Posttranslational hypusination of the eukaryotic translation initiation factor-5A regulates Fusarium graminearum virulence

Activation of eukaryotic translation initiation factor eIF5A requires a posttranslational modification, forming the unique amino acid hypusine. This activation is mediated by two enzymes, deoxyhypusine synthase, DHS, and deoxyhypusine hydroxylase, DOHH. The impact of this enzymatic complex on the life cycle of a fungal pathogen is unknown. Plant pathogenic ascomycetes possess a single copy of the eIF5A activated by hypusination. We evaluated the importance of imbalances in eIF5A hypusination in Fusarium graminearum, a devastating fungal pathogen of cereals. Overexpression of DHS leads to increased virulence in wheat, elevated production of the mycotoxin deoxynivalenol, more infection structures, faster wheat tissue invasion in plants and increases vegetatively produced conidia. In contrast, overexpression of DOHH completely prevents infection structure formation, pathogenicity in wheat and maize, leads to overproduction of ROS, reduced DON production and increased sexual reproduction. Simultaneous overexpression of both genes restores wild type-like phenotypes. Analysis of eIF5A posttranslational modification displayed strongly increased hypusinated eIF5A in DOHH overexpression mutant in comparison to wild type, and the DHS overexpression mutants. These are the first results pointing to different functions of differently modified eIF5A.

Overexpression of DHS and/or DOHH disturbed conidiation, germination and perithecia production. To assess their importance in eIF5A hypusination, targeted DHS or DOHH overexpressing mutants (DHSoex, DOHHoex) were generated by single crossover event inserting the linearized vectors pMW-DHS or pMW-DOHH (Supplementary Table S1, Supplementary Fig. S3A,B). Southern blots confirmed the replacement of the endogenous DHS and DOHH genes by homologous recombination with the PgpdA-DHS or Pgpd1-DOHH alleles, excluding any possibility of mutagenic effects by ectopic integration ( Supplementary  Fig. S3C). In addition, a double overexpressing mutant (DHSoex/DOHHoex) was generated by transforming the DHSoex overexpressing mutant with the pMW-DOHH construct. The transformants showing homologous recombination were used in further analysis. We confirmed overexpression of the DHS and DOHH genes by reverse transcription-quantitative polymerase chain reaction (RT-qPCR) performed on RNA extracted from mycelia of the overexpressing mutants and the wild type strain. In the single overexpressing mutants, expression levels of DHS or DOHH were 60-or 65-fold higher than in the wild type strain. In the double overexpressing mutant expression levels of DHS and DOHH were up-regulated 30-and 25-fold compared to the wild type. Furthermore, expression levels of EIF5A in the overexpressing mutants were unaffected as well as the levels of DOHH in the DHSoex mutant or levels of DHS in the DOHHoex mutant ( Supplementary Fig. S3D).
High levels of DHS or DOHH proteins in the respective overexpressing mutants were confirmed by Western blots using antibodies against human DHS or F. graminearum DOHH. In the DHSoex and double overexpressing mutants DHS protein levels were 55% and 45% more than in the wild type or DOHHoex mutant. DOHH protein levels were 31% and 26% more than in the wild type or DHSoex mutant ( Supplementary Fig. S4).
To characterize DHSoex and DOHHoex mutants, we first assessed two key aspects of the pathogen's life cycle, conidia production from vegetative asexual growth and perithecia production from sexual reproduction, as well as assessing the respective germination of conidia and ascospores. Conidia morphology of the overexpressing mutants did not differ from the wild type strain. However, the DHSoex mutant produced 30% more conidia, the DOHHoex produced 54% less conidia and the double overexpressing mutant showed levels of conidiation similar to the wild type. The germination rate (%) was determined by counting bipolar, unipolar, or non-germinated conidia. The DHSoex and DHSoex/DOHHoex mutants showed wild type germination rates. However, germination in the DOHHoex mutant was highly impaired: bipolar germination decreased, unipolar germination remained similar, and non-germination increased drastically compared to the wild type strain. Perithecia production was quantified by counting the number of perithecia per square cm that had formed on wheat nodes. Surprisingly, the DOHHoex mutant produced 5 times more perithecia than the wild type, while wild type, DHSoex and DHSoex/DOHoex strains formed similar numbers of perithecia. In addition, when we determined the viability of discharged mature ascospores by germinating them in water, all strains produced similar numbers of viable ascospores ( Supplementary Fig. S5).

