Haustorium formation and a distinct biotrophic transcriptome characterize infection of Nicotiana benthamiana by the tree pathogen Phytophthora kernoviae

Abstract Phytophthora species cause some of the most serious diseases of trees and threaten forests in many parts of the world. Despite the generation of genome sequence assemblies for over 10 tree‐pathogenic Phytophthora species and improved detection methods, there are many gaps in our knowledge of how these pathogens interact with their hosts. To facilitate cell biology studies of the infection cycle we examined whether the tree pathogen Phytophthora kernoviae could infect the model plant Nicotiana benthamiana. We transformed P. kernoviae to express green fluorescent protein (GFP) and demonstrated that it forms haustoria within infected N. benthamiana cells. Haustoria were also formed in infected cells of natural hosts, Rhododendron ponticum and European beech (Fagus sylvatica). We analysed the transcriptome of P. kernoviae in cultured mycelia, spores, and during infection of N. benthamiana, and detected 12,559 transcripts. Of these, 1,052 were predicted to encode secreted proteins, some of which may function as effectors to facilitate disease development. From these, we identified 87 expressed candidate RXLR (Arg‐any amino acid‐Leu‐Arg) effectors. We transiently expressed 12 of these as GFP fusions in N. benthamiana leaves and demonstrated that nine significantly enhanced P. kernoviae disease progression and diversely localized to the cytoplasm, nucleus, nucleolus, and plasma membrane. Our results show that N. benthamiana can be used as a model host plant for studying this tree pathogen, and that the interaction likely involves suppression of host immune responses by RXLR effectors. These results establish a platform to expand the understanding of Phytophthora tree diseases.


