eIF4A, a target of siRNA derived from rice stripe virus, negatively regulates antiviral autophagy by interacting with ATG5 in Nicotiana benthamiana

Autophagy is induced by viral infection and has antiviral functions in plants, but the underlying mechanism is poorly understood. We previously identified a viral small interfering RNA (vsiRNA) derived from rice stripe virus (RSV) RNA4 that contributes to the leaf-twisting and stunting symptoms caused by this virus by targeting the host eukaryotic translation initiation factor 4A (eIF4A) mRNA for silencing. In addition, autophagy plays antiviral roles by degrading RSV p3 protein, a suppressor of RNA silencing. Here, we demonstrate that eIF4A acts as a negative regulator of autophagy in Nicotiana benthamiana. Silencing of NbeIF4A activated autophagy and inhibited RSV infection by facilitating autophagic degradation of p3. Further analysis showed that NbeIF4A interacts with NbATG5 and interferes with its interaction with ATG12. Overexpression of NbeIF4A suppressed NbATG5-activated autophagy. Moreover, expression of vsiRNA-4A, which targets NbeIF4A mRNA for cleavage, induced autophagy by silencing NbeIF4A. Finally, we demonstrate that eIF4A from rice, the natural host of RSV, also interacts with OsATG5 and suppresses OsATG5-activated autophagy, pointing to the conserved function of eIF4A as a negative regulator of antiviral autophagy. Taken together, these results reveal that eIF4A negatively regulates antiviral autophagy by interacting with ATG5 and that its mRNA is recognized by a virus-derived siRNA, resulting in its silencing, which induces autophagy against viral infection.


Introduction
Eukaryotic initiation factor 4A (eIF4A), a member of the DEAD-box RNA helicase protein family, is thought to use the energy from ATP hydrolysis to unwind the mRNA structure and to prepare mRNA templates for ribosome recruitment during translation initiation, together with other components [1]. The Arabidopsis thaliana genome encodes two isoforms of eIF4A, one of which (eIF4A-1) is required for the coordination between cell cycle progression and cell size [2]. eif4a1 mutants display slow growth, reduced lateral root formation, delayed flowering, and abnormal ovule development [2]. A T-DNA mutation of eIF4A confers dwarfing in Brachypodium distachyon in a dose-dependent manner [3]. Therefore, eIF4A is essential for plant growth and development [2,3]. We previously determined that the mRNA of eIF4A from N. benthamiana (NbeIF4A) is targeted by a viral small interfering RNA (vsiRNA-4A) produced from the genomic RNA4 fragment of rice stripe virus (RSV) by mRNA cleavage [4]. Silencing of NbeIF4A caused dwarfing in plants, which is consistent with findings for B. distachyon [3,4]. However, the role of eIF4A in plant-virus interactions remains unclear.
Autophagy is an evolutionarily conserved mechanism that employs double-membrane vesicular autophagosomes to enclose and deliver cytoplasmic material for vacuolar or lysosomal degradation and recycling [5]. Autophagy-related genes (ATGs) encode key factors in this process. In plants, autophagy plays essential roles in development, reproduction, metabolism, senescence, and tolerance to abiotic and biotic stress [6,7]. Autophagy is induced in response to viral infection; increasing evidence suggests that autophagy participates in antiviral responses during plant-virus interactions [7][8][9][10]. For example, autophagy is induced during the resistance response to tobacco mosaic virus (TMV), and plants lacking ATG activity exhibit enhanced virus accumulation [11].
Several viral proteins were recently shown to be targeted by autophagy for degradation, thus limiting viral infection. For instance, the autophagy cargo receptor NEIGHBOR OF BREAST CANCER 1 (NBR1) targets the capsid protein (CP) of cauliflower mosaic virus (CaMV) and the helper component proteinase of turnip mosaic virus (TuMV) for autophagic degradation and suppresses viral accumulation [12,13]. The TuMV protein NIb is targeted by Beclin1 (also called ATG6), a core autophagy component that mediates NIb degradation [14]. The virulence factor βC1 of cotton leaf curl Multan virus (CLCuMuV) interacts with the key autophagy protein ATG8 and is degraded by the autophagic machinery [15]. We previously demonstrated that p3, a viral suppressor of RNA silencing (VSR) from rice stripe virus (RSV), interacts with NbP3IP, a cargo receptor from Nicotiana benthamiana, and is delivered to autophagic vesicles for degradation [16]. In addition, protein 2b from cucumber mosaic virus (CMV) is thought to be targeted for degradation by autophagy through the calmodulin-like protein rgs-CaM [17].
Although the antiviral function of autophagy has been well established in plants, how autophagy is induced by viral infection is not well understood [13][14][15]18]. Ismayil et al. recently reported that CLCuMuV βC1 protein interacts with the negative autophagic regulators glyceraldehyde-3-phosphate dehydrogenases (GAPCs) to induce autophagy in plants [19,20]. We previously identified RNA4, a small interfering RNA derived from RSV that targets and silences the host mRNA encoding eIF4A, thus contributing to the leaf-twisting and stunting symptoms observed during RSV infection [4]. Here we discovered that eIF4A acts as a negative regulator of antiviral autophagy in N. benthamiana, thus defining a mechanism whereby a negative regulator of antiviral autophagy recognizes a virus-derived siRNA and sacrifices itself to induce autophagy against viral infection.

