The Luteovirus P4 Movement Protein Is a Suppressor of Systemic RNA Silencing

The plant viral family Luteoviridae is divided into three genera: Luteovirus, Polerovirus and Enamovirus. Without assistance from another virus, members of the family are confined to the cells of the host plant’s vascular system. The first open reading frame (ORF) of poleroviruses and enamoviruses encodes P0 proteins which act as silencing suppressor proteins (VSRs) against the plant’s viral defense-mediating RNA silencing machinery. Luteoviruses, such as barley yellow dwarf virus-PAV (BYDV-PAV), however, have no P0 to carry out the VSR role, so we investigated whether other proteins or RNAs encoded by BYDV-PAV confer protection against the plant’s silencing machinery. Deep-sequencing of small RNAs from plants infected with BYDV-PAV revealed that the virus is subjected to RNA silencing in the phloem tissues and there was no evidence of protection afforded by a possible decoy effect of the highly abundant subgenomic RNA3. However, analysis of VSR activity among the BYDV-PAV ORFs revealed systemic silencing suppression by the P4 movement protein, and a similar, but weaker, activity by P6. The closely related BYDV-PAS P4, but not the polerovirus potato leafroll virus P4, also displayed systemic VSR activity. Both luteovirus and the polerovirus P4 proteins also showed transient, weak local silencing suppression. This suggests that systemic silencing suppression is the principal mechanism by which the luteoviruses BYDV-PAV and BYDV-PAS minimize the effects of the plant’s anti-viral defense.


Small RNA Deep Sequencing and Bioinformatic Analyses
Total RNA was extracted from barley plants infected with BYDV-PAV using the miRNeasy kit (Qiagen, Hilden, Germany), according to the manufacturer's instructions. BYDV-PAV small RNA library was prepared (Illumina small RNA sample preparation kit, Illumina, San Diego, CA, USA) and deep sequenced on the Illumina GAIIx platform (Illumina, San Diego, CA, USA) according to the manufacturer's protocols. Deep sequencing reads were preprocessed (read quality assessment, adapter removal, read size extraction) utilizing the Fastx Toolkit set of scripts (http://hannonlab.cshl. edu/fastx_toolkit/index.html) and in-house custom perl scripts. Small RNA reads between 21 and 24 nt were mapped to the BYDV-PAV reference genome using Bowtie v0.12.7 [23], with no mismatches allowed. Read distribution profile visualization and coverage analyses were performed utilizing custom perl scripts, mpileup from SAMtools [24], and the OpenOffice (in Ubuntu Linux) spreadsheet application. For comparisons of read distribution profiles, reads were normalized against the total read count (after preprocessing) per 10 million reads, and plotted as number of reads per base.

Transient Expression Assay in N. benthamiana
Transgenic N. benthamiana (line 16c), which is homozygous for the GFP transgene [25], and the Agrobacterium infiltration method have been described previously [31]. For co-infiltrations of wild-type N. benthamiana or line 16c, cultures of A. tumefaciens (strain GV3101) harboring a relevant binary plasmid, were mixed before infiltration to a final OD 600 = 0.5 each, except for L6 and CED-9 (OD 600 = 1). Dilutions of P4 PAV , P4 PAS and P4 PL were obtained by adding appropriate concentrations of a vector-less Agrobacterium culture to the infiltrate solutions. The GFP fluorescence was monitored with a hand-held UV lamp as previously described [31]. Pictures (UV and visible light) were taken with a Canon EOS 550D digital camera (Canon, Tokyo, Japan).

RNA Blot Analysis
Total RNA was extracted from agroinfiltrated leaves of N. benthamiana 16c collected at 6 dpi, with TRIzol reagent (Invitrogen, Carlsbad, CA, USA). The manufacturer's protocol was modified as an additional phenol:chloroform wash was added before an overnight precipitation step at −20 • C. For high and low molecular weight Northern blots, 20 µg and 40 µg of total RNAs were loaded and run on formaldehyde and polyacrylamide gels, respectively. RNA detection and hybridization conditions were previously described [20].