DHS and DOHH overexpression alters stress responses and ROS production.
To determine whether the overexpressing mutants have wild type-like responses to different stresses, they were subjected to phenotypic analysis on CM plates containing components such as sodium chloride (1.5 M NaCl) for osmotic stress, hydrogen peroxide (20 mM H 2 O 2 ) for oxidative stress and calcofluor white (0.5 mM CFW) for cell wall stress. The DHSoex mutant was more sensitive to oxidative stress than the wild type strain, and the DOHHoex mutant was highly sensitive to osmotic stress and cell wall stress. The phenotype of the double overexpressing mutant was similar to the wild type. In addition, the fungicides tebuconazole, which impairs ergosterol synthesis, and azoxystrobin, which prevents ATP generation in the mitochondria, were added to complete media to analyze fungicide resistance. Whereas the DHSoex mutant was more sensitive to tebuconazole, the DOHHoex mutant was more resistant to both tebuconazole and azoxystrobin than the wild type strain. The double overexpressing mutant's phenotype was similar to the wild type ( Supplementary Fig. S6 and Supplementary Table S2).
Based on the phenotypes revealed in the stress-related tests above, we compared ROS production in the overexpressing mutants and wild type strain. We qualitatively measured superoxide (O 2 − ) by staining mycelia grown on CM plates for 3 days with nitroblue tetrazolium (0.2% NBT), which is reduced upon exposure to superoxide anions to produce a dark blue water-insoluble formazan precipitate. Overexpression of DOHH caused more production of superoxide radicals than in the DHSoex and DHSoex/DOHHoex mutants or wild type strain (Fig. 1A). In addition, we quantified the production of hydrogen peroxide (H 2 O 2 ) in mycelia grown for 3 days in liquid CM using an AmplexRed-kit (Invitrogen). Similar to the above result, overexpression of DOHH caused a 3-fold over-production of H 2 O 2 . In contrast, overexpression of DHS, or both DHS and DOHH together produced levels of H 2 O 2 similar to the wild type strain (Fig. 1B).
DHSoex is hyper-virulent but DOHHoex lacks virulence towards wheat. We explored the pathogenicity towards wheat by infecting the susceptible cultivar Nandu with the overexpressing mutants and the wild type strain, and evaluate fungal virulence at 7, 14 and 21 days post inoculation (dpi). Spikes infected with DHSoex, DHSoex/DOHHoex and the wild type strains showed full infection and bleaching at 21 dpi. In contrast, infection symptoms in spikes inoculated with the DOHHoex mutant were only visible at the inoculated spikelets. Mock inoculation with water resulted in no symptoms ( Fig. 2A). To determine whether the DHSoex is indeed to determine hydrogen peroxide (H 2 O 2 ) production using the Amplex ® Red Hydrogen Peroxide/Peroxidase Assay Kit (Invitrogen). Error bars indicate standard deviations calculated from data per triplicate samples and are representative of three independent experiments. Significance with respect to wild type: *p < 0.05, **p < 0.01, ****p < 0.0001 (calculated with ANOVA-Bonferroni-Holm). more virulent than the wild type, we performed experiments on the partially resistant cultivar Melissos. Infection symptoms in wheat spikes inoculated with the DHSoex mutant expanded faster and further than in spikes infected by the wild type or DHSoex/DOHHoex strains. Again, the DOHHoex mutant only produced symptoms in the inoculated spikelets (Fig. 2B). Monitoring the progression of infection indicated that the DHSoex mutant had a slightly faster infection rate than the wild type strain in the cultivar Nandu (Fig. 2C). Furthermore, the DHSoex mutant was the only one able to fully infect the partially resistant cultivar Melissos (Fig. 2D).
