Genetic Analysis of Biosurfactant Production in Ustilago maydis

Abstract

The dimorphic basidiomycete Ustilago maydis produces large amounts of surface-active compounds under conditions of nitrogen starvation. These biosurfactants consist of derivatives of two classes of amphipathic glycolipids. Ustilagic acids are cellobiose lipids in which the disaccharide is O-glycosidically linked to 15,16-dihydroxyhexadecanoic acid. Ustilipids are mannosylerythritol lipids derived from acylated β-d-mannopyranosyl-d-erythritol. Whereas the chemical structure of these biosurfactants has been determined, the genetic basis for their biosynthesis and regulation is largely unknown. Here we report the first identification of two genes, emt1 and cyp1, that are essential for the production of fungal extracellular glycolipids. emt1 is required for mannosylerythritol lipid production and codes for a protein with similarity to prokaryotic glycosyltransferases involved in the biosynthesis of macrolide antibiotics. We suggest that Emt1 catalyzes the synthesis of mannosyl-d-erythritol by transfer of GDP-mannose. Deletion of the gene cyp1 resulted in complete loss of ustilagic acid production. Cyp1 encodes a cytochrome P450 monooxygenase which is highly related to a family of plant fatty acid hydroxylases. Therefore we assume that Cyp1 is directly involved in the biosynthesis of the unusual 15,16-dihydroxyhexadecanoic acid. We could show that mannosylerythritol lipid production is responsible for hemolytic activity on blood agar, whereas ustilagic acid secretion is required for long-range pheromone recognition. The mutants described here allow for the first time a genetic analysis of glycolipid production in fungi.

Many microorganisms produce surface-active compounds that can fulfill different roles in microbial physiology. Biosurfactants allow the adhesion of microorganisms to hydrophobic surfaces, increase the bioavailability of water-insoluble substrates, and often show antimicrobial activity (35). Many bacterial species secrete high-molecular-weight biosurfactants consisting mainly of polysaccharides, lipoproteins, or lipopolysaccharides. Low-molecular-weight biosurfactants are generally glycolipids or lipopeptides and are produced by a variety of microorganisms, including bacteria and fungi. Secreted lipopeptide antibiotics with high surface activity were first found in Bacillus subtilis, which produces surfactin (4). Extracellular glycolipids consist of different mono- or disaccharides that are either acylated or glycosidically linked to long-chain fatty acids. Although these microbial glycolipids have attracted some technological interest (12), very little is known about their regulation and biosynthesis (43). The production of rhamnolipid by Pseudomonas aeruginosa is the best-understood pathway at the genetic level. The synthesis of this glycolipid proceeds by sequential glycosyl transfer reactions. The genes involved in rhamnolipid biosynthesis are encoded on a plasmid, and their expression is regulated by a quorum sensing system (for review, see reference 28).

Production of extracellular glycolipids has also been detected in many fungal species. Sophorose lipids are secreted by Candida bombicola (21), mannosylerythritol lipids (MEL) were first isolated in the dimorphic fungus Ustilago maydis as extracellular oil with a higher density than water (19) and were later detected also in Candida antarctica, Schizonella melanogramma, and Geotrichum candidum (11, 24, 27). The primary structure of these glycolipids has been determined (6, 13, 16, 27). They consist of a disaccharide formed by mannose and erythritol, which is acylated at the mannose moiety with fatty acids of different lengths (Fig. 1C). Recently, mannosylerythritol lipids from U. maydis were identified in a screen for inhibitors of mammalian dopamine receptors (27). The authors of this study named these glycolipids ustilipids and determined the composition of the main variants of these glycolipids by mass spectrometry (27).

FIG. 1.

Glycolipid production in U. maydis. (A) Haploid cells of U. maydis strain MB215 produce large amounts of insoluble glycolipids under conditions of nitrogen starvation. Glycolipids are visible as needle-like precipitates. Length of scale bar, 20 μm. (B and C) Chemical structures of (B) ustilagic acids and (C) ustilipids.

