A cis-acting ligase ribozyme generates circular RNA in vitro for ectopic protein functioning

Abstract

Delivering synthetic protein-coding RNA bypassing the DNA stage for ectopic protein functioning is a novel therapeutic strategy. Joining the linear RNA head-to-tail covalently could be a state-of-the-art strategy for functioning longer. Here we enroll a cis-acting ligase ribozyme (RzL) to generate circular RNA (circRNA) in vitro for ectopic protein expression. The RNA circularization is confirmed by masking the 5’ phosphate group, resisting exonuclease RNase R digestion, failing for further tailing, and sequencing the RT-PCR products of the joined region. Interestingly, one internal ribosome entry site (IRES) renders circRNA translation competent, but two IRES in cis, not trans, hamper the translation. The circRNA with highly potent in translation is conferred for antiviral functioning. Accompanying specific guided RNA, a circRNA expressing ribonuclease Cas13 shows excellent potential against the corresponding RNA virus, further extending circRNA functioning in its growing list of applications.

Unlike conventional methods of generating circular RNA through RNA splicing, the authors here utilize RNA itself as a ligase ribozyme to directly link a protein-encoding RNA head-to-tail. This approach enhances the circular RNA’s antiviral potential compared to mRNAs.

Introduction

The success of introducing messenger RNA (mRNA) encoding SARS-CoV-2 spike protein as a novel vaccine strategy undoubtedly sparks the potential of RNA research in molecular therapy^1,2. Theoretically, the more half-life the delivered mRNA has in the cell, the longer functioning duration the mRNA has, and the fewer amounts of mRNA required to be delivered. Most mRNAs have 5’-cap and 3’ polyA in conjunction with various cellular proteins to reach each other and form a loop structure for efficient protein expression and avoiding exonuclease digestion^3,4. In response to elevating translation efficiency and longevity of the exogenous mRNA in cells, covalently circularizing single-stranded mRNA to generate the circular RNA (circRNA)⁵ became a new choice^6,7.

An internal ribosome entry site (IRES)-driven protein translation cassette could adapt translation initiation of the circRNA in lack of 5’-cap^7–9. circRNA could be generated in vitro by chemical^10,11 or enzymatic^11,12 RNA ligation with long-winded procedures. Alternatively, learning from the RNA self-splicing mechanism of group I intron, the permuted intron-exon (PIE) splicing became the popular circRNA formation strategy with fused partial exons flanked by intron and corresponding complementary sequences^13–16. When the RNA elements of the intron, splice site, and exon are permuted correctly, the RNA acts as an enzyme, the so-called ribozyme, catalyzes the splicing event to form circRNA via RNA breakdown and re-ligation. RNA splicing-based circularization strategies^15,17 tolerates linear RNA in large and can be performed in vitro and in vivo, but consumes quota in RNA sequence length for being the spliced waste. Despite a particular condition with GTP and Mg²⁺ accelerating PIE splicing in vitro⁶, small byproduct RNAs are matters of course after splicing required additional removal procedures to avoid the uncertainty of possible side effects^7,18.

Here we covalently circularized a single-stranded RNA by a cis-acting ligase ribozyme (RzL) selected from a systemic screening study¹⁹ that minimizes RNA sequences required for an RzL acting in trans. The RzL strategy enrolled a 5’-triphosphate (ppp)-bearing RzL to join the 3’-end resided corresponding substrate sequence directly^20,21 and autonomously during the in vitro transcription (IVT). Unlike the splicing method acquiring GTP to facilitate circRNA formation, the RzL strategy highly depends on the enzyme–substrate RNA pairing²² to generate circRNA without byproduct RNA. Despite not being well understood about the protein-encoded circRNA biophysical properties, we interrogated the protein translation properties of the bicistronic circRNA expressing two proteins, which provides important information for the experiments using circRNA for exogenous protein functioning and application.

Results

The PIE splicing strategy forms RNase R-resistant circRNA that RT-PCR could quantify

We began the circRNA study with the synthesized poR1 vector containing an IRES-driven protein expression cassette flanked by a sequence set of Anabaena catalytic intron/exon PIE system⁷ resided in a pair of homology arms under the control of T7 promoter (Supplementary Fig. 1a). The restriction enzyme SalI-linearized template contains all the elements for the PIE-mediated group I splicing to form circRNA that resists the exoribonuclease RNase R digestion (Fig. 1a, b; upper panels). In contrast, the XbaI-linearized template generates RNA lacking pairing elements and staying in linear form sensitive to RNase R (Fig. 1a, b; lower panels). Therefore, the RNA electrophoresis of samples that remained after RNase R digestion could tell the circRNA formation (Fig. 1b; right) among various RNA species derived from the SalI- but not XbaI-linearized templates (Fig. 1c). A designed RT-PCR method measuring circRNA with specific primers (Supplementary Fig. 1b) consistently showed the SalI but not XbaI group of RNA forms circRNA as expected (Fig. 1c). To improve the circularization efficiency, we heated the RNA in the presence of GTP (Fig. 1d, e; enhancement) for a better splicing activity generating circRNA as previously described¹⁴. Thus, circRNA production, qualification, and quantification by the conventional PIE splicing method were established.

