Double-Stranded RNA Binding May Be a General Plant RNA Viral Strategy To Suppress RNA Silencing

ABSTRACT In plants, RNA silencing (RNA interference) is an efficient antiviral system, and therefore successful virus infection requires suppression of silencing. Although many viral silencing suppressors have been identified, the molecular basis of silencing suppression is poorly understood. It is proposed that various suppressors inhibit RNA silencing by targeting different steps. However, as double-stranded RNAs (dsRNAs) play key roles in silencing, it was speculated that dsRNA binding might be a general silencing suppression strategy. Indeed, it was shown that the related aureusvirus P14 and tombusvirus P19 suppressors are dsRNA-binding proteins. Interestingly, P14 is a size-independent dsRNA-binding protein, while P19 binds only 21-nucleotide ds-sRNAs (small dsRNAs having 2-nucleotide 3′ overhangs), the specificity determinant of the silencing system. Much evidence supports the idea that P19 inhibits silencing by sequestering silencing-generated viral ds-sRNAs. In this study we wanted to test the hypothesis that dsRNA binding is a general silencing suppression strategy. Here we show that many plant viral silencing suppressors bind dsRNAs. Beet yellows virus Peanut P21, clump virus P15, Barley stripe mosaic virus γB, and Tobacco etch virus HC-Pro, like P19, bind ds-sRNAs size-selectively, while Turnip crinkle virus CP is a size-independent dsRNA-binding protein, which binds long dsRNAs as well as ds-sRNAs. We propose that size-selective ds-sRNA-binding suppressors inhibit silencing by sequestering viral ds-sRNAs, whereas size-independent dsRNA-binding suppressors inactivate silencing by sequestering long dsRNA precursors of viral sRNAs and/or by binding ds-sRNAs. The findings that many unrelated silencing suppressors bind dsRNA suggest that dsRNA binding is a general silencing suppression strategy which has evolved independently many times.

RNA silencing (termed RNA interference [RNAi] in animals) is an RNA-based eukaryotic gene regulatory system that plays essential roles in many biological processes. RNA silencing is induced by accumulation of double-stranded RNAs (dsRNAs). dsRNAs are first processed by an RNase III-like nuclease called DICER (in plants it is termed DICER-LIKE [DCL]) into (21-to 25-nucleotide [nt]) small dsRNAs (ds-sRNAs) having 2-nt 3Ј overhangs, and then these sRNAs incorporate into different silencing effector complexes. In the active effector complexes sRNAs are present as single-stranded molecules, which guide these complexes to the complementary nucleic acids for suppression (2,3,22,61).
In plants, different dsRNA precursors are processed by distinct DCLs into functionally different short (21-to 22-nt) and long (23-to 25-nt) sRNAs (24,25,67). Short sRNAs guide a multicomponent nuclease (RNA-induced silencing complex [RISC]) to homologous mRNAs for suppression. RISC cleaves targeted mRNA in the case of (near) perfect base pairing between mRNA and guide RNA. When the guide RNA is only partially complementary to the mRNA, RISC mediates translational repression. Short sRNAs could also provide (directly or indirectly) sequence specificity for plant-encoded RNAdependent RNA polymerase (RdRP) which transforms homologous mRNAs into dsRNAs, thus amplifying silencing (3,42,65). Long sRNAs play a role in transcriptional silencing by directing the histone and DNA methylation of homologous DNA (66,72).
RNA silencing plays important antiviral roles in plants and animals (15). In plants, double-stranded replicative intermediates of RNA viruses or strongly structured segments of viral mRNAs (40) are processed by DCL2 and perhaps by other DCL enzymes into viral sRNAs (18,24,67), short 21-nt dsRNAs having 2-nt 3Ј overhangs (also called viral short interfering RNAs [siRNAs]). Viral sRNAs incorporate and target RISC to viral mRNAs for suppression, and thereby silencing could reduce pathogen levels in the infected cells. Virus-induced silencing can also act as a systemic defense system. Viral sRNAs are supposed to spread 10 to 15 cell layers and activate silencing in still noninvaded neighboring cells, limiting the extent of virus invasion (27).
To counteract RNA silencing most plant viruses express silencing suppressor proteins (44,48,53,63,65). Since silencing suppressors operate differently in widely used silencing inhibition assays it is believed that viral suppressors target different steps of the silencing response (64). However, as dsRNAs play key roles in silencing, it was speculated that binding of a dsRNA component of the silencing machinery could be a frequently used viral suppression strategy (53). In theory, dsRNA-binding silencing suppressors could target silencing response at two different steps, by binding and sequestering the silencing inducer long viral dsRNAs and/or by binding and sequestering viral ds-sRNAs.
