Effector mining from the Erysiphe pisi haustorial transcriptome identifies novel candidates involved in pea powdery mildew pathogenesis.

Pea powdery mildew (PM) is an important fungal disease caused by an obligate biotroph, Erysiphe pisi (Ep), which significantly impacts pea production worldwide. The phytopathogen secretes a plethora of effectors, primarily through specialized infection structures termed haustoria, to establish a dynamic relationship with its host. To identify Ep effector candidates, a cDNA library of enriched haustoria from Ep-infected pea leaves was sequenced. The Ep transcriptome encodes 622 Ep candidate secreted proteins (CSPs), of which 167 were predicted to be candidate secreted effector proteins (CSEPs). Phylogenetic analysis indicates that Ep CSEPs are highly diverse, but, unlike cereal PM CSEPs, exhibit extensive sequence similarity with effectors from other PMs. Quantitative real-time PCR of a subset of EpCSEP/CSPs revealed that the majority are preferentially expressed in haustoria and exhibit infection stage-specific expression patterns. The functional roles of EpCSEP001, EpCSEP009 and EpCSP083 were probed by host-induced gene silencing (HIGS) via a double-stranded (ds) RNA-mediated RNAi approach. Foliar application of individual EpCSEP/CSP dsRNAs resulted in a marked reduction in PM disease symptoms. These findings were consistent with microscopic and molecular studies, suggesting that these Ep CSEP/CSPs play important roles in pea PM pathogenesis. Homology modelling revealed that EpCSEP001 and EpCSEP009 are analogous to fungal ribonucleases and belong to the RALPH family of effectors. This is the first study to identify and functionally validate candidate effectors from the agriculturally relevant pea PM, and highlights the utility of transcriptomics and HIGS to elucidate the key proteins associated with Ep pathogenesis.

Introduction

Fungal phytopathogens have immense economic repercussions as they cause significant yield loss in field crops (Dean et al., 2012). Powdery mildews (PMs) are filamentous ascomycetes fungi (Order Erysiphales) that infect more than 9000 dicot and 650 monocot species, including agriculturally important crops such as wheat, barley, grape, tomato and pea (Ahmed et al., 2015; Glawe, 2008). PM fungi are obligate biotrophs that depend entirely on living host plants for their nutrition and survival (Micali et al., 2008). They deliver an arsenal of virulence proteins termed effectors into plant cells, which primarily interfere with host metabolism and suppress immune signalling to facilitate successful host colonization (Thordal‐Christensen et al., 2018). Within plant cells, effector proteins act on diverse host molecules in different cellular compartments to promote infection (Khan et al., 2018). Their functions include minimizing pathogen recognition by the host immune surveillance system, inhibiting host defence‐related enzyme activities and/or signalling processes, and interfering with modifications of interacting host proteins (Uhse and Djamei, 2018). Indeed, PM effectors have been shown to target well‐known host defence‐related proteins (Pennington et al., 2016; Weßling et al., 2015; Zhang et al., 2012), and suppress defence‐related processes such as pathogen‐induced hydrogen peroxide production and programmed cell death (Martínez‐Cruz et al., 2018; Pliego et al., 2013). Recent studies have also identified several PM avirulence (Avr) effectors that, when recognized by cognate host resistance proteins, activate a robust form of plant immunity known as effector‐triggered immunity (ETI) (Bourras et al., 2015; Jones and Dangl, 2006; Lu et al., 2016; Praz et al., 2017). Therefore, the interplay between fungal effectors and their host targets can influence the outcome of plant–pathogen interactions.

To date, few PM effector candidates have been identified, and the mechanisms by which they manipulate host processes is poorly understood. Candidate secreted effector proteins (CSEPs) have been predicted from the genomes, transcriptomes and/or proteomes of PMs that infect barley [Blumeria graminis f. sp. hordei (Bgh)], wheat [Blumeria graminis f. sp. tritici (Bgt)], Arabidopsis [Golovinomyces orontii (Gor)], grape [Erysiphe necator (En)], cucurbits [Podosphaera xanthii (Pxa)] and Eucalyptus [P. pannosa (Ppa)] (Fonseca et al., 2019; Frantzeskakis et al., 2018; Jones et al., 2014; Muller et al., 2019; Pedersen et al., 2012; Spanu et al., 2010; Vela‐Corcía et al., 2016; Weßling et al., 2012; Wicker et al., 2013). Among these, the cereal PMs (Bgh and Bgt) harbour the largest repertoire of effectors (Bourras et al., 2018; Frantzeskakis et al., 2018; Menardo et al., 2017; Muller et al., 2019; Pedersen et al., 2012; Wicker et al., 2013). Analysis of Bgh CSEPs revealed that PM effectors are small proteins that exhibit high sequence diversity and generally lack homology to known proteins (Spanu et al., 2010). In addition, many Bgh CSEPs possess a conserved N‐terminal Y/F/W‐x‐C motif (Godfrey et al., 2010), the biological significance of which is hitherto unknown.

Functional characterization of PM CSEPs has been hindered by the unavailability of standard mutagenesis and genetic transformation protocols. However, a few Bgh and Pxa CSEPs have been functionally validated through host‐induced gene silencing (HIGS), an RNA interference (RNAi)‐based gene silencing method in which hairpin constructs targeting effector transcripts are transiently expressed in haustorial‐infected host cells via Agrobacterium‐mediated or biolistic transformation methods (Ahmed et al., 2015, 2016; Martínez‐Cruz et al., 2018; Nowara et al., 2010; Pliego et al., 2013; Zhang et al., 2012). In general, CSEP silencing resulted in reduced fungal penetration and haustorium formation rates, indicating that these effectors act as virulence factors to promote infection. Recently, HIGS via spray application of long double‐stranded (ds) RNAs has proven to be a viable and faster alternative to dsRNA delivery by transgenic expression (Koch et al., 2016, 2018; McLoughlin et al., 2018). For example, spraying of long dsRNAs targeting three Fusarium graminearum genes involved in the ergosterol biosynthesis pathway reduced fungal transcript levels and pathogen growth on barley leaves (Koch et al., 2016). Similarly, foliar application of dsRNAs targeting key Sclerotinia sclerotiorum pathogenesis‐related genes significantly reduced the development of disease lesions on Brassica napus (McLoughlin et al., 2018). However, foliar application of dsRNAs as a method for RNAi‐based gene silencing has not yet been tested in PMs.

