Genetic tools for the stable overexpression of circular RNAs

ABSTRACT Circular RNAs (circRNAs) are a class of non-coding RNAs featuring a covalently closed ring structure formed through backsplicing. circRNAs are broadly expressed and contribute to biological processes through a variety of functions. Standard gain-of-function and loss-of-function approaches to study gene functions have significant limitations when studying circRNAs. Overexpression studies in particular suffer from the lack of efficient genetic tools. While mammalian expression plasmids enable transient circRNA overexpression in cultured cells, most cell biological studies require long-term ectopic expression. Here we report the development and characterization of genetic tools enabling stable circRNA overexpression in vitro and in vivo. We demonstrated that circRNA expression constructs can be delivered to cultured cells via transposons, whereas lentiviral vectors have limited utility for the delivery of circRNA constructs due to viral RNA splicing in virus-producing cells. We further demonstrated ectopic circRNA expression in a hepatocellular carcinoma mouse model upon circRNA transposon delivery via hydrodynamic tail vein injection. Furthermore, we generated genetically engineered mice harbouring circRNA expression constructs. We demonstrated that this approach enables constitutive, global circRNA overexpression as well as inducible circRNA expression directed specifically to melanocytes in a melanoma mouse model. These tools expand the genetic toolkit available for the functional characterization of circRNAs.


Introduction
The last decade of RNA research revealed a myriad of noncoding RNAs that support fundamental biological processes and whose deregulation may contribute to the development of diseases such as cancer. Circular RNAs (circRNAs) are among the most recent additions to the non-coding RNA compendium and represent a peculiar class of RNA molecules. Generated from pre-mRNA via a process termed backsplicing where the 3' end of a downstream exon is spliced to the 5' end of an upstream exon, circRNAs form covalently closed ring molecules lacking 5' CAP and 3' polyA tail features. The circular structure protects also from exonuclease cleavage, rendering circRNAs extraordinarily stable. Advances in RNA sequencing led to the identification of thousands of circRNAs in humans and other species. Interestingly, circRNA expression is cell type-or differentiation state-dependent and often uncoupled from the expression of their cognate linear RNAs from which they derive [1][2][3][4][5][6][7][8][9][10]. This suggests that circRNA formation is a regulated process and that circRNAs must serve important functions in the cell. Indeed, circRNAs have been shown to act as natural microRNA sponges [3,11], associate with RNA-binding proteins [12][13][14][15][16] and chromatin [17], and regulate the transcription of their parental genes [18][19][20]. Some circRNAs even contain internal ribosome entry sitelike structures or N6-methyladenosine modifications to recruit the translation machinery and encode proteins [21][22][23][24][25][26][27]. Thus, circRNA functions are diverse and unpredictable based on their sequence alone, requiring functional characterization to reveal the underlying mode of action.
Functional analyses of circRNAs have been hindered by technical challenges. Knock-down experiments using RNAi require that the siRNA or shRNA is targeted to the backsplice junction to avoid simultaneous silencing of the cognate linear mRNA. This limits the possibilities for designing potent siRNA/shRNA and may lead to off-target effects on the cognate linear mRNA or other unrelated transcripts. Recently, a CRISPR approach using an engineered type VI-D Cas13d enzyme, CasRx, was developed to silence RNA transcripts [28]. Using an extended guide RNA placed over the backsplice junction, this CRISPR approach demonstrated superior circRNA silencing in terms of efficiency and specificity [29]. The overexpression of a circRNA of interest may be even more difficult than silencing. Circularization requires the presence of flanking introns containing repeat elements, which upon their alignment bring the appropriate splice sites into close proximity to aid in the backsplicing reaction. Mammalian expression plasmids have been generated where circRNA exons are flanked by introns derived from genes that efficiently produce circRNAs, for instance the D. melanogaster Laccase2 introns or the human ZKSCAN1 introns [30,31].
With these expression plasmids efficient ectopic circRNA expression can be achieved in mammalian or D. melanogaster cells [30]. However, such plasmids do not integrate into the genome and ectopic circRNA expression is therefore only temporary. This significantly limits long-term phenotypic analyses of cell biological effects. In addition, whether similar approaches are amenable to circRNA overexpression in genetically engineered mice for functional circRNA studies in vivo is not known.
Here, we tested stable, long-term circRNA overexpression approaches. While lentiviral circRNA delivery performed poorly, we found that using a transposon for delivery and genomic integration of circRNA expression constructs enables long-term circRNA expression in cells and a hepatocellular carcinoma mouse model. Moreover, we generated genetically engineered mice that overexpress a circRNA either constitutively and ubiquitously or inducibly in a cell-type specific manner in a melanoma model. These tools and approaches will enable more in-depth analyses of circRNA functions in vitro and in vivo.

