In Planta Functional Analysis and Subcellular Localization of the Oomycete Pathogen Plasmopara viticola Candidate RXLR Effector Repertoire

Downy mildew is one of the most destructive diseases of grapevine, causing tremendous economic loss in the grape and wine industry. The disease agent Plasmopara viticola is an obligate biotrophic oomycete, from which over 100 candidate RXLR effectors have been identified. In this study, 83 candidate RXLR effector genes (PvRXLRs) were cloned from the P. viticola isolate “JL-7-2” genome. The results of the yeast signal sequence trap assay indicated that most of the candidate effectors are secretory proteins. The biological activities and subcellular localizations of all the 83 effectors were analyzed via a heterologous Agrobacterium-mediated Nicotiana benthamiana expression system. Results showed that 52 effectors could completely suppress cell death triggered by elicitin, 10 effectors could partially suppress cell death, 11 effectors were unable to suppress cell death, and 10 effectors themselves triggered cell death. Live-cell imaging showed that the majority of the effectors (76 of 83) could be observed with informative fluorescence signals in plant cells, among which 34 effectors were found to be targeted to both the nucleus and cytosol, 29 effectors were specifically localized in the nucleus, and 9 effectors were targeted to plant membrane system. Interestingly, three effectors PvRXLR61, 86 and 161 were targeted to chloroplasts, and one effector PvRXLR54 was dually targeted to chloroplasts and mitochondria. However, western blot analysis suggested that only PvRXLR86 carried a cleavable N-terminal transit peptide and underwent processing in planta. Many effectors have previously been predicted to target organelles, however, to the best of our knowledge, this is the first study to provide experimental evidence of oomycete effectors targeted to chloroplasts and mitochondria.

Downy mildew is one of the most destructive diseases of grapevine, causing tremendous economic loss in the grape and wine industry. The disease agent Plasmopara viticola is an obligate biotrophic oomycete, from which over 100 candidate RXLR effectors have been identified. In this study, 83 candidate RXLR effector genes (PvRXLRs) were cloned from the P. viticola isolate "JL-7-2" genome. The results of the yeast signal sequence trap assay indicated that most of the candidate effectors are secretory proteins. The biological activities and subcellular localizations of all the 83 effectors were analyzed via a heterologous Agrobacterium-mediated Nicotiana benthamiana expression system. Results showed that 52 effectors could completely suppress cell death triggered by elicitin, 10 effectors could partially suppress cell death, 11 effectors were unable to suppress cell death, and 10 effectors themselves triggered cell death. Live-cell imaging showed that the majority of the effectors (76 of 83) could be observed with informative fluorescence signals in plant cells, among which 34 effectors were found to be targeted to both the nucleus and cytosol, 29 effectors were specifically localized in the nucleus, and 9 effectors were targeted to plant membrane system. Interestingly, three effectors PvRXLR61, 86 and 161 were targeted to chloroplasts, and one effector PvRXLR54 was dually targeted to chloroplasts and mitochondria. However, western blot analysis suggested that only PvRXLR86 carried a cleavable N-terminal transit peptide and underwent processing in planta. Many effectors have previously been predicted to target organelles, however, to the best of our knowledge, this is the first study to provide experimental evidence of oomycete effectors targeted to chloroplasts and mitochondria.