Silencing of NbeIF4A inhibits RSV infection in Nicotiana benthamiana
We previously determined that the mRNA of eIF4A from N. benthamiana is targeted by a viral small interfering RNA (vsiRNA-4A) produced from the genomic RNA4 fragment of RSV by mRNA cleavage [4]. Here, we tested whether the downregulation of eIF4A would have any effects on RSV infection. We silenced N. benthamiana eIF4A (NbeIF4A) in plants with the Tobacco rattle virus (TRV)-based virus-induced gene silencing (VIGS) system by inserting a DNA fragment specific for NbeIF4A into RNA2 of the VIGS vector TRV to generate the TRV:4A construct. We monitored the progression of symptoms linked to RSV infection in TRV:4A-infected and control plants. At 15 days post infiltration (dpi), plants infected with TRV:4A alone showed leaf-twisting, whereas plants infected with the empty VIGS vector TRV:00 displayed no obvious morphological changes ( Fig 1A). RT-qPCR analysis revealed a reduction in NbeIF4A transcript levels of approximately 60% in TRV:4A-infected plants relative to the TRV:00 controls ( Fig 1B). These results are consistent with a previous report [4].
Next, we inoculated NbeIF4A-silenced (TRV:4A-infected) or control (TRV:00-infected) plants with RSV. At 20 dpi with RSV, typical RSV symptoms, such as stunting and leaf-twisting with yellow mosaicism, appeared in all TRV:00 control plants (Fig 1C and 1D). By contrast, only 20% of RSV-inoculated NbeIF4A-silenced plants showed yellowing and mosaicism in leaves, which was less pronounced than that of the control (Figs 1C, 1D and S1). RT-PCR con-

