**Selected Trichothecenes in Barley Malt and Beer from Poland and an Assessment of Dietary Risks Associated with their Consumption**

#### **Edyta Ksieniewicz-Wo ´zniak 1, Marcin Bryła 1,\*, Agnieszka Wa´skiewicz 2, Tomoya Yoshinari <sup>3</sup> and Krystyna Szymczyk <sup>1</sup>**


Received: 21 November 2019; Accepted: 6 December 2019; Published: 9 December 2019

**Abstract:** Eighty-seven samples of malt from several Polish malting plants and 157 beer samples from the beer available on the Polish market (in 2018) were tested for *Fusarium* mycotoxins (deoxynivalenol (DON), nivalenol (NIV)), and their modified forms ((deoxynivalenol-3-glucoside (DON-3G), nivalenol-3-glucoside (NIV-3G), 3-acetyldeoxynivalenol (3-AcDON)). DON and its metabolite, DON-3G, were found the most, among the samples analyzed; DON and DON-3G were present in 90% and 91% of malt samples, and in 97% and 99% of beer samples, respectively. NIV was found in 24% of malt samples and in 64% of beer samples, and NIV-3G was found in 48% of malt samples and 39% of beer samples. In the malt samples, the mean concentration of DON was 52.9 μg/kg (range: 5.3–347.6 μg/kg) and that of DON-3G was 74.1 μg/kg (range: 4.4–410.3 μg/kg). In the beer samples, the mean concentration of DON was 12.3 μg/L (range: 1.2–156.5 μg/L) and that of DON-3G was 7.1 μg/L (range: 0.6–58.4 μg/L). The concentrations of other tested mycotoxins in the samples of malt and beer were several times lower. The risk of exposure to the tested mycotoxins, following the consumption of beer in Poland, was assessed. The corresponding probable daily intakes (PDIs) remained a small fraction of the tolerable daily intake (TDI). However, in the improbable worst-case scenario, in which every beer bottle consumed would be contaminated with mycotoxins present at the highest level observed among the analyzed beer samples, the PDI would exceed the TDI for DON and its metabolite after the consumption of a single bottle (0.5 L) of beer.

**Keywords:** *Fusarium* toxins; modified mycotoxins; beer; malt; risk assessment

**Key Contribution:** High number of malt and beer samples were contaminated with mycotoxins. Strong beers (with higher alcohol content) contain higher levels of mycotoxins. Risk analysis showed a low level group probable daily intake of mycotoxin from beer. DON-3G present in beer has a significant share in group exposure to mycotoxins.

#### **1. Introduction**

Barley (*Hordeum vulgare* L.) has been grown for many years and is of great economic importance [1]. Approximately 57 million tonnes of barley was produced annually (in 2018) in the European Union, while global production has reached 147 million tonnes annually [2]. Most of the harvested grain is used as feed but the highest quality barley is selected for food production, including the production of

malt. Malt is an ample source of the B-group vitamins, niacin, and minerals. It is increasingly used in the bakery and pastry industries to improve the quality of both the taste and health of their products [3]. However, beer production remains as the main application of malt [1,4]. Beer is an alcoholic beverage commonly consumed in numerous countries globally. Poland has the third largest quantity of beer production in Europe (approximately 93, 40.5, and 40.4 million hectoliters in Germany, UK, and Poland, respectively) and the fourth highest beer consumption per capita in Europe (approximately 138, 105, 101, and 97 liters in Czech Republic, Austria, Germany, and Poland, respectively [5]).

To arrive at a high-quality malt, one needs to start with a healthy grain with sufficiently high energy for germination and sufficient protein content. However, unfavorable climatic conditions during the plant vegetation season may negatively impact the quality of the grain and consequently, the decrease quality of the malt produced from that grain [6]. The most important climatic conditions are rainfall and temperature, which are two factors that mostly determine the degree to which the plants may become infected with pathogen fungi. *Fusarium* is one of the major fungal species infecting cereal grains, including barley. *Fusarium* head blight (FHB) disease caused by these fungi is a problem in various regions of the world. The fungal infection decreases crop yield, but even greater damage may result from the production of mycotoxins, which are secondary metabolites of the fungi that are toxic to humans and animals [7].

