Trichothecenes are a group of mycotoxins produced by a variety of different Fusarium species, such as F. culmorum, F. equiseti, F. graminearum, F. moniliforme, F. sporotrichioides, and F. proliferatum. In particular, F. graminearum, a plant pathogen of gramineous plants, especially wheat, is known to produce deoxynivalenol (DON), nivalenol (NIV), and ZEN. It is the most widely distributed toxinogenic Fusarium species (Miller & Trenholm, 1997).
DON, the most important type B trichothecene, occurs predominantly in grains such as wheat, barley, and maize and less often in oats, rice, rye, sorghum, and triticale. After short-term administration, DON induces decrease in feed consumption (anorexia) and emesis (vomiting in animals). DON has been detected not only in cereals (wheat, barley, rice, rye, sorghum, oats), but also in some processed food products in Europe and North America. In addition, it has been reported in barley products, such as malt and beer (Hussein & Brasel, 2001; Creppy, 2002). Under natural conditions, DON is found in wheat and barley in association with several strictly related compounds, such as 3-acetyl-deoxynivalenol (3-ADON), 15-acetyl-deoxynivalenol (15-ADON), 3,4-diacetyl-nivalenol (3,4-DA-NIV), and 3,15-diacetyl-deoxynivalenol (3,15-DA-DON). The same situation occurs for T-2 toxin (T2) and diacetoxyscirpenol (DAS), which are often found together with closely related toxins contributing to the overall toxicity. Generally, the same Fusarium strains are also producers of ZEN, a mycoestrogen, which occurs along with DON in maize, barley, and wheat (see below).
The most common methods for the determination of trichothecenes involve the use of a gas-chromatographic system with an FID, ECD or mass spectrometric detector or an HPLC system with UV or fluorescence detection (after derivatization of the molecules). The GC methods allow to get detection limits in the range 1–10 ppb for type A and 10–30 ppb for type B trichothecenes, whereas the LODs obtained by HPLC-UV methods are higher (100–200 ppb). Likely, the HPLC would be better suited for the trichothecene determination from a chromatographic point of view, but the lower sensitivities obtained by using UV detectors have always hampered the spreading of the technique. Since trichothecenes are not fluorescent, the LC/FLD can be used only after derivatization of the trichothecene with suitable fluorophores (Dall'Asta et al., 2004b). Therefore, the use of GC for trichothecene analysis is due to the necessity to reach sensitivities in the range of low ppbs, although it requires derivatization steps prior to the analysis, introducing points of uncertainty. Samples can be readily confirmed by GC/MS. As far as HPLC is concerned, MS detection is essential to selectively detect underivatized trichothecenes at a fair sensitivity (Langseth & Rundberget, 1998).
A. Gas Chromatography/Mass Spectrometry Determination of Trichothecenes
GC analysis is mostly used for the determination of type A trichothecenes, since they are not fluorescent and do not strongly adsorb in the UV-Vis range. GC is also widely employed for the determination of type B trichothecenes, although the conjugation of the carbonyl group with the double bond makes them more suited for UV detection and therefore HPLC analysis. The GC determination is mostly based on derivatization of the hydroxyl groups by trimethylsilylation or fluoroacylation to trifluoroacetyl (TF), pentafluoropropionyl (PFP), or heptafluorobutiryl (HFB) derivatives. FID detectors are commonly utilized or alternatively, ECD for the fluoroacyl derivatives. Although several GC/MS methods have been proposed as alternatives to GC/ECD or GC/FID, none seems to allow for a good quantitative determination. At the moment, no isotope dilution approach has been developed. Moreover, only few GC/MS methods allow the simultaneous detection of type A and type B trichothecenes. The GC/MS approach seems to give the best results for confirmatory analyses and also for obtaining structural information on several derivatives of the main trichothecenes, which is fundamental for toxicological and metabolic studies. The widespread possibility of using GC/MS for the development of mass spectral databases inspired some recent studies.