Silencing of NbeIF4A activates autophagy
In our recent report, autophagy is demonstrated to play a key role in plant defense against RSV infection [16]. To determine whether autophagy pathway helps prevent RSV infection in NbeIF4A-silenced plants, we investigated the expression of genes in autophagy upon NbeIF4A silencing. Results showed that genes functioning in autophagy, such as NbATG2, NbATG3, NbATG5, NbATG6, NbATG7, and PHOSPHOINOSITIDE 3-KINASE (NbP13K), were significantly upregulated in NbeIF4A-silenced plants (S3 Fig), suggesting that autophagy may be activated in these plants.
To directly test this hypothesis, we tagged N. benthamiana ATG8f with cyan fluorescent protein (CFP) at its N terminus (CFP-NbATG8f) to monitor autophagic activity in NbeIF4Asilenced cells, as previously described [20]. We observed a diffuse fluorescent signal in plants infected with the empty vector TRV:00, whereas silencing of NbeIF4A resulted in more fluorescent foci, indicative of activated autophagy in NbeIF4A-silenced cells (Fig 2A and 2B). We independently validated these results using the stain monodansylcadaverine (MDC), which accumulates in acidic autophagic vacuoles (Fig 2C and 2D). Finally, we observed multiple autophagic structures in NbeIF4A-silenced cells by transmission electron microscopy (TEM) and far fewer in TRV:00 control plants (Fig 2E and 2F). These results support our hypothesis that autophagy is activated in NbeIF4A-silenced cells.
To exclude the possibility that silencing of NbeIF4A affected general translation in cells, we examined whether protein translation was affected by silencing of NbeIF4A. For this analysis, we developed an RNAi construct expressing hairpin RNA of NbeIF4A (4A-hairpin) to silence NbeIF4A. As a negative control, we generated a construct expressing hairpin RNA of β-glucuronidase (GUS-hairpin). As a positive control, we used a construct expressing hairpin RNA of NbeIF6A (6A-hairpin), which plays key roles in protein translation [21]. These three constructs were individually infiltrated into a single leaf of a plant with construct of 35S-driven GFP via Agrobacterium-mediated infiltration. At 3 dpi, in zones expressing 6A-hairpin where the expression of NbeIF6A was silenced, GFP accumulated to a lower level than that in zones expressing GUS-hairpin or 4A-hairpin. By contrast, there was no significant difference in GFP accumulation between zones expressing GUS-hairpin and 4A-hairpin (S4 Fig). These results indicate that protein translation was not significantly affected by silencing of NbeIF4A, suggesting that the activated autophagy in NbeIF4A-silenced cells was not due to the suppression of protein translation. Meanwhile, results cannot rule out that NbeIF4A has any role in RSV translation.
To further characterize the activation of autophagy in NbeIF4A-silenced plants, we also silenced NbATG3, which encodes a key component of autophagy. Using CFP-NbATG8f as a reporter of autophagic activity, we established that plants co-silenced for both NbeIF4A and NbATG3 accumulated fewer fluorescent foci (representing CFP-NbATG8f-positive autophagosomes) compared to NbeIF4A-silenced plants, supporting our hypothesis that silencing of NbeIF4A activates autophagy (Figs 2A, 2B and S5).
NIb, a viral RNA-dependent RNA polymerase (RdRp) of TuMV, interacts with Beclin (ATG6) and is degraded by autophagy [14]. We reasoned that if silencing of We also assessed the extent of NIb degradation in NbeIF4A-silenced plants. We transiently expressed NIb tagged with GFP in NbeIF4A-silenced or control leaves by Agrobacterium-mediated infiltration. At 3 dpi, the accumulation of NIb-GFP was substantially reduced in NbeIF4A-silenced leaves relative to TRV:00 control plants (S6C Fig). These results further support the hypothesis that silencing of NbeIF4A activates autophagy.

