Control of Glucosylceramide Production and Morphogenesis by the Bar1 Ceramide Synthase in Fusarium graminearum

The contribution of plasma membrane proteins to the virulence of plant pathogenic fungi is poorly understood. Accordingly, the objective of this study was to characterize the acyl-CoA dependent ceramide synthase Bar1 (previously implicated in plasma membrane organization) in the wheat pathogen Fusarium graminearum. The role of Bar1 in mediating cell membrane organization was confirmed as ΔBAR1 mutants failed to display a distinct sterol-rich domain at the hyphal tip. The ΔBAR1 mutants were non-pathogenic when inoculated onto wheat heads, and their in vitro growth also was severely perturbed. ΔBAR1 mutants were incapable of producing perithecia (sexual fruiting structures) and only produced macroconidia (asexual spores) in the presence of NaCl. Sphingolipid analyses indicated that Bar1 is specifically necessary for the production of glucosylceramides in both F. graminearum and Aspergillus nidulans. Interestingly, glucosylceramides appear to mediate sensitivity to heat stable antifungal factor (HSAF), as, in addition to ΔBAR1 mutants, a glucosylceramide synthase deficient mutant of Yarrowia lipolytica is also resistant to HSAF.


Introduction
Fusarium graminearum (teleomorph Gibberella zeae) is the causative agent of head blight (scab) on small grains such as wheat. The effects of the disease are two-fold; infected plants produce less grain, and the grain that is produced can contain various mycotoxins, particularly deoxynivalenol [1,2]. The significant economic impact of head blight in wheat-growing areas of the United States and Europe has triggered interest in understanding the mechanisms that underlie interactions between F. graminearum and prospective host plants [3,4]. Of the several features distinguish F. graminearum from the well-characterized Magnaporthe oryzae-rice pathosystem, the most notable is the absence of appressoria production by F. graminearum during host invasion [1]. Accordingly, the identification and characterization of functions required for the pathogenicity of F. graminearum should provide a broader perspective on virulence mechanisms deployed by fungi.
We are interested in exploring the idea that lipid microdomains located on the surface of fungal plant pathogens play an important role in host interactions. In animals and yeast, signaling complexes have been shown to aggregate into lipid microdomains termed ''lipid rafts'' [5,6]. These rafts are regions of the plasma membrane rich in sterols and sphingolipids, which alter the biochemical properties of the domains and confer resistance to mild detergents [7,8]. In fungi, several proteins have been isolated from detergent-resistant membrane (DRM) domains in Saccharomyces cerevisiae, Candida albicans, and Cryptococcus neoformans. A general trend for DRM proteins appears to be the presence of a glycosylphosphatidylinositol (GPI) anchor, the lipid tail of which is presumed to interact favorably with saturated sphingolipids [9,10]. However, many transmembrane proteins are also found to be enriched in DRM fractions [11]. For example, the ATPase Pma1 has become a marker for DRM fractions in both S. cerevisiae and C. albicans [8,9]. Given their involvement in clustering membrane proteins and signaling complexes on the cell surface, lipid rafts represent attractive targets for the discovery of novel virulence factors in plant pathogenic fungi (such as those that might regulate MAP kinase and other signaling pathways; [12]). Consistent with this notion, GPI-anchored virulence determinants Sod1 (Cu/Zn superoxide dismutase) and Plb1 (lysophospholipase) localize to DRM fractions of the human pathogen C. neoformans [13].
Sterol-rich domains are crucial for polarized hyphal growth in a number of fungi, as disruption of their organization typically results in failure to maintain a discrete polarity axis. The importance of these domains has been demonstrated by the genetic characterization of several A. nidulans mutants affecting the sphingolipid biosynthesis pathway. Genetic and pharmacological depletion of serine palmitoltransferase activity, responsible for the first step in sphingolipid biosynthesis, causes a severe polarity defect, thereby suggesting that sphingolipids contribute significantly to hyphal extension [14]. Further evidence that links sphingolipids to sterol-rich domains comes from the functional characterization of the acyl-CoA dependent ceramide synthase BarA in A. nidulans. Ceramides are the ''simplest'' of the sphingolipids, in that they only contain hydrogen as the head group, and they serve as a template for the synthesis of more complex sphingolipids. BarA was originally identified as a gene product that was necessary for sensitivity to the heat-stable antifungal factor (HSAF) from the bacterium Lysobacter enzymogenes [15]. HSAF is a mixture of three structurally-related compounds, with dihydromaltophilin as the primary one [16]. The genome of A. nidulans encodes at least two ceramide synthases, barA and lagA. Deletion of BarA function causes a hyphal polarity defect and disrupts the organization of sterol-rich domains at hyphal tips, while deletion of the lagA gene did not yield a viable mutant. However, when repressed with an inducible promoter, lagAdepleted mutants displayed a severe reduction in growth compared to the barA mutant, suggesting that two pools of ceramide are produced in A. nidulans and each contributes differentially to fungal growth, with the LagA pool being essential for cell viability [15]. Indeed, two pools of ceramide have been demonstrated in other organisms, and they mainly differ based on the length of their fatty acid chain [17,18]. Importantly, the pool of ceramide produced by BarA appears to contribute specifically to membrane organization at the hyphal tip and hence to polarized growth [15].
The first objective of this study was to characterize the role of lipid microdomains in the process of host infection by F. graminearum. Our prior studies suggest that cell surface organization impacts pathogenicity [19]. Here, we exploit our previous work on the A. nidulans BarA ceramide synthase to address the role of sphingolipids in this process. In particular, we hypothesized that deletion of the F. graminearum barA homologue BAR1 would alter cell surface organization and hence disturb plant infection. The second objective of this study was to determine the nature of the ceramides generated by the BarA/Bar1 ceramide synthase. Given that different classes of ceramides are produced (e.g. different fatty acid chains, different head groups etc.), we surmised that Bar1 contributes to the production of a specific class of sphingolipid.

