FgMon1, a guanine nucleotide exchange factor of FgRab7, is important for vacuole fusion, autophagy and plant infection in Fusarium graminearum

The Ccz1-Mon1 protein complex, the guanine nucleotide exchange factor (GEF) of the late endosomal Rab7 homolog Ypt7, is required for the late step of multiple vacuole delivery pathways, such as cytoplasm-to-vacuole targeting (Cvt) pathway and autophagy processes. Here, we identified and characterized the yeast Mon1 homolog in Fusarium graminearum, named FgMon1. FgMON1 encodes a trafficking protein and is well conserved in filamentous fungi. Targeted gene deletion showed that the ∆Fgmon1 mutant was defective in vegetative growth, asexual/sexual development, conidial germination and morphology, plant infection and deoxynivalenol production. Cytological examination revealed that the ∆Fgmon1 mutant was also defective in vacuole fusion and autophagy, and delayed in endocytosis. Yeast two hybrid and in vitro GST-pull down assays approved that FgMon1 physically interacts with a Rab GTPase FgRab7 which is also important for the development, infection, membrane fusion and autophagy in F. graminearum. FgMon1 likely acts as a GEF of FgRab7 and constitutively activated FgRab7 was able to rescue the defects of the ∆Fgmon1 mutant. In summary, our study provides evidences that FgMon1 and FgRab7 are critical components that modulate vesicle trafficking, endocytosis and autophagy, and thereby affect the development, plant infection and DON production of F. graminearum.


FgMon1 plays a critical role in growth and conidiogenesis.
To investigate the roles of FgMon1 in F. graminearum, the FgMON1 gene replacement construct was generated by split marker approach ( Figure S2A) and transformed into the protoplast of wild type strain PH-1 as previously described 29 . The resulting transformants were screened by PCR and further confirmed by Southern blot analysis ( Figure S2B). We first checked the growth and colony morphology of the ∆Fgmon1 mutant. Compared to the wild type PH-1 and complemented transformant ∆Fgmon/FgMON1, the ∆Fgmon1 mutant showed less aerial hyphae and significantly reduced growth rate on V8, 5xYEG, CM and MM agar plates (Fig. 1A, Table 1). Conidial production of the ∆Fgmon1 mutant was quantified in CMC media and the number of conidia in the mutant was decreased to 17% of the wild type PH-1 (Table 1). Microscopy observation revealed that the ∆Fgmon1 mutant was also defective in conidial morphology. Conidia of the ∆Fgmon1 mutant were shorter and had fewer septa in comparison to the wild type and the complemented transformant. More than 50% conidia of the ∆Fgmon1 mutant have only one or two septa, while 91% conidia of PH-1 have three septa or more. The average conidial length of the ∆Fgmon1 mutant was 69% of that the wild type (Fig. 1B, Table 1). These results indicated that FgMon1 is important for vegetative growth, conidiation and conidial morphology.

FgMon1 is involved in conidial germination and is essential for sexual reproduction. To exam-
ine whether FgMon1 has a role in conidial germination, we examined the conidial germination in liquid YEPD media. The result showed that conidia of the ∆Fgmon1 mutant were able to germinate from the end cells but delayed to germinate from middle cells ( Fig. 2A). Even when incubation for 16 h, multipolar germination rate of the mutant conidia was remain significantly lower than that of the wild type (Table 2). After 18 h incubation, the mutant and wild type showed a similar multipolar germination rate ( Table 2). Because ascospores play a crucial role in the disease cycle of F. graminearum, we also assayed sexual reproduction of PH-1, ∆Fgmon1 mutant and ∆Fgmon/FgMON1 on carrot agar plates as previously described 13 . After 10 days of inoculation, the PH-1 and the complemented tranformant produced numerous mature perithecia. In contrast, the ∆Fgmon1 mutant failed to produce perithecia under the same conditions (Fig. 2B). These results suggested that FgMon1 is involved in conidial germination and plays an essential role in sexual reproduction in F. graminearum.

FgMon1 is important for plant infection and DON production.
