinus-specific lncRNAs contribute to speciation adaptation evolution. Furthermore, considering the cis-regulation mechanism, from 286 differentially expressed lncRNAs (DElncRNAs) and their nearby protein coding genes (PCGs) between O. thoroldii and O. kokonoricus, 128 lncRNA-PCG pairs were obtained in O. thoroldii, whereas 92 lncRNA-PCG pairs were obtained in O. kokonoricus. In addition, a total of 19 lncRNA-PCG pairs in O. thoroldii and 14 lncRNA-PCG pairs in O. kokonoricus were found to participate in different biological processes, indicating that the different expression profiles of DElncRNAs between O. thoroldii and O. kokonoricus were associated with their adaptation at different elevations on the QTP. We also found several pairs of DElncRNA nearby transcription factors (TFs), indicating that these DElncRNAs regulate the expression of TFs to aid O. thoroldii in adapting to the environment. Therefore, this work systematically identified a series of lncRNAs in Orinus, laying the groundwork for further exploration into the biological function of Orinus in environmental adaptation.

Underground Communication: Long Noncoding RNA Signaling in the Plant Rhizosphere

Article

Full-text available

Apr 2024

Long non-coding RNAs (lncRNAs) have emerged as integral gene expression regulators underlying plant growth, development, and adaptation. To adapt to the heterogeneous and dynamic rhizosphere, plants use interconnected regulatory mechanisms to optimally fine-tune gene expression governing interactions with soil biota, nutrient acquisition, and heavy metal tolerance. Recently, high-throughput sequencing has enabled the identification of plant lncRNAs responsive to rhizosphere biotic and abiotic cues. Here, we examine lncRNA biogenesis, classification, and mode of action, highlighting the functions of lncRNAs in mediating plant adaptation to diverse rhizosphere factors. We then discuss studies that revealed lncRNA significance and target genes during developmental plasticity and stress responses at the rhizobium interface. Thus, a comprehensive understanding of specific lncRNAs, their regulatory targets, and the intricacies of their functional interaction networks will provide crucial insights into how these transcriptomic switches fine-tune responses to shifting rhizosphere signals. As we look ahead, we foresee that single-cell dissection of cell type-specific lncRNA regulatory dynamics will enhance our understanding of precise developmental modulation mechanisms enabling plant rhizosphere adaptation. Overcoming future challenges through multi-omics and genetic approaches will better reveal the integral lncRNA roles governing plant adaptation to the belowground environment.

Integrative analysis of the lncRNA-miRNA-mRNA interactions in smooth muscle cell phenotypic transitions

Article

Full-text available

Apr 2024

Objectives: We previously found that the pluripotency factor OCT4 is reactivated in smooth muscle cells (SMC) in human and mouse atherosclerotic plaques and plays an atheroprotective role. Loss of OCT4 in SMC in vitro was associated with decreases in SMC migration. However, molecular mechanisms responsible for atheroprotective SMC-OCT4-dependent effects remain unknown. Methods: Since studies in embryonic stem cells demonstrated that OCT4 regulates long non-coding RNAs (lncRNAs) and microRNAs (miRNAs), making them candidates for OCT4 effect mediators, we applied an in vitro approach to investigate the interactions between OCT4-regulated lncRNAs, mRNAs, and miRNAs in SMC. We used OCT4 deficient mouse aortic SMC (MASMC) treated with the pro-atherogenic oxidized phospholipid POVPC, which, as we previously demonstrated, suppresses SMC contractile markers and induces SMC migration. Differential expression of lncRNAs, mRNAs, and miRNAs was obtained by lncRNA/mRNA expression array and small-RNA microarray. Long non-coding RNA to mRNA associations were predicted based on their genomic proximity and association with vascular diseases. Given a recently discovered crosstalk between miRNA and lncRNA, we also investigated the association of miRNAs with upregulated/downregulated lncRNA-mRNA pairs. Results: POVPC treatment in SMC resulted in upregulating genes related to the axon guidance and focal adhesion pathways. Knockdown of Oct4 resulted in differential regulation of pathways associated with phagocytosis. Importantly, these results were consistent with our data showing that OCT4 deficiency attenuated POVPC-induced SMC migration and led to increased phagocytosis. Next, we identified several up- or downregulated lncRNA associated with upregulation of the specific mRNA unique for the OCT4 deficient SMC, including upregulation of ENSMUST00000140952-Hoxb5/6 and ENSMUST00000155531-Zfp652 along with downregulation of ENSMUST00000173605-Parp9 and, ENSMUST00000137236-Zmym1. Finally, we found that many of the downregulated miRNAs were associated with cell migration, including miR-196a-1 and miR-10a, targets of upregulated ENSMUST00000140952, and miR-155 and miR-122, targets of upregulated ENSMUST00000155531. Oppositely, the upregulated miRNAs were anti-migratory and pro-phagocytic, such as miR-10a/b and miR-15a/b, targets of downregulated ENSMUST00000173605, and miR-146a/b and miR-15b targets of ENSMUST00000137236. Conclusion: Our integrative analyses of the lncRNA-miRNA-mRNA interactions in SMC indicated novel potential OCT4-dependent mechanisms that may play a role in SMC phenotypic transitions.