BYDV-PAV Elicits an RNA Silencing Response in the Phloem
BYDV-PAV is known to circularize its RNA for cap-independent translation of its genome [32]. It also produces a highly abundant non-coding sgRNA3 ( Figure 1B) [33]. We wondered if, in the absence of a P0 VSR, these attributes might in some way protect BYDV-PAV against the plant's anti-viral RNA degradation system in the phloem. Therefore, we examined the small RNA (sRNA) profiles of virus-infected plants. Total RNA from BYDV-PAV-infected barley was extracted, size-fractionated, and the sRNAs sequenced using an Illumina machine. The predominant class of sRNAs, identical or complementary in sequence to the BYDV-PAV genome, was 22 nt (502k reads) with less abundant 21 nt (328k reads) and 24 nt (146k reads) size classes ( Figure 1C). This profile of abundant siRNAs (representing 9.76% of total sRNAs) shows that the silencing machinery in phloem tissues mounts a response to BYDV-PAV infection, suggesting that circularization of BYDV-PAV genomic RNA for translation does not strongly inhibit this process, as shown for other viruses (e.g., [34]). Despite the abundance of sgRNA3, which might have acted as a decoy to saturate the plant's silencing machinery and thereby protect the viral genome from degradation, there did not appear to be a predominance of siRNAs from this region, suggesting that BYDV-PAV employs a different silencing suppression or avoidance strategy.

Assaying BYDV-PAV-Encoded Proteins for Silencing Suppressor Activity
In an attempt to identify potential VSR candidates in BYDV-PAV, we compared its ORF sequences with those of known suppressors, including rice yellow mottle virus (RYMV) P1, tomato bushy stunt virus (TBSV) P19, turnip crinkle virus (TCV) P38, cucumber mosaic virus (CMV) 2b, potato virus Y (PVY) HCPro and five P0s (data not shown). We also compared the translated ORF sequences using TBLASTn and BLASTp against accessions in the NCBI plant viral database containing the keyword "suppressor". Neither analysis produced any obvious VSR ortholog for any of the six BYDV ORFs. Therefore, the proteins P1-2, P2, P3, P4, P3-5 and P6 encoded by the BYDV-

Assaying BYDV-PAV-Encoded Proteins for Silencing Suppressor Activity
In an attempt to identify potential VSR candidates in BYDV-PAV, we compared its ORF sequences with those of known suppressors, including rice yellow mottle virus (RYMV) P1, tomato bushy stunt virus (TBSV) P19, turnip crinkle virus (TCV) P38, cucumber mosaic virus (CMV) 2b, potato virus Y (PVY) HCPro and five P0s (data not shown). We also compared the translated ORF sequences using TBLASTn and BLASTp against accessions in the NCBI plant viral database containing the keyword "suppressor". Neither analysis produced any obvious VSR ortholog for any of the six BYDV ORFs. Therefore, the proteins P1-2, P2, P3, P4, P3-5 and P6 encoded by the BYDV-PAV genome were tested for local and systemic VSR activity using the 16c system [35]. The virus genes were placed under the control of the 35S promoter in binary vectors and transiently co-expressed with a sense GFP transgene (sGFP) in fully-expanded leaves of N. benthamiana line 16c. This line has an integrated GFP transgene [25,31] and is a well stablished assay for detecting suppression activity, as a bright green fluorescent tissue is observed under the UV light 3-7 days post infiltration (dpi) if the candidate protein is able to block the locally-induced silencing triggered by the sGFP transgene. If the co-infiltrated virus gene has no systemic VSR activity, the infiltrated tissue emits a signal that spreads through the phloem and emerges in the apex and young leaves of the plant to silence the GFP; co-infiltration with a systemic VSR delays or negates the emergence of distal silencing. The percentage of plants that fail to show distal silencing after 2-3 weeks is taken as a measure of the candidate genes' systemic VSR activity [36].
