The circRNA circVAMP3 restricts influenza A virus replication by interfering with NP and NS1 proteins

Circular RNAs (circRNAs) are involved in various biological roles, including viral infection and antiviral immune responses. To identify influenza A virus (IAV) infection-related circRNAs, we compared the circRNA profiles of A549 cells upon IAV infection. We found that circVAMP3 is substantially upregulated after IAV infection or interferon (IFN) stimulation. Furthermore, IAV and IFN-β induced the expression of QKI-5, which promoted the biogenesis of circVAMP3. Overexpression of circVAMP3 inhibited IAV replication, while circVAMP3 knockdown promoted viral replication, suggesting that circVAMP3 restricts IAV replication. We verified the effect of circVAMP3 on viral infection in mice and found that circVAMP3 restricted IAV replication and pathogenesis in vivo. We also found that circVAMP3 functions as a decoy to the viral proteins nucleoprotein (NP) and nonstructural protein 1 (NS1). Mechanistically, circVAMP3 interfered with viral ribonucleoprotein complex activity by reducing the interaction of NP with polymerase basic 1, polymerase basic 2, or vRNA and restored the activation of IFN-β by alleviating the inhibitory effect of NS1 to RIG-I or TRIM25. Our study provides new insights into the roles of circRNAs, both in directly inhibiting virus replication and in restoring innate immunity against IAV infection.


Introduction
Influenza A viruses (IAVs), zoonotic pathogens that continuously circulate, pose a threat to public health and can cause human epidemics or pandemics [1,2]. IAV, which belongs to the Orthomyxoviridae family, consists of eight different negative-sense, single-stranded RNA segments encoding at least 17 proteins [3]. The virus ribonucleoprotein (vRNP) complex, which consists of viral RNA (vRNA), multiple nucleoproteins (NPs), and RNA polymerase complex polymerase basic 1 (PB1), PB2, and polymerase acidic protein (PA), plays essential roles in IAV genome replication and transcription [4,5]. The vRNP complex contains nuclear localization signal (NLS) peptides, which are recognized by the nuclear localization carrier protein importin-α in host cells, mediating entry of the vRNP complex into the nucleus. Once the viral proteins are synthesized in the cytoplasm, the self-dependent NLS is introduced into the nucleus, where vRNP complex assembly is completed [6][7][8][9]. Specifically, the viral nonstructural protein 1 (NS1) protein is widely recognized as a cofactor through which IAVs resist the host immune response [10].
The host innate immune system can be activated upon IAV infection [11]. In the process of viral infection, the antiviral immune response is triggered when the virus or its genetic material is detected by cellular pattern recognition receptors, followed by activating downstream signaling pathways to induce the expression of type I interferons (IFNs), such as IFN-β [12]. Type I IFNs promote the transcription of several genes called IFN-stimulating genes (ISGs). ISGs and other proteins with antiviral potential (restriction factors) can antagonize the viral life cycle. For example, IFN-inducible transmembrane proteins inhibit IAV entry by blocking the binding of viral envelope protein to the host membrane [13]. MxA functions as a broadspectrum antiviral protein by capturing viral components in the early stage of infection and preventing the virus from entering cells [14]. ZAP and PKR inhibit IAV replication by interfering with the translation system [15,16]. Cyclophilin A interacts with the matrix protein (M1) of IAV and inhibits the transport of M1 into the nucleus [17]. Furthermore, noncoding RNAs (ncRNAs) are also involved in antiviral innate immunity. For instance, microRNA-33a inhibits IAV replication by targeting ARCN1 and repressing vRNP activity [18]. Lnc-ISG20 was found to disrupt the inhibitory effect of miR-326 on ISG20 mRNA, followed by increasing the protein level of ISG20 to inhibit viral replication [19]. However, further elaboration on the underlying mechanisms of ncRNAs utilizing innate immune responses against IAV infection is required.
Circular RNAs (circRNAs) are a new class of ncRNAs that originate from nonsequential back splicing of exons and/or introns of precursor messenger RNAs (pre-mRNAs). CircRNAs were first discovered in RNA viruses in 1976 and described as single-stranded, covalently closed molecules [20]. Then, Hsu et al. found that endogenous circRNAs were also present in eukaryotic cells [21]. Subsequent studies have shown that circRNAs are widely distributed in various eukaryotes and exhibit highly conserved and tissue-specific characteristics [22]. Due to their circular conformation, circRNAs are more stable than mRNAs. CircRNAs interact with RNA-binding proteins (RBPs) or microRNAs (miRNAs) and thereby affecting their activities, regulating a variety of cellular processes, such as transcription and translation [23,24]. According to recent studies, circRNAs play important roles in innate immunity, suggesting that cir-cRNAs have the potential to affect IAV replication. Indeed, Qu et al. found that the upregulation of the circRNA AIVR by IAV inhibits IAV replication by sponging miRNAs and promoting the expression of CREBBP to facilitate IFN-β production [25]. In addition, RBPs are essential for circRNA generation [26]. However, the biogenesis processes, underlying functions, and potential mechanisms of circRNAs in IAV infection remain elusive.
In this study, we screened the circRNA transcriptional profiles of A549 cells infected with different IAV H1N1 strains (WSN, PR8, or CA04) to identify several circRNAs involved in IAV replication. We found a set of circRNAs that were stimulated by IAV. We then focused on the circRNA with the greatest increase upon IAV infection, circVAMP3, as it is also induced by IFN-β and inhibits IAV replication. Further analysis indicated that circVAMP3 biogenesis is mediated by QKI-5, which is upregulated by IAV and IFN-β. Mechanistically, circVAMP3 serves as a decoy to the viral NP and NS1 proteins, which interfere with vRNP complex activity and inhibit the effect of NS1 on the RIG-I signaling pathway. Together, these findings revealed that circVAMP3 plays an important role in restricting IAV replication.

Both IAV infection and IFN-β stimulation upregulate circVAMP3
To identify circRNAs involved in IAV infection, A549 cells were infected with IAV H1N1 strains (WSN, PR8, or CA04), and circRNA screening was performed by deep RNA sequencing. We found differences in the expression levels of many circRNAs when cells infected with all three strains compared to the uninfected cells (S1A Fig). Among them, the levels of 260 cir-cRNAs were significantly altered in all cells respectively infected with the three IAV strains (S1B Fig), in which 159 circRNAs upregulated and 101 circRNAs downregulated ( Fig 1A). Among these circRNAs, hsa_circ_0006354 was the most significantly upregulated upon IAV infection. We then validated that hsa_circ_0006354 was upregulated in IAV-infected cells using RT-qPCR (Fig 1B and 1C). In addition, we examined whether hsa_circ_0006354 expression is induced by other stimuli and found that hsa_circ_0006354 is also upregulated by Sendai virus (SeV) (S1C To verify the presence of hsa_circ_0006354, the convergent and divergent primers were designed for cDNA and genomic DNA (gDNA) PCR amplification (Fig 1D). Sanger sequencing of the PCR product, the head-to-tail splice junction site amplified by the divergent primers, indicated that hsa_circ_0006354 is derived from VAMP3 gene exons 3 and 4, which located on human chromosome 1 (Fig 1E), and hsa_circ_0006354 was accordingly named circVAMP3. CircVAMP3, except VAMP3 mRNA, was resistant to RNase R treatment ( Fig 1F) and had a longer half-life than VAMP3 and RIG-I mRNAs ( Fig 1G). Interestingly, we observed that VAMP3 mRNA had a longer half-life than RIG-I mRNA (Fig 1G). We then examined the subcellular localization of circVAMP3 by performing cell fractionation and RNA FISH. It revealed that circVAMP3 is mainly located in the cytoplasm but at very low levels in the nucleus ( Fig  1H and 1I). Together, circVAMP3 is expressed at high levels upon IAV infection or IFN-β stimulation and is located mainly in the cytoplasm.

