Circular RNA circSamd4a Regulates Antiviral Immunity in Teleost Fish by Upregulating STING through Sponging miR-29a-3p Free

As the host’s first line of defense against pathogen invasion, the innate immune system plays an important role in rapidly identifying viral infections and establishing an effective host defense mechanism. This process relies on the synergy of multiple pattern-recognition receptors, such as TLRs, NOD-like receptors (NLRs), and retinoid-inducible gene I (RIG-I)–like receptors (1–3). All of them can be used as viral nucleic acid sensors to recognize virus-associated pathogen-related molecular patterns and transmit signals downstream to activate innate immune responses. RIG-I–like receptors (RLRs) including RIG-I, MDA5, and LGP2 can be used as cytoplasmic sensors for viral RNA recognition (4, 5). They can recruit the downstream adaptor protein mitochondrial antiviral signaling proteins (MAVS) through the CARD–CARD domain to drive the expression of inflammatory cytokines and type I IFN (5–7). The occurrence of this process usually requires MAVS to interact with another adaptor protein that is the stimulation factor of IFN gene (STING; also known as MYPS, ERIS, or MITA); it exists on the endoplasmic reticulum and Golgi apparatus and then activates the atypical IKK family member TANK-binding kinase 1 (TBK1), which promotes interference regulation factor 3 (IRF3) phosphorylation and induces type I IFN production (7–10). Recent studies have also found that STING plays an important role in coping with various viral infections and inducing type I IFN in teleost fish infected with viruses (11, 12). Therefore, it is necessary to further explore the regulatory mechanism of STING-mediated signal transduction on the antiviral immune response in teleost fish.

Up to 80% of host gene transcription products are identified as noncoding RNA (ncRNA), including microRNA (miRNA), long ncRNA (lncRNA), and circular RNA (circRNA) (13). Among them, the existence of circRNA was discovered as early as the 1970s (14). Primitively, it is considered a nonfunctional junk product of pre-mRNA misconnection and has been largely ignored (15). However, with the development of high-throughput sequencing technology, a large number of circRNAs have been found to be widely expressed in eukaryotic cells (16). Unlike linear RNAs containing typical 5′-cap and 3′-poly(A) tails, circRNA is a covalently closed loop structure formed by connecting the 5′ and 3′ ends of linear RNA (15). This highly stable structural feature enables it to resist the degradation of endogenous RNase R and to stably exist in the cytoplasm. Existing studies have confirmed that circRNA can be widely involved in various cell biological processes, such as proliferation, invasion, and migration (17, 18). In addition, recent studies have shown that circRNA also plays a regulatory role in viral infection and immune response (19). It has been determined that circRNA functions mainly through the following ways: acting as an miRNA sponge (20), regulating gene expression at the transcription and splicing level (21), and interacting with RNA-binding protein (22), and part of circRNA is translated into proteins (23). However, the above studies are all carried out in mammals, and the function of circRNA in teleost fish is still little known.

miRNAs are small ncRNAs with 18–25 nucleotides that bind to the 3′-untranslated region (3′UTR) of mRNA to regulate the expression of target genes, leading to degradation of target transcripts or inhibition of protein translation (24). The current research indicates that miRNAs play a role in the regulation of a variety of biological processes, including proliferation, development, and immunity (25–27), although the regulatory mechanism involved by miRNA has been widely reported. However, recent studies have found that miRNAs are capable of binding to other types of RNA via miRNA response elements (MREs) (28). These MRE-containing RNAs are called “miRNA sponges,” including mRNA-3′UTR, lncRNA, pseudogenes, and circRNA (29). There is a growing body of evidence that miRNA can be combined by circRNA to form a circRNA–miRNA gene regulatory axis and compete with mRNA to bind miRNA to form a competitive endogenous RNA (ceRNA) regulatory network (30). However, there are few reports on the ceRNA regulation mechanism of circRNA in fish. Therefore, we need to further explore the function of ceRNA network in teleost fish.

As an important initial link in the evolution of vertebrates, fish have both natural and adaptive immune systems (31). Given that the natural immune-related molecular composition and functions in fish are quite similar to those of higher vertebrates, when the host is invaded by a pathogen, a cascade of signals is triggered to ward off the invading pathogen (31). Therefore, fish can be used as a good biological model for immunological research. However, the survival of fish is often attacked by a variety of pathogenic bacteria. Studies have increasingly showed that the infection of viral diseases is the main cause of fish death. As a typical rhabdovirus, Sininiperca chuatsi rhabdovirus has the characteristics of enveloped and ssRNA virus (32). After fish infection, it causes hemorrhagic septicemia and death, causing huge economic losses to aquaculture industry. Therefore, it is necessary to further study the antiviral signaling pathway in fish.

In this study, we discovered a ceRNA regulatory loop that is involved in the innate antiviral response of the teleost miiuy croaker (Miichthys miiuy). More precisely, the expression of circRNA Samd4a (circSamd4a) was upregulated after S. chuatsi rhabdovirus infection, which could act as a molecular sponge of miR-29a-3p and restore the inhibitory effect of miR-29a-3p on STING gene, thus enhancing the innate antiviral immune response of the host. Therefore, circSamd4a plays a positive role in regulating the innate immune response of teleost fish to viral infections, thereby blocking the escape of viral immunity. Our research results not only clarify the biological mechanism of the circRNA–miRNA–mRNA axis in fish antiviral immune response but also provide a reference for the study of immune regulation in lower vertebrates.

Miiuy croaker (∼50 g) was obtained from Zhoushan Fisheries Research Institute, Zhoushan, Zhejiang Province, China. Fish was acclimated in aerated seawater tanks at 25°C for 6 wk before experiments. Experimental procedures and S. chuatsi rhabdovirus infection were performed as described (33). All animal experimental procedures were performed in accordance with the National Institutes of Health’s Guide for the Care and Use of Laboratory Animals, and the experimental protocols were approved by the Research Ethics Committee of Shanghai Ocean University (no. SHOU-DW-2018-047).

