Localization of Poa semilatent virus cysteine-rich protein in peroxisomes is dispensable for its ability to suppress RNA silencing

A natural host defence response that is directed to selectively degrade invading virus RNAs is the manifestation of a general mechanism of RNA turnover, termed RNA silencing (Voinnet, 2001; Moissiard & Voinnet, 2004). Replicative intermediates of RNA-containing viruses are represented by double-stranded RNAs (dsRNAs), which are the key inducers of RNA silencing (Fire et al., 1998; Bass, 2000). dsRNAs are processed by an RNase III-like nuclease into 2126 nt double-stranded fragments, referred to as short interfering RNAs (siRNAs) (Hamilton & Baulcombe, 1999; Elbashir et al., 2001). siRNAs are incorporated into an RNA-induced silencing complex (RISC), a multicomponent RNase that is guided by siRNA to specifically cleave complementary single-stranded RNA (ssRNA) molecules (Hammond et al., 2000). Local induction of RNA silencing in plants is followed by both short-distance spread of the silenced state to surrounding cells and long-distance systemic spread of a mobile silencing signal across the whole plant (Voinnet & Baulcombe, 1997; Klahre et al., 2002; Himber et al., 2003).

Plant viruses have evolved an ability to counteract RNA silencing with silencing-suppressor proteins (Kasschau & Carrington, 1998; Voinnet et al., 1999). By now, silencing suppressors have been identified in more than 12 genera of plant viruses. The structure of viral silencing suppressors is strikingly diverse. Moreover, the ability to suppress silencing is often found as an additional activity of proteins that were previously known to have another function in the virus life cycle, such as movement protein, coat protein (CP) or a component of the viral replicase (Moissiard & Voinnet, 2004; Silhavy & Burgyán, 2004).

The molecular mechanisms of RNA silencing suppression by viral proteins are poorly understood. A remarkable exception is the tombusviral suppressor P19, which has recently been demonstrated to block the silencing machinery by specifically sequestering siRNA molecules, thereby preventing their incorporation into RISC complexes (Silhavy et al., 2002; Vargason et al., 2003; Ye et al., 2003). Many viral silencing suppressors are RNA-binding proteins and, moreover, even silencing-unrelated, dsRNA-binding proteins have been shown to exert silencing-suppression activity in plants (Lichner et al., 2003). However, siRNA binding is unlikely to be the general mechanism of suppression, as the unique structural fold formed by the P19 homodimer to selectively bind the 19 nt RNA duplex (Vargason et al., 2003; Ye et al., 2003) is apparently absent in many viral suppressors (Moissiard & Voinnet, 2004). Due to the diversity of suppressors, it is generally believed that these proteins could target different components of the silencing machinery and, therefore, interfere with different steps of RNA silencing suppression (Moissiard & Voinnet, 2004; Silhavy & Burgyán, 2004). For instance, the potyviral suppressor HC-Pro inhibits RISC effector complexes that are involved in siRNA-guided degradation of target RNAs (Kasschau et al., 2003) and interferes with dsRNA processing into siRNA (Dunoyer et al., 2004), whereas the cucumoviral suppressor 2b efficiently blocks systemic spread of the RNA silencing signal (Guo & Ding, 2002).

Recently, a silencing suppressor has been identified in the genus Hordeivirus (Yelina et al., 2002; Bragg & Jackson, 2004). Viruses of this genus contain a tripartite, positive-stranded RNA genome consisting of RNAα, RNAβ and RNAγ. RNAα is monocistronic and encodes a component of viral replicase; RNAβ encodes the CP and three movement proteins; RNAγ is bicistronic and encodes the other component of viral replicase (γa protein) and a non-structural γb protein (Jackson et al., 1989; Solovyev et al., 1996; Savenkov et al., 1998; Lawrence et al., 2000). The N-terminal γb region is able to bind ssRNA and Zn²⁺ (Donald & Jackson, 1996; Bragg et al., 2004), whereas the C-terminal γb region contains a predicted coil-coiled structure that is responsible for protein self-interactions (Bragg & Jackson, 2004). Heterologous complementation and cross-protection assays have revealed the silencing-suppressor function of γb encoded by Poa semilatent virus (PSLV) (Yelina et al., 2002). For γb of Barley stripe mosaic virus (BSMV, the type species of the genus Hordeivirus), the ability to suppress RNA silencing was demonstrated in an agrobacterium-mediated transient-expression assay (Bragg & Jackson, 2004). These findings are consistent with previous reports showing that γb is dispensable for virus replication, but influences hordeivirus genome amplification, determines viral long-distance movement and affects infection phenotype (Petty et al., 1990, 1994; Donald & Jackson, 1994).

