Apoplastic effector candidates of a foliar forest pathogen trigger cell death in host and non-host plants

Forests are under threat from pests, pathogens, and changing climate. A major forest pathogen worldwide is the hemibiotroph Dothistroma septosporum, which causes dothistroma needle blight (DNB) of pines. While D. septosporum uses effector proteins to facilitate host infection, it is currently unclear whether any of these effectors are recognised by immune receptors to activate the host immune system. Such information is needed to identify and select disease resistance against D. septosporum in pines. We predicted and investigated apoplastic D. septosporum candidate effectors (DsCEs) using bioinformatics and plant-based experiments. We discovered DsCEs that trigger cell death in the angiosperm Nicotiana spp., indicative of a hypersensitive defence response and suggesting their recognition by immune receptors in non-host plants. In a first for foliar forest pathogens, we developed a novel protein infiltration method to show that tissue-cultured pine shoots can respond with a cell death response to a DsCE, as well as to a reference cell death-inducing protein. The conservation of responses across plant taxa suggests that knowledge of pathogen–angiosperm interactions may also be relevant to pathogen–gymnosperm interactions. These results contribute to our understanding of forest pathogens and may ultimately provide clues to disease immunity in both commercial and natural forests.


Results
Dothistroma septosporum candidate effectors have sequence and structural similarity to fungal virulence factors. The D. septosporum genome has approximately 12,580 predicted genes, 397 of which encode putatively secreted proteins that are expressed during infection of P. radiata seedlings 20,27,38 . From this resource, a short list of 30 predicted apoplastic D. septosporum candidate effector proteins (DsCEs) was identified for further analysis (Fig. 1, Supplementary Table S1).
We first determined if the DsCE sequences were conserved in other fungi with available genomes. All but five of the shortlisted DsCEs (i.e. Ds19040, Ds24625, Ds72737, Ds73520 and Ds74875) had homologues in at least 10 different fungal species, while nine were predicted to have conserved functional domains (Supplementary Table S2). Furthermore, almost all of the DsCEs had a homologue in at least one of the nine pine needleassociated fungi investigated (Supplementary Table S3), with more than half of these possessing a homologue in both a pathogenic species and the pine endophyte species Lophodermium nitens. Only three DsCEs (Ds19040, Ds72737 and Ds73520) appeared to be exclusive to D. septosporum, with no apparent homologues present even in the queried pine needle-associated fungi. Ds52422 was unique in having no homologues in the pathogenic species, but one homologue in L. nitens (Supplementary Table S3).
Structural analysis was carried out because some effector protein families share structural similarity, rather than sequence similarity, with effectors from different fungi 28,29 . HHpred analysis inferred structural relationships for five of the DsCEs that had no sequence-based functional annotations. Of these five, Ds43416 and Ds70694 showed structural similarity to Alt a 1 allergen/PevD1-like proteins, Ds52422 to endoglucanase proteins, and both Ds68958 and Ds72870 to polyketide cyclase proteins (Supplementary Table S1, Fig. 1). Notably, the gene encoding Ds43416 (Alt a 1/PevD1-like) was one of the most highly expressed and up-regulated of the DsCEs, with a remarkably high level of expression at the late infection stage (Fig. 1, Supplementary Table S1). Among the other DsCEs assessed using HHpred, possible structural relationships were also identified with proteins that carry functional domains associated with fungal virulence, namely chitin binding (lectin), cerato-platanin (sub-group of endoglucanase-like proteins) and CAP family domains (Supplementary Table S1) 30 www.nature.com/scientificreports/ Taken together, the short list of 30 DsCE proteins ( Fig. 1) included 13 with sequence and/or predicted structural similarity to fungal proteins with described functions, and 17 with sequence similarity to yet undescribed proteins encoded in other fungal genomes (Fig. 1, Supplementary Tables S1 and S2). Twenty-five of the 30 DsCEs had at least four cysteine residues (Fig. 1), consistent with the need for disulphide bonds for stability in the hostile apoplast 22 . Further, an overview of the D. septosporum reference genome, which is assembled to chromosome level 20 , showed no particular clustering of the 30 DsCE genes on small chromosomes or near repetitive elements ( Supplementary Fig. S1, Supplementary Table S1).

