Invention of circRNA promoting RNA to specifically promote circRNA production

Abstract

CircRNA, an essential RNA molecule involved in various biological functions and diseases, often exhibits decreased expression in tumor tissues, playing a role as a tumor suppressor, and suggesting therapeutic potential for cancer. However, current methods for promoting circRNA production are limited. This study introduces a novel approach for enhancing circRNA biogenesis, termed circRNA promoting RNA (cpRNA). CpRNA is designed to complement the flanking sequences of reverse complementary matches (RCMs) within pre-mRNA, thereby facilitating circRNA formation through improved exon circularization. Using a split-GFP reporter system, we demonstrated that cpRNA significantly enhance circGFP production. Optimization identified the best conditions for cpRNA to promote circRNA biogenesis, and these cpRNAs were then used to augment the production of endogenous circRNAs. These results indicate that cpRNAs can specifically increase the production of endogenous circRNAs with RCMs, such as circZKSCAN1 and circSMARCA5 in cancer cells, thereby inhibiting cell proliferation and migration by modulating circRNA-related pathways, showcasing the therapeutic potential of cpRNAs. Mechanistic studies have also shown that cpRNA promotes circRNA biogenesis, in part, by antagonizing the unwinding function of DHX9. Overall, these findings suggest that cpRNA represents a promising strategy for circRNA overexpression, offering a potential treatment for diseases marked by low circRNA levels.

Graphical Abstract

Introduction

Circular RNA (circRNA) is an endogenous RNA molecule that forms a closed-loop structure. One distinguishing feature of circRNA is its lack of a 5′ cap and a 3′ poly-A tail, which makes it highly resistant to nucleases, resulting in a longer lifespan compared to other RNA molecules (1). This unique characteristic positions circRNA as a promising biomarker or therapeutic target due to its enhanced stability (2). Numerous biological functions of circRNA have been discovered thus far, including its role as a sponge for microRNAs (miRNAs), its ability to modulate transcription or splicing, its interaction with proteins, and even its potential to encode proteins (3–5). The diverse functions of circRNAs contribute to the development of various diseases, such as cancers, neurodegenerative diseases, and other ailments (6,7). In the context of cancer, extensive research has confirmed the dysregulated expression of circRNAs, underlining their critical involvement in tumor initiation, progression, invasion, and metastasis (8). Hence, manipulating anticancer circRNAs holds great promise as a novel gene therapy strategy for human cancer.

Recent studies have revealed a significant reduction in circRNA levels in tumor tissues compared to normal tissues, with even lower expression observed in immortalized cancer cell lines. Moreover, there is a negative correlation between circRNA expression and cancer cell proliferation (9). Several circRNAs have been identified to be markedly downregulated in various types of tumors, highlighting their substantial inhibitory effects on tumorigenesis and metastasis (10), including circSMARCA5 (hsa_circ_0 001 445) (11), circZKSCAN1 (hsa_circ_0 001 727) (12), circMTO1 (hsa_circ_0 007 874) (13) and others. Therefore, how to augment the expression of these anticancer circRNAs is a clinically significant issue that requires urgent attention for resolution. However, the current methods available for promoting circRNA production are still limited. The most common method is constructing vectors that contain exon sequences to overexpress circRNA. These exon-containing vectors are transfected into cells, which then produce circRNA through cellular transcription, similar to the overexpression of linear RNAs (14). However, this approach inevitably leads to the production of linear RNA isoforms (15). Moreover, circRNAs with a large number of nucleotides exhibit low transfection efficiency, resulting in poor overexpression effectiveness (16). Another strategy involves replacing weak promoters with strong ones using CRISPR genome-editing tools, which can promote circRNA production. However, the off-target effects of genome-editing still affect the accuracy of experimental results (17,18). Furthermore, circRNAs can also be synthesized in vitro through extracellular ligation (19). However, if the sequence of circRNA is excessively long, it significantly increases the synthesis difficulty and costs. Additionally, artificially synthesized circRNA produced in vitro may potentially induce significant immune reactions (18), thereby limiting its application. Therefore, the exploration for novel approaches to circRNA biogenesis is imminent.

Unlike the classical splicing of linear RNA, circRNA is generated through back-splicing, which covalently links the 5′ and 3′ ends of the exons. During this process, various mechanisms are involved, including lariat-driven circularization and intron-pairing-driven circularization (3). In the mechanism of intron-pairing-driven circularization, reverse complementary matches (RCMs) are formed to bring the splicing sites closer together, facilitating the formation of circRNA (20). According to Liang's research, as few as 30–40 nucleotides of RCMs could effectively promote exon circularization (21). Furthermore, disrupting the RCMs has been demonstrated to be an efficacious strategy for circRNA knockout (22). Several RNA binding proteins that interact with RCMs also play a role in regulating the expression of circRNAs (23,24). For instance, DEAD-box helicases, particularly DHX9, have been found to bind specifically to Alu repeats within RCMs. In the absence of DHX9, there is an increased accumulation of circRNAs (23). Additionally, nuclear factor90/110 (NF90/NF110) has been shown to accumulate on RCMs and enhance their stability, thereby promoting the formation of circRNA (24). Considering these findings, it is intriguing to explore the possibility of influencing circRNA circulation by regulating the intron pairing process. This could potentially have an impact on circRNA expression and offer therapeutic opportunities for diseases associated with circRNAs.

In this study, we utilized a specific RNA molecule known as circRNA promoting RNA (cpRNA), which is designed to be complementary to the outer sequence of the RCMs. The purpose of cpRNA is to bring the ends of the circRNA precursor sequence closer together and enhance the stable pairing of the flanking introns of the circRNA. Through this ingenious approach, we had achieved a significant improvement in the generation of the target circRNA. Remarkably, we applied cpRNAs to specifically promote the back-splicing of both split-GFP reporters and endogenous circRNAs. This significant finding highlights a novel and effective strategy to enhance circRNA production. Such advancements in circRNA production not only contribute to a comprehensive understanding of the functions of diverse circRNAs, but also hold tremendous potential for the development of therapeutic interventions for circRNA-associated diseases.

