Bean pod mottle virus: a new powerful tool for functional genomics studies in Pisum sativum

Summary Pea (Pisum sativum L.) is an important legume worldwide. The importance of pea in arable rotations and nutritional value for both human and animal consumption have fostered sustained production and different studies to improve agronomic traits of interest. Moreover, complete sequencing of the pea genome is currently underway and will lead to the identification of a large number of genes potentially associated with important agronomic traits. Because stable genetic transformation is laborious for pea, virus‐induced gene silencing (VIGS) appears as a powerful alternative technology for determining the function of unknown genes. In this work, we present a rapid and efficient viral inoculation method using DNA infectious plasmids of Bean pod mottle virus (BPMV)‐derived VIGS vector. Six pea genotypes with important genes controlling biotic and/or abiotic stresses were found susceptible to BPMV carrying a GFP reporter gene and showed fluorescence in both shoots and roots. In a second step, we investigated 37 additional pea genotypes and found that 30 were susceptible to BPMV and only 7 were resistant. The capacity of BPMV to induce silencing of endogenes was investigated in the most susceptible genotype using two visual reporter genes: PsPDS and PsKORRIGAN1 (PsKOR1) encoding PHYTOENE DESATURASE and a 1,4‐β‐D‐glucanase, respectively. The features of the ‘one‐step’ BPMV‐derived VIGS vector include (i) the ease of rub‐inoculation, without any need for biolistic or agro‐inoculation procedures, (ii) simple cost‐effective procedure and (iii) noninterference of viral symptoms with silencing. These features make BPMV the most adapted VIGS vector in pea to make low‐ to high‐throughput VIGS studies.


Introduction
Pea (Pisum sativum L.) belongs to the Fabaceae plant family. Pea is a cool-season legume crop, which produces seeds of high nutritional value for feed and food that are an excellent protein complement of cereal-based diets. In Europe and especially in France, pea is currently the major grain legume crop and is mainly cultivated for the production of forage for animal consumption. But since the end of the nineties, the development of this crop has gradually decreased due to yield inconsistency. Consequently, European countries import about 70% of their plant protein consumption. Thus, significant productivity progress has to be made and stable and high-yielding varieties are needed.
Different biotic and abiotic stresses are responsible for yield instability. First, root rot disease caused by the oomycete Aphanomyces euteiches, considerably reduces the yields of spring pea (Kraft and Pfleger, 2001). Genetic resistance to A. euteiches root rot was shown to be partial and controlled by multiple genes (Hamon et al., 2010(Hamon et al., , 2013Pilet-Nayel et al., 2002. In winter peas, early sowings can increase yields and limit the effect of late-stage stresses, but cultivars need to have a high level of frost tolerance, a complex genetic trait (Lejeune-Henaut et al., 2008). Moreover, Ascochyta blight caused by a complex of three fungal pathogens affects aerial plant parts and can have devastating effects on yield of winter peas (Kraft and Pfleger, 2001). Both major genes and quantitative trait loci, conferring resistance to Ascochyta blight, have been identified (Clulow et al., 1991;Darby et al., 1985;Fondevilla et al., 2011;Prioul et al., 2004;Tar'an et al., 2003;Timmerman-Vaughan et al., 2002, 2004. Many genes controlling frost tolerance and resistance to A. euteiches root rot and Ascochyta blight have been mapped on the consensus genetic map of pea (Hamon et al., 2010(Hamon et al., , 2013Lejeune-Henaut et al., 2008;Pilet-Nayel et al., 2002Prioul et al., 2004). To identify the underlying gene sequences, fine mapping of the corresponding regions is needed for the identification of candidate genes that must be further validated using functional genetics tools. Ultimately, the discovery of gene functions will facilitate the genetic improvement of commercial pea varieties by molecular breeding.
Furthermore, with the future availability of the whole-genome sequence of pea (Alves-Carvalho et al., 2015;McGee, 2013), it is essential for the pea research community to develop novel molecular tools for large-scale analysis of gene functions at the genome-wide level. Although pea is amenable to stable genetic transformation Svabova et al., 2005), this remains a challenge and thus hampers the application of conventional reverse genetics approaches such as T-DNA insertion mutagenesis (AzpirozLeehan and Feldmann, 1997) and RNA interference (McGinnis et al., 2005). In pea, fast neutron and TILLING mutant populations have been developed (Dalmais et al., 2008;Hofer et al., 2009) but to perform large-scale, highthroughput genomic studies in pea, a cheap and rapid system that bypasses genetic transformation is required.
