Characterization of a DCL2-Insensitive Tomato Bushy Stunt Virus Isolate Infecting Arabidopsis thaliana

Tomato bushy stunt virus (TBSV), the type member of the genus Tombusvirus in the family Tombusviridae is one of the best studied plant viruses. The TBSV natural and experimental host range covers a wide spectrum of plants including agricultural crops, ornamentals, vegetables and Nicotiana benthamiana. However, Arabidopsis thaliana, the well-established model organism in plant biology, genetics and plant–microbe interactions is absent from the list of known TBSV host plant species. Most of our recent knowledge of the virus life cycle has emanated from studies in Saccharomyces cerevisiae, a surrogate host for TBSV that lacks crucial plant antiviral mechanisms such as RNA interference (RNAi). Here, we identified and characterized a TBSV isolate able to infect Arabidopsis with high efficiency. We demonstrated by confocal and 3D electron microscopy that in Arabidopsis TBSV-BS3Ng replicates in association with clustered peroxisomes in which numerous spherules are induced. A dsRNA-centered immunoprecipitation analysis allowed the identification of TBSV-associated host components including DRB2 and DRB4, which perfectly localized to replication sites, and NFD2 that accumulated in larger viral factories in which peroxisomes cluster. By challenging knock-out mutants for key RNAi factors, we showed that TBSV-BS3Ng undergoes a non-canonical RNAi defensive reaction. In fact, unlike other RNA viruses described, no 22nt TBSV-derived small RNA are detected in the absence of DCL4, indicating that this virus is DCL2-insensitive. The new Arabidopsis-TBSV-BS3Ng pathosystem should provide a valuable new model for dissecting plant–virus interactions in complement to Saccharomyces cerevisiae.


Introduction
Tomato bushy stunt virus (TBSV) is the type member of the genus Tombusvirus in the family Tombusviridae. First isolated in 1935 in Ireland [1], this soil-borne plant pathogen for which no biological vector is known, is readily transmitted by mechanical inoculation. Symptoms induced by TBSV are largely dependent on the plant host and vary from necrotic and chlorotic lesions, to a mild or severe mosaic, or even lethal systemic necrosis in extreme cases. TBSV has a host range limited in nature mostly to dicotyledonous species including agricultural crops, ornamentals and vegetables such as the RNA-induced silencing complex (RISC), to mediate sequence-specific degradation of viral RNA. The main AGO proteins involved in defense against ssRNA viruses in plants have been shown to be AGO1 and AGO2 [32,[35][36][37]. The P19 protein of TBSV, arguably the best-studied VSR, has been shown in heterologous systems to suppress RNAi by sequestering siRNA and preventing their loading into AGO proteins [34,38]. While P19 can bind siRNA duplex products of all DCL proteins in planta [39], the very high affinity for 21 nt small RNA in vivo and in vitro [18,[39][40][41] suggests that TBSV specifically relies on sequestration of DCL4 products to suppress antiviral RNAi.
In this study, we describe an isolate of TBSV that efficiently infects A. thaliana Col-0 ecotype, resulting in rapid and consistent systemic infection of all plants tested. We provide insight into the 3D structure of peroxisome-bound vesicles harboring VRCs in infected plants. Moreover, we identify and localize host dsRNA-binding proteins associated to TBSV VRCs. Finally, by taking advantage of the genetic tools available in Arabidopsis, we characterize the non-canonical RNAi response to TBSV infection in planta.

Virus Infection and Transgene Transient Expression
Lyophilized tissue from N. glutinosa, Chenopodium quinoa and Petunia infected with TBSV BS3 [48], used as TBSV positive control in ELISA, was obtained from Bioreba, Switzerland (Bioreba, Art-No:161853). The inoculum was propagated by rub-inoculation on N. benthamiana followed by the harvest of systemically infected leaves, which were used to infect A. thaliana. This was achieved by grinding the infected tissue in liquid nitrogen and resuspending in 50 mM sodium phosphate buffer, pH 5.8. After clearing the debris by centrifugation for 2 min at 2000× g and transferring the supernatant to a new tube, the resulting inoculum was gently rubbed onto Arabidopsis leaves previously sprinkled with celite. After a few minutes, leaves were rinsed with water. The pEAQ-∆P19 plasmids expressing 35S:tRFP, 35S:DRB2:tRFP, 35S:DRB4:tRFP and 35S:tRFP:NFD2 were previously described [46]. For overexpression and TBSV infection experiments, leaves of 5-6 week-old 35S:B2:GFP/N. benthamiana were infiltrated with A. tumefaciens GV3101 carrying the pEAQ∆P19 plasmid of interest, at absorbance 600nm (A 600 ) of 0.2. Bacteria were incubated in 10 mM MES pH 5.6, 10 mM MgCl 2 , 200 µM acetosyringone for 1 h, then infiltrated with a syringe without a needle. The following day, TBSV inoculum was applied as above on the infiltrated tissues, on the adaxial side of the leaves. Molecular/microscopy analysis was carried out 3 days after infection.

Virus Isolation and HTS Sequence Analyses
TBSV was purified from infected N. benthamiana and genomic RNA extracted as described [49]. Analyses of HTS dataset was performed using Workbench 12.0 software (CLC bio Genomics, Aarhus, Denmark), as previously described [50]. Briefly, after the trimming procedure and quality check, only reads above 70 nucleotides (nts) were kept. De novo assembly was then performed using word size = 17 and contig length min = 200 as parameters. Contigs were then tested against TBSV reference sequences and against NCBI references using BlastN/BlastX (http://blast.ncbi.nlm.nih.gov/Blast.cgi, last visited 04/2020). Multiple sequence alignments comparing TBSV-BS3 and all available full length TBSV sequences were performed using CLUSTALW [51] and maximum likelihood-based phylogenetic trees were conducted MegaX software [52]. The best ML-fitted model for each sequence alignment (nucleic or amino acid) was used. Nodes in phylogenetic trees and branch validity were evaluated by bootstrap analyses (100 replicates).