Inhibition of RSV infection in NbeIF4A-silenced plants requires activated autophagy
We recently reported that autophagy counters RSV infection by degrading the RSV p3 protein.
Since silencing of NbeIF4A activated autophagy, we reasoned that the autophagic degradation PLOS PATHOGENS eIF4A negatively regulates antiviral autophagy by interacting with ATG5 of p3 should be enhanced and the activity of p3 as suppressor of RNA silencing should be hence weaken in NbeIF4A-silenced cells, which would explain the limited spreading of RSV in these plants. To test our hypothesis, we expressed Nb4A-hairpin and GUS-hairpin individually in a single leaf of N. benthamiana plants (16c) by Agrobacterium-mediated infiltration. One day later, p3 and GFP were then expressed in the same leaf. At 3 dpi with p3 and GFP expression, green fluorescence in zones expressing 4A-hairpin was weak compared to that in zones expressing GUS-hairpin control ( Fig 3A). We measured p3 and GFP accumulation by immunoblot analysis with anti-p3 and anti-GFP antibody, respectively. Both p3 and GFP accumulated to lower levels in zones expressing 4A-hairpin than in zones expressing GUS-hairpin ( Fig  3B). In the control experiment, when GFP only was expressed in zones expressing 4A-hairpin or GUS-hairpin, the accumulation of GFP was not affected by 4A-hairpin ( Fig 3C). These results demonstrate that the autophagic degradation of p3 is enhanced and the suppressor activity of RNA silencing is impaired in NbeIF4A-silenced plants. These findings support the notion that the inhibition of RSV infection in NbeIF4A-silenced plants is due to activated autophagy.
To further validate our hypothesis, we silenced the key autophagy gene NbATG3 or NbATG5, together with NbeIF4A, and monitored RSV infection in these plants. NbATG3 [16]. When NbeIF4A was co-silenced with either NbATG3 or NbATG5, typical symptoms of RSV appeared in all plants at 20 dpi, whereas only approximately 20% of NbeIF4A-silenced plants showed mild RSV symptoms (Figs 3D, 3E and S8). Based on RSV CP accumulation, we determined that RSV infection was more pronounced in co-silenced plants than in NbeIF4A-silenced plants, underscoring the compromised resistance to viral infection exhibited by the co-silenced plants (Figs 3F and S8). These results demonstrate that co-silencing of genes encoding key autophagy components compromises the inhibition of RSV infection in NbeIF4A-silenced plants, further confirming the notion that the inhibition of RSV infection in NbeIF4A-silenced plants depends on induced autophagic responses. However, we also noticed that silencing of NbATG3 was not sufficient to cancel the effect of NbeIF4A silencing on RSV accumulation, suggesting that NbeIF4A might have another function during RSV infection.

NbeIF4A interacts with NbATG5
Next, we explored the mechanism by which NbeIF4A intersects with other components of the autophagy pathway. To this end, we tested the potential interactions between NbeIF4A and ATG3, ATG5, ATG6, ATG7, and ATG8, which were previously shown to function in the interplay between viruses and autophagy by yeast-two hybrid (Y2H) and bimolecular fluorescence complementation (BiFC) assays. Only NbATG5 interacted with NbeIF4A (Figs 4A, 4B, S9A and S9B). We confirmed this interaction using coimmunoprecipitation (Co-IP) and firefly luciferase complementation imaging (LCI) assays (Fig 4C and 4D).
To map the domain(s) in NbeIF4A responsible for its interaction with NbATG5, we generated a series of truncated NbeIF4A mutants in accordance with the conserved domains in the protein: 4A (

Overexpression of NbeIF4A suppresses NbATG5-activated autophagy
Based on the finding that NbeIF4A silencing activated autophagy and that NbeIF4A interacted with NbATG5, we hypothesized that NbeIF4A likely suppresses the function of NbATG5, thereby inhibiting autophagy. We therefore examined the effects of NbeIF4A overexpression on ATG5 function during autophagy. We tagged NbATG5 with red fluorescent protein (RFP) and co-expressed NbATG5-RFP with CFP-NbATG8f in N. benthamiana leaves by Agrobacterium-mediated infiltration. At 2.5 dpi, the number of autophagic bodies increased significantly in cells co-expressing both proteins compared to cells only expressing the control protein RFP, which is consistent with the activation of autophagy (Figs 5A, 5B and S10).
Next, we individually co-expressed myc-tagged NbeIF4A; its mutant 4A(71-241), which retained the ability to interact with NbATG5; or its mutant 4A(252-413), which could no longer interact with NbATG5, with NbATG5-RFP and CFP-NbATG8f. Cells co-expressing myctagged NbeIF4A or 4A(71-241) and NbATG5-RFP produced fewer autophagic bodies than cells expressing NbATG5-RFP only. However, the number of autophagic bodies in cells coexpressing myc-tagged 4A(252-413) and NbATG5-RFP was similar to that in cells expressing NbATG5-RFP alone (Figs 5A, 5B and S10). MDC staining produced similar results (Fig 5C  and 5D). These results demonstrate that the accumulation of NbeIF4A suppresses the ability of NbATG5 to activate autophagy. Moreover, such suppression is dependent on the interaction of NbeIF4A with NbATG5.
We also examined the effects of NbeIF4A on autophagy induced by methyl viologen (MV), which results in severe oxidative stress and induces autophagy [23,24]. Cells treated with MV showed a sharp increase in the number of autophagic bodies relative to untreated cells (S11 During autophagy, ATG5 interacts with ATG12 to form ATG12-ATG5 conjugates. We wondered whether NbeIF4A would interfere with such an interaction and hence suppress autophagy. To test this possibility, we examined the effects of expressing NbeIF4A on the interaction between ATG5 and ATG12 via a BiFC assay. NbATG5 interacted with NbATG12 in the BiFC assay (S12A Fig). Furthermore, when NbeIF4A-myc was coexpressed with these proteins in a BiFC assay, fluorescent signals significantly decreased, pointing to the impaired interaction between NbATG5 and NbATG12 (S12A Fig). Consistently, in a Co-IP assay, when NbeIF4A was expressed, NbATG5 was immunoprecipitated, but it was difficult to immunoprecipitate NbATG12 in this assay (S12B Fig), indicating that expressing NbeIF4A interferes with the interaction between NbATG5 and NbATG12.
Taken together, these results demonstrate that overexpressing NbATG5 activates autophagy and that expressing NbeIF4A prevents this activation by competing with NbATG12 for interaction with NbATG5.

