The PH-1 strain was used as the wild-type strain in this study. Culture conditions were the same as described previously using complete medium (Zheng et al., 2012; Chen et al., 2016). Sensitivities to environmental stimuli were examined on complete medium supplemented with 1 mg/ml CR, 200 mg/ml Calcofluor white (CFW), 0.01% SDS, 1M NaCl, 1M KCl, and 10 mM H₂O₂.

The F. graminearum ESCRT genes were disrupted by gene replacement. To replace the FgVPS23, the 1,056-bp upstream and 816-bp downstream fragments of FgVPS23 were amplified with primer pairs 1F/1R and 2F/2R (Supplementary Table 2), respectively. And then Splicing Overlap Extension (SOE)-PCR was used to connect the resulting amplicons to the hph fragments (Catlett et al., 2003). The resulting PCR fragments containing the gene replacement cassette were transformed into protoplasts of wild-type strain PH-1 as described (Hou et al., 2002). After transformation, transformants with hygromycin-resistance were picked individually and PCR analyses with designated primer pairs 3F/3R and 4F/4R were performed to identify transformants that carried the insertion of hph at the VPS23 locus. Southern blot analysis was performed according to standard manufacturer's protocols (Digoxigenin High Prime DNA Labeling and Detection Starter Kit I; Roche) to further confirmed the mutant strains. Similar method was used to generate other ESCRT genes deletion mutants.

To generate FgVps23-GFP fusion vector, the FgVPS23 gene with its 1515-bp upstream promoter region was amplified with 29F and 29R using genomic DNA extracted from wild type (PH-1) as template. The resulting fragment was cloned into the Nde I and EcoR I sites of the PKNTG vector harboring the GFP allele and the neomycin gene as a selection marker via using the ClonExpress One Step Cloning Kit (Vazyme, Nanjing, China). Other ESCRT protein GFP fusion vectors were generated by using the same strategy. The resulting vector was verified by sequencing and then transformed into the protoplasts of corresponding ESCRT gene deletion mutant strains. The neomycin-resistant transformants were picked and then screened by PCR and GFP signal.

For asexual reproduction assays, mycelial agar blocks (5 mm in diameter) were inoculated into CMC liquid medium containing 0.1% NH₄NO₃, 0.1% yeast extract, 0.1%KH₂PO₄, 0.05% MgSO4*7H₂O, 1.5% carboxylmethyl cellulose or synthetic low-nutrient agar (SNA) medium containing 0.1% KNO₃, 0.05% MgSO₄·7H₂O, 0.02% glucose, 0.1% KH₂PO₄, 0.05% KCl, 0.02% sucrose, and 2% agar. The number of conidia were determined 6 days after incubation at 25°C under microscopy by using a hemacytometer. Since F. graminearum is homothallic, to induce sexual reproduction, mycelial agar blocks were placed on carrot agar and incubated at 28°C for a week. Then aerial hyphae were removed and the plates were gentle pressed down with 300μl of sterile 0.1% Tween 20. All of the sexually-induced cultures were incubated at 28°C for an additional 2–3 weeks under a 12 h dark/12 h black light cycle. Each experiment was repeated independently three times.

The virulence of fungal strains was determined on flowering wheat heads. Mycelial agar blocks (3 mm in diameter) of wild-type strain, mutant strains and the corresponding complementation strains were incubated at the middle spikelet of wheat. The inoculated wheat heads were enclosed in small plastic bags misted with water for 3 days to maintain the moisture. After incubation for 14 days, spikelets with typical head blight symptoms were examined and counted to estimate the disease index. All the infection assays were repeated at least three times. For determination of DON production, tested strains were cultured in liquid trichothecene biosynthesis induction (TBI) medium (Gardiner et al., 2009) in the dark for 1 week and the amount of DON was measured by using DON detection plate kit (Finder Biotech Co. Ltd, China).

The yeast two-hybrid assay was used to examine the possible protein-protein interactions among ESCRT proteins in accordance with the manufacturer's protocol (Matchmaker GAL4 Two-Hybrid System 3; Clontech). To generate vectors for yeast two-hybrid analyses, the full-length cDNA of each tested gene was amplified from first-strand cDNA of PH-1 with primers listed in Supplementary Table 2. The cDNA of each gene was cloned into the yeast GAL4 activation domain vector pGADT7 and GAL4-binding domain vector pGBKT7 as the prey vector and bait vector, respectively. The resulting bait and prey vectors were confirmed by sequencing and were co-transformed in pairs into the S. cerevisiae reporter strain AH109 (Clontech) via the lithium acetate transformation procedure (Schiestl and Gietz, 1989). The isolation and confirmation of transformants were conducted as described (Chen et al., 2008). Briefly, the Leu+ and Trp+ transformants were isolated and assayed for growth on SD-Ade-Leu-Trp-His with the addition of 40 μM X-gal to examine the HIS3 reporter gene expression and β-galactosidase activity. In all assays, yeast transformants expressing the pGBKT7-P53 bait-pGADT7-T prey and pGBKT7-Lam bait-pGADT7-T prey constructs were used as the positive control and negative control, respectively. All the yeast two-hybrid assays were performed three times to confirm the results. All primers used in this experiment were listed in Supplementary Table 2.

