Double-stranded RNA-binding proteins could suppress RNA interference-mediated antiviral defences

RNA interference (RNAi) is a double-stranded (ds)RNA-inducible, sequence-specific RNA-degradation mechanism that operates as a natural antiviral system in plants and animals. Successful virus infection requires evasion or suppression of RNAi. Indeed, RNAi suppressor proteins have been identified in plant and animal viruses, although the molecular mechanism of silencing inhibition is still poorly understood. Because many RNA viruses encode dsRNA-binding proteins (dsRBPs) and as RNAi is triggered by the accumulation of dsRNAs, dsRBPs were examined to see if they inhibit RNAi. Here, it is shown that heterologous dsRBPs suppressed RNAi in plants, indicating that in natural hostvirus interactions, pathogen-encoded dsRBPs could inactivate RNAi-mediated host defences.

Published ahead of print on 22 January 2003 as DOI 10.1099/vir.0.18987-0.

Eukaryotes have evolved many different systems to resist virus infection. Identification of specific virus-encoded molecules or recognition of nucleic acid structures that are present only in infected cells could induce antiviral responses (Plasterk, 2002). As long double-stranded (ds)RNAs do not occur in the cytoplasm of eukaryotic cells, the accumulation of ds replicative intermediates of RNA viruses activates antiviral responses as RNA interference (RNAi) or translation inhibition and apoptosis. RNAi is an ancient defence mechanism that degrades dsRNAs and cognate mRNAs in a sequence-specific manner (Hannon, 2002; Voinnet, 2001; Zamore, 2001). Viral dsRNAs are first processed by an RNase III-like nuclease (DICER) into 2126 nt dsRNAs (siRNAs) that guide another nuclease complex (RISC) to cleave homologous single-stranded (ss) viral RNAs. siRNAs also serve as guides for an RNA-dependent RNA polymerase to transform the target ssRNA into dsRNA (Lipardi et al., 2001; Sijen et al., 2001). RNAi was shown to act as an efficient antiviral system in plant (Matzke et al., 2001; Vance & Vaucheret, 2001) and insect cells (Li et al., 2002) and might also play an antiviral role in mammalian cells (Cullen, 2002). In higher plants, RNAi has evolved into a whole plant defence system. Cell-autonomous RNAi generates an unidentified mobile signal, thereby directing sequence-specific RNA degradation in distant tissues (Palauqui et al., 1997; Voinnet & Baulcombe, 1997). To inhibit the antiviral effect of RNAi, plant (Li & Ding, 2001) and insect (Li et al., 2002) viruses express different RNAi suppressor proteins. Although, the suppression of RNAi could be essential for efficient virus infection, the molecular mechanism of RNAi inhibition is still unknown.

In vertebrate cells, dsRNAs also activate RNA-dependent protein kinase (PKR)-mediated, non-specific antiviral responses, including inhibition of translation and induction of cell death. As a counterdefence strategy, many vertebrate viruses express dsRNA-binding proteins (dsRBPs) that prevent PKR activation by sequestering dsRNAs (Kaufman, 1999). As dsRNAs play a role in RNAi and since many non-vertebrate RNA viruses also express dsRBPs, it is possible that virus-encoded dsRBPs could operate as inhibitors of RNAi. To address this issue, we tested to see if dsRBPs could suppress RNAi in plants.

