Food Additives & Contaminants: Part A. ISSN: (Print) (Online) Journal homepage:

Transcription

1 Food Additives & Contaminants: Part A ISSN: (Print) (Online) Journal homepage: Validation and application of a liquid chromatography-tandem mass spectrometry based method for the assessment of the cooccurrence of mycotoxins in maize silages from dairy farms in NW Spain Thierry Dagnac, Alicia Latorre, Bruno Fernández Lorenzo & Maria Llompart To cite this article: Thierry Dagnac, Alicia Latorre, Bruno Fernández Lorenzo & Maria Llompart (216) Validation and application of a liquid chromatography-tandem mass spectrometry based method for the assessment of the co-occurrence of mycotoxins in maize silages from dairy farms in NW Spain, Food Additives & Contaminants: Part A, 33:12, , DOI: 1.18/ To link to this article: Accepted author version posted online: 6 Oct 216. Published online: 26 Oct 216. Submit your article to this journal Article views: 23 View related articles View Crossmark data Full Terms & Conditions of access and use can be found at Download by: [KU Leuven University Library] Date: 19 November 216, At: 7:26

2 FOOD ADDITIVES & CONTAMINANTS: PART A, 216 VOL. 33, NO. 12, Validation and application of a liquid chromatography-tandem mass spectrometry based method for the assessment of the co-occurrence of mycotoxins in maize silages from dairy farms in NW Spain Thierry Dagnac a, Alicia Latorre a, Bruno Fernández Lorenzo a and Maria Llompart b a Department of Animal Production, INGACAL (Galician Institute for Food Quality) CIAM (Agrarian and Agronomic Research Centre), Laboratory of Food/Feed Safety and Organic Contaminants, A Coruña, Spain; b Department of Analytical Chemistry, Nutrition and Food Science. Faculty of Chemistry, Campus Vida. University of Santiago de Compostela, Santiago de Compostela, Spain ABSTRACT The first objective of this study was the validation of an efficient multi-analyte method for the simultaneous detection and quantification of mycotoxins in maize silage, by reverse-phase liquid chromatography coupled with electrospray ionisation triple quadrupole mass spectrometry (LC- HESI-MS/MS). A simple liquid/solid extraction was performed either with clean-up on Mycospin 4 columns or without any clean-up. Almost all the target mycotoxins showed highly-suppressed signals in the presence of a matrix, emphasising the need to quantitate mycotoxins by means of matrix-matched calibrations. An alternative validation method based on ISO and on a single factor balanced design was implemented. The achieved average recoveries from spiked samples at three levels ranged from 6% to 122% with relative standard deviations (rsd) below 11%. Limits of Detection (LODs) and Limits of Quantification (LOQs) were between µg kg 1 and.6 57 µg kg 1. The calculated repeatability and within-lab reproducibility ranged from 5.2 to 23.2% and from 7.2 to 23.9%, respectively. Finally, the decision limit and detection capacity, CCα and CCβ, were calculated for all mycotoxins having regulated/recommended contents in feed. The validated method was applied to 148 samples collected over two years in 19 dairy farms from Galicia (NW Spain). Of the analysed samples, 62% contained at least one mycotoxin. Zearalenone (ZEA), deoxynivalenol (DON), fumonisins B1 and B2, roquefortine C, α-zearalenol, β-zearalenol, enniatins B and B1, andrastin A, marcfortine A, verruculogen and mycophenolic acid were quantified, the highest average detection frequency being for enniatin B (51%). DON, mycophenolic acid and ZEA plus metabolites (α-zearalenol, β-zearalenol) were the most abundant mycotoxins. ARTICLE HISTORY Received 17 June 216 Accepted 2 September 216 KEYWORDS Multi-mycotoxin method; clean-up free extraction; validation strategy; monitoring; maize silage; mass spectrometry Introduction Mycotoxins are secondary metabolites naturally produced by filamentous fungi that can cause a carcinogenic, a teratogenic, an estrogenic, a nephrotoxic, a neurotoxic, a hepatotoxic and/or an immunosuppressive response when ingested by farm animals (Van Pamel et al. 211). The presence of these fungi reduces the nutritional value of the feed and can lead to the production of mycotoxins, one of the major risks for human and animal health (Tangni, Pussemier, Bastiaanse, et al. 213). The Food and Agriculture Organization of the United Nations (FAO) has estimated that approximately 25% of crops worldwide are affected by mycotoxins, and that the loss extends to billions of dollars (Bhat et al. 21). The incidence and occurrence of mycotoxins in animal feed has mainly been studied in cereals. However, a broad range of mycotoxins is reported to be present in concentrates, pasture forage and in preserved feeding stuffs, such as silage and hay (Cheli et al. 213; Tangni, Pussemier, & Van Hove 213). In maize silage, these mycotoxins can be generated from preharvest species such as Fusarium, Alternaria and Aspergillus, or post-harvest species such as Penicillium roqueforti, P. paneum, Aspergillus fumigatus and Byssochlamys nivea (Drejer Storm et al. 214). Maize silage is one of the most important feed sources intended for ruminants in many parts of the world, constituting between 5 75% of the dry matter intake for dairy cows (Dunière et al. 213). Owing to their chemical and physical resistance, some mycotoxins can be CONTACT Thierry Dagnac Thierry.dagnac@xunta.es INGACAL (Galician Institute for Food Quality) CIAM (Agrarian and Agronomic Research Centre), Laboratory of Food/Feed Safety and Organic Contaminants, Apartado 1, Abegondo, E-158, A Coruña, Spain 216 Informa UK Limited, trading as Taylor & Francis Group

