N. benthamiana plants were grown in a greenhouse at 25°C with 16-h light/8-h dark cycles. rdr1/6 and wild-type Arabidopsis plants were grown in a greenhouse at 22°C with 16-h light/8-h dark cycles.

35S-R1, 35S-R2, 35S-R3, and 35S-RΔ2b(R2aΔ2b) were described in a previous study (Hou et al., 2011). For R2a2b_(1–76), R2a2b_(8–111), and R2a2b_(18–111) point mutant constructs, mutagenesis was introduced using QuikChange® Lightning Site-Directed Mutagenesis Kits (Agilent Technologies, 210518) according to the manufacturer's instructions. The templates were 35S-R2, 35S-R2, and R2a2b_(8–111), and the primer pairs were R2-2b77TGAF/R, R2-2b8-111F/R, and R2-2b18-111F/R, respectively (Table S1).

For the constructs used in dsRNA binding activities, pGEX-4T-2-SD2b was described in a previous study (Duan et al., 2012). The constructs pGEX-4T-2-SD2b_(8–111) and pGEX-4T-2-SD2b_(18–111) were generated with the pGEX-4T-2-SD2b template using QuikChange® Lightning Site-Directed Mutagenesis Kits (Agilent Technologies, 210518) according to the manufacturer's instructions. The primer pairs were GST2b8-111F/R and GST2b18-111F/R for the constructs pGEX-4T-2-SD2b_(8–111) and pGEX-4T-2-SD2b_(18–111), respectively (Table S1).

For the constructs used in suppression activity, pBI121-35S-SD2b was described in a previous study (Duan et al., 2012). To generate pBI121-35S-SD2b_(8–111), pBI121-35S-SD2b_(18–111), and pBI121-35S-SD2b_(1–76), the 2b8-111F/2bR, 2b18-111F/2bR, and 2b1-76F/R primers (Table S1) were used to amplify the 2b_(8–111)/2b_(18–111)/2b_(1–76) mutant fragments with the template of pBI121-35S-SD2b; the resulting fragments were cut by XbaI-SacI and inserted into the XbaI-SacI digested pBI121-35S-SD2b vector to yield pBI121-35S-SD2b_(8–111), pBI121-35S-SD2b_(18–111), and pBI121-35S-SD2b_(1–76).

For constructs used in the subcellular localization and SD2b-AGO colocalization, pBI121-35S-SD2b-EGFP and RFP-AGO4 were described in a previous study (Duan et al., 2012). The constructs pBI121-35S-SD2b_(8–111) and pBI121-35S-SD2b_(18–111)-EGFP were generated with the pBI121-35S-SD2b-EGFP template using QuikChange® Lightning Site-Directed Mutagenesis Kits (Agilent Technologies, 210518) according to the manufacturer's instructions. The primer pairs were 2B8-111-EGFPF/R and 2B18-111-EGFPF/R for the pBI121-35S-SD2b-EGFP and RFP-AGO4 constructs, respectively (Table S1).

35S-R1, 35S-R3 and different R2 mutant constructs were co-infiltrated into the leaves of 5-week-old N. benthamiana plants, as described in a previous study (Hou et al., 2011). Systemically infected leaves were harvested from pools of 15 to 20 plants for sap extraction for viral infection. Similar levels of viral RNAs in each sap sample estimated by RNA gel blotting were inoculated to N. benthamiana and A. thaliana seedlings.

Plant total RNA used for RNA gel blotting was extracted by the hot-phenol method as previously described (Fernández et al., 1997). For the detection of viral RNAs, three 1-kb fragments at the 3′-terminus of each cDNA clone (35S-R1, 35S-R2, and 35S-R3) were amplified, which were then labeled with [a-³²P] dCTP using the Rediprime II system (GE Healthcare, RPN1633) and were mixed as probes. For the detection of siRNAs, 30 mg of total RNA was separated on 17% polyacrylamide-8 M urea gels. The probes were labeled with [r-³²P]ATP using T4 PNK (NEB, M0201V). VsiRNAs were detected using mixtures of labeled DNA oligonucleotides specific to RNA3. Signal intensity was quantified using ImageQuant TL software (GE Healthcare).

For the expression of fusion proteins in Escherichia coli, recombinant plasmids were transformed into BL21 cells and induced at 0.3 mM isopropyl b-D-1thiogalactopyranoside (Sigma-Aldrich) in Luria-Bertani medium at 28°C for 3 to 6 h. GST-tagged fusion proteins were purified using Glutathione Sepharose 4B (GE Healthcare, 17-0756-01) according to the manufacturer's instructions.