QKI-5 promotes the generation of circVAMP3
It is known that splicing factors play a role in regulating the biogenesis of circRNAs [27]. To explore the mechanism by which circVAMP3 is formed, we identified two and two putative binding sites for EIF4A3 and DHX9, respectively, in the flanking region of circVAMP3 using Circinteractome (https://circinteractome.nia.nih.gov) and found three binding site motifs of ACUAAY (Y = C or U) for QKI-5 (S2A Fig). We hypothesized that if the three proteins contribute to circVAMP3 biogenesis, they would also be regulated by IAV. To test this hypothesis, we also explored the process of circVAMP3 biogenesis upon H1N1 WSN infection by conducting a proteomics analysis of A549 cells in the early stage of WSN infection. Among the eight circRNA biogenesis-associated proteins, only QKI-5 was significantly upregulated upon WSN infection (S2B Fig). We confirmed that the levels of QKI-5 expression and transcription were significantly upregulated after WSN infection, while the levels of EIF4A3 and DHX9 did not change (S2C and S2D Fig). Next, to explore whether IAV regulation of QKI-5 is viral titer dose-dependent, we used multiplicities of infection (MOIs) with different gradients to stimulate A549 cells and cell lysates were collected 4 h after infection for western blotting. We found that compared to the uninfected group, the protein level of QKI-5 increased gradually with the increase of virus titer (Fig 2A), indicating that H1N1 WSN upregulates circVAMP3 cyclingrelated protein QKI-5 in a dose-dependent manner. To investigate the effect of other IAV strains on QKI-5 expression. A549 cells were infected with the three types of H1N1 (WSN, CA04, and PR8), the avian IAV H9N2, and H3N2, individually. Consistent with WSN infection, all viruses can significantly upregulate the protein level of QKI-5 ( Fig 2B). Because both IAV and IFN-β upregulate circVAMP3 expression, we hypothesized that IFN also influences QKI-5 transcription and protein levels. To test this hypothesis, A549 cells were incubated with IFN-β for 8 h, which significantly upregulated QKI-5 transcription and protein levels in a dose-dependent manner ( Fig 2C). In conclusion, IAV can induce a significant increase in QKI-5 transcription and protein levels by up-regulating the expression of IFN.
Next, we examined the effects of QKI-5, EIF4A3, and DHX9 on the biogenesis of cir-cVAMP3. We transfected A549 cells with two specific siRNAs targeting QKI-5, EIF4A3, or DHX9 or with scrambled siRNA for 48 h, Western blotting analysis showed that specific siR-NAs treatment indeed downregulated the expression of QKI-5, EIF4A3, and DHX9 (S2E Fig). The expression level of circVAMP3 was then measured in siRNA-treated cells. We found that QKI-5 knockdown significantly decreased circVAMP3 generation, while EIF4A3 and DHX9 knockdown did not ( Fig 2D). Then, we transfected QKI-5 or EIF4A3 (negative control) in 293T cells for 48 h to detect their effects on circVAMP3 levels (S2F Fig). We found that the overexpression of QKI-5, not EIF4A3, increased circVAMP3 expression (Fig 2E). It reported that QKI acts as a dimer capable of binding two separated pre-mRNA regions and facilitating the circle-forming exons proximity, followed to circRNAs biogenesis [28,29]. To assess whether QKI binds the VAMP3 pre-mRNA, we performed RNA immunoprecipitation (RIP) assays. The result demonstrated that QKI bound to the exon-adjacent sites of VAMP3 pre-mRNA at a level comparable to NUMB, a binding site in the previously validated QKI target [30], no matter with or without IAV infection ( Fig 2F). Together, these results suggest that expression levels of circVAMP3, VAMP3 (F), and RIG-I (G) in A549 cells with or without RNase R treatment (F) or with actinomycin D treatment (G) were quantified using RT-qPCR. (H) A549 cells were infected with or without WSN at an MOI of 0.5 for 8 h. RNA was extracted from nuclear and cytoplasmic fractions and subjected to RT-qPCR to detect circVAMP3, GAPDH, and MALAT1 (H, left panel). The fractionated cell lysates were immunoblotted with anti-GAPDH and anti-Lamin B1 antibodies (H, right panel). (I) A549 cells were fixed and subjected to RNA FISH with a biotin-labeled circVAMP3 junctionspecific antisense probe (green). Nuclei were stained with DAPI (blue). Scale bar, 10 μm. Data are presented as the means ± SD; n = 3; **p < 0.01. Data are representative of results from at least three independent experiments. https://doi.org/10.1371/journal.ppat.1011577.g001

CircVAMP3 restricts IAV replication in vitro
To investigate the effect of circVAMP3 on IAV replication, the pLC5-ciR-GFP plasmid was used as the backbone to insert the cyclization splicing site and the third and fourth exon sequences of VAMP3 between the EcoRI and BamHI restriction sites. And then, 293T cells were transfected with pLC5-circVAMP3 or pLC5-ciR-GFP (control). The circVAMP3 was overexpressed in 293T cells (S3A Fig), and then the cells were infected with WSN virus to further explore the function of circVAMP3 during IAV replication. The result showed that circVAMP3 overexpression substantially reduced the viral titer (S3B Fig) and the level of the viral proteins M1 and NP (S3C Fig). However, we also found that the plasmid creates an additional 10 kDa fragment detected using VAMP3 antibody (S3C Fig). To eliminate the interference of this by-product, the circVAMP3 junction site was redesigned in this study (Fig S3D), and we verified that 10-kDa protein was no longer produced (S3E Fig). To identify the function of circVAMP3 during IAV replication, we generated the 293T-cir-cVAMP3 and A549-circVAMP3 cell lines, which stably ectopically expresses circVAMP3, and the control cell lines 293T-NC and A549-NC ( Fig 3A). To verify the overexpression efficiency, 293T and A549 cells were infected with pLC5-circVAMP3 lentivirus particles, and stable cell lines were constructed. Seventy-two hours later, the infected cells were collected, cell RNA was extracted for reverse primer amplification, the amplified products were purified, and they were sequenced. The amplification products from cells infected with pLC5-circVAMP3 lentivirus were more intense than those infected with empty lentivirus (S3F Fig), and the sequencing results were consistent with the expected new cyclic interface sequence (S3F Fig). The established cells were then infected with WSN virus. The results demonstrated that stable circVAMP3 overexpression substantially reduced the viral titer ( Fig 3B) and the level of the viral proteins M1 and NP (Fig 3C). Consistent results were observed in the two circVAMP3overexpressing stable cell lines, suggesting that circVAMP3 inhibits IAV replication. To verify whether the effect of circVAMP3 on WSN replication was dose-dependent, 293T cells were transfected according to the doses shown in S3G Fig, cell lysates were collected under the same conditions as above, and western blotting was performed. We found that transfection of 500 ng could significantly reduce the NP and M1 protein levels of WSN, and with the gradual increase of circVAMP3 transfection dose, the NP and M1 protein levels gradually decreased. These results indicate that transient circVAMP3 overexpression in 293T cells can significantly inhibit the replication of IAV in a dose-dependent manner.
To further explore the effect of endogenous circVAMP3 on IAV replication, we stably knocked down circVAMP3 in A549 cells using lentivirus-mediated short hairpin RNAs (shRNAs) targeting two different junction sites of circVAMP3, causing a reduction in cir-cVAMP3 expression but not VAMP3 mRNA levels (Fig 3D), and then infected the cells with WSN. We observed that circVAMP3 depletion significantly increased the viral titer in (Fig 3E), as well as the levels of both the M1 and NP proteins ( Fig 3F). To further demonstrate that the observed phenotype was not due to off-target effects, we transfected pLC5-ciR-GFP (EV, control) or pLC5-circVAMP3 into the two A549 stable knock-down cell lines, and then infected with WSN virus. As shown in Fig 3G, circVAMP3 levels were~10-fold higher than control cells. We found that the circVAMP3 restoration group significantly decreased the titer of WSN virus in A549 cells (Fig 3H), as well as the level of M1/NP protein (Fig 3I), compared to the empty plasmid. Taken together, these results suggest that circVAMP3 affects IAV replication in different cell types.
To further explore the effect of circVAMP3 at later time points. Therefore, we tested WSN replication in A549-circVAMP3 cells in comparison to that in A549-NC cells. The samples were harvested at 24 h, 36 h, 48 h, 60 h, and 72 h after infection. The result demonstrated that the virus titer of A549-circVAMP3 cells exhibited obvious reduction at every indicated time point (Fig 3J). Especially, the titer exhibited an approximately 0.9 log reduction at 72 hp.i. ( Fig  3J). Meanwhile, the cell lysates were harvested for the Western blotting assay, and circVAMP3 can significantly reduce the NP and M1 protein levels of WSN at later time points compared to control cells. To determine the function of circVAMP3 in cells infected with other subtype IAVs, we infected A549-NC and A549-circVAMP3 with WSN, CA04, PR8, H9N2, and H3N2, respectively. The supernatants were collected for plaque assay and the data showed that the viral titer of WSN, CA04, H9N2, and H3N2 from A549-circVAMP3 cells was significantly lower than that from control cells (Fig 3L). These results demonstrated that circVAMP3 can strongly inhibit the replication and proliferation of WSN at multiple longer time points and can inhibit the proliferation of multiple strains of IAV.
Given that VAMP3 is highly homologous in human and mice, mouse ortholog Vamp3 also has the two exons of circVAMP3. However, circVamp3 doesn't exist in mice through searching from circBase (http://www.circbase.org/). To verify no homologous circVamp3 in mice, we designed the divergent and convergent primers to amplify the circVamp3, circIpo11(positive control, which was verified to exist in mice), and Vamp3 mRNA in Lewis lung carcinoma Collectively, these results demonstrated that circVAMP3 plays an important role in IAV replication.