*Fusarium* spp. most often responsible for FHB in Poland include *F. graminearum*, *F. avenaceum*, and *F. culmorum*; however, other species are also seen in various regions of the world [8–10]. The mycotoxins produced by *Fusarium* in cereal grains include the trichothecenes, deoxynivalenol (DON), and nivalenol (NIV), and their modified forms. These toxins are also phytotoxic [11,12]. *F. culmorum* and *F. graminearum* are among the varieties that most aggressively infect plant ears [13,14]. Many of these fungi are capable of synthesizing 3- (3-AcDON) or 15-acetyl deoxynivalenol (15-AcDON), which are modified forms of DON [15]. Studies of the phytotoxic effects of DON have shown that the ability to covert DON into deoxynivalenol-3-glucoside (DON-3G) is the plant's primary defense mechanism against the toxin. Similar metabolic detoxication mechanisms help to build resistance to toxins in numerous cereal grain plants [16]. In barley, this mechanism is thought to be controlled by the QTL (quantitative trait loci)-specific region. Future studies involving deeper genetic analyses may help to develop tools to select fungal toxin-resistant plants using specific markers (marker-assisted selection; [17]). The phytotoxic effects of DON-3G are very weak compared to DON [18] and thus, it may be expected that a similar relationship holds for nivalenol 3-glucoside (NIV-3G) and NIV.

The consumption of DON- and/or NIV-contaminated food/feed may lead to disorders of the gastrointestinal tract, reproductive organs, and/or the immune system in both humans and animals. The toxicological characteristics of these toxins have been extensively described [19]. The lower levels of toxicity of DON-3G compared with DON have been confirmed in both humans and animals. In some in vitro studies and in some research on animals, it has been shown that DON-3G is not transported through the intestinal epithelium, but rather, is hydrolyzed by bacteria within the lower part of the alimentary tract [20]. Similar data are not available for NIV-3G, but it is commonly thought that the adverse effects of NIV-3G are weaker than those of NIV, as they are for DON-3G and DON.

Currently, the only European Commission regulation concerning mycotoxins in foodstuffs requires that the DON concentration in unprocessed cereal grains must not exceed 1250 μg/kg [21]. Taking into consideration the scientific evidence regarding the rapid absorption and excretion of DON, the in vivo deacetylation of 3- and 15-AcDON, and the hydrolysis of DON-3G in the lower parts of the alimentary tract; a European Food and Safety Authority (EFSA) expert panel recognized in 2017 that the toxic effects of DON-derivatives in humans may be comparable to the toxic effects of DON. Therefore, the tolerable daily intake (TDI) and reference dose (RfD) values have been recalculated as the sum of the three latter substances. Based on epidemiological data, a TDI threshold of 1 μg/kg body weight/day and an RfD dose of 8 μg/kg body weight/day have been accepted [19].

Reports on mycotoxins and their metabolites in Polish malts used in the brewing industry are very limited. The aims of this work included: (i) to assess the contamination of malts, sampled from several Polish malting plants, with selected *Fusarium* mycotoxins including their modified forms; (ii) to assess the mycotoxin contamination of beer available in 2019 on the Polish market; and (iii) to assess the risk of exposure to these mycotoxins following the consumption of beer in Poland.

#### **2. Results and Discussion**

#### *2.1. Malt*

Mycotoxins were found in the majority of the malt samples analyzed (Table 1). DON and DON-3G were found most often (in 90% and 91% of the malt samples, respectively) and at the highest levels (average of 52.9 and 74.1 μg/kg for DON and DON-3G, respectively). The percentage of samples positive for 3-AcDON was clearly lower 59% and NIV and NIV-3G were detected in the least number of samples (24% and 48%, respectively). DON-3G/DON molar ratios varied from 22% to 186% among DON-positive samples, while NIV-3G/NIV molar ratios varied from 32% to 126% among NIV-positive samples. Individual results regarding the content of individual mycotoxins in malt samples are presented in Table S1.


DON, deoxynivalenol; DON-3G, deoxynivalenol-3-glucoside; 3-AcDON, 3-acetyldeoxynivalenol; NIV, nivalenol; NIV-3G, nivalenol-3-glucoside.

In grains, DON-3G is known to be a product of the plant defense reaction to the presence of the phytotoxin, DON [22–24]. DON-3G is easily soluble and plants can easily transport it from the cytoplasm to vacuoles or the intercellular space [16]. The DON-3G/DON ratio in the grain itself does not usually exceed 30% [25,26]. However, in malt samples we observed an average DON-3G/DON ratio of 89%, with a range of 22%–186%. Relatively high values (average 65%, range 32%–126%) were also noted for the NIV-3G/NIV ratio. Some researchers have suggested that changes occur during the malting process that activate secondary detoxicating enzymes, which then catalyze the conversion of the toxins to their glycoside derivatives [27–29]. Maul et al. [29] have shown that sprouting seeds of barley, millet, oat, rye, and spelt are capable of converting DON into DON-3G by means of UDP-glucosyltransferases. In barley, approximately 50% of DON was found to be converted, mainly into DON-3G, with a similar conversion rate observed in wheat. Moreover, Lancova et al. [28] reported that, during barley grain germination, the concentration of DON may decrease by 90%, while the concentration of DON-3G may markedly increase, to a level as high or several times higher than DON. Spanic et al. [30] presented data on mycotoxin levels in wheat varieties varying in Fusarium head blight resistance; the average content of DON-3G increased from 59.9 μg/kg in grain to 163.9 μg/kg in malt.

