4-O-Glucosylation of Trichothecenes by Fusarium Species: A Phase II Xenobiotic Metabolism for t-Type Trichothecene Producers

The t-type trichothecene producers Fusarium sporotrichioides and Fusarium graminearum protect themselves against their own mycotoxins by acetylating the C-3 hydroxy group with Tri101p acetylase. To understand the mechanism by which they deal with exogenously added d-type trichothecenes, the Δtri5 mutants expressing all but the first trichothecene pathway enzymes were fed with trichodermol (TDmol), trichothecolone (TCC), 8-deoxytrichothecin, and trichothecin. LC-MS/MS and NMR analyses showed that these C-3 unoxygenated trichothecenes were conjugated with glucose at C-4 by α-glucosidic linkage. As t-type trichothecenes are readily incorporated into the biosynthetic pathway following the C-3 acetylation, the mycotoxins were fed to the ΔFgtri5ΔFgtri101 mutant to examine their fate. LC-MS/MS and NMR analyses demonstrated that the mutant conjugated glucose at C-4 of HT-2 toxin (HT-2) by α-glucosidic linkage, while the ΔFgtri5 mutant metabolized HT-2 to 3-acetyl HT-2 toxin and T-2 toxin. The 4-O-glucosylation of exogenously added t-type trichothecenes appears to be a general response of the ΔFgtri5ΔFgtri101 mutant, as nivalenol and its acetylated derivatives appeared to be conjugated with hexose to some extent. The toxicities of 4-O-glucosides of TDmol, TCC, and HT-2 were much weaker than their corresponding aglycons, suggesting that 4-O-glucosylation serves as a phase II xenobiotic metabolism for t-type trichothecene producers.


Introduction
Trichothecenes are a group of mycotoxins that contain a trichothecene skeleton (12,13epoxy-trichothec-9-ene; EPT) in their structures ( Figure 1). To date, many trichothecenes with various modifying groups have been found. They are produced by several genera of fungi, including Fusarium, Spicellum, Trichothecium, Trichoderma, and Myrothecium. Based on their chemical structures, Ueno classified trichothecenes into four types: A to D [1]. Type A trichothecenes are defined as those lacking a ketone at C-8, including T-2 toxin, to D [1]. Type A trichothecenes are defined as those lacking a ketone at C-8, including T-2 toxin, HT-2 toxin (HT-2), 4,15-diacetoxyscirpenol (DAS), neosolaniol (NEO), 8-deoxytrichothecin , and harzianum A. The former four toxins are produced by Fusarium, such as F. sporotrichioides and F. langsethiae, and the latter two toxins are produced by non-fusaria species. On the other hand, type B trichothecenes, including deoxynivalenol (DON), nivalenol (NIV), trichothecolone (TCC), and trichothecin (TCN), possess a ketone at C-8. While wheat scab fungus, such as F. graminearum and F. culmorum, contaminates grains with DON-and NIV-type trichothecenes, Trichothecium and other genera produce TCC and TCN. Type C trichothecenes have a second epoxide ring between the C-7 and C-8 positions, and type D trichothecenes are macrocyclic trichothecenes with polyketide esters between the C-4 and C-15 positions [2].

Purification and Structure Elucidation of Hexose Conjugate of d-Type Trichothecenes
To confirm the structures of 1 and 2, we carried out large-scale preparations of these putative trichothecene glucoconjugates. In total, 80 mg of TDmol was fed to 2.0 L of GYEP medium inoculated with Fs∆Tri5. After extraction with acetonitrile, the metabolite (16.2 mg) was fractionated as described in Section 3.7. Briefly, the metabolites were applied to Purif ® -Rp2 equipped with a reverse phase column; the partially purified 1 (10.2 mg) was then purified by preparative HPLC twice. The final yield of 1 was 6.0 mg from 40 mg of TDmol fed to Fs∆Tri5. NMR analysis identified 1 as TDmol-4-O-α-glucopyranoside (TDmol-4-glc) (Figures 3 and S2, Table S1). We also fed 72 mg of TCC to the strain Fs∆Tri5 inoculated in 1.8 L of GYEP medium to purify 2 in the same manner; the yield was 12 mg after acetonitrile extraction, 8 mg after Purif-Rp2 purification, and 4.8 mg after HPLC. NMR analysis identified 2 as TCC-4-O-α-glucopyranoside (TCC-4-glc) (Figures 3 and S3, Table S1). These results clearly demonstrate that 4-oxygenated d-type trichothecenes are conjugated with glucose at C-4 when fed to the culture of t-type trichothecene producers. Interestingly, Gorst-Allman et al. have reported on the structure elucidation of 15acetoxyscirpenol-4-O-α-glucopyranoside (15-AS-4-glc) purified from the maize meal culture (5 kg) of Fusarium sulphureum [37] more than 35 years ago. The metabolite was identified as giving a blue color reaction with 4-(p-nitrobenzyl)pyridine (NBP) on a TLC plate. The yield of 15-AS-4-glc was 182 mg under the condition when DAS, in the same culture, was recovered as much as 3.02 g [38]. This suggests that the C-4 glucosylation participates in a minor shunt pathway in the biosynthesis of DAS under certain culture conditions. To the best of our knowledge, this is the only report on the structural confirmation of 4-Oglucoside of a t-type trichothecene by NMR. The observation that Fusarium species show such 4-O-glucosylation activity, regardless of the type of 4-hydroxytrichothecenes, may indicate conserved features of t-type trichothecene producers to the exposure of C-4 oxygenated trichothecenes.

