Research Articles

Hierarchical Action and Inhibition of Plant Dicer-Like Proteins in Antiviral Defense

See allHide authors and affiliations

Science  07 Jul 2006:
Vol. 313, Issue 5783, pp. 68-71
DOI: 10.1126/science.1128214

This article has a correction. Please see:


The mechanisms underlying induction and suppression of RNA silencing in the ongoing plant-virus arms race are poorly understood. We show here that virus-derived small RNAs produced by Arabidopsis Dicer-like 4 (DCL4) program an effector complex conferring antiviral immunity. Inhibition of DCL4 by a viral-encoded suppressor revealed the subordinate antiviral activity of DCL2. Accordingly, inactivating both DCL2 and DCL4 was necessary and sufficient to restore systemic infection of a suppressor-deficient virus. The effects of DCL2 were overcome by increasing viral dosage in inoculated leaves, but this could not surmount additional, non–cell autonomous effects of DCL4 specifically preventing viral unloading from the vasculature. These findings define a molecular framework for studying antiviral silencing and defense in plants.

In RNA silencing, ribonuclease (RNase) III–like enzymes in the Dicer family produce short interfering (si)RNA and micro (mi)RNA from RNA with double-stranded (ds) features (1). These molecules guide RNA-induced silencing complexes (RISCs) to suppress gene expression at the transcriptional, RNA-stability, and translational levels (2). Arabidopsis thaliana has four specialized Dicer-like (DCL) proteins. DCL1 processes fold-back precursors to release miRNAs (3). DCL3 produces 24–nucleotide (nt)–long, DNA repeat–associated siRNAs guiding heterochromatin formation (4). DCL4 generates 21-nt-long siRNAs that mediate posttranscriptional silencing of some endogenous genes [trans-acting (ta)–siRNAs; (5, 6)] and of transgenes mediating RNA interference (7). DCL2 synthesizes stress-related natural-antisense-transcript (nat)–siRNAs (8), siRNAs derived from at least one virus (4), and, in dcl4 mutant plants, it alternately processed ∼22-nt siRNAs from ta-siRNA precursors (5, 6).

The observations that virus-derived siRNAs accumulate in plant and insect infected tissues and that many viruses encode suppressor proteins targeting DCL, RISC, or small RNA activities strongly suggest that RNA silencing has antiviral roles (911). In plants, one or more of the six RNA-dependent RNA-polymerase (RDR) paralogs, including Arabidopsis RDR6 and RDR1, may strengthen primary silencing responses by producing dsRNA from viral templates (12) and by amplifying mobile silencing signals conditioning antiviral immunity in non-infected tissues (7, 13). Nevertheless, the genetic bases of silencing induction and suppression by plant viruses remain unclear. Even the existence of an antiviral RISC (“slicer”) is arguable because DCL-mediated processing of virus-derived dsRNA could be, in principle, sufficient to dampen infections. It remains also uncertain how, when, and where antiviral silencing and its suppression impact susceptibility and defense in whole plants. This study addresses these issues using Arabidopsis silencing mutants and three distinct RNA viruses.

DCL4- and DCL2-dependent siRNAs recruit an antiviral RISC. Arabidopsis plants were inoculated with modified Tobacco rattle virus (TRV-PDS) (Fig. 1A) containing a fragment of the Arabidopsis phytoene desaturase (PDS) gene in place of the RNA2-encoded 2b and 2c sequences. Like TRV-infected tissues (Fig. 1B), TRV-PDS–infected tissues are free of disease symptoms, because of a strong silencing response that dramatically reduces viral titers (14), and exhibit extensive photobleaching due to virus-induced gene silencing (VIGS) of PDS (Fig. 1C) (7).

Fig. 1.

(A) Genome organization of TRV and its TRV-PDS derivative. (B and C) Asymptomatic infection (B) and extensive photobleaching (C) caused by TRV and TRV-PDS, respectively. (D and E) Analysis of (top) low- and (bottom) high-molecular-weight RNAs from TRV-PDS–infected plants carrying single (D) and double (E) dcl mutations [14 days post inoculation (dpi)]. The probe was specific for viral PDS. The numbers of infected plants showing photobleaching are from four independent experiments involving four plants each. (F to L) Disease symptoms and VIGS in dcl mutants (14 dpi). (M) TRV-PDS siRNA analysis in dcl2-dcl3-dcl4 triple mutants (14 dpi). rRNA shown by ethidium bromide staining.

