For performing mycelial growth assays, the WT F. graminearum strain PH-1, FgPEX4 deletion mutants, and other transformants generated in this study were grown on potato dextrose agar (PDA) medium in a 25°C incubator. For the aerial hyphal growth assay, all strains were inoculated into test tubes (1.5 cm diameter) containing 5 mL PDA medium and grown at 25°C for 5 days. To study the integrity of cell membranes and cell walls, all strains were grown as previously reported on CM medium supplemented with 0.01% SDS as a cell membrane-damaging agent and 0.2% Congo red as a cell wall-damaging agent (Gavric et al., 2007; Chayakulkeeree et al., 2008; Qin et al., 2015).

The Y2H assay was performed using the Matchmaker GAL4 Two-Hybrid System 3 (Clontech) according to the manufacturer’s instructions. Full-length complementary DNA (cDNA) of FgPEX22-like and FgPEX4 was PCR amplified using the primer pairs P22-AD-F/22R and P4-AD-F/R, respectively. The resulting PCR products were cloned into pGADT7 and pGBKT7 which were digested with XholI to create FgPEX22-like-AD and FgPEX4-BD as the prey vector and bait vector, respectively. A similar method was used to generate FgPEX4-AD and FgPEX22-like-BD. The resulting bait and prey vectors were confirmed by sequencing and co-transfected as pairs. To explore the key action regions of FgPEX4 and FgPEX22-like, truncated cDNA of FgPEX4 and FgPEX22-like were amplified with their respective primers (Supplementary Table 2). These shorter PCR products were cloned into pGADT7 to construct different FgPEX22-like (a–b)-ADs as the prey vector. The same method was used to create FgPEX4 (c–d)-BDs as bait vectors. The truncated bait and prey vectors were also confirmed by sequencing and were co-transfected into yeast in various combinations (Supplementary Table 2). The interaction between pGBKT7-53 (BD-3) and pGADT7-T (AD-1) was used as a positive control.

For co-immunoprecipitation assays, the intracellular region of FgPEX4 and FgPEX22-like were PCR-amplified with the primer pairs 4-GFP-F/R and 22-Flag-F/R (Supplementary Table 2), and then inserted into the vector pFL2 and pFL7 which were digested with Xhol I respectively. The resulting fusion constructs PEX4-GFP and FgPEX22-like-3 × FLAG were verified by DNA sequencing. Constructs PEX22-like-3 × FLAG and PEX22-like-3 × FLAG with PEX4-GFP were transfected into strain PH-1. For transformant selection, G418 (Geneticin) was added at a final concentration of 200 mg/mL. Total proteins were extracted and incubated with GFP beads as previously reported (Li et al., 2017). Total proteins and proteins eluted from the GFP beads (elution) (Kangti Life Technology Co., Ltd., KTSM1334) were analyzed by western blotting using monoclonal anti-FLAG antibody (Abimate medical technology (Shanghai) Co., Ltd., PA9020S) and anti-GFP antibody (Abimate medical technology (Shanghai) Co., Ltd., PA9056S), accordingly. The results were visualized using the enhanced chemiluminescent (ECL) detection system.

The mutants were generated using the split-marker method (Wang et al., 2011). For the FgPEX22-like gene (FGSG_11970), the sequence was obtained from the F. graminearum database. The upstream (931 bp) and downstream (855 bp) regions flanking the gene were PCR-amplified using primer pairs AF/AR and BF/BR, respectively (Supplementary Table 2). The plasmid pCB1003 harbored hygromycin B resistance gene (HPH) and the primer pair HPH-F/HPH-R was used to amplify HPH (Supplementary Table 2). A fusion cassette containing the FgPEX22-like flanking sequences and the HPH gene was transfected into protoplasts of PH-1 to generate the ΔPEX22-like mutant (Catlett et al., 2003). Southern blot assay for FgPEX22-like deletion mutants was performed using the digoxigenin (DIG)-labeled probe and the High Prime DNA Labeling and Detection Starter Kit I (Roche Diagnostics, Mannheim, Germany), as instructed by the manufacturer. Hybridization was performed using a DIG-labeled specific probe (Supplementary Figure 1).

