A QuEChERS method for simultaneously analyzing 24 different Fusarium and Alternaria toxins was successfully developed and validated. Before the chosen modified QuEChERS setup was finally established, dilute-and-shoot, SPE, and QuEChERS approaches were tested. Finally, the QuEChERS workup emerged as the most promising in performance, sensitivity, cost, and time requirements.

Workup optimizationQuEChERS approach

A preliminary screening of several clean-up variants revealed that simple dilute-and-shoot was the least sensitive option for most analytes. Moreover, a direct comparison of SPE and QuEChERS followed by SPE had no advantages over a simple QuEChERS procedure, with the SPE being more time-consuming as it requires a double solvent exchange. Therefore, the QuEChERS approach was extensively optimized to be competitive with our group’s previously developed SPE methods while including more analytes, being faster, and reducing the workup time and costs. The workup was inspired by other QuEChERS-based procedures [31], which showed the applicability of the QuEChERS approach for a wide variety of mycotoxins and complex matrices.

Extraction

Multiple mycotoxins are commonly extracted with different ratios of ACN and H2O, often in conjunction with acidification using FA or AA. Various ratios and volumes of those solvents were tested during the optimization process. The extraction time was varied for 30, 60, 90, and 120 min. Based on the extraction mixture of ACN and H2O described in the literature [17, 32], the FA concentration adjustment proved necessary. Concentrations below 1% FA showed a substantial loss in TeA and displayed reduced extraction efficiency for modified Alternaria toxins, while concentrations above 2% led to increased matrix coextraction (Fig. 2). For exhaustive analyte extraction, it was necessary to perform three consecutive extraction steps, with the first extraction taking 60 min and the second and third extraction taking 30 min each. The first consists of 10 mL ACN/H2O (80/20; v/v), the second of 5 mL ACN/H2O (80/20; v/v), and the third of 5 mL ACN/H2O (70/30; v/v), all containing 1% FA.

Fig. 2

Comparison of extraction yields for important mycotoxins of the ACN/H2O 80/20 extraction mixture with increasing percentages of FA. The experiment was conducted using double determination and double injection. The concentrations used for spiking were as follows: DON (20 µg/kg), 3-AcDON (10 µg/kg), Fus X (50 µg/kg), T-2 (5 µg/kg), HT-2 (5 µg/kg), ZEN (5 µg/kg), ENN B (1 µg/kg), BEA (1 µg/kg), TeA (20 µg/kg), AOH (10 µg/kg), AOH-3-G (10 µg/kg), AME (1 µg/kg), AME-3-S (1 µg/kg), TEN (10 µg/kg)

QuEChERS clean-up

To separate the matrix components and keep pH-sensitive analytes in the ACN phase during QuEChERS clean-up, the FA content was varied between 0 and 5%. Also, the three established variants of non-buffered, acetate-buffered, and citrate-buffered methods were tested [15, 25, 26], while the salt addition was based on the total H2O content of the final extraction solvent [25]. While all those methods were suitable for multi-mycotoxin workup, the approach using anh. MgSO4 and NaCl was finally applied, as no buffering was found to be necessary for the recovery of the analytes, and manually weighing only two components saves time. A surplus of QuEChERS salts did not lead to beneficial or adverse effects during the workup. Thus, the amount of salts used was adjusted to the amount of H2O left in the sample based on the initially published method [15]. To guarantee the transfer of TeA in the ACN phase, 1% FA was added before QuEChERS clean-up, leading to a final FA concentration of 2%.

d-SPE clean-up

To further clean up the extract, the following d-SPE sorbents were tested: C18, PSA, Supelclean™ ENVI Carb™, and Supel™ QuE Z-SEP. Of these, PSA is the most commonly used d-SPE sorbent. In our case, it showed an excellent clean-up without any adverse effects. C18 is the second most used sorbent. It showed advantages for most toxins while decreasing the sensitivity of TeA, the ENNs, and BEA. ENVI-carb is based on graphitized carbon black and strongly removed different matrix compounds, but unfortunately, quantitatively removed AME and the ENNs as well. The effect of the adsorbents tested was similar to reports in the literature [33]. The acidification before d-SPE clean-up solved the problem of TeA loss, as known in the literature [34]. The decision favoring the sole usage of PSA and anh. MgSO4 compared to combining C18/PSA and anh. MgSO4 was made due to the better sensitivity for TeA and DON (Fig. 3).