Southern blot
A375 cells were transfected with 0.9 µg of spGFP transposon and 0.1 µg of pCDNA-mPB (piggyBac transposase) and selected with 10 µg/mL Blasticidin for 3 weeks. Clones derived from single cells were picked using cloning cylinders and expanded. Genomic DNA was isolated with Proteinase K lysis buffer and 15 µg of DNA were digested overnight with 40 units of MspI (New England Biolabs, Cat. # R0106). Digested DNA was separated on a 0.8% agarose gel, followed by in-gel depurination, denaturation, and neutralization. DNA was capillary transferred in 10xSSC onto nitrocellulose membrane and crosslinked by baking at 70°C for one hour. A Blasticidin probe was PCR amplified, labelled with α-32 P-dCTP using DECAprime II random prime labelling kit (Invitrogen, Cat. # AM1456) and purified with MicroSpin G-50 columns (Cytiva, Cat. # 27533001) using the manufacturers' recommendations. The membrane was incubated with labelled probe and 250 µg/mL salmon sperm DNA (Roche, Cat. # 10223638103) at 65°C overnight in Perfecthyb Plus Hybridization Buffer (Millipore, Cat. # H7033-125ML), washed in 2xSSC plus 0.5% SDS, and exposed to X-ray film for 4 days.

Mouse melanoma cell line generation
Melanoma cells were isolated from two tumours from a TRE-spGFP BPP chimera, expanded in vitro for one week, and transplanted subcutaneously into athymic nude mice (The Jackson Laboratory, J:NU Stock #007850), followed by the establishment of cell lines from the secondary tumours. Melanoma cell isolation was performed as previously described [32]. The CAGS-LSL-rtTA3 allele was unrecombined in the established cell lines and to induce rtTA3 expression, cell lines were transduced with Ad5CMVCre adenovirus purchased from the University of Iowa Viral Vector Core (https://vector-core.medicine.uiowa.edu/). To induce spGFP expression, cells were cultured in the presence of 0.5 µg/mL Doxycycline for 48 hours.

Statistical analysis
Statistical analysis was performed using the unpaired twotailed t-test. All in vitro experiments were performed with three biological and three technical replicates. Unless otherwise noted, the mean ± SEM of one representative experiment is shown. A p-value below 0.05 was considered statistically significant.