3 FOOD ADDITIVES & CONTAMINANTS: PART A 1851 transferred to food produced by animals. The European Union (EU) has then adopted rules limiting the maximum levels of mycotoxins in feeding stuffs (European Commission 23); particularly aflatoxin B1, ochratoxin A, deoxynivalenol, zearalenone, fumonisins, and, more recently, T-2 and HT-2 toxins. The EU also suggested increasing the monitoring studies for the presence of these mycotoxins in cereals and products intended for animal feeding. The Food and Drug Administration (FDA) in the United States and the Canadian Food Inspection Agency in Canada have set legislated maximum tolerance levels for aflatoxins and regulatory guidelines for other mycotoxins, such as deoxynivalenol (United States and Canada) and HT-2 toxin (Canada) present in animal feed. Moreover, while the United States recommended tolerance levels for fumonisins in feedstuff, Canada has established recommendations for diacetoxyscirpenol, zearalenone, ochratoxin A, ergot and T-2 toxin (Canadian Food Inspection Agency). Since these regulations cover several mycotoxins and their simultaneous occurrence is commonplace, the European Food Safety Authority (EFSA) recommends developing multi-mycotoxin methods. Furthermore, additive and synergistic effects have been observed when mixtures of mycotoxins were present in the same commodity (Streit et al. 213). Because there are no specific regulations for mycotoxins in silages, previous legal levels for feed could be considered as guidelines for silage (Cheli et al. 213). The high diversity of physical and chemical properties of mycotoxins means that finding a suitable sample preparation and a good detection method for the simultaneous analyses of a large number of mycotoxins is a great challenge (Ates et al. 214). Several multi-mycotoxin methods based on liquid chromatography coupled with triple quadrupole mass spectrometry (LC-MS/MS), have been published for various food and feed commodities (Berthiller et al. 214) A wide variety of extraction solvent mixtures such as acetonitrile (MeCN), methanol (MeOH), acetone, chloroform, dichloromethane and ethyl acetate have been employed, MeCN/water (84:16, v/v) being the mixture most commonly used for multimycotoxin extractions (Pereira et al. 214). Several studies have described clean-up procedures in silage and feedstuffs with different types of solid phase extraction (SPE) (Garon et al. 26; Richard et al. 27) and the use of multi-mycotoxin immunoaffinity columns (IACs) (Lattanzio et al. 214) and different forms of Mycosep multifunctional columns (Ren et al. 214). However, due to the wide range of polarities and physical properties of the mycotoxins, avoiding the clean-up step should be advantageous, and also provide faster multimycotoxin analysis. A modified Quick, Easy, Cheap, Effective, Rugged and Safe (QuEChERS) extraction procedure excluding the dispersive SPE (dspe) clean-up step has been developed by several authors for multimycotoxin analysis in food and feed products (Rasmussen et al. 21; Dzuman et al. 214; Zachariasova et al. 214; McElhinney et al. 215). The co-occurrence of several mycotoxins in silages creates the need for multi-mycotoxin methods that should be properly validated for silages (Cheli et al. 213). Although multi-residue methods have been developed for the detection of mycotoxins in maize silage samples (Richard et al. 27; Boudra & Morgavi 28; Driehuis et al. 28; Mansfield et al. 28; Van Pamel et al. 211; Keller et al. 213; Streit et al. 213), due to the complexity of this matrix just a few multi-mycotoxin methods have been fully validated in maize silage (Rasmussen et al. 21; Van Pamel et al. 211; Dzumanetal.214). Van Pamel et al. (211)designedavalidatedLC-MS/MS and UHPLC-MS/MS method for the determination of 6 and 26 different mycotoxins, respectively, after a solid phase extraction. The study of Rasmussen et al. (21), using a LC-MS/MS and a QuEChERS procedure, was validated for the extraction of 27 mycotoxins in maize silage samples. More recently, Dzuman et al. (214) optimised and validated a method for analysis of a wide range of mycotoxins in wheat, complex compound feed and maizesilageusingaquechersmethodfollowedbya dispersive solid phase clean-up (C18 silica sorbent) of the extract. The aim of the current study was to develop and validate an alternative method for the simultaneous determination of multi-class mycotoxins in maize silage by LC/HESI-MS/MS detection, using a simplified extraction with clean-up on Mycospin 4 columns as a first approach and without any clean-up step as a final approach. The method was successfully validated for 24 mycotoxins in maize silage according to a simplified strategy following the ISO (2) guidelines. Different parameters such as instrument detection and quantification limits, matrix effect, recovery, repeatability, and within-lab reproducibility, decision limit (CCα) and detection capability (CCβ) were evaluated. In order to assess the multi-mycotoxin occurrence in maize silages, the validated method was applied to 148 samples collected in 19 dairy farms from Galicia (NW Spain), at two different seasons and over two years. Materials and methods Chemicals and reagents Whatman No. 4 filter papers were obtained from Whatman International Ltd. (Maidstone, UK).