There are very few reports in the literature on the co-occurrence of DON/DON-3G and NIV/NIV-3G in brewing malts, even though such data are essential for regulating food safety. In the present study, we detected these substances in both malt and beer samples. However, the DON concentration did not exceed 750 μg/kg, the maximum permissible level in malt specified in EC Regulation 1881/2006, in any of the tested malt samples [21]. Practically, malt plants in Poland do not purchase grain contaminated with DON at levels above 1 mg/kg, while the maximum permissible level in grain is 1.25 mg/kg, as per EC Regulation 1881/2006 [21]. Mitteleuropäische Brautechnische Analyskomommision [31] recommends the inspection of each batch of grain offered to a malting plant for the presence of *F. graminearum* and

*F. culmorum*. If mycelia are visible, they recommend the analysis of the grain for mycotoxins. There are some indications in the literature [32–35] that high amounts of additional mycotoxins may be synthesized in fungi-contaminated grain during the malting process, thus significantly impacting food safety.

#### *2.2. Beer*

The majority of beers marketed in Poland are light beers based on pilsner malts. However, dark ale or lager beers produced from Munich malts, usually obtained from lower quality grains [36], caramel malts or roasted pale ale malts are also popular. The two latter malts are enzymatically inactive; they are introduced in small amounts [37], to darken the beer and enhance its flavor. Wheat beers are also becoming increasingly common on the market. They are produced from barley malt, with the addition of at least 50% wheat or wheat malt. The flavor of these beers is unique, differing from the flavor of classical barley-only beers [38]. We divided our beer samples into three common categories for analysis: light, dark, and wheat beers. The percentage of mycotoxin-positive beer samples in all these groups was high (Table 2). Individual results regarding the content of mycotoxins in beer samples are presented in Table S1.


DON, deoxynivalenol; DON-3G, deoxynivalenol-3-glucoside; 3-AcDON, 3-acetyldeoxynivalenol; NIV, nivalenol; NIV-3G, nivalenol-3-glucoside.

As was the case for malt samples, DON and DON-3G were the most frequently found toxins in beer samples, being present in 96% and 98% of light beer samples, respectively, and in all the samples of dark and wheat beers. Other mycotoxins, namely, 3-AcDON, NIV, and NIV-3G were found at lower levels in 69%, 25%, and 58%; 67%, 54%, and 63%; and 43%, 25%, and 42% of the light, dark, and wheat beer samples, respectively. The maximum DON (156.5 μg/L) and DON-3G (58.4 μg/L) concentrations were found in a light and a dark beer sample, respectively. The average levels of the three remaining tested mycotoxins ranged from 0.7 to 1.5 μg/L, i.e., they were approximately 6–20 times lower than the DON levels. The average DON-3G/DON and NIV-3G/NIV molar ratios ranged from 34% to 46% and 41% to 50%, respectively. Neither the mycotoxin concentrations nor their molar ratios were dependent on the beer category.

The alcohol content of beer depends on the extent to which the yeast ferments the sugars, which largely depends on the amount of grain and malt in the fermentation batch. Stronger beer requires more grain, which results in a higher risk of mycotoxin contamination [27,39,40]. Grain extracts used for beer production contain mainly sugars but may also contain dextrins, nitrogenous compounds (proteins), mineral salts, and other compounds, depending on the recipe used by the beer manufacturer [41]. Therefore, a comparison of the level of mycotoxin contamination in beers with different extract contents must be treated only as an approximation. Therefore, we re-organized the beer samples into three different categories: mild beers (0.5–5.0% alcohol, 3.5–12.5% extract), regular beers (5.1–6.0% alcohol, 6.8–16.0% extract), and strong beers (6.1–10.0% alcohol, 8.4%–21.0% extract; Table 3).


**Table 3.** Concentrations of mycotoxins in mild, regular, and strong beers.

DON, deoxynivalenol; DON-3G, deoxynivalenol-3-glucoside; 3-AcDON, 3-acetyldeoxynivalenol; NIV, nivalenol; NIV-3G, nivalenol-3-glucoside.

The number of positive samples and the concentration of the majority of the tested mycotoxins positively correlated with alcohol content in most cases. DON and DON-3G were the predominant toxins in 94% and 98% of mild beer samples, respectively, and in all samples of regular and strong beer, with average DON concentrations of 7.1, 12.1, and 17.3 μg/L and average DON-3G concentrations of 5.6, 7.0, and 8.6 μg/L for mild, regular, and strong beers, respectively. Less clear, but similar trends were noted for the other tested mycotoxins.