Infection-defective mutants are able to invade caryopses, but cannot trespass the rachis node barrier during wheat head infection 19,20 . Thus, to analyze wheat infection at early time points and observe progression of fungal growth inside a spike, we constitutively expressed the EGFP fluorescent protein in the overexpressing mutants and wild type strain ( Supplementary Fig. S7). Analysis of infection at 3 dpi showed browning of the caryopses caused by the DHSoex, DHSoex/DOHHoex and wild type strains, but not by the DOHHoex mutant. The browning of the caryopses observed with bright field microscopy correlated with EGFP detected with the fluorescence channel, inferring that necrosis occurs at the sites of overexpressing and wild type strains infection. At 7 dpi the DHSoex/DOHHoex and wild type strains had started to infect the adjacent caryopsis. The DHSoex mutant had already colonized the adjacent caryopsis and passed the rachis node barrier. However, infection progression of the DOHHoex mutant was seen in the caryopsis, but did not trespass across the rachis node into the adjacent spikelet (Fig. 3A). By 10 dpi the wild type strain and double overexpressing mutant had fully infected the two inoculated  spikelets and had started to invade the next one. The DHSoex mutant had already reached more than three adjacent spikelets, whereas the DOHHoex mutant had not been able to cross the rachis node barrier (Fig. 3B).

DHSoex increases but DOHHoex reduces infection structures and DON mycotoxin production.
Recently, we described that F. graminearum undergoes a morphological transition from epiphytically growing hyphae to complex infection structures during early infection steps. Structures such as lobate appresoria and infection cushions form in places where stringent contact with the plant occurs and are necessary for penetration. In addition, the TRI5 gene, which encodes a trichodiene synthase, the first enzyme necessary to produce the mycotoxin deoxynivalenol (DON), is induced in such infection structures; however, lack of TRI5 does not interfere with the formation of infection structures 21 .
To investigate whether loss of virulence in DOHHoex or hypervirulence in DHSoex mutants is due to a lack or hyper-formation of infection structures as well as DON production, we screened infection during the second stage as described by Boenisch and Schäfer 21 . Detached wheat flower leaves were inoculated with the overexpressing mutants carrying a constitutive EGFP allele. Infection structure formation was determined at 8 dpi. Compared to the wild type, the DHSoex mutant developed 25% more complex infection structures. The DOHHoex mutant produced no infection structures, and only a few germinated conidia with epiphytical growth were detected. The double overexpressing mutant developed a similar number of infection structures as the wild type (Fig. 4A).
Production of DON mycotoxin in plants was measured on infected wheat spikes (cv. Nandu) at 21 dpi using a DON ELISA kit. Levels of DON production were normalized with the amount of fungal gDNA in the samples. During infection of wheat, the DHSoex mutant produced more DON (18 mg of DON/kg of Fusarium) than the wild type strain, whereas the DOHHoex mutant produced less DON under the same conditions (4 mg/kg). The double overexpressing mutant produced similar amounts of DON as the wild type strain (12 mg/kg) (Fig. 4B). In addition, levels of DON were determined in culture by inducing DON production in the wild type and overexpressing mutants grown for 8 days on minimal media containing 5 mM ammonium sulphate as the sole nitrogen source and 1% sucrose as a carbon source 19 . The DHSoex mutant produced slightly more DON in culture (6 mg/Kg) than the wild type (5 mg/kg). In contrast, the DOHHoex mutant showed no induction of DON, while the DHSoex/DOHHoex mutant produced levels of DON similar to the wild type (5.5 mg/kg). As a control, 5 mM sodium nitrate was used as a non-inductive nitrogen source (Fig. 4C).
Besides DON production, TRI4, TRI5 and TRI6 transcripts were quantified by real time quantitative polymerase chain reaction (RT-qPCR) during infection of wheat and in culture. RNA was extracted from wheat plants infected with the overexpressing mutants or the wild type strain, wheat inoculated with water was used as a negative control. β -tubulin (TUB) and ubiquitin-ligase (UBI) genes were used as normalizers in RT-qPCR. They were compared with 4 different methods to determine the best and more stable house-keeping gene for the expression assays. Both genes showed equal expression stability under the given conditions and TUB was used to normalize relative expression (Supplementary Tables S4 and S5). For in culture assays conidia were inoculated on minimal media under inductive conditions (as described above) and incubated for 3 days at 28 °C and 150 rpm. Expression levels of TRI4, TRI5 and TRI6 genes in the overexpressing mutants were compared to the transcript levels in the wild type, which were set to one. In plants, expression of the TRI genes was similar in the wild type, DHSoex and double overexpressing strains. However, with the DOHHoex mutant expression levels of the TRI genes were greatly reduced (Fig. 4D). In culture transcript levels of TRI genes were slightly higher in the DHSoex mutant and slightly lower in the DHSoex/DOHHoex mutant compared to the wild type, no TRI5 or TRI6 transcripts were detected in the DOHHoex mutant (Fig. 4E).
