Detoxification of the Fusarium Mycotoxin Deoxynivalenol by a UDP-glucosyltransferase from Arabidopsis thaliana*

Plant pathogenic fungi of the genus Fusarium cause agriculturally important diseases of small grain cereals and maize. Trichothecenes are a class of mycotoxins produced by different Fusarium species that inhibit eukaryotic protein biosynthesis and presumably interfere with the expression of genes induced during the defense response of the plants. One of its members, deoxynivalenol, most likely acts as a virulence factor during fungal pathogenesis and frequently accumulates in grain to levels posing a threat to human and animal health. We report the isolation and characterization of a gene from Arabidopsis thaliana encoding a UDP-glycosyltransferase that is able to detoxify deoxynivalenol. The enzyme, previously assigned the identifier UGT73C5, catalyzes the transfer of glucose from UDP-glucose to the hydroxyl group at carbon 3 of deoxynivalenol. Using a wheat germ extract-coupled transcription/translation system we have shown that this enzymatic reaction inactivates the mycotoxin. This deoxynivalenol-glucosyltransferase (DOGT1) was also found to detoxify the acetylated derivative 15-acetyl-deoxynivalenol, whereas no protective activity was observed against the structurally similar nivalenol. Expression of the glucosyltransferase is developmentally regulated and induced by deoxynivalenol as well as salicylic acid, ethylene, and jasmonic acid. Constitutive overexpression in Arabidopsis leads to enhanced tolerance against deoxynivalenol.

A complex of closely related species of the genus Fusarium is responsible for destructive and economically very important diseases of cereal crops (Fusarium head blight of wheat and barley) and maize (Fusarium ear rot). In years with climatic conditions that favor the development of the fungi, these infections can reach epidemic proportions (1). Diseases caused by Fusarium do not only severely reduce yield but also result in contamination of grain with unacceptably high amounts of mycotoxins, a problem of world-wide significance. A toxin class of particular concern to human and animal health is the trichothecenes, sesquiterpenoid epoxides, which are potent inhibitors of eukaryotic protein synthesis. More than 180 compounds of this class have been isolated from natural sources, predominantly from Fusarium species (2). Depending on the concentration and the substitution pattern of the trichothecene, either translational initiation, elongation, or termination are preferentially inhibited (3).
Fusarium graminearum and Fusarium culmorum are the two most common causative agents of Fusarium head blight of cereals. F. graminearum lineages present in Europe and North America predominantly produce deoxynivalenol (DON) 1 as well as the acetylated derivatives 3-acetyl-deoxynivalenol (3-ADON) and 15-acetyl-deoxynivalenol (15-ADON), whereas producers of nivalenol (NIV), which contains one additional hydroxyl group (Fig. 1A), predominate in Asia (4). While the toxicity of these trichothecenes is well studied in animal systems (5), fairly little is known about differences in phytotoxicity. In contrast to animal cells, NIV and T-2 toxin are less toxic than DON to wheat (6), which may indicate significant differences in uptake or metabolism of particular trichothecenes. Animal exposure to DON (also known as vomitoxin) has numerous adverse health effects with neural and immune system being the most sensitive targets (7). In contrast to high doses of DON, which inhibit antibody production, low doses of DON act synergistically with bacterial lipopolysaccharide to stimulate proinflammatory processes (7). To protect consumers, the United States Food and Drug Administration has established advisory levels for food. The European Community has also recently recommended action levels for DON (8). 2 The role of DON in plant disease is still a matter of discussion (9), but most of the evidence supports the hypothesis that it functions as a virulence factor. Fusarium mutants with a disrupted trichodiene synthase (Tri5) gene, involved in trichothecene biosynthesis, are still pathogenic. However, they exhibit reduced virulence on wheat (10); they are unable to spread from the infection site (11). DON can move ahead of the fungus in infested plants, suggesting a role in conditioning host tissue for colonization (12). Results of several studies indicate that the in vitro resistance of wheat cultivars toward DON correlates with Fusarium head blight resistance in the field (13). This is also the case for segregating experimental populations. 3 Increased resistance against toxic substances can result from several mechanisms, ranging from reduced uptake to bypass, overexpression, or mutation of the toxin target. Yet a very prominent process seems to be metabolic transformation often followed by compartmentation (14). An observed decline in the concentration of DON, which occurred in Fusarium-infected wheat in the field (15), suggests that the toxin may be metabolized. Two wheat cultivars differing in Fusarium resistance also differ in their ability to form a DON metabolite, suspected to be a glycoside (16). Sewald et al. (17) used radiolabeled DON to show that it was primarily conjugated to 3-␤-D-glucopyranosyl-4-deoxynivalenol in a maize suspension culture. No plant enzymes capable of modifying the trichothecene have been described so far.