For quantitative RT-PCR (qRT-PCR) assays, RNA isolation was performed with the RNA extraction kit following the instructions provided by the manufacturer. First-strand cDNA was generated using the PrimeScript^TM RT reagent kit (Takara). The resultant cDNA was used as a template for qRT-PCR. qRT-PCR was performed with SYBR Premix Ex Taq^TM (Takara), denaturation at 95°C, 32 s of annealing at 60°C for 40 cycles. The relative abundance of the transcripts of each gene were calculated by 2^−ΔΔT (Livak and Schmittgen, 2001) using F. graminearum tublin gene (FGSG_09530) as control. Primers used to amplify selected genes in qRT-PCR reactions are listed in Supplementary Table 2.

Previously, we identified the interactions within the ESCRT-0 complex, and showed that its two components FgVps27 (FGSG_08545) and FgHse1 (FGSG _08492) could directly interact with each other (Xie et al., 2016). In this study, we focused on protein-protein interactions within F. graminearum ESCRT-I, -II, and –III complexes, as well as between these complexes (Supplementary Table 1). To identify orthologs of components of ESCRT-I, -II, and -III complexes in F. graminearum, the sequences of corresponding ESCRT genes from the budding yeast S. cerevisiae were used for BLASTP searches in the F. graminearum genome database (http://www.broadinstitute.org/annotation/genome/fusarium_group/MultiHome.html). In total, 9 genes FGSG_02656, FGSG_000291, FGSG_04120, FGSG_08401, FGSG_05263, FGSG_04112, FGSG_10883, FGSG_10832, FGSG_10092 were found. They are renamed according to their orthologs in S. cerevisiae (Supplementary Figure 1). However, there are no Vps37 and Mvb12 in F. graminearum, suggesting F. graminearum may have lost the Vps37 and Mvb12 orthologs during evolution. To determine the interactions between ESCRT core components, yeast two-hybrid was performed between ESCRT components within or between complexes I, II, and III. With proper positive and negative controls (Figure 1A), we found strong interactions between individual components with high β-galactosidase signal within ESCRT-I and ESCRT-II complexes, while ESCRT-III complex also displayed several interactions, including FgVps20-FgVps32, FgVps32-FgVps24, and FgVps24-FgVps2 in addition to FgVps32 self-interaction (Figure 1B). Between the ESCRT complexes, the interaction of FgVps27 and FgVps23 connected ESCRT-0 and ESCRT-I complexes whereas the interactions of FgVps23-FgVps22, FgVps23-FgVps36, and FgVps28-FgVps36 facilitated the association of ESCRT-I and ESCRT-II. FgVps20 interacted with FgVps25 and FgVps36, linking ESCRT-II and ESCRT-III complexes (Figure 1C). Taken together, our results established the interactome of ESCRT core complexes in F. graminearum (Figure 1D).

In order to investigate the biological functions of ESCRT genes in F. graminearum, we attempted to generate gene deletion mutants for every ESCRT component of ESCRT-I, -II, and -III complexes by replacing each gene with a selective marker [the bacterial phosphotransferase (hph) gene], through transformation of protoplasts of the wild-type F. graminearum strain with the deletion constructs. The resulting hygromycin-resistant transformants of each ESCRT gene were identified by PCR analyses with gene-specific primers listed in Supplementary Table 2. For 6 genes (FgVPS23, FgVPS28, FgVPS22, FgVPS25, FgVPS36, FgVPS20), at least two independent deletion mutants were identified for each gene with similar phenotypes as described below, and further confirmed by Southern hybridization assays (Supplementary Figure 2). For the other 3 genes (FgVPS2, FgVPS24, FgVPS32), which belonged to the ESCRT-III complex, we failed to obtain gene knockout mutants after screening thousands of transformants from many independent transformation experiments, suggesting that these genes are probably essential for F. graminearum growth or development (Son et al., 2011; Wang et al., 2011; Yun et al., 2015). To further confirm that the defects displayed in ESCRT mutants were caused by the loss of corresponding ESCRT proteins, native ESCRT genes with their native promoters, were reintroduced into corresponding gene deletion mutants. The corresponding complementation strains were obtained after PCR and phenotypic screening and named as ΔFgvps23-C, ΔFgvps28-C, ΔFgvps22-C, ΔFgvps25-C, ΔFgvps36-C, and ΔFgvps20-C, respectively, in this study.