Strains and culture conditions
All F. graminearum strains used in this study were derived from strain PH-1 (NRRL 31084). The DBAR1 mutants were generated by transforming strain PH-1 as described below. Strain P2 is a derivative of PH-1 that expresses the hygromycin phosphotransferase (hph) gene from plasmid pUCH2-8 [19]. Strains DFMK1 and DMGV1 were the kind gifts of Dr. Jin-Rong Xu, Purdue University. Stocks were maintained by storing mycelia in 30% (v/ v) glycerol solution at 280uC. Strains were maintained solid V8 agar medium [19].
To assess macroconidia production, 100 ml of a 1610 4 per ml macroconidial suspension was spread inoculated onto YMA [20] or YMA+4% NaCl and incubated at room temperature for seven days. In general, we obtain greater yields of macroconidia when using this approach instead of the standard method of growth in liquid CMC media (also, some of the mutants generated in our laboratory do not sporulate well in CMC). Macroconidia were harvested in 2 ml of sterile distilled water and counted with a hemacytometer. Three readings were recorded per plate and averaged, with the data presented representing averages from five replicate plates, Macroconidia lengths and widths were measured using differential interference contrast microscopy and IPLab Imaging Software (Scanalytics, Inc).
Biomass was assessed by inoculating 50 ml liquid YMA with 5 ml of 1610 5 per ml macroconidia suspension, followed by incubation on a rotary shaker set at 28uC and 200 RPM for three days. The resulting mycelium was dried at 60uC for 16 hours and the mass was recorded. Measurements represent the average across three independent experiments.
Wheat head inoculations (variety ''Norm'') were performed as previously described [19] . Development of subcuticular and wide, invasive hyphae was observed on detached wheat glumes as previously described [21]. Sexual crosses were performed on carrot agar as previously described [19,22]. Sensitivity to heat stable antifungal factor [HSAF; 16] was tested by mixing HSAF (suspended in methanol) with molten YMA amended to 0.05% Tergitol. Equal amounts of methanol were added to each treatment group as a control.

