Wageningen Academic World Mycotoxin Journal, 2016; 9 (5): 847-861 Publishers SPECIAL ISSUE: Mycotoxins in a changing world

Challenges in the analysis of multiple mycotoxins

J. Stroka1* and C.M. Maragos2

1Joint Research Centre, European Commission, Retieseweg 111, 2440 Geel, Belgium; 2Agricultural Research Service, National Center for Agricultural Utilization Research, United States Department of Agriculture, 1815 N. University St., Peoria, IL 61604, USA; [email protected]

Received: 17 January 2016 / Accepted: 15 June 2016 © 2016 Wageningen Academic Publishers OPEN ACCESS REVIEW ARTICLE Abstract

The problems associated with different groups or ‘families’ of mycotoxins have been known for some time, and for many years certain groups of mycotoxins have been known to co-occur in commodities and foods. Until fairly recently commodities and foods were analysed for individual toxins or groups of related toxins and attempts to measure multiple groups of toxins required significant investments in terms of time, effort, and expense. Analytical technologies using both the instrument-intensive techniques, such as mass spectrometry, and screening techniques, such as immunoassays, have progressed significantly in recent years. This has led to the proliferation of techniques capable of detecting multiple groups of mycotoxins using a variety of approaches. Despite considerable progress, the challenges for routine monitoring of multiple toxins continue. Certain of these challenges, such as the need for co-extraction of multiple analytes with widely different polarities and the potential for carry-over of matrix components that can influence the results, are independent of the analytical technique (MS or immunoassay) used. Because of the wide variety of analytical platforms used for multi-toxin analysis, there are also specific challenges that arise amongst the analytical platforms. We showed that chromatographic methods with optical detection for aflatoxins maintain stable response factors over rather long periods. This offers the potential to reduce the analytical burden, provided the use of a single signal receives general acceptance once shown in practise as working approach. This must however be verified by a larger community of laboratories. For immunosensors the arising challenges include the reusability of sensors and, for chromatography-based assays they include the selection of appropriate calibration systems. In this article we seek to further describe the challenges associated with multi-toxin analysis and articulate how such challenges have recently been addressed.

Keywords: mycotoxins, plant toxins, liquid chromatography, mass spectrometry, determination, biosensors

1. Introduction The mycotoxins generally accepted to be of the greatest importance to human and animal health, and to have the While mycotoxins likely pre-date human agriculture, it greatest economic impact are those belonging to several is only fairly recently that the nature of these compounds groups of toxins (‘families’): the aflatoxins, trichothecenes, and their prevalence in agricultural products have become fumonisins, ochratoxins, and zearalenones. known. Recent history has seen dramatic changes in agronomic practices, and further changes are expected Until fairly recently, the analysis of such toxins was done as climates change throughout the world. As previous individually or as families. The decision of which toxins contributions to this issue have noted, this will likely to test for was done based upon the fungi that might be lead to changes in the distribution, if not the type, of expected to be present in the commodity and the toxins mycotoxin contamination present in our food supply. Since which they might produce. This targeted approach is logical, the discovery of aflatoxins, mycotoxins of many different particularly because the costs for monitoring can be high, types have been discovered and analytical methods for especially in relation to the value of the commodity being their detection and quantification have been developed. tested. However, new analytical technologies such as high

ISSN 1875-0710 print, ISSN 1875-0796 online, DOI 10.3920/WMJ2016.2038 847 J. Stroka and C.M. Maragos

resolution mass spectrometry and biosensors now facilitate maize, in which most the above mycotoxins have been detection of multiple toxins in a relatively simultaneous reported to occur. Consideration should also be given to fashion. This can permit monitoring for larger numbers differences in consumption patterns in different regions of of potential contaminants, permitting a less targeted the world, which when combined with exposure estimates, approach for monitoring and revealing heretofore unknown may suggest a ‘one size fits all’ approach to regulation may potential hazards. The promise of the newer technologies not be appropriate for every region or climate. is greater efficiency in detection of hazards. Increasing such efficiency and reducing monitoring costs have the In parallel to regulatory developments, progress on the potential to permit increased monitoring, further reducing availability of technologies that allow for the simultaneous the potential for exposure- a positive feedback loop. An determination of mycotoxins has occurred. On one hand additional benefit of testing for many compounds is that new immunochemical tools have progressed, allowing they provide prevalence data that may be serve as criteria the application of easy-to-use systems for end-users. for putting a new compound on the list of those for which On the other hand improved mass selective detectors targeted analysis is appropriate. have become available in many analytical laboratories, replacing to a certain extent the classical (optical) detectors The potential benefits of multi-toxin detection are clear. for mycotoxin identification and quantification by liquid There are, however, challenges associated with multi-toxin chromatography. While immunochemical methods are analysis. These include resolving conceptional and societal rather specific for a certain mycotoxin or a group of closely issues as well as the inherent technical challenges. Detecting related mycotoxins, modern mass-spectrometers can be large numbers of potential metabolites and toxins was applied to target structurally different mycotoxins in a once limited by technical constraints, but increasingly single method. Antibody-based techniques and mass the technologies to detect greater and greater numbers spectrometry are not mutually exclusive, and using them of compounds in complex commodities and foods are together (e.g. immunoaffinity column clean-up followed becoming more widely available. This has created a by LC-MS/MS) is a potent combination for assuring very central philosophical challenge to analysts: how many (and high selectivity. which) toxins to include in the test panel. In this case, social considerations both help and hinder the process for As noted, some guidance in selecting targets is obtained target selection. In a very real sense, societies declare which through the need to meet regulatory requirements. problems they regard as significant by exercising control However, the application of broader methods has the over them (regulation). Hence the presence of regulatory potential to allow for the identification of previously levels helps to determine which toxins are important targets. unknown food safety risks (Sulyok et al., 2010). This is particularly true for commodities having established A fundamental issue in multi-toxin detection is which toxins problems with known toxins. If the food supply were a should be the focus in multi-toxin tests. Certainly, the static system, limiting targets to regulated toxins might be main drivers for mycotoxin testing in commercial routine sufficient for adequate control and protection. However, the testing as well as food or feed law enforcement laboratories food supply is a very dynamic system with a tremendous are the legal obligations that apply, or trade specifications. number of variables including the environment, agronomic As a matter of fact, until the first half of the 1990’s it was practices, post-harvest treatment, processing, and shipment primarily aflatoxins, and little more, that were routinely – often to destinations far from the source of production. screened for. In 1998 the first EU legislation on aflatoxins As discussed elsewhere in this special issue of World and ochratoxin A (OTA) came into force with Commission Mycotoxin Journal, climate change may also be a factor Regulation (EC) No 1525/98 (EC, 1998) and until then the leading to a shift of fungal populations and consequently main detection techniques for liquid chromatographic a shift of mycotoxin patterns. Furthermore, new toxins methods were based on the determination of the natural emerge, and precursors and metabolites of established fluorescence of both, the aflatoxins and OTA. Now, about toxins (e.g. the ‘masked’ respectively ‘modified’ mycotoxins) two decades later, the legal framework of mycotoxin are increasingly found. The result is that the number of regulations in the EU has widened, to include those mycotoxins regarded as relevant has increased over time. previously mentioned, as well as deoxynivalenol (DON), Nonetheless, the process of looking for new possible patulin, , (ZEA), fumonisins (EC, 2006a) threats to human or animal health should be approached and T-2 and HT-2 toxins (EC, 2013). A rather unfortunate with critical reflection. Perspectives must include the circumstance is the lack of worldwide harmonised best approach to achieve protection in a cost effective regulations. While a number of countries or regions have way. This may require thinking about ways to reduce the set legislation, the extent of the toxins regulated, as well as analytical burden without compromising the intended the limits set, differ (FAO, 2004). A number of agricultural goal. Or, further, to improve service without the need for products are known candidates for the presence of several redirecting resources. One approach that is currently used of the above mentioned mycotoxins. A good example is is a two-stage process where sample screening is conducted,