Virus-induced gene silencing (VIGS) is an attractive reverse genetics tool for functional genomics in plants. VIGS exploits a plant innate antiviral defence that is based on post-transcriptional gene silencing (PTGS) (Kumagai et al., 1995). In this defence mechanism, the presence of viral-derived doublestranded RNA (dsRNA) molecules triggers Dicer-like (DCL) proteins that identify and cleave dsRNA, generating 21-23 nucleotide viral-derived short-interfering RNAs (siRNAs). These siRNAs in association with the RNA-induced silencing complex (RISC), target homologous viral RNA for degradation. In VIGS, a recombinant virus carrying a fragment of a specific host gene is first constructed and inoculated in susceptible host plants. PTGS is triggered causing the sequence-specific degradation of mRNA of the target gene resulting in a knockdown phenotype that can be used to assess gene function (Burch-Smith et al., 2004;Robertson, 2004).
To date, several VIGS vectors have been designed to address reverse genetics needs in various plants (Lange et al., 2013). In legumes, VIGS vectors were developed from several plant viruses, such as the Apple latent spherical virus (ALSV, genus Cheravirus), the Pea early browning virus (PEBV, genus Tobravirus) and the Bean pod mottle virus (BPMV, genus Comovirus, family Comoviridae) (reviewed in Pflieger et al., 2013). In pea, PEBV was originally developed for VIGS studies (Constantin et al., 2004), but its use is restricted to low-throughput analyses because delivery of the viral vector into plants is not possible by mechanical inoculation and requires agro-inoculation which may become laborious and timeconsuming for high-throughput analyses (Pflieger et al., 2013).
In legumes, the most adapted VIGS vector for high-throughput genomics studies is the 'one-step' VIGS vector derived from BPMV (Zhang et al., 2010(Zhang et al., , 2013. Indeed, delivery of the BPMV vector into the plant is possible by mechanical inoculation via direct DNA rubbing of infectious plasmid DNA, with no need for in vitro transcription, Agrobacterium transformation or coating to gold particles for biolistic delivery. This efficient, simple and costeffective plasmid-based VIGS vector successfully allowed the functional validation of many genes involved in disease resistance and plant defence in soybean (Cooper et al., 2013;Liu et al., 2011Liu et al., , 2012Meyer et al., 2009;Pandey et al., 2011;Zhang et al., 2009Zhang et al., , 2012. Recently, we showed that direct DNA rubbing of BPMV-derived infectious plasmids gives high infection rates in Phaseolus vulgaris cv. Black Valentine and induces efficient VIGS in common bean . To test whether VIGS using the 'one-step' BPMV vector was feasible in pea, we first tested the susceptibility to BPMV of six pea genotypes of interest. We also investigated the spatial and temporal infection patterns of BPMV in aerial vegetative and reproductive tissues of these six genotypes. With the goal of adapting the 'one-step' BPMV vector for high-throughput VIGS studies in pea, we next optimized the conditions for mechanical inoculation of the BPMV vector in pea plants. The efficiency of endogenous gene silencing using the BPMV VIGS vector was investigated by targeting the PHYTOENE DESATURASE (PsPDS) and the KORRIGAN-1 (PvKOR1) VIGS reporter genes in aerial parts and roots of cv. Champagne, respectively. Finally, we tested the susceptibility to BPMV of 37 different pea genotypes reflecting the phylogenetic diversity among cultivated peas.