In this paper, we studied the subcellular localization of PSLV γb protein and the possible role of its subcellular localization signal in the silencing-suppression function.

Recombinant clones.
To construct γbΔSKL with a truncation of 5 aa at the C terminus, the γb gene was amplified with specific primers and cloned as an NcoIBamHI fragment. The γb and γbΔSKL genes were subcloned under the control of a duplicated Cauliflower mosaic virus 35S promoter and Tobacco etch virus translational enhancer into NcoIBamHI-digested vector pCK-GFPS65T (Reichel et al., 1996). Expression cassettes flanked by HindIII sites were subcloned from the resulting plasmids in binary vector pLH7000 (Hausmann & Töpfer, 1999), provided by Dr L. Hausmann (Federal Centre for Breeding Research on Cultivated Plants, Germany). Similarly, a modified GFP gene called cycle 3 (GFPC3) was amplified on the 30B-GFP-C3 template (Shivprasad et al., 1999), kindly provided by William O. Dawson (Department of Plant Pathology, University of Florida, USA), cloned as an NcoIBamHI-fragment into pCK and then transferred into pLH7000. dsGF contained the 342 nt GFPC3 5'-proximal region, followed by the 3'-untranslated region of BSMV RNAβ and the same GFPC3 gene fragment in the reverse orientation. This construct was cloned in pCK and then transferred to pLH7000, as described above. To construct pRT-GFP-γb and pRT-GFP-γbΔSKL, the γb and γbΔSKL genes (as NcoIBamHI fragments) were cloned into similarly digested pRT-GFP-15K (Solovyev et al., 2000) to replace the 15K gene. To obtain PVX-PSγbΔSKL, the γbΔSKL gene was amplified by PCR and inserted between the NheI and SalI restriction sites of pPVX201 as described for PVX-PSγb (Yelina et al., 2002). BγP-γbΔSKL was constructed by substitution of the γbΔSKL gene for the γb gene in BγP (Yelina et al., 2002).

Monoclonal antibodies (mAbs).
For mouse immunization, 50 µg soluble PSLV GFPγb fusion, mixed with an equal volume of Freund's complete adjuvant, was injected intraperitoneally and boosted twice with the same dose mixed with incomplete adjuvant at 2 week intervals. The final injection of 50 µg antigen was without adjuvant. Immunized spleen cells were fused 3 days later with the mouse myeloma cell line Sp2/0 by using 45 % polyethylene glycol. Cells were cultured under selective conditions on Dulbecco's HAT MEM (Amresco) supplemented with 15 % fetal calf serum (Loewe) in the presence of mouse peritoneal macrophages as feeder cells, as described previously (Erokhina et al., 2000). Culture fluids were screened for specific antibody production by indirect ELISA. Hybridomas secreting specific mAbs were cloned twice under limiting-dilution conditions. mAbs were purified from ascitic fluids by affinity chromathography on protein ASepharose (Sigma).

Transient expression and plant analyses.
Inoculation of Nicotiana benthamiana plants with recombinant viruses and Western and Northern blot analyses were carried out as described by Yelina et al. (2002). Agrobacterium tumefaciens strain GV2260/C58C1 was used for the agroinfiltration assay. Bacteria carrying pLH7000-based vectors were cultured overnight at 28 °C with 10 mM MES and 20 µM acetosyringone, then resuspended in 10 mM MES (pH 5·5), 10 mM MgCl₂ and 150 µM acetosyringone to a final density of OD₆₀₀=2, incubated for 3 h at room temperature and infiltrated into N. benthamiana leaves. In silencing-suppression experiments, Agrobacterium cultures were mixed prior to all infiltrations. Green fluorescent protein (GFP) reporter-gene culture (1 vol.) was mixed with either (i) 2 vols culture containing empty vector, (ii) 1 vol. dsGF-containing culture and 1 vol. culture containing empty vector or (iii) 1 vol. dsGF-containing culture and 1 vol. γb- or γbΔSKL-containing culture. Total RNA was isolated as described by Verwoerd et al. (1989). High-molecular-mass RNA was precipitated with an equal volume of 10 % PEG8000/1 M NaCl. For analysis of siRNAs, 5 µg low-molecular-mass RNA fraction was separated in a 15 % polyacrylamide gel containing 8 M urea, transferred to Hybond-N membranes (Amersham Biosciences) and hybridized with a radiolabelled, negative-sense T7 transcript of the GFP gene. For subcellular fractionation, leaf tissue was ground to a fine powder with liquid nitrogen in a buffer containing 400 mM sucrose, 100 mM Tris/HCl (pH 7·5), 10 mM β-mercaptoethanol, 2 mM PMSF, 1 mM EDTA, 1 mM EGTA and 2 µg aprotinin ml¹. The slurry was filtered through two layers of Miracloth (Calbiochem). The filtrate was taken for Western blot analysis as a total fraction and the residual portion of the filtrate was centrifuged at 30 000 g for 30 min to yield pellet (P30) and supernatant (S30) fractions. GFPC3 fluorescence in infiltrated leaves was monitored under a hand-held long-wave UV lamp (UVL-56; UVP). Particle bombardment of N. benthamiana leaves was performed by using the flying-disc method with the PDS-1000 system (Bio-Rad), as described previously (Morozov et al., 1997). Cells expressing fusions of fluorescent proteins were imaged with a Leica TCS SP2 system. GFP was visualized with an argon ion laser at 488 nm and an acquisition window of 500530 nm. For imaging of co-expressed yellow fluorescent protein (YFP) and GFP constructs, argon ion laser-excitation lines (488 nm for GFP and 514 nm for YFP) were used alternately. Accordingly, the fluorescence of GFP and YFP was detected alternately by using the switching between lines' option of the confocal system in the 496510 nm acquisition window for GFP and the 560615 nm window for YFP. In this way, any cross-talk between the GFP and YFP channels was eliminated (Brandizzi et al., 2002).

