Transgenic O. sativa L. (O. s.) spp. japonica cv. Nipponbare containing the uidA (gusA) transgene from pCAMBIA 1301 (GenBank accession No. AF234297) and carrying a single copy of T-DNA 26 (C. Sallaud, unpublished data) were used. Transgenic line L4 constitutively expresses the uidA gene and the transgenic line L10 is gus-silenced (Figure 1). L10 expressed a PTGS phenotype for expression of the uidA gene on the basis of GUS staining in leaves or roots. With gus-specific siRNA detection, we correlated the L10-phenotype to PTGS and demonstrated its stability over the lifetime of vegetative plants from germination (Figure 1). L10 was used as a model in our silencing suppression studies during host-pathogen interactions. Plants were grown and maintained in controlled conditions in a confined greenhouse (CGG consent for GMO culture n° 3576) under 12 h light at 28°C, 12 h dark at 24°C and at a relative humidity of 70%. As a control, wild-type O. sativa spp. japonica cv. Taipei (Tai.) was used.

We used 10 fully sequenced RYMV isolates belonging to different serogroups and representative of viral diversity in Africa (i.e. with 9.9% maximum diversity) 1923, these were; BF1, CI63, CI4, Ni1, Ni2, Mg1, Tz3, Tz5, Tz8 and Tz11. There was a second set of 10 isolates belonging to serogroup 2, showing variable pathogenicity assessed through symptom intensity, but low sequence diversity (2%). These were; CI110, 111, 112, 113, 114, 115, 116, 121, 129, and 138 27 (Figure 2). They were independently propagated in the susceptible cultivar O. sativa spp. indica cv. IR64.

Two weeks after seeding, plants (2 leaves) were mechanically inoculated as described elsewhere 28 with diluted sap (1/10 w/v). Each isolate was independently inoculated into 20 plants that were grown separately to limit cross contamination. Viral inoculum were adjusted to a standard concentration, as estimated by DAS-ELISA assays, to prevent variation in silencing suppression due to differences in initial virus delivery. In order to avoid bias due to inoculation heterogeneity, leaf samples were collected from each inoculated plant at 35 or 40 dpi and separated into two independent samples before being ground for proteins (fluorimetry, ELISA and western) and RNA extractions.

P1 sequences with UTR (from 1 to 709 nt) 17 from five isolates representative of the diversity of P1 with a divergence up to 17.8% were selected. These isolates are representative of the diversity of the other parts of the genome as phylogenetic reconstruction from the P1, the CP and the full genome are congruent. They were cloned into the CaMV 35S constitutive promoter and terminator, that were named sTz3; sMg1; sCI63; sBF1; sTz8. PCR fragments of P1 sequences (underlined) were amplified with: forward GAATTCAAGCTT

GACAATTGAAGCTAGGAAAGGAGC

CGCGGCC GCTATCAA

Total RNA was extracted from rice leaves with TRIzol reagent (Invitrogen™) according to the manufacturers' instructions. Analysis of low molecular weight RNA was performed on 15 μg total RNA as described previously 29. Analysis of mRNA was performed with NorthernMax^®-Gly (Ambion^®) as recommended by the manufacturer. Radio-labelling of the probe and hybridization was performed as done in ref. 30. A 808 bp PCR amplified fragment (808 to 1616 bp) of GusA was used for the specific probe preparation. Radio-labelled signals were detected either by autoradiography or by scanning with Typhoon (Amersham). RNA band intensities were quantified using Image Quant software (Molecular Dynamics).

Soluble proteins were extracted from rice leaves in a sodium phosphate buffer (50 mM NaHPO₄, 10 mM Na₂EDTA, 1 mg.ml^-1 N-Laurylsarcosine, 0.1% v/v Triton ×100) by two successive centrifugations of 20000 g at 4°C. The Bradford colorimetric method (Coomassie protein assay kit, Pierce) 31 was used for protein quantification. Twenty micrograms of protein were resolved by SDS-PAGE and transferred by electroblotting onto a nitrocellulose membrane (Trans-Blot^® Transfer Medium, Biorad). A diluted rabbit polyclonal antibody directed against the P1 protein (1/500) was used and revealed by a second (1/40000), horseradish peroxidase-conjugate (Pierce) through Supersignal^® West-Pico chemiluminescence substrate. Blots were exposed to CL-XPosure™ film (Pierce).

The virus concentration was evaluated by DAS-ELISA as described previously 32. DAS-ELISA was performed with diluted (1/1000 v/v) polyclonal antiserum against an isolate from Madagascar (RYMV-Mg). Positive reactions were detected after incubation with alkaline phophatase-conjugated polyclonal antibody to RYMV and substrate, with absorbance read using a Multiskan fluorimeter (Labsystem).