In addition to MEL production, U. maydis secretes a second class of glycolipids, the cellobiose lipid ustilagic acid. Ustilagic acid was first described in 1950 by Haskins, who observed in submerged culture of U. maydis the formation of an insoluble compound with antibiotic activity (18) (Fig. 1A). The sugar moiety of ustilagic acid is the disaccharide cellobiose, which is O-glycosidically linked to the ω-hydroxyl group of the unusual long-chain fatty acid 15,16-dihydroxyhexadecanoic acid or 2,15,16-trihydroxyhexadecanoic acid (Fig. 1B). In addition, the cellobiose is esterified at several positions with either acetyl or medium-chain hydroxy fatty acids (30). Cellobiose-derived glycolipids have since been detected in other fungal species (26, 34). The fungal biocontrol agent Pseudozyma flocculosa produces flocculosin, a rare cellobiose lipid with antifungal activity (9). Production of both types of glycolipids in U. maydis occurs readily under conditions of nitrogen starvation and can reach large yields (up to 23 g/liter). The yield and ratio of both classes of glycolipids depend on the available carbon source and can be shifted towards either of these biosurfactants (42).

Although glycolipid production in fungi has been known for a long time, the genetic basis of their production and regulation is largely unknown. Here we describe the cloning of two genes involved in glycolipid biosynthesis in U. maydis. We have identified a putative glycosyltransferase, Emt1, which is required for production of mannosylerythritol lipids. Sequence similarity with mycosamine transferases from macrolide-producing actinomycetes suggests that Emt1 catalyzes the transfer of mannose to erythritol. In addition, we could abolish the production of the cellobiose-lipid ustilagic acid by deleting the P450 monooxygenase Cyp1. Since ustilagic acid contains the unusual 15,16-dihydroxyhexadecanoic acid, we assume that Cyp1 is involved in terminal and/or subterminal hydroxylation of this fatty acid. These mutants allow, for the first time, a molecular characterization of glycolipid biosynthesis in fungi.

MATERIALS AND METHODS

Strains, plasmids, and culture conditions.

U. maydis strains FB1 (a1 b1) and FB2 (a2 b2) have been described^. (5). Strain MB215 is a wild isolate and was collected in northern Germany. Its mating type was determined as a2 b13. Strain MB215 and the derived emt1 and cyp1 mutants were deposited at the German Collection of Microorganisms and Cell Cultures (DSMZ) (accession numbers: DSM17144 [MB215], DSM17145 [MB215cyp1], DSM17146 [MB215emt1], and DSM17147 [MB215cyp1emt1]). U. maydis strains were grown at 28°C in liquid YEPS (1% yeast extract, 2% peptone, 2% sucrose) or on solid potato dextrose agar (PDA). Solid medium contained 1.5% (wt/vol) Bacto-agar. For selection, PD plates containing 2 μg/ml carboxin or 200 μg/ml hygromycin were used. Glycolipid production strains were precultured at 28°C in liquid potato dextrose broth (PDB) (2.4%) to logarithmic phase and then shifted to nitrogen starvation medium containing 1.7 g/liter YNB (yeast nitrogen base without ammonium sulfate) and 5% glucose as the carbon source. Glycolipids were isolated after cultivating cells for 4 days at 28°C on the rotary shaker. Transformation of U. maydis was performed as described (38).

Escherichia coli strains DH5α and XL1-Blue MR [mcrA183 Δ(mcrCB-hsdSMR-mrr)173 endA1 supE44 thi-1 recA1 gyrA96 relA1 lac] were used for all DNA manipulations. E. coli was grown in LB medium (1% NaCl, 1% tryptone, 0.5% yeast extract) or in dYT medium (1.6% tryptone, 1% yeast extract, 0.5% NaCl).

Generation and characterization of mutants defective for glycolipid production.

The Δemt1 and Δcyp1 mutants were generated according to the published protocol (23). The upstream flanking region of emt1 was amplified by PCR with primers 5′-GCCGCTGTGGCACGTCTCTG-3′ and 5′-CACGGCCTGAGTGGCCCAGGGAGGGGGAGGGAGAGAG-3′, which introduces an SfiI site (italic). The downstream flanking region was amplified by PCR using primers 5′-GTGGGCCATCTAGGCCGGCTCCGCTCATGCATTTCCTCG-3′ and 5′-GGCTTGACCTCTGGCAACAGC-3′. The amplified fragments were cut with SfiI and ligated to the SfiI-digested hygromycin resistance cassette. The knockout construct was amplified with the distal primers and transformed into U. maydis.

