Infection cushions of Fusarium graminearum are fungal arsenals for wheat infection

Abstract Fusarium graminearum is one of the most destructive plant pathogens worldwide, causing fusarium head blight (FHB) on cereals. F. graminearum colonizes wheat plant surfaces with specialized unbranched hyphae called runner hyphae (RH), which develop multicelled complex appressoria called infection cushions (IC). IC generate multiple penetration sites, allowing the fungus to enter the plant cuticle. Complex infection structures are typical for several economically important plant pathogens, yet with unknown molecular basis. In this study, RH and IC formed on the surface of wheat paleae were isolated by laser capture microdissection. RNA‐Seq‐based transcriptomic analyses were performed on RH and IC and compared to mycelium grown in complete medium (MY). Both RH and IC displayed a high number of infection up‐regulated genes (982), encoding, among others, carbohydrate‐active enzymes (CAZymes: 140), putative effectors (PE: 88), or secondary metabolism gene clusters (SMC: 12 of 67 clusters). RH specifically up‐regulated one SMC corresponding to aurofusarin biosynthesis, a broad activity antibiotic. IC specifically up‐regulated 248 genes encoding mostly putative virulence factors such as 7 SMC, including the mycotoxin deoxynivalenol and the newly identified fusaoctaxin A, 33 PE, and 42 CAZymes. Furthermore, we studied selected candidate virulence factors using cellular biology and reverse genetics. Hence, our results demonstrate that IC accumulate an arsenal of proven and putative virulence factors to facilitate the invasion of epidermal cells.


| INTRODUC TI ON
Fusarium head blight (FHB), caused by Fusarium graminearum, is a devastating disease of cereals including wheat, barley, oats, and rye with large economic impacts (Savary et al., 2012). After infection and colonization of wheat heads, F. graminearum reduces wheat yield by interfering with kernel development and by poisoning the remaining kernels with a cocktail of mycotoxins, such as deoxynivalenol (DON) and zearalenone, rendering them unsuitable for food and feed usage (Takemura et al., 2007). To date, there are no wheat cultivars available that are fully resistant to F. graminearum infection (Mesterhazy, 1995).
Recently, comprehensive transcriptomic analyses of partially resistant and susceptible wheat cultivars inoculated with F. graminearum were performed to better understand the host molecular response to FHB (Biselli et al., 2018;Pan et al., 2018;Wang et al., 2018). Additionally, a transcriptional profiling approach separated symptomless and symptomatic aspects of the FHB infection and defined subsets of F. graminearum genes expressed in a single cereal host species or across two or more cereal hosts (Brown et al., 2017). Additional transcriptomics-based studies have been conducted, focusing on later stages of infection, using wheat coleoptiles, wheat spikes, and maize stalks (Lysøe et al., 2011;Zhang et al., 2012;Zhang et al., 2016;Kazan and Gardiner, 2018). To date, we are lacking information on the initial stages of fungal infection, from conidial germination to fungal growth on the plant surface and penetration into the plant epidermal cells. FHB starts when conidia of F. graminearum adhere to the surface of wheat spikelets with the help of hydrophobin proteins (Quarantin et al., 2019).
We previously showed that the deletion of F. graminearum adenylyl cyclase necessary for cAMP production, as well as overexpression of deoxyhypusine hydroxylase, the second activating enzyme of the eukaryotic translation initiation factor 5A, abolish the formation of IC and, therefore, the ability of the fungus to infect wheat (Bormann et al., 2014;Martinez-Rocha et al., 2016).
Appressoria of Magnaporthe oryzae and Colletotrichum species are melanized single cells emerging directly from conidial germ tubes (Perfect et al., 1999;Wilson and Talbot, 2009). They are produced due to the perception of a hydrophobic surface and are morphologically very different from F. graminearum complex IC that differentiate from specialized RH (Boenisch and Schäfer, 2011). Unicellular appressoria have been widely studied at both the histological and the molecular level, making it most probably the best examined fungal structure (O'Connell et al., 2012;Soanes et al., 2012). Several plant pathogens such as Botrytis cinerea, infecting approximately 200 plant species, Sclerotinia sclerotiorum, causing white mould mostly on vegetables, and Rhizoctonia solani, a wide host range pathogen, penetrate their host plants using complex appressoria similar to F. graminearum IC (Armentrout et al., 1987;Backhouse and Willetts, 1987;Garg et al., 2010). Although IC have been histologically described several times in recent decades (Dodman et al., 1968;Nikraftar et al., 2013), a molecular description of their development is still pending.
In this study we removed the infecting fungal mycelium from the underlying wheat floral tissue and separated RH from IC using laser capture microdissection. We compared transcriptional changes occurring in these specialized fungal cells (RH and IC) to mycelium (MY) grown in complete medium (CM). This transcriptomic analysis allowed the identification of fungal transcripts specifically detected in the specialized fungal structures for epiphytic growth (RH) and IC.
In more detail, we analysed genes encoding carbohydrate-active enzymes, putative secreted effector proteins, and secondary metabolite biosynthetic enzymes. In particular, transcripts detected in IC encode putative virulence factors prior to the invasion of epidermal cells.