Histochemical staining for GUS activity and GUS assays were performed according to Jefferson 33 and involved pieces of leaves randomly collected from two to five plants at 1 and 2 dpi (inoculated leaves), and 7, 14, 35 and 40 dpi (systemically infected leaves).

As described previously 34, proteins were incubated with 1 mM of 4-methylumbelliferyl-β-D-glucuronide (MUG, Sigma^®). Fluorescence was measured at 15 min intervals for 2 h with a Fluoroskan fluorimeter (Labsystem), with a 365 nm excitation filter and a 455 nm emission filter. Calibration was performed with quantification of fluorescence of soluble 4-methylumbelliferone (MU, Sigma^®).

We investigated the diversity of RYMV efficiency in silencing suppression and used a transgenic gus-silenced rice line (L10). As seen in Figure 1, silencing of the gus transgene, leads to an absence of GUS activity, and is accompanied with a strong accumulation of gus specific siRNA. This accumulation is maintained during the life-cycle of rice plants (Figure 1). Moreover, infection with the RYMV virus leads to a restoration of GUS activity concomitant with a decrease of gus-siRNA accumulation. Assuming that silencing suppression activity of RYMV involves restoration of the GUS phenotype and a decrease of steady-state levels of gus-siRNA, we quantified silencing suppression efficiency by evaluating the decrease in steady state levels of siRNA. We separately inoculated 10 RYMV isolates that differed in their coat protein (CP) (Figure 2). From 1 dpi, we revealed GUS reversion for all isolates by observing the blue histochemical staining patterns in both L10-infected leaves and positive controls (i.e. transgenic L4 lines that constitutively expresses the uidA transgene) whereas in non-inoculated (NI) and mock-inoculated (BI) L10 leaves, GUS activity was undetectable (Figure 3A) indicating that the reversion of GUS activity in L10-inoculated plants was only due to RYMV and not to mechanical stress caused by the inoculation.

Because histochemical staining of GUS is not a quantitative method the reversion of uidA gene expression in L10 was assayed by fluorimetric quantification of GUS activity, in systemically infected leaves. We found that RYMV isolates differed in their efficiency to reverse the constitutive silencing, as illustrated by the heterogeneity of restored GUS activity (Figure 3B). In each serogroup (Figure 3B), we detected a contrasting pattern of reactivation for the uidA transgene. In S1, the isolates CI4, Ni1 and Ni2 exhibited distinct effects on silencing suppression (Figure 3B, lanes 3–5). Similarly, the S4 Mg1 isolate weakly suppressed silencing, whereas a strong effect was obtained for the Tz8 isolate (Figure 3B, lanes 8, 10). Finally, strong suppression leading to higher GUS activity was detected for the Tz3 and Tz11 isolates belonging to the S5 serogroup (Figure 3B, lanes 11–12). This result showed that the ability to suppress PTGS was not correlated with the viral phylogeny.

Moreover, we found that viral content was not correlated with GUS activity, as illustrated by isolates belonging to the same serotype Ni2 and CI4, or, CI63 and BF1 isolates (Figure 3C, lanes 3 and 4 or 4 and 5). We also noted that the diversity of silencing suppression displayed by these RYMV isolates was not correlated with viral pathogenicity (data not shown). Indeed, the most pathogenic isolate, BF1, was not the strongest in suppressing PTGS.

Taken together, these results highlighted the prevalence of silencing suppression within the species and showed that gus-silenced gene reversion was dependent on viral isolate.

As plants infected with different RYMV isolates exhibited variable GUS activity, we investigated the diversity of this restoration at the serogroup level. Thus we considered isolates from the Ivory Coast belonging to the least variable RYMV serogroup S2 (diversity around 2% based on CP sequence) (Figure 2). Ten representative isolates of S2 were chosen for homogeneous biological characters (same collection date and geographical localisation) and were inoculated into L10 plants.