The cyp1 gene was replaced accordingly using primers 5′-CAATCTGCCGACGACGGG and 5′-CACGGCCTGAGTGGCCTCTGCAGACCTTCACATAC-3′ for the upstream flanking region and primers 5′-GTGGGCCATCTAGGCCTTCACGATTGTCTTGCCACGGG-3′ and 5′-GCCCGCCCGCAGACCACGCG-3′ for the downstream flanking region.

For Southern blot analysis of the Δemt1 and Δcyp1 mutant strains, genomic DNA was isolated according to (20). Δemt1 DNA was digested with BamHI and probed with an internal 1-kb fragment of the emt1 gene, which was amplified by PCR using primers 5′-GCAGCCACAATGGTCCGCG-3′ and 5′-CGAGTCTGTGAGGCCAGGCGTG-3′. For Southern analysis of the cyp1 mutant strain, the genomic DNA was digested with EcoRV and probed with an internal 500-bp fragment of the cyp1 gene, which was amplified by PCR using primers 5′-GAAGGTGCTCGACATGCGC-3′ and 5′-CTTGGGGATCGATGGATGC-3′.

Northern analysis.

Total RNA was isolated for Northern blot analysis (37). Cultures were grown at 28°C to logarithmic phase in YNB medium containing 5% glucose and 0.2% ammonium sulfate. After centrifugation the cells were resuspended in the same volume of fresh medium containing 5% glucose but lacking a nitrogen source. RNA was prepared at the times indicated. Northern blots were prepared (36). For the emt1 expression analysis, an internal probe of the gene was used. This 500-bp probe was amplified by PCR using primers 5′-AACCTGACATGACTAACCCTGTCATCTTGATT-3′ and 5′-ACATATGTCAATCGCTCCCAACAGTGG-3′. For the cyp1 expression analysis a 500-bp internal probe of the gene was used. This probe was amplified by PCR using primers 5′-GAAGGTGCTCGACATGCGC-3′ and 5′-CTTGGGGATCGATGGATGC-3′. Northern hybridization was performed as described (46).

Isolation of glycolipids.

Extracellular glycolipids were extracted from suspension cultures (0.5 ml) with 1 volume of ethyl acetate. The ethyl acetate phase was evaporated and glycolipids were dissolved in methanol (12). Glycolipids were analyzed by thin-layer chromatography (TLC) on silica plates (Silica gel 60; Merck) with a solvent system consisting of chloroform-methanol-water (65:25:4, vol/vol). The plate was dried thoroughly and sugar-containing compounds were visualized by spraying with a mixture of glacial acetic acid-sulfuric acid-p-anisaldehyde (50:1:0,5, vol/vol) and heating at 150°C for 3 min (14).

Phenotypic characterization of mutants.

Hemolytic activity was analyzed by spotting 10 μl of an overnight culture on blood agar plates (24 g/liter potato dextrose, 20 g/liter agar, 8 g/liter NaCl, and 5% sheep blood [Oxoid]). Hemolysis was monitored after incubation for 3 days at 28°C.

Confrontation assays were performed as described previously (40). Plant infections were essentially done as described previously (15). For infections the maize cultivar Early Golden Bantam (purchased from Olds Seed Solutions, Madison, WI) was used.

The surface activity of supernatants of strains grown under conditions of nitrogen starvation was assayed by their ability to collapse droplets on a hydrophobic surface (parafilm M laboratory film) as described (25); 25 μl of culture supernatant were pipetted after adding xylene cyanol for staining. The dye xylene cyanol has no influence on the shape of the droplets.

RESULTS

Isolation of a strain with a high level of glycolipid production.