To determine whether the growth defect of the DOHHoex mutant on detached wheat flower leaves or during infection is due to a lack of secretion of degrading enzymes, the wild type and overexpressing mutants were grown on minimal media containing different carbon sources (e.g. wheat germ oil, Supplementary Table S2). No growth differences were observed, except when they were grown on 1% linoleic acid where interestingly, the DOHHoex mutant grew best ( Supplementary Fig. S8).

DOHHoex does not infect maize.
To establish whether the overexpressing mutants have similar effects on pathogenicity in maize, we carried out infection experiments with the inbred maize line A188 22 . Maize cobs were inoculated through the silk channel with the fungal strains, or water as a negative control. The severity of the symptoms were evaluated 35 dpi using the 7-class rating scale (disease severity rating = dsr) from Reid and Hamilton 23 , where 1 = no symptoms, 2 = 1-3%, 3 = 4-10%, 4 = 11-25%, 5 = 26-50%, 6 = 51-75% and 7 = more than 75% of kernels visibly mouldy. The water control showed no infection symptoms (dsr = 1). While the wild the wild type strain. On the contrary, DOHH overexpression suppresses production of DON mycotoxin and expression of the TRI4, TRI5 and TRI6 genes. DON mycotoxin production was measured in the overexpressing mutants grown in (B) wheat plants (cv. Nandu) for 21 days or (C) mycelia grown under inductive (5 mM (NH 4 ) 2 SO 4 as the nitrogen source) and non-inductive (5 mM NaNO 3 as the nitrogen source) conditions at 28 °C, 150 rpm and 8 dpi. (D) RNA extracted from wheat plants infected with the wild type and overexpressing mutants 12 dpi was used to quantify the relative expression of TRI4, TRI5 and TRI6 genes. (E) Relative expression of TRI genes in culture was measured from mycelia grown under inductive conditions at 28 °C, 150 rpm and 3 dpi. Error bars indicate standard deviations calculated from data per triplicate samples and are representative of three independent experiments. Significance with respect to wild type: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 (calculated with ANOVA-Bonferroni-Holm).
DOHHoex leads to complete hypusination of eIF5A. To investigate hypusine biogenesis in the wild type and overexpressing mutants, we determined spermidine incorporation by supplementing complete media with 70 μCi of [1, H]-spermidine trihydrochloride. Similar amounts of total eIF5A protein were produced in the wild type and overexpressing mutants when grown for 20 or 30 hours in supplemented complete media (Fig. 6A). Incorporation of radiolabelled spermidine after 20 h of growth was also similar (Fig. 6B). However, after 30 hours of growth, the DOHHoex mutant incorporated more spermidine than the wild type or DHSoex and DHSoex/DOHHoex mutants. Relative density quantification confirmed these results, and revealed that at 30 hours the DOHHoex mutant had incorporated twice the amount of spermidine as wild type, DHSoex or DHSoex/DOHHoex strains (Fig. 6C), posing the question of whether DOHH overexpression leads to higher deoxyhypusination and/or hypusination of eIF5A.
The eIF5A protein exists in three different forms, unmodified (Lys), deoxyhypusinated (Dhp) or hypusinated (Hyp) (Fig. 6D) 24,25 , which can be distinguished by their isoelectric points. To determine the hypusination forms of eIF5A in the wild type and overexpressing strains, we performed 2D gel electrophoresis and subsequent Western blot using antibody against human eIF5A. The wild type, DHSoex and DHSoex/DOHHoex strains produced all three isoforms of eIF5A when grown on complete media for 72 hours, while only the fully hypusinated form of eIF5A was detected in the DOHHoex mutant (Fig. 6E). Determined by 2D gel electrophoresis, the isoelectric points of the unhypusinated, deoxyhypusinated and hypusinated forms were 4.65, 5.2 and 5.45, respectively. Data from three independent assays analyzed with the ImageJ program 26 indicated that the majority of antibody attached to the more basic fully hypusinated form in all strains. Nevertheless, the other two more acidic deoxyhypusinated and unmodified forms of eIF5A were detected in the wild type, DHSoex and DHSoex/ DOHHoex strains, whereas the DOHHoex mutant showed only the fully hypusinated form of eIF5A (Fig. 6F). Complete images of Westerns blots and fluorograms are given in Supplementary Fig. S9.

