RNA silencing of host transcripts by cauliflower mosaic virus requires coordinated action of the four Arabidopsis Dicer-like proteins

The 35S Leader Is a Major Source of CaMV-Derived sRNAs, Which Accumulate as Discrete, 21- to 22-nt and 24-nt Species in Infected Plants.

Turnip plants (Brassica rapa rapifera) and Arabidopsis ecotype Col-0 were inoculated with CaMV (Fig. 1A). Systemically infected tissues were harvested 21 days postinoculation (dpi), and Northern blot analyses carried out on low-molecular-weight RNA to assay for accumulation of vsRNA. By using a DNA probe corresponding to the full-length viral genome, 24-nt and 21- to 22-nt RNAs were detected in both plants, although the latter RNA species was less abundant in Arabidopsis than in turnip (Fig. 1B). To investigate the origin of those vsRNAs, low-molecular-weight RNA from infected Arabidopsis (21 dpi) was sequentially hybridized with a set of ≈1,000-bp DNA probes covering the entire CaMV genome (Fig. 1C, diagram). This analysis revealed a strong vsRNA signal originating from the 5′ end of the 35S RNA (Fig. 1C, lanes 1a and 1b); a less abundant signal was also detected with a probe corresponding to the 5′ end of the shorter 19S RNA (Fig. 1C, lane 6). Under these hybridization conditions, there were no appreciable signals from any of the remaining probes specific for the other regions of the CaMV genome (Fig. 1C, lanes 2–5, 7, and 8).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

CaMV infection in B. rapa and Arabidopsis triggers accumulation of vsRNAs derived from the 35S RNA leader sequence. (A) Organization of the double-stranded circular DNA genome of CaMV and the two 35S and 19S viral RNAs. The leader sequence is in the 5′ end of the 35S RNA. (B) RNA blot analyses of low-molecular-weight RNAs from CaMV-infected plants. The probe used covered the entire CaMV genome. 5S, ethidium bromide staining of 5S ribosomal RNA. (C) Schematics of the 35S and 19S viral RNAs and RNA blot analyses of low-molecular-weight RNAs derived from different regions of the CaMV genome (nucleotide coordinates are indicated) by using DNA probes. Only a 1,500-nt fragment between nucleotides 4700 and 6264 was not probed. The probe against the leader sequence (1b) allowed detection of mainly 24-nt vsRNAs at short exposures (≈2 h). Longer exposures (≈20 h) revealed 21-nt vsRNA accumulation. (D) Predicted strong secondary structure of the 35S RNA leader (Vienna RNA folding package). Col-0, Columbia.

The strong vsRNA signal from the 5′ end of the 35S RNA likely originated from the extensive fold-back structure of the 35S leader (Fig. 1D). To test this idea, we used an oligonucleotide probe (1a) spanning 40 nt located just upstream of the stem-loop leader structure or a 600-nt DNA probe (1b) exactly spanning the entire stem-loop sequence. The vsRNA signal was detected only with the 600-nt probe (Fig. 1C). As observed with the full-length CaMV DNA probe, the 24-nt siRNA species detected with the leader-specific probe was prominent compared with the 21- to 22-nt species, which was nevertheless evident after longer exposures (Fig. 1C, lane 1b lower). These results identify the translational leader as a major source of CaMV-derived sRNAs and suggest that this stem-loop is normally processed by distinct DCL activities.

DCL2, DCL3, and DCL4, but Not DCL1, Efficiently Process the 35S Leader into Specifically Sized vsRNAs.