INTRODUCTION
Oomycetes are a group of straminipilous organisms, which are thought to have arisen from biflagellate, free-ranging predatory protists, and are most closely related to diatoms and brown algae, although they have evolved to possess characters similar to those of true fungi (Sekimoto et al., 2008;Beakes et al., 2012Beakes et al., , 2014. Some oomycetes are highly destructive plant pathogens. For example, the potato late blight pathogen Phytophthora infestans not only caused the Irish famine in the nineteenth century but continues to affect modern agriculture (Dyer et al., 1993;Kamoun and Smart, 2005;Tyler et al., 2006;Fry, 2008;Yoshida et al., 2013).
The grapevine downy mildew pathogen Plasmopara viticola is an obligate oomycete belonging to the Peronosporaceae family, which causes severe damage to cultivated grapevines and wild species worldwide (Rouxel et al., 2013). As the most common varieties of wine and table grapes lack genetic resistance to this pathogen, control methods against P. viticola are mainly based on the frequent use of fungicides (Feechan et al., 2013). Due to the potential harmful impacts of agrochemicals on the environment and vineyard workers, finding alternative methods for the control of grapevine downy mildew is a high priority. However, the obligate biotrophic life cycle of P. viticola and the low efficiency of grapevine transformation make it difficult to investigate the molecular basis of the grapevine-downy mildew interaction (Perl et al., 1996;Kiefer et al., 2002;Kortekamp and Zyprian, 2003;Valsesia et al., 2005;Kortekamp, 2006;Dubresson et al., 2008;Kortekamp et al., 2008;Rossi and Caffi, 2012;Feechan et al., 2013;Rossi et al., 2013).
Most plant pathogens, including bacteria, fungi, oomycete, nematodes, and insects, are known to secrete a diverse array of low molecular weight proteins called effectors, which are thought to modulate plant immune responses and enhance pathogenicity (Jones and Dangl, 2006;Kamoun, 2006Kamoun, , 2007Zipfel, 2008Zipfel, , 2014Boller and He, 2009;Dodds and Rathjen, 2010;Wang et al., 2011). These effectors can be classified into two main categories, apoplastic, and cytoplasmic, based on their final locations in plant cells. The functions of most cytoplasmic effectors are poorly understood in comparison with those of apoplastic effectors, which often inhibit the activities of extracellular enzymes or interfere with the functions of host receptors (Damasceno et al., 2008;Ma et al., 2017). In contrast to most bacterial and fungal effectors, which rarely possess conserved motifs and domains, oomycete cytoplasmic effectors are modular proteins that carry N-terminal signal peptides followed by certain conserved motifs, notably the RXLR and LXLFLAK motifs. The RXLR motif defines a domain that is similar to a host-targeting translocation signal observed in the effectors secreted by malaria parasites, but the biological function of RXLR motif is still controversial (Birch et al., 2006;Dou et al., 2008a,b;Ellis and Dodds, 2011;Wawra et al., 2013Wawra et al., , 2017. These so-called RXLR effectors have been identified in a diverse range of Phytophthora and downy mildew pathogens, but appear to be absent in necrotrophic pathogens, such as members of the Pythium genus and the animal pathogens belonging to the Saprolegnia genus (Lévesque et al., 2010;Krajaejun et al., 2011Krajaejun et al., , 2014Adhikari et al., 2013;Jiang et al., 2013). A number of RXLR effectors have been shown to suppress programmed cell death in plant cells (Bos et al., 2006;Dou et al., 2008a;Oh et al., 2009). However, the function of most RXLR effectors identified in obligate biotrophic pathogens, remains unclear.
Once inside host cells, effectors are transported to distinct compartments according to their sorting signals. Most effectors have been shown to be targeted to the nucleus, cytosol, plasma membrane, or endoplasmic reticulum (Caillaud et al., 2012;Lindeberg et al., 2012;Hicks and Galán, 2013). However, a few effectors have been reported to target chloroplasts and mitochondria. For example, four type-III bacterial effectors Hopl1, HopN1, HopK1, and AvrRps4 were reported to target chloroplasts (Jelenska et al., 2007;Rodríguez-Herva et al., 2012;Li et al., 2014). HopK1 and AvrRps4 are thought to target chloroplasts via their cleavable N-terminal transit peptides (Li et al., 2014), but similar N-terminal transit peptides appear to be lacking in Hopl1 and HopN1. The ToxA effector from the fungus Pyrenophora tritici-repentis not only targets the host chloroplasts, but also has the ability to cross the plant plasma membrane via the apoplastic space (Manning and Ciuffetti, 2005;Manning et al., 2007Manning et al., , 2009. More recently, the effectors CTP1, CTP2, and CTP3 secreted by the rust fungus Melampsora larici-populina were demonstrated to carry cleavable N-terminal transit peptides, which targeted them to chloroplasts (Petre et al., 2016). In addition, it was demonstrated that the type-III effector HopG1 was targeted to mitochondria and that it altered plant development, resulting in dwarfism, increased branching, and infertility (Block et al., 2010). However, oomycete effectors targeting chloroplasts and mitochondria have rarely been reported (Schornack et al., 2010;Caillaud et al., 2012;Stam et al., 2013).
Plasmopara viticola is thought to secrete a repertoire of effectors to modulate host physiological processes during infection, in a manner similar to that of other oomycete pathogens (Mestre et al., 2012(Mestre et al., , 2016Yin et al., 2015;Xiang et al., 2016). Recently, the genome sequencing of the isolate "JL-7-2" revealed that P. viticola might potentially encode at least 100 candidate RXLR effectors (PvRXLRs) (Yin et al., 2017). The next challenge is to assess the functions of these effectors. However, as the pathogen is recalcitrant to stable transformation, it will be necessary to develop other approaches to determine the function of these P. viticola candidate effectors. The most widely used method for functional characterization of biotrophic pathogen effectors is based on a surrogate heterologous expression system, which facilitates the expression of effectors in plant cells directly via a bacterial type-three secretion system (TTSS) (Fabro et al., 2011), or agroinfiltration. Although the delivery of effectors into plant cells via the TTSS-mediated approach is efficient, the lack of post-translational modifications and the potential effect of co-expressed bacterial effectors may complicate the interpretation of results with this assay. In contrast, Agrobacterium-mediated heterologous expression in Nicotiana benthamiana allows for the transient expression of proteins in leaf cells, and a wide range of assays are available for conducting functional investigations. In this study, we successfully utilized the Agrobacterium-mediated heterologous expression strategy to investigate a total of 83 candidate PvRXLR effectors leading to a better understanding of the pathogenicity of this important pathogen.

Microbial Strains, Plants, and Culture Conditions
Escherichia coli DH5α and Agrobacterium tumefaciens GV3101 were routinely grown in Luria-Bertani (LB) media containing the appropriate antibiotics at 37 and 30 • C, respectively. DNA transformations of DH5α and GV3101 were conducted by using standard protocols for heat shock treatment and electroporation, respectively. The P. viticola isolate "JL-7-2" was maintained on detached leaves of V. vinifera cv. Thompson seedless at 18 • C over a 16 h photoperiod. Nicotiana benthamiana, V. vinifera and A. thaliana plants were grown and maintained at 22 • C in a controlled environment greenhouse with a 12 h photoperiod.

DNA and RNA Manipulation
Grapevine leaves were surface-sterilized with chlorine bleach, followed by three washes in sterile water. The leaves were sprayed with a suspension containing 1 × 10 6 sporangia/mL using an air brush, kept on 0.8% (w/v) agar plates, and incubated in a growth chamber at 18 • C with a 16 h photoperiod. Leaves were sampled 3-4 days after infection and the total DNA and RNA extracted as described previously (Zhang et al., 2008). The cDNA was obtained using a one-step gDNA removal and cDNA synthesis supermix kit (TransGen Biotech, Beijing, China), according to the manufacturer's protocols.

