Does the linear Sry transcript function as a ceRNA for miR-138? The sense of antisense [version 2; peer review: 2 approved]

Recently, the sex determining region Y ( Sry ) and the cerebellar degeneration-related protein 1 ( CDR1as ) RNA transcripts have been described to function as a new class of post-transcriptional regulatory RNAs that behave as circular endogenous RNA sponges for the micro RNAs (miRNAs) miR-138 and miR-7, respectively. A special feature of the Sry gene is its ability to generate linear and circular transcripts, both transcribed in the sense orientation. Here we remark that both sense (e.g. Sry RNA) and antisense (e.g. CDR1as ) transcripts could circularize and behave as miRNAs sponges, and importantly, that also protein-coding segments of mRNAs could also assume this role. Thus, it is reasonable to think that the linear Sry sense transcript could additionally act as a miRNA sponge, or as an endogenous competing RNA for miR-138. This article presents a short summary and review of recent work on circular RNAs and on competing endogenous RNAs, which can sequester and “sponge” microRNAs. A point made by the authors is that both sense and antisense RNAs from the same genomic region can act as miRNA sponges and that linear transcripts from the same regions that produce circular RNAs can themselves sequester miRNAs. Overall, I found the article to be provocative, though not novel. During the past several years there have appeared a number of papers describing circular RNAs of various sorts (for example, stable introns and circles resulting from splicing events commonly promoted by strong secondary structures flanking exons). There has also been a wealth of published work suggesting or showing that abundant RNAs can bind miRNAs and thus reduce their availability for the regulation of some mRNAs. Thus, this paper breaks little new ground, though it does point out that many genomic regions express both sense and antisense transcripts, and each of these can affect miRNA availability. To me, the area covered by this paper is a very interesting and important one, and it is of value to present and discuss new concepts in gene regulation. Thus, I think this manuscript would be improved by:

