Inhibition of Japanese encephalitis virus infection by the host zinc-finger antiviral protein

CCCH-type zinc-finger antiviral protein (ZAP) is a host factor that restricts the infection of many viruses mainly through RNA degradation, translation inhibition and innate immune responses. So far, only one flavivirus, yellow fever virus, has been reported to be ZAP-resistant. Here, we investigated the antiviral potential of human ZAP (isoform ZAP-L and ZAP-S) against three flaviviruses, Japanese encephalitis virus (JEV), dengue virus (DENV) and Zika virus (ZIKV). Infection of JEV but not DENV or ZIKV was blocked by ZAP overexpression, and depletion of endogenous ZAP enhanced JEV replication. ZAP hampered JEV translation and targeted viral RNA for 3′-5′ RNA exosome-mediated degradation. The zinc-finger motifs of ZAP were essential for RNA targeting and anti-JEV activity. JEV 3′-UTR, especially in the region with dumbbell structures and high content of CG dinucleotide, was mapped to bind ZAP and confer sensitivity to ZAP. In summary, we identified JEV as the first ZAP-sensitive flavivirus. ZAP may act as an intrinsic antiviral factor through specific RNA binding to fight against JEV infection.


Author summary
In addition to innate and adaptive immunities, many cellular proteins also exert antiviral activity against viral invasion. Human zinc-finger antiviral protein (ZAP) is a cellular restriction factor against many viruses but its role with regard to the flavivirus family is largely unknown. We tested the antiviral potential of ZAP against three flaviviruses and found that Japanese encephalitis virus (JEV) was ZAP-sensitive, while dengue virus and Zika virus were ZAP-resistant. ZAP specifically targets JEV viral RNA and induces translation repression and RNA degradation. Our findings highlight the ZAP-mediated anti-JEV mechanisms and extend the antiviral spectrum of ZAP to include a member of the Flavivirus genus. Introduction Zinc-finger CCCH-type containing, antiviral 1 (ZC3HAV1) also known as zinc-finger antiviral protein (ZAP) was first discovered in rats as a host antiviral protein against Moloney murine leukemia virus [1]. Later, multiple RNA and DNA viruses, like retroviruses, filoviruses, alphaviruses, and hepatitis B virus were shown to display sensitivity to ZAP [2][3][4][5][6]. However, ZAP does not induce a universal antiviral state, since viruses such as herpes simplex virus 1 (HSV-1), vesicular stomatitis virus and yellow fever virus (YFV) are resistant to ZAP [4]. Furthermore, viruses within the same family may exhibit different sensitivity to ZAP. For example, in the family Picornaviridae, ZAP inhibited coxsackievirus B3 but not poliovirus infection [4,7]. In humans, ZAP contains two major isoforms (ZAP-L and ZAP-S), which differ in the Cterminal region through alternative splicing [8]. In the N-terminus of ZAP, there are four CCCH-type zinc-finger motifs, which are required for RNA binding and antiviral property [9]. ZAP exhibits antiviral activity generally via posttranscriptional RNA regulation, such as mRNA decay and translation inhibition. ZAP recruits the RNA processing exosome complex and poly(A)-specific ribonuclease (PARN) to degrade the target RNA from the 3 0 -end [6,10]. ZAP also interacts with p72 RNA helicase to recruit decapping enzymes DCP1/DCP2 and exoribonuclease XRN1 to degrade the target RNA from the 5 0 -end [6,11]. Besides targeting RNA accumulation, ZAP can block translation of incoming viral RNA of Sindbis virus (SINV) [4] probably by interrupting the interaction between translational initiation factors eIF4A and eIF4G [12]. Moreover, ZAP-S can associate with retinoic acid-inducible gene I (RIG-I), a key sensor to recognize viral RNA, to promote the innate immune response and contribute to its antiviral potential [13].
Flaviviruses include numerous important human pathogens such as YFV, West Nile virus (WNV), Zika virus (ZIKV), dengue virus (DENV) and Japanese encephalitis virus (JEV), causing endemic or pandemic outbreaks in tropical and subtropical areas [14]. Flaviviral virions are enveloped and contain a single-stranded, positive-sense RNA genome with 5 0 -cap but not 3 0 -poly(A) tail. Flavivirus RNA encodes a single open reading frame (ORF) flanked by 5 0 -and 3 0 -untranslated regions (UTRs), which contain RNA secondary structures required for viral translation and transcription [15,16]. YFV was resistant to the antiviral activity of rat ZAP [4], but the susceptibility of other flaviviruses to ZAP is not clear. In this study, we determined the antiviral activity of ZAP against JEV, DENV, and ZIKV and found different viral responses to ZAP. We further demonstrated the antiviral mechanism of ZAP against JEV infection.