| Identification of F. graminearum epiphytic growth on wheat palea tissue
Under favourable conditions, conidia germinate and differentiate into specialized RH that epiphytically colonize the surface of wheat ( Figure 1a). RH differentiate into IC that are complex appressoria made of agglomerated hyphae (Figure 1a (Table S1), were used. The expression patterns of a set of five genes differentially expressed during infection according to RNA-Seq data were confirmed using quantitative reverse transcription PCR (RT-qPCR), proving the reliability of the experimental setup ( Figure S1). The validated genes are relevant to our study (trichodiene synthase FgTRI5, polyketide synthase FgPKS12, and the putative effector 1 FgPE1) or are required for wheat virulence (GABA transaminases FgGTA1 and FgGTA2; Bönnighausen et al., 2015).

| Global gene expression profile of F. graminearum during initial infection of wheat floral tissue
Expression patterns of 13,826 predicted genes of F. graminearum were compared between MY, RH, and IC (Data S1 and Figure 2a). A total of 12,089 (87%), 11,778 (85%), and 12,504 (90%) transcripts were detected in MY, RH, and IC, respectively (Figure 2b); 870 (6%) transcripts were not detected in any cell type (Figure 2b). A comparison of the differentially expressed genes between the three cell types was performed ( Figure 2c). Genes with a log 2 fold change (log 2 FC) above the threshold of +2 were classified as "up-regulated", while genes with a log 2 FC below −2 were "down-regulated" and genes with a log 2 FC between −2 and +2 were "nonregulated". The major differences were found between IC and MY presenting 839 up-regulated genes and 2,709 down-regulated genes ( Figure 2c).
Infection regulated genes were identified by comparing RH and IC expression of genes to MY (Table 1). In total, we identified 3,916 infection regulated genes, further dissected into infection up-regulated (982) and infection down-regulated genes (2,934).
Most of the infection up-regulated genes were up-regulated in both RH and IC (485) or in IC (354, Figure 3a and Table 1). Fewer genes were specifically up-regulated in RH (143). Similarly, most infection down-regulated genes were down-regulated in both RH and IC (1,536) or in IC (1,173) and fewer genes were specifically downregulated in RH (225, Figure 3b and Table 1 genes in RH compared to MY (Table S2) and the top 50 up-regulated genes in IC compared to MY (Table S3).
While heat maps for SMC, DH, TMR, and PE showed a diverse regulation pattern, genes grouped in nonsecreted proteins, TF, HM, and PK were under-represented in IC. Heat maps for secreted proteins, CAZymes, and ROS showed more over-represented genes in IC compared to MY or RH ( Figure S2).  (Table S4).
The highest transcriptional changes on infection regulated secreted proteins were on CAZymes, ROS, and PE (Table S4) (Table S4). In addition, 32 of the 67 known SMC were differentially expressed on infection, with 19 SMC being down-regulated and 13 SMC being up-regulated (Table 2). In conclusion, plant colonization triggers a wide range of changes in the F. graminearum transcriptome.

