Dual targeting of a virus movement protein to ER and plasma membrane subdomains is essential for plasmodesmata localization

Plant virus movement proteins (MPs) localize to plasmodesmata (PD) to facilitate virus cell-to-cell movement. Numerous studies have suggested that MPs use a pathway either through the ER or through the plasma membrane (PM). Furthermore, recent studies reported that ER-PM contact sites and PM microdomains, which are subdomains found in the ER and PM, are involved in virus cell-to-cell movement. However, functional relationship of these subdomains in MP traffic to PD has not been described previously. We demonstrate here the intracellular trafficking of fig mosaic virus MP (MPFMV) using live cell imaging, focusing on its ER-directing signal peptide (SPFMV). Transiently expressed MPFMV was distributed predominantly in PD and patchy microdomains of the PM. Investigation of ER translocation efficiency revealed that SPFMV has quite low efficiency compared with SPs of well-characterized plant proteins, calreticulin and CLAVATA3. An MPFMV mutant lacking SPFMV localized exclusively to the PM microdomains, whereas SP chimeras, in which the SP of MPFMV was replaced by an SP of calreticulin or CLAVATA3, localized exclusively to the nodes of the ER, which was labeled with Arabidopsis synaptotagmin 1, a major component of ER-PM contact sites. From these results, we speculated that the low translocation efficiency of SPFMV contributes to the generation of ER-translocated and the microdomain-localized populations, both of which are necessary for PD localization. Consistent with this hypothesis, SP-deficient MPFMV became localized to PD when co-expressed with an SP chimera. Here we propose a new model for the intracellular trafficking of a viral MP. A substantial portion of MPFMV that fails to be translocated is transferred to the microdomains, whereas the remainder of MPFMV that is successfully translocated into the ER subsequently localizes to ER-PM contact sites and plays an important role in the entry of the microdomain-localized MPFMV into PD.

Introduction but certain proteins show patchy distribution in the PM and are recognized as microdomainassociated proteins [19]. One of the microdomain-associated proteins, remorin (REM1.3), suppresses cell-to-cell movement of potato virus X (PVX) [20]. Furthermore, REM1.3 localizes also to PD and interacts with triple gene block protein 1 (TGBp1), an MP of PVX. Thus, it has been suggested that PM microdomains as well as ER-PM contact sites are important for virus cell-to-cell movement. Considering that PD localization of virus MPs is necessary for facilitating virus cell-to-cell movement, these two membrane subdomains, microdomains and ER-PM contact sites, are speculated to be involved in MP traffic to PD [16,17,20]. However, there is no direct evidence that virus MPs use these subdomains to reach PD, and a functional relationship of these subdomains in MP traffic to PD is unclear.
Signal peptides (SPs) are short sequences comprising approximately 7-30 aa that are frequently found in the N terminus of a diverse array of proteins. In general, proteins in eukaryotic cells with an SP are co-translationally recruited to the ER, and penetrate the ER membrane or are released into the ER lumen concomitant with the SP cleavage [21,22]. Some of these proteins play specific roles in the ER, whereas others are further transported to other organelles or secreted into the extracellular space. Thus, SPs are essential for proper localization and membrane targeting of proteins.

Fig mosaic virus (FMV) is a negative-strand RNA virus in the genus Emaravirus.
We showed previously that the MP of FMV (MP FMV ) localized to the PM in addition to PD, and remarkably, MP FMV was predicted to possess an N-terminal SP [23]. To the best of our knowledge, no viruses, other than the members of the genus Emaravirus, have an MP possessing an SP. In this study, we analyzed the intracellular trafficking of MP FMV focusing on the SP function in hopes of determining how the recruitment of a virus MP to the ER is involved in PD targeting. As a result, we found that the SP of MP FMV has extremely low ER translocation efficiency compared with conventional SPs of plant proteins, thereby causing abortive ER translocation of MP FMV at a high frequency. A fraction of MP FMV was translocated to the ER, whereas the remainder of MP FMV , which was not translocated to the ER, was transported to the patchy microdomains in the PM. Moreover, the ER-translocated MP FMV specifically localized to the ER-PM contact sites and played an essential role in the entry of microdomain-localized MP FMV into PD. Taken together, these findings suggest that dual targeting to two distinct subdomains in the ER and PM is essential for PD localization of MP FMV .