To characterize the genetic requirements for biogenesis of leader-derived vsRNAs, we did Northern blot analyses of low-molecular-weight RNAs extracted from CaMV-infected Arabidopsis carrying dcl knockout mutations (Col-0 ecotype, 21 dpi). We initially used the same leader-specific probe as used in Fig. 1C. The prominent 24-nt siRNA species was as abundant in dcl2 and dcl4 mutants as in WT plants. However, it was lost in dcl3 mutants, in which only 21-nt siRNA accumulated (Fig. 2A). The vsRNA pattern was unaltered in rdr2 and rdr6 mutants (Fig. 2A), suggesting a limited contribution of de novo dsRNA synthesis to vsRNA accumulation, in agreement with the extensive fold-back structure of the 35S leader (Fig. 1D).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

DCL dependencies of vsRNA patterns and CaMV accumulation in Arabidopsis silencing mutants. (A) Northern blots of low-molecular-weight RNA accumulating in CaMV-infected silencing mutants of Arabidopsis. Probe was specific for the entire leader sequence. The numbers of inoculated plants showing systemic infection are from two separate experiments involving five plants each. (B) Same as in A, but the probe was an end-labeled oligonucleotide corresponding to the sRCC1 vsRNA (sequence on top) identified through the procedure described in Fig. 3. (C) Same as in B, but in the dcl1–9 and hen1 mutants. (D) Same as in C, but with two independent alleles of dcl1. The blot was exposed longer to allow detection of vsRNA accumulation (Lower). (E) Immunoblot of CaMV CP (a product of the 35S RNA) accumulation in various RNA silencing mutants and combinations thereof. Proteins were extracted from the samples used in A–D. 5S, ethidium bromide staining of 5S ribosomal RNA; Tot prot: total protein revealed by Coomassie staining. (F) Northern blot analysis of high-molecular-weight RNA extracted from the samples used in A–D. (Right) Composite comparing effects of dcl1–9 and triple dcl2-dcl3-dcl4 mutations.

The very strong 24-nt siRNA signal from the leader-specific probe hampered further investigation of the 21-nt siRNA species. To solve this problem and improve vsRNA resolution, similar analyses were performed with a 20-nt probe specific for a small region of the 35S leader, processed as a discrete vsRNA species. The procedure for identifying this vsRNA, coined “sRCC1” (Fig. 2B, diagram), is described below. As shown in Fig. 2B, the sRCC1 probe detected a 24-nt RNA signal and a less abundant, 21-nt RNA signal in WT plants. The vsRNA patterns in various dcl and rdr mutants were similar to those detected with the probe spanning the entire leader sequence (Fig. 2, compare A with B). To further dissect vsRNA biogenesis, these experiments were repeated in double and triple dcl mutants. These analyses confirmed that biogenesis of 24-nt vsRNA was DCL3 dependent, because this species was lost in dcl3, dcl2-dcl3, dcl3-dcl4, and dcl2-dcl3-dcl4 mutants (Fig. 2B). The 21-nt siRNA accumulation was lost in dcl4 and dcl2-dcl4 mutants, suggesting that this RNA species is processed by DCL4 (Fig. 2B). Analysis of dcl3-dcl4 mutants showed strong accumulation of 22-nt RNAs, also detectable in infected WT Col-0 plants (Fig. 2B). Synthesis of the 22-nt RNA likely required DCL2 because it was lost in dcl2-dcl4, dcl2-dcl3 double mutants and in dcl2-dcl3-dcl4 triple mutants, the latter accumulating very low levels of 21-nt vsRNA (Fig. 2B). This residual processing of vsRNA could be explained by the inefficient action of DCL1 (the only active DCL in the triple dcl2-dcl3-dcl4 knockout) or by suboptimal activities of other Arabidopsis RNase-III, as proposed recently (19). These vsRNA accumulation patterns and DCL dependencies were not specific to sRCC1, because they were also detected with other 20-nt probes corresponding to other regions of the leader (Fig. 3A and data not shown). We conclude that, in natural infection contexts, processing of the 35S RNA leader results from the combined action of DCL2, -3, and -4 producing 22-, 24-, and 21-nt vsRNAs, respectively.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