Erysiphe pisi (Ep) is the main causal agent of PM on pea and is responsible for the withering of foliage, poor pod quality and c. 25–80% yield loss worldwide (Nisar et al., 2006; Warkentin et al., 1996). Despite the economic importance of Ep, effectors have not been identified and the molecular mechanisms underlying pathogenesis are unknown. In fact, the lack of transcriptomic data, and the fragmented and unannotated nature of the Ep draft genome (Spanu et al., 2010) have hindered effector prediction from Ep. Effector synthesis and secretion in biotrophic fungi are thought to occur mainly through haustoria (Chaudhari et al., 2014), which are also the sites of nutrient uptake. Therefore, the haustorial transcriptome forms a valuable resource from which candidate secreted proteins (CSPs) and CSEPs can be effectively mined. Here, we present the Ep haustorial transcriptome, from which 622 Ep CSPs comprising 167 Ep CSEPs were predicted. We determined the expression patterns of a subset of Ep CSEP/CSPs and demonstrated the role of three candidates in PM pathogenesis by using a dsRNA‐mediated HIGS approach. Further, we studied their subcellular localization, predicted structures and putative functions, and based on the results discuss how these secreted proteins may contribute to Ep pathogenicity.

Results

Sequencing and assembly of the enriched Ep haustorial transcriptome

Ep haustoria were enriched from heavily infected pea leaves [6 days post‐inoculation (dpi)] using a protocol previously described for Gor (Micali et al., 2011). Assessment of enriched Ep haustorial fractions by optical microscopy revealed that the haustoria were intact (Fig. S1). To determine the extent of plant contamination in these fractions, we performed quantitative reverse transcription PCR (qRT‐PCR) of a pea housekeeping gene, GAPDH, using cDNA synthesized from enriched haustorial RNA. We were unable to detect PsGAPDH amplification (not shown), indicating that the enriched haustorial sample had minimal plant contaminants. The haustorial RNA sample was then subjected to RNA sequencing yielding 129 031 246 raw reads (SRA accession: SRR7066906). The workflow for transcript assembly/annotation and sequencing/assembly statistics are provided in Fig. S2 and Tables S1 and 1. Briefly, 96.8% of the reads qualified the Q20 score, and c. 34% of the high‐quality (HQ) reads that mapped to the draft genome assembly of Ep (NCBI accession PRJEA50315) were assembled using a reference‐based approach. We reasoned that the low mapping rate of the HQ reads was probably due to the partial and fragmented nature of the Ep reference genome. We therefore used a de novo assembly method to extract additional transcripts from the remaining unmapped reads after filtering out contaminating plant transcripts. This resulted in a set of 36 593 non‐redundant transcripts: 21 189 from the reference‐based and 15 404 from the de novo assembly (Tables 1 and S2). Evaluation of the completeness of the assembled transcriptome [BUSCO (Benchmarking Universal Single‐Copy Orthologs) notation: C:45.5% [S:31.9%, D:13.6%], F:38.7%, M:15.8%, n:1315] revealed that the combined assembly was 11% better than that of the reference‐based assembly alone, and among the top five fungal transcriptome assemblies (Fig. S3; Simão et al., 2015). We speculated that some of the missing orthologues may belong to the set of 99 missing ascomycete core genes (MACGs), which are absent in PM genomes but present in other ascomycete fungi, including the yeast Saccharomyces cerevisiae (Spanu et al., 2010). Indeed, we found that, except for one heme biosynthesis gene (HEM4), the remaining 98 MACGs were not expressed in the Ep haustorial transcriptome (Table S3). To identify protein‐coding regions within transcript sequences, open reading frames (ORFs) were predicted using TransDecoder with a default size cut‐off ≥100 amino acids. A total of 16 078 unique proteins were predicted, of which 7319 full‐length proteins (beginning with methionine) were analysed further (Table S4).

Table 1

Summary of the assembly statistics for Erysiphe pisi haustorial transcriptome data using reference‐based and de novo methods

	Reference‐based assembly	De novo assembly
Total transcripts	21 189	15 404
GC %	40.20	40.78
N50 value	1650	597
Median transcript length (bp)	934	364
Average transcript length (bp)	1238.51	491.07
Total assembled bases (bp)	26 242 920	7 564 373
Longest transcript (bp)	11242	3615
Shortest transcript (bp)	200	200

Prediction of Ep secretome and CSEPs

Fungal effectors are predominately secreted via the classical endoplasmic reticulum/Golgi‐dependent pathway, which involves an N‐terminal signal peptide (SP) (Sperschneider et al., 2015). However, the existence of non‐classical pathways of effector secretion that do not require an SP has also been reported (Liu et al., 2014). Therefore, we used an in silico pipeline to identify all canonical (with SP) and non‐canonical secreted proteins in the set of 7319 Ep full‐length proteins (Fig. S2). A total of 622 CSPs were predicted: 308 canonical and 314 non‐canonical, collectively defining the Ep secretome (Fig. S2 and Table S5). To distinguish effectors from secreted proteins, we used EffectorP v. 2.0, a machine learning‐based prediction tool that predicts fungal effectors in secretomes (Sperschneider et al., 2016). Based on EffectorP, c. 33% (103) of canonically secreted proteins and 21% (64) of non‐canonically secreted proteins were predicted as effectors, yielding a total of 167 Ep CSEPs (Table S6). Analysis of their protein sequences revealed that none of the Ep CSEPs have homologues outside the PMs and c. 65% do not contain any PFAM domains, both characteristic features of PM effectors (Spanu et al., 2010). Furthermore, a number of conserved motifs reported from different oomycete and fungal effectors (Sonah et al., 2016) were identified in Ep CSEPs (Table S7), among which only the Y/F/W‐x‐C motif was found to be significantly enriched (P = 0.0006). However, unlike in Bgh, where c. 97% of CSEPs harbour the Y/F/W‐x‐C motif (Godfrey et al., 2010), this motif was present in merely 27% of Ep CSEPs. Within the canonical Ep CSEPs, in the first position of the motif there was an abundance of Y (tyrosine, 18) > F (phenylalanine, 9) > W (tryptophan, 2), whereas within the non‐canonical Ep CSEPs, the residues were F (10) > Y (3) > W (3) (Fig. S4).

Ep CSEPs are highly diverse but conserved across PMs

Phylogenetic analysis indicates that the 167 Ep CSEPs group into 13 distinct clusters (Fig. 1). The overall bootstrap support of the neighbour‐joining tree was low and can be attributed to the low sequence relatedness, a characteristic of the PM CSEPs (Spanu et al., 2010). Markov cluster algorithm (MCL) analysis identified only four protein families with three or more members per family. Of these, the ‘ribonuclease/ribotoxin’ family was the largest, comprising nine members. BLAST searches in the En, Gor and Bgh proteomes revealed that the vast majority of Ep CSEPs (73%) show sequence similarity with proteins from one or more PM species: 101 with En, 82 with Gor and 71 with Bgh (Fig. 1); only 45 CSEPs were unique to Ep. Notably, 40 Ep CSEPs were found to harbour sequence‐related proteins in all three PMs, likely representing a set of conserved effectors that are important for virulence.

