2b has NLS and NES motifs for nucleocytoplasmic translocation

To search for NES motif(s) in 2b, we used the NetNES v1.1 online program and found an NES motif in the C-terminal half (S1A Fig). For the analysis, 2bs of three CMV subgroup I strains, CMV-Y (Y2b), CMV-CM95R (R2b) and CMV-HL (HL2b), and a chimeric 2b, Y/HL2b (chimeric 2b between CMV-Y and CMV-HL), were used [22,43,44,45]. As shown in S1B Fig, a leucine residue (L87) of HL2b, R2b and Y/HL2b has a significantly high NES score, and a motif of residues from L77 to L87 has high values of hidden Markov model (HMM), a tool for protein domain identification [46], indicating that the residues from L77 to L87 may function as an NES. Unexpectedly, only Y2b contains a serine (S77) instead of a leucine, and its NES score was not significantly high (S1B Fig).

To verify the role of the predicted NES for 2b’s nuclear export, we generated a 2b mutant by substituting three leucine residues (L77, L85 and L87) with alanine residues (L77/85/87A) based on R2b of the CMV-CM95R strain because the leucine residues generally carry a role in NES by associating with AtXPO1 [12]. To investigate the effect of those mutations on 2b’s subcellular localization, we tagged the mutant 2bs with GFP, and these 2b-GFP constructs were transiently expressed in Nicotiana benthamiana leaves after insertion by agroinfiltration. As shown in Fig 1A, we found that accumulation of the mutant 2b was high in the nucleus and nucleolus compared to the amount of wild-type 2b. Those observations were further confirmed by western blot analysis using GFP antibodies and nuclear-rich fractions; the amount of mutant L77/85/87A in the nucleus was almost twice that of wild-type 2b (Fig 1B and 1C). We then used leptomycin B (LMB), an inactivator of XPO1 to test whether 2b’s localization was affected. LMB (40 nM) was infiltrated into the agroinfiltrated patches at 2 days post agroinfiltration, and the localization of 2b-GFP was observed at 4 h after the LMB treatment. 2b-GFP was found mainly in the nucleus, including the nucleolus, in the LMB-treated tissues, whereas 2b in the non-treated tissues accumulated in the cytoplasm and nucleus (Fig 1A). In addition, the 2b protein level in the nuclear-rich fraction was also significantly increased by the LMB treatment (Fig 1D and 1E). These results indicate the presence of NES in the C-terminal region from L77 to L87 and, thus, XPO1-mediated nuclear export.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. 2b has an NES motif for translocation from nucleus to cytoplasm.

(A-E) Effect of a mutation in NES on the subcellular localization of 2b in N. benthamiana. The GFP-tagged wild-type 2b (2b-GFP) or L77/85/87A was co-expressed with Ds-Red-fused histone 2B (H2B-DsRed), a nuclear marker protein, in N. benthamiana by agroinfiltration. LMB (40 nM) was infiltrated into the leaves at 2 days post agroinfiltration. (A) GFP fluorescence was observed with an epifluorescence microscope at 4 h after the LMB treatment. Scale bar: 50 μm (white), 5 μm (yellow). (B) Western blot of nuclear-rich fractions to detect 2b-GFP. The numbers above the blot are 5 replicates, and the identical numbers below 2b-GFP and L77/85/87A-GFP indicate the same leaves were used. The histone protein (H3) was detected using anti-H3 antibodies as a loading control. (C) Relative 2b-GFP/H3 ratios were densitometrically calculated using Multi Gauge Software (Fujifilm) and shown as a fold change. Mean values (±SE) were tested on log-transformed data for a significant difference using Student’s t-test (**P < 0.01). (D) Western blot analysis using the nuclear-rich fractions to detect 2b-GFP from the LMB-treated leaves. The identical numbers above the blot indicate the same leaves used. (E) Relative 2b-GFP/H3 ratios in the H₂O-treated and LMB-treated samples were densitometrically calculated using Multi Gauge Software (Fujifilm) and shown as a fold change. Mean values (±SE) were tested on log-transformed data for a significant difference using Student’s t-test (**P < 0.01).

https://doi.org/10.1371/journal.ppat.1010267.g001

We further investigated the nuclear export of Y2b into the cytoplasm because Y2b lacked an ideal NES when predicted by the NetNES v1.1 program. Because the CMV A1 vector has a truncated Y2b (residue 1 to 81, Y2bΔC) without NES, we examined whether the Y2b’s NES motif functions by cloning the DsRed gene into the A1 vector as a reporter for 2b accumulation (S2A Fig). As shown in S2B Fig, Y2b-DsRed was localized in both the nucleus and cytoplasm, whereas Y2bΔC-DsRed highly abundant in the nucleus and scarce in the cytoplasm, indicating that Y2b also contains a functional NES motif in the C-terminal region. By comparing Y2b and Y2bΔC, we revealed that the the C-terminal region including the NES motif was important for 2b’s nuclear localization. These results suggest that all subgroup I CMV 2bs have both NLS and NES motifs for nucleocytoplasmic translocation.

Phosphorylation on S28 is required for 2b nuclear localization

When we scanned the NLS motifs of 2b, we noticed some serine residues near the NLSs, which might be phosphorylated. Because phosphorylation-mediated nuclear import is a common mechanism to localize proteins in the nucleus of eukaryotic cells [8], we expected that phosphorylation of the residues near the NLSs might affect 2b nuclear localization. To identify phosphorylation sites on 2b, we used the NetPhos v2.0 software (http://www.cbs.dtu.dk/services/NetPhos-2.0/) and found seven residues including S40 and S42 (S3A Fig), which were previously proven to be phosphorylated [31].

To determine whether the predicted residues are phosphorylated, we assayed phosphorylation using 2b proteins with mutations at possible phosphorylation sites. For the phosphorylation assay, the 2b protein was produced in Escherichia coli as a C-terminal fusion with the maltose binding protein (MBP). We used a chimeric Y/HL2b protein (S3A Fig) because the E104 in the C-terminal region of Y2b is toxic to E. coli [47] (S1A Fig). E. coli-produced HL2b and Y/HL2b were treated with protein kinase C (PKC), separated by SDS-PAGE, and then the phosphorylated 2bs were visualized by phosphoprotein gel staining. As shown in S3B Fig, HL2b and Y/HL2b were clearly phosphorylated in vitro. For the in vitro phosphorylation using several truncated 2b constructs (Fig 2A), we found that the presence of the N-terminal half (46 amino acids) of 2b was sufficient for 2b phosphorylation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Mapping of phosphorylation sites of 2b.

