Abstract

RNA crosstalk in competitive endogenous networks

Competition between various molecular species to bind to a specific molecular target has been described in multiple contexts, and includes DNA–protein, RNA–protein, RNA–RNA and protein–protein crosstalk. This Review focuses on crosstalk involving RNA–RNA interactions and an additional regulatory layer of RNA–protein interactions.

Artificial miRNA sponges as miRNA competitors

Several years before the discovery of naturally occurring miRNA sponges, or ceRNAs, various groups described the use of artificial miRNA sponges as effective miRNA inhibitors^28–30. These sponges are usually expressed from strong promoters, contain multiple binding sites for an miRNA of interest and have been shown to derepress miRNA targets at least as effectively as chemically modified antisense oligonucleotides³⁰. The efficacy of artificial sponges has been demonstrated for multiple miRNAs both in vitro and in vivo^30–34. Artificial miRNA sponges constructed with multiple binding sites for different miRNAs may also be used to study the effect of several miRNAs simultaneously^30,32.

Intriguingly, although sponges with perfectly complementary miRNA-binding sites have been shown to be effective^29,30,33, ‘bulged sponges’, which include a central bulge and hence bind miRNAs with imperfect complementarity, have been demonstrated to sequester miRNAs with greater efficacy^{30–33,35,36}. This may be partly due to the fact that, unlike perfectly complementary targets, imperfect targets are not immediately degraded and are thus able to reduce miRNA bioavailability until the mRNA is destabilized by other factors^31,35. These artificial miRNA sponges will not only be invaluable tools for miRNA loss-of-function studies in vitro and in vivo, but may also provide a new platform for RNA-based therapeutic applications.

Natural miRNA sponges as ceRNAs

It has been proposed that naturally occurring protein-coding and non-coding RNA transcripts can act as endogenous miRNA sponges, or ceRNAs^11,12,15. ceRNAs communicate with, and co-regulate, each other by competing to bind to shared miRNAs, thereby titrating miRNA availability. Because ceRNA crosstalk can be deciphered bioinformatically and experimentally, the identification and analysis of ceRNA interactions may allow the systematic functionalization of the non-coding transcriptome, as well as a non-protein-coding function to be attributed to mRNA transcripts, which may be complementary or even distinct from their protein-coding function¹¹.

Non-coding RNAs as ceRNAs

Increasing experimental evidence supports the hypothesis that multiple non-coding RNA species, including small non-coding RNAs, pseudogenes, lncRNAs and circular RNAs (circRNAs) may possess ceRNA activity.

Pseudogene ceRNAs

In humans, the non-coding PTENP1 pseudogene has been reported to regulate levels of its cognate gene, the tumour suppressor PTEN, by competing for shared miRNAs¹⁶ (Fig. 1). PTENP1 has been shown to act as a decoy for PTEN-targeting miRNAs, to possess tumour suppressive activity and to be selectively lost in human cancers. An additional dimension to PTENP1-mediated ceRNA regulation was recently out-lined³⁷, whereby two antisense RNAs encoded by the PTENP1 locus α and β were characterized. The β isoform was shown to interact with the PTENP1 transcript through an RNA–RNA pairing interaction, affecting its stability and ceRNA activity³⁷.

Open in a separate window

Figure 1

PTEN competing endogenous RNA (ceRNA) network

Validated interactions (grey arrows) between ceRNAs have been reported, shown here with the respective validated microRNAs (miRNAs). Studies have also identified potential ceRNA interactions (blue arrows) based on the respective validated miRNAs. Validated PTEN ceRNAs described in ref. 26 are not included because the specific miRNAs were not identified.

In addition, overexpression of the KRAS pseudogene KRAS1P 3′ untranslated region (UTR) has been reported to increase KRAS transcript abundance and accelerate cell proliferation. The KRAS1P locus was amplified in various human cancers, and KRAS1P transcript levels positively correlated with KRAS transcript levels in prostate cancer. Taken together, these findings were particularly relevant because they attributed a new function to transcribed pseudogenes, lncRNAs, which have been considered to be biologically inconsequential owing to their inability to be translated into functional proteins. Although transcribed pseudogenes may be expressed at much lower levels than their cognate genes, this is counterbalanced by their high degree of shared sequence homology, which results in the conservation of multiple miRNA-binding sites and allows them to compete for the binding of many shared miRNAs simultaneously. Furthermore, it has been suggested that RNA transcripts that contain premature stop codons, such as pseudogenes, may be subjected to nonsense-mediated mRNA decay³⁸. This rapid turnover may conceivably lead to their low abundance, as well as the increased degradation of bound miRNAs, enhancing their effectiveness as miRNA sponges.

