A precise and efficient circular RNA synthesis system based on a ribozyme derived from Tetrahymena thermophila

Abstract Classic strategies for circular RNA (circRNA) preparation always introduce large numbers of linear transcripts or extra nucleotides to the circularized product. In this study, we aimed to develop an efficient system for circRNA preparation based on a self-splicing ribozyme derived from an optimized Tetrahymena thermophila group Ⅰ intron. The target RNA sequence was inserted downstream of the ribozyme and a complementary antisense region was added upstream of the ribozyme to assist cyclization. Then, we compared the circularization efficiency of ribozyme or flanking intronic complementary sequence (ICS)-mediated methods through the DNMT1, CDR1as, FOXO3, and HIPK3 genes and found that the efficiency of our system was remarkably higher than that of flanking ICS-mediated method. Consequently, the circularized products mediated by ribozyme are not introduced with additional nucleotides. Meanwhile, the overexpressed circFOXO3 maintained its biological functions in regulating cell proliferation, migration, and apoptosis. Finally, a ribozyme-based circular mRNA expression system was demonstrated with a split green fluorescent protein (GFP) using an optimized Coxsackievirus B3 (CVB3) internal ribosome entry site (IRES) sequence, and this system achieved successful translation of circularized mRNA. Therefore, this novel, convenient, and rapid engineering RNA circularization system can be applied for the functional study and large-scale preparation of circular RNA in the future.


INTRODUCTION
Cir cular RNAs (cir cRNAs) ar e a class of covalently closed RNA molecules pr oduced fr om precursor mRNA back splicing and are widely found in nature ( 1 ). Due to the lack of 5' or 3' ends, circRNAs have a certain resistance to exonuclease digestion, which makes circRNAs more stable than linear RNAs. Most natural circRNAs are noncoding, although a few of them can be translated into peptides ( 2 , 3 ). Noncoding circRNAs can serve as sponges and influence corresponding functions by binding to miRNAs and proteins ( 4 ). Endogenous cir cRNAs ar e essential in normal cell dif ferentia tion, tissue homeostasis, and disease development ( 5 ). In addition, some endo genous circRN As play a role in antiviral responses while some are associated with immune responses. Exogenous circRNAs can stimulate immune signaling in mammalian cells b y activ ating the pattern-r ecognition r eceptor RIG-I ( 6 ).
CircRNAs can be used to express various types of proteins and perform the same function as mRN As, w hich makes them a promising next-generation mRN A thera py ( 7 ). Cir cRNAs can over come some limitations of mRNA in terms of stability, nonspecific tissue expression, and imm uno genicity ( 8 ). Ubiquitous RNA enzymes in the environment threaten mRNA stability due to their open ends ( 9 ). In contrast, with their ring structur es, cir cRNAs ar e r esistant to the RNA degradation system in vivo , which leads to a longer half-life for their effects as a therapeutic drug. CircRNAs can also be designed and optimized for effecti v e expression in specific tissues and cells ( 10 , 11 ), which helps to reduce the side effects of circRNAs e v en if they are delivered into improper tissues. Furthermore, mRNAs have imm uno genicity that can e xcessi v ely acti va te inna te immunity, resulting in symptoms similar to those of viral infection, inflammation, and autoimmune diseases ( 12 , 13 ). In contrast, cir cRNAs display r elati v ely low immunogenicity e v en without base modification ( 14 , 15 ), showing more advantages in the r esear ch and de v elopment of future patented drugs and vaccines.
Curr ently, ther e ar e two main strategies for engineering RNA circulariza tion in vitr o : T4 RN A / DN A ligase ligation or ribozyme methods. The ligases used for circularization can be either T4 DNA ligase or T4 RNA ligases 1 and 2. All three enzymes are ATP-dependent and can catalyze the binding of 5'-phosphate and 3'-hydroxyl groups at the end of the RNA. Howe v er, ligase-media ted circulariza tion is not suitable for the construction of long fragments due to its low circularization efficiency, abundant byproducts with exogenous nucleotides, and complicated preparation and purification processes ( 16 ). Previously, some catalytic RNAs have been widely applied to synthesize larger circRNAs. A group of small RNAs deri v ed from satTRSV (-) RNA was designed to produce self-circularizing RNAs of known sizes ( 17 ). The hairpin ribozyme variants have also been engineered to generate circular RNA ( 18 ). Another common method called permuted intron-exon (PIE) relies on a spontaneous gr oup I intr on self-splicing system deri v ed from the thymid yla te synthase gene of the T4 phage, pretRNA Leu of Anabaena or pre-rRNA of Tetrahymena (19)(20)(21)(22). The permuted gr oup I intr on precursor RNA contains end-to-end fused exons that interrupt half intron sequences. Foreign sequences can be integrated into the exon of such a permuted self-splicing system, thus allowing the synthesis of circR-NAs. Compared with the T4 RNA ligase method, the group I intron ribozyme-based circularization method is more suitable for the synthesis of long fragment circRNA and substantially improv es efficiency. Howe v er, the latter also has many problems. For example, circRNA derived from the phage T4 thymid yla te synthase (td) gene or permuted Anabaena pretRNA group I intron cannot clip off the long splints that assist circular formation, thus introducing an additional 74 nt or 186 nt of nucleotides, respecti v ely ( 21 ).
The most common method of endogenic RNA circularization for circRNA function research in mammalian cells is mediated by flanking complementary sequences, such as ALU repeats or short intronic complementary sequences. For the flanking ICS-mediated circularization method, endonucleolytic splicing for gener ating RNA fr agments appears to be a rate-limiting step. Due to its low efficiency, the flanking ICS method may lead to the introduction of a large number of linear transcripts, which may exert nonspecific effects in the studies of biological function.
In the present study, we used an optimized ribozyme from the Tetrahymena therm ophila gr oup I intr on (23)(24)(25) to establish a novel and simplified system for RNA circularization both in mammalian cells and in vitr o , genera ting an ideal circRNA without any additional nucleotides for the ov ere xpression in cells, and therefore, this system is suitable for large-scale industrial production of circular mRNA.