Limited utility of lentiviruses for circRNA overexpression
To identify efficient approaches that enable long-term, stable expression of ectopic circRNAs, we first tested the utility of lentiviruses. To this end, we cloned a split-GFP (spGFP) reporter into the pLenti-Blasticidin lentiviral vector under the control of the CMV promoter (spGFP-pLenti). The spGFP reporter [30] contains a split enhanced GFP cDNA and an internal ribosome entry site (IRES) which allows for the production of GFP protein only when the mRNA is backspliced and a circular GFP RNA is generated. Circularization is mediated by the introns flanking the human ZKSCAN1 circRNA (Fig. 1A). We transfected the spGFP-pLenti vector along with VSVG envelope and ∆8.2 packaging plasmids into HEK293T cells for lentivirus production. We also transfected a pLenti-Blasticidin version carrying a regular GFP cDNA whose mRNA is not circularized (GFP-pLenti, Fig. 1A) into HEK293T cells for side-by-side comparison of lentiviral vectors carrying constructs that are backspliced (spGFP) or not spliced (GFP). GFP-pLenti-transfected HEK293T cells expressed GFP protein (Fig. 1B), which is expected due to translation of the viral transcript. Notably, HEK293T cells transfected with spGFP-pLenti also were GFPpositive, albeit at lower levels (Fig. 1B). In addition, circular GFP RNA was detected in spGFP-pLenti-transfected HEK293T cells by RT-qPCR using divergent primers that specifically detect backspliced GFP RNA (Fig. 1C). Circular GFP was not detected in GFP-pLenti-transfected HEK293T cells, while using convergent primers that do not distinguish between circular and linear transcripts detected GFP RNA in cells transfected with spGFP-pLenti or GFP-pLenti (Fig. 1C). We also detected GFP protein in spGFP-pLenti-transfected HEK293T cells by Western blot (Fig. 1D). These results indicate that the spGFP-pLenti viral transcript is backspliced prior to packaging into viral capsids, which may interfere with efficient virus production. Alternatively, a spliced viral RNA lacking the circRNA could be packaged into viral capsids. This spliced viral RNA would consist of the circRNAflanking sequences connected via the branch point in the ZKSCAN1 5ʹintron and therefore would be 'kinked' and likely dysfunctional. To test this, we first estimated the viral titre by measuring viral RNA by standard curve RT-qPCR in the viral supernatant. The viral titre produced by spGFP-pLenti was comparable to that produced by GFP-pLenti (Fig. 1E), indicating that backsplicing does not impair the generation of particles containing viral RNA per se. We then transduced two human melanoma cell lines, WM164 and 501Mel, with the spGFP-pLenti and GFP-pLenti viral particles and selected the cells in 10 µg/mL Blasticidin for 10 days. Notably, despite the comparable viral titres, melanoma cells transduced with the spGFP-pLenti virus preparation died significantly more during Blasticidin selection than cells transduced with GFP-pLenti (Fig. 1F). WM164 and 501Mel cells transduced with spGFP-pLenti that survived Blasticidin selection were negative for GFP by Western blot (Fig. 1G) and fluorescence imaging (Fig. 1H). Moreover, we could not detect circularized spGFP by RT-qPCR using divergent primers and only very minor amounts of GFP using convergent primers that detect circularized and non-circularized GFP (Fig. 1I). Overall, these findings suggest that lentiviruses are not conducive to the generation of stable cell lines overexpressing circRNAs because of the undesired splicing of viral mRNA in virusproducing cells.

Stable circRNA overexpression using a transposon
Backsplicing of the viral transcript in virus-producing HEK293 cells may interfere with the production of viral particles carrying an intact, non-spliced circRNA transgene. We therefore decided to test a non-viral circRNA transgene delivery method using a transposon. We modified the ATP1 transposon [33] harbouring inverted repeats for the Sleeping Beauty and piggyBac transposases to contain the spGFP reporter under the control of the EF1α promoter and a Blasticidin selection cassette ( Fig. 2A). We transfected HEK293 cells with the spGFP transposon along with expression plasmids encoding the Sleeping Beauty or piggyBac transposases. Both combinations resulted in robust expression of circularized spGFP, as determined by RT-qPCR with divergent primers (Fig. 2B). In addition, we observed expression of GFP by Western blot in these cells (Fig. 2C). Next, we tested whether circular spGFP expression is maintained over multiple passages, which would indicate that copies of the spGFP transposon are integrated into the genome. To this end, we transfected HEK293 cells with the spGFP transposon and either Sleeping Beauty or piggyBac expression vectors or empty pCDNA3 and determined the expression of circular spGFP at 2 and 10 passages post-transfection. Both in the presence and absence of Blasticidin, circular spGFP expression declined after 10 passages (Fig. 2D). However, in the presence of Sleeping Beauty and especially piggyBac, spGFP levels were maintained at higher levels compared to the empty pCDNA3transfected control (Fig. 2D). We next tested whether a similar effect is observed in human melanoma cells by transfecting A375 cells with the spGFP transposon and either Sleeping Beauty or piggyBac expression vectors or empty pCDNA3. The transfected A375 cells were cultured for 10 passages in the absence of Blasticidin and then placed on Blasticidin selection. Significantly more cells survived selection when Sleeping Beauty or piggyBac were expressed (Fig. 2E), suggesting the transposon was maintained longterm in the absence of selection. Accordingly, the presence of piggyBac or Sleeping Beauty mediated increased expression of circular spGFP in Blasticidin-selected A375 cells (Fig. 2F). Interestingly, spGFP expression levels were stable when these A375 cells were maintained for another 10 passages in the presence of Blasticidin (Fig. 2F), demonstrating that longterm circRNA expression is stable.
To test whether transposon maintenance was due to genomic integration, we isolated six Blasticidin-selected single cell A375 clones that were transfected with the spGFP transposon and piggyBac expression plasmid and isolated genomic DNA. We digested the DNA with MspI to cut within the transposon and in juxtaposed genomic regions and subjected the DNA to Southern blot analysis using a Blasticidin probe. This strategy detected multiple bands of different sizes in most clones (Fig. 2G), validating that the transposon was integrated in the genome. We also assessed the levels of circular spGFP in these six A375 clones and observed that circular spGFP levels strongly correlated with the number of integrated transposons (Fig. 2H). To confirm that the transposon is amenable for delivering other circRNA transgenes, we replaced the spGFP construct with the exons of the human circDNAJC2 and circPMS1 circular RNAs. We delivered these circRNA transposon constructs along with piggyBac to WM164 cells and selected in Blasticidin. This resulted in significant overexpression of circDNAJC2 and circPMS1 as determined by RT-qPCR using divergent primers (Fig. 3A,B). To test if circRNAs can be delivered also to murine cells using the transposon approach, we cotransfected NIH3T3 fibroblasts with piggyBac and the spGFP transposon. Circular spGFP expression was detected in NIH3T3 cells co-transfected with piggyBac and the spGFP transposon compared to cells transfected with piggyBac alone, and the spGFP expression was further increased upon selection with Blasticidin (Fig. 3C). Finally, we compared spGFP expressed achieved with the transposon in HEK293 cells to spGFP expression from the pCDNA3 plasmid used for transient transfection.
While spGFP expression was significantly lower from the transposon, replacing the EF1α promoter with a CMV promoter increased spGFP expression approximately 15fold (Fig. 3D). Taken together, stable circRNA overexpression is readily achieved in human and murine cells using a transposon as delivery vehicle.