Recent results suggest that two unrelated viral suppressors (tombusvirus P19 and closterovirus P21) inhibit silencing by binding viral ds-sRNAs (11,19,31,54). It has been shown that P19 sequesters viral sRNAs in infected plants, and thus it depletes the specificity determinant of antiviral silencing effector complexes (31). As P21 also binds ds-sRNAs in plant cells (11), it is likely that sequestering of ds-sRNAs is a silencing suppression strategy that has evolved at least twice independently. Moreover, the P19-related aureusvirus P14 silencing suppressor is also a dsRNA-binding protein (39).
In this study we wanted to test the hypothesis that dsRNA binding is a general silencing suppression strategy of plant RNA viruses. Here we show that many viral suppressors can bind dsRNAs. CP of Turnip crinkle virus (TCV), a carmovirus, like P14, binds dsRNA without obvious size selection. By contrast, P15 of Peanut clump virus (PCV), a pecluvirus, ␥B of Barley stripe mosaic virus (BSMV), a hordeivirus, P21 of Beet yellows virus (BYV), a closterovirus, and HC-Pro of Tobacco etch virus (TEV), a potyvirus, like P19, are size-specific ds-sRNA-binding silencing suppressors. These proteins efficiently form complexes with 21-nt ds-sRNA but fail to bind long dsRNA. We have also demonstrated that the structural basis of size-selective ds-sRNA binding is different for P19 and for PCV P15, BSMV ␥B, and TEV HC-Pro size-specific dsRNAbinding suppressors.
Our data strongly suggest that dsRNA binding is a general plant viral silencing suppression strategy which has evolved independently many times.
The P25 fragment was generated with forward (5Ј-TTCCATGGATATTCTC ATCAGTAGTTTG) and reverse (5Ј-CTATGGCCCTGCGCGGACATATG) primers, and then the PCR fragment was cloned into EcoRV-cleaved pBluescript KS. The forward orientation was selected, and the BamHI-XhoI fragment was cloned into BamHI-and SalI-digested BIN61S. The P15 PCR fragment was generated with the P15 forward primer (5Ј-ATCGGTACCATGCCTAAGTCG GAGTTCTTTCG) containing a KpnI site (underlined) and with the P15 reverse primer (5Ј-ATCGGATCCTTACAGTTTAGAACGAAGATT) containing a BamHI site, and then the P15 PCR product was cloned into BamHI-KpnIdigested pBluescript KS (KS-P15). Finally, the P15 insert was moved as a KpnI-XbaI fragment into BIN61S.
The CP fragment was generated (primers 5Ј-CTGGATCCATGGAAAATGA TCCTAGAG and 5Ј-CAGTCGACCTAAATTCTGAGTGCTTGC) and cloned into the SmaI site of BIN61S (BIN61S CP). Note that the underlined restriction sites (BamHI and SalI) were not used in this cloning. The orientation was determined by orientation-dependent PCR. BSMV ␥B was PCR amplified with forward (5Ј-CATTGGATCCATGGCTACTTTCTCTTGTG) and reverse (5Ј-TCACTCTAGATTACAACTTAGAAACGGAAGAAG) primers carrying a BamHI and an XbaI site, respectively. The PCR fragment was cleaved with BamHI and XbaI and cloned into BIN61S.
Plasmid pKS TAV 2b (K. Salanki, unpublished) was partially digested with KpnI and XbaI, and then the expected TAV 2b fragment was purified and cloned into BIN61S. Binary suppressor constructs were sequenced and used for agroinfiltration assays. PVXP14 and PVXP19 were previously described (39,57). To create N-terminally c-Myc-tagged binary constructs, P15, TCV CP, and BSMV ␥B suppressors were cloned into BINmyc, a derivative of BIN61S in which the multiple cloning site was replaced with the c-Myc fusion cloning sites from pMESHI (21). To create Myc-P15, the BamHI-digested P15 fragment which was amplified from KS-P15 with P15BamHI forward (5Ј-CATGGATCCATGCCTA AGTCGGAGTTCTTTCG) and P15 reverse primers, was cloned into BINmyc. Myc-CP was obtained by cloning the BamHI fragment from the BIN61S CP into BINmyc.