Using this system and the well-described local VSR, P19, from TBSV as a control, BYDV-PAV P1-2, P2, P3, P3-5 and P6 showed little or no evidence of local VSR activity at 6 dpi ( Figure 2A), even when multiple proteins were co-expressed (data not shown). This contrasts with a report that P6 from BYDV-GAV has local VSR activity [37], but this may be due to the marked sequence divergence of GAV from other BYDV isolates, especially in the P6 ORF sequence ( Figure S1). BYDV-PAV P4 (P4 PAV ), however, showed a weak, transient local VSR activity that could be detected through a mild increase in GFP fluorescence when compared to control empty vector. Assessing the spread of the silencing signal from infiltrated spots at 14 dpi revealed that P4 PAV and BYDV-PAV P6 (P6 PAV ) produce strong and moderate systemic silencing suppression, respectively, when compared to the well-known systemic VSR P1 from RYMV (P1 RY ) ( Figure 2B-D). The spread of the silencing signal was detected in only 20% of plants co-infiltrated with P4 PAV , a profile of systemic VSR activity equivalent to that of P1 RY , while 40% of plants showed systemic silencing when co-infiltrated with P6 PAV . This same moderate systemic VSR activity has also been reported for BYDV-GAV P6 [37]. PAV genome were tested for local and systemic VSR activity using the 16c system [35]. The virus genes were placed under the control of the 35S promoter in binary vectors and transiently coexpressed with a sense GFP transgene (sGFP) in fully-expanded leaves of N. benthamiana line 16c. This line has an integrated GFP transgene [25,31] and is a well stablished assay for detecting suppression activity, as a bright green fluorescent tissue is observed under the UV light 3-7 days post infiltration (dpi) if the candidate protein is able to block the locally-induced silencing triggered by the sGFP transgene. If the co-infiltrated virus gene has no systemic VSR activity, the infiltrated tissue emits a signal that spreads through the phloem and emerges in the apex and young leaves of the plant to silence the GFP; co-infiltration with a systemic VSR delays or negates the emergence of distal silencing. The percentage of plants that fail to show distal silencing after 2-3 weeks is taken as a measure of the candidate genes' systemic VSR activity [36].
Using this system and the well-described local VSR, P19, from TBSV as a control, BYDV-PAV P1-2, P2, P3, P3-5 and P6 showed little or no evidence of local VSR activity at 6 dpi ( Figure 2A), even when multiple proteins were co-expressed (data not shown). This contrasts with a report that P6 from BYDV-GAV has local VSR activity [37], but this may be due to the marked sequence divergence of GAV from other BYDV isolates, especially in the P6 ORF sequence ( Figure S1). BYDV-PAV P4 (P4 PAV ), however, showed a weak, transient local VSR activity that could be detected through a mild increase in GFP fluorescence when compared to control empty vector. Assessing the spread of the silencing signal from infiltrated spots at 14 dpi revealed that P4 PAV and BYDV-PAV P6 (P6 PAV ) produce strong and moderate systemic silencing suppression, respectively, when compared to the well-known systemic VSR P1 from RYMV (P1 RY ) ( Figure 2B-D). The spread of the silencing signal was detected in only 20% of plants co-infiltrated with P4 PAV , a profile of systemic VSR activity equivalent to that of P1 RY , while 40% of plants showed systemic silencing when co-infiltrated with P6 PAV . This same moderate systemic VSR activity has also been reported for BYDV-GAV P6 [37].  When analyzing the short-distance spread of the silencing signal at 7 dpi, it became apparent that P4 PAV , but not P6 PAV , reduces the development of the red "halo" ( Figure 3A) which normally reports the initial cell-to-cell spread of the mobile silencing signal to adjacent cells outside the infiltrated area [38][39][40]. This result shows that the VSR activity by P4 PAV delays the cell-to-cell movement of the systemic silencing signal. This assay also showed that P4 PAV , but not P6 PAV , induces necrosis of the infiltrated tissue, which, under the blue light used to excite GFP, gives a yellowish fluorescence making it difficult to visually assess the GFP expression. The necrosis induced by P4 PAV could be largely prevented by adding a CED-9-encoding construct to the P4 PAV treatment ( Figure 3A), since CED-9 is an antiapoptotic protein from C. elegans shown to suppress viral-induced necrosis in plants [41]. Inhibition of P4 PAV -triggered necrosis by CED-9 did not impact on P4 PAV VSR activity, since the delay of red halo formation at 7 dpi or 11 dpi ( Figures 3A and S2), or the strong suppression of systemic silencing by P4 PAV (Figure 3B), remained unaffected in plants agro-infiltrated with a mix containing P4 PAV and CED-9 constructs. These results suggest that the VSR activity by P4 PAV and induction of necrosis are not linked. When analyzing the short-distance spread of the silencing signal at 7 dpi, it became apparent that P4 PAV , but not P6 PAV , reduces the development of the red "halo" ( Figure 3A) which normally reports the initial cell-to-cell spread of the mobile silencing signal to adjacent cells outside the infiltrated area [38][39][40]. This result shows that the VSR activity by P4 PAV delays the cell-to-cell movement of the systemic silencing signal. This assay also showed that P4 PAV , but not P6 PAV , induces necrosis of the infiltrated tissue, which, under the blue light used to excite GFP, gives a yellowish fluorescence making it difficult to visually assess the GFP expression. The necrosis induced by P4 PAV could be largely prevented by adding a CED-9-encoding construct to the P4 PAV treatment ( Figure 3A), since CED-9 is an antiapoptotic protein from C. elegans shown to suppress viral-induced necrosis in plants [41]. Inhibition of P4 PAV -triggered necrosis by CED-9 did not impact on P4 PAV VSR activity, since the delay of red halo formation at 7 dpi or 11 dpi ( Figure 3A and Figure S2), or the strong suppression of systemic silencing by P4 PAV (Figure 3B), remained unaffected in plants agro-infiltrated with a mix containing P4 PAV and CED-9 constructs. These results suggest that the VSR activity by P4 PAV and induction of necrosis are not linked.

