A specific fungal transcription factor controls effector gene expression and orchestrates the establishment of the necrotrophic pathogen lifestyle on wheat

The fungus Parastagonospora nodorum infects wheat through the use of necrotrophic effector (NE) proteins that cause host-specific tissue necrosis. The Zn2Cys6 transcription factor PnPf2 positively regulates NE gene expression and is required for virulence on wheat. Little is known about other downstream targets of PnPf2. We compared the transcriptomes of the P. nodorum wildtype and a strain deleted in PnPf2 (pf2-69) during in vitro growth and host infection to further elucidate targets of PnPf2 signalling. Gene ontology enrichment analysis of the differentially expressed (DE) genes revealed that genes associated with plant cell wall degradation and proteolysis were enriched in down-regulated DE gene sets in pf2-69 compared to SN15. In contrast, genes associated with redox control, nutrient and ion transport were up-regulated in the mutant. Further analysis of the DE gene set revealed that PnPf2 positively regulates twelve genes that encode effector-like proteins. Two of these genes encode proteins with homology to previously characterised effectors in other fungal phytopathogens. In addition to modulating effector gene expression, PnPf2 may play a broader role in the establishment of a necrotrophic lifestyle by orchestrating the expression of genes associated with plant cell wall degradation and nutrient assimilation.

The fungus Parastagonospora nodorum causes septoria nodorum blotch (SNB) of wheat. P. nodorum uses necrotrophic effectors (NEs) to cause tissue necrosis and facilitate infection of hosts possessing dominant susceptibility genes. The genes encoding three of these NEs are known: SnToxA, SnTox1, and SnTox3. SnToxA encodes a 13.2 kDa mature protein that causes necrosis on wheat cultivars that possess the dominant susceptibility gene Tsn1 1,2 . Near-identical copies of ToxA have been found in two other wheat fungal pathogens, Pyrenophora tritici-repentis (Ptr) 3 and Bipolaris sorokiniana 4 . These may have been horizontally acquired, presumably from P. nodorum 1 . SnTox1 encodes a 10.3 kDa cysteine-rich mature protein that causes necrosis and confers virulence on wheat cultivars possessing Snn1 5 . SnTox3 is also a cysteine-rich NE. Sensitivity to the effector is conferred by either Snn3-B1 or Snn3-D1 located on wheat chromosomes 5BS and 5DS, respectively 6,7 . Genetic studies and protein purification assays indicate that P. nodorum possesses many more unidentified effectors associated with SNB 8 .
SnToxA, SnTox1 and SnTox3 are highly expressed during early infection but their expression is greatly decreases during saprophytic growth on the necrotised host tissue 9 . However, else was known about factors affecting their regulation until recently. Studies of TFs in P. nodorum have also provided some insights into effector gene regulation. Deletion of the APSES-class TF gene SnStuA in P. nodorum resulted in mutants with abnormal vegetative growth, loss of sporulation and a complete loss of virulence on wheat 10 . The expression of SnTox3 was significantly down-regulated in the mutant, though the loss in virulence is likely attributable to pleotropic effects incurred by the mutation. A C 2 H 2 zinc finger TF PnCon7 that binds to the promoter region of SnTox3 was

Results
PnPf2 is required for full hyphal proliferation during host infection. The transcriptome of the P. nodorum reference wildtype strain SN15 was compared to the PnPf2-deleted strain pf2-69 grown under two conditions. Firstly, we sampled RNA during early infection at three days in planta (ip) where PnPf2, SnToxA, SnTox1 and SnTox3 are maximally expressed. Wheat cv. Halberd (Tsn1, Snn1, Snn3) was used as a host as it is susceptible to SN15 and pf2-69 9 . Secondly, SN15 and pf2-69 were grown for three days in vitro (iv) in Fries 3 broth which is conducive for SnTox1 and SnTox3 production 9 . Vegetative growth of pf2-69 was comparable to SN15 9 . Pairedend Illumina HiSeq technology was used as an RNAseq sequencing platform. The latest SN15 genome revision produced 13,563 predicted genes 16 . Deep sequencing produced more than 90% fungal transcripts that aligned to predicted genes for all samples (Supplementary Data S1 and Table 1). In vitro and ip samples returned an average of 24 million and 290 million read pair fragments (including plant reads), respectively. Between 18 and 22 million read pairs, representing an average of 6.94% of the total reads, aligned to the SN15 genome for the SN15 ip treatment (Supplementary Table S1). Between 3.4 and 5.9 million reads (average 1.57% of total) from pf2-69 ip growth aligned to the SN15 genome. The low proportion of fungal reads from pf2-69 suggests reduced biomass during infection. Quantitative PCR of genomic DNA extracted from three days post infected wheat cv. Halberd confirmed that the biomass of pf2-69 was significantly lower than strains carrying a functional copy of PnPf2 (Fig. 1a,b).