The leader-derived sRCC1 vsRNA down-regulates accumulation of an Arabidopsis transcript and acts as a bona fide siRNA in infected turnip. (A) Discrete, leader-derived sRNAs, as detected by Northern blot of low-molecular-weight RNA from CaMV-infected Arabidopsis. Probes used were oligonucleotides complementary to indicated sequences. sRNAs in black were not found to exhibit significant complementarity to Arabidopsis transcripts as assessed by BLAST searches. sRNAs in blue were found to exhibit near-perfect complementarity to Arabidopsis transcripts (BLAST search results only). (B) Schematic of predicted sRCC1 target position in At1g76950. (C) sQPCR analysis of At1g76950 transcript accumulation in infected (I) and mock-inoculated (M) Arabidopsis. No RT, control reaction performed without reverse transcriptase; UBI, amplification control using ubiquitin-specific primers. (D) sRCC1 accumulation as assessed by Northern blot of low-molecular-weight RNA in CaMV-infected and mock-inoculated turnip leaves. (E) Schematic of sRCC1 sensor transgene constructs and their transient expression in CaMV-infected and noninfected (Mock) turnip leaves at 15 dpi. LB and RB, left and right T-DNA borders, respectively. (F) Immunoblot analysis of GFP accumulation in the tissues depicted in E. e.v., empty vector used as a negative control in infiltrations.

DCL1, but Not HEN1, Facilitates the Accumulation of All Species of Leader-Derived vsRNA.

We tested vsRNA accumulation in plants compromised for miRNA biogenesis/accumulation, employing alleles of the dcl1 and hen1 mutations, respectively [Lansberg erecta (Laer) ecotype, 28 dpi]. The levels of the 21- to 22- and 24-nt species of sRCC1 were reduced in the dcl1–9 allele (Fig. 2C). By contrast, both vsRNA species accumulated normally in the hen1 mutant, as compared with WT (Fig. 2C). dcl1–9 is a T-DNA insertion mutant in the RNA-binding domain of DCL1 (20). To rule out that reduction in vsRNA levels was a peculiarity of this allele, a missense allele of DCL1, dcl1–7, was assayed. dcl1–7 carries a Pro-to-Ser substitution within the DExH motif (21). As shown in Fig. 2D Upper, the sRCC1 levels were similarly decreased in dcl1–9 and dcl1–7 plants. Longer exposures showed that 21-, 22-, and 24-nt vsRNA were still detectable in both dcl1 backgrounds (Fig. 2D Lower), agreeing with the partial loss-of-function nature of the mutations. The difference between the effects of hen1 and dcl1 on vsRNA levels was not peculiar to sRCC1; it was also observed with a probe spanning the entire leader and with other 20-nt probes specific for other leader-derived vsRNAs (Fig. 3A and data not shown). We rule out that the particular effect of dcl1 results from pleiotropy because the hen1 mutation also causes strong developmental anomalies that overlap with those of dcl1 (22). These results suggest that DCL1, although not directly involved in the processing of leader-derived vsRNA (Fig. 2B, track dcl2-dcl3-dcl4), specifically facilitates their biogenesis by the three other DCLs.

Effects of dcl Mutations on CaMV Accumulation.

We sought to determine the effects of single and combined dcl mutations on CaMV accumulation. We monitored viral particle formation in immunoblot analyses of the 35S RNA-encoded coat protein (CP). Viral titers (21 dpi) were not significantly different between WT plants and single dcl2, dcl3, and dcl4 mutants (Fig. 2E Upper Left); the same observation was made in double dcl2-dcl3, dcl2-dcl4, and dcl3-dcl4 mutants (Fig. 2E Upper Right). However, the virion levels were higher in triple dcl2-dcl3-dcl4 mutants (Fig. 2E Upper Right). Comparisons of WT Laer plants and hen1 and dcl1 mutants (28 dpi) did not reveal any significant change in CaMV titers (Fig. 2E Lower). Northern blots of RNA extracted from the same tissues showed that 35S RNA levels were significantly higher in dcl2-dcl3-dcl4 mutants, whereas they remained similar to those of WT infected plants in all other single and combination dcl mutants (Fig. 2F and data not shown). We conclude from those analyses that the combined inactivation of DCL2, -3, and -4, which causes a near-complete loss of vsRNA processing (Fig. 2B), is necessary to promote hypersusceptibility to CaMV in Arabidopsis. By contrast, loss of DCL1 activity, which only partially reduces vsRNA accumulation (Fig. 2D), is not sufficient to promote hypersusceptibility to the virus.