CircVAMP3 restricts IAV replication and pathogenesis in vivo
To investigate the effect of circVAMP3 on IAV infection in vivo, AAV6-expressing circVAMP3 (AAV6-circVAMP3) or negative control AAV6 (AAV6-NC) were delivered into the lungs of BALB/c mice by intratracheal intubation, with PBS serving as a control ( Fig 4A). RT-PCR confirmed that exogenous circVAMP3 was expressed in the lungs of the mice at 3 weeks postinfection ( Fig 4B). We then examined the effect of circVAMP3 on viral pathogenicity in the mice. We intranasally infected BALB/c mice with WSN and measured their body weights and daily survival. The data showed that the mice treated with AAV6-circVAMP3 lost less body weight than the mice treated with AAV6-NC or PBS (Fig 4C and S1 Table). Additionally, the mortality rate of the AAV6-circVAMP3-treated mice was lower than that of the mice treated with AAV-NC or PBS. The AAV6-NC-and PBS-treated mice died within 8 dpi, while 30% of the AAV6-circVAMP3-treated mice were still alive at the end of the experiment ( Fig 4D). As shown in Fig 4E, the lungs of the AAV6-NC-or PBS-treated mice were characterized by severe red blood cell extravasation at 5 and 7 dpi, but this symptom was alleviated in the AAV6-cir-cVAMP3-treated mice. The lung index (lung weight/body weight) of the AAV6-circVAMP3- The supernatants of infected cells were collected for plaque tests to determine virus titer. (C) Cell lysates were collected and used for immunoblotting with corresponding antibodies. (D, E, and F) An A549 cell line with stable circVAMP3 knockdown was constructed, and cells constructed with sh-NC-GFP no-load lentivirus were used as a control. After infection with circVAMP3 lentivirus or control lentivirus for 72 h, A549 cells were plated in 12-well plates. The next day, the cells were infected with WSN at an MOI of 0.5 or treated with PBS. Cell supernatant and cell lysate were collected at 8 h. (D) 72 h after lentivirus infection, total RNA was collected for knockdown efficiency and off-target rate detection. (E) The cell supernatants were collected for plaque assays to determine virus titer. (F) The cell lysates collected were used for immunoblotting assays and incubated with the antibodies shown. (G, H, and I) Lip3000 was used to transfect pLC5-circVAMP3 or pLC5-ciR-GFP (as negative control) in an A549 cell line with a stable knockdown of circVAMP3, and the sh-NC-GFP stable cell line was used to transfect pLC5-ciR-GFP in the control group. Twenty-four hours after transfection of corresponding plasmids, cells were infected with WSN at an MOI of 0.5, and cell supernatants and cell lysates were collected at 8 h. (G) RT-qPCR confirmed that the level of circVAMP3. (H) The cell supernatants collected were used for plaque assays to determine virus titer. (I) The cell lysates collected were used for immunoblotting assays and incubated with the antibodies shown. (J and K) The cells were grown in the 12-well plates and then infected with WSN (MOI = 0.01) without supplement TPCK-treated trypsin, the supernatants were followed harvested at the indicated time points for plaque assays to detect the viral growth curves (J), and the cell lysates were collected for immunoblotting assay (K). (L) The A549-NC and A549-circVAMP3 cells were infected with IAV (WSN, CA04, PR8, H9N2, or H3N2) at an MOI = 0.5 for 16 h, and the cell supernatants were collected for plaque assays. Data shown in A, B, D, E, G, H, J, and L are presented as the mean ± SD. n = 3; **p < 0.01. These data represent the results of at least three independent experiments.  carrying circVAMP3 (AAV6-circVAMP3) or EV (AAV6-NC) was generated. BALB/c (4-6 weeks old, female) mice were administered PBS, AAV6-NC, or AAV6-circVAMP3 in a volume of 50 μL (5×10 10 vector genomes) through intratracheal instill. Then, the lungs of the mice were harvested at 3 weeks postinfection for RT-PCR analysis of exogenous circVAMP3 expression (B). The mice were infected with WSN (10 4 PFU/each) by nasal drip. The body weight (C) and survival rate (D) of the mice were monitored daily for 14 days (n = 10). At 0, 3, 5, and 7 dpi, a necropsy was performed (E), the gray dashed curves indicated the significant gross pathology and hemorrhage, and the lungs of the infected mice were collected and used to measure the lung index (n = 3 or 6) (F). At 3 and 5 dpi, the lungs of the infected mice (n = 3) were harvested for plaque assays (n = 3) to measure the viral titers (G) and the lung tissues were stained with anti-M1 mAb for immunohistochemistry (H) or with hematoxylin and eosin (n = 2), the red line indicated 100 μm (I). (J) Pathological scores were obtained by evaluating the pulmonary interstitial edema, alveolar edema, inflammatory invasion, and alveolar hemorrhage areas and degree of hyaline membrane development (n = 5). The data shown in C, F, G, and J are presented as the means ± SD; *p < 0.05 and **p < 0.01. https://doi.org/10.1371/journal.ppat.1011577.g004

PLOS PATHOGENS
circVAMP3 suppresses IAV replication treated mice was lower than that of the NC mice, suggesting that circVAMP3 reduced pulmonary edema (Fig 4F). To explore the viral load in the lungs, the viral titers in the lungs of the mice at 3 and 5 dpi were examined by plaque assay and immunohistochemistry (IHC) staining of the lung was performed at 0, 3, 5, and 7 dpi using anti-M1 mAb. The viral titer in the lungs of the AAV6-circVAMP3-treated mice was significantly lower than that in the AAV-NC-or PBS-treated mice at 3 and 5 dpi (Fig 4G). Less positive signals were distributed in lung tissues of AAV6-circVAMP3-treated mice than that in the AAV-NC-or PBS-treated mice at 3, 5, and 7 dpi (Fig 4H), indicating a lower viral load in the AAV6-circVAMP3-treated mice. These results indicate that circVAMP3 can inhibit the viral replication in the lungs of infected mice, thus having the potential to increase the survival of mice during IAV infection in vivo. Histological analysis of the lungs indicated that alveolar damage and interstitial inflammatory cell infiltration were less serious in the AAV6-circVAMP3-treated mice than those in the AAV6-NC-or PBS-treated mice at 3, 5, and 7 dpi ( Fig 4J). Consistently, the histopathological scores for the lungs of the AAV6-circVAMP3-treated mice were significantly lower than those for the lungs of the control mice ( Fig 4K). Taken together, these results suggest that cir-cVAMP3 reduces IAV replication and pathogenesis in vivo.