4 1852 T. DAGNAC ET AL. Polytetrafluoroethylene (PTFE) syringe filters (.45 µm) were obtained from Sartorius (Goettingen, Germany). Methanol (MeOH), acetonitrile (MeCN) and water, gradient grade, for HPLC and acetic acid 99 1% glacial (HPLC analysed) were purchased from J.T. Baker (Deventer, the Netherlands). Formic acid (98%) (LC/MS grade) and ammonium formate (LC/ MS grade) were supplied by Fluka (Basel, Switzerland). Standards of HT-2 Toxin (HT-2), zearalenone (ZEA), deoxynivalenol (DON), 3-acetyldeoxynivalenol (3-ADON) and 15-acetyldeoxynivalenol (15-ADON) were purchased from Biopure (Tulln, Austria). Penicillic acid, verruculogen (VERR), T-2 toxin (T-2) and roquefortine C (ROQ-C) each at 1 µg ml 1, aflatoxin B1 (AFB1) and aflatoxin G1 (AFG1) each at 2µgmL 1 and ochratoxin A (OTA) at 1 µg ml 1 were purchased in MeCN from Biopure (Tulln, Austria). Fumonisin B1 (FB1) and Fumonisin B2 (FB2) were obtained from Acros Organics (New Jersey, USA) and mycophenolic acid from Alfa Aesar (Karlsruhe, Germany). Standards of alternariol and penitrem A were obtained from Cayman Chemical (Michigan, USA) and andrastin A and marcfortine A from Santa Cruz Biotechnology (California, USA). The other mycotoxin standards were supplied by Sigma Aldrich (Missouri, USA): enniatin B, enniatin B1, sterigmatocystin, α-zearalenol and β-zearalenol. From the solid standards, stock solutions were prepared in MeCN at 1 mg ml 1 for HT-2, ZEA, 3-ADON, 15-ADON, penitrem A, at.1 mg ml 1 for enniatin B, enniatin B1, sterigmatocystin, andrastin A and marcfortine A, at.5 mg ml 1 for alternariol, at 5 mg ml 1 for α-zearalenol and β-zearalenol and at 1 mg ml 1 for DON, while FB1 and FB2 solutions were prepared in MeCN/H 2 O (1:1; v/v) at.5 mg ml 1.Allthesestock solutions were stored at 2 C except FB1 and FB2 that were stored at 4 C. The main standard working solution of mycotoxins was prepared by combining aliquots of each individual stock solution and diluting with MeCN to obtain the following concentrations of mycotoxins: DON 5 µg ml 1 ;penitrema,α-zearalenol and β-zearalenol 3 µg ml 1 ; 3-ADON, 15-ADON and HT-2 25 µg ml 1 ; mycophenolic acid and alternariol 2 µg ml 1 ; penicillic acid 1 µg ml 1 ;ZEA,enniatin B and enniatin B1,5 µg ml 1 ; FB1 and FB2 4 µg ml 1 ; VERR, ROQ-C and sterigmatocystin 2 µg ml 1 ;OTA 1.5 µg ml 1 ; AFB1, AFG1 and andrastin A 1 µg ml 1 ; and marcfortine A and T-2.5 µg ml 1.Inorderto prepare the calibration solutions, the previous working solution was diluted (1:1; v/v) with MeCN. Both standard solutions were stored at 2 C and in the dark. Sample preparation Since no blank certified reference materials for mycotoxins in silage were available, the validation method was performed with maize silage collected at the CIAM experimental farm. This blank material was dried (2 days at 4 C) and ground through a mill (particle size: 1 mm) until the method development and validation took place. One gram of ground maize silage was placed into a 5 ml polypropylene conical centrifuge tube and treated with 1 ml of MeCN/water (84:16 v/v) containing 1% acetic acid. The mixture was shaken for 9 min at 58 g Orbital Maxi OL3-ME (Ovan, Badalona, Spain) and the supernatant was filtered through filter paper (Whatman No. 4). In this study the efficiencies of two methods for sample preparation were evaluated. For the first procedure, 5 µl of filtered extract were evaporated to dryness under gentle nitrogen stream and reconstituted with 5 µl of water/meoh (8:2, v/v) containing 3 mmol L 1 of ammonium formate and.1% of formic acid. The obtained solution was forced through a.45 µm PTFE syringe filter. In the second procedure, a clean-up step with Mycospin 4 Multi-toxin columns (Romer Labs, Europe, Tulln, Austria) was tested. An aliquot of 8 µl of the filtered extract was transferred to the Mycospin column, shaken for one minute and centrifuged for 3 seconds at 928 g (Microcentrifuge 5415D, Eppendorf, Hamburg, Germany). Then, 5 µl of the purified extract were evaporated to dryness under gentle nitrogen stream and reconstituted with 5 µl of water/meoh (8:2, v/v) containing 3mmolL 1 ammonium formate and.1% of formic acid. Maize silage samples Samples of maize silage were collected from 19 selected dairy farms located in Galicia (Northern Spain). Sixtyfour samples were first collected between February and March (32), then between July and August (32). The following year, 84 silage samples were first collected between January and February (44) and then between July and August (4). On each farm, two samples of 1.5 kg each were taken per silo, one sample at the top and the other at the middle of the feed-out face. The collected silage samples were dried (two days at 4 C), ground through a mill (particle size: 1 mm) and stored at 18 C until the multi-mycotoxin analysis took place by means of the validated method. Each silage sample was analysed in duplicate.

