Identification of O-mannosylated Virulence Factors in Ustilago maydis

The O-mannosyltransferase Pmt4 has emerged as crucial for fungal virulence in the animal pathogens Candida albicans or Cryptococcus neoformans as well as in the phytopathogenic fungus Ustilago maydis. Pmt4 O-mannosylates specific target proteins at the Endoplasmic Reticulum. Therefore a deficient O-mannosylation of these target proteins must be responsible for the loss of pathogenicity in pmt4 mutants. Taking advantage of the characteristics described for Pmt4 substrates in Saccharomyces cerevisiae, we performed a proteome-wide bioinformatic approach to identify putative Pmt4 targets in the corn smut fungus U. maydis and validated Pmt4-mediated glycosylation of candidate proteins by electrophoretic mobility shift assays. We found that the signalling mucin Msb2, which regulates appressorium differentiation upstream of the pathogenicity-related MAP kinase cascade, is O-mannosylated by Pmt4. The epistatic relationship of pmt4 and msb2 showed that both are likely to act in the same pathway. Furthermore, constitutive activation of the MAP kinase cascade restored appressorium development in pmt4 mutants, suggesting that during the initial phase of infection the failure to O-mannosylate Msb2 is responsible for the virulence defect of pmt4 mutants. On the other hand we demonstrate that during later stages of pathogenic development Pmt4 affects virulence independently of Msb2, probably by modifying secreted effector proteins. Pit1, a protein required for fungal spreading inside the infected leaf, was also identified as a Pmt4 target. Thus, O-mannosylation of different target proteins affects various stages of pathogenic development in U. maydis.


Introduction
O-mannosylation is an essential posttranslational protein modification in fungal cells [1]. This type of protein Oglycosylation, which adds mannoses to the nascent glycoproteins at the Endoplasmic Reticulum (ER) and Golgi Apparatus (AG), is required for correct protein conformation and stabilization [2]. In pathogenic fungi, such as Candida albicans or Cryptococcus neoformans, O-mannosylation is crucial for virulence [3][4][5].
The first step of the O-mannosylation pathway is catalyzed by protein O-manosyltransferases (Pmts), which add the first mannose to hydroxyl groups of serine and threonine in ER resident target proteins [6]. In fungi, the Pmt family is grouped into three subfamilies: Pmt1, Pmt2 and Pmt4, which mannosylate specific target proteins [7,8]. The number of members in these subfamilies differs in each organism; for example, in Saccharomyces cerevisiae or C. albicans the Pmt1 subfamily contains two members [9,10], while in Schizosaccharomyces pombe, C. neoformans, Aspergillus nidulans or in Ustilago maydis, the Pmt1 family contains only one member [11][12][13][14]. PMT orthologs have also been found in humans, rat and Drosophila melanogaster but not in nematodes (Caenorhabditis elegans) or plants (Arabidopsis thaliana and Oryza sativa) [6,[15][16][17][18].
Interestingly, the Pmt4 subfamily contains only one member in all organisms analyzed [6]. The protein O-mannosyltransferase Pmt4 has been characterized in several pathogenic fungi and was shown to be essential for virulence of fungal animal pathogens such as C. albicans and C. neoformans [4,5,10,19]. In addition, Pmt4 is relevant for virulence in the phytopahogenic fungus U. maydis [11]. It has been proposed that the loss of Pmt4 in fungal pathogens causes an alteration in their ability to adhere to the host, which might be caused by affecting the glycosylation of secreted proteins and cell wall proteins [5,20].
Despite their proposed relevance, Pmt4 target proteins have only been described in S. cerevisiae and C. albicans [21][22][23]. The lack of a defined consensus sequence that is targeted by Pmt4, coupled with the transient interaction of this O-mannosyltransferase with its target proteins, makes identification of Pmt4 substrates difficult [21]. The only knowledge regarding Pmt4 target proteins is that they should contain a membrane anchor to facilitate their glycosylation in the ER lumen. This characteristic feature has been used to perform a search for Pmt4 substrates in S. cerevisiae [8]. However this type of in silico screening has not been carried out in pathogenic fungi, which would allow the identification of fungal virulence factors. In this work, we have performed an in silico screening of putative Pmt4 target proteins in the plant pathogen U. maydis with a role during fungal pathogenesis.
U. maydis is a basidiomycete fungus causing smut disease in maize. For infection two yeast-like compatible haploid cells need to mate and to form a filamentous dikaryon [24][25][26]. On the leaf surface U. maydis senses a combination of physical-chemical plantderived signals and differentiates into an appressorium, a morphogenetic structure which mediates fungal penetration into the plant tissues [27]. Appressorium formation and penetration is controlled by two MAP kinases, Kpp2/Ubc3 and Kpp6 [28][29][30]. Moreover, it is known that the perception of the physical surface signal required for appressorium formation depends specifically on the Kpp2/Ubc3 MAP kinase [27], and that its activation is likely to involve the signalling mucin Msb2 [31]. Msb2 has been placed upstream of the MAP kinase pathway, consisting of the MAPKK kinase Kpp4/Ubc4, the MAPK kinase Fuz7/Ubc5 and the MAP kinase Kpp2/Ubc3 [28,30,[32][33][34]. After penetration of the plant epidermis, U. maydis proliferates inside the plant as a mycelium. During this process the fungus produces high amounts of putative secreted proteins which are only found in related smut fungi and are important for virulence [25,35]. Many of the respective effector genes are organized in gene clusters [25]. At late stages of the fungal invasion process, U. maydis induces the formation of prominent tumors in the plant which will finally contain the diploid spores [24,36,37].
The Pmt4 protein is the only O-mannosyltransferase required for pathogenesis in U. maydis. The deletion of pmt4 causes a drastic reduction in appressorium formation and, more importantly, a total loss of fungal capacity for plant penetration [11] ( Figure S1). In this work, we present novel Pmt4 target proteins, identified by an in silico screening, with an essential role during U. maydis pathogenic development. Among them, we detect the signalling mucin Msb2 as a Pmt4 substrate, suggesting that an altered glycosylation pattern of this sensor membrane protein could be the cause for the defects in appressoria production in Dpmt4 cells.