CircVAMP3 directly interacts with both IAV NP and NS1
We prepared biotin-labeled circVAMP3 and performed an RNA pulldown assay, followed by mass spectrometry to identify circVAMP3-binding proteins and elucidate the mechanism by which circVAMP3 inhibits IAV replication (Figs 5A, S5A and S5B). The result displayed that circVAMP3 only interacts with the viral NP and NS1 proteins (Fig 5B). We then used RNA pulldown and RIP assays to further confirm whether circVAMP3 interacts with viral NP and NS1. The biotin-labeled probe was incubated with the lysates of WSN-infected A549 cells for RNA pulldown to verify this result. As shown in Fig 5C, endogenous circVAMP3, which was specifically enriched by the probe specific for circVAMP3, could interact with NP. Then, following RNA immunoprecipitation, the precipitates were used for RT-qPCR-mediated detection of circVAMP3 or GAPDH as an internal control. The results suggested the presence of endogenous circVAMP3 in the viral NP-immunoprecipitated fraction (Fig 5D). We transfected 293T cells with pcDNA3-FLAG-NP or empty vector as a control to further confirm the interaction between circVAMP3 and NP. We observed the presence of circVAMP3 in the FLAGimmunoprecipitated fraction of FLAG-NP ( Fig 5E). To characterize the interaction of cir-cVAMP3 with NS1, we transfected 293T cells with pcDNA3-FLAG-NS1 or an empty vector as a control. The cell lysates were harvested for RIP assays with a biotin-labeled circVAMP3 probe or a random probe as control. As shown in Fig 5F, NS1 was pulled down with the cir-cVAMP3 probe, indicating that NS1 can interact with circVAMP3. Then, an RNA immunoprecipitation assay was performed, and RT-PCR was used to identify circVAMP3. The data showed the presence of circVAMP3 in the FLAG-immunoprecipitated fraction of FLAG-NS1 ( Fig 5G). Moreover, to verify circVAMP3 does not bind to any IAV other proteins, we further transfected 293T cells with pcDNA-FLAG-PA, pcDNA-FLAG-PB1, pcDNA-FLAG-M1(as a negative control), and cDNA-FLAG-NP (as a positive control), respectively. The cell lysates were collected for RIP assays with a biotin-labeled circVAMP3 probe or a random probe as control. As shown in Fig 5H, among the four viral proteins, only NP was pulled down with the circVAMP3, indicating circVAMP3 does only interact with viral NP and NS1.
Next, we predicted the modeling structure using HADDOCK 2.4 and I-TASSER algorithm, which indicated that circVAMP3 can directly interact with NP and NS1 by generating structural predictions (S5C and S5D Fig). To confirm whether circVAMP3 directly interacts with NP or NS1 protein, we purified His-tagged NP, His-tagged NS1, and His-tagged M1 (as a PLOS PATHOGENS circVAMP3 suppresses IAV replication The cell lysates were harvested and subjected to a pulldown assay with biotin-labeled circVAMP3 or unlabeled circVAMP3 as the control. The bound proteins were visualized by silver staining. The arrows indicate the bands that were subjected to mass spectrometry analysis. (C) The bound protein was detected by immunoblotting with anti-NP antibody (C, left). RT-qPCR was used to PLOS PATHOGENS circVAMP3 suppresses IAV replication negative control) from Escherichia coli and prepared in vitro-transcribed and self-circularized circVAMP3. We incubated His-NP/His-NS1 and circVAMP3 with the biotin-labeled cir-cVAMP3 probe or a random probe as a control and then subjected the samples to pulldown with streptavidin beads, followed by immunoblotting with anti-NP or anti-NS1 antibody. This revealed the presence of NP or NS1 was both detected in the biotin-labeled circVAMP3 probe precipitated fraction (Fig 5I). In addition, we performed immunoprecipitation with His-beads, followed by RT-qPCR to detect circVAMP3 in the His-immunoprecipitated fraction. The presence of circVAMP3 was detected in the His-immunoprecipitated fraction of His-NP or His-NS1 but not in the M1-immunoprecipitated fraction (Fig 5J and 5K), which further confirmed that circVAMP3 can directly interact with NP or NS1 protein.
To determine the regions of NP responsible for interaction with circVAMP3, two truncated NP proteins based on the location of the RNA-binding domain (RBD) were expressed and subjected to RIP assays. Interestingly, the RT-qPCR results indicated that circVAMP3 is associated with NP truncation mutant (aa 182-498) but not NP (aa 1-181) (the RBD) ( Fig 5L). Two truncated NS1 proteins, based on the location of the RBP, were expressed and subjected to RIP assays to determine the exact domain of NS1 that interacts with circVAMP3. As expected, the RT-qPCR results indicated an interaction of circVAMP3 with NS1 truncation mutant (aa 1-84) (the RBD) but not NS1 (aa 85-230) ( Fig 5M). In addition, the four segments 1-30, 121-150, 151-180, and 181-210 nt of circVAMP3 were sufficient to bind NP ( Fig 5N). It demonstrated that circVAMP3 directly interacts with the NP (aa 182-498) region and contains multiple NP-binding sites. In addition, the four segments 1-30, 31-60, 121-150, and 181-210 nt of circVAMP3 were sufficient to bind NS1 (Fig 5N), indicating that circVAMP3 shares partial interaction binding sites with NP and NS1. Taken together, these results indicate that cir-cVAMP3 directly interacts with NS1 (aa 1-84) region and contains multiple NS1-binding sites.