Materials and methods

Design of circRNA promoting RNAs and plasmids

An IRES-derived split-GFP plasmid was used as a fluorescent reporter to mimic the back-splicing process of endogenous circRNA with reverse complementary matches (RCMs). In the reporter system, GFP is divided into two fragments, which undergo back-splicing to generate green fluorescent protein. Meanwhile, intron1 and intron2 are inserted beside the GFP fragments to mimic the RCMs (Supplementary Figure S1A). To facilitate back-splicing of split-GFP, a cpRNA was designed based on the intronic sequences besides RCMs. The most optimal cpRNA sequence was inserted into a pcDNA3.1/Zeo(+) vector to generate cpRNA-expressing plasmids (Supplementary Figure S1B). The enzyme cleavage sites for cpRNA insertion are NheI and HindIII. In order to validate the effects on endogenous circRNAs, we selected four downregulated circRNAs in HCC, including circSMARCA5, circZKSCAN1, circKCNN2, as well as circLIFR. Different cpRNAs were individually designed based on their RCMs, and the sequences were inserted into the pcDNA3.1/Zeo(+) vector between the NheI and HindIII restriction sites. All the cpRNA sequences were listed in Table 1 and were synthesized by Tsingke Biotechnology Co., Ltd (Beijing, China). The DHX9 mRNA sequence (NM_001357.4) was inserted into a pCMV3-N-HA vector between KpnI and NotI constructed by Sino Biological (Beijing, China) to construct a DHX9-overexpressing (OE) plasmid (Supplementary Figure S1C). The seed sequences in the 3′-UTR (DLC1 3′-UTR: 5′-ATTTTAAAGCTGCTTCCTGT-3′; TIMP3 3′-UTR: 5′-GAATTTTATATTCCGTGAATGTA-3′) were inserted into the pGL3-Promoter vector, which had been double-digested with XbaI and FseI, to construct the luciferase target reporter plasmids (Supplementary Figure S1D). The large-scale plasmids were extracted using the PureYield™ Plasmid Midiprep System (Promega, Madison, WI), while the small-scale plasmids were extracted using the Mini DNA Purification Kit (Dingguo, Beijing, China).

Table 1.

Sequences of probes, primer sets, and cpRNAs used in this study

Probe name	Sequences (5′ to 3′)
Biotin-cpGFP probe	Biotin-UUGGUCUUUAUUGUUUUCCAACACCGGUAGCGCUAGCGGAUC
Biotin-circSMARCA5 probe	Biotin-GATTCTGATCCACAAGCCTCCCTTTTGTTTTTCTCTATAGT
Biotin-circZKSCAN1 probe	Biotin-TGTTTCTTTATCTATTCCTCGTGACTGTAAGAGGC
Primer name	Sequences (5′ to 3′)
hsa-miR-181b-5p	AACATTCATTGCTGTCGGTGGGT
hsa-miR-873-5p	GCAGGAACTTGTGAGTCTCCT
U6	AGAAGATTAGCATGGCCCCT
Universal miRNA Reverse Primer	miRNA First Strand cDNA Synthesis (Tailing Reaction) (Sangon Biotech)
circKCNN2	F: CTGTCCGAGCTTGTGAAAGT
	R: TGGTTGTGGATGGGGCATAG
circLIFR	F: ACACCGCTCAAATGTTATCTGG
	R: CAGGATGGTCGTTTCAAACATAC
circSMARCA5	F: TGGGCGAAAGTTCACTTAGAA
	R: TCTTTGCACCTCTTTCCAAAA
circZKSCAN1	F: TCATGGACCTGAGATGCTCG
	R: GCTTCCCGTGATTCAGCAGT
circGFP	F: TCGTGACCACCCTGACCTAC
	R: GTCCTCCTTGAAGTCGATGC
FP-G primer pair	F: AGAAGAACGGCATCAAGGTG
	R: CTGGGTGCTCAGGTAGTGG
left primer pair	F: GCTACCGGTGACTGAACATG
	R: TGCCCTCAGACTTCTTTCCT
right primer pair	F: CTCTGCAGCCCTTTATCCTG
	R: AGAAGAGGCCCAATTCAAGGA
DLC1	F: CACAGGACAACCGTTGCCTCAG
	R: CTCTTCAGGGTGTTGAGATGGA
TIMP3	F: CTGCAAGGGCTGGGCATC
	R: TCCATGGCCCGGTTGGCAGTGTGGAG
β-actin	F: CTCCATCCTGGCCTCGCT
	R: GCTGTCACCTTCACCGTTCC
cpRNA Name	Sequences (5′ to 3′)
S1	UUGGUCUUUAUUGUUUUCCAA
S2	CACCGGUAGCGCUAGCGGAUC
M0	UUGGUCUUUAUUGUUUUCCAACACCGGUAGCGCUAGCGGAUC
M2	UUGGUCUUUAUUGUUUUCCAAAACACCGGUAGCGCUAGCGGAUC
M6	UUGGUCUUUAUUGUUUUCCAAAAAAAACACCGGUAGCGCUAGCGGAUC
B18	GUCUUUAUUGUUUUCCAACACCGGUAGCGCUAGCGG
B21	UUGGUCUUUAUUGUUUUCCAACACCGGUAGCGCUAGCGGAUC
B24	GAUUUGGUCUUUAUUGUUUUCCAACACCGGUAGCGCUAGCGGAUCUGA
D1	UUGGUCUUUAUUGUUUUCCAACACCGGUAGCGCUAGCGGAUC
D5	GGAUUUGGUCUUUAUUGUUUUGGUAGCGCUAGCGGAUCUGAC
D10	UUCAAGGAUUUGGUCUUUAUUCGCUAGCGGAUCUGACGGUCC
D20	AGAGGCCCAAUUCAAGGAUUUUCUGACGGUUCACUAAACCCU
D50	GGUAUGGCUGAUUAUGAUCAGGGAGUGGCGGGCCUCGGCGUG
D100	AGGGGGAGGUGUGGGAGGUUUCGUUGUGGUUGGUGAUGUCCA
cpSMARCA5-0A	CCAAAUUACUAUUCAUCCCAU GUGAAAGCUGAUAUGAAAAAA
cpSMARCA5-2A	CCAAAUUACUAUUCAUCCCAU AA GUGAAAGCUGAUAUGAAAAAA
cpSMARCA5-6A	CCAAAUUACUAUUCAUCCCAU AAAAAAAA GUGAAAGCUGAUAUGAAAAAA
cpSMARCA5-18nt	AAUUACUAUUCAUCCCAUGUGAAAGCUGAUAUGAAA
cpSMARCA5-24nt	AUCCCAAAUUACUAUUCAUCCCAUGUGAAAGCUGAUAUGAAAAAAUUU
cpSMARCA5-gap 10nt	GUUACCAUCCCAAAUUACUAUGCAUAUGAAAAAUUUUCAAAA
cpSMARCA5-gap 100nt	AUCUCUAUAUUACCAAGUAUUUAAUACUCCCCAAAUCAGCUU
cpZKSCAN1-0A	CUUUUGAGACAGGGUCCUUGC AAGUUAAACUCCGUCAAAAAA
cpZKSCAN1-2A	CUUUUGAGACAGGGUCCUUGC AA AAGUUAAACUCCGUCAAAAAA
cpZKSCAN1-6A	CUUUUGAGACAGGGUCCUUGC AAAAAA AAGUUAAACUCCGUCAAAAAA
cpZKSCAN1-18nt	UUGAGACAGGGUCCUUGC AAGUUAAACUCCGUCAAA
cpZKSCAN1-24nt	UAGCUUUUGAGACAGGGUCCUUGC AAGUUAAACUCCGUCAAAAAAAAA
cpZKSCAN1-gap 10nt	UGUAAGUAGCUUUUGAGACAGUCCGUCAAAAAAAAAAAAAUC
cpZKSCAN1-gap 100nt	AACUCCUGACCUCAAGUGAUCCCUCAAACAGGCCGACUCCAC
cpKCNN2	GGUAAAUUAUGUUCUCAGGAAAGAAAAUCAGCUUCAGUAUGG
cpLIFR	UAAUGUCAUAGGAAGAAAAUCAAUUCCAGAUGGUGUUCAGAU
siRNA	Sequences (5′ to 3′)
siDHX9-1	GUUAAUGAACGUAUGCUGA (dT)(dT)
siDHX9-2	GCAUGGACCUCAAGAAUGA (dT)(dT)