Transgene expression can also trigger RNAi. Since virus- and transgene-induced RNAi operate in overlapping pathways, virus-encoded RNAi suppressors inhibit transgene-triggered RNAi. As the mechanism of plant and animal RNAi is conserved, the Agrobacterium tumefaciens infiltration assay has been used to identify silencing suppressors encoded by both plant and animal viruses (Li et al., 2002; Voinnet et al., 1999). The infiltration of green fluorescent protein (GFP) transgenic Nicotiana benthamiana plants with A. tumefaciens carrying a vector in which the transcription of GFP is controlled by the 35S promoter (35S-GFP) not only results in transient GFP expression but also leads to the induction of GFP silencing. Cell-autonomous GFP silencing manifests as a weakening of green fluorescence, a decline in the level of GFP mRNA and an accumulation of GFP-specific siRNAs in the infiltrated patches (Brigneti et al., 1998). siRNAs accumulate in two functionally different size classes. The 2123 nt siRNA fraction guides RISC, while the 2426 nt siRNA fraction is associated with systemic silencing (Hamilton et al., 2002). If 35S-GFP is co-infiltrated with another A. tumefaciens expressing an RNAi suppressor, the levels of green fluorescence remain high, GFP mRNA levels do not decrease and siRNA accumulation is reduced in the infiltrated leaves (Voinnet et al., 2000). Escherichia coli RNase III and the mammalian reovirus outer shell polypeptide σ3 are among the best-characterized dsRBPs; therefore, we tested the RNAi suppressor capacity of these proteins and their mutants. Both proteins carry conservative dsRNA-binding motifs and bind dsRNAs in vitro and in vivo (Dasgupta et al., 1998; Denzler & Jacobs, 1994; Fierro-Monti & Mathews, 2000; Huismans & Joklik, 1976; Kharrat et al., 1995; Nicholson, 1999; Yue & Shatkin, 1997). The postulated silencing suppressor capacity of E. coli RNase III, a mutant RNase III that binds dsRNA but lacks RNA cleavage activity (Rnc70) (Dasgupta et al., 1998) and reovirus σ3 proteins were tested in the Agrobacterium co-infiltration assay. The rnc+ (encodes RNase III) and rnc70 (encodes Rnc70) genes were amplified by PCR from plasmids pACS21 and pSDF70 (Dasgupta et al., 1998) with primers RNC START (5'-ATGAACCCCATCGTAAT-3') and RNC STOP (5'-TCATTCCAGCTCCAGTT-3'). The PCR products were then cloned into the SmaI-digested Agrobacterium binary vector BIN61S (Silhavy et al., 2002) to create the constructs 35S-rnc+ and 35S-rnc70 (Fig. 1a). The S4 segment (encodes σ3) was amplified by PCR with primers S4START (5'-ATGGAGTGTTGCTTGCC-3') and S4STOP (5'-TTAGCCAAGAATCATCGG-3') from plasmid pBC12BI (Giantini & Shatkin, 1989) and cloned into the SmaI site of BIN61S to create the construct 35S-σ3. As a negative control, a 35S-Δσ3 clone was constructed by PCR, amplifying the 5' first 846 nt segment of the S4 gene with primers S4START (5'-ATGGAGTGTTGCTTGCC-3') and ΔS4STOP (5'-TTACATTTTACAGTTCCCAG-3'). Then, the PCR fragment was cloned into the SmaI-digested BIN61S plasmid. 35S-Δσ3 encodes a truncated protein that fails to bind dsRNAs (Miller & Samuel, 1992).

(31K): Fig. 1. (a) Schematic representation of constructs used in this work. dsRBD, dsRNA-binding domains; 35S and 35S poly(A), promoter and terminator regions of the 35S transcript encoded by Cauliflower mosaic virus (CaMV), respectively; NOS term, terminator region of the NOS transcript of A. tumefaciens. (b) Effect of dsRBPs on transient GFP expression. GFP transgenic N. benthamiana plants were infiltrated with 35S-GFP or co-infiltrated with 35S-GFP and dsRBPs as E. coli RNase III (35S-GFP+35S-rnc+), Rnc70 (35S-GFP+35S-rnc70) and reovirus σ3 (35S-GFP+35S-σ3). Co-infiltration of 35S-GFP with the truncated version of σ3 lacking dsRNA-binding activity (35S-GFP+35S-Δσ3) was used as a control. Photographs of infiltrated leaves were taken under UV illumination.

To examine whether dsRBPs suppress RNAi, GFP silencing was monitored in 35S-GFP infiltrated cells and in 35S-GFP+35S-rnc+, 35S-GFP+35S-rnc70, 35S-GFP+35S-σ3 and 35S-GFP+35S-Δσ3 co-infiltrated leaves of GFP transgenic N. benthamiana plants. Agrobacterium infiltration assays, GFP expression tests and GFP-specific RNA gel blot analyses were carried out as described previously (Silhavy et al., 2002). In line with previous reports (Voinnet et al., 2000), we found that, although green fluorescence was strong (Fig. 1b) and GFP mRNA expression was still high (Fig. 2a, top panel), the accumulation of GFP-specific siRNAs (Fig. 2a, bottom panel) in 35S-GFP infiltrated leaves at 3 days post-inoculation (p.i.) confirmed the early induction of GFP silencing. As expected, co-infiltration of 35S-Δσ3 with 35S-GFP did not affect GFP silencing (Fig. 1b and Fig. 2a). In contrast, co-infiltration of 35S-rnc+, 35S-rnc70 and 35S-σ3 with 35S-GFP suppressed the early effects of RNAi. GFP expression was stronger (Fig. 1b) and levels of GFP mRNA were higher (Fig. 2a, top panel), while the accumulation of GFP-derived siRNAs was reduced (Fig. 2a, bottom panel) in all three dsRBP co-infiltrated samples compared with 35S-GFP-injected and 35S-GFP+35S-Δσ3 co-infiltrated controls. These findings indicate that dsRBPs could act as RNAi suppressors. Different dsRBPs, however, suppressed transgene-induced RNAi to a different degree. GFP-derived siRNAs were not detected in 35S-GFP+35S-rnc+ or 35S-GFP+35S-σ3 co-infiltrated samples, while the presence of Rnc70 only reduced the levels of the siRNA accumulation (Fig. 2a, bottom panel). These data indicate that RNase III and σ3 are strong RNAi suppressors, whereas Rnc70 acts as a weak inhibitor of silencing. By 6 days p.i., the degree of GFP silencing was similar in 35S-GFP+35S-rnc70 co-infiltrated samples to the 35S-GFP- and 35S-GFP+35S-Δσ3-injected controls (Fig. 1b and Fig. 2b), indicating that the weak RNAi suppressor could only delay the silencing-mediated degradation of GFP. In contrast, strong GFP expression (Fig. 1b) together with very low levels of GFP-specific siRNA indicated that the strong RNAi suppressor σ3 inhibited GFP silencing in the 35S-GFP+35S-σ3 co-infiltrated leaves, at least to 6 days p.i. (Fig. 2b). As infiltration with 35S-rnc+ leads to local necrosis by 45 days p.i., 35S-rnc+ co-infiltrated leaves could not be analysed at 6 days p.i.