To obtain Myc-␥B, the BIN61S ␥B construct was cleaved with XbaI, filled in, and then digested with BamHI. The ␥B fragment was cloned into BamHI-and SmaI-digested vector GEX2T (GEX ␥B). The BamHI-EcoRI fragment was isolated from GEX ␥B and cloned into BINmyc. C-terminally hemagglutinin (HA)-tagged P21 was described previously (11). His-HC-TEV was created by PCR amplifying the His-tagged TEV HC-Pro fragment with forward (5Ј-AAC AATGCATCACCATCACCATCACGCG) and reverse (5Ј-TCATCCAACATT GTAAGTTTTCATTTC) primers from His-HC-Pro TEV-infected plants (6), and then it was cloned into SmaI-cleaved pBluescript KS. The proper orientation was determined, and then the BamHI-SalI fragment was moved into Bin61S.
RNA gel blot analysis. The same total RNA extract was used for high-and low-molecular-weight RNA gel blot analyses. RNA extraction and RNA gel blot analysis were carried out as described (39). PCR fragments labeled with the random priming method were used for Northern analyses of high-molecularweight RNAs. Radioactively labeled in vitro transcripts corresponding to the antisense strand of GFP were used as probes for Northern analyses of lowmolecular-weight RNAs.
Preparation of viral and suppressor extracts. To prepare protein extract, 0.25 g leaf tissue was ground in 1 ml band shift buffer (83 mM Tris-HCl, pH 7.5, 0.8 mM MgCl 2 , 66 mM KCl, 100 mM NaCl, and 10 mM dithiothreitol), and then this crude extract was centrifuged twice for 15 min at 15,000 ϫ g (39). The supernatant was frozen in aliquots at Ϫ70°C. Viral extracts were prepared from strongly symptomatic leaves after the accumulation of viral RNAs was confirmed by RNA gel blot assays (data not shown).
Preparation of labeled dsRNA probes. Synthetic sRNAs were labeled with T4 polynucleotide kinase, and then the polynucleotide kinase was heat inactivated. To generate ds-sRNAs, unlabeled 5Ј-phosphorylated complementary-strand sR-NAs in twofold molar excess were added to labeled single-stranded sRNAs, and then the mixture of labeled and unlabeled single-stranded sRNAs was heated for 94°C (1 min) and allowed to cool down slowly to room temperature. [␣-32 P]UTPlabeled in vitro RNA transcripts were used as long RNA probes. Labeled transcripts were produced from an M13Forward-M13Reverse pBluescript KS PCR fragment of the T7 and T3 RNA polymerase, respectively. To generate long dsRNAs, a 1:1 mixture of labeled T7 and T3 in vitro transcripts was heated (1 min, 95°C) and annealed (2 h, 37°C). To remove single-stranded overhangs, annealed long dsRNAs were treated with RNase A, extracted with phenolchloroform, and precipitated. Labeled, annealed long dsRNAs were redissolved and diluted.
Gel mobility shift assay. In a binding reaction, labeled dsRNA (0.5 nM) was incubated with extract containing ϳ2 g total protein. Binding reactions, mobility shift assays, and direct competition assays were carried out as described (39,

5748
MÉRAI ET AL. J. VIROL. 54) except that 0.02% Tween 20 was added to the binding buffer; 0.2 g total protein was used for P19, P14, and P21 direct competition assays, while ϳ2 g total protein was used for P15, TEV HC-Pro, and TCV CP competition experiments. Supershift assays were performed by adding 1 g antibody (anti-c-Mycperoxidase, Roche 1814150; anti-His, Amersham 274710-01; anti-HA, Roche 1583816) to the suppressor extracts 30 min before incubation with ds-sRNA probes.

TCV CP binds long dsRNA in vitro and prevents sRNA accumulation in hairpin transcript-induced silencing assays.