Crosstalk involving RNA-RNA interactions adds a new dimension to our understanding of complex regulatory networks and offers profound implications for the elucidation of gene function 1 .
MicroRNAs (miRNAs) are a type of endogenously expressed small regulatory non-protein-coding RNAs that negatively regulate gene expression by base-pairing (with imperfect complementarity) to miRNA response elements (MREs), which are usually located within the 3′-untranslated region (3′-UTR) of target RNA transcripts 2 . According to their number and location, it has become evident that a biological process may involve multiple miRNAs, and that a given gene may be regulated by more than one miRNA. A recently discovered molecular mechanism, named Competing Endogenous RNA (ceRNA) effect, has highlighted the importance of indirect interactions among transcript RNAs competing for the same pool of miRNAs 3 . In the case of two ceRNAs having MREs in common, both are regulated by the same set of miRNAs. Multiple classes of non-coding RNAs (lncRNAs) including circular endogenous RNA sponges (circRNAs), pseudogenes, and protein-coding mRNAs function as key ceRNAs and "super-sponges" to regulate the expression of mRNAs in plants and mammalian cells 4 . The effectiveness of a ceRNA would depend on the number of miRNAs that it can ''absorb''. This, in turn, would depend on the ceRNA's accessibility to miRNA molecules, which is influenced by its subcellular localization and its interaction with RNA-binding proteins. Furthermore, the specific cellular context in which the ceRNA is expressed would also impact its overall influence because not all microRNAs are present ubiquitously and at all times 5 . Aberrant expression of central nodes of such ceRNA networks may cause a disturbance that could contribute to disease pathogenesis 6 .
Recent bioinformatic and experimental analyses have identified thousands of circRNAs in the mammalian transcriptome, suggesting that circRNAs may in fact represent a new class of ceRNA regulators 7 . These circRNAs are produced mainly through a type of alternative RNA splicing named 'back-splicing', in which a splice donor splices to an upstream acceptor rather than a downstream acceptor 8  Hansen et al. also showed that another circular RNA molecule, transcribed from the mouse Sry gene, could also act as an endogenous sponge. They noted that this transcript contains 16 binding sites for miR-138 and demonstrated in vitro that the Sry circRNA selectively "absorbs" this specific miRNA. Recently, Kartha and Subramanian asserted, based on the report by Hansen et al., that this Sry RNA is an antisense circular transcript that functions as a miRNAs sponge 11 . Although this apparently is a typographical error (antisense instead of sense), it was also referred as such in the original report by Memczak et al. in Nature. This suggests that the circular Sry transcript is, as occurs with the CDR1as sponge, an antisense circular RNA. Although it seems obvious that sponges are antisense to the miRNA they bind to, it should not be assumed that all circRNAs are transcripts in an antisense orientation to a protein coding gene, as occurs with CDR1as.
Natural antisense transcripts (NATs) are ncRNAs transcribed from the opposite strand of a coding gene and are capable of regulating the expression of their sense gene pair or of several related genes 12 . Genomic loci that express NATs are highly abundant and sense/antisense (SAS) transcript pairs tend to be co-expressed. The most comprehensive studies predict that in human and mice 40-72% of all transcriptional units show evidence of bi-directional transcription 13 . To our knowledge, there are no reports of a circRNA whose sequence in sense and antisense orientation, possesses the ability to function as miRNA sponge. A special feature of the Sry gene is that it can generate linear as well as circular transcripts depending on the use of alternative promoters (proximal vs distal) 14 . Capel et al. reported for the first time that the circular Sry RNA is derived from a sense sequence that consists of a single exon. This molecule is formed by the processing of a longer precursor transcript that contains one inverted repeat at each end. This unusual configuration promotes the formation of a stem-loop structure that facilitates the nucleophilic attack of a donor splicing site at the 3′ end to an acceptor site at the 5′ end, which results in its circularization 14 ( Figure 1). Thus, it can be asserted that this is, in fact, a circular sense Sry mRNA. Although the notion that the Sry circRNA is derived from an antisense transcript does not alter the interpretation of the results obtained by Hansen et al., we consider that this distinction is important, because it implies that both sense (e.g. Sry RNA) and antisense (e.g. CDR1as) transcripts could be circularized and act as RNA sponges, an observation which is not acknowledged by the authors of either of the original papers. To the best of our knowledge, no antisense transcript of the murine Sry gene has been reported. Nevertheless, if the circular version of the Sry transcript can soak up miRNAs, can the Sry linear transcripts also do the same?
In this respect, there is evidence that certain miRNAs may function by targeting sites in the 5′-UTR 15 and open reading frame (ORF) regions of mRNAs 16 , suggesting that miRNAs may modulate gene expression by mechanisms different from canonical 3′-UTR target mRNA suppression. Binding of a miRNA to a ceRNA not only prevents that miRNA from binding to other MREs, but can

Amendments from Version 1
In the current version we have included additional information in response to the comments and observations made by the reviewers.
We have discussed aspects indicated by both referees about the ceRNA field, genes transcribed bi-directionally, biogenesis of circRNA molecules, requirements for ceRNA functionality, as well as the impact of ceRNA copy number on the miRNA function.