Results
Susceptibility to BPMV and spatio-temporal BPMV infection patterns of aerial tissues of six P. sativum cultivars One inherent property of VIGS is that gene silencing is effective only in genotypes in which the viral vector can multiply and spread systemically throughout the plant. To test the feasibility of BPMV VIGS in pea, we first tested the susceptibility of six pea cultivars (Champagne, FP, PI 180693, AeD99OSW-50-2-5, Isard and 552). These six genotypes were selected because they are the parental lines of recombinant inbred line (RIL) populations used to map genes involved in resistance to A. euteiches (Hamon et al., 2010(Hamon et al., , 2013Pilet-Nayel et al., 2002, Ascochyta blight (Prioul et al., 2004) and genes involved in frost tolerance (Lejeune-Henaut et al., 2008) (Table 1). The susceptibility to BPMV was tested using the BPMV vector expressing the green fluorescent protein GFP (BPMV-GFP) ( Table 2), which was considered as a marker of viral infection Zhang et al., 2010). Pea plants were mechanically inoculated using leaf sap obtained from P. vulgaris-infected plants. Indeed, in a previous study, we showed that P. vulgaris cv. Black Valentine is highly susceptible to BPMV and that direct rubinoculation of BPMV infectious plasmids results in a high infection rate . The green fluorescence of GFP was monitored by visual inspection of inoculated pea plants under UV light at different times postinoculation (pi).   For all six pea genotypes, GFP fluorescence was detected after 7 to 14 days pi (dpi) in the inoculated leaves (data not shown) attesting that the BPMV vector has entered into leaf cells and that viral proteins are effectively expressed, including GFP. Fluorescence pattern, in the form of round green spots, corresponded to the primary infection sites. At 4 weeks pi (wpi), GFP fluorescence extended to the upper uninoculated leaves and/or stems in the six genotypes, indicating that the virus is able to multiply in the inoculated leaf and to move systemically (Figures 1 and S1). Significantly, GFP fluorescence was detected as early as 3 wpi in cv. Champagne compared to 4 wpi in all other genotypes. At 8 to 10 wpi, GFP fluorescence was detected in the flowers and/or pods of five genotypes: Isard, 552, PI 180693, AeD99OSW-50-2-5, and FP (Figures 1 and S1). For cv. Champagne, the absence of flower production in our growth conditions explains the lack of data for this genotype. In cv. Isard, GFP fluorescence was visible in the seed coat, whereas no fluorescence was detected in the embryo ( Figure 1a). Overall, for the five genotypes Isard, 552, PI 180693, AeD99OSW-50-2-5 and FP,~30% of the plants inoculated with BPMV-GFP were systemically infected at 4 wpi. An infection rate of 100% was obtained for cv. Champagne at 4 wpi, and the complete aerial parts were infected at 8 wpi ( Figure 1c). Thus, according to our results, we considered that cv. Champagne was the most susceptible genotype to BPMV.

Optimization of BPMV infection efficiency in P. sativum
As efficient VIGS is directly related to efficient viral infection and regarding the low infection rate obtained for five genotypes (Isard, 552, PI 180693, AeD99OSW-50-2-5, and FP), we next optimized the mechanical inoculation procedure. This work was done on two genotypes: Isard and 552. Isard was chosen because it has an afila phenotype (leaflets are replaced by tendrils) and because this genotype possesses genes for resistance to Ascochyta blight and tolerance to frost. The genotype 552 has a normal phenotype and presents a high level of resistance to Aphanomyces euteiches. As during mechanical inoculation, the virus gets into plant cells by wounding, we tested two intensities of wounding of the upper leaf surface: scarification and abrasion.
As obtained previously, green fluorescence was visible at~14 dpi in the inoculated leaves of the two genotypes regardless of the inoculation method (data not shown). The scarification inoculation method gave the best inoculation rate for both genotypes Isard and 552 at 4, 5 and 6 wpi ( Table 3) and was thus chosen to perform all further mechanical inoculations in pea.
Virus-induced gene silencing of PsPDS in P. sativum cv. Champagne In the first part of this work, we demonstrated that the six pea genotypes of interest were susceptible to BPMV and that VIGS is thus theoretically feasible in these genotypes. However, to evaluate the efficiency of BPMV VIGS, we focused on the most susceptible pea genotype: cv. Champagne. We constructed a BPMV-PDS VIGS vector containing a 336-bp insertion derived from the 3 0 -end of the PsPDS ORF of cv. Champagne. PDS is routinely used as a reporter gene in VIGS assays in plants as silencing of this gene causes chlorophyll degradation resulting in a typical photobleached phenotype in emerging leaves (Kumagai et al., 1995). To produce a BPMV-PDS inoculum suitable for cv. Champagne inoculation, we first inoculated a mix of infectious plasmids corresponding to the BPMV-PDS construct (Table 2) in P. vulgaris cv. Black Valentine. Three wpi, typical photobleaching was observed on young uninoculated leaves of~60% of the inoculated cv. Black Valentine plants ( Figure S2). This observation confirms that the BPMV-PDS vector is functional and reveals that heterologous VIGS is induced in P. vulgaris. Indeed, a sequence alignment showed that the PDS fragment from pea shares 91% nucleic acid identity with its homologue in common bean (data not shown).