Overexpression of ZAP inhibits JEV infection
To evaluate the antiviral potential of human ZAP against members of flavivirus, we established A549 cells overexpressing ZAP-L and ZAP-S by lentiviral transduction. Cells with or without ZAP overexpression were infected with JEV, DENV or ZIKV and analyzed for viral replication. As compared with control EGFP cells, JEV infection measured by viral propagation, viral NS3 protein expression, and viral RNA replication were lower in cells ectopically overexpressing ZAP-L and ZAP-S (Fig 1A-1C). The inhibitory effect of ZAP-L/S against JEV infection was also noted in the induced pluripotent stem cells (iPSCs)-derived human neural progenitor cells (hNPCs), human microglia HMC3 cell line, as well as human neuroblastoma BE(2)C and SK-N-SH cell lines (S1 Fig). Nevertheless, no significant antiviral effect of ZAP was observed after high and low multiplicity of infection (MOI) of DENV and ZIKV (Fig 1D-1F, S2 and S3 Figs). Thus, different flaviviruses exhibit different sensitivity to the antiviral activity of ZAP.
Viruses may downregulate the expression of ZAP and/or block its functions to evade the ZAP-mediated antiviral action [17][18][19][20]. To understand why DENV was resistant to ZAP, we detected the endogenous protein levels of ZAPs in JEV or DENV-infected cells and no reduction of ZAPs was noted (S4 Fig). The antiviral activity of ZAP against SINV was also not obstructed by DENV infection (S5 Fig), implying that DENV did not antagonize the antiviral activity of ZAP. Other mechanisms, besides affecting ZAP expression and function, need to be considered.

Downregulation of ZAP enhances JEV replication
To investigate the antiviral potential of endogenous ZAP, A549-shZAP-L/S cells deprived of ZAP expression were established by transduction with lentiviral vector expressing shRNA targeting both ZAP-L and ZAP-S (Fig 2A). As compared to the control shLacZ cells, shZAP-L/S cells supported significantly higher levels of JEV infection as indicated by a 2.87-and 3.08-fold increase in viral RNA and viral progeny production, respectively (Fig 2B and 2C). Furthermore, knockdown of ZAP in human neuronal BE(2)C cells enhanced JEV replication, supporting the anti-JEV role of endogenous ZAP in neuronal cells (S6 Fig). Since JEV replication occurs at the surface of the endoplasmic reticulum (ER) membrane in the cytoplasm [21], and ZAP predominantly localizes to the cytoplasm [22], we examined whether ZAP co-localized with JEV viral RNA. A549 cells with or without JEV infection were processed for immunofluorescence analysis with anti-dsRNA antibody for viral replication complex [23,24] and anti-ZAP antibody for endogenous ZAP protein. Co-localization of dsRNA and ZAP was seen at the viral replication sites in JEV-infected cells (Fig 2D and 2E). Furthermore, we applied immunoprecipitation/RT-PCR assay using antibody against endogenous ZAP to evaluate the interaction of ZAP with JEV viral RNA. ZAP-L and ZAP-S precipitated by anti-ZAP antibody also pulled down JEV RNA (Fig 2F), indicating association between ZAP proteins and JEV RNA.