Generation of DBAR1 mutant
To identify an F. graminearum barA homologue, the BarA sequence (ANID_04332; http://www.broadinstitute.org/annotation/genome/aspergillus_group/MultiHome.html) was used in a pBLAST search against the F. graminearum proteome (http:// www.broadinstitute.org/annotation/genome/fusarium_group/Multi Home.html). For phylogenetic analysis, the amino acid sequences of other fungal ceramide synthases were retrieved from the NCBI (www. ncbi.nlm.nih.gov). MacVector software (MacVector Inc, Cary, North Carolina) was used to align the sequences and generate a neighborjoining tree, and bootstrap values were generated after 1000 iterations.
A split-marker approach was used to replace the endogenous BAR1 (FGSG_09423.3) gene of F. graminearum with a hygromycin phosphotransferase (hph) marker from plasmid pUCH2-8 [23]. The constructs necessary to perform this transformation were obtained using the following steps. First, a 1890 bp fragment upstream of BAR1 start codon was amplified from genomic DNA with primers bko1 and bko2 (Fig. 1A). Similarly, a 1989 bp fragment downstream of BAR1 was amplified using primers bko3 and bko4. Also, two separate fragments were amplified from the hph cassette of pUCH2-8; the first fragment was amplified with primers H1 and H2 and encompassed the first 1737 bp of the cassette, while the second fragment was amplified with primers H3 and H4 and encompassed the last 1626 bp of the cassette [19]. Importantly, these two fragments from the hph cassette shared 552 bp of homologous sequence [24]. Primers bko2/H1 and primers H4/bko3 had complementary tails to promote fusion PCR (see below). All fragments were then gel purified using the QIAquick Gel Extraction Kit (Qiagen). Then, a fusion PCR approach was used to fuse the upstream genomic fragment to the first hph sequence, and a separate reaction was used to fuse the second hph fragment to the downstream genomic fragment (Fig. 1A). All of the above PCR reactions (including the fusion PCR) were performed using the High Fidelity PCR kit (Roche) according to the manufacturer's instructions. The template DNA used for the fusion PCR was an equal amount of the two fragments to be fused plus primers bko1/H2 (for the upstream fusion fragment) or primers H3/bko4 (for the downstream fusion fragment).
Replacement of the endogenous BAR1 gene was accomplished by transforming protoplasts of strain PH-1 with the two fusion fragments (750 ng of each fragment) generated as described above. Transformation of protoplasts was performed as described previously [19]. Hygromycin resistant (hyg R ) colonies were collected 7-10 days later and cultured on V8 medium with hygromycin B (300 mg/ml). Proper incorporation of the hph cassette at the BAR1 locus was assessed using primers P1-P6 and Taq polymerase (Invitrogen) according to the manufacturer's instructions (Fig. 1). The DBAR1-31 strain was complemented with plasmid pBR30.1, which was generated by ligating a full-length copy of BAR1 into the XbaI site of plasmid NatXho1-1 containing a nourseothricin resistance marker.

Characterization of cell surface organization in DBAR1 mutants
Nuclei and hyphae were observed in germinated macroconidia by staining with Hoechst 33258 (Molecular probes) and Calcofluor (= fluorescent brightener 28; Sigma) respectively, as described previously [20]. In order to observe the distribution of sterol-rich lipid microdomains, germinated macroconidia were stained for three minutes in media containing 15 mg/ml Filipin (Sigma) and imaged using an Olympus BX51 fluorescent microscope. To assess cell wall defects, strains were tested on YMA containing Calcofluor white (fluorescent brightener 28; Sigma) and Congo red (Sigma) as described previously [25].
To test for presence of the Spitzenkorper, macroconidia were grown for four hours at 28uC on coverslips covered with liquid YMA. After 4 hours, coverslips were stained with 32 mM FM4-64 (Sigma) and washed 3 times in pre-warmed YMA. The coverslips were then submerged once again into liquid YMA, incubated for 1.5 more hours, then imaged using fluorescence microscopy. All images were recorded within three minutes of mounting. Germinated macroconidia were stained at least two times for each analysis.