Glucosylation of a t-Type Type A Trichothecene, HT-2, by a Tri101-Null Mutant of F. graminearum
Except for the case of the F. sulphureum large-scale culture [37], there have been no reports so far on the production of t-type trichothecene 4-O-glucosides in axenic cultures of Fusarium strains. Considering the limited accumulation of 15-AS-4-glc on the maize meal medium, it seems likely that the amounts of 4-O-glucosylated trichothecenes are marginal in a small-scale liquid culture. It is also possible that a C-3 hydroxy group, as well as C-4, must not be acylated for efficient conjugation with glucose, as the occurrence of 3,15-diacetoxyscirpenol-4-O-glucoside was not reported in the literature [37].
Since the presence of a C-3 acetyl group is essential to serve as the substrates of the trichothecene pathway enzymes, including the C-4 acetylase Tri7p [39], the lack of the 3-O-acetyl group makes it easier for non-trichothecene enzymes to act on the C-4 hydroxy group. For this reason, the glucose conjugate of trichothecenes could be made more easily by Fusarium when added exogenously to a Tri101 disruption mutant. To assess this possibility, we examined the metabolic fate of HT-2, a t-type type A trichothecene hydroxylated at both C-3 and C-4, in the feeding experiment. F. graminearum FGD5 and FGD5/101 (ΔFgtri101::hph, ΔFgtri5::neo) mutants, with or without trichothecene 3-O-acetyltransfer- Interestingly, Gorst-Allman et al. have reported on the structure elucidation of 15acetoxyscirpenol-4-O-α-glucopyranoside (15-AS-4-glc) purified from the maize meal culture (5 kg) of Fusarium sulphureum [37] more than 35 years ago. The metabolite was identified as giving a blue color reaction with 4-(p-nitrobenzyl)pyridine (NBP) on a TLC plate. The yield of 15-AS-4-glc was 182 mg under the condition when DAS, in the same culture, was recovered as much as 3.02 g [38]. This suggests that the C-4 glucosylation participates in a minor shunt pathway in the biosynthesis of DAS under certain culture conditions. To the best of our knowledge, this is the only report on the structural confirmation of 4-O-glucoside of a t-type trichothecene by NMR. The observation that Fusarium species show such 4-O-glucosylation activity, regardless of the type of 4-hydroxytrichothecenes, may indicate conserved features of t-type trichothecene producers to the exposure of C-4 oxygenated trichothecenes.