ZAP interacts with JEV RNA and exerts antiviral activity through its zincfinger motifs
To understand the anti-JEV mechanism of ZAP, we looked at whether RNA binding is required by generating constructs expressing the zinc-finger (ZF) domains deleted ZAP-L and ZAP-S (ZAP-L/S-del4ZFs) ( Fig 3A). We first tested the binding ability of wild type (WT) and ZF-deleted ZAP with JEV viral RNA by immunoprecipitation/RT-PCR assay. The V5-tagged WT ZAP-L and ZAP-S precipitated by anti-V5 antibody also pulled down JEV RNA, which was not seen in ZF-deleted ZAP and EGFP control ( Fig 3B). Furthermore, the anti-JEV activity of ZAP was greatly reduced in the ZF-deleted mutants as shown by viral NS3 protein expression, viral RNA replication, and viral progeny production (Fig 3C-3E). Thus, the zinc-finger motifs of ZAP were essential for RNA targeting and antiviral activity against JEV infection. To explore why DENV was resistant to the antiviral action of ZAP, we determined whether ZAP associated with DENV viral RNA. Interestingly, ZAP did not bring down DENV RNA while it readily pulled down the cellular TRAILR4 mRNA (S7 Fig) known to bind with ZAP [25], suggesting that RNA-binding ability may dictate the antiviral potential of ZAP.

ZAP interferes with viral translation and prevents accumulation of JEV RNA by destabilizing viral RNA
After entering the cells, JEV RNA undergoes a first round of translation to produce nonstructural proteins required for viral RNA replication. To assess whether viral translation was affected by ZAP, we detected viral protein expression in ZAP-and EGFP-overexpressing cells at early time points of JEV infection. Lower JEV NS3 protein expression was noted in cells with ZAP overexpression starting from 2 hours post-infection (hpi) (Fig 4A and 4B). We also monitored the viral RNA levels in cells with or without ZAP overexpression by RT-qPCR. No significant difference was found at 2 hpi, but ZAP significantly reduced JEV RNA since 3 hpi Total RNA and culture supernatants were harvested for RT-qPCR (B) and viral titration (C), respectively. Relative JEV RNA level normalized by GAPDH was determined using RT-qPCR. Viral titer was measured using plaque assay. Representative data from two independent experiments shown as mean ± SD (n = 3) were analyzed by two-tailed Student's t test. ÃÃ P 0.01. (D) Confocal microscopy of mock and JEV (MOI = 1) infected A549 cells at 16 hpi stained with anti-dsRNA and anti-ZAP antibodies. Cell nuclei were counterstained with DAPI. Scale bar = 20 μm. (E) Co-localization of dsRNA with ZAP was estimated by Pearson's correlation coefficient (PCC). Mean ± SD was calculated from 30 cells each group and the statistical significance was analyzed by two-tailed Student's t test. ÃÃÃ P 0.001. (F) Cell lysates from mock and JEV (MOI = 1) infected A549 cells ( Fig 4C). To further address whether the decrease of viral protein at early time point of JEV infection is through repressing translation and/or reducing viral RNA, we used replicationdead JEV replicon RNA (Fig 4D) to transfect 293T/17 cells with or without ZAP overexpression ( Fig 4E). The cell lysates were separated into two portions for measurements of luciferase activity and RNA level. Reduction of luciferase activity was noted in cells with ZAP overexpression since 1 h post-transfection when the RNA level was not yet affected (Fig 4F), suggesting repression of viral translation by ZAP. To clarify whether ZAP influenced JEV viral RNA stability, we also used replication-dead JEV replicon RNA for longer time points. A decrease in replicon RNA was noted at 3 h and 9 h post-transfection in cells with ZAP overexpression as compared to those with EGFP control (Fig 4G), suggesting that ZAP promoted JEV viral RNA decay. Thus, ZAP inhibited JEV translation and impaired viral RNA stability.
JEV suppression by ZAP is dependent on the 3 0 -5 0 RNA decay pathway ZAP modulates target RNA by recruiting cellular 5 0 -3 0 XRN1-dependent and 3 0 -5 0 RNA exosome-dependent RNA decay machineries [6,10]. To evaluate whether cellular RNA decay machineries participated in the anti-JEV activity of ZAP, we depleted the expression of XRN1 or EXOSC5 in ZAP-L, ZAP-S and EGFP-overexpressing cells by using a shRNA-targeting approach. Compared with the shLacZ control, knockdown of XRN1 did not affect the anti-JEV activity of ZAP ( Fig 5A-5C), whereas the anti-JEV effect of ZAP was hampered by knockdown of EXOSC5 (Fig 5D-5F). The data suggest that 3 0 -5 0 RNA degradation by the exosome complex is involved in the antiviral mechanism of ZAP against JEV infection.