White photobleached leaves of P. vulgaris were used as inoculum to produce leaf sap. Mechanical inoculation of pea cv. Champagne was performed using the scarification method described above. At 6 wpi, plants inoculated with BPMV-PDS displayed a clear photobleached phenotype with completely white newly emerging leaves, unlike plants infected with the empty BPMV-0 vector or mock buffer (Figure 2a). To confirm that the photobleached phenotype described above correlated with reduced endogenous levels of PsPDS, semi-quantitative RT-PCR was carried out on uninoculated leaves from each of the three treatment groups (Figure 2b). To test whether the phenotype observed in treated plants could be due to the presence of the viral vectors, the presence of BPMV RNA1 and RNA2 transcripts was also determined by RT-PCR (  Fluorescence observed on a quarter, a half or three quarters of the aerial parts of the plant, respectively.

BPMV infection of roots of six P. sativum cultivars
As one of our aims in developing BPMV-derived VIGS for pea is to identify genes involved in resistance to the root pathogen A. euteiches, we next investigated whether BPMV could be a suitable VIGS vector for silencing endogenes expressed in roots.
To examine whether BPMV infects roots, we first tested the presence of GFP fluorescence in the root systems of genotypes Isard and 552 at 6 and 5 wpi, respectively. When young lateral roots were observed with an epifluorescence microscope, GFP fluorescence was detected for the two genotypes Isard and 552 ( Figure 3a). As an alternative to microscope observations, we tested the presence of BPMV-GFP in the roots of all six genotypes using semi-quantitative RT-PCR analyses. Total RNAs were extracted from roots of BPMV-GFP-infected plants and control plants at 4 wpi. PCR using primers specific to BPMV RNA1 showed the presence of viral RNAs in the roots of BPMV-GFPinfected plants for all six genotypes, whereas no viral RNAs were detected in control plants as expected ( Figure 3b). Thus, the results confirmed the presence of BPMV viral RNAs in roots of the six genotypes of interest.
Virus-induced gene silencing of PsKOR1 in P. sativum cv. Champagne To test BPMV VIGS in roots, we constructed a BPMV-KOR1 vector containing a fragment of the KORRIGAN-1 gene of P. sativum (PsKOR1). In Arabidopsis thaliana, the KOR1 gene encodes an endo-1,4-b-D-glucanase involved in cellulose synthesis (Sato et al., 2001). Mutants carrying T-DNA insertions in the putative promoter region display an extreme dwarf phenotype with severe stunting of both stem and roots (Nicol et al., 1998). Silencing of the KOR1 gene in P. sativum using the PEBV VIGS vector resulted in a reduced plant height and a shorter root length (Constantin et al., 2004). We constructed two BPMV-KOR1 vectors. The first vector (so-called BPMV-KOR1a) contains a 345-bp length fragment derived from the 3 0 -end of the PsKOR1 ORF of cv. Champagne (Table 2). Indeed, according to Zhang et al. (2010Zhang et al. ( , 2013, silencing efficiency using the 'one-step' BPMV vector is generally higher for the 3 0 region of the target gene ORF and for a fragment size of~300-400 bp. The second vector (so-called BPMV-KOR1b) contains a 470-bp length fragment homologous to the 470-bp fragment used by Constantin et al. (2004) in the PEBV vector (Table 2). The two KOR1 fragments (345 bp and 470 bp) overlap by 181 bp on the PsKOR1 mRNA. We first inoculated the two BPMV-KOR1 constructs in P. vulgaris cv. Black Valentine by direct DNA rubbing of a mix of two infectious plasmids (Table 2). Three wpi, typical viral symptoms were observed on uninoculated systemic leaves of~60% of the inoculated cv. Black Valentine plants ( Figure S2), but infected plants did not display a reduced plant height (data not shown). This may be due to the low nucleic acid identity (<50%) between the KOR1 fragments of the PsKOR1 gene of pea and PvKOR1 expressed in roots of common bean. Infected leaves of P. vulgaris were used as inoculum to produce leaf sap for mechanical inoculation of pea cv. Champagne. Inoculated plants were put in hydroponic culture. At 4 wpi, root length of each plant was measured and representative plants were photographed (Figure 4a). Roots of plants inoculated with BPMV-KOR1a or BPMV-KOR1b were significantly shorter than roots of plants inoculated with BPMV-0 (Figure 4a and b) supporting the hypothesis that silencing of PsKOR1 had taken place. To correlate this root dwarf phenotype with the transcript extinction of PsKOR1, we performed semi-quantitative RT-PCR in roots at 4 wpi. Total RNA was extracted from plants inoculated with mock, BPMV-0, BPMV-KOR1a and BPMV-KOR1b. To test whether the phenotype observed in treated plants could be due to the presence of the viral vectors, the presence of BPMV RNA1 and RNA2 transcripts was also determined by RT-PCR ( Figure 4c). As expected, no viral RNAs were detected in the roots of the mock-treated plants, unlike in all other samples (Figure 4c). Expression levels of PsKOR1 were similar in roots of BPMV-0-inoculated plants compared to mock-treated plants suggesting that the viral infection does not affect the expression level of PsKOR1 (Figure 4c). Transcripts levels of PsKOR1 were significantly decreased in samples from roots of plants inoculated with either BPMV-KOR1a or BPMV-KOR1b, relative to the control plants (mock and BPMV-0) and relatively to PsPP2A (Figure 4c). No significant difference was detected between BPMV-KOR1a and BPMV-KOR1b samples.