Glucosylation of a t-Type Type A Trichothecene, HT-2, by a Tri101-Null Mutant of F. graminearum
Except for the case of the F. sulphureum large-scale culture [37], there have been no reports so far on the production of t-type trichothecene 4-O-glucosides in axenic cultures of Fusarium strains. Considering the limited accumulation of 15-AS-4-glc on the maize meal medium, it seems likely that the amounts of 4-O-glucosylated trichothecenes are marginal in a small-scale liquid culture. It is also possible that a C-3 hydroxy group, as well as C-4, must not be acylated for efficient conjugation with glucose, as the occurrence of 3,15-diacetoxyscirpenol-4-O-glucoside was not reported in the literature [37].
Since the presence of a C-3 acetyl group is essential to serve as the substrates of the trichothecene pathway enzymes, including the C-4 acetylase Tri7p [39], the lack of the 3-Oacetyl group makes it easier for non-trichothecene enzymes to act on the C-4 hydroxy group. For this reason, the glucose conjugate of trichothecenes could be made more easily by Fusarium when added exogenously to a Tri101 disruption mutant. To assess this possibility, we examined the metabolic fate of HT-2, a t-type type A trichothecene hydroxylated at both C-3 and C-4, in the feeding experiment. F. graminearum FGD5 and FGD5/101 (∆Fgtri101::hph, ∆Fgtri5::neo) mutants, with or without trichothecene 3-O-acetyltransferase activity, respectively, were used. When fed to the FGD5 mutant, HT-2 was metabolized to 3-acetyl HT-2 toxin (3-A HT-2) and T-2 toxin, as revealed by TLC and LC-MS/MS analyses (Figure 4a). In contrast, HT-2 fed to FGD5/101 did not yield the same metabolites on TLC. LC-MS/MS analysis of the feeding mixture revealed the emergence of a new XIC peak at 24 h (Figure 4b), which corresponded to [HT-2-hexoside (C 28 Figure 4d). Thus, HT-2 added exogenously was suggested to be metabolized along pathways differently between the FGD5 and FGD5/101 mutants (Figure 4a,b). To confirm the structure of 3, a total of 300 mg HT-2 was fed to FGD5/101 in 15 L of the rice flour (RF) medium, and 35 mg of acetonitrile extract was obtained. After fractionation by Purif ® -Rp2 (yield 24 mg), followed by repeated purification by preparative HPLC, 12 mg of pure 3 was obtained. NMR analysis confirmed that the structure of 3 was HT-2 toxin-4-O-αglucopyranoside (HT-2-4-glc) ( Figures 5 and S4, Table S2).
To examine whether 3-acetyltrichothecenes can serve as the substrates of Fusarium UGT(s), 3-A HT-2 was fed to the strains FGD5 and FGD5/101. The XIC peak corresponding to the HT-2-4-glc ammonium adduct ion was detected from the FGD5/101 culture, which could reasonably be explained by the deacetylation of 3-A HT-2 and the subsequent C-4 glucosylation of HT-2. In contrast, 3-acetyl HT-2 toxin-4-α-glucopyranoside was not found in the feeding extract of either mutant strain. We also incubated HT-2 and 3 individually with Tri101p and acetyl CoA to determine whether C-3 can be acetylated when C-4 is glucosylated. Although more than 88% of HT-2 was converted to 3-A HT-2 in 1 h, acetylated 3 could not be detected under the same condition. Only when a high concentration of Tri101p was used and the reaction time was extended to 6 h, a small amount of 3 acetylated at C-3, with a parent ion peak of m/z 646.3076 corresponding to [3-A HT-2-hexoside (C 30 H 44 O 14 ) + NH 4 ] + (theoretical m/z 646.3069), was barely detectable in MS/MS analysis ( Figure S5; upper panel). The MS/MS spectra of the putative 3-A HT-2 glucoside below an m/z value of 484.2635 was similar to that of 3-A HT-2 (theoretical m/z 484.2541) ( Figure S5; lower panel). In this way, 3 proved to be much less efficiently acetylated by Tri101p than its aglycon HT-2. The two modifications, 3-O-acetylation and 4-O-glucosylation, marginally occur simultaneously on the t-type type A trichothecene ring.