ZAP enhances innate immune responses through the RIG-I signaling pathway
To address whether ZAP boosted innate immune responses during JEV infection, we measured the mRNA levels of type I IFN and proinflammatory cytokines in A549 cells with or without ZAP overexpression. As compared with the EGFP control, ZAP-L and ZAP-S slightly increased the basal level of IFN-β, TNF-α and IL-6 ( Fig 6A-6C), as well as chemokines CXCL10 and CCL5 (S8 Fig) in uninfected cells as previously reported [13], and further increased the expression of these cytokines/chemokines in JEV-infected A549 cells (Fig 6A-6C, S8 Fig). We further determined the upstream RLRs involved in the sensing events by knocking down the expression of RIG-I or MDA5 by lentiviral transduction (Fig 6D). Compared with the shLacZ control, knockdown of RIG-I, but not MDA-5, reduced the induction of interferon-stimulated genes (ISGs), e.g., IFIT1, IFIT3, and ISG15 that was seen in ZAP overexpressing cells (Fig 6E and 6F), indicating that ZAP-enhanced innate immune responses mainly went through the RIG-I signaling pathway in JEV-infected cells. Interestingly, knockdown of MDA5 even elevated the induction of ISGs (Fig 6F), similarly to a previous report showing higher IFN-β production in MDA5 -/but not RIG -/mouse embryonic fibroblasts infected with JEV [26]. Thus, deprive of MDA5 in JEV-infected cells might trigger some compensational IFN-related signaling events through unclear mechanism. Moreover, the anti-JEV effect of ZAP measured by viral NS3 protein expression was still noted in cells deprived of RIG-I with decreased ISGs expression ( Fig 6E)  involved in the anti-JEV effect, probably because JEV can block the JAK-STAT pathway and is somewhat resistant to type I IFN [27,28].  Both RNA sequence and structure have been considered as important for ZAP recognition [29,30]; however, the common features of ZAP-responsive elements (ZRE) are still inconclusive. To elucidate the possible regions of JEV genome targeted by ZAP, we performed UV crosslinking and immunoprecipitation (CLIP) to pull down the RNA interacting with ZAP in JEV-infected ZAP-S overexpressing A549 cells. The enriched RNA population was subjected to next generation sequencing (NGS) by use of the Ion Torrent platform. The read coverages of several peaks/regions were above the average read coverage especially within the JEV 3 0 -UTR ( Fig 7A). We thus focused on 5 0 -and 3 0 -UTR, which contain conserved complementary sequences and extensive secondary structures to regulate viral translation and transcription [15,16], to elucidate the potential ZRE in the JEV genome. We designed reporter constructs containing the 5 0 197 nt. (5 0 -UTR plus the first 102 nt. of the core gene) and/or the entire 3 0 -UTR flanking firefly luciferase (Fluc) in the pGL3-promoter vector under a SP6 promoter (Fig 7B, left panel). The in vitro transcribed reporter RNA and control Renilla luciferase (Rluc) RNA were cotransfected into 293T/17 cells with or without ZAP-S overexpression. As compared with the EGFP control, ZAP did not influence the luciferase activity of the reporter RNA with 5 0 -UTR 197 , while ZAP significantly reduced those with 3 0 -UTR (Fig 7B, right panel), revealing the possible ZRE in the JEV 3 0 -UTR. In vitro RNA pull-down assay also confirmed that the interaction between ZAP-S and JEV 3 0 -UTR in a ZF domain dependent manner (Fig 7C) was stronger than that with 5 0 -UTR 197 (S10 Fig).