Cell culture and transfection

The Bel7402 (RRID: CVCL_549), Bel7404 (RRID: CVCL_6568), Hek293 (RRID: CVCL_0045), Huh7 (RRID: CVCL_0336) and Skhep1 (RRID: CVCL_0525) cell lines were obtained from the Xiangya Experiment Center (Changsha, China). All cell lines were authenticated and tested for mycoplasma contamination prior to use. Bel7402, Bel7404 and Skhep1 cells were cultured in RPMI-1640 medium (VivaCell, Shanghai, China) supplemented with 10% FBS (Cell-box, Changsha, China) and 1% penicillin-streptomycin sulfate (Solarbio, Beijing, China) at 37°C with 5% CO₂. Hek293 and Huh7 cells were grown in high-glucose DMEM (VivaCell) supplemented with 10% FBS and 1% penicillin-streptomycin sulfate at 37°C with 5% CO₂.

Lipofectamine™ 2000 reagent (Invitrogen, Carlsbad, CA) and Opti-MEM I (Invitrogen) were utilized for transfection according to the manufacturer's instructions. The circGFP reporter transfected amounted to 0.3 μg per 12-well plate. For other plasmids, the transfected quantity was 1.2 μg per 12-well plate unless otherwise specified. The amount of cpRNA and siRNA transfected was 100 nM per 12-well plate unless otherwise specified. Forty-eight hours post-transfection, total RNAs and proteins were extracted for subsequent experiments unless otherwise specified.

RNA extraction and real-time PCR

Total RNA was extracted from various cell lines using TRIzol reagent (Invitrogen). The extracted RNA was then treated with DNase I (Promega) at 37°C for 15 min, followed by heat inactivation at 95°C for 10 s to inactivate DNase I. To digest linear RNA, 1 μg of total RNA was subjected to 2 U RNase R (Beyotime, Shanghai, China) treatment in 20 μl reactions at 37°C for 30 min, followed by heat inactivation at 70°C for 10 min. The cDNA synthesis for circRNA and mRNA was performed using the Revert Aid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific, Waltham, MA). The cDNA synthesis for miRNA was performed using the SynScript® III miRNA RT SuperMix Kit (Tsingke Biotech). All cDNA products were uniformly diluted to a concentration of 100 ng/μl. Subsequently, a 1 μl sample volume was used for qPCR analysis with the Hieff qPCR SYBR Green Master Mix (Yeasen Biotech, Shanghai, China). The primers for PCR were listed in Table 1. For normalization, β-actin was used for circRNA, and U6 snRNA was used for miRNA. Each experiment was repeated three times, and the expression levels were assessed using the 2^(–ΔΔCt) method.

RNA precipitation

All Biotin-labelled probes listed in Table 1 were synthesized by Sangon Biotech (Shanghai, China), and 100 pm biotin-labeled probe was incubated with 100 μg streptavidin beads T3 (APExBIO, Houston, TX) for 1 h at 25°C to facilitate binding. Following incubation, the beads were washed with B&W Wash Buffer three times. Total RNA was extracted from cells using TRIzol reagent, and 10 mg of the total RNA was then incubated at 4°C with the probe-conjugated beads overnight. The next day, the mixture was washed three more times, reverse transcribed, and subsequently detected by qPCR.

Luciferase assay

Hek293 cells (1 × 10⁵) were seeded in 24-well plates and co-transfected with 0, 5 and 10 nM of GMR-miRTM single-stranded mimics (sequences were listed in Table 1), 200 ng of the target reporter plasmids, and 40 ng of pRL-CMV-Renilla plasmids. Approximately 48 hours post-transfection, the luciferase activities of both firefly and Renilla luciferase were assessed using the Dual-Luciferase Reporter Assay System (Promega), and the ratio between firefly and Renilla was calculated as the normalized translational activity.