A GC/MS method used for the analysis of DON and derivatives (3-ADON, 15-ADON, NIV, 7-ADON) in wheat and barley, described by Mirocha et al. (1998), is based on selected ion monitoring (SIM) detection of the mycotoxin as trimethylsilyl-ethers (TMS). The analysis involved electron impact ionization (EI) of the derivatives and monitoring of the diagnostic ions (from four to seven according to the different toxins). The sensitivity varied according to the column, the conditions of the filament, as well as the instrument tuning. Moreover, ionization turned out to be matrix dependent, so that internal calibration was often required, in particular with complex matrices such as feeds. Quantification was performed by external calibration with good precision (around 5% SD on mean values) and good recovery (around 90%), although LOD and LOQ were not reported.
Onjy et al. (1998) suggested that the derivatization step of trichothecenes before the GC/MS analysis might cause a lack of structural informations for the identification of compounds. As a possible approach, they proposed a direct analytical method to determine the native mycotoxins extracted from several cereals without any derivatization procedure, using an improved cold on-column injection equipped with a laboratory-made inlet liner directly connected to the GC capillary column. Prior to GC/MS analysis, sample clean-up procedures (zinc acetate treatment and Florisil columns) for separation of toxins from fats are indispensable. The chromatographic separation, performed for DON, 3-ADON, FUSX, DAS, scirpentriol (SCT), 15-monoacetoxyscirpenol (15-MAS), T-2, and ZEN, allowed to achieve good peak resolutions, with the EI spectra showing the molecular ions of the analytes and the specific fragmentation for each toxin. However, the direct analysis of NIV was not possible, because of the degradation of this compound into several fragments after injection. Moreover, although average recovery was 93%, two toxins (FUSX and SCT) showed a recovery of 49% and 42%, respectively, again for likely degradation of these compounds. The LODs of the method, measured for barley sample extracts, ranged from 100 to 500 ppb using an SIM mode.
A GC/MS method for the determination of eight trichothecenes of type A and B developed by Schoellenberg et al. (1998) involved the use of an ion trap mass spectrometer operating in chemical ionization mode, using isobutane as reactant gas. Also in this case, it was necessary to apply different clean-up steps to remove the co-extracted matrix components. The authors suggested that the critical point of the whole procedure, which was applied to complex matrices including heavily molded cereals, cereal foods, and feed, was the derivatization step to obtain TF derivatives, so that an internal standard was added to control the reaction. Since the ion trap system allowed to obtain the typical fragmentation for each toxin, the quantification was based both on the main fragment ion and on a secondary characteristic ion, which allowed to correctly identify the analyte. The reproducibility of the fragmentation was good, with differences between the two selected ions below 10%. The LODs and LOQs of the whole method, measured by injection of standard solutions, ranged from 2 to 12 ppb and from 6 to 38 ppb, respectively. The highest detection limit was obtained for NIV, consistently with the lower ionization properties of this compound.
Quite often GC/MS methods allow the simultaneous detection of trichothecenes and ZEN, which is also produced by Fusarium molds (see below). Recently, Tanaka et al. (2000) proposed a GC/MS method, essentially analog to that used by Mirocha (see above), for the simultaneous detection of seven trichothecenes (DON, NIV, FUSX, 3-ADON, T-2, neosolaniol (NEO), DAS), and ZEN as trimethylsilyl derivatives by single ion monitoring (SIM) of the main ion of each compound. The authors decided to detect trichothecenes as TMS derivatives according to the mass range limitation (600–700 m/z) of their bench-top type GC/MS, obtaining detection limits in cereals ranging from 5 to 10 ppb. The recovery ranged from 81% to 94% for the different toxins in the different matrices.