(A) Schematic structures of the 2b mutants expressed in E. coli and in vitro phosphorylation assay. White boxes: Y2b peptide sequence; gray boxes: HL2b peptide sequence. Black bars: serine (S) or threonine (T); red bars: alanine. When the mutant 2b backbone is the construct containing alanine-substituted residues on all the putative phosphorylation sites, the mutant names contain “[A]”. Purified mutant 2b proteins were co-incubated with (+) or without (-) PKC. Phosphorylated proteins were detected by phosphoprotein gel staining (upper panel); the same gel was stained with Coomassie brilliant blue (lower panel). (B) In vivo phosphorylation assay. White boxes: N-terminal half of Y2b; gray boxes: C-terminal half of HL2b. Black bars: putative phosphorylation residues (S, T or Y); red bars: alanine-substituted residues. Fraction numbers are above the lanes. Elution 1–5: Elution fraction 1–5. Arrows on right: phosphorylated 2bs.

https://doi.org/10.1371/journal.ppat.1010267.g002

To determine which amino acids (T9, S28, S40 and S42) were phosphorylated in this region, we created several alanine-substituted 2b mutants at the four sites. Among those mutants containing a single serine, only ΔC[A]S28 was phosphorylated by PKC (Fig 2A). Curiously, the mutants with only one serine at either S40 or S42 were not phosphorylated by PKC. On the other hand, the mutants containing both S40 and S42 were phosphorylated regardless of the alanine substitution at S28.

To confirm the phosphorylation of S28, S40 and S42 in planta, we examined in vivo phosphorylation of these serine residues. For the phosphorylation assay, we used the CMV-H1 RNA 2 vector in which the entire 2b open reading frame (ORF) was deleted and a multiple cloning site (MCS) was inserted; a stop codon for 2a was added before MCS (S4A Fig) [48]. The infection of the recombinant CMVs carrying wild-type 2b and 2b recombinants was confirmed by symptoms and western blot analysis (S4B and S4C Fig). Our western blot analysis revealed that anti-2b antibodies could efficiently detect the mutant 2bs as well as wild-type 2b, suggesting that the alanine substitution did not affect the antibody-binding to 2b (S4C Fig). Using a phospho-protein affinity column, we first isolated total phosphorylated proteins from tobacco tissues infected with CMV containing the 2b mutants (S5 Fig). Phosphorylated 2b was then detected using western blot analysis with anti-HL2b antibodies. As shown in Fig 2B, the positive control Y/HL2b was phosphorylated in vivo. We then prepared the 2b mutants containing an alanine at all the possible phosphorylation residues (Y/HL[A] in Fig 2B); in addition to the seven software-predicted amino acids, the other serine, threonine and tyrosine residues (Y58, S63, T66, S68 and T97) were all substituted with alanine. As expected, Y/HL[A] was not phosphorylated. On the other hand, Y/HL[A]S28 containing a serine at position 28 and Y/HL[A]S40/S42 containing a serine at positions 40 and 42 were phosphorylated in planta. These results clearly indicate that the three serine residues (S28, S40 and S42) are phosphorylated in planta.

Because S28 is located near the two NLSs (S3A Fig), we expected that nuclear localization of 2b might be affected by phosphorylation on S28. To determine the effect of S28 phosphorylation on 2b’s subcellular localization, the S28 residue was replaced with an alanine (S28A) to impair phosphorylation or with a glutamate (S28E) to mimic the state of phosphorylation, and then GFP was added to the C-terminus of 2b. The 2b-GFP sequences were cloned under the 35S promoter in a binary vector (pBE2113) and expressed in N. benthamiana by agroinfiltration. The wild-type 2b-GFP was detected in the nucleus and nucleolus as previously reported [37], and substantial levels were also found in the cytoplasm (Fig 3A). In contrast, the GFP signals of S28A were mainly localized in the cytoplasm and barely detected in the nucleus. Interestingly, even if the GFP signal was detected in the nucleus at a low level, it was never detected in the nucleolus (Fig 3A). When the phospho-mimetic mutant (S28E) was tested, GFP signals were found to be localized in the nucleus and cytoplasm, just like the wild-type 2b. In addition, when wild-type 2b and S28A were compared, 2b accumulation in the nuclear-rich fraction was dramatically reduced by the S28A substitution (Fig 3B and 3C). These results suggest that S28 is important for the 2b’s nuclear localization. We further analyzed the subcellular localization of S28E in two experiments using a NES mutant and with the exportin inhibitor LMB. As expected, both experiments confirmed that the impairment of the exportin-mediated 2b translocation resulted in an increase in nuclear accumulation of 2b (S6 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Effect of S28 phosphorylation on subcellular localization of 2b.

(A) Subcellular localization of 2b and the 2b mutants in N. benthamiana cells. 2b-GFP and the GFP-tagged 2b mutants (S28A-GFP and S28E-GFP) were co-expressed with H2B-DsRed in agroinfiltrated leaves. GFP fluorescence was observed at 2 days post agroinfiltration using a confocal microscope. Scale bar: 50 μm (white) or 5 μm (yellow). (B) Detection of 2b and S28A in the nuclear-rich fraction. 2b-GFP and S28A-GFP were overexpressed in agroinfiltrated leaves, and the nuclear-rich fraction was probed for the proteins in a western blot. The GFP-tagged proteins and H3 (the loading control) were detected using anti-GFP antibodies and anti-H3 antibodies, respectively. The numbers above the blot are 5 replicate leaves; identical numbers indicate the same leaves were used for 2b-GFP and S28A-GFP. (C) Relative 2b accumulation levels. The relative 2b-GFP/H3 ratios were densitometrically calculated using Multi Gauge Software and shown as a fold-change. Mean values (±SE) were tested on log-transformed data for a significant difference using Student’s t-test (***P < 0.001). (D) In vivo co-immunoprecipitation of NbIMPα1 and 2b-GFP. NbIMPα1/2b-GFP complexes were first co-precipitated using anti-FLAG antibody, and 2b-GFP, S28A-GFP and S28E-GFP in the precipitant were detected by western blot analysis using anti-GFP antibodies. 2b-GFP, S28A-GFP, S28E-GFP and NbIMPα1-FLAG in the total protein extracts are shown as the inputs. RuBisCo large subunit is shown as a loading control.