Vector construction
The acti v e ribozyme contained the sequences from 28 to 414 nt of the T. thermophila intron, which was located in the intervening region of the T. thermophila large subunit rRNA precursor. The sequences of the Tetrahymena ribozyme with an internal guide sequence (IGS), the RNAs to be circularized, and a 45-nt antisense region were chemically synthesized by Beijing Genomics Institute (Beijing, China) and cloned into a double-enzyme-digested linearized pCDH vector.
The ov ere xpression plasmids of specific circRNAs based on flanking ICSs and paired control plasmids were obtained from the Public Protein / Plasmid Library (PPL, Nanjing, China), which contains a general front and back circular frame. The indicated RNA sequences were cloned into the construct by EcoRI / BamHI restriction sites.
For the GFP reporter gene vector, a synthesized Coxsackievirus B3 (CVB3) internal ribosome entry site (IRES) fragment and an exon that encoded GFP were inserted into the abo vementioned circRNA o ver expr ession vector using the Tetrahymena ribozyme. Specifically, the exon was divided into two parts, and the split GFP gene fragments were inserted into the vector in reverse order. The CVB3 IRES was then inserted between the split GFP gene start and stop codons, resulting in a large exon downstream of the ribozyme. Similarly, we used Gibson Assembly to obtain the full-length sequence. The primers and the sequences of these constructs are listed in Supplementary Table S1.