Results
MP FMV was distributed predominantly in PD and the patchy microdomains of the PM MP FMV was fused to YFP (MP FMV :YFP) to investigate its subcellular localization in Nicotiana benthamiana. Consistent with our previous results [23], transiently expressed MP FMV :YFP localized to the punctate structures along the PM in epidermal cells (Fig 1Ai). Treatment with aniline blue, which stains callose structures including PD, showed that the punctate structures of MP FMV :YFP colocalized with PD (Pearson correlation coefficient [PCC] = 0.53 ± 0.03). Measurement of the fluorescent intensity across plasmodesma shows that the fluorescence signal of MP FMV :YFP coincided with that of aniline blue (Fig 1Aii), showing that MP FMV :YFP localized to PD.
MP FMV :YFP appeared to accumulate also on the PM. PM and PD localization can be easily distinguished in plasmolyzed cells because PM proteins are associated with Hechtian strands, which are stretched PMs connecting the retracted PM and the cell wall, whereas PD proteins are retained in PD even during plasmolysis [24]. In plasmolyzed cells, MP FMV :YFP fluorescence was observed in Hechtian strands and PD (Fig 1Bi). On the other hand, fluorescence signal was observed only in PD, but not in Hechtian strands, when cells expressing tobacco mosaic virus MP as a GFP fusion (30K:GFP) were plasmolyzed (Fig 1Bii). These results indicate that MP FMV :YFP accumulated in the PM, in addition to PD. Observation of the cell surface revealed that MP FMV :YFP shows patchy distribution throughout the PM (Fig 1Ci). This distribution pattern was different from the fluorescence pattern of the PM stained with FM4-64, an amphiphilic styryl dye which is inserted into the outer layer of the PM (Fig 1Cii). These images were taken in the abaxial surface of the abaxial epidermal cells, and PD were not contained in these patches. Since such uneven distribution in the PM is similar to the localization of microdomain-associated proteins remorines [19,20], a YFP fusion with A. thaliana REM1.3 (YFP:REM1.3) was co-expressed with MP FMV :CFP; however, YFP:REM1.3 did not substantially co-localize with MP FMV :CFP (PCC = −0.06 ± 0.12; Fig 1Ciii). MP FMV probably localized to the PM subdomains distinct from those of REM1.3. To corroborate the subcellular distribution of MP FMV observed by microscopy, we performed chemical treatment using 1% TritonX-100. 1% TritonX-100 was expected to solubilize proteins associated with the cellular membranes excluding proteins localized to detergentinsoluble domains, including PM microdomains [25]. As expected, TritonX-100 treatment solubilized the majority of a PM marker H + ATPase and an ER marker BIP ( Fig 1D). Conversely, only a small proportion of transiently expressed MP FMV :FLAG was soluble in 1% TritonX-100. In FMV-infected cells, almost all MP FMV was detected from insoluble fraction. MP FMV became more insoluble in FMV-infected cells probably due to other virus factors. Furthermore, this result was in agreement with that of YFP:REM1.3, which localizes to the Tritoninsoluble microdomains [19,20]. Given that MP FMV unevenly distributed in the PM and that did not colocalize with REM1.3 (Fig 1Ciii), MP FMV may localize to the detergent-insoluble microdomains different from those of REM1.3.

MP FMV has an ER-directing signal peptide at the N-terminus
An ER-directed SP was predicted at the N-terminus of MP FMV by SignalP software in our previous work [23]. The cleavage site was between G19 and M20, and the length of the deduced SP was 19 aa (Fig 2A). To verify the cleavage at the predicted site, MP FMV :FLAG expressed transiently in N. benthamiana was purified by immunoprecipitation using anti-FLAG antibody. Immunoblot analysis using anti-FLAG antibody and Coomassie Brilliant Blue (CBB) staining of the immunoprecipitated samples showed a protein band of approximately 37 kD, indicating that precipitated MP FMV :FLAG has a single molecular weight (Fig 2B). Amino acid sequences beginning at M20, but not at the N-terminal methionine, were found in the sequences determined by Edman degradation of the purified MP FMV :FLAG, which means that the N-terminal 19 aa was cleaved off (S1 Fig). Combined with CBB staining and immunoblot analysis, the N-terminal 19 aa was suggested to be almost perfectly processed from MP FMV : FLAG. The N-terminal 19 aa had characteristics of SPs, a central hydrophobic region which forms an α-helix and a polar C-terminal region ( Fig 2C) [21,22]. These results suggest that MP FMV has an N-terminal SP (hereafter referred to as SP FMV ), which is expected to translocate the nascent protein into the ER.