Genome sequencing projects have revealed the existence of a vast number of genes in plants that code for putative UDPglycosyltransferases (UGTs), predicted to conjugate small molecules. For instance, Arabidopsis thaliana has been shown to harbor more than 100 members of this multigene family (18) whose functions are largely unknown. Here we report the cloning of an Arabidopsis UGT by functional expression in yeast that is able to inactivate the Fusarium mycotoxin DON.
Plant Material and Growth Conditions-A. thaliana experiments were conducted with the wild-type ecotype Columbia-0 (Col-0). For propagation, seeds were sterilized, plated on standard Murashige and Skoog growth medium (22) supplemented with 1.0% sucrose and 1.0% phytagar (Invitrogen), and subjected to a 2-day dark treatment at 4°C to synchronize germination. The seedlings were grown for 2 weeks in a controlled environment of 16 h/8 h light-dark cycle (140 mol m Ϫ2 s Ϫ1 white light) at 22°C before they were transferred to soil and grown at 20°C and 55% humidity under continuous white light.
A. thaliana cDNA Library Screen in Yeast-The ATP-binding cassette transporter-deficient Saccharomyces cerevisiae strain YZGA452, which is hypersensitive to DON, was transformed with an A. thaliana cDNA library constitutively expressed under the control of the phosphoglycerate kinase (PGK1) promoter (23). A total of 10 7 transformants were selected on minimal medium lacking uracil and transferred to medium containing 180 ppm DON (kindly provided by Marc Lemmens), a dose sufficient to completely inhibit growth of yeast transformed with the empty library plasmid. Colonies that showed resistance were isolated, and the plasmid dependence of the phenotype was tested by plasmid DNA preparation and retransformation of YZGA452. The NotI fragment containing the cDNA insert of the candidate was subcloned into pBluescript SKIIϩ (Stratagene) and sequenced.
The vector pYAK7 was constructed by first inserting the doublestranded linker 5Ј-GGGATGCCCGAACAAAAGTTAATTTCAGAAGA-GGACTTATCAAAGCTTGAGGCCTCGCGA-3Ј into the SmaI site of vector pAD4⌬ (24), thus generating an N-terminal c-Myc epitope and a HindIII site into which a genomic HindIII fragment containing the yeast PDR5 was inserted in-frame.
The tagged UGT construct was verified by sequencing and used to transform the yeast strain YZGA515. The empty vector (pYAK7 HindIII ϩ NotI-digested and religated) was used as a control. Transformants were selected on synthetic complete medium lacking leucine. Exponentially growing cultures were diluted to A 600 0.05 and spotted on YPD (1% yeast extract, 2% peptone, 2% dextrose) medium containing increasing concentrations of different trichothecenes. The toxins used were: DON, 3-ADON, 15-ADON, NIV, trichothecin, T-2 toxin, HT-2 toxin, diacetoxyscirpenol, and verrucarin A. With the exception of DON, 3-ADON, and NIV, which were obtained from Biopure Referenzsubstanzen GmbH (Tulln, Austria), mycotoxins were purchased from Sigma and stored at Ϫ20°C dissolved in 70% ethanol.
For immunodetection, the extraction of proteins from yeast cells was performed as described by Egner et al. (25). Western blot analysis was conducted with a primary mouse anti-c-Myc antibody (1:5000, clone 9E10, Invitrogen).
Isolation and Analysis of DON Metabolites in Vivo-To elucidate the chemical structure of DON metabolites resulting from enzymatic transformation of the mycotoxin by DOGT1, a highly tolerant strain was constructed, and the DON metabolite was extracted from toxin-treated cells for subsequent HPLC analysis. The yeast strain YZGA515 was transformed with c-Myc-tagged DOGT1 under the control of the constitutive ADH1 promoter and with plasmid pRM561. This plasmid contains a mutant version of the ribosomal protein L3 (RPL3) of S. cerevisiae that substantially increases DON resistance when expressed in yeast. 5 Transformants were selected on synthetic complete medium lacking leucine and adenine.