To prepare the complementation strain, a 2,674 bp fragment containing the full-length FgPEX22-like gene sequence and its promoter sequence was PCR-amplified using the primer pair 22CF/22CR (Supplementary Table 2) and then amplimer inserted into pYF11 using the yeast in vivo recombination approach (Bruno et al., 2004; Zhou et al., 2011). The recombinant plasmid pYF11-PEX22-like was transfected into protoplasts of theΔPEX22-like mutant to produce ΔPEX22-like complementation (ΔPEX22-like -C) strains, which were identified by PCR analysis (Supplementary Figure 1).

To generate the FgPEX22-like and FgPEX4 double-knockout mutants, a FgPEX22-like gene replacement construct was generated with the G418 amplified from pFL2 (Zhou et al., 2011) and transfected into the FgPEX4 mutant ΔPEX4 (Table 1). Transformants resistant to both hygromycin and G418 were screened by PCR analysis (Jiang et al., 2015).

For conidiation assays, strains were inoculated into CMC medium as previously described (Hou et al., 2002). The chemical compound 4′,6-diamidino-2-phenylindole (DAPI, 10 μg/mL) and calcofluor white (CFW, 1 μg/mL) were used to stain nuclei and septa of conidia, respectively. After staining, images were captured using a fluorescence microscope (Eclipse 90i, Nikon). To determine the germination rate of conidia, freshly harvested conidia were transferred to the sterile YEPD medium (yeast extract, 10.0 g; peptone, 20.0 g; glucose, 20.0 g; and distilled water to make up the volume to 1,000 mL) for 6 h and observed under a fluorescent microscope (Eclipse 90i, Nikon). For sexual reproduction assays, 7-day-old aerial hyphae growing on specific sporulation medium were compressed in 1 mL sterile 2.5% Tween 60 solution as previously described (Bowden and Leslie, 1999; Jenczmionka et al., 2003). Perithecium formation was examined after 2–3 weeks of incubation at 25°C.

Plant infections were assayed on wheat heads and corn silks. Susceptible wheat cultivar Jimai 22 was used in wheat infection assays and was sprayed with a conidial suspension (2 × 10⁵ spores/mL) collected from 5-day-old liquid CMC medium (Qin et al., 2015). Wheat heads were photographed and assayed 14 dpi. Fresh corn silks were infected with hyphal plugs, incubated at 25°C, and examined at 5 dpi (Seong et al., 2005; Gale et al., 2007). To determine DON production, we inoculated three mycelial plugs from each strain into 5 g healthy and aseptic rice grains. After incubating at 25°C, DON was extracted at 20 dpi as described previously and quantified using a liquid chromatography-mass spectrometer/mass spectrometer (HPLC–MS/MS) system (AB Sciex 5500) as previously described (Mirocha et al., 1998).

The carbon source utilization was evaluated using minimal medium containing various carbon sources in lieu of sucrose. The following concentrations were used: 2.5 mM myristic acid (C14), 2.5 mM palmitic acid (C16), 2.5 mM oleic acid (C18), and 2.5 mM erucic acid (C22) as previously described (Leslie and Summerell, 2007). Emulsifier NP40 was added to the minimal medium containing palmitic acid, oleic acid, and erucic acid. Colony diameters were measured after 3.5 days incubation at 25°C.

Lipid droplets (LD) in hyphae were stained using Nile red (50 μg/mL) as reported previously (Lu et al., 2009). Hyphae grown on PDA plates at 25°C for 3 days were collected and subjected to ultrastructural analysis. The collected fungal mass was treated and examined by transmission electron microscopy (TEM) (JEM-1400 Plus, JEOL, Tokyo, Japan).

For subcellular localization of FgPEX4, peroxisome membrane protein 70 (PMP70), and Woronin body protein hex1 (HEX1), coding sequences of FgPEX4, FgPMP70, and FgHEX1 were PCR-amplified using the primer pair 4-GFP-F/R, PMP70-GFP-F/R, and HEX1-GFP-F/R, respectively (Supplementary Table 1). The amplimers were then inserted into pYF11 using the yeast in vivo recombination approach (Bruno et al., 2004; Zhou et al., 2011). The recombinant plasmids were transfected into protoplasts of the WT and mutant strains. The transformants were verified by PCR analysis using the appropriate primers. Transformants were observed using a fluorescence microscope.

The ROS generated in the hyphae of PH-1 and ΔPEX22-like strains was assessed using NBT (nitroblue tetrazolium chloride) following the growth of the strains on CM at 25°C for 3 days. Each plate was then stained with 20 mL of 0.2% NBT solution and incubated in the dark at 28°C for 45 min. The liquid stain was drained from the plates, which were then washed with ethanol. The plates were incubated again for 45 min in the dark at 28°C prior to imaging.

Total RNA was isolated from hyphae of the WT, mutant, and complementation strains using TransZol Up (TransGen Biotech, Beijing, China). The quantitative real-time PCR (qRT-PCR) experiments were performed as the manufacturer’s instructions (Vazyme Biotech Co., Jiangsu, China). The glyceraldehyde 3-phosphate dehydrogenase (GAPDH) gene of F. graminearum was used as the internal control. Relative expression levels for each gene were calculated using the 2^–ΔΔCT method (Livak and Schmittgen, 2001).

Each experiment was performed in three individual replicates. Data are presented as mean ± standard error values and the differences among variables were analyzed using Duncan’s multiple range test. Results with p < 0.05 were considered statistically significant.

The homologs of PEX22 (FGSG_11970) were identified using the Fusarium genome database¹. The predicted gene was 1,631 bp, encoding 363 amino acids. FgPEX22-like shared only 14.34% identity with C. orbiculare and 17.93% identity with S. cerevisiae PEX22. Therefore, we named the gene FgPEX22-like. SMART-PFAM analysis revealed the protein encoded by PEX22-like contained a transmembrane domain and a possible PEX4 interacting region similar to that in C. orbiculare and S. cerevisiae (Supplementary Figure 1A). The deduced FgPEX22-like protein was distributed in one branch in the phylogenetic tree along with other fungal PEX22 proteins (Supplementary Figure 1B).

To investigate whether FgPEX22-like interacted with FgPEX4 in F. graminearum, a yeast two-hybrid assay was used. Yeast transformants expressing binding domain (BD)-FgPEX4 as the bait and the activation domain (AD) pGADT7 alone as the prey, could not grow on the Sabouraud dextrose (SD)-Leu-Trp-His-Ade plates, excluding the possibility of self-activation. This was also true for the reverse situation. The sets of interacting pairs, BD-FgPEX4 and AD-FgPEX22-like or AD-FgPEX4 and BD-FgPEX22-like, were found to bind with each other (Figure 1A).

To further verify these results, the intracellular region of FgPEX4 was fused with pFL2 [PEX4-green fluorescent protein (GFP)] and co-transfected into F. graminearum strain PH-1 with the FgPEX22-like-3 × FLAG fusion construct. In western blot analysis of total proteins from the strain containing PEX22-3 × FLAG, a 40-kDa band was detected using the anti-FLAG antibody, while the elution proteins were not detected. This showed the beads had been washed prior to protein elution and that there were no other contaminants, confirming the subsequent results were reliable. In the transformants co-expressing FgPEX22-like-3 × FLAG and FgPEX4-GFP, 19-kDa and 40-kDa bands were detected using the anti-FLAG and anti-GFP antibodies, respectively (Figure 1B). Therefore, FgPEX22-like interacted with FgPEX4 in F. graminearum. Taken together, these results demonstrated the two proteins could interact directly with each other.

Based on the predicted secondary structure and domain structure, we explored the key functional regions of FgPEX4 and FgPEX22-like. FgPEX22-like contained an N-terminal transmembrane (TM) fragment and a large region spanning an unknown fold that was exposed in cytosol. Several truncated variants of FgPEX22-like were established and their interactions with FgPEX22-like and FgPEX4 were evaluated using the yeast two-hybrid system. Our results showed that the region spanning residues 102–282 was essential for FgPEX4 binding; moreover, residues were part of the soluble region of FgPEX22-like. Meanwhile, we also discovered that the first 11 amino acids of FgPEX4 were not necessary for FgPEX22-like binding (Supplementary Figure 2).

To further understand the relationship between the two proteins, we compared the subcellular localization of wild-type (WT) FgPEX4 to that of mutant strains. Interestingly, the results showed that a deficiency of FgPEX22-like could result in abnormal subcellular localization of FgPEX4. FgPEX4-GFP fusions was distributed in punctate patterns in transformed PH-1, but were dispersed in the cytoplasm of the PEX22-like deletion mutant ΔPEX22-like (Figure 2). These results indicated FgPEX22-like was necessary for the location of FgPEX4 and that it functioned as a rivet of FgPEX4.

To elucidate the biological functions of FgPEX22-like in F. graminearum, we generated the single-gene knockout mutant ΔPEX22-like and the double-gene knockout mutant ΔΔPEX4/22-like by split-marker polymerase chain reaction (PCR) and protoplast transformation. A schematic diagram depicting the strategy used to create the FgPEX22-like gene deletion mutant and molecular analysis of ΔPEX22-like is shown in Supplementary Figure 3. The pYF11-PEX22-like complementation construct was generated by PCR amplification using the primer pair 22hfF/22hfR (Supplementary Table 1), followed by transformation of ΔPEX22-like mutant protoplasts to produce the ΔPEX22-like complementation strain ΔPEX22-like-C.

No obvious vegetative growth or colony morphology defects were observed for ΔPEX22-like compared to that of the WT strain. The FgPEX4 and FgPEX22-like double mutants exhibited a significant reduction in growth rate compared to that of the FgPEX22-like single mutant and produced compact colonies with limited aerial hyphae (Figures 3A–D). These results suggested FgPEX22-like was not involved alone in the regulation of hyphal growth of F. graminearum.

When assayed for conidiation in carboxymethyl cellulose (CMC) medium cultures, ΔPEX22-like and ΔΔPEX4/22-like were reduced by 25.0 and 82.3%, respectively, compared with that of the WT strain (Table 1). In addition, conidia of the mutants exhibited normal conidium morphology, displaying conidia that lacked intracellular content and devoid of nuclei (Figure 4A). Phialides of the ΔPEX22-like and ΔΔPEX4/22-like mutants were rarely clustered together, unlike those of PH-1 (Figure 4B), which may have been directly responsible for reduced conidiation in the ΔPEX22-like mutant. When incubated in yeast extract peptone dextrose (YEPD) medium, the conidium germination of ΔPEX22-like and ΔΔPEX4/22-like at 6-h post-incubation was reduced by 26.2 and 31.2%, respectively, compared with that of PH-1 and the germ tubes were shorter than those of PH-1 (Figure 4C and Table 1). The ability of the FgPEX22-like mutants to undergo sexual reproduction was also investigated. We found the number of perithecia for ΔPEX22-like and ΔΔPEX4/22-like were reduced by 65.0 and 56.6%, respectively, compared with that of the WT strain at 2-wk post-fertilization, and the ascocarp is no difference between them (Figure 4D). Together, these results demonstrated that FgPEX22-like played vital roles in conidiation, conidial germination, and sexual reproduction.

The pathogenicity assays were performed to ascertain the effect of ΔPEX22-like on flowering wheat heads and corn silks. Wheat heads inoculated with either PH-1 or the ΔPEX22-like -C strain presented typical scabs, spreading from the inoculated spikelets to almost the entire head by 14 days post inoculation (dpi). In contrast, wheat heads inoculated with the ΔPEX22-like mutant presented scabs mainly on or near the inoculated spikelets (Figure 5A). The disease index of ΔPEX22-like and ΔΔPEX4/22-like exhibited approximately a 71.9 and 74.0% reduction, respectively, compared with that of PH-1 (Figure 5B). In corn silk infection, the extended length of brown lesions caused by PH-1 and the ΔPEX22-like-C strain were longer than those of ΔPEX22-like and ΔΔPEX4/22-like. The ability of ΔPEX22-like or ΔΔPEX4/22-like to infect corn silk was only 1/5 that of the WT strain at 5 dpi (Figures 5C,D).

We also assayed ΔPEX4 and ΔΔPEX4/22-like for DON production. We found that infected rice seeds inoculated with the ΔPEX22-like mutant produced significantly less DON than that of PH-1 or the ΔPEX22-like-C strain (Table 1). These results indicated FgPEX22-like had an important role in pathogenicity and DON production.

To determine the effect of FgPEX22-like on F. graminearum organelles, the ultrastructure of PH-1 and ΔPEX22-like was evaluated using TEM. Spherical peroxisomes were observed in the periphery of mycelial cells of PH-1, but not ΔPEX22-like. In addition, Woronin bodies, which were present around the mycelial septum in the WT strain, were absent in the ΔPEX22-like mutant (Figure 6A). Besides, Compared with WT strain, ΔPEX22-like accumulated more lipid droplets. And Nile Red staining revealed that the WT strain degraded most of the lipid droplets, while mutant strains still contained numerous bright lipid droplets (Figure 6B).

To confirm the existence of peroxisome and Woronin body, GFP fusion constructs with the Woronin body protein HEX1 or PMP70 were transfected into the WT and mutant strains and evaluated using laser-scanning confocal microscopy. For PMP70-GFP, ΔPEX22-like displayed a green punctate distribution in hyphal cells similar to that in the WT strain (Figure 7). However, the fusion of GFP with the Woronin body protein HEX1 resulted in a punctate distribution in the WT strain while being dispersed in the cytoplasm of ΔPEX22-like (Figure 7). These results indicated FgPEX22-like was indispensable for maintaining Woronin bodies, but not peroxisomes.

In addition, based on the TEM observations, the mutant accumulated more lipid drops, indicating the utilization rate of cellular lipid drops in ΔPEX22-like was slower than that in PH-1. Further verification was carried out by performing Nile red staining. Brighter and larger lipid droplets were observed in ΔPEX22-like compared to that of the WT strain. This was consistent with the TEM findings.

To investigate whether FgPEX22-like participated in environmental stress responses, we examined the sensitivity of mutants to the cell wall-damaging agent Congo red and the cell membrane-damaging agent sodium dodecyl sulfate (SDS). We found that ΔPEX22-like exhibited increased sensitivity to both SDS and Congo red compared to that of PH-1 (Supplementary Figures 4A,B). To verify these results, hyphae were treated with lysozyme and driselase. After incubation at 30°C for 30 min, hyphae of the ΔPEX22-like mutant were almost completely digested and had released abundant numbers of protoplasts, whereas few protoplasts were observed among hyphae of the PH-1 and ΔPEX22-like -C strains (Supplementary Figure 4C). Together, these results illustrated that deletion of FgPEX22-like resulted in reduced cell wall integrity in F. graminearum.

To investigate whether FgPEX22-like was involved in the response to oxidative stress, the tolerance of ΔPEX22-like to ROS was measured. The radial growth of ΔPEX22-like and ΔΔPEX4/22-like were inhibited by 60.0 and 56.8%, respectively, when exposed to 20 mM H₂O₂, which was greater than that of PH-1 at 42.7% (Supplementary Figures 5A,B). Cellular ROS production in the hyphae was qualitatively analyzed by staining with nitroblue tetrazolium (NBT). The assay showed that the staining of ΔPEX22-like and ΔΔPEX4/22-like hyphae were darker than those of PH-1 (Supplementary Figure 5C). Taken together, these results demonstrated the capacity of ΔPEX22-like and ΔΔPEX4/22-like to eliminate ROS was decreased.

To determine the effect of deleting FgPEX4 on fatty acid utilization, vegetative growth was assessed using minimal medium with different fatty acids as sole carbon sources. The fatty acids tested included long-chain fatty acids myristic acid (C14), palmitic acid (C16), and oleic acid (C18), and very long-chain fatty acid erucic acid (C22). After incubation with the particular fatty acids for 3 days, the radial growth of the mutants was significantly reduced on the media containing the long-chain fatty acids and very long-chain fatty acids as the sole carbon sources (Figures 8A,B). These results indicated that deletion of FgPEX22-like in F. graminearum resulted in a defect in fatty acid metabolism.