Analysis of differentially expressed (DE) genes.
Genes were considered DE in a contrast of isolate or treatment if tests of absolute log 2 fold change >1 were consistently significant (P adj < 0.05) for three test methods described below (Supplementary Data S3). Because pf2-69 ip samples had considerably fewer reads than other samples, an additional filter requiring pf2-69 ip samples to have > = 10 counts per million (CPM) for a gene to be called down-regulated was used for high-confidence DE prediction sets. For SN15 ip and iv treatments, 1,889 genes were up-regulated and 1393 were down-regulated ip (Supplementary Table S2). A total of 1,736 genes were up-regulated and 706 genes were down-regulated between the pf2-69 ip and iv treatments. For ip comparisons, 303 genes were significantly reduced whereas 449 were up-regulated in pf2-69 over SN15. Additional DE genes were observed using relaxed criteria, allowing genes where any of the three tests are significant (<3 tests) (Supplementary Table S2). The main difference between the three tests results were in how they handle contrasts involving samples with few or no reads aligned to the gene. Additional genes involving pf2-69 ip samples with <10 CPM were identified using the same relaxed criteria. In total, 269 gene were down-regulated in pf2-69 ip compared to iv growth and had fewer than 10 CPM in pf2-69 during ip growth. Similarly, 163 genes were down-regulated in pf2-69 during infection compared to SN15 and had fewer than 10 CPM in pf2-69 during ip growth (Supplementary Table S2).
A principal component analysis (PCA) plot for PC1 and PC2 was constructed based on normalised fragment counts per gene to describe the variation between and within each treatment (Fig. 1c). The biological replicates tightly clustered together, with each treatment strongly differentiated from the others. This indicates that sample treatment and sequencing did not contribute to systematic biases that could not be removed by normalisation. PC1 captured 71% of the total variance and discriminated iv from ip samples. PC2 captured 12% of the variance and discriminated SN15 from pf2-69.
We then examined SnToxA, SnTox1 and SnTox3 expression profiles (Fig. 1d). As expected, the expression of SnToxA and SnTox3 was almost abolished in pf2-69 ip. SnTox3 expression was also highly reduced in pf2-69 iv. SnToxA is poorly expressed in SN15 and pf2-69 during iv growth. SnTox1 expression was significantly higher in SN15 compared to pf2-69. SnTox1 is still strongly expressed during ip growth and had the lowest fold change difference between SN15 and pf2-69 ip compared to SnToxA and SnTox3.
PnPf2 regulates genes that encode effector-like proteins. To identify candidate effector genes positively regulated by PnPf2, we analysed genes that were down-regulated in pf2-69 that possessed a secretory signal peptide (but no transmembrane domains outside of the signal peptide) and were predicted to be effector-like by EffectorP 17 . Twelve genes that showed a similar expression profile to SnToxA (ie. down-regulated in pf2-69 ip compared to SN15 ip and up-regulated ip in both strains) were identified (Fig. 2). In contrast, SnTox1 and SnTox3 were the only effector genes categorised in their respective expression profile categories (Fig. 2). The expression profiles of these candidate effector genes in SN15 and pf2-69 three days post-infection were validated using qRT-PCR ( Supplementary Fig. S1). Apart from SNOG_10736, SNOG_13939 and SNOG_02980, the qRT-PCR-based expression profile of all other candidate effector genes between SN15 and pf2-69 was consistent with findings from the RNAseq data. The expression profiles of the 12 candidate effector genes in SN15 were examined between three and 10 days post-infection using available microarray gene expression data 18 and qRT-PCR analyses performed in this study ( Supplementary Fig. S2). SNOG_08150, SNOG_13939, SNOG_30077, SNOG_30352 and SNOG_30359 demonstrated similar expression profiles to SnToxA, SnTox1 and SnTox3 where gene expression peaked at three dpi and decreased to almost non-detectable levels at seven and 10 dpi, coinciding with host tissue necrosis.