The resultant yeast strain was grown to an A 600 0.7 in selective medium (synthetic complete lacking leucine and adenine). Cells were transferred to adenine-supplemented YPD medium (10% glucose), grown for 2 h at 30°C, harvested by centrifugation, and diluted to an A 600 of 3.0 in YPD. DON was added to 5 ml of culture starting at a dose of 200 ppm. After 3 h the concentration was increased to 400 ppm, after 6 h the concentration was increased to 600 ppm, and after 9 h the concentration was increased to 1,000 ppm. The cells were incubated for an additional 15 h at 30°C. The cells were then harvested, washed three times with ice-cold water, extracted in 2.5 ml of methanol/water (4:1), and sonicated. After centrifugation the supernatant was filtered through a glass microfiber filter (Whatman 1822 025). 500-l aliquots of the yeast extracts were concentrated to dryness under a constant stream of nitrogen and dissolved in 100 l of HPLC-grade water.
Heterologous Expression of DOGT1 in Escherichia coli-The DOGT1 protein was expressed in E. coli XL1-blue as a GST fusion. The DOGT1 gene was isolated from the yeast expression vector by HindIII digestion and Klenow fill-in followed by a NotI digest. The resulting fragment was cloned into the SmaI ϩ NotI sites of the GST gene fusion vector pGEX-4T-3 (Amersham Biosciences). The recombinant fusion protein was purified using glutathione-coupled Sepharose (Amersham Biosciences) according to the manufacturer's instructions.
To test the effect of the N-terminal GST tag on activity, the gene encoding the fusion protein was PCR-amplified using DNA polymerase with proof reading activity (Pfu polymerase, MBI) and the fusion protein-specific primers GSTDOGpYAK7-fw (5Ј-TCACCCGGGAAACAG-TAATCATGTCC-3Ј) and GSTDOGpYAK7-rv (5Ј-CGAGGCAGATCGT-CAGTCAGTC-3Ј). The PCR product was cloned into the HindIII ϩ NotI sites of the yeast expression vector pYAK7. DON detoxification ability was tested by expression in YZGA515 and application to toxin-containing medium as described above. Enzyme Assays-The glucosyltransferase activity assay mixture contained 1 g of recombinant GST fusion protein, 10 mM 2-mercaptoethanol, 50 mM Tris-HCl, pH 7.0, 0.5 mM radioactive labeled UDP-[ 14 C]glucose (4.4 ϫ 10 3 cpm, PerkinElmer Life Sciences), 0.01% bovine serum albumin, and a 1 mM concentration of acceptor substrate (dissolved in Me 2 SO in 20 mM stock solutions). The reactions were carried out in 20-l volumes at 30°C for 1 h, stopped by adding 2 l of trichloroacetic acid (240 mg/ml), frozen, and then stored at Ϫ20°C.
Analysis of reaction products was performed by TLC. An aliquot of each sample was spotted on a silica gel plate (Kieselgel 60, Merck) and developed with a mixture of 1-butanol/1-propanol/ethanol/water (2:3: The intensity of each radioactive spot was determined using a PhosphorImager (STORM 860 system, Amersham Biosciences). The plates were additionally stained with p-anisaldehyde (0.5% in methanol/H 2 SO 4 /acetic acid, 85:5: Inhibition of Wheat Ribosomes in Vitro-To analyze whether the 3-␤-D-glucopyranosyl-4-deoxynivalenol is less phytotoxic than its aglycone, we used a wheat germ extract-coupled in vitro transcription/ translation system (TNT coupled wheat germ extract, T3, Promega). After performing transcription/translation reactions for 25 min according to the manufacturer's instructions (in the presence of either DON, purified DON-glucoside (1, 2.5, 5, 10, and 20 M), or water as a control), the activity of the firefly luciferase reporter was determined (luciferase assay system, Promega) using a luminometer (Victor 2, Wallac). Analysis of mRNA Expression of DOGT1 by Reverse Transcriptase PCR-Total RNA was isolated from plant tissue ground in liquid nitrogen with TriZol reagent as recommended by the manufacturer (Invitrogen). RNA was quantified photometrically and visually on a denaturing RNA gel analyzing 5 g of total RNA. cDNA was synthesized from 1 g of total RNA (digested with DNase I) using 500 ng of an 18-mer oligo(dT) and SuperScript reverse transcriptase (Invitrogen). PCR was performed with ϳ2 l of the 1:20 diluted cDNA using primers that amplify C-terminal fragments of DOGT1 and the UBQ5 control gene: DOGRT-fw, 5Ј-ATCCGGGGTT-GAACAGCCT-3Ј;

Plant Treatment with Different Stress Response-related Compounds for Expression Analysis-For
DOGRT-rv, 5Ј-TCAATTATTGGGTTCTGCC-3Ј; UBQ5-U, 5Ј-GTCCTTCTTTCTGGTAAACGT-3Ј; UBQ5-D, 5Ј-AACC CTTGAGGTTGAATCATC-3Ј). To compare relative amounts of transcripts in the samples, DNA fragments of the UBQ5 were first amplified, and normalized sample volumes, based on the amount of products corresponding to the UBQ5 transcripts, were used for PCR.