Figure 1

CIRCOS plot summarizing select features of EpCSEPs. From the perimeter to the centre: Erysiphepisi candidate secreted effector proteins (CSEPs) identifier number; Color‐coded squares represent Markov cluster algorithm (MCL) family clustering (cyan, ribonuclease/ribotoxin; red, MULE transposase domain; orange, aspartic peptidase A1; green, alpha/beta hydrolase fold; pink, histidine phosphatase; blue, peptidase S8; grey, unknown); Color‐coded circles indicate the presence of Y/F/W‐x‐C conserved motif (blue, YxC; pink, FxC; green, WxC); Color‐coded squares indicate the presence of homologous sequences in other powdery mildew species (red, En; green, Golovinomyces orontii; blue, Blumeria graminis f. sp. horde); Colored bars represent the FPKM expression heat map in log2 scale (yellow to red = low to high expression); At the centre is a radiating neighbour‐joining tree clustering the 167 EpCSEPs into 13 clades shown in distinct colours. Bootstrap support values greater than 50% are shown next to the branches.

Ep CSEPs are preferentially expressed in haustoria and exhibit infection stage‐specific expression patterns

To determine their expression kinetics, a subset of 15 Ep CSEPs and two Ep CSPs with high transcript FPKM values (>30) and/or similarities to proteins predicted to be involved in host–pathogen interactions were selected. This set includes eight ribonuclease/ribotoxin domain‐containing proteins (EpCSEP001, 002, 009, 023, 027, 039, 045 and 068), two Egh16H‐like virulence factors (EpCSEP087 and EpCSP083), two glycoside hydrolase/chitin binding proteins (EpCSEP007 and EpCSP249), a heat‐shock protein 70 (EpCSEP019), a cutinase (EpCSEP028) and three hypothetical proteins (EpCSEP012, 018 and 035). Notably, all candidates could be PCR amplified only from Ep genomic DNA and not from pea genomic DNA, confirming that they are Ep‐specific (Fig. S5).

Since haustoria are considered to be the primary sites of effector synthesis and secretion (Chaudhari et al., 2014), we determined whether the selected Ep CSEP/CSPs are predominately expressed in haustoria. For this, we estimated their relative transcript abundance in haustorial versus epiphytic mycelial samples via qRT‐PCR. We observed that all tested Ep CSEP/CSPs, except EpCSEP027, were over‐expressed by ≥2‐fold in the haustorial samples (Fig. 2). Subsequently, we determined the expression of these Ep CSEP/CSPs during the course of Ep infection. The vast majority of the analysed Ep CSEPs exhibited ≥2‐fold induction in expression at different infection time points compared to 0 h post‐inoculation (hpi) (Fig. 3). Three of the eight ribonuclease‐like Ep CSEPs (EpCSEP001, 009 and 068) were induced during the penetration and primary haustorium formation stages (12–24 hpi) and again during the colony expansion and asexual reproduction stages (72–120 hpi). Induced expression of four other ribonuclease‐like Ep CSEPs (EpCSEP002, 023, 027 and 039) was restricted to the penetration and primary haustorium formation stages (12–24 hpi). Expression of EpCSEP012 (hypothetical protein) was induced exclusively at the appressorial stage (6 hpi) but remained unchanged or repressed at other stages of infection. The Egh16H‐like virulence factors EpCSEP087 and EpCSP083 were highly induced between 6 and 24 hpi, corresponding to the appressorial, penetration and primary haustorium formation stages. EpCSP249 (chitin‐binding protein) was induced maximally at 48 hpi, corresponding with the colony expansion stage. Expression of EpCSEP007, 018, 019, 028, 035 and 045 remained unchanged or repressed throughout the infection time‐course (Fig. S6).

Figure 2

Relative expression of select Erysiphe pisi candidate secreted effector protein (CSEP) and candidate secreted protein (CSP) genes in haustoria versus epiphytic mycelia. qRT‐PCR analysis was performed on total RNA isolated from epiphytic mycelia and haustoria‐containing pea leaves. Data represent mean ± SD of EpCSEP/EpCSP expression values normalized to that of the reference gene Eptub2 from two biological replicate experiments.

Figure 3

Temporal expression profiles of select Erysiphe pisi candidate secreted effector protein (CSEP) and candidate secreted protein (CSP) genes determined by qRT‐PCR. x‐axes represent hours post‐inoculation (hpi). Data represent mean ± SD of EpCSEP/EpCSP expression values normalized to that of the reference gene Eptub2 at different time points of infection (0–120 hpi) on pea leaves from two biological replicate experiments.

Foliar infiltration of Ep CSEP/CSP‐dsRNAs compromises Ep infection on pea

Based on InterProScan and MCL results, ‘ribonuclease/ribotoxin’ emerged as the largest protein family among the Ep CSEPs, with 16 members (Fig. 1 and Table S6). Ribonuclease‐like proteins are prevalent among cereal PM effectors and are known as RALPHs (RNase‐Like Proteins associated with Haustoria; Spanu, 2017). Hence, two highly expressed Ep CSEPs from this family, EpCSEP001 and EpCSEP009, were selected for functional characterization via HIGS. In addition, EpCSP083, an Egh16H‐like virulence factor that is conserved across PMs and other pathogenic fungi (Frantzeskakis et al., 2018; Grell et al., 2003) and expressed early in the infection process, was also selected for functional validation. For the HIGS assay, dsRNA molecules specifically targeting EpCSEP001, EpCSEP009, EpCSP083 or GFP (control) were synthesized (Fig. S7) and individually infiltrated into pea leaves. One day post‐infiltration, leaves were inoculated with Ep conidia and gene expression was assessed by qRT‐PCR at 72 hpi. This infection time point was selected based on initial assessments of the duration of dsRNA‐mediated gene silencing, which indicated that significant knockdown of Ep CSEPs occurs between 24 and 72 hpi (Fig. S8). On average, we observed that the residual amounts of EpCSEP001, EpCSEP009 and EpCSP083 transcripts were 25%, 37% and 50% in the respective dsRNA‐infiltrated leaves as compared to GFP‐dsRNA‐infiltrated control leaves at 72 hpi (Fig. 4a).

