Association between flower stalk elongation, an Arabidopsis developmental trait, and the subcellular location and movement dynamics of the nonstructural protein P3 of Turnip mosaic virus

Abstract Virus infections affect plant developmental traits but this aspect of the interaction has not been extensively studied so far. Two strains of Turnip mosaic virus differentially affect Arabidopsis development, especially flower stalk elongation, which allowed phenotypical, cellular, and molecular characterization of the viral determinant, the P3 protein. Transiently expressed wild‐type green fluorescent protein‐tagged P3 proteins of both strains and selected mutants of them revealed important differences in their behaviour as endoplasmic reticulum (ER)‐associated peripheral proteins flowing along the reticulum, forming punctate accumulations. Three‐dimensional (3D) model structures of all expressed P3 proteins were computationally constructed through I‐TASSER protein structure predictions, which were used to compute protein surfaces and map electrostatic potentials to characterize the effect of amino acid changes on features related to protein interactions and to phenotypical and subcellular results. The amino acid at position 279 was the main determinant affecting stalk development. It also determined the speed of ER‐flow of the expressed proteins and their final location. A marked change in the protein surface electrostatic potential correlated with changes in subcellular location. One single amino acid in the P3 viral protein determines all the analysed differential characteristics between strains differentially affecting flower stalk development. A model proposing a role of the protein in the intracellular movement of the viral replication complex, in association with the viral 6K2 protein, is proposed. The type of association between both viral proteins could differ between the strains.