Western blot

The cells were washed with cold PBS and lysed using RIPA buffer (Beyotime) containing protease inhibitor cocktails (Thermo Fisher Scientific). The concentration of the total protein was determined using a BCA kit (Bioss). Equal amounts of extracted proteins were separated by 8% SDS-polyacrylamide gel and then transferred to 0.45 μm PVDF membranes (Millipore, Billerica, MA). Subsequently, the membranes were blocked with 5% milk (Bioss, Beijing, China) for 1 h at 25°C, followed by incubation with primary antibodies against GFP (1:2000, TransGen Biotech, Beijing, China), Tubulin (1:5000, Proteintech, Wuhan, China), GAPDH (1:5000, Proteintech), DLC1 (1: 1000, Immunoway Biotechnology, Beijing, China) and TIMP3 (1: 1000, Immunoway Biotechnology) at 4°C overnight. Afterward, the membranes were rinsed three times with TBS-T and incubated with horseradish peroxidase (HRP)-linked secondary antibodies (1:5000, Proteintech) for 2 h. The protein bands were detected using enhanced chemiluminescence (APExBIO).

Cell growth assay

The cells were seeded in a 96-well plate at a density of 1 × 10³ cells per well and incubated at 37°C with 5% CO₂ for up to 7 days. Afterwards, 10 μl of the CCK-8 reagent (Biosharp, Beijing, China) was added to each well, and the plate was further incubated at 37°C for 2 h. The absorbance was then measured at 450 nm using a microplate reader at different time points.

Clone formation assay

Five hundred cells mixed with 1.5 ml top gel (0.35% agar in RPMI 1640 medium supplemented with 10% FBS and 1% penicillin / streptomycin) were seeded on 6-well plates covered with 2 ml bottom gel (0.5% agar in RPMI 1640 medium supplemented with 10% FBS and 1% penicillin/streptomycin), and cultured for 2–3 weeks. To visualize the anchorage independent cell growth, the colonies were fixed with 4% paraformaldehyde and stained with 0.1% Crystal Violet Staining solution (Beyotime).

Wound healing assays

Cells were initially seeded in 6-well plates at a density of 5 × 10⁵ cells per well and incubated at 37°C with 5% CO₂ for 24 h. After incubation, the adherent cell monolayers were intentionally scratched using a 200-μl pipette tip. The cells were then cultured in 2 ml RPMI 1640 medium without FBS or antibiotics. Subsequently, cell migration was observed and monitored under a microscope.

Bioinformatic analysis

For RCM pairing, we selected 66 395 circRNAs fully situated within protein-coding genes from the circbase (http://www.circbase.org/) (25) and 16 212 exonic circRNAs which are conserved (Multiple Conservation Score > 5) from circAtals3.0 (https://ngdc.cncb.ac.cn/circatlas/) (26). We then identified the sequences of the intron pairs flanking these circRNAs, using randomly paired introns as controls. Intron pair alignments were conducted utilizing the Needleman-Wunsch algorithm with the parameters ‘-gapopen 10.0 -gapextend 0.5 -endopen 10.0 -endextend 0.5’. Subsequently, we computed scores for the paired introns, along with recording their respective lengths. Intron pairs with scores below 32 were classified as unpaired and assigned a score of 16. Detailed scripts and an illustrative example are provided in supplementary file S1. The scores and lengths of introns for all circRNAs are displayed in supplementary files S2 and S3.

For predicting circRNA-miRNA interactions, we employed miRanda (https://regendbase.org/tools/miranda/)(27), while for forecasting miRNA-mRNA interactions, we utilized Starbase (28). The comprehensive list of miRNA and mRNA targets can be found in supplementary file S4.

Statistics

GraphPad Prism 8.0 software packages were performed for statistical analysis. Differences between experimental groups were conducted through Student t-test to analyze the expression changes. All P values were two-sided, and P < 0.05 indicates statistically significant differences (*P < 0.05, **P < 0.01, ***P < 0.001).

Results

Construction of circRNA promoting RNA

Previous studies have revealed that the flanking sequences of many circRNAs contain RCMs within their introns, and these RCMs are crucial for circRNA formation. Therefore, we used the Needleman-Wunsch algorithm to compare the pairing details between 66 395 pairs of flanking introns forming circRNAs and 54 181 pairs of randomly paired introns within circRNA genes from Circbase, as well as 16 212 pairs of flanking introns forming circRNAs and 13 259 pairs of randomly paired introns within circRNA genes from CircAtlas (Supplementary files S2, S3). The findings from Circbase aligned with previous studies, showing that the average pairing scores of flanking intron pairs of circRNAs were significantly higher than those of the randomly paired introns (P< 0.0001, Figure 1A). However, when compared to randomly paired introns, the average length of the flanking introns did not exhibit significant differences (front intron P= 0.004, rear intron P= 0.212, Figure 1B), indicating that the high scores of the flanking intron pairs were not due to their length but rather to their better complementary pairing. Results from CircAtlas exhibited a similar trend, with the average score of flanking intron pairs being higher than that of randomly paired introns, although no differences were observed in the length of either the front or rear intron (Supplementary Figure S2). The complementary pairing of flanking sequences can bring the downstream splice donor site and the upstream splice acceptor site closer, thus stabilizing the circular structure. This phenomenon might explain why the flanking regions of circRNAs contain numerous RCMs, and these RCMs may facilitate the biogenesis of circRNAs. Recognizing the importance of RCMs in circRNA formation, we propose the development of oligonucleotide sequences capable of binding to the outer regions of circRNA flanking RCMs. This approach would aid in the complementary pairing between circRNA flanking RCMs, specifically promoting circRNA generation.

Figure 1.

Designed cpRNA promotes the production of circGFP. (A) The pairing scores in the flanking intron pairs of circRNAs are compared to those in randomly paired introns. (B) Length distributions of introns flanking circRNA-producing exon(s), compared with the lengths of random introns. (C) Schematic diagram illustrating cpRNAs promoting circRNA formation. (D–G) The feasibility of cpRNA promoting circGFP. (D) The promotive effect of cpRNA on the biogenesis of circGFP was determined using semi-quantitative RT-PCR assay. Sequencing data of the junction site was listed below. (E) The promotive effect of cpRNA on circGFP production was examined by confocal microscopy. (F) The promotive effect of cpRNA on circGFP production was further detected by western blot. (G) The biotin-labeled cpRNA pull-down assay confirmed the binding between cpRNA and circGFP pre-mRNA. The error bars represented the standard deviation (n = 3).