Gene Cloning, RT-PCR, and Vector Construction
Full-length candidate effector genes (from ATG start codon to stop codon) were amplified from total genomic DNA by using 2 × High-Fidelity Master Mix (Tsingke Biotech, Beijing, China) ( Table S2). In brief, 20 ng of total gDNA was amplified with an initial denaturation step at 98 • C for 2 min; this was followed by 30 cycles at 98 • C for 10 s, at 55 • C for 15 s, and 72 • C for 30 s. In the last cycle, a final elongation was carried out at 72 • C for 1 min. For RT-PCR analysis, 50 ng of total RNA was used and amplified, as described above. PCR products were separated using 1.5% (w/v) TAE agarose gels and purified using a Gel Extraction Kit (Omega, Norcross, USA). The DNA fragments were inserted into a cloning vector pLB-Simple (Tiangen Biotech, Beijing, China) containing blunt-ends, and transformed into E. coli DH5α. Positive amplicons were confirmed by colony PCR. Five independent positive colonies carrying an insert of the expected size were selected for plasmid purification (Omega) and sequencing (Tsingke).
Sequences encoding mature PvRXLR proteins (i.e., without a signal peptide, in which translation is initiated from an artificial ATG codon) were introduced into linearized pGR106 (Potato virus X-based binary expression vector) (Chapman et al., 1992;Takken et al., 2000), pBI121, or pSuper1300 (Yang et al., 2010) using the In-Fusion cloning kit, according to the manufacturer's instructions (Tsingke). The gene sequence encoding the mature protein PvRXLR54 was introduced into a linearized pUC19-35S-GFP-RBS vector after Xho I and BstB I digestion, and the recombinant plasmids were purified by cesium chloride density gradient ultracentrifugation (Glisin et al., 1974). The DNA sequences of ScCOX4 1−29 and GmMan1 1−49 were synthesized by Tsingke Biological Technology and inserted into the pSuper1300 plasmid.

Yeast Signal Sequence Trap Assay
The predicted DNA fragments encoding for the PvRXLR signal peptides were amplified by PCR, and purified using an Oligo DNA purification Kit (Shangon Biotech, Shanghai, China), and introduced into the linearized pSUC2 vector (pSUC2T7M13ORI) following EcoR I and Xho I (New England Biolabs, Ipswich, USA) digestion. The recombinant plasmids were transformed into the invertase-negative yeast strain YTK12 by using the lithium acetate method. Cells were plated onto CMD-W media (

Functional Verification of Candidate PvRXLR Effectors in Planta
Four-to five-week-old N. benthamiana seedlings were used for agroinfiltration. Agrobacterium tumefaciens strain GV3101 carrying pGR106-PvRXLR effector plasmids or the negative control pGR106-GFP plasmid was cultured in LB liquid media (200 rpm, 30 • C). The cells were collected by centrifugation (2,500 × g, 5 min), and resuspended with 10 mM MgCl 2 until an OD 600 of 0.4 was achieved, and then incubated at room temperature for 3 h before infiltration. For cell death suppression assays, A. tumefaciens strains carrying effectors and control GFP plasmids were first used to infiltrate the left and right sides of N. benthamiana leaves, respectively. One day later, the same areas of the leaves were re-infiltrated with Agrobacterium cultures carrying either the pGR106-BAX or pGR106-INF1 constructs. Co-expression of effectors and p19 was achieved by adjusting the OD 600 of the two Agrobacterium cultures to 0.8, and mixing the culture solutions in a ratio of 1:1 (final OD 600 of 0.4). The Agrobacterium culture carrying the pGR106-GFP vector was infiltrated at the same time as a control.

Live-Cell Imaging
Agrobacterium tumefaciens cultures were infiltrated into N. benthamiana leaves in the manner described above with some modifications. For achieving subcellular localization of a single protein, the final OD 600 was adjusted to 0.7; for co-localization studies, the final OD 600 of two mixed A. tumefaciens cultures was adjusted to 0.5 (1.0 of each prior to mixing). Inoculated leaves were maintained under condition of low light and collected 36-72 h post-infiltration for analysis using a Leica TCS SP 8 confocal laser scanning platform (20× air lens). The excitation wavelengths were set as follows: GFP and chlorophyll (488 nm) and mCherry (561 nm). The fluorescence signals for GFP, chlorophyll, and mCherry were detected at 505-525, 580-620 and 680-700 nm, respectively. Scanning was performed in a sequential mode when required, and all the images obtained were those of single optical sections, with cells exhibiting moderate fluorescence intensity levels. Sub-nuclear localizations were named according to the naming convention described by Caillaud et al. (2012).