Figure 4

dsRNA‐mediated HIGS of EpCSEP001, EpCSEP009 and EpCSP083. (a) EpCSEP/EpCSP transcript levels measured in pea leaves infiltrated with EpCSEP/EpCSP‐ or GFP‐dsRNA at 72 hours post‐inoculation (hpi) with Erysiphe pisi (Ep). Data represent mean ± SEM of expression values normalized to that of the reference gene Eptub2 from at least ten biological replicates from two independent experiments. Significant differences in mean expression values were computed using the non‐parametric Wilcoxon matched‐pairs signed‐rank test (**P < 0.005, ***P < 0.001). (b) Ep18S rRNA transcript levels measured in pea leaves infiltrated with EpCSEP/EpCSP‐ or GFP‐dsRNA at 72 hpi. Data represent mean ± SEM of expression values normalized to that of the reference gene Pstubulin from at least four biological replicates. Significant differences between means were computed using paired t‐test (*P < 0.05, **P < 0.005). (c) Visible powdery mildew disease symptoms are reduced on EpCSEP/EpCSP‐dsRNA‐infiltrated pea leaves compared to GFP‐dsRNA controls at 72 hpi. Bar, 1 cm. (d) Ep growth on EpCSEP/EpCSP‐dsRNA and GFP‐dsRNA‐infiltrated leaves visualized by trypan blue staining at 48 hpi. UN, ungerminated; AP, appressorium; PH, primary hypha; SH, secondary hypha; bar, 50 μm. (e) The percentage of Ep conidia that reached different developmental stages at 48 hpi assessed from >250 conidia from four leaves per dsRNA treatment. Asterisk indicates significantly different values (****P < 0.0001 based on unpaired t‐test) between EpCSEP/EpCSP‐dsRNA and GFP‐dsRNA‐infiltrated leaves at the respective growth stage.

Once we established that dsRNAs could significantly silence Ep genes, we evaluated the effect of EpCSEP/CSP knockdown on PM growth and disease development. Using qRT‐PCR, we quantified the pathogen load in EpCSEP/CSP‐ and GFP‐dsRNA‐infiltrated leaves by monitoring the abundance of Ep18S rRNA. Compared to GFP‐dsRNA‐infiltrated leaves, the 18S rRNA transcript abundance was c. 35% in EpCSEP001‐, c. 44% in EpCSEP009‐ and c. 64% in EpCSP083‐dsRNA‐infiltrated leaves (Fig. 4b). Further, leaves infiltrated with EpCSEP001- and EpCSEP009‐dsRNAs displayed dramatically reduced PM symptoms at 72 hpi compared to control leaves (Fig. 4c). Reduced symptoms were also visible on leaves infiltrated with EpCSP083‐dsRNAs compared to controls (Fig. 4c), albeit less pronounced than that observed for EpCSEP001‐ and EpCSEP009‐dsRNA treatments. To investigate the effect of EpCSEP/CSP silencing on PM disease progression, fungal growth stages were observed with a microscope and quantified. Consistent with the visual symptoms, clear differences in fungal growth stages were observed in the control and respective EpCSEP/CSP‐dsRNA‐infiltrated leaves (Fig. 4d,e). A significantly larger number of conidia were arrested at the multilobed appressorial stage in EpCSEP/CSP‐dsRNA‐infiltrated leaves compared to control leaves. Additionally, an overall reduction in the number of conidia that formed secondary mycelia was observed in leaves infiltrated with EpCSEP/CSP‐dsRNAs compared to controls (Fig. 4d‐e).

In silico off‐target prediction and assessment of EpCSEP/CSP‐dsRNA specificity

To confirm the target specificity of the designed EpCSEP/CSP‐dsRNAs, we performed an in silico Ep transcriptome‐wide off‐target assessment for each dsRNA‐producing sequence (Table S10). The predicted siRNAs for EpCSEP001‐ and EpCSP083‐dsRNAs showed three significant off‐targets each (EpCSEP001: CUFF22211, CUFF219271, CUFF260901; EpCSP083: CUFF108551, CUFF31531, CUFF370461). EpCSEP009‐dsRNA showed only one significant off‐target (CUFF194411). Based on the number of siRNA hits, CUFF22211 and CUFF108551 were predicted to be the major off‐targets of EpCSEP001‐ and EpCSP083‐dsRNA‐derived siRNAs, respectively. To test this prediction, we quantified the expression of all off‐target genes in the corresponding EpCSEP/CSP‐dsRNA and GFP‐dsRNA‐infiltrated samples via qRT‐PCR. Although the data were not statistically significant, the transcript level of EpCSEP001‐dsRNA off‐target genes CUFF22211 and CUFF219271 and EpCSP083‐dsRNA off‐target gene CUFF108551 was c. 55–60% in the respective EpCSEP/CSP‐dsRNA‐infiltrated samples compared to GFP‐dsRNA‐infiltrated controls (Fig. S9). Notably, no off‐target effects were observed for the EpCSEP009‐dsRNA, indicating that it is highly specific.

Subcellular localization and homology modelling of EpCSEP/CSPs

To determine their subcellular localization, EpCSEP/CSPs were fused to GFP and transiently expressed in Nicotiana benthamiana leaf epidermal cells (Fig. 5). The fluorescence patterns indicated that GFP, from the empty vector, and EpCSEP009 localized to the cytoplasm and nucleus whereas EpCSP083 localized exclusively to the nucleus. EpCSEP001 predominately localized to the nucleus (Fig. 5); however, in a few cells, weak GFP fluorescence was also observed in the cytoplasm (not shown).

Figure 5

Subcellular localization of EpCSEP001, EpCSEP009 and EpCSP083 using a fluorescent reporter in Nicotiana benthamiana leaf epidermal cells. Green fluorescence indicates the location of EpCSEP/EpCSP fused to GFP and blue fluorescence indicates DAPI‐stained nuclei. Bars, 25 µm.