Endocytosis constitutes a fundamental eukaryotic function that internalizes fluids, solutes, and plasma membrane components into vesicles which incorporate with the endosomal system (Huotari and Helenius, 2011). Our previous study showed that FgVps27, the core component of ESCRT-0 complexes, was involved in endocytosis of F. graminearum (Xie et al., 2016). To determine whether other ESCRT components have a similar function, we followed the uptake of FM4-64, a lipophilic dye endocytosed and trafficked to the vacuolar membrane (Vida and Emr, 1995; Fischer-parton et al., 2000), to monitor endocytosis kinetics in the mycelial cells of wild-type and ESCRT gene mutants. As shown in Figure 2, upon FM4-64 application, the plasma membrane and septum were immediately stained in wild-type strain PH-1, and after 1 h incubation, this fluorescent dye was delivered to the vacuolar membrane by normal endocytosis. Similar staining pattern of FM4-64 was observed in ESCRT gene deletion mutants initially (data not shown). However, FM4-64 failed to be transported and targeted to vacuolar membranes, and instead, predominantly appeared in the plasma membrane and small punctate compartments adjacent to the vacuole in all the mutants after incubation for 1 h (Figure 2). Even for 2 h, ESCRT gene deletion mutants failed to deliver FM4-64 to the vacuolar membrane (data not shown), indicating that defects in proper endocytosis caused by the loss of ESCRT genes were not time-dependent. These results suggest that similar to ESCRT-0, ESCRT-I, -II, and -III complexes play critical roles in endocytic trafficking to the vacuole in F. graminearum.

Endocytosis plays an essential role in cellular responses to environmental stimuli, which is critical for cell survival and proper development (Murphy et al., 2005; Fan et al., 2015; Paez et al., 2016). The deletion of ESCRT genes resulted in defective endocytosis in F. graminearum, suggesting that development and differentiation of F. graminearum may also be affected in the mutants. To test this, wild-type strain, ESCRT deletion mutants, together with corresponding complementation strains, were cultured on complete medium and incubated at 25°C for 3 days. As shown in Figure 3A, the colonies of mutants were yellowish whereas those of the wild type strain, and complemented strains were pinkish, suggesting that ESCRT may be involved in red pigment formation. In addition, the loss of ESCRT components led to severe defects in vegetative growth, displaying smaller colony morphology and thinner aerial hyphae in comparison to the wild-type (Figures 3A,C,D). Microscopic examination further showed that deletion of ESCRT components resulted in more sparse aerial hyphae than PH-1 (Figure 3B). Previous studies showed that the surface hydrophobicity of aerial hyphae is closely related to hyphal formation in many fungal species (Kershaw and Talbot, 1998; Wösten et al., 1999), suggesting that the mutants with fewer aerial hyphae may have reduced hyphal hydrophobicity. To test this contention, 2.5% bromophenol blue droplets were added onto the colony surface of each strain and observed for their absorption and dispersion. While the wild-type strain and complemented strains exhibited strong hydrophobicity and the bromophenol blue droplets didn't disperse within 30 min, the droplets placed on the colony surface of all the ESCRT mutants dispersed immediately (Figure 3E), suggesting that the deletion of ESCRT genes resulted in the loss of hydrophobicity on the mycelia surface. Taken these results together, the depletion of ESCRT impairs the hyphae development of F. graminearum, possibly due to the defects in endocytosis.

Next, we investigated whether the dysfunction in endocytic pathway due to the loss of ESCRT genes also had an effect on response to various environmental stresses in F. graminearum. When strains were cultured on CM plates with 1M NaCl and 1M KCl for 3 days, the growth of these ESCRT mutants was almost completely blocked in comparison to that of the wild type (Figures 4A,C), indicating that the ESCRT core components in F. graminearum are indispensable for regulating responses to hyperosmotic stress. Furthermore, both ESCRT mutants were hypersensitive to oxidative stress and their growth was more significantly reduced by the addition of 10 mM H₂O₂ (Figures 4A,C)_. Therefore, deletion of the ESCRT core components may result in decreased expression of genes involved in ROS scavenging. We also determined sensitivity of the ESCRT mutants to a series of damaging agents of cell membrane and cell wall. As shown in Figures 4B,D, the presence of 0.01% SDS severely inhibited the growth of all mutants in comparison with that of PH-1. Likewise, the ESCRT mutants grew slower than the wild type with the treatment of Calcofluor white (CFW) and Congo Red (CR), suggesting that the ESCRT core components play an important role in the cell wall integrity (CWI) pathway in F. graminearum.