Screening for susceptibility to BPMV of a P. sativum panel of diversity In this work, we demonstrated that six pea genotypes (Champagne, FP, PI 180693, AeD99OSW-50-2-5, Isard and 552) were all susceptible to BPMV and that efficient VIGS was induced using the 'one-step' BPMV vector in leaves and roots of cv. Champagne. To further extend the application of VIGS using the 'onestep' BPMV vector in P. sativum and as VIGS technology can be used only in susceptible genotypes, we carried out a screening experiment using 37 different pea genotypes reflecting the diversity of cultivated peas (Table 4). Three plants of each genotype were tested in at least two independent experiments. Pea plantlets were inoculated at the two-leaf stage using the high-intensity inoculation method and leaf sap obtained from BPMV-GFP-infected leaves of P. vulgaris cv. Black Valentine. The presence of GFP fluorescence was scored under UV light at 3 and 4 wpi in inoculated leaves and uninoculated leaves, respectively, for the 37 pea genotypes, and results were synthesized in Table 4. Twenty-eight genotypes were infected systemically by BPMV-GFP (noted 'YES' in Table 4). Two genotypes (cv. Enduro and James) displayed GFP fluorescence in the inoculated leaves only (noted 'YES only in IL' in Table 4), indicating that systemic movement of BPMV-GFP was restricted. Seven genotypes dis-  Table 4).

Discussion
Pisum sativum is an economically important crop especially in Europe. The genome of pea is about to be completely sequenced by the International Pea Sequencing Project consor-tium (Alves-Carvalho et al., 2015;McGee, 2013). To relate the identified genes with important agronomic traits, functional genomics tool are needed at the genome-wide level. In pea, as in other legumes, stable genetic transformation is relatively inefficient Svabova et al., 2005) and VIGS has become an important reverse genetics tool for functional genomics in these species (Pflieger et al., 2013). In the present study, we investigated whether BPMV-based vector, the most widely used VIGS vector in legumes, could be effective for silencing of endogenous genes and for foreign gene expression in pea. In pea, a VIGS vector was originally developed from PEBV that induced an efficient and reliable gene silencing (Constantin et al., 2004). P. sativum cv. Scout inoculated with PEBV-PDS developed the characteristic photo-bleached phenotype on 95% of the upper leaves at 42 dpi (Constantin et al., 2004). Using the BPMV VIGS vector, we observed comparable intensity and timing of silencing induction for PsPDS as leaf photobleaching in cv. Champagne was observed at 6 wpi and this phenotype was associated with a reduction in PDS transcripts. In roots, silencing of PsKOR1 gene was shown to occur at 4 wpi both in cv. Scout (Constantin et al., 2004) and in cv. Champagne (this study). In conclusion, the one-step BPMV VIGS vector is as effective as the PEBV VIGS vector for silencing of endogenes in pea.
One critical step when making VIGS is the delivery of the recombinant viral vector into plant cells to initiate viral infection (i. e. primary inoculation). Delivery can be achieved by various techniques such as Agrobacterium-mediated infiltration (agroinoculation), mechanical inoculation (generally by leaf rubbing) of in vitro-transcribed RNA or biolistic delivery of infectious plasmid DNA (Burch-Smith et al., 2004;Robertson, 2004). In the case of PEBV, initial delivery is achieved by agro-inoculation because PEBV is not transmitted mechanically (Constantin et al., 2004). In the aim of making high-throughput VIGS studies, this technique may become laborious and time-consuming because of the step of Agrobacterium transformation (Pflieger et al., 2013). On the contrary, BPMV is a 'one-step' VIGS vector delivered into plants via direct DNA rubbing of infectious plasmid DNAs which is a rapid, simple and cost-effective procedure for making low-to high-throughput genomic studies (Pflieger et al., 2013;Zhang et al., 2010).