Extracellular glycolipids were detected in U. maydis more than 50 years ago (18). Nevertheless, very little is known about their biosynthesis and their biological function. Therefore we aimed at generating mutants defective for glycolipid production. Since the commonly used laboratory strain FB1 (5) produces only small amounts of ustilagic acid (Fig. 3, lane 1), we screened several wild isolates for enhanced glycolipid production. Cells were grown in rich medium and shifted to nitrogen starvation. After incubation for 4 days, secreted glycolipids were extracted with ethyl acetate and subjected to thin-layer chromatography on silica plates. We could identify strain MB215, which secretes large amounts of both ustilipids and ustilagic acids (Fig. 3, lane 2). We used this strain in a reverse genetics approach to identify genes that are involved in glycolipid biosynthesis in U. maydis.

FIG. 3.

Analysis of glycolipid production by wild-type and glycolipid-deficient strains. Ethyl acetate extracts of extracellular glycolipids were analyzed by thin-layer chromatography. Wild-type strains FB1 and MB215 produce both mannosylerythritol lipids (MEL) and ustilagic acids (UA). Δemt1 and Δcyp1 mutant strains are deficient for MEL and ustilagic acid production, respectively. Double emt1 cyp1 mutants produce neither of these substances. The slow-migrating band with low intensity which is visible in the Δcyp1 strains corresponds to an unidentified compound.

Putative mannosyltransferase Emt1 is required for ustilipid production.

To generate mutants of U. maydis that are unable to produce mannosylerythritol lipids (MELs), we reasoned that a major step in MEL biosynthesis would be the generation of the central mannosyl-β-d-erythritol moiety. This assumption is supported by the fact that this compound has been isolated in significant amounts from MEL-producing U. maydis cells as a water-soluble substance (7). The most likely pathway would be the transfer of activated GDP-mannose to the C-4 hydroxyl group of meso-erythritol. No enzyme has been described so far that catalyzes such a reaction. Therefore, we searched the publicly available genome database of U. maydis for putative glycosyltransferases that could be involved in the generation of this moiety.

About 40 genes encoding proteins with some similarity to glycosyltransferases were identified in the U. maydis database. The function of some of these enzymes could be derived by similarity to known glycosyltransferases involved in cell wall biosynthesis or protein glycosylation. For the remaining candidate genes, mutants were systematically generated by a PCR-based deletion strategy (23). In brief, flanking sequences were amplified with primers that introduce specific sticky ends. These were used to ligate both flanks with a hygromycin resistance cassette, which carries compatible ends. After another round of PCR amplification, the replacement constructs were transformed into protoplasts of the haploid U. maydis strains MB215 and FB1. Transformants were checked for successful deletion of the respective genes by Southern analysis (Fig. 2A). The deletion mutants were grown in rich medium and shifted to nitrogen starvation. After 4 days glycolipids were extracted with ethylacetate and subjected to thin layer chromatography. One of the mutants showed a total loss of mannosylerythritol (MEL) production as detected by TLC analysis (Fig. 3, lanes 3 and 4). Whereas in extracts from wild-type strains, several bands could be detected that correspond to the different ustilipids, no such signals were visible in extracts from the mutant strain. The lack of ustilipid secretion was confirmed by liquid chromatography/mass spectometry analysis of the ethyl acetate extract (data not shown). Thus, deletion of this putative glycosyltransferase completely blocks MEL production.

FIG. 2.

Southern and Northern blot analysis of emt1 and cyp1 genes. (A) Southern analysis of emt1 deletion mutants. Genomic DNA was cut with BamHI and probed with an internal 1-kb emt1 fragment. (B) Southern analysis of Δcyp1 mutants. DNA was digested with EcoRV and probed with an internal 500-bp cyp1 fragment. (C) Northern analysis of emt1 and cyp1 expression. Total RNA was prepared at the indicated time points after shifting cells to nitrogen starvation medium. The expression of the constitutively expressed ppi gene was used as a control. ppi encodes the U. maydis peptidyl-prolyl cis-trans isomerase (accession number EAK84904).