Peroxisome Isolation
The detailed procedure of Arabidopsis peroxisome isolation is described in [39,53].

Immunoprecipitation
Immunoprecipitation procedures were carried out as in [46,54]. Rosette leaves (0.15 g) were ground in liquid nitrogen, homogenized in a mortar with 1 mL lysis buffer (50 mM Tris-HCl, pH 8, 50 mM NaCl and 1% Triton X-100) containing a protease inhibitor (Roche, Basel, Switzerland), transferred to a tube, incubated for 15 min at 4 • C on a wheel with slow rotation. Lysate was clarified by two successive centrifugations at 12,000× g for 10 min at 4 • C, after which an aliquot of supernatant was set aside on ice as input. The remaining lysate was incubated with anti-GFP magnetic beads (µMACS purification system, Miltenyi Biotech, Bergisch Gladbach, Germany, catalog number #130-091-125) at 4 • C for 20 min. A sample was passed through the M column (MACS purification system, Miltenyi Biotech) and an aliquot of the flow-through was set aside on ice. The M column was washed twice with 500 µL lysis buffer and once with 100 µL of washing buffer (20 mM Tris-HCl, pH 7.5). The beads and associated immune complexes were recovered by removing the column from the magnetic stand and passing 1 mL of Tri Reagent (for subsequent RNA analysis-see dedicated section) or 200 µL hot 1X Laemmli buffer (for protein analysis-see dedicated section). 4X Laemmli buffer was added to input and flow-through fractions before protein denaturation for 5 min at 95 • C.

Protein Extraction and Analysis
Immunoprecipitated proteins for mass spectrometry were isolated as described above, then directly denatured for 5 min at 95 • C. Immunoprecipitated proteins from RNA IP were obtained by transferring to a new tube 400 µL of the phenolic phase following Tri-reagent/chloroform extraction (see RNA analysis section), adding 3 vol acetone, mixing by inversion and incubating at −20 • C O/N. After centrifugation (13,000 rpm, 15 min, 4 • C) the pellet was washed in 80% acetone, resuspended in 1X Laemmli and denatured for 5 min at 95 • C. Proteins were resolved by SDS-PAGE and electroblotted onto Immobilon-P PVDF membrane. This was incubated with anti-GFP polyclonal antibody and revealed with Roche LumiLight ECL kit following incubation with the secondary antibody.

RNA Extraction and Analysis
RNA extraction from total and immunoprecipitated fractions was performed with Tri-Reagent (Sigma, St. Louis, USA) according to the manufacturer's instructions. For total tissue, 0.2 g of frozen tissue were ground in liquid nitrogen and homogenized in 1 mL of Tri-Reagent, while for immunoprecipitated fractions 1 mL of Tri-reagent was passed through the columns as described above. 400 µL of chloroform was added and the sample was vortexed/shaken for 2 min. After 10 min centrifugation at 13,000 rpm, 4 • C, the supernatant was transferred to new tube and 1 vol isopropanol was added (and 1.5 µL glycogen in the case of immunoprecipitated samples) and incubated 1 h on ice (O/N at −20 • C for IP). After a 15 min spin at 13,000 rpm, 4 • C (30 min for IP), the supernatant was discarded, 350 µL of 80% ethanol was added, tubes were spun for a further 5 min, the supernatant was discarded and the pellet was dried and resuspended in water. High molecular weight Northern blot was performed by electrophoresis of 5 µg of total RNA on 1% agarose, HEPES pH 7.4, 6% formaldehyde gels followed by capillary transfer in SSC 20x onto Amersham HyBond N+ nylon membrane and UV crosslinked for 1 min. Low molecular weight Northern blot was performed by electrophoresis of 10 µg total RNA on 15% polyacrylamide, 0.5x TBE, 7.5 M urea gels followed by electro-transfer Viruses 2020, 12, 1121 5 of 21 onto Amersham HyBond NX nylon membrane and 1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide EDC chemical crosslinking [39]. miR159, miR173, TAS1 (ta-siRNA255) and snU6 were detected through DNA oligonucleotides labeled with γ-32 P-ATP using T4 PNK (Thermo Fischer, Waltham, USA). TBSV genomic and subgenomic RNAs were detected in the same way, with an oligonucleotide complementary to 3 -terminal nucleotides of the viral genome. TBSV-derived and IR71 siRNA were detected through PCR products labeled by Klenow reaction (Promega, Madison, USA) in the presence of α-32 P-dCTP. The cDNA to use for this PCR was obtained by using the SuperScriptIII kit (Invitrogen, Carlsbad, USA) with a TBSV-specific primer (same oligo used for probe) on RNA from TBSV-infected plants. Northern blots hybridization was carried out O/N at 42 • C in PerfectHyb (Sigma), followed by 3 washes of 15 min in 2X SSC, 2% SDS at 50 • C. In the Northwestern blots, dsRNA was detected through recombinant Strep-Tagged FHV B2 after migration of 5 µg total RNA at 4 • C on non-denaturing HEPES, 1% agarose gel and capillary transfer, as previously described [47]. To perform RT-qPCR, first cDNA was generated using SuperScript IV (Invitrogen) with random priming, then qPCR was performed with the SYBR-Green kit (Roche) in a LightCycler 480 Real-Time PCR System (Roche). Using UBIQUITIN 10 (UBQ10) and ACTINE2 mRNA as a reference, relative expression levels were calculated using the ∆∆Ct method. List and sequences of primers and oligonucleotides used in this study are described (Supplementary Table S1).