Four P. nodorum SN79-1087 is non-pathogenic on wheat and lacks SnToxA, 1 and 3 5 . We decided to investigate if these 12 candidate effectors are present or altered in SN79-1087 22 . BlastP and tBlastN analysis revealed five genes were identical between SN15 and SN79-1087. SNOG_02755 and 10736 are also present in SN79-1087, but both have in-frame deletions in low-complexity amino acid repeat regions. Changes in amino acid sequence were observed for seven gene homologs in SN79-1087 (Table 1). Frame shifts or premature stop codons were not observed for these genes.

PnPf2 regulates depolymerase and nutrient assimilation gene expression in planta.
To investigate changes in overall biochemical processes between SN15 and pf2-69 during iv and ip growth, we assessed DE genes for enrichment of GO terms 23 (Fig. 3). GO terms were assigned to all genes where possible using InterProScan 24 and dbCAN 25 .
During iv growth, genes categorised under oxidoreductase activities, flavin adenine dinucleotide binding and catalytic activity were significantly up-regulated in pf2-69 (Fig. 3a) revealed that differentially expressed genes associated with oxidoreductase activities are central to biological processes related to respiratory electron transport chain, steroid metabolism, redox, carbohydrate metabolism and generation of precursor metabolites and energy (Fig. 3b).
During ip growth, molecular functions (MFs) associated with hydrolase, glucosidase and peptidase activities tended to be down-regulated in pf2-69 (Fig. 3c). The MF hydrolase activity associated with hydrolysing N-glycosyl compounds consisted of 155 genes. Of these, the expression of 39 genes were significantly lower in pf2-69. Similarly, the MF associated with another hydrolase activity associated with hydrolysing N-glycosyl compounds consisted of 32 genes, of which eight genes were significantly down-regulated in pf2-69 compared to SN15. The MF hydrolase activity consisted of 1,168 genes. Of these, 64 were expressed at lower levels in pf2-69. The majority of genes annotated encode plant cell wall degrading enzymes (CWDEs) and other carbohydrate depolymerases such as β-xylosidases, acetyl xylan esterases, glucanases and glucosidases (Supplementary Data S3 & S5). Arabinose is a major constituent of the plant cell wall. GO enrichment indicates that PnPf2 regulates arabinose metabolism in P. nodorum. Of the six genes associated with α-L-arabinofuranosidase activity, five were expressed at lower levels in pf2-69 (Fig. 3c,d).
For protein degradation, 240 genes encode proteins with predicted peptidase activity were differentially expressed (GO:0008233) (Fig. 3c). Of these, 29 were down-regulated in pf2-69 compared to SN15. Additionally, 63 genes encoding proteins with putative metallopeptidase activity were identified from the genome. Of these, 14 were down-regulated in pf2-69 ip compared to SN15. The MF associated with metallocarboxypeptidase activity (GO:0004181) consisted of nine genes where the expression of seven was reduced in pf2-69. For the MF associated with serine-type peptidase activity, 22 of 131 genes were expressed at lower levels in pf2- 69  www.nature.com/scientificreports www.nature.com/scientificreports/ GO analysis revealed that cellular redox potential in pf2-69 was perturbed during ip growth in addition to a similar defect observed during iv growth. A MF associated with oxidoreductase (GO:0016491) activity was enriched in up-regulated genes in pf2-69 ip compared to SN15 (Fig. 3c). The majority of DE genes encoding oxidases, cytochrome P450s, reductases and dehydrogenases (Supplementary Data S3 and S5) are associated with a biological role in carbohydrate metabolism (Fig. 3d). In addition, MFs linked to transport activities were enriched with genes that were similarly up-regulated in pf2-69 ip compared to SN15 (Fig. 3c). Genes associated with the transport function encode sugar and amino acid transporters ( Fig. 3d; Supplementary Data S3 and S5).