Systemic Silencing Suppression and Necrosis Is Produced by P4 from Two Different Luteoviruses, but Not by P4 from a Polerovirus
To explore whether P4 proteins in other luteoviruses and poleroviruses also have VSR activity, the appropriate ORF was cloned from BYDV-PAS (P4 PAS ) and from PLRV (P4 PL ), and placed under the control of the CaMV 35S promoter. Testing the constructs in the 16c system with the standard agro-inoculum strength of 0.5 OD ( Figure 4A top and Figure 4B) showed that P4 PAS , but not P4 PL , induced a delay in red halo formation and onset of necrosis in a very similar way to P4 PAV . With the 0.5 OD agro-inoculum, the necrosis induced by P4 PAV or P4 PAS was barely discernable to the naked eye at 4 dpi and obvious at 7 dpi ( Figure 4A top and Figure 4B), and the boundaries of the spots were as devoid of red halos as the P19 spots. Diluting the P4 PAV and P4 PAS infiltrate solutions five-fold ( Figure 4A bottom and Figure 4B 1:5 dilution), with buffer containing vector-less agrobacterium, led to some halo induction and reduction of necrosis, which indicates a dose-dependent effect by both P4 proteins on red halo formation and necrosis induction. The five-fold dilution of P4 PAV and P4 PAS infiltrate solutions also produced a mild increase in GFP fluorescence in infiltrated spots at 4 dpi when compared to the control empty vector, and this difference to the control was still discernable at 7 dpi, although less obvious. This mild increase in GFP fluorescence compared to the control was also noted in spots infiltrated with P4 PL at 4 dpi, but no longer detectable at 5 dpi ( Figure 4A top and bottom). These results indicate that both P4 PAV and P4 PAS produce weak and transient local VSR activity, which is even less pronounced in the case of P4 PL .

Systemic Silencing Suppression and Necrosis Is Produced by P4 from Two Different Luteoviruses, but Not by P4 from a Polerovirus
To explore whether P4 proteins in other luteoviruses and poleroviruses also have VSR activity, the appropriate ORF was cloned from BYDV-PAS (P4 PAS ) and from PLRV (P4 PL ), and placed under the control of the CaMV 35S promoter. Testing the constructs in the 16c system with the standard agro-inoculum strength of 0.5 OD ( Figure 4A top and Figure 4B) showed that P4 PAS , but not P4 PL , induced a delay in red halo formation and onset of necrosis in a very similar way to P4 PAV . With the 0.5 OD agro-inoculum, the necrosis induced by P4 PAV or P4 PAS was barely discernable to the naked eye at 4 dpi and obvious at 7 dpi ( Figure 4A top and Figure 4B), and the boundaries of the spots were as devoid of red halos as the P19 spots. Diluting the P4 PAV and P4 PAS infiltrate solutions five-fold ( Figure 4A bottom and Figure 4B 1:5 dilution), with buffer containing vector-less agrobacterium, led to some halo induction and reduction of necrosis, which indicates a dose-dependent effect by both P4 proteins on red halo formation and necrosis induction. The five-fold dilution of P4 PAV and P4 PAS infiltrate solutions also produced a mild increase in GFP fluorescence in infiltrated spots at 4 dpi when compared to the control empty vector, and this difference to the control was still discernable at 7 dpi, although less obvious. This mild increase in GFP fluorescence compared to the control was also noted in spots infiltrated with P4 PL at 4 dpi, but no longer detectable at 5 dpi ( Figure 4A top and bottom). These results indicate that both P4 PAV and P4 PAS produce weak and transient local VSR activity, which is even less pronounced in the case of P4 PL .  Single leaves of young 16c plants were infiltrated with a cocktail of sGFP and one of the P4 or control vectors, and observed for 20 days for the emergence of systemic silencing. P4 PAV and P4 PAS , but not P4 PL , gave profiles of strong systemic suppressors ( Figure 5A), and this lack of VSR activity by P4 PL (Figure 5A,D) could be explained by the low sequence conservation (~53%) between the polerovirus P4 and the two luteovirus P4 proteins which are highly conserved (~91% sequence identity) ( Figure S1). In this experiment, both P4 PAV and P4 PAS showed stronger VSR activities than P1 RY , and using different agro-infiltrate concentrations of the P4 constructs revealed that P4 PAS is slightly weaker than P4 PAV ( Figure 5B,C), with their systemic VSR activities being dose-dependent. Single leaves of young 16c plants were infiltrated with a cocktail of sGFP and one of the P4 or control vectors, and observed for 20 days for the emergence of systemic silencing. P4 PAV and P4 PAS , but not P4 PL , gave profiles of strong systemic suppressors ( Figure 5A), and this lack of VSR activity by P4 PL (Figure 5A,D) could be explained by the low sequence conservation (~53%) between the polerovirus P4 and the two luteovirus P4 proteins which are highly conserved (~91% sequence identity) ( Figure S1). In this experiment, both P4 PAV and P4 PAS showed stronger VSR activities than P1 RY , and using different agro-infiltrate concentrations of the P4 constructs revealed that P4 PAS is slightly weaker than P4 PAV ( Figure 5B,C), with their systemic VSR activities being dose-dependent.