To determine the role of FgMon1 in plant infection, we first assayed the ∆Fgmon1 mutant on wheat germs by droplet inoculation. After incubation at 25 °C for 10 days, severe disease symptoms were observed on the wheat coleoptiles inoculated with conidia Scientific RepoRts | 5:18101 | DOI: 10.1038/srep18101 suspensions prepared from wild type and complemented transformant. In contrast, the ∆Fgmon1 mutant almost caused no symptoms on the wheat coleoptiles under the same conditions (Fig. 3A). Microscopy examination revealed that the mutant was unable to penetrate through the wheat coleoptile epidermis and no infectious hyphae were observed in plant cells, while the wild type and complemented transformant formed branching and expanded infectious hyphae in the cells (Fig. 3B). We further point-inoculated the flowering wheat heads with conidial suspensions. The ∆Fgmon1 mutant was also defective in plant infection. Wheat kernels nearby the inoculation sites remained healthy 14 days following inoculation, while all wheat kernels in the inoculated spikelets were infected by the wild type and complemented transformant (Fig. 3C). These results indicated that FgMon1 plays a critical role in plant infection in F. graminearum. Because DON was known as an important virulence factor in F. graminearum 30 , DON production was measured in the wheat kernels infected by PH-1 and the ∆Fgmon1 mutant. DON production was at a very low level in the ∆Fgmon1 mutant, only 0.016 mg of DON was detected in per milligram of ergosterol, while over 4 mg of DON in per milligram of ergosterol was detected in the wild type PH-1 (Table 1), suggesting FgMon1 has a critical role in DON production. We further test the expression level of trichothecene synthase genes TRI5 and TRI6 that are involved in DON biosynthesis. qRT-PCR analysis revealed that the expression of TRI5 (decreased to 60% of the wild type) and TRI6 (decreased to 50% of the wild type) was significantly decreased in the ∆Fgmon1 mutant. These results indicate that FgMon1 modulates DON biosynthesis by regulating the expression of TRI5 and TRI6 in F. graminearum. Expression and intracellular localization of GFP-FgMon1 in F. graminearum. To examine the expression and localization pattern of the GFP-FgMon1 proteins in F. graminearum, a GFP-FgMON1 fusion construct was generated and transformed into the ∆Fgmon1 mutant. The conidia and germ tubes from the resulting transformant were observed under a fluorescence microscopy. Strong GFP signals were present mainly in the punctate structures of the cytosol, and weak fluorescence was observed in cytosol in both conidia and germ tubes. We further stained the conidia and germ tubes with CMAC (7-amino-4-chloromethylcoumarin), a dye that labels the lumen of fungal vacuoles. The results showed that CMAC exactly stained the above punctate structures, indicating GFP-FgMon1 localized in the vacuoles in F. graminearum (Fig. 4A,B).    Table 2. Germination of the wild type, ∆Fgmon1 and ∆Fgrab7 mutants in YEPD medium. ± SD was calculated from three repeated experiments and asterisks indicate statistically significant differences (p < 0.01). NA, not assayed.  FgMon1 is indispensable for autophagy in F. graminearum. The autophagy process is regulated by multiple autophagy-related proteins. Since the autophagy process also includes membrane trafficking and fusion events, proteins involved in vesicle trafficking such as SNAREs and Rab GTPases have been reported to be essential in autophagy 23,24 . Because the ∆Fgmon1 mutant was defective in endocytosis and vacuole fusion, we suppose that the mutant might also have defect in autophagy pathway. To test this possibility, vacuoles of hyphal cells were examined with starvation induction assays. We first examined the autophagic bodies under transmission electron microscopy. After cultured in liquid MM-N medium with 2 mM PMSF for 4 h, no autophagic bodies in the vacuole of the ∆Fgmon1 mutant was observed. However, numerous autophagic bodies were observed in the vacuole of wild type PH-1 (Fig. 6A). The autophagic process could be tracing-observed by monitoring the vacuolar delivery and breakdown of GFP-Atg8 31 . Under non-induction conditions (CM medium) for 10 h, GFP-FgAtg8 was localized in the punctuate structures in both wild type and the ∆Fgmon1 mutant, while in wide-type some of the punctuate structures are delivered to the vacuole for degradation. When induced under nitrogen starvation (MM-N medium) condition in the presence of 2 mM PMSF for another 8 h, GFP-FgAtg8 accumulated in the vacuoles of the wild type. However, GFP signals remain exist in the punctuate structures of the mutant but not in the surrounding CMAC stained fragmented vacuoles (Fig. 6B). We concluded that the fusion of autophagosomes and vacuoles was impaired in the Δ Fgmon1 mutant. To further explain this observation, GFP-FgAtg8 proteolysis assay was performed. Under normal conditions, a clear full-length GFP-FgAtg8 band (40 kDa) and a slightly weak GFP band (26 kDa) was detected in the wild type with an anti-GFP antibody (Fig. 6C). When hyphae were shifted to MM-N conditions, a relatively weak full-length GFP-FgAtg8 band but a very clear GFP band was detected in the wild type. In comparison, a clear full-length band and a hardly-detected GFP band were detected in the ∆Fgmon1 mutant regardless of cultural conditions (Fig. 6C). These results implicated the ∆Fgmon1 mutant was defective in autophagy.