To obtain structural and functional insights, homology‐based models of the EpCSEP001 and EpCSEP009 proteins were constructed. The BLAST search displayed significant homology of the C‐terminus of both EpCSEP001 and EpCSEP009 to the fungal RNase superfamily (RNase T1), whose members are predominantly guanyl‐specific nucleases (Fig. 6a). The N‐terminus region of EpCSEP001 and EpCSEP009 shows no significant homology to any of the known proteins and hence could not be modelled. The RNase domains of EpCSEP001 and EpCSEP009 were independently modelled using high‐resolution (1.3Å) crystal structure of F1 ribonuclease from Fusarium moniliforme (PDB: 1FUS) (Vassylyev et al., 1993). The RNase F1 is a guanine‐specific enzyme that hydrolyses a phosphodiester bond at the 3ʹ side of guanine in single‐stranded (ss) RNA. The template shares 37% sequence similarity and 31% sequence identity with both the proteins. The models of EpCSEP001 and EpCSEP009 show good stereochemistry with main chain conformations for 90.82% and 93.48% of amino acids present in the most favoured region of the Ramachandran plot, respectively. The models show a typical RNase fold comprising of one α‐helix packed against an anti‐parallel β‐sheet and a N‐terminus β‐hairpin. The cysteines that form the two disulphide bridges, [Cys148: Cys203 and Cys130: Cys220 for EpCSEP001; Cys94: Cys149 and Cys76: Cys166 for EpCSEP009], characteristic of the RNase F1 family, are conserved in both the proteins (Fig. 6b) (Pace et al., 1991). The residues His65, Glu83, Arg101 and His116 comprise the catalytic site of the RNase F1 (Vassylyev et al., 1993). The multiple sequence alignment of EpCSEP001 and EpCSEP009 with RNase F1 reveals that the active site residues in Ep effectors are partially conserved (Fig. 6c). This is in contrast to Bgh RALPHs, where the active site residues are absent (Pedersen et al., 2012). Three out of four catalytic residues corresponding to Glu83 (Glu177), Arg101 (Arg196) and His116 (His212) are conserved in EpCSEP001, and two active site residues, Glu123 and Arg142, are conserved in EpCSEP009. However, the residue corresponding to His65 is not conserved in either of the two proteins (Fig. 6c). Superimposition of the recently determined crystal structure of BghBEC1054 (PDB: 6FMB) (Pennington et al., 2019) on EpCSEP001 and EpCSEP009 yields root‐mean‐square deviations of 5.1 and 5.8 Å, respectively (Fig. 6d). Major conformational changes are present in the two loops, one between β3 and β4 and the other between β6 and the C‐terminus of the protein. Additionally, the C‐terminus β‐strand, unique to BghBEC1054, is absent in EpCSEP001 and EpCSEP009. The chemical shift perturbations as measured by NMR for BghBEC1054 reveal conformational flexibility in the loop region and implicates them in substrate recognition (Pennington et al., 2019). The electrostatic potential surface shows a positively charged surface surrounding the predicted catalytic site in both Ep CSEPs unlike in the case of BghBEC1054 (Fig. 6e). This distribution of specific positive charge is indicative of its RNA‐binding property. Although the BLAST result displayed similarity of EpCSP083 with the DUF3129 family, it could not be modelled since it shows poor homology with the known structures in the PDB.

Figure 6

Homology modelling of EpCSEP001 and EpCSEP009. (a) Domain organization of EpCSEP001 (left) and EpCSEP009 (right) depicting the position of SPs and RNase‐like domains. (b) Multiple sequence alignment showing sequence similarity among Fusarium moniliforme RNase F1, EpCSEP001, EpCSEP009 and BghBEC1054. The conserved Cys residues are highlighted with a red asterisk. Black arrows show position of catalytic residues in F1 RNase of F. moniliforme. (c) Homology models of EpCSEP001 and EpCSEP009 are depicted in golden orange and cyan ribbon, respectively. The conserved active site residues are shown in the sticks representation. (d) Structural superimpositions of BghBEC1054 (PDB: 6FMB, light violet), with models of EpCSEP001 (golden orange) and EpCSEP009 (cyan). (e) Electrostatic surface potential of EpCSEP001 and EpCSEP009 are coloured according to the bar underneath. Potential catalytic site is marked with arrows.

Discussion

Effector proteins, primarily synthesized and secreted from the haustorium (Chaudhari et al., 2014), play important roles in determining the outcome of plant–PM interactions. Therefore, the goal of this study was to identify and characterize candidate secreted effector proteins from enriched haustoria of the pea PM Ep using transcriptomics and HIGS.

Quality of the Ep haustorial transcriptome

The haustorial enrichment procedure (Weßling et al., 2012) yielded a mixture of juvenile and mature, physically intact Ep haustoria, representing different developmental stages (Fig. S1). Although microscopic and qRT‐PCR evaluation of the enriched haustorial samples revealed minimal contamination from epiphytic fungal structures (e.g. conidia and hyphae) and plant cells, we suspected that transcripts originating from these other cell types would also be represented in the haustorial cDNA library, albeit to a lower extent. While contaminating plant sequences were filtered out at the read mapping level (14.6%) and transcript (3.2%)/protein (20%) annotation steps (Fig. S2), a similar approach could not be used to remove contaminating fungal transcripts. Therefore, it is noteworthy that the assembled haustorial transcriptome likely includes a small percentage of transcripts from other fungal cell types.

We found that our approach of using both reference genome‐based and de novo assembly methods slightly improved the coverage and comprehensiveness of the Ep haustorial transcriptome (Table 1; Fig. S3). Such integrated approaches are known to recover additional transcript fragments, especially in cases where partial genome information is available in addition to RNA‐seq reads (Jain et al., 2013). Further, gene ontology (GO) analysis revealed that biological processes enriched in the top‐expressed Ep protein‐coding transcripts were similar to those enriched in the haustorial transcriptomes/proteomes of other PMs (Bindschedler et al., 2009, 2011; Weßling et al., 2012), validating the overall quality of the Ep haustorial transcriptome (Fig. S10 and Table S12).

The Ep effector repertoire differs from cereal PM CSEPs

Our analysis of Ep CSEPs validates the previously observed differences between dicot‐adapted and cereal PM CSEPs. The total number of CSEPs (167) predicted from the Ep haustorial transcriptome was consistent with that previously reported for other dicot‐adapted PMs, including En (150), Gor (115), Pxa (53), Ppa (81) and Oidium heveae (133) (Fonseca et al., 2019; Jones et al., 2014; Liang et al., 2018; Vela‐Corcía et al., 2016; Weßling et al., 2012), but significantly lower than that reported for cereal PMs [Bgh (491) (Spanu et al., 2010); Bgt (844) (Muller et al., 2019)]. Further, MCL analysis revealed that Ep CSEP families contain very few members, similar to what is observed in dicot‐adapted PM CSEPs, but different from cereal PM CSEP families, which are known to comprise up to 38 members (Wu et al., 2018). In addition, unlike in the cereal PM CSEPs, fewer Ep CSEPs carry the Y/F/W‐x‐C motif, highlighting another difference between the two PM lineages. In general, sequence similarity between Bgh and Bgt CSEPs is limited, suggesting that effector differentiation in the cereal PMs is related to their strict host specialization (Spanu and Panstruga, 2012). In contrast, c. 70% of Ep CSEPs share sequence similarities with proteins from dicot‐adapted and/or cereal PMs (Fig. 1). The high sequence conservation and smaller effectorome in Ep may be partially attributable to the polyphagous nature of dicot‐adapted PM fungi, which may result in lower selective pressure for host specialization (Liang et al., 2018; Wu et al., 2018).