In vivo delivery of a circRNA construct using transposons
Following the establishment of a circRNA transposon expression system in cultured cells, we assessed whether the transposon approach is suitable for delivering circRNA overexpression constructs in vivo. Specifically, we examined whether a circRNA construct can be delivered to the liver of mice via hydrodynamic tail vein injection. To enrich for cells that contain the circRNA construct, we used a hepatocellular carcinoma model in which tumorigenesis is driven by delivering a Sleeping Beauty transposon harbouring a c-Myc cDNA [34]. We combined the spGFP transposon plasmid with a Sleeping Beauty transposase expression plasmid, the c-Myc transposon, and a plasmid encoding Cas9 and a sgRNA targeting p53 [34] and performed hydrodynamic tail vein injections in wildtype C57BL/6  (Fig. 4A). After ageing the mice for 4 weeks we isolated the livers and out of the 5 mice injected, 2 mice had developed hepatocellular carcinomas. We first isolated genomic DNA from 4 liver tumours and performed qPCRs using a primer pair designed over the Blasticidin cassette. This demonstrated that the spGFP transposon DNA was present in the tumours (Fig. 4B). Next, we isolated RNA from the same liver tumours and performed a RT-qPCR for circular spGFP using divergent primers. All tumours tested exhibited expression of circular spGFP, albeit at varying levels (Fig. 4C). These data demonstrate that circRNAs can be ectopically expressed in mouse liver by delivering circRNA transgene-containing transposons via hydrodynamic tail vein injection.

Global circRNA overexpression in genetically engineered mice
While the in vivo delivery of transposon DNA is readily achievable in livers, it is far more difficult to accomplish in other organs and cell types. To study gene functions in vivo, overexpression constructs are typically inserted into the mouse germline of genetically engineered mice. However, to our knowledge this approach has not been used for in vivo overexpression of circRNAs. Thus, we tested whether an ectopic circRNA knock-in allele is functional in mice. To this end, we cloned a targeting vector for recombination-mediated cassette exchange (RMCE) into embryonic stem cells (ESCs) harbouring a homing cassette (CHC) downstream of the collagen1a1 locus (Fig. 5A). This  targeting vector contains the spGFP circRNA reporter under the control of the constitutive EF1α promoter as well as a FRT site and a PGK promoter to render it compatible with RMCE (Fig. 5B). We targeted C10 v6.5 ESCs [35] with this spGFP expression construct and validated correct transgene integration by PCR (Fig. 5B). We then injected one positively targeted ESC clone into Balb/c blastocysts followed by transfer into pseudopregnant CD1 females. Three chimeras were born with ESC contribution based on coat-colour ranging from 25-80% (Fig. 5C). We euthanized the chimeras at 4 weeks of age and collected all major organs, isolated RNA, and performed RT-qPCR to detect spGFP expression with divergent primers. Notably, spGFP expression was detected in all organs (Fig. 5D). Thus, using a ubiquitous promoter and the human ZKSCAN1 introns enables expression of circRNAs in various organs of knock-in mice.