As VIGS efficiency is correlated with the virus titre in the infected plants (Senthil-Kumar and Mysore, 2011), we performed all primary inoculations of infectious plasmid DNAs in P. vulgaris cv. Black Valentine because this genotype is highly susceptible to BPMV Zhang et al., 2010). Systemically infected leaves of cv. Black Valentine are trifoliate leaves, and each leaflet constitutes a highly efficient inoculum for making secondary inoculations. This is an advantage compared to pea leaflet that are quite smaller and that would produce less inoculum material not suitable for high-throughput genomic studies. In addition, we demonstrated that the secondary infection rates in pea ranges from 100% (in cv. Champagne) to 30% (cv. FP, PI 180693, AeD99OSW-50-2-5, Isard and 552). Thus, to optimize chances to have a successful infection of pea genotypes, we recommend to use a highly efficient viral inoculum produced in P. vulgaris cv. Black Valentine.
Our results showed that BPMV infection in pea only generates mild mosaics (Figures 1 and 4, S1), therefore not interfering with the phenotype associated with silencing of the gene of interest. This feature is of great importance when making VIGS studies because viral infection may alter defence responses, development and morphology of the silenced plant. Thus, it is recommended if possible to work with mild viral strains. This is the case for the 'one-step' BPMV VIGS vector, which is derived from the IA-Di1 isolate of BPMV inducing very mild visual symptoms on infected soya bean plants. In common bean, we showed that although viral symptoms were present (mottling and distorted leaves), they do not interfere with the silencing phenotypes . Nevertheless, it is important to include a positive control (i.e. plant infected with an empty VIGS vector) in all VIGS assays to check the effect of viral infection on the silenced plant.
The ideal size of the foreign sequence inserted into the viral vector for VIGS induction ranges between an upper size limit allowing the virus to replicate and to move systemically and a lower size limit allowing efficient VIGS induction (Burch-Smith et al., 2004;Robertson, 2004). These two parameters depend on the chosen virus and on the host plant genotype. Moreover, orientation of the foreign sequence inserted into the viral vector and the position of the target region inside the target gene are critical for silencing intensity. With the 'one-step' BPMV VIGS vector, fragments in the range of 100-500 nucleotides, inserted in antisense orientation and targeting the ORF 3 0 -end of the target gene, are recommended to ensure efficient silencing (Juvale et al., 2012;Pandey et al., 2011;Zhang et al., 2013). Based on these parameters, we used an insert of 336 and 345 nucleotides selected in the ORF 3 0 -region of PsPDS and PsKOR1 genes, respectively, to construct the two BPMV vectors BPMV-PDS and BPMV-KOR1a. For PsKOR1, we also tested BPMV-KOR1b containing the same 470-nt fragment as previously used with the PEBV vector (Constantin et al., 2004). Although this insert does not comply with the choice of the ORF 3 0 -end region, the observed results showed no significant difference between silenced plants with BPMV-KOR1a and BPMV-KOR1b (Figure 4c). This can be explained by the fact that the 470-nt fragment of BPMV-KOR1b covers a portion of the 3 0 region because the whole PsKOR1 ORF has a length of 688 nt.
One limitation of VIGS is that it is possible only in genotypes where the viral vector can spread systemically. For a given species, different genotypes might react differently to viral infection, from total resistance due to specific resistance genes to various levels of susceptibility. This has been well documented for the PEBVderived vector in Medicago truncatula and Lathyrus odorata (Grønlund et al., 2008). Consequently, it is essential to test the susceptibility to BPMV of each genotype of interest. To our knowledge, the BPMV vector has never been tested on pea genotypes. In a first step, we analysed the susceptibility to BPMV of six pea genotypes of interest and found that all six were susceptible to BPMV (Figures 1 and 3, S1). In these six genotypes, BPMV spread systemically throughout the whole plant (Figures 1and 3, S1). Following BPMV-GFP infection, fluorescence was detected in primary leaves, trifoliate leaves, stems, roots, floral buds, flowers, pods and seed coats (Figures 1and 3, S1). As VIGS is more effective only in tissues infected by the viral vector, it should therefore be possible to induce gene silencing in all these organs. In a second step, we investigated 37 additional pea genotypes and found that 28 were susceptible to BPMV, two genotypes displayed fluorescence only in the inoculated leaf, and only 7 seemed resistant (Table 4). Unfortunately, the genotype Cameor, the genome of which is currently being sequenced (Alves-Carvalho et al., 2015;McGee, 2013), was found to be part of the 7 resistant genotypes. To affirm that these 7 genotypes are completely resistant to BPMV, infection with the empty BPMV vector should be tested because BPMV-GFP may be less efficient to infect these genotypes. Nevertheless, the BPMV vector is usable in the 6 + 28 pea genotypes provided further investigation to improve infection rates, viral spread and VIGS efficiency.