In vitro transcription and RNA circularization
DNA templates for in vitro transcription were amplified, and an upstream T7 promoter was added by PCR using Phanta Flash Master Mix (Vazyme, Nanjing, China). The product was identified with 1.2% agarose gel electrophoresis and purified with a TIANgel purification kit (Tiangen, Beijing, China). T7 RN A pol ymerase (NEB, MA, USA) was used for in vitro transcription reactions that lasted 2 h at 37 • C. After in vitro transcription, the RNA products were treated with DNase I (Beyotime Biotechnology, Shanghai, China) at 37 • C for 30 min to remove DNA templa tes. Each circulariza tion reaction was supplemented with 2 mM GTP and incubated at 55 • C for 15 min to induce RNA circularization, followed by quick cooling to 4 • C. The aceta te / ethanol precipita tion method was used to purify the RNA ( 26 ).

RN A e xtr action, r everse tr anscription, and polymer ase chain reaction (PCR)
Total RNA was extracted using TRIzol reagent (Invitrogen, CA, USA) according to the manufacturer's instructions and subsequently treated with DNase I (Beyotime Biotechnology, Shanghai, China) at 37 • C for 15 min, followed by treatment at 65 • C for 10 min to inactivate DNase I. The OD value 260 / 280 of the extracted RNA was between 1.9 and 2.1 and the OD value 260 / 230 was greater than 2.0. One microgram of RNA was re v erse-transcribed using Evo M-MLV RT Premix (Accurate Biology, Hunan, China) according to the manufacturer's instructions. The RT product was used for RT-PCR amplification with Phanta Flash Master Mix (Vazyme, Nanjing, China). The RT-PCR product was then separated on a 1.2% agarose gel and scanned with a scanner (Bio-Rad, CA, USA). The same cDNA template was diluted 4-fold for quantitati v e real-time PCR (qPCR) with a SYBR Green Premix Pro Taq HS qPCR kit (Accurate Biology, Hunan, China). The r eaction procedur e for qPCR was incuba tion a t 95 • C f or 30 s, f ollowed by 40 cycles of 95 • C for 5 s and 58 • C for 30 s on an iQ5 Real-Time PCR Thermal Cycler (Bio-Rad, CA, USA). The 2 − Ct method was used to calcula te rela ti v e fold changes. The gene expression le v els were normalized to the endogenous e xpression of the housekeeping gene GAPDH . The splicing efficiency value of circRNA was calculated by dividing circRNA by total RNA products (circRNA / total RNA). The error bar r epr esents the standard error of the mean (SEM) of three independent experiments. The primers used for the PCR are listed in Supplementary Table S1.

Northern blotting (NB)
In brief, RNA was resolved on formaldehyde denatured agarose gel, followed by transfer to a nylon membrane. Digoxigenin (Dig)-labelled cDNA probes were made using a DIG DNA Labelling kit (Mylab, Beijing, China). The membrane was then hybridized with specific Dig-labelled probes at 42 • C. The NB probe is listed in Supplementary  Table S1. Detection steps were performed using a DIG hybrid detection kit (Mylab, Beijing, China) according to the manufacturer's protocol.

Immunoblotting
The cells were washed twice with phospha te-buf fered saline (PBS), and total protein was extracted with a Whole Protein Extraction Kit (K eygen, Nanjing, China). Pr otein samples were separated by SDS −PAGE and then transferred to PVDF membr anes. Subsequently, the membr anes were blocked in 1 × casein buffer for 1 h at room temperature and then incubated with anti-cleaved-parp1 antibody (Santa Cruz, CA, USA), anti-cleaved-caspase 3 antibody (Cell Signaling Technology, USA), anti-GAPDH antibody (Abmart, Shanghai, China) and anti-␤-actin antibody (Abmart, Shanghai, China) at 4 • C overnight. The membranes wer e washed thr ee times with TBS-T, incubated with secondary antibodies (Abmart, Shanghai, China) for 1 h, and detected with a gel imaging analysis system (Sinsage, Beijing, China).

Cell viability assay
A total of 3 × 10 3 transfected cells were harvested and plated in 100 l of medium containing 10% FBS per well in 96-well plates. Then, 100 l of medium containing 10% Cell Counting Kit-8 (CCK-8) reagent (Meilunbio, Dalian, China) was added to each well at different time points (0, 24, 48, 72 or 96 h). Then, the cells were incubated at 37 • C for 1 h. The absorbance at 450 nm was measured by a multimode microplate reader (BioTek, Bedfordshire, UK).