| INTRODUC TI ON
Trees are long-lived organisms that must contend with prolonged exposure to attack by multiple pathogens. Some diseases of trees are particularly prominent in public awareness, such as Dutch elm disease, chestnut blight, ash dieback, and sudden oak death, caused by Ophiostoma novo-ulmi, Cryphonectria parasitica, Hymenoscyphus fraxineus, and Phytophthora ramorum, respectively.
Although first identified in the UK, P. kernoviae has also been found in the Republic of Ireland, New Zealand, and Chile, and may originate in New Zealand (Gardner et al., 2015;Sanfuentes et al., 2016;Studholme et al., 2019). There is concern both within the UK and internationally regarding the potential spread of P. kernoviae (Drake & Jones, 2017;Fichtner et al., 2012;Hyun & Choi, 2014;Tracy, 2009;Widmer, 2015). Most research on this species has been directed at understanding its potential host range, epidemiology, and improving detection methods (Gardner et al., 2015;Kong et al., 2012;Miles et al., 2015;Mulholland et al., 2015;Schwenkbier et al., 2015;Widmer, 2015).
The Phytophthora genus comprises over 140 species organized into 12 clades (Jung et al., 2017;Yang et al., 2017). Tree-infecting species are found in all clades of the genus such as P. plurivora in Clade 2, P. pseudosyringae in Clade 3, P. cinnamomi in Clade 7, and P. ramorum in Clade 8. P. kernoviae is more distantly related to these species, in Clade 10, along with P. boehmeriae, P. morindae, P. gondwanensis, P. gallica, and P. intercalaris (Yang et al., 2017). P. kernoviae causes disease and sporulates on Rhododendron leaves, and causes bleeding stem cankers on a range of tree species where it may spread via the xylem or establish asymptomatic infections in roots and leaves (Brasier et al., 2005;Brown & Brasier, 2007;Denman et al., 2009;Fichtner et al., 2011). That P. kernoviae can establish asymptomatic infections suggests that it either evades detection or actively forms a biotrophic interaction with its host plants.
Although there are draft genome assemblies for P. kernoviae strains from diverse geographical sites (Feau et al., 2016;Sambles et al., 2015;Studholme et al., 2016Studholme et al., , 2019, these resources remain to be fully exploited to advance understanding of the biology of this species. Like other Phytophthora species, the P. kernoviae genome encodes numerous candidate effector proteins (McGowan & Fitzpatrick, 2017) that may act to facilitate infection through a variety of mechanisms (Wang, Tyler, et al., 2019). Prominent among effector proteins in Phytophthora species are those that possess an N-terminal signal peptide for secretion from the pathogen cell and an RXLR peptide motif (Arg-any amino acid-Leu-Arg) located near the N-terminus of the protein. The majority of RXLR effectors studied to date also possess an EER motif (Glu-Glu-Arg) downstream of the RXLR (Whisson et al., 2016). In P. infestans and P. sojae, the RXLR-EER region directs effector translocation from haustoria into host plant cells (Dou et al., 2008;Whisson et al., 2007). Haustoria are pathogen structures that project into plant cells from the invading intercellular hyphae and are bounded by the host cell plasma membrane. Phytophthora haustoria are the sites of secretion for diverse effector proteins delivered by at least two different secretion pathways (Kagda et al., 2020;Liu et al., 2014;Meng et al., 2015;Wang et al., 2017Wang et al., , 2018 (Figure 1a). These observations suggest that P. kernoviae has a hemibiotrophic interaction with N. benthamiana, where infected tissue remains alive for up to 48 hpi before infection leads to host cell death.
We generated 13 transgenic P. kernoviae lines to constitutively express GFP, of which one (PkGFP8) expressed GFP at a high level.
PkGFP8 growth in culture was indistinguishable from the nontransgenic progenitor and readily infected N. benthamiana, so was used for confocal microscopy analyses. Infections of N. benthamiana leaves with PkGFP8 revealed infection stages typical of Phytophthora species Hardham, 2007): host invasion and initial hyphal colonization by 24 hpi, followed by extensive hyphal spread by 48 hpi, before macroscopic symptoms were visible. Digit-like projections were observed on hyphae ramifying through the leaf tissue.
To determine if these projections were haustoria, we infected transgenic N. benthamiana plants stably expressing an mOrange-LTi6b fusion that labels the plant plasma membrane (Kurup et al., 2005).
This revealed that the fluorescently labelled plasma membrane surrounded the hyphal projections and, as such, they are likely to be biotrophic haustoria (Figure 1b).
PkGFP8 was also used to infect leaves of Rhododendron ponticum and Fagus sylvatica (beech). A fluorescently labelled plasma membrane marker was not available in these hosts, but imaging of presymptomatic leaf tissue revealed haustoria on the invading hyphae of P. kernoviae (Figure 1c,d). This suggests that the interaction with host tissue in N. benthamiana is similar to that in natural hosts Rhododendron and beech, and indicates that N. benthamiana is a useful model host to study P. kernoviae pathogenicity.