848 World Mycotoxin Journal 9 (5) Challenges in the analysis of multiple mycotoxins

often with an immunoassay, followed by dedicated analysis toxin are necessary depends upon the sensitivity and of samples near or above the regulatory limits, often using selectivity of the technology that will be used in the final a quantitative chromatrographic method. determinative step (i.e. detection). Depending upon the detection technology the presence of remaining matrix is The preceding discussion has assumed that detecting addressed in a number of ways. Often these include simple multiple toxins is in the best interests of society. From dilution, matrix-matched calibration, or the use of isotopic the standpoint of human food and animal feed safety, the internal standards. The solutions for many of the issues are assumption may be valid. However, society also bears the dependent upon the technology used and are discussed increased financial burdens associated with monitoring. in more detail in the technical sections that follow. In the These include not only the direct costs of the testing, but remainder of this article the current aspects that need to be the costs to commodity producers whose products lose considered by analysts, with respect to the determination value when toxins are detected. In fairness to producers, and evaluation of results from multi-toxin methods, are commercial decisions should be based upon the levels of viewed within the context of recent developments those compounds for which there is consensus agreement on the potential for hazard, and for which regulations have 2. Challenges and new aspects using been established. While there is currently no consensus on chromatography based multi-toxin detection the target analytes, a logical starting point for a list would assays include known toxins and compounds structurally similar enough to known toxins as to potentially contribute to Is less sometimes more effective? exposure. Certain mass spectrometric technologies allow for quantitative determination of selected analytes (targeted In the field of mycotoxins the most frequently analysed analysis), and qualitative screening for additional analytes group of mycotoxins are certainly aflatoxins. In addition (non-targeted analysis). to (AFB1) alone, aflatoxins B2 (AFB2), G1 (AFG1), and G2 (AFG2) are usually also regulated as a Once targets have been selected, what remains are the group. Certainly the aflatoxins co-occur as contaminants, technical challenges associated with accurate detection, but they usually follow a certain pattern. The United States quantification and interpretation of the results. As will be Department of Agriculture – Grain Inspection, Packers and described later in this article, there are many analytical Stockyards Administration (USDA-GIPSA) for example platforms for multi-toxin detection. However despite very recognises this, and, as a result, requires producers of different platforms, there are challenges in common among kits that measure ‘total’ aflatoxins to consider a ratio of them. Many methods follow a process involving sampling, 10:1:1:1 for AFB1:AFB2:AFG1:AFG2 for their test kits extraction, clean-up, and detection. Technical challenges (USDA, 2016). In a few cases AFB1 alone is regulated, such are encountered at each of these stages. The first step is as in the European Union (EU) for animal feed, as well as sampling, which is thought to be the largest contributor cereal-based foods for infants and young children (EC, to the uncertainty of the compliance/non-compliance 2002b, 2006a), while a number of countries around the statement made upon an analytical result. world have set regulation for AFB1 alone, others regulate the sum of AFB1, AFB2, AFG1 and AFG2 or police both While some technologies, such as rapid imaging techniques, parameters (FAO, 2004). In this in depth investigation on do not rely upon extraction, most analytical methods world-wide regulations for mycotoxins the author already attempt to separate the toxins from the bulk of the matrix stresses the different views on the need to control beyond through extraction. For testing of toxins within individual AFB1 by mentioning: ‘Whether a regulatory level for the families, significant effort has been placed on finding the sum of the aflatoxins, which requires more analytical work optimum extraction conditions. However, the wide range than for aflatoxin B1 alone, contributes significantly to of polarities of the mycotoxins, from the highly polar better protection of public health than a regulatory level fumonisins to the zearalenones complicates the selection for aflatoxin B1 alone is debatable’. However, until now no of an appropriate extraction solution. Selection of an ambitions to harmonise aflatoxin regulation in this respect extraction solution is also influenced by societal forces, in are in discussion. In the following an approach offering particular the desire for ‘greener’ technologies that reduce lesser analytical burden, while aiming at maintaining the solvent consumption and the generation of hazardous maximum quality of information is presented. waste. Once extracted, many analytical platforms require either the removal of much of the remaining matrix (to Today, the most frequently used confirmatory method for minimise interferences) or concentration of the sample aflatoxin determination is still liquid chromatography with (to boost analytical signal). This step, the isolation and fluorescence detection (LC-FLD). The technique allows clean-up, is often tailored to dovetail with the particular the detection of all four aflatoxin analogues individually in analytical platform being used for detection. Typically a single assay. This, as a consequence, currently requires the extent to which isolation and concentration of the analysts to maintain reliable calibration conditions for

World Mycotoxin Journal 9 (5) 849 J. Stroka and C.M. Maragos

all four aflatoxins. Furthermore, to demonstrate this datasets are only an excerpt of the full data available within is a key element for ISO 17025 accreditation of testing RASFF. However, as data must be extracted manually, the laboratories (ISO, 2005). In a proficiency test organised selected set was considered as sufficiently representative for by the European Reference Laboratory for Mycotoxins a correlation between AFB1 and total aflatoxins. Unsound in 2007, it was demonstrated that the aflatoxin calibrants data was removed (e.g. instances where AFB1 levels were used in laboratories were a critical element that, when not reported as above those of total aflatoxins). Analytical under control, contributed to a large deviation of results. results below the legislative limit within one notification In short, the dispersion of results between laboratories was were considered as well. Data was kept pooled across all worst for AFG1, with a relative standard deviation (RSD) food types to follow a pragmatic approach common for of 23%, and best for AFB2, with an RSD of only 9% when all notifications. a laboratory’s own calibrants were measured against a common calibrant of known quality. This finding actually An evaluation of the RASFF data comparison compliance reflects the stability of aflatoxins towards degradation scenarios on the basis of AFB1 vs total aflatoxins was in solution (Stroka et al., 2007). Therefore AFB2 can be conducted using three situations. These situations reflect considered to be the most stable calibrant during practical the levels set in the European Union (EC, 2006a) for food laboratory activities, followed by AFB1 (RSD 14%) and (with the exception of almonds, pistachios and apricot AFG2 (RSD 15%). Additional in-house data over the years kernels to be subjected to sorting/processing), namely: 2008-2013 (not published) show that, as long as the same (1) AFB1 at 8 µg/kg and total aflatoxins at 15 µg/kg; (2) instrumental system was used, the relative responses of AFB1 at 5 µg/kg and total aflatoxins at 10 µg/kg; and (3) signal-to-mass-fraction remained reasonably constant AFB1 at 2 µg/kg and total aflatoxins at 4 µg/kg. For the first between aflatoxins. However, between different instruments situation (regulatory limits 8 and 15 µg/kg), 118 out of the as well as derivatisation devices, even of the same type, 4,774 data sets exceeded total aflatoxins (>15 µg/kg) but ratios differ significantly. This implies that even though not AFB1 (<8 µg/kg). This is a rate of 2.5%. Of this fraction signal-to-mass-fraction ratios are specific for a particular the average was calculated as 26 µg/kg and the median HPLC system, they were very constant. This observation was 21 µg/kg total aflatoxins. For the second situation raises the question whether it is essential to determine (regulatory limits 5 and 10 µg/kg), 139 out of the 4,774 aflatoxins by individual calibration curves, provided data sets exceeded total aflatoxins (>10 µg/kg) but not for sufficient means for identification of peaks (e.g. relative AFB1 (<5 µg/kg), that is a rate of 2.9%. Of this fraction the retention and peak width) are available. average was calculated to be 19 µg/kg, and the median was 14 µg/kg total aflatoxins. For the third scenario (regulatory In this context is it worth looking at aflatoxin data in the limits 2 and 4 µg/kg) 89 out of the 4,774 data sets exceeded European Commission’s Rapid Alert System for Food and total aflatoxins (>4 µg/kg) but not AFB1 (<2 µg/kg), that is Feed (RASFF). This database lists reported non-compliances 1.9%. Of this fraction the average was calculated to be 9 µg/ across the EU that are relevant for the common market kg, and the median was 7 µg/kg total AF. In other words, (EC, 2016). For food products classified as unfit because if only AFB1 would have been regulated this limit would of aflatoxin, 4,774 datasets in the period 2004 to 2007 were cover up 97.5%, 97.1% respectively 98.1% of all notified analysed for AFB1 and total aflatoxins (Figure 1). These cases. For the misclassified datasets the total aflatoxins content calculated to 1.4, 1.4 and 1.9 times the limit set in legislation (based on the median). This simulation shall not question any need monitoring for total aflatoxins at all, as it has been reported recently that there are regional incidences where the pattern of aflatoxins differ from what is usually observed (Matumba et al., 2014).

Nonetheless, it must be taken into account that the

RASFF notifications represent only a small fraction of the

overall number of analyses performed during aflatoxin monitoring. Further, the rate of non-compliance does usually – on average – not exceed 5%. Therefore it might be rewarding to consider whether it is fit-for-purpose to use a calibration system based upon a single calibrant. The

A calibrant chosen could be AFB1, based upon the price of the aflatoxin and the fact that AFB1 is regulated as such, Figure 1. Double log plot of 4,774 RASFF entries for aflatoxin alternatively AFB2 is the most stable AF as shown before. B1 (AFB1) reported together with total aflatoxins (as sum of Such an approach has potential for testing laboratories using aflatoxin B1, B2, G1 and G2) for food in the period of 2004-2007. LC-FLD cutting costs while not jeopardising food safety.