Mass Spectrometry Protein Analysis
Mass spectrometry procedures were carried out as in [46]. Proteins were digested with sequencing-grade trypsin (Promega) and analyzed by nanoLC-MS/MS on a TripleTOF 5600 mass spectrometer (Sciex, Framingham, MA, USA), as previously described [55]. Data were searched against the TAIR v.10 database with a decoy strategy (27,281 protein forward sequences). Peptides were identified with the Mascot algorithm (version 2.5, Matrix Science, London, UK) and data were further imported into the Proline v1.4 software (http://proline.profiproteomics.fr/). Proteins were validated on Mascot pretty rank equal to 1, and 1% FDR on both peptide spectrum matches (PSM score) and protein sets (Protein Set score). The total number of MS/MS fragmentation spectra was used to quantify each protein from at least three independent biological replicates. A statistical analysis based on spectral counts was performed using a homemade R package as described in [56]. The R package uses a negative binomial GLM model based on EdgeR [57] and calculates, for each identified protein, a fold-change, a p-value and an adjusted p-value corrected using the Benjamini-Hochberg method.

Fluorescence Microscopy
Leaf disks from N. benthamiana leaves were collected 3 days after rub inoculation (4 days after agro-infiltration), placed on standard microscopy slides, covered with coverslips, immersed in water by pipetting and placed in the vacuum chamber until air had been removed. Observations of leaf disks were carried out using Zeiss LSM700 and LSM780 laser scanning confocal microscopes. eGFP was excited at 488 nm, tRFP was excited at 561 nm. Image processing was carried out with ImageJ/FIJI, while figure panels were assembled with Adobe Photoshop and Affinity Photo.

Electron Microscopy
N. benthamiana and A. thaliana leaves were fixed overnight at 4 • C with 3% glutaraldehyde in 0.05 M phosphate buffer saline pH 7.5. Samples were then washed 3 × 10 min before post-fixation with 1.5% potassium ferrocyanide reduced, 2% osmium for 1 h at room temperature. Leaves were then treated with filtered thiocarbohydrazide (TCH) for 20 min at RT. After 3 × 10 min washes with ddH 2 O samples were stained with 2% osmium tetroxide for 30 min and with 1% uranyl acetate overnight at 4 • C. The next day, a fresh solution of en bloc Walton's lead aspartate solution was used to stain samples for 30 min at 60 • C (0.7% of lead nitrate in L-aspartic acid, pH 5.5). Leaves were then dehydrated with an ethanol series and infiltrated with EPON812 resin. Leaves were directly mounted on an aluminum 3View pin and embedded in 100% resin for 72 h at 60 • C. Samples were trimmed with an DM3 files from the 3View microscope were compiled into a TIFF stack using the ImageJ-Fiji software [58]. The area of interest was cropped and segmentation was performed manually using trakEM2 [59]. The outlines of the object of interest were drawn using a Wacom graphics tablet. The 3D representation of peroxisomes was performed with the ImageJ-Fiji volume viewer plugin. Three-dimensional rendering was achieved with the display volume mode and tricubic smoothing interpolation.

Characterization of a Tomato Bushy Stunt Virus (TBSV) Isolate Infecting Arabidopsis thaliana
Our initial aim was to identify a TBSV isolate able to infect the model plant Arabidopsis thaliana. To do so, TBSV isolate BS3 [48] initially propagated on Nicotiana glutinosa, petunia and Chenopodium quinoa was used as a source of inoculum and mechanically inoculated on Arabidopsis Col-0 ecotype. 9 days post-inoculation (dpi), newly emerged uninoculated leaves were screened by Northern blotting using as a probe the highly conserved 3 end sequence from TBSV. All three sources of inoculum displayed infectivity on Col-0 plants (Supplementary Figure S1), however only TBSV isolate BS3 initially propagated on N. glutinosa, named TBSV-BS3Ng hereafter, led to 100% infection rate (8/8 plants) accompanied by mild symptoms on systemic leaves ( Figure 1A,B) and was therefore further characterized. Firstly, the virus was fully sequenced by high-throughput sequencing (HTS). RNA from purified virions produced >2.3 × 10 6 clean reads of an average size of 136.8 nts, out of which >99% mapped the de novo assembled viral genome. Due to the very high sequencing depth (>67 K), the genome sequence and organization were easily determined, confirming the virus classification in the family Tombusviridae, genus Tombusvirus ( Figure 1C and Supplementary Figure S2). The sequence was deposited to GenBank under accession number MT856702.

