Development of a Viral RdRp-Assisted Gene Silencing System and Its Application in the Identification of Host Factors of Plant (+)RNA Virus

Gene silencing induced by hairpin RNA or virus infection expression is one of the major tools in genetics studies in plants. However, when dealing with essential genes, virus-induced gene silencing (VIGS) and transgenic expression of hairpin RNA could lead to plant death, while transient expression of hairpin RNA in leaves is often less competent in downregulating target gene mRNA levels. Here, we developed a transient double-stranded RNA (dsRNA) expression system assisted by a modified viral RNA-dependent RNA polymerase (RdRp) in plant leaves. We show that this system is more effective in inducing gene silencing than the intron-spliced hairpin RNA expression. Furthermore, by using this system, we tested the role of the early secretory pathway during infection of Soybean mosaic potyvirus (SMV). We found that key components of the coat protein complex II vesicles are required for the multiplication of SMV. Overall, this dsRNA-based gene silencing system is effective in downregulating plant gene expression and can be used to identify host genes involved in plant-virus interactions.


INTRODUCTION
Gene silencing is a cellular mechanism that acts on both transcriptional and posttranscriptional levels to regulate gene expression. Double-stranded RNA (dsRNA) is the key trigger of gene silencing and can be artificially delivered by many vector-based systems (Wesley et al., 2001;Schwab et al., 2006;Senthil-Kumar and Mysore, 2014). In plants, dsRNA in the form of an intron-spliced hairpin RNA (ihpRNA) was demonstrated to be effective in inducing the posttranscriptional gene silencing (PTGS), comparing with the single-stranded sense or antisense RNA fragments or the hairpin RNA without intron as the spacer (Smith et al., 2000;Wesley et al., 2001). However, the ihpRNA could not lead to total degradation of target mRNA and often allow detectable levels of target gene activity (Wesley et al., 2001). The ihpRNA transgenic Arabidopsis plants targeting, for example, the polyphenol oxidase encoding gene PPO or the ethylene signaling gene EIN2 showed a 70% or 66% silencing rate in the progeny plants (Wesley et al., 2001). A model system based on transient expression of ihpRNA in Nicotiana benthamiana leaves showed that the transcripts level of targeted ectopically expressed β-glucuronidase or green fluorescence protein-coding gene only reduced 76% or 64% (Yan et al., 2012). The remaining transcripts of target genes could still be translated and contribute to the phenotype, thus limiting the use of ihpRNA-induced PTGS in genetics studies.
Besides ihpRNA, plant (+)RNA viruses are often used as vectors to deliver dsRNA produced during the viral RNA replication Zhang and Ghabrial, 2006;Yuan et al., 2011). The virus-induced gene silencing (VIGS) is a robust way to downregulate the expression of target gene transiently. However, the VIGS system does have limitations. Some rescued viruses could, for example, cause symptoms on the host plants (Ratcliff et al., 2001;Zhang and Ghabrial, 2006;Yuan et al., 2011) or change the metabolism of the infected cells (Fernandez-Calvino et al., 2014), thus could potentially bring complications in the interpretation of the genetic data. Limited viral host range, compatible interaction between the virus and some cultivated varieties, and the regulation on biosafety and environmental release could also potentially hamper the use of VIGS (Purkayastha and Dasgupta, 2009).
In this study, we developed a plant-based virus-free dsRNA delivery system that takes advantage of a cytoplasm-localized viral RNA-dependent RNA polymerase (RdRp) to amplify the single-stranded RNA template into dsRNA. This system does not require a live virus to generate the silencing signal. This viral RdRp-assisted silencing (VRAS) system reduced the mRNA of the target gene to a lower level than the ihpRNA. We further showed that this VRAS system could be used to identify host genes involved in plantvirus interactions.