Immunogold labelling and electron microscopy.
Sections of N. benthamiana leaves were embedded in Epon 812 resin after fixation with 2·5 % glutaraldehyde in 0·1 M phosphate buffer, pH 7·0 (Erokhina et al., 2001). Ultrathin sections (90110 nm) were cut with a diamond knife and placed on Formvarcarbon-coated grids. Grids were preblocked by incubation for 15 min in 0·1 M sodium cacodylate buffer, pH 7·4 (CB), with 1 % BSA, washed with CB and incubated overnight at room temperature with solution containing purified mAbs (5 µg ml¹). After washing, the grids were incubated for 2 h at room temperature with goat anti-mouse IgG conjugated to 15 nm gold beads (Biocell) and washed with distilled water. Finally, grids were stained with 1 % uranyl acetate for 30 min and examined under a LEO EM-906 electron microscope.

Subcellular localization of PSLV γb and its SKL mutant
Previous sequence analysis revealed that the N-terminal region of hordeiviral γb proteins has sequence similarity to the N terminus of a cysteine-rich protein, P15, encoded by Peanut clump virus (PCV, genus Pecluvirus) (Savenkov et al., 1998). The PCV P15 C-terminal tripeptide SKL was shown to direct the protein to peroxisome-like structures (Dunoyer et al., 2002). SKL is the C-terminal tripeptide in the γb proteins of PSLV and Lychnis ringspot virus (Agranovsky et al., 1992; Savenkov et al., 1998). Interestingly, in four sequenced BSMV strains, only two (the Type and Argentina Mild strains) contain SKL at the C termini of the encoded γb proteins (Gustafson et al., 1987; Kozlov et al., 1989). In the ND18 strain, the γb protein lacks the C-terminal Leu residue and therefore ends with SK, whereas in the CV17 strain, the tripeptide SKL is replaced with SEF (Gustafson et al., 1987; Edwards, 1995). In this paper, we analysed the role of the SKL tripeptide of the PSLV γb protein.

For protein immunodetection, mAbs against PSLV γb were raised. In preliminary experiments, 11 mAbs showed a strong positive reaction with Escherichia coli-expressed γb, which was used for mice immunization (data not shown). To test the reaction of mAbs with PSLV γb that accumulated in plants during viral infection, we inoculated N. benthamiana plants with BγP, a previously described hybrid hordeivirus representing BSMV in which the native γb gene has been replaced by that of PSLV (Yelina et al., 2002). Western blotting of BγP-infected tissues with five mAbs confirmed their ability to recognize γb in plant extracts (Fig. 1a and data not shown); three mAbs (3D5, 2B2 and 1D3) were selected for further experiments. In plants, PSLV γb was detected not only as a monomer, but also as a dimer and as oligomers (Fig. 1a), which are presumably due to a putative coiled-coil region that is found in the PSLV γb sequence (Bragg & Jackson, 2004). Such a predicted coiled-coil region in BSMV γb was shown to be responsible for protein self-interactions (Bragg & Jackson, 2004).