| INTRODUC TI ON
Plant viruses can alter the development of infected plants. In addition to symptoms traditionally associated with viral infections such as leaf mosaics, yellows, or a general dwarfism, viruses often induce alterations in typical plant traits like androsterility, shape alteration in the temporal evolution of different organs, or the lack of development of reproductive organs. This development-related aspect has received less attention compared with the study of defence responses, for example. We have recently reviewed the intimate relationship between virus infections and alterations in developmental traits, and discussed the different viral and plant components identified so far in this complex relationship ). An appropriate model system for the study of developmental alterations is the pathosystem formed by the potyvirus Turnip mosaic virus (TuMV) and Arabidopsis thaliana. TuMV encapsidates a single genomic RNA encoding all viral proteins, whose expression is mostly mediated by proteolytic processing of a precursor polyprotein (Ivanov et al., 2014;Revers and García, 2015), plus two additional fusion proteins, P3N-PIPO (Chung et al., 2008) and P3N-ALT (Hagiwara-Komoda et al., 2016). Two isolates/strains of TuMV (UK 1 and JPN 1) alter development differentially (Sánchez et al., 2015), and TuMV infections in Arabidopsis are also affected by the developmental stage of the plant (Lunello et al., 2007). A particularly dramatic change affects flower stalk elongation. Plants grown under long day conditions, inoculated with virus at developmental stage 1.08 (Boyes et al., 2001), were differentially affected: UK 1-infected plants did not elongate a flower stalk, whereas JPN 1-infected did.
Other alterations in plant architecture were also found but this was the main difference, which has dramatic effects on the fertility and reproduction of the plants, depending on the infecting isolate.
The implication of virus-encoded suppressors of RNA silencing (VSRs) as the main viral factors involved in developmental alterations has been raised (Kasschau et al., 2003;Chapman et al., 2004;Chellappan et al., 2005), although this view has been also questioned (Mlotshwa et al., 2005). In the case of stalk elongation of TuMVinfected Arabidopsis plants, the major viral determinant is the protein P3, not a VSR (Sánchez et al., 2015). The determinant was mapped to the C-terminal coding region of the cistron, out of the genomic region involved in the formation of the fusion proteins P3N-PIPO and P3N-ALT.
An RNA element in the P3 cistron was found to affect viral replication and movement (Choi et al., 2005). In the case of TuMV P3, the protein has been identified as a determinant of the three responses mentioned (Jenner et al., 2002(Jenner et al., , 2003Suehiro et al., 2004;Tan et al., 2005;Kim et al., 2010;Sánchez et al., 2015;Cui et al., 2017), and the relevance of its C-terminal region has been highlighted (Suehiro et al., 2004;Tan et al., 2005;Cui et al., 2017). Potential interactions of P3 have been proposed for different potyviruses, and its subcellular location has also received attention. It has been proposed that P3 interacts (or has the capability to do so) with other viral proteins such as P1, HC-Pro, CI, NIa, NIb, andVPg (Rodríguez-Cerezo et al., 1993, 1997;Merits et al., 1999;Guo et al., 2001;Shen et al., 2010;Zilian and Maiss, 2011), or that no interactions take place with any of these proteins (Urcuqui-Inchima et al., 1999). It was also proposed to interact with itself (Choi et al., 2000) and with no other viral protein (Kang et al., 2004). Very recently, the interaction of P3 with the fusion protein P3N-PIPO has been shown for TuMV (Chai et al., 2020).
Because P3 is the determinant for stalk elongation in TuMVinfected Arabidopsis plants, we are interested in linking P3 subcellular behaviour with its impact on plant development through studies at the cellular level, including mutant analysis. The results obtained show differential subcellular behaviour of P3 between elongation-allowing and elongation-arresting TuMV strains. We also identify a mutant in a single amino acid position able to interconvert the infection phenotypes.
A time-course study was performed (16, 24, 36, and 44 hr postinoculation, hpi). At 16 hpi only a diffuse and faint fluorescence was seen in all cells, not labelling any specific structure Western blots at 44 hpi of total protein extracts from these leaves revealed that both P3-GFP fusion proteins had the correct size (approximately 68 kDa) and similar accumulation levels ( Figure S1).
To confirm that the fluorescent reticulate was ER, each fusion protein was co-expressed together with a red ER-luminal marker,  Figure S2e-h). Interestingly, cytoplasmic punctate accumulations clearly associated with the reticulum (see enlarged z-series stack images in Figure 1k,m, 360° 3D Videos S3 and S4 and enlarged scans of P3-GFP and ChFP-KDEL co-expression in Figure 1l,n).
Tobacco etch virus (TEV) P3 has been reported to form punctate inclusions in association with the Golgi apparatus (Cui et al., 2010). In our studies, P3 particles seemed to behave like they are associated with the Golgi apparatus because they were permanently associated with the ER, showing a dynamic movement (see next section).
However, in contrast to the uniform size showed by the dictyosomes, TEV P3 was also predicted to contain two transmembrane domains (Cui et al., 2010), and thus to be an integral membrane protein. The characteristic of integral or peripheral protein was addressed by selective extraction procedures. The S100 and P100 fractions (soluble and microsomal, respectively) of agroinfiltrated leaf tissue were obtained and subjected to different treatments to release the soluble luminal proteins (Peremyslov et al., 2004), to dislodge proteins weakly or peripherally associated to membranes (Schaad et al., 1997), or to release integral proteins with a nonionic detergent (Triton X-100). Western blots using an anti-GFP antibody showed that the first treatment (0.1 M Na 2 CO 3 ) did not affect P3 membrane association, indicating that P3 is not a luminal protein (Figure 3). In contrast, the Triton X-100 treatment totally dislodged P3 proteins from the membranes. The presence of the GFP-fused p7B protein of Melon necrotic spot virus (MNSV), a well-known transmembrane protein of the ER/Golgi apparatus (GFP-p7B) (Genovés et al., 2010;Serra-Soriano et al., 2014), in the S100 fraction of the Triton X-100 treatment confirmed that the extraction procedure separated soluble away from membrane proteins. However, urea treatments differentially affected both P3-GFPs with respect to the GFP-p7B. In 4 M urea, almost all GFP-p7B remained in the P100 fraction (95%) while P3-GFP was mostly recovered in the S100 fraction in both isolates (72% for P3[UK1]-GFP and 73% for P3[JPN1]-GFP), indicating that both P3-GFPs are peripheral membrane proteins, rather than integral ones.