Fig. 1. The conventional permuted intron-exon (PIE) method was used to help establish the ligase ribozyme (RzL)-catalyzed strategy in generating circular RNA (circRNA).

a, b Diagrams illustrated the RNA transcribed from the SalI- and XbaI-linearized DNA templates with elements or not for the PIE strategy (a). The precursor linear RNA enrolled the essential elements for the PIE splicing-mediated RNA circularization yields circRNA resistant to RNase R but not RNase A (b). homo homology arm, IRES internal ribosome entry site. c In vitro transcription (IVT) products from the indicated DNA templates were purified and then subjected to RNase A or RNase R digestion. RNA samples were separated by RNA electrophoresis and analyzed by RT-PCR using specific primer sets. The PCR products were revealed by agarose electrophoresis. White triangle, linear precursor; black triangle, circRNA product. Ctrl, internal control. d, e The IVT product from the SalI-linearized template was incubated with (+) or without (−) the splicing enhancement condition and then revealed by RNA electrophoresis (d). The significant band intensity was analyzed by ImageJ (e). White triangle, linear precursor; black triangle, circRNA product; gray triangle, cleaved byproducts. m.w., molecular weight. f A diagram illustrates how RNA flanked by RzL and substrate sequence (SUB) autonomously circularize the same RFP expression cassette in the IVT reaction mixture. g The IVT products from DNA templates with (+) or without (−) RzL were treated with RNase A or RNase R. Samples were revealed by the RNA electrophoresis and analyzed by RT-PCR. h The RNA products bearing RzL (+) or not (−) were analyzed by polyA-tailing assay and then sampled for the RNA electrophoresis. i, j circRNA from the indicated methods were analyzed by the polyA-tailing assay with a series of standard RNA mixtures containing the indicated % of circRNA. RNA samples were revealed by the RNA electrophoresis (i) or analyzed by quantitative RT-PCR (RT-qPCR) for circularization efficiency in percentage (j). Data are mean ± SD (n = 3, independent repeats/group) and were compared by two-tailed Student’s t test. enPIE the enhancement condition of PIE. k The designed ligation junction sequences of the RzL-catalyzed circRNA were illustrated (left). The ligation junction of the experimental products was RT-PCR amplified and sequenced (right). Source data are provided as a Source Data file.

A paired ligase ribozyme (RzL) and its substrate resided at the RNA 5’- and 3’-end autonomously catalyzes RNA circularization in vitro

Despite forming circRNA, the small species of RNA byproducts accompanying the splicing reaction were noted (Fig. 1d, e; gray triangle). We next sought to generate circRNA without the splicing byproducts. The synthesized poR2 vector contains the same protein translation cassette but is flanked by a pair of RzL and substrate sequences (Supplementary Fig. 2). We used PCR to generate a blunted-ended template to transcribe and end with the corresponding substrate RNA (SUB) fitting the RzL so that the linear RNA precursor could self-ligate directly to form circRNA (Fig. 1f). Despite indistinguishable RNA forms by electrophoresis, RNase R digested the RNA without RzL (Fig. 1g; lanes 1–3) but spared the RNA harboring both RzL and SUB (Fig. 1g; lanes 4–6), suggesting the presence of RzL and SUB in cis forms circRNA. Meanwhile, circRNA-specific RT-PCR consistently supported the conclusion from RNase R digestion (Fig. 1g; lower panel).

Because the RzL-mediated circularization seemed efficient, we enrolled the polyA polymerase. This template-independent RNA polymerase forms a polyA tail by consistently adding AMP to the linear RNA 3’ end (Supplementary Fig. 3) to trace the little uncircularized RNA precursors to visualize better and understand a high-efficiency RNA circularization. As expected, the RNA without RzL was polyA-tailed, migrated slower, and ultimately left to a higher molecule weight position in RNA electrophoresis (Fig. 1h; lanes 1 and 3), whereas most RNA containing RzL and SUB seemed unaffected (Fig. 1h; lanes 2 and 4). Using the linear RNA control (as 0% circRNA) and RNase R-treated circRNA (as 100% circRNA), a series of RNA dilution mixtures containing various percentages of circRNA could serve as the control standard for measuring the RNA circularization efficiency in percentage (Supplementary Fig. 4). With the control standard RNAs side-by-side, we found the RzL-mediated circularization efficiency was approximately 75-100% of the RNase R-treated (100%) control (Fig. 1i; lanes 1–3). In comparison, PIE-mediated efficiency was around 50% or more (Fig. 1i; lane 7), reaching more than 75% while enhanced by heat and GTP (Fig. 1i; lane 8). Because the RT-qPCR standard curve hardly regressed the circularization efficiency in percentage precisely (Fig. 1j) as a matter of manual operation, we turned to the polyA assay for the following evaluation of RNA circularization. Still, RT-PCR was used to assist in the analysis of circularization results and checked the RNA head-tail junction ligated as expected (Fig. 1k).