Theoretically, plant viruses can interfere with the silencing response by sequestering two different dsRNA components of the silencing system, long dsRNAs and ds-sRNAs. Long-dsRNA-binding proteins might compete with DCLs, thus reducing or delaying the accumulation of ds-sRNAs. Indeed, we have previously shown that PoLV P14 binds long dsRNA in vitro and prevents the accumulation of sRNA in hairpin-induced silencing assays. To test whether other viruses also express long-dsRNA-binding proteins, we prepared crude extracts (viral extracts) from plants which were infected with different well-characterized viruses, and then the long-dsRNAbinding activity of these viral extracts was tested in gel mobility shift assays. Confirming our previous results (39), we could detect long-dsRNA-binding activity in the PoLV extract (Fig.  1A, lane 2) but not in extracts that were isolated from mockinoculated plants or in the CymRSV extract (lanes 1 and 4). As Fig. 1A shows, while BYV, PCV, PVY, PVX, CMV, TMV, TAV, TEV, and BSMV extracts failed to bind long dsRNAs, TCV extract bound long dsRNA efficiently (Fig. 1A, lane 7). These data suggest that TCV expresses or induces the expression of a long-dsRNA-binding protein.
We assumed that the coat protein (CP), the silencing suppressor of TCV (45,60,71), provided the long-dsRNA-binding activity for the TCV extract. If it is true, similar long-dsRNAbinding activity should be detectable in TCV viral extracts and in an extract that is prepared from leaves, in which CP is expressed alone. Therefore, TCV CP was agroinfiltrated into N. benthamiana leaves, and then the long-dsRNA-binding activity of the extract (suppressor extract) prepared from the infiltrated patch was analyzed. As controls, the long-dsRNA- binding activity of suppressor extracts of other viruses (BYV P21, PCV P15, TEV HC-Pro, and BSMV ␥B) was also tested. PoLV P14 and reovirus Sigma3 were used as positive controls. As Fig. 1B shows, in addition to the positive controls, only the TCV CP extract bound long dsRNAs (Fig. 1B, lane 3). Moreover, the binding characteristics of TCV viral and CP suppressor extracts were similar. Both extracts caused accumulation of distinct shifted bands, suggesting that more than one CP molecule could bind to a single dsRNA.
To confirm that TCV CP also binds long dsRNA in vivo, we analyzed the effect of TCV CP on sRNA accumulation in hairpin-induced silencing assays. As a control, the BYV P21 suppressor was used. Coinfiltration of an Agrobacterium sp. that expresses an inverted repeat-GFP construct (GFP-IR) with another Agrobacterium sp. that expresses GFP (GFP-IRϩ35SGFP) prevents transient GFP activity (Fig. 1C, upper  panel), because GFP-IR-derived hairpin transcripts (GFP-ir) are rapidly processed to sRNAs, and then these sRNAs direct early degradation of GFP mRNAs (hairpin-induced silencing) (39). Coinfiltration of GFP-IRϩ35SGFP with the TCV CP or P21 suppressor resulted in strong green fluorescence (data not shown) and led to the accumulation of GFP mRNAs (Fig. 1C), indicating that both suppressors inhibited hairpin-induced silencing.
However, CP and P21 inhibit hairpin-induced silencing differently. In line with previous results (11,43,45), we found that in TCV CP-coinfiltrated samples GFP-ir accumulated to easily detectable levels, while sRNAs could not be detected (Fig. 1C). By contrast, in P21-coinfiltrated leaves GFP-ir transcripts could not be detected, whereas sRNAs were abundant (Fig. 1C). These data suggest that TCV CP binds hairpin transcripts, the long dsRNA precursors of sRNAs. These data are also consistent with previous suggestions that P21 suppresses silencing by binding hairpin-derived ds-sRNAs (11).
To further support the idea that CP directly inhibits hairpinderived sRNA accumulation and that P21 is unable to prevent sRNA generation from hairpin transcripts, we coinfiltrated leaves with GFP-IR and with CP or with P21. As expected in GFP-IR-and CP-coinfiltrated leaves hairpin transcripts were detected, while sRNAs could not be found. By contrast, in GFP-IR-and P21-coinfiltrated samples sRNA accumulated to high levels, while hairpin transcripts could not be detected (Fig. 1D).
We also tested the effect of some other suppressors that were unable to bind long dsRNAs (PCV P15, TEV HC-Pro, and BSMV ␥B) in hairpin-induced silencing assays. We found that these suppressors operated like P21; they inhibited hairpin-induced silencing but failed to prevent the accumulation of hairpin-derived sRNAs (data not shown). These data suggest that the PCV P15, TEV HC-Pro, and BSMV ␥B suppressors, like P21, target hairpin-induced silencing downstream from the sRNA generation step.
Many plant viral silencing suppressors bind 21-nt ds-sRNA. Plant viruses can also inhibit silencing by expressing ds-sRNAbinding proteins. To test if ds-sRNA binding is a general silencing suppression strategy, the 21-nt ds-sRNA-binding activity of different viral extracts was analyzed.