https://doi.org/10.1371/journal.ppat.1010267.g003

Wang et al. (2004) previously demonstrated that AtIMPα binds to the NLSs of 2b for nuclear transport [27] and that 2b enters the nucleus, perhaps in an AtIMPα-dependent manner. Because the S28 residue is just adjacent to the two NLSs, and the results of our subcellular localization study indicated that phosphorylation of S28 is required for 2b’s nuclear localization, we considered that phosphorylation of S28 was important for the 2b–IMPα interaction. Because the IMPα-mediated nuclear import of 2b had not been unequivocably proven, we first analyzed the involvement of IMPα in nuclear localization of 2b by silencing two AtIMPα homologs in N. benthamiana (NbIMPα1 and NbIMPα2), which share ~80% amino acid sequence identity with AtIMPα (S7A Fig). By virus-induced gene silencing (VIGS) using the PVX vector, both IMP genes were silenced by coinoculation with PVX-NbIMPα1 and PVX-NbIMPα2 (S7B and S7C Fig), while their viral RNA levels were not significantly different (S7D Fig). To examine whether IMPα-silencing affects nuclear localization of 2b, we used agroinfiltration to transiently express 2b-GFP in the IMPα-silenced N. benthamiana plants. As a result, 2b-GFP was clearly detected in both the nucleus and cytoplasm in the PVX vector-infected plants, and 2b nuclear localization was reduced in the IMPα-silenced plants (S8A Fig). Consistent with the microscopic observation, the level of 2b-GFP in the nuclear-rich fraction was slightly reduced (~12% reduction) in the IMPα-silenced N. benthamiana compared to that in the PVX-infected control plants (S8B and S8C Fig). When the free GFP was examined as a control, its nuclear localization was not affected by silencing the IMPα genes (S8D–S8F Fig). We anticipated that the 12% reduction in 2b nuclear localization was due to the recovery of IMPα expression by 2b’s RSS activity in the IMPα-silenced plants. To reduce 2b’s RSS activity, which may affect IMPα silencing, we then used the 2b mutant R46C, which lacks RSS activity and is localized in both the nucleus and cytoplasm [22] in the IMPα-silenced plants. As expected, the level of nuclear-localizing R46C-GFP was greatly decreased (~73% reduction) by the IMPα-silencing (S8G–S8I Fig). These data together suggest that the nuclear import of 2b is mediated by NbIMPα.

To determine whether the phosphorylation of S28 affects the interaction between 2b and NbIMPα, we performed a co-immunoprecipitation assay using NbIMPα1-FLAG and GFP-tagged 2bs (2b-GFP). The NbIMPα1-FLAG/2b-GFP complexes were first precipitated using anti-FLAG immunoglobulin G (IgG), then the 2b-GFP precipitants were detected by western blot analysis using anti-GFP IgG. As shown in Fig 3D, the wild-type 2b-GFP protein co-precipitated with NbIMPα1, indicating that 2b directly interacts with NbIMPα1. While S28A was not detected in the precipitants, the S28 phospho-mimetic mutant (S28E) was clearly detected. These results together suggest that phosphorylation at S28 is important for the 2b–NbIMPα1 interaction and thus 2b nuclear localization.

Phosphorylation of 2b negatively affects RSS activity

To analyze the relationship between the phosphorylation of 2b and its RSS activity, we generated a series of 2b mutants (alanine substitution at the phosphorylation sites and glutamate/aspartate substitution to mimic phosphoserine residues). We then examined the RSS activities of these 2b mutants in a conventional agroinfiltration assay in N. benthamiana. The 2b mutants were cloned into pBE2113 (Fig 4A), and RNA silencing of the GFP gene was induced by an inverted repeat construct of GFP (GFP-IR). Silencing of the GFP gene was efficiently suppressed by 2b and S28A; S28A showed even stronger RSS activity compared to wild-type 2b (Fig 4B). On the other hand, S40A/S42A and S28E had clearly weaker RSS activity, while S28A/S40A/S42A, S40D/S42D, S40E/S42E and S28E/S40E/S42E had no RSS activity. Because the sRNA-binding ability of 2b is important for suppressing RNA silencing [21,22], we tried to elucidate whether the phosphorylation of 2b affects its binding with sRNAs by examining the sRNA-binding ability of 2b in vitro using the phospho-mimetic 2b mutant (S28E/S40E/S42E) (S9 Fig). The MBP-fused 2b and the phospho-mimetic 2b mutant were synthesized in E. coli, and the purified MBP-2bs were incubated with biotinylated 21-nt siRNAs. The 2b–siRNA complexes were co-precipitated with streptavidin-coupled beads. The precipitated 2bs were separated by SDS-PAGE and visualized by Coomassie brilliant blue (CBB) staining. As a result, wild-type 2b was co-precipitated with biotinylated siRNAs, whereas S28E/S40E/S42E mutant was barely detectable in the gel (S9 Fig). These results suggest that phosphorylation of the three serine residues (S28, S40 and S42) negatively affects 2b’s RSS activity by interfering with the 2b–siRNA interaction. To estimate how this phosphorylation would affect siRNA-binding in the 2b’s conformation, we next analyzed the tertiary structures of the 2b mutants using homology modeling. As shown in S10A Fig, we noted that a hydrogen (H)-bond was formed between S42 and the backbone phosphate of siRNA in the nonphosphorylated 2b. All the mutant models were consistent with our experimental observations on 2b’s RSS activity (Fig 4B). The H-bond disappeared after phosphorylation or glutamate/aspartate substitution of S42 (S10A Fig). We further calculated the distance between S42 and cytosine 14 of siRNA in the 2b-siRNA binding models. All the 2b mutants without RSS activity are located farther from the siRNAs (S10B Fig) than all the 2b mutants with RSS activity. We thus consider that this H-bond formation at S42 somehow contributes to 2b’s siRNA binding. On the other hand, in a previous report of Chen et al. (2008) [49], R28 of tomato aspermy virus (TAV) was predicted to form H-bond with siRNA. Because the R28 of TAV 2b is corresponding to S28 of CMV 2b and it is located inside of RNA binding motifs, we expected that the phosphorylation of S28 may also affect 2b’s RNA-binding ability. To determine whether the S28 phosphorylation affects 2b’s sRNA-binding, another in vitro sRNA-binding assay was performed. As shown in S11A Fig, the S28E substitution dramatically reduced 2b’s sRNA-binding ability suggesting that the S28 phosphorylation also negatively affect 2b’s RSS activity by disturbing sRNA-binding.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. RSS activity of 2bs with mutations at phosphorylation sites.

(A) Schematic structure of expression vector constructs containing mutant 2bs. The mutant 2bs were created by substituting S28, S40 and S42 with alanine (A), glutamic acid (E) or aspartic acid (D). (B) RSS activity of the mutant 2bs in the agroinfiltration transient expression assay. Agrobacterium cultures carried a 5:1:5 ratio of the GFP-sense (GFP), GFP-inverted-repeat (GFP-IR), and the 2b constructs. N. benthamiana leaves were agroinfiltrated with bacterial suspensions (~1.0 OD₆₀₀). GFP was examined at 2 days post-infiltration. Inoculum containing GFP and GFP-IR was used as a control for GFP silencing. (C–E) Comparison of RSS activity between 2b and S28A in N. benthamiana. The Agrobacterium suspension of 2b or S28A was serially diluted to OD₆₀₀ = 0.5, 0.25 and 0.13. The inocula were a mixture of Agrobacterium cultures carrying the GFP, GFP-IR and the 2b constructs. (C) Leaves were examined for GFP fluorescence at 2 days post-infiltration. (D) Real-time RT-PCR to compare GFP mRNA levels between 2b and S28A. The agroinfiltrated patches, where GFP, GFP-IR and 2b were simultaneously expressed, were excised. Mean values (±SE) are fold-changes when the value in 2b was set to 1.0. The values were analyzed on log-transformed data by Student’s t-test (**P < 0.01). (E) GFP in agroinfiltrated patches was detected by western blot analysis using anti-GFP antibodies. RuBisCo large subunit is shown as a loading control. The ratios of GFP/RuBisCo were densitometrically calculated when the value in 2b was set to 1.0. (F and G) Comparison of RSS activity between 2b and S28A in onion epidermal tissues. Onion bulb scales were infiltrated with a mixture of Agrobacterium suspension of GFP, GFP-IR and 2b. The plant experimental vector pBE2113 (GUS) was used as a negative control. GFP florescence was observed with an epifluorescence microscope. GFP fluorescence intensity was measured for three different onion scales using the LAS AF program (Leica). Scale bar: 50 μm. The values of GFP florescence are fold-changes when the value in GUS was set to 1.0. Means (±SE) were tested on log-transformed data for a significant difference using Tukey’s multiple comparison test (P < 0.05); different letters above the bars indicate a significant difference.