Identification of DNA motifs enriched in the promoters of PnPf2-regulated genes. We hypothesised that a shared pf2-69 DE patterns implied a common transcriptional regulator. Therefore, promoters of these gene sets may harbor over-represented motif(s) functioning as potential PnPf2 transcription factor binding site(s) (TFBS). Analysis of the promoters from the respective pf2-69 DE gene groupings revealed three such motifs (Fig. 4). The motif WMGGVCCGAA, enriched in pf2-69 iv and ip down-regulated gene promoters, is similar to an enriched motif associated with AbPf2 down-regulated genes in A. brassicicola 12 and is characteristic of a Zn 2 Cys 6 TFBS 26,27 . A second motif resembling a C 2 H 2 TFBS (RTSYGGGGWA) was significantly enriched in pf2-69 ip down-regulated gene promoters. The third motif (CTGYGCCGCA) also resembled a C 2 H 2 TFBS and was enriched in pf2-69 iv up-regulated gene promoters. The identification of unique enriched motifs in the separate datasets suggests that PnPf2 may act as an indirect regulator or its binding site specificity can be influenced by other regulators of target genes.

Absence of interaction between PnPf2 and the putative consensus motif on SnToxA and SnTox3
promoters. Inspection of the SnToxA and SnTox3 promoter region revealed at least one occurrence of the WMGGVCCGAA motif consensus sequences that was absent from SnTox1. For SnToxA, the consensus sequence was identified at 218, 364 and 416 bp upstream of the transcriptional start site. The consensus sequence was also observed at two sites in the PtrToxA promoter of Ptr. For SnTox3, the consensus sequence was identified at 679 bp upstream of the transcriptional start site. This consensus sequence was not observed in the promoter region of SnTox1. Therefore, it was hypothesised that WMGGVCCGAA functions as a PnPf2 binding site (Pf2BS). A yeast 1-hybrid (Y1H) assay was performed in order to determine whether PnPf2 can directly interact with the putative binding site represented in the SnToxA promoter. No significant interaction was observed between PnPf2 and four tandem repeats of the Pf2BS (Fig. 5a). Western blot analysis confirmed the presence of the PnPf2 protein indicating that the absence of Y1H interaction was not the result of the lack of protein (Fig. 5b).

Identification of DE TF genes.
We then screened for putative TF genes that were DE between SN15 and pf2-69 from the high confidence DE gene set to explore the possibility that PnPf2 operates indirectly. We limited our search to genes that encode proteins with TF domains found in fungi 28 . A total of 20 DE putative TFs were identified covering both iv and ip treatments. Based on distinct InterPro classifications 24 , this set consisted of five basic leucine zippers, one zinc knuckle, one myc-type, one CCHC-type, one p53-like, one C 2 H 2 , one homeodomain-like, six fungal specific Zn 2 Cys 6 and three unspecified fungal TFs (Table 2). A BLAST search of PHIbase 19 revealed that seven of these DE TF genes have strong matches to other fungal TFs associated with virulence ( Table 2). Three of these belong to the fungal-specific Zn 2 Cys 6 class (SNOG_03490, 07307 and 08440), one homeodomain-like (SNOG_08237) and three basic-leucine zippers (SNOG_04486, 13689 and 16487).  www.nature.com/scientificreports www.nature.com/scientificreports/ RNAseq confirmed SnToxA and SnTox3 down-regulation in pf2-69 but the expression of SnTox1 was significantly higher in SN15 than pf2-69 than our previous observation 9 . The possibility that PnPf2 plays a minor regulatory role in SnTox1 regulation requires further investigation. Culture filtrates derived from pnpf2 mutants caused chlorosis on Snn1 wheat lines although the symptom was slightly weaker than with SN15 9 . Nevertheless, SnTox1 is still strongly expressed in pf2-69 during infection and is sufficient to produce detectable SnTox1 activity in the culture filtrate and confer virulence on Snn1 wheat lines 9 .