SP FMV has extremely low translocation efficiency
To confirm whether or not SP FMV translocates a protein into the ER lumen as those of plant proteins, we constructed GFPs flanked with the N-terminal SPs derived from MP FMV or plant proteins (sporamin A, calreticulin or CLAVATA3) and a C-terminal ER retention signal ( Fig  3A; SP FMV GFP:HDEL, SP spo GFP:HDEL, SP cal GFP:HDEL and SP clv GFP:HDEL, respectively). We selected these three SPs because the SP activity was empirically assessed in previous studies [26][27][28]. These four SPs including SP FMV have no sequence homology (S2 Fig). As controls, we prepared free GFP and GFP:HDEL. Here, it is noted that GFP:HDEL does not have an Nterminal SP. Unexpectedly, transiently expressed SP FMV GFP:HDEL was distributed throughout the cytosol, but not to the ER (Fig 3B). The fluorescence pattern appeared to be the same as that of free GFP or GFP:HDEL. In contrast, SP spo GFP:HDEL, SP cal GFP:HDEL and SP clv GFP:HDEL predominantly localized in the ER, as expected. These results indicate the functional difference in ER translocation between SP FMV and the other SPs.
We further investigated the ER translocation ability of these SPs using N-glycosylation as an indicator. We constructed fusion proteins in which the 3× N-glycosylation sequon was Cterminally fused to GFP, SP FMV GFP, SP spo GFP, SP cal GFP and SP clv GFP (Fig 3C; GFPglc, SP FMV GFPglc, SP spo GFPglc, SP cal GFPglc and SP clv GFPglc, respectively) [29]. If an SP successfully translocates the GFP fusion into the ER lumen after SP cleavage by an ER membranebound SP peptidase on the lumenal side, glycosylation of asparagine residues in the sequon by an ER-resident enzyme, oligosaccharyl transferase, occurs and results in an increase in the molecular weight of the translocated GFPglc relative to that of the non-translocated GFPglc. Immunoblot analysis of total proteins extracted from leaves expressing GFPglc, SP FMV GFPglc, SP spo GFPglc, SP cal GFPglc or SP clv GFPglc using anti-GFP antibody showed that almost all of the SP cal GFPglc or SP clv GFPglc molecules were glycosylated (91.9% and 93.0%, respectively), and that more than half (70.9%) of the SP spo GFPglc molecules were glycosylated (Fig 3D and  3E). However, compared with these measurements, a dramatically lower proportion (4.3%) of the SP FMV GFPglc molecules were glycosylated. No glycosylation was detected in GFPglc, which lacks an SP. Thus, SP FMV had much lower translocation efficiency compared with conventional SPs, suggesting that only a small proportion of MP FMV molecules were translocated to the ER. In the case of SP FMV GFP:HDEL (Fig 3B), the fluorescence of ER-translocated GFP was thought to be masked by GFP fluorescence in the cytosol because SP FMV translocated only a small fraction of GFP molecules into the ER.

The subcellular localization pattern of MP FMV is altered depending on the SP
To gain insight into the role of ER translocation in the intracellular trafficking of MP FMV , we constructed an MP FMV mutant whose SP was not expected to be cleaved (ncMP). In this mutant, two substitutions, L7P and V11P, were introduced into the central α-helix region of SP FMV to break the helix [21]. We verified that an SP was no longer predicted in the ncMP sequence by SignalP (    Fig 4Aii). Also, Trun:YFP did not specifically localize to aniline blue-stained PD (PCC = 0.21 ± 0.05; Fig 4Aiii). The fluorescent signal of Trun:YFP was reduced in the region corresponding to the center of PD (Fig 4Aiv), unlike MP FMV :YFP (Fig 1Aii). Furthermore, we confirmed that Trun:YFP was not retained in PD in plasmolyzed cells (Fig 4Av right panel). These results indicate that the SP-deficient mutant did not localize to PD and that the SP FMV is essential for the PD localization of MP FMV , but dispensable for targeting the PM microdomains.
Next, SP FMV function was compared with SPs derived from plant proteins, sporamin A, calreticulin or CLAVATA3, by analyzing localization of SP chimeras in which the SPs of these proteins were fused to the N-terminus of Trun:YFP (SP spo Trun:YFP, SP cal Trun:YFP and SP clv Trun:YFP, respectively). SP spo Trun:YFP localized to the PM microdomains (Fig 4Bi, 4Bii and 4Bv) and PD (PCC = 0.47 ± 0.08; Fig 4Biii and 4Biv), similar to MP FMV :YFP. However, unlike MP FMV :YFP, cytoplasmic aggregations were observed in the SP spo Trun:YFP-expressing cells (Fig 4Bi). Co-expression with the ER marker ER-CFP showed that these aggregations were formed in the ER (Fig 4Bii). On the other hand, SP cal Trun:YFP and SP clv Trun:YFP were associated with nodes in the ER network (PCC: 0.55 ± 0.05 and 0.52 ± 0.06, respectively; Fig 4Ci, Cii, 4Di and 4Dii). Although some of these punctate spots were located in close proximity to PD, many of them did not co-localize with PD (PCC: 0.18 ± 0.03 and 0.27 ± 0.07, respectively; Fig  4Ciii and 4Diii). Fluorescence intensity measurements confirmed that fluorescence signals of SP cal Trun:YFP and SP clv Trun:YFP did not coincide with that of aniline blue (Fig 4Civ and  4Div). Hechtian strands were invisible when cells expressing SP cal Trun:YFP or SP clv Trun:YFP were plasmolyzed (Fig 4Cv and 4Dv), indicating that these chimeras were not distributed to the PM. These observations suggest that SP FMV plays an essential role in MP FMV localization.