Results
Pmt4 substrate screening on the U. maydis proteome Since the role of Pmt4 in virulence must be mediated via its target proteins, we performed a search for such substrates in order to explain the role of Pmt4 in appressorium formation and penetration. Thus, we devised a bioinformatic approach to identify putative Pmt4 targets from the 6787 proteins contained in the U. maydis proteome. As Pmt4 mediates O-mannosylation of Ser/Thrrich membrane-associated proteins [8], we carried out three sequential in silico searches. First, we identified membrane proteins containing a region of at least 20 aa in which the percentage of Ser/Thr was $40%. With these parameters we found 826 proteins (12.17% of the U. maydis proteome) (Table S1). Then, we performed a more restrictive search on the U. maydis proteome for membrane attached proteins containing a wider region of 40 aa, where the percentage of Ser/Thr was $40%. With this bioinformatic approach we identified 306 proteins from the U. maydis database (4.51% of the U. maydis proteome) (Table S2). Since it has been described that the presence of a Ser/Thr rich region in the amino terminal part of single-pass transmembrane proteins triggers the recognition by Pmt4 [8], this was the search criterion for the third round. 64 out of the 306 previously identified sequences met these criteria (0.94% of the U. maydis genome) (Table S3) (see Figure 1A).
The obtained putative Pmt4 substrates were categorized based on their probable functions using FunCatDB from the MIPS U. maydis database (http://mips.helmholtz-muenchen.de/genre/ proj/ustilago/). We found that the initially identified group of 826 proteins was mainly enriched in proteins associated with transport, i.e. ion transport, detoxification by export as well as homeostasis of cations. The more restricted group of 306 proteins contained mostly unclassified proteins (52.1% of the proteins contained in the list). Other enriched functional categories were associated with inorganic chemical agent resistance and detoxification by export. Finally, the most restrictive group of 64 proteins was also mainly enriched in unclassified proteins ( Figure S2).
In the genome of U. maydis many of the genes encoding small secreted proteins are organized in gene clusters, most of them implicated in virulence [25,35]. Interestingly, we found in our screening four proteins included in these clusters: Um01235, Um03746 and Um03749 (identified in group 1, Table S1); and Um01374 (belonging to group 2, Table S2). The Um01235 protein is part of the cluster 2A. The deletion of all eight genes from cluster 2A causes a hypervirulent phenotype [25]. On the other hand, Um03746 and Um03749 reside in cluster 10A, which contains 10 genes for effector proteins in total. The deletion of cluster 10A reduces virulence [25]. Finally, Um01374 (Pit1) is one of four proteins encoded by the pit (proteins important for tumors) cluster. The deletion of the transmembrane gene pit1 or the effector gene pit2 strongly reduces pathogenic development at the level of tumor formation [38].

In silico screening validation
To confirm our bioinformatic results, we performed a biochemical approach to study the role of Pmt4 in posttranslational modification of its putative target proteins by western blot. For this purpose we isolated candidate proteins from wild-type and Dpmt4 strains and compared the mobility of these proteins during SDS polyacrylamide gel electrophoresis (SDS-PAGE). Given the critical role of cluster 10A during U. maydis virulence, we decided to analyse the role of Pmt4 in the posttranslational modification of Um03746 and Um03749. In addition we included the protein Pit1. Finally, we also randomly selected an until now unchar-

Author Summary
The O-mannosyltransferase Pmt4 is essential for virulence of animal and plant pathogenic fungi. This protein attaches one mannose at serine/threonine residues of cell wall and secreted proteins modulating their location and function. Thus, the crucial role of Pmt4 in fungal pathogenic development is probably caused by a defective glycosylation of its target proteins altering hostfungus interaction. In this paper, we performed a screen for Pmt4 target proteins employing the fungus Ustilago maydis, which causes smut disease in maize plants. This allowed identifying novel Pmt4 target proteins having a crucial role on its virulence. One of these targets is the signalling mucin Msb2, a conserved protein which acts upstream of MAP kinase cascades in various fungi and regulates early pathogenic development in U. maydis. We propose that Pmt4-dependent glycosylation of the extracellular domain of Msb2 is required for Msb2 activity and hence pathogenic development of U. maydis. This is divergent to the situation in S. cerevisiae where the mannosylated extracellular region of Msb2p possesses a negative regulatory function. In addition, we demonstrate important roles of Pmt4 during later stages of plant infection and identified Pmt4 target proteins which could be responsible for the virulence defect of pmt4 mutants during tumor formation.
acterized protein, Um04580, which contains a stretch of 40 aa where the percentage of Ser/Thr is $40%. The deletion of um04580 did not affect U. maydis virulence ( Figure S3).
Since some of the candidate genes are expressed exclusively during biotrophic development, i.e. um03746, um03749 and um01374 [25,38], we placed the open reading frames (ORFs) of the candidate genes under the control of the constitutive active otef promoter [39]. A C-terminal fusion to gfp allowed visualization of the produced proteins by western blot. The constructs were integrated into the ip locus of the solopathogenic SG200 strain and its derivate SG200Dpmt4. Cells were grown in YEPSL liquid medium, total cell extracts prepared and proteins were separated by SDS-PAGE. In western blot analysis Um03749-GFP, Pit1-GFP and Um04580-GFP, but not Um03746-GFP showed differences in mobility when isolated from the two strains ( Figure 1B). The higher mobility observed when pmt4 is deleted suggests that these proteins could be direct Pmt4 targets. Thus, in analogy with S. cerevisiae, the bioinformatic approach is a useful tool to identify new Pmt4 target proteins.
In order to substantiate the role of protein glycosylation on the O-mannosylated virulence factors identified, we investigated if the novel Pmt4 target proteins, Pit1 and Um03749, are also Nglycoproteins. N-glycosylation consists in the addition of an oligosaccharide core to nascent protein in the consensus sequence Asn-X-Ser/Thr (where X could be any acid amino except proline). Similarly to O-mannosylation, N-glycosylation is also required for full virulence of U. maydis [40]. In Um03746 and Um03749 no putative glycosylation sites could be identified. In Pit1 we identified the putative N-glycosylation site NGTF (amino acids 330-333). To know if Pit1 is also modified by Nglycosylation we analysed the Pit1-GFP protein in the SG200Dcwh41 background. Cwh41 encodes glucosidase I, an upstream component of the N-glycosylation machinery in U. maydis [41]. Pit1-GFP isolated from Dcwh41 strains showed a similar mobility shift during SDS-PAGE as Pit1-GFP isolated from Dpmt4 strains. This observation indicates that the glycosylated isoform of Pit1 is also N-glycosylated ( Figure S4).