5 FOOD ADDITIVES & CONTAMINANTS: PART A 1853 LC-MS/MS equipment and conditions LC-MS/MS analysis was carried out using a Thermo Fisher Scientific (San Jose, CA, USA) instrument consisting of a Dionex Ultimate 3 pump and an Accela autosampler and a Quantum Access triple quadrupole mass spectrometer equipped with a heated electrospray ionisation (HESI) source. Twenty millilitres of standard solution or silage extract were injected into the Kinetex C18 column (1 2.1 mm, 2.6 µm) (Phenomenex, Torrance, CA, USA). The column and the sample temperature were maintained at 25 C and 1 C, respectively. The analytical separation was performed using a gradient elution of water (mobile phase A) and methanol (mobile phase B), both with 3 mmoll 1 ammonium formate,.1% formic acid (v/v): 1 min, isocratic step at 8% A, 1 1 min to 1% B, 1 12 min isocratic step at 1% B; finally, the column was equilibrated to initial conditions for 1 min. The flow rate was set to.2 ml min 1 giving a total run time of 33 min, with positive and negative modes running separately. The HESI-MS/MS interface was working with the following parameters: sheath gas (38 au (arbitrary unit)), auxiliary gas (5 au), skimmer offset (4 V), capillary temperature (35 C) and ion sweep cone gas (3 au); all in both positive and negative modes. In the positive mode: the vaporiser temperature was 25 C and the spray voltage was 3 V. In the negative mode: the vaporizer temperature was 7 C and the spray voltage was 25 V. The MS parameters and selective reaction monitoring (SRM) transitions were optimised by direct infusion of standard solutions (1 µg ml 1 ). All the SRM transitions for each analyte, along with the retention times, are detailed in Table 1. The peak width was.7 Da in Q 1 and Q 3, and the argon pressure in the collision cell was set to.2 mbar. The scan time was set to 5 ms for all the transitions acquired. The Thermo Scientific LC QUAN Quantitative Software was used to process the quantitative data obtained from the calibration standards and samples. Evaluation of the method performance parameters The matrix-matched calibration curves were prepared by adding appropriate volumes of the mixture working solution to the blank sample extracts obtained (see the extraction procedure previously described), whereas the standard calibration curves were constructed by evaporation to dryness of the appropriate volumes of the working standard solution and subsequently reconstituted with 5 µl of water/meoh (8:2, v/v) containing 3 mmol L 1 ammonium formate and.1% of formic acid. Both the matrix-matched and standard calibration curves were performed at eight concentration levels in the ranges: µg L 1 for andrastin A;.1 11 µg L 1 for AFB1 and AFG1; 1 22 µg L 1 for T-2;.1 21 µg L 1 for marcfortine A;.7 69 µg L 1 for OTA; 1 2 µg L 1 for ZEA; µg L 1 for DON; µg L 1 for ADON; 5 15 µg L 1 for FB1; µg L 1 for FB2; µg L 1 for Table 1. Retention times and optimised parameters for the analysis of the target mycotoxins by LC-HESI-MS/MS. Mycotoxin RT (min) a Ionisation Precursor Ion Precursor ion (m/z) Product ions (m/z) b Pennicillic acid 4.7 HESI + [M + H] /125.2/153.4 DON 2.7 HESI + [M + H] /23./231.1/249.1/279. AFB1 8.3 HESI + [M + H] /285.1 AFG1 7.7 HESI + [M + H] /214.2/ ADON 6.5 HESI + [M + H] /22.9/213.1/231./261.1/321. ROQ-C 9.2 HESI + [M + H] /198.1/322.1 OTA 1.5 HESI + [M + H] /238.9/358. HT HESI + [M + NH4] /145.2/263.1/425.4 T HESI + [M + NH4] /169./35.2 Verrculogen 11.1 HESI + [M - H2O] /227.1/352.1 FB2 1.5 HESI + [M + H] /336.1/354.4 FB1 9.5 HESI + [M + H] /352.1 Enniantin B 12.4 HESI + [M + H] /196.1/214.2 Enniantin B HESI + [M + H] /21./214. Sterigmatocystin 1.8 HESI + [M + H] /39.9 Marcfortine A 8.3 HESI + [M + H] /176.1/188.2/45.1 alfa-zol 1.3 HESI - [M H] /159.9/173.9/275.1 beta-zol 9.7 HESI - [M H] /159.9/173.9/275.1 Penitrem A 11.7 HESI - [M H] /545.9/586.3 Andrastin A 11.1 HESI - [M H] /166.8/18.8 Alternariol 9.5 HESI - [M H] /156.7/188.6/212.6 ZEA 1.4 HESI - [M H] /159.6/175./187. Mycophenolic acid 9.7 HESI - [M H] /243.3/245.1/275.1 a RT, retention time. b In bold, ion used for quantitation.

6 1854 T. DAGNAC ET AL. HT-2; 2 9 µg L 1 for VERR and ROQ-C; µg L 1 for penicillic acid; 6 93 µg L 1 for mycophenolic acid;.1 13 µg L 1 for enniatin B;.1 11 µg L 1 for sterigmatocystin;.1 23 µg L 1 for enniatin B1; µg L 1 for penitrem A; µg L 1 for α-zearalenol; and 3 16 µg L 1 for β-zearalenol. The matrix effect (ME) was evaluated for each mycotoxin by comparing the slopes obtained with the matrix-matched standard calibration curve (n = 3) (a matrix ) with that of the standard calibration curve (a standard ) for the same concentration range. The ME was calculated with the following formula: 1 1 ME ð% Þ¼ a matrix =a standardþ In addition, the linearity of the external calibrations was checked for all mycotoxins. Limits of Detection (LODs) and Limits of Quantification (LOQs) were determined using the standard calibration curves (y = ax + b) according to the formulas: LOD ¼ b þ 3s b =a LOQ ¼ b þ 1s b =a where b is the intercept of the standard calibration curve, s b is the standard deviation of the b value and a is the slope of the standard calibration curve. Description of the method validation study The European Commission (22) describes the validation protocol and the performance characteristics for methods related to foodstuff control. Following this procedure, a very large amount of analytical determination has to be performed in order to obtain the required quality parameters, meaning high solvent consumption, material and time expenses, and limitations related to the laboratory work. The recoveries are given by the ratio of the observed concentration value obtained from the matrix-matched calibration curves, to the theoretical concentration value. The repeatability and within-lab reproducibility at each level were calculated by analysis of variance (ANOVA, Statsgraphics Plus 5.1) with a single factor balanced design (27 experiments) with three replicates on the same day at the three different concentration levels for the repeatability, and with the three replicates at the three different concentration levels performed on the three separate days for the within-lab reproducibility. Degrees of freedom of the design are equal or higher than with conventional validation strategies. An alternative validation method, proposed by Van Loco and Beernaert (23) and based on the ISO (2) (in the case of