Discussion
Here we demonstrate the repercussions of unbalanced expression of the eIF5A activating enzymes DHS and DOHH, and the tight regulation of eIF5A hypusine modification in the necrotrophic fungus F. graminearum during both vegetative and sexual reproduction, as well as during plant infection.
EIF5A, DHS and DOHH genes exist in all eukaryotes and their protein sequences are highly conserved among different species. Although two or more eIF5A isoforms have been identified in all eukaryotes, including some fungi 27,28 , in the plant pathogen ascomycetes we found only one protein containing the hypusine domain. The second protein with similarity to eIF5A does not contain a hypusine domain and has evolved a different function, namely the woronin body major protein Hex1, unique for filamentous fungi (Supplementary Fig. S1) 29 . This indicates that plant pathogenic ascomycetes possess a unique copy of the eIF5A activated by hypusination on a specific lysine (Lys 56 in F. graminearum), making the hypusine-eIF5A system an excellent target for proliferation control.
The DHS and DOHH genes exist as a single highly conserved copy in most eukaryotes, including ascomycetes 10,11 . In S. cerevisiae deoxyhypusine synthase Dys1 is essential for cell survival; on the contrary deoxyhypusine hydroxylase Lia1 as well as its orthologue Mmd1 from Schizosaccharomyces pombe are not essential 30,31 . Unlike the yeast, DHS and DOHH gene deletions are lethal, suggesting full hypusination of eIF5A in F. graminearum is essential for cell viability. Surprisingly, overexpressing DHS or DOHH individually resulted in opposing  phenotypes. In general, DHS overexpression led to an increase in virulence traits, whereas overexpression of DOHH reduced virulence. The simultaneous expression of both enzymes restored wild type phenotypes.
Expression of EIF5A gene was unaffected in the overexpressing mutants. Similarly the expression of the DHS gene was unchanged in the DOHHoex mutant and expression of DOHH gene was unchanged in the DHSoex mutant. These results indicated that no transcriptional co-regulation occurs among the genes necessary for hypusine biosynthesis and their substrate eIF5A. The regulatory effect of DHS and DOHH towards eIF5A is only at a translational level under the studied conditions.
Conidia are the infectious agents of fungi, essential for survival, proliferation and virulence 32 . The observed increased or impaired production of asexual conidia as a result of overexpressing DHS or DOHH would lead to a significant advantage or disadvantage in the field. Besides propagation, fungal pathogens must overcome hostile environments that include stress insults deriving from the host during invasion 33,34 , as well as toxic external compounds or by products of its own metabolism, such as ROS generated during aerobic respiration or infection 35,36 . Interestingly, the DOHHoex mutant was more sensitive to hyperosmotic and cell wall stresses, and notably produced extremely high levels of ROS.
The DOHHoex mutant exhibited both decreased production of asexual conidia and increased production of perithecia. Although the harmful effect of ROS due to uncontrolled oxidizing reactions affecting DNA, RNA, proteins and lipids have been documented, ROS has been found to be essential for sexual development 35,36 . In Botritis cinearea as well as in F. graminearum, deletion of the NADPH oxidase NoxA results in low ROS and no perithecia production, emphasizing the essential role of ROS during sexual reproduction 37,38 .
The pleiotropic stress responses could be due to direct involvement of eIF5A in diverse signalling pathways. In yeast, overexpression of protein kinase C (PKC1), PKC1 regulators WSC1, WSC2 and WSC3, and diverse regulators of actin organization restored the growth of the thermo-sensitive mutant tif51A (orthologue of F. graminearum EIF5A), suggesting a role for eIF5A in cell integrity and establishing actin polarity 4,5 . Moreover, lack of equilibrium in ROS production is one of the main consequences of deleting the gene FgOS2, which encodes a MAP kinase necessary for stress responses, or the downstream transcription factors FgATF1 and FgAP1 in F. graminearum [39][40][41] . Therefore, eIF5A hypusination is likely to directly or indirectly influence signalling pathways.