| The ER-movement of P3 particles is actomyosin-dependent
Transiently expressed P3-GFPs forming ER-linked aggregates increasing in size suggested movement along the ER. To visualize the movement, videos were taken from agroinfiltrated tissues. At 1 dpi small and medium-sized P3[UK1]-GFP particles were highly ER-mobile (Video S5 and particle tracking in Figure S3a). Some particles oscillated around the same ER location, but others were displaced fast. At 2 dpi, most of the fluorescence was in large immobile aggregates (Video S6). Some small and medium-sized particles still remained, but the high mobility was lost. For P3[JPN1]-GFP, the mobility of the particles was still high at 2 dpi (Video S7 and particle tracking in Figure S3b), similarly to the UK 1 protein at 1 dpi.
In plants, organelle and particle ER-associated movement (ER streaming) is mostly recognized as actin-dependent through myosin motors (Nebenfuhr et al., 1999;Ueda et al., 2010). We therefore used latrunculin B (Lat B), a well-known actin microfilament disruptive agent (Holzinger and Blaas, 2016). LatB treatments require optimization. In our case, 1 µM Lat B treatments for 10 min did not affect ER structure, whereas 2 µM affected ER integrity, thus 1 µM was used in subsequent experiments. Treatment with 1 µM Lat B at 1 dpi (UK 1) or 2 dpi (JPN 1) stopped the advancement of ER-associated aggregates, which just kept oscillating (Videos S8 and S9, respectively). Moreover, co-expression of the P3-GFPs with an actin microfilament marker (dsRFP-Talin) showed their alignment along microfilaments ( Figure S4a-f). This alignment was not so obvious when both P3-GFPs were co-expressed with a microtubule marker (αTubulin-ChFP) ( Figure S4g-l). Based only on this lack of co-localization we cannot rule out the possible involvement of microtubules in P3-GFP movement. However, treatments using colchicine, a microtubule disrupting agent, did not affect P3-GFP movement (data not shown). Taken together, these results show that the ER-associated movement of P3-GFP proteins was dependent on the integrity of the actomyosin cytoskeleton.

| P3 3D structural models predict differences between both isolates in C-terminal regions
TEV P3 has been predicted to contain two transmembrane domains (TMDs), typical features of integral membrane proteins F I G U R E 2 Confocal laser microscopy z-series stack projection of epidermal cells co-expressing P3[UK1]-GFP (a) or P3[JPN1]-GFP (e), and the Golgi marker STtmd-ChFP (b) and (f). Overlay images of red and green channels (c) and (g). 2D histograms showing no correlation between pixel intensities over space of the two colour channels and Pearson correlation coefficient (PCC) values (d) and (h). Scale bars are 20 µm F I G U R E 3 Peripheral endoplasmic reticulum (ER)-membrane association of P3-GFP proteins determined by differential treatments of the microsomal fraction. Extraction procedures differentiate luminal (Na 2 CO 3 ), peripheral (urea), and integral (Triton X-100) membrane proteins. The figure shows a western blot with anti-GFP antibody. S100 and P100 refer to the 100,000 × g supernatant and pellet. Paired numbers correspond to signal intensity percentages showing the protein fraction distribution (Cui et al., 2010). TuMV P3 is a membrane peripheral protein, and is thus not very likely to contain TMDs. Nevertheless, exploratory searches with different computer programs to predict TMDs from sequence rendered TMDs for both TuMV P3 proteins ( Figure   S5). We therefore took a different approach to find out if there are TMDs in P3 by predicting full 3D structures. Conventional TMD prediction is based on sequence and its reliability depends on the similarity of the query sequence with sequences in the data sets used. In the case of TuMV P3 proteins, no experimental structures for homologous transmembrane proteins exist in the Protein Data Bank (PDB), thus the reliability of TMD prediction is limited and homology-based 3D modelling is not possible.
Consequently, we decided to apply I-TASSER (Roy et al., 2010), a method ranking top in critical assessment of protein structure prediction experiments (Moult et al., 2014). I-TASSER estimates the quality of predicted structural models through a confidence score (C-score) that typically ranges from −5 to +2, with higher values indicating structural models with a higher confidence (Roy et al., 2010;Yang et al., 2014).
C-scores for 3D models of TuMV P3 were −2.07 (UK 1) and −2.04 (JPN 1), a medium level reliability. They are shown in Figure 4 with a consensus representation of subdomains.
Secondary structure was identified with the Dictionary of Protein Secondary Structure (DSSP) (Kabsch and Sander, 1983;Touw et al., 2015), which is the PDB-recommended standard method.