Colony formation assay
A total of 1.0 × 10 3 transfected cells were seeded in each well of a 6-well plate and cultured at 37 • C for 14 days. On the final day in culture, the cells were fixed for 15 min with methanol and then washed twice with PBS. Then, the cells were stained for 30 min with a 0.5% crystal violet solution. The colonies were photo gra phed and counted with ImageJ software. Alternati v ely, acetic acid was used to wash off the crystal violet, and was taken the eluent to measure the OD value at 570 nm on a multimode microplate reader (BioTek, Bedfordshire, UK).

T r answ ell assay
A 24-w ell transw ell insert with an upper chamber was used. Briefly, 1.0 × 10 5 transfected cells in 200 l of serum-free medium were seeded in the upper chamber, and 600 l of medium containing 10% FBS was added to the lower chamber. After incubation at 37 • C for 24 h, the cells on the upper membrane surface were removed with a cotton s wab , and the cells that had migrated to the bottom surface of the upper membrane were fixed in methanol for 30 min and then stained with 0.5% crystal violet for 30 min. The number of migrating cells was counted with ImageJ software. Alternati v ely, acetic acid was used to wash off the crystal violet, and the eluent was taken to measure the OD value at 570 nm on a multimode microplate reader (BioTek, Bedfordshire, UK).

Wound healing assay
The transfected cells were plated in 12-well plates and grown to confluence. The cell monolayer was wounded using a 200l pipette tip. The wounds were imaged at 0, 24 and 48 h after wounding with a phase-contrast microscope (Olympus, Tok yo , Japan).

Statistical analysis
GraphPad Prism version 8.0 was used to analyze all the experimental da ta. Dif ferences between experimental groups were ev aluated b y a two-tailed Student's t test or one-way ANOVA. P values < 0.05 were considered statistically significant.

Establishment of a novel engineered T. Thermophila ribozyme-based RNA circularization system
The desired RNA sequence to be circularized was inserted downstream of the optimized ribozyme and a complementary antisense region was added to the pCDH backbone upstream of the ribozyme to facilitate the folding and stabilization of the ribozyme. The catalytic domain in the 5' portion of the ribozyme was fused with an adjustable 13nt-long IGS. The first 9 nt of the 3' terminus of the IGS sequence was complemented by pairing with the 3' terminus of the circRNA and a U residue was inserted in the 3' terminus of the circRNA. In contrast, the last 6 nt of the 5' terminus of the IGS was designed for incomplete complementary pairing with the 5' terminus of the target circRNA sequence to form helix P10. The secondary structure of RNA pr ecursor was pr edicted by RNAFold (Supplementary Figure S1A).
Cir cDNMT1 (hsa cir c 0049224) was chosen as a target circRNA to evaluate the established T. thermophila ribozyme-based RNA circularization system (Figure 1 A). As a result, a free guanosine was covalently attached to the 3' terminus of DNMT1 to produce circularized DNMT1 (cir cDNMT1) (Figur e 1 B). The cir cDNMT1 over expr ession plasmid and negati v e control plasmid were transfected into HeLa cells. Sanger sequencing re v ealed that the target RNA was circularized at the 'U' residue in the cir-cDNMT1 natural splice junction (Figure 2 A). qRT-PCR showed that the e xpression le v el of circDNMT1 was approximately 100 times higher than that of the negati v e contr ol gr oup in both the RNase R-treated group and the nontr eated group (Figur e 2 B), and the cir cularization system achie v ed a 300-600-fold ov ere xpression of circDNMT1 compared to the negati v e contr ol gr oup in HEK293T and MCF-7 cells (Figure 2 C). Furthermore, the splicing efficiency of the RNA circularization method based on the T. thermophila ribozyme was a pproximatel y 80% in vitro (Figure 2 D). Similar to the results of in vitro , T. thermophila ribozyme-based RNA circularization system realized a high splicing efficiency of close to 80% in HEK293T cells (Figure 2 E). CircDNMT1 e xpression le v els using the flanking ICS-mediated circularization method only increased 4-5-fold compared to those of the negati v e control group (Figure 2 F).