| The P. kernoviae transcriptome reveals effectors expressed during early infection stages
The P. kernoviae transcriptome was explored by RNA sequencing (RNA-Seq) of cultured mycelium, mixed asexual spores (sporangia and zoospores), and three different infection stages in N. benthamiana leaves (24, 48, and 72 hpi). Illumina sequencing generated 1,206 million raw reads. After removal of low-quality and contaminant reads, and trimming, a total of 1,102 million reads remained, yielding an average of 55 million reads per sample (Table S1). An average of 86% of the reads from mycelium and spore libraries were successfully mapped to the P. kernoviae reference genome sequence.
The proportion of reads from P. kernoviae in the infection samples ranged from an average of 0.3% (24 hpi) to 2.5% (72 hpi), which was supplemented by increased sequencing depth for infection samples (average 65 million reads).
All sequence reads were initially assembled into 12,559 transcripts (Table S2) (Table 1 and Figure S1c). The greatest number of DEGs was identified when comparing in vitro cultured mycelia with spores (Table 1) probably due in part to the spore samples comprising mixed transcriptomes of both sporangia and zoospores. We identified 60 differentially expressed candidate RXLR effectors and 164 carbohydrate active proteins (CAZymes) (Tables S5 and S6). Mapping DEGs onto pathways revealed the greatest number of genes were involved in core cellular processes such as protein processing, ribosome function, endocytosis, and RNA transport ( Figure 2). Metabolic pathways most enriched for DEGs were oxidative phosphorylation, inositol phosphate metabolism, glycine, serine and threonine metabolism, glycerophospholipid metabolism, and tyrosine metabolism.
The most abundant transcripts in all samples typically encompassed genes encoding ribosomal proteins, but with some F I G U R E 2 KEGG pathway enrichment for Phytophthorakernoviae differentially expressed genes (DEGs). The richness factor indicates the ratio of numbers of DEGs annotated for a specific pathway term, compared to the number of all genes annotated for that pathway. The size of the circles represents the number of DEGs annotated for that pathway or process Protein processing in endoplasmic reticulum -Ribosome -Endocytosis -RNA transport -Peroxisome -RNA degradation -Oxidative phosphorylation -mRNA surveillance pathway -Phosphatidylinositol signaling system -Phagosome -Inositol phosphate metabolism -Glycine, serine and threonine metabolism -Glycerophospholipid metabolism -Nucleotide excision repair -Proteasome -DNA replication -Tyrosine metabolism -Protein export -Base excision repair -One carbon pool by folate -Homologous recombination -Mismatch repair -Arachidonic acid metabolism -Sphingolipid metabolism -Ether lipid metabolism -Nicotinate and nicotinamide metabolism -Folate biosynthesis -Basal transcription factors -Phenylalanine metabolismalpha-Linolenic acid metabolism -Other glycan degradation -Ubiquinone and other terpenoid-quinone biosynthesis -Steroid biosynthesis -Glycosylphosphatidylinositol(GPI)-anchor biosynthesis -Riboflavin metabolism -Non-homologous end-joining -SNARE interactions in vesicular transport - additional genes, depending on the samples (Table S2). In cultured mycelium, two of the most strongly expressed genes encoded glycolytic enzymes glyceraldehyde-3-phosphate dehydrogenase (TCONS_00003986.p1) and fructose bisphosphate aldolase (TCONS_00007488.p1). In spore samples, the gene encoding CDC14 phosphatase (TCONS_00005250.p1), involved in sporangium formation in other Phytophthora species (Ah Fong & Judelson, 2003;Zhao et al., 2011), was the most highly expressed functionally annotated gene (Table S2). A homolog of the Phytophthora cell wall transglutaminase (TCONS_00009334.p1), an elicitor of plant defence responses (Brunner et al., 2002), was also highly expressed in spore samples, although this P. kernoviae homolog does not possess an intact PEP13 elicitor sequence. At 24 hpi, a gene encoding heat shock protein 70 (Hsp70) (TCONS_00007412.p1) was the most highly expressed, while at 72 hpi it was a gene encoding ATP synthase (TCONS_00008180.p1).
To approximate the expression for each CAZy protein family identified, the expression values were summed for each family (Table S7), as performed previously for P. infestans and P. parasitica (Ah Fong, Kim, et al., 2017;Ah Fong, Shrivastava, et al., 2017;Blackman et al., 2015). This revealed that CAZy family GH17 (β-1,3glucosidase) was among the most highly expressed in all P. kernoviae samples. In addition to GH17 expression, in cultured mycelium the most highly expressed CAZy families were GH30_1 and GH16, comprising glucosylceramidase and β-glucan synthesis-associated activities, respectively. In spores, the most highly expressed CAZy families were families GH6 and GH5_20, both comprising endoβ-1,4-glucanase (cellulase) activity. At both 24 and 48 hpi, the most highly expressed CAZy families were families AA2 and AA7 comprising haem-peroxidase and glucooligosaccharide oxidase (annotated as berberine-like) activities. At 72 hpi, families GH3 and GH1 were the most highly expressed, both comprising β-glucosidase activity.