Imbalances in hypusination of eIF5A have not been investigated until now, nor the role of eIF5A, DHS and DOHH in fungal pathogenicity. Fusarium Head Blight disease is initiated by airborne spores landing on open florets during anthesis 13,14 . Open florets allow F. graminearum to make contact with young anthers, developing caryopses and adaxial surfaces of lemma, palea and glumes. Subsequent access is provided through vascular tissue crossing the rachis node to the rachis. From the rachis, F. graminearum then extends to other florets, aggravating the disease 16,19 . Point inoculation of wheat spikelets during anthesis and silk inoculation of maize with the single overexpressing mutants resulted in strikingly opposite phenotypes. DHS overexpression led to hyper-virulence in wheat and wild type infection rates in maize. On the contrary, DOHH overexpression abolished infection symptoms on wheat and maize. The DHSoex/DOHHoex mutant induced symptoms similar to wild type on both wheat and maize.
Analysis of infection progression on wheat using constitutive EGFP-tagged strains demonstrated that DHS overexpression drives faster progression than the wild type, while DOHH overexpression led to weak colonization of the inoculated spikelet, resembling typical type II resistance, with progression stopping at the base of the rachis node.
Production of the trichothecene mycotoxin DON is necessary for full F. graminearum invasion of wheat but not maize 15 . The rachis node was reported to induce expression of the TRI5 gene, which encodes a trichodiene synthase, a key enzyme in the trichothecene pathway 19 . Moreover, a strain lacking the TRI5 gene is prevented from moving into the rachis by thickening of the cell walls, demonstrating that the rachis node is a key component of wheat resistance against F. graminearum 16 . Our virulence results could at least partially be explained by the elevated production of DON associated with DHS overexpression and the low levels of DON production with DOHH overexpression during plant infection. The differences in DON regulation occur at the transcriptional level, with an up-regulation of genes in the DHSoex mutant and a dramatic down-regulation in the DOHHoex mutant.
Although DON production partly explains the hyper and hypo-virulence phenotypes of DHSoex and DOHHoex mutants in wheat, DON mycotoxin is not essential for maize infection. Therefore, the avirulence of the DOHHoex mutant in maize may be due to the increased ROS production in this mutant. As mentioned above, deletion of the stress activated protein kinase FgOS-2 in F. graminearum led to a massive release of ROS after mild osmotic stress, drastically reduced DON production, and avirulence on maize 39 , closely resembling the phenotypes of DOHH overexpression. This indicates the potential involvement of eIF5A in signalling pathways and the importance of ROS production.
F. graminearum initially colonizes the surface of wheat florets where conidia germinate epiphytically without immediate penetration. After colonization of the leaf surface, fungal hyphae undergo morphological differentiation and form infection structures, whose complexity increases with time 21 . Indeed, the hypervirulent DHSoex mutant exhibited profuse mycelia growth and infection structure formation. DOHH overexpression was associated with less dense mycelia growth and no infection structures forming on the flower leaf surface. Interestingly, the development of infection structures as well as the pathogenesis of F. graminearum is regulated by the cAMP cyclic response pathway 42 and the mitogen-activated protein kinase (MAPK) gpmk1 43 . Again, these results are in agreement with our hypothesis of eIF5A being involved in diverse signalling pathways.
In the rice blast fungus M. oryzae, a ΔcpkA mutant lacking protein kinase A and a Δgpmk1 mutant lacking the MAP kinase involved in the cascade that responds to hard, hydrophobic surfaces, or the absence of exogenous nutrients, are unable to develop appressoria, and consequently, to invade rice leaves 44 . In the southern corn blight pathogen Cochliobolus heterostrophus pmk1 is necessary for appressoria development 45 ; in F. oxysporum the pmk1 Scientific RepoRts | 6:24698 | DOI: 10.1038/srep24698 orthologue fmk1 is required for pathogenicity 46 . In F. graminearum the Δgpmk1 mutant is defective in degrading host cell wall material, and fails to induce secreted lipolytic activity 47 .
An additional reason for the differences in infection structure formation and pathogenicity could be that cell cycle progression from the G1/S phase is impaired in the DOHH overexpressing mutants 1,2 . In M. oryzae G1/S cell cycle progression has been implicated in appressoria formation. Conditional nim1 I327E mutants that arrest entry into S-phase failed to differentiate into functional infection cells, i.e. appressoria on hydrophobic surfaces. Instead, germlings continue growing with a filamentous morphology 48 . As mentioned before, eIF5A has been implicated in cell-cycle transition from G1 to S-phase by allowing selective translation of mRNAs encoding proteins specialized in initiating and enabling eukaryotic proliferation 1,2 . Results from overexpressing DHS or DOHH suggest that the imbalance of these activating enzymes causes differences in hypusine formation that impair or activate the mRNA transport and translation necessary for the morphological transition during early infection of wheat.