sincFold: end-to-end learning of short- and long-range interactions for RNA folding

Preprint

Full-text available

Oct 2023

Motivation Coding and non-coding RNA molecules participate in many important biological processes. Non-coding RNAs fold into well-defined secondary structures to exert their functions. However, the computational prediction of the secondary structure from a raw RNA sequence is a long-standing unsolved problem, which has stagnated in the last decades. Traditional RNA secondary structure prediction algorithms have been mostly based on thermodynamic models and dynamic programming for free energy minimization. More recently deep learning methods have shown competitive performance compared with the classical ones, but still leaving a wide margin for improvement. Results In this work we present sincFold, an end-to-end deep learning approach that predicts the nucleotides contact matrix using only the RNA sequence as input. The model is based on hierarchical 1D-2D residual neural networks that can learn short- and long-range interaction patterns. We show that structures can be accurately predicted with minimal physical assumptions. Extensive experiments on well known benchmark datasets were conducted, comparing sincFold against classical methods and recent deep learning models. Results show that sincFold can outperform state-of-the-art methods on all datasets assessed. Availability The source code is available at https://github.com/sinc-lab/sincFold/ (v0.13) and a webdemo is provided at https://sinc.unl.edu.ar/web-demo/sincFold Contact lbugnon@sinc.unl.edu.ar

Exploration of the Noncoding Genome for Human-Specific Therapeutic Targets—Recent Insights at Molecular and Cellular Level

Article

Full-text available

Nov 2023

While it is well known that 98–99% of the human genome does not encode proteins, but are nevertheless transcriptionally active and give rise to a broad spectrum of noncoding RNAs [ncRNAs] with complex regulatory and structural functions, specific functions have so far been assigned to only a tiny fraction of all known transcripts. On the other hand, the striking observation of an overwhelmingly growing fraction of ncRNAs, in contrast to an only modest increase in the number of protein-coding genes, during evolution from simple organisms to humans, strongly suggests critical but so far essentially unexplored roles of the noncoding genome for human health and disease pathogenesis. Research into the vast realm of the noncoding genome during the past decades thus lead to a profoundly enhanced appreciation of the multi-level complexity of the human genome. Here, we address a few of the many huge remaining knowledge gaps and consider some newly emerging questions and concepts of research. We attempt to provide an up-to-date assessment of recent insights obtained by molecular and cell biological methods, and by the application of systems biology approaches. Specifically, we discuss current data regarding two topics of high current interest: (1) By which mechanisms could evolutionary recent ncRNAs with critical regulatory functions in a broad spectrum of cell types (neural, immune, cardiovascular) constitute novel therapeutic targets in human diseases? (2) Since noncoding genome evolution is causally linked to brain evolution, and given the profound interactions between brain and immune system, could human-specific brain-expressed ncRNAs play a direct or indirect (immune-mediated) role in human diseases? Synergistic with remarkable recent progress regarding delivery, efficacy, and safety of nucleic acid-based therapies, the ongoing large-scale exploration of the noncoding genome for human-specific therapeutic targets is encouraging to proceed with the development and clinical evaluation of novel therapeutic pathways suggested by these research fields.