| A conserved amino acid residue in the P3 C-terminal region is the main determinant for flower stalk elongation
To finely map the stalk elongation determinant within the P3 C-terminal region (subdomain in the model), we tested some additional sequenced TuMV isolates to classify them as stalkallowing or -arresting. Ten isolates were individually inoculated in Arabidopsis and stalk development was assessed. Five isolates allowed elongation and five did not. Isolate identification and alignments of the stalk-relevant region are shown in Figure S6.
Only three positions (268, 279, and 280; locations indicated in Figure 4) correlated perfectly between stalk development and a specific amino acid combination. All isolates arresting stalk development had the combination M 268 , K 279 , M 280 ; the nonarresting ones were I 268 , T 279 , I 280 . We generated single amino acid mutants in the infectious clones (Sánchez et al., 1998;López-González et al., 2017). All three single mutants were made for UK 1, and positions 279 and 280 in JPN 1. Double mutants were also made at positions 279-280. Following inoculation, stalk elongation was assessed at 15 dpi ( Figure 5), a time at which no significant differences in viral titres between the two strains were found previously (Manacorda et al., 2013). Of the single mutants, only those in position 279 (UK 1 K279T; JPN 1 T279K) were able to interchange the stalk growth pattern (Figure 5d However, a greater change is seen in the orientation of side chains, more protuberant and exposed to the solvent when residue 279 is lysine (Figure 7), something all the more reasonable considering that lysine is positively charged and has a longer chain than threonine.
Therefore, from a purely structural standpoint the effect on stalk elongation and movement/accumulation of P3 can be associated with a larger exposition of lysine to the aqueous environment.
However, in the study of protein-membrane interactions electrostatic potentials weigh more than local structural changes, or even protein global shapes (Davis and McCammon, 1990;Honig and Nicholls, 1995). So, we computed electrostatic potentials for both P3 proteins and their mutants by solving the nonlinear Poisson-Boltzmann (PB) equation with the program APBS (Davis and McCammon, 1990;Baker et al., 2001). The results are shown in  (Figures 4 and 7a,b).

| Transiently expressed fluorescent TuMV P3 co-expressed with fluorescent 6K2
P3 has been reported to co-localize with the 6K2 protein, the main one responsible for the extensive endomembrane rearrangement and chloroplast association in potyvirus infections, even when expressed alone (Wei and Wang, 2008;Cotton et al., 2009;Grangeon et al., 2012;Cui et al., 2017). To address the co-localization issue of P3 and 6K2 in UK 1 and JPN 1, N. benthamiana leaves were agroin- and S16). 6K2 forms chloroplast-interacting ER-derived vesicles (Wei et al., , 2013Grangeon et al., 2012;Cui et al., 2017), and UK 1 P3 is present in these membranous structures, as described for TEV and TuMV by other authors (Cui et al., 2010(Cui et al., , 2017Chai et al., 2020), but the behaviour of JPN 1 P3 in relation to its 6K2 partner has not been previously described.