Mapping the ZRE in JEV 3 0 -UTR
To further evaluate the possible ZRE in the 3 0 -UTR of JEV genome, CLIP-seq results within the three defined domains of JEV 3 0 -UTR: domain I (variable region), domain II (dumbbell structure) and domain III (3 0 conserved sequence and terminal stem-loop) [31] are shown (Fig 8A). We measured the binding ability of these RNA domains with ZAP by using biotinylated RNA. The proteins pulled down by streptavidin were subjected to western blotting with anti-V5 antibody for ZAP-S level. Domain II (showing a 45% binding ability of the full-length) had the strongest interaction with ZAP when compared to domain I (12% of the full-length) and domain III (3% of the full-length) of the JEV 3 0 -UTR RNA (Fig 8B). Interestingly, domain II contained high frequency of CG dinucleotide (Fig 8A), which was recently reported to confer ZAP binding and recognition [32]. Furthermore, the binding of ZAP to domain I+II RNA containing most of the CG dinucleotides reached 88% of the full length 3 0 -UTR RNA (Fig 8B). To verify the regions targeted by ZAP, we generated five reporter RNAs containing domain I, II, III, I+II, and II+III, respectively (Fig 8C, left panel). Significant reduction of luciferase activity by ZAP-S was noted in the reporters with WT, domain II, domain I+II and domain II+III, indicating the importance of domain II in conferring the sensitivity to ZAP (Fig 8C, right panel). Thus, domain II of JEV 3 0 -UTR containing dumbbell structures and high content of CG dinucleotide may function as ZRE and contribute to ZAP sensitivity. points, cells were collected and separated into two portions for the measurements of luciferase activity and RNA level, respectively. The relative luciferase activity and RNA level of Renilla luciferase reporter normalized with transfection control firefly luciferase are shown as the percentage to that of EGFP at 1 h post transfection. (G) 5 0 -capped RdRP-dead JEV replicon RNA and control firefly luciferase RNA cotransfected 293T/17-EGFP and -ZAP-S cells were harvested at 3 and 9 h post-transfection to determine the replicon RNA level. The relative replicon RNA level normalized with that of firefly luciferase is shown. Representative data are shown as mean ± SD from 3 independent experiments and analyzed by two-tailed Student's t test. Ã P 0.05; ÃÃ P 0.01; ÃÃÃ P 0.001.
https://doi.org/10.1371/journal.ppat.1007166.g004 Discussion ZAP exhibits antiviral activity against a variety of viruses, but flaviviruses were not known to be sensitive to ZAP until this study. Here, we demonstrate that overexpression of human ZAP (isoforms ZAP-L and ZAP-S) inhibited JEV infection and downregulation of endogenous ZAP enhanced JEV replication, indicating the intrinsic antiviral potential of ZAP. We also found that, similar to YFV [4], DENV and ZIKV are resistant to ZAP, supporting the notion that ZAP is not a universal antiviral factor even for viruses within the same family.
Viruses may use different strategies to evade the ZAP-mediated antiviral action, but it is still not fully understood what determines the sensitivity of a virus to ZAP. Influenza A virus NS1 protein antagonizes ZAP-S by interrupting its binding to target mRNA [19]. Murine gammaherpesvirus 68 RTA inhibits the antiviral activity of ZAP by disrupting the N-terminal intermolecular interaction of ZAP [20]. HSV-1 UL41 endoribonuclease was identified as an antagonist of human ZAP by degrading its mRNA [17]. Enterovirus 71 3C protease mediates the cleavage of ZAP protein [18]. In our hands, DENV neither downregulated the expression of ZAP nor blocked the antiviral activity of ZAP (S4 and S5 Figs). However, we were able to detect the interaction of ZAP with JEV RNA (Figs 2F and 3B) but not with DENV RNA (S7 Fig), consistent with the antiviral potential of ZAP against JEV but not DENV. Thus, the resistance of DENV to ZAP might be resulted from a failure of ZAP to bind with DENV viral RNA.
Recently, viral RNA with high CG dinucleotide content was found to be targeted by ZAP, and ZAP bound directly and selectively to RNA sequences containing CG dinucleotides [32]. The known ZAP-sensitive virus genomes contained relatively higher CG frequencies, such as ZAP-sensitive SINV (0.9) vs. ZAP-resistant YFV (0.38) [32]. Interestingly, the ZAP-sensitive JEV genome also has a relatively higher CG frequency (0.61) as compared to that of ZAP-resistant ZIKV (0.46) and DENV (0.36). Moreover, we detected the interaction of ZAP with JEV RNA (Figs 2F and 3B) but not with DENV RNA (S7 Fig). Thus, JEV RNA enriched with CG dinucleotides may be targeted by ZAP binding to trigger the antiviral effect. Our crosslinkingimmunoprecipitation-sequencing assay showed several peaks of ZAP binding in the JEV 3 0 -UTR, and domain II of JEV 3 0 -UTR with high frequencies of CG dinucleotide was further mapped as the major ZRE (Figs 7 and 8). Since domain II also contains stem-loop and pseudoknot structures [31,33], the contribution of RNA secondary structures in addition to CG dinucleotides in ZAP recognition cannot be ruled out.
ZAP-S serves as a key regulator of RIG-I-mediated innate immune responses for type I IFN production to limit the infection of influenza A virus and Newcastle disease virus. Besides, ZAP-S also enhances NF-κB and IRF3 signaling downstream of RIG-I for the induction of proinflammatory cytokines, like TNF-α and IL-6 [13]. In this study, we showed that ZAP promoted the production of IFN-β, TNF-α and IL-6 during JEV infection (Fig 6A-6C). It has been speculated that elevated IFNs, TNF-α, and IL-6 during viral infection correlated with the severity and outcome of viral diseases. Similarly, IFNs and proinflammatory cytokines including TNF-α and IL-6 were elevated in patients with acute Japanese encephalitis [34][35][36]. Thus, the induction of IFN and proinflammatory cytokines/chemokines by ZAP may contribute to host defense responses as well as the JEV-induced inflammatory states and pathogenesis during JEV infection.
ZAP restricted JEV infection mainly by posttranscriptional regulation such as blocking protein translation and enhancing RNA degradation. Through interaction with XRN1 and exosome components, like Rrp46 (EXOSC5), Rrp40 (EXOSC3) and Rrp42 (EXOSC7) [6,10], ZAP utilizes cellular RNA decay machineries to destabilize both viral and cellular RNAs. ZAP binds with 5 0 -UTR of HIV-1 nef mRNA and 3 0 -UTR of cellular TRAILR4 mRNA and then requires RNA exosome, and potentially XRN1, to degrade the target RNAs [6,25]. The anti-JEV effect of ZAP was diminished by knockdown of the exosome component (Fig 5D-5F), indicating the involvement of 3 0 -5 0 RNA decay in the ZAP antiviral pathway. Thus, ZAP can bind with JEV RNA and target the viral RNA to the 3 0 -5 0 RNA exosome complex for RNA degradation. Furthermore, XRN1-dependent RNA decay is known to generate subgenomic flavivirus RNA (sfRNA) [37,38], which can then block XRN1 activity and alter host mRNA stability [39]. Interestingly, XRN1 was not involved in the anti-JEV activity of ZAP (Fig 5A-5C), probably due to the interplay between XRN1 and sfRNA in JEV-infected cells prevailing the involvement of XRN1 in the antiviral action of ZAP. Since ZAP not only restricts viral infection but also regulates cellular mRNA abundance [25], we cannot exclude the possibility that the blocking effect of ZAP on JEV is due to the altered cellular mRNA and protein expression. Overall, we have identified JEV as the first flavivirus sensitive to human ZAP and provide insight about the antiviral mechanism of ZAP against viral RNA without 3 0 -poly(A) tail.