PLOS PATHOGENS
eIF4A negatively regulates antiviral autophagy by interacting with ATG5 To test this hypothesis, we expressed vsiRNA-4A via the same method used to express artificial miRNAs in N. benthamiana leaves [4]. At 2.5 dpi, cells co-expressing vsiRNA-4A and CFP-NbATG8f showed a 40% decrease in NbeIF4A transcript levels relative to control cells coexpressing a control small RNA and CFP-NbATG8f, demonstrating the silencing of NbeIF4A by vsiRNA-4A (S13 Fig). Using confocal microscopy, we observed autophagosomes and autophagic bodies in cells co-expressing vsiRNA-4A and CFP-NbATG8f, but fewer of these structures in control cells (Fig 6A and 6B). We independently validated these results by MDC staining (Fig 6C and 6D). These results demonstrate that expressing vsiRNA-4A induces autophagy.
To further support our hypothesis that the induction of autophagy depends on the downregulation of NbeIF4A, we co-expressed vsiRNA-4A and CFP-NbATG8f in the leaves of transgenic N. benthamiana overexpressing NbeIF4A generated in our previous study [4]. The expression of vsiRNA-4A failed to reduce NbeIF4A transcript levels in transgenic plants to lower levels than observed in the wild type and did not induce autophagy (Fig 6E and 6F). We obtained similar results using MDC staining (Fig 6G and 6H). These results demonstrate that the expression of vsiRNA-4A induces autophagy via the downregulation of NbeIF4A.

Rice eIF4A interacts with OsATG5 and suppresses OsATG5-activated autophagy
Finally, we looked for the rice (Oryza sativa) ortholog of NbeIF4A. OseIF4A (Os06g0701100) shows 81% nucleotide sequence identity with NbeIF4A, and the encoded protein shares 74.6% amino acid sequence identity with NbeIF4A (S14A and S14B Fig). In addition, OseIF4A mRNA is also targeted by vsiRNA-4A for cleavage (S15 Fig). We therefore tested whether OseIF4A performs similar functions to NbeIF4A. Indeed, OseIF4A interacted with rice ATG5 (OsATG5) (Fig 7A and 7B). OsATG5 shares higher sequence identity with NbATG5 in the Nterminus of the protein (S16 Fig). Transient expression of OsATG5 in N. benthamiana activated autophagy (Fig 7C-7E). Moreover, co-expression of OseIF4A effectively suppressed autophagy activated by OsATG5 or NbATG5 (Figs 7C-7E and S17). These results suggest that OseIF4A plays a similar role to NbeIF4A, i.e., its transcript is targeted by vsiRNA-4A and suppresses autophagy by interacting with OsATG5. These findings highlight the functional conservation of eIF4A in the regulation of autophagy in N. benthamiana and rice.

