Dynamin-Like Proteins of Endocytosis in Plants Are Coopted by Potyviruses To Enhance Virus Infection

It is well known that animal viruses enter host cells via endocytosis, whereas plant viruses require physical assistance, such as human and insect activities, to penetrate the host cell to establish their infection. In this study, we report that the endocytosis pathway is also involved in virus infection in plants. We show that plant potyviruses recruit endocytosis dynamin-like proteins to support their infection. Depletion of them by knockout of the corresponding genes suppresses virus replication, whereas overexpression of them enhances virus replication and intercellular movement. We also demonstrate that the dynamin-like proteins interact with several viral proteins that are essential for virus replication and cell-to-cell movement. We further show that treatment of a dynamin-specific inhibitor disrupts endocytosis and inhibits virus replication and intercellular movement. Therefore, the dynamin-like proteins are novel host factors of potyviruses. The corresponding genes may be manipulated using advanced biotechnology to control potyviral diseases.

V iruses are obligate intracellular parasites that infect all living organisms and cause serious diseases worldwide. Different from animal viruses that utilize endocytosis to enter cells, plant viruses initiate their infection through mechanical inoculation or insect-mediated injection to pass through the cell wall barrier. The majority of known plant viruses are positive-sense, single-stranded RNA viruses which have a relatively small genome with limited coding capacity. These RNA viruses typically encode a single-digit number of viral proteins and, thus, must usurp cellular machineries and factors to survive and replicate in their host cells. In the past decade, a large number of host proteins have been identified for viruses to establish their infection in plants (1)(2)(3). Many of these host proteins are recruited to form the virus replication complex (VRC) that is associated with modified cellular membranes for the reproduction of their progeny.
Endocytosis involves the internalization or uptake of plasma membrane (PM) proteins or extracellular materials into the cell via a series of vesicle compartments (4). Endocytosis may be clathrin mediated (CME) or clathrin independent. The former is relatively better characterized, whereas the latter remains largely to be understood. As in animal cells, CME is a predominant internalization pathway in plant cells (5) and initiates from the formation of clathrin-coated pits (CCPs), followed by cargo selection. The recognition and recruitment of cargo proteins are mediated by adaptor protein complexes (6). Then, mature clathrin-coated vesicles (CCVs) detach from the PM by dynamin (7) and enter the cytosol, where they are targeted to and fused with the early endosome and recycling endosome trans-Golgi network (TGN/EE). Subsequently, these vesicles are recycled back to the PM/cell wall/cell plate via different pathways or delivered into multivesicular bodies, also named the late endosome (LE), which will sort PM proteins to vacuoles for degradation (8,9).
Dynamin and dynamin-related proteins (DRPs) are multidomain large GTPases that function in many cellular processes underlying membrane dynamics (such as endocytosis) and also act in post-Golgi network clathrin-mediated trafficking in animals (7,10). Based on the sequences, all dynamin members contain a conserved N-terminal GTPbinding domain, a middle domain that mediates dimerization during self-assembly, and a GTPase-effector domain that modulates GTPase activity. In addition, they have a pleckstrin homology (PH) domain and a C-terminal proline-rich domain (PRD) that interact with Src homology 3 domain-containing proteins, which is critical to the recruitment of dynamin to CCPs (11). Most land plants have six types of DRPs (DRP1 to -6) based on domain structure (12), and only two of them, DRP1 and DRP2, have been reported to participate in CCV formation during endocytosis (13) and post-Golgi network trafficking (14,15).
In Arabidopsis thaliana, the DRP1 subfamily consists of five members (DRP1A to DRP1E) with a high degree sequence identity (63% to 82%) at the amino acid level (16). The Arabidopsis DRP1 homologs lack a PH domain and a PRD and show highly variable and tissue-type expression profiles (17,18). The DRP2 subfamily in Arabidopsis comprises only two members, DRP2A and DRP2B (13). The amino acid sequences of DPR2A and DPR2B are highly identical (93%) and harbor all five domains (13,17). DPR2A and DPR2B are expressed throughout development in Arabidopsis (19). The drp2a drp2b double mutant shows gametophyte-lethal phenotypes; however, no loss-of-function phenotypes have been detected in either drp2a or drp2b mutants, suggesting that their two proteins have some functional redundancy in plant development (19).