PD localization is necessary for exerting MP FMV functions
Our previous study showed that the expression of MP FMV complements cell-to-cell movement of movement-deficient PVX mutant (PVXΔTGBp1-GFP) [23]. We investigated whether MP FMV mutants (Trun, SP spo Trun and SP cal Trun) facilitate virus cell-to-cell movement using this system. SP clv Trun was not used in this experiment because its transient expression for more than 3 days caused cell death. The fluorescence of PVXΔTGBp1-GFP spread to adjacent cells when MP FMV or SP spo Trun was co-expressed (Fig 5A), indicating that MP FMV and SP spo-Trun complemented cell-to-cell movement of PVXΔTGBp1-GFP. Quantitative analysis of fluorescence area suggested significant differences between MP FMV or SP spo Trun and β-glucuronidase (GUS) control (Fig 5B). Categorizing the fluorescence area by the cell number per fluorescent spot also showed significant difference between MP FMV or SP spo Trun and GUS control (p < 0.01 by Fisher's exact test). In contrast, the fluorescence of PVXΔTGBp1-GFP was almost confined to a single cell when co-expressed with Trun or SP cal Trun like GUS control ( Fig 5A). Comparable expression levels of MP FMV and its mutants were validated by Western blot analysis ( Fig 5C). Thus, SP spo Trun, an SP chimera which has the ability to reach PD, complements cell-to-cell movement of PVXΔTGBp1-GFP, indicating that PD localization of MP FMV is necessary to facilitate virus cell-to-cell movement. Most virus MPs are known to move to adjacent cells autonomously [30]. We assessed whether or not these MP FMV mutants were able to move to adjacent cells. MP FMV :YFP and SP spo Trun:YFP spread to adjacent cells when expressed in a single cell ( Fig 6A). Conversely,  Trun:YFP, SP cal Trun:YFP and SP clv Trun:YFP, which do not have the PD-targeting ability, did not move to adjacent cells. Quantitative analysis suggested a significant difference between these two groups in the ability to move to adjacent cells (Fig 6B). These results are in good accordance with virus cell-to-cell complementation assay (Fig 5).
SP-deficient MP FMV localizes to PD when co-expressed with MP FMV or SP chimeras The experiments described above suggested that the translocation efficiencies determine the subcellular distribution and function of MP FMV . SP cal Trun and SP clv Trun, whose SPs have high translocation efficiencies, localized to the ER, whereas Trun, which does not have an SP, localized to the PM (Table 1). SP spo Trun, which have moderate translocation efficiency, were able to localize to PD in addition to the ER and the PM. Given that SP FMV has low translocation efficiency, a small fraction of MP FMV was probably recruited to the ER. These results allowed us to speculate that co-existence of ER-translocated and non-translocated MP FMV is required for PD localization; in other words, ER-translocated MP FMV and microdomain-localized MP FMV act cooperatively to reach PD.
We tested this hypothesis by co-expressing SP-deficient MP FMV fused with CFP (Trun: CFP) and MP FMV :YFP, Trun:YFP, SP spo Trun:YFP, SP cal Trun:YFP or SP clv Trun:YFP. Localization pattern was not altered when Trun:YFP and Trun:CFP were co-expressed ( Fig 7B). By contrast, Trun:CFP was found to localize to punctate structures along the PM when coexpressed with MP FMV :YFP, SP spo Trun:YFP, SP cal Trun:YFP and SP clv Trun:YFP (Fig 7A and  7C-7E). Aniline blue staining confirmed that these punctate structures formed by Trun when co-expressed with MP FMV and SP chimeras coincided with PD (PCC: 0.50 ± 0.06, 0.60 ± 0.06, 0.61 ± 0.05 and 0.56 ± 0.15, respectively; S4 Fig). YFP-fused MP FMV and SP chimeras did not alter their localization by the expression of Trun:CFP. Detailed views showed that SP cal Trun: YFP and SP clv Trun:YFP localized in close proximity to, but did not substantially colocalize with Trun:CFP (Fig 7D and 7E; PCC: 0.31 ± 0.11 and 0.29 ± 0.03, respectively). This localization pattern was similar to those observed when SP cal Trun:YFP or SP clv Trun:YFP was expressed alone (Fig 4Ciii and 4Diii). These results suggest that Trun, which normally localized to the PM, was transported to PD by the function of MP FMV or SP chimeras, which were at least partially translocated to the ER.
The fact that MP FMV or SP chimeras changed Trun localization raised the possibility of the physical interaction between microdomain-localized MP FMV and ER-translocated MP FMV . We have now investigated the interaction between MP FMV and Trun or SP clv Trun using bimolecular fluorescence complementation (BiFC). We first co-expressed the basic leucine zipper transcription factor bZIP63 fused with N-terminal half of YFP (bZIP63:NYF) and with C-terminal half of YFP (bZIP63:CYF) as a control [31]. In this combination, strong fluorescence was observed in the nuclei (S5 Fig). Next, we tested the interaction between MP FMV :NYF and