Cloning of Plant Overexpression and GUS Fusion Constructs-For constitutive overexpression of c-Myc-tagged DOGT1 protein in Arabidopsis, the vector pBP1319 was constructed. It is derived from a modified version of the plant expression vector pPZP221 (27). The promoter cassette, consisting of two copies of the 35S promoter and the polyadenylation signal of cauliflower mosaic virus strain Cabb B-D, came from the vector p2RT, a modified version of pRT100 (28).
The c-Myc-tagged DOGT1 fragment was released by SmaI ϩ NotI digestion (Klenow-filled) from the yeast expression vector and cloned into ClaI ϩ SmaI-digested pBluescript SKIIϩ, which was treated with Klenow enzyme. In the next step, the gene was excised from the resulting plasmid as a SalI ϩBamHI fragment and inserted in the XhoI ϩ BamHI sites of p2RT. The obtained 2x35S c-Myc-DOGT1 cassette was isolated by PstI digestion and cloned into the unique PstI site of pPZP221 after the multiple cloning site in that vector had been destroyed by digesting the plasmid with EcoRI ϩ SalI, filling the sites with Klenow, and religating it (p235a). 6 In the resulting vector pBP1319, the 2x35S c-Myc-DOGT1 cassette is orientated in the opposite direction than the 2x35S gentamycin resistance marker.
For construction of a transcriptional DOGT1-GUS fusion, the GUS vector pPZP-GUS.1, which originates from pPZP200 and contains the GUS gene from pBI101.1 (inserted by HindIII ϩ EcoRI digestion into the MCS), was used (29). The DOGT1 promoter region was PCRamplified from genomic DNA using a DNA polymerase with proof reading activity (Pfu polymerase, MBI) and specific primers (DOGP-GUS-fw, 5Ј-GTTAAAAGCTTACATGTGCATTACGGTCTGTGTGAA-TA; DOGP-GUS-rv, 5Ј-TTTCGGATCCCATG ATTCAACCTTAGTAAG-AAACTCTC). The resulting product was cloned in-frame with the GUS gene by HindIII ϩ BamHI digestion into the pPZP-GUS.1 vector. Constructs were confirmed by DNA sequencing.
Generation and Analysis of Transgenic A. thaliana-For all plant transformations, the recA-deficient Agrobacterium tumefaciens strain UIA143 (30), which harbors the helper plasmid pMP90 (31), was used. For analysis of DON resistance, seeds of homozygous lines exhibiting high DOGT1 expression and Col-0 as control were germinated on Murashige and Skoog media containing different concentrations of DON (5-30 ppm). Seedlings were grown for 5 weeks before the phenotype was documented. GUS activity was analyzed by staining seedlings or organs of adult plants in 5-bromo-4-chloro-3-indolyl-␤-D-glucuronic acid (X-Gluc) solution for 2-4 h at 37°C (33).

Isolation of the DOGT1 by Heterologous Expression in
Yeast-To identify plant genes that contribute to mycotoxin resistance, an unbiased functional screen based on heterologous expression of cDNAs in yeast was set up. Wild-type S. cerevisiae is highly resistant to DON. To reduce the amount of toxin necessary for the screen, we generated a strain deficient in four ATP-binding cassette transporters, which are to a large extent responsible for pleiotropic drug resistance in yeast (34). Strain YZGA452 (snq2⌬::hisG pdr5⌬::TRP1 pdr10⌬::hisG yor1⌬::hisG) is hypersensitive to a wide range of different xenobiotic substances and natural products, including DON (data not shown). YZGA452 was transformed with a cDNA expression library of A. thaliana (23) in which cDNAs are constitutively expressed under the control of the yeast phosphoglycerate kinase promoter. Ten million transformants were generated, and diluted pools of transformants were plated on DON-containing medium. After selection of DON-resistant yeast colonies and confirmation of the plasmid dependence of the phenotype, the insert was subcloned and sequenced.