Mycotoxin contamination of beer has been studied by numerous groups (Table 4). However, data on the co-occurrence of DON, DON-3G, 3-AcDON, NIV, and NIV-3G in beer are scarce. The scope of most reported studies has been restricted to DON, DON-3G, and 3-AcDON, with a few studies also including NIV. Typically, the reported concentrations of the predominant DON have not exceeded 100 μg/L [27,42–45]. The findings from the present study mostly agree with those from previous studies (because the fraction of positive samples may depend on the LOD and LOQ of the method used). Higher concentrations of DON have been found mainly in beers originating from non-European countries, including craft beers from Brazil (127–501 μg/L; [46]), traditional African beers from Cameroon (140–730 μg/L; [47]), and Busaa-type beers from Kenya (200–360 μg/kg [48]). However, relatively high DON concentrations (104–182 μg/L) have also been found in strong (>8% alcohol) Norwegian Imperial Stout beer [49]. In this study, we found a high DON concentration (156.5 μg/L) only in one strong (>8% alcohol) sample of a light beer.


**Table 4.** Selected literature data on mycotoxins in beer.


**Table 4.** *Cont.*


**Table 4.** *Cont.*

Some of the beer samples tested had a higher concentration of DON-3G than DON. Similar DON-3G/DON molar ratios have been reported in the literature, with averages of 0.56 (range 0.11–1.25 [43]) and 0.79 (range 0.1–2.6 [49] and 0.7–1.0 [26]. As can be seen, the DON-3G/DON molar ratios in beer are similar to those in malt.

#### *2.3. Dietary Exposure Assessment*

The following group TDI values were used in the assessment of risk of exposure to mycotoxins following beer consumption: 1 μg/kg body weight/day of the sum of DON, DON-3G, 3-AcDON, and 15-AcDON [20] and 1.2 μg/kg body weight/day of the sum of NIV and NIV-3G [50]. The average beer consumption in Poland is 97 L per capita annually, i.e., 0.27 L per capita per day [5]. In three considered scenarios, it was assumed that consumed beer contained mycotoxins at a level equal to: (i) the median, (ii) the third quartile, or (iii) the maximum concentration found in our samples (the worst-case scenario). It was assumed that the average adult in Poland weighs 70 kg. The results of the calculations are shown in Table 5. PDI values remained a small fraction of TDI values in the first and second scenarios (5.1% and 7.9%, respectively, for DON and its derivatives and 0.32% and 0.61%, respectively, for NIV and its derivatives). In the improbable third scenario (worst case), the PDI would reach 65.2% of the TDI for DON and its derivatives and 2.41% of the TDI for NIV and its derivatives.


**Table 5.** Group probable daily intake and its share of the total daily intake calculated in three scenarios, in which different concentrations of mycotoxins were assumed in the consumed beer.

DON, deoxynivalenol; DON-3G, deoxynivalenol-3-glucoside; 3-AcDON, 3-acetyldeoxynivalenol; NIV, nivalenol; NIV-3G, nivalenol-3-glucoside; PDI, probable daily intake; TDI, total daily intake; \* PDI = *<sup>C</sup>*∗*Cd <sup>b</sup>*.*w*. , where C is concentration of the mycotoxin in the contaminated beer, Cd is the average daily consumption of beer in Poland, and b.w. is mean body weight. \*\* If the measurement for any analyte was below the LOQ, the median and 3rd quartile were calculated assuming that the analyte was present at the level of LOQ/2.

The average consumption of 0.27 L of beer per day assumed in the above dietary exposure assessment does not reflect the real situation, since beer consumers rarely drink less than one bottle (0.5 L) per day. The PDI for persons drinking 0.5 L of beer daily would be approximately twice the values calculated above, in which case the TDI of DON and its derivatives would exceed the worst-case scenario by approximately 30%. Each additional beer bottle consumed per day would double the above calculated PDI values. It is also worth noting that the analytical method developed here was not efficient at detecting 15-AcDON. However, since 3-AcDON was detected at very low levels, one can expect that the contribution of 15-AcDON to the PDI is insignificant.

Of course, beer is not the main source of DON and its derivatives (the most important trichothecenes from a food safety point of view) in the human diet. Greater levels of exposure come from the consumption of bakery products, corn flakes, pasta, and other grain-based foodstuffs that are consumed daily, not only by beer consumers. Considering the exposition, bakery products and pastas are in Europe more and more often indicated as a possible quite serious threat to human health [50]. Studies of markers in urine have shown that chronic exposure to DON and its derivatives is greater than the accepted TDI [51–53]. Therefore, the consumption of beer may increase the risk of excessive mycotoxin exposure.

Data on the risks associated with the consumption of mycotoxin-contaminated beers exist only with respect to officially regulated toxins. It is a common observation that DON is the greatest risk factor, but beer is not generally considered an important source of dietary mycotoxin exposure. Even if the maximum detected DON concentrations are taken into account, the PDI values remain a small percentage of the TDI values, regardless of the country of origin of the beer. For example, the PDI is 14.0–20.8% of the TDI in Poland [54]; 18% of the TDI in Brazil [46]; 0.15–6.14% of the TDI in Spain, where the average consumption is just half of that in Poland [55]; 0% of the TDI in Cyprus and 10% of the TDI in Ireland [56].