As Fig. 2A shows, ds-sRNA-binding activity was identified in many but not in all viral extracts. Confirming our previous results (39), we could detect 21-nt ds-sRNA-binding activity in CymRSV and PoLV extracts ( Fig. 2A, lanes 2 and 4) but not in extract that was isolated from mock inoculated plants (Fig. 2A,  lane 1). Importantly, 21-nt ds-sRNA-binding activity could be also identified in BYV, PCV, TCV, TMV, TEV, and BSMV extracts ( Fig. 2A, lanes 5, 6, 7, 13, 15 and 16). By contrast, extracts prepared from leaves infected with PVY, PVX, CMV, and TAV failed to bind ds-sRNA ( Fig. 2A, lanes 8, 9, 12, and  14). Different viral extracts caused different mobility shifts, and therefore we concluded that the ds-sRNA-binding proteins were expressed by the viruses rather than by the infected hosts.
An unlikely alternative explanation would be that these suppressors activate host-encoded ds-sRNA-binding proteins in the infected cells and in the infiltrated patches. To distinguish between these two possibilities, antibody supershift assays were carried out with extracts isolated from leaves, in which epitopetagged suppressors were expressed. Extracts were prepared from leaves that were infiltrated with agrobacteria expressing the HA-tagged P21 (P21-HA) or His-tagged TEV HC-Pro (His-HC-TEV) suppressor and then the 21-nt ds-sRNA-binding activity of these extracts was tested in the absence of antibody, in the presence of HA antibodies, and in the presence of His antibodies.
In the absence of antibody, both the P21-HA and His-HC-TEV extracts efficiently bound 21-nt ds-sRNA. As Fig. 2C shows, HA antibodies but not His antibodies caused a supershift on P21-HA extracts, while His antibodies but not HA antibodies resulted in a supershift on His-HC-TEV extracts. Supershift assays were also conducted with extracts prepared from leaves in which Myc-tagged TCV CP (Myc-CP), P15 (Myc-P15), or BSMV ␥B (Myc-␥B) suppressors were expressed. In the absence of antibody, the Myc-CP and Myc-␥B extracts efficiently bound 21-nt ds-sRNA, whereas the Myc-P15 extract weakly bound 21-nt ds-sRNA. As Fig. 2D shows, the presence of Myc but not His antibodies resulted in strong supershifts for all three extracts (compare Fig. 2D, lane 2 to 3, lane 5 to 6, and lane 8 to 9). These data indicate that the suppressor proteins themselves bound the 21-nt ds-sRNA in the suppressor extracts.
P19 binds ds-sRNA as a dimer (62,68). As other suppressords-sRNA complexes (P14, P15, P21, etc.) also migrated more slowly than expected from the monomer, it is possible that these proteins also bind ds-sRNAs as dimers or as oligomers.
As the TCV CP and P14 suppressors bind both long dsRNAs and 21-nt ds-sRNAs, they will be referred to as size-independent dsRNA-binding suppressors. The P19, P21, P15, BSMV ␥B, and TEV HC-Pro suppressors bind 21-nt ds-sRNA but fail to bind long dsRNAs, and therefore these proteins will be referred to as size-selective ds-sRNA-binding suppressors. We could also detect a size-selective ds-sRNA-binding activity in TMV extract. As the replicase proteins are supposed to act as silencing suppressors in tobamoviruses (16,30), we postulate that the TMV replicase is also a size-selective ds-sRNA-binding silencing suppressor.
We failed to detect dsRNA-binding activity in the extracts of PVX, CMV, TAV, and PVY. It is possible that (some of) these viruses also express dsRNA-binding protein, but these proteins are not present in an active form in either the viral or suppres-sor extract. For instance, it is likely that different HC-Pro suppressors inhibit silencing by targeting an identical step. Indeed, like TEV HC-Pro, Turnip mosaic virus HC-Pro is also a ds-sRNAbinding protein, although it binds ds-sRNA relatively weakly (E. Chapman and J. Carrington, personal communication). Therefore we assume that PVY HC-Pro could be a ds-sRNA-binding suppressor.