To develop and assess the new circRNA modulation tools, we engineered a split-GFP reporter plasmid containing two reverse GFP fragments capable of undergoing reverse splicing to produce circGFP, thereby expressing functional GFP protein. Additionally, we introduced a pair of reverse complementary sequences within the flanking introns to bring the splicing sites closer, mimicking the RCM sequences found in endogenous circRNAs. Subsequently, we designed an oligonucleotide sequence known as cpRNA, which can complementarily pair with both the upstream and downstream flanks of the RCM (Figure 1C). After co-transfecting the split-GFP reporter plasmid and cpRNA into Bel-7402 cells, we used semi-quantitative RT-PCR to assess the levels of circGFP after 48 h. The results showed a significant increase in circGFP production due to cpRNA, without affecting linear GFP expression. Furthermore, validation using Sanger sequencing confirmed that the amplified circRNA levels induced by cpRNA indeed originated from circGFP (Figure 1D). Additionally, confocal microscopy results revealed heightened green fluorescence intensity in cpRNA-transfected cells compared to the negative control, indicating an increase in circGFP expression (Figure 1E). Western blot analysis further supported this finding, showing a substantial increase in circGFP expression in cells co-transfected with cpRNA compared to the negative control (Figure 1F). To confirm the binding between cpRNA and the RCM flanks of circGFP, we used biotinylated cpRNA probes to pull down pre-mRNA transcribed by the split-GFP reporter plasmid. The pulled-down RNA was then measured using three pairs of primers targeting the upstream, downstream, and middle segments of circRNA, with β-actin serving as a negative control that did not bind to the cpRNA. The results indicated that cpRNA could co-precipitate with pre-mRNA through complementary pairing with the upstream and downstream flanks of the RCM. The precipitation efficiency for all three primer pairs was around 50%, suggesting that the binding of cpRNA to pre-mRNA did not cause the breakage or degradation of pre-mRNA (Figure 1G).

Optimization of the circRNA promoting RNA

In order to achieve the optimal promotion efficacy and cost efficiency, we carried out optimization experiments for cpRNA using the split-GFP reporter system, manipulating factors such as the number of intermediate spacer bases in cpRNA, the lengths of the binding sites on both sides of cpRNA, the distance from the binding site to RCMs, as well as the concentration and duration of action.

At a concentration of 100nM, we tested cpRNAs with varying numbers of intermediate spacer bases, finding that a smaller number of these bases in the two sequences targeting the front and back of the circRNA junction site resulted in a more potent promotion effect for circGFP biogenesis. The most significant effect was observed when there were no intermediate spacer bases. Consequently, we opted for a single-stranded cpRNA with a total of 0 intermediate spacer bases. Additionally, we observed that separated RNA fragments targeting the front and back of the circRNA junction site did not effectively promote circGFP formation (Figure 2A), suggesting that the promotion of circRNA circularization by cpRNA relies on the proximity effects of the introns at both ends of the circRNA junction site. Subsequently, we optimized the lengths of the binding sites on both sides of cpRNA. Results revealed that cpRNAs with binding site lengths of 18, 21 and 24 bp can all demonstrated the ability to enhance circGFP expression. Notably, the most significant promotion of circGFP biogenesis was observed when each side had a binding length of 21 bp (Figure 2B). Therefore, we chose the length of 21bp as the optimal binding length on each side. To explore the impact of the distance from RCMs on the promotion effect of cpRNA, we designed cpRNAs at various distances from the end of RCMs. We found that as the distance from RCMs increased, the promotion effect of circGFP biogenesis weakened. Specifically, when the distance was within 20 bp, cpRNA exhibited a notable promotion effect on circRNA production. However, once the distance exceeded 50 bp, the promotion effect almost disappeared (Figure 2C). Consequently, we chose to use cpRNA seamlessly targeting the end of RCMs.

Figure 2.

Optimization of the cpRNA. (A) Optimization of intermediate spacer bases of the cpRNA. M0, M2, M6 represent the cpRNAs with 0 adenine, 2 adenines and 6 adenines in their intermediate spacer bases, respectively. S1 + S2 denotes that the cpRNAs are two separate RNAs without a direct linkage. (B) Optimization of the lengths of the binding sites on both sides of cpRNA. B18, B21, B24 represent the lengths of the binding sites of cpRNAs, which are 18 nt, 21 nt and 24 nt, respectively. (C) Optimization of the distance from the binding site to RCMs. D1, D5, D10, D20, D50, D100 represent the distances from the binding site of cpRNAs to RCMs, which are 1 nt, 5 nt, 10 nt, 20 nt, 50 nt and 100 nt, respectively. (D) Optimization of reaction concentration of cpRNA. (E) Optimization of reaction time of cpRNA. (F) Optimization reaction concentration of 2′-O-Me modified cpRNA. (G) Optimization of the amounts of cpRNA-containing plasmid. The error bars represented the standard deviation (n = 3).

Following this, we co-transfected a fixed amount of the circGFP reporter gene with varying concentrations of cpRNA. After two days, we observed that the production of circGFP increased to a certain extent with the rise in cpRNA concentration. The optimal concentration was found to be 100 nM (Figure 2D), therefore, we selected 100 nM as the best concentration for cpRNA transfection. Additionally, to assess the time dependency of cpRNA action, we transiently co-transfected a fixed amount of circGFP at a final concentration of 100nM and checked the levels of circGFP at 24, 48, 72 and 96 h post transfection. As expected, cpRNA promoted the production of circGFP in a time-dependent manner, reaching its peak at 48–72 h (Figure 2E). Furthermore, due to the easy degradation of free single-stranded oligonucleotides in vivo, their duration and efficiency are inherently limited. To ensure sustained and effective regulation, the optimal cpRNA sequence was either modified with 2′-O-methylation or integrated into pcDNA3.1/Zeo(+) to enhance the regulatory efficiency. The results demonstrated that both the 2′-O-Me modified cpRNA and the cpRNA expression plasmid notably enhanced circularization efficiency. Specifically, the 2′-O-Me modified cpRNA amplified circGFP production by 15.0 ± 0.8 times at a concentration of 100 nm (Figure 2F), while the cpRNA expression plasmid increased circGFP levels by 19.0 ± 1.2 times at a concentration of 1.2 μg (Figure 2G). These enhancements far surpassed those achieved with unmodified RNA oligonucleotides, which yielded approximately 6–8 times increase at 100nM. Consequently, for subsequent endogenous experiments, we opted to utilize the cpRNA expression plasmid.