Discussion
We previously reported that the downregulation of NbeIF4A transcript by vsiRNA-4A contributes to viral symptoms in RSV-infected N. benthamiana. Here, we uncovered the biological function of downregulated NbeIF4A: it regulates the autophagic responses that normally inhibit RSV infection, pointing to the complex roles of NbeIF4A during RSV infection. eIF2α, a member of the eukaryotic translation initiation factor (eIF) family, is associated with endoplasmic reticulum (ER) stress-induced autophagy in mammalian cells [25,26]. eIF2α can be phosphorylated by the PKR-like ER kinase PERK [25]. PERK/eIF2α phosphorylation is critical for the conversion of microtubule-associated protein 1 (MAP1) light chain 3 (LC3) from LC3-I to LC3-II, which plays a key role in autophagy [26]. eIF5A was recently identified as a key factor required for the lipidation of members of the ATG8 family of proteins as well as autophagosome formation via translation of the E2 ubiquitin ligase-like ATG3 protein in mammalian cells [27,28]. Yet, little is known about the roles of eIFs is plant autophagy. Our results provide the first evidence that eIF4A functions in regulating antiviral autophagy in plants.
eIF4A is thought to use the energy from ATP hydrolysis to unwind mRNA structures and, together with other components, prepare mRNA templates for ribosome recruitment during translation initiation [1]. The current results suggest that eIF4A functions as a negative regulator of autophagy that plays a role in protein metabolism by degrading proteins. These findings point to the intricate roles of eIF4A in protein production: not only does it function in translation initiation, but also in inhibiting autophagy, a mechanism for protein degradation. We also noticed that although silencing of NbeIF4A caused a dwarf phenotype in N. benthamiana, which is consistent with findings for B. distachyon [3], it did not significantly affect protein Several studies have shown that autophagy is regulated during viral infection. Xu et al. demonstrated that plant Bax inhibitor-1 (BI-1) interacts with ATG6 to regulate autophagy [29]. Silencing of BI-1 reduced the autophagic activity induced by N gene-mediated resistance to TMV, and overexpressing plant BI-1 increased autophagic activity [29]. The cytosolic glyceraldehyde-3-phosphate dehydrogenase (GAPC) negatively regulates autophagy by interacting with ATG3 [20]. Silencing of GAPCs significantly activated ATG3-dependent autophagy, while overexpressing GAPCs suppressed autophagy in N. benthamiana plants [20]. In the current study, we determined that NbeIF4A suppresses ATG5 activity by interfering with the ATG5-ATG12 interaction, thus demonstrating its role as a new negative regulator of autophagy in plants. In addition, the current and previous results suggest that autophagy factors may be targeted by other host factors to regulate autophagy [16,20,29]. It is expected that more host factors will be identified in the future, which will help further dissect the mechanisms regulating autophagy.