Tissue-specific and inducible circRNA expression in a mouse melanoma model
Having achieved global expression of the spGFP reporter circRNA in mice, we assessed whether ectopic circRNA expression can be induced and directed to a specific cell type in vivo. To this end, we used our ESC-genetically engineered mouse modelling (ESC-GEMM) melanoma platform [32] consisting of ESCs derived from GEMMs harbouring alleles to efficiently induce melanoma formation, integrate transgenes by RMCE, and control transgene expression with the Tet-ON system. Specifically, we used the BPP model, which harbours LSL-Braf V600E and Pten FL/FL driver alleles, a melanocyte-specific, 4OH-Tamoxifen (4OHT)-inducible Tyr-CreERt2 recombinase allele, the Cre-dependent CAGs-LSL-rtTA3 transactivator, and the CHC homing allele. When chimeric mice generated with the BPP model are topically treated with 4OHT on their back skin, melanomas form within 6 weeks. Moreover, expression of inducible transgenes is activated specifically in melanoma cells by switching chimeras to a Dox-containing diet. This model is thus ideal to rapidly test if ectopic circRNA expression can be directed to melanoma cells.
To generate an inducible spGFP circRNA reporter targeting vector, we replaced the EF1α promoter with a Doxycycline-inducible TRE promoter (Fig. 6A). We targeted BPP ESCs with the TRE-spGFP construct by RMCE (Fig. 6A) and produced chimeric mice via blastocyst injection (Fig. 6B). At weaning, we topically applied to the shaved back skin either several 1 µL drops of 25 mg/mL 4OHT in DMSO or enough 4OHT solution (2.5 mg/mL) to cover the back skin. After 5-6 weeks when melanomas had formed, we switched chimeras from a regular diet to a 200 mg/kg Dox diet for one week. We then harvested melanomas (Fig. 6C) and isolated RNA. As there is no selective pressure to recombine the CAGs-LSL-rtTA3 allele in contexts where the inserted transgene does not accelerate melanomagenesis [32], we first validated CAGs-LSL-rtTA3 recombination. To do so, we assessed expression of rtTA3 in melanomas by RT-qPCR. Of the 9 isolated tumours, 5 had the CAGs-LSL-rtTA3 allele recombined and expressed rtTA3, while 4 tumours did not (Fig. 6D). We then determined if circular spGFP was expressed in melanomas by RT-qPCR using divergent primers. Notably, circular spGFP was detected in Dox-treated tumours that also express rtTA3 (Fig. 6E) while circular spGFP expression was not detected in Dox-treated tumours that lacked rtTA3 expression or in livers (Fig. 6E). To further test the inducibility of circRNA expression, we established two melanoma cell lines from TRE-spGFP chimeras. When cultured in the presence of 0.5 µg/mL Dox, circular spGFP expression was induced in both cell lines (Fig. 6F). Thus, ectopic circRNA expression can be spatiotemporally controlled in vivo by using the appropriate Cre strain and the Tet-ON system.

