Flippases play specific but distinct roles in the development, pathogenicity, and secondary metabolism of Fusarium graminearum

Abstract The membrane trafficking system is important for compartmentalization of the biosynthesis pathway and secretion of deoxynivalenol (DON) mycotoxin (a virulence factor) in Fusarium graminearum. Flippases are transmembrane lipid transporters and mediate a number of essential physiological steps of membrane trafficking, including vesicle budding, charging, and protein diffusion within the membrane. However, the roles of flippases in secondary metabolism remain unknown in filamentous fungi. Herein, we identified five flippases (FgDnfA, FgDnfB, FgDnfC1, FgDnfC2, and FgDnfD) in F. graminearum and established their specific and redundant functions in the development and pathogenicity of this phytopathogenic fungus. Our results demonstrate that FgDnfA is critical for normal vegetative growth while the other flippases are dispensable. FgDnfA and FgDnfD were found crucial for the fungal pathogenesis, and a remarkable reduction in DON production was observed in ΔFgDNFA and ΔFgDNFD. Deletion of the FgDNFB gene increased DON production to about 30 times that produced by the wild type. Further analysis showed that FgDnfA and FgDnfD have positive roles in the regulation of trichothecene (TRI) genes (TRI1, TRI4, TRI5, TRI6, TRI12, and TRI101) expression and toxisome reorganization, while FgDnfB acts as a negative regulator of DON synthesis. In addition, FgDnfB and FgDnfD have redundant functions in the regulation of phosphatidylcholine transport, and double deletion of FgDNFB and FgDNFD showed serious defects in fungal development, DON synthesis, and virulence. Collectively, our findings reveal the distinct and specific functions of flippase family members in F. graminearum and principally demonstrate that FgDnfA, FgDnfD, and FgDnfB have specific spatiotemporal roles during toxisome biogenesis.