FgMon1 plays an important role in response to vesicular transport inhibitor and cell wall perturbing agents. Because the ∆Fgmon1 mutant was defective in vacuole fusion, we speculated that this defect might affect the vesicular transport pathway and cell wall integrity. Therefore, the wild type PH-1, ∆Fgmon1 mutant and complemented transformant were inoculated onto the CM plates with cell wall perturbing agents (0.03% CFW, 0.01% SDS) and the drug that interferes with intracellular protein transport processes (0.0001% monensin). After 3 days incubation, the ∆Fgmon1 mutant showed an extremely small colony in comparison to that of the wild type on the plates (Fig. 7A). The growth inhibition rate of the mutant was increased 2-, 3-and 1.5-fold on CFW, SDS and monensin plates, respectively (Fig. 7B), indicating the ∆Fgmon1 mutant was hypersensitive to vesicular transport  inhibitor and cell wall perturbing agents. It also implicates that FgMon1 had a role in vesicular transport pathway as well as maintenance of the cell wall integrity in F. graminearum.
Constitutively activate FgRab7 could rescue the defects of the ∆Fgmon1 mutant. Mon1-Ccz1 complex was known as the GEF of Rab7 homolog Ypt7 in yeast 25 . A latest study reported that Rab GTPase FgRab7 was essential for membrane trafficking-dependent growth and plant infection in F. graminearum 23 . Our independent work also showed that FgRab7 was important for the development of infection related morphogenesis, vacuole fusion and autophagy that is similar to the biological functions of FgMon1 ( Figures S3 and S4). To figure out whether FgMon1 was a GEF of FgRab7 in F. graminearum, a construct encoding a constitutively activated FgRab7 Q67L (GTP hydrolysis defective) was transformed into the protoplast of the ∆Fgmon1 mutant. The resulting transformant ∆Fgmon1/FgRAB7 Q67L was confirmed by qRT-PCR and showed 3.2-fold increased expression of FgRAB7 compared to the wild type PH-1. Phenotype analysis revealed that the ∆Fgmon1/FgRAB7 Q67L transformant displayed normal vegetative growth, conidiation, conidial morphology and virulence as the wild type PH-1 (Fig. 8A,B). Furthermore, the ∆Fgmon1/FgRAB7 Q67L transformant also showed normal endocytosis and vacuole fusion by cytological examination (Fig. 8C,D). These results indicated that constitutively activated FgRab7 could rescue the defects of the ∆Fgmon1 mutant. In addition, we also transformed pYF11-FgRAB7 Q67L into the protoplast of wide-type PH-1. The resulting transformants WT/FgRAB7 Q67L were confirmed by qPCR, and showed a 3.3-fold increase of FgRAB7 expression compared with the wild-type. Phenotype analysis revealed that WT/FgRAB7 Q67L showed no obvious changes on vegetative growth, conidiation, conidial morphology, pathogenicity as well as vacuole morphology and endocytosis (Table 1, Fig. 8A-D).
FgMon1 physically interacts with FgRab7, FgRab7 Q67L and FgRab7 T22N . To clarify the relationship between FgMon1 and FgRab7, yeast two hybrid (Y2H) and in vitro GST-pull down assays were carried out to test whether they interact with each other. The pGBKT7-FgMon1 bait and pGADT7-FgRab7 prey constructs were generated and co-transformed into yeast cell AH109. The result showed that FgRab7 interacts with FgMon1 in Y2H assay (Fig. 9A). This interaction was further confirmed by GST-pull down assay using GST-FgRab7 and His-FgMon1 fusion proteins (Fig. 9B), suggesting the direct association between FgRab7 and FgMon1. To further analyze the relationship between FgMon1 and FgRab7, we constructed GTP-associated version FgRab7 Q67L and GDP-associated version FgRab7 T22N . Both Y2H and in vitro GST-pull down assays showed that FgMon1 interacts with FgRab7 Q67L and FgRab7 T22N , respectively (Fig. 9C,D).