Fig. 3

Comparison of recoveries for important mycotoxins when using different d-SPE clean-up materials. The experiment was conducted using double determination and double injection. The concentrations used for spiking were as follows: DON (20 µg/kg), 3-AcDON(10 µg/kg), Fus X (50 µg/kg), T-2 (5 µg/kg), HT-2 (5 µg/kg), ZEN (5 µg/kg), ENN B (1 µg/kg), BEA (1 µg/kg), TeA (20 µg/kg), AOH (10 µg/kg), AOH-3-G (10 µg/kg), AME (1 µg/kg), AME-3-S (1 µg/kg), TEN (10 µg/kg)

Reconstitution solvent

A low volume of 200 µL revealed the best intensity and signal-to-noise ratio for all analytes, while matrix contamination was still acceptable. The low reconstitution volume is possible by the good clean-up achieved by QuEChERS and d-SPE. Ratios of 6/4 (v/v) for MeOH/H2O gave the best results for the re-solvation of analytes, matrix reduction, and peak shape. To prevent matrix precipitation during storage or at the autosampler temperature of 4 °C, the reconstituted extract was frozen for 30 min. Afterward, the sample was filtered with a PVDF filter before LC–MS/MS analysis. PVDF and PTFE filters were compared for the latter filtration. At the applied solvent ratio of 6/4 (v/v) MeOH/H2O, the PVDF filter showed better results and no specific analyte loss compared to the observations described in the literature [35].

Development of the LC–MS/MS method

Concerning the expected occurrence and quantitative amounts of the mycotoxins in our study, a clear focus was set on the Alternaria toxins TeA as the major toxin of these fungi and the benzopyrones AOH and AME due to their toxicity. It is essential to get an insight into the occurrence of these Alternaria toxins and to obtain LODs and LOQs as low as possible to circumvent left-censored data for accurate exposure studies. Therefore, the development of the multi-method was targeted towards these aims, especially as TeA is not included in multi-methods in general, which is a general drawback we wanted to solve in our study.

Column selection

Selecting a suitable column was essential to achieve a reasonable separation of the analytes, especially the structural isomers. Different columns were initially tested based on availability and experience reported in the literature. Among those were various modified C18 columns, such as Acquity BEH C18 (Waters), Acquity CSH C18 (Waters), Acquity TSS T3 (Waters), Gemini® C18 (Phenomenex), HyperClone C18 (Phenomenex), Triart C18 (YMC), and Shim-pack Velox PFPP (Shimadzu), and others like biphenyl and phenylhexyl variants. Due to the number of different analytes with similar properties, it appeared straightforward to use modern UHPLC columns that allow for sharper peaks and better separation. It was quickly apparent that C18 columns showed the overall best separation results, as they are applicable to a wide range of analytes. The only disadvantage of the tested C18 columns was the inability to separate 3-AcDON from its isomer 15-AcDON. However, this disadvantage was accepted because 3-AcDON occurs more frequently in foods. After the first tests, the best results were achieved using three columns, namely the PFPP, BEH C18, and HSS T3 columns. The BEH C18 and HSS T3 columns showed promising results in direct comparison. Therefore, the chromatographically challenging molecule TeA that performs best in strong acidic or basic eluents determined the final column choice. As the latter two columns have a different pH working range, the final column selection was highly influenced by the choice of solvent. As the BEH C18 column is stable from pH 1 to 12, it was preferred and showed the best peak shape for TeA while still being compatible with all other analytes.

Solvents

To achieve the aim of separating all analytes within only one chromatographic run, the solvents required special attention. In particular, for the implementation of TeA into the method with a reasonably sharp peak, two options, according to literature and own experiments, are possible: the measurement either with 1% AA both in H2O and in ACN or the measurement with MeOH and H2O with NH4CH3CO2 at pH 9 [32, 36]. A range of pH values and buffer concentrations were tested for NH4CH3CO2 and NH4HCO2, and the best peak shape for TeA was obtained with 5 mM NH4HCO2 at pH 9. Additionally, NH4HCO2 favored the formation of NH4+ adducts in the ion source of the MS for better sensitivity of ENNs, BEA, T-2, and HT-2 without compromising the sensitivity of other toxins.