| INTRODUC TI ON
Fusarium head blight (FHB), caused predominantly by Fusarium graminearum, is an economically devastating disease of a wide range of cereal crops, including wheat and barley (Dean et al., 2012). This disease not only reduces yield and seed quality, but also poses a great risk to human and animal health owing to its ability to contaminate grains with mycotoxins such as deoxynivalenol (DON), which remains the most frequently detected mycotoxin in contaminated cereal grains (Audenaert et al., 2014;Chen et al., 2019). The major approach for controlling FHB today is the use of chemical fungicides due to the unavailability of resistant wheat cultivars. Although the application of some commercial fungicides such as azoxystrobin is effective in controlling FHB, the chemicals trigger DON biosynthesis at sublethal concentrations, while others like tebuconazole and the novel cyanoacrylate fungicide phenamacril (JS399-19) effectively suppress DON production in addition to their FHB-controlling property (Simpson et al., 2001;Magan et al., 2002;Chen and Zhou, 2009;Zhang et al., 2015). However, there are limited varieties of effective fungicides used in controlling FHB and the available ones are likely to bring a high risk of fungicide resistance (Chen and Zhou, 2009;Yin et al., 2009;Willyerd et al., 2012). As such, there is the need to uncover more target pathways for developing more effective fungicides and reducing their subsequent resistance by the pathogen. A clear understanding of the key regulatory processes for DON production and for F. graminearum pathogenicity is therefore important for efficient management of this disease.
There have been extensive genetics and biochemical studies on the biosynthesis of DON and its derivatives in Fusarium (Proctor et al., 2018). In F. graminearum, the biosynthetic enzymes required for DON production are encoded by 15 TRI genes, which are located on different chromosomes, including a gene cluster consisting of 12 core TRI genes on chromosome 2, two TRI genes (TRI1 and TRI16) on chromosome 1, and a single gene (TRI101) on chromosome 3 (Merhej et al., 2011;Tang et al., 2018). However, only a few studies so far have addressed the cellular processes involved in DON biosynthesis and export in F. graminearum. Previous studies in Penicillium chrysogenum and Aspergillus parasiticus showed that these secondary metabolite (SM)-producing fungi possess a conserved and compartmentalized SM biosynthetic pathway (Chanda et al., 2009;Fernandez-Aguado et al., 2014;Kistler and Broz, 2015). This compartmentalization has also been established recently in F. graminearum in relation to DON biosynthesis, where the enzymes involved in DON production have been analysed (Menke et al., 2013;Boenisch et al., 2017;Tang et al., 2018). Menke et al. first demonstrated that Tri4 and Tri1 (the proteins involved in the early and late steps of DON biosynthesis) colocalize in a vesicle called a toxisome that is presumed to be the site of trichothecene biosynthesis (Menke et al., 2013). Boenisch et al. found that growing F. graminearum in trichothecene biosynthesis induction (TBI) medium reorganizes the fungal endoplasmic reticulum (ER) to form perinuclear and peripheral structures, and Tri1 and Tri4 colocalize on these structures, suggesting that toxisomes are formed from the ER (Boenisch et al., 2017;Chen et al., 2019). Tri12, a major facilitator superfamily (MFS) transporter in F. graminearum, localizes to the plasma membrane, vacuole, and small (c.1 μm) motile vesicles in the fungal cells in TBI medium; the motile vesicles containing Tri12 may accumulate DON and transport it to the vacuole for storage or the plasma membrane for export via exocytosis (Menke et al., 2012(Menke et al., , 2013. Based on the above reports, we hypothesize that in F. graminearum the membrane trafficking system is important for DON biosynthesis and secretion. In  (Tang et al., 2018). Zheng et al. have characterized all the 11 F. graminearum Rab GTPase proteins by live-cell imaging and genetic analyses, and shown that they are involved in DON production (Zheng et al., 2015). In another study, the SNARE homolog FgVam7 was found to positively regulate the expression of the DON biosynthesis genes TRI5, TRI6, and TRI101, and subsequently DON production .
A striking aspect of eukaryotic membranes is the uneven distribution of different kinds of phospholipids (membrane asymmetry) across the bilayer, which is essential for proper architecture of the biological membranes (Graham, 2004;Panatala et al., 2015). Flippases are responsible for the formation and adjustment of membrane asymmetry and the proteins responsible for flippase activity are type IV P-type ATPases (P4-ATPases) (Lee et al., 2015). The human genome contains 14 flippases, and mutations in some flippases result in some genetic disorders (Lee et al., 2015). In Arabidopsis thaliana, 12 proteins constitute the lipid flippase family, and they are responsible for the plant's adaptation to temperature changes, defence responses, and so on (Nintemann et al., 2019). Saccharomyces cerevisiae has five flippases that mediate a number of steps in membrane trafficking, including vesicle budding, charging, and protein diffusion within membranes (Pomorski et al., 2003;Takeda et al., 2014).
Despite poor understanding of the roles of flippases in filamentous fungi, some evidence has shown that the proteins may be critically important for fungal growth and pathogenicity. In Aspergillus nidulans, Schultzhaus et al. found that the flippase AnDnfD is essential for conidiation, and that AnDnfA and AnDnfB work complementarily in the regulation of growth and phosphatidylserine asymmetry K E Y W O R D S DON production, flippase, Fusarium graminearum, pathogenicity, secondary metabolism (Schultzhaus et al., 2015(Schultzhaus et al., , 2019. In the opportunistic fungal pathogen Cryptococcus neoformans, the flippase Apt1 is involved in stress tolerance, polysaccharide secretion, and virulence (Hu and Kronstad, 2010;Rizzo et al., 2014). In the rice blast fungus Magnaporthe oryzae, the biological functions of two flippases, MoPde1 and MoApt2, have been characterized, and both proteins are involved in fungal virulence (Balhadere and Talbot, 2001;Gilbert et al., 2006). In F. graminearum recent studies identified FgDnfB and FgNeo1 (FgDnfD homologs) as flippases, and FgDnfB plays a minor role in fungal vegetative growth, polarity maintenance, and conidiation , while FgNeo1 is important for asexual/sexual developments and virulence in F. graminearum (Li et al., 2019), but the remaining family proteins remain unknown. In addition, the roles of flippases in the biosynthesis of secondary metabolites have not been established in filamentous fungi.
In the present study we carried out a BLAST search using the amino acid sequences of the S. cerevisiae flippases Dnf1, Dnf2, Drs2, Dnf3, and Neo1 against the F. graminearum genome and identified five flippases that were named FgDnfA, FgDnfB, FgDnfC1, FgDnfC2, and FgDnfD. Subsequently, we constructed both single and double gene deletion mutants for the various flippase genes and systematically analysed their functions. Our findings reveal not only the active involvement of the flippases in growth, development, and pathogenesis, but also their distinct regulatory roles in DON biosynthesis of F. graminearum.

