Encoding activities of non-coding RNAs

Abstract

The universal expression of various non-coding RNAs (ncRNAs) is now considered the main feature of organisms' genomes. Many regions in the genome are transcribed but not annotated to encode proteins, yet contain small open reading frames (smORFs). A widely accepted opinion is that a vast majority of ncRNAs are not further translated. However, increasing evidence underlines a series of intriguing translational events from the ncRNAs, which were previously considered to lack coding potential. Recent studies also suggest that products derived from such novel translational events display important regulatory functions in many fundamental biological and pathological processes. Here we give a critical review on the potential coding capacity of ncRNAs, in particular, about what is known and unknown in this emerging area. We also discuss the possible underlying coding mechanisms of these extraordinary ncRNAs and possible roles of peptides or proteins derived from the ncRNAs in disease development and theranostics. Our review offers an extensive resource for studying the biology of ncRNAs and sheds light into the use of ncRNAs and their corresponding peptides or proteins for disease diagnosis and therapy.

Introduction

The genome is a cryptic store of genetic information, which encodes the blueprint of life. The product, such as peptide or protein molecules, is a key player in pathological and physiological processes ¹. The central dogma is a widely accepted classical rule in which DNA and protein are correlated by messenger RNA (mRNA) through transcription and translation ². However, only ~1.5% of human genome transcripts can be further translated to generate about 300,000 protein molecules, which govern the unlimited activities in human development ³. Most of the remaining transcripts were once considered useless ^{4, 5}. A large, but as of yet undetermined, number of non-coding RNAs (ncRNAs) have recently been uncovered. We, along with other groups, have demonstrated that these ncRNAs can regulate normal development and disease occurrence through controlling RNA maturation and protein synthesis ^6-12. Notably, the number of ncRNAs is more than that of protein-coding genes within the genome ^{13, 14}.

More shockingly, it has recently come to light that ncRNAs might have encoding capacity (Figure 1). Many ncRNAs do not overlap with canonical genes and some also contain small open reading frames (smORFs, containing <100 codons), which were previously thought unable to be translated into peptides or proteins ^15-18. The studies of mammalian genomes reveal that hundreds of non-annotated smORFs can be translated into small peptides in mammalian genomes according to Encyclopedia of DNA Elements (ENCODE) ^19-21. Besides 201 cleavage sites of signal peptides and 198 N-termini of proteins, mass spectrometry (MS) analysis showed a total of 808 newly annotated regions within the human genome. These regions are involved in the translation events of 44 novel smORFs, 140 pseudogenes, 106 novel coding regions/exons that originally belonged to the annotated genes, as well as elongation of 110 gene/protein/exons ²². More importantly, several peptides were detected, which corresponded to 9 transcripts annotated as ncRNAs ²³. Couco et al. further reported that two types of smORFs, which could bind a few ribosomes and go through translation, were mostly found in ncRNAs and 5' untranslated regions (UTRs) ²⁴.

Figure 1.

Kinds of ncRNAs-derived small peptides involved in theranostics. Some genome genes are transcribed to be ncRNAs including housekeeping ncRNAs (tRNAs, rRNAs and snRNAs) and regulatory ncRNAs. The functions of these ncRNAs are relatively clear. Regulatory ncRNAs include miRNAs, lncRNAs, circRNAs and transcripts from repeat sequences. As illustrated, by binding with ribosomes in the cytoplasm, these ncRNAs are further translated into small peptides. Circ-ZNF609 is translated and the produced small peptides function in myogenesis. These small peptides are involved in regulating muscle performance, suppressing colon cancer growth, promoting embryo cells internalization/migration, leading to FTLD/ALS and accumulation of mature miRNAs. (FTLD/ALS: frontotemporal lobar degeneration and amyotrophic lateral sclerosis; lncRNAs: long non-coding RNAs; miRNAs: microRNAs; ncRNAs: non-coding RNAs;rRNAs: ribosomal RNAs; snRNAs: small nuclear RNAs; tRNAs: transfer RNAs).

Today the major question and concern lies in whether ncRNAs will be translated into peptides or proteins and whether the translated products will be functional. In particular, whether these peptides or proteins are useful in the diagnosis and treatment of disease, such as cancer, remains to be studied. Here, we summarize the recent progresses within the scope of our comprehension of the coding potential of ncRNAs and discuss the underlying molecular mechanisms. We further provide our perspective about this field in terms of the research tools and methods as well as the role of peptides or proteins derived from ncRNAs in the diagnosis and treatment of disease.