Six candidate effectors of D. septosporum induced cell death in non-host plants. In order to
identify DsCEs that elicit cell death and are potentially recognised by non-host plant immune receptors, the 30 DsCE genes listed in Fig. 1 were transiently expressed in N. tabacum and N. benthamiana. In addition to DsEcp2-1, five other DsCEs induced cell death, indicative of a hypersensitive defence response (HR) (Fig. 2).
The DsCEs differed in their ability to consistently elicit a response in N. tabacum. Three (DsEcp2-1, Ds70057 and Ds131885) caused cell death in at least 90% of the infiltration spots within 6 days post-infiltration (dpi), while Ds70694 (66%) and Ds71487 (58%) were less consistent (Fig. 2). To gauge whether the culture density of infiltrated A. tumefaciens played a role in plant responses to the DsCEs, we trialled Agrobacterium tumefaciensmediated transient transformation assays (ATTAs) at six optical densities (ODs; 0.05-1.0) with the three most consistent cell death-inducing DsCEs. Whilst DsCEs Ds70057 and Ds131885 triggered strong cell death across all ODs, DsEcp2-1 showed less cell death at lower ODs ( Supplementary Fig. S2).
The six DsCEs also differed in their ability to elicit cell death in the non-host plant N. benthamiana. Ds74283 was unique in eliciting consistent and strong cell death in N. benthamiana, despite showing no response in N. tabacum. Only two of the five effectors that induced cell death in N. tabacum (Ds70694 and Ds70057) also consistently induced cell death in N. benthamiana, although it was often not uniform across the infiltrated zone (Fig. 2). Of the other DsCEs, Ds71487 and Ds131885 only induced weak cell death along the perimeter of infiltration zones in less than 50% of infiltrations into N. benthamiana, while DsEcp2-1 caused no cell death (Fig. 2).
To support the premise that the cell death responses were due to cell death elicitation by DsCEs in the apoplast, we cloned them for ATTAs without a secretion signal peptide. As expected, all of the non-secreted versions of these proteins failed to induce cell death (Fig. 2). The lack of cell death was not due to lack of DsCE production in the plant, but rather a lack of DsCE secretion to the apoplast, as each DsCE protein could be detected in infiltrated leaves by western blotting (Supplementary Fig. S3).
Dothistroma septosporum cell death inducers are conserved among fungi. We queried the sequences of each DsCE protein that elicited cell death in Nicotiana spp. to further investigate their similarities to other fungal proteins. Some of these DsCEs are broadly conserved across different fungal classes, and even outside the Ascomycetes (Table 1, Supplementary Table S4). Sequences similar to Ds70057 were the most numerous, with more than half of them found in Eurotiomycete species; at the other extreme, Ds70694 only had homologues in 14 Dothideomycete species. However, both Ds70057 and Ds70694 have multiple paralogues in some species, as well as in the D. septosporum genome, suggesting they are each members of a multi-gene family.
Among the cell-death elicitors, two were notable for their similarity to cell-death inducing proteins that have been described as PAMPs in other fungi. Based on reciprocal BLAST analyses, Ds71487 appears to be an orthologue of RcCDI1 from the barley pathogen Rhynchosporium commune 33 , with a pairwise identity of 41% www.nature.com/scientificreports/ and four conserved cysteine residues between the two proteins. Ds131885 is orthologous to VmE02, a crosskingdom PAMP from the apple pathogen Valsa mali 34 , with 60% amino acid identity and ten conserved cysteine residues in both proteins.

Infiltration of a DsCE protein into pine shoots induces cell death. Screening candidate effectors for
their ability to elicit cell death in model plants is indicative of their potential functions, but screening in a pine host is critical for understanding their actual roles in disease. As a first step to achieving this, we developed a reliable, straight-forward methodology to deliver effector proteins into pine needles (Fig. 3). A positive control that elicits cell death when infiltrated into clonal shoots of P. radiata was required to confirm the efficacy of the new method. Consistent with previous studies in which the Botrytis cinerea BcSSP2 effector induced strong cell death in a range of plant species 35 , vacuum infiltration of purified BcSSP2 protein produced by P. pastoris elicited cell death in three different genotypes of pine shoots within three days, providing an effective positive control.