P4 PAV and P4 PAS Reduce the Production or Accumulation of siRNAs
To investigate the mechanism of the systemic VSR activity of P4 PAV and P4 PAS , the production of siRNAs was measured by Northern blots. Agroinfiltration of 16c plants with the 35S:GFP transgene is known to induce the production of siRNAs against the GFP mRNA by an RNA-dependent RNA

P4 PAV and P4 PAS Reduce the Production or Accumulation of siRNAs
To investigate the mechanism of the systemic VSR activity of P4 PAV and P4 PAS , the production of siRNAs was measured by Northern blots. Agroinfiltration of 16c plants with the 35S:GFP transgene is known to induce the production of siRNAs against the GFP mRNA by an RNA-dependent RNA polymerase-mediated process and these secondary siRNAs guide the destruction of the GFP mRNA [42,43]. Total RNA extracts were made from 16c leaves infiltrated with different cocktails of agrobacterium cultures containing constructs for the expression of GFP and candidate or control VSR genes, and collected at 6 dpi. The RNA extracts were size fractionated, run on gels, transferred to membranes and probed with the 300-nt 3 sequence of the GFP gene. The blots (Figure 6) show that the GFP mRNA level dropped from its endogenous level (16c) in tissues infiltrated with the 35S:GFP transgene (Vec) and became elevated in leaf tissues co-infiltrated with the transgene and P19 or P0 PL . Tissues co-infiltrated with 35S:GFP and P1 RY , P4 PAS , or P4 PL showed GFP mRNA levels similar to those of the GFP transgene alone. Interestingly, P4 PAV -infiltrated tissues had slightly increased GFP mRNA accumulation, which could be almost nullified by a five-fold dilution of the P4 PAV agro-inoculum concentration, confirming the weak local VSR activity by P4 PAV . As expected, the enhancement of GFP mRNA levels by P19 and P0 were mirrored by reduced levels of GFP-specific siRNAs. This is consistent with their local VSR activities, since P19 affects siRNA stability and production, while P0 affects the production of siRNAs [12,20,44,45]. The siRNA levels of all size classes were also considerably reduced in the tissues infiltrated with P4 PAV and P4 PAS , with a minor reduction in tissues treated with P4 PL or the diluted P4 PAV inoculum.