Msb2 is a Pmt4 substrate in U. maydis
To understand the specific role of pmt4 during U. maydis appressorium formation and penetration we searched our list of putative Pmt4 target proteins for those for which a link to appressoria had previously been established. Interestingly, in group 3 we identified Um00480 (Msb2), a transmembrane mucin, which plays an important role during early pathogenic development in U. maydis [31]. Recently, homologs of Msb2 have been also linked to virulence in the plant pathogenic fungi Magnaporthe oryzae and Fusarium oxysporum [42,43]. Msb2 meets the most restrictive criteria for Pmt4 targets by displaying a window of 40 aa with a percentage of $40% Ser/Thr located in the amino terminal part of the protein (Table S3). Furthermore, S. cerevisiae Msb2p has been shown to be glycosylated by Pmt4p [44]. Although U. maydis Msb2 is unable to functionally complement the MSB2 mutant of S. cerevisiae and displays only weak similarity to S. cerevisiae Msb2p [31], we took this as a strong indication that Pmt4 glycosylates Msb2 in U. maydis.
In S. cerevisiae, Msb2p is cleaved upstream of the transmembrane domain releasing an extracellular N-terminal part and a cellular C-terminal fragment [45]. To analyse the Msb2 protein in U. maydis we used a differentially tagged protein, which was Cterminal fused to GFP and carried an internal HA-epitope between amino acids 709 and 710 in the extracellular region ( Figure 2A). The msb2-HA-GFP construct was placed under the control of the otef promoter and integrated into the ip locus of and Um04580 (,36 KDa) proteins tagged with GFP (,27 KDa) in the SG200 (WT) and SG200Dpmt4 backgrounds. Um03749, Pit1 and Um04580 suffer a differential glycosylation processing in the pmt4 mutant since we did not observe the bands that probably correspond to the glycosylated fraction of the proteins. The Um03746 processing does not seem to be affected by the loss of pmt4. doi:10.1371/journal.ppat.1002563.g001 SG200Dmsb2. Western blot analysis revealed that the Msb2 protein is processed into two distinct fragments. The extracellular domain, detected by the anti-HA-antibody, migrated at a molecular weight of .250 kDa, while the anti-GFP-antibody detected a product of approximately 65 kDa. The size of the latter fragment allowed us to predict a cleavage site, situated between the Ser/Thr rich region and the transmembrane domain ( Figure 2). In S. cerevisae, cleavage of Msb2 results in the secretion of the N-terminal extracellular domain [45]. Similar to the situation in S. cerevisiae we could detect the extracellular N-terminal domain of U. maydis Msb2 in the culture supernatant, while the C-terminal fragment was exclusively detected in the cellular fraction ( Figure S5).
To confirm that Msb2 is a Pmt4 target protein in U. maydis, we deleted pmt4 in SG200Dmsb2/msb2-HA-GFP and analysed whether pmt4 deletion influences the migration of the Msb2 protein in SDS-PAGE. The extracellular part of Msb2 isolated from Dpmt4 strains migrated faster in polyacrylamide gel electrophoresis than the extracellular part of Msb2 isolated from wild-type strains ( Figure 2B). Thus, we can conclude that Msb2 is a mannosylated substrate of Pmt4 and that mannosylation occurs in the extracellular domain of Msb2. However, the extracellular fragments from wild-type and Dpmt4 strains migrated at positions .250 kDa, which exceeded by far the predicted size of the full length Msb2 protein of ,145 kDa. This indicates that Msb2 is also postranslationally modified by Pmt4-independent mechanisms. We analysed the Msb2 protein in the N-glycosylation deficient Dcwh41 strain. We found that in SDS-PAGE the extracellular part of Msb2 migrated similarly in both, wild-type and Dcwh41 background ( Figure S4). Thus, our data suggest that, similarly to the situation described in budding yeast [44], Msb2 in U. maydis is subject to a non-characterized posttranslational modification process.