DOGT1 Is a Member of the UDP-glycosyltransferase Family of A. thaliana and Exhibits High Similarity to Salicylic Acidand Wound-inducible UGTs of Other Species-The cDNA insert conferring resistance was 1.75 kb in length and contained an open reading frame of 1488 bp encoding a putative UDPglycosyltransferase. The identified DOGT1 corresponds to gene UGT73C5 (locus At2g36800) and belongs to subfamily 73C in group D of A. thaliana UGTs (35). Arabidopsis UGTs constitute a very large gene family that has been divided into 14 distinct groups believed to have originated from common ancestors (18).
DOGT1 is located on chromosome II in a cluster with five other members of the subfamily 73C (bacterial artificial chromosome clone F13K3, European Molecular Biology Laboratory (EMBL) accession number AC006282). All six tandemly repeated genes contain no introns and are highly similar to each other (77-89% identity at the amino acid level). The similarity is also very high in the intergenic promoter regions. A data base search using the deduced DOGT1 amino acid sequence revealed high similarity to glucosyltransferases from tobacco (TOGT1, Ref. 36 (40). Expression of the glucosyltransferases from tobacco and tomato has been shown to be elevated following treatment with SA, fungal elicitors, or wounding (36,38,39). Two putative, uncharacterized glucosyltransferases from Vigna angularis (ADGT-9) and Oryza sativa, which have similarity to DOGT1, are also included in the amino acid alignment shown in Fig. 2. Regions of high similarity were observed in both N-and C-terminal domains of the sequences. Indicated in Fig. 2 are the hypothetical acceptor substrate binding region (40) and the UGT consensus sequence (35).

The Expression of DOGT1 Is Developmentally Regulated and Induced by DON and Other Stress Response-related Com-
pounds-To investigate whether DOGT1 expression is regulated similarly to that described for the related genes of other plant species, we constructed a reporter by placing the open reading frame of the ␤-glucuronidase reporter gene behind the DOGT1 promoter (P DOGT1 -GUS). The tissue-specific expression of the transcriptional GUS fusion was examined histochemically in transgenic Arabidopsis homozygous for the fusion gene. The results shown in Fig. 3 demonstrate that DOGT1 expression is regulated developmentally and is overall rather low. In seedlings, GUS activity was observed to be rootand hypocotyl-specific with the strongest expression in the vascular system, in meristematic tissue of the root tips (in the primary root as well as in lateral roots), and in the vasculature of the hypocotyl right after germination. Staining in the vasculature decreased significantly later in development, and a patchy staining pattern appeared in epidermal root cells. In adult plants GUS activity was detected in late stages of flower development in petals and in abscission zones (Fig. 3A).
Exposure of seedlings to either DON (5 ppm or 16.9 M for 4 h, Fig. 3B) or the ethylene precursor ACC (2 M for 1 h, not shown) was found to induce P DOGT1 -GUS expression. No induction of expression of the reporter was detected upon SA treatment (200 M for 12 h) or treatment with JA (50 M for 1 h). Semiquantitative reverse transcriptase PCR was used to validate the results obtained from GUS reporter analyses by detecting changes in mRNA levels of DOGT1 after treatment with the same concentrations of DON, SA, ACC, and JA.
As shown in Fig. 3C the results of the reverse transcriptase PCR confirmed that DOGT1 expression was induced by DON as observed previously with the reporter construct. An increase in the amounts of transcript was apparent after only 1 h of incubation with the toxin, reaching a peak after 4 h before declining between 6 and 12 h. After SA treatment, DOGT1 expression was evident at low levels at 4 h and increased slightly by 12 h. It must be noted that the applied 200 M SA induced expression rather weakly. Jasmonic acid as well as treatment with ACC also led to weak induction of expression of DOGT1 after 1 h of treatment but rapidly declined with no transcript accumulation detectable after 2 h of exposure to the compounds (Fig. 3C).