CircRNA promoting RNA stimulates the biogenesis of endogenous circRNAs

After establishing the optimal conditions for cpRNA regulation, we used cpRNAs to facilitate the production of endogenous circRNAs. The following four endogenous circRNAs were individually selected: circSMARCA5, generated from exons 15 to 16 (Supplementary Figure S3A), has been reported to be markedly downregulated in HCC and closely associated with poor prognosis in HCC patients(11). CircZKSCAN1, produced from exons 2 to 3 (Supplementary Figure S3B), inhibits the growth, migration, and invasion of HCC through various signaling pathways, and its expression is significantly reduced in HCC tissues(12). CircKCNN2, produced through exon 3 splicing (Supplementary Figure S3C), exhibits lower expression levels in HCC tissues, and low levels of circKCNN2 are significantly linked to poor overall survival (OS) and recurrence-free survival (RFS) in HCC(29). CircLIFR, generated from exons 8 to 11 (Supplementary Figure S3D), is notably downregulated in HCC(30).

We utilized previously optimized conditions for cpRNA and designed specific cpRNAs for these four endogenous circRNAs, cloning their sequences into pcDNA3.1/Zeo(+). Subsequently, we transfected these plasmids carrying cpRNAs into Bel-7402, Bel-7404, Huh7, SK-Hep1 and Hek293 cells. As anticipated, semi-quantitative RT-PCR confirmed that cpRNAs increased the expression of circSMARCA5 and circZKSCAN1 in all five cell lines (Figure 3A,B). Sanger sequencing further validated the existence of endogenous circSMARCA5 and circZKSCAN1 sequences by confirming the presence of circRNA junction sites in sequences (Figure 3A, B). However, for circKCNN2 and circLIFR, two circRNAs without obvious RCMs, cpRNAs did not exhibit a specific stimulatory effect (Figure 3C, D). To assess the robustness of this approach, we varied the distance between the two binding sites, the length of the binding sites, and the distance from the RCM of cpRNA for the endogenous circRNA in Bel7402 and Bel7404 cells. The outcomes revealed that cpRNAs effectively facilitated the biogenesis of endogenous circSMARCA5 and circZKSCAN1 when the distance between the two binding sites ranged from 0 to 2 adenines. However, as this distance expanded to 6 adenines, the efficacy notably decreased (Figure 3E). When the length of the cpRNA binding sites was between 21 and 24 nucleotides, the cpRNAs effectively promoted the biogenesis of endogenous circSMARCA5 and circZKSCAN1. Conversely, a reduction in the binding site length to 18 nucleotides led to a decrease in the promotional effect (Figure 3F). The proximity to the RCM appeared crucial for cpRNAs, with only those cpRNAs situated close to RCM binding sites demonstrating efficacy in promoting circRNA generation. CpRNAs positioned at a moderate distance (around 10 nucleotides) or distal locations (approximately 100 nucleotides) from the RCM exhibited minimal to no effect on enhancing circRNA production (Figure 3G). In summary, our results demonstrate that the application of cpRNAs not only stimulates the biogenesis of exogenous circGFP reporter genes but also promotes the production of endogenous circRNAs. This promotion of circRNA generation appears to function by influencing RCMs, as only circRNAs containing RCMs can be promoted in expression by cpRNA, while circRNAs lacking RCMs cannot be enhanced by this method. In this method, the distance between the two binding sites and the length of the binding sites of the cpRNAs exhibit a certain level of robustness, while the proximity to the RCM appears to be a fixed factor that cannot be altered.

Figure 3.

The impact of cpRNA on the biogenesis of endogenous circRNA production: (A, B) Real-time PCR results demonstrated an increase in the expression of circSMARCA5 and circZKSCAN1 (both containing RCMs) following transfection with cpRNA-containing plasmids. Sequencing data of the junction site is provided below. (C, D) Real-time PCR results revealed no significant alteration in the expression of circKCNN2 and circLIFR (lacking RCMs) after cpRNA-containing plasmids transfection. (E) The promotion effect of cpRNA-containing plasmids with varying linker bases on endogenous circRNAs. (F) The promotion effect of cpRNA-containing plasmids with different lengths of the binding sites on endogenous circRNAs. (G) The promotion effect of cpRNA-containing plasmids with varying distances from the RCMs on endogenous circRNAs. The error bars represented the standard deviation (n = 3).

CircRNA promoting RNA inhibits the proliferation and migration of HCC cells by regulating the circRNA-related pathway

In order to validate the direct impact of enhanced cpRNA-induced circRNAs on the proliferation and migration of HCC cells, we separately transfected cpZKSCAN1 and cpSMARCA5 express plasmid into Bel-7402 and Bel-7404 cells. Following this, we assessed the growth rate, colony formation ability, and migration rate of the HCC cells post cpRNA transfection. Initially, we conducted CCK8 experiments to measure cell proliferation, revealing a decreased proliferation rate in cells transfected with cpZKSCAN1 and cpSMARCA5 compared to the control group (Figure 4A, B). Furthermore, the results from the colony formation assay showed fewer and smaller colonies in cells transfected with cpZKSCAN1 or cpSMARCA5 in comparison to the control cells (Figure 4C,D). Subsequently, scratch healing experiments were utilized to examine cell migration, unveiling a significantly slower healing rate in cells transfected with cpZKSCAN1 or cpSMARCA5 compared to the control group (Figure 4E,F). Collectively, these findings demonstrated that the heightened levels of circZKSCAN1 and circSMARCA5 induced by the transfection of cpZKSCAN1 and cpSMARCA5 inhibited the proliferation and migration of HCC cells.

Figure 4.

CpRNA inhibits the proliferation and migration of HCC cells. (A, B) CCK-8 assay demonstrated a reduced growth rate of cells transfected with cpRNAs compared to those transfected with the empty vector. (C) The colony formation assay indicated a lower number of colonies formed by cells transfected with cpRNAs compared to the control group. (D) Statistical analysis of the colony formation assay. (E) Results from the wound healing assay showed a slower healing speed in cells transfected with cpRNA. (F) Statistical analysis of the wound healing assay. The error bars represented the standard deviation (n = 3).