850 World Mycotoxin Journal 9 (5) Challenges in the analysis of multiple mycotoxins

To facilitate a deeper look into the matter the Institute for quantification. They used either a less abundant PA (Kempf Reference Materials and Measurements recently launched et al., 2011), with a different backbone structure, as the an open enquiry on its website (IRMM, 2016) calling for internal standard or an isotopically labelled retronecine the collection of calibration data from a larger pool of (Cramer et al., 2013). These are certainly rather special laboratories. For laboratories using mass spectroscopy cases, but illustrate the effectiveness of such approaches. (MS) this simplified approach was not followed because of In a PT carried out for PA in honey and hay materials anticipated issues with controlling matrix effects between (Tamosiunas et al., 2013), laboratories used the sum samples. This difficulty was demonstrated in a proficiency parameter method proposed by Kempf et al. (2011) as well test that highlighted that matrix-matched calibration, or as single parameter methods. Interestingly, independent calibration using isotopolouges, were essential for precise from the methods used, participants found comparable mycotoxin estimation by LC with mass selective detection levels in artificially fortified samples (containing a defined (De Girolamo et al., 2013). and limited number of commercial available PA). A different scenario occurred for naturally contaminated samples, Similar scenarios surely exist for several other areas of which likely contained a more complex composition of PA. contaminant determination. For example the measurement Data from the naturally contaminated samples (Figure 2), of ergot alkaloids (EA) requires determination of both demonstrated that methods which determined the sum the ‘-in’ and ‘-inin’ forms for each alkaloid, as both forms parameter (as reduced to a single measurand) tended to have been identified in extracts and therefore need to report higher values relative to those reporting individual be considered. Both forms of EA are characterised by a PAs. In a study that followed, These et al. (2013) identified stereoisomeric difference on the position 8 of lysergic acid. 121 new PA in plant material, substantiating the validity of A lack of the calibrants for the ‘-inin’ forms in a proficiency the differences observed in the PT. test (PT) (DLA 15-2011 Ergot-alkaloids in cereal product) led to the pragmatic approach of assuming equivalent The above PA scenario reflects a situation where the sum fluorescence response factors for both forms (in order parameter represents a single measurand and not the to be able to report figures for the sum of all EA). This arithmetic sum of individual parameters. The arithmetic was based on the assumption that the structure of the sum is far more challenging for the interpretation of results, corresponding racemates did not influence the fluorescence especially when the occurance of analytes do not follow response. The z-scores obtained in the study indicated that a given pattern. For example, European legislation (EC, this approach lead to sufficiently valid results. The z-scores 2006a, 2013) regulates the sum of individual parameters, for the combined parameter (‘-in’ and ‘-inin’ form of each such as fumonisins B1 (FB1) and B2 (FB2), as well as the EA) were -0.9, 0.3, -0.6, -1.1, -0.8, -0.6, for ergometrine, sum of T-2 and HT-2 toxins, as it does for the sum of AFB1, ergosine, ergotamine, ergocornine, ergocryptine, and AFB2, AFG1 and AFG2 in foods. This requires that the ergocrystine, respectively. For the sum of all EA the z-score combined result must take note of the individual limits was -1.0, with a reported value of 3,352 µg/kg being rather of quantification (LOQ) for each contributing toxin. close to the consensus value of 3,853 µg/kg calculated by Further, in order to avoid misinterpretation for exposure the PT provider. Despite the fact that response factors are assessments, a harmonised procedure for combining these not exactly the same, they were sufficiently ‘equivalent’ for values, namely applying either ‘upper bound’ or ‘lower quantification. Thus such strategies remain an interesting approach for simplifying multi-analyte methods. A general validity of the approach was however not further investigated. Simple and fit-for-purpose calibration approaches that prove to yield equivalent precision and trueness compared to complex ones are of strong interest. This is especially true in instances where calibrants are subject to temporarily unavailability, where prolonged periods are common for their acquisition or in instances where there is a limited shelf life. μ PA

Where the sum is more than the collection of the individual parts S Other examples where simplification can bring N improvements are the determination of pyrrolizidine alkaloids (PA) (Cramer et al., 2013; Kempf et al., 2011). Figure 2. Amount of total pyrrolizidine alkaloids reported based Investigators managed to reduce the most toxicologically on the number of pyrrolizidine alkaloids (PA) in the scope of relevant PA to a common backbone structure for the method for naturally contaminated plant material.

World Mycotoxin Journal 9 (5) 851 J. Stroka and C.M. Maragos

bound’ reporting is important. The analytical impact of from collaborative trials published after 1997 for mycotoxins multi-analyte challenges for mycotoxins becomes apparent at levels below 100 µg/kg. Focusing on this data he proposed recognising the number of ZEA metabolites discussed an adjustment for the mass fraction range below 120 µg/kg. for a health-based guidance value in the recent scientific As a result he proposed truncating the Horwitz function, opinion by the European Food Safety Authority (EFSA, to 22% RSDR as the target standard deviation for a HorRat 2016). Furthermore, this particular scenario with more than of 1. The data used by Thompson were derived from single 15 different ZEA metabolites appears to be a good candidate analyte methods or those closely related (such as aflatoxins). for biosensors based estimations provided a sufficient cross This led to the question what performance can be expected reactivity to the relevant metabolites is achievable. for multi-analyte methods. These usually cannot follow analyte specific procedures but require compromises in The re-cast of Commission Regulation (EC) No 1881/2006 extraction, clean-up, or dilution to allow a simultaneous (EC, 2006a) that is due in 2016 will address such issues and determination of mycotoxins with different chemical provide the basis for proper interpretation of analytical properties. Another important issue is the working range results in the future. However, proper interpretation of of such methods for each analyte. A current example for analytical finds requires effort on the part of the analyst this is found in the EU legislation for animal feed. While and remains a burden and can lead to disputes. This AFB1 is regulated at 2 µg/kg in feed for dairy cattle (EC, remains a challenge in cases where no fixed guidelines 2002b), a regulatory level 30,000-fold higher (60 mg/kg) exist for signal interpretation, such as for plant toxins. The pertains to the sum of FB1 and FB2 (EC, 2006b). Certainly question that remains, when using fragmentation in order such high levels of fumonisins are rather uncommon, to identify substances beyond a reasonable doubt, is how however, a multi-mycotoxin method targeting not only specific the fragmentation needs to be. In Europe, clear the ‘presence’ of mycotoxins, but aiming to serve as tool for requirements exist for residues in products of animal origin compliance testing, needs to address these large working (EC, 2002a). Guidance for mycotoxins is being developed. range difference between analytes. This self-evident results The basis for such guidance was presented by the chair mostly in the strategy of incorporating extract dilutions of the working group on identification criteria within the to stretch the working ranges. In addition this also limits network of European mycotoxin reference laboratories the amount of mycotoxins needed during calibration an during a workshop held in 2014 (Mol, 2014). important pre-requisite for limiting costs when working with isotope labelled mycotoxins. Method performance guidance for multi-analyte methods – Quo vadis? A recently finalised collaborative trial for the determination of AFB1, OTA, ZEA, DON, T-2, HT-2-toxin, FB1 and FB2 in The system for predicting method performance developed compound animal feed by LC-MS followed the approach of by Horwitz et al. (1980) still serves as a good benchmark extract dilution to cover a larger working range (Kujawski for newly developed methods and is widely accepted. More et al., 2016). The obtained precision data was used to recently, Thompson (2000) re-processed performance data investigate how such a method performs in the context of expectable method performance. Furthermore a similar collaborative trial study on DON, ZEA, T-2 and HT-2 in animal feed and cereals by LC-MS (Breidbach et al., 2013) S S was taken for this method performance investigation. The S S reproducibility (RSDR) values were plotted against the mass fraction for each mycotoxin for visual inspection (Figure 3). The plot visually corroborates the Horwitz prediction for reproducibility. Nonetheless, it must be noted that this plot,

S even though yielding many precision parameters, is only based on two studies that used a rather straight forward ‘dilute and shoot’ approach, similar to what other authors proposed (Al-Taher et al., 2013; Mol et al., 2008; Sulyok et al., 2007; Varga et al., 2012). In order to further investigate how standardised methods in the EU fit into this picture, and to take advantage of Figure 3. Repeatability (RSDr) and reproducibility (RSDR) data collaborative trials published after the proposed truncation from two LC-MS multi-mycotoxin methods for aflatoxin B1, of the Horwitz curve by Thompson (2000), mycotoxin /B2, deoxynivalenol, zearalenone, ochratoxin A methods published by the European Committee for or T-2/HT-2 toxins test in a collaborative trial and subject to Standardization (CEN) as standards (CEN, 2001, 2004, standardisation by CEN. 2007, 2009a,b,c, 2010a,b, 2011a,b,c) were analysed. The