M. miiuy spleen cells (MspC), M. miiuy kidney cells (MKC), M. miiuy muscle cells, M. miiuy gonadal cells, M. miiuy brain cells (MBrC), and M. miiuy intestine cells (MIC) were cultured in L-15 medium (HyClone) supplemented with 15% FBS (Life Technologies), 100 U/ml penicillin, and 100 μg/ml streptomycin at 26°C. Epithelioma papulosum cyprini (EPC) cells were maintained in medium 199 (Invitrogen) supplemented with 10% FBS, 100 U/ml penicillin, and 100 mg/ml streptomycin at 28°C in 5% CO₂. For stimulation experiments, MBrC and MIC were challenged with S. chuatsi rhabdovirus at a multiplicity of infection (MOI) of 5 and harvested at different times for RNA extraction. S. chuatsi rhabdovirus was isolated as described, and the replication of S. chuatsi rhabdovirus was detected by quantitative PCR (qPCR).

miRNAs capable of binding circSamd4a were predicted using miRanda (34), TargetScan (35), and RNAhybrid (36). Then, TargetScan, miRanda, and miRInspector (37) were used to predict the targets for miR-29a-3p. The following filtering restrictions were used: 1) total energy < 20 kcal/mol, and total score ≥ 140; 2) minimum free energy ≤ −25 kcal/mol; and 3) number of estimated binding sites ≥ 1. We considered the intersection of the three database predictions.

To construct the STING-3′UTR reporter vector, the 3′UTR region of M. miiuy STING gene as well as Sciaenops ocellatus and Nibea diacanthus STING-3′UTR were amplified using PCR and cloned into pmirGLO luciferase reporter vector (Promega). Meanwhile, the sequences of M. miiuy STING-3′UTR were inserted into mVenus-C1 vector (Invitrogen), which included the sequence of enhanced GFP. To construct the STING expression plasmid of M. miiuy, the full length of coding sequence region and 3′UTR of M. miiuy STING gene were amplified by specific primer pairs and cloned into pcDNA3.1 vector (Invitrogen). To construct circSamd4a overexpression vector, the full-length circSamd4a cDNA was amplified by specific primer pairs and cloned into pLC5-circ vector (Geneseed Biotech), which contained a front and back circular frame to promote RNA circularization. Also, the circSamd4a overexpression vectors of S. ocellatus and N. diacanthus were constructed by synthesizing the full-length circSamd4a cDNA of S. ocellatus and N. diacanthus, respectively. The empty vector with no circSamd4a sequence was used as negative control (NC). The mutated forms with point mutations in the miR-29a-3p binding site were synthesized using Mut Express II Fast Mutagenesis Kit V2 with specific primers. The miR-29a-3p sensor was created by inserting two consecutive miR-29a-3p complementary sequences into psiCHECK vector (Promega). The correct construction of the plasmids was verified by Sanger sequencing and extracted through EndoFree Plasmid DNA Miniprep Kit (Tiangen Biotech). To build pLC5-MS2, the MS2 fragment was inserted into the pLC5-circ vector, and then the MS2 sequence was inserted into any position in the circSamd4a sequence in the pLC5-circSamd4a vector, except for the binding site of miR-29a-3p. The sequences of all primers are listed in Supplemental Table I.

The miR-29a-3p mimics are synthetic dsRNAs with stimulating naturally occurring mature miRNAs. The miR-29a-3p mimics sequence was 5′- UAGCACCAUUUGAAAUCGGUUA-3′. The miR-29a-3p mimics mutant sequence was 5′- TGTAGAAGTTTGAAATCGGTTA-3′. The NC mimics sequence was 5′-UUCUCCGAACGUGUCACGUTT-3′. miRNA inhibitors are synthetic ssRNAs that sequester intracellular miRNAs and block their activity in the RNA interfering pathway. The miR-29a-3p inhibitor sequence was 5′-UAACCGAUUUCAAAUGGUGCUA-3′. The NC inhibitors sequence was 5′-CAGUACUUUUGUGUAGUACAA-3′. The small interfering RNA (siRNA) for circSamd4a are as follows: si-circSamd4a-1 (si-circ-1) sequence was 5′- GGCUGAUCACCCUGTACUCGCTT-3′, and si-circ-2 sequence was 5′- CAUUGGCUGAUCACCCCUGGUTT-3′. The scrambled control RNA sequences were 5′- GGCUGAUCACCCCUGGTGUUATT-3′.

Transient transfection of cells with miRNA mimic, miRNA inhibitor, or siRNA was performed in 24-well plates using Lipofectamine RNAiMAX (Invitrogen), and cells were transfected with DNA plasmids performed using Lipofectamine 3000 (Invitrogen) according to the manufacturer’s instructions. For functional analyses, the overexpression plasmid (500 ng per well) or control vector (500 ng per well) and miRNA mimics (100 nM), miRNA inhibitor (100 nM), or siRNA (100 nM) were transfected into cells in culture medium and then harvested for further detection. For luciferase experiments, miRNA mimics (100 nM) or miRNA inhibitor (100 nM) and pmirGLO (500 ng/well) containing the wild or mutated plasmid of STING-3′UTR were transfected into cells.

For the isolation and purification of both cytoplasmic and nuclear RNA from MIC, the Cytoplasmic & Nuclear RNA Purification Kit has been used according to the manufacturer’s instructions (Norgen Biotek). Total RNA was isolated with TRIzol Reagent (Invitrogen), and the cDNA was synthesized using the FastQuant RT Kit (Tiangen), which includes DNase treatment of RNA to eliminate genomic contamination. The expression patterns of each gene were performed by using SYBR Premix Ex Taq (Takara). The small RNA was extracted by using miRcute miRNA Isolation Kit (Tiangen), and miRcute miRNA FirstStrand cDNA Synthesis Kit (Tiangen) was applied to reverse transcription of miRNAs. The expression analysis of miR-29a-3p was executed by using the miRcute miRNA qPCR Detection Kit (Tiangen). Real-time PCR was performed in an Applied Biosystems QuantStudio 3 (Thermo Fisher Scientific). GAPDH and 5.8S rRNA were employed as endogenous controls for mRNA and miRNA, respectively. Primer sequences are displayed in Supplemental Table I.