PLOS PATHOGENS
eIF4A negatively regulates antiviral autophagy by interacting with ATG5 degradation of ARGONAUTE 1 (AGO1), a key component of the antiviral RNA silencing pathway [31]. VPg of TuMV mediates the degradation of SUPPRESSOR OF GENE SILENC-ING 3 (SGS3) via ubiquitination and autophagy to promote virus infection [32]. βC1 of Tomato yellow leaf curl China virus (TYLCCNV) regulates CaM expression, thereby mediating the autophagic degradation of NbSGS3 in N. benthamiana [33]. Virus-induced vesicles enriched with ATG8f provide an alternative site for Bamboo mosaic virus (BaMV) RNA replication or a shelter from the host silencing mechanisms [34]. Together, these findings reveal that autophagy can be manipulated or even exploited by viruses for infection, highlighting the complex roles of autophagy in plant-virus interactions [9,10].
Our previous and current results demonstrate that autophagy also functions in defense against RSV infection, illustrating the complex role of autophagy during plant-RSV interactions [16]. Fu et al. reported that the RSV-encoded NSvc4 protein interferes with the S-acylation of remorin, which normally prevents viral movement, thus inducing its degradation through the autophagy pathway [35]. This observation also demonstrates that autophagy can function indirectly during plant-RSV interactions to benefit RSV infections. This might also represent a common way in which autophagy participates in plant-virus interactions. For example, AGO1 and SGS3 are degraded via autophagy during Polerovirus, TuMV, or TYLCCNV infection [31][32][33]. This observation could also help explain why RSV retained a sequence that can induce autophagy. Perhaps the induced autophagy benefits RSV by facilitating its movement.
Infection with CLCuMuB, TuMV, and RSV induces autophagy, although the underlying mechanism is not well understood [14,15]. Upregulated expression of ATGs it thought to be associated with the induction of autophagy in TuMV-and RSV-infected plants [12]. Ismayil et al. reported that CLCuMuB βC1 interacts with the negative autophagic regulator GAPC to induce autophagy in plants, suggesting a mechanism by which CLCuMuB induces autophagy [19,20]. Here we demonstrated that NbeIF4A negatively regulates antiviral-associated autophagy in N. benthamiana. Moreover, a vsiRNA that target sNbeIF4A mRNA for cleavage induces autophagy. These findings suggest a mechanism in which a negative regulator of antiviral autophagy (NbeIF4A) sacrifices itself to induce autophagy against viral infection by allowing its transcripts to be recognized and cleaved by vsiRNAs.
Mounting evidence indicates that vsiRNAs play roles in the interaction between viruses and their host plants by regulating host gene expression [36][37][38][39][40]. For example, vsiRNA-20 from Chinese wheat mosaic virus (CWMV) affects pyrophosphate hydrolysis and H(+) concentrations in CWMV-infected wheat (Triticum aestivum) cells by regulating the mRNA accumulation of vacuolar H + -PPases to create a more favorable cellular environment for CWMV replication [41]. The current results support the indirect participation of vsiRNAs in the regulation of autophagy.
Based on these observations, we propose a model for the possible role of autophagy in RSV infection. According to this model, eIF4A functions as a negative regulator to suppress autophagy in plants by competitively interacting with ATG5 to interfere with its interaction with ATG12. Upon RSV infection, eIF4A mRNA becomes targeted by vsiRNAs for cleavage. The resulting downregulation of eIF4A releases ATG5 to activate autophagy, which in turn prevents RSV infection by degrading the p3 protein, a suppressor of antiviral RNA silencing (Fig  8). Perhaps this mechanism could be exploited to design ways to improve plant virus resistance, for example, by editing eIF4A.
For RSV inoculation, we mechanically inoculated the middle of each N. benthamiana leaf with sap from RSV-infected rice. Symptoms were photographed with a Canon 550D digital camera at the indicated times. For TuMV-GFP inoculation, we infiltrated an Agrobacterium cell suspension carrying the TMV-GFP plasmid into N. benthamiana leaves as previously described [25,26]. GFP was imaged under UV light.
To construct a TRV-based recombinant VIGS vector containing NbeIF4A, NbATG3, or NbATG5, we PCR-amplified a partial fragment for each gene using the appropriate primer pair and cloned the PCR product into the pTRV2-lic vector. Two fusion DNA fragments for NbeIF4A+NbATG3 and NbeIF4A+NbATG5 were obtained by overlapping PCR and cloned into the pTRV2-lic vector.
The primers used for these vectors are listed in S1 Table.

Bi-molecular Fluorescence Complementation (BiFC) assays
For the BiFC experiments, we infiltrated fully expanded leaves of three-week-old N. benthamiana plants with Agrobacterium strain (GV3101) containing constructs at a cell density OD 600 = 0.5. At 36-72 h after Agrobacterium-mediated infiltration, we examined the epidermal cells of 1-cm infiltrated leaf explants by confocal microscopy (Leica TCS SP5, Mannheim, Germany).