Coding potential of ncRNAs

In terms of length and structure, ncRNAs include two basic types: housekeeping and regulatory ncRNAs ^{25, 26} (Table 1). Generally, housekeeping ncRNAs are composed of small nuclear RNAs (snRNAs), transfer RNAs (tRNAs) and ribosomal RNAs (rRNAs). The functions of these ncRNAs are relatively clear. The regulatory ncRNAs include long non-coding RNAs (lncRNAs), microRNAs (miRNAs), circular RNAs (circRNAs) and the other transcripts far from known protein-coding regions in the genome ^{27, 28}. In general, except for mRNAs, transcripts from genomes can be considered as ncRNAs. Such transcripts, from pseudogenes and repeat sequences, were once called “junk genes” due to their non-coding or unproved coding roles. As far as the protein coding was concerned, ncRNAs were generally thought to lack any protein-coding potential. However, lately, a considerable amount of literature has been published on the study of the coding potential of ncRNAs. The epitope tags, such as FLAG, human influenza hemagglutinin (HA), and green fluorescent protein (GFP), have been used to validate the endogenous expression of the predicted smORFs in ncRNAs by detecting the fusion products through western blot (WB) ^{17, 29}. Anderson et al. inserted FLAG epitope tag into the myoregulin (MLN) locus by CRISPR/Cas9-mediated homologous recombination. Afterwards, the MLN-FLAG fusion peptide was detected by WB directly ³⁰. For the purpose of testing the in vivo translation and subcellular location of potential peptides or proteins, Magny et al. generated terminal GFP-tagged fusions within predicted smORFs ³¹. The start codon of the predicted smORFs could also be mutated with the same approaches to inhibit ncRNA translation ³². The polyclonal antibodies for predicted smORFs were further raised and then detected by MS ³³. Based upon these approaches, it has been consistently found that ncRNAs can be translated into peptides or proteins (Table 2).

Table 1.

Types of ncRNAs and their basic features.

Name	Abbreviation	Size	Location	Functions	Examples	Database	References
Ribosomal RNAs	rRNAs	6.9 kb	rDNA	Composition of ribosome	\*	NONCODE database	14
Transfer RNAs	tRNAs	<0.1 kb	tDNA	Amino acid transport	\*	NONCODE database	15
Small nucleolar RNAs	snoRNAs	60 to 300 bp	Intronic	Important function in the maturation of other ncRNAs/Association with development of some cancers	U50, S NORD	Starbase, PITA	26
Small nuclear RNAs	snRNAs	100 to 200 bp	\*	mRNAs splicing	U3, U6	miRanda	42
MicroRNAs	miRNAs	19 to 24 bp	Widespread loci	Targeting of mRNAs and many others/ Initiation of various disorders including many cancers	let-7	miRmap, miRSNP, MirBase	36
Transcribed ultraconserved regions	T-UCRs	>200 bp	Widespread loci	Regulations of miRNA and mRNA levels/Possible involvement in tumorigenesis	UCR106	ncRNA.org	112
PIWI-interacting RNAs	piRNAs	26 to 31 bp	Clusters, intragenic	Transposon repression or relationship with diseases has not been discovered	MIWI, MILI, MIWI2	piRNA BANK	23
Circular RNAs	circRNAs	~500 bp	Widespread loci	As endogenous competitive RNA/Association with many cancers and diseases	ciRs7, cANRIL	circRNAbase	25
Long non-coding RNAs	lncRNAs	>200 bp	Widespread loci	Scaffold DNA-chromatin complexes, X-chromosome inactivation/Involved in tumorigenesis and cancer metastasis	H19, XIST, HOTAIR, TSIX	lncRNA database, 3.0, LNCipedia	19

*No reference data.

Table 2.

Peptides or proteins derived from their corresponding ncRNAs.

ncRNAs	Peptides or Proteins	Length	Reference
ENOD40	2 small peptides	12 and 24 aa	45
pri	4 small peptides	11 to 32 aa	48
LOC100506013	Toddler	58 aa	52
LINC00948	MLN	46 aa	30
LOC100507537	DWORF	34 aa	54
LINC00961	SPAR	90 aa	32
pri-miR171b	miPEP171b	9 aa	58
pri-miR165a	miPEP165a	18 aa	58
circ-ZNF609	ZNF609 peptide	250 aa	70
Repeats of GGGGCC	GGGGCC-encoded DPR*	56 to 290 aa	84, 85
lncRNA HOXB-AS3	HOXB-AS3	53 aa	117

*DPR: dipeptide-repeat protein.