Toward a better understanding of molecular potyvirus-host interactions and identification of novel host factors of potyviruses, we employed gel fractionation and liquid chromatography-tandem mass spectrometry (LC-MS/MS) methods to characterize host proteins associated with SMV virions. We identified an endocytosis protein dynamin, Glycine max (soybean) endocytosis dynamin-like protein 5A (GmSDL5A). We show that this protein interacts with potyviral proteins and is essential for potyvirus infection. endosome localization (24), our data indicated that the interacting complex of GmSDL5A with VPg or CI was present in the endosomes labeled by FM4-64 (Fig. 1C). No positive interactions were found in the leaf cells coinfiltrated with the empty BiFC vector and GmSDL5A (Fig. 1D). Transient expression of GmSDL5A or SMV CP and CI proteins (with an N-terminal yellow fluorescent protein [YFP]) alone in N. benthamiana leaf cells stained by FM4-64 showed that GmSDL5A, CP, or CI subcellularly localized to the endosomes labeled by FM4-64 ( Fig. 2A). They showed similar subcellular distribution patterns when YFP was tagged to their C termini (Fig. 2B). When expressed alone, VPg predominantly targeted the nucleus (Fig. 2B). Knockdown of GmSDL5A and GmSDL12A in soybean suppresses SMV infection. To investigate whether GmSDL5A is essential for SMV infection, we used an improved bean pod mottle virus (BPMV)-based virus-induced gene silencing (VIGS) vector (25,26) to knock down GmSDL5A expression in soybean. Since GmSDL5A and GmSDL12A are highly similar (Ͼ98% identity) at both the amino acid and nucleotide levels and, thus, are likely functionally redundant, we generated a vector named BPMVR2::GmSDL5A/12A to target the conserved nucleotide sequence region shared between them (Fig. 3A). This modified VIGS vector was expected to simultaneously knock down both GmSDL5A and GmSDL12A. Soybean plants (Glycine max cv. Williams 82) were bombarded with the empty BPMV vector or BPMVR2::GmSDL5A/12A constructs. At 14 days postbombardment (dpb), treated plants showed typical symptoms induced by BPMV, and the presence of BPMV was confirmed by reverse transcription-PCR (RT-PCR) and Western blotting. We evaluated gene silencing efficacy in these plants by quantitative RT-PCR (qRT-PCR) (Fig. 3B). We found that at 21 dpb, the expression levels of GmSDL5A and GmSDL12A were significantly reduced (ϳ70% for GmSDL5A and ϳ60% for GmSDL12A) in the soybean plants bombarded with BPMVR2::GmSDL5A/12A, in comparison with the control plants bombarded with the BPMV empty vector (Fig. 3B). Then, both GmSDL5A-and GmSDL12A-silenced and control plants were challenged with the SMV-L isolate (27). SMV infection was monitored at both the RNA and protein levels at 7, 14, and 21 days postinoculation (dpi). At all three time points, the SMV RNA level in the new upper leaves of GmSDL5A-and GmSDL12A-silenced soybean plants was significantly lower than that in the control plant treated with the BPMV empty vector (Fig. 3C). SMV CP protein was not detectable in any of the plants tested at 7 dpi. At 14 and 21 dpi, the SMV CP protein levels were substantially reduced in the GmSDL5A-and GmSDL12Asilenced plants compared to those in the control plants (Fig. 3D). The symptoms shown by the GmSDL5A-and GmSDL12A-silenced plants were also much milder than those in the control plants at 10, 17, and 22 dpi (Fig. 3E). In addition, we also investigated whether SMV infection affects the mRNA level of GmSDL5A and GmSDL12A at 7, 14, and 21 dpi in soybean plants. No significant difference was found between SMV-infected and control plants at 7 dpi (Fig. 3F). However, the expression levels of GmSDL5A were reduced in SMV-infected soybean plants at 14 and 21 dpi, and GmSDL12A was also downregulated at 14 dpi (Fig. 3F). These data suggest that GmSDL5A/12A is required for SMV infection and that downregulation of GmSDL5A and GmSDL12A might be part of induced resistance in response to SMV infection in soybean.
GmSDL5A/12A homologs in Arabidopsis are essential for TuMV infection. Since soybean is recalcitrant to genetic transformation and the soybean genome is a partially diploidized tetraploid, we switched to the Arabidopsis-TuMV pathosystem to further understand the role of dynamin in potyviral infection. BLAST searches against the NCBI database identified AtDRP1A (AT5G42080), which shares the highest amino acid similarity (86%) with GmSDL5A/12A in the Arabidopsis genome. As indicated in the introduction, AtDRP1A belongs to the AtDRP1 subfamily, which consists of five homologs (AtDRP1A to -E). Since AtDRP1 and AtDRP2 have almost the same function and share three conserved domains (15), the only two members (AtDRP2A and AtDRP2B) of the AtDRP2 subfamily were investigated in this study. To examine whether TuMV infection requires AtDRP1 and AtDRP2, we obtained AtDRP1 and AtDRP2 transfer DNA (T-DNA) lines from the Arabidopsis Biological Resource Centre (ABRC). These lines included SALK_061139, SALK_069077, and CS16362 for atdrp1a (28), CS401637 for atdrp1b, SALK_085100 for atdrp1c, SALK_043327 for atdrp1d, SALK_060080 for atdrp1e (29), SALK_071036 for atdrp2a (30,31), and SALK_134887C (30,31) and SALK_150606 for atdrp2b (31). We further performed PCR-based genotyping (32,33) and obtained homozygous knockout mutants except for the AtDRP1C and AtDRP1D lines, in which no T-DNA insertion was detected (Fig. 4A). All these atdrp mutants showed growth and developmental phenotypes very similar to those of wild-type Arabidopsis plants. The Arabidopsis mutants and wild-type plants were mechanically inoculated with TuMV inoculum sampled from N. benthamiana leaves infected with an infectious pCambiaTuMV::GFP clone (32). As expected, the wild-type plants showed typical symptoms caused by TuMV infection (32,33). Single atdrp1 mutants (atdrp1a, atdrp1b, and atdrp1e mutants) displayed similar symptoms (Fig. 4B) and accumulated similar amounts of viral RNA (data not shown). However, homozygous atdrp2 mutants (the SALK_071036 line for atdrp2a and the SALK_134887C line for atdrp2b) showed much milder symptoms than wild-type plants (Fig. 4C). The viral RNA level in both atdrp2 mutants was reduced by more than 80% at 16 dpi (Fig. 4D). These results suggest that knockout of either AtDRP2A or AtDRP2B inhibits TuMV infection.