Configuration of the Viral RdRp-Assisted Gene Silencing System
To circumvent the use of infectious virus but still exploit the ability of viral RdRp in amplifying the RNA template, we have developed a two-vector-based system to initiate the synthesis of dsRNA in vivo for the induction of gene silencing. First, we cloned the C-terminal part of the p82 (p82C) RdRp from the tobacco necrosis virus A (TNV-A) Chinese isolate (Xi et al., 2008). The p82 protein is produced by translational readthrough at the amber stop codon of the first ORF. The C-terminal part of p82 includes the amino acids downstream of the amber stop codon and contains the RdRp motif (Xi et al., 2008). TNV-A is a type member of the Alphanecrovirus genus in the Tombusviridae family. Its replication is initiated by the recognition of viral RdRp to the 3 -end of viral (+) or (−) RNA sequence, also known as the viral promoter (Song and Simon, 1994). We cloned the p82C RdRp under the control of a 35S promoter for plant expression ( Figure 1A). The sequence of the gene targeted for silencing was then placed between a 35S promoter and a TNV-A promoter ( Figure 1A). A Hepatitis delta virus antigenomic ribozyme (Rz) (Ferre-D'Amare et al., 1998) was placed at the downstream end of the TNV-A promoter to ensure the release of the promoter. This configuration of double promoter allows the initial RNA polymerase II (POL II)-mediated synthesis of the target gene sequence and the following TNV-A promoterdriven p82C RdRp-mediated synthesis of the complementary sequence, thus producing plenty of dsRNA for targeted gene silencing ( Figure 1B).

Optimization of the Viral RdRp-Assisted Gene Silencing System
We optimized this system in two aspects: to find the more active viral promoter sequence and to choose the optimal subcellular localization of the viral RdRp. The amount of products made by viral RdRp is affected by the context of the promoter sequences that form different RNA structures (Song and Simon, 1994;Koev et al., 2002;Rajendran et al., 2002;Sun and Simon, 2006). The promoter for positive-strand RNA synthesis is often more active than the promoter for negative-strand RNA synthesis (Koev et al., 2002;Rajendran et al., 2002). We have cloned both positive-and negative-strand initiation promoters. Also, since the sense-strand of the target sequence is synthesized in the nucleus by RNA polymerase II, a p82C fused with nuclear localization sequence (p82C-NLS) was designed to test whether this nucleus-localized version is more efficient in the induction of gene silencing.
A 436-bp sequence from the Nicotiana benthamiana phytoene desaturase (PDS) open reading frame was cloned and inserted into the vector containing TNV-A promoter for either positiveor negative-strand viral RNA synthesis [pTNV(+)-PDS or pTNV(−)-PDS]. Each of these two vectors was agroinfiltrated into N. benthamiana leaves alone, with p82C-NLS, or with p82C ( Figure 1C). After ∼2 days, the total RNAs were extracted from the infiltrated leaves and subjected to firststrand cDNA synthesis using oligo(dT)18 as a primer. The level of PDS mRNA accumulation in the total RNAs was detected by semiquantitative reverse transcription polymerase chain reaction (semiquantitative PCR). The housekeeping gene encoding α-tubulin was used as an internal control. The results showed that coexpression of p82C and mRNA containing PDS sequence driven by TNV-A positive-strand initiation promoter greatly reduced the PDS mRNA level ( Figure 1C, lanes 19-21). In the absence of p82C, the transcription of PDS target sequence could only moderately reduce the PDS mRNA level ( Figure 1C, lanes 4-9). The target gene degradation in the absence of p82C might be due to the ribozyme cleavage that produced a poly(A)-lacking mRNA, which can be recognized by endogenous plant RdRp (RDR6) to trigger gene silencing (Baeg et al., 2017). Moreover, we observed no additional effect for co infiltration of vectors carrying p82C-NLS and the template RNA for the target gene ( Figure 1C, lanes 10-15 vs. lanes 4-9), suggesting that the nucleus localized p82C could not induce gene silencing. At 10 days postinfiltration (dpi), the "pTNV(+)-PDS + p82C"-infiltrated leaves display obvious albino phenotype, suggesting successful silencing of the PDS gene ( Figure 1D).