Analysis of sphingoid long chain bases
Long chain bases (LCBs) of sphingolipids were analyzed by hydrolyzing the amide bond between the LCB and the fatty acid using a protocol described previously [26]. The hydrolyzed bases were then conjugated to ophthaldialdehyde and analyzed with high performance liquid chromatography (HPLC) with a C18 column. The production of glucosylceramide was further tested by enriching lipid fractions for glucosylceramides and thin layer chromatography as described previously [26].
Generation of the Yarrowia lipolytica DGCS1 mutant. The URA3 gene (YALI0E26741g) of Y. lipolytica was used for targeted integrative disruption of the GCS1 (YALI0B09669g) gene. A ,1.0 kb DNA fragment upstream of the ATG start codon of the GCS1 open reading frame was amplified by the PCR using the primers GCSHind III-A and GCSBamHI-B. The resulting fragment was digested with Hind III and BamHI and inserted into the corresponding sites of the vector pUC19 to generate pUC19UP. Then a ,1.0 kb DNA fragment downstream of the stop codon of the open reading frame was amplified using the primers GCSBamHI-C and GCSEcoRI-D. This fragment was digested with BamHI and EcoRI and inserted into the vector pUC19UP to produce the vector pUC19UPDS. Then a ,4.8 Kb DNA fragment containing the Yarrowia lipolytica URA3 gene was amplified with the primers URA5' and URA3'. This fragment was digested by BamHI and ligated into the BamHI site of the vector pUC19UPDS. Then a ,6.8 Kb DNA fragment containing the URA3 gene flanked by the upstream and downstream of the GCS sequence was amplified by primes GCSHind III-A and GCSEcoRI-D. The PCR product was gel purified and used to transform Y. lipolytica strain ATCC90811 (leu2-35 lys5-12 ura3-18) using the frozen-EZ yeast transformation II kit ( Zymo Research, Orange, CA). Transformants were selected on URA dropout media [0.17% (w/v) yeast nitrogen base, 2% (w/v) glucose, 0.5% (w/v) ammonium sulfate, and 0.08% (w/v) CSM-URA (MP Biomedicals, Solon, OH)]. Genomic DNA obtained from transformants was analyzed by PCR to confirm deletion. For this analysis, the primers GCSup (located ,1.1 Kb upstream of the GCS ATG start codon) and GCSdown (located internal of the DNA fragment containing the URA3 genomic gene) were used. Potential deletion mutants were further confirmed by thin layer chromatographic analysis of lipid extracts as previously described [26].

Identification and deletion of BAR1 in Fusarium graminearum
We have previously described the existence of two distinct clades of acyl-CoA-dependent ceramide synthases in filamentous fungi [15]. This analysis revealed the presence of ceramide synthases related to both LagA (FGSG_05525) and BarA (FGSG_09423 and FGSG_03851) in F. graminearum. Here, we present the characterization of FGSG_09423, now designated Bar1, which possesses 41% identity and 58% similarity (Evalue = 1e 292 ) with BarA. We note that FGSC_03851, which has  slightly weaker homology to BarA (39% identity, 58% similarity, E-value = 2e 276 ), will be the subject of future studies. The expression of the BAR1 and FGSG_03851 genes remains relative unchanged during F. graminearum infection of barley as well as during nutrient starvation. However, whereas the BAR1 transcript increases approximately 10-fold during the first 2 hours after macroconidia germinate, the FGSG_03851 transcript decreases approximately 4 orders of magnitude during this same time period (Fig. S1). On the other hand, BAR1 and FGSG_03581 are expressed at comparable levels during latter stages of in vitro growth (i.e., 72-96 hrs.) as well as during in vivo sexual development (dikaryotic hyphae to mature perithecia).
In order to characterize the function of Bar1 in F. graminearum, a split-marker approach was used to replace the BAR1 gene in wildtype strain PH-1 with a hygromycin phosphotransferase marker from plasmid pUCH2-8. Various primers were designed to confirm the complete replacement of the endogenous BAR1 gene with the hph marker (Fig. 1). Two independent transformants were collected and displayed similar phenotypes (strains DBAR1-21 and DBAR1-31). DBAR1 phenotypes (see below) were complemented by transforming the DBAR1-31 strain with plasmid pBR30.1, which contains an endogenous BAR1 coding sequence plus a nourseothricin-resistance maker (Fig. 1C).
Growth and sporulation in DBAR1 mutant DBAR1 mutants displayed a severe colonial phenotype (Table 1; Fig. 1C). In addition, the biomass of strain DBAR1-31 was lower than that of the wild-type and ectopic strain after three days of growth in liquid YMA (Table 1). Nevertheless, macroconidia of DBAR1 germinated at similar time points compared to control strains ( Fig. 2A). These data suggest that deletion of BAR1 triggers growth defects at later stages of colony formation. Accordingly, whereas germlings of DBAR1 mutants looked 'healthy' immediately after germ tube release, they were heavily vacuolated and abnormal in appearance upon extended incubation in liquid media (Fig. 2B). One potential explanation for this observation is that accumulation of a metabolite was inducing cell death (i.e. growth rate was hindered only after a period of time that allowed accumulation of a deleterious metabolite). In A. nidulans, the sphingoid ceramide precursors dihydrosphingosine and phytoshphingosine induce programmed cell death [PCD]; [27]. In order to determine if PCD was occurring in the DBAR1-31 strain, germinated macroconidia were stained with an in situ cell death detection kit (i.e. TUNEL assay; Roche). While some nuclei of the DBAR1 mutants stained positive, the absence of extensive staining (e.g., as we previously observed in wild-type F. graminearum hyphae exposed to farnesol; [28]) implies that PCD is not the cause of cell deterioration (data not shown).
While attempting to harvest macroconidia for microscopic observation, it was noticed that DBAR1 mutants sporulate very poorly. No macroconidia were observed when strains were grown on solid YMA or liquid CMC ( Fig. 3B; data not shown). This sporulation defect was partially alleviated by amending YMA with 4% NaCl (Fig. 3A,B). Even in this case, the macroconidia of DBAR1 strains displayed morphological abnormalities in that they were significantly shorter and did not display the slender, canoeshape morphology typical of F. graminearum macroconidia ( Fig. 3A; Table 1). In addition to the defect in asexual reproduction, the DBAR1 mutant was not capable of forming perithecia, the sexual fruiting structures of F. graminearum (Fig. 3C). These data suggest that Bar1 contributes significantly to sexual and asexual sporulation in F. graminearum.
TheDBAR1-31 strain showed no symptoms when inoculated onto wheat heads (unpublished data). However, the severe growth phenotype exhibited by DBAR1 strains made the interpretation of this result difficult, as any defect in host tissue colonization could not be separated from its severe in vitro phenotype. However, the morphological defects displayed in sexual and asexual reproduc-tion made us question whether or not the DBAR1 mutants would be defective in differentiating infection-related hyphae [21]. Despite their inability to efficiently differentiate sexual and asexual structures, DBAR1 mutants were able to differentiate both subcuticular hyphae and bulbous infection hyphae when inoculated onto detached wheat glumes ( Fig. 4; Table 1). These data suggest that Bar1 is not absolutely required for infection-related hyphal development in F. graminearum.