To further investigate whether AtDRP1 or AtDRP2 is associated with the TuMV virions, we generated transgenic Arabidopsis plants (in the ecotype Col-0 background) expressing AtDRP1A or AtDRP2B fused with a flag-4ϫ c-myc tag and then inoculated these transgenic plants with TuMV-green fluorescent protein (GFP). The protein fusion AtDRP1A-flag-4ϫ c-myc or AtDRP2B-flag-4ϫ c-myc (with their respective protein complexes) in TuMV-infected tissues was subjected to coimmunoprecipitation (co-IP) with anti-flag M2 gel and detected by immunoblotting. We found that TuMV CP was coimmunoprecipitated with AtDRP2B-flag-4ϫ c-myc but not with AtDRP1A-flag-4ϫ c-myc (Fig. 4E), suggesting that the TuMV CP is associated with AtDRP2B. To confirm this result, we further performed an immunocapture (IC)-RT-PCR assay (34). Transgenic AtDRP2B-flag-4ϫ c-myc or AtDRP1A-flag-4ϫ c-myc plants infected by TuMV were sampled and captured by anti-myc polyclonal antibody (PAb) or anti-flag monoclonal antibody (MAb), and the purified complexes were analyzed by using RT-PCR to detect the viral RNA. Consistently, the viral genomic RNA was detected in the sample immunocaptured from the AtDRP2B-flag-4ϫ c-myc transgenic plants infected by TuMV (Fig.  4F). A trace amount of TuMV viral RNA was also detected in the sample immunocaptured from the transgenic AtDRP1A-flag-4ϫ c-myc Arabidopsis plants infected by TuMV. As AtDRP2 showed much greater affinity to TuMV virions than AtDRP1 and is essential for TuMV infection, we thus focused on AtDRP2.
Overexpression of AtDRP2 promotes TuMV infection. We investigated the effect of transient overexpression of AtDRP2 on TuMV accumulation. N. benthamiana leaves were coinfiltrated with a TuMV full-length infectious clone, pCambiaTuMV::6K2-GFP, together with an AtDRP2B expression construct or an empty vector as a control. We found that coexpression of AtDRP2B remarkably enhanced the fluorescent intensities of GFP directly resulting from TuMV infection (Fig. 5A). Virus accumulation was assessed in infiltrated leaves at 2 days postagroinfiltration (dpa) by qRT-PCR. Consistently, we found a significant increase (Ͼ20-fold) of TuMV genomic RNA when plants were coinfiltrated with AtDRP2B compared to that of control plants (Fig. 5B). To further analyze the effect of AtDRP2 on TuMV infection, transgenic Arabidopsis lines overexpressing AtDRP2B were generated. A total of 15 independent T0-positive lines were obtained and the overexpression of AtDRP2B in these lines was identified by immunoblotting (Fig. 5C). Two independent T2 lines (35S:AtDRP2Boe#3 and 35S: AtDRP2Boe#10) (Fig. 5D) were challenged with TuMV-GFP. In comparison with the wild-type plants, the transgenic plants overexpressing AtDRP2B were more susceptible to TuMV infection and developed more-severe symptoms (yellowish and dwarf phenotypes) ( Fig. 5E and F). Consistently, higher levels of the viral genomic RNA were also detected in the AtDRP2B overexpression plants (Fig. 5G). These data support the possibility that AtDRP2 positively regulates TuMV infection.
AtDRP2 interacts with TuMV 6K2, VPg, CP, and CI. We then investigated whether AtDRP2 interacts with TuMV proteins by conducting an mYTH assay. Among the 11 viral proteins, 6K2, VPg, CP, CI, and P1 showed a positive interaction with AtDRP2A (Fig. 6A). BiFC assays were further carried out to confirm these interactions in N. benthamiana. Confocal microscopy detected positive signals in tissues coexpressing AtDRP2A and the four viral proteins, including 6K2, VPg, CP, and CI (Fig. 6B). No positive interaction signals were detectable between AtDRP2A and seven other viral proteins (data not shown). Similar results were also found for AtDRP2B (Fig. 7A). Interestingly, TuMV 6K2 interacted with either AtDRP2A or AtDRP2B in the cytoplasm and formed ring-like structures ( Fig. 6B and 7A). The interaction of AtDRP2A or AtDRP2B with the three remaining viral proteins apparently led to the formation of numerous punctate structures in proximity to the PM ( Fig. 6B and 7A).