CircVAMP3 inhibits vRNP activity by impairing the interaction of NP with PB1 or PB2
NP is an important component of the vRNP complex and is critical for vRNP activity [31][32][33]. We hypothesized that vRNP activity could be alleviated through the expression circVAMP3 decoying NP. To test our hypothesis, we checked whether circVAMP3 affects viral polymerase activity by interacting with NP. We performed a luciferase assay in 293T transfected with the vRNP promoter-reporter (driven by NS promoter), or in 293T-IAV-Luc cells (vRNP reporter stable expressed in 293T cells, driven by NP promoter) to observe the vRNP activity. As shown detect the efficiency of the pulldown with the probe (C, right) (n = 4). (D) The cell lysates were immunoprecipitated with anti-NP antibody or IgG as the control. (E) Cell lysates were harvested for immunoprecipitation with FLAG beads and subjected to RT-qPCR to detect circVAMP3 or actin as an internal control (n = 2). (F) The bound NS1 protein was detected using immunoblotting. (G) The cell lysates were immunoprecipitated with FLAG beads. The bound NS1 was detected by immunoblotting (left panel). The bound circVAMP3 was quantified using RT-qPCR ( Fig 6A, relative luciferase activity was significantly decreased in circVAMP3-overexpressing cells compared to the NC control, suggesting that circVAMP3 inhibits viral polymerase activity. We then utilized the two types of A549-circVAMP3 stable knock-down cells for the luciferase assays. We found that the vRNP activity was increased when the expression of circVAMP3 was depleted, indicating that endogenous circVAMP3 inhibits the transcription and replication of the viral genome, while circVAMP3 reconstitution rescued the inhibition phenomenon ( Fig  6B). To further determine the steps of viral transcription and replication that are inhibited by circVAMP3 expression, we measured the vRNA, cRNA, and mRNA levels of M1 and NP in WSN-infected (8 h) 293T cells with or without circVAMP3 overexpression. We found that cir-cVAMP3 overexpression reduces the vRNA, cRNA, and mRNA levels of M1 and NP (Fig 6C). Furthermore, we suspected that the vRNP activity was impaired due to the retention of vRNP and NP in cytoplasm caused by circVAMP3 expression. Therefore, we explored whether circVAMP3 has an inhibitory effect on the nuclear import of NP with a cell fractionation assay. The pLC5-circVMAP3 or pLC5-ciR-GFP were transfected in 293T cells, followed infected with WSN at an MOI of 5. The cells were then lysed, and the nuclear and cytoplasmic fractions were separated and subjected to western blotting. As shown in S6C Fig, the marker proteins Lamin B1 and GAPDH were only detected in the nucleus and cytoplasm, respectively. At 1 and 2 hp.i., NP levels were not significantly changed in either the nucleus or cytoplasm of the circVAMP3-overexpressing cells compared to control cells. However, at 3 and 4 hp.i., NP was significantly downregulated in both the nucleus and cytoplasm of the circVAMP3-overexpressing cells compared to the control cells. These results suggested that circVAMP3 indeed inhibits IAV replication at an early stage, but it can not affect incoming vRNP and NP nuclear import and virus entry. To further verify whether circVAMP3 affects viral attachment or entry, we measured the effect of circVAMP3 on viral attachment to cell membranes. As shown in S6D Fig, there was no significant difference in the amounts of infectious virus attached on the surface between the A549-circVAMP3 cells and A549-NC cells. Meanwhile, we also harvested the cells for immunoblotting assay. As shown in S6E Fig, overexpression circVAMP3 did not adversely affect the attachment of WSN to the A549 cell surface. We then verified whether overexpressed circVAMP3 influences the internalization of IAV. The amount of internalized viral M1 protein between A549-circVAMP3 and A549-NC has a minimal difference (S6F Fig). Taken together, circVAMP3 inhibits IAV replication at the first replication cycle by impairing the activity of the vRNP complex, but circVAMP3 is not required for IAV attachment to the cell surface and the internalization of IAV into cells.
To explore the molecular mechanism by which circVAMP3 regulates IAV vRNP activity, we detected the effect of circVAMP3 on impairing the function of NP during vRNP complex 1-PB2-vRNA as a positive control. The data shown in B and G are presented as the means ± SD; n = 3 **p < 0.01. Data are representative of the results from at least three independent experiments. https://doi.org/10.1371/journal.ppat.1011577.g006

PLOS PATHOGENS
circVAMP3 suppresses IAV replication assembling. Previous studies report that NP aa 182-498 contains the NP self-polymerization region [34] and PB1 or PB2 binding region [33,35]. Since NP self-polymerization is important for vRNP activity, we wondered whether circVAMP3 affects NP self-polymerization. We cotransfected 293T cells with pcDNA-FLAG-NP and pcDNA-Myc-NP along with the pLC5-cir-cVAMP3 or IAV-PB2-vRNA (Ben1, as a negative control). The cell lysates were then harvested for immunoprecipitation and immunoblotting. We found that circVAMP3 did not influence NP self-polymerization ( Fig 6D). Next, we sought to determine whether circVAMP3 interfered with the interaction of NP and other subunits of the vRNP complex. To this end, we co-transfected 293T cells with pcDNA-Myc-NP, pcDNA-FLAG-PB1 or pcDNA-FLAG-PB2, and pLC5-circVAMP3. The cell lysates were harvested for immunoprecipitation and immunoblotting to detect the association of NP with PB1 or PB2. The overexpression of circVAMP3 reduced the interaction of NP with PB1 or PB2 (Fig 6E and 6F). Meanwhile, we investigated whether circVAMP3 influence NP interact with vRNA. We co-transfected 293T cells with pcDNA-Myc-NP, IAV-M-vRNA (Ben7), and pLC5-circVAMP3 or IAV-PB2-vRNA (Ben1, as positive control), the cell lysates were collected for Myc-immunoprecipitation. The samples were harvested for immunoblotting and RT-qPCR detection, respectively. The data showed that overexpression of circVAMP3 reduced the interaction of NP with vRNA ( Fig 6G). Taken together, circVAMP3 inhibits IAV polymerase activity by interfering with the interaction of NP with other vRNP subunits.

CircVAMP3 alleviates the inhibitory effect of NS1 on IFN-β activation antagonizing NS1 binding to RIG-I
NS1 plays an important role in inhibiting the RIG-I-mediated antiviral immune response [36,37]. We hypothesized that circVAMP3 can abrogate the inhibitory effect of NS1 on the antiviral immune response. To test this hypothesis, we detected the expression of the IFN-β mRNA in A549-circVAMP3 with or without WSN infection. The level of IFN-β mRNA was significantly increased in A549-circVAMP3 cells compared with control cells, indicating that IAV-infected cells produced more IFN-β upon increasing circVAMP3 (Fig 7A). In addition, 293T cells expressing pLC5-circVAMP3 or control cells were transfected with poly (I: C) alone or together with the pcDNA-Myc-NS1. The RNA was prepared for RT-qPCR to detect the level of the IFN-β mRNA. We found that circVAMP3 did not affect the IFN-β mRNA induced by poly(I:C). However, circVAMP3 significantly enhanced the IFN-β mRNA expression, which was suppressed by NS1 protein (Fig 7B, left panel). Consistently, we observed that cir-cVAMP3 also antagonized the inhibitory effect of NS1 on the activation of IFN-β induced by SeV infection (Fig 7B, middle panel). To further determine the effect of circVAMP3 on the RIG-I signaling pathway, 293T cells were transfected with RIG-I-N, a constitutively active form of RIG-I. Consistent with other stimuli, circVAMP3 overexpression significantly rescued the decrease in IFN-β expression induced by NS1 protein (Fig 7B, right panel). Meanwhile, we also observed that circVAMP3 overexpression antagonized the inhibitory effect of NS1 on the expression of IFN-β in A549-stable-overexpressing cells (Fig 7C).
Since we already identified that circVAMP3 interacts with NS1 1-84 aa region (RBD domain) and NS1 RBD domain can interact with RIG-I and TRIM25 [36,38]. Next, we sought to determine whether circVAMP3 affected the association of NS1 with RIG-I or TRIM25. Firstly, FLAG-RIG-I and Myc-NS1 were co-expressed with or without pLC5-circVAMP3, and cell lysates were subjected to a coimmunoprecipitation assay. We found that circVAMP3 reduced the association of Myc-NS1 with FLAG-RIG-I (Fig 7D). Meanwhile, FLAG-NS1 and Myc-TRIM25 were co-expressed with or without circVAMP3, and cell lysates were subjected to a coimmunoprecipitation assay. We found that circVAMP3 reduced the association of Myc-