Protoplast Transformation and Production of Transgenic Plants
Protoplast transient expression assays were performed essentially according to the method described by Yoo et al. (2007) and Zhao et al. (2016), with some modifications. Grapevine leaves were cut into 0.5 mm strips and incubated in an enzyme solution (1.5% [w/v] Cellulase RS, 0.75% [w/v] Macerozyme R-10, 0.5 M mannitol, 10 mM MES at pH 5.7, 10 mM CaCl 2 and 0.1% [w/v] BSA) for 5 h in the dark with gentle shaking (50 rpm, 25 • C). Protoplasts were filtered through Miracloth and washed with W5 solution (154 mM NaCl, 125 mM CaCl 2 , 5 mM KCl, and 2 mM MES, pH 5.7). The pellets were collected by centrifugation at 100 × g for 5 min and resuspended in MMG solution (0.4 M mannitol, 15 mM MgCl 2 and 4 mM MES, pH 5.7) at a concentration of 1 × 10 6 cells/mL. The plasmid pUC19-35S-PvRXLR54-GFP (20 µg) was purified by ultracentrifugation and mixed with 200 µL protoplasts, followed by the addition of an equal volume of freshly prepared PEG solution (40% [w/v] PEG 3350, 0.4 M mannitol and 100 mM CaCl 2 ). After incubation in PEG for 10 min at room temperature, the protoplasts were collected and washed with W5 and re-suspended in 1 mL WI solution (0.5 M mannitol, 20 mM KCl, 4 mM MES, pH 5.7), and incubated for 16 h at room temperature in the dark. Confocal microscopy was performed as described above.
The A. tumefaciens strain GV3101 carrying the pBI121-35S-PvRXLR54-GFP plasmid was used for stable transformation. Col-0 A. thaliana plants were transformed using the dipping method (Clough and Bent, 1998). The seeds were selected after cultivation in MS (without sucrose) medium containing 0.4% (w/v] phytagel and 50 µg/mL kanamycin. Transgenic plants were confirmed by GFP microscopy, transferred to soil and seeds were collected. T1 seedlings were analyzed by confocal microscopy. Nicotiana benthamiana (same strain as that used for agroinfiltration) was transformed using the leaf disc method (Horsch et al., 1985). The leaf discs were cultured on MS medium containing 0.8% (w/v) agar, 0.5 µg/mL IAA, 2 µg/mL NAA, 50 µg/mL kanamycin, and 200 µg/mL Timentin. Three weeks later, the shoots were transferred to MS medium containing 0.4% (w/v) phytagel, 0.2 µg/mL IAA, 10 µg/mL kanamycin, and 100 µg/mL Timentin. The regenerated seedlings were transferred to soil and cultivated for 1 week, after which confocal microscopy was performed. The reagents for plant tissue culture were purchased from Caisson Laboratories Company (Smithfield, UT, USA). The enzymes and other reagents used for achieving protoplast expression were purchased from Yakult (Minatoku, Tokyo, Japan) and Sigma-Aldrich (St. Louis, MO, USA), respectively.

Protein Isolation and Immunoblotting
Chloroplast protein extraction was carried out using the method described by Petre et al. (2016). Leaves were frozen in liquid nitrogen and ground into a powder. The lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM KCl, 1 mM EDTA, 0.5% [v/v] Triton X-100, 1 mM DTT, 1 mM PMSF, 5 µM MG132, 1 × Protease Inhibitor Cocktail) was added to the mixture and it was kept on ice for 20 min. The extracts were centrifuged and supernatant was collected (14,000 × g, 20 min, 4 • C). Protoplasts were collected (200 g, 2 min), mixed with 100 µL lysis buffer, and kept on ice for 5 min. Protoplast proteins were extracted as described above. The total proteins were loaded onto 12% SDS-PAGE gels and run for 1.5 h at 150 V. Gels were blotted onto a nitrocellulose membrane (PALL, Port Washington, NY, USA) for 1 h at 250 mA, and stained with Ponceau solution to confirm protein loading and transfer. Membranes were blocked in 5% [w/v] fat-free milk in 1 × TBST (0.5% [v/v] Tween-20), after which the primary monoclonal GFP antibody (TransGen) or polyclonal Rubisco large subunit (rbcL) antibody (Agrisera, Vännäs, Sweden) was added at a dilution of 1:10,000. The membrane was washed with TBST three times before the secondary anti-mouse or anti-rabbit IG-HRP raised in a goat (Sigma-Aldrich) was added at a 1:10,000 dilution. Blots were developed using the SuperSignal West Femto ECL kit (Thermo Scientific, Rockford, IL, USA).

Cloning Effector Genes From the P. viticola
Isolate "JL-7-2" A total of 83 candidate PvRXLR genes were successfully cloned from the P. viticola "JL-7-2" genome. Sequence analysis confirmed that 76 of these genes had the same length compared with the published genome sequence (Yin et al., 2017). In addition, three PvRXLR genes had insertions, which resulted in longer ORFs; and four PvRXLR genes had premature stop codons, which resulted in shorter ORFs. The presence of inconsistent ORFs was probably attributable to these effectors having highly similar paralogous genes, or a mismatched genome assembly. None of the cloned effector genes were found to contain introns. Of the 76 effectors with verified sequences, 49 encoded proteins which were identical to previously predicted sequences, while the remaining 27 exhibited SNPs, resulting in differences in one or several amino acids ( Table S1). The predicted length of the PvRXLR proteins ranged from 76 to 720 aa, and the average length was 260 aa (Figure 1). Of the 83 cloned PvRXLR genes, 45 were previously shown to be expressed in infected grapevine leaves by RNA-Seq analysis (Yin et al., 2015). To determine whether the remaining 38 candidate PvRXLR genes were transcribed, RT-PCR analysis was carried out on the total RNA extracted from "JL-7-2" infected grapevine leaves. The results confirmed that 33/38 could be amplified from cDNA ( Figure S1). Thus, in total, 78 out of the 83 predicted PvRXLR genes (∼94%) in the P. viticola "JL-7-2" genome have been confirmed to be expressed during grapevine leaf infections.