Afterwards, we delved into the downstream pathways of circRNA. Drawing from previous literature(11,31), we pinpointed circZKSCAN1/miR-873-5p/DLC1 and cpSMARCA5/miR-181-5p/TIMP3 as the downstream pathways of circZKSCAN1 and circSMARCA5, respectively. This selection was based on reports indicating that circZKSCAN1 directly targets miR-873-5p, resulting in the suppression of DLC1 (31), which subsequently inhibits the proliferation and autophagy of HCC(32). Similarly, circSMARCA5 was discovered to target miR-181-5p directly, leading to the suppression of TIMP3. The upregulation of TIMP3 expression and the downregulation of miR-181-5p have been linked to the progression and poor prognosis of HCC (11). This information was then validated through bioinformatic analysis, RNA precipitation, and luciferase assay. The MiRanda website revealed that circZKSCAN1 partially binds to miR-873-5p, while circSMARCA5 partially binds to miR-181-5p. Additionally, the Starbase database showed that miR-873-5p could bind to the 3′-UTR of DLC1, and miR-181-5p could bind to the 3′-UTR of TIMP3 (Figure 5A, Supplementary file S4). RNA precipitation demonstrated that miR-873-5p co-precipitated with circZKSCAN1, while miR-181-5p co-precipitated with circSMARCA5 (Figure 5B). Luciferase assays illustrated that miR-873-5p silenced the activity of the luciferase reporter with the 3′-UTR of DLC1 in a dose-dependent manner, while miR-181-5p silenced the activity of the luciferase reporter with the 3′-UTR of TIMP3 in a dose-dependent manner (Figure 5C). These results collectively confirmed the existence of these two pathways as reported previously. Subsequently, the expression of circZKSCAN1 and circSMARCA5 and their downstream miRNAs and mRNAs in cells transfected with cpZKSCAN1 and cpSMARCA5 was measured using semi-quantitative RT-PCR. The data revealed an upregulation of circZKSCAN1, a downregulation of miR-873-5p, and an upregulation of DLC1 mRNA in cpZKSCAN1-transfected cells. Similarly, an upregulation of circSMARCA5, a downregulation of miR-181-5p, and an upregulation of TIMP3 mRNA were observed in cpSMARCA5-transfected cells (Figure 5D). Furthermore, the impact of cpRNA transfection on these miRNA target genes was evaluated by assessing the protein levels of DLC1 and TIMP3 using western blotting. The results indicated an increase in DLC1 expression in cpZKSCAN1-transfected cells and an increase in TIMP3 expression in cpSMARCA5-transfected cells (Figure 5E).

Figure 5.

CpRNA regulates the downstream pathways of circRNAs. (A) The predicted binding sites of the circZKSCAN1-miR-873-5p-DLC1 signaling pathway and the circSMARCA5-miR-181b-5p-TIMP3 signaling pathway. (B) The co-precipitation results confirmed the interaction between circRNA and its target miRNA. (C) The luciferase assay confirmed the silencing effect of miRNA on its target mRNA. (D) Real-time PCR results demonstrated alterations in the expression levels of circZKSCAN1 and circSMARCA5, as well as their respective downstream miRNAs and mRNAs following transfection with cpZKSCAN1 and cpSMARCA5 containing plasmids. (E) Western Blot results indicated an increase in the expression of downstream proteins after transfection with cpRNA-containing plasmids. The error bars represented the standard deviation (n = 3).

In summary, these findings indicate that the introduction of cpRNA disrupts miRNA-mediated gene silencing by altering the levels of circRNA.

CircRNA promoting RNA promotes circRNA biogenesis in part by antagonizing the unwinding function of DHX9

Since cpRNA can only promote the biogenesis of circRNA containing RCMs, we speculate that the promotion of circRNA by cpRNA is related to its assistance in RCM helix formation. To further investigate the mechanism by which it affects RCMs, we decided to explore the proteins acting on RCMs. Previous studies have indicated that DHX9, an important member of RNA helicases, inhibits circRNA expression by unwinding the ALU complementary sequences (23). Therefore, we hypothesize that cpRNA can antagonize DHX9, inhibit DHX9’s unwinding action on RCMs, and thereby promote circRNA generation. To address this, we separately constructed cell lines with DHX9 overexpression (OE) and knockdown (KD), and transfected cpRNA and circGFP plasmids into each. The results showed that DHX9 OE led to a downregulation of circGFP expression. However, this inhibitory effect of DHX9 was reversed when cpRNA was introduced, although the expression levels were still lower than the DHX9 non-OE control group (Figure 6A). This suggests that cpRNA’s promotive effect may be achieved by weakening DHX9 function. Subsequently, we knocked down DHX9 and obtained consistent results: circGFP expression increased following DHX9 KD, and in the DHX9 KD group, introducing cpRNA led to increased circGFP expression. However, the promotive effect of cpRNA in the KD group was not significantly different from that in the non-KD group, indicating that only part of cpRNA’s promotive effect on circRNA biogenesis is achieved by antagonizing DHX9 function (Figure 6B). There are other mechanisms through which cpRNA can promote circRNA biogenesis. Figure 6C depicts the regulatory mechanism by which circRNA is influenced by cpRNA and DHX9.

Figure 6.

CpRNA promotes circGFP biogenesis in a DHX9-dependent manner. (A) Over-expression of DHX9 inhibited circGFP expression, which could be reversed by cpRNA. (B) Knockdown of DHX9 promoted the production of circGFP, but had little effect on cpRNA. The upper panel showed the circRNA level, while the lower panel showed the protein level. (C) The schematic diagram illustrates the regulatory mechanism by which circRNA is influenced by cpRNA and DHX9. The error bars represented the standard deviation (n = 3).

Discussion

In recent years, the exploration of circRNA biology has expanded significantly, leading to the discovery of a large number of highly abundant circRNAs in mammalian transcriptomes(33). However, the functional roles of many circRNAs remain unknown due to the lack of techniques and experimental methods, particularly limitations in in vivo overexpression strategies(14). Hence, exploring methods to promote circRNA production should provide great hope for both basic and translational research. It is worth noting that in the mechanism of intron-pairing-driven circularization, the formation of reverse complementary matches (RCMs) brings the splice sites closer together, thereby facilitating the formation of circRNA (20). Based on this, we have developed a specific RNA molecule called circRNA-promoting RNA (cpRNA), which is designed to be complementary to the outer sequence of the RCMs. The purpose of cpRNA is to bring the ends of the circRNA precursor sequence closer together and enhance the stable pairing of the flanking introns of the circRNA, ultimately facilitating circRNA production.