21/24-bp ds-siRNA and 55-bp ds-RNA, which were described in a previous study (Duan et al., 2012), were radiolabeled in 50-pmol quantities with 0.3 mM [r-³²P]ATP and 20 units of T4 PNK (NEB, M0201V). Binding reactions were performed with 1 ng of radiolabeled ds-siRNA and 1 nmol of protein in binding buffer (20 mM Tris-HCl, pH 7.5, 5 mM MgCl2, 300 mM NaCl, 0.1% Nonidet P-40, and protease inhibitor cocktail. After 40 min at room temperature, 1 mL of 50% glycerol and dye was added, and protein RNA complexes were resolved on 6% native polyacrylamide gel. The gels were then exposed to a storage phosphor screen (GE Healthcare).

The subcellular localization of 2b mutants and their colocalization with AGO proteins were determined by infiltrating binary plasmids of pBI121-35S-SD2b-EGFP mutants and RFP-AGO4 into 5-week-old N. benthamiana leaves, which were maintained for 3 days at 25°C (16-h light/8-h dark). The nuclei were stained with 100 ng/mL 4′,6-diamidino-2-phenylindole (DAPI) for 10 min before confocal microscopy. Confocal fluorescence of GFP, RFP, and DAPI were captured with a Leica SP8 confocal microscope.

We previously constructed infectious clones of SD-CMV with the viral genomic RNA1, RNA2, and RNA3 under the 35S promoter (Figure 1A) and a chimeric RNA2 (Δ2b) infectious clone, in which 2b protein expression was abolished by nucleotide substitution in the start codon ATG, as well as four other ATG codons in the 2b coding sequence, but 2a protein expression was unaffected, designated as CMVΔ2b (Figure 1A; Hou et al., 2011). To investigate the biological functions of the different biochemical properties of the 2b protein in the context of CMV infection in plants, in this study, we further created mutations of 2b of the genomic RNA2 according to the two main biochemical properties of double-stranded RNA (dsRNA) and AGO binding activities (Duan et al., 2012). As shown in Figure 1A, in addition to the above Δ2b, three 2b mutants were created by deleting the C-terminal 35 amino acids (aa) and creating a stop codon in the 2b coding sequence to yield 2b_(1–76) without affecting the overlapping portion of the 2a polymerase. Another two mutants were created by deleting the N-terminal 7 or 17 aa to yield 2b_(8–111) and 2b_(18–111) by nucleotide substitution in the 1st or both the 1st and 8th “ATG” codons of the 2b coding sequence without affecting the 2a protein. To obtain viral sources of chimeric CMV with different 2b mutant, each of these constructs was transformed into Agrobacterium for the infiltration of N. benthamiana in the presence of 35S-RNA1 and 35S-RNA3 to examine the infectious properties of the chimeric CMV with 35S-2bx (x represents different 2b mutations shown in Figure 1A), and the related viruses were referred to as wild-type CMV, CMVΔ2b, CMV2b_(1–76), CMV2b_(8–111), and CMV2b_(18–111) (Figure 1B and Supplementary Figure S1).

Typical symptoms were observed in CMV-, CMV2b_(1–76)-, and CMV2b_(8–111)-inoculated plants, whereas CMVΔ2b- and CMV2b_(18–111)-inoculated plants did not develop visible symptoms (Supplementary Figure S1). However, as shown in Figure 1B, RNA gel blot analysis confirmed that the three CMV genomic RNA1, RNA2, and RNA3 transcripts, as well as both subgenomic RNA4 and 4A transcripts, were accumulated in non-inoculated systemic leaves for all chimeric viruses detected using the 3′-untranslated region (3′-UTR) as a probe (Figure 1B). This indicated that the infiltration of the mixture could sustain the replication of both subgenomic RNA4 and 4A. RT-PCR and sequencing analysis of viral RNAs isolated from infected plants confirmed that all chimeric 2bx mutations were genetically stable in N. benthamiana (Nb) plants. Similar symptom development and the accumulation of viral RNAs were obtained in the Nb plants inoculated with sap extracted from these each chimeric CMV-infected Nb leaves (Figure 1B).