RNA modifications affect the RzL–SUB circRNA formation

The elements of RzL and SUB working in cis were analyzed (Fig. 2a, b) for influences of common RNA modifications on RzL-mediated circularization. To demonstrate the 5’-ppp group is the principle for the RzL-catalyzed circRNA formation that is fundamentally different from the PIE splicing method, we used m7(3’OMeG)(5’)ppp(5’)(2’OMeA)pG (Cap1 structure) or guanosine 5’-monophosphate (GMP) to reduce the free 5’-ppp and then withdrew its circularization ability. With equal amounts of the IVT products, we found that increasing the cap (Fig. 2c) or GMP (Fig. 2d) in the reaction mixture dramatically inhibited the RNA products from undergoing circularization, as evidenced by increased vulnerability to the RNase R, enhanced susceptibility to the polyA tailing, and decreased circRNA-specific RT-PCR products. Thus, RzL utilizes a mechanism independent of the PIE strategy to catalyze RNA self-circularization.

Fig. 2. Modifying RNA affects RzL-catalyzed circRNA formation.

The RzL-SUB pair was highlighted with dark circles to indicate possible m6A (a) or m5C (b) modifications. Note that the content of A in the pair is higher than C. The IVT reaction mixture contained a serial ratio of cap analog:GTP (0:1, 5:1, or 10:1) (c) or GMP:GTP (0:1, 1:1, or 3:1) (d). RNA products of the IVT were tested for RNase R resistance, polyA tailing susceptibility, and circRNA formation. 5’-cap the Cap1 analog m7(3’OMeG)(5’)ppp(5’)(2’OMeA)pG, GMP the Guanosine 5’-monophosphate. e Different ratios of N6-methyladenosine (m6A)- or 5-methylcytosine (m5C)-modified A or C were added in the IVT whose RNA products were analyzed for their RNase R resistance and polyA tailing susceptibility. f The RNA without (−) or with (+) RzL were treated with the indicated ratios of m5C-modified C during IVT, and then analyzed by RNase R and polyA tailing assay. g The RzL-SUB pair was highlighted with red circles to indicate the mutated nucleotides in the TG/GC mutant of RzL. h The prototype and TG/GC mutant constructs were sequenced and aligned with each other. Gray background, the mutated nucleotides. Proto prototype. i IVT products from the prototype RzL and TG/GC templates were analyzed by RNase R resistance, polyA tailing, and circRNA formation as indicated. Source data are provided as a Source Data file.

Since modifications in the RzL and SUB regions could likely affect the circularization and the contents of A and C were not equivalent within the region (Fig. 2a, b), we enrolled two common eukaryotic RNA modifications, N6-methyladenosine²³ (m6A) and 5-methylcytosine²⁴ (m5C) to understand whether modified RNA is tolerated in the RzL-mediated circRNA formation. As low as 1% A or C modification marginally affecting the circRNA formation, which was perfectly blocked by replacing all the A with m6A (Fig. 2e). Thus, the RNA with 100% m6A modification remained linear, migrated slower in the electrophoresis (Supplementary Fig. 5), and was inert in protein expression (Supplementary Fig. 6). In contrast, replacing all the C with m5C just marginally prohibited the circRNA formation (Fig. 2e; lane 5) despite a certain level of RNase R and end tailing resistance found in the m5C-modified linear control RNA (Fig. 2f; lane 2). Nevertheless, the m5C-modified circRNA remained incapable of protein expression (Supplementary Fig. 6), probably because the m5C hampers IRES structures required for translation or because of the m5C-mediated translation inhibition²⁵.

Besides modification, sequence replacement could also affect circularization. To test, the two nucleotides, T and G, which resided at the RzL catalytic core, were replaced with G and C¹⁹ (TG/GC mutant; Fig. 2g, h). Even though a very trace circRNA survived after RNase R digestion and was detected by circRNA-specific RT-PCR, the TG/GC mutant almost blocked the RNA ligation activity of the RzL (Fig. 2i). Moreover, while blunted ended PCR product as the IVT template, T7 RNA polymerase added an extraneous nontemplated nucleotide²⁶ at the very end of 3’, which would interfere with the RNA circularization (Supplementary Fig. 7). Therefore, as for the enzyme–substrate relationship, the corresponding SUB RNA sequences would be ideally better if they ended as designed.