See referee reports
REVISED also repress translation from the coding segment of the ceRNA 17 . A study of the pseudogene of the phosphatase and tensin homolog PTEN, PTENP1, provided the first experimental evidence for the cross-talk between coding and non-coding RNAs 18 . Tay et al. found that several endogenous protein-coding transcripts, such as serine incorporator 1 (SERINC1), vesicle-associated membrane protein associated protein A (VAPA), CCR4-NOT transcription complex and subunit 6-like (CNOT6L), act as PTEN ceRNAs, which regulate PTEN tumor suppressor levels in a miRNA-dependent manner 18 . This clearly suggests that mRNAs can function as ceRNAs and we propose that the mouse linear Sry sense transcript could also behave as a miRNA sponge, or as a ceRNA for miR-138.
Recently, Denzler et al. questioned the biological relevance of ceRNAs in terms of the abundance of these molecules which would be required to induce derepression of the targets of specific miRNAs 19 . However, Memczak et al. and Hansen et al. shown that circRNA behaving as miRNA sponges selectively bind miRNAs forming complexes with Ago proteins, which raises the possibility that ceRNAs modulate gene expression not only by capturing miRNAs but also through the depletion of the pool of available effector molecules of the miRNA pathway. Additionally, Denzel et al. based their calculations for target abundance on sites present in transcriptome 3′ UTRs, however, they were unable to rule out that unidentified highly abundant and regulated non coding RNAs (including circRNAs) might substantially contribute to the pool of available binding sites, a limitation acknowledged in their paper 19 . This may be of particular importance in the adult testis, which express the circular Sry transcript and also has been shown to provide a permissive environment for transcription initiation, a phenomenon that has been called "transcriptional promiscuity" 20 . Denzel et al. also state that their findings in liver tissue can be generalized to other tissue and disease states, given that target abundance did not show large changes in the presence of insulin signaling or liver disease, conditions know to modify gene expression in such tissue. However, the authors also discuss that during cellular processes such as differentiation (like the spermatogenesis in the adult testis), expression of coding and noncoding RNAs changes dramatically, potentially making these systems more amenable to ceRNA-mediated gene regulation 19 .
Shortly after the emergence of circRNAs, the first public circRNA database (circBase version 0.1) was developed by the Rajewsky laboratory as a compendium of thousands of circRNAs sequences that are expressed in eukaryotic cells 21 . Access to this resource allows us to use the information in order to validate those circRNAs that are probably involved in many important cellular processes. Nevertheless, the precise molecular mechanisms that underlie post-transcriptional repression by circRNAs remain still largely unknown, but their discovery demonstrates the importance of this distinct type of non-protein-coding regulatory RNAs for the elucidation of gene function. The extent to which other animal or human antisense or sense circRNAs also behave as miRNA sponges will doubtlessly be a subject of intense research. Moreover, due to their longer half lives in vivo, circRNAs may possess a great potential for therapeutic intervention. Thus, manipulating miRNA function, either by mimicking or inhibiting ceRNAs implicated in several disorders such as cancer, could provide a novel strategy to interfere with disease initiation and/or progression. The antisense modulation of circRNAs/ceRNA→miRNAs→mRNAs→protein regulatory networks could offer ingenious decoy combinations (antisense technology) as well as delivery platforms for concurrently target multiple miRNAs in abnormal or undesired conditions 22 .
Author contributions JTGR and GAJ contributed extensively to this work and were involved in the critical revision of the manuscript. Both authors have agreed to the final version of the manuscript.

Competing interests
No competing interests were disclosed. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Amy Pasquinelli
Division of Biological Sciences, University of California San Diego, La Jolla, CA, USA This short commentary points out that both sense and antisense transcripts with miRNA target sites could serve to titrate the miRNAs and, thus, release other target RNAs from repression. Recently, there has been much interest in these so called competing endogenous RNAs (ceRNAs) as a form of regulating miRNA activity. While ceRNA activity has been attributed to pseudogenes and transcripts that circularize (circRNAs), theoretically any RNA, including protein coding mRNAs, could serve this function. The correspondence proposes that the linear Sry sense, as well as the previously reported antisense, transcript could function as sponges for miR-138. A major criticism of the proposal that ceRNA activity plays a function in regulating miRNA availability for target regulation is the consideration of cellular RNA concentrations. In many cases, the ceRNAs, even the circRNAs which can be more stable, are expressed at levels far below the miRNA or even its target mRNA. In fact, after this correspondence by Granados-Riveron & Aquino-Jarquin was originally published in April 2014, the hypothesis that ceRNAs can alter miRNA function in vivo was rigorously tested and the conclusion was published in May that most ceRNAs simply are not abundant enough to act as competitive inhibitors of miRNA binding to target mRNAs (Denzler et  al., 2014). Thus, the authors need to include discussion of this concern over how generally miRNA activity might actually be affected by the presence of any kind of ceRNA, sense, antisense or circular.