The visible GFP fluorescence of P4 PAV and P4 PAS infiltrated leaves ( Figure 6) and the GFP mRNA levels in samples taken from them ( Figure 6, GFP mRNA blot) were not greatly elevated, unlike those treated with P19 or P0. However, the GFP-specific siRNAs in the P4 PAV and P4 PAS samples were nearly as low as in the P19 or P0 treated samples ( Figure 6, siRNA blot). This suggests that systemic VSR activities of P4 PAV and P4 PAS are related to their capacity to reduce the local production or accumulation of secondary siRNAs, which could be explained, in part, by their weak local VSR activities. However, P4 PAV and P4 PAS probably operate differently to either P19 or P0 since they lack strong local silencing suppression activity. The different increase in GFP mRNA accumulation at 6 dpi ( Figure 6, GFP mRNA blot) confirms that P4 PAS is slightly weaker than P4 PAV in its VSR activity. polymerase-mediated process and these secondary siRNAs guide the destruction of the GFP mRNA [42,43]. Total RNA extracts were made from 16c leaves infiltrated with different cocktails of agrobacterium cultures containing constructs for the expression of GFP and candidate or control VSR genes, and collected at 6 dpi. The RNA extracts were size fractionated, run on gels, transferred to membranes and probed with the 300-nt 3′ sequence of the GFP gene. The blots ( Figure 6) show that the GFP mRNA level dropped from its endogenous level (16c) in tissues infiltrated with the 35S:GFP transgene (Vec) and became elevated in leaf tissues co-infiltrated with the transgene and P19 or P0 PL . Tissues co-infiltrated with 35S:GFP and P1 RY , P4 PAS , or P4 PL showed GFP mRNA levels similar to those of the GFP transgene alone. Interestingly, P4 PAV -infiltrated tissues had slightly increased GFP mRNA accumulation, which could be almost nullified by a five-fold dilution of the P4 PAV agro-inoculum concentration, confirming the weak local VSR activity by P4 PAV . As expected, the enhancement of GFP mRNA levels by P19 and P0 were mirrored by reduced levels of GFP-specific siRNAs. This is consistent with their local VSR activities, since P19 affects siRNA stability and production, while P0 affects the production of siRNAs [12,20,44,45]. The siRNA levels of all size classes were also considerably reduced in the tissues infiltrated with P4 PAV and P4 PAS , with a minor reduction in tissues treated with P4 PL or the diluted P4 PAV inoculum. The visible GFP fluorescence of P4 PAV and P4 PAS infiltrated leaves ( Figure 6) and the GFP mRNA levels in samples taken from them ( Figure 6, GFP mRNA blot) were not greatly elevated, unlike those treated with P19 or P0. However, the GFP-specific siRNAs in the P4 PAV and P4 PAS samples were nearly as low as in the P19 or P0 treated samples ( Figure 6, siRNA blot). This suggests that systemic VSR activities of P4 PAV and P4 PAS are related to their capacity to reduce the local production or accumulation of secondary siRNAs, which could be explained, in part, by their weak local VSR activities. However, P4 PAV and P4 PAS probably operate differently to either P19 or P0 since they lack strong local silencing suppression activity. The different increase in GFP mRNA accumulation at 6 dpi ( Figure 6, GFP mRNA blot) confirms that P4 PAS is slightly weaker than P4 PAV in its VSR activity. . P4 PAV was also infiltrated at a five-fold dilution (*P4 PAV ). Non-infiltrated plants (16c) were also included as negative controls. Pictures were taken under the UV light and leaf samples collected for Northern blots at 6 dpi. Total RNA (20 µg) was blotted and probed for the presence of GFP mRNA. rRNA was used as a loading control for high-molecular-weight RNA. Total RNA (40 µg) from the agro-infiltrated tissues was also blotted and probed for the presence of GFP siRNAs. U6 was used as a loading control for low-molecular weight RNA. RNA size markers are indicated.

Induction of Cell Death by P4 PAV and P4 PAS Does Not Prevent Viral Replication and Spread
As both BYDV and the only other described luteovirus, bean leafroll virus, are known to induce phloem necrosis in their hosts [46][47][48], it is possible that the P4s of these viruses could be the elicitors of this response. This suggests that these viruses would have evolved to not only fight the anti-viral silencing machinery but to also escape from cells undergoing host-induced cell death in the . P4 PAV was also infiltrated at a five-fold dilution (*P4 PAV ). Non-infiltrated plants (16c) were also included as negative controls. Pictures were taken under the UV light and leaf samples collected for Northern blots at 6 dpi. Total RNA (20 µg) was blotted and probed for the presence of GFP mRNA. rRNA was used as a loading control for high-molecular-weight RNA. Total RNA (40 µg) from the agro-infiltrated tissues was also blotted and probed for the presence of GFP siRNAs. U6 was used as a loading control for low-molecular weight RNA. RNA size markers are indicated.

Induction of Cell Death by P4 PAV and P4 PAS Does Not Prevent Viral Replication and Spread
As both BYDV and the only other described luteovirus, bean leafroll virus, are known to induce phloem necrosis in their hosts [46][47][48], it is possible that the P4s of these viruses could be the elicitors of this response. This suggests that these viruses would have evolved to not only fight the anti-viral silencing machinery but to also escape from cells undergoing host-induced cell death in the vasculature.
For this to be true, the virus has to replicate and egress from the infected cell before the cell dies from necrotic processes induced by P4.
