A novel sweet potato potyvirus open reading frame (ORF) is expressed via polymerase slippage and suppresses RNA silencing

Summary The single‐stranded, positive‐sense RNA genome of viruses in the genus Potyvirus encodes a large polyprotein that is cleaved to yield 10 mature proteins. The first three cleavage products are P1, HCpro and P3. An additional short open reading frame (ORF), called pipo, overlaps the P3 region of the polyprotein ORF. Four related potyviruses infecting sweet potato (Ipomoea batatas) are predicted to contain a third ORF, called pispo, which overlaps the 3′ third of the P1 region. Recently, pipo has been shown to be expressed via polymerase slippage at a conserved GA6 sequence. Here, we show that pispo is also expressed via polymerase slippage at a GA6 sequence, with higher slippage efficiency (∼5%) than at the pipo site (∼1%). Transient expression of recombinant P1 or the ‘transframe’ product, P1N‐PISPO, in Nicotiana benthamiana suppressed local RNA silencing (RNAi), but only P1N‐PISPO inhibited short‐distance movement of the silencing signal. These results reveal that polymerase slippage in potyviruses is not limited to pipo expression, but can be co‐opted for the evolution and expression of further novel gene products.


INTRODUCTION
Several viruses of the genus Potyvirus (family Potyviridae) infect sweet potato (Ipomoea batatas). Currently, most fall within the monophyletic 'Sweet potato feathery mottle virus (SPFMV) group' (Untiveros et al., 2008). These include SPFMV, Sweet potato virus C (SPVC), Sweet potato virus G (SPVG) and Sweet potato virus 2 (SPV2), but not Sweet potato latent virus. SPFMV is the most common virus infecting sweet potatoes worldwide. In mixed infections with Sweet potato chlorotic stunt virus (SPCSV; genus Crinivirus), it is associated with a severe sweet potato virus disease (SPVD), causing heavy yield losses (Gibson et al., 2004;Tairo et al., 2005). SPVD is particularly important in sub-Saharan Africa, where sweet potato is the second most important root crop after cassava.
The complete genome sequences of several SPFMV-group viruses have been determined Pardina et al., 2012;Sakai et al., 1997;Untiveros et al., 2010;Yamasaki et al., 2010). Although all members of the genus Potyvirus encode a P1 protein, SPFMV-group viruses encode the largest P1 among currently sequenced potyviruses ( Fig. 1a) (Sakai et al., 1997;Valli et al., 2007). P1 is a serine protease that cleaves at its own C-terminus, but is not known to be involved in other polyproteinprocessing events. Apart from its protease activity, the functions of P1 are not well established. It appears to play a variety of roles in different potyviruses, including enhancement of the RNA silencing suppression (RSS) activity of the helper component proteinase (HCpro), and roles in genome amplification and host specificity (reviewed in Roho zkov a and Navr atil, 2011; Valli et al., 2007). Although HCpro is generally considered to be the RSS protein in potyviruses (Anandalakshmi et al., 1998;Brigneti et al., 1998;Kasschau and Carrington, 1998), in members of some other potyvirid genera, P1 has been shown to be a potent RSS. In Sweet potato mild mottle virus (SPMMV; genus Ipomovirus), Triticum mosaic virus (genus Poacevirus) and Wheat streak mosaic virus (genus Tritimovirus), P1, but not HCpro, suppresses silencing (Giner et al., 2010;Tatineni et al., 2012;Young et al., 2012). Cassava brown streak virus (genus Ipomovirus) lacks HCpro, and P1 suppresses silencing (Mbanzibwa et al., 2009). Cucumber vein yellowing virus (genus Ipomovirus) also lacks HCpro, but has two P1 serine protease proteins, P1a and P1b, where P1b suppresses silencing (Valli et al., 2006). In Potato virus Y (PVY), P1 may be needed to stabilize the RSS protein HCpro (Tena Fern andez et al., 2013). However, macluraviruses lack any P1 protein (Revers and Garcia, 2015).