| Secreted apoplastic proteins that interact with host cells to influence infection
From the P. kernoviae transcripts detected, 1,052 were predicted to encode secreted proteins without a predicted transmembrane domain (Table S2), including homologs of microbe-associated molecular patterns (MAMPs) or apoplastic effectors in other Phytophthora species (Table S8). Elicitins, ubiquitous secreted proteins from many p1), and one at 24 hpi (TCONS_00001592.p1).
p2 and TCONS_00006350.p1, peak transcript accumulation for these genes occurred during infection. TCONS_00005245.p1 exhibited strong up-regulation and high expression during infection.
Expression of two candidate cysteine protease inhibitors was also detected in P. kernoviae, but neither exhibited up-regulation during infection (Table S8).
Secreted carbonic anhydrases and berberine proteins have been proposed as effectors in P. infestans and other Phytophthora species (Hosseini et al., 2015;Raffaele et al., 2010). Transcripts for three candidate berberine proteins accumulated to high levels during infection (

| RXLR effectors
We identified 193 candidate RXLR-EER effectors from P. kernoviae genomic sequences (Table S10); candidates with/without predicted signal peptides were included to allow for mispredicted gene start The relative expression of 18 predicted RXLR effectors in P. kernoviae was validated by quantitative reverse transcription PCR. Transcript levels were compared to cultured mycelium, which was normalized to a value of 1. Error bars shown are standard error. Two additional independent biological replications are shown in Figure S3 codons or missing 5′ ends.  King et al., 2014;Qiao et al., 2013;Wang, McLellan, et al., 2019). In these assays, an increase in disease lesion diameter relative to an expressed GFP control suggests that the effector acts to facilitate infection. We selected nine P. kernoviae effectors that were observed to be up-regulated during infection, compared to spores, in either RNA-Seq or 89,96,97,98,111,123,132,136,and 138), and two that were not significantly differentially expressed (PkRXLR99 and 105) for ATTAs. PkRXLR89 caused a rapid cell death response in N. benthamiana and was not considered further. Two effectors, PkRXLR98 and 136, did not enhance infection.
The remaining nine effectors all significantly enhanced P. kernoviae infection when expressed in N. benthamiana (Figure 4). were produced for all RXLR effectors tested ( Figure S4).

| Localization of P. kernoviae RXLR effectors inside plant cells
Nuclear and nucleolar labelling observed for PkRXLR79, 99, and 97 was investigated further using known plant nuclear and nucleolar markers. Labelling the nucleoplasm with a coexpressed mRFP-Histone2B (H2B) fusion showed that the GFP-PkRXLR97 fusion uniformly labelled the nucleolus ( Figure S5a). In contrast, GFP-PkRXLR79 and 99 showed fluorescence in a ring surrounding the nucleolus in cells with low GFP-effector expression ( Figure S5a,b).
Effector PkRXLR99 appeared to disrupt the nucleus even with low levels of expression, forming protein aggregates that excluded mRFP-H2B fluorescence, but were separate to the nucleolus ( Figure S5b).
Localization of GFP-PkRXLR132 and 123 to the plasma membrane was confirmed by coexpression with the mOrange-LTi6b fusion in transgenic N. benthamiana plants ( Figure S6). Effector PkRXLR123 also exhibited labelling of the nucleus and nucleolus; the peptide sequence of this effector contains a predicted nuclear localization signal (Table S10).  (Avrova et al., 2008;Boevink et al., 2020;Evangelisti et al., 2017;Wang et al., 2011Wang et al., , 2013, P. kernoviae forms digit-like projections from intercellular hyphae during infection that are most likely to be haustoria because they are enveloped by the plant plasma membrane. P. kernoviae also appears to form haustoria in beech and