The wild-type of circSamd4a and the mutant devoid of miR-29a-3p binding site were cotransfected with miR-29a-3p mimics into EPC cells. At 48 h posttransfection, reporter luciferase activities were measured using the dual-luciferase reporter assay system (Promega). To determine the functional regulation of circSamd4a, cells were cotransfected STING overexpression plasmid or circSamd4a overexpression plasmid together with NF-κB, IL-1β, IFN-1, and IRF3 luciferase reporter gene plasmids (33); phRL-TK Renilla luciferase plasmid; and either miR-29a-3p mimics or NCs. At 48 h posttransfection, the cells were lysed for reporter activity using the dual-luciferase reporter assay system (Promega). miR-29a-3p sensor was cotransfected with miR-29a-3p mimics or circSamd4a overexpression plasmid. At 48 h posttransfection, the cells were lysed for reporter activity. All the luciferase activity values were achieved against the Renilla luciferase control. Transfection of each construct was performed in triplicate in each assay. Ratios of Renilla luciferase readings to firefly luciferase readings were taken for each experiment, and triplicates were averaged.

Cellular lysates were generated by using 1 × SDS-PAGE loading buffer. Proteins were extracted from cells and measured with the BCA Protein Assay kit (Vazyme Biotech), then subjected to SDS-PAGE (8%) gel and transferred to PVDF (MilliporeSigma) membranes by semidry blotting (Bio-Rad Trans Blot Turbo System). The membranes were blocked with 5% BSA. Protein was blotted with different Abs. The Ab against STING was diluted at 1:500 (Abcam), anti-Flag and anti-Tubulin mAb were diluted at 1:2000 (Sigma-Aldrich), and HRP-conjugated anti-rabbit IgG or anti-mouse IgG (Abbkine) at 1:5000. The results were the representative of three independent experiments. The immunoreactive proteins were detected by using WesternBright ECL (Advansta). The digital imaging was performed with a cold charge-coupled device camera.

For RNase treatment, the RNAs (10 μg) from MBrC and MIC were treated with RNase R (3 U/μg; EpiCenter) and incubated for 30 min at 37°C. For actinomycin D treatment, MBrC and MIC were treated with 5 μg/ml actinomycin D (Sigma-Aldrich) and collected in a series of time intervals. Then, the expression of circSamd4a and the linear mRNA was detected by quantitative RT-PCR.

The cDNA and genomic DNA (gDNA) PCR products were investigated using 2% agarose gel electrophoresis with Tris acetate–EDTA running buffer. DNA was separated by electrophoresis at 100 V for 30 min. The DNA marker was Super DNA Marker (100–10,000 bp) (Comwin-Biotech). The bands were examined by UV irradiation.

The MS2 fragment was cloned into plasmid pLC5-circ vector and plasmid pLC5-circSamd4a to construct pLC5-MS2 and pLC5-MS2-circSamd4a plasmids that could produce MS2 protein. pLC5-MS2-GFP (Addgene) can be combined with MS2 fragment and can express GFP-MS2 fusion protein. The pLC5-MS2, pLC5-MS2-circSamd4a, and pLC5-MS2-GFP (Addgene) plasmids were cotransfected into MBrC. After 24 h of transfection, the nuclei were stained blue with DAPI. Because the expression of GFP protein produces green fluorescence, we can observe the localization of circSamd4a.

circSamd4a and circSamd4a–sequence mutated in the miR-29a binding site (mut) with miR-29a-3p binding sites mutated were transcribed in vitro. The two transcripts were biotin labeled with the T7 RNA polymerase and Biotin RNA Labeling Mix (Roche Diagnostics), treated with RNase-free DNase I, and purified with the RNeasy Mini Kit (QIAGEN). The whole-cell lysates from MBrC (∼1.0 × 10⁷) were incubated with purified biotinylated transcripts for 1 h at 25°C. The complexes were isolated by streptavidin agarose beads (Invitrogen). RNA was extracted from the remaining beads, and qPCR was used to evaluate the expression levels of miRNAs.

To conduct pulldown assay with biotinylated miRNA, MBrC was harvested at 48 h after transfection, then incubated on ice for 30 min in lysis buffer (20 mM Tris [pH 7.5], 200 mM NaCl, 2.5 mM MgCl₂, 1 mM DTT, 60 U/ml Superase-In, 0.05% IGEPAL, and protease inhibitors). The lysates were precleared by centrifugation for 5 min, and 50 μl of the sample was aliquoted for input. The remaining lysates were incubated with M-280 streptavidin magnetic beads (Sigma-Aldrich). To prevent nonspecific binding of RNA and protein complexes, the beads were coated with RNase-free BSA and yeast tRNA (both from Sigma-Aldrich). The beads were incubated for 4 h at 4°C and washed twice with ice-cold lysis buffer, three times with the low salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl [pH 8], and 150 mM NaCl), and once with the high-salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl [pH 8], and 500 mM NaCl). RNA was extracted from the remaining beads with TRIzol Reagent (Invitrogen) and evaluated by qPCR.

RNA immunoprecipitation (RIP) assay was performed by using the Magna RIP RNA-Binding Protein Immunoprecipitation Kit (MilliporeSigma) following the manufacturer’s protocol. The Argonaute (Ago)–RIP assay was conducted in MBrC (∼2.0 × 10⁷), transfected Ago2-flag, or pcDNA3.1-flag and miR-29a-3p mimics or control mimics. After 48 h transfection, the cell extract was incubated with magnetic beads conjugated with IgG and anti-Flag Ab (Sigma-Aldrich). RNA was extracted from the remaining beads, and qPCR was used to evaluate the expression levels of circSamd4a.