The corresponding gene of this mutant encodes a 628-amino-acid polypeptide (accession number XP_400732) that displayed the highest similarity to a predicted protein from Aspergillus nidulans (accession number XP_412272) whose function has not been determined (Fig. 4). More informative is the significant similarity to a group of actinomycete glycosyltransferases, which are involved in the production of macrolide antibiotics (for review, see reference 3). These prokaryotic enzymes have been proposed to transfer GDP-mycosamine, which is derived from GDP-mannose, to the polyketide structure of macrolides. Therefore we assume that U. maydis uses GDP-mannose to transfer a mannosyl residue to the C-4 hydroxyl group of meso-erythritol. Thus, we termed the corresponding gene emt1 (Erythritol-mannosyl-transferase).

FIG. 4.

Sequence alignment of the U. maydis Emt1 glycosyltransferase. The amino acid sequence of Emt1 was aligned with the Emt1 homologue AN8135.2 from A. nidulans (accession number EAA58772) and with the PimK protein from Streptomyces natalensis, involved in pimaricin biosynthesis (accession number CAC20918).

Emt1 expression is highly induced under conditions of glycolipid production.

Secretion of ustilipids is observed only in medium that contains a carbon source but lacks nitrogen. Therefore we analyzed the expression pattern of emt1 under these conditions. By Northern analysis we could not detect any expression of emt1 in rich medium. Under conditions of nitrogen starvation, expression of emt1 is visible after 4 h and reaches its maximum at 24 h. The time point of emt1 expression correlates perfectly with the onset of glycolipid production (Fig. 2C).

Next we asked whether overexpression of Emt1 stimulates glycolipid production. We placed the open reading frame of Emt1 under control of the arabinose-inducible crg promoter (8). If this construct was transformed into the Δemt1 mutant strain, restoration of MEL production was observed. Although the crg promoter is known to result in high levels of expression, only weak MEL production could be observed in arabinose-containing medium. This is most likely due to low glycolipid production in the presence of arabinose as a carbon source since wild-type strains also produce only small amounts of glycolipids if arabinose is the only available carbon source (data not shown).

Isolation of mutants defective in ustilagic acid production.

All of the glycosyltransferase mutant strains that we generated were still able to secrete the cellobiose lipid ustilagic acid. Therefore we decided to search for other candidate genes involved in the production of this cellobiose lipid. Ustilagic acid contains an unusual fatty acid which caries two vicinal hydroxyl groups at its distal end (15,16-dihydroxyhexadecanoic acid or 2,15,16-trihydroxyhexadecanoic acid). We reasoned that these hydroxy fatty acids are most probably derived from nonhydroxylated precursors. Such hydroxylation reactions are often catalyzed by cytochrome P450 enzymes (for review, see reference 17); 23 genes in the genome database of U. maydis are annotated as P450 enzymes (http://mips.gsf.de/genre/proj/ustilago/).

We have concentrated on two of these genes that code for proteins related to the CYP94A1 family of P450 enzymes. Members of this family have been identified in the plant Vicia sativa as fatty acid omega-hydroxylases (44). The U. maydis genes cyp1 (accession number EAK87170) and cyp2 (accession number EAK82744), both encoding P450 enzymes, were deleted by gene replacement (Southern analysis shown for cyp1 in Fig. 2B). The resulting mutants were tested for glycolipid secretion. Whereas the production of glycolipids was not affected in the cyp2 deletion mutant, no cellobiose lipids could be detected in the ethyl acetate extract of cyp1 mutant cells (Fig. 3).

The predicted Cyp1 polypeptide sequence consists of 628 amino acids and contains a single intron. Cyp1 is very similar to another U. maydis P450 enzyme (Um06473, accession number EAK87284) and a P450 enzyme from the related fungus Rhodotorula sp. strain CBS 8446 (accession number AY316198) (Fig. 5). In a BLAST search with the Cyp1 sequence, many P450 enzymes of plant origin were found to be very similar to Cyp1. Of these plant enzymes, only the members of the CYP94A1 family from Vicia sativa have been characterized on the molecular level. CYP94A1 catalyzes the ω-hydroxylation of epoxy and hydroxy fatty acids, probably as part of a plant defense reaction (33). The highly related P450 enzyme CYP94A2 of V. sativa hydroxylates medium-chain fatty acids at their terminal or subterminal end, depending on the length of the fatty acid. Thus, we suggest that Cyp1 may catalyze the terminal or subterminal hydroxylation of hexadecanoic acid to generate the precursor for ustilagic acid biosynthesis.