Long ncRNAs

Transcription of lncRNAs occurs by RNA polymerase II without long and conserved ORFs. Such transcription often occurs in the genomic regions that are far away from the already known protein-coding genes ^{34, 35}. At the post-transcriptional level, lncRNAs regulate corresponding gene expression. They are also involved in modulating epigenetics and many TF proteins ³⁶. Notably, some lncRNAs have exogenous features comparable to those of annotated coding genes: they may be capped and polyadenylated. Also, they finally accumulate in the cytoplasm to become functional ^{19, 37}.

It was found that some lncRNAs with smORFs could be translated into peptides ³⁸. For instance, Jorge Ruiz-Orera et al. found that the translational efficiency of transcripts as both coding RNAs and lncRNAs with smORFs, was significantly increased compared to 3'UTR with smORFs ³⁹. Furthermore, the sequences of lncRNAs with smORFs presented distinctly higher coding scores than the normal ORFs that were randomly obtained from any kind of ncRNAs. The score was calculated by related software such as phylogenetic analysis of codon substitution frequencies based on sequence alignment (PhyloCSF), coding potential calculator (CPC), sORF finder, coding region identification tool invoking comparative analysis (CRITICA), micro-peptide detection pipeline (micPDP) and so on ^{20, 40, 41}. More importantly, a number of studies about developmental biology have pointed out that a high fraction of lncRNAs are bound with one or more ribosome ⁴², indicating that lncRNAs can be translated into peptides or proteins ⁴³. Bioinformatics analysis indicates that, different from the quintessential protein-coding genes, lncRNA binding with ribosomes tends to be more conserved across different species and overlap with more exon regions. In addition, they are expressed in low abundance and show similar codon usage with annotated genes ⁴⁴.

Increasing studies have measured the protein-coding output from lncRNAs. For example, John et al. discovered that a conserved sequence located at the 5' terminal of the lncRNA was transcribed from gene ENOD40 having smORFs and encoded two small peptides. Both of these small peptides were specially bound to the same sucrose synthase to control sucrose use ⁴⁵. Takefumi Kondo et al. reported one lncRNA and named it as pri (polished rice) in Drosophila ⁴⁶. In fact, it contains some evolutionarily conserved smORFs that encode four similar peptides, ranging from 11- to 32-aa in length. These small peptides were found to have essential roles in epithelial morphogenesis ^{47, 48}. In zebrafish, Toddler is annotated as one lncRNA at the very beginning, namely ENSDARG00000094729. The same annotation is found both in mice (Gm10664; also known as Ende ⁴⁹), and humans (LOC100506013). Moreover, LOC100506013 is raised in two lncRNA contents ⁵⁰. Nevertheless, Pauli et al. found that the ncRNA Toddler is highly conserved among many vertebrates, including homo sapiens. It contains a 58-aa smORF, which can be further translated as a signal peptide ^{49, 51}. They proved that this small peptide, named the Toddler, was an important signaling molecule expressed during zebrafish embryogenesis. Both the inhibition and overexpression of Toddler reduced the directional movement ability of mesendodermal cells in the period of zebrafish's gastrula stage (Figure 2) ⁵².

Figure 2.

Toddler is an embryonic signal that promotes cell internalization and migration during the embryogenesis of zebrafish. Toddler is annotated as a lncRNA ENSDARG00000094729. It contains a 58-aa smORF and is also capped and polyadenylated, so it can be caught by ribosome and produce peptides named Toddler. The normal expression of Toddler can promote cell internalization and migration during the embryogenesis of zebrafish via the G-protein-coupled APJ/Apelin receptor. However, the overexpression or inhibition of Toddler will inhibit cell internalization and migration.

Studies on the whole-genome scale have indicated that lncRNAs can be translated into many functional small peptides for mammlians. For humans, lncRNA LINC00948 is a skeletal muscle-specific RNA. It was demonstrated that this lncRNA encodes a conserved 46-aa small peptide named MLN. MLN bears a function and structure identical to phospholamban (PLN) and sarcolinpin (SLN), which depress sarco endoplasmic reticulum Ca²⁺-ATPase (SERCA). SERCA is a kind of membrane pump for monitoring the relaxation of muscle by modulating the uptake of calcium ions into the sarcoplasmic reticulum (Figure 3). PLN and SLN are over-expressed in cardiac cells but almost absent in skeletal muscle cells. Unlike them, MLN is stably expressed in skeletal muscle cells ⁵³. Recently, a putative muscle-specific lncRNA situated in the membrane of SR, which encodes a 34-aa small peptide termed dwarf open reading frame (DWORF), was discovered. Like PLN, SLN and MLN, DWORF could enhance the activity of SERCA by substituting some particular inhibitors of SERCA (Figure 4) ⁵⁴. Furthermore, among those previously annotated and intergenic ncRNAs, more than half of them contain predicted longer ORFs (>100 codons), such as HOTAIR and Xist ⁵⁵. Another recent study conducted by Matsumoto et al. showed that one putative lncRNA-LINC00961 harboring three smORFs encoded one functional and conserved polypeptide, which is named “small regulatory polypeptide of amino acid response” (SPAR). By localizing to the lysosome/endosome and interacting with the lysosomal v-ATPase, the novel polypeptide SPAR negatively regulated the stimulation of mammalian target of rapamycin complex 1 (mTORC1) (Figure 5) ³². These data illustrate that many transcripts currently annotated as lncRNAs encode peptides with important biological functions ³⁰. Moreover, these findings may be useful for the diagnosis and treatment of myopathy.