Involvement of the RdRp in the dsRNA Synthesis
The obtained data also showed that the promoter for positivestrand RNA synthesis [pTNV(+)] is more efficient in inducing PDS silencing than the promoter for negative-strand RNA synthesis [pTNV(−)] when coexpressed with p82C ( Figure 1C, lanes 19-21 vs. lanes 16-18). We proposed that the positivestrand initiation promoter of TNV-A is more robust in driving dsRNA synthesis. To test this hypothesis, we performed the in vitro RNA synthesis assay. An N-terminally maltose-binding protein (MBP) tagged p82C was expressed in Escherichia coli and purified by affinity chromatography using amylose resin (Figure 2A). The recombinant protein MBP-p82C was then incubated with in vitro-transcribed RNA template containing PDS gene fragment followed by either positive or negative promoter of TNV-A. The complementary strand of the RNA template was synthesized by MBP-p82C in the presence of radioactively labeled rNTPs. The resulting products were either heated to denature the dsRNA or unheated to keep the dsRNA formation and then separated by non-denaturing polyacrylamide gel electrophoresis (PAGE) and imaged using a phosphorimager. Upon heat treatment, the dsRNA unwinded, and the radioactively labeled RNA products migrated as singlestranded RNA corresponding to the template size ( Figure 2B, heated). The MBP-p82C synthesized ∼4 times more products using a TNV-A positive-strand initiation promoter-driven template ( Figure 2B, lane 1 vs. lane 2). While in the unheated samples, most of the newly synthesized ssRNA products hybridized with the template as dsRNA (Figure 2B, unheated). The remaining RNA products kept in ssRNA format were probably due to weak terminal transferase activity copurified with MBP-p82C that can label the template RNA or due to insufficient dsRNA winding ( Figure 2B, lanes 3-4, ssRNA band). Nevertheless, ∼5 times more dsRNAs were produced when using the viral positive-strand initiation promoter ( Figure 2B, lane 3 vs. lane 4). Overall, the positive-strand initiation promoter supported 4∼5-fold higher complementary RNA synthesis than the promoter for the negative strand. These results demonstrated that the higher silencing efficiency in the "pTNV(+)-PDS + p82C" combination is due to more active promoter activity that leads to more dsRNA production.
Since the p82C-NLS cannot effectively induce PDS silencing ( Figure 1C), thus we tested whether the nuclear localization sequence (NLS) alone has any effect on the RdRp activity of p82C. We found that the recombinant protein containing the nuclear localization sequence (MBP-p82C-NLS) has comparable in vitro RdRp activity with MBP-p82C ( Figure 2C), suggesting that the nuclear localization rather than the attachment of NLS caused ineffective PDS silencing.

Comparison of VRAS and ihpRNA-Induced Gene Silencing
To compare the VRAS system with the intron-spliced hairpin RNA in the efficiency of inducing gene silencing, we constructed two binary vectors that drive the transcription of ihpRNA under the control of a 35S promoter. One contains a 190-bp-long intron of castor bean catalase gene (CAT1) (Tanaka et al., 1990). The other vector contains a petunia chalcone synthase A (chsA) intron that is 1,349-bp long in length (Kerschen et al., 2004). The same 436 bp gene fragment of PDS used for p82C-mediated silencing was cloned in both sense and antisense orientations on each side of the intron loop to form the typical ihpRNA structure. Each of these PDS-ihpRNA expression vectors, the VRAS vectors "pTNV(+)-PDS + p82C, " and the empty vector were agroinfiltrated into N. benthamiana leaves. Infiltrated leaves were collected ∼3 days after agroinfiltration and subjected to mRNA analysis using semiquantitative PCR and quantitative PCR (qPCR). The results showed that the PDS mRNA reduced to 41% in the leaf samples infiltrated with the VRAS vectors compared with the empty control at 3 dpi. In the PDS-ihpRNA silenced leaf samples, PDS mRNA reduced to 52 or 50%, suggesting that the VRAS system is more efficient in silencing the PDS gene (Figures 3A,B).
We also tested the silencing of the SECRETION-ASSOCIATED RAS 1 (SAR1) encoding an Arf family small GTPase. SAR1 regulates protein export at the ER export sites (ERES) and is essential in plant growth and development (Hanton et al., 2008;Liang et al., 2020). The VRAS system reduced the SAR1 mRNA level to 36% at 3 dpi ( Figure 3D), while the SAR1 mRNA only reduced to 69% or 73% in SAR1-ihpRNA-expressed leaves. Thus, we conclude that the VRAS is more robust than the ihpRNA in inducing gene silencing when transiently expressed.