To test whether the processes induced by P4 PAV and P4 PAS might allow viral replication and spread to occur before the death of infected cells, a PVX-GFP infectious clone was agroinfiltrated in various combinations with the P4 and control constructs into wild-type N. benthamiana leaves using the concentrations that induce the cell death response. As expected, agroinfiltration of the PVX-GFP construct with P4 PL , P1 RY , or vector backbone, which do not induce necrosis, produced PVX-GFP that, at 5 dpi, was able to replicate and spread from the infiltration spots to the nearest veins, then pass via the petiole ( Figure 7A) to the apex of the plant ( Figure 7B). Despite inducing tissue necrosis, the PVX-GFP/P4 PAV and PVX-GFP/P4 PAS infiltrations with agro-inoculum strength of 0.5 OD also generated PVX that was able to spread systemically at this same time point (Figure 7A,B). Moreover, even the co-infiltration of PVX-GFP with a fungal resistance gene (L6), which also induces cell death, allowed PVX to replicate and spread systemically at 6 dpi.
This assay with a GFP-encoding PVX construct also identifies local VSRs, since PVX-GFP replicates to a higher level and expresses its GFP gene more strongly and for longer in the presence of a VSR [20]. The co-infiltration of PVX-GFP with P4 PAV or P4 PAS constructs led to a slight increase in GFP fluorescence in infiltrated spots when compared to P4 PL , P1 RY , or vector backbone ( Figure 7A,C), which confirms the weak local VSR activity by P4 PAV and P4 PAS . Although the two luteovirus P4 proteins did not display strong local silencing suppression activity during PVX replication when compared to P0 PL , P0 CY and P19 ( Figure 7C), the replicated virus was still able to exit before the death of the infected cell, induced by P4.
Viruses 2017, 9, 294 11 of 16 vasculature. For this to be true, the virus has to replicate and egress from the infected cell before the cell dies from necrotic processes induced by P4.
To test whether the processes induced by P4 PAV and P4 PAS might allow viral replication and spread to occur before the death of infected cells, a PVX-GFP infectious clone was agroinfiltrated in various combinations with the P4 and control constructs into wild-type N. benthamiana leaves using the concentrations that induce the cell death response. As expected, agroinfiltration of the PVX-GFP construct with P4 PL , P1 RY , or vector backbone, which do not induce necrosis, produced PVX-GFP that, at 5 dpi, was able to replicate and spread from the infiltration spots to the nearest veins, then pass via the petiole ( Figure 7A) to the apex of the plant ( Figure 7B). Despite inducing tissue necrosis, the PVX-GFP/P4 PAV and PVX-GFP/P4 PAS infiltrations with agro-inoculum strength of 0.5 OD also generated PVX that was able to spread systemically at this same time point (Figure 7A,B). Moreover, even the co-infiltration of PVX-GFP with a fungal resistance gene (L6), which also induces cell death, allowed PVX to replicate and spread systemically at 6 dpi.
This assay with a GFP-encoding PVX construct also identifies local VSRs, since PVX-GFP replicates to a higher level and expresses its GFP gene more strongly and for longer in the presence of a VSR [20]. The co-infiltration of PVX-GFP with P4 PAV or P4 PAS constructs led to a slight increase in GFP fluorescence in infiltrated spots when compared to P4 PL , P1 RY , or vector backbone ( Figure 7A,C), which confirms the weak local VSR activity by P4 PAV and P4 PAS . Although the two luteovirus P4 proteins did not display strong local silencing suppression activity during PVX replication when compared to P0 PL , P0 CY and P19 ( Figure 7C), the replicated virus was still able to exit before the death of the infected cell, induced by P4.

Discussion
RNA silencing is the main anti-viral mechanism in plant tissues and the viral counter-defence usually involves avoidance or inactivation of the mechanism. As a survival strategy, the P0 proteins of some poleroviruses and the enamovirus have been reported to be strong local VSRs, acting by destabilizing AGO1 [20,44,45,49,50]. The viruses of the family Luteoviridae are strongly associated with the phloem tissues of their host plants [51], and the possession of strong locally-acting VSRs by the enamovirus and the poleroviruses might suggest that cells of the vasculature possess highly active RNA silencing machinery. The levels of virus-derived siRNAs detected by deep-sequencing PLRV-and cotton leafroll dwarf virus-infected materials (representing 15% and 6% of total sRNAs, respectively) support this [52,53]. However, the luteovirus BYDV-PAV replicates in cereals to similarly high levels as the polerovirus cereal yellow dwarf virus (CYDV)-RPV [33], and its siRNA profile revealed that it too is attacked by the RNA silencing machinery (this report). Yet, BYDV-PAV does not have an ORF in a counterpart genomic location to the enamovirus or poleroviruses, and testing all of the proteins expressed from the genome failed to identify any strong locally-acting VSR.