Implementation of BPMV VIGS in pea further supports the use of this VIGS vector in other important legumes for which stable transgenic techniques are not yet available. The extent to which the 'one-step' BPMV vector can be used in other legume species (except soybean, common bean, and pea) is unknown, and it would be interesting to test the susceptibility to BPMV in species such as Cajanus cajan, Cicer arietinum, Vigna radiata whose genomes have recently been sequenced (Kang et al., 2014;Varshney et al., 2012Varshney et al., , 2013, as well as in the model legume Medicago truncatula (Young et al., 2011).
Plant virus-based vectors are also valuable tools for heterologous protein expression in plants. These vectors are fast-acting, cost-efficient and high yield, and significantly, they can be used in a variety of genetic backgrounds provided that the selected genotype is susceptible to the viral vector. Here, we show that the BPMV vector is a suitable virus-based system that is effective for stable expression of a foreign protein in pea. We found that a recombinant BPMV-GFP vector gives an intense GFP expression in a large range of vegetative and reproductive organs in several susceptible pea genotypes (Figures 1 and S1). Further investigations are needed to determine the maximal insert size compatible with virus infectivity and systemic movement for a given genotype. Previous studies have estimated the BPMV RNA2 vector capacity for foreign gene insertion to be between 1.4 and 1.8 kb (Zhang and Ghabrial, 2006;Zhang et al., 2010). Although the average length of a plant gene is~3 kb, it could be possible to express short proteins such as viral or fungal effectors involved in plant immune responses.
In this study, we found that seven pea genotypes were resistant to BPMV-GFP. Resistant genotypes have also been identified previously in common bean (Mozzoni et al., 2009;Pflieger et al., 2014). To our knowledge, no resistance to BPMV has been reported so far in soybean (Mozzoni et al., 2009;Wang et al., 2005;Zheng et al., 2005). This result suggests (i) that pea has an R gene conferring resistance to BPMV that could be an orthologue of R-BPMV (the R gene present in common bean cv. BAT93)  or (ii) that pea may have one or several other different R genes conferring resistance to BPMV. As the resistant pea genotypes do not belong to a single group of diversity (data not shown), we favour the latter hypothesis. However, genetic analysis will be necessary to confirm whether one or several R genes are involved in BPMV resistance in P. sativum.

Plant material and growth conditions
Seeds of Phaseolus vulgaris cv. Black Valentine used for primary inoculations in this study were obtained from multiplications performed at IPS2, Orsay (France). All seeds of Pisum sativum genotypes were obtained from multiplications performed at INRA Rennes (France) and INRA Dijon (France). The thirty-seven genotypes used for the screening for susceptibility to BPMV belong to a panel of diversity of cultivated P. sativum established at INRA Dijon (unpublished results). Seedlings of common bean and pea genotypes were cultivated as described previously (Pflieger et al., , 2015. Before inoculation, pea seedlings for PsKOR1 silencing assays were placed individually in hydroponic culture using a nutrient solution as substrate (Pflieger et al., 2015). At 14 dpi, roots were cut back to approximately 3 cm and root growth was allowed to re-initiate as described by Constantin et al. (2004) in PEBV VIGS experiments.