Although not intended for food analysis, Nielsen & Thrane (2001) proposed a useful GC/MS/MS data bank for the characteristic fragmentation of 21 trichothecenes from Fusarium and Stachybotrys fungi as PFP derivatives. The use of PFP anhydride as derivatizing agent allowed to achieve the highest sensitivity by GC/MS analysis, if a mass spectrometer with a range higher than 800 amu is available. The disadvantage of this reagent, as mentioned by the authors, was the isomerization of type B trichothecenes during the reaction. Moreover, the derivatization reaction itself was also very slow. The method used an ion trap detector for MS/MS analysis. Indeed, MS/MS spectra were much more reproducible than simple MS spectra, where a concentration dependence of the intensity was observed, a typical problem of the instrumentation utilized. In general, trichothecenes were detected by using electron impact (EI) ionization in a positive ion mode, although in some cases, a chemical ionization (CI) in the negative ion mode with methane as ionizing gas was used for peak confirmation.
A database obtained by GC/MS analysis (EI ionization) of the TMS-derivatized trichothecenes was also recently proposed (Rodrigues-Fo et al., 2002). The authors studied the spectral behavior of DON and several DON-related compounds isolated from moulds (NIV, 7-ADON, 15-ADON, 3-ADON, 3-lactyl-DON, 4,15-DA-ADON, 3-ANIV). The 3-acetyl-nivalenol was identified for the first time. Structures of the main ions at m/z 103, 117, 147, and 191 were elucidated by high resolution MS (HRMS) experiments and a general fragmentation scheme for the DON-derivatives was suggested. The relative abundances of the assigned ions in the mass spectra of trichothecenes allowed a fast structural investigation in the analysis of complex matrices and could be used for database identification of the investigated compounds.
A summary of the main recent GC/MS methods for trichothecene determination is reported in Table 5. The last two methods are not reported in this table since they are intended neither for food nor for feed.
Table 5. Main recent GC/MS methods for trichothecene determination
aRecovery range: 42–108%.
bCold on-column injection.
cExtracted chromatograms corresponding to the molecular ions and to the main fragments were used for the quantification.
B. Liquid Chromatography/Mass Spectrometry Determination of Trichothecenes
LC/MS analysis is becoming the method of choice for the determination of trichothecenes, which do not contain characteristic chromophores or fluorophores, thus avoiding derivatization steps and achieving more robust and reproducible results. Moreover, the LC/MS analysis also allows to reduce the sample preparation step: different studies showed that the direct analysis of the sample extract can be performed without the need of the immunoaffinity purification. The injection of raw extracts requires, however, an extremely careful calibration: it was widely demonstrated that the matrix effect on ionization is particularly significant for trichothecenes. Thus, the most important limitation, at the moment, is the difficulty to find a proper internal standard, being the isotope-labeled compounds not commercially available or very expensive.
Another important area of research concerns the development of multiresidual methods, able to detect a great number of trichothecene toxins in many different matrices: in this case, not only the ionization but also the extraction method needs to be carefully studied and completely validated. Quite interestingly, almost all published methods make use of the less common APCI interface, rather than of the more common ESI. Several reasons probably justify this preference: the ionization yield in the ESI interface is likely to be more affected from matrix effects, it usually does not allow the simultaneous detection of type A and type B trichothecenes with the same ionization mode and generally both classes of toxins show a poor ESI response.