A circRNA with two IRES hampers the protein expression of each other

To have a circRNA with multiple functions, we wondered whether circRNA could express two proteins. Three strategies were enrolled for the bicistronic expression of FLuc and GFP: fuse the two in a single protein (FG), separate the two by 2a-mediated ribosome skipping (F2aG and G2aF), or introduce another IRES between the two (FiresG and GiresF) (Fig. 3a). We found GFP fluorescence was decreased by fusion with FLuc (Fig. 3b; GFP vs. FG). Separating the FLuc from GFP with a 2a peptide, neither F2aG nor G2aF could rescue the fluorescence as the single GFP circRNA did (Fig. 3c; upper panel). Consistently, no matter whether FG, F2aG, or G2aF failed to reach the FLuc activity as high as the single FLuc circRNA (Fig. 3d). The consistent trends of protein expression level evaluated (Fig. 3e), suggesting translation efficiency more than the protein misfolding that affects GFP fluorescence (Supplementary Fig. 8) or FLuc enzyme activity, would be considered. Still, neither possibility should be excluded.

Fig. 3. An extraneous IRES in cis, not trans, hampered the circRNA translation.

a Mono- and bicistronic circRNA designs for expressing the indicated reporter proteins were illustrated with nomenclature. N18 cells were cotransfected with RLuc mRNA (100 ng) and the indicated circRNA (1 μg) for 24 h and harvested for fluorescence microscopy (b), flow cytometry (c), FLuc activity (normalized to that of RLuc activity, data are mean ± SD, n = 3, independent repeats/group) (d) and Western blotting (e). FLuc firefly luciferase, RLuc renilla luciferase. Source data are provided as a Source Data file.

We next checked the last bicistronic strategy of two identical IRES within a circRNA. Unexpectedly, the dual-IRES strategy rendered FLuc activity (Fig. 3d; FiresG and GiresF) and protein expression (Supplementary Fig. 9; lanes 7 and 8) also lower than the single FLuc circRNA. Consistently, flow cytometry (Fig. 3c; lower panel) and fluorescence microscopy (Fig. 3b) analysis of the GFP expression showed that it was not because of variations in transfection efficiency but because of different expression capabilities. To ascertain whether insufficient spacing²⁷ or self-competition of the two identical IRES shrinks circRNA translation capability, one of the FiresG’s CVB3 IRES was replaced with the EV71 IRES to have the dual cassettes circRNA ieFiresG, which was further added untranslated sequences to form another circRNA, namely ieFu-iresGu with more spacing between the two cassettes (Supplementary Fig. 10a). However, the GFP expression declined with ieFiresG and deteriorated with ieFu-iresGu (Supplementary Fig. 10b–d). Even though having two identical CVB3 IRES-driven GFP cassettes, the GiresG circRNA failed to have equivalent GFP expression as the single GFP circRNA (Fig. 3; GiresG). To overcome this, we cotransfected two separated circRNA in which either FLuc or GFP was driven by a single IRES (FLuc & GFP) and found they worked well for each other in the transfected cells (Fig. 3). Thus, the IRES-driven circRNA translation could be interfered with by another IRES in cis (Supplementary Fig. 11), not trans.

Highly potent translation efficiency could bestow circRNA antiviral potentials

To understand how competent the circRNA is in translation, we enrolled a recombinant JEV harboring GFP (JEV-GFP) whose 5’-capped RNA genome undergoes continuous replication and efficient translation. The NS3 viral protein together with the GFP viral reporter were readily detected at 24 h of the infection with or without circRNA transfection (Fig. 4a, b). Both Western blotting and fluorescence microscopy consistently showed marginal effects of viral infection (green) on circRNA translation (red) in the experimental model of infection–transfection and vice versa (Fig. 4a–c). Thus, the delivered circRNA remains capable of IRES-driven translation while the cell gets infected, which is the prerequisite against the virus successfully via expressing antiviral proteins.

Fig. 4. An IRES-driven circRNA expressing erCas13 could be antiviral against the recombinant RNA virus JEV-GFP.

N18 cells were infected with JEV-GFP virus (moi 10) for 18 h and followed with circRNA (1 μg) transfection as indicated (a), which was analyzed by Western blotting (b) and fluorescence microscopy (c). d, e N18 cells were infected with the JEV-GFP virus, followed by transfection of the circRNA harboring viperin (Vip) or FLuc for the indicated time (d). The GFP+ rate was then analyzed by flow cytometry (e). f A diagram illustrates the design of circRNA encodes an ER-targeting Cas13. g N18 cells were cotransfected with DNA plasmid encoding erGFP and circRNA encoding Flag-tagged Cas13 or erCas13 as indicated. Refer to Supplementary Fig. 12a for details. h–j N18 cells were treated with erCas13 RNA of either mRNA or circRNA form, infected with JEV-GFP, and followed by sgRNA treatment at the time and dose indicated (h). An RNA electrophoresis showed the mRNA and circRNA used (i). The N18 cells were sampled at p.i. 48 h (post erCas13 RNA transfection 66 h) for flow cytometry analysis (j). CDS coding sequence, + sgRNA targeting GFP, − N18 cellular RNA. Source data are provided as a Source Data file.

We next sought to use a circRNA overexpressing the known broad-spectrum antiviral protein viperin for virus inhibition^28,29 (Fig. 4d). The GFP positive rate reached approximately 76% at 48 h after JEV-GFP infection with low moi (Fig. 4e; left panel). Introducing control circRNA (FLuc) reduced a certain level of GFP, which was further attenuated by introducing the circRNA encoding viperin (Fig. 4e; right panel). Thus, the delivery of a natural antiviral protein by circRNA could effectively inhibit the virus.