The Ser/Thr rich region of Msb2 is required for U. maydis virulence
Considering that the extracellular domain of Msb2 is mannosylated by Pmt4, we assumed that this domain might be needed for the function of Msb2. We therefore deleted the coding region of the Ser/Thr rich domain of Msb2 and integrated the construct into the ip locus of SG200Dmsb2 generating the variant SG200Dmsb2/msb2DSTR-GFP ( Figure 3A). The loss of this region did not alter the normal localization of Msb2 in the plasma membrane as revealed by treatment with the endocytosis inhibitor Latrunculin A [31] ( Figure 3B). Western blot analysis revealed that processing of the truncated Msb2 protein still occurred since the C-terminal 65 kDa fragment could be detected for both full length Msb2-HA-GFP and truncated Msb2DSTR-GFP ( Figure 3C). However, in plant infections no complementation was observed when the truncated Msb2 protein was expressed, while virulence could be completely restored when Msb2 was expressed as fulllength protein ( Figure 3D). This indicates that, contrary to the situation in S. cerevisiae, where the Ser/Thr rich region of Msb2p possesses a negative regulatory function [46], the mannosylated Ser/Thr-rich region of Msb2 in U. maydis has a positive role during virulence.
Genetic interaction analysis between pmt4 and msb2 suggests that both regulate appressorium differentiation by the same pathway In U. maydis, appressorium formation requires a set of combined physical-chemical plant-derived signals [27]. Dmsb2 strains are capable of recognizing the chemical signal (hydroxy fatty acids) but show defects in responding to the physical stimulus (hydrophobic surface). Hence, the msb2 mutant is severely affected in appressorium differentiation [31]. To investigate whether the defect in appressorium formation described for the pmt4 mutant might be due to a defective sensing activity of Msb2, we analysed the capability of Dpmt4 cells, as well as the double mutant Dpmt4Dmsb2 to respond to physical-chemical signals.
To test the chemical signal perception, we used 16-hydroxy hexadecanoic acid, a main component of the maize cuticle. Exposure of SG200 to this fatty acid induces filamentous growth [27]. Filament formation of SG200Dpmt4, SG200Dmsb2 and SG200Dmsb2Dpmt4 cells in response to 16-hydroxy hexadecanoic acid was indistinguishable from SG200 cells ( Figure 4A). Thus, the defect in appressorium formation described for Dpmt4 cells is not likely to be linked to a defective perception of the chemical signal.
To study the perception of the physical signal, we added 16hydroxy hexadecanoic acid to U. maydis cell suspensions and The HA-epitope is integrated at amino acid 709 and the C-terminus is fused to GFP. The Ser/Thr rich region as well as the predicted proteolytic cleavage site is indicated. B. Western Blot analysis of Msb2-HA-GFP isolated from SG200Dmsb2/msb2-HA-GFP (WT) and two independent clones of SG200Dmsb2Dpmt4/msb2-HA-GFP (#1 and #2, respectively). SG200 was used as a control. The first gel (above) contains 6% of polyacrylamide and a-HA antibody was used to detect the N-terminal part of Msb2. The other gel (below) contains 10% of polyacrylamide and was treated with a-GFP antibody to detect the C-terminus of Msb2. Equal amounts of proteins of total cell extracts were loaded in each lane. doi:10.1371/journal.ppat.1002563.g002 sprayed the cells onto parafilm M, which constitutes a hydrophobic and hard surface inducing differentiation of hyphae into appressoria [27]. To facilitate quantification of appressoria, we used strains harbouring a gfp reporter, AM1, which is specifically expressed in those hyphae that have formed an appressorium [27]. We observed comparable filament formation of SG200AM1 and the respective derivative mutants. However, with respect to appressorium formation, msb2, pmt4 and the double msb2 pmt4 mutants were severely reduced ( Figure 4B). Interestingly, appressorium formation in the various mutants was impaired to a similar extent ( Figure 4C) suggesting that the defect in appressorium development in the Dpmt4 strain might be a consequence of a defect in glycosylation of the Msb2 protein.
Pmt4 and Msb2 are likely upstream components of the MAP kinase pathway In U. maydis Msb2 has been placed upstream of the MAP kinases Kpp2 and Kpp6 implicated in appressorium development and penetration respectively [27,29,31,34]. In fact, expression of a hyperactive allele of the MAP kinase kinase Fuz7, Fuz7DD [34], which activates Kpp2 and Kpp6, restores plant cuticle penetration in sho1 msb2 double deletion strains [31], while the sho1 msb2 double deletion strain is non-pathogenic and unable to form appressoria on the plant surface [31]. The genetic interaction analysis between msb2 and pmt4 may suggest a role of pmt4 upstream of the MAP kinase cascade. However, we cannot exclude the possibility that Pmt4 is acting downstream of the MAP kinase pathway controlling glycosylation of proteins implicated in the appressorium morphogenesis. To test our hypothesis that Pmt4 acts above the MAP kinase cascade we used the hyperactive allele fuz7DD cloned under the control of the crg promoter, which is repressed by glucose and induced by arabinose [31,47]. This construct was integrated in the ip locus of SG200, SG200Dpmt4 and SG200Dmsb2. To analyse appressorium formation, we infected leaves from seven days old maize seedlings with SG200, SG200fuz7DD, SG200Dpmt4, SG200fuz7DDDpmt4, SG200Dmsb2 and SG200fuz7DDDmsb2 in the presence of arabinose. To quantify appressorium formation, cells were stained with calcofluor white 15 hours after plant inoculation. Induction of the fuz7DD allele restored appressoria production in the pmt4 and msb2 mutants ( Figure 5A and Figure  S6A). We also analysed the appressorium-mediated penetration under these conditions. To this end, we used the chlorazol black E staining procedure to visualize invading hyphae [29]. Remarkably, the appressorium-mediated penetration capability of Dpmt4 cells was also restored upon expression of the constitutively active fuz7DD allele, including the formation of clamp-like cells, the structures associated with fungal progression inside the plant [48] (Figure 5B, Cells of SG200Dmsb2/msb2-HA-GFP and SG200Dmsb2/msb2DSTR-GFP were grown to mid log phase in YEPSL and analyzed microscopically without addition (top row) or after addition of latrunculin A (bottom row). The fluorescent signal corresponding to GFP is shown. C. Total protein extracts of the indicated strains were subjected to western analysis with a-HA antibody (upper panel) and a-GFP antibody (middle panel). Tubulin served as loading control and was detected with a-tubulin antibody (bottom panel). The fusion proteins and processed fragments are indicated by arrowheads on the right. The molecular weight ruler is depicted on the left. D. Disease symptoms caused by SG200 (WT), SG200Dmsb2, and SG200Dmsb2 complemented with either Potef:msb2 or Potef:msb2DSTR were scored 12 dpi. N indicates the total number of plants evaluated in each case. doi:10.1371/journal.ppat.1002563.g003 5C and 5D). Analysis of disease progression 12 days after infection revealed an increased virulence capacity of SG200fuz7DDDpmt4 and SG200fuz7DDDmsb2 when compared to SG200Dpmt4 and SG200Dmsb2, respectively ( Figure 6 and Figure S6B). Thus, our data suggest that the role of Pmt4 on appressorium biology might take place upstream of MAP kinase cascade.
If Pmt4 and Msb2 activate the MAP kinase cascade during appressorium formation, Kpp2 phosphorylation in the respective deletion mutants is expected to be reduced during this developmental stage. To examine the phosphorylation of Kpp2 we incubated SG200, SG200Dpmt4 and SG200Dmsb2 for 10 h on parafilm M. At this time point about 15% of the SG200 filaments begin to form appressoria (data not shown). As shown in Figure  S7, the hydrophobic surface induced Kpp2 phosphorylation in SG200 as well as in the mutants. We observed a slight reduction in Kpp2 phosphorylation in the Dpmt4 (84611%) and Dmsb2 (7667%) strains when compared to SG200 (100%) ( Figure S7). However, compared to SG200 only the Dmsb2 strain showed a statistically significant reduction in Kpp2 phosphorylation (P,0.05), while differences in the Dpmt4 strain were not statistically significant (P.0.05; Figure S7). This experiment shows that both, Pmt4 and Msb2, are marginally involved in the activation of the MAP kinase Kpp2.
Scanning Electron Microscopy suggests that the deletion of pmt4 causes post-appressorial outgrowth of hyphae due to a penetration defect We could previously observe that the penetration of Dpmt4 appressoria was arrested [11]. To find out more about this phenotype, we carried out an analysis of appressorium penetration of wild-type strains and pmt4 mutants using scanning electron microscopy (SEM). With this aim, we infected maize plants with a mixture of the compatible FB1 and FB2 as well as compatible FB1Dpmt4 and FB2Dpmt4 strains. 15 hours after infection maize leaves were fixed and visualized by SEM. In infections with compatible wild-type strains filaments differentiated into appressoria and penetrated inside the plant tissue. By contrast, the few appressoria formed by the Dpmt4 strain were unable to penetrate the cuticle (Figure 7), and instead the hyphae continued to grow on the plant surface. In order to verify that this aberrant morphology is a consequence but not the cause of the deficiency to penetrate the plant cuticle, we studied appressorium formation of FB1 and FB2, as well as FB1Dpmt4 and FB2Dpmt4 strains on a nonpenetrable surface such as parafilm M. In the in vitro conditions post-appressorial outgrowth of hyphae was observed in both, pmt4 mutants and wild-type strains, resembling the phenotype of Dpmt4 strains on the plant surface (Figure 7). This suggests that the Dpmt4 appressoria are unable to penetrate the plant cuticle and as consequence continue to grow on the plant surface.