Importantly, infiltration of the negative controls (protein purification elution buffer or bovine serum albumin protein) did not cause any damage or visible stress responses, as expected, even with higher concentrations of BSA (Fig. 4). In contrast, Pinus radiata shoots infiltrated with purified protein of Ds70057 consistently showed cell death within seven days across all three P. radiata genotypes tested (S6, S11 and R4) (Fig. 4, Supplementary  Fig. S4), with these responses near-identical to the positive control BcSSP2.

Discussion
Dothistroma needle blight (DNB) remains a threat to both commercial and native pine forests worldwide, despite decades of research into this disease and its management 6,8 . Although screening with pathogen effectors based on genome data (effectoromics) has fast-tracked identification of resistant angiosperm crop varieties with corresponding immune receptors, it is not known whether a similar effectoromics-based approach could be applied to improve forest health given the longer life-cycles of forest trees, and the complexity of factors affecting disease resistance 36 . However, a recent transcriptomics study of Pinus contorta responses to infection with D. septosporum showed up-regulation of pine genes involved in Ca 2+ and MAPK defence signalling pathways as well as pathogen-specific defence responses 37 . Signatures of selection were identified in candidate immune receptors (R genes), suggesting adaptation to the pathogen and the importance of specific interactions between pathogen effectors and host immune receptors in pine defence to D. septosporum.
We built on a prior study 38 to identify and refine a small set of secreted proteins that are likely to play a role as effectors in the infection of P. radiata by D. septosporum. A major selection criterion for these proteins was high expression and/or up-regulation during the infection 27 . We mainly adhered to size and cysteine content thresholds (≤ 300 aa, ≥ 4 Cys) used in studies of angiosperm pathogens 39 and used the EffectorP 3.0 40 tool to help classify the DsCE candidates. The lack of clustering of the 30 DsCE genes on small chromosomes or near repetitive elements ( Supplementary Fig. S1, Supplementary Table S1) supports earlier suggestions that, in contrast to several other Dothideomycetes, 'pathogenesis-associated gene classes' in D. septosporum were not enriched in these often-hyper-evolving regions 41 .
Here we used an effectoromics screening approach in two non-hosts to explore the effector arsenal of D. septosporum. Of the 30 small secreted D. septosporum candidate effector (DsCE) proteins that were expressed by transient transformation in N. tabacum or N. benthamiana, six induced cell death. Four of these six proteins are very highly conserved across a broad range of fungal taxa, suggesting some commonality in angiosperm and gymnosperm effectors, as also seen in some Phytophthora pathogens 42 . Most effectoromics studies include www.nature.com/scientificreports/ screening on host as well as non-host plants but, to the best of our knowledge, there were no methodologies in place to enable effector protein screening on pines or indeed any other gymnosperms. Thus, we developed a novel method that involves infiltration of purified effector proteins into clonal pine shoots. We validated the system using protein of a highly expressed DsCE, Ds70057, that elicited cell death in N. tabacum, and found that it also triggered cell death across three independent clonal genotypes of P. radiata. Importantly, we showed that P. radiata needles can be screened for responses to potential effectors, suggesting that an effectoromics screen to identify host resistance or susceptibility genes may be possible even in gymnosperm trees. Functional predictions and screens for homology usually rely on comparisons of primary amino acid sequences using BLASTp. In our study, we also used HHpred to infer possible structural relationships between the 30 DsCEs and proteins with characterised tertiary structures. Notably, HHpred predictions largely supported the BLASTp results, but also uncovered potential functions for five additional DsCEs. One of these five, Ds70694, induced cell death in Nicotiana spp. and was predicted to have structural similarity to proteins with the major human allergen Alt a 1 fold 43 , with the top HHpred hit to the PAMP-like effector protein PevD1 from Verticillium dahliae 44 . Another protein with an Alt a1 fold, the Magnaporthe oryzae HR-inducing protein 1 (MoHrip1), is highly induced during the infection of its rice host and has been implicated in promoting disease 45 . Notably, it has been shown that, like Ds70694, both PevD1 and MoHrip1 can trigger cell death in Nicotiana species [44][45][46] . This is despite the fact that Nicotiana species are not hosts for D. septosporum, V. dahliae or M. oryzae, suggesting that all three proteins share a conserved function in promoting cell death and/or they possess a conserved epitope that is recognised as a PAMP by these plant species.