Discussion
In this study we established new genetic tools for the stable, long-term overexpression of candidate circRNAs in cultured mammalian cells in vitro and in mice. We used a transposon approach to deliver a circRNA expression cassette to cell lines, where it is integrated into the genome upon transient coexpression of a transposase. We further show that circRNA expression cassettes can be delivered to the livers of mice via hydrodynamic tail vein injections. In addition, using a recombination-mediated cassette exchange approach, we inserted a circRNA expression cassette in the genome of mouse ESCs, enabling the generation of mice that ubiquitously express an ectopic circRNA. We used the same approach to generate a melanoma mouse model in which ectopic circRNA expression can be induced specifically in melanoma cells. These tools and approaches will pave the way for functional analyses of circRNA function and overcome some of the limitations of current circRNA expression systems. In particular, we are unaware of any attempts to overexpress circRNAs in vivo and our work demonstrates the feasibility of generating circRNA transgenic mice. This will stimulate further research of circRNA function in normal physiology and disease.
We observed that circRNA expression cassettes in lentiviral constructs are backspliced in the virus-producing cells. This may interfere with packaging the viral transcript into viral particles. Moreover, viral transcripts that have undergone backsplicing and are packaged into viral particles would lack the circRNA expression cassette. These considerations offer possible explanations as to why we observed a high viral titre with the spGFP-pLenti lentiviral construct but failed to obtain appreciable Blasticidin resistance or circular spGFP expression in melanoma cells transduced with the viral supernatant. The extent of backsplicing in virus-producing HEK293cells is likely dictated, at least in part, by the introns used to mediate circularization. The human ZKSCAN1 introns used in our construct efficiently promote circularization [30], which in this case is a detriment. The ZKSCAN1 introns could be replaced by introns that less efficiently mediate circularization to prevent backsplicing in HEK293 cells. However, this would also reduce the extent of circularization in transduced cells and increase the abundance of unspliced linear viral transcript containing the circRNA exons and flanking introns. It may therefore be challenging to discern whether any observed phenotype is elicited by the circRNA or the linear transcript. One could use RNAi or inhibitors targeting the splicing machinery to transiently reduce splicing in virusproducing cells. However, this may negatively affect HEK293 cells and thereby lower the quality of the produced virus. We found the transposon approach to be an easy-to-use alternative to lentiviruses for stable circRNA overexpression.
Given that the transposon is not delivered to target cells via an RNA intermediate, this approach is not hindered by the shortcomings of lentiviral circRNA delivery. However, as the transposon and transposase expression plasmids are cotransfected, this approach may be limited in difficult-totransfect cell types. Nevertheless, we observed long-term expression of the circular spGFP reporter in melanoma cell lines using two different transposases, Sleeping Beauty and piggyBac. Interestingly, we observed that circular spGFP expression decreased in the first 10 passages posttransfection. This is likely explained by the presence of donor plasmids that have retained the transposon. Such plasmids contribute to spGFP expression but are lost from the cells over time. The modular nature of the transposon allows for the easy exchange of the promoter, the circularizing introns, or the selection marker, allowing for the generation of custom-built transposons that fit one's need. For instance, we showed that replacing the EF1α promoter with a CMV promoter increased spGFP expression 15-fold. One may also wish to optimize the DNA delivery and transposon: transposase plasmid ratios to control for the number of integrated transposons per cell in a polyclonal cell population. As the transposon copy number correlates with the extent of ectopic circRNA expression, normalizing integration events would lead to more homogeneous expression levels. As an alternative, individual single cell clones may be selected that exhibit the desired transposon copy number and/or circRNA expression levels. Finally, to ensure that any observed phenotypes are not due to insertional mutagenesis but rather expression of the circRNA of interest, strategies to activate or inactivate ectopic circRNA expression using Tet-ON/OFF or Cre/loxP approaches could be used in combination with transposon-mediated delivery. Thus, the approaches and controls for generating stable cell lines with lenti-or retroviruses must also be applied when using transposons as a delivery method.
We successfully generated genetically engineered mice in which an ectopic circRNA was expressed constitutively in all cells or inducibly in melanocytes/melanoma cells. Given that ectopic circRNA expression was possible with both approaches, we surmise that most, if not all, gene targeting approaches to produce overexpression mouse strains are compatible with circRNA expression constructs. circRNA depletion via genetic knock-out, as has been done for CDR1as [36], is feasible and useful for transcripts that are primarily circularized like CDR1as [3,37]. However, most circRNAs are encoded by genes that also encode protein-coding mRNAs. The knock-out of circRNA exons in those genes will therefore also affect the encoded protein, which complicates the analysis of the observed phenotypes. Sophisticated methods that target the flanking introns to prevent circularization without affecting linear splicing could be developed. For instance, one could employ a CRISPR approach to mutate or delete AluI sites or other repeat sequences that mediate circularization. Similarly, binding sites of RNA-binding proteins that facilitate backsplicing could be mutated using the same approach. The efficiency of these approaches remains to be tested and is likely going to be locus dependent. Thus, circRNA overexpression in genetically engineered mice to study circRNA function in vivo is currently the more practical approach. In summary, the tools and approaches developed in this study expand the genetic toolkit to overexpress circRNAs in vitro and in vivo and will stimulate studies to functionally characterize circRNAs.