P21 and P15 suppressors bind 21-nt ds-sRNAs preferentially. In direct competition assays P14 binds both 26-nt ds-sRNA (having 24-nt duplex and 2-nt 3Ј overhangs) and 21-nt ds-sRNA (having 19-nt duplex region and 2-nt 3Ј overhangs), while P19 preferentially binds 21-nt ds-sRNA (39,62,68). As suppressors with different ds-sRNA-binding preferences can interfere with short-and long-sRNA-mediated silencing pathways differently (4,62), we wanted to define whether other viral suppressors could discriminate between 21and 26-nt ds-sRNAs. and used as probes in binding studies. The 21-nt ds-sRNAs were incubated with crude viral extract. Extracts prepared from mock-inoculated (Ϫ) plants and extracts isolated from plants that were infected with P⌬14, a mutant PoLV that was unable to express P14, were used as negative controls. The ds-sRNA-binding activity of extracts prepared from PVX P14 and PVX P19 recombinant PVXs, which express the P14 and P19 suppressor, respectively, shows that ds-sRNA-binding proteins are also active if they are expressed by heterologous viruses. * indicates a host-encoded nonspecific ds-sRNA-binding activity that can be frequently detected in extracts prepared from wild-type, mock-inoculated, agroinfiltrated, or virus-inoculated plants. To this aim, direct competition assays were carried out with BYV P21, PCV P15, and TCV CP suppressor extracts. Labeled 26-nt ds-sRNAs were incubated with suppressor extracts and with increasing molar concentrations of unlabeled 21-nt and 26-nt ds-sRNA competitors (Fig. 3A, B, and C, upper panels). Competition experiments were repeated with labeled 21-nt ds-sRNAs (Fig. 3A, B, and C, bottom panels). As Fig. 3 shows, TCV CP bound ds-sRNA without obvious size specificity, because approximately the same molar excess of unlabeled 21-nt ds-sRNAs and 26-nt ds-sRNAs was required to outcompete either labeled 21-or 26-nt ds-sRNA (Fig. 3C). By contrast, P21 and P15 extracts bound 21-nt ds-sRNAs preferentially, and in both cases the 21-nt ds-sRNAs were much more efficient unlabeled competitors than the 26-nt ds-sRNAs ( Fig. 3A and B).
It has been reported that Escherichia coli-expressed P21 is a general nucleic acid binding protein whose binding affinity increases with the length of the RNA (69). By contrast, our binding experiments strongly suggest that in planta-expressed P21 is a size-selective dsRNA-binding protein. As at a low P21-to-RNA ratio bacterially expressed P21 also acts as sizespecific ds-sRNA-binding protein (Chapman and Carrington, personal communication), it is likely that these apparently conflicting results might be due to different assay conditions. Moreover, hairpin-inducing silencing experiments (Fig. 1C and D) suggest that in plant cells P21 fails to bind long dsRNAs but binds ds-sRNAs.
Suppressors show different structural requirements for ds-sRNA binding. P19 binds ds-sRNAs as a homodimer. The central and C-terminal ␤-strands of P19 contact the stem of ds-sRNA, while the N-terminal ␣-helixes (reading head) interact with the 5Ј end of sRNAs. Due to the limited flexibility of the reading head, efficient ds-sRNA binding occurs only on dsRNAs having 19-nt duplex regions. On the other hand, 3Ј overhangs are not required for ds-sRNA binding. P19 binds 19-nt blunt-ended dsRNAs (19-nt dsRNAs) more efficiently than 21-nt ds-sRNAs (4,62,68). It was demonstrated that TEV HC-Pro, another size-selective ds-sRNA-binding suppressor, binds 21-nt ds-sRNA more efficiently than 19-nt dsRNA (L. Lakatos et al., submitted for publication). These findings have been interpreted to mean that size-selective ds-sRNAbinding requires interaction with the end of the ds-sRNA molecules, and that different suppressors can identify different structural elements of ds-sRNA ends. To test whether other suppressors also require 3Ј 2-nt overhangs for ds-sRNA binding, labeled 19-nt dsRNA was incubated with different suppressor extracts, and then dsRNA binding was analyzed in gel mobility shift assays.
The P14 and TCV CP suppressors bound 19-nt dsRNAs and 21-nt ds-sRNAs equally well (compare Fig. 2B, lanes 3 and 5, with Fig. 4A, lanes 3 and 5), suggesting that these suppressors did not require 3Ј overhangs for ds-sRNA binding. By contrast, the PCV P15 and BSMV ␥B extracts failed to bind 19-nt dsRNA (Fig. 4A, lanes 4 and 7), suggesting that the P15 and ␥B suppressors required the presence of 3Ј overhangs for efficient ds-sRNA binding. To confirm these conclusions direct competition assays were carried out.