Another of the cell-death inducing DsCEs, Ds131885, is an orthologue of VmE02, a cross-kingdom PAMP from the necrotrophic apple pathogen Valsa mali 34 that also elicits cell death in Nicotiana spp. VmE02 mutants had reduced numbers of pycnidia, suggesting the protein may be involved in pathogen conidiation. A similar function for Ds131885 would concur with its upregulated expression at the mid and late stages of infection. Recently a receptor-like protein (RLP) in N. benthamiana was identified that mediates VmE02-triggered immunity to fungal and oomycete pathogens 47 .
Ds71487 is homologous to a Rhynchosporium commune protein, RcCDI1, which induced cell death in N. benthamiana, N. sylvestris, and two other Solanaceae species, but not in its host barley 33 . RcCDI1 homologues in three other pathogens also induced cell death in N. benthamiana, and the authors concluded that the protein is a PAMP 33 . Given this context, the induction of cell death in N. tabacum by Ds71487 is not surprising and this response might also be found in other plants. However, it is unclear why this protein only triggered a weak response in N. benthamiana.  (1, 2). The proteins were purified by immobilised metal affinity chromatography (IMAC) (3) then vacuum-infiltrated into clonal microshoots of Pinus radiata (4). Infiltrated shoots were returned to LPch media for up to 7 days after infiltration (5). 1% neutral red was infiltrated to assess the efficiency of vacuum infiltration (6) in the different pine genotypes Sus6 (S6), Sus11 (S11) and Res4 (R4). www.nature.com/scientificreports/ The D. septosporum effector protein DsEcp2-1 was already known to elicit cell death in N. tabacum from a previous study 25 . In that work, DsEcp2-1-deficient mutants showed increased virulence on P. radiata compared to an Ecp2-1-containing wild-type strain, suggesting a possible avirulence role for DsEcp2-1 in eliciting a defence response by the plant. This suggestion supports the concept that D. septosporum effectors could, through adaptation of the host, become a warning signal (avirulence effector) similar to those of biotrophic pathogens 48,49 .
The remaining DsCEs that caused cell death in non-host plants in our study were not affiliated with any known functions, but were highly conserved across a broad range of fungal taxa (Table 1). Effectors with conserved core functions shared among plant pathogens could be important in elicitation of plant resistance as they are likely to be essential for the microbes' survival. Mutation or loss of essential conserved pathogen effector genes to escape recognition by a specific plant host immune receptor would most likely lower pathogen viability. It is possible that highly conserved ' core' effectors may have been overrepresented in our results because we screened gymnosperm pathogen effectors in angiosperm non-host plants. However, our results strongly affirm an earlier study that suggests conservation of core effectors between pathogens of angiosperms and gymnosperms 42 . The matching cell-death responses of pine and Nicotiana sp. to infiltration with Ds70057 and the control B. cinerea SSP2 proteins in our study also suggest similarities in the physiological responses of these disparate plant groups. Similarly, Chen et al. 50 found that infiltration of the cerato-platanin protein (CPP) HaCPL2 from Heterobasidion annosum induced cell death in the host Pinus sylvestris when applied to roots in vitro as well as in N. tabacum leaves.
Effectors can sometimes have opposing roles depending on the host context. Whilst they are mainly exploited in effectoromics studies on the basis of a role in avirulence, in which their recognition by a specific plant immune receptor elicits defence, they can have important roles in virulence that may be indispensable for fitness, causing conflicting selection pressures. At the same time, the induction of cell death by the plant host as part of its defence mechanism is beneficial for necrotrophic pathogens, such as Parastagonospora nodorum that causes Septoria blotch of wheat 16 . The P. nodorum-wheat interaction is regarded as a model pathosystem for understanding the 'inverse gene-for-gene' relationship between effectors and their corresponding plant targets, which are termed susceptibility factors. For both biotrophic and 'inverse' necrotrophic interactions, germplasm can be screened www.nature.com/scientificreports/ with effector proteins for a resistant or susceptible reaction, respectively, and accordingly selected or deselected for breeding 17 . However, the situation with hemibiotrophs, such as that of D. septosporum, is more complex and requires accurate understanding of the disease. The DsCE Ds70057 is of interest due to its very high expression and up-regulation in the mid and late (necrotrophic) stages of disease caused by D. septosporum on P. radiata 27 and its ability to elicit cell death in three genotypes of P. radiata as well as in N. tabacum. We suggest it may be a necrotrophic effector triggering cell death that assists the pathogen by eliciting the destruction of needle tissue. This action would augment that of dothistromin which is required for disease lesion expansion, but not for initial elicitation of cell death 51 . Further studies with Ds70057, including construction and analysis of D. septosporum knockout mutants, will help to determine if Ds70057 is acting as a necrotrophic effector, in which case we would expect these pathogen mutants to have decreased virulence.