The MS2-RIP assay was also conducted in MBrC (∼2.0 × 10⁷) transfected with pLC5- MS2, pLC5-MS2-circSamd4a, pLC5-MS2-circSamd4a-mut, or pLC5-MS2-GFP (Addgene). pLC5-MS2-GFP plasmid can express GFP protein and can be detected by anti-GFP Ab (Abcam). After 48 h transfection, the MBrC were used in RIP assays via the Magna RIP RNA-Binding Protein Immunoprecipitation Kit (MilliporeSigma) and an anti-GFP Ab following the manufacturer’s protocol. RNA was extracted from the remaining beads, and qPCR was used to evaluate the expression levels of miRNAs.

Cell viability was measured at 48 h after transfection in MBrC and MIC with CellTiter-Glo Luminescent Cell Viability assays (Promega) according to the manufacturer’s instructions. The 5-ethynyl-2′-deoxyuridine (EdU) assay was performed to assess the proliferation of cells by using BeyoClick EdU cell Proliferation Kit with Alexa Fluor 555 (Beyotime Biotechnology) following the manufacturer’s instructions. The EdU cell lines were photographed and counted under a Leica DMiL8 fluorescence microscope and evaluated by Thermo Scientific Varioskan LUX. These experiments were repeated for three times.

Data are expressed as the mean ± SE from at least three independent experiments’ replicates. Student t test was used to evaluate the data. The relative gene expression data were acquired using the 2^−ΔΔCT method, and comparisons between groups were analyzed by one-way ANOVA, followed by Duncan multiple comparison tests (38). A p value <0.05 was considered significant.

In recent years, circRNA has been found to play a key role in the antiviral immune response of mammals (16). However, the role of circRNA in the innate immune response of lower vertebrates is still unclear. In previous studies, we used RNA sequencing data to detect the expression level of circRNA after S. chuatsi rhabdovirus infection (39). Among them, circSamd4a, as one of the differentially expressed circRNA (Supplemental Table II), was significantly upregulated after S. chuatsi rhabdovirus infection. To confirm the reliability of this result, we conducted in vivo and in vitro experiments to detect the changes in the expression level of circSamd4 under S. chuatsi rhabdovirus stimulation. Considering that circRNA is spliced from linear RNA, we also tested the expression of Samd4a. The results of qPCR experiments showed that compared with linear Samd4a, circSamd4a was significantly upregulated in miiuy croaker spleen tissue and MBrC treated with S. chuatsi rhabdovirus at different time points (Fig. 1A, 1B).

The transcriptome sequencing revealed that circSamd4a was 598 bp in length. Using miiuy croaker whole-genome library to perform blast analysis on the Samd4a gene, it was found that Samd4a gene was located on chromosome 2, which was composed of 10 exons, and circSamd4a was formed by self-cyclization of exon 2. First, circSamd4a divergent primers were designed for RT-PCR amplification, and the amplified products were subjected to Sanger sequencing to confirm that circSamd4a was spliced from the head to the tail (Fig. 1C). We assessed the expression levels of circSamd4a in MSpC, MIC, M. miiuy muscle cells, MBrC, M. miiuy gonadal cells, and MKC (Fig. 1D). The expression levels of circSamd4a in MBrC and MIC were determined as high and low, respectively. Therefore, we chose these two cell lines to investigate the function and regulatory mechanism of circSamd4a. Next, we performed circSamd4a identification in MBrC and MIC, respectively, using the cDNA and gDNA of MBrC and MIC lines as templates, using divergent primers to amplify circSamd4a gene, convergent primers to amplify Samd4a, and agarose gel electrophoresis to assay RT-PCR–amplified products. The results in (Fig. 1E show that only the divergent primers amplified circSamd4a (an expected 158-bp fragment) from the cDNA, but no amplified product was found from the gDNA. Given that the stability is considered to be a key feature of circRNAs, we thus performed RNase R digestion employed to confirm the cyclization properties of circSamd4a. The results from the analysis of RT-PCR and agarose gel electrophoresis assay showed that circSamd4a was significantly resistant to RNase R digestion, whereas the linear mRNAs of Samd4a and GAPDH were degraded by RNase R (Fig. 1F). Moreover, the half-time of the circSamd4a transcript was significantly longer than Samd4a mRNA after treatment with actinomycin D, which suppressed RNA transcription (Fig. 1G). In addition, we analyzed the subcellular distribution of circSamd4a mainly confined to the cytoplasm (Fig. 1H, 1I), indicating that circSamd4a mainly plays a regulatory role at the posttranscriptional level. Accordingly, all data indicated that circSamd4a is a circRNA with important functions stably expressed in different miiuy croaker cell lines.