Functional Validation of Predicted Signal Peptides of PvRXLR Proteins
To functionally validate signal peptide predictions at the N-terminus of PvRXLR effectors, we used a genetic assay based on the requirement that yeast cells need to secrete invertase to be able to grow on raffinose media (Klein et al., 1996;Jacobs et al., 1997;Lee et al., 2006;Oh et al., 2009). To perform this assay, signal sequences of 58 PvRXLR effectors with S-score values ranging from 0.812 to 0.995 were randomly selected and cloned into the yeast invertase vector pSUC2. The pSUC2 vector carrying the signal peptide of the RXLR effector Avr1b from Phytophthora sojae (Shan et al., 2004) was used as the positive control (CK +), and the untransformed YTK12 was used as negative control (CK−). The results indicate that 49 out of the 58 (∼85%) predicted PvRXLR effector signal peptides supported invertase secretion, and allowed the YTK12 strain to grow on YPRAA medium (Figure S2). These results confirm that the majority of candidate PvRXLR effectors are likely to be secreted from the pathogen into the host plant cell.

PvRXLR Effectors Suppress Programmed Cell Death in N. benthamiana
To identify PvRXLR effectors possibly having a role in suppressing plant immunity, we first infiltrated A. tumefaciens cultures carrying the pGR106 expression vector containing the 83 different PvRXLR gene sequences (minus their predicted signal peptide sequence) into N. benthamiana leaves. After 24 h, these leaves were re-infiltrated with cultures carrying PCD-inducing BAX (a mammalian pro-apoptotic member of the Bcl-2 family, which induces apoptosis-like cell death in N. benthamiana; Lacomme and Santa Cruz, 1999) or INF1 (a 10-kD extracellular elicitin from P. infestans, which induces a hypersensitive response in N. benthamiana) (Kamoun et al., 1998). The cell death phenotype was scored after 4-5 days.
The results revealed that 52 PvRXLR effectors (∼63%) could suppress INF1-and BAX-induced cell death completely ( Figure 2B and Figure S3), while 10 effectors could only partially suppress cell death, as indicated by the weaker necrosis ( Figure 2C and Figure S4A). However, 21 PvRXLR effectors were unable to suppress the necrosis triggered by the expression of INF1 and BAX (Figures 2D,E and Figures S4B,C). These results suggest that a majority of the PvRXLR effectors play a role in suppressing host immunity (Figure 2A and Figure S5).

Certain PvRXLR Effectors Trigger Programmed Cell Death in N. benthamiana
The necrotic areas observed for 13 out of the 21 effectors that were unable to suppress INF1-triggered cell death were larger than those observed for the INF1 control ( Figure 2E and Figure S4C). This result suggested that these PvRXLR effectors themselves trigger cell death. To test this hypothesis, A. tumefaciens strains carrying the pGR106 expression vector containing these 13 different PvRXLR effector gene sequences FIGURE 1 | Analysis of amino acid length in 83 cloned candidate PvRXLR effectors. The dashed line indicates the average length of all effectors. Red bars indicate effectors whose expression had been previously detected by RNA-Seq analysis (Yin et al., 2015), blue bars indicate effectors whose expression is detected by RT-PCR analysis (in this study), and green bars indicate effectors whose expression during grapevine infection is unconfirmed.
Frontiers in Plant Science | www.frontiersin.org were agroinfiltrated into N. benthamiana leaves. The results showed that 10 out of the 13 non-suppressing effectors could trigger cell death independently, but that the speed of induction of necrosis varied markedly ( Figure 3A). Five effectors induced clear necrosis on day 5, four at day 6, and one at day 9 after inoculation. It has previously been reported that the induction of cell death in N. benthamiana, by the P. infestans RXLR effector PexRD2,was dose-dependent (Oh et al., 2009). To determine if the rate of onset of necrosis induced by the different PvRXLR effectors was related to protein expression levels, the five weaker cell death effectors (PvRXLR95, 117, 122, 111, and 138) were co-expressed with p19, a suppressor of posttranscriptional gene silencing, which is known to increase gene expression in the agroinfiltration assay (Voinnet et al., 2003). The results showed that the onset of necrosis was accelerated for these PvRXLR effectors, when tested in the presence of p19 (Figure 3B), suggesting that effector protein levels were associated with the speed of initiation of programmed cell death. Interestingly, PvRXLR35 induced necrosis much more strongly than any of the other PvRXLR effectors, as indicated by the fact that strong necrosis could still be observed when very low concentrations of Agrobacterium cultures (i.e., OD600 of 0.0005) were used (Figure 3C). The amino acid sequences of these 10 PCD-inducing PVRXLR effectors were analyzed using Protein BLAST, but no conserved domains or homologous proteins were found, suggesting that these effectors were specific to P. viticola.