Identification and Structural Aspects of G-Quadruplex-Forming Sequences from the Influenza A Virus Genome

Article

Full-text available

Jun 2021
INT J MOL SCI

Influenza A virus (IAV) causes seasonal epidemics and sporadic pandemics, therefore is an important research subject for scientists around the world. Despite the high variability of its genome, the structure of viral RNA (vRNA) possesses features that remain constant between strains and are biologically important for virus replication. Therefore, conserved structural motifs of vRNA can represent a novel therapeutic target. Here, we focused on the presence of G-rich sequences within the influenza A/California/07/2009(H1N1) genome and their ability to form RNA G-quadruplex structures (G4s). We identified 12 potential quadruplex-forming sequences (PQS) and determined their conservation among the IAV strains using bioinformatics tools. Then we examined the propensity of PQS to fold into G4s by various biophysical methods. Our results revealed that six PQS oligomers could form RNA G-quadruplexes. However, three of them were confirmed to adopt G4 structures by all utilized methods. Moreover, we showed that these PQS motifs are present within segments encoding polymerase complex proteins indicating their possible role in the virus biology.

A functional motif of long noncoding RNA Nron against osteoporosis

Article

Full-text available

Jun 2021

Long noncoding RNAs are widely implicated in diverse disease processes. Nonetheless, their regulatory roles in bone resorption are undefined. Here, we identify lncRNA Nron as a critical suppressor of bone resorption. We demonstrate that osteoclastic Nron knockout mice exhibit an osteopenia phenotype with elevated bone resorption activity. Conversely, osteoclastic Nron transgenic mice exhibit lower bone resorption and higher bone mass. Furthermore, the pharmacological overexpression of Nron inhibits bone resorption, while caused apparent side effects in mice. To minimize the side effects, we further identify a functional motif of Nron. The delivery of Nron functional motif to osteoclasts effectively reverses bone loss without obvious side effects. Mechanistically, the functional motif of Nron interacts with E3 ubiquitin ligase CUL4B to regulate ERα stability. These results indicate that Nron is a key bone resorption suppressor, and the lncRNA functional motif could potentially be utilized to treat diseases with less risk of side effects.

Activity-regulated synaptic targeting of lncRNA ADEPTR mediates structural plasticity by localizing Sptn1 and AnkB in dendrites

Article

Full-text available

Apr 2021

Activity-dependent structural plasticity at the synapse requires specific changes in the neuronal transcriptome. While much is known about the role of coding elements in this process, the role of the long noncoding transcriptome remains elusive. Here, we report the discovery of an intronic long noncoding RNA (lncRNA)—termed ADEPTR—that is up-regulated and synaptically transported in a cAMP/PKA-dependent manner in hippocampal neurons, independently of its protein-coding host gene. Loss of ADEPTR function suppresses activity-dependent changes in synaptic transmission and structural plasticity of dendritic spines. Mechanistically, dendritic localization of ADEPTR is mediated by molecular motor protein Kif2A. ADEPTR physically binds to actin-scaffolding regulators ankyrin (AnkB) and spectrin (Sptn1) via a conserved sequence and is required for their dendritic localization. Together, this study demonstrates how activity-dependent synaptic targeting of an lncRNA mediates structural plasticity at the synapse.

Evolutionary conservation of RNA sequence and structure

Article

Full-text available

Mar 2021

Elena Rivas

An RNA structure prediction from a single‐sequence RNA folding program is not evidence for an RNA whose structure is important for function. Random sequences have plausible and complex predicted structures not easily distinguishable from those of structural RNAs. How to tell when an RNA has a conserved structure is a question that requires looking at the evolutionary signature left by the conserved RNA. This question is important not just for long noncoding RNAs which usually lack an identified function, but also for RNA binding protein motifs which can be single stranded RNAs or structures. Here we review recent advances using sequence and structural analysis to determine when RNA structure is conserved or not. Although covariation measures assess structural RNA conservation, one must distinguish covariation due to RNA structure from covariation due to independent phylogenetic substitutions. We review a statistical test to measure false positives expected under the null hypothesis of phylogenetic covariation alone (specificity). We also review a complementary test that measures power, that is, expected covariation derived from sequence variation alone (sensitivity). Power in the absence of covariation signals the absence of a conserved RNA structure. We analyze artifacts that falsely identify conserved RNA structure such as the misuse of programs that do not assess significance, the use of inappropriate statistics confounded by signals other than covariation, or misalignments that induce spurious covariation. Among artifacts that obscure the signal of a conserved RNA structure, we discuss the inclusion of pseudogenes in alignments which increase power but destroy covariation. This article is categorized under: RNA Structure and Dynamics > RNA Structure, Dynamics and Chemistry RNA Evolution and Genomics > Computational Analyses of RNA RNA Evolution and Genomics > RNA and Ribonucleoprotein Evolution