A Leader-Derived vsRNA Induces Down-Regulation of an Endogenous Gene Belonging to the Regulator of Chromosome Condensation (RCC1) Family.

Several mammalian DNA viruses were shown to produce sRNAs from discrete, genomic stem-loop structures (reviewed in ref. 23). Those molecules appear to play important roles in cis by regulating viral genome expression, and, in at least one example, they were also shown to inhibit expression of host transcripts in trans (24). To test whether, likewise, some CaMV-derived vsRNA could inhibit expression of specific host transcripts, we subjected the entire 35S RNA leader sequence to a BLAST search against Arabidopsis cDNAs and ESTs, looking for possible 18- to 25-nt microhomologies. By using a procedure allowing a maximum of two mismatches, three Arabidopsis transcripts were retrieved (Fig. 3A), among which was the mRNA for At1g76950. The predicted product of At1g76950 awaits functional characterization, but it contains a zinc-finger motif and a domain characteristic of the members of the RCC1 family of proteins that bind to chromatin and interact with the GTP-binding factor RAN. Expression profiling indicates that At1g76950 is mainly expressed in flower buds (TAIR gene expression resource). Sequence alignment revealed a near-perfect complementarity between the 5′ UTR of At1g76950 and a 20-nt sequence located in the descending arm of the 35S leader (Fig. 3A, section 4, and B). Using a complementary oligonucleotide probe, we confirmed, by Northern blot analysis, the accumulation of a corresponding vsRNA (sRCC1) in CaMV-infected but not in mock-inoculated Arabidopsis (Fig. 2 B–D and Fig. 4A). The sequence alignment (Fig. 3B) suggested that sRCC1 could act as a bona fide siRNA to promote cleavage of At1g76950 transcripts during infections. Semiquantitative RT-PCR (sQPCR) analyses indeed revealed that At1g76950 accumulation was dramatically reduced in CaMV-infected but not in mock-inoculated Arabidopsis (Fig. 3C).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Tens of Arabidopsis transcripts are potentially targeted by sRCC1 and other leader-derived vsRNAs. (A) sRCC1 accumulation in infected tissues of the At1g76950 T-DNA insertion mutant, as assessed by Northern blot of low-molecular-weight RNA. The last panel shows CaMV CP accumulation assessed by immunoblot. (B) sQPCR analysis of At1g75330 and At3g52500 transcript levels in leaves of CaMV-infected (I) and mock-inoculated (M) plants at 21 dpi (33 amplification cycles, Upper). (Lower) QPCR analysis of At1g75330 transcript levels in WT and dcl2-dcl3-dcl4 mutant backgrounds. For each sample, mRNA levels were normalized to that of Actin2 (At3g18780). Error bars represent standard deviation from three independent experiments involving triplicate PCRs. (C) The sRCC1 sequence was queried for complementarity to Arabidopsis transcripts by using miRU settings with decreasing stringencies. Mism, mismatch; G:U, G:U wobbles allowances. The transcript sets retrieved under high stringency conditions were always contained within those obtained under lower stringency settings. The target transcript number reached a plateau of 90. Sequence alignments and identities of predicted targets are in Table 1 and Data Set 1, which are published as supporting information on the PNAS web site.