We designed the two siRNA against circSamd4a, and the overexpression plasmid of circSamd4a (oe-circ) was constructed to explore the biological function of circSamd4a in S. chuatsi rhabdovirus–infected MBrC and MIC (Fig. 2A, 2B). We transfected si-circ-1/2 into MBrC, and the results showed that two siRNAs remarkably reduced the expression of circSamd4a but did not affect the expression level of linear Samd4a mRNA in MBrC. In addition, si-circ-1 can induce higher inhibitory efficiency; thus, we choose si-circ-1 for the subsequent experiment (Fig. 2C). Moreover, the oe-circ was successfully constructed because it significantly increased the circSamd4a expression levels rather than linear Samd4a mRNA in MIC (Fig. 2C). Considering that the IFN genes (ISGs) are important antiviral effectors, we focused on the study of the function of circSamd4a in regulating the expression of IFN, ISGs and inflammatory cytokines upon viral infection. As shown in (Fig. 2D, knockdown of circSamd4a could significantly decrease the expression levels of IFN-1, ISG15 and Mx1, and TNF-α and IL-1β in MBrC treated with S. chuatsi rhabdovirus; si-circraSamd4a-1 showed a time-dependent effect on the expression of inflammatory cytokine and ISG gene. Conversely, overexpression of circSamd4a increased the expression levels of these genes in MIC treated with S. chuatsi rhabdovirus (Fig. 2E). These results suggest that circSamd4a may play an important role in responding to RNA virus infection. Furthermore, to investigate the biological significance of circSamd4a in S. chuatsi rhabdovirus–induced host cells, we examined the effect of circSamd4a on S. chuatsi rhabdovirus replication. qPCR results showed that circSamd4a knockdown upregulated S. chuatsi rhabdovirus replication (Fig. 2F), whereas circSamd4a overexpression moderately decreased S. chuatsi rhabdovirus replication (Fig. 2G). These data suggest that circSamd4a regulates S. chuatsi rhabdovirus replication. Meanwhile, we sought to understand whether circSamd4a can affect the cell proliferation and viability following infection with S. chuatsi rhabdovirus. When we explored the effect of circSamd4a on cell viability using luminescent cell viability assay, we found knockdown of circSamd4a reduced cell viability upon S. chuatsi rhabdovirus infection (Fig. 2H), whereas overexpression of circSamd4a increased cell viability (Fig. 2I). We conducted EdU-555 assay to measure the cell proliferation in MBrC and MIC. Knockdown of circSamd4a led to considerably decreased percentages of EdU-positive cells, whereas overexpression of circSamd4a exhibited the opposite effect, indicating that circSamd4a promote the proliferation of miiuy croaker cell lines (Fig. 2J). These data suggest that circSamd4a is a positive regulator and is involved in regulating antiviral immunity by regulating inflammatory cytokines and antiviral genes as well as cell proliferation and viability.

Given that circSamd4a is abundant in the cytoplasm, most reports have suggested that circRNA has the function of acting as miRNA sponge. Therefore, next we probed the ability of circSamd4a to bind miRNAs. To this end, we performed RNA immunoprecipitation for Ago2 in MBrC by transfecting Ago2-flag or pcDNA3.1-Flag. The results of qPCR analysis showed that endogenous circSamd4a could be pulled down by Ago2-Flag (Fig. 3A); this suggests that circsamd4a may act as miRNAs sponge. To further search for miRNAs combined with circSamd4a, we selected six candidate miRNAs through overlapping the prediction results of miRNA recognition elements in circSamd4a sequence by miRanda, TargetScan, and RNAhybrid (Fig. 3B). Subsequently, we compared the expression levels of these candidate miRNAs by transfecting si-circ-1 or NC (si-NC) in MBrC and MIC transfected with circSamd4a overexpression plasmid or pLC5-ciR vector. Among the six candidate miRNAs, the abundance of miR-29a-3p was markedly enhanced to circSamd4a inhibition and significantly reduced upon overexpression of circSamd4a (Fig. 3C, 3D). To understand whether circSamd4a affects the activity of miR-29a-3p. We constructed an miR-29a-3p sensor and miR-15a-5p sensor (Fig. 3E), along with miR-29a-3p, plc5-circ vector, or circSamd4a overexpression plasmid cotransfected MIC. The reduction in luciferase activity caused by miR-29a-3p was restored when cotransfected with circSamd4a overexpression plasmid, suggesting that circSamd4a specifically sponges miR-29a-3p, thus preventing it from repressing luciferase activity (Fig. 3F). In conclusion, these data suggest that circSamd4a can regulate the expression and activity of miR-29a-3p, and circSamd4a may be a sponge of miR-29a-3p.

To investigate whether circSamd4a could directly bind miR-29a-3p, we analyzed the sequences of circSamd4a and noted that circSamd4a possessed a complementary sequence to miR-29a-3p seed region (Fig. 4A). Next, Luciferase reporters were constructed by inserting either the wild-type circSamd4a sequence (wt) or the mut into the pmirGLO vector. Luciferase assays showed that upregulation of miR-29a-3p could remarkably decrease the luciferase activities of the wild-type of circSamd4a report plasmid, but not the mutant (Fig. 4B). It was subsequently confirmed that miR-29a-3p mimics inhibited luciferase activity in dose-dependent ways (Fig. 4C). In addition, wild or a mutant type of circSamd4a sequence were inserted into mVenus-C1 vector and detected whether cotransfecting with miR-29a-3p could suppress the levels of GFP. The results revealed that miR-29a-3p could significantly inhibit the levels of GFP (Fig. 4D, 4E), which suggested that miR-29a-3p could directly interact with circSamd4a by complementary seed region.

Given that miRNAs inhibit translation and degrade mRNAs in Ago2-depedent manner by binding to their targets, we further tested the ability of circSamd4a to bind to miR-29a-3p. To this end, we conducted RIP assays in MBrC by cotransfecting Ago2-flag and miR-29a-3p or Ago2-flag and miR-15a-5p. The qPCR analysis results indicated that circSamd4a pulled down with Ago2-flag was markedly enriched when transfected with miR-29a-3p mimics, compared with controls and miR-15a-5p (Fig. 4F). To further confirm the direct interaction between circSamd4a and miR-29a-3p, we conducted RNA pulldown detected with biotin-labeled circSamd4a probe or biotin-labeled miR-29a-3p. The qPCR analysis results revealed that miR-29a-3p could be pulled down by biotin-labeled circSamd4a, but not circSamd4a-mut (Fig. 4G). Additionally, biotin-labeled miR-29a-3p captured more circSamd4a than the NC, whereas biotin-labeled mutant type of mir-29a-3p did not (Fig. 4H). Furthermore, we cloned an MS2 fragment into pLC5 vector, pLC5-circSamd4a, and pLC5-circSamd4a-mut plasmids to construct plasmids capable of producing MS2 protein. In addition, we cotransfected pLC5-MS2-GFP plasmid, which can produce GFP-MS2 fusion protein that could bind with the MS2 fragment and be identified by anti-GFP Ab. Hence, miRNAs interacting with circSamd4a could be pulled down by the GFP-MS2-circSamd4a compounds. The results from qPCR assays showed that miR-29a-3p was significantly enriched for pLC5-circSamd4a compared with pLC5-circSamd4a-mut or empty vector (Fig. 4I). Collectively, these data suggest that circSamd4a can directly bind to miR-29a-3p, and circSamd4a acts as sponge of miR-29a-3p.