| Transiently expressed fluorescent TuMV P3 in the context of TuMV infections
The differential behaviour of the P3 proteins when co-expressed

| D ISCUSS I ON
P3 was identified as the main mediator of flower stalk elongation or arrest in TuMV-infected Arabidopsis plants. Its C-terminal domain contains the determinant of this developmental trait (Sánchez et al., 2015). Recent work (Cui et al., 2017) has shown the implication of the protein in viral intercellular movement, virus replication, co-localization with the other viral membrane protein (6K2), and formation of perinuclear chloroplast-bound 6K2 vesicles. All these P3-mediated processes are influenced by determinants in the P3 C-terminal region. We also recently identified a molecular motif within this region as determinant of the apparent nonhost resistance of Ethiopian mustard to the JPN 1 isolate (Sardaru et al., 2018). All these findings provided an adequate context for P3 comparisons of When P3-GFP fusion proteins were transiently expressed in N. benthamiana, they associated with the ER, highlighting the reticulate but also moving along and forming cytoplasmic puncta of aggregated protein. Both fusion proteins associated peripherally with the ER. An association of the P3-GFP fusion proteins with the Golgi apparatus was not found. Both fusion proteins formed large fluorescent aggregates, but at different rates. UK 1 P3 already formed large aggregates 1 day after inoculation, whereas it took one more day for the JPN 1 isolate to do so. In both cases most aggregates formed close to the cell periphery. In our view, this behaviour of the P3 protein is related to its proposed role as a protein involved in the viral intercellular movement. That the protein alone is able to move by ER streaming towards the cell periphery and aggregate at certain locations close to the periphery speaks of a probable role in transporting viral components. In this regard, the recent proposal of a role for specific ER subdomains, such as cortical microtubule-associated ER sites (cMERs) or ER-plasma membrane contact sites, as hubs for the control of viral intercellular movement through plasmodesmata (Pitzalis and Heinlein, 2017), would fit nicely with the behaviour of TuMV P3, which would move towards these structures, aggregating in them. An important difference between the P3 proteins of both isolates would be the rate of doing so. We previously  (Manacorda et al., 2013), probably related to the intracellular movement rate of each virus. Both P3 proteins differ in the presence/absence of a lysine residue. Remarkably, a lysine residue has been shown as critical in the ER lateral translocation of a viral movement protein (Serra-Soriano et al., 2014).
Association of peripheral proteins with membranes is looser than integral proteins and may involve lipids and/or other membrane proteins (Monje-Galván and Klauda, 2016). The 3D structure of peripheral proteins must govern the type of association with their membrane counterparts, but the TuMV P3 3D structure has not been resolved. In the absence of this information, refined modelling should shed light on the differential ER-association of both P3 proteins. We  (Cui et al., 2018), an integral ER membrane protein.
A final relevant difference was found in the subcellular behaviour of the two P3 proteins when transiently expressed together with another potyviral membrane protein, 6K2. For UK 1 P3 we confirmed the behaviour found previously for TuMV P3, a co-localization of both membrane proteins in association with ER-derived 6K2induced vesicles and chloroplasts in the close periphery of the nucleus, in addition to its normal ER reticulate location. In this case, P3 would be "dragged" by 6K2 to co-localize, although interestingly no direct interaction has been so far found between the two proteins.
The presence of P3 in the 6K2-induced membranous structures has been related to a critical role of P3 in the formation and functioning of the VRC (Cui et al., 2017). In the case of JPN 1 P3, the protein was not found co-localizing with perinuclear 6K2 vesicles. Interestingly, the P3 domain mediating the interaction with the vesicles is the C-terminal domain, which also contains the determinant of stalk elongation and subcellular movement dynamics of the protein. This result opens up the possibility of alternative ways through which P3 would move 6K2-containing VRCs along the ER, but this aspect should be further investigated.
Interestingly, when the fluorescent P3 proteins were transiently expressed in cells infected with their corresponding TuMV strain, their localizations and dynamics were highly similar to those found when co-expressed with their corresponding 6K2 proteins, thus highlighting both the differential behaviours of the two P3 proteins within cells and the relevance of 6K2 in relation to P3 in the infection process.
Taking together the findings of this comparative study, a model based on them can be proposed for the role of P3 in TuMV infections, although still an incomplete one. This viral membrane protein peripherally associated with the ER membrane, possibly through a direct interaction with some ER integral protein, would undergo a process of cytoplasmic streaming. This would be an actomyosin-dependent process leading the protein towards specific locations close to the cell periphery. Most likely this process is a main mediator of the intracellular movement of VRCs on their way towards plasmodesmata. This model implies that the ER-mobile P3 must interact with the 6K2-containing VRCs, the viral complex spreading the infection to the neighbour cells (Movahed et al., 2017), which would be the cargo towards the cell periphery. The specifically dedicated intercellular movement viral proteins (CI and P3N-PIPO) would then play their role as mediators of the movement through plasmodesmata. P3 may or may not play a role in the formation of the VRC, but it would be its carrier, although at different rates depending on viral strains. Very recently a model for the role of TuMV P3 protein in the viral cell-to-cell movement has been proposed (Chai et al., 2020).
Although not specifically mentioned in the publication, the TuMV isolate on which this model is based is most probably of the UK 1 type. The main novelty of this model is that it shows a direct interaction of P3 and P3N-PIPO via the shared N-terminal domain of both proteins, an interaction that would allow the connection between CI inclusions and the 6K2-containing vesicles to be established. We discarded previously a possible role of P3N-PIPO as a determinant of flower stalk elongation (Sánchez et al., 2015), so we did not study this movement-dedicated fusion protein in our work. However, the proposed model is fully compatible with the one described in this paper. P3 would interact with the 6K2 vesicles forming a complex also containing P3N-PIPO, although the major localization of P3N-PIPO to the cell surface suggests that the main incorporation of P3N-PIPO into the vesicles would occur after the P3-mediated ER streaming of the vesicles. The presence of some P3N-PIPO in the ER-moving vesicles is also possible, but this would not affect our main proposal of a differential moving rate between both isolates.
Finally, the link of the subcellular behaviour of the protein and its 3D structure with the alterations in the developmental phenotype of the infected plant is still not fully resolved. The phenotypes found for each isolate (flower stalk arrest in UK 1 and creeping growing habit in JPN 1) point to different types of cell wall alterations depending on virus strain. Work in progress is currently devoted to the characterization of these alterations.