linear calibration with constant standard deviation) and ISO (1994) standards, was developed. ISO (2) provides the equations for the calculation of CCα and CCβ defined as the critical value of the net state variable and as the minimum detectable value of the net state variable, respectively. CCα, Limit from which a sample can be declared noncompliant with a statistical certainty equal to 1 α (α = 5% for substances with maximum level) and CCβ, Limit from which the analyte can be detected, identified, and/ or quantified (according to the needs) with a statistical certainty of 1 β (β = 5% for substances with maximum level), were calculated according to the equations provided by the ISO (2). The results (not corrected for recovery) are linear regressed (y = ax + b) versus the spiked concentrations (Van Loco & Beernaert 23). CCα is directly obtained from the regression slope (b), the intercept (a) and the critical value of the response variable (y c ), equations (1) and (2). CCβ is calculated using an iterative procedure, equation (3). sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi y c ¼ a þ bx MRL þ t df ;α Sy 1 þ 1 1:J þ ðx MRL xþ 2 P ðxi xþ 2 (1) yc a CC α ¼ (2) b sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi y c ¼ a þ bcc β t df ;β Sy 1 þ 1 ðccβ xþ2 þ P 1:J ðxi xþ 2 (3) with, x MRL concentration at the regulated limit, t df the 95%-quantile of the t-distribution with 25 degrees of freedom (here t df = 1.78), S y the estimate of the residual standard deviation of the regression model, I the number concentration levels (here I =3),J the number of replicates per concentration level (here J = 9), x average of the 27 real concentration values (x i ). Since some of the target mycotoxins do not have maximum levels in feed, a cut-off level (CoV) was established for each of them prior to the validation (Monbaliu et al. 21). Preliminary experiments aimed at evaluating the actual limits of quantification in silage matrix have been conducted to evaluate the CoV levels. Nine blank silage samples were spiked with the mycotoxin mixture solution at concentrations of.5, 1 and 1.5 times the CoV level and they were analysed in triplicate and on three different days. The equilibration period of spiked samples was set to 3 min.

7 FOOD ADDITIVES & CONTAMINANTS: PART A 1855 Results and discussion Evaluation of the extraction and clean up procedures Maize silage is a complex matrix that contains not only characteristic components from the plant such as chlorophylls and carotenoids, but also products of the fermentation process (Dzuman et al. 214). Owing to the wide range of polarities and physical properties of mycotoxins, a simple sample treatment without any previous clean up would likely offer the best compromise. Previous methods have been based on direct injection for the determination of several mycotoxins in different commodities, including crude plants. In all of them, the clean-up step was avoided, but the extract had to be diluted before injection in order to achieve good recoveries (Sulyok et al. 26; Spanjer et al. 28; Soleimany et al. 212). In terms of extraction solvent, several authors have shown the improved efficiency of the use of MeCN compared with MeOH. Furthermore, as mentioned in some published reports, acidification of the extraction solvent improves the recoveries of acidic mycotoxins such as fumonisins, mycophenolic acid or OTA (Garon et al. 26; Sulyok et al. 26; McElhinney et al. 215). In addition, extraction at low ph is especially important in such a ph-sensitive matrix as silage (Rasmussen et al. 21; Dzuman et al. 214). Taking all of this into account, two procedures were assessed for the extraction of the studied mycotoxins: (a) direct injection without any clean-up step, and (b) a clean-up process with a multi-toxin column containing a combination of adsorbents (Mycospin 4) to purify the extract prior to the injection. In both procedures, the sample extracts were not diluted prior to the LC- MS/MS analysis. The mixture acetonitrile/water (84:16 v/v) containing 1% acetic acid was used as an extraction solvent in the current study. In order to evaluate the efficiency of both extraction methods and the necessity of a cleanup step, recovery experiments were carried out by spiking, prior to the extraction, blank maize silage samples at three concentration levels (.5 CoV, 1 CoV and 1.5 CoV) in triplicate and on three separate days. Recoveries obtained with the two extraction methods are shown in Tables 2 and 3. Without using Mycospin columns, mean recoveries for the tested mycotoxins were in the range of 6 127% (except for ROQ-C and enniatins) with all relative standard deviations (rsd) below 11%, except one (AFB1, 15%). On the other hand, the use of Mycospin columns provided largely acceptable recoveries (>97%) for 14 mycotoxins, but for OTA (36.8%), mycophenolic acid (24.2%) and penitrem A (195.5%), the recoveries were unacceptable. It has also been observed that the use of Mycospin columns for the purification of silage extracts did not allow recoveries of fumonisin and alternariol. They should be retained on the adsorbent material and completely lost when the extract passed through the Mycospin column. However, it has to be underlined that Mycospin 4 Multitoxin columns from Romer Labs had only been optimised for several regulated Table 2. Mycotoxin recoveries achieved without using Mycospin 4 cartridges at three fortification levels. Mycotoxins.5 CoV (μg kg 1 ) Recovery (%) RSD (%) CoV (µg kg 1 ) Recovery (%) RSD (%) 1.5 CoV (μg kg 1 ) Recovery (%) RSD (%) _ADON Enniantin B Enniantin B Sterigmatocystin Marcfortine A Alternariol α-zearalenol β-zearalenol Andrastin A Penitrem A Penicillic acid AFB AFG DON FB FB HT OTA ROQ-C T VERR ZEA Mycophenolic acid