PLOS PATHOGENS
circVAMP3 suppresses IAV replication TIRM25 with FLAG-NS1 (Fig 7E). And then, based on previous studies reported that NS1 specifically inhibits TRIM25-mediated RIG-I CARD ubiquitination, hence suppressing RIG-I signal transduction [38,39]. Therefore, we investigated whether circVAMP3 affected the inhibition of NS1 on the association of RIG-I with TRIM25. We transfected Myc-NS1, Myc-TRIM25, FLAG-RIG-I, with or without pLC5-circVAMP3, and cell lysates were subjected to a coimmunoprecipitation assay. We found that circVAMP3 expression impaired the inhibition of Myc-NS1 on the association of FLAG-RIG-I with Myc-TRIM25 (Fig 7F). Finally, we observed K63-linked RIG-I ubiquitination to observe RIG-I activation. As shown in Fig 7G, circVAMP3 restored RIG-I K63 ubiquitination that was suppressed by NS1. These data confirmed that circVAMP3 rescues RIG-I activation by diminishing the interaction of NS1 with RIG-I. To further clarify if circVAMP3 can play an antiviral role through the RIG-I signaling pathway, RIG-I-depleted 293T cells were used to detect the effect of circVAMP3 upon IAV infection. As shown in Fig 7H and 7I, in 293T cells without RIG-I depletion, overexpression of circVAMP3 significantly inhibited viral titers and IAV protein levels, consistent with previous findings. However, the RIG-I deficient 293T cells had significantly higher titers and protein levels than wild-type cells. Although circVAMP3 overexpression still significantly inhibited the proliferation of influenza virus, the inhibition rate was reduced in RIG-I-deficient 293T cells.
In conclusion, we identified that a circular RNA, circVAMP3, is upregulated by IAV infection or IFN-β treatment. Additionally, IAV-or IFN-β-mediated upregulation of QKI-5 mediates circVAMP3 biogenesis. Functional in vitro and in vivo analyses indicate that circVAMP3 plays a vital role in restricting IAV replication. Mechanistically, we first discovered that cir-cVAMP3 inhibits vRNP activity by disturbing the interaction of NP with PB1, PB2, or vRNA and restores the activation of IFN-β by alleviating the inhibitory effect of NS1 on the RIG-I signaling pathway (Fig 8).

Discussion
In this report, we reveal a novel circRNA-mediated antiviral mechanism in which circVAMP3 directly inhibits viral replication by acting as a decoy to IAV NP and NS1 proteins. We present evidence that circVAMP3 inhibits the interaction of NP with PB1, PB2, or vRNA to impair vRNP activity, as well as antagonize NS1-mediated suppression of RIG-I activation to restore antiviral innate immunity.
CircVAMP3 plays multiple functions in different biological processes. CircVAMP3 is highly expressed in alveolar rhabdomyosarcoma cells and potentially regulates cell cycle progression through AKT-related pathways [40]. Conversely, circVAMP3 is significantly downregulated in hepatocellular carcinoma tissues, which inhibiting proliferation by facilitating stress granule formation [41]. However, the well-known function of circVAMP3 is still limited to neoplasm regulation. In this study, we unveil a novel molecular mechanism of circVAMP3 during viral infection. We reported that circVAMP3 can regulate the replication of IAV. Contrary to the widely reported role of circRNAs as miRNAs sponges to indirectly regulated gene expression, we did not observe the upregulated VAMP3 protein levels upon circVAMP3 overexpressed. Consistently, VAMP3 mRNA remained unaffected in both cricVAMP3 knocking-down or overexpression cell lines [42], suggesting that circVAMP3 does not act as a miRNA sponge. The mechanisms of circRNAs generation, stability, and degradation remain uncleared. QKI belongs to the signal transduction and activation of RNA (STAR) family of KH domain-containing RBP with three major isoforms: QKI-5, QKI-6, and QKI-7 [43]. In our study, we observed that overexpressing QKI-5 enhances circVAMP3 generation, while QKI knockdown suppressed circVAMP3 generation. It's consistent with the result of the QKI knockdown transcriptome [43]. Additionally, IAV infection and IFN-β stimulation upregulate QKI-5 to promote circVAMP3 biogenesis. Since IAV infection induces dose-dependent upregulated of IFN-β gene expression [44], IAV infection may promote circVAMP3 generation by inducing IFN-β expression. Notably, QKI is frequently downregulated in lung cancer and associated with a poorer prognosis [30]. Thus, QKI-5 may also facilitate circVAMP3 formation in cancerrelated processes. Moreover, it is unclear whether IFN-β facilitates VAMP3 pre-mRNA expression, suggesting circVAMP3 may serve as an ISG product or another mechanism promoting

PLOS PATHOGENS
circVAMP3 suppresses IAV replication its generation. What's more, we observed that linear VAMP3 mRNA and circVAMP3 had longer half-lives than RIG-I mRNA. It has been reported that VAMP3 mRNA contains m6A modifications, which may stabilize its stability. However, circVAMP3 is insensitive to m6Amediated regulation [40]. Previous studies have shown a rapid decrease in circRNAs abundance after poly (I:C) treatment or virus infection [45]. However, there are conflicting reports on the regulation of circRNAs levels during these treatments [46]. Similarly, we found both upregulation and downregulation of circRNAs during IAV infection. The underlying mechanisms controlling the levels and half-life of specific circRNAs need further investigated.
Like other ncRNAs, circRNAs can work as RNA decoys for proteins. The biological function and detailed mechanisms of circRNAs remain unclear. Bioinformatics analysis has been used to predict the potential functions of circRNAs, but it has shown that circRNAs do not tend to function as miRNA sponges or bind RBPs like linear mRNAs [58,59]. Nevertheless, interactions between circRNAs and proteins or miRNAs have been reported in various studies [23]. Our study also revealed that circVAMP3 can directly interact with viral proteins. Recent studies have focused on how circRNAs binding to host proteins indirectly affects the response to viral infection [60]. We were the first to report that circRNA can act as an RNA decoy by directly interacting with viral proteins to inhibit their function. The concept of using decoy ncRNAs to sequester proteins originated from bacteria with the discovery of small RNAs acting as decoys [61]. Long ncRNAs have also been identified as decoy molecules that bind and sequester proteins, inhibiting the proteins' normal functions [62,63]. As a new member of long ncRNA, circRNAs have been reported to act as RNA decoys for host proteins [64]. In our study, we found that circVAMP3 acts as a decoy for viral proteins NP and NS1. Our results demonstrated that ncRNAs act as RNA decoys not only for host proteins but also for viral proteins.
The structure of ncRNAs plays a crucial role in their interaction with proteins. RNA typically interacts with protein through base stacking, hydrogen bonding, electrostatic interactions, and hydrophobic interactions [65]. The ncRNAs were thought to predominantly adopt secondary or tertiary structures, suggesting that the primary sequence of RNA is critical for protein binding, while the secondary structure enhances the affinity of RNA-protein interaction [61]. Due to the unique tertiary structure of circRNAs, their ability to bind proteins may be more complex than previously investigated. In our study, we found that circVAMP3 preferentially binds to the C-terminal region of NP rather than the RNA-binding domain, interrupting NP binding with vRNA. It would be worthwhile to investigate in future studies which portion of the unique tertiary structure of circVAMP3 determines its interaction with NP. These findings indicate that traditional nucleotide sequencing-based methods used to analyze linear mRNAs binding to protein may not be suitable for studying circRNA-protein interactions. Furthermore, our study demonstrated that circVAMP3 antagonized the association of NP with PB1 and PB2, but not NP self-polymerization. This is consistent with previous reports, as RNA (vRNA and 24-nucleotide ssRNA) binding itself does not affect NP self-polymerization [5,66]. Interestingly, although all four segments of circVAMP3 significantly bind to NP or NS1 protein, none of them significantly inhibit IAV replication compared to cir-cVAMP3 itself (S7 Fig). This suggests that the primary structure of circVAMP3 may support strong binding, while the tertiary structure may play a role in its inhibition function. However, the dynamic tertiary structure of circRNAs can be influenced by various factors in the cellular environment, leading to variations in different cell lines, tissues, and developmental stages [67]. Studying circRNAs in vitro may cause changes in their tertiary structures, making it challenging to reveal their true dynamic interactions in situ. Therefore, more research is needed to understand how circRNAs bind to different proteins.
CircRNAs hold promise as a new type of molecular drug, but further research is needed. CircVAMP3, as a stable antiviral factor with multiple binding sites for viral proteins, has demonstrated suppression of various IAV strains by inhibiting viral NP and NS1 protein functions. Further evaluation of circVAMP3 potential as an antiviral drug is warranted. CircRNAs have attracted significant attention due to their involvement in various biological processes. However, many circRNAs functions during viral infection remain unknown [68,69], and further investigations are needed to understand their interaction with host immune system in the antiviral defense response. It is also important to consider the spatiotemporal regulation of cir-cRNA function, as some circRNAs may work by changing localization patterns or over time.
Exploring the physiological function of circRNAs will enhance our understanding and provide potential therapeutic strategies for influenza virus infection.