Figure 3.

MLN (from LINC00948) is a skeletal muscle-specific small peptide that regulates muscle performance by modulating intracellular calcium handling. MLN shares structural and functional similarity with PLN and SLN, which inhibit SERCA, the membrane pump that controls muscle relaxation by regulating Ca²⁺ uptake into SR. (KO: knock out; MLN: myoregulin; PLN: phospholamban; ; SERCA: sarco endoplasmic reticulum Ca²⁺ -ATPase; SLN: sarcolinpin; SR: sarcoplasmic reticulum; WT: wild type). Reproduced with permission from ³⁰, copyright 2015 Elsevier.

Figure 4.

Working model for DWORF (from LOC100507537) function. DWORF localizes to the SR membrane, where it enhances SERCA activity by displacing the SERCA inhibitor PLN. (DWORF: dwarf open reading frame). Adapted with permission from ⁵⁴, copyright 2016 Springer.

Figure 5.

Small peptide SPAR derived from LINC00961 involved in working model of mTORC1 activation and signaling with SPAR. With the stimulation of amino acids (aa), Ragulator is released from v-ATPase and then interacts with Rags to facilitate mTORC1 recruitment. Rag proteins are mostly activated by Rheb, but can also be regulated through additional mechanisms involving the aa leucine and arginine. SPAR interacts with v-ATPase to promote and stabilize the interaction between the v-ATPase. (aa: amino acids; AKT: protein kinase B; GDP: guanosine diphosphate; GTP: guanosine triphosphate; SPAR: small regulatory polypeptide of amino acid response). Reproduced with the permission from ³², copyright 2016 Springer.

microRNAs (miRNAs)

miRNAs are the most investigated ncRNAs. They are hairpin-derived RNAs transcribed by RNA polymerase II. miRNAs are composed of ~20-24 nucleotides (nt) and regulate post-transcriptional silencing of genes through interacting with the 3' UTRs of mRNA ⁵⁶. Generally, miRNA coding regions in the genome can be found in both the intergenic regions and introns ⁵⁷, so they seemed to be noncoding. However, Lauressegues et al. evaluated the coding potential of pri-miRNAs and demonstrated that pri-miRNAs can produce some peptides that can regulate the accumulation of miRNAs. In this study, some pri-miRNAs in the plant kingdom were found to contain smORFs, which encode regulatory peptides termed miPEP (mi-peptides). Therein, pri-miR171b and pri-miR165a were translated into peptides miPEP171b and miPEP165a, respectively. These miPEPs promote the accumulation of their homologous miRNAs by activating transcriptional activators of their corresponding pri-miRNAs. This resulted in the expression of target genes being down-regulated. Other primary transcripts of some well-studied miRNAs of M. truncatula and A. thaliana were also encoded for functional miPEPs, indicating that miPEPs or the functional peptides may be common ⁵⁸.

It should be noted that RNA polymerase II transcribes a vast majority of pri-miRNAs, followed by capping, polyadenylatation, and accumulation in the cytoplasm. Their translation may possess the same regulatory patterns of the well-annotated protein-coding genes ⁵⁹. Based on these findings, there is no doubt that the traditional concept of miRNAs needs to be redefined in terms of coding potential.

circular RNAs (circRNAs)

circRNAs was first discovered in the 1990s, and had been ignored for a long time until recently ⁶⁰. Recently, the abnormal expression of circRNAs was found to be involved in many cell biological behaviors like apoptosis, angiogenesis, epithelial-to-mesenchymal transition, and drug resistance ⁶¹. Fu et al. identified over 500 differentially expressed circRNAs in hepatocellular carcinoma (HCC) tumors with respect to the adjacent tissues. Two circRNAs derived from SMYD4 and FAM53B were found to be related to HCC clinicopathological processing ⁶². Generally speaking, circRNAs are very stable. They behave like sponges of miRNAs, sponges or scaffolds of proteins and regulators of splicing activities ⁶³. Thus, circRNAs may be useful therapeutic targets for cancer and other diseases ⁶⁴. Moreover, many circRNAs are highly conserved, tissue-specifically expressed, and can be detected in many kinds of body fluids ⁶⁵. Accordingly, circRNAs have potential as non-invasive therapeutic biomarkers for diseases.