Sexual and asexual reproduction play a critical role in the infection cycle of F. graminearum since ascospores are the primary inoculum and conidia are required for host colonization and disease spreading (Trail et al., 2002; Dgiii et al., 2005; Yang et al., 2015). To test the role of ESCRT components in sexual development, wild-type strain, deletion mutants, and corresponding complemented strains were cultured on carrot agar plates. Abundant perithecia were found on the plates of wild-type strain and complemented strains 14 days post-fertilization. In contrast, mutants were unable to produce any perithecium under the same conditions (Figure 5). For asexual reproduction, fresh mycelial plugs taken from wild-type strain, mutants and corresponding complemented strains were inoculated into liquid carboxymethylcellulose (CMC) media (Cappellini and Peterson, 1965). After incubation for 6 days, the ESCRT mutants failed to produce any conidia while the wild type produced (127.86 ± 12.64) × 10⁴ macroconidia per milliliter. To rule out the possibility of medium dependency in conidiation, the conidiation of these strains were also investigated on synthetic low-nutrient agar (SNA) plates, and similar results were obtained. These results indicate that ESCRT components are critical for sexual and asexual development in F. graminearum.

To determine whether ESCRT components play a role in plant infection, the virulence of all strains were evaluated on flowering wheat heads by point inoculation. As shown in Figure 6A, the wild-type strain PH-1 and the complementation strains colonized the host and spread rapidly from inoculated spikelet to other spikelets through rachis 14 days post-inoculation, resulting in typical scab symptoms in the wheat. However, mutants were unable to cause any symptom on the inoculated wheat kernel and failed to spread to adjacent rachis and neighboring spikelets, suggesting that ESCRT components are required for plant infection.

Deoxynivalenol (DON), a potent mycotoxin, has been identified as an important virulence factor produced by F. graminearum during plant infection (Seong et al., 2008; Hallen-Adams et al., 2011). We speculated that the impaired virulence of ESCRT mutants was partially due to the reduction in DON production. To test this contention, we measured the amount of DON production in the ESCRT mutants, in comparison with wild-type strain PH-1 and the complementary strains by using ELISA based DON detection plate kit. Consistent with significant reduction in virulence, DON production in ESCRT mutants was almost undetectable whereas more than 2,000 ppm DON was produced in the wild-type strain and the corresponding complementary strains (Figure 6B).

In addition, we also determined the transcription levels of the DON biosynthesis-related genes TRI5 and TRI6, the trichothecene efflux pump gene TRI12 by qRT-PCR. Consistently, the expression levels of these three genes, especially TRI5 and TRI12, were decreased dramatically in the ESCRT mutants in comparison with that of PH-1 (Figure 6C). Taken together, ESCRT components positively regulate DON production in F. graminearum.

In Aspergillus nidulans, Vps23 had been reported to localized in endosomes (Calcagno-Pizarelli et al., 2011; Galindo et al., 2012). Moreover, our previous study showed that ESCRT-0 component FgVps27 was localized to punctate structures adjacent to the vacuole labeled by endosomal marker FM 4-64 (Xie et al., 2016). Given that ESCRT core complexes can be sequentially recruited and assembled in other organisms (Saksena et al., 2007; Otegui and Spitzer, 2008; Hurley, 2010; Henne et al., 2011) and that our yeast two-hybrid assay have identified the interactions between different subcomplexes in this study, the components of F. graminearum ESCRT-I, ESCRT-II, ESCRT-III should have similar localization pattern as FgVps27. To test this contention, each native ESCRT gene driven by native promoter was fused in-frame with the green fluorescent protein (GFP) and transformed into corresponding gene deletion mutant. After screening by PCR and GFP signal, positive transformants were obtained and the localization pattern was determined by confocal fluorescence microscopy. Similar to FgVps27, these GFP-ESCRT components were mainly found in the mobile punctate structures adjacent to the vacuolar membrane in conidia and mycelia (Figure 7 and Supplementary Figure 3 and Supplementary Videos S1–S6), which co-localized with the endocytic marker FM 4-64 labeling endosomes and vacuolar menbranes (Figure 7).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.