Gradient

The flow rates were tested in the range between 0.2 and 0.4 mL/min, and a flow rate of 0.3 mL was elucidated as optimal as it allows sharp peaks while ensuring reduced pressure. Moreover, a column temperature of 40 °C gave the best performance. Although baseline separation for all analytes was not achievable, the specificity of MS/MS detection allowed us to quantify them separately. However, we had to compromise for separating 3-AcDON and 15-AcDON as the PFPP column capable of separating these isomers was omitted as it did not yield a reasonable peak shape for TeA (Fig. 4A). The alkaline mobile phase also allowed us to resolve the isomer pairs AOH-3-G/AOH-9-G and AOH-3-S/AOH-9-S (Fig. 4C) that are not differentiable by MS. Special attention had to be paid to the starting conditions: starting at 5% of the organic component B sharpened the peaks of TeA and NIV and in combination with the shallow gradient slope from 18 to 25% B allowed for the baseline separation of DON and DON-3-G (Fig. 4B). This is particularly important because DON-3-G shows in-source fragmentation to DON and might otherwise add to the DON signal. The other toxins did not require special attention, and the linear gradient achieved a sufficient resolution.

Fig. 4

Excerpts of a multi-analyte chromatogram highlighting TeA (A), DON and DON-3-G (B), and the isomer pairs AOH-3-G/AOH-9-G and AOH-3-S/AOH-9-S (C). The concentrations used for the measurements were DON (0.1 µg/mL), DON-3-G (0.2 µg/mL), AOH-3-G (0.005 µg/mL), AOH-9-G (0.01 µg/mL), AOH-3-S (0.005 µg/mL), and AOH-9-S (0.005 µg/mL)

ESI polarity switching

During initial ESI tuning, it became apparent that some analytes could be detected more sensitively in the negative mode and others in the positive mode despite the variation of pH and mobile phase. This pointed to the need for polarity switching, which displayed similar sensitivity and stability as the singular ionization modes. With the implementation of this feature, we were able to reliably and sensitively measure 24 different mycotoxins, including their modifications in one method, as displayed in Fig. 5.

Fig. 5

Full chromatogram of LC–MS/MS separation of the following mycotoxins in order of increasing retention time in the respective concentrations: NIV (1; 0.5 µg/mL), TeA (2; 0.2 µg/mL), DON (3; 0.1 µg/mL), DON-3-G (4; 0.2 µg/mL), AOH-3-G (5; 0.005 µg/mL), Fus X (6; 0.05 µg/mL), AOH-3-S (7; 0.005 µg/mL), 3-AcDON (8; 0.01 µg/mL), AOH-9-S (9; 0.005 µg/mL), AOH-9-G (10; 0.01 µg/mL), AOH (11; 0.03 µg/mL), AME-3-S (12; 0.0005 µg/mL), ATX I (13; 0.15 µg/mL), AME-3-G (14; 2 µg/mL), HT-2 (15; 0.02 µg/mL), TEN (16; 0.02 µg/mL), T-2 (17; 0.005 µg/mL), ZEN (18; 0.02 µg/mL), AME (19; 0.002 µg/mL), ENN B (20; 0.0004 µg/mL), BEA (21; 0.0004 µg/mL), ENN B1 (21; 0.0004 µg/mL), ENN A1 (23; 0.0004 µg/mL), ENN A (24; 0.0004 µg/mL)

Source parameter optimization and polarity switching

Source optimization was conducted by manually modifying the desolvation line, heat block, and injector port temperature. Here, a quite low desolvation line temperature of 150 °C showed the best impact on the occurrence of NH4+ adducts while having a positive to no effect on the other toxins. The chosen heat block temperature of 450 °C and interface temperature of 350 °C displayed the highest intensity of analytes on average.

Comparison of the chromatographic performance

The newly developed method allows combining the two previously applied workup methods and four different LC–MS/MS methods previously used by our group in one LC–MS/MS run [17, 18, 37]. Analyzing Alternaria toxins under unusual basic conditions enables a high sensitivity for TeA. This is already described in the literature, but its compatibility with all other analytes and introducing Fusarium toxins in an Alternaria-focused method is a new feature of our method [20, 32, 34]. Moreover, optimizing all parameters allowed for low LOQs significantly below the current ML and IL in cereals and cereal-based foods (see next chapter).