8 1856 T. DAGNAC ET AL. Table 3. Mycotoxin recoveries achieved using Mycospin 4 cartridges at three fortification levels. Mycotoxins.5 CoV (μg kg 1 ) Recovery (%) RSD (%) CoV (μg kg 1 ) Recovery (%) RSD (%) 1.5 CoV (μg kg 1 ) Recovery (%) RSD (%) _ADON Enniantin B Enniantin B Sterigmatocystin Marcfortine A Alternariol ND ND ND α-zearalenol β-zearalenol Andrastin A Penitrem A Penicillic acid AFB AFG DON FB1 ND ND ND FB2 ND ND ND HT OTA ROQ-C T VERR ZEA Mycophenolic acid ND: not detected. mycotoxins such as ZEA, T2, HT2, diacetoxyscirpenol (DAS), neosolaniol (NEOS), DON, 3- and 15-ADON, nivalenol (NIV), fusarenon-x (FX), aflatoxins and OTA, but not for the other mycotoxins studied in this work, such as fumonisins, alternariol, andrastin A or enniatin(s). Furthermore, unacceptably high variability (>2%) of mycotoxin recoveries were obtained after this cleanup procedure for 11 mycotoxins. These findings suggested that the use of Mycospin columns did not provide an actual improvement for the simultaneous extraction of all the target mycotoxins in maize silage. However, the average mycotoxin recovery was higher with Mycospin columns, although with less precision. Optimization of the chromatographic conditions The ESI positive and negative modes were evaluated, observing that whereas the majority of target mycotoxins were better detected in positive mode, others such as ZEA, mycophenolic acid, α-zearalenol, β-zearalenol, penitrem A, alternariol and andrastin A were only detected in negative mode. Both modes were shown to be effective for the ionisation of FB1, FB2 and OTA, but due to the lower signal and the larger matrix effect for FB1 and FB2 in the negative mode (65.8% for FB1 and 36.7% for FB2), ESI in positive mode was selected for these mycotoxins. The suitable composition of the mobile phase was also investigated and optimised. All the mycotoxins studied have shown better responses by using the solvent mixture MeOH/water instead of MeCN/water as a mobile phase. Other authors have pointed out the same trends that might be attributed to the protic nature of MeOH, which enhances the response of [M + H] + in the positive mode (Pereira et al. 214). Due to the presence of four carboxylic groups in the molecular structure of fumonisins, acidic chromatographic conditions should be required (Berthiller et al. 214). Thus, the acidification of the mobile phases with formic acid has provided better signal intensities for FB1, FB2 as well as for OTA. Additionally, ammonium formate should be added to avoid the formation of unwanted stable adducts (e.g. M+Na + ) for trichothecenes type A such as T-2 and HT-2 toxins. Ammonium adducts [M +NH4] + were formed as precursor ions for these trichothecenes. Therefore, water and methanol both with 3 mmol L 1 ammonium formate and.1% of formic acid were selected as mobile phases. Method validation procedure No directive or guidance is currently established for the validation of analytical methods for the multi-mycotoxin determination in food and feed. In this context, some rules defined in the SANTE (ex SANCO) document (SANTE/11945/215) devoted to pesticide residue analysis in food can be implemented. The method validation of the present study was then conducted on the basis of the following parameters: recoveries, repeatability, within-lab reproducibility, CCα (decision limit) and CCβ (detection capability). Different experiment sets were implemented to

9 FOOD ADDITIVES & CONTAMINANTS: PART A 1857 evaluate the detection and quantification limits and matrix effects, as well. Matrix effect assessment The presence of matrix components that co-elute with the analyte(s) of interest can have an adverse effect on their analytical responses resulting in an ion suppression (or enhancement) due to competition from the available charge. Figure 1 exhibits the ME for all studied mycotoxins in maize silage. According to current guidelines, the ME could be considered tolerable in a range of between 2% if related to ion suppression, and +2% if related to signal enhancement. It was observed that for the vast majority of the target mycotoxins, ME values were outside this no-matrix effect range. Thus, the results indicated that the largest ion suppression was for AFB1, AFG1, OTA, ROQ-C, mycophenolic acid, enniatin B, enniatin B1, sterigmatocystin, alternariol, α-zearalenol, β-zearalenol, penitrem A and ZEA (7 12%). In contrast, a signal enhancement for both fumonisins was observed, but with FB2 being only slightly affected by ME (11.9%). Even when including a purification step, strong matrix effects for many mycotoxins were also reported in maize silage samples (Van Pamel et al. 211). This was confirmed by our experiments performed with Mycospin columns (Figure 1) since ion suppressions ranged from around 1% (AFB1 and ROQ C) to 23% (DON and OTA). These values were often similar to those achieved without using Mycospin cartridges. However, in the case of OTA, andrastin A and mycophenolic acid, ion suppression was strongly reduced by including the clean-up step. Therefore, the presence of pigments, lipids or fermentation products abundantly present in maize silages was compensated by using matrix matched standard calibration curves for quantification purposes. Another possible strategy for minimising ME involves the use of isotope labelled internal standards (IS). However, labelled compounds for each mycotoxin would result in a high analytical cost and a single IS could not compensate for the ME for every compound in the considered matrix. The use of matrix-matched standard calibration avoids the need for isotope labelled internal standards. Linearity, LOD and LOQ Coefficients of determination (R 2 ) and linear ranges are presented in Table 4. As shown in this table, LODs and LOQs were between µg kg 1 and.6 57 µg kg 1, respectively. Trueness and precision: recovery, repeatability and within-lab reproducibility Parameters for trueness and precision were assessed at three concentration levels following the protocol described above. As regards the study performed without Mycospin cartridges (Table 4), with the exception of enniatins, average recoveries for the tested mycotoxins varied from 6.5% to 127%, with relative standard deviations (rsd) below 16%. By comparison, Rasmussen et al. (21) reported reduced recoveries of FB1 (6%) and FB2 (13%) with a QuEChERS method without clean-up step for the determination of multiple 6 4 Without Mycospin With Mycospin 2 Matrix effect (%) Figure 1. (colour online) Matrix effect assessment in 1% of blank maize silage for each target mycotoxin (1 to 23) with and without using Mycospin 4 cartridges. In both cases, extractions were performed with a mixture acetonitrile/water (84:16 v/v) containing 1% acetic acid. 1 (OTA); 2 (ZEA); 3 ( Acetyldon); 4 (DON); 5 (Fum B1); 6 (Fum B2); 7 (T-2); 8 (HT-2); 9 (Verruculogen); 1 (Roq C); 11 (Penicillic acid); 12 (Mycophenolic acid); 13 (Enniatin B); 14 (Enniatin B1); 15 (Sterigmatocystin); 16 (Marcfortine A); 17 (Alternariol) 18 (α-zearalenol); 19 (β-zearalenol); 2 (Andrastin A); 21 (Penitrem A); 22 (AFB1); 23 (AFB2).