Sphingolipid production
Ceramides are required for a number of cellular processes, including signaling and membrane integrity [10,29]. Our previous analyses suggested that polarized hyphal extension may depend upon two distinct pools of ceramide; one generated by BarA, and the other by LagA. Notably, many fungi are capable of producing both inositolphosphoceramides and glucosylceramides [17], leading to the idea that each type is specifically synthesized by one of the ceramide synthases. To better understand the nature of the ceramide pools generated by BarA homologues, o-phthaldialdehyde derivatives of the long chain sphingoid bases from the DBAR1 mutant were analyzed by high performance liquid chromatography (HPLC) [30]. Preliminary data suggested that the DBAR1 mutant of F. graminearum and the barA1 mutant of A. nidulans were specifically defective in glucosylceramide (GlcCer) production (Fig. 5A). In order to further test this hypothesis, lipid fractions were isolated from F. graminearum strains, enriched for GlcCer, and analyzed by TLC. These analyses confirmed that the DBAR1 mutant fails to generate GlcCer (Fig. 5B).

HSAF sensitivity
The barA1 mutation in A. nidulans was originally identified as a mutation that conferred resistance to HSAF [15]. Similarly, DBAR1 mutants of F. graminearum are resistant to HSAF (Fig. 6A). An F. graminearum mutant defective in GlcCer production  conferred resistance to the plant defensins MsDef1 and RpAFP2 [26]. Since Bar1 is necessary for the production of glucosylceramides, we hypothesized that HSAF may share the same or similar targets as these plant defensins. Accordingly, a readily available DGCS (glucosylceramide synthase) mutant of the dimorphic fungus Yarrowia lipolytica (M. Chen and E. Cahoon, unpublished results) was tested for sensitivity to HSAF. Plating assays indicated that GlcCer is required for HSAF sensitivity, as the DGCS1 mutant is resistant to HSAF (Fig. 6C). The ability of F. graminearum to cope with MsDef1 is mediated via the Gpmk1 and Mgv1 MAPKs, as mutants in these two genes are more sensitive to this defensin [31]. To test the possibility that these two MAPKs are also involved in the response to HSAF, the DGPMK1 and DMGV1 mutants were tested for their sensitivity. The DMGV1 mutant exhibited an increased susceptibility to HSAF, whereas the sensitivity of the DGPMK1 mutant was similar to that of wild-type strain PH-1 (Fig. 6B). These data suggest that HSAF and the plant defensins MsDef1/RpAFP2 potentially target fungi via related mechanisms that involve glucosylceramide.