The consumption of mycotoxin-contaminated beer results in negligible risk of exposure to NIV and NIV-3G. EFSA has reported that even the consumption of bakery products and pasta is safe in terms of exposure to these toxins [57]. In view of the low concentrations of NIV and NIV-3G, the PDI values are far below the TDI values, even for foodstuffs that are consumed in relatively large quantities, such as bakery products and pasta.

#### **3. Conclusions**

The data presented here on the co-occurrence of DON, NIV, and their metabolized (masked) forms in brewing malts and beers available on the Polish market are among the first reported in the literature. Mycotoxins were found in the majority of the barley malt and beer samples tested. DON and its metabolite, DON-3G, were found most frequently (in more than 90% of samples), although at safely low levels. NIV and its metabolite, NIV-3G, were found at lower levels in malt and beer samples. Because of the low mycotoxin levels, none of the tested beers were regarded as unsafe from a toxicological point of view. However, in the worst-case scenario, the PDI would exceed the TDI for DON and its metabolites after drinking just one bottle (0.5 L) of beer.

#### **4. Materials and Methods**

#### *4.1. Reagents and Standards*

Certified reference standards of DON, 3-AcDON, and NIV (100 μg/mL in acetonitrile), and DON-3G (50 μg/mL in acetonitrile:water, 50:50, *v*/*v*), were purchased from Romer Labs (Tulln, Austria). NIV-3G (110 μg/mL) was isolated from wheat, according to the procedure described by Yoshinari et al. [58]. Acetonitrile, methanol, and LC/MS-grade water were purchased from Witko (Łód´z, Poland). Ammonium formate and formic acid (LC-MS grade) were obtained from Fisher Scientific (Millersburg, PA, USA). DON-NIV wide-bore (WB) immunoaffinity columns and PBS buffer solutions were purchased from Vicam (Watertown, NY, USA).

#### *4.2. Research Material*

One hundred and fifty-seven beer samples and 87 barley malt samples were analyzed. Various brands of light, dark, and wheat beers (mild, regular, and strong) were purchased in 2019 from local supermarkets in Poland. Malt was sampled from various malt plants located throughout the country, in line with the guidelines specified within EC Regulation 519/2014 (February 23, 2006) [59], which describes sampling and analysis methods for the official control of mycotoxin levels in foodstuffs. All the acquired samples belonged to the most common Pilsner malts, which are used to produce pale straw-colored ale and lager beers [36]. Malt samples, each with a mass of approximately 1 kg, were ground in a Knife Mill Grindomix GM 200 grinder (Retsch GmbH, Haan, Germany).

#### *4.3. Sample Preparation*

Malt and beer samples were prepared for analysis using a method previously described by our research team [42,60]. After extraction and homogenization (for malt extraction in Unidrive 1000 homogenizer, CAT Scientific Inc., Paso Robles, CA, USA), each sample was passed through a DON-NIV WB immunoaffinity column at a speed of 1–2 drops/s. The column was rinsed with 10 mL of PBS and 10 mL of de-ionized water. Analytes were washed out of the column, first with 0.5 mL of methanol and then with 1.5 mL of acetonitrile and were collected into a reaction vial. The solvent was evaporated in a stream of nitrogen. The residues were re-dissolved in 300 μL of 30% methanol and analyzed by liquid chromatography-mass spectrometry (LC-MS). Samples were analyzed at three replications.

#### *4.4. LC-MS Analysis*

An H-class liquid chromatograph coupled to a mass spectrometer with a time-of-flight analyzer (UPLC-TOF-HRMS; Waters, Milford, MA, USA) was used to analyze mycotoxins. Analytes were separated on a 2.1 × 100 mm, 1.6 μm UPLC C18 Cortecs chromatographic column (Waters) with an appropriate pre-column, operated with a gradient regime. Phase A was 90:10 *v*/*v* methanol:water, phase B was 10:90 *v*/*v* methanol:water. Both phases contained 0.2% formic acid and 10 mM ammonium formate. The flow rate was 0.3 mL/min, with the following flow gradient: 0–2 min, 100% B; 3–6 min, 50% B; 22–23 min, 100% A; and 25–28 min, 100% B. Five microliters of each sample was injected onto the column. The mass spectrometer was operated in the positive/negative electrospray ionization mode, with an ion source temperature of 150 ◦C and a desolvation temperature of 300/350 ◦C for positive/negative ionization, respectively. The nebulizing gas (N2) flow rate was 750 L/min and the cone gas flow rate was 40 L/min. The capillary bias was 3200 V. Ion optics was operated in V mode and the instrument was calibrated using a leucine-enkephalin solution.