Pmt4 is required for cell adhesion in U. maydis
We have shown that the deletion of pmt4 affects the perception of physical signals required for appressorium development as well as the appressorium-mediated penetration on plant surface. In the rice blast fungus M. oryzae, cell adhesion to surfaces has been shown to be needed for the detection of the physical plant-derived signals and for the penetration of the plant cuticle [49,50]. Thus we wondered whether pmt4 mutants were able to properly adhere to solid surfaces. To analyse adhesion capacity of Dpmt4 cells, we spotted exponentially growing cells of FB1 wild-type and FB1Dpmt4 strains on starch agar plates and incubated them for three days at 28uC. After this time period plates were washed with water. As shown in Figure 8, deletion of pmt4 significantly reduces . The pmt4 and the double msb2 pmt4 mutants respond to hydroxy fatty acid but not to hydrophobic surfaces. A. Chemical plant-derived signals reception does not depend on Pmt4. SG200 (WT) and its derivative mutants were grown to exponential phase and incubated in liquid medium (2% YEPSL in water) supplemented with 100 mM of 16-hydroxyhexadecanoic acid dissolved in ethanol or ethanol (control) for 18 hours at 28uC. Filament formation was analyzed. All data are shown as mean values 6SEM of three independent repetitions quantifying 200 cells in each case. B. Physical signals reception depends on Pmt4. SG200AM1 (WT) and its derivative mutants were sprayed as cell suspension (2% YEPSL) on parafilm M with or without 100 mM 16-hydroxyhexadecanoic acid and incubated for 18 hours at 28uC. Cells were stained with calcofluor white (CFW) and appressorium formation was scored visualizing the expression of the GFP reporter gene. Appressorium formation was reduced in all analysed mutants. Scale bar represents 50 mm. C. Quantification of appressoria production. The assay was performed as described in B. In three independent experiments, approximately 100 filaments per strain and experiment were analyzed with respect to appressorium formation. All data are shown as mean values 6SEM. Asterisk indicates statistically significant differences between wild-type (control) and each mutant. Single asterisk means P value#0.001. doi:10.1371/journal.ppat.1002563.g004 U. maydis cellular adhesion capacity since the colonies of FB1Dpmt4 were easily removed from the plates after a gentle washing. By contrast, wild-type colonies as well as Dpmt1 (control) colonies remained attached to the surface. This result shows that Pmt4 plays a role in cellular adhesion in U. maydis which may be connected to the appressorium formation and/or penetration defects. To distinguish between both possibilities, we studied the cellular adhesion capacity in FB1Dmsb2 strain. We observed a similar behaviour to wild-type cells ( Figure S8). Because the loss of Msb2 affects to appressorium formation but not penetration, our data suggest a link between cellular adhesion capacity and appressorium penetration in pmt4 mutant cells.

Pmt4 plays a role after appressorium penetration
We have observed that although the plant infections with the SG200fuz7DDDpmt4 strain were able to induce small tumors, the virulence of this strain was severely reduced when compared to SG200fuz7DD suggesting a role for Pmt4 during fungal progression inside the plant tissues. To confirm this result, we cloned the pmt4 ORF under the control of the otef promoter which is active during early pathogenic development, but turned off during the late biotrophic stage [51]. This construct was integrated in the ip locus of the SG200Dpmt4 strain. To address if Pmt4 plays a role after plant penetration, we inoculated maize plants with SG200, SG200Dpmt4 and SG200Dpmt4P otef :pmt4. Disease progression was scored 12 days after infection. Although appressorium penetration was completely restored in plants infected with SG200Dpmt4P otef :pmt4, tumor formation was severely reduced compared to infections with SG200 ( Figure 9). This demonstrates, that Pmt4 is not only needed for early pathogenic development, but has additional roles during the in planta development, most likely by posttranslational modification of fungal effectors required for tumor formation.

