Interaction of HSP20 with a viral RdRp changes its sub-cellular localization and distribution pattern in plants

Small heat shock proteins (sHSPs) perform a fundamental role in protecting cells against a wide array of stresses but their biological function during viral infection remains unknown. Rice stripe virus (RSV) causes a severe disease of rice in Eastern Asia. OsHSP20 and its homologue (NbHSP20) were used as baits in yeast two-hybrid (YTH) assays to screen an RSV cDNA library and were found to interact with the viral RNA-dependent RNA polymerase (RdRp) of RSV. Interactions were confirmed by pull-down and BiFC assays. Further analysis showed that the N-terminus (residues 1–296) of the RdRp was crucial for the interaction between the HSP20s and viral RdRp and responsible for the alteration of the sub-cellular localization and distribution pattern of HSP20s in protoplasts of rice and epidermal cells of Nicotiana benthamiana. This is the first report that a plant virus or a viral protein alters the expression pattern or sub-cellular distribution of sHSPs.

sHSP was induced in tobacco as a defense response to bacterial infection 12 . Stress granules are induced and then sometimes dispersed in some animal cells in response to virus infection 13 but there have been no reports that SHSPs are involved in response to virus infection. In contrast, there are several reports that plant HSP70 proteins play roles during infection by various viruses including geminiviruses 14,15 and potyviruses 16 .
Rice stripe disease is one of most devastating viral diseases of rice in East Asia 17,18 . Infected plants often have chlorotic stripes or mottling and necrotic streaks in the newly expanded leaves and growth is stunted 17,19,20 . The causal agent, Rice stripe virus (RSV), is one of best-studied rice viruses and is the type member of the genus Tenuivirus 21 . RSV is transovarially transmitted by small brown planthopper (SBPH) (Laodelphax striatellus) in a persistent and circulative-propagative manner 22,23 . Although RSV only infects rice and other poaceous plants in nature 22 , it can be transmitted experimentally to the model dicot Nicotiana benthamiana 24,25 .
The RSV genome consists of four single-stranded RNA segments, designated RNA1 to RNA4 in the decreasing order of their molecular weight, which encode seven ORFs using a negative or ambisense coding strategy 26 . RNA1 (~9 kb) is the largest RNA segment and has a single ORF in the viral-complementary sense, encoding a 337-kDa protein that is an RNA-dependent RNA polymerase (RdRp) 27,28 . The other three segments (RNA2, 3.5 kb; RNA3, 2.5 kb; RNA4, 2.2 kb) each have two open reading frames (ORFs) one on the viral RNA (vRNA) and the other on the viral complementary strand (vcRNA) [29][30][31] . RSV vRNA2 encodes a membrane-associated protein p2 that reportedly acts as an RNA silencing suppressor 32 . The vcRNA2 encodes a glycoprotein pc2 of unknown function but which moves from the ER to the Golgi bodies in a manner strictly dependent on COP I and COP II early secretion pathways 33,34 . RNA3 encodes an RNA silencing suppressor p3 from the vRNA and a nucleocapsid protein pc3 from the vcRNA 29,35 . RSV vRNA4 encodes a disease-specific protein p4 that accumulates in both infected plant and insect cells 36 . The protein encoded by vcRNA4 (pc4) has been identified as a cell-to-cell movement protein 24 . Infection by RSV selectively modifies the expression of host genes and establishes a complex interaction with host cell components to block cellular defense mechanisms and hijack host cell machinery 20,32,37,38 . RSV p2 can bind to OsSGS3, a rice host protein 32 , PsbP, a 23 kDa oxygen-evolving complex protein of plants, has been shown to interact with p4, and its silencing resulted in more severe symptoms and the accumulation of viral RNAs 37 . Recently, host HSP70 has been shown to interact with RSV RdRp, and RSV RNAs were reduced in HSP70-silenced Nicotiana benthamiana 38 .
In this study we investigated the interactions between sHSPs and RSV-encoded proteins and found that the expression and sub-cellular localization of a host small heat shock protein 20 (HSP20) was significantly affected by RSV infection and that this was caused by its interaction with the viral RdRp.

Expression of a conserved small heat shock protein (sHSP) was regulated by viral infection.