An interesting LC/MS method based on ion trap MS for the determination of nine trichothecenes in wheat using LC-APCI-MSn in positive ionization mode was described by Berger, Oehme, & Kuhn (1999). Clean-up procedures with Mycosep columns were performed allowing for fair recovery (≥80%). The method allowed the separation of the trichothecenes (NIV, DON, FUSX, NEO, 3-ADON, 15-ADON, DAS, HT-2, and T-2) in less than 12 min using a MRM detection mode, with LODs of 10 ppb for NIV and 15-ADON, 6 ppb for DON, 4 ppb for NEO, 3 ppb for FUSX, 3-ADON, and T-2, and 1 ppb for DAS and HT-2. The method did not allow for the separation of the two acetylated DON-derivatives, but they could be distinguished by the specific fragmentation performed in an ion trap mass spectrometer. The quantification was based on the addition of an internal standard (verrucarol, a semisynthethic trichothecene) to the sample before the extraction and was carried out using the [M + H]+ ions or the main fragment ions. Moreover, the authors presented a scheme to unequivocally identify trichothecenes by online MS/MS and ion adduct formation with ammonium acetate. For the ion adduct formation experiments, ammonium acetate was added connecting a T-tube between the HPLC column and the ionization chamber. The addition of ammonium acetate resulted in increased sensitivity for the type A trichothecenes as [M + NH4]+ adducts and for type B toxins as [M + CH3COO]− ions. MS2–MS6 experiments were performed to study the fragmentation in the ion trap. Compared to the type A trichothecenes, type B easily lost the side chain C15, likely resulting in a quinoic structure followed by the rearrangement into a stable aromatic alcohol. Online HPLC/MS/MS experiments provided information about the substituents present. Since the hydroxyl groups were cleaved off as water and the ester as their corresponding acids, every lost substituent generated a double bond equivalent. Type B trichothecenes with three or four substituents usually generated fragments of 200 m/z, while a higher energy was required to break the 3-ring backbone generating a fingerprint profile typical of both type A and type B trichothecenes. The LODs and the LOQs of the whole method ranged from 1 to 10 ppb and from 10 to 100 ppb, respectively.
An isocratic HPLC method combined with negative APCI-MS detection for the determination of NIV and DON in wheat was published by Razzazi-Fazeli, Bohm, & Luf (1999). The instrument employed was a single quadrupole mass spectrometer and the detection of each compound was based on the molecular ion. All the ionization parameters were optimized for the analysis, to obtain the best sensitivity. The best flow was found to be 1 mL/min, without splitting. The effect of the nebulizing temperature on the sensitivity and on the fragmentation pattern was studied, showing the need of high temperatures (>300°C) to allow high ionization yields. The cone voltage was found to have a strong effect on fragmentation: the intensity of the deprotonated ion [M−H]− and of the adducts decreased by increasing the cone voltage to values higher than 20 V. For both molecules, by using a vaporizer temperature of 400°C, the loss of 30 mass units, together with other fragments, was observed, which was explained by the authors as the cleavage of the epoxy group yielding fragments [M−H-CH2O]− at m/z 265 (DON) and 281 (NIV). The authors also studied the effect of buffer addition on the ionization of DON and NIV. In particular, a decrease in the signal-to-noise ratio was observed by the use of modifiers to the mobile phase such as ammonium acetate or acetic acid, whereas addition of ammonia resulted in no visible effect. With the optimized method and using a SIM mode of acquisition on the deprotonated molecular ions, LOQs (LOD not reported) of 40 ppb for DON and of 50 ppb for NIV (in matrix) with fair recoveries (70–86%) were obtained.
More recently, the same group (Razzazi-Fazeli et al., 2003) extended the proposed negative ion mode APCI-MS method to the detection of the acetylated derivatives of DON (3-ADON and 15-ADON) and to the de-epoxy metabolite of DON called DOM-1. This metabolite is formed from DON by gut and rumen bacteria through reduction of the epoxide group of DON, and it was found in some mammalian tissues and in cow's milk. DOM-1 was proposed as a marker for the exposition to DON in animals. The type B trichothecenes separation involved the use of a linear gradient from H2O-CH3CN-CH3OH 82:9:9 to H2O-CH3CN 40:60 as mobile phase and a Polar-RP standard bore column (150 mm × 4.6 mm, 4 μm). The quantification was performed using the [M−H]− ion, while the identification was based on the main fragments. This is the first method, which allows the separation of 3-ADON and 15-ADON without changing the flow rate, by choosing a more polar column than the commonly used C18. The detection limits for type B trichothecenes were 70 ppb for NIV, 50 ppb for DON, 3-ADON, FUS-X, and 150 ppb for 15-ADON.