852 World Mycotoxin Journal 9 (5) Challenges in the analysis of multiple mycotoxins

resulting plot (Figure 4) is characterised by two clouds of as the highest point for OTA and two highest values for dots, the left one representing methods for aflatoxins and ZEA were removed, as these indicate measurements outside OTA in a lower µg/kg range (in food) and the right cloud the working range. The t-test for comparison of means mycotoxins that occur usually in higher mass factions like, suggests that they are equivalent (19.2±1.7 for <120 µg/kg ZEA, DON and fumonisins (in food and feed). vs 18.6±1.3 for >120 µg/kg). The comparison of standard deviations comes to the same conclusion indicating they are It must be noted that most of the single analyte (or single statistically equivalent (39.1% relative standard deviation group) method performances (Figure 4) are based on for RSDR values <120 µg/kg vs 35.7% for those above). This an analyte specific clean-up and enrichment procedure finding shall not undermine the overall prediction made (e.g. immunoaffinity clean-up) and rather robust optical by Horwitz. However, it must be taken into account that detection systems where it was not necessary to optimise modern LC-MS methods are developed with the aim of parameters in order to obtain sufficient signal quality. diluting sample solutions for highly abundant mycotoxins However, the multi-mycotoxin methods that used LC-MS (such as fumonsins or DON) to mass fractions that permit (Figure 3) made use of stable isotope analogs, an important economic use of calibrants. Therefore, evaluation concepts tool to improve method performance, as previously should consider that instrumental measurements are usually concluded by De Girolamo et al. (2013). In addition to made at lower concentrations in the injection solution. these specific method features, it must be noted that in all cases the calibration range was optimised to the lowest mass Another aspect in favour of a fixed, rather than a dynamic, fraction possible not jeopardising method performance. guidance value for RSDR for multi mycotoxin methods is This means that the calibration ranges did not consequently that, independent from the amount of mycotoxins present correlate with instead of between mycotoxin levels in test (either >10,000 µg/kg or <100 µg/kg), all analytes will have material. As a result, highly contaminated samples had to undergone the same method principle. This, however, does be diluted and re-analysed to allow the analytical working not mean that there are no mycotoxin specific properties range to be covered. influencing the individual measurement capacity. A closer look at Figure 5, shows that for each mycotoxin plotted Figure 5 shows the summarised RSDR data points shown in alone specific correlations for the RSD and mass fraction Figures 3 and 4 for individual mycotoxins over the whole exist with values above 44%, reflecting the individual end mass fraction range (<1 µg/kg to >10,000 µg/kg). The visual of the lower working range, that is, values that reached or evaluation allows the conclusion that a method performance fell below the LOQ for a mycotoxin (in some or all of the guidance of 22% reproducibility (as previously proposed by laboratories). This challenge to include contamination Thompson for levels below 120 µg/kg) actually appears to levels at the expected LOQ for the majority of participants also be a practical guidance for levels higher than 120 µg/kg. was part of the design for the two multi-mycotoxin studies, This is further supported by statistical analysis comparing as the ambition was to stress the method capacity and the standard deviation and the means of the reproducibility demonstrate it experimentally. values below vs above 120 µg/kg. For this comparison, the two extreme values above 44% for T-2/HT-2 toxin as well This performance observation is also in line with a similar performance requirement mentioned in Commission Regulation (EC) No 401/2006 (EC, 2006c), the ‘fit-for- purpose’ function. This function specifies maximum S S S S acceptable levels of standard uncertainty (k=1) for different contaminant concentrations in relation to the limit of detection (LOD) of the method used. Compliance with this criterion has to be demonstrated for single laboratory

validation and is intended to be used for those cases where S collaborative trial data is lacking. The function converges to approx. 20% acceptable standard uncertainty till 500 µg/ kg if measurements are done at 10 times or higher the LOD of the method.

The aspect in the study design of Kujawski et al. (2016) extending the working range as much as possible to lower concentrations aimed making the method a useful tools Figure 4. Repeatability (RSDr) and reproducibility (RSDR) data for verification of compliance, as well as for estimating from mycotoxin specific methods of analysis (aflatoxin B1, exposures based upon monitoring data limiting the fraction fumonisin B1/B2, deoxynivalenol, zearalenone, ochratoxin A of left censored data in exposure assessments. Therefore, or T-2/HT-2 toxins), standardised by CEN. organisers of future collaborative trials might consider

World Mycotoxin Journal 9 (5) 853 J. Stroka and C.M. Maragos

A O A S

Figure 5. Summary of the reproducibility (RSDR) results from Figure 3 and 4 visualised for performance comparison based on the type of mycotoxin for both, multi-mycotoxin and single (respectively specific mycotoxin group) mycotoxin methods.

the general benefit of including levels that fall below the modern multi-mycotoxin methods has only an upside for previously established ‘robust’ working range into the precise quantification has to be individually judged. A study design. In conclusion, a general guidance value of critical issue extracting aflatoxins with acetonitrile-water 22% for reproducibility could simplify a multi-mycotoxin has been highlighted by Stroka et al. (1999). method evaluation by applying a pragmatic performance guidance value over a large concentration range. In any Other approaches include a more specific immunoaffinity case, sound method evaluation always requires expert clean-up (Tanaka et al., 2010) and have been nicely judgment, as any guidance value only serves as a general described as ‘Heracles battling the Hydra’ (Kokkonen, 2011). rule for benchmarking. Desmarchelier et al. (2014) conducted a study in which they compared the effect of sample clean-up procedures How much clean-up is enough for LC-MS? with the aim to improve analytical results. In a similar way, Gonçalves and Stroka (2016) compared immunoaffinity The strategies for clean-up are a very broad field and with a generic solid phase extraction (SPE) clean-up for include ‘dilute and shoot’ approaches (Malachova et al., the LC-MS determination of compound animal feed. The 2015) and other clean-up procedures like for example authors showed that the LOQ of an LC-MS method for 8 QuEChERS (Lacina et al., 2012). While QuEChERS was mycotoxins (AFB1, FB1, FB2, DON, ZEA, OTA, T-2 and initially developed for pesticides, specific modifications HT-2 toxins) in difficult matrices, such as compound animal have been made depending on the mycotoxins targeted. feed, could be lowered by a factor of about 20 using an IAC Sospedra et al. (2010) used it for type-A and type-B clean-up. However, immunoaffinity clean-up has shown little trichothecenes following the original procedure proposed additional benefit for test materials that have lower matrix by Anastassiades et al. (2003), while Zachariasova et al. effects, e.g. cereals. The improved measurement capacity (2010) included fumonisins in their method scope and was thought to be result of matrix effects, and the resulting subsequently had to exclude the use of the primary- injection-to-injection variability, adding significantly to the secondary-amine (PSA) sorbent that would compromise overall dispersion signal magnitudes of repeated analyses the quantitative determination of the rather polar at lower levels. fumonisins. When including a wide range of mycotoxins into a method scope this requires limiting the specificity The dilemma of ‘too much’ information of the clean-up, which on one hand still can reduce the total load (e.g. oligopeptides) onto the LC column while Certainly, LC-MS methods designed for multiple analytes the purification effect beneficial for interference free have the advantage of affording the possibility of a wider scope mycotoxin quantification will remain limited. In this respect of analytes with the same or only little additional analytical the question whether a forced separation of acetonitrile- effort. This has been demonstrated nicely by Sulyok et al. water mixtures used during extraction in most of the (2010) who identified numerous compounds other than the

854 World Mycotoxin Journal 9 (5) Challenges in the analysis of multiple mycotoxins

regulated mycotoxins in spoiled food products. However, this biosensors is very small. This highlights the challenges feature can also have a flip side depending on the point of view. associated with preparing and using multiplexed biosensors. Contract laboratories that use multi-toxin methods, especially Biosensors do not directly measure a physical property those that also act as contractors for official control, might of a toxin, rather they indirectly measure the effect of a run into a situation where more, potentially relevant, results toxin upon a recognition element. Two common formats (not explicitly requested by the customer) are revealed. This for binding assays, immobilised antigen and immobilised is not an analytical issue, but certainly an aspect laboratories antibody are depicted in Figure 6. In the figure the binding applying multi-analyte LC-MS multi-methods need to be element is represented by an antibody, but other types aware of. As a result there is a need to establish clear guidelines of recognition elements could also be used. Recognition on how to deal with it by regulators, as during laboratory elements include, but are not limited to, polymers made from audits dilemmas are likely if no official rules are set. This amino acids (peptides, antibodies, and their fragments), is certainly the case where potential carcinogens (e.g. like nucleic acids (aptamers), carbohydrates (cyclodextrins) sterigmatocytin) are co-identified at levels of concern while and synthetic subunits (molecularly imprinted or non- consignments would classify complaint for the initial target imprinted polymers) (Bazin et al., 2012; Choi et al., 2009; mycotoxin(s) asked for by the customer. Liu et al., 2012; Ton et al., 2015; Wang et al., 2013a). The use of recognition elements other than antibodies opens 3. Challenges in multi-toxin analysis using up the possibility of materials with greater stability and biosensors therefore potentially useful under conditions that might denature antibodies and interfere with immunoassays. The In stark contrast to the wide availability of many commercial development and selection of new recognition elements products for single mycotoxin testing (ELISAs, lateral has the potential to expand the utility of ‘immunoassays’ flow strips, etc.), the number of commercial multi-toxin into novel applications.

A

S C

S C A

Figure 6. Two common assay formats widely used in mycotoxin biosensors. The recognition component depicted here is an antibody. (A) Immobilised recognition element (a.k.a. ‘direct’ format). (B) Immobilised toxin or toxin-protein conjugate (a.k.a. antigen immobilised or ‘indirect’ format). Depending upon the exact assay format, additional steps (washing, addition of substrate or stop reagents etc. may be needed).

