Fusarium mycotoxins in total mixed rations for dairy cows

Mycotoxins produced by certain fungal species of the Fusarium genus are frequently found as contaminants in cereals and feedstuffs. Fumonisins (FBs), deoxynivalenol (DON) and zearalenone (ZEN) are of special concern relative to animal health and productivity. The aim of this work was to analyse the levels of Fusarium mycotoxin contamination in samples of total mixed rations (TMRs) for dairy cows. To accomplish this analysis, an HPLC-MS/MS multi-mycotoxin method was developed and validated. The relation between the formulation of TMR samples and the presence of mycotoxins was also studied. From February 2016 to January 2018, a total of 193 TMR samples for dairy cows collected from farms located in different areas of Spain were analysed for the presence of FBs, ZEN, DON and their metabolites. In total, 112 samples (58%) were contaminated with at least one mycotoxin, and 38 samples (20%) presented more than one mycotoxin. FBs were the mycotoxins most frequently found (34% positive samples). DON was detected in 17% of samples, and ZEN was detected in 16% of samples. Among the metabolites analysed, only deoxynivalenol-3-glucoside (DON-3-Glc) and 15-acetyldeoxynivalenol (15-ADON) were detected. The levels of all the Fusarium mycotoxins studied were always below the values recommended by the European Commission for feedstuffs. The wide variety of ingredients used in the formulation of the analysed samples made it difficult to reach definite conclusions, although it seemed that some cereal silages and concentrates such as cereals or compound feed used as ingredients of the TMR may be related to the presence of mycotoxins.


Introduction
Fusarium mycotoxins, which are usually synthesized before harvest, are frequently found as contaminants in cereals and feeds (Driehuis et al. 2008a). Among these Fusarium mycotoxins, tricihothecenes, zearalenone (ZEN) and fumonisins (FBs) are of special concern regarding animal health and productivity (Streit et al. 2012). Ruminants are considered to be relatively resistant to Fusarium mycotoxins since these compounds are degraded by rumen microorganisms into less toxic metabolites (Pleadin et al. 2017). Moreover, the carryover of these mycotoxins into milk has been reported to be low, although further research is needed (EFSA 2004a;EFSA 2005;Prelusky et al. 1987).
Deoxynivalenol (DON) is a type B trichothecene that is mainly produced by F. culmorum and F. graminearum. This mycotoxin inhibits protein synthesis and has cytotoxic and immunotoxic effects (Pestka 2007). Some studies have reported that this compound may affect the milk yield (EFSA 2017). ZEN is primarily produced by F. graminearum and is able to bind to oestrogen receptors as its chemical structure is similar to oestrogens, causing alterations in reproductive organs (CAST 2003). FBs are synthesized by F. proliferatum and F. verticillioides (Gallo et al. 2015), have hepatotoxic effects and can affect the production of milk (Fink-Gremmels 2008). The EU has recommended guidance values for DON, ZEN and FBs in products intended for animal feeding (EC 2006). Moreover, the European Food Safety Authority (EFSA) has published several opinions related to the presence of these mycotoxins in food and feed (EFSA 2004a;EFSA 2004b;EFSA 2005). Ruminants are likely to be exposed to a wide range of mycotoxins due to the heterogeneous composition of feeds. The total mixed ration (TMR) is a method of feeding dairy cattle based on the use of different known proportions of forages, commodities or by-products (e.g., cottonseeds), grains and protein supplements, minerals and vitamins. Some of these materials are frequently contaminated by different Fusarium species; thus, they are potential sources of DON, ZEN and FBs, among other mycotoxins. FB 1 and FB 2 are mainly present in maize (Tevell Åberg et al. 2013), and ZEN and DON are frequently found as contaminants in maize, wheat, triticale and oats (Döll and Dänicke 2011). DON can also be present in forages, especially in maize forage (Whitlow and Hagler Jr 2005;Driehuis et al. 2008b;Ogunade et al. 2018). Consequently, mycotoxins usually co-occur in animal feed, resulting in toxic effects that could be antagonistic, additive or synergistic (Speijers and Speijers 2004).
Due to the complexity of feed matrices and the low level of contamination usually found in these samples, it is necessary to develop reliable analytical methods that enable the simultaneous detection of small amounts of different mycotoxins with high sensitivity. In recent years, high-performance liquid chromatography (HPLC) coupled to mass spectrometry (MS) has become the most popular tool for multiple mycotoxin detection and quantification in food and feed due to its high specificity and selectivity (Zhou et al. 2016). As TMR is a mixture of different kinds of crops and materials, sample preparation and clean-up procedures are critical steps in the development of analytical methods, since it is necessary to minimize the matrix interferences as much as possible.
Several studies have been conducted to analyse the occurrence of mycotoxins in different feed ingredients, including in cereals and silages (Bottalico 1998;Placinta et al. 1999;Rasmussen et al. 2010;Pereira et al. 2014;Gallo et al. 2015;McElhinney et al. 2016;Pleadin et al. 2017;Panasiuk et al. 2019). However, very little research has been carried out on the incidence of mycotoxins in TMR intended for dairy cows (Pereyra et al. 2012;Yalçin et al. 2016;Cogan et al. 2017).
The aim of the current study was to determine the frequency and levels of Fusarium toxins in TMR samples used as dairy cattle feedstuffs collected over 2 years. For this purpose, an easy and reliable method for the simultaneous detection of various mycotoxins by HPLC-MS/MS was developed and validated. The possible relation between the ingredients used for the formulation of TMR samples and the presence of the mycotoxins analysed was also assessed.