First, P19 and TEV HC-Pro extracts were studied to validate the system. We found that P19 extract bound 19-nt dsRNA preferentially (Fig. 4B), while TEV HC-Pro extract formed complexes more efficiently with 21-nt ds-sRNA (Fig. 4C). These findings were in line with the results obtained with purified P19 (62,68) and TEV HC-Pro proteins (L. Lakatos et al., submitted for publication), indicating that suppressor extract can be used to define whether a suppressor requires 3Ј overhangs for efficient ds-sRNA bindings or not. As Fig. 4 shows, competition assays confirmed that P14 and TCV CP bound 19-nt dsRNAs and 21-nt ds-sRNAs with similar efficiency ( Fig. 4D and E) and that PCV P15 and BSMV ␥B extracts bound 21-nt ds-sRNAs much more efficiently than 19-nt dsRNAs (Fig. 4F and G). These data suggest that TCV CP and P14 interact with the stem but not with the ends of ds-sRNAs. By contrast, the P15 and ␥B suppressors might interact directly with the 3Ј 2-nt overhangs of ds-sRNAs as these proteins require the presence of 3Ј overhangs for efficient ds-sRNA binding.

DISCUSSION
Here we show that many unrelated plant viral RNA silencing suppressors bind dsRNAs. These data strongly support the hypothesis that dsRNA binding is a widely used silencing suppression strategy which has evolved independently many times.
dsRNA-binding RNA silencing suppressors. In vitro binding studies revealed that extracts of CymRSV, PoLV, BYV, TEV, TMV, PCV, and BSMV plant RNA viruses bind dsRNAs. As in each case studied the suppressor extract binds dsRNAs similarly to the viral extract, we concluded that the silencing suppressors themselves form complexes with dsRNAs. The findings that the purified P19 (31,54,62,68), P14 (39), P21 (11,69), and TEV HC-Pro (L. Lakatos et al., submitted for publication) proteins can bind ds-sRNAs in vitro and that antibodies specifically affect 21-nt ds-sRNA binding of tagged suppressor extracts (Fig. 2C and D) support this conclusion.
Competitive dsRNA-binding-based silencing suppression. Our findings that many unrelated suppressors bind dsRNAs suggest that dsRNA binding is a key element of silencing suppression. We propose that dsRNA-binding suppressors inhibit silencing by sequestering a dsRNA component of the silencing response and that size-selective and size-independent dsRNA-binding suppressors target different dsRNAs. In plants, silencing generates only 21-nt ds-sRNAs from RNA viruses (40,53), and hence expression of a protein that preferentially binds these molecules appears to be an ideal viral counterdefensive strategy. We suggest that size-selective dsRNA-binding suppressors interfere with silencing by sequestering ds-sRNAs, preventing the formation of active silencing effector complexes. Many different observations support the idea that all these suppressors act by binding ds-sRNAs. For instance, in the GFP-IR coinfiltration assay the P19, P21, P15, BSMV ␥B, and TEV HC-Pro suppressors act similarly, and these proteins inhibit hairpin-induced silencing but fail to prevent ds-sRNA generation. Moreover, transgenic expression of P19, P21, HC-Pro, and P15 results in similar phenotypes (11,19), likely because the endogenous sRNA (microRNA)-mediated silencing pathways were inhibited in each transgenic plant.
The binding data suggest that the P19, P21, HC-Pro, and P15 suppressors inhibit the formation of active RISCs in transgenic plants by binding the ds-sRNA intermediate molecules of mi-croRNA biogenesis. Finally, the findings that in heterologous (Drosophila melanogaster or mammalian) systems, the P19 (31), P15 (19), P21 (L. Lakatos et al., submitted for publication), and TEV HC-Pro (46) suppressors inhibit both dsRNAand ds-sRNA-induced silencing are also consistent with suggestions that these proteins act by sequestering ds-sRNAs. It has been experimentally proven that P19 can sequester viral ds-sRNAs in CymRSV-infected cells (31). Therefore, it is likely that other size-selective ds-sRNA-binding suppressors also inhibit virus-induced silencing by preventing the incorporation of viral sRNAs into effector complexes. Indeed, it has been shown that HC-Pro can also bind viral sRNAs in TEVinfected plants (L. Lakatos et al., submitted for publication).