RALPH‐like effectors and an Egh16H‐like virulence factor are required for Ep virulence

A scan for protein domains in the Ep CSEPs revealed that certain classes of effectors are conserved across dicot‐adapted and cereal PMs, hinting that they may be derived from a common ancestor. For instance, the ribonuclease‐like effectors, which constitute the largest effector family in cereal PMs (Bourras et al., 2018; Liang et al., 2018), is also prevalent in Ep (Table S6). Therefore, we focused on two ribonuclease‐like Ep CSEPs (EpCSEP001 and EpCSEP009) and an Egh16H‐like virulence factor (EpCSP083) for further functional validation. All three candidates are preferentially expressed in haustoria, but differences in infection stage‐specific expression patterns were observed. EpCSEP001 and EpCSEP009 exhibit a biphasic expression pattern, expressing at early and late stages of infection whereas EpCSP083 expression is restricted to early infection stages (Fig. 3). This may imply that EpCSEP001 and EpCSEP009 are important throughout the infection cycle whereas EpCSP083 is critical during early infection stages. Further, their subcellular localization patterns corroborate the in silico prediction using the Localizer tool (Fig. 5; Table S6), suggesting potential host target sites for these proteins.

To further examine the role of these CSEP/CSPs in Ep pathogenesis, we individually silenced each candidate in infected pea leaves using a dsRNA delivery‐based HIGS approach. Although the exact mechanism of HIGS is unknown, it is proposed that the in planta expressed dsRNAs are processed into small RNAs by the plant RNAi machinery, which, when taken up by the fungus, regulate the expression of target genes in a sequence‐specific manner. Alternatively, the precursor dsRNA can also be taken up and processed into small RNAs by the fungal RNAi machinery (Qi et al., 2019). We found that infiltration of EpCSEP/CSP‐dsRNAs into pea leaves resulted in silencing efficiencies of 50–75%, which is consistent with earlier studies in which dsRNAs were delivered through spraying (Koch et al., 2018; McLoughlin et al., 2018). Further, pea leaves silenced for EpCSEP001, EpCSEP009 or EpCSP083 exhibited reduced PM symptoms, fungal penetration and colony growth compared to their respective controls, indicating that all three candidates contribute to full virulence. This suggests that dsRNA delivery via infiltration could be a powerful and alternative strategy for high‐throughput screening of gene functions, particularly in fungi having obligate biotrophic lifestyles.

Macroscopic, microscopic and fungal quantification data revealed that the reduced PM growth phenotype was stronger in leaves infiltrated with EpCSEP001‐ or EpCSEP009‐dsRNAs compared to those infiltrated with EpCSP083‐dsRNAs. This may be related to the higher silencing efficiencies of these EpCSEP‐dsRNAs (Fig. 4a) and/or the functional requirement of these CSEPs at multiple infection stages (discussed above). Analysis of EpCSEP001‐dsRNA off‐targets revealed that the expression of two additional genes was affected in EpCSEP001‐dsRNA‐infiltrated leaves (Fig. S9). Although these data were not significant, reduced expression of these off‐targets may partly contribute to the observed PM growth phenotype on EpCSEP001‐dsRNA‐infiltrated leaves. One of the off‐targets, CUFF22211, encodes the ribonuclease‐like effector EpCSEP002, highlighting the importance of this class of effectors in Ep virulence.

The Egh16H‐like virulence factor EpCSP083 is expressed between 6 and 24 hpi (Fig. 3), before effective silencing is typically achieved via HIGS (Fig. S8). Therefore, it is possible that a significant knockdown of EpCSP083 at later infection time points (Fig. 4a), particularly when its function is not as critical, may not impact Ep virulence as severely as EpCSEP001 or EpCSEP009‐dsRNAs (Fig. 4b–e). Off‐target expression analysis revealed that, in addition to EpCSP083, another Egh16H‐like gene, CUFF108551, showed reduced expression in EpCSP083‐dsRNA‐infiltrated leaves (Fig. S9). This might explain the significant yet less pronounced difference in PM growth phenotype on EpCSP083‐ and GFP‐dsRNA‐infiltrated leaves despite the lower level of EpCSP083 silencing in these leaves. Reduced fungal virulence was previously observed when Egh16H homologues, GAS1 and GAS2, were deleted in the rice blast fungus Magnaporthe grisea (Xue et al., 2002), highlighting the conserved role of Egh16H‐like virulence factors in fungal pathogenesis.

We used homology modelling to gain further insights into the putative functions of Ep CSEPs. The pattern of disulphide bonds in the model of EpCSEP001 and EpCSEP009 and sequence similarity places them in the RNase F1 family. The sequence and structural analysis reveals that the residues responsible for catalysis are partially conserved in the two proteins. Thus, Ep CSEPs may harbour specific RNA cleavage activity, although the residue corresponding to His65 is not conserved in EpCSEP001 and EpCSEP009, and the one corresponding to His116 is Asn158 in EpCSEP009. It has been shown that the extracellular nuclease from Bacillus intermedius (binase, RNase Bi) retains its guanine‐specific cleavage activity even though it lacks the histidine corresponding to His65. In addition, mutation of the second His to Asn in binase does not result in abrogation of its activity (Okorokov et al., 1997), indicating that these two residues may not be essential for catalysis. This is in contrast to the RALPH‐CSEPs from Bgh, which lack all of the catalytic residues and only have RNA binding activity (Pedersen et al., 2012; Pennington et al., 2019). EpCSEP001 and EpCSEP009 may therefore represent RALPH effectors that also possess RNA cleavage activity. Although a few Bgh RALPHs, including BghBEC1054, function as virulence factors (Pennington et al., 2019; Pliego et al., 2013), based on our modelling data it is tempting to speculate that the mechanism(s) by which the Ep CSEP‐RALPHs and BghBEC1054 interfere with their targets may be different.

In conclusion, our data shows that EpCSEP001, EpCSEP009 and EpCSP083 represent novel virulence factors that are important for pea PM pathogenesis. Future investigations into their host targets will provide deeper functional insights that can be eventually used to engineer durable resistance against PM in legumes.

Experimental Procedures

A detailed description of the experimental procedures is provided in supplementary Text S1.