ER-translocated MP FMV specifically localizes to ER-PM contact sites
Given that SP cal Trun:YFP and SP clv Trun:YFP localized in the close proximity to PD (Fig 4Ciii  and 4Diii and Fig 7D and 7E), translocated MP FMV appeared to localize in the specific region of the ER. We noticed that the distribution pattern of SP cal Trun:YFP and SP clv Trun:YFP was similar to that of the ER-PM contact site-associated protein, synaptotagmin1 (SYTA) [14,32], which is known to be involved in virus cell-to-cell movement [15][16][17]. Prior to the co-expression with MP FMV :YFP, we first analyzed localization of SYTA. Co-expression of SYTA:CFP and ER-YFP showed that SYTA:CFP was predominantly distributed to nodes of the ER (PCC = 0.61 ± 0.01; S6A Fig) consistent with the previous reports [14,15]. Aniline blue staining showed that SYTA: CFP was in close proximity to PD (PCC = 0.21 ± 0.01; S6B Fig), which was similar to those seen in SP cal Trun:YFP and SP clv Trun:YFP (Fig 7D and 7E). The close proximity of ER-PM contact sites and PD was reported also in a previous study [32]. To assess whether translocated MP FMV localize to ER-PM contact sites, SYTA:CFP was co-expressed with SP clv Trun:YFP or MP FMV : YFP. SP clv Trun:YFP co-localized almost perfectly with SYTA:CFP, probably in the ER nodes (PCC = 0.54 ± 0.08; Fig 8A). A large fraction of MP FMV :YFP was distributed to the PM and PD, but a small fraction of punctate structures colocalized with SYTA:CFP (Fig 8B). These results show that ER-translocated MP FMV specifically localized to ER-PM contact sites.
To obtain more information about the role of ER-PM contact sites in MP FMV trafficking, we constructed a c-myc tagged SYTA ΔC2B (SYTA ΔC2B :myc), which is a dominant-negative form lacking the C-terminal 177 aa of SYTA [16]. When MP FMV :YFP was expressed together with SYTA ΔC2B :myc, localization of MP FMV :YFP was apparently affected (Fig 8C and 8D). Although a small fraction of MP FMV :YFP was still retained in PD, a large proportion of MP FMV :YFP excessively accumulated next to PD (PCC = 0.20 ± 0.13; Fig 8C). MP FMV :YFP substantially colocalized with SYTA:CFP in the cortex; instead, fluorescence from MP FMV that localized in the PM microdomains became weaker (PCC = 0.72 ± 0.10; Fig 8D). This result suggests that MP FMV :YFP that normally localized to the PM microdomains aberrantly accumulated in ER-PM contact sites by the expression of SYTA ΔC2B :myc.
We also investigated whether SYTA ΔC2B affects cell-to-cell movement of MP FMV :YFP. Expression of SYTA ΔC2B :myc inhibited MP FMV :YFP movement to adjacent cells compared with when expressed with GUS (Fig 8Ei and 8Eii). Immunoblot analysis using anti-myc antibody confirmed the expression of SYTA ΔC2B :myc (Fig 8Eiii). Taken together, these data suggest that translocated MP FMV localized to ER-PM contact sites and played an essential role in cell-to-cell movement. To see whether MP FMV interacts with SYTA, BiFC was carried out. Fluorescent signal was not observed when MP FMV :NYF, Trun:NYF and SP clv Trun:NYF were co-expressed with SYTA:CYF (S7 Fig). Thus, the interaction between SYTA and MP FMV or MP FMV mutants was not suggested by BiFC.

MP FMV localization is not affected by inhibiting COPII transport
As the MPs of several viruses use the secretory pathway [9,33], involvement of COPII transport in MP FMV trafficking has been verified by BFA treatment or expression of a dominant-negative form of Sar1 [Sar1(H74L)] [34]. We first confirmed that BFA treatment and Sar1(H74L) expression caused retention of the Golgi marker ManI:CFP [35] in the ER, as expected (S8 Fig). PD localization in cells expressing MP FMV :YFP was not affected by BFA treatment or Sar1(H74L) expression. Similarly, inhibiting COPII transport did not affect localization of Trun:YFP in the PM. These results suggest that COPII transport is not involved in the subcellular localization of MP FMV and that MP FMV uses pathways different from BFA-sensitive MPs [9,33].