We showed that TCV CP, like PoLV P14, is a size-independent dsRNA-binding protein. This explains apparently conflicting previous observations that CP prevents ds-sRNA accumulation in hairpin-induced silencing assays (11,43,45), while it suppresses ds-sRNA-induced silencing in mammalian cells (19). We suggest that size-independent dsRNA-binding suppressors as TCV CP and P14 prevent the generation of sRNAs from hairpin transcripts because they compete with DCLs for long dsRNAs, and they inhibit ds-sRNA-mediated silencing because they can also sequester ds-sRNAs. It is still not known if CP or P14 suppresses silencing by sequestering viral precursors of ds-sRNAs in virus-infected cells or inhibit silencing by binding ds-sRNAs.
As silencing also plays an antiviral role in animals, it is likely that many animal viruses suppress silencing by expressing dsRNA-binding proteins (32). Indeed, it has been demonstrated that size-independent dsRNA-binding proteins could act as silencing suppressors in insect and mammalian cells (5,9,14,28,33,55). The flock house virus B2 suppressor is a dsRNA-binding protein which interacts with the backbone of dsRNA as a dimer (10,35). B2 binds longer dsRNA more efficiently than ds-sRNA and inhibits antiviral silencing in C. elegans upstream of AGO, likely by binding the precursor of ds-sRNA (38). Thus, the molecular basis of silencing suppres-sion might be similar for TCV CP, PoLV P14, and B2. Interestingly, a size-specific ds-sRNA-binding animal viral silencing suppressor has not been identified yet.
Evolution of dsRNA-binding silencing suppressors. Many dsRNA-binding suppressors are evolutionarily unrelated, suggesting that dsRNA-binding silencing suppressors have evolved independently many times. Convergent evolution results in remarkably similar dsRNA-binding characteristics of unrelated suppressors. For instance, both BSMV ␥B and TEV HC-Pro are size-selective ds-sRNA-binding silencing suppressors that require the presence of 3Ј overhangs for efficient ds-sRNA-binding even though no significant sequence similarity can be found between these two proteins. Moreover, BSMV ␥B is a small protein whose only function is silencing inhibition, while TEV HC-Pro is a multifunctional protein.
Interestingly, although BSMV ␥B and HC-Pro are very dissimilar, both suppressors contain a cysteine-rich region that might form a zinc finger motif (1,7,17,29,52). It is tempting to speculate that zinc finger domains play important roles in the ds-sRNA binding of BSMV ␥B and HC-Pro. Moreover, predicted zinc fingers were also identified in suppressors which are distantly related to BSMV ␥B, such as PCV P15 (20), furovirus P19 (58), and tobravirus P16 (29,46,70), and in other suppressor proteins which are not related to either BSMV ␥B or HC-Pro, such as P23 of Citrus tristeza virus, a closterovirus (36,37), or P10 of Grapevine virus A, a vitivirus (13,23). These zinc finger-containing suppressors could also act by binding dsRNAs size independently or size selectively. Indeed, it was shown that citrus tristeza virus P23 binds dsRNA in vitro, although its size selectivity was not studied. Of course, not all dsRNA-binding suppressors contain zinc fingers: for example, zinc fingers could not be identified in P19, P21, or TCV CP.
Suppressors have evolved to target silencing with as little damage as possible to the host. Size-selective ds-sRNA binding is likely the result of such a dual selection, and these suppressors sequester sRNAs that play a role in antiviral response, while they do not interfere with endogenous long ds-sRNAs or long structured RNAs (4,62). Due to this remarkable specificity of these suppressors, they can be expressed in all stages of infections. Interestingly, the expression of P14 and TCV CP size-independent silencing suppressors in infected plants might be downregulated by the virus (49,71). The level of Sg2 RNA, from which P14 is translated, declines after 2 to 3 d.p.i. (39,49,50). In TCV-infected cells, the formation of capsid reduces available suppressor levels. It has been proposed that these are viral strategies to reduce symptoms, because infection with a mutant PoLV which expresses P14 constitutively or with a mutant TCV which expresses CP that fails to form capsid results in strong symptom accentuation (49,71).
One can speculate that ancient dsRNA-binding suppressors were size-independent dsRNA-binding proteins that evolved in two ways to reduce unnecessary host damage. They could evolve into size-selective ds-sRNA-binding proteins or control the expression of size-independent dsRNA-binding suppressors.