Another example of a pseudogene ceRNA was described by Marques et al., who focused their attention on ‘unitary pseudogenes’, which retain protein-coding capability in the human lineage but lose protein-coding function in the rodent lineage. This allowed the RNA-mediated function of RNAs to be disentangled from their ancestral protein-coding function. The mouse Pbcas4 transcript was validated as a conserved ceRNA for human BCAS4, which downregulated expression levels of the protein-coding transcripts Bcl2, Il17rd, Pnpla3, Shisa7 and Tapbp by acting as a miR-185 decoy³⁹. The preservation of miRNA response elements (MREs) and thus miRNA sponge function of this unitary pseudogene after the loss of its protein-coding function lends support to the hypothesis that ceRNA interactions represent a conserved and biologically relevant mechanism of post-transcriptional gene regulation, which confers an additional non-protein-coding role to protein-coding transcripts.

mRNAs as ceRNAs

The gene with perhaps the most extensively characterized ceRNA network is the important tumour suppressor PTEN. Aside from the non-coding pseudogene PTENP1 (discussed earlier), the PTEN ceRNA network includes multiple protein-coding transcripts (Fig. 1, Table 2).

A combined computational and experimental approach has been used to identify CNOT6L and VAPA as protein-coding transcripts that regulate PTEN transcript and protein expression in a Dicer-dependent manner, antagonize downstream PI(3)K signalling and possess growth-and tumour-suppressive properties¹⁷. These genes were also coexpressed with PTEN mRNA in several human cancers and displayed copy number loss in colon cancer. The study demonstrated that previously uncharacterized transcripts could be functionalized, partly through the identification of their ceRNA interactors, and presented a framework for the prediction and validation of ceRNA interactions that is widely applicable to any potential transcript of interest.

One of the first examples linking aberrant ceRNA expression to tumorigenesis in vivo came from the discovery of a significant enrichment of predicted PTEN ceRNAs among genes whose loss accelerated melanomagenesis in a Sleeping Beauty insertional mutagenesis screen in vivo. ZEB2 was subsequently validated as a bona fide PTEN ceRNA that modulated PTEN protein levels in an miRNA-dependent, protein-coding-independent fashion¹⁸. Decreased ZEB2 transcript levels activated downstream signalling, enhanced cell transformation and were found to occur frequently in human melanoma and other cancers with low PTEN mRNA expression. These data suggest that the dysregulation of PTEN expression owing to the loss of its ceRNA ZEB2 contributes to the development of melanoma both in vitro and in vivo.

Analysis of gene expression data in glioblastoma in combination with matched miRNA profiles validated 13 PTEN ceRNAs or miR program-mediated post-transcriptional regulatory (mPR) regulators whose locus deletions were predictive of decreased PTEN expression, downregulated PTEN in a 3′ UTR-dependent manner and increased tumour cell growth rates²⁶. When the analysis was significantly extended beyond the binary ceRNA associations described in most other studies, the PTEN ceRNA interactions were found to be part of a post-transcriptional regulatory layer comprising more than 248,000 miRNA-mediated interactions.

The VCAN 3′ UTR has been reported to modulate PTEN levels by sequestering the shared miRNAs miR-144 and miR-136, freeing PTEN mRNA for translation⁵⁰. In addition to its role as a PTEN ceRNA, VCAN has been shown to modulate the expression of several other genes through miRNA competition (Fig. 1). VCAN was also demonstrated to act as a ceRNA for the cell-cycle regulator RB1 by regulating miR-199a-3p and miR-144 levels, upregulating the expression of this crucial tumour suppressor both in vitro and in vivo⁵⁰. CD34 and FN1 were validated as two additional VCAN ceRNAs that communicate through interactions with miR-133a, miR-199a-3p, miR-144 and miR-431 (refs ^{27, 51}). As overexpression of the VCAN 3′ UTR in vivo was shown to induce organ adhesion followed by hepatocellular carcinoma development at a later time point, these studies were the first to demonstrate that the perturbation of a non-coding RNA transcript with validated ceRNA function has physiological and pathophysiological relevance in vivo.