| Transcriptional changes specific for RH and IC
A total of 573 genes were differentially expressed in IC compared to RH, of which 238 were down-and 335 up-regulated in IC ( Figure 2c).
Some 248 genes were infection and IC up-regulated compared to RH, while only 44 were infection and RH up-regulated genes ( Table S5).
The main gene families corresponding to infection up-regulated genes in IC were SMC (38 genes, corresponding to 16 clusters), CAZymes (42), ROS (34), and PE (33) ( Table 3). This comparison shows that IC expressed a specific set of putative virulence factors (CAZymes, PE, and SMC), which we further investigated.

| IC are enriched in plant cell walldegrading enzymes
CAZymes are proteins involved in cleavage, modification, or synthesis of glycosidic bonds (Lombard et al., 2014). The  (Table S7). Among the 140 infection up-regulated degradative CAZymes, 35 PCWDC and four FCM were specifically up-regulated in IC, while only three genes (with yet unknown pathway annotation) were specifically up-regulated in RH (Table S8).
To test whether the production of PCWDC by IC has an impact on

| IC are enriched in infection up-regulated putative effectors
Secreted fungal effector proteins modulate host immune response to facilitate infection (Petre and Kamoun, 2014). Here, PE proteins were defined as secreted proteins, without transmembrane domains and a maximum size of 1,000 amino acids. Using this definition, 524 PE were identified (Data S4 and Figure 5a). Furthermore, 199 PE smaller than 200 amino acids and with a cysteine content higher than 2% were identified (Table S9). PE were classified as known effectors (PE: 44) containing previously identified domains and/or were defined as effectors in fungi or bacteria (Table S10) Fusarium-specific and 10 F. graminearum-specific (Table S11). Genes were categorized as infection down-regulated, infection up-regulated or not regulated by comparing runner hyphae (RH) and infection cushions (IC) to mycelium grown in complete medium (MY). The genes were also separated into differentially expressed genes in RH and IC. To determine the subcellular localization, FgPE1 was translationally fused to mCherry (FgPE1 Prom ::FgPE1::mCherry, Figure S4a) and transformed into a wild-type-(WT) strain expressing cytoplasmic green fluorescent protein (GFP) constitutively. A high level of FgPE1-mCherry was found in conidia produced on wheat medium ( Figure S4b,c). In addition, FgPE1-mCherry localized around old hyphae but not in young growing hyphae, both on palea (6 dpi; Figure   S4f-h) and in wheat medium (1 and 3 dpi; Figure S4i FgPE1 was deleted by gene replacement (Figure S6a,b). In virulence assays on wheat the deletion mutants as well as an ectopic mutant and WT exhibited full infection, indicating that FgPE1 is dispensable for virulence ( Figure S6c,d).

| Wheat infection triggers secondary metabolite production
In this study, 53 of 67 SMC (Sieber et al., 2014) were expressed in any of the tested conditions ( Table 2). The remaining 14 SMC, including those involved in biosynthesis of zearalenone (C15), butenolide (C49), fusarin C (C42), and fusarielin (C60), were neither expressed in CM nor during infection (  Figure 7a) and fusaoctaxin A (C64, wheat spikes revealed less mycelial growth inside spikelets inoculated with the ∆tri5 mutant compared to the ones inoculated with the WT (Figure 7b). For molecular quantification, fungal DNA from inoculated spikelets was extracted and the relative amount measured by quantitative PCR (qPCR) as previously described (Voigt et al., 2007). Results revealed that ∆tri5 grew 60% and 75% less than the WT at 3 and 5 dpi, respectively ( Figure 7c). Hence, DON facilitates rapid colonization of plant tissues at the initial stage of infection.

| Aurofusarin is a RH-specific antibiotic active against a wide range of microorganisms
The single SMC specifically up-regulated in RH is involved in aurofusarin biosynthesis (C13) and is strongly down-regulated in IC (Table 2 and Figure 8a). To test whether aurofusarin could act as an antibiotic against microbial competitors growing on the wheat floral tissue's surface, the toxicity of aurofusarin on bacteria and fungi was assessed using mycelium extracts from either WT or the aurofusarindeficient Δpks12 mutant ( Figure 8b and Table S12). The presence or absence of aurofusarin in extracts was determined by LC-MS ( Figure   S7). WT extract was highly toxic to the gram-positive bacteria Bacillus subtilis NG234 were insensitive (Figure 8b). WT extract was also highly toxic to Pyrenophora teres (100% GI), Candida albicans (100% GI), and Pichia pastoris (85% GI), while moderately toxic to Candida parapsilosis (50% GI) and Saccharomyces cerevisiae (50% GI). F. graminearum and, closely related, Nectria haematococca were totally insensitive to WT extracts ( Figure 8b). Extracts from Δpks12 were completely nontoxic to all tested organisms, demonstrating that aurofusarin is indeed responsible for the observed toxicity of WT extract to bacteria and fungi.