Glucosylation of t-Type Type B Trichothecenes by a Tri101-Null Mutant of F. graminearum
To determine whether fusaria can also glucosylate t-type type B trichothecenes oxygenated at C-4, FGD5 and FGD5/101 were fed with NIV and 15-acetylnivalenol (15-ANIV). The mutant strains were also fed with 3-acetyltrichothecenes, 3-acetylnivalenol (3-ANIV), 3,15-diacetylnivalenol (3,, and 3,4,15-triacetylnivalenol (3,4,. When these trichothecenes were fed to FGD5, they were all metabolized to 4,15-diANIV and 4-ANIV, the final products of F. graminearum MAFF 111233 in liquid culture. Hexoside conjugates of NIV derivatives were not detected by LC-MS/MS analysis. In contrast, a very weak activity to conjugate glucose became detectable in the Tri101 disruption mutant. FGD5/101 yielded an extremely limited amount of probable hexose conjugate(s) by each feeding, except for the 3,4,15-triANIV feeding, as described below. 35 mg of acetonitrile extract was obtained. After fractionation by Purif ® -Rp2 (yield 24 mg), followed by repeated purification by preparative HPLC, 12 mg of pure 3 was obtained. NMR analysis confirmed that the structure of 3 was HT-2 toxin-4-O-α-glucopyranoside (HT-2-4-glc) ( Figures 5 and S4, Table S2).   To examine whether 3-acetyltrichothecenes can serve as the substrates of Fusarium UGT(s), 3-A HT-2 was fed to the strains FGD5 and FGD5/101. The XIC peak corresponding to the HT-2-4-glc ammonium adduct ion was detected from the FGD5/101 culture, which could reasonably be explained by the deacetylation of 3-A HT-2 and the subsequent C-4 glucosylation of HT-2. In contrast, 3-acetyl HT-2 toxin-4-α-glucopyranoside was not found in the feeding extract of either mutant strain. We also incubated HT-2 and 3 individually with Tri101p and acetyl CoA to determine whether C-3 can be acetylated when C-4 is glucosylated. Although more than 88% of HT-2 was converted to 3-A HT-2 in 1 h, acetylated 3 could not be detected under the same condition. Only when a high concentration of Tri101p was used and the reaction time was extended to 6 h, a small amount of 3 acetylated at C-3, with a parent ion peak of m/z 646.3076 corresponding to [3-A HT-2hexoside (C30H44O14) + NH4] + (theoretical m/z 646.3069), was barely detectable in MS/MS analysis ( Figure S5; upper panel). The MS/MS spectra of the putative 3-A HT-2 glucoside below an m/z value of 484.2635 was similar to that of 3-A HT-2 (theoretical m/z 484.2541) ( Figure S5; lower panel). In this way, 3 proved to be much less efficiently acetylated by Tri101p than its aglycon HT-2. The two modifications, 3-O-acetylation and 4-O-glucosylation, marginally occur simultaneously on the t-type type A trichothecene ring.
In a previous study, NIV-3-O-β-glucopyranoside (NIV-3-glc) was isolated from NIVcontaminated wheat and its structure was elucidated by NMR [40]. Thus, we applied the NIV-3-glc standard to LC-MS/MS and compared the spectrum of the ammonium adduct ion with that of the NIV-hexoside obtained in our feeding experiments ( Figure S6). The MS/MS spectra were indistinguishable, probably because fragmentation of the parent ion predominantly occurs at the glycoside linkage between NIV and hexose, resulting in the formation of the same major fragments. However, the retention times were slightly different; NIV-hexoside obtained in our feeding assay was eluted at 5.70 min, which is earlier than the NIV-3-glc standard eluted at 5.95 min by the same HPLC system. The result demonstrates that the NIV-hexoside is not the known NIV-3-glc, which supports the above consideration of its structure.

Stability of the Glucoside Conjugates of Trichothecenes
Trichothecene 3-O-glucosides escape detection by the methods validated for trichothecene analysis [28]. Thus, the development of chemical and enzymatic hydrolysis of the modified mycotoxins has been attempted [41]. However, acidic and alkaline conditions were inefficient in cleaving the glucosidic bond of the modified mycotoxins. In addition, except for a specific glucosidase produced by a strain of Bifidobacterium adolescentis, no commercially available enzymes release trichothecene aglycon by hydrolysis [42]. With regard to the 4-O-glucosidic bond of trichothecenes, we also observed a similar stability of the glucoconjugates against a variety of enzymes obtained from Amano Enzyme Inc. (Nagoya, Japan). In this study, the chemical stabilities of 1, 2, and 3 values were studied at various pH.   /z 268.1907) was not detected in the information-dependent acquisition (IDA) mode, demonstrating that the glycosidic linkage in 1 was not cleaved, but the aglycon of 1 was transformed into a new compound at pH 2. In contrast, 2, a C-8 oxygenated derivative of 1, was quite stable. These results suggest that the glycosidic linkage at C-4 of 1 and 2 was stable over a wide pH range but TDmol, aglycon of 1, was rather unstable at a low pH.   (Figure 7c) was similar to that of 4 below an m/z value of 269.1736 (Figure 7b), suggesting that the aglycon of 1 was unstable at pH 2, irrespective of the presence or absence of the sugar moiety. Similar transformation was not observed for TCC, an aglycon of 2.
To elucidate the structure of 4 by NMR, 4 mg of 1 was incubated at pH 2 for 21 d, when the transformation of 1 into 4 was completed. Two mg of highly purified 4 was obtained by purification with preparative HPLC. NMR analysis revealed the aglycon portion and overall structure of 4 to be 2,4,13-trihydroxyapotrichothecene (THA) [43] and THA-4-O-αglucopyranoside (THA-4-glc) (4), respectively ( Figure 8 and Figure S7, Table S3). Thus, it was clarified that the trichothecene skeleton of 1 was opened and recyclized to yield an apotrichothecene skeleton of 4. In short, the trichothecene skeleton and side-chain ester linkages were sensitive to pH changes, as exemplified by the aglycons of 1 and 3. However, cleavage of the C-4 glucosidic linkage does not proceed easily under wide pH ranges.