https://doi.org/10.1371/journal.ppat.1010267.g004

According to a previous study by Du et al. (2014), the cytoplasmic 2b is important for 2b’s RSS activity [36]. Because the S28A mutant, which lacks phosphorylation at S28, was primarily localized in the cytoplasm (Fig 3), we anticipated that the S28A mutant might have even stronger RSS activity than wild-type 2b. To compare RSS activity between 2b and S28A, we performed a comparative study on RSS activity by diluting the Agrobacterium inoculum. Agrobacterium suspensions containing either wild-type 2b or S28A construct was serially diluted and infiltrated together with GFP- and GFP-IR-expressing bacterial inocula. As expected, the GFP signal was indeed stronger in the tissues infiltrated with S28A compared to those with wild-type 2b; S28A was ~2.5 times higher (Fig 4C–4E).

To confirm the difference in GFP fluorescence between 2b and S28A, we then used onion epidermal tissues where GFP fluorescence is more easily observed than in N. benthamiana. As shown in Fig 4F and 4G, the GFP signal was ~2.5 times higher in the tissues with S28A than in those with wild-type 2b. These results therefore suggest that the S28 phosphorylation provides some negative effect on RSS activity by enhancing 2b’s nuclear localization and/or structurally inhibiting sRNA-binding.

Because S28A has stronger RSS activity than wild-type 2b (Fig 4), we considered that S28A might contribute to an increase in CMV accumulation by strongly suppressing antiviral RNA silencing. A recombinant CMV RNA 2 containing S28A (H1:S28A) was constructed and used with RNAs 1 and 3 of CMV-Y to co-inoculate N. benthamiana. The plants infected with CMV expressing S28A developed a systemic mosaic at 10 days post inoculation (dpi), earlier than those infected with CMV containing wild-type 2b (H1:2b), and the viral RNA level was ~4 times higher in the inoculated leaves and ~30 times higher in the upper systemic leaves than in the controls (Fig 5A–5C). Furthermore, compared to H1:2b, H1:S28A caused a very distinct yellowish mosaic in systemically infected leaves at 21 dpi, with ~65 times higher viral RNA level (Fig 5D and 5E). These data thus suggest that phosphorylation at S28 may reduce the level of 2b in the cytoplasm, and viral accumulation is thus lower due to the weaker RSS activity. To exclude the effect of other putative phosphorylation sites on viral disease severity, we used the 2b backbone, which includes alanine residues on all the putative phosphorylation sites except S28, S40 or S42, and three mutant 2bs ([A]S28, [A]S40/S42 and [A]S28/S40/S42) to test their ability to induce symptoms (S12A Fig). N. benthamiana plants inoculated with the CMV-H1 vector expressing [A]S28 were asymptomatic at 11 dpi, and the viral RNA level was ~20% of the level in plants infected with CMV-[A]S28/S40/S42 (S12B and S12C Fig). The plants infected with CMV expressing [A]S28/S40/S42 had relatively mild symptoms and an early symptom recovery at 11 dpi with a lower viral RNA level compared to the plants infected with CMV expressing [A]S40/S42 (S12B and S12D Fig). At 21 dpi, the plants infected with CMV-[A]S40/S42 had distinct mosaics, earlier than the plants with CMV-[A]S28/S40/S42, and the viral titer was significantly higher in the systemically infected leaves (S12E and S12F Fig), suggesting that S28A can enhance not only viral accumulation but also disease severity.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Effect of alanine substitution at S28 on CMV disease severity.

(A) CMV symptoms on N. benthamiana 10 days post inoculation (dpi) with CMV containing S28A at early stage of infection. Wild-type 2b and S28A genes were inserted into the CMV-H1 vector (H1:2b and H1:S28A). CMV RNA levels based on real-time RT-PCR in inoculated (B) and systemically infected (Systemic) leaves (C). Means (±SE) are fold-changes, calculated when the value of H1:2b is set to 1.0, and tested on log-transformed data for a significant difference using Student’s t-test (*P < 0.05 or **P < 0.01). (D) Disease severity induced by 2b and 2b-S28A and (E) CMV RNA levels based on real-time qRT-PCR in systemically infected upper leaves at 21 dpi. Mean values (±SE) are fold-changes and statistically analyzed as above (****P < 0.0001).

https://doi.org/10.1371/journal.ppat.1010267.g005

NES-mediated nuclear export of 2b promotes 2b’s RSS activity

When we aligned the amino acid sequences of the 2bs of several CMV strains, we noticed that all of the 2b proteins from CMV subgroup II strains lacked the corresponding NES motif (Fig 6A), that is characteristic of all subgroup I strains. Because the 2bs of subgroup II had less efficient RSS activities [18,28], we further examined the effect of the mutations at NES on viral pathogenicity to determine whether nuclear accumulation of 2b indeed reduces 2b’s RSS activity. Three leucine residues in NES (L77, L85 and L87) were substituted with alanines in the backbone of R2b and its S28E mutant (L77/85/87A and S28E-L77/85/87A, respectively), which localized in the nucleus and the cytoplasm (Figs 1 and 3). Wild-type 2b strongly suppressed GFP gene silencing while S28E showed low RSS activity (Fig 6B and 6C). Unexpectedly, the RSS activity of 2b(L77/85/87A) was greatly reduced, and S28E-L77/85/87A were totally impaired, indicating that 2b’s RSS activity might be enhanced by the export of 2b to the cytoplasm (Fig 6B and 6C). In addition, our sRNA-binding assay revealed that the amino acid substitution in the NES site did not affect 2b’s RNA-binding ability (S11B Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. NES motif determines 2b’s RSS activity and CMV accumulation.