Discussion
In S. cerevisiae, the Mon1-Ccz1 complex was found to function in cytoplasm to vacuole targeting (Cvt) pathway and autophagy pathway 7 . In addition, the complex also plays a role in endosomal membrane fusion machinery 8 . Recent studies reported that the Mon1-Ccz1 complex serve as the Rab7 GEF 25 . However, in our study only the MON1 homolog but not CCZ1 were found in F. graminearum, indicating other components might replace CCZ1 or CCZ1 might function redundantly in this pathogen. Besides, phylogenetic analysis showed Mon1 was well conserved in filamentous fungi. In addition, we generated the FgMON1 gene deletion mutants of F. graminearum. The multiple defects of the ∆Fgmon1 mutant implicate that FgMon1 is a key protein for the development, infection and DON production of F. graminearum. Since DON is an important virulence factor in F. graminearum, in addition to its reduced growth rate, defects of the ∆Fgmon1 mutant in DON biosynthesis may also contribute to its defects in plant infection. In the rice blast fungus M. oryzae, MoMon1 is known to be involved in conidial morphology 28 . Similar to this, we found conidial morphology of the Δ Fgmon1 mutant was also changed when compared with the wide type PH-1. Besides, the ∆Fgmon1 mutant was hypersensitive to cell wall perturbing agents, indicating FgMon1 might have a role in cell wall integrity maintenance. Many genes involved in membrane fusion of the endomembrane system have been reported to have important roles in the development and pathogenicity of phytopathogens. Deletion of MON1 in M. oryzae resulted in defects in vegetative growth, sporulation, autophagy, appressoria formation, pathogenicity and massive vacuole fragmentation 28 . SNARE proteins MoVam7 and MoSec22, play crucial roles in hyphal growth, conidiation, vacuole morphology, virulence and endocytosis 32,33 . Recent studies showed that FgVam7, FgYpt7, MoYpt7 plays roles similar to those of MoVam7, MoSec22 and MoMon1 13,23,24 , indicating that the proteins function in endomembrane system are important for the correct regulation of infection-related morphogenesis in different fungi. In yeast, GFP-tagged Mon1 and Ccz1 mutants were found in punctate structures, which probably represent endosomes 7,25,34 , while in Arabidopsis, GFP-Mon1 mainly showed a cytosolic and endosomal localization 15 . Our data show that GFP-Mon1 localized in the cytosol and vacuoles in F. graminearum. This would be compatible with the recruitment of Mon1 to some membranes, as expected for a function in vacuolar fusion of Cvt vesicles and autophagosomes. In M. oryzae and F. graminearum, Rab7 proteins are thought to be localized to the vacuolar membrane, similar to that of in S. cerevisiae and Arabidopsis 15,23,24,35,36 . In yeast and animal cells, maturation of late endosomes from early endosomes require the conversion of Rab5-to-Rab7. Mon1-Ccz1 complex, the effectors of Rab5, could be recruited to the membrane by activated Rab5 protein and then bind Rab7. In addition, the Mon1-Ccz1 complex also influence Rab7 activation, and acts as an important link between Rab5 and Rab7 25,37,38 . Theoretically, wild-type Rab7 should contain a GDP-bound form that can be recruited by the Mon1-Ccz1 complex. However, the data in yeast and Arabidopsis revealed that the complex only interact with the GDP-locked version of Rab7 T22N , which could due to the possible transient interaction between wild-type Rab7 and the Mon1-Ccz1 complex 15,34 . Our results showed that FgMon1 directly interacts with FgRab7 which showed similar phenotypes to that of FgMon1. In addition, we observed that FgMon1 specific interacts with GTP-associated version FgRab7 Q67L and GDP-associated version FgRab7 T22N in both Y2H and in vitro GST-pull down assays. This result is consistent with what have been found in Caenorhabditis elegans 37 . The Mon1:Ccz1 complex appears to facilitate the displacement of GDI (Rab Guanine Nucleotide Dissociation Inhibitors) from Rab7 and promote GTP loading of Rab7 37 . Wild-type FgRab7 should contain a GDP-bound form and GTP-bound form that can interact with FgMon1. Therefore, we speculate that FgMon1 may facilitates GTP loading of FgRab7 as well as GDI release, and a dynamic balance likely exist between FgMon1, FgRab7 Q67L and FgRab7 T22N , which might regulate the activity of Rab7. Besides, constitutively activated FgRab7 could rescue the defects of the ΔFgmon1 mutant, indicating that FgMon1 likely is a GEF of FgRab7, which is similar to that found in yeast. However, constitutively activated FgRab7 in the wide-type PH-1 caused no phenotypic changes, indicating the active form and the negative form of FgRab7 coexistence in the wide-type PH-1, and the transformation between both forms might be transient and in a dynamic balance.