Biological material and growth conditions

Susceptible Pisum sativum cv. AP‐3 (pea) seeds were grown in Conviron growth chambers at 22 °C, 70% relative humidity and a 16/8‐h photoperiod with photosynthetically active radiation of 170 µmol m−2 s−1. Leaves of 10–12‐day‐old plants were inoculated with conidia of Ep isolate Palampur‐1 (Banyal et al., 2014) from heavily infected pea leaves (5–7 dpi) using a fine‐hair brush. Nicotiana benthamiana plants were grown at identical conditions for the localization assays.

Isolation of Ep haustoria, RNA extraction, cDNA library synthesis and sequencing

Ep haustoria were isolated from heavily infected pea leaves (6 dpi) by isopycnic centrifugation (Micali et al., 2011) and checked for their integrity under a confocal microscope (SP8, Leica, Wetzlar, Germany) after staining with Calcofluor White (Sigma, St Louis, MO, USA). Total RNA was extracted from two independently isolated haustorial pellets using RNAiso Plus (Takara, Shiga, Japan), pooled and purified using the RNeasy Plant Miniprep Kit (Qiagen, Hilden, Germany). The cDNA library was synthesized from ~1 µg rRNA‐depleted RNA and 100 million 101 bp paired‐end reads were sequenced using a HiSeq‐2500 system (Illumina, San Diego, CA, USA). The quality of sequenced reads was assessed using FastQC (Andrews, 2010) and adapters and low‐quality sequences (Phred scores < 30) were removed using Trimmomatic (Bolger et al., 2014).

Ep haustorial transcriptome assembly, ORF prediction and GO enrichment

HQ paired‐end reads were mapped to the Ep draft genome (NCBI BioProject: PRJEA‐50315) using Tophat (v. 2.1.1) (Kim et al., 2013) and assembled into transcripts via a reference‐based approach using Cufflinks (v. 2.2.1) (Trapnell et al., 2012). Reads that did not map to the Ep genome were mapped to the pea transcriptome using Tophat to remove any contaminating plant sequences. Remaining HQ reads were de novo assembled using Trinity (v. 2.3.2) (Haas et al., 2013). Completeness of the assembled transcriptome was evaluated using BUSCO (Simão et al., 2015). Protein coding regions were predicted from the non‐redundant transcripts using TransDecoder (v. 3.0.1) (Haas et al., 2013). Candidate ORFs were validated using the HMMER module hmmscan (Potter et al., 2018) with Pfam database as input. GO terms were predicted as per Weßling et al. (2012).

Prediction of the Ep secretome and EpCSEPs

For the secretome prediction, two approaches were followed on the ORFs beginning with methionine. To categorize canonically secreted proteins, SPs were predicted using SignalP (v. 3.0/4.1) (D‐score cut‐off ≥ 0.5) (Bendtsen et al., 2004a; Petersen et al., 2011). Proteins lacking the SP were analysed for non‐canonical secretion using SecretomeP (v. 2.0) (NN score ≥ 0.6) (Bendtsen et al., 2004b). TargetP (v. 1.1) (Emanuelsson et al., 2007) was used for the prediction of subcellular localization of the proteins and the presence of the transmembrane (TM) domain was predicted by TMHMM (v. 2.0) (Krogh et al., 2001). Proteins having localization signal ‘S: Secreted’ and no TM/TM within the SP were retained. To remove the proteins having a GPI modification site, Big‐Pi Fungal predictor (Eisenhaber et al., 2004) was used. The prediction of the subcellular localization of Ep CSPs in the plant cell was also made using Localizer (Sperschneider et al., 2017). EffectorP (v. 2.0) was used to predict Ep Ep CSEPs from the Ep secretome. Ep CSEP homologues were identified by performing BLASTP alignments (NCBI‐BLAST suite, v. 2.7.1) with protein sequences of En (ftp://ftp.ensemblgenomes.org/pub/fungi/release-37/fasta/fungi_ascomycota2_collection/erysiphe_necator), Gor (Weßling et al., 2012), Bgh (ftp://ftp.ensemblgenomes.org/pub/fungi/release-37/fasta/blumeria_graminis) and Ascomycetes (RefSeq) using cut‐off E‐value score ≤0.001 and query coverage ≥50%. The amino acid sequences of Ep CSEPs were screened for the presence of conserved motifs from different classes of fungal effectors using FIMO (Find Individual Motif Occurrence) (Grant et al., 2011) and their enrichment was determined using AME (Analysis of Motif Enrichment) (McLeay and Bailey, 2010) at MEME suite (v. 4.12.0). Protein domain/family analysis was done using InterProScan (v. 5.26‐65.0) (Quevillon et al., 2005).

Phylogeny and CIRCOS analysis

The evolutionary relatedness of Ep CSEPs was inferred by the neighbour‐joining tree and distances were computed using the p‐distance method with MEGA‐X (Kumar et al., 2018). The confidence of the clustering pattern was estimated with bootstrap analysis (Felsenstein, 1985) of 100 replicates. The protein families were predicted via MCL analysis (Enright et al., 2002). The CIRCOS standalone tool (Krzywinski et al., 2009) was used to visualize the data for expression (FPKM), protein family, sequence similarity with proteins from other PMs and presence of Y/F/W‐x‐C motif.

Haustorial‐specific and infection‐dependent expression profiling

Relative transcript abundance of EpCSEP/CSPs in haustorial and epiphytic fungal material was analysed by qRT‐PCR. For the temporal expression analysis, infected leaf tissues were harvested at 0, 6, 12, 24, 48, 72 and 120 hpi. First‐strand cDNA was synthesized from total RNA using iScript reverse transcriptase (Bio‐rad, Hercules, CA, USA). qRT‐PCR was performed using 5x HOT FIREPol® EvaGreen® qPCR Supermix (Solis BioDyne, Tartu, Estonia) in a QuantStudio Flex 6 ABI system (Thermo Fisher Scientific, Waltham, MA, USA). LinRegPCR was used to calculate mean PCR efficiencies per amplicon (Ruijter et al., 2009). Efficiency‐corrected relative expression values normalized to the reference Ep β‐tubulin gene (Eptub2, NCBI accession: X81961) were calculated as per Pfaffl (2001).