World Mycotoxin Journal 9 (5) 855 J. Stroka and C.M. Maragos

The fact that biosensors are indirect detection methods strength, solvent strength, and the presence of sample introduces certain challenges that are different from components that modulate the interaction between the methods, such as LC-MS/MS. Some of the challenges toxin and the antibody. Matrix interferences can cause associated with biosensors are not unique to biosensors, artificially high, or low, responses leading to false positive but are rooted in the fact that they are based upon or false negative results. As with instrumental assays, there biological recognition and are therefore susceptible to are multiple ways to reduce matrix effects. The simplest is some of the same issues as traditional immunoassays. dilution, which is possible provided the assay has sufficient While immunoassays have been used successfully for sensitivity. In quantitative assays centrifugation or filtration years, there are challenges that must be addressed during of sample extracts, followed by dilution, is often combined their development and application. Chiefly these fall with matrix-matched calibration (Anderson et al., 2010; into three categories: sensitivity, selectivity, and matrix Beloglazova et al., 2014; Oswald et al., 2013; Peters et al., effects. Sensitivity and selectivity are often addressed 2013, 2014; Wang et al., 2013b; Zangheri et al., 2015). during development of the recognition element, typically Unfortunately, with qualitative or semi-quantitative assays, an antibody. For multiplexed assays there is another aspect literature reports often neglect to describe whether the cut- to selectivity: namely, the possibility for cross-reaction off values were established with toxin standards in buffer between assays for the individual toxins (e.g. does FB1 or in matrix. The problems caused by sample matrix are interfere with the assay for ZEA, etc.). The recognition well known to most analysts. However, not all matrices are elements used in multiplexed biosensors must be evaluated equivalent and validation in a cereal grain, for example, is for cross-reactivity not only towards congeners of the target not equivalent to validation in silage. Usage on matrices for toxin, but also to the other toxins within the multiplexed which test kits have not been validated is a known problem. system. This is an additional validation that is often not There is every reason to believe that multiplexed biosensors, conducted for assays that detect single, or closely related, as they become more widely used, will also face this issue. toxins. The degree of selectivity (the inverse of cross- Manufacturers of commercial test kits address the problem reactivity) towards closely related toxins is determined by through specific directions within their instruction sheets the application. For example, because of how the antibodies and through appropriate training. The same, if not greater, were produced many DON immunoassays are highly training and awareness will be needed for the application sensitive to either 3-acetyl-DON (3-ADON) or 15-acetyl- of multiplexed biosensors. DON (15-ADON) (Maragos et al., 2006; Sanders et al., 2014). If the purpose of the assay is to quantify only DON, Because it impacts the ability of a sensor to be re-used, the presence of these congeners may lead to an inaccurate the non-specific adsorption of matrix to either the sensor measurement. However, if the purpose is to qualitatively surface or to the reagents (fouling) represents a major or semi-quantitatively screen for DON then the selectivity challenge. It is this aspect which is likely responsible for may be acceptable or even beneficial. As climates change the large disparity seen between the number of mycotoxin there may be alterations in the patterns of distribution of biosensor formats that have been reported as ‘proof of toxigenic fungi, with resulting changes in the patterns of principle’ in buffered solution and the number that have distribution of their toxins. For example, should the existing been applied to actual commodities or foods. While there are distributions of 15-ADON and 3-ADON producing strains many potential solutions to fouling the two most common of F. graminearum change, there may be a corresponding are to either (1) make the sensors single-use (disposable) need to change the immunoassay screening that is used. or (2) replenish the reagents that have been consumed. These aspects highlight a central insight into bioassays: Examples of disposable devices include multiplexed lateral whether cross-reactivity is beneficial or not is determined flow devices (LFDs) and certain electrochemical sensors by the local application (purpose). Fortunately, it is relatively based upon inexpensive electrodes. Disposable devices, easy to tailor the selectivity of immunoassays, which such as LFDs are often (but not always) qualitative, and provides the flexibility to accommodate such changes as represent an economical solution. Rather than dispose of the need arises. the entire device, many of the more instrument-intensive technologies re-use the device and simply replace reagents The carryover of matrix into the test solution is another that have been consumed. This solution is often used challenge common to all recognition-based assays. In with the microbead (suspension array or planar array) everyday parlance ‘matrix effects’ refers to the effect of formats. Certain of the formats go a step further and re-use residual sample on the results of the assay. Whether a the sensor for multiple tests by regenerating the surface particular assay is susceptible to matrix effects depends between assays. However, in all of the multiplexed assays upon a host of considerations and is not, as commonly the device is either discarded after single use, or reagents believed, solely due to the quality of the antibody used. are replenished in some fashion. For biosensors, the term ‘matrix effects’ actually embodies many different factors that can contribute to making the Significant effort has been made to address the challenge of assay inaccurate. These can include extremes of pH, ionic developing re-usable, quantitative, assays for multiplexed

856 World Mycotoxin Journal 9 (5) Challenges in the analysis of multiple mycotoxins

analyses. The underlying format of the assay affects the been overcome. Biosensors based upon enzymatic labels ease with which it can be made re-usable. Formats with generally include washing steps to remove unbound label the recognition element immobilised can be more difficult and matrix that might interfere with enzymatic activity. to regenerate than antigen-immobilised formats (Figure Sensitive and inexpensive optical detectors have also 6B). The reason lies in the strong solution conditions such contributed significantly to the widespread acceptance as pH extremes or solvent strength, needed to disrupt of these methods. In multiplex environments there has the binding interaction. To maintain activity, antibodies been a trend to miniaturise the ELISAs and, as noted that have been denatured must be re-natured. The same earlier, a commercial assay has used this approach using a is not necessarily true for immobilised toxin-protein coloured end product. However, ELISAs can also be used conjugates, which may still display the bound toxins even to generate electrochemically active end products, which when denatured. For antibody-immobilised formats fresh can be detected by a number of techniques, including those ‘toxin’, in a form that is generally labelled, is provided with using amperometric, potentiometric, condimetric, and each iteration of the assay. That is, the antibody is either impedimetric approaches. Progress in electrochemical reused or it is removed from the surface and replenished. sensors, for which there is an extensive literature, was The latter requires replenishing two reagents: the antibody recently reviewed (Vidal et al., 2013). A challenge to and the labelled toxin. For antigen-immobilised formats certain of the electrochemical devices can be the presence the antigen is reused and the antibody is replaced with of materials, other than the enzymatic reaction products, each iteration. It is easier to maintain consistent activity that can affect the redox status of the solution and therefore by providing fresh antibody. This is one reason why most affect the signal, which is a form of matrix effect. To reduce of the regenerable biosensors use an immobilised antigen the time required for the substrate incubation step, many format. However, there are widespread efforts underway to assays use non-enzymatic labels such as fluorescent develop other types of binding reagents, such as aptamers tags or colloidal gold. The latter is frequently used in (Bonel et al., 2011; Castillo et al., 2015; Cruz-Aguado and immunochromatographic devices (LFD), and the signal Penner, 2008; Evtugyn et al., 2013; Park et al., 2014; Shim can be further enhanced with silver (Anfossi et al., 2013; Yu et al., 2014; Zhu et al., 2015). Improving the capacity of the et al., 2015). With LFD many of the challenges are associated binding element to regenerate, or improving the ability to with construction of the devices to minimise the influence remove and replenish the binding element, could result in of matrix and to permit reproducible signal development more applications where the selective binding agent (i.e. for quantification. The commercially available multiplexed aptamer or other) is immobilised rather than the antigen. LFD for ZEA, T-2/HT-2 toxins, DON, and fumonisins is an excellent example of where the challenges have been met While the preceding discussion has focused on recognition (Lattanzio et al., 2012, 2014). Investigations also continue and binding interactions, these interactions must be into the use of labels other than those involving fluorescence detected, and in some cases amplified, in order to generate (Mak et al., 2010; Malhotra et al., 2014; Tang et al., 2009). an observable response. Signal transduction is often facilitated by using reagents with enzymatic, coloured, or There have been efforts for many years to develop ‘label- fluorescent labels to improve detection by optical, electrical, free’ approaches to multiplexed mycotoxin detection. There or acoustic means. Signal transduction and detection, are are many such technologies but the greatest effort, at least also areas where obstacles are encountered. The challenges with the individual toxins, has been based on the technology inherent from a specific platform derive from the underlying of surface plasmon resonance (SPR) (Li et al., 2012; Meneely technology upon which the platform is based. For example, and Elliott, 2014; Van der Gaag et al., 2003). Recently a platforms that rely upon detecting a fluorescent response of variant of SPR, imaging SPR (iSPR) has also been used microbeads in solution can have different issues from those (Dorokhin et al., 2011; Hu et al., 2014). Imaging SPR may based upon changes in the refractive index at the surface of have an advantage over ‘traditional’ SPR in that multiple a sensor chip. As such, the issues that arise from detection assays can be conducted simultaneously in a single flow cell, platforms are intertwined with the type of label used. as opposed to the traditional approach that has involved using separate flow cells for each analyte. Individual The multitude of biosensor formats makes generalisations mycotoxin assays have also been reported using the difficult and unfortunately even a simple listing of recent technology of biolayer interferometry (BLI), where assays publications in this area would take more space than what are conducted at the ends of fibre optic sensors (Maragos, is available. In order to highlight some of the challenges 2012; McGrath et al., 2013). Even though BLI is readily that are encountered, we have classified biosensor formats amenable to multiplexing, such assays have not yet been by their underlying technologies. These include those that reported. For SPR and BLI approaches a challenge has been use enzymatic labels in ‘ELISA-type’ assays, those that the effect of residual solvent or matrix upon the refractive use non-enzymatic labels, and those that are considered index of the solution, which can affect signal intensity. ‘label-free’. Because of the extensive work that has been Often this is handled by using matrix-matched calibration done with ELISAs, many of the obstacles to their use have (Maragos, 2012). There are currently widespread efforts