(50K): Fig. 1. Immunodetection of PSLV γb with mAb 3D5 in plant extracts. (a) Western blot analysis of N. benthamiana leaves inoculated with BγP or mock-inoculated. Samples were taken at 6 d.p.i. Positions of molecular mass markers are indicated. (b) Subcellular fractionation of extracts from leaves agroinfiltrated with γb and γbΔSKL constructs. Lanes loaded with fractions (S30 and P30) or total protein are indicated below the gel.

To analyse the potential role of the C-terminal tripeptide SKL, a derivative of the γb gene, designated γbΔSKL, was constructed to encode a protein with its five C-terminal amino acid residues truncated. Cultures of A. tumefaciens harbouring binary vectors with γb- and γbΔSKL-expression cassettes were infiltrated into N. benthamiana leaves. Western blot analysis of agroinfiltrated leaf patches revealed that the mutant had an altered electrophoretic mobility and migrated in the gel as two major bands (Fig. 1b). In contrast to γb, the dimers were barely detected for γbΔSKL (Fig. 1b). One can speculate that the deletion of SKL, which is located close to the predicted coil-coiled region that is presumably responsible for protein oligomerization (Bragg & Jackson, 2004), reduced the oligomer stability.

Crude fractionation of tissues expressing γb and γbΔSKL demonstrated that γb was localized mostly in the P30 fraction, which contained cell membranes (Fig. 1b); this was in agreement with the previously reported association of PCV P15 with the P30 fraction (Dunoyer et al., 2002). On the other hand, γbΔSKL was detected mainly in the S30 fraction (Fig. 1b), in accordance with earlier reports showing that γb encoded by the ND18 strain of BSMV (which lacks the C-terminal tripeptide SKL) was found predominantly in the S30 fraction (Donald et al., 1993; Donald & Jackson, 1994).

Subcellular localization of GFP-fused PSLV γb and its SKL mutant
To analyse the subcellular localization of PSLV γb and to verify that its C-terminal tripeptide SKL could serve as a peroxisomal targeting signal, we used particle bombardment-mediated transient expression of GFP-tagged proteins. In these experiments, the expression vectors pRT-GFP-γb and pRT-GFP-γbΔSKL were used. Confocal laser-scanning microscopy of bombarded N. benthamiana epidermal cells revealed that GFP-fused γb was localized to punctate structures (Fig. 2a), resembling those in which GFP-fused PCV 15K was localized in virus-infected protoplasts (Dunoyer et al., 2002). The latter structures were proposed to represent peroxisomes (Dunoyer et al., 2002). Therefore, to resolve the nature of γb-containing structures, we used peroxisome-targeted fluorescent marker proteins.

(65K): Fig. 2. Confocal laser-scanning microscopy of N. benthamiana epidermal cells transiently expressing fluorescent fusion proteins 24 h after particle bombardment with 35S promoter-based vectors. (a) GFPγb. (b) GFPγbΔSKL. (c) Co-expression of GFPγb (left panel) with YFPPTS1 (middle panel); right panel shows an overlap of images in the left and middle panels. (d) Co-expression of GFPγb (left panel) with PTS2YFP (middle panel); right panel shows an overlap of images in the left and middle panels. In (c) and (d), YFP signal (middle panels) was pseudocoloured digitally with red to facilitate interpretation of merged images. Bars, 20 µm.

There are two known peroxisomal targeting signals (PTSs) that direct post-translational transport of peroxisomal matrix proteins, synthesized in the cytosol, to peroxisomes. A non-cleavable signal, PTS1, is located at the C-terminal end of the proteins, whereas PTS2 is contained within a cleavable N-terminal presequence (Hayashi & Nishimura, 2003; Reumann, 2004). Following the cloning scheme described by Mano et al. (2002), we constructed two chimeric genes encoding fusions of YFP to the ten C-terminal residues of pumpkin hydroxypyruvate reductase, which contain PTS1, or to the 48 N-terminal residues of pumpkin citrate synthase, which contain PTS2. It has been shown that these PTSs target reporter proteins to peroxisomes in higher plants (Kato et al., 1996; Hayashi et al., 1997; Mano et al., 2002). The fusions YFPPTS1 and PTS2YFP were co-expressed individually with GFP-fused PSLV γb in N. benthamiana epidermal cells by particle bombardment. In both cases, confocal laser-scanning microscopy revealed perfect colocalization of the GFP and YFP signals in punctate structures (Fig. 2c and d).

In contrast to GFPγb, GFP-fused γbΔSKL was distributed diffusely in the cytoplasm and in the nucleus (Fig. 2b), thus resembling the GFP distribution in plant cells (Reichel et al., 1996). These data show that PSLV γb is directed to peroxisomes and that its C-terminal signal, SKL, is implicated in this targeting. Interestingly, a GFP fusion to the BSMV ND18 strain γb, which lacks the C-terminal tripeptide SKL, was also distributed diffusely in plant cells (Lawrence & Jackson, 2001a, b).