As one kind of ncRNAs, circRNAs were also thought unable to be translated. Most of the circRNAs are spliced from exons and are also located in the cytoplasm ⁶⁶. Under some certain conditions, many special and endogenous circRNAs with internal ribosome entry sites (IRES) and AUG sites elements have been shown to be translated in some tissues ⁶⁷. For example, Yang et al. revealed the widespread N⁶-methyladenosine (m⁶A)-driven translation of circRNA, which was further proved by MS, computational prediction and polysome profiling ⁶⁸. Meng et al. designed an integrated tool named CircPro to detect circRNAs with protein-coding potential ⁶⁹. It was shown that a series of circRNAs was actually translated in a cap-independent manner ⁶⁷. The produced small peptides or proteins have specific domains. Legnini et al. profiled the expression of circRNAs to study the differentiation of myoblasts derived from murine and human. They showed a circRNA with smORFs named circ-ZNF609. It can be translated into a small peptide to control myoblast proliferation (Figure 1) ⁷⁰. Their studies served as valuable references for the study of circRNA translation. Nevertheless, the study of circRNA translation and the possible function of produced small peptides or proteins is still in its infancy. Furthermore, the cancer relevance of this translation remains to be examined.

The transcripts of pseudogenes, repeat sequences and untranslated regions (UTRs)

Compared with their homologous functional genes, pseudogenes are referred to as the same sequences with some defunct genomic loci. Owing to the existence of disruptive mutations (e.g., some frame shifts or premature stop codons), pseudogenes have been thought to be untranslated ⁷¹. The transcripts of pseudogenes are also defined as a sub-class of ncRNAs, and participate in regulating the post-transcriptional process of their homologous genes ⁷². However, their connection with RNA polymerase II still needs to be further proved. Studies have shown that pseudogenes can perform crucial roles in the development of some diseases, in particular, cancer. Pandolfi et al. reported that specific transgenic mice were subjected to malignancy analogous to human diffuse large B cell lymphoma due to the overexpression of pseudogenes such as Braf-rs1. Their study indicated the oncogenic potential of pseudogenes ⁷³. Recently, emerging studies indicated that pseudogenes have a coding potential and can indeed become translated into some reliable peptides or proteins. NANOG is a pluripotent transcript factor and plays a crucial role in self-renewal of embryonic stem cells (ESCs). In cancer cells, however, NANOG mRNA variants include 9 retrotransposed genes, and 8 out of the 9 genes are pseudogenes. The pseudogene NANOG (NANOGP8, gi 47777342) is processed from a retrogene locus and has no structural defects ⁷⁴. The expression level of NANOGP8 and NANOG protein is observed to be high in the putative cancer stem cell (CSC) populations, and to have unknown function in the development of tumors ⁷⁵. Phosphoglycerate mutase 3 (PGAM3) is an intronless pseudogene and is situated in the Menkes disease gene's first intron. Its transcript has the same length as the mRNA of the homologous gene. PGAM3 also possesses a very short poly-A tail (16bp) at the ending of the 3'UTR. Further, Betran et al. reported that PGAM3 indeed produced a functional protein with the pressure of positive-selection ⁷⁶. Moreover, many novel peptides or proteins in the annotated pseudogenes group were searched using the protein basic local alignment search tool (BLAST) ⁷⁷. These data indicate that pseudogenes might be not only transcribed but also translated ⁷⁸.

The repeat sequences occupy about half of the human genome and have often been considered neutral, with no phenotypic consequences ⁷⁹. As the main part of non-coding regions in the genome, repeat sequences are also often transcribed as ncRNAs ⁵⁵. Recent data indicates that repeat sequences also have the potential to produce proteins for developing frontotemporal lobar degeneration and amyotrophic lateral sclerosis (FTLD/ALS) ⁸⁰. In the upstream region of C9orf72, expanding GGGGCC repeat sequences often results in FTLD/ALS ^81-83, although the mechanism of pathogenesis remains largely unknown. For these FTLD/ALS patients, researchers also found that the intracellular contents of one misfolded protein are characteristic of C9orf72-associated pathology ⁸⁴. Mori et al. found that the majority of these intracellular contents contained not only poly-(Gly-Ala) protein, but also dipeptide-repeat proteins (poly-(Gly-Arg) and poly-(Gly-Pro)) in a small portion. These proteins were assumed to be generated from the expanded GGGGCC repeat sequences by non-ATG-initiated translation. To a large extent, these discoveries directly link the predominant pathology of FTLD/ALS with the coding potential of C9orf72 hexanucleotide expansion ⁸⁵.