CDR1as
(hsa circ 0001946) , circFOXO3 (hsa circ 0006404), and circHIPK3 (hsa circ 0000284) were used to determine whether the T. thermophila ribozyme-based RNA circularization system was suitable for general cir cRNA over expr ession. To avoid adding an extra uridine residue ('U') at the 5' splice site, we adjusted the IGS sequence to correctly hybridize the 5' and 3' terminal sequences of the rearranged circRNA sequence to form the P1 and P10 helices (Figure 3 A-C, right). This adjustment causes circRNA to be spliced at the artificial site rather than the natural splicing site of cir cRNA (Figur e 3 A-C, left) but perfectly mimicked the nati v e sequence of the circRNA.
Di v ergent primers were designed for detecting circRNA. Convergent primers were designed for detecting total RNA including the forms of pr ecursor RNA and cir cRNA. Linear primers were designed for detecting linear RNA that r epr esented the RNA form that is not circularized in the pr ecursor RNA (Figur e 3 D, upper panel). Negati v e control with the empty plasmid, linear expression constructs and ribozyme-based cir cular expr ession constructs wer e transfected into HEK293T cells. The endogenous or ov ere xpr essed cir cRNAs in HEK293T cells were detected in all groups by RT-PCR (Supplementary Figure S1B). The results showed that our ribozyme-based circular constructs strongly enhanced the production of cir cRNA compar ed with endogenous le v els. Furthermor e, compar ed with the linear expression constructs, the circular expression constructs achie v ed higher circRNA e xpression in the case of plasmid transfection of the same quality ( The results showed that the T. thermophila ribozyme-based RNA circularization system had overw helming ad vantages over the commonly used flanking ICS method.

CircRNAs synthesized by T. thermophila ribozyme-based RNA circularization system performed biological functions consistent with those of natural noncoding circRNAs
Both DU145 and PC-3 cells were transfected with circ-FOXO3 linear and circular expression constructs separately to observe the intracellular function of the T. thermophila ribozyme-based circularization system (Figure 5 A). As expected, the high expression of circFOXO3 suppressed the viability (Figure 5   CDR1as, which is associated with the occurrence of various cancers, has more than 60 binding sites for miR-7 . It has been suggested that CDR1as may function as a miR-7 sponge ( 27 ). We investigated whether CDR1as maintains regulation of miR-7 function when ov ere xpressed using the T. thermophila ribozyme-based c ircularization system in cells expressing miR-7 ( 28 ). The CCK8 assay, transwell migration assay, and colony formation assay showed that CDR1as ov ere xpression reduced miR-7 tumor suppressi v e function (Supplementary Figure S2), which was in line with similar observ ations b y Weng et al. ( 28 ), highlighting that the system successfully mimics nati v e CDR1as. Howe v er, it should be noted that it has been shown that CDR1as is not expressed in the cancer cells in colon tumors from patients ( 29 ).