Immunogold studies of PSLV γb in virus-infected cells
To verify whether PSLV γb is localized to peroxisomes in virus-infected cells, we used immunogold electron microscopy of thin sections of N. benthamiana leaves collected 6 days post-inoculation (d.p.i.) with BγP. mAbs 3D5, 2B2 and 1D3 gave similar results, therefore only data obtained with mAb 3D5 will be presented below. In BγP-infected cells, immunogold labelling was found predominantly in oval structures (Fig. 3) that were classified, according to their size, shape and electron density, as peroxisomes (Nishimura et al., 1996; Pastori & del Río, 1997). Gold particles were distributed uniformly over the matrix of a given labelled peroxisome; however, not all observed peroxisomes appeared to be labelled (Fig. 3). Examination of other subcellular structures revealed that labelling of the chloroplasts was comparable to the background labelling in healthy cells (Table 1). On the other hand, low but statistically significant labelling was found in the cytoplasm and nucleus of BγP-infected cells (Table 1, Fig. 3). However, this labelling did not exhibit the consistent localization that is characteristic of proteins associated with specific cytoplasmic or nuclear structures. Additionally, no specific immunolabelling was observed in control sections of BγP-infected leaves treated with mAbs specific to Beet yellows virus methyltransferase and leader papain-like proteinase (Erokhina et al., 2001; Zinovkin et al., 2003) (data not shown).

(106K): Fig. 3. Immunogold labelling and electron microscopy of PSLV γb protein in sections of N. benthamiana leaves infected with the hybrid hordeivirus BγP. Solid arrows point to labelled peroxisomes and open arrows to non-labelled ones. The insert shows a higher magnification of a cell region containing both labelled and non-labelled peroxisomes. CW, Cell wall; CH, chloroplast; V, vacuole. Bars, 1 µm (main image); 0·3 µm (insert).

Table 1. Immunogold labelling of various ultrastructures in BγP-infected and healthy N. benthamiana cells probed with mAb 3D5 to PSLV γb Data are no. gold particles µm2±SD, calculated from 25 examined fields.

Effects of the SKL deletion on infection phenotypes of recombinant potexvirus and hordeivirus
We have previously described PVX-PSγb, a recombinant Potato virus X (PVX) that carries the PSLV γb gene (Yelina et al., 2002). Expression of the γb protein had a dramatic effect on the phenotype of PVX infection, converting its mild systemic symptoms (Fig. 4b) into an extensive necrosis that resulted in rapid death of infected plants (Fig. 4c; Yelina et al., 2002). To analyse the contribution of the γb C-terminal tripeptide SKL to the phenotype of PVX-PSγb, the γbΔSKL gene was subcloned into PVX cDNA and the resulting recombinant, PVX-PSγbΔSKL, was inoculated onto N. benthamiana plants in parallel with the construct PVX-PSγb. Observation of plants at 12 d.p.i. revealed that γb and the γbΔSKL mutant had similar effects on the PVX infection phenotype (Fig. 4a). As a direct comparison of PVX, PVX-PSγb and PVX-PSγbΔSKL accumulation in systemically infected leaves was precluded by the fast leaf necrotization induced by PVX-PSγb and PVX-PSγbΔSKL (see Fig. 4a and c), we analysed the inoculated leaves of infected plants. Western blotting with PVX-specific antiserum revealed similar levels of viral CP for PVX, PVX-PSγb and PVX-PSγbΔSKL (Fig. 4g). Thus, the enhanced virulence of PVX associated with expression of γb or γbΔSKL was not due to enhanced virus accumulation. Moreover, the C-terminal tripeptide SKL did not contribute the to γb-induced severe phenotype of the recombinant PVX.

(94K): Fig. 4. Effects of γbΔSKL on infection phenotypes of recombinant potexvirus and hordeivirus. (ac) Influence of γb and γbΔSKL on PVX phenotype in N. benthamiana at 12 d.p.i. (b) PVX-infected control plant. PVX-PSγbΔSKL-infected plants (a) and PVX-PSγb-infected plants (c) exhibit necrosis of systemically infected leaves. (df) Deletion of SKL has little effect on the phenotype of recombinant hordeivirus BγP in N. benthamiana. (d) Symptoms of BγP-γbΔSKL on inoculated leaves at 5 d.p.i. Plants infected systemically with BγP (e) and BγP-γbΔSKL (f) were imaged at 20 d.p.i. (g) Western blot analysis of viral CP accumulation in leaves inoculated with PVX, PVX-PSγb and PVX-PSγbΔSKL (10 d.p.i.). (h) Western blot analysis of viral CP accumulation in leaves infected systemically with BγP and BγP-γbΔSKL (20 d.p.i.). M, Mock-inoculated plant.