| FgDnfA is crucial for vegetative growth in F. graminearum
We generated five hits (FGSG_08595, FGSG_06743, FGSG_09020, FGSG_00595, and FGSG_05149) from a BLAST search using the amino acid sequences of the S. cerevisiae flippase proteins Dnf1, Dnf2, Drs2, Dnf3, and Neo1, respectively, against the F. graminearum genome (https://blast.ncbi.nlm.nih.gov/Blast.cgi). To use the canonical format of gene nomenclature, we renamed each of the hits FgDNFA, FgDNFB,FgDNFC1,FgDNFC2,and FgDNFD,respectively,in accordance with a phylogenic analysis and naming convention in A. nidulans (Schultzhaus et al., 2015). Our phylogenetic analysis suggests that flippases from S. cerevisiae and other filamentous fungi, including A. nidulans, F. graminearum, M. oryzae, and Neurospora crassa, could be classified into four subgroups ( Figure S1a). Of these, each subgroup contains one F. graminearum ortholog except subgroup 3, which has the two FgDnfC members, suggesting that the FgDNFC gene has undergone duplication relative to its yeast homolog. To investigate the function of the five flippase genes in F. graminearum we used a homologous recombination strategy to generate their respective gene deletion mutants, except for FgDNFB, which has been generated from our previous study . The resulting hygromycin-resistant transformants were screened by PCR (Table S1) and Southern blot ( Figure S1b). In addition, we generated a complemented strain for each of the single-gene deletion mutants by transforming the full DNA sequences (tagged with green fluorescent protein [GFP] at their C-termini) of the deleted genes into the protoplasts of the respective mutants.
Of all the five single-gene deletion mutants generated, only the ΔFgDNFA mutant grew significantly more slowly than the wild-type strain on both complete medium (CM) and minimal medium (MM) ( Figure 1a, Table 1). Deletion of FgDNFB (ΔFgDNFB) also resulted in a slight but insignificant reduction in growth rate compared to the wild-type strain. However, deletion of the other three flippase genes did not display any clear growth defects on the CM or MM plates ( Figure 1a). These results show that among the five flippase proteins in F. graminearum, FgDnfA is crucial for vegetative growth of the fungus.

| FgDnfA and FgDnfB are important for cell membrane-associated stress response in F. graminearum
Because previous studies have shown that flippases may be involved in phosphatidylserine asymmetry, and that phosphatidylserine is a core component of the cell membrane and is necessary for sensing environmental changes (Hankins et al., 2015;Schultzhaus et al., 2015), we decided to investigate the growth of the flippase gene deletion mutants and their corresponding complemented stains under cell membrane stress conditions. From these assays, we found that both ΔFgDNFA and ΔFgDNFB mutants showed increased tolerance to membrane stress due to sodium chloride (NaCl), Congo red (CR), and calcofluor white (CFW) compared to the wild-type strain and the other flippase gene deletion mutants ( Figure S2). In addition, the ΔFgDNFA mutant showed higher tolerance to sodium dodecyl sulphate (SDS)-induced stress ( Figure S2). These results suggest that FgDnfA and FgDnfB are important for cell membrane-associated stress responses in F. graminearum.