3D Electron Microscopy of TBSV Replication Vesicles on Peroxisomes from Infected Plants
All positive-strand (+) ssRNA viruses replicate their genomes by co-opting intracellular membranes and subverting host components, thereby enabling the assembly of the replication organelles. The structure, composition and formation of replication organelles vary greatly between different Viruses 2020, 12, 1121 7 of 21 viruses [61][62][63]. The TBSV replication organelle has been extensively studied in various host plants and Saccharomyces cerevisiae [15,61,64]. Replication per se takes place inside numerous separate membrane invaginations called spherules that derive from peroxisomes and the endoplasmic reticulum [12,15,65]. To test if TBSV-BS3Ng replication is also associated to the production of spherules, we first infected N. benthamiana constitutively expressing the dsRNA-binding sensor protein B2:GFP [47]. Confocal and electron microscopy analyses of these plants confirmed that TBSV-BS3Ng replicates in association with severely modified peroxisomes containing numerous spherules (Supplementary Figure S3 and [46]). Confocal and electron microscopy analyses of constitutively-expressing B2:GFP A. thaliana Col-0 plants infected with TBSV-BS3Ng ( Figure 2) revealed similar cytopathic structures than those observed in N. benthamiana (Supplementary Figure S3). At low magnification, B2:GFP was redistributed from a nucleo-cytoplasmic localization ( Figure 2A) to numerous fluorescent aggregates ( Figure 2B) upon infection with TBSV-BS3Ng. Confocal analysis of these aggregates at higher magnification revealed numerous clustered ring-like structures very similar to those observed in N. benthamiana corresponding to peroxisomes (compare Figure 2C,D to Supplementary Figure S3A). Arrays of individual serial block face SEM (SBEM) images taken from 35S B2:GFP A. thaliana plants infected with TBSV-BS3Ng at 9 dpi revealed altered peroxisomes containing numerous spherules ( Figure 2E,F). Three-dimensional reconstruction of a cytopathic peroxisome revealed that all spherules were embedded inside the lumen of the organelle ( Figure 2G) similarly to N. benthamiana (Supplementary Figure S3E). Altogether it was concluded that TBSV-BS3Ng was able to infect A. thaliana in a systemic manner and its replication occurred on peroxisome-associated spherules. Our results also indicate that B2:GFP had access to the dsRNA present within spherules.

Immunoprecipitation of TBSV Replication Complexes
After having established that the TBSV-BS3Ng isolate forms replication vesicles on the membranes of peroxisomes in Arabidopsis, we sought to isolate and characterize the VRC proteome directly from purified peroxisomes. To do so, and taking advantage of our experience with the peroxisome isolation procedure [34,39], we first isolated peroxisomes from non-infected and TBSV-infected dcl2-1/dcl4-2 mutant plants. Mass spectrometry (MS) analysis of peroxisomes isolated from healthy and TBSV-infected plants revealed numerous hits corresponding to peroxisomal proteins as expected (Supplementary Table S2). However, although peptides from the TBSV P41 and the P19 VSR were detected in the peroxisome fraction from infected plants only, the infected peroxisome extracts contained no peptides matching the TBSV replicase, the hallmark component of VRCs. The absence of replicase peptides led us to conclude that this experiment did not achieve sufficient isolation of VRCs. It is likely that the cytopathic peroxisomes ( Figure 2) display different biophysical properties that prevent them from being isolated by density gradient fractionation using conventional procedures [34,39]. In addition, the large number of host proteins present in peroxisomal fractions (Supplementary Table S2 and [34]) may interfere with the detection of possibly small quantities of viral proteins by mass spectrometry.
Given these results, and considering B2:GFP localizes to TBSV replication complexes in N. benthamiana (Supplementary Figure S3 and [46]) and A. thaliana (Figure 2), we decided to isolate VRCs using the technique we have recently developed for Tobacco rattle virus (TRV) [46,54]. This procedure enables the identification of TRV dsRNA-associated proteome using B2:GFP as bait for immunoprecipitation (IP). Molecular analysis of the immunoprecipitated fractions revealed that TBSV RNA ( Figure 3A) and dsRNA ( Figure 3B) were co-immunoprecipitated with B2:GFP, but not with GFP. Interestingly, contrarily to TBSV-infected plant expressing GFP only in which vsiRNA were abundantly produced, B2:GFP-expressing plants accumulated little if any detectable levels of TBSV-derived siRNA ( Figure 3C). Finally, the Western blot analysis of IP fractions using GFP antibodies revealed that GFP and B2:GFP were specifically immunocaptured as expected ( Figure 3D). We then performed anti-GFP IP on B2:GFP/TBSV-infected leaves in triplicate, and analyzed the immunoprecipitated proteins by mass spectrometry. Nine controls from three genotypes/conditions, Viruses 2020, 12, 1121 9 of 21 comprising of anti-GFP IPs performed in GFP/TBSV-infected leaves and IPs performed in non-infected B2-GFP or GFP leaves were included in the analysis. The comparison of B2:GFP IPs performed in TBSV-infected leaves to the 9 controls revealed a total of 11 proteins significantly enriched in the IPs with an adjusted p-value < 0.05 (Supplementary Table S3) as shown in the volcano plot representation ( Figure 3E). As expected, these include the TBSV proteins P19, P92 and P41 ( Figure 3E, blue spots) as well as eight host proteins: AT1G24450 (NFD2), AT3G55410, AT3G62800 DRB4), AT5G55070, AT1G52400, AT4G26910, AT1G30120 and AT2G28380 (DRB2). Remarkably, among these significantly-enriched proteins, NFD2 (Nuclear Fusion Deficient 2), DRB2 (Double-stranded RNA Binding 2) and DRB4 (Double-stranded RNA Binding 4; Figure 3E, red spots) belong also to the TRV dsRNA-associated proteome [46].