An aim of our work was to develop a method to screen putative effectors of conifer pathogens directly in their host using a protein infiltration approach, with the long-term goal of identifying corresponding immune receptors for the effectors, and thus assisting breeding for disease resistance. There were several challenges associated with protein delivery into pines. In an earlier study on P. sylvestris, roots took up purified protein from a root pathogen via filter paper strips, resulting in root cell death 50 . By submerging pine shoots into solutions containing purified candidate effector proteins, and applying a vacuum pressure, we were able to develop a robust and reliable infiltration process that did not harm the pine needle tissue. We identified a positive control that can be used as a reference in future studies and showed that a secreted protein of D. septosporum induced cell death symptoms in pine shoots as well as Nicotiana leaves.
This novel method of effector delivery to gymnosperm tissue would be well suited for medium-scale screening of multiple pine genotypes with a small number of effector proteins. If the effector-screening process was to be scaled up for high-throughput commercial use with a large number of effector proteins, the time used to grow and prepare fresh purified effector proteins in the P. pastoris system as used here could be a limiting factor. An alternative delivery method might be through Agrobacterium-based transformation of pines, in which the effector genes rather than the proteins are delivered to the host plant. Agrobacterium-based transformation of P. radiata and other Pinus spp. has been achieved 52,53 , but it is a slow, technically challenging process. Despite this, future efforts to establish reliable transient transformation of pines may be worthwhile.
This study has broad implications for our understanding of fungal pathogens of gymnosperm trees and how these species interact with each other at the molecular level. Importantly, the high level of conservation of effectors across fungal taxa, and conservation of plant responses across angiosperms and gymnosperms suggests that much of what the scientific community has learned about plant-pathogen interactions involving angiosperm systems may also be relevant to gymnosperms. A similar conclusion was reached by studying pine responses to D. septosporum infection 37 . There also appear to be common ' core' interactions that could be exploited to give broad host-range resistance. Another implication of our work relates to the pine screening method which could be adapted for use with other gymnosperm tree species and other plant hosts that are not amenable to the usual screening methods. Finally, this type of screening method has the potential, along with expression data and pathogen gene knockout studies, to identify necrotrophic effectors that could in turn identify susceptibility gene targets, leading to durable resistance.

Methods
Bacterial strains and plants. Escherichia coli DH5α was used for gene cloning and plasmid propagation.
A. tumefaciens GV3101 54 was used, in conjunction with N. tabacum Wisconsin 38 and N. benthamiana, for ATTA experiments. P. pastoris GS115 was used for heterologous protein expression in liquid culture. P. radiata clonal shoots without roots, grown from embryo cotyledon tissue under sterile conditions on LPch agar 55 , were used for protein infiltration experiments. P. radiata material used for development of the clonal shoots was obtained with permission from the New Zealand Radiata Pine Breeding Company (RPBC) for the purposes of this research. This study complies with institutional, national and international guidelines and legislation. DsCE identification and homology searches. Genes encoding conserved candidate secreted apoplastic effector proteins were identified in the D. septosporum genome 20 , following the bioinformatic pipeline of Hunziker et al. 38 in conjunction with transcriptomics data to determine which genes were highly expressed and/ or upregulated during infection of P. radiata 27 . We also used EffectorP 3.0 40 and comparisons with the characterised apoplastic secretome of C. fulvum 22 to generate a short list of 30 predicted apoplastic D. septosporum candidate effector proteins (DsCEs).