Cytotoxicity of the Glucoside Conjugates of Trichothecenes
In animal systems, glycoconjugates of toxic compounds produced by phase II reactions are generally less toxic than aglycones. In fact, trichothecene-3-glucosides are less toxic than their aglycones [28,44], although the risk of toxic aglycons being released must be considered. To date, nothing is known about the toxicity of trichothecene-4-glucosides.
By using HL-60, human myeloma leukemia cells, toxicities of the glucosides and aglycons were evaluated ( Figure 9a). As expected, glucoconjugates 1, 2, and 3 showed marginal or much less growth inhibition, while the corresponding aglycones inhibited growth with IC50 of 4.04 ± 0.41, 9.00 ± 0.87, and 0.0183 ± 0.002 µM, respectively. No aglycones were detected in the cell culture medium. The toxicity of apotrichothecene glucoside 4 was marginal even at 10 µM. In short, the trichothecene skeleton and side-chain ester linkages were sensitive to pH changes, as exemplified by the aglycons of 1 and 3. However, cleavage of the C-4 glucosidic linkage does not proceed easily under wide pH ranges.

Cytotoxicity of the Glucoside Conjugates of Trichothecenes
In animal systems, glycoconjugates of toxic compounds produced by phase II reactions are generally less toxic than aglycones. In fact, trichothecene-3-glucosides are less toxic than their aglycones [28,44], although the risk of toxic aglycons being released must be considered. To date, nothing is known about the toxicity of trichothecene-4-glucosides.
Interestingly, 3 was much less toxic than 3-A HT-2, which showed IC50 as low as 0.295 ± 0.26 µM and 3.45 ± 2.31 µM against HL-60 cells and the trichothecene-hypersensitive yeasts, respectively. The higher toxicity of 3-A HT-2 may be attributed to the instability of the 3-O-acetyl group as we previously reported [15]. We also performed a toxicity assay using a highly sensitive yeast bioassay for trichothecene detection using a Saccharomyces cerevisiae triple gene-deletion mutant [45]. As shown in Figure 9b, each glucoconjugate was much less toxic than its aglycone (IC 50 values; TDmol, 3.46 ± 0.30; TCC, 5.01 ± 0.57; HT-2, 0.102 ± 0.020 µM), and all the glucoconjugates were nontoxic at 10 µM.

Possible Role of Trichothecene 4-O-Glucosylation by Fusarium
Interestingly, 3 was much less toxic than 3-A HT-2, which showed IC 50 as low as 0.295 ± 0.26 µM and 3.45 ± 2.31 µM against HL-60 cells and the trichothecene-hypersensitive yeasts, respectively. The higher toxicity of 3-A HT-2 may be attributed to the instability of the 3-O-acetyl group as we previously reported [15].

Possible Role of Trichothecene 4-O-Glucosylation by Fusarium
The 3-O-acetylation of trichothecenes by Tri101p plays a critical role in the self-defense against trichothecene-producing Fusarium. In view of the activity of the Tri101-disruption strain, but not the wild-type strain, to conjugate glucose at C-4 in the liquid culture, the modification could be interpreted as Fusarium possessing a phase II xenobiotic metabolism of 4-O-glucosylation against C-3 unacylated trichothecenes. In mammals, phase II reactions involve conjugation with sulfate, glucuronide, glutathione, and glycine, but reports of glucoside conjugation are limited [46,47]. However, compared with the mammalian phase II reactions, glucoside conjugation is a well-known metabolic pathway for filamentous fungi. For example, Cunninghamella elegans can conjugate glucose to various polycyclic aromatic hydrocarbons after hydroxylation by phase I reactions [48]. Other fungi, including Rhizopus sp. [49], Phanerochaete chrysosporium [50], Pleurotus ostreatus [51], Penicillium canescens [52], Mucor circinelloides [53], Phlebia radiata [54], and Mucor plumbeus [55] are also known to have glucose-conjugation activity against xenobiotics.
The Fusarium phase II glucoconjugation system of trichothecenes demonstrated in this study is similar to the plant detoxification system against exogenously added trichothecenes [56]. However, the fungal and plant mechanisms differ; while the Fusarium system is based on the formation of α-glucosidic linkages mainly at C-4 of d-type trichothecenes oxygenated at C-4, the plant system targets t-type trichothecenes by means of 3-O-glucosylation. In addition, the stereochemistry of glucoconjugation varies among plants. While UGT forms glucoside conjugates with β-glucosidic linkage in rice and barley [27], the enzyme in oats forms glucoside conjugates of T-2 toxin with α-glucosidic linkage [57].
Although 3-acetyltrichothecenes were not conjugated with glucose at C-4 by fungal UGTs, the final trichothecene metabolites of the NIV chemotype cultured on solid substrate are 3-hydroxytrichothecenes that may be conjugated with sugars. Given that a small amount of NIV within the fungal cells undergoes glucoside conjugation under certain conditions, NIV-4-O-glucoside in cereal grains may also need to be monitored to ensure food safety.