Immunoprecipitation of TBSV Replication Complexes
After having established that the TBSV-BS3Ng isolate forms replication vesicles on the membranes of peroxisomes in Arabidopsis, we sought to isolate and characterize the VRC proteome directly from purified peroxisomes. To do so, and taking advantage of our experience with the peroxisome isolation procedure [34,39], we first isolated peroxisomes from non-infected and TBSVinfected dcl2-1/dcl4-2 mutant plants. Mass spectrometry (MS) analysis of peroxisomes isolated from immunoprecipitation (IP). Molecular analysis of the immunoprecipitated fractions revealed that TBSV RNA ( Figure 3A) and dsRNA ( Figure 3B) were co-immunoprecipitated with B2:GFP, but not with GFP. Interestingly, contrarily to TBSV-infected plant expressing GFP only in which vsiRNA were abundantly produced, B2:GFP-expressing plants accumulated little if any detectable levels of TBSV-derived siRNA ( Figure 3C). Finally, the Western blot analysis of IP fractions using GFP antibodies revealed that GFP and B2:GFP were specifically immunocaptured as expected ( Figure 3D).

Host Double-Stranded RNA-Binding Proteins Associate to TBSV Replication Complexes
Considering NFD2, DRB2 and DRB4 are associated to TRV replication complexes [46] and were also found to be significantly enriched upon dsRNA IP from TBSV-infected plants ( Figure 3E), we further investigated their association with TBSV replication complexes by confocal microscopy. To do so, we transiently expressed tRFP-tagged candidates in 35S:B2:GFP/N. benthamiana leaves and subsequently infected them with TBSV, as previously described [46]. In accordance with our previous report on TRV infection [46], tRFP localization was unaffected upon TBSV infection, showing a constant nuclear-cytoplasmic distribution, while B2:GFP surrounded clustered peroxisomes ( Figure 4A, Supplementary Figure S4A). As expected, DRB2:tRFP colocalized with B2:GFP at the periphery of peroxisomes upon infection with TBSV (Supplementary Figure S5 and [46]). Similarly, striking relocalization of DRB4:tRFP from the nucleus in healthy cells (Supplementary Figure S4B and [46]) to B2-labeled peroxisome clusters was observed upon TBSV infection ( Figure 4B). This relocalization of DRB4 from the nucleus to VRCs is in line with published data [46,66]. Observation at a higher magnification of the "bunch of grape"-like structures corresponding to clustered peroxisomes revealed two types or localization for B2:GFP, bright peripheral rings as well as dim intraperoxisomal punctate structures ( Figure 4B, bottom panels). From our electron microscopy data it is likely that the peripheral B2 labeling corresponds to spherules connected to the outer peroxisomal membrane while internal labeling likely points to intraperoxisomal spherules ( Figure 2F,G and Supplementary Figure S3D,E). Remarkably, DRB4:RFP colocalized only with the peripheral spherules (that appeared yellow upon merging of the channels) and not with the intraperoxisomal spherules (arrowheads, Figure 4B). This suggests that B2:GFP and DRB4:tRFP differ in their capacity to access spherule-associated dsRNA within peroxisomes. However, the significance of this difference remains to be determined.
Finally, as with DRB2 and DRB4, TBSV infection also strongly influenced the intracellular localization of NFD2, a putative RNAse III-Like protein also known as RTL4 [67], which concentrated in the matrix surrounding the grape-like peroxisome clusters ( Figure 4C and Supplementary Figure  S4C). In this case, no colocalization between NFD2 and B2 was observed, suggesting that NFD2 is associated with the larger viral factory but not with the spherule-restricted replication complexes per se in which genomic viral dsRNA accumulate.
Next, we wondered whether transient overexpression of tRFP-tagged DRB2, DRB4 and NFD2 in 35S:B2:GFP/N. benthamiana has any effect on TBSV accumulation. Northern blot analysis performed on RNA extracted from two independent pools of agroinoculated and infected tissues showed a drastic reduction in TBSV RNA accumulation in tissues expressing DRB2:tRFP ( Figure 5A). This is well in agreement with analogous results we obtained in wild-type N. benthamiana [46]. In contrast, while no clear difference was observed upon DRB4:tRFP and tRFP overexpression, a slight increase in TBSV accumulation was detected upon tRFP:NFD2 overexpression ( Figure 5A). We then wondered whether knock-out of these proteins could affect TBSV accumulation, viral dsRNA patterns or TBSV-derived siRNA accumulation. Unfortunately, since NFD2 homozygous mutation leads to sterility [68], mutants were not available. We therefore restricted our analysis to drb2-1, drb4-1 and drb2-1/drb4-1 mutants [42]. Northern analysis of total RNA from systemically infected leaves at 10 dpi revealed a moderate increase in TBSV RNA accumulation in all drb2-1, drb4-1 and drb2-1/drb4-1 mutants compared to wild-type plants ( Figure 5B). Northwestern blot revealed no evident changes in the accumulation or patterns of dsRNA ( Figure 5C). Finally, a moderate decrease in TBSV-derived siRNA was observed in lines defective in DRB4 function (drb4, Figure 5D), consistently with the known role of DRB4 as a player in antiviral siRNA biogenesis and a cofactor of DCL4 [42,66,69,70].