In addition, smORFs also can be found in the 5' UTR of mRNAs and were named as upstream ORFs (uORFs) ⁸⁶. Although they have low average conservation, uORFs had been reported in many species ⁸⁷. Same as ncRNAs, uORFs in UTR were once also considered to be non-coding. But, a 31-mer peptide translated from the uORF of gene Chop was proved to inhibit the translation of CHOP protein by blocking the peptide exit tunnel of ribosomes ⁸⁸. Starck et al. also found notable translation events from uORFs of binding immunoglobulin protein (BiP) mRNA through tracing translation during the integrated stress response (ISR) for T cells. They proposed that the peptides translated from uORFs could serve as primary histocompatibility complex class I ligands to make specified cells identified by the immune system ⁸⁹.

The underlying coding mechanisms of ncRNAs

Emerging evidence has demonstrated the coding capacity of ncRNAs. However, the underlying mechanisms that result in such coding capacities remain unclear though mRNAs can be translated into peptides or proteins in a relatively known manner. Many ncRNAs such as lncRNAs are also observed to be similar to mRNAs ⁹⁰. What's more, most ncRNAs contain at least one smORF ⁹¹. The association of lncRNAs with ribosomes indicates that ncRNAs may be a possible source of novel peptides or proteins ⁹². Ingolia et al. reported that, like the coding sequences discussed earlier, some lncRNAs in ESCs can be efficiently associated with one or more ribosome, and, to some extent, their coding capacity was much higher than that of mRNA 3'UTRs ⁹³.

The translation of some smORFs was triggered by mammalian Ste20-like kinase (MST1)-mediated eIF4E phosphorylation. Once phosphorylated, eIF4E could barely interact with the 5' cap of mRNA and then make polyribosomes associated with lncRNAs ⁹⁴. Other smORFs used non-ATG start codons ⁹⁵. Zu et al. has proved that translational initiation of some ncRNAs, which often occurs at the CAG.CTG expansion sites, is very common ⁹⁶. By delaying the binding of 40S ribosomal subunit, it is thought that hairpin structures can initiate translation at some undefined regions ⁹⁷. Furthermore, hairpin structures can recruit both the initiation factors and ribosomal subunits to areas known as IRESs. The resultant IRESs-hairpins complexes could act as tRNA_i^Met and trigger the translation initiation at non-AUG codons. These complexes are stabilized due to the existence of IRES translation-associated factors ⁹⁸. But for lncRNAs, they seldom contain highly structured regions like hairpins. They are thought to contain several translatable smORFs ranging from an AUG codon to a stop codon. Occasionally, AUG codons in such smORFs will be replaced as promoters by Kozak sequences that will also initiate the translation ^{99, 100}. Notably, two remarkable non-AUG-initiated translation mechanisms have been proposed. One mechanism is the translation initiation from an entire Kozak sequence (substitute for the AUG codon) ^101-103. Another is the triggering of non-AUG-associated translation for some repeat sequences, like CAG repeats in ataxin 8 (ATXN8), by hairpin-formation (substitute for the AUG codon) in repeat regions (Figure 6) ¹⁰⁴. For spinocerebellar ataxia type 8 (SCA8) patients, gene ATXN8 encodes a native poly-Q inclusion as a result of the existence of repeat expansion ^{96, 105, 106}.

Figure 6.

The assumed mechanism for translation initiation with special structure. (A) The classical mechanism for translation initiation. The complex composed of eIF4E, eIF4G, eIF4A, binds to the 5' cap of target RNA molecules. The poly A-binding protein (PABP) is associated with eIF4G to circularize the target mRNA molecules. Then, the eIF4F complex recruits the 43S pre-initiation complex (PIC), composed of the 40S ribosomal subunit, 30S ribosomal subunit, and the ternary complex, consisting of initiator methionine-tRNA and GTP. Next, the PIC and the components of the eIF4F complex scan through the 5'UTR in the 5' to 3' direction until encountering an AUG start codon, at which point the translation activity will be triggered by the present AUG codon. (B-C) Studies have shown that some Kozak sequences with AUG codon (B) and hairpin-structures such as (GGGGCC)n (C) can substitute the AUG codon and trigger the translation activity.