We previously uncoupled the 2b domain requirements for dsRNA binding and nucleolar targeting from the physical interaction with AGO proteins (Duan et al., 2012). The 61 aa N-terminal end [2b_(1–61)], which contains the complete α1-linker-α2 structure involved in dsRNA binding, retained the wild-type 2b ability to bind 21- and 24-nt siRNA duplexes, whereas 2b_(13–111) exhibited weak affinity for 21-nt siRNA but showed no detectable affinity for 24-nt siRNA (Duan et al., 2012). To characterize the dsRNA binding activities of 2b_(8–111) and 2b_(18–111), which corresponded to the deletion mutants constructed in the context of the CMV RNA2 genome (Figure 1A), we performed electrophoretic mobility shift assays (EMSAs) with the full-length of 2b, 2b_(8–111), and 2b_(18–111) expressed and purified as GST fusion proteins (Figures 2A,B). Similar to previous results (Duan et al., 2012), 2b exhibited high affinity for either small 21- and 24-nt siRNA duplexes or long dsRNA (Figure 2B). 2b_(8–111) was almost as active as 2b in binding to the 21- and 24-nt siRNA duplex but exhibited reduced affinity for long dsRNA (Figure 2B). Deletion of 17 aa from the N terminus abolished both of the siRNA dsRNA binding activities (Figure 2B), which indicated that further deletion of 4 aa (from 13 to 17 aa) in the N-terminal α1 helix completely abolished the weak affinity for 21-nt siRNA of 2b_(13–111). Consistently, 2b_(13–111) retained the partial silencing suppression activity detected using an Agrobacterium coinfiltration assay (Duan et al., 2012); however, 2b_(18–111) failed to suppress GFP silencing as indicated by the lack of green fluorescence in the co-infiltrated leaves, in which 2b_(8–111) was almost as active as 2b in the suppression of GFP silencing (Figure 2C).

We next investigated the subcellular localization of 2b_(8–111) and 2b_(18–111) and their possible interaction with the AGO protein in N. benthamiana leaf epidermal cells. 2b_(8–111) and 2b_(18–111) fused at their C-termini with enhanced green fluorescent protein (EGFP) were transiently expressed in N. benthamiana via Agrobacterium-mediated infiltration. The 2b-EGFP (Duan et al., 2012) was used as a control. Similar to 2b-EGFP, both 2b_(8–111)-EGFP, and 2b_(18–111)-EGFP were mainly detected in the nucleus with dense fluorescence in the nucleoli (Figure 3A), consistent with both deletion mutants containing NoLS from the 13 to 37 region, including both NLSs (Figure 3C). This allowed for the accumulation of the fusion protein in the nucleoli and nucleus (Duan et al., 2012).

Coexpression of 2b-EGFP and RFP-AGO4 resulted in their colocalization in the nucleus with dense fluorescence in nucleus-associated bodies (Figure 3B), which was consistent with our previous finding that the redistribution of 2b-EGFP in the nucleus when coexpressed with AGO proteins (Duan et al., 2012). Similar nuclear colocalization and redistribution was observed for 2b_(8–111)-EGFP, but not 2b_(18–111), when they were coexpressed with RFP-AGO4 (Figure 3B). Intriguingly, the densities of fluorescence were greatly reduced for both 2b_(18–111)-EGFP and RFP-AGO4 when they were coexpressed compared to that of each when expressed alone (cf. Figures 3A,B). One of the possible explanations might be that the interaction of 2b_(18–111)-EGFP and RFP-AGO4 caused the instability of both proteins due to 2b_(18–111)'s lack of binding to ds-siRNA and suppression of silencing (Figures 2B,C). Nevertheless, these data demonstrate that the deletion of 17 aa from the N terminus of the 2b protein completely abolished both the long dsRNA and ds-siRNA binding activity, whereas the deletion of 7 aa from the N terminus retained the wild-type 2b abilities in binding both ds-siRNA and AGO protein but reduced the affinity for binding long dsRNA compared to wild-type 2b.

To investigate the biological functions of the different biochemical properties of the 2b protein in the context of CMV infection in plants, wild-type Arabidopsis plants were inoculated with sap extracted from the above N. benthamiana leaves infected with CMV, CMVΔ2b, CMV2b_(1–76), CMV2b_(8–111), or CMV2b_(18–111). The development of disease symptoms in Arabidopsis plants was monitored. At 9 days post-inoculation (dpi), plants infected with wild-type CMV displayed severe developmental defects, including reduced leaf size and a shortened petiole, and all new leaves were aggregatecd in the center of the plants as observed at 21 dpi (Figure 4A). Plants inoculated with CMVΔ2b displayed no symptoms and normal growth compared to mock infection plants. CMV2b_(1–76)-infected plants displayed severe developmental defects similar to CMV-infected plants, and the development of both CMV- and CMV2b_(1–76)-infected plants was arrested with short and defective inflorescence at 21 dpi (Figure 4A). CMV2b_(8–111)-infected plants also showed typical defects in development and inflorescence, albeit less stunting. However, CMV2b_(18–111)-infected plants exhibited very mild symptoms, and plant growth was not arrested (Figure 4A). RT-PCR and sequencing analysis of viral RNAs isolated from infected plants confirmed that all chimeric 2bx mutations are genetically stable in Arabidopsis.