TRV-PDS–specific siRNAs accumulated as discrete 21-nt and 24-nt species in wild-type (WT) Arabidopsis (Fig. 1D), a pattern unchanged in rdr1, rdr2, rdr6 [supporting online material (SOM), fig. S1], and dcl2 mutants (Fig. 1D). However, the 24-nt and 21-nt siRNAs were undetectable in dcl3 and dcl4 mutants, respectively. Loss of 21-nt siRNAs coincided with appearance of 22-nt siRNAs in dcl4 mutants (Fig. 1D). Identical siRNA patterns were detected with an RNA2(TRV)-specific probe, whereas probes specific for cellular PDS sequences absent in TRV-PDS yielded no signal, which indicated that all siRNA species detected were of viral origin (fig. S1). Viral RNA accumulation was not altered in any of the single dcl or rdr mutants, nor was the extent and/or consistency of VIGS compared with WT-infected plants (Fig. 1, D and F to I) (fig. S1), which suggested redundancy among those factors in mediating antiviral silencing.

dcl combination mutants were then infected. Like dcl3, dcl2-dcl3 double mutants accumulated only 21-nt siRNAs and had unaltered TRV-PDS levels and VIGS phenotype (Fig. 1, E and J). Similar infection and VIGS phenotypes were detected in dcl3-dcl4 double mutants, although they accumulated exclusively 22-nt siRNAs (Fig. 1, E and K). By contrast, VIGS was abolished in dcl2-dcl4 plants accumulating only 24-nt siRNAs. This coincided with higher virus RNA levels and stronger disease symptoms (Fig. 1, E and L), also observed in dcl2-dcl3-dcl4 triple mutants, in which low-abundance, DCL1-dependent siRNAs were detected (Fig. 1M; fig. S1).

Although the DCL3-dependent, 24-nt siRNA accumulated to the same high levels as 21-nt and 22-nt siRNAs (Fig. 1, D and E), it was neither necessary (Fig. 1D) nor sufficient (Fig. 1E) to mediate VIGS and defense against TRV. Thus, the DCL3-dependent dicing reaction alone could not limit virus infection. The additional requirement for an siRNA-loaded RISC was evidenced by the fact that loss of slicer activity (i.e., PDS VIGS), appearance of disease symptoms, and high virus accumulation were inherently correlated in dcl2-dcl4. We conclude that the respective 21-nt and 22-nt siRNA products of DCL4 and DCL2 (Fig. 1, D and E) guide an antiviral RISC to promote defense against TRV. DCL2 likely acts as a DCL4 substitute because its activity was contingent on DCL4 loss of function.

TCV-encoded P38 suppresses DCL4. To assess the generality of these findings and the significance of a DCL4 substitute, we analyzed infection by Turnip crinkle virus (TCV, Fig. 2A). Unlike TRV, TCV causes disease symptoms in WT Arabidopsis (fig. S2) and encodes a strong silencing suppressor, the P38 capsid protein (15). In WT Arabidopsis, TCV-derived siRNAs accumulated as a single, 22-nt species. This profile was not changed in dcl3, dcl4, or dcl3-dcl4 mutants (Fig. 2A) and not in rdr1, rdr2, or rdr6 backgrounds (fig. S2). By contrast, 22-nt siRNA levels were strongly reduced in dcl2 mutants, as has been reported (4), and in dcl2-dcl3 and dcl2-dcl4 double mutants (Fig. 2A). A time-course analysis revealed that all single, double, and triple mutants with dcl2 mutations accumulated statistically higher TCV RNA levels than did WT plants or mutants without dcl2 (Fig. 2B). However, these effects were modest (up to twofold), as also revealed by P38 immunoblots and symptom evaluation (Fig. 2A and fig. S2), and were disproportionate compared with the near-loss of 22-nt siRNAs in dcl2 and its derivatives.

Fig. 2.