Another gene with validated protein-coding ceRNAs is CD44, which encodes a transmembrane glycoprotein that is involved in a wide range of cellular functions⁵². The CD44 3′ UTR was shown to inhibit cell proliferation, colony formation, tumour growth and enhance cell motility, invasion and adhesion^52,53. CD44 regulated levels of CDC42, a Rho-GTPase involved in the control of cell morphology, migration and cell-cycle progression, by binding and sequestering three miRNAs, miR-216a, miR-330 and miR-608. It was also shown to modulate levels of Col1α1 through miR-328 binding and FN1 through miR-512-3p, miR-491 and miR-671 binding. The ceRNA interactions between CD44 and FN1 link CD44, CDC42 and Col1α1 to the broader miRNA–ceRNA network revolving around PTEN and VCAN. This adds another level of complexity to these post-transciptional regulatory associations between multiple transcripts, which play important parts in tumorigenesis (Fig. 1).

A recent study outlined the role of Hmga2’s ceRNA function in promoting lung carcinogenesis⁵⁴. Hmga2, a non-histone chromosomal high mobility group protein, was shown to be highly expressed in metastatic lung adenocarcinoma and to promote lung cancer progression by sequestering the abundant let-7 family of miRNAs. The TGF-β co-receptor Tgfbr3 was identified as a putative Hmga2 ceRNA, and Tgfbr3-driven TGF-β signalling was demonstrated to be largely necessary for Hmga2-mediated lung-cancer progression. The discovery that the primary function of a protein-coding transcript is to act as an oncogenic ceRNA for a very abundant miRNA, largely independently of its protein-coding function, provides further support for the hypothesis that the ceRNA function of multiple protein-coding transcripts is of fundamental importance in cancer progression.

A viral mRNA that functions as a natural miRNA sponge in host cells has also been described. Intriguingly, this mRNA was found to sequester the same miRNA reported to be downregulated by HSUR1-binding in H. saimiri, suggesting that ceRNA-mediated regulation of this miRNA may be a conserved mechanism for controlling host gene expression in multiple viruses. Two studies independently showed that the highly abundant murine cytomegalovirus mRNA m169 acted as a natural miRNA sponge in host cells by binding to miR-27 through a single site in its 3′ UTR, hence directing its degradation^55,56. Importantly, m169-mediated miR-27 regulation is another example of the importance of ceRNA function in vivo: disruption or replacement of the miRNA target site resulted in significant viral attenuation in multiple organs, suggesting that this miRNA-sponge interaction is crucial for efficient replication in vivo.

Predicting ceRNA crosstalk

ceRNA crosstalk depends on the MREs located on each transcript, which combinatorially form the foundation of these co-regulatory interactions¹¹. Prediction of ceRNA crosstalk is thus dependent on the identification of MREs on the relevant transcripts of interest. Several miRNA-target prediction algorithms, including TargetScan, miRanda, rna22 and PITA, have been successfully used to identify ceRNA interactions. However, miRNA target identification is challenging owing to the imperfect nature of base pairing between an miRNA and its target, and the rules of targeting are not completely understood⁵⁷. Further development and optimization of these algorithms will undoubtedly improve subsequent predictions of ceRNA interactions.

A database of predicted ceRNA interactions (ceRDB), which was generated by examining the co-occurrence of MREs on a genome-wide basis⁵⁸, has recently been compiled. Predicted miRNA–mRNA target interactions were obtained from TargetScan release 5.2, and interaction scores were defined for each mRNA by adding the total number of MREs that overlap with the miRNAs for the mRNA of interest; these interaction scores were then used to rank the predicted potential ceRNAs. Although this analysis is limited to the 3′ UTRs of protein-coding transcripts, it still represents a useful tool for the identification of putative ceRNA crosstalk.