RNA gel blot analysis confirmed that three CMV genomic (RNA1, RNA2, and RNA3) and both subgenomic (RNA4 and 4A) transcripts were accumulated in the systeimic leaves of these infected plants (Figure 4B). Consistent with the degree of disease, minimal and small quantities of viral RNAs in CMVΔ2b- and CMV2b_(18–111)-infected plants were detected (Figure 4B). Intriguingly, similar severities of disease symptoms were observed for CMV-, CMV2b_(1–76)-, and CMV2b_(8–111)-infected plants; however, the level of viral RNAs in CMV2b_(1–76)-infected plants was obviously lower than that in CMV- and CMV2b_(8–111)-infected plants (Figure 4B). 2b_(1–76) lacked the AGO binding domain but retained the dsRNA-binding domain and the silencing suppression activity (Figure 4C). Therefore, we examined the production of viral siRNAs (vsiRNAs) in these infected plants. High levels of vsiRNAs were detected in CMV2b_(1–76)-infected plants compared to wild-type and those infected with other chimeric viruses (Figure 4B), which was consistent with the low level of viral RNAs in these plants. No major differences in the accumulation of miR173 were detected following infection with either wild-type or each mutant CMV, which supported an earlier observation that CMV infection does not alter miRNA accumulation (Diaz-Pendon et al., 2007). These results demonstrate that N terminal dsRNA binding activity is responsible for the induction of the virulence of CMV, which does not necessarily correlate with the accumulation of viral RNAs, and 2b-AGO binding is likely required for CMV to suppress the silencing of viral RNAs in Arabidopsis plants.

Previous studies showed that the 2b gene of the CMV Fny strain is required for interference with RDR1- and RDR6-dependent antiviral silencing (Diaz-Pendon et al., 2007; Dong et al., 2016). To examine whether the accumulation of CMV-derived vsiRNAs in CMV2b_(1–76)-infected plants was dependent on RDR proteins, we inoculated Arabidopsis with double mutants of RDRs (RDR1 and RDR6) with wild-type CMV, CMVΔ2b, CMV2b_(1–76), CMV2b_(8–111), or CMV2b_(18–111). Trans-acting siRNA tasiR255 was absent in the tested rdr1/6 plants, verifying that the RDR mutant alleles (Figure 5B). CMVΔ2b remained defective in inducing virulence in rdr1/6 mutant; CMVΔ2b-infected plants displayed no symptoms and normal growth compared to mock infection plants (Figure 5A). Similar to infected wild-type Arabidopsis plants, rdr1/6 plants infected with CMV-, CMV2b_(1–76)-, and CMV2b_(8–111) displayed severe developmental defects, and whole plant development was arrested, although less stunting was observed in CMV2b_(8–111)-infected rdr1/6 plants (Figure 5A). CMV2b_(18–111)-infected rdr1/6 mutant plants exhibited mild but clear disease symptoms (Figure 5A). The inflorescences of CMV2b_(18–111)-infected rdr1/6 plants were shorter and defective compared to those of CMV2b_(18–111)-infected wild-type plants (Figure 4A), which demonstrated that 2b_(18–111) retained partial activity in the suppression of RDR6- and/or RDR1-mediated antiviral defense.

We further conducted RNA gel blot hybridizations to compare the accumulation of viral RNAs in these infected rdr1/6 mutant plants. Consistent with the asymptomatic phenotype, minimal levels of viral RNAs in CMVΔ2b-infected plants were detected (Figure 5B). Interestingly, a similar severity of disease symptoms was observed for CMV-, CMV2b_(1–76)-, and CMV2b_(8–111)-infected rdr1/6 plants; however, the accumulation level of viral RNAs in CMV2b_(1–76)-infected plants was clearly higher than that in CMV- and CMV2b_(8–111)-infected mutant plants (Figure 5B), opposite of that detected in infected wild-type Arabidopsis plants, in which a lower level of viral RNAs accumulated in CMV2b_(1–76)-infected plants than in CMV- and CMV2b_(8–111)-infected plants (Figure 4B).

We then examined the production of vsiRNAs. Remarkably, vsiRNAs in CMV2b_(1–76)-infected rdr1/6 plants (Figure 5B) were greatly reduced compared to CMV2b_(1–76)-infected wild-type Col-0 plants (Figure 4B). These results indicated that RDR6 and/or RDR1 play role(s) in partially silencing CMV RNAs disrupted by 2b-AGO binding activity. Taking into account of the low level of viral RNAs and the high level of vsiRNAs in CMV2b_(1–76) infected wild-type plants, our data demonstrate that the 2b-dsRNA binding activity is insufficient to suppress the host degradation of viral RNA that requires the functions of RDR6 and/or RDR1.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.