(A) Alignment of 2b amino acid sequences belonging to subgroup I and II. Amino acid positions are indicated above the alignment. Red arrows indicate leucine residues important for 2b nuclear export. (B) RSS activity in 2b and NES mutant after infiltration of N. benthamiana leaves with Agrobacterium suspensions carrying GFP, GFP-IR and 2b at a ratio of 5:1:5. GFP florescence was observed at 2 days post-infiltration. (C) Western blot analysis to compare the GFP levels. GFP in the agroinfiltrated patches was detected by western blot analysis using anti-GFP antibodies. RuBisCo large subunit is shown as a loading control. The ratios of GFP/RuBisCo were densitometrically calculated when the value in GUS was set to 1.0. (D–G) Effect of a mutation in the NES motif on (D) CMV cyclic symptoms and (E–G) viral accumulation at 10, 15 and 21 dpi. N. benthamiana plants were inoculated with CMV containing either the wild-type 2b or the NES mutant (L77/85/87A). Viral RNA accumulation levels were detected by real-time RT-PCR. Means (±SE) are fold-changes calculated when the value in H1:2b was set to 1.0 and were tested on log-transformed data for a significant difference using Student’s t-test (**P < 0.01, *P < 0.05).

https://doi.org/10.1371/journal.ppat.1010267.g006

Although the RSS activity of 2b(L77/85/87A) was very low, it has been previously reported that a C-terminal deletion mutant without NES had RSS activity comparable to wild-type 2b [37]. To elucidate this discrepancy between the two observations, we here investigated the RSS activity of a series of C-terminal deletion mutants (S13A Fig). As the C-terminal amino acids were gradually trimmed on the background of 2b(L77/85/87A), the RSS activity suddenly recovered to a level similar to that of wild-type 2b between the 23-AA deletion [L77/85/87A(1–87)] and the 10-AA deletion [L77/85/87A(1–100)]. This result suggests that the 23 AA residues downstream of the NES site would affect the 2b’s RSS activity. Our western blot analysis showed that all the C-terminal deletion mutants were stably expressed in the agroinfiltrated tissues (S13C Fig).

To determine whether a lack of NES affects viral pathogenicity, we compared the disease severity between CMV with the intact 2b (H1:2b) and CMV with L77/85/87A (H1:L77/85/87A). At an early stage of infection (10 dpi), symptom severity and viral accumulation were not significantly different between H1:2b and H1:L77/85/87A (Fig 6D and 6E), but at 15 and 21 dpi, H1:2b had developed a systemic cyclic symptom on the upper leaves with a higher level of viral RNA accumulation (Fig 6D, 6F and 6G), unlike plants with H1:L77/85/87A. These results suggest that the nuclear export of 2b is necessary for efficient suppression of host antiviral RNA silencing.

2b is dephosphorylated in the nucleus

Although S28A was not efficiently transported into the nucleus, it had very strong RSS activity. On the other hand, S28E was abundant in the nucleus, and S28E-L77/85/87A was also present in the nucleolus (S6 Fig). We thus wondered which form will predominate in the nucleus, the phosphorylated or nonphosphorylated 2b. To start to answer this question, we compared the accumulation levels of phosphorylated 2b in the nucleus with those of wild-type 2b and L77/85/87A. To isolate 2b from the nucleus, FLAG-tagged 2b and L77/85/87A were expressed in N. benthamiana leaves by agroinfiltration, and the nuclear fractions were isolated by sucrose density centrifugation. As expected from the results shown in Fig 1, the level of L77/85/87A-FLAG was higher than 2b-FLAG in the nucleus (Fig 7A). When we analyzed 2b isolated from the cytoplasmic fraction (S14 Fig), wild-type 2b was more abundant than L77/85/87A (Fig 7B). Then, we purified the phosphorylated 2b in the nuclear-rich fraction using a phosphoprotein-specific column (S5 Fig). In this experiment, we fractionated nuclei from the cells of agrobacterium-infiltrated tissues and tried to recover total protein containing 2b, but the recovery efficiency of 2b was not good enough to clearly detect phospho-2b by western blot using phosphoprotein staining. We thus decided to enrich phosphorylated proteins first by column chromatography and detect 2b in the eluted fractions using anti-FLAG antibody. As shown in the review by Schmidt et al. (2007) [50], the separation and enrichment of phosphoproteins using columns has been well studied, and excellent kits are now commercially available. The purified phospho-2bs were then detected by western blot analysis. As a result, we found that the phosphorylated 2b was indeed much more abundant than the phosphorylated L77/85/87A in the nucleus (Fig 7C). Considering that the nuclear export of L77/85/87A was impaired, we infer that phospho-2bs are likely dephosphorylated in the nucleus.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Comparison of levels of phosphorylated 2b in nucleus.

(A) Western blot analysis to detect nuclear-localized 2b-FLAG protein in the nuclear-rich fraction from leaves of N. benthamiana expressing the FLAG-tagged 2b and L77/85/87A. Histone H3 was detected using anti-H3 antibodies for a loading control. Values below the blot for 2b are the relative ratio of 2b/H3. (B) Detection of cytoplasmic 2b by immunoprecipitation. The cytoplasmic fractions were separated from the total plant extracts by differentiated centrifugation. The 2b-FLAG was immunoprecipitated with anti-2b antibodies and detected in the western blot using anti-FLAG antibody. (C) Detection of phosphorylated 2b in the nuclear fraction. The level of phosphorylated 2b-FLAG in the nuclear-rich fraction was isolated using a phospho-protein enrichment column. Phosphorylated 2b in each elution (E1–5) was detected by western blot using anti-FLAG antibody.

https://doi.org/10.1371/journal.ppat.1010267.g007

Phosphorylation at S28 regulates 2b’s nuclear transport and RSS activity

In this study, we demonstrated that phosphorylation of S28 alone can determine 2b’s interaction with NbIMPα1 and its nuclear localization (Fig 3). According to Nardozzi et al. (2010), phosphorylation can either activate or inactivate IMP-mediated nuclear import [8]. For example, SV40 T-antigen has phosphorylation sites adjacent to the NLS, and the phosphorylation of SV40 T-antigen enhances NLS recognition by IMPα [2,51,52]. Because S28 is close to the two NLSs of 2b but not inside the NLS motifs, we hypothesized that 2b’s nuclear transport may be mediated by IMPα1 in the same fashion as that of SV40 T-antigen.

We here found that 2b could shuttle between nucleus and cytoplasm by the importin- and exportin-mediated transport systems. We also found that the phosphorylation of S28 was important for nuclear translocation. However, we consider that the role of the other phosphorylation sites (especially S40 and S42) in 2b’s subcellular localization is also important, and that combination of the phosphorylation of S28 and that of the other sites might determine the balance between the importin-mediated nuclear import and exportin-mediated nuclear export. For example, Neme et al. (2017) previously reported that S40D/S42D was localized to the cytoplasm [31]. On the other hand, because we here observed that S28A was localized to the cytoplasm, it is unclear whether the localization of 2b to the cytoplasm is due to the fact that S28 is not phosphorylated, or whether it is due to the fact that S40 and S42 are phosphorylated. One explanation for this is the balance between the 2b’s affinities for importin and exportin depending on the combination of phosphorylation at the three sites. If S28 is not phosphorylated, 2b cannot be transported to the nucleus by importin and thus localizes to the cytoplasm. On the other hand, S40D/S42D, in which S28 can be phosphorylated, will be transported to the nucleus by importin and then may be quickly exported from the nucleus by exportin. In other words, the phosphorylation state of 2b inside of the nucleus will be also important for exportin-mediated nuclear export. The relationship between the combination of phosphorylation at these three sites and the shuttling of 2b between nucleus and cytoplasm will be our next subject for further examination.

Phosphorylation at S28 also has a profound effect on the function of 2b. The S28 phosphorylation negatively regulated 2b’s RSS activity and CMV pathogenicity, perhaps due to a decreased level of cytoplasmic 2b (Figs 3–5). These results are in accordance with the previous reports in that 2bs in the cytoplasm play the main role in suppressing RNA silencing [21,22,36]. In addition, the ability of glutamate or aspartate substitution mutants (S28E, S40D/S42D, S40E/S42E and S28E/S40E/S42E) to suppress RNA silencing was reduced or totally impaired, suggesting that phosphorylation at those three serine residues can compromise 2b’s RSS activity (Fig 4B). However, we believe that the S28 amino acid residue is more influential than S40 and S42 in the regulation of 2b’s RSS activity via phosphorylation because it has a great effect on 2b’s subcellular localization. This belief is based on the observation that the 2b’s RSS activity was enhanced by an alanine substitution at S28 compared to that of wild-type 2b (Fig 4B), while the S40A/S42A mutant, which has a serine at S28, had greatly decreased RSS activity (Fig 4B). The two NLS regions and S42 have already been demonstrated to be important for 2b’s siRNA-binding affinity [21,31,49]. Because phosphorylation of amino acids confers a negative charge on the side chains, phosphorylation in proximity to RNA binding sites negatively affects 2b’s sRNA binding. In silico analysis based on the 2b-siRNA binding model suggests that the H-bond formation between 2b and siRNA is destroyed by S42 phosphorylation (S10 Fig), and in fact, our in vitro siRNA binding assay using wild-type 2b and its phospho-mimetic mutant (S28E and S28E/S40E/S42E) showed a negative correlation between the phosphorylation of 2b and its siRNA-binding activity (S9 and S11 Figs). These results have similarities with Chen et al. (2008) in that R28 and S42 of TAV 2b, which correspond in position to S28 and S42 of CMV 2b, are predicted to form H-bond with uracil 10 and 15, respectively [49].

The phosphorylation of 2b likely regulates its binding to some components in the RNA silencing pathway. For example, phosphorylation of AGO2 strongly reduces the association between AGO2 and sRNAs in Caenorhabditi elegans [40]. While AGO2 is also thought to be translocated to the nucleus by IMP [53,54], it is not yet known whether phosphorylation is really necessary for the translocation of AGO2. However, it is worth noting that AGO2 is similar to 2b in that it loses its ability to bind siRNAs upon phosphorylation.

The subcellular localization of 2b is determined by a balance of importin/exportin-mediated intracellular transport

The translocation of 2b from the nucleus to the cytoplasm is mediated by exportin in that (i) the treatment by exportin inhibitor, LMB, promoted 2b’s accumulation in the nucleus (Figs 1A and S6) and that (ii) the mutations at the leucine residues in the predicted NES motif induced high levels of 2b accumulation in the nucleus and nucleolus (Figs 1, S2 and S6). Only a few cases have been reported on the nuclear export of plant viral proteins. A mutation in the NES motif of BNYVV p25 induced attenuated symptoms in quinoa plants inoculated with the virus containing the mutant p25 [17]. XPO1-dependent nuclear export of the C4 protein of TLCYnV has been found to be critical for developmental abnormalities in virus-infected N. benthamiana [16]. On the basis of these observations, plant viruses likely utilize not only nuclear import but also nuclear export to enable their own proteins to function in either the nucleus or cytoplasm.

We were then interested in whether the 2bs that had been transported from the nucleus to cytoplasm could still function as an RSS. We thus analyzed the RSS activity of the mutant L77/85/87A. As shown in Fig 6B, L77/85/87A was highly concentrated in the nucleus and did not efficiently suppress RNA silencing. Therefore, we hypothesized that 2b exported from nucleus might be recycled in the cytoplasm to work again as an efficient RSS. It has been reported that 2bs from subgroup II strains predominantly accumulate in the nucleus and are scarcely detectable in the cytoplasm [36] and that 2b of strain CMV-Q (subgroup II) has very weak RSS activity compared to that of 2b of strain CMV-SD (subgroup I) [55]. These observations thus suggest that the presence or absence of the NES site is the major factor underlying the RSS characterizations of CMV subgroups I and II.

The NES-amino acid substitution mutant, L77/85/87A showed very low RSS activity while some C-terminal deletion mutants without NES had RSS activity comparable to that of wild-type 2b. To elucidate this discrepancy, we found a key in the results of our 2b’s RSS activity assay using a series of C-terminal deletion mutants. The RSS activity suddenly recovered to a level similar to that of wild-type 2b just between L77/85/87A(1–87) and L77/85/87A(1–100); we thus considered that 2b’s siRNA-binding activity might be higher in the former than in the latter. We then conducted in silico analysis to predict 3D structure models between siRNA and those deletion mutants. As shown S15 Fig, siRNA can make a complex more stably with L77/85/97A(1–87) than with L77/85/97(1–100), based on the H-bond distance and contact surface area between the two molecules. Considering that 2b’s RSS activity in the agroinfiltration assay must be mainly that of cytoplasmic 2b before it is transported to the nucleus, it is conceivable that the C-terminal deletion mutants without NES could exhibit high RSS activity before they are translocated to the nucleus, and that relatively low RSS activity of 2bs with NES including wild-type 2b will be compensated by that of the recycled 2b exported from the nucleus. Although this hypothesis must be further examined, we believe that we could provide one possible explanation for the discrepancy on the 2b’s RSS activity between the C-terminal deletion mutants and L77/85/87A.

In addition to importin and exportin, 2b also interacts with AGOs and other host proteins including plant catalase 3 (CAT3) and rgs-CaM [28,33,56,57]. Duan et al. (2012) previously suggested that the 2b–AGO interaction may affect the intranuclear localization of both proteins [37]. Because 2b’s NLSs overlap with the sRNA binding site, the abundance of sRNAs may affect the phosphorylation of 2b by host kinase(s) and, thus, the 2b–IMP interaction. Therefore, intracellular localization of 2b will be determined by a balance in the accumulation levels among sRNAs, importins, exportins and other host 2b-binding proteins.

2b may be recycled to function as an efficient RSS in the nucleus and the cytoplasm by dephosphorylation

When we analyzed the level of phosphorylated 2b in the nuclear-rich fraction, the phosphorylated L77/85/87A was very low compared to the wild-type 2b (Fig 7), suggesting that 2b might be dephosphorylated in the nucleus. In the nucleus, there are various Ser/Thr protein phosphatases that regulate phosphoproteins in the nucleus [58–62]. After 2b is dephosphorylated in the nucleus, it might then be exported from the nucleus to the cytoplasm for recycling. On the other hand, considering that the phospho-mimetic mutant 2b (S28E and S28E/S40E/S42E) lost its sRNA-binding ability (S9 and S11 Figs), the dephosphorylation of 2b might facilitate 2b binding to sRNAs in the nucleus. These observations are also consistent with a previous report that 2b can bind AGO4-related 24-nt siRNAs in the nucleus to interfere with RNA-dependent DNA methylation (RdDM) [33].

On the other hand, we also demonstrated that S28 phosphorylation may be necessary for 2b’s IMP-mediated nucleolar localization (Fig 3). Among the few studies on IMP-mediated nucleolar transport of plant viral proteins, one is the potato mop-top virus triple gene block1 (TGB1) protein, which interacts with two NbIMPα proteins to be translocated into the nucleolus in N. benthamiana [63], suggesting that IMPα is necessary for the protein transport into the nucleolus. The NLS-mediated nuclear/nucleolar localization of the potato virus A viral genome-linked protein (VPg), which is an RSS, was found to be important for interacting with the protein, suppressor of gene silencing 3 [64,65].

Although the main function of the nucleolus is the synthesis of ribosomal components [66,67], accumulating evidences suggest that the nucleolus is also involved in other cellular functions including gene silencing (reviewed by Kalinina et al. 2018 [68]). In addition to ribosomal components, transfer RNAs, small nucleolar RNAs (snoRNAs) and some mature miRNA precursors are present in the nucleolus [69,70]; some miRNAs are processed from snoRNA precursors [71,72]. snoRNA-derived sRNAs were reported to be associated with AGOs in Arabidopsis and human [73,74]. These results thus suggest that 2b may function as an RSS not only in the nucleus but also in the nucleolus by regulating its own phosphorylation.

Similarities between 2b and AGOs in their subcellular localization and interactions

Bologna et al. (2018) demonstrated that AGO1 of Arabidopsis has both NLS and NES motifs, and its transport from the nucleus is mediated by XPO1 [39]. They proposed a revised model of the miRNA pathway in plants; mature miRNAs are loaded on AGO1 in the nucleus, for transport across the nuclear membrane with the help of XPO1 into the cytosol [39]. Thus, AGO1 is shuttled between the nucleus and the cytoplasm to induce miRNA-mediated PTGS. Considering that both 2b and AGO1 depend on XPO1 for their export from the nucleus, they are likely to compete for XPO1. Therefore, 2b may be able to inhibit the step where AGO1 transports mature miRNAs from the nucleus to the cytoplasm by competition for XPOI. This step could be another operational point for 2b to suppress RNA silencing.

On the other hand, a human AGO1 promotes gene expression in human cancer cells by interacting with RNA polymerase II [75]. Such positive gene regulation has also been reported for Arabidopsis AGO1 in response to hormones and biotic/abiotic stresses [38]. Therefore, it is conceivable that 2b interacts with AGO1 in the nucleoplasm to negatively regulate transcription of the genes related to some antiviral responses. In addition, 2b also interacts with AGO4 and AGO4-related 24-nt siRNAs, which work in RdDM [30,33]. Pontes et al. (2006) previously reported that the core players of RdDM—RNA-dependent RNA polymerase 2, Dicer-like 3, AGO4, the chromatin remodeler and the de novo cytosine methyltransferase—are all colocalized with siRNAs in the nucleolus [76], suggesting that some processes of siRNA generation and RISC formation may occur in the nucleolus. Therefore, there is a good possibility that 2b also works as an RSS in the nucleolus. We believe that 2b is shuttled between the nucleolus and non-nucleolar space in the nucleus, so that 2b can chase AGOs and sRNAs wherever they go in the nucleus.

Plant materials and virus inocula

N. benthamiana and N. tabacum plants were grown at 25°C with 16 h light, 8 h dark. N. benthamiana plants were used for agroinfiltration assays and symptom observations. N. tabacum plants were inoculated with the CMV constructs containing the mutant 2bs for in vivo phosphorylation assays. Transcripts of CMV RNAs 1 and 3, and RNA 2 containing mutant 2bs were synthesized in vitro as described before [77]. Plants were mechanically inoculated with a mixture of three transcripts (RNAs 1–3) using carborundum [22].

Construction of 2b recombinants and CMV infectious vectors

The mutant 2b constructs, which have point mutations at possible phosphorylation sites, were created by recombinant PCR using the primers listed in S1 Table, and then inserted between the StuI and SpeI sites in the CMV-H1 vector to generate RNA 2 harboring the mutant 2b [47]. All the mutant 2b clones in CMV-H1 were sequenced to confirm their point mutations before use. Additional sequencing analyses were conducted to check whether the introduced 2b mutations were still maintained (S16 Fig).

Prediction of NES in 2b proteins

The NES motif in the 2b proteins of four CMV strains were analyzed by NetNES 1.1 Server (http://www.cbs.dtu.dk/services/NetNES/). Residues with a NES score higher than the threshold and its nearby amino acids, where a peak score in the HMM or artificial neural network (NN) was found, were predicted as the NES motif.

Agrobacterium-mediated transient overexpression

For the transient expression of 2b, the 2b genes were integrated into pBE2113 between the XbaI and SacI sites. For subcellular localization studies, the 2b-GFP and histone 2B-DsRed (H2B-DsRed) constructs were generated using the primers listed in S1 Table. A. tumefaciens strain KYRT1 was transformed with the vector construct by the conventional freeze–thaw method. The transformed KYRT1 was cultured in yeast extract peptone (YEP) medium (1% yeast extract, 1% peptone and 0.5% NaCl, w/v) containing 10 μM of acetosyringone. The harvested cells were resuspended in the Agrobacterium resuspension buffer (10 mM MgCl₂, 10 mM MES and 0.2 mM acetosyringone) at ~1.0 optical density at 600 nm (OD₆₀₀), then allowed to sit for 2–3 h, N. benthamiana leaves were then infiltrated with the inoculum (100–200 μL for each patch) using a 1-mL syringe without a needle. H2B-DsRed was co-expressed by agroinfiltration as a nuclear marker. Leaves were examined for GFP signal at 2 days post agroinfiltration by fluorescence microscopy (Leica DMI 6000B with filter block L5 containing a 480/40 nm excitation filter, a 505 nm dichroic mirror and a 527/30 nm barrier filter for GFP fluorescence), and images were taken using Leica LAS AF software.

Leptomycin B treatment

The leptomycin B (LMB) treatment was carried out following the method described in Mei et al. (2018) [16] with slight modifications. GFP-tagged 2bs were overexpressed by agroinfiltration in N. benthamiana. The leaf tissues in the agroinfiltrated patches were infiltrated with LMB (40 nM) at 2 days post agroinfiltration, and the leaves were excised at 4 h after the LMB treatment for microscopic observation. The subcellular localization of 2b-GFP was then observed by fluorescence microscopy (Leica DMI 6000B).