| D ISCUSS I ON
Although FHB is a devastating cereals disease that occurs worldwide, The colour-coded heat map considers the total of infection up-regulated genes regarding their regulation in infection cushions (IC) and runner hyphae (RH) from blue (highest) to red (lowest). Non-SP, nonsecreted proteins; SP, putative secreted proteins; TF, transcription factors; TP, transporter proteins; HM, histone-modifying proteins; PK, protein kinases/phosphatases; DH, dehydrogenases; CAZyme, carbohydrate-active enzymes; PE, putative effector proteins; ROS, proteins related to reactive oxygen species; TMR, transmembrane receptors; no, genes without any annotation. Due to several genes belonging to more than one gene family, there is a higher total number of genes in this table than in the annotated genome (Data S1 and Figure 2b,c).
The majority of up-regulated genes in RH or IC compared to MY encode hypothetical proteins with unknown function, indicating a requirement for characterization of such proteins (Tables S2 and   S3). The next more up-regulated genes encode for CAZyme degradative enzymes, suggesting preparation to break the plant cell wall (Tables S2 and S3) propanoid pathway, which are also known to be involved in plant defence (Dixon and Paiva, 1995;Lang et al., 1991). Cell wall-bound ferulic acid is the major substance causing blue light emission in grasses like wheat (Lichtenthaler and Schweiger, 1998 Zhang et al., 2012). DON inhibits the eukaryotic translational machinery and is essential for colonization of the spike, with massive induction during colonization of the developing caryopses and the rachis node (Ilgen et al., 2009). Importantly, both SMC are up-regulated in IC clearly in preparation for the following colonization steps. DON-deficient mutants fail to cross the rachis node (Proctor et al., 1995;Maier et al., 2006;Ilgen et al., 2009), which is accompanied by plant cell wall thickening and jasmonate-related F I G U R E 5 Fusarium graminearum secretome prediction and effectors selection. (a) The secreted proteins were predicted using TargetP and SignalP for classic secretion (CS), or secretomeP and wolfPSORT for non-classic secretion (NS). Proteins without transmembrane domains (w/o TMHs) were predicted using TMHMM. We identify putative effector proteins (PE) similar to effectors with known domains using IPRO or PFAM prediction domains (Table S10). Proteins with unknown domains, not belonging to a secondary metabolism gene cluster, with fewer than 1,000 amino acids, were defined as putative unknown effectors (Table S11) The cluster for aurofusarin biosynthesis is the only one specifically induced in RH. Aurofusarin is a red pigment produced by different Fusarium species, belonging to polyphenol, more accurately bis-naphthopyrone pigments (Frandsen et al., 2006;Xu et al., 2019). In a previous study, deletion of the F. graminearum pks12 gene (FGSG_02324) led to a loss of red pigment, a higher growth rate, and 10-fold more conidia production than WT but had no impact on pathogenicity on wheat and barley (Malz et al., 2005). Recently, aurofusarin has been described to inhibit Lactobacillus and Bifidobacterium, but not E. coli (Sondergaard et al., 2016). Excitingly, it has been described as an antifeedant that accumulates in high amounts to protect Fusarium fungi from a wide range of insects (Xu et al., 2019). The microbiology of the phyllosphere is, in general, not very well understood, but it seems safe to assume that RH of F. graminearum ward off other microbes during colonization of the palea's surface. Among the bacteria found in the microbiome of wheat spikes are Pseudomonas, Bacillus, Janthinobacterium, and Actinomycetes (Chen et al., 2018). In this study, we showed that aurofusarin is an inhibitor of different bacterial and fungal species, among them yeast, including most notably the widespread human pathogen C. albicans. Another polyphenol pigment found in F. graminearum is bostrycoidin purpurfusarin, which is also known to have antibiotic properties against C. albicans (Frandsen et al., 2016). Further research will show if these secondary metabolites could improve the fight against this widespread human pathogen.
Up-regulated in RH and IC are two infection up-regulated iron-chelating siderophores, triacetyl fusarin and malonichrome, which are necessary for virulence (Oide et al., 2015). A third infection up-regulated iron-chelating siderophore, ferricrocin, is important for sexual development but not for virulence (Oide et al., 2015), explaining why we found this metabolite down-regulated during infection. Fourteen SMCs were not at all expressed, including zearalenone (C15), fusarin C cluster also shows a significant up-regulation in IC and RH compared to mycelium grown in complete medium (MY). Asterisks indicate significant up-regulation for log 2 FC(FPKM + 1) ≥ 2 or down-regulation log 2 FC(FPKM + 1) ≤ −2. (b) Cross-section images show wild-type (WT)-green flourescent protein (GFP) mycelia copiously grown in inoculated spikelet and across the rachis node, while Δtri5-GFP mutant grows scarcely and is unable to cross the rachis node. Images are a composition of three pictures taken at 5 days post-inoculation (dpi).
White arrows indicate the rachis node. Scale bar = 2 mm. (c) Δtri5-GFP fungal growth is 60% and 75% less than the WT-GFP at 3 and 5 dpi, respectively. Error bars indicate ±SD calculated from data of three independent experiments and three experimental replicates (n = 9). Significance with respect to WT, ****p < .0001 (calculated with ANOVA-Bonferroni-Holm) 2012). F. graminearum is a pathogen with a variety of hosts such as wheat, barley, oats, rye, maize, and soybean (Savary et al., 2012;Sella et al., 2014), and many transcripts not detected might be necessary for specific colonization of such hosts (Harris et al., 2016). For instance, the SMC DON necessary for wheat infection, and highly transcribed under our study conditions, does not seem to act as a virulence factor on barley (Maier et al., 2006). Therefore, the lack of transcript detection could be due to specificity or redundancy on their protein function.