FIG. 5.

Comparison of P450 enzymes similar to U. maydis Cyp1. The conserved domain characteristic of cytochrome P450 enzymes is shaded. The number indicates the percentage of identical amino acids compared to Cyp1. The following sequences were aligned to Cyp1 using the BLAST algorithm (2): U. maydis Um06473 (accession number EAK87284), Rhodotorula sp. strain CBS 8446 cytochrome P450 gene (accession number AY316198), and Vicia sativa CYP94A2 (accession number AAG33645).

Expression of cyp1 is induced by nitrogen starvation but is detectable only 24 h after shifting cells to the new medium (Fig. 2C). This correlates with the time course of glycolipid production in U. maydis, which reaches significant levels after 2 to 4 days of cultivation (42).

Phenotype of glycolipid-defective mutants.

To create a strain which is unable to secrete any glycolipids, we crossed mutants of compatible mating type by infecting corn seedlings. The haploid progeny were analyzed for glycolipid production. We could easily detect strains that secreted neither of the two glycolipids (Fig. 3). We checked the double mutants for their mating type and selected strains with compatible a and b mating types (SH6Δemt1Δcyp1 and SH9Δemt1Δcyp1). With the glycolipid-defective mutants at hand, we were able to analyze the phenotype of these strains in more detail.

To test whether glycolipid secretion is required for cell recognition during mating, Δemt1, Δcyp1, and Δemt1/Δcyp1 mutants were tested for mating on charcoal-containing agar plates (10). None of these mutants showed a significant mating defect in combination with either wild-type or mutant strains (data not shown). To test whether pheromone recognition over larger distances is affected in the mutants, we used the confrontation assay (40). In this assay compatible strains were placed close to each other on water agar and incubated under liquid paraffin. In combinations of wild-type cells, the induction of conjugation hyphae growing towards the pheromone source is observed predominantly in cells of the a2 mating type (Fig. 6, WT). Deletion of the emt1 gene did not interfere with the formation of conjugation hyphae, but cells appeared to stick to each other in the hydrophobic environment (Fig. 6, Δemt1). Loss of ustilagic production in Δcyp1 mutants, however, abolishes conjugation tube induction, indicating that this glycolipid is required for pheromone recognition over large distances. The phenotype of double mutants appears to be a combination of that of single mutants (Fig. 6).

FIG. 6.

Confrontation assay to test for pheromone recognition. Compatible combinations of cells with the indicated genotype were placed in close proximity on water agar and covered with liquid paraffin. After cultivation for 16 h, cells were observed under the microscope. Formation of conjugation hyphae occurs only in combinations of wild-type (WT) and Δemt1 strains. The mating type of the strains is indicated at the left. Length of scale bar, 30 μm.

Next we tested whether the loss of glycolipid production affects the ability of the fungus to infect corn plants. Mutants were mixed with compatible wild-type or mutant strains and injected into corn seedlings. Tumor formation occurred with all combinations with similar incidences, and the symptoms induced were indistinguishable from those induced by wild-type strains (Table 1). Thus, glycolipid production is not essential for the virulence of U. maydis.

TABLE 1.

Pathogenicity assay

Cells	No. of infected plants	No. with tumors
FB1 × MB215	15	6
FB1Δemt1 × MB215Δemt1	25	10
FB1Δcyp1 × MB215Δcyp1	16	9
SH6Δemt1Δcyp1 × SH9Δemt1Δcyp1	32	12

Hemolytic activity of U. maydis is caused by mannosylerythritol lipid secretion.