(A) TCV genome organization; TCV siRNA (top) and coat-protein (P38) accumulations (bottom) in single or double dcl mutants (14 dpi). Prot, total protein staining. (B) Time-course analysis of TCV genomic RNA accumulation. Variance analysis of the data from 15 dpi produced five statistical groups (a to e). (C) Analysis of high (top) and low (second from top) molecular-weight RNAs derived from the inverted-repeat (IR) CHS locus in nontransgenic (NT), P38-transgenic (left) or TCV-infected plants (right). (Bottom) Accumulation of TAS255.

Previous studies suggested that DCL2 functions redundantly with at least one other DCL (4), and we envisaged that P38 effectively masked the effects of this second TCV-antagonizing activity. To test this hypothesis, we analyzed P38-expressing Arabidopsis that also expressed a second inverted-repeat (IR) transgene producing siRNAs against the chalcone synthase (CHS) mRNA (Fig. 2C). Transgenic P38 restored CHS accumulation, significantly reduced 21-nt CHS siRNA levels, and triggered accumulation of less abundant, 22-nt siRNAs (Fig. 2C, left panel). This resembled the TRV-PDS siRNA patterns in DCL4-deficient mutants (Fig. 1, D and E), which suggested that P38 suppresses DCL4. Accordingly, endogenous DCL4-dependent ta-siRNAs were specifically lost in P38 plants (Fig. 2C, left). TCV infection recapitulated all these DCL4-suppressing effects (Fig. 2C, right). These data suggested that both DCL4 and DCL2 mediate TCV silencing, with DCL2 providing redundant siRNA-processing functions when DCL4 is suppressed by P38.

DCL4 and DCL2 redundantly silence P38-deficient TCV. To fully appreciate DCL4 effects on TCV silencing, we used a recombinant virus (in vitro transcript inoculum of 500 ng/leaf) in which the green fluorescent protein (gfp) reporter-gene replaced the P38 sequence [TCV-GFPΔP38, Fig. 3A and (16)]. TCV-GFPΔP38 siRNAs accumulated as a 21-nt species in inoculated leaves of WT plants and single dcl mutants, except in dcl4, in which siRNAs were 22 nt long (Fig. 3A). Viral RNA levels were low in dcl2 and dcl3 mutants, but higher in dcl4 and dcl3-dcl4 plants (Fig. 3B), which suggested a greater antiviral contribution of DCL4 alone in TCV-GFPΔP38 than in TRV-PDS infections (Fig. 1D). TCV-GFPΔP38 levels were extremely high in dcl2-dcl4 (Fig. 3B), consistent with results of dcl2-dcl4 and dcl2-dcl3-dcl4 mutants infected with wild-type TCV (Fig. 2B). Although 24-nt siRNAs were undetectable in WT-infected plants, they accumulated in dcl2-dcl4 plants (Fig. 3C), as in TRV-PDS–infected plants (Fig. 1E). Also as with TRV-PDS, very low siRNA levels were detected in dcl2-dcl3-dcl4, likely reflecting inefficient DCL1 activities (Fig. 3C). We conclude that a similar DCL consortium affected siRNA production and virus levels in plants infected with TRV-PDS or TCV-GFPΔP38, and that DCL4 and DCL2 redundantly mediate defense against P38-deficient TCV.

Fig. 3.

(A) Genome organization of TCV-GFPΔP38 and TCV-GFPΔP38 siRNA analysis in dcl mutants (in vitro transcripts; 500 ng/leaf; 7 dpi). (B) TCV-GFPΔP38 RNA accumulation in single and double dcl mutants. The probe was GFP-specific. (C) TCV-GFPΔP38 siRNA analysis in double and triple dcl mutants and P38-expressing plants. (D) TRV-PDS–induced photobleaching occurs in WT, but not P38-expressing plants (14 dpi). (Right) TRV-PDS RNA accumulation. The numbers of infected plants showing photobleaching are from four independent experiments involving three plants each.