Sampling
Samples of 193 different feedstuffs for dairy cows were collected from different areas of Spain (Cantabria, Castilla-León, Cataluña and Galicia) in the period from February 2016 to January 2018. TMRs were formulated by the farmers at each farm at the beginning of the day using a "unifeed system". Each farm provided a bulk sample of 4-5 kg of TMR generated by pooling elementary samples of 100 g that were taken from the daily batch as it was delivered by specialized technicians, as a representative fraction of the whole. Farmers were also asked to provide a report that contained the composition of the TMR, as well as the constituents of the compound feed used as one of the ingredients of the TMR. Out of 193 TMR samples received, a total of 163 included a report with the specific composition of the sample. The samples were stored at −18°C until analysis.

Chemicals, reagents and materials
Acetonitrile and methanol, both HPLC grade, and n-hexane were obtained from Scharlab (Sentmenat, Spain). Glacial acetic acid and ammonium formate were sourced from Fischer Scientific (Loughborough, UK). Water was purified on a Milli-Q system (Millipore, Billerica, MA, USA). Mycotoxin stock solutions were made by diluting the standard solutions in methanol for FB 1 , FB 2 , DON, ZEN, α-ZEN and β-ZEN and in acetonitrile for 15-ADON, 3-ADON, DON-3-Glc, 13 C 15 -DON, 13 C 18 -ZEN and 13 C 34 -FB 1 . The concentrations in the stock solutions were checked by UV spectroscopy according to AOAC official methods of analysis (Horwitz and Latimer 2006). Working solutions of each family of mycotoxins and one of the labelled mycotoxins at different concentrations were prepared using the appropriate solvent. All solutions were stored at − 4°C.

Sample preparation
Feed samples were dried over 24 h at 60°C, ground into fine powder and stored at − 18°C until analysis. An in-house validated method based on methods by Monbaliu et al. (2010) and Tevell Åberg et al. (2013) was used. About 2 g of feed sample were weighted into 50-mL polypropylene centrifuge tubes, spiked with 250 μg/kg of 13 C 15 -DON (used as the internal standard) and allowed to stand for 30 min. Then, 30 mL of extraction solution consisting of acetonitrile/water/acetic acid (79:20:1, v/v/v) was added. The samples were horizontally shaken for 1 h at 200 rpm and centrifuged for 10 min at 1008 g. Then, 15 mL of the supernatant were defatted by extraction with 10 mL of n-hexane for 10 min on a rotatory shaker. Finally, after 10 min of centrifugation at 1008 g, the hexane layer was removed and discarded.
To extract all the mycotoxins, two different clean-up pathways were performed. On the one hand, to recover DON, ZEN and their metabolites, 7 mL of the defatted extract was passed through a Multisep® 226 Aflazon+ Multifunctional column, and 3 mL of the clean extract was collected in a vial. On the other hand, with respect to FB 1 and FB 2 , 6 mL of the defatted extract was filtered through a glass microfibre filter, and 2 mL of the filtrate was combined with the Multisep® 226 eluate. The mixture was evaporated to dryness by nitrogen gas at 40°C, and the residue was reconstituted in 1 mL of the mobile phase consisting of methanol/5 mM ammonium formate in water (50:50, v/v). Before the injection into the chromatographic system, the final extract was filtered through a 0.22-μm PTFE disposable syringe filter (Kinesis, Cambridgeshire, UK).