Subcellular Localization of PvRXLR Effectors in Planta
To characterize the subcellular localizations of these candidate PvRXLR effectors, C-terminal GFP-tagged constructs were generated and expressed in N. benthamiana leaves via agroinfiltration. Live-cell imaging indicated that 76 out of the 83 PvRXLR effectors examined generated informative fluorescence signals in plant cells (Figure 4A and Figure S6). Of these, 29 (∼38%) PvRXLR effectors were targeted specifically to the nucleus (Figure 4B and Figure S7), and 34 (∼45%) were distributed in both the nucleus and the cytosol (Figure 4B and Figure S8). In silico analysis (Motif Scan) indicated that only 24 of the analyzed effectors had canonical nuclear-localization signals (NLS). However, most of the PvRXLR effectors contained NLS-like stretches of amino acids enriched in positively charged residues (arginine and lysine), which were predicted to target the proteins to the nucleus (Table S1) by the LOCALIZER prediction program.
Interestingly, we noted that most PvRXLR effectors targeted to the nucleus showed complex patterns of sub-nuclear distribution. Specifically, 14 effectors were targeted to the nucleolus, of which three (PvRXLR45, 91, and 138) were accumulated at the margin of the nucleolus, but were absent within it ( Figure 4B and Figure S9A). Eleven effectors accumulated in irregular structures within the nucleus; of these, one (PvRXLR40) localized to bubble-like structures, one (PvRXLR128) localized to fiberlike structures, four (PvRXLR24, 83, 111, and 115) localized to speckle-like structures and four (PvRXLR76, 81, 134, and 149) localized to bulk/chunk-like structures (Figure 4B and Figure S9B).
In addition to the nuclear-targeted effectors, nine PvRXLR effectors were targeted to the plant membrane system. Of these, PvRXLR80 and 151 were associated solely with the endoplasmic reticulum (ER), five (PvRXLR85, 90, 143, 153, and 154) were associated solely with the plasma membrane, and three (PvRXLR47 and 126) were found to be associated with other membranes (Figures 4B and Figure S10). Amino acid sequence analysis of these 9 membrane-associated PvRXLR effectors revealed that most of them were predicted to be potential targets of post-translational N-myristoylation modification, which facilitates protein-lipid interactions and plays an essential role in membrane targeting (Sessa et al., 1993;Johnson et al., 1994).
In order to validate the observed subcellular distribution of PvRXLR effectors in N. benthamiana transient assays, 15 effectors were randomly selected to generate stable Arabidopsis transgenic lines. All of the PvRXLR effectors showed the same subcellular localization in stably transformed Arabidopsis transgenic lines, as that observed in transient expression assays (Figure S11), demonstrating that protein localization data derived from N. benthamiana agroinfiltration were highly reliable.

Selected PvRXLR Effectors Are Targeted to Chloroplasts
Pathogen effectors targeted to host cell organelles are relatively uncommon (Petre et al., 2016). In this study, the following four PvRXLR effectors that were localized to chloroplasts were identified: PvRXLR54 was found to accumulate in chloroplasts, nuclei, and cytosolic punctate structures; PvRXLR61 and PvRXLR161 were both localized in chloroplasts and nuclei; PvRXLR86 was targeted specifically to chloroplasts ( Figure 5A). All four of these chloroplastic PvRXLR effectors (54, 61, 161, and 86) are small molecular proteins containing 104, 102, 142, and 198 amino acids, respectively. Amino acid composition analysis revealed that all were rich in leucine (L) and arginine (R), which are known to be important components of chloroplast transit peptides (Bruce, 2000). However, three different software programs (WoLF PSORT, ChloroP, and LOCALIZER) predicted that only PvRXLR86 carried a transit peptide and was potentially targeted to chloroplasts ( Table 1). These four effectors showed a very low similarity with each other, and only PvRXLR161 had a predicted RNase H-like domain in the C-terminal region. Protein BLAST results showed that PvRXLR161 was similar to a candidate RXLR effector of Hyaloperonospora arabidopsis, but no homologous proteins were identified in the database for the other three PvRXLR effectors.
To determine if these effectors carried cleavable chloroplast transit peptides and underwent processing in planta, we performed anti-GFP western blotting with total proteins and chloroplast proteins extracted from N. benthamiana leaves expressing PvRXLR effector-GFP fusion proteins. Figure 5B shows that the positive signals for PvRXLR54-GFP, PvRXLR61-GFP, and PvRXLR161-GFP were in accordance with their predicted molecular weights of 37.5, 37.8, and 41.1 kDa, respectively. However, the positive signal for the PvRXLR86-GFP fusion protein was ∼37 kDa, which is significantly smaller than its predicted size of 47.9 kDa. These results suggest that only PvRXLR86 contains an N-terminal cleavable transit peptide and can be processed in plant cells, indicating that these effectors utilize different pathways to target chloroplasts.

PvRXLR54 Is Targeted to Chloroplasts and Mitochondria
In order to further identify the cytosolic puncta of PvRXLR54, the Golgi and mitochondria marker proteins GmMan1 1−49 (the first 49 aa of Glycine max α-1,2-mannosidase I) and ScCOX4 1−29 (the first 29 aa of Saccharomyces cerevisiae cytochrome c oxidase IV) were selected for labeling these structures, respectively (Köhler et al., 1997;Saint-Jore-Dupas et al., 2006).
The co-expression of PvRXLR54 20−104 -GFP and GmMan1 1−49 -mCherry in N. benthamiana leaf cells showed that there was no overlap between the fluorescent signals ( Figure 6A), indicating that the punctate structures were not Golgi bodies. However, when ScCOX4 1−29 -GFP and PvRXLR54 20−104 -mCherry were co-expressed in N. benthamiana, the green and red fluorescence signals overlapped to form an orange-yellow fluorescent signal (Figure 6A), demonstrating that PvRXLR54 was not only targeted to chloroplasts, but was also targeted to mitochondria. Furthermore, mitochondrial targeting of PvRXLR54 was also observed in transformed N. benthamiana epidermal cells which lack chloroplasts ( Figure 6B). It should be noted that both PvRXLR54 20−104 -GFP and PvRXLR 20−104 -mCherry showed the same subcellular localization, indicating that the fluorescent tags did not influence the subcellular targeting of PvRXLR54.
It has previously been reported that the observed subcellular localization of specific effector proteins might varied, depending on the methods of transformation and the plant species used. For example, AvrRps4-GFP was reported to be localized to the cytoplasm and nucleus in Agrobacterium-mediated N. benthamiana transient assays, but was found to be localized to chloroplasts in transgenic Arabidopsis (Sohn et al., 2012;Li et al., 2014). To test whether the dual-targeting of PvRXLR54 to the chloroplasts and mitochondria was species-specific, PvRXLR54 20−104 -GFP was transiently expressed in grapevine protoplasts, and stably transformed into N. benthamiana and A. thaliana. The results showed that the full-length PvRXLR54 20−104 -GFP protein (37.5 kDa) was present in these plant species, and was targeted to both the chloroplast and mitochondria under these conditions (Figures 6C-F). This suggests that the targeting of the PvRXLR54 effector protein to these organelles utilizes a conserved transport pathway in plant cells.