To analyse possible effects of the SKL deletion on hordeivirus infection, we modified BγP to replace the PSLV γb gene with γbΔSKL. Both BγP and BγP-γbΔSKL displayed yellow lesions on the inoculated N. benthamiana leaves at 5 d.p.i., the phenotype that is characteristic of BγP infection (Fig. 4d). The systemic-infection phenotype of BγP-γbΔSKL was identical (in terms of timing and symptoms) to that of BγP (Fig. 4e and f). Western blot analysis revealed similar levels of viral CP accumulation in leaves of plants infected systemically with BγP and BγP-γbΔSKL (Fig. 4h). These observations suggest that the C-terminal γb tripeptide SKL does not contribute to the hordeivirus infection phenotype.

Ability of the γb protein and its SKL mutant to suppress RNA silencing in an agroinfiltration assay
Recently, BSMV γb was reported to exhibit silencing-suppression activity in an agrobacterium-mediated transient-expression system (Bragg & Jackson, 2004). To compare the abilities of γb and γbΔSKL to suppress RNA silencing, we employed a similar assay. This approach involved simultaneous transient co-expression of a reporter gene with a dsRNA silencing inducer and a candidate silencing suppressor (Johansen & Carrington, 2001). In this study, GFPC3 (Crameri et al., 1996) was used as a reporter. As an inducer of GFPC3-targeted RNA silencing, we used a construct, referred to as dsGF, that contained two copies of the 5'-proximal part of the GFPC3 coding sequence in the sense and antisense orientations separated by a spacer, so that transcription of this construct in plant cells gave rise to an RNA that formed a long RNA duplex.

To test the silencing-suppression activity of PSLV γb, A. tumefaciens harbouring binary vector pLH-γb was mixed with A. tumefaciens containing the binary vectors pLH-GFPC3 and pLH-dsGF. A similar mixture was prepared with agrobacteria harbouring binary vector pLH-γbΔSKL. The mixtures were infiltrated into N. benthamiana leaves. In controls, infiltrations were carried out with either a bacterial culture carrying pLH-GFPC3 or a mixture of cultures carrying pLH-GFPC3 and pLH-dsGF. For simplicity, we shall refer below to each A. tumefaciens culture used in the infiltrations by the name of the gene that it harbours in expression cassettes of the binary vectors. Observations of leaves under a long-wave UV light revealed bright GFP fluorescence in leaf areas infiltrated with GFP, but not with GFP+dsGF (Fig. 5a), confirming that dsRNA efficiently induced silencing of homologous target RNA (Johansen & Carrington, 2001). In leaf areas infiltrated with GFP+dsGF+γb or GFP+dsGF+γbΔSKL, GFP fluorescence was restored to a level similar to that in the areas expressing GFP only (Fig. 5a), suggesting that both PSLV γb and γbΔSKL efficiently suppressed RNA silencing in the agroinfiltration assay.

(54K): Fig. 5. Silencing-suppression activity of PSLV γb and γbΔSKL in an agroinfiltration assay. (a) Infiltrated N. benthamiana leaves imaged under long-wave UV light at 4 d.p.i. Infiltrated constructs are indicated below the images. (bf) Accumulation of GFP and GFP-specific RNAs in the infiltrated areas analysed at 2, 4 and 6 d.p.i. Infiltrated constructs are shown above the panels. (b) Western blot detection of GFP accumulation. (c) Northern blot analysis of GFP mRNA accumulation. (d) Loading control for (c). (e) Detection of GFP-specific siRNA. Positions of marker oligoribonucleotides of 21 and 25 bases are indicated. (f) Loading control for (e).