F I G U R E 5 Phytophthorakernoviae
Rhododendron leaves, indicating that this is a feature of its pathogenesis in natural hosts. In P. infestans, haustoria are major sites of protein secretion, and different classes of proteins involved in pathogenesis, such as CAZymes, protease inhibitors, elicitins, and RXLR effectors, are secreted from these structures (Kagda et al., 2020;Wang et al., 2017Wang et al., , 2018. Our RNA-Seq data provide the first insight into gene expression during infection for P. kernoviae, a tree pathogen more distantly related to other tree-infecting species for which transcriptome analyses have been carried out during plant infection (Blackman et al., 2015;Evangelisti et al., 2017;Hayden et al., 2014;Meyer et al., 2016).
As in other Phytophthora pathosystems (Evangelisti et al., 2017;Jupe et al., 2013), we observed up-regulation of genes encoding candidate pathogenicity factors and MAMPs such as Hmp1, elicitins, necrosis and ethylene-inducing proteins, glucanase and protease inhibitors, degradative enzymes (CAZymes, cutinases), RXLR effectors, and additional proposed effectors such as carbonic anhydrases. While we did identify expression of 12 candidate CRN proteins, none were predicted to possess signal peptides for secretion and thus are unlikely to be active effectors in P. kernoviae.
Invasion of plant tissues requires pathogens to breach physical barriers such as the cuticle, plant cell walls, and junctions between cells. Plant pathogens use enzymes such as cutinases, cellulases, and pectic enzymes to degrade these barriers (Kubicek et al., 2014;Nandi et al., 2018;Toth et al., 2003). Like other Phytophthora species (Brouwer et al., 2014;Ospina-Giraldo et al., 2010), the P. kernoviae genome encodes many CAZymes, which may have a variety of roles in pathogen biology. We detected 211 of these within our RNA-Seq experiment and found individual genes within the CAZyme families that exhibited strong up-regulation during plant infection. However, because many CAZyme families were represented by multiple genes in P. kernoviae, we summed the expression values from each family to approximate the activity of each CAZyme family in each sample type.
Pectin-degrading enzymes such as pectin/pectate lyase, pectinesterase, polygalacturonase, rhamnogalacturonase, and arabinan endo-1,5αl-arabinosidase were expressed at lower levels than cellulose or hemicellulose active CAZymes in P. kernoviae. A surprising finding was that, although P. kernoviae possesses and expresses multiple GH28 polygalacturonases that could degrade the homogalacturonan backbone, these enzymes were only expressed at low levels.
Our data suggest that rhamnogalacturonan may be degraded before homogalacturonan in the early stages of infection, in conjunction with enzymes that digest the side chains. In P. parasitica infection of lupin, the pectin backbone may be degraded first, followed by side chains (Blackman et al., 2015).
As a tree pathogen, P. kernoviae may deploy lignin-modifying enzymes such as laccases and lignin peroxidases to promote colonization of host tissues. While two laccases were found in the transcriptome, neither was expressed at detectable levels during infection. Lignin peroxidases fall within CAZy Auxiliary Activity family 2 (AA2). P. kernoviae expresses a single predicted extracellular AA2 peroxidase (TCONS_00004027.p1); another AA2 protein (TCONS_00002817.p3) is annotated as a mitochondrial protein.
While TCONS_00004027.p1 is annotated as a catalase peroxidase, the predicted inclusion in family AA2 suggests it may function as a lignin peroxidase and will require further biochemical characterization to verify the substrate.
These may be translocated inside host cells to modify target proteins and promote infection, as they are in more extensively studied species such as P. infestans (Wang et al., 2017(Wang et al., , 2018Whisson et al., 2016). In our RNA-Seq data, 69 putative P. kernoviae RXLR effectors were detectably expressed during infection of N. benthamiana.
Because RXLR effectors often function to inhibit pattern-triggered immunity (He et al., 2020), they are required from the earliest stages to facilitate infection. Consistent with this, we found that 48 of the  Haas et al. (2009) was detected at 12 hpi, which is within the biotrophic phase of infection (Yin et al., 2017).
We used transient expression of 12 selected effectors in plant tissue, followed by pathogen infection, to gain a broad overview of whether only the most highly expressed RXLR effector-coding genes led to increased pathogen colonization or if many more have the potential to contribute to pathogenicity. One RXLR effector resulted in cell death after infiltration, but nine of the remaining 11 effectors tested resulted in significantly increased lesion sizes. These were similar effects to those seen for these assays involving P. infestans effectors (e.g., Wang, McLellan, et al., 2019) in terms of the proportion that led to increased infection and the extent to which they increased infection.
Because RXLR effectors are known to be translocated into plant cells, the subcellular destinations of these have been determined by N-terminal fusion to GFP or other fluorescent proteins. RXLR effectors from P. infestans and the downy mildew Hyaloperonospora arabidopsidis have shown localization to most cellular organelles and structures (with the possible exception of the actin cytoskeleton) (Caillaud et al., 2012;Wang, McLellan, et al., 2019). The 11 P. kernoviae RXLR effectors tested for their contribution to pathogenicity (above) localized to the cytoplasm, nucleus, nucleolus, and plasma membrane. This information is a valuable resource for interpreting outcomes from future experiments to identify the host proteins targeted by these effectors.
PkRXLR89 caused cell death in infiltrated leaves. Cell death phenotypes have also been observed for RXLR effectors from P. parasitica, P. infestans, and P. agathidicida (Guo et al., 2020;Huang et al., 2019;Wang, McLellan, et al., 2019). These effectors are either toxic when overexpressed in plant cells or are recognized by resistance proteins or other immune receptors, as for the PpE4 effector from P. parasitica or PaRXLR24 from P. agathidicida (Guo et al., 2020;Huang et al., 2019). PkRXLR89 is highly expressed during the biotrophic stage of P. kernoviae infection. If PkRXLR89 triggers an immune response through recognition by the host plant, then it is possible that another effector suppresses this cell death response, as for grapevine downy mildew Plasmopara viticola (Yin et al., 2019) and P. agathidicida (Guo et al., 2020). To date, RXLR effectors are the only oomycete effectors known to be recognized by host resistance proteins (Anderson et al., 2015). That P. kernoviae expresses RXLR effectors suggests that it may be possible to identify tree germplasm with genetic resistance to this pathogen, limiting the impact of this disease in forestry and informing (re)planting strategies to control disease.