#### *4.5. Method Validation*

Linearity ranges, limits of detection (LOD, the concentration at which the signal:noise ratio was 3), limits of quantification (LOQ, the concentration at which the signal:noise ratio was 10), recovery rates (R), and repeatability/precision (expressed as the relative standard deviation [RSD]), were determined

using calibration curves that were constructed using separate blank samples for each mycotoxin of interest in the beer and malt matrices. The blanks were prepared in the same way as the analytes, except that the respective amount of standard mixture was added just prior to finally dissolving it in 30% methanol, after which the solvent was removed in a dry nitrogen stream. Each calibration curve consisted of eight points. The concentrations covered for the malt samples (in μg/kg) were: 5.0–1028 for DON; 4.0–516 for DON-3G; 2.0–1028 for 3-AcDON; 8.0–1050 for NIV; and 5.0–565 for NIV-3G. The concentrations covered for the beer samples (in μg/L) were: 3–68.6 for DON; 2.1–34.4 for DON-3G; 0.9–68.6 for 3-AcDON; 2.1–70.1 for NIV; and 1.6–37.7 for NIV-3G. The results of the analytical method validation experiment are shown in Tables 6 and 7.

**Table 6.** Limits of detection, limits of quantification, and determination coefficients for individual analytes determined in malt and beer samples.


DON, deoxynivalenol; DON-3G, deoxynivalenol-3-glucoside; 3-AcDON, 3-acetyldeoxynivalenol; NIV, nivalenol; NIV-3G, nivalenol-3-glucoside; LOD, limit of detection; LOQ, limit of quantification; *R2*, determination coefficient.


**Table 7.** Recovery rates and relative standard deviations for individual analytes determined in malt and in beer samples spiked at different fortification levels.

DON, deoxynivalenol; DON-3G, deoxynivalenol-3-glucoside; 3-AcDON, 3-acetyldeoxynivalenol; NIV, nivalenol; NIV-3G, nivalenol-3-glucoside; R, recovery rate; RSD, relative standard deviation.

Since all analytes of interest belonged to the trichothecenes group, we assessed the performance of the method for DON analysis using the following specifications listed in EC Regulation 519/2014 [59]: recovery rates 60%–110% or 70%–120%, depending on the fortification level and RSD ≤20%. These

criteria were met in 34 out of 35 analyte/fortification level combinations. In one case, the RSD was above 20%.

This validated method was then used to analyze DON, DON-3G, 3-AcDON, NIV, and NIV-3G in the malt and beer samples.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6651/11/12/715/s1, Table S1: Individual results of mycotoxin concentrations in the analyzed beer and malt samples.

**Author Contributions:** Conceptualization, E.K.-W. and M.B.; methodology, M.B. and A.W.; formal analysis, M.B. and E.K.-W.; performed isolation of the NIV-3G analytical standard, T.Y.; performed the manuscript preparation, M.B., E.K.-W. and T.Y.; supervised the research, A.W., K.S.

**Funding:** This research was financially supported from the Polish National Science Centre project 2016/21/D/NZ9/02597.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Acrylamide Reduction Strategy in Combination with Deoxynivalenol Mitigation in Industrial Biscuits Production**

#### **Michele Suman 1,\*, Silvia Generotti 1,2, Martina Cirlini <sup>2</sup> and Chiara Dall'Asta <sup>2</sup>**


Received: 12 August 2019; Accepted: 22 August 2019; Published: 27 August 2019

**Abstract:** Acrylamide is formed during baking in some frequently consumed food products. It is proven to be carcinogenic in rodents and a probable human carcinogen. Thus, the food industry is working to find solutions to minimize its formation during processing. To better understand the sources of its formation, the present study is aimed at investigating how acrylamide concentration may be influenced by bakery-making parameters within a parallel strategy of mycotoxin mitigation (focusing specifically on deoxynivalenol—DON) related to wholegrain and cocoa biscuit production. Among Fusarium toxins, DON is considered the most important contaminant in wheat and related bakery products, such as biscuits, due to its widespread occurrence. Exploiting the power of a Design of Experiments (DoE), several conditions were varied as mycotoxin contamination levels of the raw materials, recipe formulation, pH value of dough, and baking time/temperature; each selected treatment was varied within a defined range according to the technological requirements to obtain an appreciable product for consumers. Experiments were performed in a pilot-plant scale in order to simulate an industrial production and samples were extracted and analysed by HPLC-MS/MS system. Applying a baking temperature of 200 ◦C at the highest sugar dose, acrylamide increased its concentration, and in particular, levels ranged from 306 ± 16 μg/Kg d.m. and 400 ± 27 μg/Kg d.m. in biscuits made without and with the addition of cocoa, respectively. Conversely, using a baking temperature of 180 ◦C in the same conditions (pH, baking time, and sugar concentrations), acrylamide values remained below 125 ± 14 μg/Kg d.m. and 156 ± 15 μg/Kg d.m. in the two final products. The developed predictive model suggested how some parameters can concretely contribute to limit acrylamide formation in the final product, highlighting a significant role of pH value (correlated also to sodium bicarbonate raising agent), followed by baking time/temperature parameters. In particular, the increasing range of baking conditions influenced in a limited way the final acrylamide content within the parallel effective range of DON reduction. The study represents a concrete example of how the control and optimization of selected operative parameters may lead to multiple mitigation of specific natural/process contaminants in the final food products, though still remaining in the sensorial satisfactory range.