Co-immunoprecipitation (Co-IP) and immunoblot analysis
We harvested infiltrated leaves at 48-72 h after transient Agrobacterium-mediated infiltration and extracted total proteins from frozen, ground tissue by mixing with 3 Laemmli buffer at a 1:1 ratio (w/v) for 5 min. Cell debris was collected by centrifugation at 12,000 g for 5 min at 4˚C, after which we determined the protein concentration in the supernatant by Bio-Rad Protein assay. Equal amounts of total protein were aliquoted, brought to 40 μL with protein extraction buffer, and mixed with 10 μL 5× SDS-PAGE loading buffer. After boiling the samples for 10 min at 100˚C, we quickly chilled the total protein lysate on ice for 5 min, centrifuged the sample at 12,000 g for 1 min at 4˚C, and separated the proteins on SDS-PAGE gels. We performed immunoblotting with anti-myc (Sigma, St. Louis, USA), anti-RFP (Sigma, St. Louis, USA), and anti-GFP (Sigma, St. Louis, USA) primary antibodies and anti-rabbit or anti-mouse (Sigma, St. Louis, USA) secondary antibody at 1:10,000 dilution. Antibodies of TuMV CP, RSV CP and P3 proteins were prepared in our laboratory. For IP, protein extracts were incubated with anti-GFP antibody for 4 h at 4˚C. The beads were washed six times with ice-cold IP buffer at 4˚C. The IP samples were analyzed by SDS-PAGE, immunoblotted using anti-myc antibody, and detected using Pierce ECL Western Blotting Substrate (Amersham Image 680).

Luciferase Complementation Imaging (LCI) assays
For the LCI assays, we introduced the plasmid combinations into N. benthamiana leaves by Agrobacterium-mediated infiltration. We kept infiltrated leaves in the dark for 24 hpi and then treated them with white light for 48 hpi. We detached the leaves at 72 hpi and sprayed PLOS PATHOGENS eIF4A negatively regulates antiviral autophagy by interacting with ATG5 them with 1 mM luciferin. After keeping the materials in the dark for 5 min to quench the chlorophyll auto-fluorescence, we collected the luciferase signal with a low-light cooled CCD imaging apparatus.

RNA extraction and RT-qPCR
We extracted total RNA from N. benthamiana leaf tissues using TRIzol reagent and treated the RNA with RNase-free DNase I (TaKaRa, China) to remove potential DNA contamination. We performed first-strand cDNA synthesis using 1 μg of total RNA with oligo(dT) 12-18 primers using a Titanium One-Step RT-PCR Kit (TaKaRa, Japan). Primers used for RT-qPCR are listed in S1 Table.

Confocal Microscopy and Transmission Electron Microscopy (TEM)
We performed confocal microscopy using a Leica TCS SP5 (Mannheim, Germany). We introduced all combinations tested into N. benthamiana leaves by Agrobacterium-mediated infiltration. For GFP, RFP, CFP, and YFP imaging, the fluorescent proteins were excited using the LD laser line at 488 nm (GFP), 514 nm (RFP), 405 nm (CFP), and 514 nm (YFP). Detection bands were optimized for each fluorophore group to avoid emission bleeding.
For monodansylcadaverine (MDC) staining, we infiltrated leaves with 100 mM E-64d (Sigma) and incubated the leaves in the dark for 8-12 h. We excised the infiltrated parts of leaves and immediately vacuum-infiltrated them with 50 mM MDC (Sigma) for 8-14 min, followed by two washes with phosphate buffered saline (PBS). MDC was excited at a wavelength of 405 nm and detected from 450 to 550 nm. Chlorophyll autofluorescence was excited at 543 nm and detected from 580 to 700 nm.
To observe autophagosomes using the autophagy marker CFP-ATG8f, we introduced CFP-ATG8f into leaves of 5-6 leaf stage N. benthamiana plants by Agrobacterium-mediated infiltration. After 48 h, we infiltrated the leaves with 100 mM E-64d (Sigma) and incubated them in the dark for 8-12h. We observed the leaves by confocal microscopy with an excitation of 405 nm and the emission captured at 454 to 581 nm (Leica TCS SP5, Mannheim, Germany).
For electron microscopy, we cut leaves into small fragments (1-2 mm 2 ) and infiltrated them with 0.1 M PBS buffer containing 2.5% glutaraldehyde for fixation. We post-fixed the samples in 2% OsO 4 , followed by dehydration in ethanol and acetone, before embedding in Spurr resin (SPI Supplies). We cut the sections with a diamond knife on an ultramicrotome (EM UC7; Leica, Germany) and collected them on copper grids. The sections were doublestained with uranyl acetate and lead citrate before examination.