BLASTP was used to identify the top homologues of the 30 DsCE proteins in the National Center for Biotechnology Information (NCBI) and Joint Genome Institute (JGI) databases, with an Expected (E)-value threshold of 1E-02. This included a specific query against the foliar pine pathogens  Table S3). BLASTp was also used to determine the level of conservation for each of the cell death-eliciting DsCEs across fungal species in JGI, with an E-value threshold of 1E−05. Here, only one isolate per species was examined, with all paralogues from each species retained. Conserved functional domains in the DsCE proteins were identified using the Conserved Domain Database (CDD) in NCBI.
HHpred 56 was used to infer possible structural relationships between the 30 DsCEs and proteins of characterised tertiary structure present in the RCSB (Research Collaboratory for Structural Bioinformatics) protein www.nature.com/scientificreports/ databank. Here, only the top hit was retained. Structural relationships were deemed to be significant if they had a probability score of 95%, an E-value score < 1E−03, and an overall score of > 50. Despite not meeting these criteria, Ds43416 and Ds70694 were also deemed to have significant structural relationships with homologous proteins of C. fulvum through additional support from previous structural predictions 22 .

Generation of DsCE expression vectors for Agrobacterium tumefaciens-mediated transient transformation assays (ATTA).
DsCE ATTA expression vectors were generated using a method described previously 25 . Briefly, DsCE genes, without their native signal peptide sequence, were PCR-amplified from D. septosporum NZE10 genomic DNA or cDNA (derived from in vitro or in planta-grown fungus 27 ) using custom primers (Supplementary Table S5; Integrated DNA Technologies) with BsaI recognition sites and overhangs required for modular assembly using the Golden Gate approach 57 . PCR amplicons were then directly used as entry modules for cloning into the ATTA expression vector pICH86988, which contains a CaMV 35S promoter and octopine synthase terminator flanking BsaI insertion sites 58 , using Golden Gate assembly, along with an entry module encoding an N-terminal PR1α signal peptide for secretion into the apoplast and an N-terminal 3 X FLAG tag for detection by western blotting. Cell death-eliciting DsCEs were also assembled into expression vectors as above without a PR1α signal peptide. DsCE ATTA expression vectors were then transformed into E. coli, and inserts confirmed by sequencing. Finally, the DsCE ATTA expression vectors, along with the INF1 ATTA expression vector (an extracellular elicitin from Phytophthora infestans 59 ), were transformed into A. tumefaciens by electroporation as described previously 25 . To verify the presence of DsCE proteins in agro-infiltrated N. benthamiana leaves, western blotting was carried out as described previously 42 . Here, total protein extracts from the leaves were separated by SDS-PAGE (12% bis-tris-acrylamide) and transferred onto PVDF membranes. Mouse anti-FLAG(R) antibodies (Sigma Aldrich) and chicken anti-mouse antibodies conjugated with horseradish peroxidase (Santa Cruz Biotechnology, Dallas, TX, USA) were used for detection by an Azure c600 imager (Azure Biosystems, Dublin, CA, USA) after incubation with SuperSignal™ West Dura Extended Duration Signal Substrate (Thermo Fisher Scientific, Waltham, MA, USA).

Development of a protein delivery method for pine tissue. Challenges involved in developing this
method included producing DsCE proteins of interest in sufficient purity and quantity, developing an effective protein delivery method, and identifying a positive control. Trials involved in the development of methods to address these challenges were discussed in detail by Hunziker 60 , and are briefly summarised here.
To produce the DsCE proteins of interest, we first trialled existing ATTA expression constructs and whole N. benthamiana leaves as expression systems for protein production. However, collection of apoplastic wash fluid (AWF), containing secreted DsCEs from infiltrated leaves, did not yield enough protein for replicated studies with pines. We then used heterologous protein expression and secretion in Pichia pastoris, which can be more easily up-scaled. Culture filtrates from P. pastoris transformed with an empty vector (negative control) elicited cell death in some P. radiata genotypes. Thus, to avoid this problematic background response, we purified histidine-tagged DsCEs from culture filtrates of P. pastoris using immobilised metal ion affinity (IMAC).
To deliver purified proteins into pine needles, we trialled several methods. Pine needles were not amenable to syringe infiltration, so vacuum infiltration, with pine tissue immersed in a DsCE protein solution, was used. After infiltration, detached needles, or groups of needles (fascicles), from pine seedlings quickly deteriorated, even when placed in moist conditions, so we chose different plant material for protein delivery assays. Here, clonal shoots of P. radiata produced by tissue culture without roots and maintained on agar medium, were employed 55 . The shoots were vacuum-infiltrated, then returned to the agar medium for 7-10 days. Vacuum infiltration of whole shoots with a neutral red dye solution suggested good uptake efficiency.