The P1 protein of SPFMV-group viruses appears to contain two distinct protein domains (Fig. 1a): the C-terminal P1-pro, which is homologous to the P1 serine protease encoded by all potyviruses, and an N-terminal extension, designated P1-N, which is absent from other potyviruses, but bears homology to the N-terminal region of the P1 protein of the ipomovirus SPMMV (Untiveros *Correspondence: Email: aef24@cam.ac.uk; j.kreuze@cgiar.org; jari.valkonen@helsinki.fi †These authors contributed equally to this work. NCBI GenBank.   Valli et al., 2007). The P1-N domain appears to have been transferred between ancestors of the two viruses, presumably during co-infection of the common host. P1-N and P1-pro are separated by a hypervariable region (Untiveros et al., 2010). SPMMV P1-N is an RSS which functions by binding Argonaute 1 (AGO1) (Giner et al., 2010). This is thought to be mediated via WG/GW motifs, and such motifs are a common feature of AGObinding proteins (Azevedo et al., 2010;El-Shami et al., 2007;Giner et al., 2010;Till et al., 2007). Wild-type SPFMV contains only one WG/GW motif in P1-N, which was not found to confer any RSS activity (Szab o et al., 2012), whereas three WG/GW motifs are found in SPMMV P1-N. Thus, despite the homology, it is unclear whether P1-N serves a similar RSS function in SPFMVgroup viruses.
It was long believed that the potyvirus gene expression strategy was based on a large, single, open reading frame (ORF) encoding a polyprotein, which is processed to 10 mature proteins by three viral proteinases. However, Chung et al. (2008) discovered that there is another, short ORF in a different reading frame embedded within the region encoding P3. This ORF, designated pipo ('pretty interesting Potyviridae ORF'), is conserved throughout the Potyviridae family. Polymerase slippage at a highly conserved GA 6 sequence at the 5 0 end of pipo results in the insertion of an extra A into 1%-2% of transcripts (Olspert et al., 2015;Rodamilans et al., 2015), leading, after translation and proteolytic cleavage, to the production of P3N-PIPO, a 'transframe fusion' of the N-terminal part of P3 (P3N) with the product (PIPO) of the pipo ORF. An additional, long alternative-frame ORF in the region encoding P1-pro was observed in the 'SPFMV group' of potyviruses and, in analogy with pipo, was named pispo ('pretty interesting sweet potato potyvirus ORF') ( Fig. 1a) (Clark et al., 2012). The presence of multiple WG/GW motifs within the translation of pispo was noted by Clark et al. (2012), who suggested a potential role in RSS. Here, we present computational analyses of the pispo ORF that support its coding status, high-throughput sequencing data demonstrating that pispo can be expressed as part of a transframe protein P1N-PISPO via polymerase slippage at a conserved G 2 A 6 slippery sequence, and experimental analyses that confirm that P1N-PISPO is a potent suppressor of RNA silencing.

Comparative genomic analysis of the overlapping pispo ORF
The polyprotein ORFs from the 28 full-length SPFMV-group sequences currently available in GenBank, plus three additional sequences determined in this study (see Experimental procedures), were extracted, translated, aligned as amino acid sequences and back-translated to a nucleotide sequence alignment. Next, the positions of stop codons in the 11 and 12 reading frames relative to the polyprotein ORF, in all 31 sequences, were plotted. This confirmed the conserved presence of a long 12-frame ORF overlapping the P1 region of the polyprotein ORF (Fig. 1b). The ORF has 230 codons in SPFMV and SPVC, and 228 codons in SPVG and SPV2 (measured from a ubiquitous 5 0 G 2 A 6 sequence; see below). The statistical significance of such a long ORF being conserved by chance over such a degree of phylogenetic divergence (estimated >900 phylogenetically independent substitutions within pispo in the SPFMV-group sequence alignment) was assessed as described previously (Fang et al., 2012). Using this method, the probability of such a long 12 frame ORF occurring by chance at this location is P < 1 3 10 216 .