Then, we attempted to explore the biological function of miR-29a-3p in host antiviral responses. To this end, we detected changes in the expression of miR-29a-3p in the spleen tissues treated with S. chuatsi rhabdovirus at different time points, and the results showed that S. chuatsi rhabdovirus infection upregulated the expression of miR-29a-3p (Fig. 5A). Therefore, miR-29a-3p may be involved in the host’s antiviral immune response. To further explore the regulatory role of miR-29a-3p in the production of antiviral genes and inflammatory cytokines, we first measured the effects of synthetic miR-29a-3p mimics and miR-29a-3p inhibitors on the expression of miR-29a-3p. As expected, miR-29a-3p mimics enhanced miR-29a-3p expression, whereas miR-29a-3p inhibitors decreased miR-29a-3p expression (Fig. 5B). Furthermore, the results indicated that inflammatory cytokines and antiviral genes, including TNF-α, IFN-1, ISG15, Mx1, and TNF-α, were significantly reduced by the introduction of miR-29a-3p mimics upon S. chuatsi rhabdovirus infection. On the contrary, miR-29a-3p inhibitor significantly enhanced these gene expression compared with transfection of control inhibitors (Fig. 5C). To investigate the biological significance of miR-29a-3p in S. chuatsi rhabdovirus–induced host cells, we tested the effect of miR-29a-3p on S. chuatsi rhabdovirus replication. As shown in (Fig. 5D and 5E, overexpression of miR-29a-3p increased the expression of S. chuatsi rhabdovirus mRNA in the intracellular and supernatant from the infected cells (Fig. 5D), and the opposite was true after miR-29a-3p was inhibited (Fig. 5E). Next, we tried to study whether the cell viability and proliferation ability after S. chuatsi rhabdovirus infection can be regulated by miR-29a-3p. Overexpression of miR-29a-3p reduced cell viability (Fig. 5F), whereas inhibition of miR-29a-3p increased cell viability (Fig. 5G). At the same time, when we investigated the effect of miR-29a-3p on cell proliferation upon S. chuatsi rhabdovirus infection, overexpression of miR-29a-3p after S. chuatsi rhabdovirus infection has an inhibitory effect on cell proliferation (Fig. 5H), whereas inhibition of miR-29a-3p expression leads to an effective increase in cell proliferation (Fig. 5I). Overall, these data indicate that the inducible miR-29a-3p can inhibit the antiviral response and promote S. chuatsi rhabdovirus replication while reducing cell viability and proliferation rate.

We predicted the potential target genes of miR-21-3p using miRNA prediction programs, including TargetScan, miRanda, and microInspector. Among all possible target genes, we discovered the gene STING that has been reported to be involved in innate antiviral responses (40). Therefore, we investigated whether STING is direct target gene of miR-29a-3p. By prediction analysis, STING contained the standard target sequence of miR-29a-3p at the nucleotide of its 3′UTR (Fig. 6A). To obtain direct evidence that STING-3′UTR is the target of miR-29a-3p, we integrated the STING-3′UTR containing the target sequence or target site mutation fragment into the luciferase reporter vector (Fig. 6A). We cotransfected luciferase reporter plasmid with miR-29a-3p mimics into EPC cells. As shown in (Fig. 6B, miR-29a-3p mimics significantly reduced the luciferase activity of transfected wild-type STING-3′UTR cells, but the luciferase activity of transfected mutant STING-3′UTR cells had no effect. Because mature miRNAs are spliced out from pre-miRNAs, we constructed pre–miR-29a plasmids. After cotransfecting of EPC cells with the luciferase reporter plasmid, we found that overexpression of premiR-29a also significantly reduced luciferase activity, whereas no change was observed in cells transfected with the mutant STING-3′-UTR (Fig. 6B). For further verification, wild-type or mutant STING-3′UTR was inserted into the mVenus-C1 vector to investigate whether miR-29a-3p mimics can inhibit the expression of GFP. As shown in (Fig. 6C and 6D, compared with the transfection control mimic, miR-29a-3p can significantly suppress GFP expression (Fig. 6C), whereas the fluorescence intensity of the cells transfected with the mutant STING-3′UTR remains unchanged (Fig. 6D). These results clearly indicate that miR-29a-3p can directly target STING. Given that miRNAs regulate the expression of target genes through mRNA degradation or translational repression, we transfected MIC with miR-29a-3p mimics and inhibitors, respectively. The results of Western blotting and qPCR assays showed that miR-29a-3p mimics reduced the expression level of STING (Fig. 6E). In contrast, miR-29a-3p inhibitor significantly increased the expression level of STING (Fig. 6F). Considering that miR-29a-3p targets STING and regulates its expression, we wanted to test whether STING-mediated activation of NF-κB and IRF3 signaling pathways is regulated by miR-29a-3p. The results showed that the activation of NF-κB, IL-1β, IFN-1, and IRF3 luciferase reporter genes was impeded by miR-29a-3p mimics (Fig. 6G). Taken together, these results suggest that STING is a direct target of mir-29a-3p, which is involved in the regulation of antiviral responses by posttranscriptional regulation of STING expression.