As an alternative to in silico prediction strategies, recently developed high-throughput biochemical techniques, which identify endogenous miRNA–target interactions can be used. Examples of these experimental methods include high-throughput sequencing of RNA isolated by crosslinking immunoprecipitation (HITS-CLIP)⁵⁹ and photoactivatable- ribonucleoside-enhanced crosslinking and immunoprecipitation (PAR-CLIP)⁶⁰. HITS-CLIP analysis of argonaute-bound miRNA–mRNA complexes has generated genome-wide interaction maps for specific miRNAs⁶¹. Recently, integration of HITS-CLIP data was shown to enhance miRNA target predictions by more than 20-fold over computational approaches alone, and mRNA–miRNA predictions have been used to identify candidate ceRNA interactions in the regenerating liver⁶². PAR-CLIP is a modification of the HITS-CLIP methodology with improved RNA recovery, which is able to indicate the exact targeting site more precisely. A complementary approach to these is MS2-tagged RNA affinity purification (MS2-TRAP), which can be used to identify all miRNAs associated with a target transcript in a particular cellular context⁶³. Harnessing these experimental techniques will provide further insight into ceRNA regulation beyond that which is possible with in silico target predictions.

Additional considerations for ceRNA activity

Although multiple examples of ceRNA interactions have been described, little is known about the molecular conditions necessary for optimal ceRNA activity. Considerations such as the abundance of key players, potential interplay with RBPs and RNA editing may have profound effects on ceRNA crosstalk.

Abundance of miRNAs, ceRNAs and argonaute

Various factors, including miRNA- and ceRNA-expression levels and subcellular localization, the number of shared miRNAs and MREs, as well as binding affinity of the shared MREs have been suggested to contribute to ceRNA effectiveness^10,11,64 (Fig. 2). For example, it has been hypothesized that miRNA–RNA competition would apply only to a small subset of miRNAs whose abundance and corresponding target abundance fell within a narrow range, as the expression of competitor RNAs would have little impact on regulation by highly abundant miRNAs, and low abundance miRNAs would be unlikely to contribute much to gene regulation because a minimal number of targets would be bound at any given time¹³.

Open in a separate window

Figure 2

Variable factors that may influence competing endogenous RNA (ceRNA) effectiveness

a, Steady state levels of ceRNA1 (protein-coding region shown in red) and ceRNA2 (protein-coding region shown in blue). Active repression by miRNAs is shown in light grey. b, If expression of ceRNA2 increases, this will increase expression of ceRNA1. c, ceRNA3 has a different subcellular localization from ceRNA1, and may be a less effective ceRNA than ceRNA2. d, Increased expression of the shared miRNA will increase repression of both ceRNA1 and ceRNA2. e, ceRNA4 contains miRNA response elements (MREs, purple and orange) for multiple shared miRNAs, and may be a more effective ceRNA than ceRNA2. f, ceRNA5 contains more MREs than ceRNA2 and may be a more effective ceRNA.

Recent studies have further refined the dynamics and constraints of ceRNA crosstalk. Optimal conditions for ceRNA activity in silico have been determined using a mathematical mass-action model, and these were confirmed experimentally using PTEN and its validated ceRNA VAPA⁶⁵. In addition, a minimal rate equation-based model to describe crosstalk between ceRNAs at steady state has been developed⁶⁶. The relative abundance of both ceRNAs and miRNAs, their stoichiometry, the number of shared MREs and indirect interactions were shown to be crucial determinants of ceRNA cross-regulation. For instance, both studies found that ceRNA crosstalk was optimal when the transcript abundance of miRNAs and ceRNAs within a network were near equimolarity. ceRNA crosstalk would be minimal when ceRNA transcript abundance vastly exceeded that of miRNAs owing to the limited number of miRNAs available to repress their targets, as well as when miRNA transcript abundance vastly exceeded that of ceRNAs owing to maximal repression of most target transcripts.

Strong correlations in network connectivity that enhance ceRNA crosstalk have also been observed⁶⁶: ceRNA interactions may be mediated by a large number of miRNA species that individually have weak effects on their respective targets; and ceRNA regulation may be symmetrical (two ceRNAs reciprocally regulate each other) or asymmetrical (one ceRNA regulates another but not vice versa). Intriguingly, a separate study⁶⁵ found that ceRNA and transcription factor networks were intricately intertwined and, with optimal molecular conditions, changes in the levels of a single ceRNA could affect integrated ceRNA and transcriptional networks.

Another potential rate-limiting determinant of ceRNA crosstalk is the abundance of argonaute, the catalytic component of the RNA-induced silencing complex (RISC). A mathematical model constructed to analyse changes in gene expression caused by competition between small RNAs, showed that competition for argonaute binding occurs within an intermediate range of argonaute abundance, whereby lower amounts of argonaute result in stronger competition⁶⁷. Surprisingly, this model also revealed that when argonaute is highly abundant, the overexpression of one miRNA decreases expression levels of its own targets as well as the targets of other miRNAs. This is a result of the overexpressed miRNA binding to targets that share MREs for that miRNA and other miRNAs, freeing the other miRNAs to regulate other target genes.