Mice and ethics statement
Female BALB/c mice (4-6 weeks old) were purchased from Beijing Vital River Animal Technology Co., Ltd. Mice were maintained under specific pathogen-free (SPF) conditions. Animal studies were approved by the Committee on the Ethics of Animal Experiments of the Institute of Microbiology, Chinese Academy of Sciences (IMCAS) and conducted in compliance with the recommendations in the Guide for the Care and Use of Laboratory Animals by the IMCAS Ethics Committee (permit number PZIMCAS2019001). The animals were housed and bred in SPF mouse facilities in compliance with ethical guidelines, and all animal experiments were approved by the Experimental Animal Ethics and Welfare Committee of IMCAS.

Antibodies and reagents
The rabbit anti-NP polyclonal antibody and mouse anti-M1 monoclonal antibody were generated as previously described [71]. The rabbit anti-NS1 polyclonal antibody was generated as previously described [72].

RNA sequencing analysis
A549 cells were infected with WSN, CA04, or PR8 at an MOI of 1 or incubated with PBS (mock-infected) for 8 h. Total RNA was extracted with TRIzol reagent (Thermo Fisher Scientific, USA). The purity of the RNA samples was evaluated using agarose gel electrophoresis and a NanoDrop One spectrophotometer (Thermo Fisher Scientific, USA). RNA with an RNA Integrity Number (RIN) > 7, as assessed with an Agilent 2200 bioanalyzer (Agilent Technologies, USA), was screened. The rRNA was removed with a Ribo-Zero rRNA Depletion Kit (RiboBio, China), and linear RNA was removed by RNase R treatment. In addition, 1 μg of PLOS PATHOGENS circVAMP3 suppresses IAV replication purified RNA was used for circRNA library construction. After removing reads that contained adapter sequences and low-quality reads, circRNAs were identified from the remaining reads using CIRI2 software. The remaining junction reads were aligned to human circRNAs in the circBase database (http://www.circbase.org/). Data from both the IAV-and mock-infected cells were used to identify differentially expressed circRNAs based on back splicing junction reads per million mapped reads (RPM) with a fold-change > 2 and p-value < 0.05. All raw data from these RNA-seq data were deposited in the National Center for Biotechnology Information (NCBI) Short Read Archive (SRA) database with accession number PRJNA750806.

RT-PCR and RT-qPCR
Total RNA was isolated from cells with TRIzol reagent, treated with or without RNase R (1 U/ μg), and subjected to RT-qPCR to detect circVAMP3. First-strand cDNA was synthesized with Hifair II 1st Strand cDNA Synthesis SuperMix for qPCR (Yeasen, China). RT-qPCR was performed with Hieff qPCR SYBR Green Master Mix (Yeasen, China) according to the manufacturer's protocol. Relative gene expression was quantified with GAPDH as the internal control to normalize.

Nuclear and cytoplasmic fractionation
Cytoplasmic and nuclear fractionation was performed with nuclear and cytoplasmic extraction reagents (Applygen, China) according to the manufacturer's instructions. Briefly, A549 cells were lysed in CEB-A buffer on ice for 15 min. After centrifugation at 2,000 × g for 10 min, the supernatant was collected as the cytoplasmic extract. The pellet was suspended in PBS on ice and washed three times. After centrifugation at 2,000 × g for 10 min, the pellet was suspended in CEB-B buffer on ice, incubated for 30 min, and subsequently centrifuged at 12,000 × g for 5 min, after which the supernatant was carefully collected as the nuclear extract.

RNA fluorescence in situ hybridization (FISH)
A biotin-labeled oligonucleotide probe for circVAMP3 was synthesized by Tsingke Co., Ltd. (China). Cultured cells on slides were washed in PBS and fixed in 4% polyformaldehyde at room temperature for 30 min. The fixed cells were then treated with 0.5% Triton X-100 and then incubated with 50% formamide/2×SSC for 20 min. Hybridization was performed at 42˚C for 16 h with the biotin-labeled circVAMP3 probe. After washing with 50% formamide/2×SSC, the slides were hybridized with FITC-conjugated streptavidin (Bioss, China) at 37˚C for 2 h in a dark chamber and washed. Then, the slides were incubated with 4% BSA, stained with DAPI (Thermo Fisher Scientific, USA), and sealed with Vectashield (Vector Laboratories, USA). Images were acquired using fluorescence microscopy (Leica SP8 confocal microscopy, Leica, Germany).

Transfection
Cells were plated and transfected with the indicated plasmid using Lipofectamine 2000 Transfection Reagent (Thermo Fisher Scientific, USA) for 24 h according to the manufacturer's instructions. A549 cells were plated and transfected with 50 nM of the indicated siRNA using RNAiMax (Invitrogen, USA).

Plaque assay
MDCK cells were seeded in 12-well plates, grown in monolayers, and infected with the indicated virus in serum-free DMEM. After adsorption, the cells were washed twice with PBS and PLOS PATHOGENS circVAMP3 suppresses IAV replication overlaid with MEM containing 2 μg/μL TPCK-treated trypsin and 1% low-melting-point agarose at 4˚C for 10 min, after which they were incubated at 37˚C with 5% CO 2 for 72 h. Then, the cells were fixed with a 4% formaldehyde solution overnight at RT and stained with a 0.05% (w/v) crystal violet solution. Viral plaques were counted, and viral titers were calculated.

Generation of stable overexpression and knockdown cell lines
Stable overexpression and knockdown cell lines were generated by transfecting 293T cells with the lentiviral packaging vector (pRSV, pVSVG, and pMDIG) and plc-ciR-circVAMP3 or pSIH-circVAMP3-shRNA for 48 h. Then, the virus-containing supernatant was collected and used to infect A549 cells for 72 h.

RNA synthesis and purification
QKI, EIF4A3, or DHX9 siRNAs and truncated fragments of biotinylated circVAMP3 were synthesized by Sangon Biotech (China). CircVAMP3 was generated in vitro using the permuted intron-exon (PIE) method, and the sequence was designed to contain autocatalytic splicing td introns [75]. CircVAMP3 was synthesized with a MEGAscript T7 transcription kit (Ambion, USA) at 37˚C for 12 h. Biotin-labeled circVAMP3 was synthesized, and dNTP was replaced with Biotin RNA Labeling Mix (Roche, Switzerland). Transcribed circVAMP3 and biotin-cir-cVAMP3 were purified with RNeasy Mini columns (Qiagen, Germany) and then treated with RNase R at 37˚C for 2 h to degrade linear RNA byproducts. After purification with RNeasy Mini columns, RNA quality was assessed on a 5% acrylamide/8 M urea gel. The goal band was extracted from the gel with a Zymoclean Gel Recovery Kit (Zymo Research, USA).