Given that circSamd4a could interact with miR-29a-3p and miR-29a-3p targets STING and regulates its expression, we tested whether circSamd4a is able to regulate STING. Transfection of the overexpression plasmids of circSamd4a into MIC significantly upregulated the protein level of STING, whereas knockdown of circSamd4a in MBrC markedly decreased STING protein level (Fig. 7A). Furthermore, overexpression of circSamd4a resulted in an increase in the mRNA expression level of STING in S. chuatsi rhabdovirus–treated cells, whereas knockdown of circSamd4a decreased the mRNA expression of STING (Fig. 7B). Next, we tested whether circSamd4a regulates STING expression via miR-29a-3p. To this end, we cotransfected the STING-3′UTR with miR-29a-3p, wild-type, and mutant circSamd4a expression plasmids, and the dual-luciferase assays showed that wild-type circSamd4a could counteract the inhibitory effect of miR-29a-3p on the STING-3′UTR, in contrast to mutant circSamd4a (Fig. 7C). Notably, circSamd4a could also counteract the effect of miR-29a-3p on STING protein level (Fig. 7D). These results suggest that circSamd4a regulates STING expression through miR-29a-3p. Given the miR-29a-3p and STING involvement in the regulation of NF-κB, IL-1β, IRF3, and IFN-1 luciferase reporters, we thus investigated the function of circSamd4a in the regulation of these reporters. The results showed that circSamd4a could counteract the negative effect of miR-29a-3p on the luciferase activity of NF-κB, IL-1β, IFN-1, and IRF3 luciferase reporters (Fig. 7E). These data suggest that circSamd4a is a ceRNA for miR-29a-3p to regulate STING expression. Furthermore, we tried to explore the effect of circSamd4a/miR-29a-3p regulatory loop on S. chuatsi rhabdovirus replication, and the results showed that we found that circSamd4a could counteract the promoting effect of miR-29a-3p on S. chuatsi rhabdovirus replication (Fig. 7F). Meanwhile, we explored the regulatory of circSamd4a/miR-29a-3p loop on cell viability and cell proliferation after viral infection. The results showed that overexpression of circSamd4a could counteract the negative effects of miR-29a-3p on cell viability and cell proliferation after S. chuatsi rhabdovirus infection (Fig. 7G, 7H). Collectively, these data suggest that circSamd4a served as a ceRNA for miR-29a-3p to regulate STING expression.

To illustrate the generalizability of our findings, we first checked the sequence alignment of pre–miR-29a across different species. As shown in (Fig. 8A, mature miR-29a-3p shows high conservation from mammals to fish. In addition, the miR-29a-3p binding site in the STING-3′UTR also exhibited a high degree of conservation in fish (Fig. 8B). To obtain direct evidence that miR-29a-3p targets the STING-3′UTR in other species, we combined the constructed wild-type and mutant S. ocellatus and N. diacanthus STING-3′UTR dual-luciferase reporter genes and miR-29a-3p–cotransfected EPC cells. Notably, miR-29a-3p mimics significantly reduced the luciferase activity of cells transfected with wild-type S. ocellatus STING-3′UTR and N. diacanthus STING-3′UTR, but the cells transfected with its mutant Luciferase activity has no effect (Fig. 8C, 8D). In addition, we also verified whether the regulation of circSamd4a by miR-29a-3p exists in other species. First, we found that circSamd4a sequences are highly conserved in different fish species by comparing circSamd4a sequences among different species. In particular, miR-29a-3p of different species is also highly conserved in the binding sites of circSamd4a (Fig. 8E). Then, to examine whether circSamd4a from other species could also interact with miR-29a-3p, we constructed luciferase reporters of S. ocellatus and N. diacanthus circSamd4a with mutants with mutated binding sites for miR-29a-3p. Luciferase assays showed that miR-29a-3p suppressed the luciferase activity of the wild-type circSamd4a luciferase plasmid in both species but had no effect in the mutant (Fig. 8F). Furthermore, to test whether circSamd4a from S. ocellatus and N. diacanthus would affect the activity of miR-29a-3p, we performed luciferase assays and found that circSamd4a from S. ocellatus and N. diacanthus both counteracted the inhibitory effect of miR-29a-3p on the miR-29a-3p sensor (Fig. 8G). These results suggest that circSamd4a may act as an endogenous sponge RNA to interact with miR-29a-3p among different teleost (Fig. 9).

Viral infections often cause high mortality in aquatic animals and cause great economic losses to the aquaculture industry. Unlike mammals, which have a mature acquired immune system, the acquired immune system of teleost fish is not perfect and mainly relies on innate immune response to resist virus invasion (40). In teleost fish, there are sets of conserved immune-related genes and signal regulation networks to respond to virus invasion and maintain immune homeostasis (40). However, this conservative innate immune system usually requires different levels of regulatory factors and signal interaction networks to achieve antiviral effects. In this study, we report the circSamd4a–miR-29a-3p–STING axis, a ceRNA regulatory network related to the antiviral immune response. We found that circSamd4a can upregulate the expression of the antiviral gene STING by adsorbing miR-29a-3p, activate the NF-κB/IRF3 pathway, promote the production of inflammatory factors and antiviral genes, and enhance the antiviral immune response. Meanwhile, miR-29a-3p can be used as a negative regulator of antiviral immune response by targeting the expression of STING to inhibit the activation of the NF-κB/IRF3 pathway, thereby facilitating virus replication and escape. Therefore, the circSamd4a–miR-29a-3p–STING regulatory loop is conducive to maintaining the immune homeostasis of the organism and ensuring that the virus is eliminated while avoiding excessive immune responses. This discovery is of great significance for in-depth exploration of the ceRNA regulatory mechanism in maintaining the innate antiviral response and moderate inflammation in teleost fish.