Implications for human disease

Many of the genes with known ceRNA interactors identified so far have been implicated in human disease. For example, PTEN is a potent tumour suppressor gene that is frequently disrupted in multiple cancers and governs multiple cellular processes, including survival, proliferation, energy metabolism and cellular architecture⁸⁴; HULC is the most upregulated gene in HCC and has been shown to regulate several genes involved in liver cancer^25,85; T cells that are infected with HSUR1-expressing H. saimiri cause aggressive leukaemias and lymphomas¹⁴ and linc-MD1 expression in primary muscle cells isolated from patients with Duchenne muscular dystrophy has been shown to result in partial mitigation of the correct timing of the differentiation program²⁴. This suggests that ceRNA crosstalk is not only of fundamental importance in physiological conditions, but is also crucially relevant in various pathophysiological states.

In addition to the genomic amplifications, deletions, mutations and epigenetic modifications that dysregulate crucial gene function and result in disease pathogenesis, aberrant changes in ceRNA regulation may also contribute to disease initiation and progression. First, changes in the expression levels of ceRNA transcripts represent an additional mechanism for regulating the levels of key disease genes, such as tumour suppressors or oncogenes. ceRNA crosstalk may thus allow the functional characterization of transcripts with no other known function on the basis of their ceRNA activity. This is of particular importance because although many lncRNAs (for example, HULC) are expressed at low levels in steady state conditions, their expression is known to be dysregulated in various diseases⁶. Second, it has been shown that, compared with non-transformed cell lines, cancer cells in vitro often express mRNAs with shorter 3′ UTRs, which result from alternative cleavage and polyadenylation⁸⁶. This phenomenon has also been reported in primary breast and lung cancer tumours⁸⁷. Transcripts with shorter 3′ UTRs are not only subject to reduced miRNA regulation in cis, but will have diminished ability to act as ceRNAs through miRNA crosstalk in trans. Third, the altered expression of transcript variants of many genes through alternative splicing or differential use of initiation and/or stop codons has been linked to disease^11,88,89. These variants may produce proteins with different structures and functions, but may also introduce or eliminate MREs, and thus affect ceRNA activity. Fourth, chromosomal translocations are frequent events in cancer that may contribute to tumorigenesis. These events result in the production of various fusion proteins, but could also potentially affect ceRNA regulation owing to misexpression of 3′ UTRs under the control of non-endogenous promoters¹¹. Finally, SNPs (single-nucleotide polymorphisms) associated with various diseases may create, modify or destroy MREs. For example, SNPs in genes related to Alzheimer’s disease may potentially dysregulate miRNA and ceRNA networks and thus contribute to aberrant gene expression in Alzheimer’s disease⁹⁰. These observations suggest that ceRNA network interactions may have a role in determining the effectiveness of RNA-directed therapies⁹¹.

Future perspectives

As ceRNA research is still in its infancy, there are only a handful of validated examples of each existing class of ceRNAs, such as mRNAs, small non-coding RNAs, lncRNAs, pseudogenes and the recently described circRNAs (Tables 1 and ​and2).2). Although research should certainly focus on the identification of more ceRNAs in these categories to ascertain whether ceRNA crosstalk represents a widespread network of RNA regulation, another exciting avenue for future work is the discovery of new classes of ceRNAs. For example, argonaute proteins have been found to interact with rRNAs and tRNAs, raising the possibility that these abundant, stable, small RNA species may also act as ceRNAs⁹². Intriguingly, 3′ UTRs that are expressed independently of their associated protein-coding sequences have been described in humans, mice and flies, suggesting that they may represent another new class of ceRNAs⁹³.