(36K): Fig. 2. Early (a) and late (b) effects of dsRBPs on transgene-induced RNAi. (a) Levels of GFP mRNA (top panel) and GFP-specific siRNA (bottom panel) in samples taken from infiltrated patches of 35S-GFP-injected or 35S-GFP+35S-rnc+, 35S-GFP+35S-rnc70, 35S-GFP+35S-σ3 or 35S-GFP+35S-Δσ3 co-injected leaves of GFP transgenic N. benthamiana plants. The same concentration of total RNA in the same volume (∼5 µg) was used for mRNA and siRNA gel blot analyses. Ethidium bromide-stained rRNA is shown as a loading control. A radioactively labelled GFP PCR fragment generated by random priming was used as the probe for mRNA gel blot assays, while a labelled in vitro transcript corresponding to the antisense strand of GFP was used as the probe in Northern blot analyses of siRNAs.

Cell-autonomous GFP silencing generates signals that lead to systemic GFP silencing in non-infiltrated tissues of GFP transgenic N. benthamiana plants. Systemic GFP silencing can be monitored easily because chlorophyll autofluorescences red when no GFP is expressed. The formation of red fluorescence around the infiltrated area by 56 days p.i. in the 35S-GFP infiltrated leaves of GFP transgenic N. benthamiana plants showed the induction of systemic GFP silencing (data not shown) (Voinnet & Baulcombe, 1997). Because the accumulation of the long 2426 nt GFP-specific siRNA fraction correlates with systemic silencing (Hamilton et al., 2002) and because dsRBPs reduce the levels of both short and long siRNAs (Fig. 2, bottom panels), we expected that co-infiltration of dsRBPs with 35S-GFP would interfere with systemic silencing. Indeed, the development of red fluorescence was delayed by 12 days in 35S-GFP+35S-rnc70 co-infiltrated leaves and by 23 days in 35S-GFP+35S-σ3 co-infiltrated leaves (data not shown). As expected, co-infiltration of 35S-Δσ3 with 35S-GFP did not have an affect on systemic GFP silencing (data not shown).

dsRBPs inactivate PKR by depleting dsRNAs. If RNAi suppression of dsRBPs is also based on dsRNA sequestering, RNase III, Rnc70 and σ3 should effectively bind dsRNAs in plant cells, thereby preventing the silencing-mediated degradation of dsRNAs. To test this hypothesis, silencing-mediated degradation of dsRNA was analysed in the presence and absence of dsRBPs. GFP transgenic N. benthamiana leaves were infiltrated with Agrobacteria carrying a GFP inverted repeat (Fig. 1a); thus, the expressed mRNAs formed hairpin structures with a long stem (35S-IR) and could be digested by DICER. In line with previous reports (Johansen & Carrington, 2001) at 3 days p.i., siRNAs were very abundant in 35S-IR-infiltrated cells of GFP transgenic N. benthamiana (Fig. 3a, bottom panel), indicating that 35S-IR induced strong RNAi. As shown in Fig. 3(a), co-infiltration of 35S-Δσ3 with 35S-IR did not influence RNAi-mediated dsRNA degradation, while dsRBPs inhibited 35S-IR-induced RNA silencing. siRNAs were not detected (35S-IR+35S-rnc+) or they accumulated to low levels (35S-IR+35S-rnc70 and 35S-IR+35S-σ3) in 35S-IR and dsRBP co-infiltrated tissues (Fig. 3a). The accumulation of a higher molecular mass mRNA fraction that corresponds to IR mRNA in samples taken from 35S-IR+35S-rnc70 and 35S-IR+35S-σ3 co-infiltrated leaves (Fig. 3a, top panel) suggests that dsRBPs prevented the degradation of IR dsRNA. The lack of this RNA fraction in control samples (Fig. 3a, top panel) could reflect the activity of DICER and other dsRNases. IR mRNAs were also absent in 35S-IR+35S-rnc+ co-infiltrated samples, even though siRNAs were not detected (Fig. 3a). These data suggest that E. coli RNase III degraded the co-expressed IR mRNAs.