HPLC-MS/MS analysis
HPLC-MS/MS analysis was performed with an Agilent series 1290 RRLC system (Agilent, Santa Clara, CA, USA) equipped with a binary pump (G4220A) and an autiosampler thermostat (G1330B) coupled to an Agilent triple quadrupole mass spectrometer (6460A) with an electrospray ionization (ESI) source (G1958-65138). Chromatographic separation was achieved using a Zorbax Plus C18 (1.8 μm × 2.1 × 100 mm) column from Agilent (San Jose, CA, USA). The MaissHunter software (Agilent, Santa Clara, CA, USA) was used for optimization and quantification. Gradient elution was established with a mobile phase consisting of methanol (eluent A) and 5 mM ammonium formate in water (eluent B) at a flow rate of 0.2 mL/min. The gradient elution started with a linear increase of A, from 25% to 100% during 3.75 min, maintaining these conditions up to 6.0 min. The composition of the mobile phase returned to the initial conditions in 0.5 min and was maintained during 1 min, resulting in a total run time of 7.5 min. The temperature of the column was set at 25°C, and the injection volume was 5 μL.
MS analyses were carried out using selected reaction monitoring (SRM) mode with positive and negative electrospray ionization (ESI ± ). The source gas temperature and sheath gas temperature were set at 325°C and 400°C, respectively, with flows of 5 L/min and 11 L/min, respectively. The nebulizer was set to 45 psi. The capillary and nozzle voltage were 3500 V and 500 V, respectively. The retention time and MS/ MS parameters are shown in Table 1.

Method development
The analytical method used was assessed for precision, calculating the apparent recovery, which is the ratio between the measured concentration and the spiked concentration. For this purpose, eleven blank feed samples were spiked with known amounts of mycotoxin mixtures at three different concentration levels (low, medium and high) at least in triplicate. Moreover, 13 C 15 -DON, 13 C 34 -FB 1 and 13 C 18 -ZEN were added to every sample in the same concentration (250 μg/kg) to be used as internal standards. As each TMR sample had a unique composition, it was not possible to validate the method for every feed sample. Accordingly, a TMR sample, with an average composition, was considered a blank sample to calculate the apparent recovery of the method.
The matrix effect was studied by comparing the slopes when matrix-matched calibration solutions and solvent calibration solutions were prepared at concentrations between 0.5 and 250 μg/L. The matrix effect was calculated using the following equation: ME (%) = [(slope matrix/slope solvent) -1] × 100. For the estimation of the LOD and LOQ, extracted blank samples were spiked at low concentrations (from 0.1 to 50 μg/L). Both LOD and LOQ were calculated as the concentrations for which signal-to-noise (S/N) ratios were 3 and 10, respectively. The method performance for DON, DON-3-Glc, 3-ADON, 15-ADON, ZEN, α-ZEN, β-ZEN, FB 1, and FB 2 is summarized in Table 2.

Statistical analysis
Data processing was performed using JMP Pro 13.1.0. Analysis of variance (ANOVA, significance level set at p < 0.05) and multivariate partial least square (PLS) regression were applied to the data. In addition, the correlation among the concentration of mycotoxins under study and the components of the TMR samples was studied. Other calculations were performed using Microsoft Office Excel 2016.