Searches in both databases of rice expression profiles (http://cdna01.dna.affrc.go.jp and http://rice.plantbiology.msu.edu) showed that the expression of a rice small heat shock protein 20 (OsHSP20), encoded by Os03G026700, was significantly changed by infection with at least seven rice viruses, including RSV and also rice black-streaked dwarf virus (RBSDV), rice dwarf phytoreovirus (RDV), rice galled dwarf phytoreovirus (RGDV), rice grassy stunt tenuivirus (RGSV), rice tungro bacilliform virus (RTBV), and rice transitory yellowing virus (RTYV) (See Supplementary Materials Figure S1). Two published data sets with both Agilent and Affymetrix rice genome microarrays consistently showed that OsHSP20 accumulated in rice plants after RSV infection 20,39 , suggesting that OsHSP20 might be involved in the host response to RSV infection. We then performed quantitative RT-PCR (qPCR) and confirmed that OsHSP20 was significantly up-regulated by RSV infection in our rice plants (Fig. 1). N. benthamiana is an excellent experimental host of RSV that provides a simple viral inoculation system for further study of RSV-plant interactions 24,25 and we confirmed that its HSP20 homologue was also up-regulated by RSV infection in N. benthamiana plants (data not shown). We then cloned both full-length coding regions of the OsHSP20 gene from rice and its homolog (NbHSP20) from N. benthamiana by RT-PCR with the respective primer pairs Os20F/Os20R and Nb20F/Nb20R ( Table 1). Members of the HSP20 subfamily have 65-86% identity to one another and contain a conserved motif in the N-terminal part in addition to the characteristic sHSP alpha-crystallin domain (ACD) at the C-terminal region (See Supplementary Materials Figure S2). OsHSP20 and NbHSP20 self-interacted in yeast cells. The full-length coding regions of OsHSP20 and NbHSP20 were separately cloned into the GAL4 binding domain vector pGBKT7 and activation domain vector pGADT7 and then combinations of plasmids expressing bait proteins BD-OsHSP20/ AD-OsHSP20 or BD-NbHSP20/AD-NbHSP20 were co-transformed into S. cerevisiae Y2HGold cells after eliminating the autoactivation and toxicity. The resultant transformants were selected on SD/-Ade/-His/-Leu/-Trp/X-α -Gal/AbA medium. Both transformants of BD-OsHSP20/AD-OsHSP20 and BD-NbHSP20/AD-NbHSP20 as well as the positive control grew well and turned blue. In contrast, no growth was observed in the negative controls (Fig. 2). These results demonstrated that these HSP20 subfamily proteins self-interacted in yeast cells. OsHSP20 and NbHSP20 formed granules in vivo in the cytoplasm when expressed alone. To examine the sub-cellular localization of both OsHSP20 and NbHSP20 proteins, constructs expressing OsHSP20 or NbHSP20 fused with eGFP at their C terminus (pCV-OsHSP20-GFP and pCV-NbHSP20-GFP) were constructed and introduced into N. benthamiana epidermal cells by Agrobacterium infiltration. In confocal microscopy 2 days post-infiltration (dpi) GFP fluorescence was localized to numerous granules of various sizes in the cytoplasm of cells expressing OsHSP20-GFP or NbHSP20-GFP. No fluorescence was seen in the nucleus. In the control, the non-fused GFP was distributed generally in the cytoplasm and nucleus, which indicated that the moiety GFP did not affect the localization of OsHSP20-GFP or NbHSP20-GFP (Fig. 3A). The same results were obtained when the plasmids were delivered into rice protoplasts via polyethylene glycol (PEG) transfection (Fig. 3B).
In confocal microscopy, the OsHSP20-GFP and NbHSP20-GFP granules moved in the cytoplasm. To record the movement of OsHSP20-GFP or NbHSP20-GFP granules, four sequential photographs were taken over a period of 30 s (Fig. 3C). In addition, a video recording the movement of OsHSP20-GFP and NbHSP20-GFP under the GFP channel was taken. The granules moved at different speeds and often moved into another focal plane (See Supplementary Materials Videos S1 and S2).
Distribution patterns of both OsHSP20 and NbHSP20 were affected by RSV infection. To further determine the relationship between the HSP20s and RSV infection, plasmids expressing OsHSP20-GFP or NbHSP20-GFP were introduced into healthy or RSV-infected N. benthamiana epidermal cells by Agrobacterium infiltration. Both proteins were localized in the cytoplasm of RSV-infected cells (Fig. 4) and no fluorescence was seen in the nucleus, as in healthy cells (Fig. 3). However, while numerous granules with a variety of sizes were observed in the cytoplasm of non-infected cells almost no GFP granules were detectable in the cytoplasm of RSV-infected cells (Fig. 4).

Both HSP20s interacted with RSV RdRp.
To investigate the interaction between the HSP20 and viral proteins, the full-length coding regions of OsHSP20 and NbHSP20 were cloned into the GAL4 binding domain vector pGBKT7 as bait for screening a prey library of RSV cDNA in yeast two-hybrid (YTH) assays. More than 10 independent clones were recovered following growth on selective media and sequences of all these clones were 100% identical to RSV genomic segment RNA1 encoding the viral RdRp. Combinations of plasmids expressing bait proteins BD-OsHSP20 or -NbHSP20 and prey protein AD-RdRp were then co-transformed into S. cerevisiae using the plasmid combinations BD-Lam/AD-T, BD/AD-RdRp, BD/AD -OsHSP20 or BD/AD -NbHSP20 as negative controls and BD-53/AD-T as a positive control. Only transformants of BD-OsHSP20/AD-RdRp, BD-NbHSP20/AD-RdRp and the positive control grew well on the selective medium and turned blue. In contrast, no growth was observed in the negative controls (Fig. 5). These results consistently indicated a strong interaction between the HSP20 proteins and RSV RdRp.
The N-terminal part of RSV RdRp was responsible for its interactions with OsHSP20 and NbHSP20 in vivo and in vitro. The large RdRp protein of RSV has multiple functional motifs and domains 40 . Five fragments of the RdRp, divided on the basis of the conserved domains (RdRp 1-296 , RdRp 297-987 , RdRp 988-1810 , RdRp 1811-2387 and RdRp 2388-2920 ), were then tested separately by cloning into the yeast vectors to determine which part was involved in the interaction with HSP20 (Fig. 6A). Only transformants expressing BD-RdRp 1-296 /AD-OsHSP20, BD-RdRp 1-296 /AD-NbHSP20 and the positive control grew well on the selective medium and turned blue, suggesting that the N-terminal part of RdRp should be crucial for the interaction between the HSP20s and the viral RdRp (Fig. 6B,C).

Primers
Sequences (5′-3′)  Interactions between the N-terminus of RSV RdRp and host HSP20 were further confirmed through in vitro pull-down assays. Equal amounts of in vitro translated c-Myc-RdRp 1-296 and OsHSP20-GFP or NbHSP20-GFP were mixed and then pulled down with the GFP-Trap M beads followed by western blot assays using an anti-c-Myc or anti-GFP antibody. In these experiments, combinations of OsHSP20-GFP and c-Myc-p53, NbHSP20-GFP and c-Myc-p53, GFP and c-Myc-RdRp 1-296 , and GFP and c-Myc-p53 were used as controls. Immunoblot analyses using an anti-c-Myc antibody demonstrated that c-myc-RdRp 1-296 was pulled down by OsHSP20-GFP and NbHSP20-GFP, but not by the control GFP ( Fig. 8, top), although the anti-GFP antibody did pull-down OsHSP20-GFP, NbHSP20-GFP and control GFP proteins (Fig. 8, bottom). These experiments clearly demonstrated that RdRp 1-296 bound to both OsHSP20-GFP and NbHSP20-GFP in vitro. Thus these results consistently indicated that the N-terminal part of RSV RdRp was responsible for its interactions with OsHSP20 or NbHSP20 in vivo and in vitro.

Restriction sites (underlined)
Both sub-cellular localization and distribution patterns of OsHSP20 and NbHSP20 were affected by the N-terminal part of RSV RdRp. To determine whether the sub-cellular alteration of HSP20s was associated with that of RSV RdRp, plasmids expressing RdRp 1-296 -GFP and RdRp 1-296 -mCherry were first constructed and introduced into N. benthamiana epidermal cells by Agrobacterium infiltration. Fluorescence microscopy indicated that both RdRp 1-296 -GFP and RdRp 1-296 -mCherry resulted in a pattern of diffuse and uniform fluorescence in the cytoplasm and nucleus at 2 dpi (Fig. 9A). When RdRp 1-296 -mCherry was co-expressed with either OsHSP20-GFP or NbHSP20-GFP, the two proteins co-localized in a pattern identical to that formed by RdRp 1-296 -mCherry alone (Fig. 9B) and no GFP granules were seen in the cytoplasm. Control combinations ruled out the possibility that GFP or mCherry expression might have some aberrant effects on the distribution of RdRp1 -296 -mCherry, OsHSP20-GFP or NbHSP20-GFP. The co-localization of RdRp 1-296 and HSP20 confirmed that the interaction between RSV RdRp and HSP20 mediated a dramatic effect on the distribution of HSP20. The sub-cellular distribution of GFP-fused OsHSP20 or GFP-fused NbHSP20 when co-expressed with RSV RdRp was very similar to that in RSV-infected cells, suggesting that RSV RdRp was responsible for both sub-cellular localization and distribution patterns of OsHSP20 and NbHSP20 in infected plants.

Discussion
sHSPs, defined by possessing a conserved α -crystallin domain (ACD), are the most abundant and complex subset of HSPs in plants 41 . A key function of the sHSPs is to prevent aggregation of denatured proteins. By forming a soluble complex with substrate proteins, they can create a transient reservoir of substrates for subsequent refolding by ATP-dependent chaperone systems 6,8,10,11 . Rice has a total of 14 sHSPs subfamilies, which are predicted to locate to various cellular organelles, including the cytosol, nucleus, chloroplasts, mitochondria, endoplasmic reticulum, and peroxisomes 42 . Organelle-targeted sHSPs are unique to plants and their diverse functions are thought to be associated with their sub-cellular localization 43 although little is known about their distinct functions 41 . Here both OsHSP20 and NbHSP20 were shown to be localized within the cytosol (Fig. 3), which was consistent with the predicted localization of the Class I (CI) subfamily of sHSPs 42 , the subfamily of which both HSP20s appear to be members. Sequence analysis showed that such HSP20s from a variety of plant species contained a highly conserved motif in the N-terminal part that has not previously been described in addition to the characteristic ACD domain of sHSPs 43 . The novel conserved motif "DPFSLDVWDPF" may act as a signal involved in their Figure 2. Self-interaction of OsHSP20 or NbHSP20 in yeast cells. Yeast colonies co-expressing BD-OsHSP20 and AD-OsHSP20, or BD-NbHSP20 and AD-NbHSP20 grew well and turned blue on SD/-Ade/-His/-Leu/-Trp/X-α -Gal/AbA medium as did yeast colonies expressing BD-53 with AD-T, which was used as the positive control. Yeast co-transformed with BD-Lam and AD-T, BD-OsHSP20 and AD, BD and AD-OsHSP20, BD-NbHSP20 and AD, or BD and AD-NbHSP20 were used as negative controls. localization. Within the cytosol, both GFP-fused OsHSP20 and NbHSP20 formed fluorescent granules of different sizes and shapes when expressed alone in N. benthamiana epidermal cells or rice protoplasts (Fig. 3) and strong self-interactions were always observed in yeast colonies co-expressing BD-OsHSP20/ AD-OsHSP20, or BD-NbHSP20/AD-NbHSP20 (Fig. 2), suggesting that both HSP20 proteins could function in vivo as dimers or larger oligomers. This was consistent with many previous reports showing that some sHSPs formed large oligomers in vitro from multiple subunits 41,44,45 . Interestingly, the fluorescent granules formed by GFP-fused OsHSP20 and NbHSP20 were able to move within the cytosol, suggesting that they might interact with other components of the plants. Most likely, these cytoplasmic granules were aggregates consisting mainly of oligomerized sHSPs, which could be, to a degree, similar in structure and composition to those of heat shock granules formed by accumulation of HSPs under hyper-thermic condition 9 . Further work is needed to identify those components and help understand the significance of these granules.
Interactions between host plant proteins and viral components are presumed to play an important role in the RSV life cycle or in viral pathogenicity. In addition to interactions between RSV or specific RSV gene products and host proteins that have already been reported 32,37,38 , we have now shown that the expression pattern and sub-cellular distribution of an OsHSP20 gene was regulated by RSV infection. This is the first report that sHSPs could be manipulated by a plant virus. A series of in vivo and in vitro protein-protein interaction assays further showed that the HSP20 interacted with the RSV RdRp and is the first report of a sHSP-plant viral RdRp interaction.
In the YTH assays, host HSP20 proteins interacted specifically with the N-terminus (residues 1-296) of RSV RdRp (Fig. 6). This segment contains a viral ovarian tumour (OTU) domain near its N-terminus which is thought to have deubiquitination activity 40,46 . The deubiquitylating function of viral RdRp has been suggested to provide a way to interfere with the proteolytic pathway in host plants during viral infection 40,46 . This might be a novel strategy of viral pathogenicity considering that sHSPs function as important players in regulating cellular proteostasis 47 . A previous study reported that RSV pc4

Figure 5. Interactions of RSV RdRp with the OsHSP20 or NbHSP20 in co-transformed yeast cells grown on SD/-Ade/-His/-Leu/-Trp/X-α-Gal/AbA.
Yeast colonies expressing BD-RdRp with AD-OsHSP20 or AD-NbHSP20 were able to grow well and turned blue on SD/-Ade/-His/-Leu/-Trp/X-α -Gal/AbA medium as did yeast colonies expressing BD-53 with AD-T, which was used as the positive control. Yeast co-transformed with BD-Lam and AD-T, BD-RdRp and AD, BD and AD-OsHSP20, or BD and AD-NbHSP20 were used as negative controls. interacted with a small heat shock protein (EU325986) from Wuyujing 3, a susceptible japonica rice cultivar 48 , in yeast cells. However, in our YTH assays, only interaction between the OsHSP20 and viral RdRp was detected. To determine whether the HSP20 interacted with RSV pc4, we also investigated the pc4-OsHSP20 or pc4-NbHSP20 interaction by YTH assays, but no interaction was detected between the two proteins ( Figure S3). Nucleotide sequence analysis indicated that pc4 gene sequences varied among RSV isolates from different geographical origins 49 , which might explain why pc4 did not interact with OsHSP20 or NbHSP20 in our experiments.
In transient expression systems, GFP fluorescence appeared as granular structures in the cytoplasm when HSP20 was expressed alone in N. benthamiana epidermal cells or rice protoplasts (Fig. 3). However, no GFP granules or reconstituted YFP-fluorescent granules were detected when the N-terminus of RSV RdRp and HSP20 were co-expressed in the leaves of N. benthamiana or in rice protoplasts (Figs 7 and 9), and the RdRp 1-296 /HSP20 complex had a diffuse distribution pattern in the cytoplasm and nucleus (Fig. 7), similar to that in N. benthamiana epidermal cells co-expressing RdRp 1-296 -mCherry/OsHSP20-GFP or RdRp 1-296 -mCherry/NbHSP20-GFP (Fig. 9). These results consistently suggested that the strong interactions between the N-terminus of RSV RdRp and HSP20 significantly affected the sub-cellular localization and distribution pattern of host HSP20 proteins. We hypothesize that the hetero-interaction between the HSP20 and viral RdRp destroyed the self-interaction of HSP20, leading to the disappearance of fluorescent granules and the diffuse distribution pattern of HSP20. If sub-cellular localization and self-interaction in vivo is important for its chaperone activities, it would be expected that this distributional change would affect the function of HSP20. In addition, RSV infection also affected the sub-cellular distribution of host HSP20 proteins. The GFP granules were not present if OsHSP20-GFP or NbHSP20-GFP were expressed in the cytoplasm of RSV-infected N. benthamiana epidermal cells (Fig. 4). Previous research has revealed that some plant HSPs (especially HSP70s) were involved in the viral life cycle, including cell entry, virion assembly, the transfer of the viral genome segments into the nucleus, replication of the viral genome, morphogenesis of the virion particles and transformation of the cell 4,50 . However, the biological function of plant sHSPs in viral infection has been unknown. We have now demonstrated that the host HSP20 proteins interacted with the viral RdRp of RSV and that the N-terminal part of the RdRp was crucial for this interaction. Further experimental work will be needed to understand the significance of this interaction but the fact that the viral RdRp interferes with the self-interaction of the host HSP20 suggested that the virus might suppress the formation of stress granules that could potentially be an anti-viral response. Stress granules are initially induced but then dispersed in animal cells infected by poliovirus 51 and this could be a broadly similar effect. Alternatively (or additionally), it is interesting that a previous study showed that HSP70 is necessary for RSV infection and that HSP70 also interacts with the N-terminal part of RSV RdRp. This may suggest that interactions between various HSPs and the viral RdRp play an important role in viral replication. These findings are therefore a step forward in understanding a virus that causes a seriously damaging disease of one of the most important crop plants in the world.

Materials and Methods
Plant materials and growth conditions. Nicotiana benthamiana plants were grown in 10-cm pots filled with a mixture of 60% vermiculite and 40% meadow soil and mechanically inoculated at the six-leaf stage with crude extracts from RSV-infected O. sativa leaves, as described previously 52 . Rice (O. sativa) was grown by germinating seeds on mesh supported in plastic containers containing ½ strength Murashige and Skoog (MS) nutrient solution. All plants used in this study were grown in a growth chamber at 25 °C with 16 h light/8 h dark and 70% r.h.

Cloning and sequencing of a rice sHSP and its homolog in Nicotiana benthamiana. The cod-
ing sequence of intact OsHSP20 was amplified by PCR with the primer pair Os20F/Os20R (Table 1) from a leaf cDNA library of rice cv. Nipponbare plants, ligated into the pGEM-T vector (Promega, Madison, WI, USA), and then transformed into competent Escherichia coli DH5α . The recombinant plasmid DNA (pGEM-OsHSP20), used for sequencing was prepared using the QIAprep spin mini prep kit (Qiagen), and the inserts were sequenced entirely on both strands using the BigDye Terminator v3.1 Cycle Sequencing Kit (Perkin Elmer Applied Biosystems, Foster, USA) on an ABI PRISM 3730 DNA Sequencer (Perkin Elmer Applied Biosystems, Foster, USA) with universal primers T7 and SP6. The sequence of HSP20 gene from N. benthamiana was searched on the website of the SOL genomic network (http://solgenomics.net). The full-length coding sequence of a homologue of OsHSP20 in N. benthamiana (Niben.v0.4.2.Scf54613, named as NbHSP20 in this study) was cloned with the primer pair Nb20F/Nb20R (Table 1) from N. benthamiana leaf cDNA and the resultant recombinant plasmid DNA (pGEM-NbHSP20) was sequenced as described above. Sequence assembly and analysis was performed using the DNAman version 6.0 program (Lynnon BioSoft, Quebec, Canada). Yeast two-hybrid (YTH) assays. The yeast GAL4 binding domain vector pGBKT7 and GAL4 activation domain vector pGADT7 (Clontech, Palo Alto, CA) were used for yeast two hybrid (YTH) assays. RSV-Zhejiang isolate and its cDNA library were from our laboratory 40 . To construct plasmids for YTH analysis, the coding sequence of OsHSP20 was digested from the recombinant plasmids pGEM-OsHSP20 with the restriction enzymes NdeI and BamHI and then inserted into the NdeI/BamHI sites of yeast GAL4 binding domain pGBKT7 and activation domain pGADT7 vectors, creating the recombinant bait plasmid BD-OsHSP20 and prey plasmid AD-OsHSP20, respectively. The coding sequence of NbHSP20 was digested from the recombinant plasmid pGEM-NbHSP20 with restriction enzymes NdeI and EcoRI, and then inserted into the NdeI/EcoRI sites of pGBKT7 and pGADT7 vectors, creating the recombinant plasmids BD-NbHSP20 (bait) and AD-NbHSP20 (prey).
These PCR amplifications used LA Taq polymerase (TaKaRa Bio, Dalian, China), and the PCR amplification program as follows: preheating for one cycle of 3 min at 94 °C; 30 cycles of 30 s at 94 °C, 40 s at 58 °C, 1-3 min at 72 °C; and a final extension at 72 °C for 10 min. All clones derived from the PCR products were verified by sequencing, and the recombinant plasmids were confirmed by restriction analyses.
The in vitro pull-down assay was performed using a TNT Coupled Wheat Germ Extract System following the manufacturer's recommendations (Promega). The plasmids for the generation of c-Myc-p53 fusion protein and GFP protein were used as negative controls in the experiments. In vitro translated c-Myc-RdRp 1-296 (15 μ L) and OsHSP20-GFP (15 μ L) or NbHSP20-GFP (15 μ L) were mixed and incubated at 4 °C overnight, then 20 μ L of GFP-Trap M beads (ChromoTek, Martinsried, Germany) were added and incubated for another 6 h. The beads were washed four times with 0.5 ml washing buffer (0.05 M Tris-HCl, 0.15 M NaCl) and the bound complexes were eluted by boiling with 1 × protein loading buffer for 5 min. Samples were analyzed by western blotting as described elsewhere 38 . Briefly, Protein samples were separated by 12% SDS-PAGE gel electrophoresis and transferred by electroblotting to PVDF membrane (Millipore, Bedford, MA, USA). Transferred proteins were detected with anti-c-Myc or anti-GFP (monoclonal antibody; 1:2,000 dilution; Quanshijin, Beijing, China) primary antibodies and an anti-mouse horseradish peroxidase-conjugated secondary antibody (1:5,000 dilution; Kangweishiji, Beijing, China).