To further confirm the protein-protein interactions, we conducted a co-IP assay. TuMV viral proteins were fused with a 3ϫ flag tag, and AtDRP2A was tagged by an N-terminal YFP. These fusions were transiently coexpressed in N. benthamiana leaves, followed by co-IP. As shown in Fig. 6C, the four TuMV proteins 6K2, VPg, CP, and CI could be immunoprecipitated with the antibodies against YFP-AtDRP2A. Taken together, these data suggest that AtDRP2A or AtDRP2B interacts with TuMV 6K2, VPg, CP, and CI in vivo. TuMV VPg and CI are associated with endosomes. As the punctate structures resulting from interactions of AtDRP2A or AtDRP2B with the viral proteins VPg, CI, and CP near the PM may be involved in endocytosis, we examined this possibility using a styryl dye, FM4-64, which is an established marker for PM internalization to localize endocytosis and endosomes (24). N. benthamiana leaves coexpressing AtDRP2B-YN and each of the four viral proteins fused with YC, e.g., 6K2-YC, VPg-YC, CP-YC, and CI-YC, were stained with FM4-64. As shown in Fig. 7B, the punctate structures labeled by the interaction complex of AtDRP2B and VPg or CI were stained by FM4-64, and many of the AtDRP2B/VPg colocalization signals were on or near the PM. At higher magnification, some of the relatively larger VPg-containing punctate structures were distant from the PM (Fig. 7B, right panel of the second row). Almost all the AtDRP2B/CI-containing punctate structures were on the PM. In the case of the AtDRP2B and 6K2 complex or AtDRP2B and CP complex, the BiFC signals did not colocalize with FM4-64 staining (Fig.  7B). These data indicate that TuMV VPg and CI, but not 6K2 and CP, are associated with the endosomes.
AtDRP2 is recruited to the 6K2-containing VRC in TuMV-infected cells. Since AtDRP2 behaves like a host factor of TuMV, we checked the subcellular localization of AtDRP2A and AtDRP2B. Regardless of the fluorescent protein fused to their N or C terminus, AtDRP2A and AtDRP2B were apparently distributed in the cytoplasm along with the PM (Fig. 8A). Further costaining with FM4-64 revealed that AtDRP2 did not exactly overlap but was in proximity to the FM4-64 staining (Fig. 8B). To investigate whether AtDRP2 is recruited by the VRC for TuMV infection, we coexpressed AtDRP2A or AtDRP2B in N. benthamiana leaf cells infected by a recombinant TuMV, pCambiaTuMV::6K2-mCherry. The modified virus contains an extra copy of 6K2 tagged by mCherry, so 6K2-containing VRCs may be visualized by confocal microscopy (35). We found that AtDRP2 was associated or colocalized with 6K2containing VRCs during TuMV infection (Fig. 8C). In addition, when a double-stranded RNA (dsRNA) sensor (36) was coexpressed with AtDRP2A or AtDRP2B tagged by CFP in the leaf cells infected by pCambiaTuMV::6K2-mCherry, dsRNA signals were colocalized with the VRCs containing mCherry-tagged 6K2 and CFP-tagged AtDRP2 (Fig. 8D). These results suggest that AtDRP2 is recruited by TuMV VRCs.
Treatment of a chemical inhibitor of endocytosis suppresses TuMV replication and intercellular movement. To confirm the possible role of AtDRP2 in viral replication, we conducted protoplast transfection assays. Protoplasts were isolated from wild-type Arabidopsis, atdrp2a and atdrp2b mutants, and AtRDP2 overexpression lines and then transfected with pCambiaTuMV::GFP (32), followed by qRT-PCR to monitor

Endocytosis Is Involved in Potyvirus Infection
Journal of Virology viral RNA accumulation at 16, 24, and 36 h posttransfection (hpt). TuMV viral RNA levels significantly decreased in both the atdrp2a and atrdp2b mutants, compared with those in the wild-type control, at all the three time points examined (Fig. 9A). In contrast, significantly higher levels of viral RNA were detected in the protoplasts from AtDRP2B overexpression transgenic lines (Fig. 9B).
To address the role of endocytosis in TuMV infection, we used two chemical inhibitors of endocytosis, dynasore and Tyrphostin A23 (Tyr A23) (37)(38)(39). Dynasore is known as a noncompetitive and reversible inhibitor which specifically targets dynamin to disrupt its GTPase activity (40), and Tyr A23 inhibits internalization of the transferrin receptor by perturbing the interaction between tyrosine motifs and the medium- chain subunit of the AP2 (41). N. benthamiana leaves were infiltrated with either Tyr A23, dynasore, or dimethyl sulfoxide (DMSO) and agrobacterial cultures containing PCB301TuMV-GFP//mCherry-HDEL. At 58 h postagroinfiltration (hpa), inoculated leaves were harvested to quantify the viral genomic RNA level by qRT-PCR. We found that treatment of dynasore significantly inhibited TuMV RNA accumulation (Fig. 9C). A similar result was also obtained from the TuMV RNA replication assay using Arabidopsis protoplasts (Fig. 9D). These data support the finding that endocytosis is involved in TuMV infection.