| Construction of binary vectors for fluorescenttagged P3 recombinant proteins
Gene sequences were amplified by PCR. The resulting fragments were purified and digested with NheI and NcoI. Digested fragments were cloned between the CaMV 35S promoter and the PoPit terminator into modified pBSIIKS(+), which allows cloning PCR products to directly generate GFP and ChFP fusion proteins. The resulting clones, described in detail in the Results section, were digested with HindIII and cloned into pMOG800 binary vector (Knoester et al., 1998). In the case of 6K2 fusion proteins, XhoI/EcoRI or HindIII enzymes were used to liberate the cassette from pBSIIKS (+). At the same time, pMOG800 vector was digested by the same enzymes.
The primer sequences used for cloning are listed in Table S1.

| Confocal laser scanning microscopy
Subcellular localization of fluorescent proteins was performed with an inverted Zeiss LSM 780 confocal microscope. eGFP and ChFP/ dsRFP fluorescence was visualized by 488 and 561 nm laser excitation, respectively. The emission detection windows were 492-532 and 590-630 nm, respectively. Chlorophyll excitation wavelength was 488 nm, fluorescence was detected above 700 nm.
Plant leaves transiently expressing GFP-p7B were used as a source of transmembrane protein of the ER and Golgi apparatus (Genovés et al., 2010). Large cellular debris was removed by centrifugation at 8,000 × g for 10 min, and the supernatant was Miracloth-filtered. The total protein fraction was aliquoted in five aliquots, which were ultracentrifuged at 100,000 × g for 30 min to generate the soluble (S100) and microsomal (P100) fraction. The S100s were stored and the P100s resuspended in 2 ml of lysis buffer supplemented with 0.1 M Na 2 CO 3 , pH 11.5, 4 M urea, 8 M urea, or 0.1% Triton X-100. Treatments were performed on ice for 30 min. After incubation, aliquots were ultracentrifuged at 100,000 × g for 30 min and pellets resuspended in 400 µl of lysis buffer. Samples from each fraction were analysed by 12% SDS-polyacrylamide electrophoresis (PAGE), and subsequently transferred to polyvinylidene difluoride membranes for immunoblotting with a polyclonal antibody to Nt-GFP.