Phenotypic Determination of the Trichothecene Resistance Spectrum in Yeast-Trichothecene toxicity in animals depends on the hydroxylation pattern as well as on the position, num-ber, and complexity of esterifications (5). The basic trichothecene structure and the numbering of its carbon atoms is shown in Fig. 1A. Members of subclass B (e.g. DON and NIV) contain a keto group at carbon 8, while type A trichothecenes (e.g. the highly toxic T-2 toxin produced by Fusarium sporotrichoides) do not. Extremely toxic also are the macrocyclic trichothecenes, like verrucarin A, which contain a macrocyclic ring with ester bonds bridging carbon 4 and carbon 15. Yeast pdr5 mutants are hypersensitive to all the trichothecenes tested so far, allowing us to investigate the ability of DOGT1 and other genes in the cluster to confer resistance to various members of the trichothecenes.
The DOGT1 gene was expressed in yeast strain YZGA515 as a fusion protein tagged with an N-terminal c-Myc epitope (see "Experimental Procedures"). Several independent yeast transformants were spotted on media containing increasing concen-trations of various trichothecenes; transformants containing the empty expression vector were used as controls.
As shown in Fig. 1B, DOGT1 had the ability to protect against DON and 15-ADON but did not protect against other tested trichothecenes (data not shown). As expected, we observed no increased resistance to trichothecin, which has no free hydroxyl group. Trichothecin was therefore used as a negative control. It should be emphasized that DOGT1 did not protect against 3-ADON, which differs from DON only in that the C-3 hydroxyl group is already blocked by an acetyl group. In contrast, the acetyl group at the C-15 hydroxyl group did not interfere with DOGT1 protection against 15-ADON. DOGT1 expression did not protect against NIV, which differs from DON only in a single additional hydroxyl group (see Fig. 1A).
In Vivo and in Vitro Analysis Prove That DOGT1 Catalyzes the Transfer of Glucose from UDP-glucose Specifically to the 3-OH Position of DON-The protection against 15-ADON and the inability to confer resistance against 3-ADON suggest that DOGT1 may catalyze the formation of a DON-3-O-glucoside. To test this hypothesis, we first chemically synthesized DON derivatives with the glucose moiety attached either to the C-3 or C-15 hydroxyl group. The two products were characterized with LC-MS/MS. The DON-3-O-glucoside and DON-15-O-glucoside eluted at 12.43 and 12.68 min, respectively, and the mass spectra of the glucosides showed characteristic differences in their fragmentation pattern (Fig. 4, A and B). With these tools at hand we were able to directly determine which glucoside was formed in yeast cells. Yeast strain YZGA515 is incapable of converting DON into 3-ADON 7 due to a deletion of the yeast acetyltransferase gene AYT1. This strain was transformed with both the DOGT1 expression vector and a plasmid containing a gene encoding a trichothecene-insensitive mutant ribosomal protein L3 to increase DON tolerance of yeast cells. (Ribosomal protein L3 is the target of trichothecenes. 5 ) 7 G. Adam, unpublished. After the resulting strain had been incubated with DON, reaching a final concentration of 1000 ppm in the medium, the DON metabolite was extracted from the cells. As shown in Fig.  4C, it was then identified as the expected 3-O-glucopyranosyl-4-deoxynivalenol (Fig. 5A) by HPLC and mass spectrometry. Fig. 4 shows the fragmentation pattern of the synthesized reference substances and the product peak from yeast expressing DOGT1. As expected, the metabolite was not present in the control strain lacking DOGT1 activity (data not shown).
To further verify substrate specificity, a GST-DOGT1 fusion was constructed. This GST-DOGT1 fusion gene conferred DON resistance like wild-type DOGT1 when expressed in yeast. To facilitate in vitro testing, the gene product was expressed in E. coli and affinity-purified. The reaction products generated in vitro during incubation of either the DOGT1 fusion protein or GST with UDP-[ 14 C]glucose and DON were analyzed using TLC. A spot with the same R F value as the synthesized DONglucoside was observed in the reaction containing the GST-DOGT1 fusion protein but not in the control (data not shown).

Glucosylation of 4-Deoxynivalenol Greatly Reduces Its
Toxicity-To analyze whether the 3-␤-D-glucopyranosyl-4-deoxynivalenol is less phytotoxic than DON, we used a wheat germ extract-coupled in vitro transcription/translation system. As shown in Fig. 5B, 1 M DON in the reaction mixture inhibited protein translation significantly. Reporter enzyme activity was only 36.8% that of control. 5 M toxin resulted in only 3.1% luminescence remaining, whereas a 20 M concentration of the synthesized DON-3-glucoside inhibited luciferase activity only by 8%. These results demonstrate that glucosylation of DON is a detoxification process.