To further examine whether AtDRP2 affects viral intercellular movement, we introduced the vector PCB301TuMV-GFP//mCherry-HDEL (42) (using a low optical density [OD] of agrobacterial culture) into N. benthamiana leaves coexpressing AtDRP2A or AtDRP2B. This vector contains two expression cassettes, including one for the generation of a recombinant TuMV tagged by GFP and one for the expression of mCherry-HDEL. The primary-infection sites emit both red and green fluorescences, and the secondary-infection cells emit only green fluorescence. Therefore, this vector can be used to distinguish primary and secondary infections. In the control leaf cells infiltrated with PCB301TuMV-GFP//mCherry-HDEL and buffer (mock infection), the inoculated cells emitted GFP and RFP fluorescence at 3 dpa (Fig. 9E), indicating that viral intercellular movement did not occur at this time point. However, at the same time point, coexpression of either AtDRP2A or AtDRP2B facilitated the cell-to-cell movement of the recombinant virus into secondarily infected cells emitting green fluorescence only (Fig.  9E). We also conducted an inhibitor assay using the chemical effectors dynasore and Tyr A23 in N. benthamiana. At 92 hpa, in the DMSO-or Tyr A23-treated leaf cells, viral intercellular movement occurred, while in the dynasore-treated cells, no viral cell-to-cell movement occurred (Fig. 9F). Taken together, these results suggest that overexpression of AtDRP2 promotes both virus replication and intercellular movement, which can be suppressed by the endocytosis inhibitor dynasore.
TuMV proteins VPg and CI are delivered into both the EE and the LE, which can be inhibited by the dynamin inhibitor dynasore. To further characterize the endosomal vesicles labeled by the interaction complexes of AtDRP2 and viral proteins, we used fluorescently tagged AtSYP61 (mCherry-AtSYP61) (43) and AtARA6 (mCherry-AtARA6) (41) as the EE and LE markers. In Arabidopsis protoplasts, we observed that a fraction of the EE and LE were labeled by TuMV VPg and CI ( Fig. 10A and B). Interestingly, CP was evident in the LE but not in the EE. Since the TuMV CP-and AtDPR2-interacting complex did not colocalize with the endosomal structures labeled by FM4-64 (Fig. 7B), TuMV CP might hijack a different pathway to reach the LE. TuMV 6K2-GFP did not colocalize with either the EE or the LE marker.
To further confirm that TuMV proteins VPg and CI traffic in the endocytic pathway, we conducted a pharmacological interference assay with dynasore. In transfected protoplasts, treatment with dynasore changed the distribution pattern of YFP-VPg and YFP-CI ( Fig. 10C and D). Fewer FM4-64-labeled endosomes were visible upon dynasore treatment. YFP-VPg and YFP-CI were concentrated on the PM and rarely labeled with FM4-64, compared with what occurred with DMSO treatment. These results support the finding that TuMV proteins VPg and CI are delivered to the EE and LE via the endocytic pathway.

DISCUSSION
Proteomic techniques are widely used to identify the protein components of macromolecular complexes that participate in different biological processes, and proteome-based studies on pathogenesis of several plant viruses have made important contributions to the elucidation of molecular plant-virus interactions (44). The use of nano-LC for the high-resolution separation of digested peptides in samples has greatly increased detection capacity, making it easy to analyze low-copy-number proteins. In this study, we used an LC-MS/MS method to identify the proteins associated with SMV virions purified in virus-infected soybean leaves. We detected two dominant proteins from purified SMV virions (Fig. 1). These two proteins showed a slight size difference but are the expected size of CP. Our top-down proteomics analysis confirmed that both bands correspond to the CP of SMV ( Fig. 1; Table 1). The different migration rates of SMV CP might be due to nondetected phosphorylations, proteolysis, or partial degradation. This observation is consistent with those of a previous report on another potyvirus, PPV (45). In addition, we also detected SMV CI. Previous works suggest that, in addition to CP being a virion component, VPg, CI, and likely HC-Pro are physically associated with potyvirus virions (45)(46)(47)(48). In this study, HC-Pro was not detected, possibly because only a small amount of HC-Pro was associated with SMV virions. We did not analyze the proteins with a molecular mass lower than that of the CP, which may explain why VPg was not detected in our purified virion samples.
We identified five novel host proteins associated with the virions (Table 1; see also Table S1 in the supplemental material). In addition to these proteins, other host proteins were present in small amounts in our purified virion preparations and are not listed here due to the conservative criteria employed in our study. The presence of these five host proteins could be attributed to copurification of cellular components during virion preparations or to purified virions sticking to cellular proteins. However, it is equally possible that some of these cellular proteins, like ribulose bisphosphate carboxylase, are present in virions merely because of their high abundance in the cytosol. Among the five identified soybean proteins, the dynamin protein GmSDL5A, which is involved in endocytosis and vesicle trafficking, was chosen for further study.