**Keywords:** acrylamide; deoxynivalenol; multiple mitigation strategies; design of experiments; bakery food processing; biscuits

**Key Contribution:** The study here reported represents a concrete example of how the control and optimization of selected operative industrial parameters may lead to multiple mitigation of specific natural/process contaminants in the final food products: in this specific case Deoxynivalenol & Acrylamide within the contest of biscuits production.

#### **1. Introduction**

Process-related mitigation strategies are among the most promising tools for controlling and minimizing mycotoxins, i.e., Fusarium mycotoxins, in cereal-based products [1]. Among Fusarium mycotoxins, deoxynivalenol (DON), along with its modified forms, is considered the most important contaminant in wheat and related bakery products, such as biscuits, due to its widespread occurrence [2]. It has been shown that milling and baking/roasting may effectively reduce the amount of DON in the bakery. At this purpose, Generotti et al. reported on the strategic mitigation of DON and its related compounds deoxynivalenol-3-glucoside (DON3Glc) and culmorin, during two different biscuit processes, exploiting the synergistic effects of recipe formulation and thermal treatment [3]. A significant reduction was achieved while not negatively impacting product quality and identity. However, increased time and temperature during baking might favor the formation of process-derived contaminants, such as acrylamide, in the final product.

Besides often reported among the most studied contaminants in cereal-based products, mycotoxins and acrylamide are almost unavoidable compounds, being the former natural toxins accumulated in crops under natural field conditions, and the latter process-related compounds formed in food during manufacturing, as a consequence of chemical reaction triggered by processing.

Although their toxic effects have been deeply studied as single compounds, little to nothing is known about combined effects exerted in animals and humans. Nonetheless, the scientific community is posing increasing emphasis on health concern related to the exposure to chemical mixtures, as recently reported by the European Food Safety Authority [4]. Therefore, the identification of possible strategies for the simultaneous mitigation of mycotoxins and acrylamide, are of upmost interest for the agro-food sector.

Concerning process-related compounds, food industry and the European Commission have undertaken extensive efforts since 2002, when scientists from the Swedish National Food Authority and the University of Stockholm reported high levels of acrylamide in normally cooked starch-rich food (compared to what had been reported earlier in other food commodities), in order to investigate pathways of formation and to reduce its levels in processed food.

Acrylamide is typically formed from an amino acid, primarily asparagine, and a reducing sugar such as fructose or glucose in starchy food products during high temperature cooking, including frying, baking and roasting through a series of reactions, known as Maillard reactions. Its formation starts at temperatures around 120 ◦C and peaks at temperatures between 160 and 180 ◦C [5,6]. Due to its toxicity and possible carcinogenic effects to humans (IARC—International Agency for Research on Cancer [7]), the European Commission has established mitigation measures and benchmark levels for its concrete reduction in food [8,9].

Several attempts have been made so far to develop mitigation strategies for acrylamide formation in bakery products, mainly focused on the processing stage and recipes [10–17]. Each strategy could present limiting factors in their applicability depending on the product type and industrial settings as feasibility and compatibility with processing, formulation, impact on sensory and nutritional characteristics, regulatory compliance and costs.

Acrylamide formation is favored by a high baking temperature and time treatment [17–23], therefore a decreasing of thermal input represents an effective way of mitigation. Reduction can be obtained by applying prolonged heating at lower temperatures, or at lower pressure than the atmospheric one [16] or by optimizing the oven temperature profile. On the other hand, a decreased thermal input may significantly affect the achievement of both appropriate hygienic properties and sensorial acceptance of the final product.

Moving from our previous studies [24–27], the present investigation is aimed at verifying how acrylamide concentration in bakery products, such as wholegrain and cocoa biscuits, is affected by modifications of technological parameters (recipe formulation and baking time/temperature) during biscuit-making process (Figure 1), while at the same targeting potential DON reduction and without affecting the sensory properties.

**Figure 1.** Scheme of wholegrain and cocoa biscuit production.

In this regard, predictive models would represent a time and cost-saving tool for finding the most suitable conditions for minimizing both natural occurring and process-related contaminants. In the present work, we have exploited them to estimate acrylamide evolution in a considered system and to manage the industrial process to optimize the role of involved parameters with regard to its formation.