Next, the alignment was analysed for enhanced conservation at polyprotein-frame synonymous sites (an indicator of (c) Conservation at synonymous sites in the alignment, using a 231-codon sliding window (green) or a 15-codon sliding window (red and brown). The upper panels (green, red) depict the probability that the degree of conservation within a given window could be obtained under a null model of neutral evolution at synonymous sites, whereas the bottom panel (brown) depicts the absolute amount of conservation as represented by the ratio of the observed number of substitutions within a given window to the number expected under the null model. The broken grey lines indicate a P 5 0.05 threshold after applying a rough correction for multiple testing (namely 3519 codons/15-or 231-codon window size).
overlapping functional elements), as described previously (Firth, 2014). The analysis revealed regions of enhanced synonymous site conservation where the polyprotein ORF is overlapped by the pipo ORF (Chung et al., 2008), and at the 3 0 end of the polyprotein ORF, where overlapping replicational RNA elements have been mapped in other potyvirus species (Fig. 1c) (Haldeman-Cahill et al., 1998). Enhancement of polyprotein-frame synonymous site conservation within the region overlapped by pispo was modest, suggesting that the PISPO amino acid sequence is not subject to strong purifying selection. Nonetheless, the mean rate of synonymous substitutions in the region of the polyprotein ORF overlapped by pispo was 10% below the genome average, with a corresponding total P value of 0.01 (Fig. 1c).
At the 5 0 end of the pispo ORF, there is a conserved G 2 A 6 sequence (Figs 2a and S1, see Supporting Information). In SPFMV and SPVC isolates, this takes the form GG_AAA_AAA, where underscores separate polyprotein codons. In SPVG and SPV2 isolates, it takes the form G_GAA_AAA_A, displaced a few codons upstream from the sequence in SPFMV and SPVC (Fig. 2a). A similar highly conserved sequence, G 1-2 A 6-7 (G 2 A 6 in SPFMV-group viruses), is present at the 5 0 end of the pipo ORF in nearly all potyvirids (Chung et al., 2008). The presence of the same motifs at the 5 0 end of pispo suggests that pispo is expressed by the same slip-page mechanism as pipo. This would result in the production of a P1N-PISPO fusion protein (Fig. 2b). In the Ruk73 strain of SPFMV, P1 has a predicted mass of 77.1 kDa, whereas P1N-PISPO has a predicted mass of 75.7 kDa. Analysis of additional sequences with only partial coverage of the polyprotein ORF, but with coverage of the 5 0 end of pispo and/or the entire pispo ORF, confirmed the conserved presence of the potential slippage site (14 of 14 sequences) and the conserved presence of the pispo ORF (nine of nine sequences).
High-throughput sequencing reveals polymerase slippage at the pispo and pipo G 2 A 6 sites The occurrence of the G 2 A 6 sequence at the 5 0 end of the pispo and pipo ORFs in different phases with respect to the polyprotein reading frame suggests that pispo and pipo are unlikely to be expressed via translational frameshifting, although, in principle, both are compatible with 21 ribosomal frameshifting (reviewed in Firth and Brierley, 2012). Further, using RNA-folding software, such as alidot (Hofacker et al., 2002) and pknotsRG (Reeder et al., 2007), we failed to predict a convincing 3 0 RNA secondary structure at a canonical spacing (i.e. 5-9 nucleotides) downstream of the slippery sites for the stimulation of 21 ribosomal frameshifting. To evaluate whether frameshifting could occur instead during RNA synthesis, we first utilized high-throughput small RNA sequencing data obtained from SPFMV-infected sweet potato plants and SPV2-and SPVG-infected Ipomoea setosa plants. At both the pispo and pipo G 2 A 6 sequences, a fraction of small RNA reads contained an additional 'A' insertion and, for most samples, insertions at these sites occurred well above background frequencies ( Fig. 3; Table 1). In these data, 'A' insertions were found to occur at a frequency ranging from 0.9% to 11.4% at the pispo G 2 A 6 site and 0.0% to 1.0% at the pipo G 2 A 6 site, depending on the sample. The 0% values are probably a result of low coverage and the lower rate of insertions at the pipo site relative to the pispo site (see below). In comparison, average single nucleotide insertions at other sites across the genome were estimated to occur at a frequency ranging from 0.001% to 0.01% per nucleotide, depending on the sample (Table 1).