Discussion
In this study, we analyzed the intracellular trafficking of MP FMV , focusing on SP function, and found that MP FMV targets two subdomains in the ER and PM as well as PD. MP FMV , which has an N-terminal SP, was distributed mainly in PD and patchy microdomains of the PM (Fig  1 and Fig 2). Investigation of ER translocation efficiency revealed that SP FMV has much lower translocation efficiency compared with those of SP spo , SP cal and SP clv (Fig 3). The SP-deficient mutant (Trun) exclusively localized to the PM microdomains (Fig 4A), whereas two SP chimeras (SP cal Trun and SP clv Trun) exclusively localized to the ER (Fig 4C and 4D). SP spo Trun was distributed in the ER, PM microdomains and PD similar to MP FMV , even though a portion of SP spo Trun aggregated in the ER (Fig 4B). The results so far indicated that MP FMV dually targets the ER and PM due to the inefficient SP; a fraction of MP FMV was successfully translocated into the ER, whereas the remainder of MP FMV , which failed to be translocated, is transferred to the microdomains. This finding led us to speculate that both ER-translocated MP FMV and microdomain-localized MP FMV are necessary for PD localization. Consistent with this notion, the SP-deficient mutant entered into PD by the expression of MP FMV or the SP chimeras, which are able to, at least partially, translocate into the ER (Fig 7). Furthermore, we showed that translocated MP FMV specifically localized to ER-PM contact sites (Fig 8A and 8B), and dominant-negative inhibition of SYTA affected PD localization and cell-to-cell movement of MP FMV (Fig 8C-8E). These results suggest that MP FMV localized to ER-PM contact sites plays an essential role in the entry of microdomain-localized MP FMV into PD. PD localization of MP FMV is necessary to facilitate cell-to-cell movement as shown in the virus movement complementation assay (Fig 5) and the cell-to-cell movement assay (Fig 6). Altogether, we propose a new model for the intracellular trafficking of a viral MP. A substantial proportion of MP FMV , which failed to be translocated, is directly transferred to the microdomains, whereas the remainder of MP FMV , which successfully translocated into the ER, subsequently localizes to ER-PM contact sites and functionally interact with PD to enable microdomain-localized MP FMV to enter into PD (Fig 9).
Dual targeting of MP FMV to the ER and the PM is explained by the low translocation efficiency of the SP FMV (Fig 3). In general, proteins with an SP are co-translationally recognized by the signal recognition particle in the cytosol and recruited to the signal recognition particle receptor on the ER membrane. Then, the Sec61p complex, a channel integrated into the ER membrane, translocates these preproteins into the ER, and subsequently the SP is cleaved by a membrane-bound SP peptidase on the lumenal side [36]. Therefore, the finding that MP FMV was translocated to the ER at a lower rate (approximately 5% ; Fig 3) despite the cleavage of SP FMV (Fig 2) is surprising. A substantial proportion of MP FMV molecules probably abort ER translocation after SP cleavage and are released into the cytosol, but further studies are needed to reveal the detailed mechanism of the abortion. In animal cells, it has been reported that an inefficient SP of an ER chaperone calreticulin regulates the ratio of translocated and nontranslocated populations [37]. However, the inefficiency of the animal calreticulin SP is quite moderate, and it generates only a small nontranslocated population. In this regard, to our knowledge, this is the first report of an SP with extremely low efficiency that controls the subcellular distribution of the nascent protein. This low-efficiency SP FMV generates only a small ER-translocated population, but it is probably sufficient for the entry of microdomain-localized MP FMV into PD considering that SP spo Trun:YFP, whose SP has medium translocation efficiency, excessively accumulated in the ER compared with the case of MP FMV :YFP (Fig 4B). SP FMV likely regulates the distribution of MP FMV between the PM and ER in the appropriate proportion.
How is MP FMV associated with two different types of the cellular membrane, the ER membrane and PM? Curiously, a transmembrane domain in MP FMV has not been predicted by SOSUI (http://harrier.nagahama-i-bio.ac.jp/sosui/) and TMHMM server v. 2.0 (http://www. cbs.dtu.dk/services/TMHMM/). Localization of the MP FMV to the PM microdomains is explained as a peripheral membrane protein. Given that the SP-deficient MP FMV was exclusively associated with the PM microdomains (Fig 4A), even ER translocation is not required to localize to PM microdomains. Microdomain-associated proteins, remorins and flotillins, which do not have SPs and transmembrane domains similarly, are suggested to be peripherally associated with the PM [38][39][40]. In other words, a transmembrane domain and penetrating the membrane are dispensable for the localization to PM microdomains. The association with the ER membrane is explained by the abnormality of viral proteins. Recent studies on the topology of viral MPs revealed that MPs of tobacco mosaic virus and tomato spotted wilt virus are suggested to be associated with the ER membrane using unusual hydrophobic regions [41][42][43]. MP FMV may also establish such unconventional hydrophobic regions to be associated with the ER membrane.
Our results raise the question of how two populations of MP FMV that localize to the ER and PM subdomains act cooperatively to gain access to PD. According to BiFC analysis (S7 Fig), physical interaction of these two population is not likely. Although only limited information is available on these membrane subdomains, one possible explanation is that MP FMV in ER-PM contact sites functionally interact with PD, and this might allow the microdomain-localized MP FMV to access into PD (Fig 9). This hypothesis is corroborated by the facts that ER-PM contact sites are spatially close to PD (S6B Fig) [32] and that expression of SYTA ΔC2B , a dominantnegative form of SYTA, affected PD and PM localization of MP FMV (Fig 8C and 8D). In accord with our results, a previous study showed that, although the localization of 30K was not affected, the cell-to-cell movement of 30K was suppressed in an Arabidopsis syta mutant [16]. This study also showed that SYTA ΔC2B inhibited the formation of endosomes, suggesting that SYTA regulates endocytosis [16]. From these facts, we suspect that MP FMV in the patchy microdomains is transported into PD through endosomal trafficking regulated by SYTA (Fig  9).
The intimate relationship between PD and patchy domains in the PM is implied in this and in other studies. In our study, relocation of the microdomain-localized MP FMV to PD (Fig 7) indicates a functional connection between the patchy microdomains and PD. One previous study about TMV 30K reported that dominant-negative inhibition of class VIII myosins affected PD localization of 30K and induced a patchy distribution in the PM, which did not merge with that of REM1.3 similar to the results from this study [44]. Originally, PM passing through PD is also recognized as a type of microdomain, as it is functionally and spatially distinguished from the surrounding PM [2,45]. Taken together, these two types of PM microdomains, the patchy domains and PM passing through PD, might be functionally connected by ER-PM contact sites.
This study also presents a functional differentiation of MP FMV between the two populations, MP FMV in ER-PM contact sites and PM microdomains. The functional differentiation of a virus MP is reminiscent of the mechanism of cell-to-cell movement regulated by more than one protein. For example, triple gene block movement proteins, which are encoded by viruses belonging to the Virgaviridae, Alphaflexiviridae and Betaflexiviridae families, have specialized functions and perform different tasks for virus cell-to-cell movement: delivering viral factors to PD, interacting with host factors, and increasing PD permeability [46,47]. MP FMV plays multiple roles in cell-to-cell movement using the SP FMV with low translocation efficiency, probably to avoid splitting into modules. This concept may be true also in plant proteins. Although a number of studies have shown that a diverse array of plant proteins have N-terminal SPs, their ER translocation abilities have seldom been investigated. Our findings raise the possibility that SPs can potentially regulate subcellular distribution of the nascent proteins and contribute to protein function in plant cells.