Approaches to Identify and Characterise the Post-Transcriptional Roles of lncRNAs in Cancer

Article

Full-text available

Mar 2021

It is becoming increasingly evident that the non-coding genome and transcriptome exert great influence over their coding counterparts through complex molecular interactions. Among non-coding RNAs (ncRNA), long non-coding RNAs (lncRNAs) in particular present increased potential to participate in dysregulation of post-transcriptional processes through both RNA and protein interactions. Since such processes can play key roles in contributing to cancer progression, it is desirable to continue expanding the search for lncRNAs impacting cancer through post-transcriptional mechanisms. The sheer diversity of mechanisms requires diverse resources and methods that have been developed and refined over the past decade. We provide an overview of computational resources as well as proven low-to-high throughput techniques to enable identification and characterisation of lncRNAs in their complex interactive contexts. As more cancer research strategies evolve to explore the non-coding genome and transcriptome, we anticipate this will provide a valuable primer and perspective of how these technologies have matured and will continue to evolve to assist researchers in elucidating post-transcriptional roles of lncRNAs in cancer.

CLIP and complementary methods

Article

Full-text available

Mar 2021

RNA molecules start assembling into ribonucleoprotein (RNP) complexes during transcription. Dynamic RNP assembly, largely directed by cis-acting elements on the RNA, coordinates all processes in which the RNA is involved. To identify the sites bound by a specific RNA-binding protein on endogenous RNAs, cross-linking and immunoprecipitation (CLIP) and complementary, proximity-based methods have been developed. In this Primer, we discuss the main variants of these protein-centric methods and the strategies for their optimization and quality assessment, as well as RNA-centric methods that identify the protein partners of a specific RNA. We summarize the main challenges of computational CLIP data analysis, how to handle various sources of background and how to identify functionally relevant binding regions. We outline the various applications of CLIP and available databases for data sharing. We discuss the prospect of integrating data obtained by CLIP with complementary methods to gain a comprehensive view of RNP assembly and remodelling, unravel the spatial and temporal dynamics of RNPs in specific cell types and subcellular compartments and understand how defects in RNPs can lead to disease. Finally, we present open questions in the field and give directions for further development and applications. Ule and colleagues discuss cross-linking and immunoprecipitation (CLIP) methods for characterizing the RNA binding partners of RNA-binding proteins and explore the data analysis workflows, best practices and applications for these techniques. The Primer also considers methods for characterizing the protein binding partners of specific RNAs and discusses how data from these complementary methods can be integrated into CLIP workflows.

Single-molecule long-read sequencing reveals a conserved intact long RNA profile in sperm

Article

Full-text available

Mar 2021

Sperm contributes diverse RNAs to the zygote. While sperm small RNAs have been shown to impact offspring phenotypes, our knowledge of the sperm transcriptome, especially the composition of long RNAs, has been limited by the lack of sensitive, high-throughput experimental techniques that can distinguish intact RNAs from fragmented RNAs, known to abound in sperm. Here, we integrate single-molecule long-read sequencing with short-read sequencing to detect sperm intact RNAs (spiRNAs). We identify 3440 spiRNA species in mice and 4100 in humans. The spiRNA profile consists of both mRNAs and long non-coding RNAs, is evolutionarily conserved between mice and humans, and displays an enrichment in mRNAs encoding for ribosome. In sum, we characterize the landscape of intact long RNAs in sperm, paving the way for future studies on their biogenesis and functions. Our experimental and bioinformatics approaches can be applied to other tissues and organisms to detect intact transcripts.