Detection of 2b in the nuclear-rich fraction

GFP- or FLAG-tagged 2b was expressed in N. benthamiana leaves after agroinfiltration, and the nuclear-rich fractions were isolated as previously described [35]. 2b-GFP or 2b-FLAG in the nuclear-rich fraction was then detected using western blot analysis with anti-GFP antibodies or anti-FLAG antibody. Histone H3 (H3) was detected using anti-H3 antibodies as a loading control. To compare the levels of phosphorylated 2b in the nuclear-rich fraction, the phosphoproteins in the nuclear-rich fraction were concentrated using the TALON PMAC phosphoprotein enrichment kit (Clontech, Tokyo, Japan) and the manufacturer’s instruction (S5 Fig); the phosphorylated 2b was detected by western blot analysis using anti-2b antibodies.

In vitro phosphorylation assay

Seven putative phosphorylation sites including one threonine and six serines were predicted by the NetPhos v2.0 program (http://www.cbs.dtu.dk/services/NetPhos-2.0/). Truncated mutant 2bs were generated by PCR and inserted between the EcoRI and SalI restriction enzyme sites in pMAL-c2x (NEB, Beverly, MA, USA). The predicted phosphorylation sites were replaced with alanine using the primer pairs listed in S1 Table. Full-length mutant 2bs were also created by recombinant PCR using the primer pairs listed in S1 Table, then cloned into pMAL-c2x using the EcoRI and SalI restriction enzyme sites. BL21 E. coli competent cells were transformed with the pMAL-c2x vectors containing the mutant 2bs. The MBP-2b fusion proteins were expressed in E. coli and harvested as previously described [47]. The extracted proteins were then purified using the pMAL Protein Fusion and Purification System (NEB, Beverly, MA, USA) according to the instructions. The components in the elution buffer were exchanged with the kinase reaction buffer (10 mM MgCl₂, 1 mM DTT and 20 mM HEPES [pH 7.5]) using a Slide-A-Lyzer MINI Dialysis Unit (Thermoscientific, Rockford, IL, USA). Purified proteins were incubated in the kinase reaction buffer (total 15 μL reaction volume) containing 37 nM ATP, 0.25% Complete Mini Protease Inhibitor Cocktail (Roche, Basel, Switzerland) and 0.5 μL of protein kinase C (Promega, Madison, WI, USA). The in vitro phosphorylated proteins were then separated by SDS-PAGE and detected using the ProQ-Diamond Phosphoprotein Gel Stain kit (Invitrogen, Carlsbad, CA, USA).

In vivo phosphorylation assay

Total protein extracts were prepared from infected leaves of N. tabacum plants at 6 days post inoculation. Phosphorylated proteins were concentrated using the TALON PMAC phosphoprotein enrichment kit (Clontech) according to the manufacturer’s instructions (S5 Fig). Equal amounts of total protein extracts were loaded on a phosphoprotein enrichment column, then the phosphoproteins were eluted from the column with the included elution buffer. Phosphorylated 2bs were then separated by 14% SDS-PAGE and detected with western blot analysis using anti-2b antibodies [47].

Co-immunoprecipitation assay

The CDS region of the NbIMPα1 gene (GenBank: ABM05487.1) was amplified, and the FLAG tag (NbIMPα1-FLAG) was added to the C-terminus by RT-PCR using the primer pairs listed in S1 Table. The NbIMPα1-FLAG sequence was then cloned in the pBE2113 binary vector between the XbaI and SacI restriction enzyme sites. The Agrobacterium strains KYRT1 carrying the NbIMPα1-FLAG and 2b-GFP constructs were used to co-infiltrate into N. benthamiana leaves at a 1:1 ratio with OD₆₀₀ = 1.0. Infiltrated leaf patches (0.2 g) were homogenized in 1 mL of the lysis buffer (20 mM HEPES pH 7.5, 1 mM EDTA, 40 mM KCl, 0.1% (v/v) Triton-X 100, 1 mM PMSF, 10% (v/v) glycerol) containing 1X protease inhibitor cocktail (Sigma Aldrich) [78], and the soluble fraction of the homogenates was extracted by two times centrifugation at 10,000 × g for 10 min. The total protein extract in the soluble fraction was co-incubated with by shaking. After washing, the bound proteins were eluted (Dynabeads Protein G Immunoprecipitation Kit; Invitrogen). The precipitant was then detected by western blot analysis using anti-GFP antibodies (MBL).

Detection of 2b in the cytoplasmic fraction

The cytoplasmic fraction was isolated as shown in S14 Fig. Agroinfiltrated N. benthamiana leaf tissues were homogenized in 5 mL of the extraction buffer (0.35 M sucrose, 50 mM Tris-Cl, pH 7.5, 5 mM MgCl₂, 2 mM 2-mercaptoethanol) per gram of tissue. Nuclei, plastids and cell wall components were removed by centrifugation at 2500 × g for 10 min, 4°C, and the supernatant was then centrifuged at 6000 × g for 10 min, 4°C to eliminate mitochondria. The final supernatant after centrifugation was used for immunoprecipitation. Cytoplasmic 2b was detected using the Dynabeads Protein G Immunoprecipitation kit (Invitrogen) with anti-2b antibodies and the manufacturer’s instructions. The precipitant was then detected by western blot analysis using anti-FLAG antibody (Fujifilm).

Analysis of 2b’s RSS activity

The conventional agroinfiltration assay for RSS activity was conducted as previously described [22]. The GFP and GFP inverted-repeat (GFP-IR) constructs were inserted in the pBE2113 vector between the XbaI and SacI sites. The pBE2113 expression vectors containing 2b, GFP and GFP-IR were introduced into A. tumefaciens strain KYRT1. Inocula consisted of a combination of Agrobacterium cultures carrying a 5:1:5 ratio of GFP, GFP-IR and 2b constructs. N. benthamiana leaves were agroinfiltrated with bacterial suspensions (~1.0 OD₆₀₀). The suspension of Agrobacterium cells that carried either 2b or S28A was serially diluted to OD₆₀₀ = 0.5, 0.25 and 0.13. GFP signals were observed at 2 days post infiltration under UV light. For infiltrating onion epidermis cells, the bacterial resuspensions were adjusted to OD₆₀₀ = 0.2 for each GFP, GFP-IR and 2b construct. GFP fluorescence was observed at 3 days post infiltration in Leica DMI 6000B, and the intensity was calculated using the Leica LAS AF software.

Quantitative real-time RT-PCR

Total RNAs were isolated using Trizol (Roche) and treated by DNase I to remove DNA contaminants. cDNAs were synthesized using the PrimeScript RT Reagent Kit (Takara, Tokyo, Japan), then used for real-time RT-PCR with the StepOnePlus Real-Time PCR system (Applied Biosystems, Foster City, CA, USA). Primer pair CMV-DET-5/CMV-DET-3 was used to quantify the levels of CMV RNAs 3 and 4; primer pair Nb-L23-5/Nb-L23-3 was used as a reference for amplification of a partial mRNA sequence of 60S ribosomal protein L23 in N. benthamiana (S1 Table). For amplification of GFP transcripts, primer pair S65T-5-168/S65T-3-168 was used (S1 Table).