Although these results support the interpretation that polymerase slippage occurs at the pispo and pipo G 2 A 6 sites, low coverage, variability between samples and low accuracy of insertion calling as a result of short read lengths mean that the slippage frequency is likely to be poorly estimated by this approach. Therefore, a separate sequencing experiment was conducted in which both the pispo and pipo sites in viral RNA were specifically targeted for amplification and sequencing. This experiment was conducted with RNA extracted from sweet potato or I. nil infected with SPFMV-Ruk73, and led to sequencing depths ranging from 6.1 to 32.5 million reads for each slippage site region ( Fig. 4; Table 2). Single 'A' insertions were observed within the pispo G 2 A 6 sequence for 5.0% and 5.5% of sweet potato and I. nil reads, respectively. Within the pipo G 2 A 6 sequence, single 'A' insertions were observed for 0.9% and 1.0% of reads, respectively. Control DNAs for the same regions showed no insertions at all in more than eight million reads, indicating that insertions did not occur during amplification, library preparation or sequencing. Parallel work with the Turnip mosaic virus G 2 A 6 pipo slippage site indicated that any insertional slippage occurring during reverse  transcription is less than 0.05% (Olspert et al., 2015). At a lower level, deletions were also observed to occur (Fig. 4), although these could be an artefact of deletional slippage during reverse transcription (Olspert et al., 2015).

PISPO contains GW/WG motifs
The amino acid sequence of PISPO is less conserved than the amino acid sequence of the part of P1 that it overlaps (Table S1, see Supporting Information), e.g. there is 29%-40% amino acid identity in PISPO, but 48%-67% identity in P1-pro between potyviruses that encode PISPO. Application of BLASTP, CS-BLAST, Inter-ProScan, hhpred and phyre-2 to PISPO revealed no significant homology to previously annotated or characterized proteins, as expected, as PISPO clearly evolved by 'overprinting' the ancestral P1-pro domain (Altschul et al., 1990;Angerm€ uller et al., 2012;Kelley and Sternberg, 2009;Quevillon et al., 2005;S€ oding et al., 2005). In addition, 60% of PISPO was predicted to be disordered according to PONDR-FIT (Xue et al., 2010) (Fig. S2, see Supporting Information), again typical of many overlapping genes (Rancurel et al., 2009). Further inspection of the amino acid sequence revealed WG/ GW ('AGO-hook') motifs. Such motifs play a key role within the P1-N domain of SPMMV P1 (Giner et al., 2010) and some other plant viral silencing suppressors (Azevedo et al., 2010;de Ronde et al., 2014). Among 31 sequenced isolates, the PISPO sequences contain between one and three GW/WG motifs, whereas fulllength P1 and full-length P1N-PISPO contain between two and four GW/WG motifs (Fig. 5). It should be noted, however, that WG/GW di-amino acids are not particularly uncommon in generic protein sequences.

P1 and P1N-PISPO are RNA silencing suppressors with distinguishable phenotypes
To test whether P1 and/or P1N-PISPO have RSS activity, we used a standard silencing assay, whereby transgenic Nicotiana benthamiana line 16c plants that constitutively express green fluorescent protein (GFP) were agroinfiltrated with recombinant Agrobacterium tumefaciens strains harbouring binary expression constructs for GFP (pBIN:GFP) and a test protein (e.g. pBIN:P1). Overexpression of GFP mRNA from pBIN:GFP induces sensemediated silencing of GFP and results eventually in a red leaf patch under UV light, unless the test protein suppresses silencing, in which case the leaf patch will show enhanced green fluorescence under UV light. We tested various potyvirus proteins in this system (Fig. 6). At 4 days post-infiltration (dpi), tissues coinfiltrated for expression of GFP and P1 or P1N-PISPO showed slightly brighter GFP fluorescence than those infiltrated for expression of SPFMV HCpro, Sweet potato latent virus (SPLV) P1 Fig. 4 Transcriptional slippage at the pispo and pipo G 2 A 6 sequences in Sweet potato feathery mottle virus (SPFMV)-infected plants. Results from targeted high-throughput sequencing of systemically infected Ipomoea batatas and I. nil. The