To have a specific-targeting antiviral, we used circRNA to express the RNA-guided RNA nucleases Cas13³⁰ (Supplementary Fig. 12a; left panel) first and then introduced the GFP-targeting single guide RNA (sgRNA) (Supplementary Fig. 12b). While GFP and JEV NS3 were readily detected after JEV-GFP infection, a certain level of antiviral activity resulting from Cas13 circRNA delivery was noticed (Supplementary Fig. 12c; lane 2 vs. 3). Although Cas13 circRNA with sgRNA lowered the NS3 level more than the control RNA (Supplementary Fig. 12c; lane 3 vs. 4), the difference was marginal by blotting GFP. Regarding the orthoflavivirus RNA that may replicate and be concealed in the ER-associated membrane structures³¹, the Cas13 was further modified to target ER (erCas13) (Fig. 4f, g and Supplementary Fig. 12a; right panel). When the GFP(+) rate reached ~81% in the group without any RNA delivery (Supplementary Fig. 12d; left panel), delivery of the erCas13 circRNA alone showed a basal antiviral activity suppressing the virus infection to ~63% low of GFP(+) rate (Supplementary Fig. 12d; right panel). Applied the sgRNA under such circumstances, erCas13 could further suppress the infection toward as low as ~45% GFP(+), which worked better than the parental Cas13 for a functional antiviral activity against the JEV-GFP virus (Supplementary Fig. 12c, d), no matter whether erCas13 gets the right place or has the better protein expression level (Supplementary Fig. 13) to target viral RNA, or both.

Since RFP encoded in the forms of mRNA and circRNA got different protein expression kinetics (Supplementary Fig. 14), we asked whether erCas13 mRNA could work as well as its circRNA format under the same experimental condition (Fig. 4h). To prepare the erCas13 mRNA, RNA coding region sequences of the Flag-tagged erCas13 were 5’-capped and then 3’ polyA-tailed as the mRNA version (Fig. 4i). As the GFP(+) rate reached ~95% in the group without any RNA delivery, a certain level of antiviral activity consistently emerged after transfection of either erCas13 mRNA or circRNA by observing the GFP-positive rates ranging from 82.9 to 84.3 (Fig. 4j). However, erCas13 circRNA rather than mRNA could translate more erCas13 (Supplementary Fig. 15a) to assist sgRNA for a better antiviral activity (Supplementary Fig. 15a and the far right panel of Fig. 4j). Although the cap-dependant translation competition might exist between the viral RNA and erCas13 mRNA, the circRNA translated more erCas13 than the mRNA without virus infection (Supplementary Fig. 15b). Therefore, a long-lived and protein-translation highly-competent RNA in circular form would be the prerequisite for the sgRNA-guided, erCas13-dependent antiviral strategy.

Discussion

circRNA manufacturing has become an emerging technic focus because circRNAs have been implicated in various biological processes, diseases, and protein expression applications^32,33, but with limited methods to produce circRNA for relevant studies. Here we provided evidence showing that an RNA transcript could autonomously undergo covalent circularization in vitro by a cis-acting RzL, which was previously believed inappropriate for ligating relatively long RNA molecules^10,34. Because of the relationship as enzyme–substrate, the RzL-mediated circularization efficiency is undoubtedly affected by the RNA sequence integrity and probably could be influenced by the dynamic RNA structures complicated in proximity and spatial arrangement between RzL and SUB, despite it having been shown to act well by acting in trans^19,35. Even though being affected by RNA modifications, the cis-acting RzL utilizes its 5’-ppp group to catalyze circRNA formation directly without byproduct RNAs.

Both the PIE and RzL are strategies repurposed for producing circRNA. The PIE took advantage of the RNA splicing mechanism and reorganized exon-intron elements in order, which is well-established in circRNA production. In contrast, RzL emerged from the RNA world hypothesis³⁶ that asks RNA for RNA polymerase activity to ligate RNA building blocks in early life RNA replication. Besides supporting the RNA world hypothesis, the functional applications of RzL remained a few despite being synthesized in various forms and minimized¹⁹. Unlike PIE splicing, which reduces RNA residues to form circRNA that can be distinguished from yet-to-be-circularized precursors by electrophoresis, the novel strategy of RzL-catalyzed circularization cannot. Despite different limitations among the three methods evaluating circRNA production, polyA tailing seemed more suitable than RT-PCR to detect the uncircularized precursor in the RzL system. Although m5C could slightly affect RNase R digestion, the conventional method remained a good choice to remove linear precursors in circRNA production rather than quantification. However, without RNase R digestion, the uncircularized RNA 5’ end of RzL should leave a triphosphate group, which in turn became an advantage being the agonist activating host innate immunity³⁷ even if the translation was inert.