Preparation of Trichothecene Substrates for the Feeding Experiment
To prepare 8-deTCN and TCN, strains JCM 8964 and NBRC 31647, respectively, were cultured on YS_60 medium with gyratory shaking at room temperature for 7 d. TDmol and TCC were prepared by treating 8-deTCN and TCN with 2.8% ammonium solution, respectively, at 37 • C for 2 d.
The hyphae of NBRC 113183 were inoculated into YS_60 medium and incubated at room temperature with continuous shaking for 7 d, yielding mixtures of 3-ANIV and 3-ADON.
The conidia of NBRC 9955 were inoculated into the GYEP medium and incubated at 28 • C for 5 d to produce T-2 toxin. HT-2 was obtained by incubating T-2 toxin with the crude cell extract of Bacillus spp. LC466619, containing the deacetylase specific to C-4. For production of 3-A HT-2, rTri101p was used to acetylate T-2 toxin and HT-2, respectively, as previously described [34]: the reaction mixture contained approximately 2.0 µM rTri101p, 100 µmol acetyl CoA trilithium salt, and 10 mg each of T-2 toxin (21.5 µmol) or HT-2 (23.6 µmol) in 20 mL of 50 mM Tris-HCl buffer (pH 7.5). The reaction was performed at 30 • C for 3 h.

Extraction and Purification of Trichothecenes
The transformation of each toxin in the feeding experiment was confirmed using TLC. An aliquot of each fungal culture was extracted with an equal volume of ethyl acetate. For the extraction of hydrophilic sugar conjugates, an equal volume of acetonitrile was added, followed by the addition of a small amount of 5 M NaCl aqueous solution to separate the organic phase. The extract was dried, dissolved in ethanol, and subjected to TLC. The TLC plates were developed with ethyl acetate/toluene (3:1) or ethyl acetate/toluene/hexane (2:2:1). The toxins were visualized using the NBP/TEPA method [64,65].
For the large-scale extraction of hydrophobic trichothecenes, an equal volume of ethyl acetate was added to the culture and the extraction was repeated. With regard to the extraction of the less hydrophobic NIV transformed from 4-ANIV, an equal volume of acetonitrile was added to the solution following the evaporation of the ammonium in the reaction mixture by the continuous flow of N 2 . A small amount of 5 M NaCl (aq) was then added to separate the solutions into two phases.
The organic phase was concentrated using a rotary evaporator or Smart Evaporator (BioChromato, Inc., Kanagawa, Japan) and the condensate was dissolved in ethyl acetate and applied to Purif-Rp2 equipped with Purif-Pack SI 25 (Shoko Scientific, Kanagawa, Japan). A portion of each fraction was applied on TLC, and the fractions containing each trichothecene were collected and concentrated again. The condensate was dissolved in ethanol and filtered through a 0.22 µm polytetrafluoroethylene (PTFE) syringe filter (ANPEL Laboratory Technologies Inc., Shanghai, China). The filtrate was applied to preparative HPLC (LC-4000 series, JASCO, Corp., Tokyo, Japan; UV detection at 254 nm or 195 nm) equipped with a C 18 column (Pegasil ODS SP100 10ϕ × 250 mm; Senshu Scientific Co., Ltd., Tokyo, Japan). Trichothecenes were eluted in a mobile phase of H 2 O-acetonitrile using the optimized gradient mode for individual toxins at 40 • C at a flow rate of 3 mL/min. The purification step using preparative HPLC was repeated until the purity of the target compound was >98%.

HPLC and LC-MS/MS Analysis
Each purified trichothecene was applied to analytical HPLC (LC-2000 plus series; JASCO, Corp., Tokyo, Japan; UV detection at 254 nm or 195 nm) equipped with a C 18 column (Pegasil ODS SP100 4.6ϕ × 250 mm). Trichothecenes were eluted in a mobile phase of H 2 O-acetonitrile using an appropriate gradient mode at 40 • C at a flow rate of 1 mL/min. The concentration of each trichothecene was calculated from the corresponding peak area based on the standard curve, if available. If the standard curve was not available, each purified toxin was dried completely using a freeze dryer (TITEC Corp., Saitama, Japan) and weighted. In some cases, quantitative NMR was performed, as previously described [66].
For LC-MS/MS analysis, samples were separated using an Eksigent ekspert™ ultraLC 100-XL system (Dublin, CA, USA) with a C 18 reverse phase column (PEGASIL ODS SP100-3; 2ϕ × 100 mm) according to the following steps: a linear gradient of 10-95% of acetonitrile in 5 min in 0.1% ammonium formate at a flow rate of 0.3 mL/min. The ultraLC was connected to an AB SCIEX Triple TOF 4600 system (Framingham, MA, USA), with a DuoSpray source operated in electrospray ionization mode. The IDA and/or time-of-flight (TOF)-MS methods were used to obtain MS/MS spectra in the positive ion mode. Data were analyzed using the PeakView software version 1.2.0.3 (AB SCIEX).