DISCUSSION
In this study, we have used the heterologous N. benthamiana expression system to investigate the biological activity and subcellular localization of candidate RXLR effectors from P. viticola. A total of 83 candidate PvRXLR genes were successfully cloned from P. viticola genomic DNA and functionally characterized. In a previous study, we concluded on the basis of RNA-Seq analysis, that 45 of these PvRXLR genes were expressed during infection of grapevine leaves (Yin et al., 2017). We have now confirmed using RT-PCR that a further 33 PvRXLR genes are expressed in infected grapevine leaf tissues, indicating that most of the candidate PvRXLR genes predicted within the P. viticola genome are expressed during infection. The remaining five candidate PvRXLR genes are either not expressed or are expressed at levels that could not be detected by RT-PCR. Alternatively, these predicted PvRXLR genes might be expressed during other times in the life cycle of this oomycete pathogen, for e.g., during sporulation.

FIGURE 5 | Chloroplastic localization of PvRXLR effectors. (A)
Effector-GFP fusion constructs were agroinfiltrated into N. benthamiana leaves, and the accumulation and distribution pattern of fluorescent protein-tagged effectors has been analyzed by confocal microscopy 36-72 h post-infiltration. Merged images of GFP (green), chlorophyll (red), and bright fields are shown. The yellow fluorescence in the merged image indicates that these effectors are localized in chloroplasts. (B) Total leaf protein and chloroplast protein extracts of transformed N. benthamiana leaves are separated by using 12% SDS-PAGE, and are transferred onto a nitrocellulose membrane. Immunodetection was performed using an anti-GFP antibody. The ponceau stain and the rubisco (anti-rbcL) are used as loading controls for analyzing the total protein content and chloroplast proteins, respectively. Protein sizes are indicated on the left-hand side of the image in kDa. Asterisks indicate the protein bands discussed in the text. n, nucleus. Scale bar = 20 µm.
The N-terminal secretory signal peptide is a common feature of all effectors, and is generally used as a first pass filter for narrowing down a whole proteome dataset into a shortlist of potential effector candidates. Various techniques have been used to functionally validate the predicted N-terminal secretory signal peptides of pathogen effectors including the adenylate cyclase (CyaA) fusion assay, culture filtrate immunoblotting, and live-cell imaging (Sory et al., 1995;Khang et al., 2010;Ribot et al.,  2013). However, it is not feasible to use above methods with obligate biotrophs such as P. viticola. Therefore, in this study, we utilized the yeast signal sequence trap assay and demonstrated that ∼85% of PvRXLR effector signal peptides were able to facilitate the secretion of invertase from yeast cells, confirming that in silico prediction tools are capable of predicting the secretome of oomycete biotrophic pathogens such as P. viticola with a high degree of accuracy. The suppression of plant innate immunity is thought to be the primary function of bacterial effectors, and is likely to also be an important activity of effectors secreted by oomycetes, fungi and nematodes. For instance, Fabro et al. (2011) reported that 77% of H. arabidopsidis effectors tested could increase Pseudomonas syringae pv. tomato DC3000-LUX (Pst-LUX) growth in Col-0 and suppress Pst CEL-induced callose deposition. Germain et al. (2017) concluded that the majority of poplar rust fungus candidate effectors examined were capable of promoting H. arabidopsidis growth in Arabidopsis. Our results with PvRXLR effectors indicate that over half of the effectors tested could suppress cell death induced by INF1 and BAX in N. benthamiana, which is in line with previous results obtained by Xiang et al. (2016) on a much smaller subset of PvRXLR candidate effectors.
A number of pathogen effectors that cause plant cell death have previously been identified. The best example of plantspecific recognition is the recognition of avirulence effectors either directly or indirectly by nucleotide-binding-leucine-richrepeat receptors (NLRs), which results in the activation of plant programmed cell death, thus halting the further growth of the pathogens (Whitham et al., 1994;Tang et al., 1996;Dodds et al., 2004;Rehmany et al., 2005;Cesari et al., 2013). Phylogenetic analysis revealed that Phytophthora and Plasmopara are closely related groups belonging to the order Peronosporales, and share RXLR-type effectors. Thus, some effectors of Phytophthora and Plasmopara possibly recognize the same protein in N. benthamiana and grapevine species. NLR-mediated cell death is generally subtle and rapid; in this study, a very low level of the PvRXLR35 effector could trigger cell death. Hence, we thought that it might be recognized by an unknown NLR that was conserved in N. benthamiana and grapevine species. The remaining nine effectors might be recognized by NLRs in N. benthamiana as well. Alternatively, one possible explanation is that the high-level expression of these nine PvRXLR proteins in the agroinfiltration assay could lead them to aberrantly bind to proteins, other than their intended targets, leading to unspecific cell death.
Recent studies have demonstrated that many oomycete effectors are targeted to the plant nucleus. For example, 66% of HaRxLR effectors examined were targeted to either the nucleus specifically or to the nucleus and cytoplasm (Caillaud et al., 2012). Schornack et al. (2010) also showed that a subset of Crinkler (CRN) effectors from P. infestans were targeted to the nucleus. Our findings are in line with these observations, as over half of the PvRXLR effectors investigated were targeted to the nucleus. In addition, we observed that multiple PvRXLR effectors showed sub-nuclear localizations with different features. Although the mechanism of protein sub-nuclear distribution remains unknown, it has been reported that many pre-mRNA splicing factors, kinases, and phosphatases could localize to nuclear speckles, and were involved in regulating transcription and pre-mRNA processing (Lamond and Spector, 2003;Spector and Lamond, 2011). Thus, we speculate that these PvRXLR effectors might be involved in the regulation of gene expression in the host.
To date, no oomycete effectors have been confirmed to target chloroplasts or mitochondria, despite the fact that several largescale subcellular localization screening experiments have been conducted with a range of oomycete pathogens. In this study, we identified four PvRXLR effectors that were localized to chloroplasts, including PvRXLR54 dual-targeted to chloroplasts and mitochondria. One explanation for this might be that previous studies used N-terminally tagged GFP-effector fusions, which might have impaired the proper translocation of these effector proteins into the organelles, whereas our experiments used C-terminal effector-GFP fusions. It is unclear how these PvRXLR effectors are transported into host organelles. Most plant chloroplast proteins utilize classical cleavable transit peptides to facilitate import into organelles, while some proteins lacking cleavable transit peptides might target the chloroplasts through the endomembrane system (Chen et al., 2004;Kleffmann et al., 2004;Villarejo et al., 2005;Nanjo et al., 2006).
It will be interesting to examine how the recognition and uptake of PvRXLR54 by the transport systems of both chloroplasts and mitochondria are possible. One explanation is that the effector contains a multipartite transit peptide, in which different segments are recognized by different receptors in each of the mitochondria and plastid import machineries (Carrie and Small, 2013;Baudisch et al., 2014). For example, alpha-MPP2 (one of the two isoforms of the substrate-binding subunit of mitochondria-processing peptidase in Arabidopsis) is composed of three functionally separated domains and carries a dual-target signal. The 29 N-terminal residues of alpha-MPP2 mediate mitochondrial targeting, whereas chloroplast transport requires the entire N-terminal region (Baudisch and Klösgen, 2012). Further studies are necessary to identify the different motifs within PvRXLR54 that control the targeting of this effector to different subcellular organelles. In conclusion, this study has increased our understanding of the molecular mechanisms of pathogenicity of P. viticola. The identification of PvRXLR effectors targeted to chloroplasts and mitochondria have also expanded our understanding of the biological activity of oomycete effectors. Future studies will focus on the identification of the host targets of these PvRXLR effectors, to understand how these secreted pathogenic proteins are able to modulate the innate immune responses of the plant.