Viral vectors
Names of the BPMV constructs used in this study and the corresponding RNA1-and RNA2-derived plasmids are listed in Table 2. The pBPMV-IA-R1M, pBPMV-IA-V1 and pBPMV-GFP2 DNA plasmids were kindly supplied by C. Zhang from Iowa State University, USA (Zhang et al., 2010). All other RNA2-derived plasmids were constructed in our laboratory by insertion of P. sativum target gene fragments in the BamH1 restriction site of pBPMV-IA-V1. Sequences of 1077 bp corresponding to PDS mRNA of P. sativum (GenBank accession AJ621573.1, cv. Scout) and 688 bp corresponding to KOR1 mRNA (GenBank accession AJ621355.1, cv. Scout) were used as matrix to determine the PCR primers using the software Primer3 (Untergasser et al., 2012). Target gene fragments were first amplified by PCR using specific primers (Table S1), P. sativum cDNAs as template and a highfidelity polymerase (Advantage â -HF2, Clontech, Mountain View, CA, USA). DNA plasmids pBPMV-IA-PDSrev336pb, pBPMV-IA-KORrev-345pb and pBPMV-IA-KORrev-470pb (Table 2) were constructed by cloning cohesive-end fragments of 336 bp of the PsPDS gene, and 345 and 470 bp of the PsKORRIGAN1 (KOR1) gene, respectively, into BamH1-digested and dephosphorylated pBPMV-IA-V1. The orientation of the cloned inserts was determined by PCR using a combination of vector-specific and fragment-specific primers (Table S1). Clones thought to contain inserted fragments in the antisense orientation were sequenced (GATC, Konstanz, Germany) to confirm their identity, and a maxipreparation of plasmid DNA was performed using the 'QIAGEN Plasmid Purification' (Hilden, Germany) kit.

Viral inoculation procedure
Primary inoculations with BPMV-derived infectious plasmids were performed in common bean cv. Black Valentine as described previously (Pflieger et al., , 2015. For secondary inoculations of pea genotypes, an infected leaf of P. vulgaris cv. Black Valentine harvested at 4 to 5 wpi (fresh leaf or frozen leaf conserved at À80°C) was ground in a mortar with mock buffer to make leaf sap. Mechanical inoculations were performed by rubbing of the leaflets and stipules of two leaves of pea plantlets at the two-leaf stage of development (approximately 10-to 12day-old plantlets in our growth conditions). Plants inoculated with either inoculation buffer without virus (Mock) or empty BPMV vector (BPMV-WT) were used as controls. After viral inoculation, the potted plants or hydroponic-cultured plants were placed in a growth room at 19°C under a 16-h light/8-h dark cycle under a humidity of 70% and watered with nutrient solution (Pflieger et al., 2015).

Optimization of conditions for mechanical inoculation of leaf sap in P. sativum
To increase the infection rate in pea, two intensities of mechanical inoculation were tested on the two younger leaves and stipules of 10-to 12-day-old plants of cv. Isard and 552: abrasion and scarification. For abrasion, the leaf was dusted with carborundum and rubbed with miracloth soaked with leaf sap. For scarification, a scalpel blade was used to make superficial incisions in the waxy layer and in the upper epidermis of the leaf. Depending on the size of the leaflet surface, four to six parallel incisions were made on one leaflet. The wounded upper leaf surface was then rubbed as in the abrasion inoculation method. Six seedlings of both cv. Isard and 552 were inoculated with BPMV-GFP, with three plants per inoculation method. Inoculum consisted of leaf sap derived from P. vulgaris cv. Black Valentine plants infected with BPMV-GFP. After inoculation, the GFP fluorescence was scored in the inoculation leaf at 14 dpi and in the upper systemic leaves at 4 and 5 wpi using a quantitative scale of notation.

GFP imaging
GFP expression in aerial organs and roots was examined using a UV lamp (High-intensity 100-Watt long-wave UV lamp, UVP, USA) and photographed using a Canon PowerShot A2300 digital camera. Microscopic observations of roots were made using an Axioskop microscope (Zeiss, Germany) and photographed using a RTKE camera (SPOT imaging, Sterling Heights, MI, USA).

Expression analysis by semi-quantitative RT-PCR
Total RNA and cDNA syntheses were performed as described in Richard et al. (2014). Primer sequences are listed in Table S1. Primers for RT-PCR analyses of PsPDS and PsKOR1 genes were designed to anneal outside the region targeted for silencing so that only the endogenous gene was probed. Primers PsPP2A-fwd and PsPP2A-rev were used to amplify PsPP2A as an internal constitutively expressed mRNA control (Die et al., 2010). After 25 cycles, PCR products were analysed using analytical agaroseethidium bromide electrophoresis.

Statistical analysis
Statistically significant differences were analysed by Student's t-test. Significance was set at P < 0.01.

Supporting information
Additional Supporting information may be found in the online version of this article: