Genome-wide functional analysis reveals that autophagy is necessary for growth, sporulation, deoxynivalenol production and virulence in Fusarium graminearum

Autophagy is a conserved cellular recycling and trafficking pathway in eukaryotic cells and has been reported to be important in the virulence of a number of microbial pathogens. Here, we report genome-wide identification and characterization of autophagy-related genes (ATGs) in the wheat pathogenic fungus Fusarium graminearum. We identified twenty-eight genes associated with the regulation and operation of autophagy in F. graminearum. Using targeted gene deletion, we generated a set of 28 isogenic mutants. Autophagy mutants were classified into two groups by differences in their growth patterns. Radial growth of 18 Group 1 ATG mutants was significantly reduced compared to the wild-type strain PH-1, while 10 Group 2 mutants grew normally. Loss of any of the ATG genes, except FgATG17, prevented the fungus from causing Fusarium head blight disease. Moreover, subsets of autophagy genes were necessary for asexual/sexual differentiation and deoxynivalenol (DON) production, respectively. FgATG1 and FgATG5 were investigated in detail and showed severe defects in autophagy. Taken together, we conclude that autophagy plays a critical role in growth, asexual/sexual sporulation, deoxynivalenol production and virulence in F. graminearum.

Sexual reproduction plays a critical role in FHB epidemic and disease cycle, thus we determined the sexual development of the wild-type strain and two ATG mutants on self-mating carrot agar cultures. The wild-type strain PH-1 produced abundant perithecia after two-week self-fertilization. By contrast, ∆Fgatg1 and ∆Fgatg5 mutants completely failed to form any perithecia under the same culture conditions (Fig. 2B). These results suggest that autophagy is important for asexual/sexual sporulation in F. graminearum. The wild-type strain PH-1 and ATG mutants were grown on PDA plates. Photographs were taken after incubation on PDA plates at 25 °C for 3 days. Linear bars in each column denote standard errors of three experiments. An asterisk indicates significant difference of colony diameter (P < 0.05).
SCientifiC RePoRTs | 7: 11062 | DOI:10.1038/s41598-017-11640-z Autophagy is required for full virulence in F. graminearum. Virulence assays were performed by point inoculation of flowering wheat heads with mycelial plugs from the wild-type PH-1 and ATG mutants. At 14 days post-inoculation (dpi), PH-1 caused typical scab symptom on inoculated and nearby spikelets. The rate of infected spikelets inoculated with most ATG mutants was significantly reduced compared to PH-1 (P < 0.01), while FgATG17 in Group 2 was fully virulent (Fig. 3). Atg17, as a component of the Atg1 complex, is important for induction of autophagy by starvation in S. cerevisiae 41 . Therefore, we determined the role of FgATG17 in autophagy and DON production in F. graminearum. We found that GFP-FgAtg8 proteolysis was not blocked in the ∆Fgatg17 mutant after induction of autophagy in MM-N liquid medium with 2 mM PMSF for 4 h (Fig. S4A).
To determine whether DON production in the ∆Fgatg17 mutant was impaired, an enzyme-linked immunosorbent assay (ELISA) was performed. The results showed that DON production between PH-1 and the ∆Fgatg17 mutant had no significant difference (Fig. S4B). These data suggested that FgATG17 is not essential for autophagy involved in FgAtg8 and DON production in F. graminearum. Furthermore, we observed that ∆Fgatg1, ∆Fgatg3, ∆Fgatg6, ∆Fgatg7, ∆Fgatg14, ∆Fgatg15, ∆Fgatg20, ∆Fgatg24 in Group 1 and ∆Fgatg2, ∆Fgatg12, ∆Fgatg13, ∆Fgatg16 in Group 2 only caused mild infection in point-inoculated spikelets but did not spread to nearby spikelets (Fig. 3A,C). Although scab symptoms developed in nearby spikelets inoculated with the other ATG mutants, the percentage of diseased spikelets was dramatically decreased by contrast to PH-1 (Fig. 3B,D). In addition, the phenotypic defects of ∆Fgatg1, ∆Fgatg7, ∆Fgatg13 and ∆Fgatg20, such as growth, conidiation and virulence, could be fully complemented by re-introduction of the corresponding ATG genes, respectively. For instance, in Fig. S3, the defects in vegetative growth of ∆Fgatg1, ∆Fgatg7, ∆Fgatg13 and ∆Fgatg20 could be complemented by re-introduction of FgATG1, FgATG7, FgATG13 and FgATG20 respectively (Fig. S3). These results indicate that autophagy is required for full virulence to wheat by F. graminearum.
Introduction of FgATG1 and FgATG5 into M. oryzae ∆Moatg1 and ∆Moatg5 complements their phenotypic defects respectively. It has been reported that MoATG1 and MoATG5 are necessary for conidiation, normal development and pathogenicity in the rice blast fungus M. oryzae 18,25 . To determine whether FgATG1 and FgATG5 can functionally complement defects in ∆Moatg1 and ∆Moatg5 mutants, the full length coding sequence of FgATG1 (under the control of the MoATG1 native promoter) and FgATG5 (under the control of the MoATG5 native promoter) were transformed into ∆Moatg1 and ∆Moatg5 mutants, respectively. The complementation transformants, ∆Moatg1/FgATG1 and ∆Moatg5/FgATG5, were identified. Phenotypic analysis showed that ∆Moatg1 and ∆Moatg5 mutants cultured on CM plates for 12 days formed sparse aerial mycelium, but ∆Moatg1/FgATG1 and ∆Moatg5/FgATG5 strains produced dense aerial hyphae, which were identical to the wild-type strain Guy11 (data no shown). Compared to (3.42 ± 0.41) × 10 7 conidia per plate produced by the wild-type strain Guy11, conidiogenesis in ∆Moatg1 and ∆Moatg5 mutants was significantly reduced (P < 0.05), only produced (7.73 ± 1.57) × 10 4 and (5.35 ± 1.14) × 10 5 conidia respectively. The ∆Moatg1/FgATG1 strain produced (2.62 ± 0.38) × 10 7 conidia per plate approximately 76.6% of that of the wild-type strain (Fig. 5A). Under the same culture conditions, the ∆Moatg5/FgATG5 strain produced (2.24 ± 0.16) × 10 7 conidia per plate,  approximately 65.5% of that of the wild-type strains (Fig. 5A). The results suggest that introduction of FgATG1 and FgATG5 can restore the growth and conidiation of ∆Moatg1 and ∆Moatg5 mutants. To determine the pathogenicity of the ∆Moatg1/FgATG1 and ∆Moatg5/FgATG5 strains, the susceptible barley cv ZJ-8 was inoculated with mycelial plugs from each strain. Like the wild-type Guy11, the ∆Moatg1/FgATG1 and ∆Moatg5/FgATG5 strains caused severe disease on the barley leaves, while the ∆Moatg1 and ∆Moatg5 mutants failed to cause lesions (Fig. 5B). These results indicate that introduction of FgATG1 and FgATG5 can functionally complement the defects of ∆Moatg1 and ∆Moatg5 mutants, respectively.
Autophagy is blocked in ∆Fgatg1 and ∆Fgatg5 mutants. To determine whether autophagy was affected by deletion of ATG genes in F. graminearum, the ∆Fgatg1 and ∆Fgatg5 mutants were used to observe autophagic bodies in vacuoles of hyphal cells by transmission electron microscopy. When strains were cultured in MM-N liquid medium in the presence of 2 mM PMSF for 4 h, autophagic bodies were observed clearly in the vacuoles of the wild-type strain PH-1 (Fig. 6A), while no autophagic bodies, or a few autophagosome-like structures were seen in vacuoles of ∆Fgatg1 and ∆Fgatg5 mutants (Fig. 6B,C). The results are consistent with the autophagic pathway being blocked in ∆Fgatg1 and ∆Fgatg5 mutants.
The GFP-Atg8 processing assay can be used to monitor autophagosome delivery 44 . The GFP-FgAtg8 fusion protein was constructed and expressed in the PH-1 strain, ∆Fgatg1 and ∆Fgatg5 mutants, respectively. When the strains were cultured in rich-nutrient conditions (CM liquid medium) for 24 h, we observed that GFP-FgAtg8 was expressed and dispersed throughout the cytoplasm and GFP-FgAtg8 signals were localized in the punctate structures close to vacuoles in both PH-1 and ∆Fgatg1 and ∆Fgatg5 mutants (Fig. 7A,B and C). However, when strains were transferred to nitrogen starvation conditions (MM-N liquid medium) with 2 mM PMSF and incubated for 4 h, we observed that GFP-FgAtg8 accumulated in vacuoles of the wild-type PH-1 (Fig. 7A), while in hyphal cells of the ∆Fgatg1 and ∆Fgatg5 mutants, GFP-FgAtg8 remained outside vacuoles in punctate structures (Fig. 7B,C). These results indicate that autophagosomes in hyphal cells of ∆Fgatg1 and ∆Fgatg5 mutants are defective in fusion with vacuoles and that the autophagic pathway is blocked in ATG mutants.
To further investigate impairment of the ATG mutants in autophagic processing, we tested processing of GFP-FgAtg8 using proteolysis assays. Under non-induction conditions (0 h), a clear full-length GFP-FgAtg8 band (40 kDa) and a free GFP band (26 kDa) were detected in both PH-1 and ∆Fgatg1 and ∆Fgatg5 mutants by immunoblotting with an anti-GFP antibody (Fig. 7D). We observed that weak free-GFP bands were present in the ∆Fgatg1 and ∆Fgatg5 mutants (Fig. 7D), which suggested that the autophagy process was not completely blocked in the ∆Fgatg1 and ∆Fgatg5 mutants. When hyphae of the wild-type PH-1 were shifted to nitrogen starvation conditions, an increasingly weaker GFP-FgAtg8 band, and an increasingly stronger free GFP band were detected as the induction time proceeded. By contrast, levels of full-length GFP-FgAtg8 in ∆Fgatg1 and ∆Fgatg5 mutants did not increase when the time of nitrogen-starvation was extended (Fig. 7D). This assay confirmed that GFP-FgAtg8 proteolysis was impaired in ∆Fgatg1 and ∆Fgatg5 mutants. Taken together, we conclude that the autophagy pathway is impaired in the ∆Fgatg1 and ∆Fgatg5 mutants and that this process is critical for fungal pathogenesis.

Discussion
In this study, we identified 28 putative autophagy-related genes in F. graminearum using the known genome-wide set of S. cerevisiae Atg protein-encoding genes as a guide. Using targeted gene deletions, the predicted F. graminearum ATG genes were functionally characterized at a global scale. We found that autophagy is required for vegetative growth, asexual/sexual sporulation, DON production and pathogenicity in F. graminearum. In addition, we investigated the biological function of FgATG1 and FgATG5 and found that the autophagic process was blocked in ∆Fgatg1 and ∆Fgatg5 mutants. Taken together, we conclude that autophagy plays a critical role in many important physiological functions of F. graminearum.
Autophagy is a conserved process from budding yeasts to mammalian cells, and a major cellular pathway for degradation of long-lived proteins and cytoplasmic organelles [45][46][47] , important for maintaining homeostasis of cells. Autophagy is triggered by starvation stress, and leads to rearrangement of subcellular membranes to sequester cargo for delivery to the lysosome or vacuole, where sequestered material is then degraded and recycled 45,48 . In addition to its homeostatic functions, autophagy is necessary for many aspects of development in multicellular organisms 49 . Previously, it has been observed that ATG mutants are impaired in mycelial growth and pathogenicity in several fungal species 17,18,50,51 . In F. graminearum, deletion of FgATG8 results in impairment of mycelial growth, usage of storage lipid droplets, formation of asexual/sexual spores and infection 20 . Similarly, FgATG15 is involved in aerial hyphal growth, conidiogenesis, lipid droplet degradation and virulence 31 . In this study, the mutants of the 28 autophagy-related genes in F. graminearum were classified into two groups according to the rate of vegetative growth on PDA medium. Deletion of any of the genes in Group 1 resulted in a statistically significant difference in linear growth on the PDA plates compared to the wild-type strain PH-1 (Fig. 1A,B). However, the mutants of the ATGs in Group 2 had similar growth rates compared to an isogenic wild-type strain (Fig. 1C,D). This suggests that genes associated with Group 1 play wider roles in cellular viability and hyphal growth than those in Group 2. However, since multiple gene deletion mutants of Group 2 have not been generated and analyzed in the study, it is unknown whether genes of group 2 are redundant or functionally overlap. Therefore, we reasoned that further clarification of gene function in Group 2 proteins, for example in cellular viability and hyphal growth in future, would be necessary. Genes required presumptively for the selective pathways in F. graminearum are also found in both groups, such as FgATG11, FgATG20, FgATG23, FgATG24 and FgATG33 in Group 1 and FgATG26, FgATG27, FgATG28 and FgATG37 in Group 2, indicating that the two groups with different growing patterns had no close relation to selective and non-selective types of autophagy. Conidial production of 11 tested ATG mutants was analyzed statistically and showed that these mutants produced significantly less or even no conidia ( Fig. 2A). Also, we found that perithecium development of ∆Fgatg1 and ∆Fgatg5 was completely blocked (Fig. 2B). Sporulation is associated closely with the energy metabolism of organisms 52 . Since autophagy provides nutrients by recycling cytoplasmic materials, a deficiency in this process probably reduces the production of conidia in consequence of the lack of energy and nutrients. Based on our pathogenicity assays, we found that most ATG mutants, but not FgATG17, exhibited a decrease in infection of wheat spikelets (Fig. 3). This affected not only symptom expression but also the ability of the fungus to spread to new spikelets. This suggests that the ability to colonise plant tissue is impaired, as well as the ability to produce spores to infect new hosts. Given these defects, it seems likely that autophagy is absolutely necessary for success of F. graminearum in the field and that the fungus would be unable to survive and cause disease without the operation of autophagy. Consistent with this loss of ability to colonise wheat tissue, we found that DON production of 13 tested ATG mutants was significantly reduced in comparison with PH-1 (Fig. 4), indicating that autophagy is necessary for fueling secondary metabolism in F. graminearum. However, the analysis of dynamic DON production for each of the ATG mutants was not carried out in the study. We cannot preclude at this stage that the lower levels of DON are associated with a delay in biosynthesis, rather than a reduction in the ability to synthesize the secondary metabolite. To date, at least 15 TRI genes in F. graminearum encoding trichothecene biosynthetic enzymes and regulators have been identified 53 . We found that the expression levels of TRI5 (trichodiene synthase), TRI6 (transcription regulator) and TRI10 (transcription regulator) genes in mycelium cultured for 24 h were significantly decreased in ∆Fgatg20 and ∆Fgatg24 mutants (data not shown). The data suggest that reduction of DON biosynthesis of ATG mutants is associated with the low expression of the TRI genes in these mutants. Since deoxynivalenol (DON) was an important virulence determinant in F. graminearum, and DON is necessary to suppress plant defense enabling the pathogen to break through the rachis node 54 , the reduction of DON production in these mutants may be a cause of their loss of virulence or may be associated with the relative lack of infection ability of the mutants.
In M. oryzae autophagy genes can be classified into those predicted to be required for nonselective autophagy and those necessary for pexophagy, mitophagy, or the Cvt pathway 17 . Loss of any of the 16 genes necessary for nonselective macroautophagy leads to M. oryzae being unable to cause rice blast disease, but the 6 genes necessary only for selective autophagy are dispensable for appressorium-mediated plant infection 17 . However, in the present study, deletion of any of the genes required presumptively for pexophagy, mitophagy, or the Cvt pathway (FgATG11, FgATG20, FgATG23, FgATG24, FgATG26, FgATG27, FgATG28, FgATG33 and FgATG37) and genes necessary for the nonselective macroautophagy (FgATG1, FgATG2, FgATG3, FgATG4, FgATG5, FgATG6,  FgATG7, FgATG8, FgATG9, FgATG10, FgATG12, FgATG13, FgATG14, FgATG15, FgATG16, FgATG18, FgATG22 and FgATG29) in F. graminearum leads to the significant reduction in virulence (Fig. 2). This suggests that both processes may be necessary for infection by F. graminearum. The functional difference of the ATGs in the two groups in plant infection between M. oryzae and F. graminearum may be due to DON production of these mutants in F. graminearum, which is significantly reduced and which plays such a critical role in plant infection. Interestingly, ∆Fgatg17 mutants were fully pathogenic, suggesting that this has a non-essential function in autophagy in F. graminearum. In S. cerevisiae, ∆atg17 mutants were almost completely defective in autophagy and produced few small autophagosomes that were less than half the normal size upon starvation 55,56 . However, we found that FgATG17 is not essential for GFP-FgAtg8 proteolysis and DON production in F. graminearum (Fig. S4).
In S. cerevisiae, Atg1 as a serine/threonine kinase involved in regulation of autophagy by protein phosphorylation 57 . Atg1 forms complexes with Atg13 and Atg17, which is required for the induction of autophagy 32 . Atg5 with Atg12 and Atg16 forms the Atg12-Atg5-Atg16 conjugation complex system and plays an essential role in the formation of autophagosomes [58][59][60] . In filamentous fungi, such as Podospora anserina, Aspergillus fumigatus and M. oryzae, homologs of ATG1 and ATG5 genes have been identified and characterized 18,25,61,62 . We found that re-introduction of FgATG1 and FgATG5 into M. oryzae Moatg1 and Moatg5 mutants functionally complemented the phenotypes of the mutants, respectively, suggesting that Atg1 and Atg5 proteins in filamentous fungi probably play a conserved role in regulation of conidiogenesis and pathogenicity. We found that autophagic bodies in the ∆Fgatg1 and ∆Fgatg5 mutants were absent or obviously decreased under starvation conditions (Fig. 6). Moreover, disruption of FgATG1 and FgATG5 prevented movement of GFP-FgAtg8 to the vacuolar lumen when autophagy was induced (Fig. 7B,C). Consistent with this, impairment of autophagy in the ∆Fgatg1 and ∆Fgatg5 mutants was confirmed by GFP-FgAtg8 proteolysis assays (Fig. 7D). These results suggest that the autophagic pathway was mostly blocked in ∆Fgatg1 and ∆Fgatg5 mutants. However, the detailed mechanism of how FgATG1 and FgATG5 are involved in autophagy in F. graminearum requires further study. In M. oryzae, deletion of MgATG1 gene influences the number of lipid bodies, and lipid storage in conidia in a ∆Mgatg5 mutant is reduced 18,25 . In Aspergillus oryzae, AoAtg1 is involved in the Cvt pathway 28 . Hence, further studies will be necessary to reveal the relationship between autophagy and the lipid metabolism in F. graminearum.
In summary, we conclude that autophagy-related genes (except FgATG17) are involved in regulating vegetative growth, aerial mycelium formation, asexual/sexual sporulation, DON production and virulence in F. graminearum.

Materials and Methods
Fungal strains and culture conditions. The wild-type F. graminearum strain PH-1 and all derivative mutants in this study were cultured on PDA (potato dextrose agar, 200 g potato, 20 g dextrose, 20 g agar per 1 L water) plates at 25 °C to assess mycelial growth and colony characteristics. Conidiation assays of all strains were performed after growing 4 days in 1% mung bean liquid (MBL) medium (10 g mung beans boiled in 1 L water for 20 min). Cultures in PDB (PDA without agar) were used for genomic DNA isolation. Complete medium (CM) and minimal medium without the nitrogen source (MM-N) were used for autophagy assays.
Generation of ATG deletion mutants. The DNA cassettes used for the gene deletions were constructed as described previously 63 . All PCR primers used in this study were listed in Table S2. The putative gene deletion mutants were identified and confirmed by PCR amplification and Southern blotting assays. Southern blot analysis was performed by the digoxigenin (DIG) high prime DNA labeling and detection starter Kit I (Roche, Mannheim, Germany).
SCientifiC RePoRTs | 7: 11062 | DOI:10.1038/s41598-017-11640-z Phenotypic analysis. For the vegetative growth of each colony, 5-mm plugs cut from the edge of a 3-day-old colony of each strain were placed on PDA plates and incubated at 25 °C. After 3 days, the diameter of each strain was measured and recorded. For the conidiation assay, five 5-mm plugs of each strain from the edge of 3-day-old colony were inoculated in 20 mL 1% MBL. After 4 d cultivation in a 180 rpm-shaker, conidia of PH-1 were harvested by filtering through cheesecloth and directly counted using a haemocytometer. The conidia of mutants were harvested using the same method and then centrifuged at 8000 rpm for 10 min. The harvested conidia of mutants were re-suspended in 1 mL sterile distilled water and subsequently counted with the haemocytometer. Each experiment with three replicates was independently repeated three times. For self-crossing assays, perithecium formation was assayed on carrot agar (CA) medium as previously described 64 . Pathogenicity assays. Since the F. graminearum ATG mutants produced few conidia, mycelial plugs taken from the PDA plates were applied for pathogenicity assays. Agar plugs with mycelium from 3-day-old PDA plates of PH-1 or mutants were scraped off with dentiscalprum, and the middle spikelet of flowering wheat heads of the susceptible cultivar Jimai33 was inoculated. Symptomatic spikelets and quantification of infected spikelets among the whole wheat heads were determined after incubation for 14 days.
DON production assays. After immersion in water for 12 h, 50 g wet wheat kernels were sterilized for 3 times and inoculated with ten mycelial plugs (5 mm in diameter) of each strain from the edge of 3-day-old colony and incubation at 25 °C for 25 days. The inoculated wheat kernels were dried at 37 °C for 24 h, and then broken into powder with juicer. The powder was delivered to the company named Pribolab biological engineering co., LTD in Qingdao which performed the DON production assays.

Western blot analysis.
A piece of agar blocks with mycelia of each tested strain was introduced into 20 mL of liquid CM medium. The suspension was shaken at 25 °C, 180 rpm for 24 h, and then transferred into the MM-N medium in the presence of 2 mM PMSF (phenymethylsulfonyl fluride) for 4 h, 8 h and 12 h. Hyphae of each sample were harvested, washed with sterile distilled water, ground into powder in liquid nitrogen, and then suspended in the protein lysis buffer (50 mM Tris-Cl (pH = 7.4), 0.15 M NaCl, 1 mM EDTA, 1% Triton × 100) added 1% 100 × Protease Inhibitor Cocktail for Fungal/Yeast Cell (Sangon, Shanghai, China). The lysate was centrifuged at 13200 rpm for 20 min at 4 °C after lysis for 30 min. Afterwards, 50 μL supernatant was mixed with isovolumetric 2 × protein loading buffer and boiled for 5 min and then cooled on the ice immediately. 10-15 μL of each sample was taken for loading on 10% SDS-PAGE gels. GFP-Tag (7G9) Mouse mAb (Abmart, Shanghai, China) as primary antibody was used at a 1:5000 dilution. HiSec TM HRP-conjugated Goat Anti-Mouse IgG (H + L) (Vazyme, Nanjing, China) was applied to immunoblot analysis at a 1:10000-1:50000 dilution. FD TM FDbio-Femto Ecl chemiluminescent substrate (Fdbio science, Hangzhou, China) was used for antigen antibody detections.
Complementation analysis of Moatg1 and Moatg5 by introducing FgATG1 and FgATG5 respectively. The pCB1532-FgATG1 and pCB1532-FgATG5 vectors for the complementation of Moatg1 and Moatg5 mutants respectively were constructed as described previously 63 .

Confocal microscopy and transmission electron microscopy assays. Hyphae express-
ing the GFP-Atg8 fusion protein were cultured in CM medium at 25 °C, 180 rpm for 24 h, and then transferred into the MM-N medium in the presence of 2 mM PMSF for 4 h at 25 °C in a 180 rpm shaker. CMAC (7-amino-4-chloromethycoumarin) (Invitrogen, USA) was used for vacuole staining. Photographs were taken under a confocal laser scanning microscopy. Transmission electron microscopy was implemented as described previously 18 . Data Availability. All data generated or analysed during this study are including in this published article (and its Supplementary Information files).