Rice protoplast isolation and transfection.
Rice protoplasts were isolated from 2-week-old seedlings as described 54 with minor modifications. Briefly, young leaves and sheaths were chopped and dipped in enzyme solution (0.5 M mannitol, 1.5% cellulose RS (Yakult Honsa, Tokyo, Japan), 0.75% macerozyme R10 (Yakult Honsa), 1 mM CaCl 2 , and 0.1% BSA). This mixture was incubated on a shaking incubator (60 rpm) for 4 to 5 h at room temperature in the dark then filtered through Miracloth. Protoplasts were pelleted by centrifugation for 5 min at 200 g and resuspended in an equal volume of W5 solution (154 mM NaCl, 125 mM CaCl 2 , 5 mM KCl, and 1.5 mM MES, adjusted to pH 5.7). Protoplasts were centrifuged and re-suspended in MMG solution (0.4 M mannitol, 15 mM MgCl 2 , and 4.7 mM MES, adjusted to pH 5.7). Plasmid DNA (10 or 20 μ g) was added to the protoplast solution and transfected with 40% polyethylene glycol (PEG) solution (40% PEG 4000, 0.4 M mannitol, and 100 mM Ca (NO 3 ) 2 ) for 20min at room temperature. W5 solution was added stepwise to dilute the PEG solution and discarded. Transfected protoplasts were incubated overnight at room temperature and then observed under confocal microscopy.

Sub-cellular localization and BiFC assay. The binary expression vectors pCV-GFP-N1
and pCV-mCherry-N1, and the BiFC vectors pCV-nYFP-C and pCV-cYFP-C (for split YFP N-terminal/C-terminal fragment expression) were previously constructed in our laboratory 55 . For BiFC assays, the coding sequence of the N-terminal fragment (residues 1-296) of RSV RdRp was amplified using primer pair B1-1F/B1-1R (Table 1) and cloned into pCV-nYFP-C as a fusion with the N-terminal fragment of YFP via the BamHI/SacI sites, forming pCV-nYFP-RdRp 1-296 . The full-length coding sequences of OsHSP20 and NbHSP20 were amplified by PCR using primer pairs BOs20F/BOs20R and BNb20F/BNb20R (Table 1), respectively. PCR products were cloned into the BamHI/SacI or KpnI/SacI sites of pCV-cYFP-C as a fusion with the C-terminal fragment of YFP, resulting in pCV-cYFP-OsHSP20 and pCV-cYFP-NbHSP20, respectively.
The recombinant binary constructs above were introduced into Agrobacterium tumefaciens strain C58C1 by electroporation (Bio-Rad Gene Pulser, 0.2 cm cuvettes, 25 microF, > 2.5 kV). Agroinfiltration was done as described 56 with a few modifications. Briefly, cultures of C58C1 harbouring a relevant binary plasmid were grown in YEP medium containing rifampicin (50 μ g ml −1 ) and kanamycin (100 μ g ml −1 ) at 28 °C for 16 h. For the BiFC assay, C58C1 strains containing the BiFC plasmids were re-suspended and adjusted to an OD 600 of 0.8:0.8 with infiltration medium (10 mM MES, pH 5.6, 10 mM MgCl 2 , 200 mM acetosyringone) before leaf infiltration. For sub-cellular localization, Agrobacterium cultures containing the pCV derivatives were re-suspended and diluted to an OD 600 of 0.6 before leaf infiltration. The cell suspensions were incubated at room temperature for 2 to 4 h and then used to infiltrate 5-to 6-week-old N. benthamiana leaves. Expression of fluorescent proteins was examined at 48 h post agroinfiltration 57 .
Confocal microscopy. Fluorescence analysis was performed using a Leica TCS SP5 confocal laser scanning microscope (Leica Microsystems, Heidelberg, Germany). GFP was excited at 488 nm and the emitted light captured between 500-550 nm, YFP was excited at 514 nm and the emitted light captured between 530-600 nm, and mCherry was excited at 561 nm and emission light captured between 570-630 nm. For analysis of co-localization assays, multi-tracking was used to prevent emission cross-talk between the channels. Images were captured digitally and handled using the Leica TCS software. Post-acquisition image processing was done with Adobe Photoshop version 7.0 software (Adobe Systems Inc., San Jose, CA, USA).