T. Thermophila ribozyme-based RNA circularization system promoted RNA circularization in vitro and performed the same protein expression function as mRNA
To verify whether circRNA could be used as a tool for protein translation based on the T. thermophila ribozymebased circularization system, a gene reporter containing two GFP fragments in a re v ersed or der was engineered. CVB3 IRES from previous research was engineered upstream of the start codon of GFP to dri v e cap-independent protein synthesis (Figure 6 A) ( 30 ). A schematic represen-tation of the ribozyme and split GFP complex is shown in Figure 6 B. The circulariza tion ef ficiency of the circular sequence encoding GFP was significantly higher than that in the negati v e contr ol gr oup (Figure 6 C). We performed the time-course experiment to re v eal the in vitro splicing process with northern blot analysis, as shown in Figure  6 D. Under the optimized splicing conditions as reported, we compared the multiple circularization conditions (55 • C for 0 min, 15 min and 30 min, respecti v ely). The results showed that most transcribed RNAs had been circularized during in vitro transcription, suggesting that the transcription initiated splicing and that a further 15-min incubation was sufficient after the transcription, which is consistent with the reaction conditions in most studies ( 26 , 31 ). Similar ly, we tr ansfected the CVB3IRES-GFP ov ere xpression plasmid into HEK293T cells and used NB to compare these two methods. The ratio of cir cRNA to pr ecursor linear RNA of the T. thermophila ribozyme was much higher than that of the classic flanking ICS method, suggesting higher splicing efficiency of the ribozyme system ( Figure  6 E). GFP fluorescence was observed from the transfection of circRN A formed, w hich proved that the CVB3 IRES and split GFP sequences could be joined together by backsplicing (Figure 6 F). Moreover, 5-methylcytosine and 2'-O-methyladenosine were introduced into synthetic circR-NAs, demonstra ting tha t nucleoside modifica tions would not disrupt the production of circRNA (Supplementary Figure S3). The product with the T7 promoter was amplified from the circular or linear expression construct by RT-PCR using specific primers. The PCR products were transcribed in vitro using T7 RN A pol ymerase according to the manufacturer's instructions. The splicing efficiency of circDNMT1 with the linear or cir cular expr ession construct was calculated. The linear expression construct does not contain the ribozyme structure. The value was calculated by dividing the circular DNMT1 expression le v el by the expression le v el of all DNMT1 products (circular DNMT1 ÷ total DNMT1 ). ( E ) After transfection of equal amounts of circular and linear expression constructs into HeLa and HEK293T cells, the splicing efficiency was analyzed by the same method. ( F ) The relati v e e xpression of circDNMT1 in HEK293T and HeLa cells using the flanking ICS-mediated circularization method. The values of all data shown are the means of three biological replicates. Error bars r epr esent the standard devia tions. Sta tistical significance was assessed using Student's t test (* P < 0.05, ** P < 0.01, *** P < 0.001).