In past studies, it has been confirmed that STING plays an important role in coping with various viral infections and in inducing the production of IFN-1. For example, STING-deficient mice are susceptible to lethal infections of HSV1 and vesicular stomatitis virus (41, 42). In grass carp reovirus– and infectious hematopoietic necrosis virus–infected teleost fish, STING participates in the antiviral immune response as a key mediator of the host’s innate immune response (11, 12). In view of the importance of STING in resisting viral infections, the antiviral signal transduction pathway mediated by STING has been deeply studied. Some proteins, such as UBXN3B and TRIM32, have been identified as positive regulators of STING-mediated signal transduction and are involved in the regulation of STING-mediated signal transduction pathways (42, 43). USP44 or USP20 interacts with STING to enhance the cellular antiviral response (44, 45). To facilitate its own survival, the virus has also evolved a mechanism to regulate the host’s STING-mediated signal transduction to escape the host’s innate immune system. For example, USP49 interacts with STING and downregulates the STING expression after HSV-1 infection, thereby reducing the cells antiviral response (46). Hepatitis C virus NS4B can reduce the production of IFNI by targeting STING and evading the immune system (47). In addition, some ncRNA regulators have also been reported in teleost fish, which are involved in the STING-mediated fish antiviral immune response (33, 40). In this study, we found that two regulatory factors miR-29a-3p and circSamd4a negatively and positively regulate STING, respectively, and thus play a key regulatory role in the host antiviral response after RNA virus infection in teleost.

The ceRNA hypothesis reveals that RNA transcripts including mRNA, lncRNA, pseudogenes, and circRNA can compete with each other and regulate miRNA expression by competing for shared MREs, thereby forming a complex posttranscriptional regulatory network (48). As the first reported circRNA to act as an endogenous competitive inhibitor, CiRS-7 participates in the occurrence of a variety of cancers by reducing the inhibitory effect of miR-7 on target mRNA (49). Subsequently, more and more studies have confirmed that the ceRNA mechanism is an important way for circRNA to exercise its biological functions. For example, the circRNA hsa_circ_0001368 serves as the ceRNA of miR-6506-5p to regulate the miR-6506-5p/FOXO3 axis and slow down the growth process of gastric cancer (50). circRNA_09505 can act as an miR-6089 sponge in macrophages through the IκBα/NF-κB signaling pathway, thereby promoting AKT1 expression and aggravating arthritis and inflammation in collagen-induced arthritis mice (51). Although the functional research on the ceRNA mechanism of circRNA in the antiviral immune response of teleost fish has just begun and there is a lack of infection models for most viruses in teleost fish, the S. chuatsi rhabdovirus has been found to be sensitive and pathogenic to miiuy croaker cells. In this study, we found that circSamd4a contains the binding site of miR-29a-3p and circSamd4a can be located in the cytoplasm, which is the basis for its biological function. Therefore, we hypothesized that circSamd4a may play a role in the innate immune response to S. chuatsi rhabdovirus infection through the ceRNA mechanism. Subsequently, dual-luciferase reporter gene experiments, RNA immunoprecipitation, and RNA pulldown experiments confirmed that circSamd4a can directly bind to miR-29a-3p. After confirming that circSmd4a can inhibit miR-29a-3p, further exploring its influence on the downstream antiviral gene STING is the key to ensuring that circSmd4a exerts an antiviral immune response through the ceRNA mechanism. Subsequent experiments fully confirmed that circSmd4a can be used as a sponge of miR-29a-3p to indirectly regulate the expression of STING, promote the production of antiviral genes, and resist virus invasion and replication. circRNA has high stability in vitro because of its special splice formation mode, and its plate decay period in vitro is much longer than most ncRNAs, which means that circSmd4a will have great potential to be developed into drugs with excellent regulatory effects. These results indicate that circSmd4a can be used as a new antiviral immunomodulatory molecule to participate in the regulation of the immune response of teleost fish, and it may also become a potential drug for the treatment of teleost fish viral diseases.

In the past decade, a large number of miRNAs have been discovered in teleost fish. These miRNAs are highly conserved compared with other species. This may be the basis for miRNAs to exist and function widely in different species. So far, many miRNAs have been reported that can act as key regulators of innate immune response in teleost fish, such as miR-731, miR-217, and miR-210; they regulate antiviral immune response by targeting antiviral genes (33, 52, 53). The latest research has found that miRNAs can be adsorbed by other ncRNAs and participate in the regulation of immune responses. For example, the inhibitory effect of fish miR-27c-3p on IRAK4 can be offset by lncRNA IRL to avoid excessive activation of the immune response (54). miR-15a-5p can be adsorbed by circular Dtx1 to maintain the stability of the antiviral response (55), which indicates that miRNA and ncRNA can construct a complex regulatory network in lower vertebrates. The key element of this regulatory network is the presence of different MREs on mRNA or ncRNA. Therefore, the same gene can be regulated by multiple miRNAs (56). In this study, we found that miR-29a-3p can negatively regulate the expression of STING, inhibit the antiviral immune response mediated by STING, and promote virus replication. This negative regulatory mechanism may be a survival strategy for RNA viruses to resist the host’s antiviral immune response. Interestingly, studies have shown that the regulatory role of miR-29a-5p in mammals is opposite to that of miR-29a-3p in teleost fish, and miR-29a-5p can targeted IRF7 and promoted the expression of STING (57).

In conclusion, in this study, we found a circRNA circSamd4a related to the antiviral immunity of bony fish, which can form a circSamd4a–miR-29a–STING axis with miR-29a-3p and STING to participate in the innate antiviral immunity of bony fish by regulation of immune response. Specifically, we found that miR-29a-3p can act as a negative regulator of STING and weaken the antiviral immune response mediated by STING. Further research found that circSamd4a can act as the ceRNA of miR-29a-3p and positively regulates the activation of STING-mediated IRF3/NF-κB pathway, thereby enhancing antiviral immune response and weakening virus replication. In addition, we also found that the structure and function of circSamd4a are highly conserved among different bony fishes. In general, our research shows that circRNA plays a key role in fish antiviral immune response, which will help to further broaden the understanding of vertebrate innate immune regulatory networks.

The authors have no financial conflicts of interest.