Cell surface organization in the DBAR1 mutant
In A. nidulans, barA1 mutants display altered pattern of filipin staining, suggesting a defect in organizing sterol-rich domains at the hyphal tip [15]. Accordingly, F. graminearum DBAR1 mutants were examined and found to display a similar pattern of filipin localization, in which hyphal tips did not stain brighter compared to distal sections of germ tubes (Fig. 7A). Also, the germ tubes of DBAR1 mutants were wider and failed to consistently display a Spitzenkorper when stained with FM4-64 (Fig. 7B). These data suggest that Bar1 is essential for maintaining a defined polarity axis during hyphal growth.
In addition to organization at the membrane surface, we also hypothesized that DBAR1 mutants would exhibit cell wall defects. Sterol-rich domains may represent membranous areas that are rich in lipid microdomains, which have been shown to aggregate membrane proteins such as glycosylphosphatidylinositol (GPI)anchored proteins [8,10]. Whereas GPI-anchored proteins are exclusively membrane proteins in other kingdoms, fungi have the ability to cleave them from the membrane and covalently attach them to the carbohydrate backbone of cell walls [32]. Therefore, a defect in membrane organization may also result in a defect in cell wall organization. To test for cell wall defects, DBAR1 mutants were grown on media containing calcofluor white and Congo red. Calcofluor white and Congo red are two compounds that bind to chitin in fungal cell walls and are useful indicators of cell wall defects [25]. The DBAR1 mutant was more sensitive to both compounds, whereas PH-1 and the complemented mutants were resistant, even up to 10 mg/ml for Congo red (Fig. S2; data not shown). These data suggest that Bar1 contributes to the cell wall structure of F. graminearum, Nevertheless, TEM analysis failed to identify any gross structural changes in the cell wall of DBAR1 mutants (Fig. S2).

Discussion
Previous work has implicated fungal acyl coA-dependent ceramide synthases of the BarA class in the regulation of polarized hyphal growth. Here, we present the first characterization of BarA in a plant pathogenic fungus, with a view toward determining its importance for virulence. We also describe biochemical experiments designed to reveal the nature of the ceramides generated by BarA. Our results demonstrate that the F. graminearum BarA homologue, Bar1, plays an unexpectedly significant role in growth and development. In addition, we find that BarA homologues Figure 6. HSAF sensitivity of DBAR1, DGCS1, and DMAPK mutants. A. DBAR1 mutants are resistant to HSAF. 7 ml of macroconidial suspensions (serially diluted from 10 5 to 10 2 per ml) were spotted onto plates containing different concentrations of HSAF. B. MAPK mutant DMGV1 is hypersensitive to HSAF, but DGPMK1 is unchanged in sensitivity. C. Glucosylceramides contribute to HSAF sensitivity, as a Yarrowia lipolytica DGCS1 mutant is resistant to HSAF. Plates in panels B and C were inoculated as described in A. doi:10.1371/journal.pone.0019385.g006 from F. graminearum and Aspergillus specifically generate glucosylceramides. Collectively, our results emphasize the importance of glucosylceramides as key regulators of growth and development in filamentous fungi.