Silencing suppression showed by the restoration of GUS activity was followed by histochemical staining of GUS at 1, 2, 7, 15, 17, 21, 27, 31, and 40 dpi. At 40 dpi, steady-state siRNA levels were monitored with Northern blots. Gus-specific siRNA were not detected in systemically infected leaves in contrast to NI or BI L10 leaves (Figure 4A). This revealed that uidA gene silencing had been reverted by all S2 isolates tested. To analyze more closely the qualitative effect of the different isolates, we compared GUS activity levels measured by fluorimetry (Figure 4B) to viral amounts measured by ELISA (Table 1) in systematically infected leaves. An analysis of covariance revealed that at 40 dpi a large part of the variability in silencing suppression was due to an isolate effect, although part of the reversion was due to an effect of virus content (Table 1). We thus revealed highly heterogeneous responses in GUS activity reversion following viral infection (Figure 4B). Isolate CI110, is considered to be strong suppressor of constitutive silencing in the iudA gene, (Figure 4B). The weaker suppressors were isolates CI113 and 114, with less than 20% maximal activity restoration (Figure 4B). Western blot analyses showed that the efficiency of silencing suppression was not correlated with RYMV-P1 protein suppressor accumulation in leaves (Figure 4C). As all isolates belong to the same serogroup, this quantitative variation of P1 accumulation should not be linked to polyclonal antibody affinity. In addition, sequences analysis of P1 proteins found no obvious correlation between the efficiency of silencing suppression and genetic diversity of the P1 protein (data not shown) suggesting that silencing suppression when analyzed at the virion scale is more complex than previously thought. As noted above, the efficiency of silencing suppression is not correlated to symptoms expressed by these isolates (data not shown).

Taken together our results at the virion scale demonstrate that the accumulation and genetic diversity of the P1 suppressor protein were not correlated to heterogeneity in the reversion of GUS activity that we observed.

As the effect of RYMV on uidA silencing suppression was not linked to viral diversity, pathogenicity or P1 accumulation, we attempted to assess the particular effect of the P1 protein identified as the RYMV silencing suppressor protein 9. Different P1s representative of the viral diversity were cloned under control of the CaMV 35S promoter. This was done in order to determine whether the sequence diversity of this suppressor reflects functional diversity and to accurately determine the qualitative effect of the P1 protein on silencing suppression.

We demonstrated a lack of effect due to biolistic delivery and the expression vector by using empty 35S (Figure 5A and 5B lane 5). At 1 day post delivery (dpd), we revealed the reversion of GUS activity, except for sTz8 for which GUS activity was only observed at 2 dpd (Figure 5A). We thus demonstrated for the first time in a natural host, that all P1s from different RYMV isolates were silencing suppressors. In addition, we observed the reversion of GUS activity over the entire leaf, even where the P1 protein had not been delivered, suggesting strong movement of this protein (Figure 5A).

To further examine whether P1 acts differently in relation to variation in its sequence, we evaluated the steady-state level of gus-specific siRNA by Northern blot. Every P1 induced a decrease in the steady-state level of gus-siRNA, as compared to siRNA detected in L10 controls (ND, BD and 35S) (Figure 5B). Moreover, as shown in Figure 5B, the different P1s exhibited contrasting activity in silencing suppression, as revealed by the different siRNA patterns. These results highlight the qualitative effect of different P1s and that they could be grouped into two classes according to their efficiency of suppression. Indeed, we observed that GUS activity restoration with sCI63, sTz3 and Tz8 was accompanied by a dramatic decrease in siRNA levels. Consequently, we concluded these proteins are highly efficient in transgene-induced silencing suppression in rice. Conversely, in case of sBF1 and sMg1, we observed a slight decrease in siRNA levels (Figure 5B), and they were thus classified as weak silencing suppressors.

Furthermore, as previously demonstrated at the virion scale, the efficiency of these silencing suppressors is not correlated with viral phylogeny. Indeed, according to RYMV evolutionary history 19, P1s from isolates belonging to the same phylogenetic group can be either strong (sCI63) or weak silencing suppressors (sBF1) (Figure 2 and 5B). In addition, P1s from distant isolates were similar in their silencing suppression efficiencies, as illustrated by the weak suppressors sBF1 and sMg1 (Figure 2 and 5B).

As we found a qualitative effect of the different P1s, and as it has been demonstrated that silencing suppression and cell-to cell movement of Potato potexvirus X (PVX) are linked 35, we investigated the involvement of cell-to-cell movement of P1 on the qualitative variation in silencing suppression (Figure 6A). We carried out an experiment to analyze P1 protein accumulation at the site of delivery (del) and in distal (dis) areas of L10-inoculated leaves. To do so, we co-bombarded 20 pieces of L10 leaves with every sP1 construct described above and associated with a 35S-GFP plasmid. Under UV illumination at 2 dpd, we were able to discriminate the delivery area from the distal area by GFP fluorescence. We could detect green areas corresponding to the delivery zone (del) and red areas due to chlorophyll fluorescence corresponding to the distal zone (dis) (Figure 6A). To limit any contamination by P1 outside from the delivery zone, we enlarge the delivery zone by an additional 1.5 cm which leads, in addition to a 6 cm flight distance, to an incompatible zone to receive any particles.