Reverse transcription, RT-PCR and real-time quantitative PCR (RT-qPCR)
Total RNA was extracted using RNeasy Mini Kit (Qiagen). For RT-PCR, 1 μg of RNA was reverse-transcribed by random hexamer primer with SuperScript III First-Strand Synthesis System (Invitrogen). PCR was then performed by using the primers described in S1 Table. For real-time qPCR, random hexamer primer was used for reverse transcription and qPCR was performed by ABI-Prism 7500 real-time PCR system (Applied Biosystems). The relative RNA levels of specific RNA were normalized with GAPDH or firefly luciferase, and calculated by the comparative threshold cycle (ΔΔCt) method. The TaqMan Universal PCR Master Mix with UNG (Invitrogen) and commercial probes for human GAPDH (Hs02758991), IFN-β (Hs01077958), IL-6 (Hs00985639), TNF-α (Hs01113624), CCL5 (Hs00982282), CXCL10 (Hs01124251) and firefly luciferase (Mr03987587) (Applied Biosystems) were used in qPCR reactions. The JEV viral RNA primers for qPCR have been described previously [48].
Protein-RNA complexes were washed three times by RNA binding buffer without heparin and yeast tRNA. After washing, 30 μl of 2X SDS-PAGE sample buffer was added and incubated for 10 min at room temperature. The pull-down proteins were further analyzed by western blot.