Transient protein expression
Plasmids expressing GUS, MP FMV , MP FMV :YFP and MP FMV :CFP were prepared as described earlier [23]. Expression vectors of ER-CFP and ER-YFP were purchased from the Arabidopsis Biological Resource Center (Stock numbers CD3-953 and CD3-957, respectively). MP FMV mutant whose SP is not cleaved (ncMP FMV ) was generated by an PCR using primers containing substitutions to introduce L7P and V11P mutations. ncMP FMV was cloned into pEarley-Gate 101 (ncMP FMV :YFP) using Gateway technology [48,49]. Trun sequence, which lacks the N-terminal 19 aa of MP FMV , was amplified by PCR and cloned into pEarleyGate 100 (Trun), pEarleyGate 101 (Trun:YFP) and pEarleyGate 102 (Trun:CFP). SP chimera sequences were amplified by PCRs using primers containing each SP sequence, followed by cloning into pEar-leyGate 100 (SP spo Trun, SP cal Trun and SP clv Trun) or 101 (SP spo Trun:YFP, SP cal Trun:YFP and SP clv Trun:YFP). In these SP chimeras, the SP region of MP FMV was replaced with the N-terminal sequences of Ipomoea batatas sporamin A (M16861; 24 aa) [26], Nicotiana tabacum calreticulin (EU984501; 28 aa) [27] or A. thaliana CLAVATA3 (AF126009; 22 aa) [28], each of which contains an SP sequence and one aa downstream of the cleavage site. MP FMV :FLAG, in which MP FMV was fused to a FLAG epitope tag immediately downstream of its C terminus, was amplified by PCRs using primers containing FLAG sequence, and cloned into pEarleyGate 100. A 3× N-glycosylation sequon [29] or an ER-retention signal [50] were introduced to the GFP sequence in their C terminus (GFPglc and GFP:HDEL) as was the case with MP FMV : FLAG. The GFP sequence was derived from pEarleyGate 103. The SPs of MP FMV , sporamin A, calreticulin or CLAVATA3 were N-terminally added to GFPglc (SP FMV GFPglc, SP spo GFPglc, SP cal GFPglc and SP clv GFPglc) or GFP:HDEL (SP FMV GFP:HDEL, SP spo GFP:HDEL, SP cal GFP: HDEL and SP clv GFP:HDEL) as was the case with SP chimeras. The REM1.3 (At4g36970) and SYTA (At2g20990) sequences were amplified by PCR from total DNA of A. thaliana. REM1.3 was cloned into pEarleyGate 104 (YFP:REM1.3), and SYTA was cloned into pEarleyGate 101 (SYTA:YFP) and pEarleyGate 102 (SYTA:CFP). SYTA ΔC2B , a dominant-negative form of SYTA, was amplified by PCR according to a previous study [16], and cloned into pEarleyGate Cmyc (SYTA ΔC2B :myc). pEarleyGate Cmyc is an in-house expression vector built from pEar-leyGate 101 to introduce a myc tag at the C terminus of a cloned gene. Vectors for BiFC analysis were constructed as shown in the previous study [31]. The TMV MP:GFP sequence (30K: GFP) was amplified from pTMV-MP:GFP [51], and cloned into the pBI121 vector using SalI and BamHI sites.