DISCUSSION
Efficient and precise preparation of circRNAs is important for the functional study of circular mRNA pharmaceutics in the future. Previous strategies for circRNA synthesis did not meet the r equir ements due to their inefficiency and the introduction of additional nucleotides. Ther efor e, it is necessary to further simplify the procedure and improve production efficiency to achie v e industrial-scale production. In our current study, the T. thermophila ribozyme-based circularization system achie v ed a circulariza tion ef ficiency of 80% in vitro and in mammalian cells, and the circRNA produced by this system was completely consistent with the designed sequence without introducing any additional nu-cleotides. Due to the complexity of the intracellular environment in mammalian cells, the splicing efficiency of ribozyme varies with gene sequence and cell type. In recent years, the ribozyme deri v ed from the T. thermophila group I intron has been widel y a pplied in gene deli v ery, mRNA repair, and mRNA detection ( 32 , 33 ). Lieshout et al. previously used the Tetrahymena ribozyme to synthesize long circular RNA in E. coli based on the PIE strategy, and this circular transcript could be translated into an acti v e ␤-glucosidase ( 34 ). Howe v er, due to suboptimal assemb ly of the ribozyme catalytic domain in the PIE strategy, intron truncation may lead to a lower splicing efficiency. In our present study, the new system inserts the target RNA  sequence downstream of the optimized ribozyme, which is simple to perform and retains the complete ribozyme sequence and high splicing efficiency. Second, our system simulated three endogenous circRNAs and verified their biological functions in mammalian cells. The results showed that this system could spontaneously catalyze the synthesis of endogenous circRNA in mammalian cells and maintain its original biological function.
Because of its covalently closed circular structure, circular mRNA is more resistant to endonuclease degradation and more stable than linear RNA ( 35 ), thus producing higher le v els of and longer-lasting proteins ( 15 ). Therefor e, engineering cir cRNA for translation as mRNA may be an effecti v e protein expression tool that can be applied to clinical disease treatment. The products transcribed by our circularization system can be ra pidl y self-catal yzed in vitro to produce a large number of pr ecise cir cRNAs as protein ov ere xpression v ectors through simplified reaction and purification processes. We used the GFP reporter system to further verify the possibility of our RNA circularization system acting as an mRNA for translation. The coding region of GFP was divided into two parts: the first part containing the initiation codon was preceded by an IRES se-quence, which could act as an mRNA cap structure in mammalian cells. The second part of GFP was located in front of the IRES and behind the self-splicing ribozyme. Without the introduction of additional nucleotides, the coding sequence of GFP will not frameshift and green fluorescence will be expressed under the mediation of IRES. For any cir-cRNA synthesized using the ribozyme described here, it is necessary to pre v ent the end-joining sequence from becoming a stop codon to ensure the translation of the full-length circRN A. Recentl y, it has been reported that exo genous cir-cRNA was pr epar ed by the PIE strategy in vitro to produce neutralizing antibodies against SARS-CoV-2 ( 31 ). In contr ast, the circRNA gener ated by our ribozyme circularization system does not contain redundant sequences, which may reduce unexpected deleterious effects in vivo .
Recent r esear ch has indica ted tha t the nucleotide modification of circRNAs attenuates the innate immune response and may enhance the future applicability of circR-NAs in treatment (36)(37)(38). Therefore, nucleoside modification should be incorporated into the production strategy of circRNA production. In our current study, we also demonstra ted tha t nucleoside modifica tions did not af fect the circulariza tion ef ficiency of circRN A production, w hich may Figure 5. Ov ere xpression of circFOXO3 using our strategy regulated cell prolifer ation, migr ation, and apoptosis of prostate cancer cells. ( A ) The relati v e e xpression le v els of circFOXO3 in DU145 and PC3 cells transfected with negati v e control (pCDH v ector), linear and circular e xpr ession constructs wer e analyzed by qRT-PCR. Statistical significance was assessed using one-way ANOVA (* P < 0.05, ** P < 0.01, *** P < 0.001). ( B ) Cell viability was measured using Cell Counting Kit-8 (CCK-8) assays. Statistical significance was assessed using one-way ANOVA (* P < 0.05, ** P < 0.01, *** P < 0.001). ( C ) CCK-8 assays were performed to examine the proliferation ability of DU145 and PC3 cells. Statistical significance was assessed using Student's t test (* P < 0.05, ** P < 0.01, *** P < 0.001). ( D ) The transfected DU145 and PC3 cells were subjected to clone f ormation assa ys f or 14 da ys. Left, typical images of clone forma tion. Right, quantifica tion of clone formation re v ealing that circFOXO3 ov ere xpression suppressed cell proliferation. The number of cell clones was normalized to the number of cells counted at the endpoint. Statistical significance was assessed using one-way ANOVA (* P < 0.05, ** P < 0.01, *** P < 0.001). ( E ) The transfected DU145 and PC3 cells were subjected to tr answell migr ation assays for 24 h. Left, typical images showing typical cell migration after 24 h. Right, quantification of transwell migration assays showing that the expression of circFOXO3 inhibited cell migration. The number of migrating cells was normalized to the number of cells counted at the endpoint. Statistical significance was assessed using one-way ANOVA (* P < 0.05, ** P < 0.01, *** P < 0.001). ( F ) Wound healing assay showing that ov ere xpression of circFOXO3 inhibited the migration of DU145 and PC3 cells. ( G ) Western blots showing that the ov ere xpression of circFOXO3 increased the expression of apoptosis-related proteins induced by cisplatin in DU145 and PC3 cells. All data are presented as the means of three biological replicates. Error bars represent the standard deviations.