| FgDnfA, FgDnfB, and FgDnfD play specific roles in regulating sexual and asexual reproductions in F. graminearum
To investigate the roles of the flippases in fungal reproduction, we tested conidiation and perithecia formation of the mutants as compared to the wild-type and complemented strains. Again, we found that only ΔFgDNFA produced a significantly lower amount of conidia (0.03 × 10 6 conidia/ml) than the wild-type strain (1.32 × 10 6 conidia/ ml) in carboxymethylcellulose (CMC) medium (Table 1), while deletion of the other four flippase genes did not affect the number of conidia of the mutants when compared to the wild-type (Table 1).
F. graminearum conidia are formed in clusters on bottle-shaped phialides or singly formed on short hyphal branches where the latter style is less efficient than the former (Wang et al., 2011;Chen et al., 2016). In this study, we tracked the conidiogenesis in each mutant on Spezieller Nährstoffarmer agar (SNA) and found that ΔFgDNFA was unable to produce clustered conidia on phialides but formed conidia directly on short hyphal branches, which could account for the reduced conidiation in ΔFgDNFA (Figure 1b). We checked the morphology of the conidia obtained from the various strains and found that the conidia produced by ΔFgDNFA, ΔFgDNFB, and ΔFgDNFD  were smaller, with fewer septa than the wild-type and complemented strains ( Figure 1b and Table 1), suggesting that the flippases FgDnfA, FgDnfB, and FgDnfD regulate conidial morphology. The germination ability of the conidia from the various strains was tested in 2% sucrose water. The conidial germination of ΔFgDNFA was delayed when compared to the wild-type and the complemented strains ( Figure S3), indicating that FgDnfA plays important roles not only in conidiation and conidial morphology, but also in temporal conidial germination in F. graminearum.
To investigate the roles of the flippases in the sexual reproduction of F. graminearum, the wild-type, mutants, and complemented strains were grown on carrot medium plates under black-light conditions to induce sexual reproduction, which is evident by perithecia formation. The results showed that ΔFgDNFA only produced several small, nonascus perihelia on the plates (Figure 1c). ΔFgDNFB also produced similar but a bit larger perithecia than ΔFgDNFA, but the size and number of these perithecia were still smaller and fewer than the perithecia from the wild-type strain, and no asci were found in them (Figure 1c,d). ΔFgDNFD showed better sexual reproduction ability than ΔFgDNFA and ΔFgDNFB, but it produced significantly smaller perithecia than the wild-type and complemented strains Activities of three virulence-related extracellular enzymes, including endoglucanase, amylase, and polygalacturonase in the wild-type, ΔFgDNFA, ΔFgDNFB, ΔFgDNFC1, ΔFgDNFC2, ΔFgDNFD, and complemented (-C) strains after 7 days of induction in Czapek's medium with bran were detected by the 3,5-dinitrosalicylic acid method. One unit of enzymatic activity is defined as 1 μg/min reducing glucose released from the substrate at pH 4.6 and 50°C

| FgDnfA and FgDnfD play crucial roles in the pathogenicity of F. graminearum
To analyse the roles of the flippases in F. graminearum pathogenicity, infection assays on wheat heads and wheat coleoptiles were conducted. As shown in Figure 2a, deletion of FgDNFA almost abolished virulence on both wheat heads and wheat coleoptiles. ΔFgDNFD appeared more virulent than ΔFgDNFA, but the virulence was highly reduced as compared to the wild-type and complemented strains.
The other three flippase gene deletion mutants displayed similar virulence to the wild-type strain. Similar results were also observed on wheat coleoptiles (Figure 2a), suggesting that FgDnfA and FgDnfD are important for F. graminearum pathogenesis.
Extracellular enzymes and DON are important virulence effectors for the pathogenicity of F. graminearum (Ma et al., 2013), we thus first detected the secreted endoglucanase, amylase, and polygalacturonase activity in wild-type, flippase mutants, and complemented strains. Our results showed that deletion of FgDNFA led to reduced endoglucanase and amylase activities but did not affect the activity of polygalacturonase, while deletion of other flippases did not affect the activities of these three enzymes (Figure 2b). Therefore, these data indicate that the fillipase FgDnfA has a specific role in the regulation of some extracellular enzymes that are responsible for the virulence of F. graminearum.

| FgDNFA and FgDNFD positively regulate DON production while FgDNFB is a negative regulator in F. graminearum
To determine the roles of the flippases in DON biosynthesis, DON production was induced and quantitatively assayed in the wild-type, mutants, and complemented strains by ELISA. Deletion of FgDNFA and FgDNFD caused a significant decrease in DON production, where the mutants were able to produce only 1.6% and 12.5% of the DON produced by the wild-type strain, respectively (Table 1).