Roles of Bar1 in growth, development, and pathogenicity
Unlike A. nidulans, F. graminearum possesses two BarA homologues; FGSG_09423 (Bar1) and FGSG_03851 [15]. Because Bar1 possesses slightly greater homology to A. nidulans BarA, and is expressed at much higher levels during the period following spore germination, our initial efforts focused on its characterization. Surprisingly, deletion of BAR1 caused significant defects in growth, formation of macroconidia, and sexual development. By contrast, in A. nidulans, null barA mutants are capable of forming relatively normal sporulating colonies despite defects in maintaining axes of hyphal polarity [15]. Furthermore, defects caused by the absence of Bar1 in F. graminearum only became apparent several hours after spore germination. The abnormal appearance of these ''older'' hyphae raised the possibility that they had undergone PCD, but no evidence for apoptosis could be detected using TUNEL assays. Although poor growth of the DBAR1 mutant could reflect the accumulation of a toxic metabolite that triggers a necrotic response, an alternative explanation is that Bar1 is required for hyphal maturation. In other filamentous fungi (i.e., Ashbya gossypii and A. nidulans), growing germlings mature into hyphae that extend at a much faster rate via a mechanism that requires microtubules and PAK kinases [33,34]. In DBAR1 mutants, failure of hyphae to mature may lead to slow growth and accumulation of vacuoles.
When tested using a detached wheat glume assay, DBAR1 mutants were able to differentiate both subcuticular hyphae and bulbous infection hyphae. Accordingly, the early stages of infection-related morphogenesis do not appear to require functional BarA. Nevertheless, DBAR1 mutants did not trigger any symptoms when inoculated onto wheat heads. This observation is presumably a reflection of the overall growth defect caused by absence of BarA, as opposed to a specific role in pathogenesis. Results from other studies, however, do implicate glucosylceramides as important factors in the virulence of both plant and animal pathogens [26,35]. Notably, glucoylceramides were recently identified as important virulence factors in the human pathogen Candida albicans despite the absence of any overt role in hyphal morphogenesis [36].
The severity of the phenotypes caused by deletion of BAR1 was somewhat surprising given the presence of a second homologue, FGSG_03851. The relatively poor expression of this homologue during hyphal growth implies that Bar1 is the dominant ceramide synthase of this class during this phase of the life cycle. On the other hand, FGSG_03851 is expressed at the same levels as Bar1 during sexual development, yet the BAR1 deletion still confers a significant developmental defect. This observation suggests that these two BarA homologues do not simply function redundantly. One interesting possibility is that FGSG_03851 might be associated with a toxin biosynthetic cluster. Indeed, the predicted functions of nearby genes coincide with the functions of many mycotoxin biosynthesis genes (Tri8 orthologue, p450 oxidases, transcription factor etc; Table S1). Also, in other Fusarium species (i.e., F. verticillioides), BarA homologues (i.e., Fum18) flank the fumonisin biosynthetic gene cluster [37], where they may play a role in altering plant or fungal ceramide pools in the presence of fumonisins.