Using VRAS to Study Plant-Virus Interactions
To test whether VRAS can be applied in genetics studies in plantvirus interactions, we tested the effect of VRAS-mediated SAR1 silencing on the replication of Soybean mosaic potyvirus (SMV). SMV belongs to the potyvirus genus, potyviridae, the replication of which takes place on the cytoplasmic surface of reorganized intracellular membranes (Wei and Wang, 2008;Cabanillas et al., 2018;Wu et al., 2020). Membrane trafficking of the early secretory system was shown to be required for the replication of several (+)RNA viruses (Belov et al., 2007;Wei and Wang, 2008;  To test the role of SAR1 in SMV replication, VRAS vectors expressing pTNV(+)-SAR1 and p82C were coinfiltrated with the infectious clone of SMV into N. benthamiana leaves. Empty vector or ihpRNA containing SAR1 coding region and chsA intron (SAR1-ihpRNA) were also coinfiltrated with SMV, respectively, as controls. Five days after infiltration, leaf samples were collected for protein and RNA analysis. We found that during SMV replication, SAR1 mRNA reduced to 17% in VRAStreated samples at 5 dpi comparing with the empty vector control (Figures 4A,B), while ihpRNA-mediated SAR1 silencing could only reduce the SAR1 mRNA to 76%. Accordingly, SMV accumulation reduced to 26% in samples silenced for SAR1 by VRAS but kept unchanged in samples expressing ihpRNA ( Figure 4A). These results demonstrated that SAR1 is required for SMV replication. The unchanged viral replication in SAR1-ihpRNA-infiltrated leaves may be because SAR1 expression contributed by the 76% non-silenced SAR1 mRNA is enough for the optimal SMV replication ( Figure 4B). In addition, we have shown that p82C expression alone does not affect SMV replication ( Figure 4C).
Furthermore, we expressed SAR1 dominant-negative (DN) mutant in N. benthamiana replicating SMV via agrobacteriummediated transient expression. The SAR1-DN can block the wild-type SAR1 function in regulating the early secretory pathway (Takeuchi et al., 2000). The expression of SAR1-DN (AtSAR1-H74L) led to reduced SMV accumulation (Figure 4D), confirming the essential role of SAR1 in SMV replication.
The potyviral Hc-Pro is a well-documented viral suppressor of gene silencing (Valli et al., 2018). We found that transiently expressed Hc-Pro under the control of 35S promoter strongly inhibited both VRAS-or ihp-RNA-mediated SAR1 silencing ( Figure 4E). In contrast, SAR1 can be silenced during SMV infection ( Figure 4B). It is likely that SMV-driven Hc-Pro expression is much weaker in suppressing gene silencing, thus allowing the induction of gene silencing for the gene function study in SMV replication.
Since SAR1 is required for the formation of coat protein complex II (COPII)-coated vesicles, other COPII components or proteins that regulate COPII vesicle transport might also affect SMV replication. To test this hypothesis, genes encoding Frontiers in Microbiology | www.frontiersin.org guanine-nucleotide exchange factor (SEC12), inner-coat of COPII (SEC23 and SEC24), the outer-coat of COPII (SEC13 and SEC31), a peripheral membrane residing scaffold protein required for COPII vesicle formation (SEC16), or Ras-related protein RAB1 GTPase that regulate COPII to Golgi transport (Takeuchi et al., 2000) were silenced by using the VRAS vectors in the presence of SMV infection. SMV accumulation reduced to less than 50% in the leaf tissues silenced for SEC13, SEC31, and RAB1 (Figures 5B,D). Silencing of SEC12, SEC16, SEC23, and SEC24 also reduced SMV accumulation to ∼65-70% (Figures 5A-C). Semiquantitative PCR analysis revealed that the target genes encoding for the COPII components were successfully silenced by the VRAS vectors, while the level of the α-tubulin coding gene was not changed (Figure 5, lower panels). Simultaneous silencing of SAR1, SEC13, and SEC31 also reduced SMV accumulation to ∼35%, similar to the level of SAR1 silencing (Figure 5E). These results demonstrated that the COPII components are required for efficient SMV infection, and the VRAS vectors can be used for the genetics studies of virus-plant interactions.