| FgDnfA, FgDnfB, and FgDnfD are involved in toxisome biogenesis
Previous reports demonstrated that Tri4 and Tri1 proteins (which are cytochrome P450 oxygenases) colocalize in some spherical structures called toxisomes, which emanate from reorganized ER during trichothecene induction and are presumed to be the sites for trichothecene biosynthesis (Boenisch et al., 2017;Chen et al., 2019). We therefore decided to check the subcellular localization of Tri1-GFP in ΔFgDNFA, ΔFgDNFB, ΔFgDNFD, and the wild-type strains, respectively. We expressed the Tri1-GFP construct in the protoplasts of the above strains and, after incubation in TBI medium for 48 hr, the cellular location of FgTri1-GFP was observed in the transformants of each strain. In the wild-type strain, the FgTri1-GFP localized to some spherical and crescent structures in the fungal hyphae, and these structures colocalized with the ER (marked by ER-specific dye) (Figure 4a), which is consistent with the character-

| The different flippases have different cellular localizations in F. graminearum
Protein domain prediction analysis showed that each flippase in F. graminearum has more than seven transmembrane motifs distributed along the protein ( Figure S4), which is consistent with their predicted roles as transporters on the membrane.  (Table 1). By evaluating the phenotypes of these double-gene deletion mutants, we found that the mutants showed similar phenotypes to the single-gene deletion mutants of FgDNFA, FgDNFB, or FgDNFD in vegetative growth, reproduction, virulence, and DON production process (Table 1 and Figure S7). Double-gene deletion mutants of FgDNFC1 and FgDNFC2 also showed similar phenotypes to the wild-type strain (Table 1 and Figure S7). This  (Figure 6a) and even when compared to the single deletion mutants of the two genes ( Figure 1a).
We also observed that hyphae from the ΔFgDNFBD double mutant were highly branched and curled (Figure 6a). In addition, we found that ΔFgDNFBD could not produce conidia and perithecia, and almost lost virulence on wheat heads (Figure 6b). DON production was also highly reduced in this mutant (Table 1)