World Mycotoxin Journal 9 (5) 857 J. Stroka and C.M. Maragos

to develop other label-free technologies for the individual Anderson, G.P., Kowtha, V.A. and Taitt, C.R., 2010. Detection of toxins, which may eventually find use in multiplexed devices fumonisin B1 and ochratoxin A in grain products using microsphere- (Grow et al., 2003; Nguyen et al., 2013; Park et al., 2014; based fluid immunoassays. Toxins 2: 297-309. Radi et al., 2009; Ricciardi et al., 2013; Spinella et al., 2014). Anfossi, L., Di Nardo, F., Giovannoli, C., Passini, C. and Baggiani, C., 2013. Increased sensitivity of lateral flow immunoassay for 4. Conclusions ochratoxin A through silver enhancement. Analytical and Bioanalytical Chemistry 405: 9859-9867. With respect to chromatographic methods it could be Bazin, I., Andreotti, N., Hassine, A.I.H., De Waard, M., Sabatier, J.M. shown that existing method proposals for aflatoxins have and Gonzalez, C., 2012. Peptide binding to ochratoxin A mycotoxin: the potential to minimise the calibration effort based on a new approach in conception of biosensors. Biosensors and the use of response factors. The example of pyrrolizidine Bioelectronics 40: 240-246. alkaloids showed that suitable sample preparation allows Beloglazova, N.V., Speranskaya, E.S., Wu, A., Wang, Z., Sanders, M., the determination of a single parameter with the benefit of Goftman, V.V., Zhang, D., Goryacheva, I.Y. and De Saeger, S., 2014. reducing the calibration effort. Also the targeted standard Novel multiplex fluorescent immunoassays based on quantum deviation for multi analyte methods is an issue that will dot nanolabels for mycotoxins determination. Biosensors and need discussion in the future, taking into account recently Bioelectronics 62: 59-65. generated data during collaborative studies. Bonel, L., Vidal, J.C., Duato, P. and Castillo, J.R., 2011. An electrochemical competitive biosensor for ochratoxin A based From the discussion on biosensors it should be clear on a DNA biotinylated aptamer. Biosensors and Bioelectronics that there is no single ‘perfect’ immunoassay solution for 26: 3254-3259. multiplexed detection of mycotoxins. Assays based upon Breidbach, A., Bouten, K., Kroeger-Negiota, K., Stroka, J. and Ulberth, molecular recognition are, by their nature, susceptible to F., 2013. LC-MS based method of analysis for the simultaneous conditions that alter the recognition event. Further, while determination of four mycotoxins in cereals and feed. JRC, Geel, many technologies have been developed to detect molecular Belgium. Available at: http://tinyurl.com/h868xn5. interactions, each has an inherent weakness associated with Castillo, G., Spinella, K., Poturnayová, A., Šnejdárková, M., Mosiello, the technology that represents a challenge to the use of L. and Hianik, T., 2015. Detection of aflatoxin B1 by aptamer-based that technology. Despite this, improvements to biosensors biosensor using PAMAM dendrimers as immobilization platform. continue to be made both in the recognition elements that Food Control 52: 9-18. are used and in the transduction and detection technologies. Choi, S.W., Chang, H.J., Lee, N., Kim, J.H. and Chun, H.S., 2009. As a result, assays based upon molecular recognition, Detection of mycoestrogen zearalenone by a molecularly imprinted such as immunoassays, continue to improve and there is polypyrrole-based surface plasmon resonance (SPR) sensor. Journal every reason to believe that the challenges to multiplexed of Agricultural and Food Chemistry 57: 1113-1118. detection of mycotoxins will be met with the development Cramer, L., Schiebel, H.-M., Ernst, L. and Beuerle, T., 2013. of fit-for-purpose methods. Pyrrolizidine alkaloids in the food chain: development, validation and application of a new HPLC-ESI-MS/MS sum parameter method. Disclaimer Journal of Agricultural and Food Chemistry 61: 11382-11391. Cruz-Aguado, J.A. and Penner, G., 2008. Determination of ochratoxin Mention of trade names or commercial products in this A with a DNA aptamer. Journal of Agricultural and Food Chemistry article is solely for the purpose of providing specific 56: 10456-10461. information and does not imply recommendation or De Girolamo, A., Solfrizzo, M., Lattanzio, V.M.T., Stroka, J., Alldrick, endorsement by the U.S. Department of Agriculture or A., Van Egmond, H.P. and Visconti, A., 2013. Critical evaluation the European Commission. USDA is an equal opportunity of LC-MS-based methods for simultaneous determination of provider and employer. deoxynivalenol, ochratoxin A, zearalenone, aflatoxins, fumonisins and T-2/HT-2 toxins in maize. World Mycotoxin Journal 6: 317-334. References Desmarchelier, A., Tessiot, S., Bessaire, T., Racault, L., Fiorese, E., Urbani, A., Chan, W.-C., Cheng, P. and Mottiera, P., 2014. Combining Al-Taher, F., Banaszewski, K., Jackson, L., Zweigenbaum, J., Ryu, D. the quick, easy, cheap, effective, rugged and safe approachand clean- and Cappozzo, J., 2013. Rapid method for the determination of up by immunoaffinity column for the analysis of 15 mycotoxins by multiple mycotoxins in wines and beers by LC-MS/MS using a isotope dilution liquid chromatographytandem mass spectrometry. stable isotope dilution assay. Journal of Agricultural and Food Journal of Chromatography A 1337: 75-84. Chemistry 61: 2378-2384. Dorokhin, D., Haasnoot, W., Franssen, M., Zuilhof, H. and Nielen, M., Anastassiades, M., Lehotay, S.J., Stajnbaher, D. and Schenck, F.J., 2011. Imaging surface plasmon resonance for multiplex microassay 2003. Fast and easy multiresidue method employing acetonitrile sensing of mycotoxins. Analytical and Bioanalytical Chemistry extraction/partitioning and dispersive solid-phase extraction for the 400: 3005-3011. determination of pesticide residues in produce. Journal of AOAC International 86: 412-431.

858 World Mycotoxin Journal 9 (5) Challenges in the analysis of multiple mycotoxins

European Committee for Standardization (CEN), 2001. CEN EN 13585 European Food Safety Authority (EFSA), 2016. Scientific Opinion –

– Foodstuffs – Determination of fumonisins B1 and B2 in maize Appropriateness to set a group health-based guidance value for – HPLC method with solid phase extraction clean-up. European zearalenone and its modified forms. EFSA Journal 14: 4425. Committee for Standardization, Brussels, Belgium. European Commission (EC), 1998. Commission Regulation (EC) No European Committee for Standardization (CEN), 2004. CEN EN 1525/98 of 16 July 1998 amending Regulation (EC) No 194/97 of

14352 – Foodstuffs – Determination of fumonisin B1 and B2 in maize 31 January 1997 setting maximum levels for certain contaminants based foods – HPLC method with immunoaffinity column clean in foodstuffs. Official Journal of the European Union L 201: 43-46. up. European Committee for Standardization, Brussels, Belgium. European Commission (EC), 2002a. Commission Decision 2002/657/ European Committee for Standardization (CEN), 2007. CEN EN 14123 EC of 14 August 2002 implementing Council Directive 96/23/ – Foodstuffs – Determination of aflatoxin B1 and the sum of aflatoxin EC concerning the performance of analytical methods and the

B1, B2, G1 and G2 in hazelnuts, peanuts, pistachios, figs, and paprika interpretation of results. Official Journal of the European Union powder – High performance liquid chromatographic method with L 221: 8-36. post-column derivatisation and immunoaffinity column cleanup. European Commission (EC), 2002b. Directive 2002/32/EC of European Committee for Standardization, Brussels, Belgium. the European Parliament and of the Council of 7 May 2002 on European Committee for Standardization (CEN), 2009a. CEN EN undesirable substances in animal feed. Official Journal of the 14132 – Foodstuffs – Determination of ochratoxin A in barley and European Union L 140: 10-27. roasted coffee – HPLC method with immunoaffinity column clean- European Commission (EC), 2006a. Commission Regulation (EC) up. European Committee for Standardization, Brussels, Belgium. No 1881/2006 of 19 December 2006 setting maximum levels for European Committee for Standardization (CEN), 2009b. CEN EN certain contaminants in foodstuffs. Official Journal of the European 15791 – Foodstuffs – Determination of deoxynivalenol in animal Union L 364: 25-31. feed – HPLC method with immunoaffinity column clean-up. European Commission (EC), 2006b. Commission Recommendation European Committee for Standardization, Brussels, Belgium. 2006/576/EC of 17 August 2006 on the presence of deoxynivalenol, European Committee for Standardization (CEN), 2009c. CEN EN zearalenone, ochratoxin A, T-2 and HT-2 and fumonisins in 15792 – Animal feeding stuffs – Determination of zearalenone in products intended for animal feeding. Official Journal of the animal feed – High performance liquid chromatographic method European Union L 229: 7-9. with fluorescence detection and immunoaffinity column clean- European Commission (EC), 2006c. Commission Regulation (EC) No up. European Committee for Standardization, Brussels, Belgium. 401/2006 of 23 February 2006 laying down the methods of sampling European Committee for Standardization (CEN), 2010a. CEN EN and analysis for the official control of the levels of mycotoxins in 15850 – Foodstuffs – Determination of zearalenone in maize based foodstuffs. Official Journal of the European Union L 70: 12-34. baby food, barley flour, maize flour, polenta, wheat flour and cereal European Commission (EC), 2013. Commission Recommendation based foods for infants and young children – HPLC method with 2013/165/EU of 27 March 2013 on the presence of T-2 and HT-2 immunoaffinity column cleanup and fluorescence detection. toxin in cereals and cereal products. Official Journal of the European European Committee for Standardization, Brussels, Belgium. Union L 91: 12-15. European Committee for Standardization (CEN), 2010b. CEN EN European Commission (EC), 2016. Rapid alert system for food and 15891 – Foodstuffs – Determination of deoxynivalenol in cereals, feed (RASFF) portal. Available at: http://ec.europa.eu/food/safety/ cereal products and cereal based foods for infants and young rasff/portal/index_en.htm. children – HPLC method with immunoaffinity column cleanup Evtugyn, G., Porfireva, A., Sitdikov, R., Evtugyn, V., Stoikov, I., and UV detection. European Committee for Standardization, Antipin, I. and Hianik, T., 2013. Electrochemical aptasensor for Brussels, Belgium. the determination of ochratoxin A at the Au electrode modified with European Committee for Standardization (CEN), 2011a. CEN EN 16006 Ag nanoparticles decorated with macrocyclic ligand. Electroanalysis – Animal feeding stuffs – Determination of the Sum of fumonisin 25: 1847-1854.