To identify a positive control, an effector protein from the broad host-range pathogen Botrytis cinerea, BcSSP2, was trialled as it was previously shown to induce strong cell death in N. benthamiana as well as in other plants 35 . Heterologous candidate effector protein expression in P. pastoris and protein purification. One of the DsCE proteins, Ds70057, was investigated for its capacity to trigger cell death in pine. www.nature.com/scientificreports/ sion vector pPic9-His 6 (Invitrogen, Carlsberg, CA, USA) behind the α-factor signal peptide sequence using SmaI/EcoRI restriction enzymes and T4 DNA ligase (New England Biolabs, Beverly, MA, USA). Here, the PCR 5′ primer was designed to incorporate a flag-tag for subsequent detection by western blotting. P. pastoris expression vectors were subsequently transformed into E. coli, and their sequence confirmed, as described above. Finally, the expression cassettes were linearised with SacI or SalI restriction enzymes (New England Biolabs) and transformed into P. pastoris according to Kombrink 61 . Production of DsCE proteins in liquid culture using P. pastoris was performed according to Weidner et al. 62 . Expression of candidate proteins was induced by incubation of P. pastoris in 200 mL of BMMY (Buffered Methanol-complex Medium) 62 for 72 h, with the successive addition of methanol every 24 h to increase the final concentration by 0.5% (v/v). Following incubation, P. pastoris cells were collected by centrifugation at 4500g for 30 min. The supernatant, containing the secreted protein of interest, was sterilised through a 0.22 µm filter (ReliaPrep, Ahlstrom, Helsinki, Finland) and adjusted to pH 8 through the addition of NaOH.
Secreted candidate proteins were purified by immobilised metal ion affinity (IMAC) using Ni Sepharose 6 Fast Flow (GE Healthcare, Chicago, IL, EUA) according to the manufacturer's protocol. Before loading, the column (Glass Econo-Column, BioRad, Hercules, CA, EUA) was packed with 5 mL of resin and was equilibrated by washing with binding buffer (20 mM sodium phosphate, 0.5 M NaCl, pH 7.4). The culture filtrate was added to the column and passed through at 1 mL/min. The protein was eluted with an elution buffer (20 mM sodium phosphate, 0.5 M NaCl and 500 mM imidazole, pH 7.4). The elution fractions were mixed, and elution buffer was added to a final volume of 50 mL to obtain enough volume to vacuum-infiltrate pine shoots. A western blot was performed, as above, to determine the presence of the proteins in P. pastoris culture filtrate and after purification.
Pine infiltration with purified proteins. An overview of the pine infiltration method is shown in Fig. 3.
Clonal rootless pine shoots were produced by adventitious shoot production from cotyledons 55 . The shoots were kept in LPch agar 55 in glass jars, with each jar containing between six and eight shoots and three jars used for each treatment. For our experiment, two susceptible (S6 and S11) and one resistant (R4) genotype (based on field data) were used. The shoots were completely submerged in approx. 50 mL of each purified DsCE or BcSSP2 protein solution, or in elution buffer or bovine serum albumin (BSA) (Thermo Fisher Scientific) as negative controls 63 . Samples were exposed to vacuum in a glass chamber for 5 min, prior to gentle release of the vacuum to allow infiltration of the solution. Shoots were then rinsed in sterile MQ water, briefly air-dried, then placed back in the agar medium in the glass jar. The shoots were maintained in a 22 °C room with natural light. Photographs were taken 7 days after infiltration (dai). Fig. 1 was generated using heatmap.2 function of the gplots R package 64 . The photographs in Figs. 2 and 4 were taken using a Nikon D7000 camera using standard automatic exposure settings under a custom-made lighting rig, then cropped as required.

Data availability
The datasets analysed during the current study are available in the Joint Genome Institute and NCBI databases repository [https:// genom es. jgi. doe. gov and https:// www. ncbi. nlm. nih. gov/ genba nk]. All data generated during this study are included in this published article and its Supplementary Information files.