Discussion
In this paper we have performed in silico screening for putative Pmt4 target proteins in the phytopathogen U. maydis. We have found a high percentage of proteins containing Ser/Thr rich regions with most of them having unknown functions. A screening approach searching for transmembrane anchored proteins, containing a stretch of 20 aa in which the percentage of Ser/ Thr was $40%, identified 826 proteins (12.17% of the U. maydis proteome). By contrast, using similar parameters in S. cerevisiae 51 candidates (0.87% of the proteome) could be identified [8]. This difference may be due to the high number of putative secreted proteins present in the U. maydis proteome, which may be linked to the biotrophic relationship with the host plant [25]. In order to obtain a more restrictive group of Pmt4 target proteins we screened the proteome of U. maydis for single-pass transmembrane proteins containing a Ser/Thr-rich region facing the ER lumen. Under these conditions, we found 64 proteins. The validation of our screening retrieved novel Pmt4 substrates such as Um04580,  . The pmt4 mutant appressorium shows outgrowing as a consequence of its incapability to penetrate the plant cuticle. Crosses of the sexually compatible strains FB1 and FB2, and FB1Dpmt4 and FB2Dpmt4 were inoculated into maize seedlings. Leaves were fixed and analysed by Scanning Electron Microscopy (SEM) one day post infection. In these images (left), we can observe appressorium formation over a plant stomata (pointed with a red arrow) penetrating the plant surface in the wild-type. However, in a similar scenario, with the appressorium developed over plant stomata, the pmt4 mutant is unable to penetrate the plant cuticle, producing hyphal outgrowing. This morphologic structure seems to be a consequence of its incapability to penetrate this surface as we can deduce from wildtype phenotype of the in vitro appressorium formation on the nonpenetrable surface of parafilm M (right). The in vitro analysis was performed by fluorescence microscopy staining the cells with calcofluor white 15 hours after being sprayed on parafilm M (see Methods for additional details). Scale bar on the in vitro conditions represents 20 mm. doi:10.1371/journal.ppat.1002563.g007 or Pit1, which is essential for pathogenic development of U. maydis [38]. Moreover, we have found that Um03749, which belongs to the cluster 10A, required for tumor formation [25], is also a novel Pmt4 target protein.
One of the most intriguing phenotypes associated with the pmt4 deletion in U. maydis is the strong reduction in appressorium formation and appressorium-mediated penetration of the plant cuticle [11]. Both processes are controlled by a MAP kinase pathway, consisting of the MAPKK kinase Kpp4/Ubc4, the MAPK kinase Fuz7/Ubc5 and the MAP kinases Kpp2/Ubc3 and Kpp6 [28][29][30][32][33][34]. While Kpp2 is needed for appressorium formation [30], Kpp6 has a specific role during the penetration process [29]. Our results suggest that the role of Pmt4 during appressorium formation consists probably not in the glycosylation of downstream components of this pathway rather Pmt4 modifies proteins placed upstream of the MAP kinase pathway. In U. maydis the signalling mucin Msb2 is likely an activator of the MAP kinase cascade and the deletion of msb2 leads to a strong reduction in appressorium formation due to defects in perception of plant surface signals [31]. We have identified Msb2 in our in silico screening and confirmed that it is substrate of Pmt4 in U. maydis, as it has been described for S. cerevisiae [44]. Our epistatic studies comprising msb2 and pmt4 suggest that during appressorium development both proteins act in the same pathway. A critical question was if Pmt4 and Msb2 are needed for the activation of the MAP kinase Kpp2, which is essential for filamentation and appressorium formation in U. maydis [30]. Here we could show for the first time that Kpp2 is phosphorylated during appressorium formation in U. maydis. However, in pmt4 and msb2 mutants phosphorylation of Kpp2 was only slightly reduced. In view of the fact that under our experimental conditions only a minor fraction of U. maydis filaments differentiate appressoria and that pmt4 and msb2 mutants exhibit normal filament formation, a process that depends also on Kpp2, we did not expect major differences in Kpp2 phosphorylation. We do currently not know in which spatial and temporal context Kpp2 phosphorylation induces either filament formation or appressorium formation. It is conceivable that appressorium formation requires a temporal coordinated threshold of MAP kinase activity and that this process involves Msb2 and its glycosylation status. A strong indication that Msb2 and Pmt4 act above the MAP kinase cascade is inferred from our observation that expression of the constitutive active fuz7DD allele restored appressorium formation in the Dmsb2 as well as Dpmt4 strains. Overall our data suggest that Msb2 and Pmt4 are needed for the activation of the MAP kinase pathway, although formal proof for this needs further experimentation.
In gel mobility shift analysis we found that Pmt4 specifically glycosylates the extracellular domain of Msb2. In S. cerevisae the extracellular domain of Msb2p has a negative regulatory function, since deletion of this domain leads to the activation of the FG (filamentous growth)-pathway specific MAP kinase Kss1p [46]. It is therefore assumed, that in yeast the release of the extracellular glycosylated domain of Msb2p activates the FG-pathway [45,46]. This release can be mimicked by the deletion of pmt4, which leads to underglycosylated Msb2p and to the activation of the FGpathway in an Msb2p-dependent manner [44]. Thus, in S. cerevisiae underglycosylated Msb2p activates the MAP kinase pathway. Our data from U. maydis demonstrates that the extracellular domain of Msb2 has a positive function, since deletion of the Ser/Thr-rich region causes a decrease in pathogenicity similar to the deletion of the entire msb2 gene. This observation is consistent with the reduced appressorium formation in pmt4 deletion mutants. Accordingly, Pmt4-mediated mannosylation of Msb2 seems to be necessary for activation of the MAP kinase cascade and the reduced virulence of Dpmt4 strains can be partially explained by defective glycosylation of the Msb2 sensor. Recent studies showed that the Ser/Thr-rich extracellular domain of M. oryzae Msb2 is like in U. maydis Msb2 and in contrast to Msb2p from S. cerevisae also essential for function [42]. This suggests that the model proposed for the Pmt4-Msb2 relationship, regulating appressorium development is conserved among phytopathogenic fungi (see Figure S9).
The defect in appressorium formation can be explained by a defective Msb2 glycosylation. However, since the few appressoria produced in Dmsb2 strains are not affected in penetration [31], the penetration defect of Dpmt4 appressoria is likely caused by insufficient glycosylation of yet unidentified Pmt4 target proteins. Our observations using scanning electron microscopy showed that the few appressoria produced by the Dpmt4 strain were unable to penetrate the plant cuticle and instead continued growth of the hyphae on the plant surface. This phenotype has also been observed in strains carrying a non-phosphorylatable allele of the MAP kinase Kpp6 [29]. These findings, together with the observation that expression of the fuz7DD allele restores the penetration defect of Dpmt4 appressoria suggest that Pmt4 target proteins implicated in appressorium-mediated penetration are also likely to be placed upstream of the MAP kinase pathway ( Figure 10).
To better analyse the outgrowing of the hypha observed in the pmt4 mutant, we studied appressoria developed by wild-type strains on non-penetrable surfaces such as parafilm M, finding a similar behaviour. This data strongly suggest that this hyphal  Figure 10. Model of the Pmt4 roles developed during U. maydis pathogenic development. Pmt4 has a major function in appressorium biology: Pmt4 O-mannosylates Msb2, a membrane-sensor that regulates appressorium development upstream of the MAP kinase cascade. The deletion of pmt4 severely affects appressorium formation, probably due to a deficient glycosylation of Msb2. During appressorium-mediated penetration Pmt4 also probably acts above the MAP kinase cascade. Furthermore, Pmt4 plays a role in planta by Omannosylation of proteins required for tumor formation. doi:10.1371/journal.ppat.1002563.g010 outgrowth is the consequence but not the cause of its incapability to penetrate the plant cuticle. The penetration failure of Dpmt4 appressoria could be caused by an insufficient adhesion of the fungus to the plant surface. For the rice blast fungus M. oryzae it has been demonstrated that proper adhesion is critical for appressoria to penetrate [50]. Interestingly, we have shown that Pmt4 plays an important role in cellular adhesion to surfaces which might be related to the appressorium penetration defect.
Our previous studies as well as this work demonstrated a prominent role of Pmt4 in the appressorium biology. In addition, we have now shown that Pmt4 is also necessary for tumor induction in planta. We observed that pmt4 expression under the control of the otef promoter, which is turned off during the infection process [51], is not sufficient to restore normal tumor formation in the SG200Dpmt4 strain. Likewise, expression of fuz7DD in Dpmt4 strains did not rescue the virulence defect, although initial appressorium formation and penetration were restored. This suggests that the Pmt4 protein is also required during fungal proliferation inside the plant tissue. For C. albicans it has been demonstrated that the deletion of pmt genes is associated with defects in cell wall integrity and secretion of fungal effectors, probably disturbing the interaction with the host [5,10,52,53]. In U. maydis, the biotrophic interaction between the plant and the fungus is also mediated by fungal effectors [24]. The role of Pmt4 during this developmental stage could be related to the Omannosylation of Pit1 and Um03749 (see Figure 10), since both are important for the biotrophic development [25,38]. Um03749 is a putative secreted protein which is directly modified by Pmt4. The underglycosylation of Um03749 in Dpmt4 strains might affect the function of this effector. Pit1 is a plasma membrane protein that is genetically linked to the secreted effector Pit2. Both proteins are specifically needed for tumor formation [38]. Although the connection between these two proteins is currently unclear, it is assumed, that they act in the same pathway [38]. Therefore, mannosylation of Pit1 by Pmt4 could indirectly affect the function of the secreted effector Pit2. This points to a role of Pmt4 in the regulation of secreted effector proteins, needed for the biotrophic development. In this context, considering the large number of putative Pmt4 targets obtained in this study, it is tempting to speculate that more Pmt4 targets exist, which contribute to the establishment of biotrophic interaction between plant and fungus.
The bioinformatic search for Pmt4 target proteins in other pathogenic fungi might highlight the role of Pmt4 during fungal pathogenic development and allow the identification of novel Pmt4 substrates required for fungus-host interactions.