We compared the steady-state level of gus-siRNA and mRNA, in the delivery and distal areas for each sP1 protein previously selected for a range of variation in their efficiency of silencing suppression (Figure 6B). For the sTz3 protein, accumulation of gus-siRNA and mRNA were similar in del and dis areas (lanes 4–5). For the sCI63 protein, we observed a weak accumulation in the both areas for gus-siRNA, and strong accumulation mainly in del area for gus-mRNA (lanes 8–9). For the sMg1 protein, we observed a contrasting situation between the del and dis areas; lower accumulation of gus-si-RNA and accumulation of mRNA in the del area, and an absence of gus-mRNA in the dis area (Figure 6B, lanes 6–7). Then, for the same samples, we investigated the presence of P1 in the del and dis areas by western blot. Similar P1 protein accumulation was revealed in del and dis zones for sTz3 and sMg1 (Figure 6C, lanes 4–7), and only in the dis zone for sCI63 (Figure 6C, lanes 8–9). Although we cannot rule out the bombardment of P1 outside of GFP area, the strong accumulation of P1 detected in distal areas seems to favour the idea that of there being cell-to-cell movement from delivery to distal area of P1 proteins in rice leaves. This would help clarify our previous observation of GUS activity being detected across entire leaves (see Figure 5A, lanes 2 and 6 compared to 8 or 9). Furthermore, we observed differences in P1 accumulation as shown for sCI63 (only detected in the dis area, Figure 6C lanes 8–9), where a weak accumulation of gus-siRNA was correlated with gus-mRNA expression, as compared to levels observed for sTz3 and sMg1. It therefore seems likely for sCI63 that the faster movement of this P1 protein is associated with its high efficiency in silencing suppression.

In contrast, for sMg1 we observed a different situation with the accumulation of P1 in the dis area which was not correlated with a decrease in gus-siRNA and the expression of gus-mRNA (Fig 6B and 6C, lanes 6–7). For sMg1, although movement of P1 occurred its lack of effects on gus-siRNA was likely caused by its weaker efficiency in silencing suppression.

As we demonstrated major differences in the silencing suppression efficiencies of different P1 proteins (Tz8, BF1, CI63, Mg1 and Tz3), we attempted to assign amino acid diversity to functional effects with ClustalW alignments 36 (Figure 7A).

First, we identified the putative eukaryotic Zn-finger motif (C64(x)₂C67(x)₂₄C92(x)₂C95) which is conserved across the RYMV phylogeny (Figure 7A). This kind of motif has been reported to be crucial for silencing suppression activity 37. We introduced two independent single mutations in the Tz3 Zn-finger with PCR amplification in cysteine 64 and 95 (Cystein changed by Serine, C64S et C95S). Secondly, when comparing sequences of the strongest silencing suppressor (sTz3) and the weaker silencing suppressor (sMg1) we detected two amino acids specific to sTz3 (i.e. N19 and F88) (Figure 7A). We used PCR amplification to introduce two independent single mutations in the sTz3 sequence at asparagine 19 (Asparagine changed by Isoleucine, N19I) and phenylalanine 88 (Phenylalanine changed by Tyrosine, F88Y). The efficacy of these mutants to abolish silencing of the uidA transgene in L10 was tested following biolistic delivery. The effects of these mutations were compared with wild-type sTz3 and sMg1 silencing suppressor proteins.

At the histochemical and molecular levels (i.e. gus-specific si- and mRNA detection), we demonstrated that none of the four mutations completely abolished the suppression effect of sTz3. Nevertheless, the single amino acid mutation C95S impaired the ability of P1 to suppress uidA silencing in the distal part of leaves (high level of gus-siRNA), suggesting that this mutation affects the cell-to-cell movement of P1 (Figure 7B, lanes 12–13). Conversely, mutations F88Y or C64S both decreased silencing efficiency, as shown by the steady-state level of gus-siRNA in the del and dis zones being higher than observed with sTz3 (Figure 7B, lanes 4–7). Finally, mutation N19I did not affect the silencing suppression efficiency of the P1 protein relative to sTz3, as shown by similar pattern of gus-siRNA and gus-mRNA (Figure 7B, lanes 8–9). Together these results demonstrate that mutations highly influence the quantitative (Figure 7B, lanes 12–13) and qualitative (Figure 7B, lanes 6, 8, 10) expression of P1 activity. Finally, we found that the cystein 95 belonging to the putative zing finger is essential for the cell-to-cell movement of P1 and cystein 64 and phenylalanine 88 were involved in the efficiency of silencing suppression of the P1-Tz3 protein.