| D ISCUSS I ON
The present study systematically analysed the functions of the five flippases in F. graminearum using genetic, biochemical, and cell bi- Flippases are ATP-dependent transporters belonging to the P4 subfamily of P-type ATPases (P4-ATPases), which are only present in eukaryotic organisms (Daleke, 2007). Among the five flippase genes in S. cerevisiae, only NEO1 is essential for viability; deletion of any one of the other four genes is not lethal, but the quadruple mutant (∆DRS2∆DNF1∆DNF2∆DNF3) is not viable, indicating that they have overlapping functions (Daleke, 2007). In filamentous fungi, four flippases have been found in the genome of A. nidulans and all of them, including the homolog gene for NEO1, are not essential genes (Schultzhaus et al., 2015(Schultzhaus et al., , 2019. However, double-gene deletion for AnDNFA (homolog gene of DNF1 and DNF2) and AnDNFB (homolog gene of DRS2) or AnDNFB and AnDNFD (homologue gene of NEO1) was lethal for the growth of A. nidulans, suggesting that functional redundancy of flippases also exists in A. nidulans (Schultzhaus et al., 2015(Schultzhaus et al., , 2019. In this study, five flippase genes were identified in the genome of F. graminearum. We also found functional redundancy among these genes. Similar to A. nidulans, FgDNFA ( Study of the flippases in A. nidulans revealed their functions in the regulation of vegetative growth and asexual sporulation (Schultzhaus et al., 2015(Schultzhaus et al., , 2019. Here, we also confirmed the involvement of flippases in growth and conidia production in F. graminearum ( Figure 1 and Table 1) is required for ascospore discharge in an ion-dependent manner (Li et al., 2019). However, in this study we found that the deletion mutant of FgDNFD produces smaller perithecia than the wild-type strain and does not produce ascospores (Figure 1), indicating that FgDnfD is essential for maturation of perithecia in F. graminearum.
A previous work demonstrated that perithecia formation requires lots of energy and precursors, and lipids may be the key resources supporting this cellular process (Lee et al., 2011). In the metabolic The wild-type strain, ΔFgDNFB, ΔFgDNFD, and ΔFgDNFBD mutants were treated with 7-nitro-2-1,3-benzoxadiazol-4-yl-phosphatidylserine (NBD-PS) and observed under a confocal microscope. Bar = 10 μm pathway of the neutral lipid triacylglycerol in S. cerevisiae, phosphatidylinositol, including phosphatidylserine, phosphatidylethanolamine, and phosphatidylcholine, are the intermediate products (Wang, 2015). In this study, we found that ΔFgDNFA mutant displays suppressed growth on CM but the growth improved markedly on MM ( Figure 1a and Table 1). This is similar to the growth pattern of the deletion mutant of Fg10302, a homolog of phosphatase gene (NEM1) involved in the triacylglycerol metabolic pathway in S. cerevisiae (Yun et al., 2015), suggesting that deletion of FgDNFA may also affect the metabolism of triacylglycerol. Based on these results, we hypothesize that flippases in F. graminearum may be involved in the biosynthesis of phospholipids, which in turn makes them important for sexual reproduction in the phytopathogenic fungus.
In fungal pathogens, flippases are known to be required for effective pathogenesis. Documented reports in the rice blast fungus M. oryzae, the opportunistic fungal pathogen C. neoformans, and F. graminearum suggest that flippases have conserved roles in the regulation of pathogenicity, but different homologs display functional diversity in different fungal pathogens (Balhadere and Talbot, 2001;Gilbert et al., 2006;Hu and Kronstad, 2010;Rizzo et al., 2014;Li et al., 2019). The homologs of Drs2 of S. cerevisiae were named as MoPde1 and Apt1 in M. oryzae and C. neoformans, respectively.
MoPde1 is involved in regulating host penetration and hyphal development in M. oryzae while loss of Apt1 in C. neoformans led to decreased survival of the mutant in the lungs of infected mice and the inability of the mutant to colonize brain tissues, indicating that MoPde1 and Apt1 are required for invasive growth in the respective organisms (Balhadere and Talbot, 2001;Hu and Kronstad, 2010;Rizzo et al., 2014). However, loss of FgDnfB, the homolog of Drs2, does not affect the virulence of F. graminearum (Figure 2).
We found that FgDnfA and FgDnfD function as positive regulators while FgDnfB takes the negative role in DON biosynthesis, which is the first evidence to establish the relationship between flippases and secondary metabolism in filamentous fungi. Filamentous fungi produce a diverse range of secondary metabolites including mycotoxins, which have negative impacts on food safety and animal health but which are potentially important for fungal pathogenesis (Kistler and Broz, 2015). In A. nidulans, deletion of AnDNFA or AnDNFD resulted in mutants that produce unpigmented conidia (Schultzhaus et al., 2015(Schultzhaus et al., , 2019. With the exception of these reports, there was no work that relates the functions of flippases to secondary metabolism. Research on the three well-studied fungal secondary metabolite biosynthetic pathways (penicillin G, aflatoxin, and DON synthesis pathways) showed that co-compartmentalization of secondary metabolism enzymes is important in promoting pathway efficiency and sequestering intermediates and products from the rest of the cell (Chanda et al., 2009;Kistler and Broz, 2015;Boenisch et al., 2017). For DON biosynthesis, formation of the toxisome from ER is necessary to ensure compartmentalization of the process (Chen et al., 2019). In this study, we found that deletion of the flip-

| Strains and culture conditions
The wild-type strain PH-1 and all transformants used in this study were stored as mycelial suspensions in 20% glycerol solution at −80°C. CM and MM were used for mycelial growth tests. CMC medium and SNA were used for conidiation assays (Leslie and Summerell, 2006). For conidial germination, fresh mycelial plugs of each strain were inoculated on SNA plates at 28°C for 7 days. Sexual reproduction was induced on carrot medium as described previously (Zheng et al., 2015). TBI medium was used for the induction of DON (Menke et al., 2012).