| P. kernoviae culture conditions
P. kernoviae (SCRP1055) was obtained from the Phytophthora culture collection at the James Hutton Institute, UK. Cultures were maintained on modified V8 juice agar (10% V8 juice, 1 g/L CaCO 3 , 2.5 mM KOH) and incubated in the dark at 20 °C.

| RNA sequencing
The P. kernoviae transcriptome was analysed using RNA-Seq of samples from cultured mycelium, spores, and three time points during infection of N. benthamiana. Spore samples were prepared by flooding 10-day-old V8 agar cultures of P. kernoviae with sterile distilled water, rubbing with a glass spreader rod, and gravity filtering through a 35 µm nylon filter. We found that P. kernoviae zoospores were released within the brief period needed to harvest sporangia and so a mixed spore sample (zoospores and sporangia) was prepared. The collected spores were centrifuged at 1,000 × g for 5 min, the supernatant discarded, then the recovered spores frozen in liquid nitrogen and stored at −70 °C until used for RNA isolation.

| Quantitative RT-PCR analysis
RNA extraction, removal of DNA, first-strand cDNA synthesis, and RT-qPCRs were performed as described by Wang et al. (2018); primers are listed in Table S11. All assays were biologically replicated in triplicate and each technically replicated in triplicate; "no template" controls were included. Data analysis was based on that described in Avrova et al. (2003). The P. kernoviae actinA gene (TCONS_00001238.p1; GenBank KAF4323451) was used as an endogenous control as it exhibited less than 2-fold variation in expression in RNA-Seq and is homologous to the ActA gene used as an endogenous control for P. infestans.