Dynamin participates in endocytosis in both plants and animals (13,49). Since endocytosis is considered a fundamental process in biology, it is conceivable that endocytosis is involved in virus infection. In a previous study, clathrin, which is required in endocytic vesicle formation, was found to be associated with the human cytomegalovirus particles (50). More recently, components of endocytic and post-Golgi pathways have also been implicated in infection by several plant viruses (51)(52)(53)(54)(55)(56). To explore whether GmSDL5A plays a role in SMV infection, we detected and confirmed that GmSDL5A interacted with several SMV proteins in yeast and in planta (Fig. 1). Interestingly, unlike TuMV P1, SMV P1 did not interact with GmSDL5A in yeast cells (Fig. 1, Fig.  6). It is well known that of the 11 known potyviral proteins, P1 is a hypervariable protein that is least conserved among potyviruses (17,57). This may account for the different abilities of SMV P1 to bind to GmSDL5A and TuMV P1 to AtDRP2. In addition to GmSDL5A interacting with SMV VPg or CI, GmSDL5A alone localized to the endosome structures labeled by FM4-64, suggesting a role for GmSDL5A in the endocytosis of SMV proteins (Fig. 2). SMV CP or CI, but not SMV VPg, could enter into endosomal compartments in N. benthamiana cells (Fig. 2). To address the biological importance of GmSDL5A in SMV infection, we used a BPMV vector to knock down GmSDL5A and its highly conserved homologous gene GmSDL12A in soybean and found that knockdown of them significantly inhibits SMV infection (Fig. 3). Due to limitations of the soybean/ SMV pathosystem, we used Arabidopsis/TuMV as a model to further investigate the role of dynamin proteins in potyvirus infection.
Computer-assisted analyses identified AtDRP1 and AtDRP2 in Arabidopsis that share high sequence identity with GmSDL5A. Logically, we checked whether knockout of AtDRP1A or AtDRP2 affects TuMV infection. There are five isoforms of AtDRP1 and two isoforms of AtDRP2. None of the knockout mutants corresponding to these five AtDRP1 genes displayed obviously different symptoms and accumulated significantly different levels of viral RNA after TuMV infection. However, knockout of either AtDRP2A or AtDRP2B significantly impaired TuMV infection (Fig. 4). Clearly, AtDRP2 plays important roles in TuMV infection. Since the double atdrp2 mutant is gametophyte lethal (31), we cannot further evaluate the effect of a knockout of both AtDRP2A and AtDRP2B on TuMV infection. We thus, investigated the effect of overexpression of AtDRP2 on TuMV infection. We found that that either transient or transgenic overexpression of AtDRP2 promoted viral accumulation (Fig. 5). These data led us to suggest that AtDRP2 positively regulates TuMV infection.
To understand whether AtDRP2 affects virus infection through virus replication, intercellular movement, or both, we conducted protoplast transfection assays and confocal microscopy. We found that knockout of AtDRP2A or AtDRP2B in Arabidopsis inhibited viral RNA accumulation in protoplasts but that its overexpression enhanced viral RNA accumulation (Fig. 9A and B). Coexpression of either AtDRP2A or AtDRP2B also facilitated TuMV intercellular movement in N. benthamiana (Fig. 9E). Since more-robust viral replication is expected to accelerate viral intercellular movement, it is possible that the expedited spread of TuMV by the overexpression of AtDRP2 actually resulted from enhanced viral replication. Treatment with dynasore, a dynamin-specific chemical inhibitor, suppressed viral replication in Arabidopsis (Fig. 9D) and N. benthamiana (Fig. 9C) and intercellular movement in N. benthamiana (Fig. 9F). These data further support the assumption that AtDRP2 is a host factor that positively regulates virus replication (and possibly cell-to-cell movement) likely via endocytosis pathways.
To verify this assumption, we examined the effect of two chemical inhibitors of endocytosis, dynasore and Tyr A23, on TuMV infection. Treatment of dynasore indeed inhibited TuMV replication and intercellular movement significantly (Fig. 9D and F). As mentioned earlier, dynasore specifically targets dynamin to disrupt its GTPase activity and inhibits endocytosis (40). The antiviral role of dynasore corroborates the requirement of endocytosis and the dynamin-like proteins GmSDL5A and AtDRP2 in potyvirus infection. However, treatment of Tyr A23 did not affect TuMV replication and cell-to-cell movement (Fig. 9C, D and F). This is consistent with a previous report that Tyr A23 does not affect TuMV intercellular movement (58). As an endocytosis inhibitor, Tyr A23 inhibits the recruitment of endocytic cargo into clathrin-coated vesicles formed at the PM (41). These data suggest that dynamin may regulate viral infection independently of the clathrin-mediated endocytic pathway. Consistently, we also found that TuMV VPg and CI could enter into the EE and LE and that dynasore treatment inhibited their trafficking into endosomes, which further confirm the role of endocytosis of TuMV VPg and CI by AtDRP2 in TuMV infection.
To explore the molecular mechanism underlying the functional role of AtDRP2 in virus infection, we conducted several types of protein-protein interaction assays (mYTHS, BiFC, and co-IP) to screen for and confirm TuMV proteins that interact with AtDRP2. We found that AtDRP2 interacted with TuMV proteins 6K2, VPg, CP, and CI (Fig.  6). Since VPg, CP, and CI are all involved in virus movement, it is possible that AtDRP2 promotes virus cell-to-cell movement through its interaction with these viral proteins, particularly CI. It is well established that CI forms conical structures at the PD and is associated with viral particles (59). The facts that AtDRP2 is associated with virions ( Fig.  4E and F) and the AtDRP2/CI interaction takes place along the PM (Fig. 7B) support this assumption. VPg and CI are multifunctional proteins, and both of them are the critical components of VRCs (17). Our confocal microscopy revealed that AtDRP2 was recruited to the VRC in TuMV-infected leaf cells (Fig. 8). Since virus replication requires remodelling of cellular members for the formation of membrane-bound VRCs and AtDRP2 plays an essential role in membrane remodelling and fusion, it is possible that the TuMV VRCs coopt AtDRP2 for VRC assembly. The exact mechanical role of AtDRP2 in the VRC remains to be further elucidated.