Despite RNA sequences and their modification could tell whether circRNA self or foreign, the exogenous circRNA was somehow detected by the cellular sensor RIG-I³⁸, which elicits IFN-induction signaling and, eventually, expression of functional proteins against viruses. Thus, circRNA encodes antiviral proteins that were not induced by circRNA or IFN, potentiating the native competence of circRNA against virus infection. That could be the reason why delivery of circRNA alone was antiviral, whereas the combination of sgRNA and circRNA encoded erCas13 showed the most therapeutic potential here against the virus in a cell model. Alternatively, since multicistronic circRNA seemed impossible, a multivalent pool composed of various circRNA encoding different antiviral proteins could be another choice. The more types of antiviral proteins in the collection, the better and broader antiviral effects the circRNA pool should have, whereas adjusting the proportion of particular antiviral protein composition might specialize the circRNA pool against specific viruses. However, how to deliver the circRNA together with sgRNA specifically to the infection site in an organism remains a great challenge before its application.

For the circRNA encodes protein to function, an IRES is required. The viral³⁹ or cellular IRES⁴⁰ could be defined as a specialized RNA sequence allowing for cap-independent translation initiation, but it is challenging to prove. A conventional strategy uses two reporter genes placed in a single mRNA separated by IRES, so the first reporter is translated through a cap-dependent mechanism, while IRES drives the second translation^41,42. However, the possibility of translational readthrough effect, saying that ribosomes continue translation past the stop codon of the first reporter, became the disadvantage of this approach. Since that, the exonuclease-resistant and IRES-required circRNA could be a more sensitive and better platform than linear RNA in defining IRES sequences. Interestingly, the two identical IRES resided in the same circRNA, deteriorating the translation potential of each other was unexpected. That might result from complex sequence pairing or interactions between the two. Even though an RNA element may facilitate translation in one or certain cases (e.g., single IRES), any RNA sequences could be translationally inhibitory by interacting with RNA elements elsewhere, both proximal and distal, in the same circRNA molecule. The spacer, UTR, and IRES sequences that affect translation would most likely be case-by-case and highly depend on the overall RNA designs. Alternatively, more than one entry site at circRNA might render ribosomes failing to keep in order getting in the translation cycle, like traffic jams by getting onto the freeway at multiple intersections without traffic lights. Despite the inert translation mechanism of the two-IRES circRNA waiting to be uncovered, the circRNA strategy was undoubtedly a better-than-ever tool to interrogate translation competition among multiple IRES both in trans and cis in the future.

Beyond protein expression for functioning, circRNA could be non-coding⁴³, bearing a specific RNA sequence or structure for functions by interacting with particular cellular factors, including DNA, RNA, or proteins⁴³. No matter whether coding or non-coding, circRNAs have emerged as essential players in gene regulation and many aspects of applications, especially RNA-mediated gene therapy⁴⁴ and RNA vaccine development⁴⁵, making them a topic of great interest and research focus. The cis-acting ribozyme thus provided a new and easy-to-handle choice circularizing RNA precisely without splicing-related RNA byproducts that could be applied to various scientific fields broadly.

Methods

Plasmids

The poR1 and poR2 vectors, both containing a coxsackievirus B3 IRES (Accession No. M33854)-driven protein expression cassette, genes of Cas13³⁰, erCas13, and sgRNA targeting GFP were all synthesized by Protech Technology, Taipei, Taiwan. The poR1 comprises a set of PIE sequences from Anabaena catalytic intron/exon⁷, whereas the poR2 series harbors a pair of RzL and substrate sequences¹⁹. The initial reporter synthesized in vectors was RFP, which was then replaced, as mentioned in need, with the GFP, FLuc, Viperin, Cas13, or erCas13 by general gene cloning methods with enzymes from New England Biolabs (NEB). Primer sets used in PCR amplification (Cat#71086, Merk) of the genes of interest for each construct were listed in Supplementary Table 1. Protech Technology also synthesized the poR2 vector containing GiresF, ieFiresG, and ieFu-iresGu, the latter two of which harbored another IRES from human enterovirus 71 (Accession No. U22521). All the constructs were confirmed by restriction enzyme digestion and DNA sequencing.

Template preparation for in vitro transcription (IVT)

The poR1 vector was linearized by XbaI or SalI for the RNA products in linear or circular. For the poR2 series, linear DNA templates were all obtained by PCR with the primer sets provided in Supplementary Table 2. To suppress the heterogeneity at the 3’ ends of the IVT products as previously described⁴⁶, the two 5’-terminal nucleotides of the reverse primer were 2’-O-methylated (customized service, Genomics). All enzyme-linearized and PCR-amplified DNA products were purified as templates by either ethanol precipitation or a Gel Extraction Kit (Cat#28706, QIAGEN).