Long noncoding RNA NIHCOLE promotes ligation efficiency of DNA double-strand breaks in hepatocellular carcinoma

Article

Jul 2021

Long noncoding RNAs (lncRNA) are emerging as key players in cancer as parts of poorly understood molecular mechanisms. Here, we investigated lncRNAs that play a role in hepatocellular carcinoma (HCC) and identified NIHCOLE, a novel lncRNA induced in HCC with oncogenic potential and a role in the ligation efficiency of DNA double-stranded breaks (DSB). NIHCOLE expression was associated with poor prognosis and survival of HCC patients. Depletion of NIHCOLE from HCC cells led to impaired proliferation and increased apoptosis. NIHCOLE deficiency led to accumulation of DNA damage due to a specific decrease in the activity of the nonhomologous end-joining (NHEJ) pathway of DSB repair. DNA damage induction in NIHCOLE-depleted cells further decreased HCC cell growth. NIHCOLE was associated with DSB markers and recruited several molecules of the Ku70/Ku80 heterodimer. Further, NIHCOLE putative structural domains supported stable multimeric complexes formed by several NHEJ factors including Ku70/80, APLF, XRCC4, and DNA ligase IV. NHEJ reconstitution assays showed that NIHCOLE promoted the ligation efficiency of blunt-ended DSBs. Collectively, these data show that NIHCOLE serves as a scaffold and facilitator of NHEJ machinery and confers an advantage to HCC cells, which could be exploited as a targetable vulnerability. Significance This study characterizes the role of lncRNA NIHCOLE in DNA repair and cellular fitness in HCC, thus implicating it as a therapeutic target. See related commentary by Barcena-Varela and Lujambio, p. 4899

Long noncoding RNAs in cancer metastasis

Article

May 2021
NAT REV CANCER

Metastasis is a major contributor to cancer-associated deaths. It is characterized by a multistep process that occurs through the acquisition of molecular and phenotypic changes enabling cancer cells from a primary tumour to disseminate and colonize at distant organ sites. Over the past decade, the discovery and characterization of long noncoding RNAs (lncRNAs) have revealed the diversity of their regulatory roles, including key contributions throughout the metastatic cascade. Here, we review how lncRNAs promote metastasis by functioning in discrete pro-metastatic steps including the epithelial–mesenchymal transition, invasion and migration and organotrophic colonization, and by influencing the metastatic tumour microenvironment, often by interacting within ribonucleoprotein complexes or directly with other nucleic acid entities. We discuss well-characterized lncRNAs with in vivo phenotypes and highlight mechanistic commonalities such as convergence with the TGFβ–ZEB1/ZEB2 axis or the nuclear factor-κB pathway, in addition to lncRNAs with controversial mechanisms and the influence of methodologies on mechanistic interpretation. Furthermore, some lncRNAs can help identify tumours with increased metastatic risk and spur novel therapeutic strategies, with several lncRNAs having shown potential as novel targets for antisense oligonucleotide therapy in animal models. In addition to well-characterized examples of lncRNAs functioning in metastasis, we discuss controversies and ongoing challenges in lncRNA biology. Finally, we present areas for future study for this rapidly evolving field. This Review discusses how long noncoding RNAs influence metastasis by functioning in discrete pro-metastatic steps including the epithelial–mesenchymal transition, invasion and migration and organotrophic colonization, and by influencing the tumour microenvironment. Diagnostic and therapeutic potential as well as controversies and ongoing technical challenges are discussed.

High‐resolution mapping of function and protein binding in an RNA nuclear enrichment sequence

Article

May 2021

The functions of long RNAs, including mRNAs and long noncoding RNAs (lncRNAs), critically depend on their subcellular localization. The identity of the sequences that dictate subcellular localization and their high-resolution anatomy remain largely unknown. We used a suite of massively parallel RNA assays and libraries containing thousands of sequence variants to pinpoint the functional features within the SIRLOIN element, which dictates nuclear enrichment through hnRNPK recruitment. In addition, we profiled the endogenous SIRLOIN RNA-nucleoprotein complex and identified the nuclear RNA-binding proteins SLTM and SNRNP70 as novel SIRLOIN binders. Taken together, using massively parallel assays, we identified the features that dictate binding of hnRNPK, SLTM, and SNRNP70 to SIRLOIN and found that these factors are jointly required for SIRLOIN activity. Our study thus provides a roadmap for high-throughput dissection of functional sequence elements in long RNAs.

Discovering functional motifs in long noncoding RNAs

Abstract

Recommended publications

lncRNAs in development and differentiation: from sequence motifs to functional characterization

Conserved RNA-binding protein interactions mediate syntologous lncRNA functions

Systematic analysis of the target recognition and repression by the Pumilio proteins

Genetic Tools to Dissect Functions of Long Noncoding RNAs