MATERIALS AND METHODS
Plant materials and growth conditions. Soybean Williams 82 plants were used and grown in an insect-free growth chamber under a cycle of 16 h of light at 22°C and 8 h of darkness at 18°C. Plants were inoculated by biolistic bombardment with SMV infectious clone SMV-L (G2 strain) (60). Virus detection was carried out by enzyme-linked immunosorbent assay (ELISA) and RT-PCR as previously described (27).
All the Arabidopsis mutants used in this study were obtained from the Arabidopsis Biological Resource Centre. Homozygous plants of each line were screened as described previously (33). Transgenic Arabidopsis (ecotype Col-0) plants overexpressing AtDRP1A or AtDRP2B were obtained by the floral-dip method (61). Transformants were screened by direct spraying of solutions containing 20 mg/liter glufosinate-ammonium. Unless stated otherwise, all Arabidopsis plants were grown in a growth chamber with a 14 h photoperiod and a relative humidity of 75% at 23°C/21°C (light/dark). The N. benthamiana plants were grown in a growth room with a 16-h photoperiod and a relative humidity of 75% at 22°C.
RNA extraction, synthesis of the first-strand cDNA, PCR, and qPCR. Total RNA extraction from G. max cultivar Williams 82, Arabidopsis, and N. benthamiana tissues, synthesis of the first-strand cDNA, and PCR and qPCR were performed essentially as described previously (32,62). Information on the primers used in this study is available upon request.
Gene cloning and plasmid construction. In this study, the Phusion high-fidelity DNA polymerase (New England Biolabs, USA) was used to amplify all DNA sequences, and Gateway technology (Thermo-Fisher Scientific, USA) was employed for plasmid construction. Coding sequences of GmSDL5A from soybean and AtDRP1A, AtDRP2A, and AtDRP2B from A. thaliana Columbia-0 were amplified from soybean or Arabidopsis cDNA as described above. Coding sequences of SMV and TuMV genes were amplified from the infectious clones SMV-L and pCambiaTuMV::GFP (60, 63), respectively. These cloned genes were recombined into pDONR221 prior to final recombination to plant expression vectors by LR reactions. AtDRP2A and AtDRP2B were recombined into pEarleyGate102, pEarleyGate103, and pEarl-eyGate104, giving C-terminal CFP, C-terminal GFP, and N-terminal YFP, respectively. pEarleyGate vectors are described in reference 64. For the BiFC assay, the SMV or TuMV genes and host genes were transferred into the modified pEarleyGate201-nYFP or pEarleyGate202-cYFP vector (65). The pBA-FLAGϪ4ϫ Myc-DC vector (66) was used to yield a FLAGϪ4ϫ Myc-tagged construct. All constructs were verified by sequencing to ensure that no errors were introduced by PCR amplification. All plant expression cassettes were electroporated into Agrobacterium tumefaciens strain GV3101. The GV3101 cells harboring the expression constructs were resuspended with the infiltration buffer and infiltrated into N. benthamiana leaves as previously described (67,68).
Virus-induced gene silencing assay in soybean. To knock down the expression of GmSDL5A and GmSDL12A in soybean, a DNA-based BPMV-modified vector was used as previously reported (25,26). Transcript levels of both GmSDL5A and GmSDL12A were evaluated by RT-qPCR. The soybean ATPbinding cassette transporter (ABCT) gene was used as an internal control (69).
SMV virion purification. For LC-MS/MS analysis, the SMV virions were prepared as previously reported, with some modifications (70). Briefly, 100 g systemically infected leaves of Glycine max were harvested 3 weeks postinoculation and homogenized in a Waring blender with chilled 0.5 M phosphate buffer, pH 7.0, containing 0.1% 2-mercaptoethanol and 0.01 M EDTA (4 ml of buffer for 1 g of leaf tissue). Homogenized tissue was filtered through two layers of cheesecloth, followed by low-speed centrifugation in a Beckman L8-80M ultracentrifuge (6,000 ϫ g for 20 min) to remove plant debris. Clarified extracts were brought to 0.2 M NaCl containing Triton X-100 (0.1%, vol/vol), and then 6 g of polyethylene glycol with a molecular weight of 6,000 (PEG 6000) was added per 100 ml of extract. After the mixture was stirred for 0.5 h and incubated for 6 h, the precipitated virus was collected by low-speed centrifugation at 8,000 ϫ g for 10 min. The precipitated virus was resuspended in 20 ml of 0.5 M phosphate buffer, pH 7.0, containing 0.1% 2-mercaptoethanol, 0.01 M EDTA, and 0.5 M urea, followed by low-speed centrifugation (8,000 ϫ g, 10 min), and this step was repeated twice more. The virus was further concentrated by high-speed centrifugation for 2 h at 90,000 ϫ g through a 10-ml cushion of 20% sucrose in an 80Ti rotor. Birefringence was checked by swirling the resuspended virus solution between two crossed polarizing prisms and observing the resulting light patterns. The virus-containing band was removed from the tube carefully and resuspended in 2 ml 0.1 M phosphate buffer, pH 8.0. After low-speed centrifugation at 10,000 ϫ g for 5 min, the supernatant containing the purified virions was collected and stored long term at Ϫ80°C and short term at Ϫ20°C. Protein concentration was measured using NanoDrop 2000c spectrophotometers (Thermo Scientific).