However, some minor drawbacks of the new method also have to be mentioned: Two Alternaria toxins, ALTP and ATX II, are not quantifiable in contrast to one of our previous methods [17]. ATX II proved to be unstable during workup, and ALTP frequently overlapped with interfering matrix compounds. We also were not able to baseline separate 3-AcDON from its isomer 15-AcDON. We can detect the predominant isomer based on different mass transitions, but as their transitions overlap, we cannot quantify both simultaneously. The additional transitions for 15-AcDON were still included in the LC–MS/MS measurements, but we did not detect 15-AcDON in significant ratios in any of the 136 samples measured. This is reasonable, as Fusarium species usually only produce one major isomer. However, if a sample may contain a significant share of 15-AcDON, the respective extract could be measured with our previously published method using the PFPP stationary phase to separate the two isomers [37].

Method validationCalibration and quantification

Linearity of the SIDA response functions of the analytes AOH, AME, TeA, DON, DON-3-G, 3-AcDON, T-2, HT-2, ENN A, ENN A1, ENN B, ENN B1, and BEA in relation to their LIS was verified with Mandel’s fitting test [27]. The linear range encompassed molar ratios from 0.01 to 100 for all toxins except DON-3-G and ENN A1. For ENN A1 the linear range covered molar ratios n(A)/n(LIS) between 0.01 and 20. For DON-3-G, the range was deliberately reduced (0.1 to 50), as the respective LIS is quite expensive and added to the sample in relatively low amounts. However, this was no problem for DON-3-G, as ratios below 0.1 did not appear in the analyzed samples. For ENN A1, the range had to be reduced as the curve lacked linearity in ratios higher than 20. However, this is not critical as ENN A1 does not occur naturally in high amounts, and a ratio exceeding 20 did not seem probable from the perspective of the analyzed samples.

MMC curves were obtained for all samples by spiking the blank matrix with eight to ten concentrations. Linearity was again confirmed using Mandel’s fitting test [27]. The LOQ was used as the lowest spiking level, while the highest spiking level was at least ten times higher, resulting in the following calibration ranges: 0.1–30 µg/kg TEN, 1.0–20 µg/kg ATX I, 0.3–20 µg/kg AOH-3-S, 0.2–20 µg/kg AME-3-S, 0.2–20 µg/kg AOH-3-G, 0.2–20 µg/kg AOH-9-G, 1.0–20 µg/kg AME-3-G, 5.0–100 NIV, 2–100 µg/kg Fus X, and 0.2–50 g/kg ZEN.

The modified Alternaria toxin AOH-9-S was only included qualitatively in the method because the available amounts of this toxin were insufficient to generate an MMC for quantification.

LODs and LOQs

LODs and LOQs were determined according to the literature [28]. Accordingly, an analyte-free blank matrix (potato starch) was spiked in four concentrations. The results are summarized in Table 2. The LODs ranged from 0.004 to 7.99 µg/kg. The ENNs and AME showed remarkably low LODs, which may be caused by a good ionization efficiency without many matrix interferences. In the case of the ENNs, especially the formed NH4+ adducts are sensitively detected in the MS. However, DON-3-G and NIV revealed high LODs as their ionization efficiency is worse, and they might dissolve partly in the H2O phase during QuEChERS clean-up. These trends have already been described in the literature [38]. Also, DON-3-G and AME-3-G showed reduced sensitivity, as both toxins tend towards in-source fragmentation, also already described in the literature [17, 32, 38].

Table 2 Limits of detection (LODs), limits of quantification (LOQs), relative standard deviation (RSD) values (precision), and recoveries for all toxins in starch as the blank matrix. Recovery values of each spiking level were determined as the mean value of three replicates in triple injection

The determined LODs and LOQs are comparable to our group’s previously developed single-species methods [17, 18], which we attempted to combine in the present study. The LOQs of the present method proved to be lower or similar for most analytes. Only ATX I showed a reduced sensitivity, which might be caused by the alkaline measuring conditions and high matrix-generated noise.

Our method displays competitive sensitivity compared to other multi-mycotoxin methods for a similar set of analytes. Setting a priority on TeA proved beneficial as the LOQs obtained were lower than those reported for other multi-methods [38, 39]. The sensitivity for the other Alternaria toxins was good and similar to those reported in the literature [17, 32, 38].