Feeding Experiments and Purification of Trichothecene Metabolites
For feeding experiments of the d-type trichothecenes, conidia of Fs∆Tri5 were preincubated in GYEP medium at 28 • C with reciprocal shaking; mycelial plugs of FGD5 were incubated in RF medium at 20-22 • C with gyratory shaking. After 2-4 d of incubation, the freshly prepared hyphae were washed in water, and the Fs∆Tri5 and FGD5 hyphae were added to 50 mL of GYEP and RF media, respectively. One milligram of each d-type trichothecene (TDmol, TCC, 8-deTCN, and TCN in 100 µL ethanol) was immediately added and incubated with shaking. In parallel, the vehicle (ethanol) and 1 mg ITDmol in 100 µL ethanol were fed as the negative and positive controls, respectively. Two aliquots of 2 mL medium were taken at 0, 3, 6, 24, and 48 h; one aliquot was extracted with ethyl acetate for TLC analysis, and another was extracted with acetonitrile and NaCl (aq) for LC-MS/MS analysis. The developing solvents for TLC of metabolites of Fs∆Tri5 and FGD5 were ethyl acetate:toluene:hexane (2:2:1) and ethyl acetate:toluene (3:1), respectively.
For feeding experiments of the t-type type A trichothecenes, the hyphae of F. graminearum FGD5 and FGD5/101 were prepared as described above, and 1 mg each of HT-2 and 3-A HT-2, respectively, was fed to 50 mL of RF medium containing freshly prepared hyphae in each flask. The strains FGD5 and FGD5/101, used for the feeding experiments, were confirmed to be metabolically active. FGD5 rapidly transformed ITDmol into NIV-type trichothecenes [34], and FGD5/101 similarly transformed TDmol into TDmol-4-glc [33]. The time course experiment was performed in a manner similar to that of the d-type trichothecene feedings. The extracts were analyzed by TLC and LC-MS.
For feeding experiments of the t-type type B trichothecenes, the hyphae were prepared in the same way as described above. The substrates fed to the hyphae were as follows: NIV, 3-ANIV, 15-ANIV, 3,15-diANIV, and 3,4,15-triANIV. The hyphae were incubated with 1 mg of each substrate in 50 mL of RF medium and each culture medium was sampled on days 0, 2, 5, 7, and 14. The extracts were applied to TLC and LC-MS.
To prepare sufficient amounts of trichothecene glucoside conjugates for NMR analysis, we performed large-scale feeding experiments. For the production of 1 and 2, 8 mg each of TDmol and TCC was added to 200 mL of GYEP media in 500 mL Erlenmeyer flasks, respectively, and freshly prepared Fs∆Tri5 hyphae were inoculated onto the media. The hyphae were incubated at 28 • C for 2 d with vigorous shaking. A total of 2 L and 1.8 L culture medium were collected for purification 1 and 2, respectively. For the production of 3, 4 mg of HT-2 was added to 200 mL of RF media in 500 mL flasks and freshly prepared FGD5/101 hyphae were inoculated. The hyphae were incubated at 20-22 • C for 16 d with vigorous shaking. A total of 15 L of culture medium was collected for the production of 3.
For the extraction and purification of the glucoside conjugates 1, 2, and 3, acetonitrile equal in volume to that of each culture medium was added, filtered through filter paper (Toyo Roshi Kaisha, LTD., Tokyo, Japan), and mixed with a small amount of 5 M NaCl aq to separate into two phases. The aqueous phase was extracted again with acetonitrile, and the organic phase was collected and concentrated. The condensate of each glucoside conjugate was dissolved in ethanol and applied to Purif-Rp2 equipped with Purif ® -Pack ODS-25 size 60. The trichothecene metabolites were eluted with a mobile phase of acetonitrile and water in an appropriate gradient mode and the fractions containing each target conjugate were concentrated.
The condensate was dissolved in ethanol and applied to a preparative HPLC (UV detection at 195 nm). This purification step was repeated three times for 1 and 2, and five times for 3, following the procedure described in Section 3.5. Each purified conjugate (1, 2, and 3) was dried completely using a freeze dryer (TITEC Corp., Saitama, Japan), weighed, and subjected to NMR spectroscopy.