RNase digestion and polyA tailing

Purified RNA was buffered with RNase R (final conc.: 2 U/μg RNA, Cat#ab286929, abcam) or 2 mg/ml RNase A (Cat#12091021, Invitrogen) for a 1 h incubation at 37 °C. For the 3’ end tailing of polyA, equal amounts of purified RNA were buffered in 1X reaction buffer with 1 mM ATP and 5 U of E. coli PolyA Polymerase (Cat#M0276S, NEB) in a 20 μl reaction. The tailing was carried out at 37 °C for 1 h and then stopped by adding EDTA to a final concentration of 10 mM. RNA digestion and tailing quality were checked by separating samples on 1% denaturing formaldehyde agarose gel.

RNA gel electrophoresis

RNA was separated on 1% formaldehyde agarose gel containing 1X MOPS buffer (20 mM MOPS, 2 mM Sodium acetate, 1 mM EDTA; Cat#GB33, ZGENEBIO) and 2 M formaldehyde (Cat#F1635, Sigma). RNA samples (100–200 ng) and markers (200 ng, Cat#N0362S, NEB) were denatured by mixing with an equal volume of RNA loading dye (Cat#B0363S, NEB), heating at 70 °C for 5 min, and then placed on ice for 3 min. Electrophoresis was performed at 100 V for 40 min in the 1X MOPS running buffer. The formaldehyde agarose gel was soaked in water for 15 min to remove the formaldehyde, allowing RNA to renature for staining. Agarose gel was then post-stained with GelRed Nucleic Acid Gel Stain (1:3300 dilution in water, Cat#41003, Biotium) for 30 min. The image was taken by VILBER with Quantum CX5 Gel Imaging System.

Synthesis of circRNA and linear mRNA

RNA from IVT was synthesized by RiboMAX™ Large Scale RNA Production Systems (Cat#P1300, Promega) with linearized templates aforementioned at 37 °C for 1.5 h. To reduce 5’ triphosphate-bearing RNA transcripts, the transcription was performed with m7(3’OMeG)(5’)ppp(5’)(2’OMeA)pG Cap analog (CleanCap AG) (Cat#N-7413, TriLink BioTechnology) or guanosine 5’-monophosphate (GMP) (Cat#G8377, Sigma). For base-modified RNA synthesis, N6-Methyladenosine-5’-Triphosphate (m6A) (Cat#N-1013, TriLink BioTechnology) and 5-Methylcytidine-5’-Triphosphate (m5C) (Cat#N-1014, TriLink BioTechnology) were added at mentioned ratios (1% and 100%). After IVT, reactions were treated with DNase I (Cat#M0303S, NEB) at 37 °C for 30 min and were then column-purified by a MEGAclear Transcription Clean-up kit (Cat#AM1908, Thermo Fisher Scientific). For the enhancement condition of the PIE splicing, purified RNA was buffered at 1X T4 RNA Ligase Reaction buffer (Cat#B0216L, NEB), adding additional GTP to a final concentration of 2 mM for another 15 min at 55 °C. To remove the uncircularized RNA, 100 μg of column-purified RNA and 10 μl of 10X RNase R buffer (Cat#ab286929, abcam) were diluted in DEPC-treated water, and then incubated with 20 U RNase R (Cat#ab286929, abcam) for 1 h at 37 °C. For linear mRNA, the 5’ capped IVT products (performed with the CleanCap AG) were further polyA-tailed following the stop codon by E. coli PolyA Polymerase (Cat#M0276S, NEB) at 37 °C for 1 h. The fully processed mRNA were column purified. RNAs were checked by 1% denaturing formaldehyde agarose gel or by capillary electrophoresis (Supplementary Fig. 16).

circRNA measurement

For the PCR-based measurement of circRNA, sequences of the primers are provided elsewhere as Supplementary Table 3. Briefly, the RT primer-primed cDNA was reverse transcribed from a 100 ng RNA sample by a SuperScript III RT kit (Cat#18080-044, Invitrogen). PCR was carried out with the KOD DNA polymerase (Cat#71086, Merck) by the specific primer sets: F1 and R1 were for the circRNA, F2 and R1 were for the Ctrl. The PCR products were then subjected to 2% agarose TBE gel. To quantitate the circularization efficiency, qPCR was performed using LightCycler 480 instrument (Roche) and the Fast SYBR Green Master Mix (Cat#4312704, Applied Biosystem) according to the manufacturer’s instructions with the same primer sets as PCR. For the end-tailing-based circularization assessment, 1.25 μg RNA samples were used for 1 h of the polyA tailing assay aforementioned. Examples of the assay evaluating RzL-mediated circularization of RNA with various lengths can be found elsewhere (Supplementary Fig. 17). In Fig. 2e, f, i, the tailing reaction was 2 h.

Viruses and cell lines

JEV-GFP from Dr. Yi-Ling Lin (Institute of Biomedical Sciences, Academia Sinica) were propagated in C6/36 mosquito cells (ATCC, CRL-1660) and quantified by plaque-forming assay using hamster kidney BHK-21 cells (ATCC, CCL-10). C6/36 cells were grown in RPMI (Cat#SH30027.02, Hyclone) containing 10% fetal bovine serum (FBS) and were cultured at 28 °C and 5% CO₂. BHK-21, and murine neuroblastoma N18 cells were cultured at 37 °C and 5% CO₂ in RPMI (Hyclone) supplemented with 5% FBS.