We tested how the expression rate of the different hypusine-forming enzymes influences final hypusine biogenesis in eIF5A by incorporating [1, H]-spermidine. The DOHHoex mutant incorporated 2-fold more spermidine after 30 hours of growth, which turned out to be due to full hypusine formation. All forms, unmodified, deoxyhypusinated, or hypusinated eIF5A proteins were found in the wild type, DHSoex and DHSoex/DOHHoex strains, with the fully hypusinated protein being the major form. However, DOHHoex mutant extracts only contained the fully hypusinated-eIF5A protein. This could be due to the non-reversible DOHH reaction with its substrate Dhp-eIF5A, exacerbated by amplification of the second enzymatic step, locking eIF5A into its hypusinated form 11 . In contrast, the DHS reaction is reversible and amplification of the first enzymatic step does not lock eIF5A in its deoxyhypusinated form (Fig. 6D) 49 .
The observed phenotypes may be due to the highly hypusinated eIF5A, and/or to the absence of the unhypusinated and deoxyhypusinated forms of eIF5A ( Table 1). The latter would point to additional biological functions of these proteins. We and others used inhibitors of DHS to abolish hypusination of eIF5A and inhibit a pathogenic phenotype. However, cells very often overcome inhibitors by overexpressing the genes encoding the inhibited enzyme 50 . Here we demonstrated that in such a case DHS overexpression will result in unwanted hypervirulence. In contrast, DOHH overexpression leads to hypovirulence of F. graminearum. Intriguingly, inhibitors directed at DOHH may offer a way to imbalance eIF5A hypusination; selection for resistance, and hence overexpression would lead to the desired phenotype even upon removing the inhibitor.  Table S3). The PCR products were cloned into the pJET1.2 vector (Fermentas-Thermo Scientific, Germany) according to the manufacturer's instructions and sequenced. For overexpression of DHS, pMW-DHS (9277 bp) was constructed by inserting the PCR product into the BamH I site of the pCWHyg-PgpdA vector, between the gpdA promoter and trpC terminator from Aspergillus nidulans, using standard protocols 52 . For overexpression of the DOHH gene, encoding a HEAT-repeat protein 53 , the plasmid pMW-DOHH (6971 bp) was constructed by inserting the ORF of DOHH into the SacI and XbaI restriction sites located between the gpd1 promoter and trpC terminator from Cochliobolus hetrostophus of the pII99-Pgpd1 vector containing the geneticin resistance cassette (Supplementary  Table S1). F. graminearum was transformed with the pMW-DHS vector linearized with ClaI, which cuts the DHS sequence once at position 328, or the pMW-DOHH plasmid linearized with Bbs I, which cuts the DOHH sequence once at position 556, to induce a single crossover event. Gel electrophoresis, restriction enzyme digestion, Southern blots and sequencing were performed using standard procedures. The double overexpressing mutant was generated by introducing the pMW-DOHH plasmid linearized with Bbs I into the DHSoex mutant (Supplementary Table S1 and Fig. S3). Wheat spike and detached wheat flower leaves infection assays. Spike infection and detached wheat flower leaves infection assays were prepared according to Boenisch and Schäfer 21 . Briefly, for intact plant infection, wheat spikes of the susceptible cultivar Nandu and partially resistant cultivar Melissos were inoculated at two of the middle florets between lemma and palea with 10 μl of a 2 × 10 4 conidia per ml suspension; infection was analyzed 7, 14 and 21 days post inoculation (dpi) or at indicated time points. The degree of infection was determined as follows: % Infection = (number of infected spikelets * 100)/ total number of spikelets. Ten spikes were used for each assay. Each assay was repeated at least three times. For the detached wheat flower leaves assay, single glumes from the susceptible cultivar Nandu were isolated and placed on 2% water-agar plates and inoculated with 5 μl of a 2 × 10 4 conidia per ml suspension and incubated in a growth chamber with a 16 h light photoperiod at 16 °C-18 °C. Infection symptoms were analyzed 8 dpi. Sterile water was used as a negative control. Four Petri dishes contained 8 biological replicates of one floret organ and represented one independent experiment.