To validate our hypothesis, we initially used a split-GFP reporter system containing two reverse GFP fragments capable of reverse splicing to generate circGFP. Concurrently, we introduced a pair of reverse complementary sequences into the flanking introns of circGFP to simulate the RCMs found in endogenous circRNAs. Next, we designed a short nucleotide sequence, termed cpRNA, which could complementarily pair with both the upstream and downstream sides of the RCMs in the split-GFP reporter system. We co-transfected the split-GFP reporter plasmid and cpRNA into Bel-7402 cells, observing that cpRNA indeed enhanced circGFP production. Moreover, we optimized the cpRNA sequences through a series of experiments and determined the optimal conditions for cpRNA to promote circRNA biogenesis. Significantly, we found that cpRNAs can also specifically promote the production of endogenous circRNAs with RCMs, such as circZKSCAN1 and circSMARCA5.

Given the involvement of circRNA dysregulation in various diseases like cancer, cardiovascular disease, and neurodegenerative disorders, the therapeutic potential of circRNA is currently under extensive investigation (14). Particularly in cancer, recent studies have revealed a significant reduction in circRNA levels in tumor tissues compared to normal tissues. The expression of many circRNAs is negatively correlated with cancer cell proliferation(9) and has substantial inhibitory effects on tumorigenesis and metastasis (13). Therefore, addressing how to increase the expression of these tumor suppressors is a clinically important issue that requires urgent attention. Hence, the primary application potential of cpRNA likely lies in its ability to elevate circRNA expression levels in tumors. CircZKSCAN1 and circSMARCA5 have been identified as tumor suppressors in HCC (11,12), and we utilized cpRNA to specifically boost the production of circZKSCAN1 and circSMARCA5 in several liver cancer cell lines. This resulted in the inhibition of proliferation and migration of HCC cells through the regulation of circZKSCAN1 and circSMARCA5-associated pathways, demonstrating the therapeutic potential of cpRNAs. Furthermore, our mechanistic studies suggest that the promoting effect of cpRNA on circRNA biogenesis is partly achieved through antagonizing the unwinding function of DHX9.

Currently, the most popular strategy for circRNA overexpression involves transfecting cells with circRNA-producing plasmids or chemically synthesizing circRNAs in vitro(14). In comparison to these overexpression methods, the design of cpRNAs is simpler and more cost-effective. Importantly, it reduces interference from linear transcripts. However, the cpRNA strategies we have proposed come with certain limitations. Firstly, cpRNAs are only effective against circRNAs with RCMs in their flanking introns. Since few circRNAs contain RCMs in their flanking introns, the scope of application is significantly limited. Secondly, there is a lack of specificity for circRNAs originating from the same gene, as long as they share similar junction sites. Unspecific circRNAs may be produced as a byproduct with the promotion of cpRNA. Thirdly, the efficiency and precision of nucleic acid delivery have consistently posed challenges in the realm of gene therapy (34). As a result, our future research endeavors will focus on developing circRNA promotion strategies that provide broader applicability, heightened efficiency, and enhanced specificity to overcome these issues and enhance the therapeutic potential of cpRNAs.

Apart from intron-pairing-driven circularization, RBP-mediated circularization is a crucial mechanism in circRNA biogenesis. However, since different RBPs have distinct recognition sequences and diverse impacts on circRNA formation (35), designing cpRNA that target these sequences may entail entirely different mechanisms. Understanding the impact of cpRNAs on the function of these RBPs and designing specific regulators for them will be the central focus of our upcoming work. Basically, these RBPs are categorized into three groups. The first group comprises proteins that bind to the flanking intronic sequences, bringing splice sites into proximity, including MBL/MBNL1 (36), QKI (37,38), NF90/NF110 (24) and others. We hypothesize that cpRNA may either facilitate or have no effect on this process. The second group consists of RNA helicases that disrupt RCM structures, such as DHX9 (23) and others. Our findings indicate that cpRNA can counteract this process and enhance circRNA biogenesis (Figure 6). The final group encompasses RNA splicing regulators like HnRNPs (39), SR Proteins (39), RBM20 (40), SFPQ (41), CPSF (42,43) and more. These proteins modulate circRNA expression through alternative mechanisms that may not be related to the cpRNA mechanism.

In summary, the development of cpRNA presents a novel approach to selectively enhance circRNA production, offering substantial implications for both fundamental research and therapeutic endeavors. Nevertheless, being short single-stranded nucleotides, its applicability is limited, necessitating improvements in specificity, targeting, and stability. Future investigations should prioritize enhancing broader applicability, increasing efficiency, and refining specificity to address these challenges and amplify the therapeutic efficacy of cpRNAs. Additionally, research efforts should focus on bolstering the stability of cpRNAs and investigating delivery methods to surmount these obstacles and maximize the therapeutic potential of cpRNAs.

Data availability

The data underlying this article are available in the article and in its online supplementary material. All cpRNA sequences are presented in Table 1. We adhere to the ‘Minimal Information for Publication of Quantitative Real-Time PCR Experiments’ (MIQE) guidelines, and the primer sequences used in the quantitative Real-Time PCR are provided in Table 1.

Supplementary data

Supplementary Data are available at NAR Online.

Acknowledgements

We are grateful to Professor Rui Xiao at Wuhan University for providing the RCMs analysis in flanking introns of circRNAs.

Author contributions: Zhilin He: Conceptualization, Validation, Methodology, writing original draft, formal analysis. Haofei Ji: Methodology, data curation. Bei Xia: Methodology, data curation. Xiuen Cao: Formal analysis. Ying Huang: Formal analysis. Qubo Zhu: Conceptualization, writing original draft, review, editing.

Funding

National Natural Science Foundation of China [C0602-32270609]; Hunan Provincial Natural Science Foundation of China [2021JJ30916, 2021JJ80078]; Hunan Provincial Natural Science Foundation for Distinguished Young Scholars [2022JJ10091]. Funding for open access charge: National Natural Science Foundation of China.

Conflict of interest statement. All authors declare that they have no conflicts of interest.