To validate that sRCC1 could directly guide sequence-specific transcript cleavage in natural infection contexts, GFP-sensor constructs were engineered in which the 3′ UTR contained either the authentic target sequence found within At1g76950 or a modified version carrying five nucleotide substitutions predicted to abolish annealing and cleavage by sRCC1 (Fig. 3E). The constructs were delivered by Agrobacterium leaf infiltration into CaMV-infected turnip plants in which accumulation of sRCC1 was confirmed at 15 dpi (Fig. 3D). Five days after agroinfiltration, GFP expression was monitored under UV light, and its accumulation was quantified by immunoblot analyses (Fig. 3 E and F). These levels were consistently higher with GFP-RCC1mut, carrying mismatches, than with GFP-RCC1wt, bearing the authentic sRCC1target (Fig. 3F Left). Both constructs were similarly expressed in uninfected turnip plants (Fig. 3F Right). Addition of the sRCC1-complementary region of At1g76950 was sufficient to promote sequence-specific down-regulation of a reporter mRNA during CaMV infection. These findings strongly suggest that reduction of At1g76950 levels in Arabidopsis results from direct action of the CaMV-derived sRCC1.

Leader-Derived sRNAs Potentially Down-Regulate a Large Number of Arabidopsis Transcripts.

To examine whether down-regulation of At1g76950 affected virus accumulation, we infected a T-DNA insertion mutant of At1g76950 with CaMV. The infection kinetics and CaMV accumulation (21 dpi) were unchanged in the mutant compared with WT-infected plants, as were vsRNA levels (Fig. 4A and data not shown). Therefore, the At1g76950 knockout did not impact the infection, suggesting that either targeting of At1g76950 by sRCC1 had no incidence (at least under laboratory conditions) or there is functional redundancy among At1g76950 homologs. Another possibility was that sRCC1 and other vsRNAs normally target other host genes in addition to At1g76950, so that their combined inactivation would have been required to appreciably affect viral titers. To test this idea, we subjected the deduced 20-nt sequence of sRCC1 to the miRU algorithm, designed to identify putative Arabidopsis target transcripts of queried, 19- to 28-nt RNAs. miRU takes into account the thermodynamics and 5′ seed-pairing requirements for optimal miRNA and siRNA activities in plants (25). By allowing a maximum of two nucleotide mismatches and only one possible G:U wobble, an additional set of 14 putative sRCC1 mRNA targets was retrieved (see Table 1), with At1g75330 and At3g52500 (encoding, respectively, an ornithine carbamoyltransferase/ornithine transcarbamylase and an aspartyl protease) ranking the highest scores. sQPCR analyses indicated that both transcripts were down-regulated in CaMV-infected but not mock-inoculated Arabidopsis (Fig. 4B Upper). This down-regulation likely resulted from the specific effects of CaMV-derived siRNAs (i.e., sRCC1) because quantitative RT-PCR (QPCR) analyses showed that the decrease in At1g75330 transcript levels was much less pronounced between infected and noninfected dcl2-dcl3-dcl4 mutant plants than between infected and noninfected WT plants (Fig. 4B Lower), consistent with the fact that accumulation of leader-derived vsRNA is strongly reduced in dcl2-dcl3-dcl4 (Fig. 2B).

The full spectrum of host mRNA potentially targeted by sRCC1 was revealed by using miRU searches with increasing tolerance for base mismatches and G:U wobbles, giving a total of 90 candidates, 65% of which carried the sRCC1 target sequence within 5′ or 3′ UTRs (Fig. 4C, Table 1, and Data Set 1). Gene ontology analysis showed that the majority of identified targets are involved in basic cell metabolism, with a preponderance of chloroplast-targeted products (Fig. 5, which is published as supporting information on the PNAS web site). A similar gene set was retrieved if the queried sRCC1 sequence had an additional 3 nt on either 5′ or 3′ ends, taking into account the fact that the sRNA also accumulates as 22- and 24-nt species (data not shown). Further miRU analyses (low stringency settings) identified >30 potential hits with similar ontology and UTR biases for vsRNA no. 2; there were no or few hits with the other vsRNAs detected by Northern blot (Fig. 3A and data not shown). We conclude that discrete vsRNA species derived from the 35S RNA leader may account for down-regulation of large numbers of Arabidopsis transcripts through 20- to 24-nt stretches of sequence complementarity.