Strains, growth conditions and plasmids
Escherichia coli DH5a and pGEM-T easy (Promega) were used for cloning purposes. pING, a derivative of pONG [31] where mcherry is exchanged by gfp, was used for cloning of ORFs from pmt4, um03746, um03749 and um04580. To clone these ORFs, U. maydis genomic DNA was used as template amplifying with the primers Pmt4ORF-5 and Pmt4ORF-3, 03746ORF-5 and 03746ORF-3, 03749ORF-5 and 03749ORF-3, 04580ORF-5 and 04580ORF-3, respectively (sequences in Table S4). The fragments were cut with SfiI and ligated with the 6.3 kb SfiI fragment of pING. Expression of ORFs from pING derivatives is controlled by the otef promoter.
To generate the msb2-HA-GFP construct, the primer combinations oDL81/oDL125 and oDL82/oDL124 (Table S4) using U. maydis genomic DNA as template generated two PCR products, 1.4 kb and 2.1 kb in length, respectively. Both fragments were cut with SfiI and ligated with the 6.3 kb SfiI fragment of pING, resulting in pPotef-msb2-HA-GFP. In this plasmid the msb2 gene with an internal HA tag (corresponding to amino acid 709) is Cterminally fused to GFP. Expression of the fusion gene is driven by the otef promoter.
To generate the truncated msb2DSTR-GFP allele, the primer combinations oDL124/oDL171 and oDL204/oDL125 using pPotef-msb2-HA-GFP as template generated two PCR products, 0.1 kb and 1.3 kb in length, respectively. Both fragments were cut with SfiI and XmaI and ligated with the 6.3 kb SfiI fragment of pING resulting in pPotef-msb2DSTR-GFP. In this plasmid the entire Ser/Thr-rich region of msb2, corresponding to amino acids 33 to 713 is deleted. All plasmids were linearized with SspI and integrated into the ip locus.
Pathogenicity assays were performed as described [25]. U. maydis cultures were grown to exponential phase and concentrated to an OD 600 of 3, washed two times in water and injected into one week old maize (Zea mays) seedlings (Golden Bantam). Disease symptoms were quantified 7 to 25 days post infection. Virulence tests were repeated at least three times using the number of plants indicated in each figure.

DNA procedures
Molecular biology techniques were used as described by [57]. U. maydis DNA was isolated following the protocol of [58]. Standard U. maydis transformation procedure was used [59].

In silico screening analysis
The U. maydis proteome was downloaded from the MIPS FTP server at ftp://ftpmips.gsf.de/ustilago/Umaydis_valid/Umaydis_ valid_orf_prot_121009, and the identifiers of proteins annotated as transmembrane were downloaded from the MIPS web server. A program to filter the protein sequences was written using the Perl programming language. The program searches for proteins with a fixed frequency of Ser/Thr within a window of selected length. The most restrictive screening was carried out selecting only transmembrane proteins, avoiding GPI-anchored and multipass membrane proteins.

Western blot analysis
Protein extracts were prepared from exponentially growing cells, collected by centrifugation, washed with stop buffer (0.9% NaCl, 1 mM NaN3, 10 mM EDTA, 50 mM NaF), and frozen on dry ice. All manipulations were done on ice or in the cold room (4u). For isolation of total protein extracts from surface-attached hyphae, cell suspensions were incubated on parafilm M as described below. Cells not attached to the surface were removed in a water bath. Attached hyphae were harvested in Thorner buffer (8 M urea, 5% SDS, 0.1 mM EDTA, 0.01% bromophenol blue, 100 mM DTT, and 100 mM Tris-HCl, pH 6.8) using a cell scraper (Greiner, Frickenhausen/Germany). Total protein extracts were prepared by Fast-prep vortexing with glass beads (Sigma) [60]. For isolation of secreted proteins, U. maydis cells were grown in 50 mL of CM medium supplemented with 1% glucose to an OD 600 of 0.4. Free-cell culture supernatant was obtained by centrifugation at 3.000 g for ten minutes at 4uC. Secreted proteins were isolated by incubation for 30 minutes on ice with 0.02% sodium deoxycholate and further addition of 10% Trichloroacetic acid. Samples were incubated overnight at 4uC and precipitated proteins were collected by centrifugation at 12.000 g for 15 minutes at 4uC. The pellet was air dried and resuspended in 200 ml of loading buffer 2X (0.1 mM Tris-HCl pH 6.8, 4% SDS, 20% glycerol, 2.98 mM bromophenol blue, 0.2 M DTT). Purified proteins were separated on SDS-PAGE (6-12% of polyacrylamide). Blots were probed with anti-HA antibodies (Roche) or anti-GFP mouse IgG1 k antibodies (Roche). To detect phosphorylated Kpp2 anti-phospho-p42/44 antibody (cell signaling, Danvers/USA) was used. To detect total Kpp2 a polyclonal anti-Kpp2 antibody was generated in rabbits using the Kpp2 antigen sequence N-CLTFSPRKRITVEEAL-C (Eurogentec, Cologne/Germany). Anti-alpha tubulin was routinely used as loading control. As secondary antibodies horseradish peroxidaseconjugated anti-mouse IgG (Sigma) or anti-rabbit IgG (Cell Signaling) were used. Supersignal (Pierce) was used to detect the proteins analysed. Quantification of western blots was performed using a chemocam imager (INTAS, Göttingen/Germany) and ImageJ software.