Overexpression of DOGT1 in A. thaliana Increases DON Resistance-Transgenic A. thaliana constitutively expressing DOGT1 under the control of a tandem 35S promoter were generated, and the amount of recombinant protein in transformants was determined by Western blotting utilizing the Nterminal c-Myc epitope tag. Seeds of the homozygous line 1319/2, which was found to have a high DOGT1 expression level (Fig. 5C), and wild-type Col-0 as control were germinated on Murashige and Skoog media containing 5-30 ppm DON. After 5 weeks of growth, the phenotype was analyzed.
The main phytotoxic effect observed was slow germination of wild-type plants. Root formation failed to occur at all; cotyledons developed but began to bleach before true leaves could be formed. After 5 weeks of DON exposure, most of the wild-type seedlings had lost all green color and ceased growth. DOGT1 overexpression lines also showed a clear delay in development. Compared with wild type, germination occurred earlier, roots were formed, cotyledons did not bleach, and true leaves appeared. These differences in DON sensitivity were most apparent on medium containing 15 ppm of toxin (Fig. 5D). Control transformants with empty vector (containing solely the gentamycin resistance gene) and lines expressing low amounts of DOGT1 were as sensitive as wild-type Col-0 (not shown). DISCUSSION We have cloned a plant UGT that confers resistance to the Fusarium mycotoxin deoxynivalenol by heterologous expression in yeast. Cloning by function, and in particular heterologous expression of plant UGT genes in yeast, is a valuable complementary approach to the widely used E. coli expression systems. This is especially true when the respective chemicals have targets in eukaryotes only. One problem, however, is that wild-type yeast cells are frequently "impermeable" to the substances of interest. Inactivating several ATP-binding cassette transporters in our host strain was a prerequisite for selection of cDNAs conferring resistance to Fusarium toxins. In the case of trichothecenes this approach allowed phenotypic detection of detoxification activity by a simple plate assay. In principle, this approach could be adopted for many other substances.
Biotechnological Relevance-Fusarium diseases of wheat and barley are of high economic significance for countries around the world. The United States for instance has been severely hit by Fusarium epidemics in the last decade. Direct losses to wheat producers in the United States due to Fusarium head blight have been estimated to average about $260 million annually, and over the period 1998 -2000 the combined economic losses for small grain cereals were estimated at $2.7 billion (41). Deoxynivalenol contamination of large portions of a harvest can lead to high human intake of the mycotoxin. Children are the population group most at risk to exceed the tolerable daily intake level for DON. In the problematic year 1998 80% of 1-year-old children in the Netherlands exceeded the tolerable daily intake (8). 2 Thus, methods to prevent and/or treat Fusarium diseases in grains are extremely important. This high importance of Fusarium diseases clearly justifies research on mycotoxin inactivation, although it is very likely that pathogen resistance is a polygenic trait in crop plants, and toxin resistance is just one of its components.
The production of trichothecenes is the only virulence factor of F. graminearum (Gibberella zeae) for which the mode of action is known. It is therefore a prime target for biotechnological approaches to combat the fungus. Other genes required for full virulence have been reported (42)(43)(44). Yet mutations in mitogen-activated protein kinases are pleiotropic, affecting not only virulence but also conidiation, perithecia formation, vegetative growth, and mycotoxin production. In the case of the recently identified CPS1 gene (44), it is unclear whether the hypothetical adenylate-forming enzyme encoded by the gene is involved in biosynthesis of an unidentified metabolite or necessary for some unknown aspect of stress tolerance, in particular in plants mounting a defense response.
Our finding that overexpression of DOGT1 leads to increased deoxynivalenol resistance in transgenic Arabidopsis is signifi- cant. It may open new possibilities for biotechnological approaches aiming to antagonize the fungal virulence factor. So far, such attempts rely on a Fusarium acetyltransferase (45) that converts DON into 3-ADON, which is ϳ2-fold less toxic to laboratory animals and nearly as toxic as DON to wheat (6). The phenotype of our transformed yeast and Arabidopsis and data obtained using in vitro translation assays demonstrate that the DON-glucoside also exhibits reduced toxicity. In general, glycosylation converts reactive and toxic aglycones into stable and non-reactive storage forms, thereby limiting their interaction with other cellular components. The attached sugar blocks the reactive site of the substance and consequently reduces toxicity for the plant. Furthermore such modifications are thought to provide access to membrane-bound transporters, thus opening exit pathways from the cytosol to, for example, the cell wall or the vacuole (46).