| Vector construction and A. tumefaciens transient assays
Candidate P. kernoviae RXLR effectors were cloned without their predicted signal peptides to retain them inside plant cells.
Sequences were PCR amplified from P. kernoviae genomic DNA using gene-specific primers that included Gateway recombination sites (Table S11), then introduced into pDONR201 using Gateway cloning (Invitrogen) to yield entry clones. Entry clones were recombined with destination vector pB7WGF2 (Karimi et al., 2005), containing the enhanced green fluorescent protein (eGFP) gene to yield N-terminal GFP-effector fusions, because the C-terminus of RXLR effectors contains the functional domain. Destination vectors containing GFP-effector sequences were electroporated into A. tumefaciens AGL1. Transformed A. tumefaciens strains were prepared for leaf infiltration as described in Wang, McLellan, et al. (2019).
ATTAs were carried out essentially as described by Kunjeti et al. (2016)  Lesion diameter was measured at 4 dpi. Four sites were inoculated on each leaf and nine leaves were used for each of three replicates.
A one-way analysis of variance (ANOVA) Student-Newman-Keuls test was performed to identify statistically significant differences compared to the control (Wang, McLellan, et al., 2019).

| Immunoblotting
Leaf disks (1 cm diameter) were harvested at 2 dpi after Agrobacterium infiltration with constructs expressing GFP-RXLR effector fusion proteins. Leaf disks were ground in liquid nitrogen and resuspended in 100 µl of GTEN buffer . Sample preparation was as described in Wang, McLellan, et al. (2019). Electrophoresis was performed as in Wang et al. (2018). Gel electroblotting, Ponceau staining, membrane blocking, and washing steps were carried out as described by McLellan et al. (2013). The αGFP primary antibody (Chromotek) was used at 1:2,000 dilution. Secondary antibody, anti-mouse IgG horseradish peroxidase (Chromotek), was diluted 1:8,000. Immunoblotting membrane was processed with ECL substrate (Thermo Scientific Pierce) using the manufacturer's protocol and imaged with Amersham Hyperfilm ECL, developed with an Xograph imaging system, compact X4 developer.

| Transformation of P. kernoviae
A previously published plasmid vector for expression of eGFP (pTor-eGFP) in Phytophthora was used for transformation of P. kernoviae (Wang et al., 2017). P. kernoviae was transformed using the method described by Judelson et al. (1991); the full method can be found at https://oomyc etewo rld.net/Proto plast %20tra nsfor mation.pdf.
Transformed lines of P. kernoviae were selected on rye agar containing 10 µg/ml geneticin and later maintained on rye agar with 20 µg/ ml geneticin.

| Confocal imaging
N. benthamiana leaf cells transiently expressing effector fusions were observed at 24-48 hpi on a Nikon A1R confocal microscope, with images obtained using water-dipping objectives. GFP was imaged using an excitation wavelength of 488 nm and the emissions were collected between 500 and 530 nm. mRFP and mOrange were excited with 561 nm light and emissions collected between 570 and 620 nm. The pinhole was set at 1.2 airy units for the longest wavelength fluorophore of any combination and coexpressed fluorophores were imaged sequentially to mimimize cross-talk. Typically, cells expressing a low level of fluorescence were imaged to minimize overexpression artefacts. Each effector fusion was examined on multiple occasions on independent plants and in cells widely distributed across each infiltration zone to gauge the most typical fusion protein localization.
N. benthamiana, Rhododendron, and beech leaves infected with transgenic P. kernoviae expressing GFP were imaged using the confocal microscopy settings described above.
All images were processed with Nikon NIS Elements confocal software v. 4.30 and figures were compiled with Adobe Photoshop.

ACK N OWLED G EM ENTS
The authors acknowledge financial support from the Biotechnology

DATA AVA I L A B I L I T Y S TAT E M E N T
Illumina sequence data were deposited in the NCBI Sequence Read