Filament and appressorium induction on artificial surfaces
The in vitro system for inducing filaments and appressoria in U. maydis was applied as describes previously [27] with minor modifications [31]. Cell cultures (2% YEPSL) were sprayed (EcoSpray Labo Chimie, France) on Parafilm M (Pechiney Plastic Packaging, Chicago, USA) and, if applicable, treated with 100 mM (f.c.) 16-hydroxyhexadecanoic acid (Sigma). To quantify filaments relative to yeast-like cells samples were directly analysed by light microscopy. To quantify appressoria, surfaces were rinsed with

Scanning Electron Microscopy
The SEM analyses were performed on PHILIPS XL30 microscope. Infected leaves from maize two days post-infection were fixed with 4% glutaraldehyde in cacodylate (0.1 M pH 7.4) at 4uC. The samples were washed several times in cacodylate. Afterwards, a second fixation in 1% tetraoxide osmium was carried out for two hours at 4uC. This was washed in cacodylate containing 7.5% of sucrose. Dehydration with acetone at 4uC: 30 minutes with 70% acetone, 30 minutes with 80% acetone, 30 minutes 90% acetone and finally, one hour with 100% acetone. The samples were then introduced into an SPI-Dry Critical Point Drying apparatus and coated in gold.

Cellular adhesion assays
To ascertain the effects of the deletion of pmt4 and msb2 on cellular adhesion, we grew the strains to exponential phase in YEPSL at 28uC and the cells were spotted on starch agar plates. Starch medium contains, 0.25% starch from potato, 0.1% ammonium sulphate, 0.1% sucrose and phosphate buffer 25 mM pH 7. Plates were incubated for three days at 28uC and cellular colonies were rinsed with water to analyse their adhesion properties.

Calcofluor white and chlorazol black E staining
To analyze the pre-penetration stages of U. maydis using fluorescence microscopy, cells were stained with calcofluor white. Post-penetration stages were studied by optical microscopy of chlorazol black E stained leaf samples as previously described [29].

Microscopy
Cells were examined using a Leica fluorescence microscope, equipped with a PlanApo 6100 lens. Analysis of the prepenetration stages was done using a Deltavision widefield microscope (Applied Precision, Issaquah, WA). Image deconvolution was performed using z-series of between 7 and 23 focal planes, acquired at 0.5 mm intervals. Image processing was carried out using Adobe Photoshop CS5 and ImageJ.  Figure S1 The U. maydis life's pathogenic cycle. The sexual pathogenic cycle of U. maydis starts with the mating between two sexually compatible strains on the plant surface to form a dikaryon filament. A set of combined physical-chemical plantderived signals leads to hyphae differentiation into appressoria which mediate plant penetration (1). Once inside the plant, the fungus proliferates as mycelium developing the clamp-like cells which ensure the maintenance of the dikaryotic state (2). During this infection process, U. maydis induces the tumor formation in the plant maize (3). Pmt4 is required for appressorium formation and penetration, and thus the Dpmt4 strain is unable to proliferate inside the plant tissues neither induce tumors. (TIFF) Western Blot analysis of Msb2-HA-GFP isolated from SG200Dmsb2/msb2-HA-GFP culture supernatant (S) and total extract (TE). SG200 was used as a negative control. The first gel (above) was used to detect the N-terminal part of Msb2 with a-HA antibody. The other gel (below) was treated with a-GFP antibody to detect the C-terminus of Msb2. The extracellular N-terminal domain of U. maydis Msb2 was identified in the culture supernatant, while the C-terminal fragment was exclusively detected in the cellular fraction. (TIFF) Figure S6 Expression of fuz7DD partially restores the msb2 mutant phenotypes. A. Seven days old maize seedlings were infected with the strains SG200 (WT), SG200fuz7DD, SG200Dmsb2 and SG200fuz7DDDmsb2 scoring appressoria production 15 hours later (.100 filaments in each case). Data are shown as mean values 6SEM. Asterisk indicates statistically significant differences between wild-type (control) and Dmsb2 strains, P value#0.001. B. Symptoms in infected plants with the strains indicated were scored 12 days post-infection. N indicates the total number of plants evaluated in each case. (TIFF) Figure S7 Kpp2 phosphorylation in Dpmt4 and Dmsb2 strains. A. The WT (SG200), Dpmt4 and Dmsb2 strains were incubated on parafilm M with 100 mM 16-hydroxyhexadecanoic acid for 10 h. Total proteins isolated before (left) and after incubation on parafilm M (right) were subjected to western blot analysis. The phosphorylated form of Kpp2 (P-Kpp2) and total Kpp2 were detected using a-phospho-p44/42 antibody and a-Kpp2 antibody, respectively. Asterisk denotes an unspecific background signal. B. The relative Kpp2 phosphorylation from three independent experiments. Kpp2 phosphorylation in WT (SG200) was set to 1. The datasets from Dpmt4 and Dmsb2 strains were compared with the wild-type dataset to calculate p-values (ttest) given above column. Error bars indicate standard deviation. (TIFF) Figure S8 Msb2 is not required for cellular adhesion to solid surfaces. The strains indicated were grown to A600 = 0.5 in YEPSL liquid medium and then were spotted on starch medium plates and incubated for three days at 28uC. The deletion of msb2 does not affect the fungal cell adhesion to solid surfaces in U.