Design and synthesis of dsRNA

The siRNA vulnerable regions from the coding sequences of EpCSEP001, EpCSEP009 and EpCSP083 excluding the SP were determined using online tools, amplified and cloned in the pGEMT‐easy vector system (Promega, Madison, WI, USA). The recombinant clones were verified and amplified with M13 primer pair to extract the predicted siRNA vulnerable sequence along with T7 and SP6 promoter sequences. The amplified products were run on a 1% agarose gel, purified and 1 μg of the eluted PCR product was taken as the template for in vitro transcription using an SP6 and T7 MEGAscript transcription kit (Thermo Fisher Scientific) to yield ssRNA transcripts. For the generation of dsRNA, ssRNA transcripts were annealed at 65 °C followed by gradual cooling at room temperature.

dsRNA‐mediated HIGS

Second leaves of uninfected 12‐day‐old pea plants were infiltrated with 100 parts per million (ppm) EpCSEP/CSP‐dsRNA using a needleless syringe. Leaves infiltrated with GFP‐dsRNA acted as control. Two to three independent experiments with at least five replicates per experiment were conducted for the evaluation of each Ep CSEP/CSP. Following dsRNA infiltration, leaves were infected with Ep conidia. At 72 hpi, transcript abundance of the corresponding Ep CSEP/CSP was evaluated via qRT‐PCR and pathogen load was quantified by measuring Ep 18S rRNA levels. qRT‐PCR data analysis was performed as described for the infection‐dependent expression profiling. Eptub2 (NCBI accession: X81961) and pea tubulin (Pstub) served as the reference genes. PM symptoms were visualized in both the GFP‐dsRNA and EpCSEP/CSP‐dsRNA‐infiltrated leaves at 72 hpi. To visualize and quantify differences in Ep growth stages, infected leaves of control and EpCSEP/CSP‐dsRNA were harvested at 48 hpi, stained with trypan blue and visualized under a ×20 objective in a bright field microscope (Carl Zeiss, Germany).

In silico assessment of transcriptome‐wide off‐targets of EpCSEP/CSP siRNAs and qRT‐PCR expression analysis

Off‐target analysis was performed using the web‐based server pssRNAit (https://plantgrn.noble.org/pssRNAit/) with default parameters. The dsRNA‐producing region of each EpCSEP/CSP was used as the query sequence for off‐target prediction in the Ep haustorial transcript library. Significant off‐target genes were annotated using BLASTX against the fungal RefSeq database with a cut‐off E‐value score ≤0.001. The expression of in silico‐predicted off‐target genes was assessed via qRT‐PCR in Ep CSEP/CSP‐ and GFP‐dsRNA‐infiltrated pea leaves at 72 hpi and analysed as described above.

Subcellular localization

Coding sequences of select EpCSEP/CSPs lacking the SP were introduced into the pSITE2CA vector (Chakrabarty et al., 2007) to create a GFP fusion protein and introduced into Agrobacterium tumefaciens strain GV3101 for transient expression in N. benthamiana leaves. The localization signal was observed via confocal microscopy.

Homology modelling of EpCSEP001 and EpCSEP009

The template, RNase F1 structure from F. moniliforme in apo form (PDB code: 1FUS), used for modelling of EpCSEP001 and EpCSEP009, was selected on the basis of resolution of the structure and homology. The homology model was generated using Discovery Studio 3.5 (BIOVIA). The sequence of 1FUS structure was aligned against the target sequence to identify the matched regions. Based on the atomic coordinates of the template, the homology model of the target protein was constructed and verified by the Ramachandran plot in Coot (Emsley et al., 2010). Figures were prepared using PyMol (Delano, 2002). The electrostatic potential surface was calculated using APBS plugin in PyMol. Multiple sequence alignment was performed by T‐Coffee server and results were generated using ESPript (Robert and Gouet, 2014). BEC1054_Bgh (N1JJ94) and F. moniliforme RNase F1 (P10282) sequences were obtained from UniProt.

Author Contributions

D.C. conceived the study. D.C., G.S. and R.A. designed the experiments. G.S., R.A. and A.G. performed the experiments. D.S. performed the bioinformatics analyses. D.C., G.S. and R.A. analysed the data. D.J. and P.B. directed and performed the molecular modelling and prepared Fig. 6. D.C., G.S. and D.J. wrote the manuscript with minor contributions from R.A. All authors have read and approved the final manuscript.

Conflict of Interest Statement

The authors declare no conflict of interest.

Fig. S1 Microscopic images showing intact Ep haustoria.

Click here for additional data file.

Fig. S2 Workflow for Ep transcriptome assembly, annotation and secretome prediction.

Click here for additional data file.

Fig. S3 BUSCO analysis.

Click here for additional data file.

Fig. S4 Abundance and position of YxC, FxC and WxC forms of the Y/F/WxC‐motif.

Click here for additional data file.

Fig. S5 Agarose gel images showing PCR amplification of select EpCSEP/ CSPs.

Click here for additional data file.

Fig. S6 qRT‐PCR profiles of low expressed EpCSEPs.

Click here for additional data file.

Fig. S7 Analysis of amplified EpCSEP/CSP sequences and in vitro synthesized ss and dsRNA.

Click here for additional data file.

Fig. S8 The timing of dsRNA‐mediated RNAi silencing of Ep CSEP/CSP.

Click here for additional data file.

Fig. S9 qRT‐PCR analysis of off‐target effects of EpCSEP001, EpCSEP009 and EpCSP083 siRNAs.

Click here for additional data file.

Fig. S10 Gene ontology enrichment of Ep haustorial proteins.

Click here for additional data file.

Table S1 Summary of statistics for the output of Ep haustorial transcriptome sequencing.

Click here for additional data file.

Table S2 List of assembled Ep transcripts.

Click here for additional data file.

Table S3 Missing ascomycete core genes in the Ep haustorial transcriptome.

Click here for additional data file.

Table S4 List of predicted Ep full‐length proteins.

Click here for additional data file.

Table S5 List of proteins in the Ep secretome.

Click here for additional data file.

Table S6 List of Ep candidate secreted effector proteins.

Click here for additional data file.

Table S7 Number of EpCSEPs harbouring known conserved motifs.

Click here for additional data file.

Table S8 List of primers used for the qRT‐PCR expression profiling.

Click here for additional data file.

Table S9 List of primers used for the synthesis of select EpCSEP/CSP‐dsRNAs, expression profiling of Ep CSEP/CSPs and off‐targets, and Ep quantification.

Click here for additional data file.

Table S10 In silico prediction of off‐target effects for EpCSEP001, EpCSEP009 and EpCSP083 ‘diced’ siRNAs using the pssRNAit web server.

Click here for additional data file.

Table S11 List of primers used for Gateway cloning (for N‐terminal GFP tag).

Click here for additional data file.

Table S12 List of enriched GOSlim biological process categories and predicted proteins.

Click here for additional data file.

Text S1 Supplementary experimental procedures and references.

Click here for additional data file.