It is not known if A. brassicicola uses effectors to modulate host infection. However, Cho et al. 12 identified eight genes that encode small-secreted proteins with effector-like hallmarks positively regulated by AbPf2. Candidate effector genes were identified in this study that showed the same differential expression patterns as SnToxA and have effector-like properties. One of the effector candidates displayed significant sequence similarities to a known effector and pathogenicity factor. SNOG_01146 possesses a glycosyl hydrolase 61 domain and showed amino acid  www.nature.com/scientificreports www.nature.com/scientificreports/ similarity to MoCDIP4. MoCDIP4 was identified as an apoplastic effector secreted by M. oryzae that causes cell death in rice 20 . Moreover, MoCDIP4 also induces cell death in non-host eudicots. In addition, these effectors are small, cysteine rich and expressed highly during early infection. SNOG_15270 is an homolog of Xyn11A which encodes an endo-ß-1,4-xylanase in B. cinerea. Deletion of Xyn11A in B. cinerea caused a significant reduction in virulence and growth on xylan 21 . All 12 candidate effector genes are also present in SN79-1087. Seven of these candidate proteins encode altered protein sequences in SN79-1087, which may explain some difference in pathogenicity. Five proteins possess changes in amino acid residues. It was previously observed that ToxA isoforms differ greatly in necrosis-inducing activities on Tsn1 wheats and affect the speed of asexual sporulation 29 . It is interesting to note that SNOG_02755 and 10736 polypeptides contain short amino acid sequence repeats that are partially deleted in SN79-1087. Several well-studied fungal and oomycete effectors contain repeats that possess functional roles in cellular localisation, host recognition and plant cell wall binding 30 . Additionally, recent studies have indicated that differential expression of effector genes between P. nodorum isolates affect their contributions to SNB of wheat 15,31,32 . The expression of these candidate genes in SN79-1087 requires further study.
GO enrichment revealed that PnPf2 functions as a positive regulator of a large subset of plant CWDEs and proteases during infection. Additionally, the removal of PnPf2 resulted in a general up-regulation in expression of nutrient transporter genes during infection. It is still not known whether this change is caused directly by the absence of PnPf2, or indirectly via another mechanism regulated by PnPf2. Comparative transcriptomic analysis of A. brassicicola identified only 13 genes that encode hydrolytic enzymes including two pectate lyases, were regulated by AbPf2 12 . Deep sequencing used in this study provided a higher resolution insight into CAZyme regulation exerted by the Pf2 Zn 2 Cys 6 class. Quantifying the contributions of plant CWDEs to phytopathogenicity is difficult because many fungal phytopathogens possess expanded gene families that result in functional redundancies 33 . For example, early studies on the causal agent of northern leaf spot of maize Cochliobolus carbonum (eg. [34][35][36][37] ) did not find a clear role for CWDEs in fungal virulence. This is not to imply that CWDEs are dispensable for fungal virulence. It was reported that feruloyl esterases from Valsa mali 38 , a AbPf2-regulated pectate lyase from A. brassicicola 39 and an endo-β-1,4-xylanase from B. cinerea 21 function as virulence factors. Since plant CWDEs deconstruct the plant cell wall and liberate simple carbohydrates for assimilation and growth, it remains to be determined if SN15 can outcompete pf2-69 during co-infection on Snn1 wheats as the former can express a much larger repertoire of extracellular hydrolytic enzymes. RNAseq read counts suggested that pf2-69 accumulated much less biomass than SN15 at three dpi. This is surprising as pf2-69 retained the ability to causes lesions on Snn1 wheat lines comparable to SN15 as previously observed 9 . It is probable that SnTox1 secreted by pf2-69 during infection is the main cause of necrosis rather than the accumulation of fungal biomass at the lesion.