Development of the sample preparation method
The complexity of some animal feed matrices, coupled with the chemical diversity of the analytes, makes sample preparation challenging for the development of a multi-mycotoxin analytical method.
Based on the literature, different extraction procedures were evaluated in this work. Relative to the extraction solvent, various combinations of organic solvents, acetonitrile or methanol and water were used. In agreement with previous studies, better results were obtained using acetonitrile than with methanol (Garon et al. 2006;Sulyok et al. 2006;Zhao et al. 2015). Moreover, it has been proved that the acidification of the extraction solvent improves the recovery of FBs (Bennett and Klich 2003;Sulyok et al. 2006) and prevents the degradation of certain analytes when acetonitrile is used (Mol et al. 2008). For these reasons, the extraction solvent was acidified with acetic acid. For the extraction method, different strategies were studied. It was not possible to develop a method without a clean-up step, owing to the matrix interferences. A defatting step was necessary to obtain sufficiently clean extracts. QuEChERS was the first approach because it is a relatively fast, cheap and easy method and because it has been frequently used for mycotoxin analysis (Berthiller et al. 2018). For the clean-up, C18, MgSO 4 and activated charcoal were tested as sorbents. ZEN could not be recovered in that way, and low recoveries were obtained for the rest of the mycotoxins. Finally, the chosen method was based on the methods developed by Monbaliu et al. (2010) and Tevell Åberg et al. (2013). A multifunctional column, Multisep®226 Aflazon+, was used for the clean-up, since these columns had been proven to work well for DON and ZEN in grains and animal feed (Tevell Åberg et al. 2013). Moreover, a second pathway using a glass microfibre filter was necessary to recover FBs (Monbaliu et al. 2010). Therefore, a suitable multi-mycotoxin method that allowed the analysis of TMR samples with only one LC-MS/MS injection was developed ( Table 2).

Analysis of TMR samples
The results of the analysis for the presence of mycotoxins in TMR samples are summarized in Table 3. In total, 112 TMR   difference was not significant in the case of DON. ZEN was not detected in any of the samples analysed that were taken from farms located in Galicia (area 08). Regarding FBs, these toxins were the most frequently detected, especially in samples received from Cantabria (area 04) and Cataluña (area 25) (Table 4). In addition, two specific farms from these two areas presented an especially high incidence of FBs in their samples, with 92 and 100% of the samples positive for the presence of FBs and with concentrations ranging from 45 to 1100 μg/kg and from 500 to 1000 μg/kg, respectively. Looking at the composition of these specific TMR samples, we observed that those taken from the farm in the area 04 contained a high percentage of maize as an ingredient of the compound feed, which also represented a high portion of the total composition (25.6-33.9%). For the samples taken from the farm of the area 25, all of them contained maize at a relatively high percentage (13.1-17.1%) compared to the total set of samples. Maize silage, bagasse and barley were always present as ingredients of these samples. FBs are frequently found contaminating maize (Marin et al. 2013;Tevell Åberg et al. 2013), so a greater use of maize in these samples may be linked with the presence of FBs.
In some of the studies conducted involving a wide variety of feedstuffs, FBs were also the most frequently detected mycotoxins. Pleadin et al. (2017) analysed the presence of Fusarium mycotoxins in maize silage and concentrated feeds. DON was detected in 77%, ZEN in 66% and FBs in 80%. In contrast, DON appeared to be the most common mycotoxin for Kocasari et al. (2012); 45% of samples of dairy cattle feed contained DON, 35% ZEN and 10% FBs. In a similar manner, Driehuis et al. (2008a) analysed different materials used as feedstuffs for dairy cattle. DON was the most frequently detected mycotoxin, and the incidence was particularly high in compound feed and silage samples (53 and 54% positive samples, respectively). ZEN was detected in 28% of compound feed samples and 17% of silage, and FBs were detected in only 2 samples, one of compound feed and one of corn gluten feed sample.
DON and ZEN metabolites could also be detected with the analytical method developed. α-ZEN and β-ZEN were not found in any of the TMR samples. DON derivatives, as the conjugated compound DON-3-Glc and the acetylated forms 3-ADON and 15-ADON, have been reported to co-occur with DON in cereal commodities (Berthiller et al. 2013). In the current study, 18 samples contained 15-ADON (mean 28 μg/kg). However, 3-ADON was not detected in any of the TMR samples. As far as DON-3-Glc was concerned, only three samples presented detectable levels of this compound (mean 46 μg/kg). Moreover, only one positive sample for DON presented DON-3-Glc, and 15-ADON co-occurred with DON in only three samples. In contrast to our results, Monbaliu et al. (2010) reported a higher incidence of 3-ADON and 15-ADON, pointing out that these metabolites did not always co-occur. Abdallah et al. (2017) found DON-3-Glc in 10% of the feed samples, and Driehuis et al. (2008a) did not find any sample containing 3-ADON or 15-ADON. Streit et al. (2013) found that DON-3-Glc always co-occurred with DON, and 75% of the samples presented this toxin.  (Payros et al. 2016). Moreover, the fact that none of the samples presented 3-ADON could be affected by the lower percentage of recovery of the method employed ( Table 2).
The co-occurrence of DON and ZEN in several crops and commodities has been widely reported (Driehuis et al. 2008a;Njobeh et al. 2012;Zachariasova et al. 2014;Abdallah et al. 2017). Both mycotoxins are produced by F. graminearum and F. culmorum, and the requirements to produce them are quite similar (Njobeh et al. 2012). Nevertheless, our outcomes showed that only four samples contained DON and ZEN at the same time, and any sample that presented DON, ZEN and FBs simultaneously was never found.