10 1858 T. DAGNAC ET AL. Table 4. Method performance parameters issued from the validation study performed without using Mycospin 4 cartridges. LODs LOQs LODs LOQs Linearity (range) CCα CCβ Repeatability Within-lab reproducibility Average recovery a ML in feed b Mycotoxins ng ml 1 ng ml 1 μgkg 1 μgkg 1 r 2. (ng ml 1 ) μg kg 1 μgkg 1 (r. %) (Rw. %) % rsd (%) μg kg _ADON (4 227) /12 Penicillic acid (25 32) AFB (.1 11) /5 AFG (.1 11) /5 DON (11 172) /12 FB (5 15) /6 FB (11 15) OTA (.7 69) /25 ROQ-C (2 9) HT (1 113) T (1 22) VERR (2 9) ZEA (1 2) /3 Mycophenolic acid (6 93) Enniantin B (.1 13) Enniantin B (.1 23) Sterigmatocystin (.1 11) Alternariol (3 93) Marcfortine A (.1 21) Alpha Zearalenol (5 113) /3 Beta Zearalenol (3 16) /3 Andrastin A (3.6 46) Penitrem (12 131) Method performance parameters issued from the validation study performed using Mycospin 4 cartridges _ADON (4 227) /12 Penicillic acid (25 32) AFB (.1 11) /5 AFG (.1 11) /5 DON (11 172) /12 OTA (.7 69) /25 ROQ-C (2 9) HT (1 113) T (1 22) VERR (2 9) ZEA (1 2) /3 Mycophenolic acid (6 93) Enniantin B (.1 13) Enniantin B (.1 23) Sterigmatocystin (.1 11) Marcfortine A (.1 21) alpha Zearalenol (5 113) /3 Beta Zearalenol (3 16) /3 Andrastin A (3.6 46) Penitrem (12 131) a Average from the three fortification levels (CoV). b Lowest (used for calculation of CCα and CCβ) and highest Maximum Level (ML) in feed according to the European guidelines.

11 FOOD ADDITIVES & CONTAMINANTS: PART A 1859 mycotoxins in maize silage. Furthermore, the recoveries achieved in our study for AFB1 (9%), AFG1 (92%) and FB1 (111%) were significantly higher than those reported in two recent studies in animal feed and cereals (Warth et al. 212; Lattanzio et al. 214): (74/ 72% for AFB1, 78/61% for AFG1 and 79/51% for FB1), and also implemented a simple extraction method without applying any clean-up procedure. Therefore, satisfactory recoveries were obtained with the present method, highlighting the suitability of the proposed procedure for extracting mycotoxins from maize silage. The calculated repeatability and within-lab reproducibility ranged from 5.2 to 23.2% and from 7.2 to 23.9%, respectively, highlighting the high robustness of the methodology proposed. In a recent study, Lattanzio et al. (214) obtained higher values of within-lab reproducibility while determining mycotoxins in cereals (especially fumonisins and aflatoxins), a matrix much less complex than silage. Our results are in agreement with the last published study dealing with the determination of mycotoxins in various feed matrices (Dzuman et al. 214). However, these authors employed a two-stage QuEChERS method, including a clean-up step. They did not apply a rigorous validation scheme such as the alternative strategy described in the present study. Regarding the study performed with Mycospin cartridges (Table 4), the average recoveries for the tested mycotoxins varied from 24% to 128%. Many target mycotoxins (e.g. ROQ-C, HT-2, T2, Enniatin(s), and so on) showed better recoveries than those achieved without using Mycospin. However, the relative standard deviations were often higher than 2% and FB1, FB2 and alternariol were no longer detected after Mycospin clean up. The calculated repeatability and within-lab reproducibility ranged from 9 to 42% and from 7 to 47%, respectively. These global results underline the low precision achieved with the Mycospin related method compared with the clean-up-free method. Decision limit (CCα) and detection capability (CCβ) The EU has set maximum levels of AFB1 (European Commission 23) and recommended maximum limits for OTA, DON, ZEA and the sum of fumonisins (European Commission 26) in animal feed. For all these mycotoxins, CCα and CCβ values were calculated by taking into account the lowest limit for each mycotoxin (Table 4). Recommendation 213/165/EU (European Commission 213) sets indicative levels for the sum of the T2 and HT2 toxins in cereals and cereal products, which should trigger investigations. These are not enforcement limits, but they could be used to improve the available data on factors affecting the occurrence of these two mycotoxins. Therefore, according to this recommendation, an indicative level of 25 µg kg 1 (compound feed) is considered here for the calculation of the detection capability. From a global viewpoint, CCα values range from 1.2 to 5776 µg kg 1 and CCβ values from 15.2 to 6842 µg kg 1. Most specifically, as regards AFB1, CCα and CCβ values were 12.3 µg kg 1 and 19.2 µg kg 1, respectively. Application of the validated method to real silage samples Taking into account the method performance parameters, especially those related to the poor trueness and precision (Table 4), the very low recovery for OTA and mycophenolic acid and the total loss of FB1, FB2 and alternariol, all the samples were extracted without using Mycospin cartridges. The validated method was put into practice with the 148 maize silage samples collected from 19 dairy farms in Galicia in two different seasons and over two years. The final content of mycotoxins in silage samples were calculated after applying the corresponding average recovery correction for each mycotoxin. The obtained results are summarised in Table 5, and Figure 2 exhibits a LC-MS/MS chromatogram corresponding to a maize silage sample in which