In yeast cells several different transport pathways converge upon the vacuole and the Cvt process overlaps with macroautophagy, which non-selectively deliver cytosolic proteins and organelles to the vacuole for degradation and recycling 39 . Therefore, we speculated that fragmented vacuoles of the ΔFgmon1 mutant probably account for the delayed endocytosis, thereby influencing endosomal membrane fusion. In autophagy pathway, assembled autophagosome transported toward the vacuole and fuses with the membrane of vacuole to release the inner membrane structure and cargo 40 . We wonder whether the defect in vacuole morphology of ΔFgmon1 could influence the autophagy pathway or not. Autophagy is a process that cytoplasmic components and organelles of a cell are delivered to lysosomes for degradation. Under nutrient-deprived conditions, autophagy can be induced for cell survival. It is also a conserved mechanism from yeast to humans 41,42 . For example, autophagic dysfunction is associated with cancer, neurodegeneration, microbial infection and ageing 43 . In fact, many studies have been carried out using the N-terminal GFP-tagged Atg8 to monitor autophagy in yeast, mammals and filamentous fungi 40,44 . In our study, GFP-FgAtg8 cannot be delivered to fragmented vacuoles to degradation. In previous studies, Rab7 is also reported to be required for the fusion of autophagosome to the vacuole in yeast and other species 24,45,46 , and combined with our results, we concluded that the ΔFgmon1 mutant showed a defective in the fusion of autophagosomes and vacuoles which may also due to the fragmented vacuoles of the mutant. Fusion of autophagosomes with the vacuole and breakdown of the single-membrane autophagic body in the vacuole are critical steps in the autophagy pathway 24 . Therefore, we conclude that deletion of FgMon1 affects vacuole fusion, thus influence endocytosis and autophagy, and eventually affects the development and infection of the F. graminearum.
Taken together, we have identified and characterized FgMON1, a gene encoding a vacuolar fusion protein in F. graminearum, is important for hyphal growth, sexual reproduction, pathogenesis, vacuole fusion, endocytosis and autophagy. We also provide evidences that FgMon1 might act as a GEF of FgRab7 and directly interaction with FgRab7 in F. graminearum. However, relationships and interaction mechanisms between FgMon1 and FgRab7 need further studies.

Methods
Fungal strains and growth conditions. The wild type F. graminearum strain PH-1 and all other strains generated in this study were cultured on V8 juice agar plates at 25 °C. Cultures for genomic DNA and RNA isolation, conidiation in CMC medium and growth assays on CM, MM, 5xYEG media were performed as previously described 32 . Complete medium (CM) with SDS, CFW or monensin was used for stress response assays. For sexual reproduction, aerial hyphae of 10-day-old carrot agar cultures of the indicated strains were pressed down with 300 μ l of sterile 0.1% Tween 20 as described 47 .

Plant infection and DON production assays.