Protein expression and purification
For the purification of His-NP, His-NS1, and His-M1 proteins, the pET-21a-His-NP, pET-21a-His-NS1, and pET30a-His-M1 plasmids were transformed into E. coli Rosetta (DE3) cells, respectively. Transformants were cultured at 37˚C on LB medium containing 50 μg/mL kanamycin for 14 h. Single colonies were picked, inoculated into LB (K + ) medium, grown to an optical density at 600 nm of 0.6-0.8, and then protein expression was induced with 0.2 mM IPTG at 16˚C for 24 h. Harvested cells were lysed by sonication, and then the proteins were purified as previously described [76]. Purified proteins were detected using SDS-PAGE.

Immunoblotting and immunoprecipitation
The cells were lysed with ice-cold lysis buffer (20 mmol/L HEPES, pH = 7.4, 150 mmol/L NaCl, 10% glycerin, 1% Triton X-100, and 1 mmol/L EDTA) supplemented with a protease inhibitor cocktail and RNase inhibitor. Protein separation was conducted by 10% SDS-PAGE, and the proteins were transferred onto a PVDF membrane that was then immunoblotted with the indicated antibodies. The blots were visualized with ECL substrate (Tanon, USA). ImageJ software (National Institutes of Health, USA) was used to perform the gradation analysis. For immunoprecipitation, the cells were lysed with lysis buffer supplemented with protease inhibitor cocktail and RiboLock RNase inhibitor on ice for 30 min. The lysates were immunoprecipitated with the indicated antibodies and Protein A-Sepharose (Sigma-Aldrich, USA) or Ni-NTA beads (TransStart, China) and anti-c-Myc agarose affinity gel (Myc beads) or anti-FLAG M2 affinity gel (FLAG beads) for 5 h. The beads were rinsed five times with ice-cold wash buffer (20 mmol/L HEPES, pH 7.4, 500 mmol/L NaCl, 10% glycerin, 1% Triton X-100, and 1 mmol/L EDTA), the bound RNA was eluted and subjected to RT-qPCR and normalized with 1% input the indicated RNA, and the proteins were detected using immunoblotting.

Virus attachment and internalization assay
The cells in 12-well plates were washed with PBS (pH = 7.2) three times and then were infected with WSN virus (MOI = 5 or 10) on ice at 4˚C for 1 h. After washing with ice-cold PBS (pH = 7.2) three times to remove unbound viruses, the cells were frozen and thawed three times to release the bound viruses of the cells' surface into the supernatant. Using 1 mL DMEM collected the viruses and then titrated by a plaque assay on MDCK cells. Or after the second washing, the cells were harvested with lysis buffer for immunoblotting assay with anti-M1 mAb. Both assays help to detect the virus attachment. Or after the first washing, the cells were infected with WSN (MOI = 10) and cultured at 37˚C for 15, 30, and 60 min to allow for internalization. The cells were then washed with ice-cold PBS-HCl (pH = 1.3) five times to remove the attached but not-yet-internalized virions. And then, subjected to immunoblotting with anti-M1 mAb.

RNA pulldown assay
A biotin-labeled circVAMP3 probe targeting the unique circVAMP3 back splicing junction (BSJ) and control probe were purchased from Sangon Biotech (Shanghai, China). They were incubated with streptavidin C1 magnetic beads (Invitrogen, USA) for 15 min at room temperature and then incubated with cell lysates or in vitro transcribed circVAMP3 and purified NP or NS1 protein at 4˚C for 4 h. The bound proteins were then detected by immunoblotting.

Silver staining and mass spectrometry
The RNA pulldown samples were separated by 10% SDS-PAGE, after which the gels were stained with a Pierce Silver Stain for Mass Spectrometry Kit (Thermo Fisher Scientific, USA). The densest bands were subjected to analysis using nanoscale liquid chromatography-tandem mass spectrometry (nano-LC-MS/MS, LCQ Deca XP Plus; Thermo, USA). The data were analyzed by performing a MASCOT search against the SWISS-PROT human database and IAV database.

Animal modeling and in vivo infection
The mice were anesthetized with tribromoethanol, after which AAV6 expressing circVAMP3 purchased from Hanheng Company (Hanheng Biotechnology, China) was intratracheally instilled into the mice (5×10 10 vg/mL per mouse). After 21 days, the mice were challenged with the WSN strain at 10 4 PFU per mouse by nasal drip. Body weights and survival were recorded up to 14 days post-infection (d p.i.). The dry and wet lung weights of the mice were recorded on days 0, 3, 5, and 7. The lungs of the mice were harvested and placed in 10% buffered formalin for histological sectioning or 1 mL of PBS for detection of the viral titer. The fixed samples were then embedded in paraffin and used for immunohistochemistry (IHC) with the anti-M1 mAb or hematoxylin and eosin (H&E) staining. And then, the pathological scores of H&E were evaluated in a blinded manner by five graders who were unaware of the specific group under review. The sections were evaluated based on five indices, including PLOS PATHOGENS circVAMP3 suppresses IAV replication pulmonary interstitial edema, alveolar edema, inflammatory cell invasion, alveolar hemorrhage areas, and hyaline membrane development, and the score of each index ranged from 0 to 3 points.

Quantification and statistical analysis
Statistical analyses were performed using Prism version 8.0 software (GraphPad Software, USA). The statistical significance of differences between groups was determined using unpaired Student's t-test. Data are presented as the means ± SD. ns, not significant. A significant difference was determined when *p < 0.05 or **p < 0.01. The 293T cells were transfected with pLC5-circVAMP3 or pLC5-ciR-GFP for 24 h, followed by being infected with WSN (MOI = 1), and cell lysates were collected at 2 hp.i. and 4 hp.i. for RT-qPCR to detect the level of viral M1 RNAs. (C) Transfected the pLC5-circVAMP3 or pLC5-ciR-GFP into 293T cells at 12-well plates. And then, infected with WSN (MOI = 5), and cell lysate was harvested at the indicated time points. One-tenth of the cell lysate was taken for total protein level detection, and the rest was separated according to the operation instructions. And then, immunoblotted with the indicated antibodies. (D to F) The A549 stable cell lines were infected with WSN on the ice at 4˚C for 1 h, followed by a neutral wash (ice-cold PBS = 7.2). E and F MOI = 10. The cells were harvested for plaque assay (D) or immunoblotting (E). (F) After a neutral wash, the cells were cultured with DMEM at 37˚C for the indicated time points, followed by an acidic wash (ice-cold PBS-HCl = 1.3) before cell lysis. The internalized WSN was detected by immunoblotting. The data shown in A, B, and D are presented as the means ± SD; n = 3; ns is no significant difference, *p < 0.05, **p < 0.01. (TIF) S7 Fig. The single segment of circVAMP3 can't inhibit IAV replication. (A and B) The 293T cell were spread into 12-well plates for 16 h, and then transfected with pLC5-circVAMP3 or pLC5-ciR-GFP for 16 h. The rest of the cells were transfected with indicated 100 pmoL RNA segments for 8 h, respectively. Finally, all cells were infected with WSN at an MOI of 0.5 for 12 h. (A) The supernatants of infected cells were collected for plaque tests to determine the virus titer. (B) Cell lysates were collected and used for immunoblotting with corresponding antibodies. (TIF) S1 Table. Excel spreadsheet containing, in separate sheets, the underlying numerical data and statistical analysis for Figs panels 1B, 1C, 1F, 1G, 1H, 2C, 2D, 2E, 2F, 3A, 3B, 3D, 3E,  3G, 3H, 3J, 3L, 4C, 4D, 4F, 4G, 4J, 5C, 5D, 5E, 5F, 5G, 5J, 5K, 5L, 5M, 6A, 6B, 6C, 6G, 7A,  7B, 7C, 7H, S1C, S1D, S1E, S2D, S2E, S3A, S3B, S4C, S4E, S6A, S6B