Analysis of the pf2-69 DE gene sets identified three distinct over-represented motifs (Fig. 4). The most notable of these is the WMGGVCCGAA motif associated with genes under PnPf2 positive regulation, as this motif was  www.nature.com/scientificreports www.nature.com/scientificreports/ observed at multiple sites along the Sn/PtrToxA and SnTox3 promoters and also enriched in AbPf2-regulated gene promoters 12 . We hypothesised that it functions as a PnPf2 binding site as it resembles a Zn 2 Cys 6 TFBS 26,27 . However, Y1H assay indicated that PnPf2 did not bind to the motif. This suggests either PnPf2 does not function as a direct regulator of SnToxA, SnTox3 and DE genes through interaction with the WMGGVCCGAA motif, or that necessary PnPf2 post-translational modifications/interactions are not compatible with the Y1H system. It was noted however that six other Zn 2 Cys 6 -type TF genes were differentially expressed between pf2-69 and SN15 (Table 2). Of these, only two were down-regulated but may serve as alternate candidates for direct regulation targeting the WMGGVCCGAA motif. A BLAST search of these against PHIbase revealed pathogenicity-associated functions in fungal homologues. SNOG_03490 is 52% identical to GzZC232 of Fusarium graminearum, the causal agent of fusarium head blight of wheat and is required for full virulence 40 . SNOG_08440 is homologous to a Zn 2 Cys 6 -type TF gene AtrR of Aspergillus fumigatus, an opportunistic fungal pathogen of mammals 41 . AtrR is a regulator of ergosterol biosynthesis pathway genes most notably Cyp51, a target for fungicide control. Deletion of AtrR resulted in impaired fungal growth and attenuated virulence on mice 41 . The other enriched motifs were characteristic of C 2 H 2 binding sites 27 however, only one DE TF of this class was identified -SNOG_15627 (Table 2). SNOG_15627 expression was up-regulated in pf2-69 under iv condition but remained unchanged during ip growth. SNOG_15627 demonstrated weak similarity to a characterised TF in F. graminearum shown to be dispensable for pathogenicity on wheat 40 . As the CTGYGCCGCA motif was enriched in the pf2-69 iv up-regulated gene promoters, it is possible that SNOG_15627 functions as a direct regulator. PnCon7 is the only characterised C 2 H 2 TF in P. nodorum involved in SnTox3-mediated disease and direct regulation 11 . However, the cis-regulatory element of PnCon7 differs to both predicted C 2 H 2 binding sites observed in this study.
We propose a model to explain the role of PnPf2 during early host infection based on evidence observed in this study (Fig. 6). The removal of PnPf2 drastically diminishes effector expression and so restricts the number of hosts on which P. nodorum is virulent 9 (Fig. 6a). Both mutant and wild type strains are able to infect but the reduced ability to produce effectors and cell wall degrading enzymes means that pf2-69 is delayed in accessing bulk nutrients that come from the early stages of cell necrosis (Fig. 6b). The mutant has reduced access to nutrients stored as complex carbohydrates or compartmentalised in plant cells leading to a reduction in growth during host infection. Increased expression of transporter proteins may be an attempt to scavenge freely available nutrients possibly from the apoplastic space 42 (Fig. 6c). In addition, we have identified candidate effector genes that are homologous to virulence factors and effectors in other phytopathogens. It is evident that PnPf2 functions to coordinate the expression of a subset of DE genes identified in this study through other TFs. Studies are currently under way to functionally characterise effector candidates and DE TF genes for their role in effector regulation and pathogenicity on wheat.

Methods
Infection assays. Whole plant infection assay on two week-old wheat seedlings was performed as previously described 43 . Disease severity was visually determined and scored. A score of zero indicates no disease symptoms. A score of nine indicates a fully necrotised plant. Detached leaf infection assays on two-week old wheat cv. Halberd leaves was performed as previously described 43,44 . Biomass analysis using quantitative (Q)-PCR. Q-PCR was to determine fungal biomass from infected wheat. Wheat cv. Halberd was infected with P. nodorum pycnidiospores as described above. Infection was allowed to develop for three days prior to sampling. The inoculated leaf section was excised and collected. Following this, gDNA was extracted using a Biosprint genomic DNA extraction kit (Qiagen, Venlo, Netherlands). Q-PCR was essentially carried out as described in Brouwer et al. 45 using the primer pair alTubulinqPCRf/r and ActinqPCRhp2F/R (Supplementary Table S3).