(42K): Fig. 3. (a) Effects of dsRBPs on dsRNA-induced RNAi. RNA samples were isolated from GFP transgenic N. benthamiana plants infiltrated with Agrobacteria expressing hairpin dsGFP transcripts (35S-IR) or co-infiltrated 35S-IR with RNase III (35S-IR+35S-rnc+), Rnc70 (35S-IR+35S-rnc70), reovirus σ3 (35S-IR+35S-σ3) and the truncated version of σ3 lacking dsRNA-binding activity (35S-IR+35S-Δσ3). Levels of IR mRNAs (IR), endogenous GFP mRNAs (GFP) (top panel) and siRNAs (bottom panel) were analysed by RNA gel blots. The same concentration of total RNA in the same volume (∼5 µg) was used for mRNA and siRNA gel blot analyses. Ethidium bromide-stained rRNA is shown as the loading control. A radioactively labelled GFP PCR fragment generated by random priming was used as the probe for mRNA gel blot assays, while a labelled in vitro transcript corresponding to the antisense strand of GFP was used as the probe in Northern blot analyses of siRNAs. (b) Effects of dsRBPs on accumulation of miRNAs. Labelled antisense oligonucleotides were used as probes for detecting levels of miR157 and miR171 in RNA samples isolated from infiltrated patches of wild-type N. benthamiana plants injected with 35S-GFP, 35S-σ3, 35S-Δσ3 and 35S-rnc+.

In addition to siRNA, DICER also generates 2125 nt long ss micro (mi)RNAs, which play a role in developmental regulation (Hutvagner et al., 2001; Ketting et al., 2001; Llave et al., 2002; Reinhart et al., 2002). miRNAs are produced from hairpin precursor RNAs transcribed from endogenous genes (Lee et al., 2002). We examined the effect of heterologous dsRBPs on miRNA accumulation in the infiltrated leaves of N. benthamiana plants. Antisense oligonucleotides corresponding to miR157 (miR157 ANTISENSE, 5'-GTGCTCTCTATCTTCTGTCAA-3') and miR171 (miR171 ANTISENSE, 5'-GATATTGGCGCGGCTCAATCA-3') (Reinhart et al., 2002) were radioactively labelled by T4 polynucleotide kinase and used as probes. RNA gel blot analysis revealed that miR157 and miR171 accumulated to equal levels in non-infiltrated controls (data not shown) and in 35S-GFP, 35S-Δσ3- and 35S-σ3-infiltrated leaves (Fig. 3b), while miRNA accumulation was reduced in 35S-rnc+ infiltrated samples (Fig. 3b). These data suggest that σ3 dsRBP failed to sequester miRNA precursors, although σ3 could sequester long dsRNA precursors of siRNAs. Indeed, σ3 binds dsRNAs efficiently only if they are longer than 3245 bp (Yue & Shatkin, 1997). Because RNase III cleaves structured ssRNAs (Nicholson, 1999), it might also bind miRNA precursors, thereby reducing the accumulation of miRNAs. It is possible that certain virus-encoded dsRBPs, like RNase III, interfere with miRNA accumulation, thus contributing to the symptoms of virus infection.

It is likely that certain virus suppressors target conserved elements of the RNAi machinery. Tombusvirus p19 RNAi suppressor binds ds siRNAs, thus inhibiting virus-induced systemic silencing in plants (Silhavy et al., 2002). Other RNAi suppressors might target another conserved elements of RNAi, long dsRNAs. Indeed, we showed that heterologous dsRBPs could effectively suppress RNAi, presumably by sequestering dsRNAs. We propose that many virus-encoded dsRBPs play important roles in pathogenicity by interfering with RNAi-mediated cell-autonomous and systemic host defences. As effective silencing suppression likely requires early, abundant cytoplasmic expression of virus-encoded dsRBPs, we think that only a subset of virus-encoded dsRBPs could operate as natural RNAi suppressors. For instance, in reovirus- or vaccinia virus-infected mammalian cells, the expression of σ3 or E3L might lead to inactivation of RNAi-mediated defences in addition to inhibition of PKR-mediated responses (Kaufman, 1999).

To confer broad-spectrum virus resistance, dsRNA-specific ribonucleases were expressed in transgenic plants (Sano et al., 1997; Watanabe et al., 1995). RNase III- and Rnc70-expressing transgenic plants have shown virus resistance against viruses with segmented genomes (Langenberg et al., 1997; Zhang et al., 2001). However, finding that both RNase III and Rnc70 suppress RNA silencing suggests that the RNAi defence system of these transgenic plants could be compromised; therefore, these transgenic plants might be more susceptible to certain viruses.