Protein separation and mass spectrometry. Purified SMV virions were denatured by boiling the reaction mixture for 5 min and then loading it onto SDS-PAGE gels. The gels were then stained with a Coomassie brilliant blue solution (0.05% Coomassie brilliant blue G250, 50% methanol, 10% acetic acid) for 2 h and then destained (50% methanol, 10% acetic acid) for 5 h. The Coomassie blue-stained bands were excised from a gel rinsed 20 times in distilled water, destained, reduced, alkylated, and digested using the Thermo in-gel tryptic digestion kit (number 89871) according to the manufacturer's protocol. The peptide digests were analyzed using an Easy-nLC 1000 nano system with a 75-m by 15-cm Acclaim C 18 PepMap column (Thermo Scientific) coupled to a Q Exactive Orbitrap mass spectrometer (Thermo-Fisher Scientific) (71). The flow rate was 300 nl min Ϫ1 , and 10 l of the protein digest was injected. The C 18 column was equilibrated with 98% mobile phase A (water with 0.1% formic acid) and 2% mobile phase B (acetonitrile with 0.1% formic acid) and eluted with a linear gradient from 2% to 30% phase B over 18 min, followed by elution with a 30% to 98% gradient over 2 min, and then maintained for 10 min. The nanospray voltage was set at 2.1 kV, the capillary temperature was 275°C, and the S-lens RF level was 55. The Q-Exactive was operated in top 5 data-dependent acquisition mode with a full-scan mass range of 400 to 2,000 m/z at a resolution of 70,000, with an automatic gain control (AGC) of 1e 6 and a maximum injection time (IT) of 250 ms. The MS/MS scans were acquired at a 17,500 resolution, an AGC of 2e 5 , a maximum IT of 50 ms, an intensity threshold of 8e 4 , a normalized collision energy of 27, and an isolation window of 1.2 m/z. Unassigned peptides and peptides that were singly charged and charged Ͼ4 times were not selected for MS/MS, and a 20-s dynamic exclusion was used. The Thermo .raw files were converted to .mgf using ProteoWizard v2, and the MS/MS scans were searched against the target/reverse UniProt Glycine max database and 11 proteins encoded by SMV (GenBank accession no. AEP04405.1) by using the X! Tandem search algorithm operated from the SearchGUI v.2.35 interface and processed in PeptideShaker v1.3.6. A 3-ppm precursor ion mass error and a 0.02-Da product ion error were used, along with carbamidomethylation as a constant modification and oxidation of methionine as a variable. A 1% false-discovery rate (FDR) was used at the protein, peptide, and peptide spectrum match levels.
Confocal microscopy. Confocal microscopy was performed by using a Leica TCS SP2 inverted confocal microscope (Leica Microsystems, Wetzlar, Germany) equipped with a 63ϫ water-corrected objective in multitrack mode (72,73). Collected images were analyzed with the Leica Application Suite for Advanced Fluorescence (LAS AF; version 2.35) software (Leica Microsystem).
co-IP. Coimmunoprecipitation (co-IP) experiments were performed as previously described, with some modifications (74). The virus genes were overexpressed in N. benthamiana leaves by using the modified high-expression vector pSK104 with a C-terminal 3ϫ flag tag (75). The host genes were overexpressed in N. benthamiana leaves by using the pEarleyGate-104 vector with an N-terminal YFP tag. Immunoprecipitation was performed from N. benthamiana leaves at 2.5 days postinfiltration by using anti-Flag M2 affinity gel (Sigma, USA). Relevant antibodies were purchased from Sigma.
Protoplast isolation and TuMV replication assay. Mesophyll protoplasts were prepared from 4-week-old Arabidopsis leaves by the procedure described previously (76). About 1 ϫ 10 5 protoplasts were transfected with 30 g pCambiaTuMV::GFP in 40% PEG 4000 in 0.8 M mannitol and 1 M CaCl 2 at room temperature for 20 min. Transformed protoplasts were then washed and resuspended in W5 buffer (154 mM NaCl, 125 mM CaCl 2 , 5 mM KCl and 2 mM MES, pH 5.7) and incubated for the virus replication.
FM4-64 staining and treatment of endocytosis chemical inhibitors. N. benthamiana leaves were infiltrated with 40 M FM4-64 (Invitrogen) and then transferred to the microscope for imaging.
Co-IP-derived virion binding assay. A virion binding assay similar to that previously reported (77) was performed, with some modifications. Monoclonal antibody directed against the Flag tag was used as trapping antibodies to capture AtDR1A-4ϫ c-myc-flag or AtDR2B-4ϫ c-myc-flag and its binding complexes. TuMV-infected crude plant extract was added. Polyclonal rabbit antibodies against TuMV virions were used for virion detection.
ImageJ analysis of data. ImageJ was used to quantify average integrated density values of bands on immunoblots (78). The scanned images were saved in 16-bit .tiff format. The relative densities of the peaks with selected areas were calculated by the ImageJ software per the manual instructions.