| Point mutations in TuMV infectious clones
Mutations were introduced in P3 by overlap extension PCR cloning. To introduce each mutation, three PCRs were performed. The first two PCRs were done with an external primer (E1/E2) and internal primer (I1/I2) carrying the mutation(s) to be introduced. The products of these PCRs were mixed and used as templates for the third PCR, with the E1 and E2 primers (Table S1). All PCRs were performed using Pfu DNA polymerase. PCR fragments were cloned into Zero Blunt Topo and sequenced in an external sequencing service (Secugen). Vectors carrying the mutations were digested with SnaBI and BtgI and cloned into the infectious clone of UK 1 (previously digested with SnaBI and BtgI) in position 3305-4011 of the clone (Sánchez et al., 1998). The same procedure was followed for JPN 1, but in this case we introduced the mutated fragment in position 2781-4167 using the restriction enzyme SalI (López-González et al., 2017). Stability of the mutations after Arabidopsis infection was confirmed by immunocapture reverse transcription PCR (IC-RT-PCR) (Nolasco et al., 1993) and sequencing. The monoclonal antibody used for IC-RT-PCR was Anti-Poty and the primers are listed in Table S1.

| Plant growth and virus inoculation
Arabidopsis plants were grown in controlled chambers and inoculated with crude sap from cDNA-infected plants, as previously described (Sánchez et al., 1998;Lunello et al., 2007). Growth conditions were 21 °C (day) and 18 °C (night) in 16 hr/8 hr cycles. Inoculations were performed at stage 1.08 of Arabidopsis development (Boyes et al., 2001) and symptoms of infection started to become visible approximately 5 dpi. They were observed daily and recorded.

| Latrunculin B treatments
A stock solution of Latrunculin B was prepared in dimethyl sulphoxide (DMSO) and diluted in water. For P3 movement experiments, N. benthamiana leaf discs transiently expressing P3-GFP fusion proteins and dsRFP Talin (Genovés et al., 2010) were immersed into 1 μM LatB for 10 min. Controls were leaf discs from plants transiently expressing dsRFP-talin and calreticulin sequence encoding KDEL at the 3′ end of GFP (GFP-KDEL) (Genovés et al., 2010) immersed into DMSO. Tissue was visualized using an inverted Zeiss LSM 780 confocal microscope.

| Image and statistical analysis
Image analysis was performed using FIJI software (US National Institutes of Health). Quantification of intracellular fluorescence was performed by manually drawing regions of interest (ROI). Particle area measure was performed using the analyse particles option.
Projection of z-series stack was done using the standard deviation method and 3D projection was done using the interpolation tool to eliminate the gaps for the final 3D outcome. Western blot intensity signal was measured on files from a Fujifilm LAS-3000 Imager using Fuji Image Gauge v. 4.0 software. For the statistical analysis, two-way ANOVA and Tukey's post hoc test at the 95% confidence level (α = .05) for pairwise comparison was applied to the data using GraphPad Prism v. 6.01 software.

ACK N OWLED G M ENTS
This work was funded by several INIA grants. Silvia López-González was funded by a predoctoral FPI-INIA fellowship/contract. P.S. was the recipient of an EU fellowship from an EU-India bilateral agreement (BRAVE Program). We thank Professor John Walsh (Warwick University, UK) for his generous gift of virus isolates. The great technical assistance of Lucía Zurita is also acknowledged. We thank the Spanish Ministry of Science for the Severo Ochoa Excellence Accreditations to the CBGP (SEV-2016-0672).

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.