RNA extraction and handling. RNA isolation and in planta gene expression analyses were performed as described in Rybak et al. 9 using three day post-infected lesions excised from detached wheat. Library construction and sequencing was performed by the Ramaciotti Centre for Genomics (The University of NSW, Australia). Briefly, the TruSeq Stranded mRNA-seq method was used to prepare all libraries. Following this, sequencing was performed on an Illumina HiSeq. 2500 platform (San Diego, CA, USA) to generate 125 bp paired-end reads. Deep sequencing of all in planta samples were carried on individual lanes in the flowcell to ensure maximum sequence data was obtained from low fungal biomass. Samples derived from in vitro growth conditions were multiplexed into a single lane. The experiment was performed with four biological replicates.
RNAseq QC and read trimming. The quality of reads in the FastQ files were assessed using FastQC v0.11.5 (http://www.bioinformatics.babraham.ac.uk/projects/fastqc) before trimming adapter sequences using cutadapt v1.12 46 . Adapter trimmed reads were then filtered into sets belonging to SN15 and wheat using BBSplit v36.67 (https://sourceforge.net/projects/bbmap) using the P. nodorum genome 16 . Fungal reads were aligned to the SN15 genome using STAR v2.5.0a 47 . Novel splice sites were identified in a first pass alignment of the adapter-trimmed reads of all samples combined. Sample reads were then aligned individually using the novel splice sites identified in the first pass.
Determining differential gene expression in RNAseq. Fragments overlapping annotated features in the genome were counted using the SubRead featureCount v1.5.1 program using the union mode 48  www.nature.com/scientificreports www.nature.com/scientificreports/ performed against the null hypothesis LFC 0 ≠ , to be used where greater sensitivity (but lower confidence) was required. Unless otherwise specified, all results refer to thresholded tests ( > LFC 1). Genes that were determined to be DE from tests by all three programs were taken as high-confidence DE sets. For contrasts involving samples with fewer than 10 million fragments (pf2-69 ip), genes from these samples were required to have a minimum of 10 CPM to be considered as differentially expressed in the high confidence sets. Normalised FPKM statistics were obtained using Cufflinks v2.2.1 52 . Genes with mean FPKM > 100 were considered to be highly expressed.
Functional annotation. Functional annotations for existing genes were determined using InterProScan v5.19-58.0 24 . Additionally, carbohydrate active enzymes were predicted using HMMER v3.1b2 (hmmer.org) and dbCAN v5 25 . GO terms 23 for each gene were found from combined dbCAN and InterProScan results, including matches from: Pfam 53 , TIGRFAM 54 , SMART 55 , PIRSF 56 , PANTHER 57 , HAMAP 58 , Prosite 59 , ProDom 60 , PRINTS 61 , and CATH-Gene3D 62 . Likely protein locations were determined using SignalP v4.1 63 , TargetP v1.1 64 , and TMHMM v2.0c 65 . Predicted proteins with a signal peptide and no transmembrane domains outside of the first 27 amino acids were considered to be secreted. Proteins with effector-like properties were determined using EffectorP v1.0 66 and were considered to be effector-like if they were also predicted to be secreted using the criteria above. Candidate genes were searched for in SN79-1087 (NCBI, GCA_002267025.1) using Spaln v2.3.3 67 . Overlapping SN79-1087 genes were extracted and protein sequences were aligned using the needle command using EMBOSS 68 . Functional enrichment of differentially expressed genes. Over-representation of GO terms in high-confidence differentially expressed gene sets were performed using the R package Goseq v1.26.0 69 . Due to differences in the ability of DESeq. 2, EdgeR, and Limma to handle features with few aligned fragments;  Table S3). Repeats of binding sites were cloned into pINT1-HIS3NB. Each construct was linearised, transformed into the yeast strain Y187 (Clontech, CA, USA) and selected on YPAD supplemented with G418. Bait strains were grown on selective media (-His) containing 3-amino-1,2,4-triazole (Sigma-Aldrich, MO, USA). Mating of the yeast bait strains with the prey strains was conducted by mixing the two strains together and grown on YPAD medium. Confirmation of the specific interaction between the bait sequence and the target protein was performed by reintroduce the prey construct into the bait strain. The prey construct pGADT7-p53 was built by cloning partial p53 from pGBKT7-53 (Clontech, CA, USA) into pGADT7. Similarly, PnPf2 was amplified from cDNA using the primer pair Pf2-F2/R3 and ligated into pGADT7.

Data availability
All data generated or analysed during this study are included in this published article (and its Supplementary  Information files)