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.
Author Response 06 Nov 2014

Guillermo Aquino-Jarquin, Hospital Infantil de México Federico Gómez, Mexico City, Mexico
This short commentary points out that both sense and antisense transcripts with miRNA target sites could serve to titrate the miRNAs and, thus, release other target RNAs from repression. Recently, there has been much interest in these so called competing endogenous RNAs (ceRNAs) as a form of regulating miRNA activity. While ceRNA activity has been attributed to pseudogenes and transcripts that circularize (circRNAs), theoretically any RNA, including protein coding mRNAs, could serve this function. The correspondence proposes that the linear Sry sense, as well as the previously reported antisense, transcript could function as sponges for miR-138. A major criticism of the proposal that ceRNA activity plays a function in regulating miRNA availability for target regulation is the consideration of cellular RNA concentrations. In many cases, the ceRNAs, even the circRNAs which can be more stable, are expressed at levels far below the miRNA or even its target mRNA. In fact, after this correspondence by Granados-Riveron & Aquino-Jarquin was originally published in April 2014, the hypothesis that ceRNAs can alter miRNA function in vivo was rigorously tested and the conclusion was published in May that most ceRNAs simply are not abundant enough to act as competitive inhibitors of miRNA binding to target mRNAs (Denzler et al., 2014). Thus, the authors need to include discussion of this concern over how generally miRNA activity might actually be affected by the presence of any kind of ceRNA, sense, antisense or circular.
Firstly, we thank Dr Pasquinelli for her insightful comments. Secondly, we would like to clarify that the previously reported circular mouse Sry transcript reported by Capel et al. 1 , now known to behave a sponge for mir-138 2,3 , is, in fact, not derived from an antisense transcript but from a sense transcript, whose expression is directed by a promoter which is different to the promoter of the canonical linear Sry transcript involved in mammalian sexdetermination. To the best of our knowledge, no antisense transcript of the murine Sry gene has been reported and a recent search of our own in publicly available EST data supports this assertion. Actually, one of our motivations for publish this piece of correspondence was to clarify this misconception (that the circular Sry RNA is an antisense transcript), which was first asserted in the report by Memczak et al. in Nature 3 and later reproduced in a review of the subject by Kartha and Subramanian. 4 We agree that recent findings by Denzler et al. question the biological relevance of ceRNAs in terms of the abundance of these molecules which would be required to induce derepression of the targets of specific miRNAs 5 . However, Memczak et al. and Hansen et al. shown that miRNA sponges selectively bind miRNAs forming complexes with Ago proteins, which raises the possibility that ceRNAs modulate gene expression not only by capturing miRNAs but also through the depletion of the pool of available effector molecules of the miRNA pathway. Additionally, Denzel et al. based their calculations for target abundance on sites present in transcriptome 3´UTRs, however, they were unable to rule out that unidentified highly abundant and regulated non coding RNAs (including circRNAs) might substantially contribute to the pool of available binding sites, a limitation acknowledged in their paper 5 . This may be of particular importance in the adult testis, which express the circular Sry transcript and also has been shown to provide a permissive environment for transcription initiation, a phenomenon that has been called "transcriptional promiscuity" 6 . Denzel et al. also state that their findings in liver can be generalized to other tissues and disease states, given that target abundance did not show large changes in the presence of insulin signaling or liver disease, conditions know to modify gene expression in hepatocytes. However, the authors also discuss that during cellular processes such as differentiation (like the spermatogenesis in the adult testis), expression of coding and noncoding RNAs changes dramatically, potentially making these systems more amenable to ceRNA-mediated gene regulation 5 .
We are grateful to Dr. Gordon Carmichael for all the comments made to our manuscript.
A recently discovered molecular mechanism, named Competing Endogenous RNA (ceRNA) effect, has highlighted the importance of indirect interactions among transcript RNAs competing for the same pool of miRNAs 1 . ceRNAs share one or more miRNA response elements (MREs) and compete for a restricted pool of common miRNAs. Thus, ceRNAs form complex regulatory networks of miRNAs and MRE-containing transcripts, both proteincoding and noncoding, which ensure a tight control of many biological processes. Aberrant expression of central nodes of such ceRNA networks may cause a disturbance that could contribute to disease pathogenesis 2 . A discussion along these lines has been included in our new submission.