TMR composition in relation to the mycotoxin concentration
Feeds for dairy cows are formulated with a wide variety of ingredients, with forages and concentrates being the main components of the ruminant diets. These materials could be potential sources of mycotoxins. Therefore, it is an important issue to identify which ingredients used in the formulation of the TMR samples contribute to the total mycotoxin load. For this purpose, PLS regression and correlation analysis were carried out. PLS is a multiple regression method that produces a model for each mycotoxin by combining all TMR components at a time, while correlation analysis was carried out with one single variable at a time. Thus, the results of both analyses can be different but complementary.

Fumonisins
TMR samples that showed detectable levels of FBs usually contained compound feed, maize silage, straw, grass silage, bagasse, defatted soya flour, dehydrated ryegrass and wheat silage more frequently than the total set of samples, which includes positive and negative samples for FBs (Table S1).
Considering only the 60 FB-positive samples (with TMR composition available), higher concentrations of FBs were found in samples that contained wheat silage (mean 775 μg/kg), sorghum silage (mean 758 μg/kg), barley (mean 733 μg/kg) and also dehydrated ryegrass, vetch silage and vetch hay (Table S2). Accordingly, the concentration of FBs in the positive samples may be linked to the use of different kinds of silages, plus barley, mainly, although barley was usually used in higher amount in samples also containing wheat and sorghum silage.
Moreover, the correlation analysis between FBs concentrations and the amount of TMR components showed a significant positive correlation between the level of FBs and the presence of wheat silage (r = 0.31), barley (r = 0.24), straw (r = 0.19) and maize (r = 0.16) in the TMR samples. However, PLS regression also pointed to the presence of barley, straw and maize and to immature corn silage as well. Both correlation analysis and PLS are somewhat dominated by the samples with higher FB concentrations. Compound feed was present in the four samples with higher concentrations. Wheat silage was rarely used by the farmers, but when used, the resulting samples contained FBs. In conclusion, as the components indicated by the correlations are different from those linked to the presence/absence of FBs, we can conclude that compound feed and silages are responsible for the presence of FBs in most samples, while, in particular, cereal silages may be related to the highest levels observed.