Glucosylation of trichothecenes represents a detoxification process for plants. Yet in the digestive tract of humans and animals the mycotoxin-glucoconjugates could easily be hydrolyzed, regenerating the toxin. The extent to which the DONglucoside is transported to the vacuole or to the apoplast is currently unknown. Before one attempts to use transgenic plants overexpressing a DON-glucosyltransferase, it should be investigated whether only the vacuolar glucoside or also toxin bound to cell wall material as "insoluble residue" is a source of "masked mycotoxin" (47). Mycotoxin-glucosides, which escape traditional analytical techniques, may be of much higher toxicological relevance as currently assumed. The analytical tools and reference materials developed during this study should be useful in addressing these questions.
Inactivation of toxins by glycosylation could be a prominent natural mechanism for resistance, enabling plants to cope with the enormous diversity of toxic microbial metabolites they may encounter in nature. DOGT1 is one of 118 UGT genes in A. thaliana predicted to conjugate small molecules. We are currently investigating whether some of these enzymes with closely related sequence also inactivate DON. 8 The C-terminal signature motif of plant UDP-glycosyltransferases is structurally very similar to those of mammalian UGTs that use UDP-glucuronic acid instead of UDP-glucose as donor substrate. The mammalian enzymes play a central role in metabolism and detoxification of chemicals like carcinogens or hydrophobic drugs. Higher plant UGTs have been found to be involved in a parallel range of activities, also modifying xenobiotic substances such as herbicides and pesticides (48,49). One open question is whether these detoxification reactions are side activities of UGTs that are normally involved in conjugating endogenous plant compounds. Recent publications are in favor of that hypothesis (50). Enzymatic activity toward different hydroxycoumarins has been shown for 73C cluster members (51), but the question of whether in vitro activity toward model substrates reflects the in vivo situation requires further work to be solved.
Variation in the substitution pattern of the trichothecene backbone may be an important mechanism to escape detoxification. The finding that DOGT1 protects against DON but not NIV, which possesses one additional hydroxyl group at residue 4 ( Fig. 1A), is consistent with this hypothesis. No product peak was observed in vitro with NIV after incubation with the purified DOGT1-GST fusion protein (data not shown). It is unknown whether plants have enzymes with the ability to glycosylate NIV, a compound with higher toxicity than DON in animal systems (5) but apparently much less toxicity than DON to wheat (6). The lack of activity against NIV may speak against the use of a single DOGT1-like gene to confer Fusarium resistance in transgenic plants since NIV-producing chemotypes of F. graminearum might gain a selective advantage.
Regulation of Gene Expression-Genes with high sequence similarities to DOGT1 from tobacco and tomato have been shown to be induced by salicylic acid or wounding (36,38,39). The analysis of expression of DOGT1 showed elevated mRNA levels in response to SA, JA, and the ethylene precursor ACC. The inducibility of gene expression by SA, JA, or ethylene is considered to be indicative for a possible role of the up-regulated gene product in plant stress or defense responses (52).
Using analysis of mRNA levels and a GUS-reporter construct we were able to show that DOGT1 transcription in wild-type A. thaliana is developmentally regulated and is rapidly and significantly induced in response to DON exposure. It would be interesting to clarify whether the inducibility is compoundspecific or represents a general response to protein biosynthesis inhibitors. The "ribotoxic stress response" is currently being actively investigated by researchers in human cell systems (53). In these cells, the mycotoxin induces expression of cyclooxigenase-2, a key enzyme in the synthesis of inflammatory response mediators, via mitogen-activated protein kinase-mediated signaling. The DON-inducible expression of DOGT1 is an attractive starting point for investigating whether similar mechanisms exist in plants.
In summary, we propose that members of the huge gene family of UGTs could play an important role in plant-pathogen interaction by participating in detoxification of metabolites produced by microbes to increase their virulence on hosts. The selective pressure to escape such glycosylation reactions may be a driving force in evolution of microbial biosynthetic reactions leading to a wide spectrum of toxin structures as observed for trichothecenes.