For plant infection assays, conidia from 3-day-old CMC cultures were harvested and resuspended to 10 6 /ml or 10 5 /ml in sterile distilled water with 0.2% gelatin. Wheat germs were inoculated with 2 μ l conidial suspensions (10 6 /ml) and examined at 10 dpi. Flowering wheat heads of cultivar Annong 8455 were drop-inoculated with 10 μ l of conidium suspensions (10 5 /ml) at the sixth spikelet from the base of the spike. 10 μ l of 0.2% gelatin served as controls. Symptomatic spikelets were examined and counted 14 dpi. For each treatment, 15 wheat heads were inoculated. For DON production assay, 50 g healthy wheat kernels was sterilized and inoculated with five mycelial plugs of each strain and incubation at 25 °C for 20 d. DON extraction and DON production quantification was performed as previously described 48 . qRT-PCR analysis. Total RNA samples were isolated from vegetative hyphae of PH-1 and ∆Fgmon1 mutant cultured in liquid YEPD for 2 days, and used for cDNA synthesis with the HiScript Q Select RT SuperMix for qPCR kit (Vazyme Biotech, Nanjing, China) following the instructions. The RT2 PCR Real-Time SYBR Green/ ROX PCR master mix (TaKaRa, Dalian, China) was used for qRT-PCR analysis. Primer pairs TRI5QF/TRI5QR and TRI6QF/TRI6QR 13 were used to amplify the TRI5 and TRI6 genes, respectively. The relative quantification of each transcript was calculated by the 2 -ΔΔCT method 49 with the F. graminearum beta-tubulin gene TUB2 as the internal control. For each gene, qRT-PCR assay repeated three times with three biological replicates.
Generation of the GFP-FgMON1, GFP-FgRAB7 and FgRAB7 Q67L constructs. Fragment including the entire FgMON1 or FgRAB7 gene and its native promoter region, was amplified by PCR with primers from PH-1. The product was then cloned into pYF11 by the yeast gap repair approach 50 . The resulting plasmids were confirmed by sequencing analysis to contain the in-frame fusion constructs and transformed into the ∆Fgmon1 mutant, respectively. The resulting zeocin-resistant transformants were screened by PCR or confirmed by the presence of GFP signals. The primers are listed in Table S1.
Yeast two hybrid and in vitro pull down assays. To examine the interaction between FgMon1 and FgRab7 using yeast two hybrid assays, the coding sequence of each tested gene was amplified from the cDNA of PH-1. The cDNA fragment of FgMON1 was inserted into pGBKT7 as the bait construct, while the cDNA fragment of FgRAB7, FgRAB7 Q67L and FgRAB7 T22N were cloned into pGADT7 as the prey construct. The pairs of the plasmids were co-transformed into the yeast strain AH109. In addition, a pair of plasmids, pGBKT7-53 and pGADT7-T, served as a positive control. The following pairs of plasmids were used as negative controls: pGBKT7-Lam and pGADT7-T; pGBKT7 and pGADT7-FgRab7; pGBKT7 and pGADT7-FgRab7 Q67L ; pGBKT7 and pGADT7-FgRab7 T22N ; pGADT7 and pGBKT7-FgMon1. Transformants were grown at 30 °C for 3 d on SD-Leu-Trp medium, and then transferred to the medium SD-Leu-Trp-Ade-His medium and containing 50 mM 3-aminotriazole (3-AT) to assess binding activity. The interaction was further examined by performing β-galactosidase activity using X-α -gal (80 μ g/L). For the in vitro GST pull-down assay, the full-length cDNA of FgRAB7, FgRAB7 Q67L , FgRAB7 T22N and FgMON1 was inserted between the EcoRI and XhoI sites of vector pGEX-4T-2 and pET-32a, respectively. The resulting plasmids GST-FgRab7 and His-FgMon1 were separately introduced into the E. coli strain BL21. Soluble proteins were incubated with 30 μ l glutathione agarose beads (Invitrogen) for 4 h at 4 °C. The beads were washed three times and then incubated with an equal amount of bacterial lysates containing His-FgMon1 for another 4 h at 4 °C. The beads were washed three times again, and the presence of His-FgMon1 was detected by immunoblot (IB) using anti-His antibody.
Confocal microscopy and transmission electron microscopy assays. For endocytosis assay, hyphae were cultured in liquid YEPD medium for 12 h and stained by FM4-64 (N-3-triethylammoniumpropyl-4-p-diethylamino-phenyl-hexa-trienyl pyridinium dibromide) (Molecular Probes, USA) following the procedures described previously 51 . For vacuole staining, hyphae, conidia and germinated conidia were stained by CMAC (7-amino-4-chloromethylcoumarin) (Molecular Probes, USA) as described 52 . Photographs were taken under a confocal laser scanning microscopy. For autophagy assay, mycelium cultured in liquid CM medium for 10 h and then transferred to the nitrogen-limiting medium (MM-N) in the presence of 2 mM PMSF for 8 h. Transmission electron microscopy was carried out as previously described 32 .