Many biosurfactant-producing microorganisms can be detected by their hemolytic activity on blood agar plates (47). Therefore we tested wild-type and mutant strains for hemolytic activity. On blood agar plates, the wild isolate MB215 induces a prominent clear zone around the growing cells, which is significantly larger than that caused by the laboratory strain FB1 (Fig. 7A). Deletion of the P450 enzyme involved in ustilagic acid production does not cause a significant reduction of hemolytic activity. Loss of mannosylerythritol production, however, results in complete lack of hemolytic activity (Fig. 7A). Next we tested whether the hemolytic activity is mainly due to the surface activity of secreted glycolipids. The effect of glycolipid secretion on the surface tension of culture supernatants was assessed by estimating their ability to cause the collapse of droplets of culture supernatants on hydrophobic surfaces (25). The presence of cellobiose lipids resulted in only a small reduction of surface tension, whereas the presence of mannosylerythritol lipids in the medium caused a significant collapse of the droplet (Fig. 7B). The effect of glycolipids was additive since the surface tension of culture supernatants from the double mutants appeared to be higher than that of the single mutants (Fig. 7B).

FIG. 7.

Determination of surface activity of secreted glycolipids. (A) Hemolytic activity of U. maydis strains is observed on blood agar plates after incubation for 2 days. A clear zone of hemolysis is observed only for wild-type (WT) and Δcyp1 mutant strains. (B) Secretion of amphipathic glycolipids reduces the surface tension of the culture supernatant. The effect of glycolipid secretion on the surface tension of the culture medium was visualized by placing 25 μl of culture supernatant on the hydrophobic surface of parafilm. Fresh culture medium was used as a control. The pictures were taken directly from above the droplets (upper panel) and from the side (lower panel).

DISCUSSION

We have identified two genes required for glycolipid production in the dimorphic fungus Ustilago maydis. To our knowledge, this is the first molecular characterization of fungal mutants that are affected in the production of glycolipid biosurfactants. In the fungus Pseudozyma flocculosa, which is used as a biocontrol agent, random mutants have been generated by insertional mutagenesis and screened for their antagonistic properties (9). One of these mutants was defective for the production of a cellobiose glycolipid (flocculosin). However, the mutant has not been characterized at the molecular level (9). Thus, it is not known whether structural genes of biosurfactant synthesis are affected in this mutant.

Synthesis of mannosylerythritol lipids.

The Emt1 protein required for mannosylerythritol lipid production displays a high level of sequence similarity with glycosyltransferases involved in actinobacterial macrolide antibiotic production. In the synthesis of candicidin, amphotericin- and pimaricin-related enzymes have been proposed to transfer mycosamine to the macrolide chain (3). Mycosamine is a derivative of mannose and is generated by the activity of several enzymes that modify the activated GDP-mannose. Therefore we suggest that the U. maydis Emt1 protein may use GDP-mannose for the generation of the mannosylerythritol moiety of ustilipids.

In the secreted mannosylerythritol lipids, the mannose is acylated by one long-chain and several short-chain fatty acids. In principle these alterations could occur at the level of GDP-mannose; however, MEL-producing U. maydis cells contain significant amounts of water-soluble mannose erythritol (19). Thus, we assume that Emt1 directly transfers mannose to the C-4 atom of meso-erythritol. This reaction has to be stereospecific, since only mannosyl-d-erythritol is generated. We are currently investigating whether this reaction can be followed in vitro using Emt1 protein overexpressed in E. coli cells. If this reaction pathway could be confirmed, it would constitute a novel enzymatic activity which may have some interesting biotechnological applications, since mannosylerythritol compounds have been used for the production of diverse products (12).

P450 enzyme is required for cellobiose lipid production.

We have identified the cytochrome P450-encoding gene cyp1, which is essential for the production of the cellobiose lipid ustilagic acid. A main characteristic of ustilagic acid is the presence of unusual hydroxy fatty acids which are O-glycosidically linked to cellobiose at their terminal hydroxyl group. Whereas many enzymes have been detected that are able to hydroxylate terminal or subterminal carbon atoms of long- and medium-chain fatty acids, no enzymatic activity has been described so far that would generate a pair of vicinal hydroxyl groups at the distal end of long-chain fatty acids.