BarA homologues generate glucosylceramides
Our prior studies in A. nidulans strongly suggested the presence of two distinct ceramide pools that regulate growth and morphogenesis in filamentous fungi [15,38]. In particular, BarA was proposed to generate a specialized ceramide pool that ensures the maintenance of hyphal polarity, whereas LagA is responsible for bulk ceramides that contribute to growth. Here, our analysis of sphingolipids in both A. nidulans and F. graminearum establish that BarA homologues generate C18 glucosylceramides. Furthermore, the phenotypes of barA and DBAR1 mutants implicate glucosylceramides as important regulators of hyphal morphogenesis and development. For example, they may help to promote the formation of membrane domains at hyphal tips that recruit the Spitzenkorper and other components needed for rapid hyphal extension.
The non-glucosylceramide-producing yeast S. cerevisiae encodes two ceramide synthases, lac1 and lag1, which are functionally redundant and prefer C:26 fatty acids as a substrate for ceramide production [39,40]. Ceramides with C:26 fatty acids are ultimately converted to inositol phosphorylceramides that serve several essential functions in cells. However, in the glucosylceramide-producing yeast species Figure 8. Neighbor-joining phylogenetic analysis of Lag1, Lac1, and Bar1 protein sequences in Ashbya gossipii (Ag), A. nidulans (An), Candida albicans (Ca), Candida glabrata (Cg), Debaryomyces hansenii (Dh), F. graminearum (Fg), Kluyveromyces lactis (Kl), Vanderwaltozyma polyspora (Vp), and Yarrowia lipolytica (Yl). HsLass1p/6p are ceramide synthase sequences from Homo sapiens used as out-groups. This analysis suggests that ScLac1 (and Lac1 genes from other non GlcCer-producing yeast strains) likely represents a Lag1 paralog, whereas Lac1 from GlcCerproducing yeast strains is an ancestral Bar1 orthologue that was lost in non GlcCer-producing strains. Bootstrap values calculated from 1000 iterations. doi:10.1371/journal.pone.0019385.g008 Kluyveromyces lactis, Lac1 is specifically necessary for the production C:18 sphingolipids, including glucosylceramide [17]. Phylogenetic analysis of Lag1 and Lac1 amino acid sequences among several yeast species revealed the Lag1 and Lac1 from non-producing strains cluster together into a single clade, whereas Lag1 and Lac1 sequences from glucosylceramide-producing strains separate into two distinct clades (one 'Lag' clade and one 'Lac' clade) [17]. These data suggest the Lag1 and Lac1 in non-producing strains are actually paralogous, that is, they represent duplicates of an ancestral gene ( = lag1). In contrast, Lac1 proteins in glucosylceramide-producing strains likely represent orthologs of an ancestral ceramide synthase distinct from Lag1 (i.e. presumably BarA). Accordingly, we speculate that 'Lac1' proteins in glucosylceramide-producing yeast strains are actually Bar1 orthologues (which have been lost in non GlcCer-producing strains) required for production of C:18 ceramides and ultimately, as demonstrated in this study, glucosylceramides. Phylogenetic analyses support this hypothesis, as F. graminearum Bar1 and A. nidulans BarA form a distinct clade with 'Lac1' proteins from glucosylceramideproducing yeasts, while Lac1 proteins from non-producing strains cluster within the Lag1 clade (Fig. 8).

Glucosylceramides as targets for anti-fungal compounds
Like its A. nidulans orthologue BarA, Bar1 is required for sensitivity to the antifungal HSAF. The observations that DBAR1 mutants are resistant to HSAF and do not produce glucosylceramides raises the possibility that HSAF targets these sphingolipds. Consistent with this notion, HSAF is ineffective against S. cerevisiae, which does not typically generate glucosylceramide [18,41]. Further support for this notion was demonstrated in this study, as a glucosylceramide synthase (GCS) deletion mutant in the dimorphic yeast Yarrowia lipolytica is resistant to HSAF. Interestingly, glucosylceramides have been implicated in the sensitivity to other antifungal compounds. For example, DGCS1 mutants of F. graminearum are resistant to the plant defensins MsDef1 and RpAFP2 [26]. The interactions of these defensins with glucosylceramide may be direct, as the defensin RpAFP2 was shown to directly bind to glucosylceramides of fungal origin [42]. F. graminearum MAPK mutants DMGV1 and DGPMK1 are hypersensitive to MsDef1, suggesting that their signaling cascades mediate basal resistance to this plant defensin [31]. Similarly, DMGV1 mutants are more sensitive to HSAF, whereas DGPMK1 mutants do not differ from wild type in their sensitivity. Our data illustrate similarities, but also differences, between how HSAF and MsDef1 target fungal cells, though both clearly involve glucosylceramide.

Conclusions
We have demonstrated that the BAR1/barA ceramide synthase genes in F. graminearum and A. nidulans are specifically required for the production of glucosylceramides. Their failure to generate this class of ceramides is likely responsible for their resistance to HSAF, as a Y. lipolytica DGCS1 mutant is also resistant to HSAF. The failure to produce glucosylceramides, coupled with the fact that C:18-ceramide is the typical substrate from which glucosylceramides are synthesized, suggest that Bar1 and its orthologs specifically catalyze the condensation of C:18 fatty acids with sphingoid bases. Interestingly, the Bar1 enzyme appears to have been lost in some yeast species and replaced with a duplicate copy (i.e. paralog) of the C:26-ceramide producing Lag1. Presently, the 'non-Lag' ceramide synthases of several yeast species are designated 'Lac1', though they more closely resemble Bar1 orthologs (Fig. 8). Our data also provide some preliminary evidence as to the role that glucosylceramides (or other C:18ceramides) may play in filamentous fungi, such as their importance in sexual reproduction and cell wall organization. Future work should focus on the specific function of C:18-ceramides at cell surface organization and vesicle transport at the hyphal tip.