However, due to their unique structure, the translation of circRNAs was promoted by m⁶A. This is very different from linear ncRNAs. Yang et al. discovered some short sequences in most of circRNAs containing m⁶A sites ⁶⁶. Moreover, studies have reported that m⁶A in 3'UTRs or 5'UTRs could promote cap-independent translation ^{107, 108}. Thus, they reported that the translation of circRNAs was triggered by m⁶A reader YTH domain family protein 3 (YTHDF3). They also found that translation could be initiated by eukaryotic translation initiation factor 4 gamma 2 (eIF4G2) and this process was enhanced by methyltransferase-like 3/14 (METTL3/14) but depressed by demethylase fat mass and obesity-associated protein (Figure 7) ⁶⁸.

Figure 7.

Schematic diagram of circRNA translation driven by m⁶A. The circRNAs here are those with m⁶A motifs. This m⁶A driven translation requires initiation factor eIF4G2 and m⁶A reader YTHDF3, and is enhanced by methyltransferase METTL3/14, inhibited by demethylase FTO. (m⁶A: N⁶-methyladenosine; YTHDF3: YTH domain family protein 3; eIF4G2: eukaryotic translation initiation factor 4 gamma 2; METTL3/14: methyltransferase-like 3/14; FTO: fat mass and obesity-associated protein). Adapted with the permission from ⁶⁸, copyright 2017 Springer.

Using a chromatin immunoprecipitation-exonuclease (ChIP-exo) method, Venters et al. identified transcription initiation complexes, as many as 160,000 copies, in the human genome. However, only ~5% of the initiation complex was associated with mRNA genes with the rest belonging to ncRNAs genes. It shows from a side view that ribosomes may occupy any accessible sites in the genome ¹⁰⁹. Hence, translation of “ncRNAs” into peptides or proteins may be fundamentally different from that of mRNAs. Yet, it remains to be studied whether the occupancy of ribosomes on ncRNAs is one of the key features of eukaryotic translation.

Unknown worlds for peptides originating from ncRNAs are probably more fantastic than previously assumed. Exploring such worlds may stimulate us to investigate the precise functions of the prevalent translation of ncRNAs. To elucidate the precision functions, we need to not only study the underlying mechanisms by which ncRNAs code for peptides or proteins, but also identify the precise roles of the resultant peptides or proteins, in particular, in disease initiation and development. Knowing these answers will improve our understanding of this biological process and help us develop new therapeutics for disease prevention and therapy.

Importance of peptides or proteins derived from ncRNAs in theranostics

It has been observed that RNA polymerases tend to trigger transcription at the reputed promoter regions, suggesting that the probability of triggering transcription on reachable DNA sequences is low ¹¹⁰. However, many random DNA sequences can promote transcription by recruiting TFs. This is consistent with the fact that most of the nucleosome-free DNA could be transcribed in vivo ¹¹¹. Like mRNAs, some lncRNAs are capped, spliced and polyadenylated ¹¹². So, they also have the chance to be caught by ribosomes and translated, though less frequently than most mRNAs ¹¹³. Hence, there is sufficient evidence to believe that ncRNAs may serve as one kind of substrate for the basal translational machinery and could be translated into a random peptide ¹¹⁴.

How novel peptides contribute to normal development and disease will open up new avenues for this field. It is well known that cancer develops by accumulation of driver mutations. Many such mutations occur in non-coding regions of the genome ¹¹⁵. It is possible that short peptides arising from the so-called non-coding regions may indeed have a higher level of expression in cancers. Moreover, they may also function in the occurrence and development of tumors ¹¹⁶. An improved understanding of ncRNAs and their corresponding peptides may result in the discovery of new methods to diagnose and treat cancer. For example, Huang et al. reported that a 53-aa peptide encoded by lncRNA HOXB-AS3 is the regulator of pyruvate kinase M (PKM) for alternative splicing and cancer metabolism reprogramming. They found reduced HOXB-AS3 peptide levels were related with a weak prognosis in colon cancer (CRC) patients and the loss of this peptide is one critical oncogenic event. Their findings uncovered a complex regulatory mechanism of cancer metabolism reprogramming orchestrated by a lncRNA-encoded peptide (Figure 8) ¹¹⁷. Also, ncRNAs and their corresponding peptides may play a significant, but yet to be characterized, role in other diseases.

Figure 8.

Working model for HOXB-AS3 peptide. Instead of functioning by lncRNA directly, the peptide derived from HOXB-AS3 competitively binds to the arginine residues in RGG motif of hnRNP A1 and antagonizes the hnRNP A1-mediated regulation of pyruvate kinase M (PKM) splicing by blocking the binding of the arginine residues in RGG motif of hnRNP A1 to the sequences flanking PKM exon 9, ensuring the formation of lower PKM2 and suppressing glucose metabolism reprogramming. (hnRNP: heterogeneous nuclear ribonucleoprotein; HOXB-AS3: HOXB cluster antisense RNA 3; PKM: pyruvate kinase M; RGG: Arg-Gly-Gly; TCA cycle: tricarboxylic acid cycle). Adapted with the permission from ref 117, copyright 2016 Elsevier.