Clues about the function of Cyp1 may come from its similarity to the plant enzyme CYP94A2 from Vicia sativa which hydroxylates medium-chain fatty acids (29). Characteristically, the position of the introduced hydroxyl group depends on the length of the fatty acid precursor. Whereas short molecules like C₁₂ fatty acids are predominantly hydroxylated at their C-terminal ends, longer fatty acids with 14 or 16 carbon atoms are hydroxylated at their subterminal position (29). To explain this peculiar regioselectivity, it has been suggested that the enzyme recognizes the carboxyl end of the fatty acid and hydroxylates the position, which is reached by the catalytic core (29). In accordance with this hypothesis, it was found that CYP94A2 is unable to hydroxylate alkanes. In addition, a mutant of CYP94A2 was identified in which a single amino acid substitution has altered this regiospecificity (22).

Thus, we suggest that production of the dihydroxyhexadecanoic acid in U. maydis occurs by two subsequent hydroxylation events. Whether subterminal hydroxylation follows the terminal hydroxylation or vice versa remains to be elucidated. In addition, it has to be clarified whether Cyp1 alone is sufficient to catalyze both steps or whether another specific P450 enzyme is required. A good candidate for such an additional enzyme could be the U. maydis enzyme encoded by the gene Um06473, which is highly similar to Cyp1 (Fig. 6). We are currently generating mutants to test whether deletion of this gene also affects cellobiose lipid production.

In principle, the vicinal hydroxyl groups of the proposed ustilagic acid precursor could also be the result of epoxide-hydrolase-catalyzed hydrolysis of a terminal epoxide (39). The reactive epoxide could be the result of epoxidation of a corresponding unsaturated fatty acid derivative. However, neither such an unsaturated fatty acid nor the necessary desaturase activity has ever been found in U. maydis or in any other fungus.

Biological function of glycolipid production in U. maydis.

We observed that the expression of both genes involved in glycolipid production was strongly induced by nitrogen starvation. Although the molecular basis of nitrogen regulation is well known in some fungi (for review, see references 31 and 32), similar studies have not yet been performed in U. maydis. However, inspection of the genome sequence revealed the presence of a GATA-like transcription factor (accession number XP_401867) which is highly similar to AreA, the central regulator of nitrogen metabolism in Aspergillus nidulans. AreA and the related GATA-like transcription factors recognize binding sequences which carry the characteristic DNA motif GATA. Inspection of the promoter sequences of both emt1 and cyp1 revealed the presence of several GATA elements. Some of these are located as pairs within 30 bp and thus are potential strong binding sites for GATA transcription factors. However, whether emt1 and cyp1 are directly controlled by a GATA factor or whether the observed induction is indirect remains to be tested.

The observed expression pattern correlated with the time course of glycolipid production. The fact that expression of these genes cannot be detected under normal conditions and that deletion of these genes is not lethal imply that these glycolipids are typical secondary metabolites (45). U. maydis is the only fungus known so far that produces two different classes of glycolipids simultaneously in significant amounts. Both compounds display surface activity, but the effect of the mannosylerythritol lipid appears to be significantly higher. Thus, it might be that both biosurfactants fulfill distinct requirements during the life cycle of U. maydis.

Since we have generated mutant strains that produce only either one of these glycolipids or none, we are now able to dissect the different contributions of these biosurfactants. Secretion of glycolipids occurs in the haploid phase, when U. maydis grows unicellular by budding. We could demonstrate that the cellobiose lipids are required for pheromone recognition over large distances during mating. The pheromones secreted by U. maydis are hydrophobic lipopeptides that are nearly insoluble in water (41). We could imagine that the biosurfactants produced by the fungus might form a thin layer at the interface between hydrophobic and hydrophilic surfaces. Two-dimensional diffusion of pheromones within this film will then greatly enhance the chance to be detected by cells that are at a distance (1). The observation that mutants are still able to infect corn plants indicates that the experimental conditions used to infect corn seedlings in the greenhouse may not be sensitive enough to detect slight differences in pheromone recognition. In particular, large amounts of cells (>10⁷ cells) are injected into the stem of corn seedlings and thus are likely to come into direct contact within the plant tissue.

The isolation of mutants that are specifically affected in glycolipid production will help us to understand the molecular mechanisms involved in the biosynthesis of these interesting natural compounds. At the same time, these mutants will allow further phenotypic assays and competition experiments to be devised to determine the biological function of extracellular glycolipids in their natural environments.