Maize infection assay.
To ensure good pollination, silks of maize cobs from the inbred line A188 were pollinated manually once per day for 3 days. Pollinated cobs were infected through the silk channel by injecting 4 ml of a 2 × 10 4 conidia per ml suspension. 4 ml of sterile water were used as a negative control. Disease symptoms were analyzed 35 dpi 23 .
Microscopy of infected wheat spikes and detached wheat flower leaves. Wheat heads were examined using a Zeiss MZFLIII microscope (Leica Microsystems, Switzerland) 3, 7 and 10 dpi. Fungal growth and plant necrosis were detected using bright field microscopy, and GFP was detected with a GFP2 filter, which allows visualization of chlorophyll auto-fluorescence in red and mycelia growth in green using the same filter. eIF5A detection assay. For analysis of eIF5A protein levels, macroconidia were inoculated in 20 ml complete media at a final concentration of 5 × 10 5 per ml and germinated for 20 h or 30 h at 28 °C and 150 rpm. Mycelia were filtered, frozen in liquid nitrogen, ground in a mortar and resuspended in a protein extraction buffer containing 10% sucrose, 50 mM Tris-HCl, pH 7.5, and a cocktail of proteases inhibitors for fungi extracts (Sigma-Aldrich, Germany). Samples were homogenized and incubated for 10 min on ice, centrifuged for 5 min at 4 °C to pellet cell debris and the protein concentration of the supernatant was determined using Bradford reagent (Sigma-Aldrich, Germany). 30 μg of total protein were separated on a 15% SDS-polyacrylamide gel and blotted onto a PVDF-Immobilon ® membrane (Millipore, Darmstadt, Germany) using standard protocols [52][53][54][55][56][57] . Membranes were blocked with 5% non-fat skimed milk (Fluka) in TBS-T for 1 h. eIF5A protein was detected using the eIF5A1 antibody (GTX111013-rabbit anti-human eIF5A1, Acris-antibodies GmbH (Herford, Germany) and a secondary anti-rabbit IgG, HRP-linked antibody (New England Biolabs GmbH, Frankfurt am Main, Germany). Detection was performed using Lumiglo reagent (New England Biolabs GmbH, Frankfurt am Main, Germany) following the manufacturer's instructions.
Detection of eIF5A precursors and the hypusine-eIF5A protein. Cell lysate (250 μg of protein) for two-dimensional gel electrophoresis were prepared as described above. The first dimension isoelectro-focusing was performed in the PROTEAN ® IEF Cell from BioRad, using 7 cm (pH range 5.5-6.8) ReadyStrip TM IPG Strip gel (Bio-Rad Laboratories, Munich, Germany). The proteins in the first dimension Strip gel were further separated in the second dimension by SDS-PAGE on a 15% polyacrylamide gel, and transferred on to a PVDF-Immobilon ® membrane (Millipore, Darmstadt, Germany) to detect the various forms of eIF5A. Western blotting was performed as described above.
Hypusine formation assay. A conidial suspension of 5 × 10 5 conidia was incubated in 20 ml complete media containing 70 μCi of [ 3 H]-Spermidine Trihydrochloride (Perkin Elmer, Hamburg, Germany) with 150 rpm shaking at 28 °C. After 24 h (or indicated time points) of incubation the mycelia were harvested using Whatman filter paper. To produce protein crude extracts the samples were resuspended in protein extraction buffer (given above) followed by sonication on ice for 15 seconds using a 5000 micro tip (Branson Sonifier B12). 50 μg of protein extract were separated by 14% SDS-PAGE. The proteins were fixed by soaking the gel in 25% isopropanol/10% acetic acid (30 min), washed twice with distilled water and then soaked in 25 ml of amplify solution (GE Healthcare, Freiburg, Germany) for 30 min. The dried gel was exposed to an ultrasensitive X-ray film for 1 to 6 weeks. The percentage density of the radio labelled and Western blot signals were calculated with the ImageJ program 26 , and the relative density was determined by dividing the percentage density of eIF5A signals detected on the Western blot by the percentage density of incorporated spermidine detected on the autoradiogram.