RNA prediction and transfection

RNA structure was predicted by using RNAfold (http://rna.tbi.univie.ac.at/)⁴⁷. For RNA transfection of cells seeded in a 12-well plate, 1 μg/well circRNAs or mRNAs were transfected with Lipofectamine MessengerMAX (Cat#LMRNA015, Thermo Fisher Scientific) according to the manufacturer’s protocol.

Reporter assay

Cells were contransfected with 1 μg of indicated circRNAs and 100 ng of RLuc mRNA for 24 h. Cell lysates were harvested and analyzed with the Dual-Luciferase Assay System (E1960, Promega) using the GLOMAX Multi+ Microplate Multimode Reader (Promega). Relative luciferase activities were normalized with the internal control renilla luciferase.

Fluorescence assay

N18 cells were treated with circRNA or JEV-GFP at the time and with the dose indicated. The expression of RFP and GFP was observed by fluorescence microscopy (Olympus IX73). For immunofluorescence assay, N18 cells were cotransfected with 0.5 μg plasmid DNA encoding erGFP and 0.5 μg circRNA expressing either Cas13 or erCas13 using lipofectamine 2000 (Cat#11668500, Thermo Fisher Scientific). After 24 h post-transfection, N18 cells were fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS; Cat#BF203-5L, Protech Technology) for 30 min at room temperature, then permeabilized in PBS containing 0.5% Triton X-100 for 10 min. After blocking with skim milk in PBS for 30 min, cells were probed with primary antibodies diluted in skim milk in PBS overnight, then incubated with corresponding secondary antibodies for another hour at room temperature, followed by nuclear DAPI counterstaining (0.25 ng/ml, 7 min). Cells were photographed under a confocal fluorescence microscope (FV3000, Olympus) with a 100X objective.

Flow cytometry

Cells were washed with warm PBS, detached by PBS-EDTA at 37 °C for 10 min, and resuspended in FACS buffer (PBS, 0.1% NaN3, and 2% FBS). The cell suspension was centrifuged at 200 xg for 5 min at 4 °C and washed with FACS buffer two times. Samples were resuspended in the FACS buffer with appropriate volume and then analyzed using CytoFLEX (Beckman Coulter) and FlowJo VX software.

Western blot analysis

Cells were lysed with RIPA buffer (10 mM Tris, pH 7.5, 5 mM EDTA, 150 mM NaCl, 0.1% SDS, 1% Triton X-100, 1% sodium deoxycholate) containing a cocktail of protease inhibitors and phosphatase inhibitors (Cat#04693116001 and #04906837001, Roche). Samples were prepared by 5X loading dye and a 5 min heat-denature at 95 °C was followed if needed. Equivalent amounts of proteins determined by the DC Protein Assay Kit (Cat#5000113, #5000114, and #5000115, Bio-Rad) were separated on SDS-PAGE and transferred to a nitrocellulose membrane (Cat#10600003, Amersham). Non-specific antibody binding sites were blocked with skim milk in PBS with 0.1% Tween 20 (PBS-T), then reacted with primary antibodies against FLuc (1:1000; Cat#GTX125848, GeneTex), RLuc (1:1000; Cat#GTX125851, GeneTex), RFP (1:2000; Cat#GTX636956, GeneTex), GFP (1:2000; Cat#GTX26556, GeneTex), NS3 (1:2000; GTX125868, GeneTex), Flag (1:2000; Cat#F1804, Sigma), or actin (1:10,000; Cat#NB600-501, NOVUS). After incubation with appropriate horseradish peroxidase-conjugated secondary antibodies (1:2500; Cat#115-035-146 and #115-035-144, Jackson ImmunoResearch), signals were revealed by Chemiluminescence HRP Substrate (Cat#WBKLS0500, Millipore) and detected by UVP (ChemiDoc-It Imaging system, Analytik, Jena, Germany).

Statistics and reproducibility

Data are shown as mean ± SD. A two-tailed Student’s t test was used for comparing two groups as described in figure legends. p < 0.05 was considered statistically significant. For Figs. 1c, d, g–i, 2c–f, i, 3b, e and 4b, c, g, i, the experiments were independently repeated at least three times with similar results.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

C.I.S., Z.S.C., C.T.S., H.I.W. and Y.T.K. performed the experiments. C.Y.Y. and C.I.S. designed the circRNA constructs. C.Y.Y., C.I.S. and Z.S.C. performed data analysis. C.Y.Y. and C.I.S. wrote the initial manuscript draft. All authors contributed to the final manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

The data supporting the findings of this study are available from the corresponding authors upon request. Source data for the figures and Supplementary Figs. are provided as a Source Data file. Source data are provided with this paper.

Competing interests

National Health Research Institutes has filed patent applications based on this work in which C.Y.Y. is named as the inventor. The remaining authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-024-51044-y.