The same authors also published a study for the LC/APCI-MS determination of type A trichothecenes T-2, HT-2, Acetyl-T-2, DAS, 15-acetoxyscirpenol (15-AS), and NEO (Razzazi-Fazeli et al., 2002). A C18 narrowbore column (200 mm × 2.1 mm, 5 μm) and a linear gradient consisting of 1 mM aqueous ammonium acetate–acetonitrile mobile phase was used. The MS experiment was performed in the APCI positive ion mode. The quantitative determination of each compound was based on the molecular protonated ions, on the ammonium adducts, or on the main fragments. In contrast to type B trichothecenes, the response in positive ion mode for all the investigated type A toxins was found to be distinctly more sensitive than in the negative ion mode. The mass spectra of the compounds often showed a [M + NH4]+ adduct more intense than the [M + H]+ ion. The loss of the acetyl group could be observed for all the tested molecules, while the formation of very intense ammonium adducts was typical of those with ester functions at the C15 or C4 positions. Although optimal flow was found to be around 1 mL/min, the use of a narrow bore column required a flow of 0.3 mL/min. LODs were not reported, but LOQs in matrix ranged from 50 to 85 ppb. As internal standard deuterated T2 toxin (D3-T2) was used. Nevertheless, the precision was nevertheless not very high (up to 25% of interday variability).
As a good example of methods using an LC/ESI-MS, Plattner (1999) reported the analysis of DON in food matrices in the negative ion mode with a single quadrupole analyzer. The sensitivity was typically low (LOD >2 ppm in matrices without clean-up treatment), although comparable with UV detection. However, the author stated that the ESI interface was more rugged and forgiving than the APCI when dirty samples were analyzed and thus more suited for routine determinations.
In a very recent study, the analysis of four type B trichothecenes (NIV, DON, FUSX, 3-ADON) was performed in maize by an LC/MS/MS technique, comparing the commonly used Atmospheric Pressure Chemical Ionization (APCI) interface to the TurboIonSpray (TISP) interface in both negative and positive ion modes (Laganà et al., 2003). The most valuable spectra for each trichothecene were obtained using the turbo ion spray TISP-ESI system in the negative ion mode, gradient elution from water/methanol/acetonitrile 90:3:7 to water/methanol/acetonitrile 20:24:56, with no modifier. The APCI responses were definitely lower than those obtained for ESI. Moreover, the ESI interface was again found to be more rugged, maintaining sensitivity after many injections of dirty samples and requiring less maintenance than the APCI interface. In the negative ion mode, the type B trichothecenes showed a very specific fragmentation, although the [M−H]− ion was the base peak for all the tested compounds. The MRM quantification was based on the [M−30]− ion, assigned by the authors to the cleavage of the epoxy group. The best results were found by TISP in the negative ion mode and in the MRM mode for the quantification, obtaining detection limits ranging between 1.5 and 10 ppb.
Very recently, we succeeded in developing a method for the simultaneous determination of several type A (NEO, DAS, T-2, HT-2) and several type B trichothecenes (DON, 15-Ac-DON, NIV, FUSX) at good sensitivity with an ESI interface and a single quadrupole analyzer. The determination was carried out for all trichothecenes in the positive ion mode by using a water/methanol 35:65 mixture as isocratic eluent. To enhance the naturally poor response of the trichothecenes in the ESI interface, we started from the observation that this class of molecules forms very stable adducts with the sodium ion. The addition of sodium chloride to the eluent (100 μM) not only simplified the spectra, which only showed the sodiated molecular ions, but also enhanced the response for all trichothecenes of about one order of magnitude. The LODs obtained with this method by monitoring the sodiated molecular ion by a SIM acquisition ranged from 20 to 50 ppb, as determined in matrices, extracted, and analyzed without any clean-up treatment. The method was tested by determining DON in a certified reference material, giving very consistent results (Dall'Asta et al., 2004c).
A summary of the main recent LC/MS methods for trichothecene determination is reported in Table 6.
Table 6. Main recent LC/MS methods for trichothecene determination
aThe LOD without Mycosep clean-up was 2 ppm.
bNaCl was added as cationization agent to the HPLC eluent.