B1 and B2 in compound animal feed with immunoaffinity clean-up Food and Agricultural Organisation (FAO), 2004. Worldwide and RP-HPLC with fluorescence detection after pre- or postcolumn regulations for mycotoxins in food and feed in 2003. FAO, Rome, derivatisation. European Committee for Standardization, Brussels, Italy. Available at: http://www.fao.org/docrep/007/y5499e/y5499e00. Belgium. htm European Committee for Standardization (CEN), 2011b. CEN EN Gonçalves, C. and Stroka, J., 2016. Food contaminants and human 16007 – Animal feeding stuffs – Determination of ochratoxin health – challenges in analysis of multiple chemicals. In: Alvito, A in animal feed by immunoaffinity column clean-up and High P., Assunção, R., Louro, H., Silva, M.J. and Vasco, E. (eds.) Food Performance Liquid Chromatography with fluorescence detection. contaminants and human health – challenges in analysis of chemical European Committee for Standardization, Brussels, Belgium. mixtures. National Institute of Health Doutor Ricardo Jorge, Lisbon, European Committee for Standardization (CEN), 2011c. CEN/TS Portugal, pp. 11-16. Available at: http://tinyurl.com/zppt3mp.

16187 – Foodstuffs – Determination of fumonisin B1 and fumonisin Grow, A.E., Wood, L.L., Claycomb, J.L. and Thompson, P.A., 2003. B2 in processed maize containing foods for infants and young New biochip technology for label-free detection of pathogens and children – HPLC method with immunoaffinity column cleanup their toxins. Journal of Microbiological Methods 53: 221-233. and fluorescence detection after precolumn derivatization. European Committee for Standardization, Brussels, Belgium.

World Mycotoxin Journal 9 (5) 859 J. Stroka and C.M. Maragos

Horwitz, W., Kamps, L.R. and Boyer, K.W., 1980. Quality assurance in Maragos, C., Busman, M. and Sugita-Konishi, Y., 2006. Production and the analysis of foods and trace constituents. Journal of the AOAC characterization of a monoclonal antibody that cross-reacts with International 63: 1344-1354. the mycotoxins nivalenol and 4-deoxynivalenol. Food Additives Hu, W., Chen, H., Zhang, H., He, G., Li, X., Zhang, X., Liu, Y. and Li, and Contaminants 23: 816-825. C.M., 2014. Sensitive detection of multiple mycotoxins by SPRi with Maragos, C.M., 2012. Signal amplification using colloidal gold in a gold nanoparticles as signal amplification tags. Journal of Colloid biolayer interferometry-based immunosensor for the mycotoxin and Interface Science 431: 71-76. deoxynivalenol. Food Additives and Contaminants Part A Institute for Reference Materials and Measurements (IRRM), 2016. 29: 1108-1117. Available at: https://ec.europa.eu/eusurvey/runner/Aflatoxins_ Matumba, L., Sulyok, M., Njoroge, S.M.C., Ediage, E.N., Van Poucke, Survey. C., De Saeger, S. and Krska, R., 2015. Uncommon occurrence ratios

International Organization for Standardization (ISO), 2005. ISO of aflatoxin B1, B2, G1 and G2 in maize and groundnuts from Malawi. 17025 – General requirements for the competence of testing and Mycotoxin Research 31: 57-62. calibration laboratories. ISO, Geneva, Switzerland. McGrath, T., Campbell, K., Fodey, T., O’Kennedy, R. and Elliott, C., Kempf, M., Wittig, M., Reinhard, A., Von der Ohe, K., Blacquière, 2013. An evaluation of the capability of a biolayer interferometry T., Raezke, K.P., Michel, R., Schreier, P. and Beuerle, T., 2011. biosensor to detect low-molecular-weight food contaminants. Pyrrolizidine alkaloids in honey: comparison of analytical methods. Analytical and Bioanalytical Chemistry 405: 2535-2544. Food Additives and Contaminants Part A 28: 332-347. Meneely, J.P. and Elliott, C.T., 2014. Rapid surface plasmon resonance Kokkonen, M., 2011. The challange of LC/MS/MS multitoxin analysis immunoassays for the determination of mycotoxins in cereals and − Heracles battling the hydra? PhD thesis, University of Helsinki, cereal-based food products. World Mycotoxin Journal 7: 491-505. Helsinki, Finland. Available at: http://tinyurl.com/zewjakz. Mol, H., Plaza-Bolanõs, P., Zomer, P., De Rijk, T., Stolker, A. and Kujawski, M., Mischke, C. and Stroka, J., in press. Collaborative study Mulder, P., 2008. Toward a generic extraction method for report: determination of multiple mycotoxins in feed materials and simultaneous determination of pesticides, mycotoxins, plant compound feed by LC-MS/MS. JRC, Geel, Belgium. toxins, and veterinary drugs in feed and food matrixes. Analytical Lacina, O., Zachariasova, M., Urbanova, J., Vaclavikova, M., Cajka, T. Chemistry 80: 9450-9459. and Hajslova, J., 2012. Critical assessment of extraction methods Mol, H., 2014. Requirements for identification in residue analysis. for the simultaneous determination of pesticide residues and Video presentation at the enlargement and integration workshop: mycotoxins in fruits, cereals, spices and oil seeds employing analysis of food and feed contaminants – the legal and scientific ultra-high performance liquid chromatography–tandem mass framework behind, 24-26 September 2014, Zagreb, Croatia. spectrometry. Journal of Chromatography A 1262: 8-18. Available at: https://ec.europa.eu/jrc/en/eurl/mycotoxins/resources. Lattanzio, V.M., Nivarlet, N., Lippolis, V., La Gatta, S., Huet, A., Nguyen, B.H., Tran, L.D., Do, Q.P., Nguyen, H.L., Tran, N.H. and

Delahaut, P., Granier, B. and Visconti, A., 2012. Multiplex dipstick Nguyen, P.X., 2013. Label-free detection of aflatoxin M1 with immunoassay for semi-quantitative determination of Fusarium electrochemical Fe3O4/polyaniline-based aptasensor. Materials mycotoxins in cereals. Analytica Chimica Acta 718: 99-108. Science and Engineering: C, Materials for Biological Applications Lattanzio, V.M.T., Von Holst, C. and Visconti, A., 2014. Collaborative 33: 2229-2234. study for evaluating performances of a multiplex dipstick Oswald, S., Karsunke, X., Dietrich, R., Martlbauer, E., Niessner, R. immunoassay for Fusarium mycotoxin screening in wheat and and Knopp, D., 2013. Automated regenerable microarray-based maize. Quality Assurance and Safety of Crops and Foods 6: 299-307. immunoassay for rapid parallel quantification of mycotoxins in Li, Y., Liu, X. and Lin, Z., 2012. Recent developments and applications cereals. Analytical and Bioanalytical Chemistry 405: 6405-6415. of surface plasmon resonance biosensors for the detection of Park, J., Byun, J., Mun, H., Shim, W., Shin, Y., Li, T. and Kim, M., 2014. mycotoxins in foodstuffs. Food Chemistry 132: 1549-1554. A regeneratable, label-free, localized surface plasmon resonance Liu, J.L., Hu, Z.Q., Xing, S., Xue, S., Li, H.P., Zhang, J.B. and Liao, Y.C., (LSPR) aptasensor for the detection of ochratoxin A. Biosensors 2012. Attainment of 15-fold higher affinity of a Fusarium-specific and Bioelectronics 59: 321-327. single-chain antibody by directed molecular evolution coupled to Peters, J., Thomas, D., Boers, E., De Rijk, T., Berthiller, F., Haasnoot, phage display. Molecular Biotechnology 52: 111-122. W. and Nielen, M., 2013. Colour-encoded paramagnetic microbead- Mak, A.C., Osterfeld, S.J., Yu, H., Wang, S.X., Davis, R.W., Jejelowo, based direct inhibition triplex flow cytometric immunoassay for O.A. and Pourmand, N., 2010. Sensitive giant magnetoresistive- ochratoxin A, fumonisins and zearalenone in cereals and cereal- based immunoassay for multiplex mycotoxin detection. Biosensors based feed. Analytical and Bioanalytical Chemistry 405: 7783-7794. and Bioelectronics 25: 1635-1639. Peters, J., Cardall, A., Haasnoot, W. and Nielen, M.W.F., 2014. 6-Plex Malachova, A., Sulyok, M., Beltran, E., Berthiller, F. and Krska, R., microsphere immunoassay with imaging planar array detection for 2015. Multi-toxin determination in food – the power of ‘dilute and mycotoxins in barley. Analyst 139: 3968-3976. shoot’ approaches in LC-MS-MS. LCGC Europe 28(10). Available Radi, A.E., Munoz-Berbel, X., Lates, V. and Marty, J.L., 2009. Label-free at: http://www.chromatographyonline.com/print/300375?page=full impedimetric immunosensor for sensitive detection of ochratoxin Malhotra, B.D., Srivastava, S., Ali, M.A. and Singh, C., 2014. A. Biosensors and Bioelectronics 24: 1888-1892. Nanomaterial-based biosensors for food toxin detection. Applied Biochemistry and Biotechnology 174: 880-896.