Discussing a bit more the well known fact that sense and antisense transcripts are commonly expressed in cells, though not necessarily from the same locus at the same time.
CDR1as is a circRNA that functions as a sponge for miR-7, deriving from an antisense transcript of the CDR1 protein-coding gene. On the other hand, the transcript of the male sex-determining gene Sry, transcribed in sense orientation 3 , is a second circRNA proposed to act as a sponge, in this case, for miR-138. However, Memczak et al. stated that "Perhaps the best known circRNA is antisense to the mRNA transcribed from the SRY (sexdetermining region Y) locus and is highly expressed in testes" 4 . Recently, Kartha and Subramanian asserted, based on the report by Hansen et al., that this Sry RNA is an antisense circular transcript that functions as a miRNAs sponge 5 . Although this apparently is a typographical error (antisense instead of sense), it was also referred as such in the original report by Memczak et al. in Nature 4 . This would suggest that the circular Sry transcript is, as occurs with the CDR1as sponge, an antisense circular RNA. Although it seems obvious that sponges are antisense to the miRNA they bind to, it should not be assumed that all circRNAs are transcripts in an antisense orientation to a protein coding gene, as occurs with CDR1as. We conducted an in silico analysis taking the antisense sequence of the Sry mRNA to make a search for potential binding sites for miR-138, using the fuzzy bioinformatics tool (EMBOSS Suite) and we found none. When we used the Sry ORF (GenBank NM_011564) for the miR-138 binding site (CACCAGCA), we found 14 potential sites which is consistent with the number of sites found by Hansen and colleagues. For this reason, we remark in our current submission that both sense and antisense Sry transcripts could circularize and behave as miRNAs sponges, and importantly, that protein-coding segments of Sry mRNA could also assume this role.
To our knowledge, there are no reports of a circRNA whose sequence in sense and antisense orientation possesses the ability to function as miRNA sponge. On the other hand, a particularly interesting family of non-protein coding RNAs consists of natural antisense transcripts (NATs). NATs are ncRNAs transcribed from the opposite strand of a coding gene and are capable of regulating the expression of their sense gene pair or of several related genes 6 . Genomic loci that express NATs are highly abundant and sense/antisense (SAS) transcript pairs tend to be co-expressed. The most comprehensive studies predict that in human and mice 40-72% of all transcriptional units show evidence of bi-directional transcription 7 . The regulatory activity of SAS pairs in human tissues has been postulated on protein expression at different levels, such as alternative splicing, post-transcriptional regulation, transport and epigenetic imprinting as well as transcriptional and translational interference through annealing to complementary sequences 8

. As requested, we have included this information in the version of our submission.
3. Mentioning the several ways that stable circular RNAs can be produced. Recent bioinformatic and experimental analyses have identified thousands of circRNAs in the mammalian transcriptome, suggesting that circRNAs may in fact represent a new class of ceRNA regulators 9 . These circRNAs are produced mainly through a type of alternative RNA splicing named 'back-splicing', in which a splice donor splices to an upstream acceptor rather than a downstream acceptor 10  4. Finally, discussing the issue lacking from the current version that, in order to effectively act as a miRNA sponge, an RNA must not only be stable, but also in the proper cellular compartment and containing a molar concentration of miRNA binding sites that is high enough to sequester a biologically significant fraction of the endogenous miRNA of interest.
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 9 . The effectiveness of a ceRNA would depend on the number of miRNAs that it can ''absorb'' This, in turn, would depend on the ceRNA's accessibility to miRNA molecules, which is influenced by its subcellular localization and its interaction with RNA-binding proteins. Furthermore, the specific cellular context in which the ceRNA is expressed would also impact its overall influence because not all microRNAs are present ubiquitously and at all times 11 . We have included a discussion of the subject in the current version.