The loss of DCL2 had only limited impact on TCV susceptibility (Fig. 2, A and B). Likewise, CHS silencing was released in P38-expressing and TCV-infected plants, despite accumulation of 22-nt siRNAs (Fig. 2C). This suggested that 22-nt siRNAs had suboptimal antiviral activity in the presence of P38. Accordingly, transgenic P38 prevented VIGS, promoted strong disease symptoms and high accumulation of TRV-PDS (Fig. 3D; fig. S2), effects that specifically required the combined inactivation of DCL2 and DCL4 in nontransgenic plants (Fig. 1, E and L). We conclude that besides primary DCL4-antagonizing activities, P38 also likely suppresses the action of DCL2-dependent siRNAs, consistent with major antiviral roles for both enzymes. Further demonstrating the key antiviral functions of DCL4 and DCL2, Cucumber mosaic virus (CMV) levels were three to four times those in dcl2-dcl4 plants compared with WT or single dcl2 or dcl4 mutants. Unlike TCV and TRV, however, CMV silencing was dependent on RDR6 (14), loss of which resulted in CMV levels that were indistinguishable from those in dcl2-dcl4 double mutants (fig. S3).

DCL2-DCL4–dependent block to systemic virus movement. We exploited the GFP tag to follow TCV-GFPΔP38 infection in WT and dcl mutant plants, initially using a moderate in vitro transcript inoculum (500 ng/leaf). TCV-GFPΔP38 moved from cell to cell to form scattered primary lesions in WT-inoculated leaves (Fig. 4A, inlay) but consistently failed to spread systemically and produced no disease symptom (Fig. 4, B and C) (16). Those defects were rescued in P38 transgenic plants (Fig. 4, D to F), which exhibited large, confluent primary lesions in inoculated leaves (Fig. 4D). The aborted systemic spread of TCV-GFPΔP38 could have resulted from its failure to counteract the DCL4-DCL2 antiviral effects, owing to the lack of P38. However, because P38 is also the viral capsid, it could have simply resulted from an inability to form virions, a prerequisite to systemic infection of many plant viruses (17).

Fig. 4.

(A) WT Arabidopsis leaf challenged with moderate TCV-GFPΔP38 inoculum (in vitro transcripts, 500 ng/leaf; 7 dpi). Chlorophyll fluoresces red under UV. (Inlay) Microscopic primary lesion. (B and C) Lack of symptoms (B) and of systemic virus movement (C) in the plant in (A). The numbers of infected plants showing systemic viral GFP are from three independent experiments involving five plants each. (D to F) Same as (A to C), but in P38-expressing plants. (G) GFP silencing in the amplicon (AMP, top) is released by transgenic P38 (AMP-P38) and TCV infection (middle and bottom). (H) Analysis of high- and low-molecular-weight RNAs from nontransgenic (NT), AMP, and AMP-P38 and from m1 and m2 mutants recovered after AMP-P38 mutagenesis. (Bottom) P38 immunoblot analysis. (I) Amino acid substitutions in m1 and m2 alleles of P38. (J) Confluent primary lesions and lack of systemic viral movement in m1 after high-dose inoculation (2.5 μg/leaf; top). TCV-GFPΔP38 is 100% transmitted from m1 to P38-expressing plants through sap extracts (bottom). (K) Same as (A to C) but in dcl4 mutants. (L) Same as (A to C) but in dcl2-dcl4 (14 dpi).

To address these issues, we used the Arabidopsis amplicon line (AMP, in SOM Text) in which replication of a GFP-tagged RNA virus triggers silencing of viral transcripts and low GFP accumulation (Fig. 4, G and H). Transgenic P38 (line AMP-P38) and TCV infection suppressed amplicon silencing [Fig. 4, G (middle and bottom), and H] and inhibited accumulation of endogenous ta-siRNA255 (Fig. 4H). A genetic screen for loss of the AMP-P38 phenotype identified two mutants, m1 and m2, that exhibited reduced amplicon RNA accumulation and GFP expression and also contained WT ta-siRNA255 levels (Fig. 4H). Linkage analysis, DNA sequencing and immunoblot assays revealed that m1 and m2 were stable, point-mutation alleles of P38 (Fig. 4, H and I). The data presented are for m1; similar results were obtained with m2.