860 World Mycotoxin Journal 9 (5) Challenges in the analysis of multiple mycotoxins

Ricciardi, C., Castagna, R., Ferrante, I., Frascella, F., Luigi Marasso, These, A., Bodi, D., Ronczka, S., Lahrssen-Wiederholt, M. and S., Ricci, A., Canavese, G., Lorè, A., Prelle, A., Lodovica Gullino, Preiss-Weigert, A., 2013. Structural screening by multiple reaction M. and Spadaro, D., 2013. Development of a microcantilever-based monitoring as a new approach for tandem mass spectrometry: immunosensing method for mycotoxin detection. Biosensors and presented for the determination of pyrrolizidine alkaloids in plants. Bioelectronics 40: 233-239. Analytical and Bioanalytical Chemistry 405: 9375-9383. Sanders, M., Guo, Y., Iyer, A., Ruiz Garcia, Y., Galvita, A., Thompson, M., 2000. Recent trends in interlaboratory precision at Heyerick, A., Deforce, D., Risseeuw, M.D.P., Van Calenbergh, S., ppb and sub-ppb concentrations in relation to fitness for purpose Bracke, M., Eremin, S., Madder, A. and De Saeger, S., 2014. An criteria in proficiency testing. Analyst 125: 385-386. immunogen synthesis strategy for the development of specific Ton, X.A., Acha, V., Bonomi, P., Tse Sum Bui, B. and Haupt, K., 2015. anti-deoxynivalenol monoclonal antibodies. Food Additives and A disposable evanescent wave fiber optic sensor coated with a Contaminants Part A 31: 1751-1759. molecularly imprinted polymer as a selective fluorescence probe. Shim, W., Mun, H., Joung, H., Ofori, J.A., Chung, D. and Kim, M., 2014. Biosensors and Bioelectronics 64: 359-366. Chemiluminescence competitive aptamer assay for the detection of United States Department of Agriculture (USDA), 2016. Design aflatoxin B1 in corn samples. Food Control 36: 30-35. criteria and test performance specifications for quantitative aflatoxin Sospedra, I., Blesa, J., Soriano, J.M. and Manes, J., 2010. Use of test kits administration. USDA – Grain Inspection, Packers and the modified quick easy cheap effective rugged and safe sample Stockyards, Washington, DC, USA. Available at: http://www.gipsa. preparation approach for the simultaneous analysis of type A- and usda.gov/fgis/metheqp/aflatoxins_criteria.pdf. B-trichothecenes in wheat flour. Journal of Chromatography A Van der Gaag, B., Spath, S., Dietrich, H., Stigter, E., Boonzaaijer, G., 1217: 1437-1440. Van Osenbruggen, T. and Koopal, K., 2003. Biosensors and multiple Spinella, K., Mosiello, L., Palleschi, G. and Vitali, F., 2014. Development mycotoxin analysis. Food Control 14: 251-254. of a QCM (Quartz Crystal Microbalance) biosensor to detection of Varga, E., Glauner, T., Köppen, R., Mayer, K., Sulyok, M., Schuhmacher, mycotoxins. Lecture Notes in Electrical Engineering 268: 195-198. R., Krska, R. and Berthiller, F., 2012. Stable isotope dilution assay Stroka, J., Petz, M., Jörissen, U. and Anklam, E., 1999. Investigation of for the accurate determination of mycotoxins in maize by UHPLC-

various extractants for the analysis of aflatoxin B1 in different food MS/MS. Analytical and Bioanalytical Chemistry 402: 2675-2685. and feed matrices. Food Additives and Contaminants 16: 331-338 Vidal, J.C., Bonel, L., Ezquerra, A., Hernandez, S., Bertolin, J.R., Cubel, Stroka, J., Breidbach, A., Doncheva, I. and Kroeger, K., 2007. Report C. and Castillo, J.R., 2013. Electrochemical affinity biosensors for of the first inter-laboratory comparison test organised by the detection of mycotoxins: a review. Biosensors and Bioelectronics

community reference laboratory for mycotoxins – aflatoxins B1, 49: 146-158. B2, G1 and G2 in acetonitrile. CRL Mycotoxins, IRRM, Geel, Belgium. Wang, Y., Li, P., Majkova, Z., Bever, C.R.S., Kim, H.J., Zhang, Q., Available at: http://tinyurl.com/guw8nmw. Dechant, J.E., Gee, S.J. and Hammock, B.D., 2013a. Isolation of Sulyok, M., Berthiller, F., Krska, R. and Schuhmacher, R., 2007. alpaca anti-idiotypic heavy-chain single-domain antibody for the Development and validation of a liquid chromatography/tandem aflatoxin immunoassay. Analytical Chemistry 85: 8298-8303. mass spectrometric method for the determination of 39 mycotoxins Wang, Y., Ning, B., Peng, Y., Bai, J., Liu, M., Fan, X., Sun, Z., Lv, Z., in wheat and maize. Rapid Communication in Mass Spectrometry Zhou, C. and Gao, Z., 2013b. Application of suspension array for 20: 2649-2659. simultaneous detection of four different mycotoxins in corn and Sulyok, M., Krska, R. and Schuhmacher, R., 2010. Application of an LC– peanut. Biosensors and Bioelectronics 41: 391-396. MS/MS based multi-mycotoxin method for the semi-quantitative Yu, Q., Li, H., Li, C.L., Zhang, S.X., Shen, J.Z. and Wang, Z.H., determination of mycotoxins occurring in different types of food 2015. Gold nanoparticles-based lateral flow immunoassay with

infected by moulds. Food Chemistry 119: 408-416. silver staining for simultaneous detection of fumonisin B1 and Tamošiūnas, V., Mischke, C., Mulder, P.P.J. and Stroka, J., 2013. Report deoxynivalenol. Food Control 54: 347-352. on the 2012 proficiency test on pyrrolizidine alkaloids in honey Zachariasova, M., Lacina, O., Malachova, A., Kostelanska, M., Poustka, and hay. IRRM, Geel, Belgium. Available at: http://tinyurl.com/ J., Godula, M. and Hajslova, J., 2010. Novel approaches in analysis of gmxm2w5. Fusarium mycotoxins in cereals employing ultra performance liquid Tanaka, H., Takino, M., Sugita-Konishi, Y., Tanaka, T., Leeman, D., chromatography coupled with high resolution mass spectrometry. Toriba, A. and Hayakawa, K., 2010. Determination of Fusarium Analytica Chimica Acta 662: 51-61. mycotoxins by liquid chromatography/tandem mass spectrometry Zangheri, M., Di Nardo, F., Anfossi, L., Giovannoli, C., Baggiani, C., coupled with immunoaffinity extraction. Rapid Communication in Roda, A. and Mirasoli, M., 2015. A multiplex chemiluminescent Mass Spectrometry 24(16): 2445-2452. biosensor for type B-fumonisins and aflatoxin B1 quantitative Tang, D., Sauceda, J.C., Lin, Z., Ott, S., Basova, E., Goryacheva, I., detection in maize flour. Analyst 140: 358-365. Biselli, S., Lin, J., Niessner, R. and Knopp, D., 2009. Magnetic Zhu, Z., Feng, M., Zuo, L., Wang, F., Chen, L., Li, J., Shan, G. and Luo, nanogold microspheres-based lateral-flow immunodipstick for S., 2015. An aptamer based surface plasmon resonance biosensor

rapid detection of aflatoxin B2 in food. Biosensors and Bioelectronics for the detection of ochratoxin A in wine and peanut oil. Biosensors 25: 514-518. and Bioelectronics 65: 320-326.

World Mycotoxin Journal 9 (5) 861