F I G U R E 8
The aurofusarin cluster C13 is up-regulated in runner hyphae (RH) and responsible for toxicity to bacteria and fungi. (a) The aurofusarin gene cluster C13 (from left to right: FGSG_02320 -FGSG_02330) is specifically up-regulated in RH compared to infection cushions (IC) or mycelium grown in complete medium (MY). No significant difference in gene expression of C13 cluster was found between IC and MY. Differential expression of genes is given by log 2 FC(FPKM + 1) values. Asterisks indicate significant up-regulation for log 2 FC(FPKM + 1) ≥ 2 or down-regulation log 2 FC(FPKM + 1) ≤ −2. (b) Bioactivity assay using extracts of wild-type (WT) and aurofusarindeficient mutant ∆pks12. For bacteria and yeasts, OD 595 was measured after 16 hr of growth in medium supplemented with either WT extracts, extracts of the aurofusarin-deficient ∆pks12 mutant, or phosphate buffer as a control. For filamentous fungi, dry weight was calculated. The highest value of buffer controls was set to 100% of growth. For every condition and organisms tested, three biological replicates were performed Besides CAZymes and SMC, we identified a large number of putative effector proteins (PE) up-regulated in IC. Previously, Lu and Edwards (2016) identified 190 small (≤200 amino acids) and cysteine-rich (≥2%) secreted proteins as candidate effectors in F. graminearum. Here, we identified 199 proteins with such characteristics (Table S9). However, several studies reported effectors as secreted proteins with very diverse sizes or cysteine content (Kulkarni et al., 2003;Rooney et al., 2005;Djamei et al., 2011;Frías et al., 2011;Sperschneider et al., 2013;Blümke et al., 2014;Tanaka et al., 2014;Jashni et al., 2015;Quarantin et al., 2016). From the 524 identified PEs, 88 were infection up-regulated and of these, 33 were specifically up-regulated in IC. Four PEs specifically up-regulated in IC contained LysM or CFEM domains, being indicative of key proteins necessary during penetration of the host cell by suppressing fungal recognition and manipulating host functions (Mentlak et al., 2012;Zhang et al., 2012;Takahara et al., 2016). The hemibiotroph C. higginsianum transcribes effectors in consecutive waves associated with the transitions in the pathogen's lifestyle .
Hence, it is likely that F. graminearum expresses a different set of effectors during later stages of plant colonization. The study of the 80 unknown plant-induced fungal effector-like proteins (Table S11) could lead to the discovery of new targets for fungal control and even specifically F. graminearum control. In a different approach using a comparative genome analysis, a set of 2,830 F. graminearum genes, presumably associated with pathogenicity, were identified (Sperschneider et al., 2013). We found that roughly 3.9% of these genes (111)  Alternaria alternata (Chruszcz et al., 2012). Recently, an Aa1-like protein, PevD1 from Verticillium dahliae, has been found to interact and inhibit the antifungal activity of GhPR5 cotton plant protein as a strategy to fight the plant defence and promote fungal infection (Zhang et al., 2018). Here, expression results indicated a transcriptional regulation of FgPE1 depending on plant factors. A previous report showed that FgPE1 is highly expressed at different time points (4, 12, 24, 48, and 72 hpi, and 8 and 14 dpi) during infection of wheat spikes and at 8 days in old minimal medium culture (Lu and Edwards, 2016), supporting the hypothesis that FgPE1 expression is regulated by plant factors and nutrient availability and that it is present not only during early infection but all along the infection process.
Replacement of FgPE1 in F. graminearum did not affect virulence according to our infection assays ( Figure S6). We assume a high degree of functional redundancy within the PE gene family. This is in accordance with studies, for example in U. maydis, that, for the most part, failed to identify novel, virulence-specific genes, for example within a pool of potential effector genes (Kämper et al., 2006). FgPE1 is localized at the fungal cell wall. During IC formation, FgPE1 is secreted and at first localized at plant cell walls in the close vicinity around the IC. Once the fungus grows beneath the plant surface, FgPE1 is localized at the fungal-host interface. Effectors with such localization such as ChEC34 and ChEC89 of C. higginsianum  or the CFEM1 protein (FGSG_02077) from F. graminearum (Zhang et al., 2012) usually are necessary for eliciting or suppressing the plant recognition depending on their lifestyle.
Taken together, we found major transcriptional changes between epiphytical and in-culture hyphae. A complex set of virulence-associated factors, comprising plant cell wall-degrading enzymes, secondary metabolites, and effector proteins among others, are synthesized in IC, preparing the fungal hyphae for successful penetration and subsequent colonization of plant tissue. Therefore, IC are arsenals of fungal combat and the genes expressed in them could provide potentially novel targets for Fusarium control.