Deoxynivalenol
Samples that presented detectable levels of DON usually contained maize silage, straw, ryegrass silage, alfalfa silage, oat silage, rye silage and rapeseed as ingredients more frequently than the total set of samples (Table S1).
Considering only the samples that contained DON (29) (with a known TMR composition), we noticed that higher concentrations of toxin were determined in samples that contained cottonseed (mean of 418 μg/kg), immature corn silage (mean of 406 μg/kg), dehydrated alfalfa (mean 363 μg/kg) and also rapeseed, barley and grass silage (Table S2). Thus, DON contamination both in terms of incidence and concentration level seems to be linked to the use of different kinds of silages and cottonseed. However, cottonseed inclusion was correlated with the presence of immature corn silage, barley, alfalfa hay and dehydrated alfalfa. Thus, it is not possible to determine which of them contributed to the contamination. Similarly, the analysis of the correlation between DON concentration and levels of the different components in the TMR showed a significant positive correlation between the level of DON and the level of dehydrated alfalfa (r = 0.40), immature corn silage (r = 0.40) and cottonseed (r = 0.35) and in a lower level straw, compound feed and rapeseed. As far as the PLS regression was concerned, the components correlated to DON were compound feed, cottonseed, maize and rapeseed, apart from immature corn silage. The three samples with higher concentrations contained dehydrated alfalfa and compound feed in concentrations higher than average, which is relevant data. Cottonseed and immature corn silage were rarely used by farmers, but they were present in one of the most contaminated samples. However, as the components related to the level of contamination are different from the ones linked to the presence/absence of DON, we can conclude that although maize, ryegrass or alfalfa silages could affect the presence of DON in most samples, other components like immature corn silage, compound feed, rapeseed, cottonseed or straw may be linked to the higher contamination.

Zearalenone
Samples that contained detectable levels of ZEN had in common that they usually contained compound feed, maize, grass silage, ryegrass silage, defatted soya flour and oat silage more frequently than the total set of samples (Table S1).
Considering only the 25 ZEN-positive samples (with TMR composition available), we observed that higher concentrations of ZEN were determined in samples that contained molasses (mean 269 μg/kg), cottonseed (mean 216 μg/kg), sugar beet pulp (mean 200 μg/kg), oat silage (mean 185 μg/kg), alfalfa hay (mean 128 μg/kg) and maize (mean 122 μg/kg) and also compound feed, maize silage and straw (Table S2). Therefore, contamination both in terms of incidence and concentration level seems to be linked to compound feed, maize, defatted soya flour and different kinds of silages.
Analysis of the correlation between ZEN concentration and levels of the different components in the TMR showed a significant positive correlation between the level of ZEN and the levels of oat silage (r = 0.45), molasses (r = 0.31), maize silage (r = 0.25) and straw (r = 0.18). Additionally, PLS regression pointed to dehydrated alfalfa, straw, oat silage, maize silage and sugar beet pulp. The two samples with higher concentration contained cereal silage plus compound feed or maize as main dry ingredients. However, the components related to the highest levels of toxin are slightly different from the ones linked to the presence/absence of ZEN because grass and ryegrass silages and defatted soya flour were not present in highly contaminated samples.
Examining the results, it was difficult to reach definite conclusions about the influence of the ingredients of TMR on the level of contamination. Other authors such as Cogan et al. (2017) stated than in some samples, non-forage feeds appeared to contribute to the mycotoxin contamination, whereas in other samples, the contributor was maize silage. For Romera et al. (2018), wheat bran, maize or barley seed seemed to be related to the presence of FBs, DON and ZEN in cattle feed samples.
In conclusion, a reliable analytical method for the detection of multiple mycotoxins in a complex matrix was successfully developed and validated. This method was applied for the detection of Fusarium mycotoxins and their metabolites in TMR samples for 2 years. A total of 58% of the samples presented at least one mycotoxin, with FBs being the most frequently found. Moreover, in 20% of the samples, more than one mycotoxin was detected. The level of contamination was always below the guidance levels established by the EU. However, the presence of mycotoxins in animal feed and in feed ingredients should be regularly monitored to prevent health hazards in animals and humans. In addition, special attention must be paid to the co-occurrence of mycotoxins due to their possible synergistic and additive effects. To our knowledge, few studies have been carried out analysing TMR. Thus, this study provided useful information about the level of mycotoxin contamination in final products, as they are consumed by animals. Although it is difficult to draw definite conclusions, the use of some silages and certain concentrates in the formulation of TMR is probably related to the presence of Fusarium mycotoxins.