Plant materials and transient expression
The upper leaves of four-week-old N. benthamiana plants were used for the transient expression assays. Transient expression was mediated by infiltration of Agrobacterium tumefaciens as described previously [23].

Confocal imaging
Cells expressing fluorescent protein fusions were imaged using a Leica TCS SP5 laser-scanning confocal microscope. An HCX PL Apo 63×/1.4-0.6 oil CS lens was used for imaging subcellular localization of fluorescent protein fusions and an HC PL Apo 10×/0.4 CS lens was used for imaging cell-to-cell movement of fluorescent protein fusions. Cells expressing CFP and/or YFP fusions were visualized as described earlier [52]. GFP was excited at the 488-nm argon laser line, and the emission was visualized at 500 to 600 nm. For PM staining, leaves were infiltrated with 50 μM FM4-64 in distilled water, and observed at 1 hpi. FM4-64 was excited at the 543-nm helium/neon laser line, and the emission was visualized at 580 to 650 nm. For PD staining, leaves were infiltrated with 0.1% (w/v) aniline blue in 50 mM sodium phosphate buffer (pH 9.0), and cells were observed at 2 hpi. Aniline blue was excited at the 405-nm laser line, and the emission was visualized at 425 to 480 nm. All the images were acquired at room temperature. Confocal images were processed with LAS AF software version 2.7.3 and Adobe Photoshop CS4. For deconvolution image analysis, between 3 and 8 z-section images at 0.15 μm intervals were captured and processed with Leica Hyvolution system. Fluorescence intensity graphs were generated using the LAS AF quantify intensity tool. PCCs were measured using an Fiji Colocalization plugin [53], and the mean values and standard deviations were calculated from three different images. Generally, PCC values of 0.2-0.4 indicate weak positive correlations and PCC values above 0.5 indicate strong positive correlations [54].

Plasmolysis and inhibition assay
Leaves expressing fluorescent protein fusions were immersed in 4% (w/v) NaCl for 15 min for plasmolysis. For the inhibition assay, leaves were treated with 50 μg/ml BFA in 0.5% (v/v) dimethyl sulfoxide at 18 hpi and observed at 6 h after the treatment.

Quantitative analysis of cell-to-cell movement
A GFP-tagged movement-defective mutant of the PVX infectious clone (PVXΔTGBp1-GFP) [55] was used in the virus movement complementation experiment. Two Agrobacterium cultures harboring the binary plasmid expressing GUS, MP FMV or an MP FMV mutant and PVXΔTGBp1-GFP were resuspended and mixed to final concentrations of OD 600 = 0.4 and 0.0002, respectively. Leaves at 5 dpi were observed under an M165 FC fluorescence stereomicroscope (Leica Microsystems) with an ET GFP filter. Images were captured by a Leica DFC 310 FX camera and LAS software version 4.4.0. The areas of fluorescent foci were measured using ImageJ software version 1.40 (National Institutes of Health). In the assessment of cell-tocell movement, an Agrobacterium culture harboring the binary plasmid expressing MP FMV or its mutants was resuspended and diluted to OD 600 = 0.0002. When cell-to-cell movement under condition of dominant-negative inhibition of SYTA was investigated, Agrobacterium cultures harboring the binary plasmid expressing SYTA ΔC2B :myc or GUS and MP FMV :YFP were resuspended and mixed to final concentrations of OD 600 = 1.0 and 0.0002, respectively. Leaves at 3 dpi were observed under the laser-scanning confocal microscope as described above.

Prediction of signal peptides
SignalP 4.1 software was used to predict SPs [57]. The accession number of MP FMV sequence is BAM13816.

Investigation of ER translocation efficiency
Deglycosylation using endoglycosidase H (Endo H; New England Biolabs) was carried out according to the manufacturer's instructions. Leaves transiently expressing GFPglc, SP FMV GFPglc, SP spo GFPglc, SP cal GFPglc or SP clv GFPglc at 30 hpi were homogenized in 1× glycoprotein denaturing buffer (0.5% SDS and 40 mM DTT), and incubated at 65˚C for 15 min. After removal of cellular debris by centrifugation, the supernatant was suspended in 1× GlycoBuffer 3 (50 mM sodium acetate, pH 6.0) followed by the addition of distilled water or Endo H. After incubation at 37˚C for 1 h, Endo H was inactivated at 75˚C for 10 min. These samples were analyzed by immunoblotting using anti-GFP antibody (Roche). Signal intensity of each band was quantified using ImageJ software. Writing -review & editing: Kazuya Ishikawa, Yasuyuki Yamaji, Shigetou Namba.