To confirm the visual observations, we analysed accumulation of GFP and GFP-specific RNAs in infiltrated leaf areas at three time points after infiltration. At 2 and 4 d.p.i., both γb and γbΔSKL restored levels of GFP (Fig. 5b) and prevented degradation of GFP mRNA induced by dsGF (Fig. 5c), confirming that γb and its mutant efficiently suppressed silencing. Interestingly, at both 2 and 4 d.p.i., the GFP mRNA level was higher in the patches infiltrated with GFP+dsGF+γbΔSKL than in those infiltrated with GFP+dsGF+γb (Fig. 5c). Moreover, at 6 d.p.i., GFP mRNA was still detectable in γbΔSKL-containing, but not in γb-containing, infiltrated areas (Fig. 5c). This may indicate that γbΔSKL is more a potent RNA silencing suppressor than γb. Analysis of GFP-specific siRNA revealed that siRNA levels increased between 2 and 6 d.p.i. (Fig. 5e) and that siRNA accumulation correlated with degradation of GFP mRNA (Fig. 5c and e). However, γb, as well as γbΔSKL, had little effect on siRNA accumulation (Fig. 6e). Thus, in the agrobacterium-mediated transient-expression assay, γbΔSKL had a suppression activity similar to, or even higher than, that of native γb. BSMV γb protein has been found to act in vitro as an RNA- and Zn²⁺-binding protein and in planta as a pathogenicity determinant of viral infection and a suppressor of RNA silencing in transient-expression assays (Donald & Jackson, 1994; Donald & Jackson, 1996; Bragg & Jackson, 2004; Bragg et al., 2004). PSLV γb has also been shown to suppress RNA silencing in viral infection (Yelina et al., 2002). In this paper, we studied the subcellular localization of PSLV γb and the relation between its biological activity and localization.

In experiments with transient expression of GFP-fused γb, we found that the protein was localized in punctate structures that were dispersed in transfected cells (Fig. 2a). To resolve the nature of the γb-containing structures, we employed both fluorescent microscopy and immunoelectron microscopy. Transient co-expression of GFPγb with peroxisomal markers, either YFPPTS1 or PTS2YFP, revealed their precise colocalization in punctate bodies (Fig. 2c and d). These data demonstrate that PSLV γb is localized in peroxisomes and are consistent with the proposed peroxisomal localization of a GFP fusion of PCV cysteine-rich protein P15 (Dunoyer et al., 2002), which exhibits sequence similarity to hordeiviral γb genes (Savenkov et al., 1998).

The occurrence of γb in peroxisomes of virus-infected cells was not uniform. Only a subpopulation of these organelles observed by immunoelectron microscopy was found to be labelled by γb-specific mAbs (Fig. 3). On the other hand, in transient-expression experiments, all peroxisomes that contained the peroxisomal marker proteins, either YFPPTS1 or PTS2YFP, also accumulated GFP-fused γb (Fig. 2c and d), showing that GFPγb localized only to peroxisomes that actively import matrix proteins. Presumably, hordeivirus-infected cells contain both functional peroxisomes and import-inactive peroxisomes or peroxisome-like structures. It is not yet known whether healthy plant cells contain such structures in parallel with normal peroxisomes. In animal cells, accumulation of peroxisome-like structures, known as ghost peroxisomes, is associated with deficiencies in peroxisome import (Wilson, 1991; Santos et al., 1992). One can speculate that hordeiviral infections induce changes in peroxisomal metabolism, resulting in the appearance of subpopulations among leaf peroxisomes, as has been described for plant cells under stress conditions (Nishimura et al., 1996; Pastori & del Río, 1997).

In the patch agroinfiltration assay, PSLV γb effectively suppressed silencing of GFP, in agreement with similar experiments with BSMV γb (Bragg & Jackson, 2004), but had little effect on GFP-specific siRNA accumulation (Fig. 5). The latter observation is consistent with the finding that PCV P15, which is structurally similar to γb, is unable to prevent processing of dsRNA into siRNAs (Dunoyer et al., 2004). However, unlike PSLV γb, PCV P15 significantly reduced the siRNA level (Dunoyer et al., 2004). PCV 15K has been proposed to exert its suppression activity at a step downstream of siRNA production, e.g. 15K may block incorporation of siRNA into RISC. This may cause an instability of RISC-excluded siRNA that is manifested by reduced siRNA levels (Dunoyer et al., 2004). Our observations indicate that PSLV γb may also act as a silencing suppressor downstream of siRNA production. On the other hand, as siRNAs accumulate in the presence of γb, one can speculate that γb does not prevent incorporation of siRNA into RISC, but instead may block RISC activity.