ZEA (247.2 µg kg 1 ); DON (847.5 µg kg 1 ); FB2 (29.6 µg kg 1 ); mycophenolic acid (3159 µg kg 1 ); ROQ-C (1439 µg kg 1 ); andrastin A (821 µg kg 1 ); and β-zearalenol (1721 µg kg 1 ) have been quantified. From a global point of view, mycotoxins were detected in 62% of the silage samples. Silage samples were completely free of toxins only in two farms, and just the second year. Five out of the 148 samples contained more than four mycotoxins, with one sample containing as many as seven. Thirteen out of the 24 mycotoxins assayed have been successfully detected and quantified in the maize silage samples (mycophenolic acid, ROQ-C, FB1, FB2, DON, enniatin B, enniatin B1, marcfortine A, VERR, ZEA, α-zearalenol, β-zearalenol and andrastin A). The most frequent mycotoxins were enniatin B, ZEA and FB2, constituting 51.4%, 21.5% and 21.3% of the total positive samples analysed, respectively. The maximum detected levels found were 6686 µg kg 1 for DON and 3679 µg kg 1 for mycophenolic acid. It must be underlined that, as expected, ZEA was present in all maize silage samples that also contained DON. This simultaneous

12 186 T. DAGNAC ET AL. Table 5. Average, minimum and maximum concentrations (µg kg 1 ) of different mycotoxins found in the 148 maize silage samples collected in 19 dairy farms over two years (year 1: 64 samples. year 2: 84 samples). Mycotoxins Year 1 Average Min Max Detection frequency (%) DON Enniantin B Fumonisin B Fumonisin B Marcfortine A ZEA Andrastin-A ADON ND Penicillic acid ND HT-2 ND T-2 ND OTA ND Sterigmatocystin ND Alternariol ND Penitrem A ND Aflatoxin B1 ND Aflatoxin G1 ND Mycotoxins Year 2 Average Min Max Detection frequency (%) DON Enniantin B Enniantin B Fumonisin B Fumonisin B Marcfortine A ROQ-C Verruculogen Mycophenolic acid ZEA alfa-zol beta-zol Andrastin-A ADON ND Penicillic acid ND HT-2 ND T-2 ND OTA ND Sterigmatocystin ND Alternariol ND Penitrem A ND Aflatoxin B1 ND Aflatoxin G1 ND ND: not detected occurrence of ZEA and DON in corn silage has already been reported (Driehuis et al. 28; Tangni, Pussemier, & Van Hove 213). Both mycotoxins may occur in the field and then be present in silage. The guidance value for ZEA in complete feedstuffs (5 µg kg 1 ) was exceeded in two of the samples (68.7 and 82.2 µg kg 1 ). Furthermore, two derivatives of ZEA, α-zearalenol and β-zearalenol, were also present in the maize silages. Although detected in just three and four of the samples, respectively, high concentration values, ranging from 66 µg kg 1 to 2889 µg kg 1 for α-zearalenol, and from 326 µg kg 1 to 1721 µg kg 1 for β-zearalenol have been found. Whereas β-zearalenol shows less estrogenic activity than ZEA, it is accepted that α -zearalenol presents higher levels of estrogenicity (Keller et al. 215). Penicillium toxins, such as ROQ-C, mycophenolic, andrastin A and marcfortine A, are considered some of the most prevailing post-harvest metabolites found in maize silage and the second most common group of mycotoxins (Drejer Storm et al. 214; Gallo et al. 215). In our study, mycophenolic acid and ROQ-C were present in four and ten samples, respectively, with concentrations ranging from 56. to 3678 µg kg 1 for mycophenolic acid and 29.6 to 1651 µg kg 1 for ROQ-C. These two mycotoxins were shown to occur simultaneously in the four mycophenolic acid positive samples. A study conducted with 21 maize silages from ten dairy farms in Belgium (Tangni, Pussemier, Bastiaanse, et al. 213) addressed a similar situation. However, the levels of mycophenolic acid (up to 2 µg kg 1 ) were much higher than those observed in our current survey. Two other post-harvest mycotoxins, andrastin A and marcfortine A, have also been detected in our silages. In particular, andrastin A, with which 11.5% of the collected samples were contaminated at concentrations ranging from 9.1 to µg kg 1. Marcfortine A was detected in 7.7% of the silages at an average contamination of 392 µg kg 1, with a maximum level up to 244 µg kg 1. Regarding the fumonisins, FB1 and FB2 were detected in 9% and 21% of the samples, respectively; the average concentration being higher for FB1 (282.6 µg kg 1 ) than FB2 (58.8 µg kg 1 ). The occurrence of non-extractable forms of fumonisins was assessed in all these samples, by implementing an alkaline hydrolysis of the primary extraction residue (Latorre et al. 214). However, no hidden non-covalent forms could be observed in the silage samples, probably due to the low concentration of fumonisins. FB1 and FB2 were detected simultaneously in only five of the positive samples. The most common mycotoxin was enniatin B, found in 51% (two-year average) of the samples. In a similar study with Danish maize silage (Drejer Storm et al. 214) enniatin B was found to be one of the most prevalent mycotoxins, although less than in our survey with only 28% incidence. On the contrary, its analogue, enniatin B1, was detected in just 3.4% of our analysed samples. DON and andrastin A concentration levels were consistent with those detected in the Danish study. However, the average concentrations of enniatin B (157 µg kg 1 compared with 53 µg kg 1 ) and ZEA (226.7 µg kg 1 compared with 66 µg kg 1 ) were significantly higher in our maize silage samples. Foralltheassayedsamples,thelevelsoftworegulated mycotoxins (DON and FB1+FB2) were found to be below