In Vitro Trichothecene 3-O-Acetyltransferase (Tri101p) Activity Assay
To evaluate the 3-O-acetylation activity of Trip101p against HT-2 and its glucoconjugate, 20 µg of HT-2 and 27.6 µg of HT-2-4-glc (each being 47.1 nmol) were incubated with acetyl CoA (200 nmol) and rTri101p (0.2 and 2.0 µM) in 10 mM Tris-HCl buffer (pH 7.5) at 30 • C for 1 and 6 h. rTri101p inactivated in advance by boiling was used as a negative control. The product was extracted with acetonitrile by the addition of 5 M NaCl aq, dissolved in ethanol, and filtered through a 0.22 µm PTFE syringe filter. Each filtrate was subjected to LC-MS/MS analysis.

Stability of the Glucoside Conjugates of Trichothecenes and Trichothecene Aglycons
The pH of each solution was adjusted to pH 2 to pH 13 at every pH unit by adding HCl or ammonium solution, and 5 µg of each glucoside conjugate (1, 2, and 3) was added to 300 µL of each solution. After incubation at 37 • C for 3 d, the reaction mixture was diluted 10 times with ethanol. Two microliters of each sample was applied to LC-MS/MS, and data were acquired in IDA mode. To investigate the acid stability of the aglycon TDmol, we also added 5 µg of each trichothecene (TDmol and TCC) into 300 µL of solution (pH 2), and incubated at 37 • C for 3 d. The solution was diluted 10 times with ethanol, and 2 µL each was applied to LC-MS/MS. The data were acquired in the TOF/MS mode.
To prepare a sufficient amount of apotrichothecene glucoside conjugate (4) for NMR analysis, we added 4 mg of 1 to 16 mL of acetic acid solution (pH 2.0) and incubated it at 37 • C for 3 weeks. Under the condition, all of the compound 1 disappeared and transformed into compound 4. The reactant was concentrated using a rotary evaporator, dissolved in 50% ethanol, and filtered through a 0.22 µm PTFE syringe filter. The filtrate was subjected to a preparative HPLC, as described in Section 3.5. Purified compound 4 was dried completely using a freeze dryer, weighed, and subjected to NMR spectroscopy.

NMR
All NMR spectra were recorded on a JEOL JNM-ECX500 (500 MHz) spectrometer in deuterated methanol (MeOH-d 3 ) calibrated with a solvent peak at 3.31 ppm for the measurement of H nucleus and 49.3 ppm for that of C nucleus. Purified trichothecenes conjugates were identified by some analyses of 1 H NMR, 13 C NMR, correlation spectroscopy (COSY), heteronuclear multiple quantum coherence (HMQC), and heteronuclear multiple bond coherence (HMBC). If necessary, total correlation spectroscopy (TOCSY) or 1D-nuclear overhauser effect (1D-NOE) analysis was performed.
For the cytotoxic assay, cells in the logarithmic growth phase were prepared. To each well of a 96-well microtiter plate, 95 µL of culture cells (2.5 × 10 4 /well) were seeded. These experiments were performed in triplicate. For a highly sensitive yeast bioassay for trichothecenes, the triple null mutant strain, pdr5∆ erg6∆ rpb4∆ of S. cerevisiae BY4742 (MATα his3∆1 leu2∆0 lys2∆0 ura3∆0, Horizon Discovery Ltd., Cambridge, UK) was preincubated in 5 mL of YPD medium (1% yeast extract, 2% peptone, and 2% glucose) for 18-48 h with shaking at 30 • C. The OD 620 of the cell culture was adjusted to 0.1 in YPD medium containing 0.003% sodium dodecyl sulfate (SDS) and 196 µL of cell culture was transferred to each well of the microtiter plate. Thereafter, 4 µL of inhibitor solution (10% ethanol as a vehicle) was added to each well and the yeast cells were incubated at 30 The OD 620 for trichothecene-treated and vehicle-treated cultures was measured using a Multiskan FC Microplate Photometer and the value was subtracted from the corresponding OD 620 at a time point of zero. The relative growth rate was calculated as the OD 620 ratio of the trichothecene-treated samples to the untreated controls. The percentage of inhibition of growth was calculated as: where ∆A 620 is the average change in absorbance of each well (end point A 620 and start point A 620 ). These experiments were performed in duplicate.

Statistical Analysis
The log-logistic model was applied using R version 3.5.0 (R project for statistical computing) for the calculation and statistical analysis of IC 50 values.

Conclusions
The trichothecene-producing Fusarium species proved to have a fungal phase II reaction against exogenously added trichothecenes. Although the 4-O-glucosylation was limited in comparison to the major resistance mechanism of 3-O-acetylation, the modification significantly reduced the toxicity compared to the corresponding aglycons. Therefore, 4-O-glucosylation may serve as a phase II xenobiotic metabolism against 4hydroxytrichothecenes for t-type trichothecene producers.