DISCUSSION
The ihpRNA vectors are among the most popular tools in genetics study. The ihpRNA is transcribed into pre-mRNA under the control of a promoter and then subsequently spliced into an intron-less hairpin RNA in the nucleus. The splicing event was thought to help align the arms of the hairpin RNA into the dsRNA form, thus improving the efficiency in gene silencing Frontiers in Microbiology | www.frontiersin.org (Smith et al., 2000). Alternatively, splicing was thought to create a loop-less duplex RNA that is probably more nuclease-resistant and stable (Smith et al., 2000). However, a later study used a hpRNA vector containing a spacer region plus an intron between the arms and showed similar silencing efficiency compared with regular intron-containing ihpRNA, which disagreed with the earlier hypotheses (Wesley et al., 2001). Despite the debatable reason, the intron-splicing event undeniably plays a positive role in inducing gene silencing.
In this study, we developed a VRAS system. The VRAS vectors were demonstrated to be more robust in achieving a higher degree of silencing than the ihpRNA (Figures 3, 4). We propose that the VRAS system is advantageous due to the following reasons. In contrast to ihpRNA that forms the RNA duplex in the nucleus, the TNV-A p82C RdRp produces the dsRNA in the cytoplasm. Accordingly, we also observed that the nucleus-located p82C-NLS was inefficiently inducing gene silencing ( Figure 1C). The transport of mRNA in the nucleus requires association of RNA binding or modifying proteins during the pre-mRNA processing and consumes ATP (Vargas et al., 2005). It is possible that nuclear export of hairpin structure containing pre-mRNA or dsRNA is the rate-limiting step, and only the exported RNAs can be subjected to the dicer cleavage during the induction of gene silencing.
Although ihpRNA can be successfully applied to investigate gene functions in plant development, it was less frequently used for genetics studies in virus-plant interactions comparing with other methods, especially when transiently delivered (e.g., VIGS or transient expression of dominant-negative mutant). Our study found that the transient expression of ihpRNA cannot sufficiently induce silencing of SAR1 during SMV infection. In contrast, transient expression of VRAS-based SAR1 silencing vectors or a SAR1 dominantnegative mutant (H74L) led to inhibited SMV replication and showed advantages over the ihpRNA-based method. The employment of VRAS is currently demonstrated by using the agrobacterium-mediated transient expression assay, which could only be done in a limited number of plants. Further exploration on the stable genetic transformation of VRAS components would expand the application of this system to more plant species.
During the COPII vesicle budding process, GDP-bound SAR1 is activated by its guanine nucleotide exchange factor (GEF) SEC12 into GTP-bound form and becomes membrane associated at ERES (Barlowe and Schekman, 1993). SAR1 forms complex with SEC23 and SEC24 to attract cargo proteins and then recruits SEC13 and SEC31 to form the COPII coat in order to bud from ER (D' Arcangelo et al., 2013). Rab1 GTPase labels the uncoated COPII vesicles and facilitates the Golgi targeting (Westrate et al., 2020). A previous study showed that SAR1 is required to form Tobacco etch potyvirus (TEV) 6K2-positive replication vesicles, and expression of SAR1 dominant-negative mutant inhibited TEV infection (Wei and Wang, 2008). Furthermore, SEC23 and SEC24 can colocalize with TEV 6K2 replication vesicles (Wei and Wang, 2008), and SEC24 interacts with Turnip mosaic potyvirus (TuMV) 6K2 to facilitate systemic viral movement (Jiang et al., 2015). However, whether SEC23, SEC24, or other COPII components are required for potyvirus replication had not been measured. Based on our genetic study using VRASbased gene silencing, we found that these COPII components are required for optimal SMV replication. Overall, our work demonstrated that the VRAS system described is a valuable tool in inducing gene silencing and can be applied to the genetics study of virus-plant interactions.
To construct the ihpRNA-containing plasmids, the 436-bp sequence from the gene coding for PDS was amplified from the N. benthamiana cDNA using PCR with primer pair 1113/1114. The resulting PCR product was digested with XbaI/EcoRI and inserted into AvrII/EcoRI-digested pGD-35S-CAT1 or pGD-35S-chsA, respectively. The resulting plasmids were digested with PstI/XbaI and ligated with PDS gene fragment that was amplified with primer pair 1113/1272 using PCR, resulting in pGD-35S-CAT1-NbPDS-hairpin or pGD-35S-chsA-NbPDS-hairpin.