| E XPERIMENTAL PROCEDURE S
Detailed experimental procedures are described in Methods S1.

| Fungal growth and conidia production
F. graminearum wild type (WT; Fg-8/1; Miedaner et al., 2000) and mutants used and produced in this study were grown, cultured, and transformed as described before (Jansen et al., 2005).

| Preparation of wheat-infected tissue for laser capture microdissection
Detached wheat palea infection assay was prepared according to Boenisch and Schäfer (2011). Mycelium samples were prepared by inoculating 750 conidia of the WT-GFP strain in 50 ml CM and incubated for 3 days. A mycelium piece of about 1 mm in diameter was used for RNA extraction, amplification, and construction of cDNA libraries.

| Laser capture microdissection
Paleae containing RH and IC were prepared by cutting off their upper and lower ends and immediately transferred to absolute ethanol on ice according to previous studies (Goldsworthy et al., 1999;Clément-Ziza et al., 2008). RH and IC were prepared and dissected as mentioned in Methods S1.

| RNA extraction, amplification, and cDNA library construction
RNA extraction, amplification, and cDNA library construction of IC, RH, or MY were performed according to Lê et al. (2005). See Methods S1.

| Purification of cDNA libraries
For the removal of primers, enzymes, and other substances of the process from the cDNA libraries the NucleoSpin Gel and PCR Clean-up Kit (Macherey-Nagel) was used according to the manufacturer's instructions.