Refinements in ceRNA prediction strategies will facilitate the discovery of additional ceRNA interactions. One limitation of current ceRNA prediction strategies is that they do not factor in miRNA and potential ceRNA expression levels. This is especially crucial as miRNAs and ceRNAs are known to have temporal, spatial and disease-specific expression patterns, and it has been suggested that ceRNA crosstalk applies mainly to a subset of ceRNAs and miRNAs whose cellular concentration and target abundance fall within a specific range of values^10,11,13. A precise understanding of miRNA and ceRNA abundance is thus essential for deconvoluting any ceRNA network of interest in a particular context. Although paired miRNA and mRNA expression profiles have been integrated with bioinformatic target predictions to refine potential ceRNA interactions^17,18,26, they have been based mainly on relative quantitation methods such as microarray and quantitative PCR (polymerase chain reaction) analysis that are subject to variations in probe binding and affinity. Techniques such as RNA sequencing that allow the quantitation of absolute levels of both ceRNA and miRNA transcript abundance will allow a more precise determination of ceRNA effectiveness. In addition, the integration of methods such as HITS-CLIP and PAR-CLIP, which allow the biochemical identification of endogenous miRNA–target interactions on a transcriptome-wide level in specific cellular contexts, will further refine ceRNA prediction strategies that currently depend mainly on in silico bioinformatic miRNA predictions.

Another limitation of current ceRNA prediction methods is their focus on crosstalk between canonical MREs that are located exclusively on the 3′ UTRs of target transcripts. A number of recent studies have suggested that miRNA regulation is not limited to canonical 3′ UTR targets. For example, several groups have demonstrated that a significant proportion of MREs in humans are non-canonical^94,95, and others have shown that effective miRNA targeting can occur outside the 3′ UTR^96,97. Furthermore, it has been shown that miRNA binding does not always result in target downregulation⁹⁸; 5′ UTR structure has been demonstrated to be a crucial determinant of miRNA-dependent regulation⁹⁹; and, for dynamically recruited miRNAs, changes in overall miRNA expression were found to not always be correlated with miRNA recruitment to the RISC⁶². These new insights into miRNA function should be incorporated into future ceRNA prediction analysis, and may potentially add further complexities to the dynamics of ceRNA regulation and significantly expand the realm of potential ceRNA activity.

Although most of the ceRNA interactions identified so far have been between binary partners, there is increasing evidence that ceRNAs crosstalk in large interconnected networks; and that, in addition to direct interactions through shared miRNAs, secondary indirect interactions may also have a profound effect on ceRNA regulation^26,65. For example, PTEN expression has been shown to be regulated by ceRNA interactions between PTEN and its ceRNAs PTENP1, CNOT6L, VAPA, VCAN and ZEB2 (Fig. 2). In addition, indirect ceRNA crosstalk (for example, between VCAN and FN1), as well as secondary interactions between PTEN ceRNAs (for example, between PTENP1, CNOT6L and VAPA mediated by the miR-17 and miR-19 families) may also contribute significantly to the effectiveness of ceRNA regulation. Future ceRNA research should extend beyond the identification of binary ceRNA interactions and include network analysis of potential intertwined miRNA and ceRNA networks.

The conservation of ceRNA crosstalk in multiple organisms, including plants, zebrafish, mice, humans and viruses, suggests that it may represent an important widespread layer of RNA regulation. As already discussed, this regulation is not limited to non-coding RNAs, it also encompasses protein-coding mRNAs. Recently, the genome of the carnivorous bladderwort Utricularia gibba was found to contain a typical number of protein-coding genes, but significantly less non-genic DNA¹⁰⁰, providing one example whereby a large non-coding genome is not essential for complex life. This presents an interesting opportunity to study potential RNA regulatory networks that involve mainly mRNAs and a much smaller non-coding transcriptome that consists of non-coding RNAs encoded in genic DNA such as circRNAs and antisense non-coding RNAs.

In summary, analysis of ceRNA interactions and crosstalk in intertwined networks may represent a robust platform to methodically functionalize non-coding transcripts on the basis of competition for shared miRNAs. Furthermore, the analysis sheds light on the non-coding function of protein-coding transcripts, and may uncover regulatory networks that have been overlooked by conventional protein-coding studies. To fully understand and appreciate the impact of ceRNA crosstalk on physiological and pathophysiological conditions, it will be of crucial importance to integrate these miRNA–RNA competitive networks with other competitive regulatory networks such as protein–RNA crosstalk. Although the full extent of ceRNA networks remains to be determined, this miRNA–RNA competition is a new post-transcriptional layer of endogenous competitive gene regulation, which has exciting implications for multiple basic biological systems, pathophysiological conditions and the development of new therapeutic approaches for cancer and other diseases.

Acknowledgments

Footnotes

References