AUTHOR CONTRIBUTIONS
All authors contributed to this work. JL conceived and supervised the research, and SS assisted the project operation. XL amplified effector genes, constructed the vectors, transformed Agrobacterium, and performed yeast signal peptide trap assay. JX cultivated N. benthamiana seedlings. XL and YL generated stable transgenic plants and performed agroinfiltration, protoplasts expression experiments, confocal microscopy, western blotting, and data analysis. ID, LY, and JQ performed bioinformatics analysis. YL, XL, and JL wrote the manuscript. JL, ID, and SS edited the manuscript.

ACKNOWLEDGMENTS
We thank the following individuals and research groups for providing materials and advice: Dr. Zhongyue Wang (Chinese Academy of Agricultural Sciences) for providing us with the yeast signal trap system, Dr. Daolong Dou (Nanjing Agricultural University) for providing us with the pGR106 vectors and helpful advice, Dr. Chaozu He (Hainan University) for providing us with the pUC19-35S-GFP-RBS plasmid, Dr. Guiqin Qu (China Agricultural University) for providing us with the pEZS-NL-mCherry plasmid to construct the pSuper1300-mCherry vector and Dr. Junping Ao (State key laboratory of oncogenes and related genes, Shanghai) for his keen support and helpful advice on confocal microscopy and ultracentrifugation.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018. 00286/full#supplementary-material Figure S1 | RT-PCR expression analysis of candidate PvRXLR genes. Fifty nanograms of RNA were isolated from infected leaves and transcribed to cDNA after DNase treatment. Full-length primers were used to amplify effector genes from cDNA; the corresponding RT-PCR results are shown. White asterisks indicate that the bands are specific for effectors, while red asterisks indicate non-specific bands. The housekeeping gene Vvactin was used as a control for cDNA synthesis. M, DNA marker. Figure S2 | Functional validation of predicted signal peptides of PvRXLR effectors in yeast. The predicted effector signal peptide sequences were fused in frame to the invertase gene in the pSUC2 vector, and transformed into yeast YTK12 (invertase negative). Yeast is able to grow on YPRAA only if the invertase protein is secreted. CK-, untransformed YTK12. CK+, Avr1b. Figure S11 | Live-cell imaging of effector-GFP fusions in transgenic A. thaliana seedlings. Confocal images of leaf epidermal cells of 10 days old transgenic plantlets expressing candidate effectors fused to GFP in a Col-0 genetic background. PvRXLR47, 126, and 151 are labeled to the endoplasmic reticulum (ER) (around the nucleus); PvRXLR85, 143, and 154 are labeled to the plasma membrane; the remaining effectors showed nucleocytoplasmic localizations. Scale bar = 10-20 µm.
Table S1 | Amino acid sequences of the selected 83 effectors.