DCL2 Does not Effectively Process TBSV dsRNA
The bulk of genetic data available on TBSV infection was generated on surrogate host S. cerevisiae, which does not possess the RNA interference molecular machinery. Although the RNAi machinery has been artificially reconstituted in yeast [71] and used to investigate anti-TBSV RNAi [30], genetic data are missing in plants. We therefore sought to take advantage of the genetic tools available in A. thaliana, along with the body of data generated on antiviral RNAi in this species, to characterize the RNAi response to TBSV in planta. Considering that DCL2 and DCL4 proteins are among the principal actors in host antiviral defense [31], we infected Col-0, dcl2-1, dcl4-2 and dcl2/4 with TBSV and analyzed total RNA from systemic leaves ( Figure 6). Northern blot analysis of RNA from independent pools of plants revealed a moderate increase of TBSV genomic RNA accumulation in the dcl4 and dcl2/4 lines compared to wild-type plants ( Figure 6A). As upon overexpression of tRFP:NFD2 ( Figure 5A), whether this observed increase in TBSV accumulation is statistically significant remains to be determined. However, the viral increase in the absence of both dicers is consistent with observations on other viruses expressing strong VSR proteins [31][32][33][34]72]. The fact that the strongest effects on TBSV genomic RNA accumulation was observed in dcl4 and dcl2/4 suggests that DCL2 may not participate significantly in RNAi against TBSV. Moreover, Northwestern blot analysis revealed that in the absence of DCL4 a distinct dsRNA species over-accumulated (arrow, Figure 6B). The relatively small size of this dsRNA suggested that it could be derived from one of the TBSV subgenomic RNAs produced from the 3 Viruses 2020, 12, 1121 12 of 21 portion of the virus. Since this dsRNA species was barely detected in the presence of DCL4, we reasoned that it could be a prime target for siRNA generation by DCL4. We therefore analyzed siRNA derived from subgenomic RNAs (@P19) and the genomic RNA only (mid, Figure 6C) and found that siRNA from both regions could be readily detected, suggesting that this dsRNA species is not the only substrate of DCL4. Sequencing experiments coupled with isolation and sequencing of long dsRNA are required to better elucidate the nature and processing of this dsRNA species.
Surprisingly, but in agreement with the results described above, virtually no antiviral siRNA, in particular the DCL2-derived 22 nt siRNA species could be detected upon DCL4 knock-out ( Figure 6C). This indicates that DCL2 does not process TBSV dsRNA, in contrast with the hierarchical action of DCL4 and DCL2 described for other viruses [31][32][33][34].
In an attempt to confirm these findings and gain further insight into this atypical RNAi response, we repeated the TBSV infections in duplicate and included mutants for the two key Argonaute proteins involved in defense against RNA viruses, AGO1 and AGO2 [73]. Northern and Northwestern blot RNA analysis revealed again enhanced TBSV replication ( Figure 7A) and the apparition of novel dsRNA species ( Figure 7B, arrow) in dcl4 and dcl2/4 mutants. An increase in TBSV systemic accumulation could also be observed in the ago1-27 and ago2-1 mutants ( Figure 7A), suggesting that both AGO1 and AGO2 participate in RNAi against TBSV.
Analysis of TBSV-derived siRNA also confirmed the previous results ( Figure 6C), with the near complete depletion in siRNA species in plants defective in DCL4 ( Figure 7C, upper panel). The highly ineffective dicing by DCL2 upon DCL4 knock-out suggests that DCL2 could be inactive during TBSV infection. To test this hypothesis, RT-qPCR quantification of mDCL2 gene expression was performed in different dcl genetic backgrounds upon infection. As expected, the wild-type DCL2 mRNA was expressed in all lines except in dcl2 and dcl2/4 mutants ( Figure 7D). Since an antibody to efficiently detect DCL2 is not available, the accumulation of the protein in infected tissues remains to be determined. To partially bypass this issue, we tested DCL2 functionality by assessing the accumulation of 22 nt siRNA from IR71, an endogenous substrate of DCL2 [74,75]. In line with DCL2 being expressed and active upon TBVS-BS3Ng infection, IR71-derived 22 nt siRNA accumulated in healthy and infected Col-0, as well as in infected dcl4 mutants, but not in dcl2 and dcl2/4 mutants ( Figure 7C). In contrast, a dramatic increase in IR71 processing by DCL3 leading to the overaccumulation of 24 nt siRNA was observed with all genotypes upon TBSV infection (compare samples from infected plants to healthy Col-0 plants). Interestingly, this phenotype strongly resembles that of drb4 mutants [75] and may result from the virus-induced relocalization of DRB4 to TBSV replication complexes during infection [46,66]. Altogether it is concluded that DCL2 is expressed and functional in TBSV-BS3Ng-infected Arabidopsis but is ineffective as a surrogate to DCL4 to trigger an antiviral response against TBSV-BS3Ng.
Further analysis of endogenous small RNA showed that TBSV had no effect on ta-siRNA processing by DCL4, and that smaller additional species of miR159 and miR173 accumulated in infected tissues ( Figure 7C). These "trimmed" miRNA molecules are highly reminiscent of those observed in transgenic plants expressing the CymRSV and TBSV P19 proteins in N. benthamiana and A. thaliana, respectively [39,76]. Finally, accumulation of TBSV-derived siRNA was reduced in the hypomorphic ago1-27 mutant compared to wild-type and ago2-1 mutant ( Figure 7E). This could be due to changes in dicing, small RNA stability and/or AGO1 loading, for example. Further experiments on a battery of RNAi mutant combinations will hopefully provide further insight into the molecular events taking place during the atypical RNAi response to TBSV in planta. The lower row of three acquisitions is shown to highlight the intraperoxisomal B2 foci (white arrowheads). For the sake of simplicity, the labeling was omitted on these but is identical to that in the upper row. (C) Observation of tissues expressing 35S:tRFP:NFD2. All acquisitions were performed with a 63× objective. Scale bars indicate 5-10 µm, as indicated. Lower magnification acquisitions from these tissues, along with the respective non-infected controls, can be found in Supplementary Figure S4. Viruses 2020, 12, x FOR PEER REVIEW 13 of 21 siRNA was observed in lines defective in DRB4 function (drb4, Figure 5D), consistently with the known role of DRB4 as a player in antiviral siRNA biogenesis and a cofactor of DCL4 [42,66,69,70]. Northern blot analysis of low molecular weight RNA to detect TBSV-derived siRNA. As loading control, the same membrane was separately hybridized to miR159-and snU6-specific probes. In all the blots a non-infected control (n.i.) was included on the far left. Except for the non-infected control, two samples were analyzed per genotype/transgene (1 and 2 in the labels).