| Strain construction
For construction of gene deletion mutants, the upstream and downstream fragments of the target gene were amplified by the primers listed in Table S1. Double-joint PCR was used to build a gene replacement construct (Yu et al., 2004), which was transformed into the protoplasts of the wild-type strain to generate the gene deletion mutants (Hou et al., 2002). The resulting transformants were screened by PCR using the primers shown in Table S1 and further verified by Southern blot. Hygromycin (100 mg/ml) or geneticin (150 mg/ ml) was used as a selective marker for single-or double-gene deletion, respectively. For complementation, the entire target gene (without stop codon), including its promoter region, was amplified by PCR using the set of primers listed in Table S1 and transformed with XhoI-digested pYF11 using the yeast gap repair approach (Zhou et al., 2011). The resulting target gene with a GFP fusion construct carrying the geneticin resistance gene was introduced into the corresponding mutant's protoplasts, and the resulting transformants were selected in geneticin (150 mg/ml)-containing media. For observation of the Tri1-GFP location, the generated FgTri1-GFP fusion vector (Adnan et al., 2020) was transformed into the wild type and the corresponding mutants' protoplasts and geneticin (150 mg/ml) was used as a selective marker.

| Pathogenicity and DON production assays
Pathogenicity assays on wheat spikelets were conducted as described previously (Yun et al., 2014). In brief, 10 μl of conidia suspension (10 6 conidia/ml) or a mycelial block (3 mm in diameter) of each strain was inoculated in the middle of spikelets of wheat flowers, and then the inoculated wheat head was covered with a plastic bag to keep it humid for 2 days. It was observed 2 weeks after inoculation. For wheat coleoptile infection assays, 10 μl of conidial suspensions (4 × 10 5 conidia/ml) were inoculated and symptoms were observed 8 days after inoculation. For DON production assays, each strain was grown in TBI at 28°C for 7 days in the dark. The liquid and mycelia were then collected. The liquid solution was tested quantitatively for DON using a Vomitoxin ELISA kit (Finder Biotech Co.) (Zheng et al., 2018), while the mycelia were dried and weighed for quantification. Each experiment was repeated three times.

| Measurement of extracellular enzyme activity
For detection of the activity of extracellular enzymes, three fresh mycelial plugs (5 mm in diameter) from each strain were inoculated in a 250-ml flask containing 100 ml of Czapek's medium at 25°C for 7 days. Mycelia were completely removed by filtration, and the culture filtrates were used for the measurement of extracellular enzyme activities. The activities of endoglucanase, amylase, and polygalacturonase were determined using the 3,5-dinitrosalicylic acid method with slight modifications, as previously described (Miller, 1959). The dry weights of the harvested mycelia were measured for normalizing the enzyme activities.

| Live-cell imaging assay
Fresh conidia were collected after 4 days of incubation in CMC medium and then stained with 10 μg/ml CFW for 2 min. The cell walls and septa of the conidia were observed under an A1 confocal microscope (Nikon). Fresh mycelia of each strain were stained with 2 μM FM4-64 and observed for endocytosis under an A1 confocal microscope. ER-Tracker Red (Beyotime Biotechnology) was used to label the ER. NBD-PS/PC/PE (Avanti Polar Lipids) was used for phospholipid staining as previously described (Hanson and Nichols, 2001). In brief, young mycelia were suspended in ice-cold MM-S medium (MM medium without sucrose but with 2% sorbitol) with 10 μl of lipid dye and incubated at 30°C for 30 min, then washed with ice-cold MM-S and observed under an A1 confocal microscope. The wavelengths of excitation/emission used for NBD-PS/PC/PE were 488 nm/500-550 nm.

| RT-qPCR analysis
Total RNA of each strain was isolated from mycelia harvested from 3-day-old TBI cultures or 3-day-old CM cultures using TRIzol. To detect the relative expression levels of the target genes, SYBR Premix Ex Taq II (Takara) was used for RT-qPCR. The tubulin gene (FGSG_09530) of F. graminearum was used as the endogenous control, and the relative expression levels of the target genes were calculated using 2 −ΔΔCt formula (Livak and Schmittgen, 2001). The experiments were repeated three times.

ACK N OWLED G M ENTS
This research was supported by the Natural Science Foundation of China (31601583 and 31870136).

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.

E N D N OTE S
1 Virulence on wheat heads was scored (4, normal; 0, markedly reduced virulence compared to wild-type strain).