A feature of BYDV-PAV that is not shared with poleroviruses is the production of a highly abundant subgenomic RNA3. It seemed possible that BYDV-PAV, in lieu of a strong locally-acting VSR, might express this non-coding RNA as a decoy for the RNA silencing machinery. CaMV has been shown to use such a strategy, producing an RNA from the leader sequence of its 35S transcript that attracts the plant's siRNA generating machinery [54]. However, despite sgRNA3 accumulating to levels approaching 100-fold higher than the genomic RNA [33], there was not a dramatic overabundance of siRNAs produced from this genomic region. This shows that either sgRNA3 does not play this role or acts by augmenting the number of RNAs that AGO must scan within the cell, thus quenching its activity.
Remarkably, the protein previously identified as a movement protein, encoded by BYDV-PAV ORF 4, suppressed the spread of the silencing signal in N. benthamiana assays; a capacity also displayed by BYDV-PAS P4, but not polerovirus PLRV P4. The systemic VSR activity by the two luteovirus P4 proteins was accompanied by a reduced production of local siRNAs and a delay in the short-distance spread of the silencing signal, and these effects were dose-dependent. The systemic VSR activity by P0 proteins of the poleroviruses PLRV (P0 PL ), CYDV (P0 CY ) and sugarcane yellow leaf virus (P0 SC ) and the enamovirus PEMV-1 (P0 PE ) also induce similar effects, although they also have strong local VSR activity [20,49]. In the case of P0s, the destabilization of AGO1 provides a mechanism for the suppression of systemic spread of the silencing signal, since AGO1 appears to be required for systemic silencing [20,42,55,56]. Because luteoviruses lack an F-box-like P0 protein, it is possible that P4 PAV and P4 PAS operate differently to P0 to reduce the production or accumulation of siRNAs and suppress systemic silencing. The lack of strong local VSR activity by P4 PAV and P4 PAS compared to polerovirus and enamovirus P0s also supports this hypothesis. The degree of systemic silencing suppression by P4 PAV and P4 PAS was similar to RYMV P1, which is another example of a movement protein with effective systemic VSR activity and very weak local silencing suppression, but its suppression mechanism still remains largely unknown [57,58].
The expression of P4 PAV and P4 PAS in N. benthamiana elicits a response by the plant that leads to cell death in the infiltrated tissues, and the degree of necrosis directly correlates with P4 titers. Similar observations were recently reported for BYDV P4 [59]. Different poleroviral P0 proteins (P0 PL , P0 CY , P0 SC and beet western yellows virus P0) have also been reported to trigger necrosis in infiltrated tissues [20,49]. It could be hypothesised that P4 PAV and P4 PAS caused the infiltrated cells to die before the sGFP construct could be transcribed to high levels and before secondary siRNAs could be amplified from GFP mRNA, which would explain the low levels of GFP siRNAs in these tissues ( Figure 6) and the reduction in red halo formation (Figures 3 and 4). However, the inhibition of necrosis in infiltrated tissues, by concomitantly expressing the antiapoptotic protein CED-9 with P4 PAV (Figures 3 and S2), did not affect the P4 PAV VSR activities of delaying the short-distance movement of the silencing signal (seen as reduced red halo) and displaying strong suppression of systemic silencing, nor did it significantly increase GFP fluorescence in the infiltrated spots. This suggests that the VSR activity displayed by P4 PAV and P4 PAS on the reduction of secondary siRNA production or accumulation is independent of the necrotic process.
Luteoviruses are known to induce phloem necrosis in their hosts and it is possible that the P4 protein of these viruses could act as elicitors of this response. Nevertheless, the necrosis triggered by luteovirus P4 in infiltrated tissues did not inhibit P4's systemic VSR activity against the propagation and spread of anti-viral siRNAs, nor did it prevent the replication and egress of PVX. Although lacking a strong local silencing suppressor, the strong systemic VSR activity by BYDV-PAV P4 and BYDV-PAS P4 and the weaker VSR activity by P6 PAV , suggest that suppression of systemic silencing is the principal mechanism by which the luteoviruses BYDV-PAV and BYDV-PAS minimize the effects of the plant's anti-viral defense.