DCL2 Does not Effectively Process TBSV dsRNA
The bulk of genetic data available on TBSV infection was generated on surrogate host S. cerevisiae, which does not possess the RNA interference molecular machinery. Although the RNAi machinery has been artificially reconstituted in yeast [71] and used to investigate anti-TBSV RNAi [30], genetic data are missing in plants. We therefore sought to take advantage of the genetic tools available in A. thaliana, along with the body of data generated on antiviral RNAi in this species, to characterize the RNAi response to TBSV in planta. Considering that DCL2 and DCL4 proteins are among the principal actors in host antiviral defense [31], we infected Col-0, dcl2-1, dcl4-2 and dcl2/4 with TBSV and analyzed total RNA from systemic leaves ( Figure 6). Northern blot analysis of RNA from independent pools of plants revealed a moderate increase of TBSV genomic RNA accumulation in the dcl4 and dcl2/4 lines compared to wild-type plants ( Figure 6A). As upon overexpression of tRFP:NFD2 ( Figure 5A), whether this observed increase in TBSV accumulation is statistically In (A-C) EtBr gel staining was used as a loading control. (D) Northern blot analysis of low molecular weight RNA to detect TBSV-derived siRNA. As loading control, the same membrane was separately hybridized to miR159-and snU6-specific probes. In all the blots a non-infected control (n.i.) was included on the far left. Except for the non-infected control, two samples were analyzed per genotype/transgene (1 and 2 in the labels).
Viruses 2020, 12, x FOR PEER REVIEW 14 of 21 significant remains to be determined. However, the viral increase in the absence of both dicers is consistent with observations on other viruses expressing strong VSR proteins [31][32][33][34]72]. The fact that the strongest effects on TBSV genomic RNA accumulation was observed in dcl4 and dcl2/4 suggests that DCL2 may not participate significantly in RNAi against TBSV. Moreover, Northwestern blot analysis revealed that in the absence of DCL4 a distinct dsRNA species over-accumulated (arrow, Figure 6B). The relatively small size of this dsRNA suggested that it could be derived from one of the TBSV subgenomic RNAs produced from the 3′ portion of the virus. Since this dsRNA species was barely detected in the presence of DCL4, we reasoned that it could be a prime target for siRNA generation by DCL4. We therefore analyzed siRNA derived from subgenomic RNAs (@P19) and the genomic RNA only (mid, Figure 6C) and found that siRNA from both regions could be readily detected, suggesting that this dsRNA species is not the only substrate of DCL4. Sequencing experiments coupled with isolation and sequencing of long dsRNA are required to better elucidate the nature and processing of this dsRNA species. Surprisingly, but in agreement with the results described above, virtually no antiviral siRNA, in particular the DCL2-derived 22 nt siRNA species could be detected upon DCL4 knock-out ( Figure  6C). This indicates that DCL2 does not process TBSV dsRNA, in contrast with the hierarchical action of DCL4 and DCL2 described for other viruses [31][32][33][34].
In an attempt to confirm these findings and gain further insight into this atypical RNAi response, we repeated the TBSV infections in duplicate and included mutants for the two key Argonaute proteins involved in defense against RNA viruses, AGO1 and AGO2 [73]. Northern and Analysis of TBSV-derived siRNA also confirmed the previous results ( Figure 6C), with the near complete depletion in siRNA species in plants defective in DCL4 ( Figure 7C, upper panel). The highly ineffective dicing by DCL2 upon DCL4 knock-out suggests that DCL2 could be inactive during TBSV infection. To test this hypothesis, RT-qPCR quantification of mDCL2 gene expression was performed in different dcl genetic backgrounds upon infection. As expected, the wild-type DCL2 mRNA was expressed in all lines except in dcl2 and dcl2/4 mutants ( Figure 7D). Since an antibody to efficiently detect DCL2 is not available, the accumulation of the protein in infected tissues remains to be determined. To partially bypass this issue, we tested DCL2 functionality by assessing the