To study whether γb peroxisomal localization was related to its function as an RNA-silencing suppressor, we constructed the mutant γbΔSKL, which has a deletion of five C-terminal amino acid residues that include the C-terminal tripeptide SKL. Localization of pecluviral P15 has been shown to depend on the C-terminal tripeptide SKL (Dunoyer et al., 2002), which represents the canonical peroxisome targeting signal PTS1, directing transport of cellular proteins into the matrix of peroxisomes (Olsen, 1998; Subramani et al., 2000; Hayashi & Nishimura, 2003; Reumann, 2004). Therefore, we envisaged blockage of γbΔSKL transport to peroxisomes. To verify this prediction, we carried out a crude subcellular fractionation of N. benthamiana leaf tissues expressing either wild-type γb or γbΔSKL. Similarly to PCV 15K (Dunoyer et al., 2002), γb was found predominantly in the P30 fraction containing cell membranes, whereas most γbΔSKL was detected in the S30 fraction containing soluble cytoplasmic proteins (Fig. 1b). Accordingly, GFP-fused γbΔSKL showed no association with peroxisomes and was found to localize diffusely in the cytoplasm and nucleus (Fig. 2b). Thus, as expected, localization of γb in cell peroxisomes depended on the C-terminal signal SKL and we used γbΔSKL in further experiments to study the functional significance of such protein targeting.

Firstly, we analysed the effect of γbΔSKL deletion on the infection phenotype of recombinant hordeivirus BγP, which represented a previously described chimera of BSMV carrying the PSLV γb gene in place of the native γb gene (Yelina et al., 2002). For this purpose, the mutant γbΔSKL gene was introduced into BγP to replace the full-length γb gene. Inoculation of N. benthamiana plants revealed that timing and phenotype of BγP-γbΔSKL infection (both local and systemic) were similar to those of BγP (Fig. 4e, f and h). Secondly, we found that the dramatic effect of PSLV γb on the phenotype of PVX infection, manifested as extensive necrosis resulting in fast death of infected plants (Fig. 4c; Yelina et al., 2002), could also be induced by γbΔSKL expressed in the PVX background (Fig. 4a). These observations suggest that deletion of the γb subcellular targeting signal has little effect on the ability of the protein to influence viral infection phenotype.

Direct comparison of γbΔSKL silencing-suppression potential with that of γb revealed that, although the mutant lost its subcellular localization, it fully retained suppression activity in an agroinfiltration assay (Fig. 5). Thus, localization of γb in peroxisomes is dispensable for its ability to suppress RNA silencing and is thus consistent with the general view that RNA silencing is a cytoplasmic process (Voinnet, 2001; Moissiard & Voinnet, 2004) and argues in favour of the suggestion that the cytoplasm, not peroxisomes, is the site of γb silencing-suppression activity. This suggestion is supported by an earlier finding showing that the peroxisome localization motif is not required for silencing suppression mediated by pecluviral P15 (Dunoyer et al., 2002). However, when the SKL motif in P15 was deleted, local and systemic transport of PCV was suppressed dramatically (Dunoyer et al., 2002), whereas γbΔSKL supported hordeivirus movement as effectively as the native protein (Fig. 4e, f and h).

Interestingly, the tobraviral cysteine-rich protein shows sequence similarity to hordeiviral γb, but lacks the C-terminal SKL tripeptide (Savenkov et al., 1998). However, hordeiviral γb can substitute functionally for the tobraviral protein (Liu et al., 2002), supporting the hypothesis that peroxisomes are not the site of γb action. This conclusion is supported by earlier observations that, among four BSMV strains, only two have the SKL tripeptide at the γb C terminus (Gustafson et al., 1987; Kozlov et al., 1989; Edwards, 1995). One can speculate that peroxisomes contain a sequestered excess of PSLV γb that is produced during viral infection, whereas the functional form of the protein can exist in cells in low amounts and/or temporarily. Depositions of presumably excessive amounts of a silencing suppressor have been reported for PVX p25, which is implicated in both silencing suppression and viral movement. Immunoelectron microscopy revealed that p25 localized in specific cytoplasmic inclusions consisting of filamentous material (Davies et al., 1993), but never in the cytoplasm, plasmodesmata or other sites where it would be expected to perform its suppression and movement functions.

At least one viral silencing suppressor is known to accumulate only to very limited levels during infection. Indeed, high levels of the suppressor P0 were shown to be unfavourable for Beet western yellows virus (BWYV, genus Polerovirus) and P0 synthesis in BWYV infection is downregulated to undetectable amounts by conservation of the non-optimal context of the translation initiation codon of the P0 gene (Pfeffer et al., 2002). Sequestering of γb in peroxisomes might be another way to reduce the effective concentration of silencing suppressor at the sites of its action. It remains to be investigated whether small amounts of PSLV γb detected in the cytoplasm and nucleus (Table 1, Fig. 4) represent the functional form of the protein.

This work was supported in part by INTAS (grant 01-2379), the Royal Swedish Academy of Sciences, Russian Foundation for Basic Research (RFBR) grants 04-04-49356 and 04-04-49126-a, the GermanRussian Inter-Governmental Program for Cooperation in Agricultural Sciences and a Grant of the President of the Russian Federation (MD-130.2003.04).