Small peptides can be used in a variety of ways such as antibacterial agents, cell signaling molecules, and cytoskeletal modulators ¹¹⁸. For example, dominant-negative peptides are as short as ~100-aa long. Another short peptide of 119-168 aa, termed helix-loop-helix (HLH)-like peptides, could capture basic proteins to form biologically active forms. Specifically, the peptides can modulate organ development, control stem and cancer cell behaviors and even modify circadian rhythms in humans ¹¹⁹. The study of Huang et al. opened a new prospect of the study of lncRNAs, especially in human development and tumorigenesis. Thus, there are reasons to believe that an increasing number of small peptides, which have been ignored in current clinical medicine and basic biomedical investigations, will be characterized in future studies.

We believe that exploring the pathological and physiological effects of new peptides generated by ncRNAs may unlock a new scientific world. The rules about biogenesis and actions should be more complicated than what we originally believed. We are witnessing a substantial improvement in our understanding of what was once thought to be a fully characterized biological process. With this in mind, researchers should thoroughly study the new coding mechanisms across all genes in the genome and develop new means to purify and identify peptides originating from the ncRNAs. Functional analyses in the future should also focus on the way in which peptides encoded by the ncRNAs with smORFs contribute to the various basic biological processes and mechanisms of some important diseases like cancer. These will very likely lead to new breakthroughs in the development of life science and medicine.

Concluding remarks and future perspectives

The discovery of ncRNAs translation and smORFs in the transcriptome highlights the need for experimental maps of pervasive translation ^{31, 120}. To date, however, very little, if any, is clear about the biological mechanisms of coding “ncRNAs” as well as the understanding of the biological function of the peptides or proteins derived from them. Future studies need to be performed to enhance our understanding of coding ncRNAs and the role of their corresponding peptides or proteins in disease development and theranostics. The following questions have yet to be answered about translation of putative ncRNAs. First, what is the exact mechanism of smORFs encoding for functional peptides or proteins? Second, what are the functions of these peptides and how do they work? Third, how can we computationally or experimentally identify the translated peptides? And finally, how are new peptides or proteins translated from ncRNAs?

Abbreviations

aa

amino acid

AKT

protein kinase B

ATXN8

ataxin 8

BiP

binding immunoglobulin protein

BLAST

protein basic local alignment search tool

CPC

coding potential calculator

CRC

colon cancer

CRITICA

coding region identification tool invoking comparative analysis

CSC

cancer stem cell

DWORF

dwarf open reading frame

eIF4G2

eukaryotic translation initiation factor 4 gamma 2

ENCODE

Encyclopedia of DNA Elements

ESC

embryonic stem cell

FTLD/ALS

frontotemporal lobar degeneration and amyotrophic lateral sclerosis

FTO

fat mass and obesity-associated

GDP

guanosine diphosphate

GFP

green fluorescent protein

GTP

guanosine triphosphate

HA

human influenza hemagglutinin

HCC

hepatocellular carcinoma

hnRNP

heterogeneous nuclear ribonucleoprotein

HOXB-AS3

HOXB cluster antisense RNA 3

IRESs

internal ribosome entry sites

ISR

integrated stress response

lncRNAs

long non-coding RNAs

m⁶A

N⁶-methyladenosine

METTL3/14

methyltransferase-like 3/14

micPDP

micro-peptide detection pipeline

miRNAs

microRNAs

miPEP

mi-peptides

MLN

myoregulin

mRNAs

messenger RNAs

MS

mass spectrometry

ncRNAs

non-coding RNAs

PGAM3

Phosphoglycerate mutase 3

PhyloCSF

phylogenetic analysis of codon substitution frequencies based on sequence alignment

PKM

pyruvate kinase M

PLN

phospholamban

RGG

Arg-Gly-Gly

rRNAs

ribosome RNAs

SCA8

spinocerebellar ataxia type 8

SERCA

sarco endoplasmic reticulum Ca²⁺ -ATPase

SLN

sarcolinpin

sm ORFs

small open reading frames

snRNAs

small nuclear RNAs

SPAR

small regulatory polypeptide of amino acid response

SR

sarcoplasmic reticulum

TCA cycle

tricarboxylic acid cycle

tRNAs

transfer RNAs

uORFs

upstream ORFs

UTR

untranslated region

YTHDF3

YTH domain family protein 3

WB

western blot.