Discussion
In this study we described a new isolate of TBSV that infects A. thaliana. This should prove to be an extremely valuable asset in the genetic investigation of host-Tombusvirus interactions, which for the moment have been investigated essentially using yeast and N. benthamiana. The localization of the VRCs on Arabidopsis peroxisomes suggests that their formation and function may follow similar pathways as those dissected using yeast, reinforcing the potential of this TBSV-BS3Ng isolate as an investigative tool. One aspect of TBSV-host interaction that cannot be thoroughly studied in yeast is antiviral RNAi, which is a central player in plant antiviral defenses. Our experiments have shown that the Arabidopsis RNAi reaction to TBSV is unique when compared to other RNA viruses. While DCL2 has been shown to act as a surrogate to DCL4 against TRV, TCV, TuMV and CMV [31][32][33], or in parallel to DCL4 against PCV [34], during TBSV infection DCL2 is unable to efficiently generate vsiRNA. Since DCL2 is present in TBSV-infected tissues and able to process endogenous targets, we proposed that either (i) DCL2 is unable to access TBSV dsRNA and/or to process it, or (ii) TBSV inhibits a putative antiviral pool of DCL2 that is distinct from that processing IR71. Given that DCL2 is arguably the least known among the Arabidopsis dicer proteins, there is little data to help explain these observations. However, future investigation into this phenomenon may shed light on the activity of DCL2 as an antiviral effector.
The possibility of TBSV-BS3Ng infection inhibiting the antiviral pool of DCL2 is compelling. It would imply that TBSV has evolved a layered RNAi suppression mechanism, where the primary DCL4-dependent RNAi reaction is neutralized through P19-mediated sequestration of 21 nt vsiRNA, while the "back-up" DCL2-dependent RNAi is also somehow neutralized upstream of dicing likely by a TBSV-encoded protein that remains to be identified. Furthermore, the exclusive processing by DCL4 may explain why P19 has evolved highest binding affinity for 21 nt-long RNA [19,[39][40][41]. Along with the observation that PCV, which is abundantly processed by DCL2, has evolved to efficiently sequester 22 nt siRNA through its P15 VSR [34], these findings suggest that the VSR proteins of different viruses adapt to neutralize the specific dicer products driving antiviral RNAi.
The inability of DCL2 to generate 22 nt TBSV vsiRNA is in agreement with the absence of vsiRNA in TBSV-infected 35S:B2:GFP/Col-0 plants. In these plants, we have previously shown that infection by TRV leads to the production of DCL2-dependent 22nt vsiRNA, as opposed to wild-type plants where vsiRNA are produced mainly by DCL4, suggesting an inhibitory activity of B2:GFP on DCL4 [46]. Upon TBSV infection, DCL4 inhibition by B2 and DCL2 inability to generate vsiRNA likely leads to the observed absence of 21-22 nt-long vsiRNA. The identification of DRB2 and DRB4 associated to VRCs in 35S:B2:GFP/Col-0 plants suggests that their binding of dsRNA is vsiRNA-independent.
Little is known about the viral dsRNA substrates of DCL4. Many works have investigated the distribution of vsiRNA along viral genomes in many species (reviewed in [77]), but these may be heavily biased toward those siRNA species that are highly stable. As we have shown, the homeostasis of viral long dsRNA in relation to dicer proteins and other RNAi components can be easily assessed by Northwestern blotting. Our results indicate that during infection DCL4 is responsible for the depletion of a dsRNA of reduced length, presumably through dicing into vsiRNA of a double-stranded subgenomic RNA. Alternatively, this could be a product of host-encoded RDR proteins that is stabilized by the absence of DCL4. However, since Northwestern blotting does not allow one to distinguish RNA sequence, it cannot be ruled out that the dsRNA species observed is actually host-encoded. This could explain why no subgenomic RNA was seen by Northern blot to over-accumulate in dcl4 mutants compared to wild-type plants. This discrepancy between Northern and Northwestern blots could also be explained if DCL4 were depleting dsRNA while not destroying the corresponding ssRNA, for example through exclusive helicase activity. Further work will hopefully unravel the molecular mechanisms behind these observations.
Our results obtained with B2:GFP IPs followed by mass spectrometry further validate the potential of this experimental approach to characterize VRC proteomes in planta. Our microscopy experiments on DRB2, DRB4 and NFD2 are highly consistent with what we observed during TRV infection [46], and provide compelling and detailed snapshots of the TBSV VRCs on cytopathic peroxisomes. Furthermore, the results obtained with DRB2 and DRB4 are in agreement with previous reports [66,77]. We have previously described the broad-ranged antiviral activity of Arabidopsis DRB2 [46]. The results we obtained through overexpression of DRB2 are in contrast with similar experiments on N. benthamiana DRB2B, which when overexpressed increases the accumulation of PVX [78]. However, the authors of this study also show that silencing of DRB2B causes an increase in systemic infectivity of PVX, uncovering an antiviral activity of this protein that is well in agreement with our observation on Arabidopsis DRB2. Our observations that DRB4 was not strictly necessary for DCL4-dependent production of TBSV-derived siRNA is in agreement with published studies on TuMV and TCV [42,70]. However, DRB4 has been shown to be mandatory for DCL4 activity during infection by TSWV and TYMV [42,66], suggesting that DRB4-DCL4 interplay is highly dependent on virus species. Interestingly, the moderate increase in TBSV accumulation observed in drb4 mutants is reminiscent of that observed in dcl4 mutants, suggesting that either (i) the vsiRNA produced by DCL4 in the absence of DRB4 are not functional in mediating antiviral RNAi or (ii) DRB4 plays an antiviral role independently of DCL4, or both.
Our identification of NFD2 in association with dsRNA during infection by both TRV and TBSV, and its localization in close proximity to VRCs, suggests that this protein may play important roles in the viral life cycle as a putative RNAseIII. The observation that NFD2 overexpression leads to increased TBSV accumulation is in agreement with the report that overexpression of RTL1, a related RNAse III-like protein, leads to increased accumulation of TCV, TVCV and TYMV in Arabidopsis [79], most likely by depriving Dicer enzymes of substrate dsRNA.
The fact that B2:GFP blocks dicer activity on TBSV dsRNA, as commented above, means that B2:GFP IPs are limited in their use to investigate the association of RNAi factors to TBSV VRCs. This problem could be circumvented by developing the use of a recombinant B2 to isolate VRCs from wild-type infected plants ex vivo. Furthermore, the gradient fractionation of cytopathic peroxisomes, which was not further pursued in this study, could be optimized to allow molecular analysis of the VRC-containing vesicles present on their surface.