frequencies of transcripts with a single 'A' insertion at the G 2 A 6 sequence are shown in blue; the frequencies of transcripts with two or more inserted 'A' nucleotides are shown in orange; and the frequencies of transcripts with one or more 'A' nucleotides deleted are shown in yellow. Controls from plasmid template were included to assess the variability introduced during amplification and sequencing. or a b-glucuronidase (GUS) control (Fig. 7a), indicating that P1 and P1N-PISPO have RSS activity. Leaf tissues expressing Potato virus A (PVA) HCpro exhibited the strongest GFP signals (Fig. 7a). These phenotypic differences correlated positively with the accumulation of GFP (Fig. 7b) and GFP mRNA (Fig. 7c), and were negatively correlated with the accumulation of GFP mRNA-derived small interfering RNA (siRNA) in the leaf tissue (Fig. 7c). The expression of P1 and P1N-PISPO was verified by western analysis using peptide antibodies (Abs) generated against an epitope (Fig. 2c) common to P1 and P1N-PISPO (blot 'P1' in Fig. 7b) and peptide Abs specific to P1N-PISPO (Fig. 7b). The observed bands were of slightly different sizes than expected, possibly as a result of post-translational modifications. The suppression of silencing by P1 and P1N-PISPO was quite apparent at 6 dpi ( Fig. 7d) and at 8 dpi (Fig. 7e), as indicated by GFP fluorescence in the co-infiltrated leaf tissue, whereas the leaf spots infiltrated for the expression of SPFMV HCpro, SPLV P1 and GUS had lost GFP fluorescence (Fig. 7e). However, P1 and PIN-PISPO differed in that a red halo developed around the leaf tissue infiltrated for the expression of P1. The red halo was most apparent at 6 dpi (Fig. 7d,e, leaf spot 1), indicating short-distance movement of the silencing signal. A red halo also developed around the leaf spots infiltrated for the expression of SPFMV HCpro, SPLV P1 and GUS (Fig. 7e). In contrast, only a faint red halo, if any, developed around the leaf tissue infiltrated for the expression of P1N-PISPO (Fig. 7d,e, leaf spot 2), indicating interference with the short-distance signalling for silencing.
As no Abs were available for detection of SPFMV HCpro and SPLV P1, these proteins were also expressed tagged with the N-proximal part (YN) of yellow fluorescent protein (YFP) (Fig. 6) and tested by western blot using anti-YN Abs. YN-P1 (SPLV) was detected at 2 dpi (Fig. S3, see Supporting Information), but only occasionally at later time points (3-5 dpi), whereas YN-HCpro could not be detected at any time point (2-10 dpi) in eight independent experiments.
As it is possible that the P1 construct could also express P1N-PISPO via translational or transcriptional frameshifting, we generated another P1 expression construct, P1DPISPO, in which expression of the pispo ORF was prevented by the insertion of a stop codon (Fig. 6). As with P1 (leaf spot 1; Fig. 8a), P1DPISPO (leaf spot 3; Fig. 8a) suppressed GFP silencing and was associated with the formation of a red halo.  The WG motif at the N-terminus of P1 is pivotal for suppression of RNA silencing The completely conserved WG motif near the N-terminus of SPFMV-group P1 (residues 25-26) aligns with the first WG motif in SPMMV P1 (Fig. 5), which has been shown to be required for SPMMV P1 RSS activity (Giner et al., 2010). To test whether this WG motif is also required for the RSS activity of SPFMV P1 and/or P1N-PISPO, we mutated the tryptophan residue (W 25 ) to alanine and tested the mutated P1 and P1N-PISPO for RSS activity, as above. In contrast with P1 and P1N-PISPO (Fig. 8b, leaf spots 1 and 3, respectively), the mutated constructs for P1(DWG 1 ) and P1N-PISPO(DWG 1 ) (Fig. 6) were unable to suppress silencing (Fig. 8b, leaf spots 2 and 4, respectively). Similarly, when all four WG motifs were mutated [P1N-PISPO(DWG 1,2,3,4 )] (Fig. 6), no suppression of silencing was observed (Fig. 8b, leaf spot 6). In contrast, similar mutations introduced in all three WG motifs in the PISPO domain [P1N-PISPO(DWG 2,3,4 )] (Fig. 6) did not prevent the suppression of silencing (Fig. 8b, leaf spot 5). Taken together, these results indicate that the WG motif near the N-terminus of P1 is essential for suppression of RNAi by both P1 and P1N-PISPO.

DISCUSSION
Potyviruses were believed to use a single polyprotein gene expression strategy until the recent discovery of the overlapping pipo ORF (Fig. 1a) (Chung et al., 2008). The pipo ORF has a highly conserved G 1-2 A 6-7 sequence at its 5 0 end. Previously, it was proposed that the expression of P3N-PIPO might depend on translational or transcriptional frameshifting occurring at this site (Chung et al., 2008). However, the true nature of the expression mechanism was not confirmed until recently, when, during the preparation of this article, both we and others observed the  (6) b-glucuronidase (GUS) (control). The leaves were photographed at 4 days post-infiltration (dpi) (a), 6 dpi (d) and 8 dpi (e). Scale bars indicate 2 cm. (b) Expressed proteins were detected at 4 dpi using antibodies to GFP, peptide antibodies generated against an epitope common to P1 and P1N-PISPO, and peptide antibodies specific to the PISPO domain of P1N-PISPO. In the P1 panel, lane 2 was loaded in a different order in the gel, and spliced electronically afterwards (indicated by black vertical lines), for consistency with other panels. Staining of total proteins by Coomassie blue was used as a loading control (shown for GFP gel). (c) Northern blot analysis for the detection of GFP mRNA and gfpderived small interfering (si)RNA in leaf tissues co-infiltrated with pBIN:GFP and the viral constructs in (A) at 4 dpi; gfp was used as a probe and 28S and 5S ribosomal RNAs (rRNA) were used as loading controls. Note the different order of samples 5 and 6 in panels (b) and (d). (d) The infiltrated leaf tissue expressing P1 was surrounded by a red halo (left), in contrast with the tissue expressing P1N-PISPO (right), which was most apparent at 6 dpi. (e) Suppression of silencing by P1, P1N-PISPO and PVA HCpro was indicated by GFP fluorescence in the co-infiltrated leaf tissue at 8 dpi, whereas no suppression of silencing was observed with SPFMV HCpro and SPLV P1.
production of virus transcripts with single nucleotide insertions at the pipo G 1-2 A 6-7 sequence in various potyviruses, thus allowing the expression of P3N-PIPO (this study; Olspert et al., 2015;Rodamilans et al., 2015). The ubiquitous presence of a G 2 A 6 sequence at the 5 0 end of the pispo ORF, conserved in all SPFMV-group sequences, suggested that pispo might be expressed via the same polymerase slippage mechanism, leading to a P1N-PISPO transframe fusion protein. The potential for P1N-PISPO to be expressed in this way was confirmed by the observation of edited transcripts in our small RNA sequencing data for SPFMV, SPV2 and SPVG, and our targeted high-throughput sequencing of SPFMV RNA. The latter data also indicate that slippage at the SPFMV pispo slip site (5.0%-5.4%) is significantly more efficient than at the SPFMV pipo slip site (0.9%-1.0%). Whilst this article was in preparation, a similar difference was also observed by others (Rodamilans et al., 2015). The reasons for the differences between the two sites are unknown, but may be a result of the flanking nucleotide context which may affect the stability of the template:nascent RNA duplex adjacent to the slippage site (Hausmann et al., 1999). Although we have currently been unable to detect P1N-PISPO in infected plants by western analysis (perhaps because of low expression, masking of the Ab epitope and/or the kinetics of virus infection and protein turnover), the presence of edited transcripts implies that P1N-PISPO can indeed be expressed during virus infection.
P1 is the most variable potyvirus protein in terms of length and sequence (Adams et al., 2005;Valli et al., 2007). The discovery of PISPO adds a new dimension to the functional variability occurring in the P1 region. Unlike PIPO, which is found universally in the Potyviridae family, PISPO is restricted to potyviruses in the SPFMV group. It does not occur in the related SPLV, nor in the ipomovirus SPMMV with which SPFMV-group viruses share a related P1-N domain. The PISPO sequence is highly variable (Table S1)  and less conserved than the overlapping region of P1-pro, in clear contrast with PIPO and the region of P3 that it overlaps (Chung et al., 2008). Differences in the evolutionary plasticity of pipo and pispo may reflect the functional importance of the corresponding transframe proteins. P3N-PIPO is an essential protein, and appears to be of ancient origin, as attested by its conservation throughout the Potyviridae family, whereas P1N-PISPO seems to be a more recent evolutionary development and may therefore not yet have acquired a critical role, allowing more sequence plasticity, as has been observed for other recently evolved overlapping viral genes (Sabath et al., 2012). PISPO is predicted to be largely disordered (Fig. S2). The function of such an intrinsically unstructured protein region may be largely independent of most of the primary sequence, allowing for rapid evolution and exploration of sequence space (Gitlin et al., 2014).
Although it is clear that PISPO is biologically relevant, as an ORF of this length would not be conserved by chance (P < 1 3 10 216 ) over the given phylogenetic divergence, at present there is no infectious clone of SPFMV, and so it is not possible to verify the functional relevance of P1N-PISPO during infection. However, given the importance of WG/GW domains in other silencing suppressors (Giner et al., 2010;de Ronde et al., 2014), the evaluation of P1N-PISPO for RSS activity was an obvious course. Surprisingly, we found that both P1N-PISPO and P1 are RNA silencing suppressors in SPFMV. Intrinsic silencing suppression has not been described previously for P1 proteins of potyviruses (Rajam€ aki et al., 2005;Tena Fern andez et al., 2013), nor was it found herein for P1 of SPLV (which clusters as an immediate outgroup to the SPFMV group). Most recent reports assign potyvirus P1 a role for increasing the accumulation of HCpro and consequently enhancing its silencing activity (Tena Fern andez et al., 2013). We also showed that P1 and P1N-PISPO seem to have different silencing activities in SPFMV, a strategy described for different RSS proteins of other viruses as well (Lu et al., 2004;Senshu et al., 2011). P1 operates only at a single-cell or local level, as evidenced by the formation of the red halo around the patch (Fig. 7), as also reported for SPMMV P1 and a few other virus silencing suppressors (Baumberger et al., 2007;Giner et al., 2010). The effect of P1 on silencing was a result of the sole action of this protein, and participation of traces of P1N-PISPO can be discarded as the phenotype was identical to P1DPISPO (Fig. 8), in which PISPO is made untranslatable by introducing a stop codon. However, P1N-PISPO transiently inhibited RNA silencing at both local and shortrange levels, as reported for suppressors that sequester siRNA, including PVA HCpro and tombusvirus P19 (Lakatos et al., 2004;Shiboleth et al., 2007). The elimination of the three PISPO WG/ GW motifs did not remove the suppressor capacity of P1N-PISPO (Fig. 8). However, the elimination of the single WG near the N-terminus of P1-N completely eliminated RSS activity of both P1 and P1N-PISPO. Thus, the differing C-termini of P1 and P1N-PISPO modulate the RSS activity resident within P1-N.
The results of our experiments differ from those of a previous study in which P1 of SPFMV was reported not to suppress RNA silencing (Szab o et al., 2012), probably because a different experimental system and SPFMV isolate were used.
The functional characterization of P1 and P1N-PISPO requires further experimental work, in particular during SPCSV synergisms. SPCSV RNase3 protein mediates these synergisms through the suppression of antiviral defence in sweet potato plants, allowing potyviruses to enhance titres (Cuellar et al., 2009). Nevertheless, symptom severity, even among SPFMV-group viruses, varies greatly (Gutierrez et al., 2003;Kokkinos and Clark, 2006;Untiveros et al., 2007), and this is not necessarily related to the enhancement of viral replication (Kokkinos and Clark, 2006). This suggests that additional viral determinants may be required for the development of unusually severe disease in sweet potato plants co-infected with SPFMV and SPCSV. Such determinants might include P1 and P1N-PISPO. Therefore, the results of this study may have important implications for understanding the devastating synergistic diseases caused by these viruses in sweet potato.

EXPERIMENTAL PROCEDURES
Viruses SPFMV isolate Ruk73 from Uganda (Tugume et al., 2010) was used to generate all the SPFMV-derived expression vectors. In addition, a Taiwanese isolate of SPLV (Wang et al., 2013), available from the International Potato Center (CIP) reference virus collection, was used to generate the corresponding P1 expression construct. SPFMV-Piu3 , and SPVG-LSU1 and SPV2-LSU2 (kindly provided by Dr Christopher A. Clark, Louisiana State University, Baton Rouge, LA, USA), available from the CIP reference isolate collection, were used for small RNA sequencing.

Computational analysis
The complete genomes of SPFMV-Ruk73, SPVG-LSU1 and SPV2-LSU2 were determined in this study (see above). Additional SPFMV-group nucleotide sequences in GenBank were identified using TBLASTN (Altschul et al., 1990) (see Table S2 for accession numbers). Coding sequences were extracted, translated, aligned and back-translated to produce nucleotide sequence alignments using EMBOSS and CLUSTALW (Larkin et al., 2007;Rice et al., 2000). Synonymous site conservation was calculated as described previously (Firth, 2014). The probability of the conserved absence of stop codons in the pispo ORF occurring by chance (i.e. if pispo was actually non-coding) was assessed via P1-frame alignment codon column shuffling, as described previously (Fang et al., 2012).

Plasmids
A schematic diagram of the vectors used in this study is displayed in Fig. 6. Briefly, DNA fragments containing the complete coding regions of SPFMV P1 and HCpro (isolate Ruk73), and SPLV P1 (KC443039.1), were described in Olspert et al. (2015), except that reverse transcriptionpolymerase chain reaction (RT-PCR) was performed with primers to amplify the regions surrounding the putative slippage sites for pispo (target site, excluding primer and adaptor sequences: ACTCGTTTGG AAAAAACTGGAC, corresponding to SPFMV-Ruk73 nucleotides 1448-1469) and pipo (target site, excluding primer and adaptor sequences: CTCATGGAAAAAATTTGGGAT, corresponding to SPFMV-Ruk73 nucleotides 4017-4037) High-throughput sequencing of small RNA was performed on RNA extracted from infected sweet potato (SPFMV-Piu3) or I. setosa (SPVG-LSU1 and SPV2-LSU2) plants as described above. Short-RNA reads, both forward and reverse sense, were mapped to the respective reference genomes using BWA (Li and Durbin, 2009), and positions of insertions were determined from the BWA CIGAR and TAG report fields. Initial background single nucleotide insertion rates were calculated by dividing the observed number of insertions (excluding those at the pispo and pipo G 2 A 6 sites) by the product of the mean read depth and genome length. As a result of the short length of reads (21-24 nucleotides) and BWA not reporting insertions or deletions (indels) that occur within 5 nucleotides of the termini of reads, these initial values were adjusted for the potentially unreported indels by multiplying by the correction factor P i P i i/(i -10), where i ranges over the read lengths (i.e. 21-24 nucleotides), P i is the fraction of mapped reads that have length i, and i/(i -10) is the reciprocal of the proportion of insertions occurring within a read of length i that would be detected by BWA (i.e. are not within 5 nucleotides of the read termini).

SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article at the publisher's website: Fig. S1 Predicted pispo G 2 A 6 slippage sites. Predicted pispo G 2 A 6 slippage sites (orange highlighted) in all available fulllength Sweet potato feathery mottle virus (SPFMV)-group sequences. SPV2, Sweet potato virus 2; SPVC, Sweet potato virus C; SPVG, Sweet potato virus G. Fig. S2 PONDRV R disorder predictions for P1 and P1N-PISPO. Common regions predicted to be unstructured or disordered include a small region of 25 residues within the N-terminus of P1N, and the hypervariable region (residues 200-300). In P1, a third region of predicted disorder occurs around residues 459-522, whereas, in P1N-PISPO, a third region of predicted disorder encompasses almost the entire PISPO domain starting around residue 444. Predicted non-disordered domains have scores no higher than 0.5. short-distance movement of the silencing signal. Scale bar indicates 2 cm. (b) Expressed YN-P1(SPLV) protein was detected at 2 dpi using anti-YN antibodies, but not at 3, 4 or 5 dpi. The upper band indicated with an arrow (2d) corresponds to the expected size (72 kDa) of YN-P1(SPLV). The small arrow on the left indicates the position of the 70-kDa protein marker. Staining of total proteins by Coomassie blue was used as a loading control. Table S1 Mean amino acid inter-and intra-species identities. Identities (%) calculated for the PISPO (italics; upper left) and P1-pro (roman; lower right) domains. Table S2 Virus species and accession number for sequences used for the computational analysis of the pispo open reading frame (ORF). Sequences of Sweet potato virus G (SPVG)-LSU1 and Sweet potato virus 2 (SPV2)-LSU2 were determined in this study, and partial sequences with coverage of pispo also included are indicated in a separate column. Table S3 Primer sequences used in this study.