Despite extensive cell-to-cell movement in inoculated leaves to levels resembling those of virion-proficient P38 plants (Fig. 4D), long-distance spread of TCV-GFPΔP38 was prevented in plants having m1 or m2 mutations in which the AMP locus had been segregated away (Fig. 4J, top). However, both P38 alleles formed virions, because TCV-GFPΔP38 was transmitted from infected plants with m1 or m2 mutations to healthy, P38-expressing plants through sap inoculation, a procedure whereby viral RNAs that are not encapsidated are rapidly degraded (Fig. 4J, bottom). Thus, m1 or m2 genetically uncoupled the virion-packaging and suppression of silencing functions of P38. Therefore, the lack of systemic TCV-GFPΔP38 P38 movement in WT plants was not caused by its inability to form virions, but rather, by its likely failure to counteract the DCL4-DCL2 antiviral effects. To test this idea, a moderate TCV-GFPΔP38 inoculum (500 ng/leaf) was applied to dcl2, dcl4, and dcl2-dcl4 mutants. Whereas the infection phenotype of dcl2 mutants was indistinguishable from that of WT plants, dcl4 mutants exhibited large primary lesions (Fig. 4K, left) but failed to support systemic movement (Fig. 4K, right). By contrast, there were confluent primary lesions, systemic movement, and disease symptoms in dcl2-dcl4 plants (Fig. 4L). We conclude that, with a moderate inoculum, the combined loss of DCL4 and DCL2 functions was necessary and sufficient to recapitulate the P38-mediated rescue of P38-deficient virus infection.

DCL4 imposes a barrier to vascular exit of TCV-GFPΔP38. The continuum of infection phenotypes in WT, dcl4, and dcl2-dcl4 plants suggested that the primary effect of DCL2 and DCL4 was to restrict the virus to inoculated leaves (Fig. 4, A, K, and L). A fivefold increase in TCV-GFPΔP38 inoculum (5×, 2.5 μg/leaf) restored systemic movement in dcl4-, dcl2-dcl4–, and dcl2-dcl3-dcl4–inoculated plants, but in none of the other single or combination mutants (Fig. 5, A to C). Moreover, primary infection foci in dcl4-inoculated leaves were confluent, as in P38-expressing plants (Fig. 5C). By contrast, and despite formation of macroscopic primary foci, the same (2.5 μg/leaf, Fig. 4D) or much higher-dose (25×, 12.5 μg/leaf) inocula failed to promote systemic movement in WT plants, even though the virus consistently entered the petiole vasculature (Fig. 5, D and E, arrows). Therefore, increasing the inoculum strength could bypass the antiviral effects of DCL2 but not those of DCL4.

Fig. 5.

(A) Systemic (syst) TCV-GFPΔP38 movement in dcl4 plants challenged with high-dose inoculum (5×, 2.5 μg/leaf; inoc). (B) TCV-GFPΔP38 unloading in systemic dcl4 leaves. (C) Confluent primary lesions and uniform GFP distribution (arrow) in petioles of dcl4-inoculated leaves. (D) Same as (C) but in WT plants. (E) The TCV-GFPΔP38 inoculum was five times that shown in (A) (25×, 12.5 μg/leaf). (F) Magnified view of the petiole in (E). Vasc, vascular bundles. (G and H) Transverse sections of (F). (I) Longitudinal section of (G) showing GFP confined into phloem cells. (J) Same as (F) but in dcl4-inoculated leaves. (K) Transverse section of (J). (L) The inverted-repeat (IR) SUC-SUL system. SUC2, phloem-specific promoter. (M) Transgenic P38 expression inhibits SUL silencing movement. (N) Loss of silencing movement in young tissues of TCV-infected SUC-SUL plants. (O) Magnified view of the apex in (N). Arrow, typical sink- (bottom leaf part, containing TCV) to-source (upper leaf part, not containing TCV) transition, showing that suppression of silencing movement is contingent on viral invasion. (P) SUL siRNA analysis in tissues depicted in (O). (Q) Model for dcl4-dependent vascular restriction of TCV-GFPΔP38 and its suppression by P38.