Measurement of replication-defective replicon RNA stability
5 0 -capped Renilla luciferase containing JR2A-NS5mt replicon RNA and control firefly luciferase RNA generated by in vitro transcription were cotransfected into EGFP, ZAP-L, and ZAP-S overexpressing 293T/17 cells for 3 and 9 h. Total RNA was collected for detection of incoming RNA by RT-qPCR. The relative JEV replicon RNA levels were normalized with that of firefly luciferase.

Reporter assay
5 0 -capped firefly luciferase containing reporter RNA and control Renilla luciferase RNA by in vitro transcription were cotransfected into EGFP and ZAP-S overexpressing 293T/17 cells. The relative luciferase activity was assessed by dual-luciferase reporter assay system.

Cross-linked immunoprecipitation followed by next generation sequencing (CLIP-seq)
CLIP assay was performed based on a previously published protocol [49]. A549-ZAP-S-V5 cells infected with JEV (MOI = 5) for 24 h were UV-crosslinked at 254 nm with 200 mJ/cm 2 using UV Stratalinker 2400 (Stratagene). Cells were lysed with CLIP lysis buffer containing protease inhibitor cocktail (50 mM Tris-HCl [pH7.4], 100 mM NaCl, 1% NP-40, 0.1% SDS, and 0.5% sodium deoxycholate), treated with 10 μL of RQ1 DNase (Promega) and 10 μL of RNase A/T1 (1:500 dilution) (EN0551, Thermo Fisher Scientific) on a shaker at 1000 rpm for 5 min at 37˚C. After treatment, cell lysates were immunoprecipitated with preincubated anti-V5 antibody-protein G complex (Anti-V5 antibody, R960-25, Invitrogen; Protein G Mag Sepharose Xtra, 28-9670-70, GE Healthcare) for 4 h at 4˚C. Samples were then washed with CLIP lysis buffer containing 500 mM NaCl and CLIP lysis buffer twice, respectively. Protease K (4 mg/mL) in protease K buffer (100 mM Tris-HCl [pH7.4], 50 mM NaCl, and 10 mM EDTA) was preincubated for 20 min at 37˚C to ensure digestion of RNase. The RNA-proteinantibody immune complexes were incubated on a shaker at 1000 rpm for 20 min at 37˚C. After digestion by protease K, the bound RNA was purified by Direct-zol RNA kit (Zymo Research). The sequence library was constructed using Ion Total RNA-seq Kit v.2 (Thermo Fisher Scientific) and the library was sequenced by Ion Torrent PGM system (Thermo Fisher Scientific). Downstream data analysis was performed by use of the CLC Genomics Workbench 11.0.1 (Qiagen). The alignment of trimmed reads to JEV RP-9 reference sequence (GenBank accession: AF014161) was based on the following parameters: match score = 1; mismatch cost = 2; insertion/deletion cost = 3; length fraction = 0.9; similarity fraction = 0.9; Non-specific match handling = ignore. Read coverage showed the read numbers of each position mapping to JEV genome divided by the total read numbers mapping to JEV genome. The CLIP-seq data have been deposited in the NCBI GEO data repository with accession code GSE115747.

Statistical analysis
Two-way AVONA was used to estimate the statistical significance of viral replication kinetics data. Two-tailed Student's t test was used to estimate the statistical significance between two groups. Representative data from repeated independent experiments are shown as mean ± standard deviation (SD) with triplicate samples (n = 3). P 0.05 was considered statistically significant. Ã P 0.05; ÃÃ P 0.01; ÃÃÃ P 0.001; NS, not significant.