Agroinfiltration of Nicotiana benthamiana Leaves
Overnight cultures of the Agrobacterium strain EHA105 carrying appropriate vectors grown in a 29 • C incubator were collected by centrifugation at 4,500 rpm for 15 min. The collected cells were resuspended in MMA media (10 mM MgCl 2 , 10 mM MES, and 200 µM acetosyringone) to reach an optical density measured at a wavelength of 600 nm (OD 600 ) of 1.0. The cultures were mixed according to the requirement of each experiment and infiltrated to the underside of leaves of 4-week-old N. benthamiana plants using a 1-ml syringe without a needle.

RNA Extraction and Reverse Transcription Polymerase Chain Reaction
The total RNAs were extracted from N. benthamiana leaves. The leaf disks were grounded to fine powders in a 1.5-ml tube with a pestle after freezing in liquid nitrogen. Then 400 µl RNA extraction buffer (0.1 M glycine, 0.1 M NaCl, 0.1 M EDTA pH 8.0, 1% SDS) and 400 µl RNA phenol were added into the tube, briefly mixed, and centrifuged at 15,000 rpm for 10 min at 4 • C. About 300 µl upper aqueous phase was transferred into a 1.5-ml tube containing 300 µl water-saturated phenol-chloroform (1:1), mixed thoroughly, and centrifuged at 15,000 rpm for 10 min at 4 • C. Then, 200 µl upper aqueous phase was transferred into a 1.5-ml tube containing 500 µl 100% ethanol with 2% (V/V) 3M sodium acetate (pH 5.6). The precipitated total RNA was subjected to centrifugation at 15,000 rpm for 30 min at 4 • C, washed with 70% ethanol, and then dissolved in 40 µl dH 2 O. The concentration of the total RNA was measured with an ultraviolet spectrophotometer and then adjusted to 250 ng/µl. About 1 µg RNA was subjected to reverse transcription with oligo(dT)18 by using the M-MLV RTase (Takara, Code No. 639575) followed by semiquantitative PCR analysis. The cDNA was also subjected to qPCR analysis using the ChamQ SYBR Color qPCR Master Mix (Vazyme, Cat. # Q411-02). Relative gene expression levels were calculated using the 2 − Ct method. Primer pairs 1045/1044, 1480/1482, 1483/1485, 1486/1488, 1489/1491, 1492/1494, and 1495/1497 were used for detection of the mRNA levels of SAR1, SEC12, SEC13, SEC16, SEC23, SEC24, and SEC31, respectively.
The in vitro RNA-dependent RNA polymerase reaction was performed in a 50-µl reaction mix containing 50 mM Tris-HCl at pH 8.2, 10 mM MgCl 2 , 10 mM dithiothreitol, 100 mM potassium glutamate, 1.0 mM each ATP, CTP, and GTP, 0.01 mM UTP, 0.2 µl [32-P] UTP, 0.2 µl RNase inhibitor, 0.3 µg MBP-p82C, and 0.3 µg RNA transcripts. The reaction mix was incubated at 25 • C for 2 h and terminated by the addition of 100 µl stop buffer (1% SDS, 0.05 M EDTA at pH 8.0). After extraction with water-saturated phenol-chloroform (1:1), precipitation in isopropanol containing 10% (V/V) 10 M ammonium acetate, and a washing step in 70% ethanol, the reaction product was mixed with loading buffer. The samples were either heated at 85 • C for 5 min or kept on ice, and then analyzed by non-denaturing 5% polyacrylamide-8 M urea gel electrophoresis ran at 300 V for 6 h at 4 • C. The gel was then imaged using a phosphorimager (Amersham Typhoon scanner IP, Cytiva).

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.