Sensitivity for Fusarium toxins that form NH4+ adducts like T-2, HT-2, ENN A, ENN A1, ENN B1, ENN B, and BEA was better than in other methods [18, 38, 40]. Moreover, the LOQs for DON, NIV, 3-AcDON, Fus X, and ZEN were low and in a similar range compared to methods in the literature. The LOQ of DON-3-G was slightly higher but still sufficiently low [38]. Overall, the method proved to be a sensitive quantification approach for all analytes.

Recovery

The recovery was determined by spiking every analyte in triplicates at three to four concentrations in the blank matrix. The lowest concentration was at the LOQ, while for the highest concentration, a reasonably high amount was chosen to establish the working range for the concentrations to be expected in the samples. As can be seen from Table 2, recoveries were between 84.0 and 108.3% for all analytes and thus met the criteria for recovery, staying between 70 and 120% [28]. Recoveries of around 100% are to be expected for all analytes determined by SIDA, but also for the other quantification methods, our results proved to be very satisfactory.

Precision

Intra-day, inter-day, and inter-injection precision were determined by calculating the RSD of every analyte after a defined number of repeated measurements. Inter-day precision was evaluated by preparing one sample in triplicate on the same day. Inter-day precision was generated by analyzing one sample in triplicates weekly for three weeks. Inter-injection precision was calculated after 10 times repeatedly injecting a toxin mix in solvent (MeOH/H2O, v/v) containing all analytes. Inter-injection RSD was between 1.2 and 9.8%, thus showing the stability of the system for most analytes. The relatively high variations for AME-3-S and AME-3-G might be caused by in-source fragmentation of the analytes. All obtained precisions are shown in Table 2.

Trueness

A CRM was used to prove the trueness of the method. However, the availability of CRMs with suitable analyte/matrix combinations was scarce. We decided to use a CRM of DON in wheat because this combination was predominant in our analyses of real samples. The measured DON content of 840 ± 67 µg/kg was well in line with the reference value of 825 ± 248 µg/kg.

Sample analysis — application to samplesWheat and pseudocereals

At first, we applied the combined Fusarium and Alternaria method to 50 classical baking and food grain products such as flours of wheat, rye, and spelt, along with oats and oat flakes, millet, and maize flours. We also analyzed the pseudocereals quinoa, amaranth, and buckwheat (Table 3). As expected, the bread cereal flours were dominated by DON with a maximum content of 200 µg/kg in wheat flour. Interestingly, the plant conjugate DON-3-G was only detected in the wheat samples, although spelt or rye samples also revealed significant contents of DON (see ESM). The ratio of DON-3-G to DON was 15% in wheat samples. The other trichothecenes 3-AcDON, NIV, T-2, and HT-2 were found sporadically and only in low concentrations in these cereal products. When comparing the different groups of cereals, clustering of T-2 and HT-2 in oat samples was observed, whereas the highest content of 6.6 µg/kg T-2 was found in a millet sample. Regarding ZEN, we found this toxin in 70% of all samples outlined before, with a maximum content of 3.2 µg/kg in a corn flour sample. ENNs and BEA were also found frequently in all of these samples but with contents below 15 µg/kg. No sample exceeded the current ML for these mycotoxins in the EU [3].

Table 3 Fusarium and Alternaria toxin contents in 50 classical baking and food grain products

Of the Alternaria toxins, the most prevalent compound, TeA, showed an interesting distribution. The wheat samples were hardly affected, but the other cereal samples were frequently contaminated and showed partly significant contents. The highest amount was found in a millet sample, which was not surprising when considering our previous reports on this cereal being particularly susceptible to TeA contamination [41, 42]. TEN was also frequently found among the other Alternaria toxins, but at minor contents, mostly below 10 µg/kg. AOH and AME were found sporadically, with no preference for a specific cereal. The respective sulfate conjugates were often found parallel with the two benzopyrones.

The pseudocereals quinoa, amaranth, and buckwheat showed minor contaminations with the mycotoxins under study except for TeA, for which most of these samples also revealed detectable contents. To the best of our knowledge, this is the first report of contaminations of DON, T-2, HT-2, AME, AME-3-S, and AOH-3-S in buckwheat samples. In quinoa and amaranth, we could only detect minor ENNs and BEA contamination along with TeA, TEN, AME, and AME-3-S.