To gain insight into additional antiviral roles of DCL4, TCV-GFPΔP38–inoculated leaves were examined in greater detail. GFP was evenly distributed in petioles of dcl4 plants (Fig. 5C, arrow), whereas in similar WT tissues it appeared as thin stripes (Fig. 5, D and E, arrows), identified as vascular bundles (Fig. 5, F to I). In contrast, TCV-GFPΔP38 could readily exit the vasculature of dcl4 petioles, invading parenchyma and epidermal cells (Fig. 5, J and K). Thus, although the virus penetrated the phloem of WT plants, its unloading was specifically prevented by a DCL4-dependent mechanism. This mechanism was likely not cell autonomous, because it was also manifested in petioles of high-dose inoculated WT leaves (12.5 μg inoculum per leaf) in which virus accumulation was likely as elevated as in dcl4 tissues (Fig. 5, E compared with C).

We recently identified DCL4 as an essential component of non–cell autonomous RNA silencing in a genetic screen using the SUC-SUL system, whereby a vascular-specific IR transgene triggers SUL silencing movement, resulting in vein-centered chlorosis [(7), Fig. 5L]. Of the two SUC-SUL–derived siRNAs species (21-nt and 24-nt; Fig. 5L), only DCL4-dependent 21-nt siRNAs were found required for cell-to-cell signaling (7). We reasoned that the vein-restriction of TCV-GFPΔP38 could have resulted from its failure to suppress the effects of an antiviral, DCL4-dependent silencing signal (Fig. 5Q). If so, P38 and TCV were expected to inhibit vascular exit of silencing signals. Transgenic P38 and TCV infection indeed alleviated silencing movement in SUC-SUL plants (Fig. 5, M to O). Moreover, the 21-nt siRNA levels were reduced in TCV-infected tissues (Fig. 5P), recapitulating the effects of dcl4 knockout mutations that specifically prevented SUL silencing movement (7). Collectively, these results support the ideas that DCL4 exerts some of its antiviral effects by producing a silencing signal that restricts virus exit from vascular bundles, and that vascular production of P38 inhibits this signaling during normal infections (Fig. 5Q).

Conclusions. Although virus-derived dsRNA is accessible to each of the four DCLs under appropriate genetic circumstances, DCL4 and DCL2 exhibit specific, hierarchical antiviral activities. The primary sensor is DCL4, which produces 21-nt siRNAs that guide a virus-antagonizing RISC. Secondarily, DCL2 forms 22-nt siRNAs with antiviral activities, but these are manifested prominently when DCL4 is genetically removed or suppressed. Hyper-susceptibility was only evident when both enzymes were inactivated, which revealed their combined action in defense. Despite detection of DCL3 and, to a limited extent, DCL1 activities on virus-derived substrates, these were not associated with antiviral defense. The hierarchical DCL access to virus-derived dsRNAs is similar to that detected with ta-siRNA precursors: Ta-siRNA biogenesis involves the preferential activity of DCL4, but both DCL3 and DCL2 gain access to precursors in dcl4 mutants (5, 6). Such specificity and hierarchical action possibly reflect contrasted DCL affinities for distinct dsRNA substrates and are likely to apply to viruses impacted by RDR6, such as CMV, and to viruses with DNA genomes (18).

Previous studies suggested that P38 suppresses one DCL (15), and we show that DCL4 is the primary target. However, DCL4 suppression was rescued by DCL2, although the antiviral activity of 22-nt siRNAs was in turn compromised by P38. Have DCL4 and DCL2 evolved primarily for optimal, redundant processing of various forms of pathogenic dsRNA? The existence of endogenous ta-siRNAs and nat-siRNAs that involve DCL4 and DCL2, respectively, complicates this question (5, 6, 8). However, exposure to highly diverse viral suppressors that impact DCL functions may explain why the DCL family has proliferated in plants and why DCL4 and DCL2 diverge in sequence much faster than the miRNA-specific DCL1 (SOM Text). Nevertheless, several silencing suppressors directly sequester DCL products, and these may differently influence evolution of silencing components (19).

Finally, this study indicates that a full appreciation of plant antiviral silencing will require analysis of viral genomes with disabled silencing suppressor functions. This notably uncovered the specific effects of DCL4 on virus vascular exit. The finding that TCV movement required the suppressor function of P38 independently of its virion-packaging function sheds a new light on systemic movement by offering a molecular explanation as to why the bundle sheath–phloem interface usually acts as a key boundary against long-distance transport of most plant viruses (17).

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S3

References and Notes

References and Notes

View Abstract

Navigate This Article