| Finalization of cDNA libraries (end-it reaction)
To provide 5′-phosphorylated, blunt-ended cDNAs, the End-It DNA End-Repair Kit (Biozym Biotech Trading GmbH) was used according to a modified protocol. One microgram of the final cDNA libraries of three independent replicates of mycelia, RH, and compound appressoria, respectively, were sent for RNA-Seq analysis.

| RNA-Seq mapping and quantification
RNA-Seq reads were mapped on the reference genome using tophat2 v. 2.0.8. The interval for allowed intron lengths was set to minimum 20 nt and maximum 1 kb (Trapnell et al., 2009). Three highly correlating replicates were used according to the Pearson correlation test (Table S1). We used cufflinks to determine the abundance of transcripts in FPKM (fragments per kilobase of exon per million fragments mapped) and calculated differentially expressed genes using cuffdiff (Trapnell et al., 2009;Trapnell et al., 2012). The gene models were included as raw junctions. The uncorrected p value and the FDR-adjusted p value of the test statistic (q value) were calculated, p and q values per each gene are given in the general part of Data S1. Any given gene of interest can be evaluated by its fold change of transcription and by the resulting p and q values. Genes with a minimum of four-fold increase or decrease in expression (|log 2 of the FPKM values + 1| ≥ 2) between two experimental conditions were considered as regulated.

| Annotations and databases used
The transcriptome data discussed in this publication have been deposited in NCBI's Sequence Read Archive (SRA; https://www.ncbi.

| Generation of knock-out, expression, and localization constructs for FgPE1 mutants
All plasmids were constructed using the yeast recombination method (Colot et al., 2006) and the pRS426 background plasmid (Christianson et al., 1992). Amplification of the ORF, and 5′ and 3′ flanks of the genes of interest was performed using primers shown in Table S15 and genomic DNA extracted from the WT strain. The final constructs were excised with the respective restriction enzymes (Table S16) and used to transform F. graminearum WT or WT-GFP strains. At least two independent mutants were generated and examined. For details see Methods S1.

| Virulence assay: wheat spikes point inoculation and wheat palea infection
Virulence assays where prepared according to Boenisch and Schäfer (2011) and Frandsen et al. (2006). For details see Methods S1.

| Quantification of fungal material within inoculated wheat spikes using qPCR
Genomic DNA of inoculated wheat spikes was isolated using the CTAB method and according to Voigt et al. (2007). For details see Methods S1.

| Fluorescence microscopy
Histological studies of WT-GFP, ∆tri5-GFP, and mutants generated in this study were performed as previously described in Boenisch and Schäfer (2011). For details see Methods S1.

| Scanning electron microscopy
Scanning electron microscopy (SEM) was done with SEM LEO 1525 at 6 kV using detached palea of wheat cultivar Nandu inoculated with 5 µl of 2 × 10 4 conidial suspension of WT-GFP strain and prepared as described (Boenisch and Schäfer, 2011). To identify penetration pores, infection structures were removed from the plant surface of critical point dried paleae using adhesive tape and processed for SEM as previously described (Bormann et al., 2014).

| Extraction of aurofusarin from F. graminearum WT and aurofusarin-deficient mutant Δpks12
Fungal material of the WT strain and the aurofusarin-deficient mutant was harvested after 4 days from 50 ml CM liquid cultures. The respective mycelium was harvested using Miracloth, washed with 100 ml double-distilled water (ddH 2 O) and semi-dried using a filter paper. Around 1 g of mycelium was transferred into a 2 ml tube and supplemented with 1 ml potassium phosphate-buffer (50 mM, pH 7). After addition of two metal pearls (3 mm diameter), the solution was ground for 15 min using a Retsch mill. After centrifugation at 13,000 rpm for 15 min, the extracted supernatant was filter sterilized using a 0.22 µm Millex GP filter.

| Analysis of fungal extracts via LC-MS
The extracts of WT strain and Δpks12 mutant were analysed as described in Methods S1.

| Bioactivity assay
Liquid cultures of the organism listed in Table S12 were used for bioactivity assays. Assays were performed as described in Methods S1.

ACK N OWLED G M ENTS
The project was partially funded by the University of Hamburg, the The authors have declared that no conflict of interests exists.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data produced in this publication have been deposited in NCBI's Sequence Read Archive (SRA), https://www.ncbi.nlm.nih.gov/sra/; accession SUB3191581.