There have been numerous reports on Fusarium and Alternaria mycotoxin contamination in cereal products. A comparison with literature data is quite difficult due to the dependence of the mycotoxin occurrence on the respective climatic and geographic growth conditions. Therefore, the following literature discussion is based on European Food Safety Authority (EFSA) opinions and comparisons to data from Germany.

The latest review published by EFSA on DON and its modified forms [43] stated a maximum DON content in European grains intended for human consumption of over 20 mg/kg from Finland, which was deemed unusually high compared to other data. The other contents from European countries peaked at 4130 µg/kg DON and 1070 µg/kg DON-3-G. Compared to our results, the higher content is not unexpected as complete grains generally contain more outer parts, which are usually more contaminated with mycotoxins. The ratio between DON-3-G to DON was reported to average around 0.2 [43], which is slightly higher than our results.

Fusarium toxins in German wheat flours were already assessed in 2002 [44]. We could confirm the reported ubiquitous contamination frequency with DON. However, the mean DON content of 290 µg/kg reported by the latter authors exceeded our mean of the wheat samples of 78.8 µg/kg by a factor of almost three. The reported occurrence and level of the other Fusarium toxins 15-AcDON, 3-AcDON, NIV, T-2, HT-2, and ZEN [44] were similar to our results.

The recent EFSA review on Alternaria toxins [45] revealed in wheat or grain milling products mean concentrations of 3.9, 0.7, 27.6, and 2.1 µg/kg for AOH, AME, TeA, and TEN, respectively. For AOH, the latter data are significantly higher than our results, whereas the AME data are quite similar. The low occurrence of TeA in the wheat flours of our study was unexpected and somehow contradictory to the results reported in the literature [45]. Similarly, our mean value for TEN of 0.6 µg/kg is also below the reported values.

Reports on Fusarium and Alternaria toxin occurrence in pseudocereals are rare. One could not detect any DON, 15-AcDON, 3-AcDON, NIV, T-2, HT-2, T-2 tetraol, and ZEN in the three pseudocereals we also analyzed [46]. EFSA reported only values for buckwheat [45]. We could not detect any AOH, whereas AME showed a maximum content of 1.7 µg/kg for one of the three samples we analyzed. These results were contradictory to the latter reports as these revealed mean contents of 30.5 and 10.6 µg/kg for AOH and AME, respectively, in buckwheat [45]. For TeA, we found contents of 118 and 22 µg/kg in the two positive samples, which is in good accordance with the latter authors. For TEN, we found only one positive sample of about 1 µg/kg in buckwheat samples, which does not differ from the mean content of 1.3 µg/kg in buckwheat milling products reported by EFSA [45].

Rice and rice products

The next type of cereals we looked at were rice grains, of which we analyzed an extended variety of 58 samples, including white, brown, fragrant, and organically grown rice (Table 4). As to be expected, the occurrence of the trichothecenes was very minor. Only about 40% of the analyzed samples contained DON at a mean of 33 µg/kg in the positive samples. The only exception was one long corn rice showing a DON content of 525 µg/kg. Our results showed somewhat lower contents than data in the literature that reported a mean DON content of 107 µg/kg in German samples [47] and 139 μg/kg in Korean samples [48]. The latter authors also reported a mean of 18.9 μg/kg for Nigerian samples, which aligns with our results. In the EU regulation [3], there are no legal limits for DON in rice, but our results show that rice still has to be screened for this toxin.

Table 4 Fusarium and Alternaria toxin contents in 58 rice samples

Similarly to DON, half of all rice samples were contaminated with ZEN at a slightly higher mean concentration around 1 µg/kg compared to the other cereals and a similar maximum content of 3.4 µg/kg. Thus, the ZEN contamination in rice was lower than published by EFSA [49], where a 2.0 – 3.7 μg/kg mean content range was reported.

The distribution of the depsipeptides was interesting, as the contamination with ENNs was lower than for the other cereals. Again, the contamination of ENNs was much lower than stated by EFSA [50], where a mean concentration range for the sum of ENNs of 20.6 - 21.3 μg/kg was reported for grains for human consumption. However, the BEA content was significantly higher at an incidence of about 70%, with a mean content of about 1 µg/kg and a maximum of more than 17 µg/kg in one particular black rice sample. This concentration is even higher than the maximum concentration of 11.7 μg/kg for grains for human consumption given by EFSA [50]. As the only other black rice sample did not contain any BE