Starting from naturally DON contaminated raw material, the experiments were performed using statistical Design of Experiment (DoE) schemes to explore the relationship between the analytical responses and independent variables conducting to an overall optimization of the baking process [28]. In particular, this approach was conducted in order to consider only those modifications that can be really applied to the industrial scale, obtaining a final product appreciable by consumers and remaining in an acceptable technological range.

#### **2. Results and Discussion**

The present study was carried out to better understand whether and how acrylamide evolution could be influenced by selected technological factors within a strategy of DON mitigation related to wholegrain and cocoa biscuit production. Different parameters were considered and a variation range was defined, as shown in Table 1. The statistical model required 19 single experiments per technological process; acrylamide level was measured by LC-MS/MS according to a previously published method with slight modifications [29].


**Table 1.** Wholegrain and cocoa biscuit experiments—pilot plant processing conditions.

#### *2.1. Statistical Elaboration of the Experimental Model*

At the end of the analysis performed on the collected samples, MODDE software extracted an answer concerning robustness and prediction capability of the method, evaluating model efficiency by fitting (R2) and prediction (Q2) values. Replicate values of acrylamide reduction per experiment were expressed on dry matter basis and averaged for the statistical elaboration. Partial least-squares (PLS) was chosen as the statistical regression treatment. The two statistical models (wholegrain and cocoa biscuit model) gave high fitting (*R*2 > 0.7) and good prediction value (Q2 ≥ 0.5), referring to the MODDE software output settings. Model robustness was also confirmed by ANOVA plot (Figure 2), being standard deviation of the regression much larger than standard deviation of the residuals.

**Figure 2.** Design of Experiments on acrylamide levels within cocoa biscuit-making—Model Robustness by ANOVA plot.

#### *2.2. Acrylamide Evolution within Biscuit-Making Technology*

All values were collected in a Variable Importance Plot (VIP, Figure 3) that enables (and illustrates in an effective and condensed way) an understanding of the effect of each factor in terms of influence on acrylamide response.

In the present case, VIP would suggest that the pH value, has the most relevant effect on the final acrylamide level: in fact pH increase is responsible for an acceleration of the reaction between asparagine and the reducing sugars, followed by baking time/temperature parameters.

With regard to the other minor ingredients, dextrose (or glucose) content confirm (as reported in previous scientific literature findings) to contribute to the overall acrylamide increase. When high dextrose level and higher thermal input are employed (200 ◦C for 8 min) an acrylamide increase up to 120% was observed (data not shown) with respect to the central point. On the other hand, a combination of lower dextrose content and moderate thermal input (180 ◦C for 8 min) may lead to a reduction up to 77%.

This is consistent with literature studies in which it was demonstrated that increasing the quantity of sugars in cookies formulation, as glucose and sucrose, the concentration of acrylamide raised up especially when a temperature of 205◦ was applied for 11 min. Moreover, the choice of the sugar resulted crucial for keeping acrylamide formation under control. Using glucose, a reducing compound, and applying the same baking conditions, acrylamide content increased more in respect to sucrose. So, the authors suggested to use sucrose in order to obtain a reduction of about the 50% of acrylamide production [30,31]. Similar results were showed by Vass et al. [32] whom replaced invert sugar syrup with sucrose in wheat crackers, obtaining an acrylamide reduction by 60%. In addition, the same studies speculated that applying baking temperatures of 160◦C the formation of acrylamide remains

below 150 μg/Kg [30]. Also in this study, the measured amount of acrylamide presented values below or close to 150 μg/Kg when the baking temperature did not exceed 180 ◦C (Table 2).

**Figure 3.** Variable Importance Plot (VIP) obtained for the data referred to cocoa biscuit-making process: influence of each main factor with respect to acrylamide response: pH, ct (cooking time), Temperature (Temp), Egg.


**Table 2.** Analytical results for acrylamide levels throughout several wholegrain and cocoa biscuit-making process trials.

\* Data expressed as mean value ± standard deviation. § Data reported and discussed elsewhere [3].

Furthermore, some authors showed how acrylamide mitigation can be achieved by adding amino acids or protein-based ingredients to food, which may influence the reaction pathway or favor AA degradation [33–35]. In our study, no significant mitigation was obtained probably due to the close variation range related to milk/egg content in the recipe formulation.

Basically, acrylamide content seems to be affected by the food matrix, being higher for the cocoa biscuits than for the wholegrain biscuits. This could be related to the additional acrylamide load related to cocoa beans roasting (Table 2). It was indeed demonstrated that acrylamide could be present in roasted cocoa beans in the order of mg/Kg and the content may depend form the temperature used during roasting [36].

With regard to the sodium bicarbonate content and the correspondent pH variation, a potential reduction in acrylamide level higher than 50% could be achieved in the finished product, remaining within an acceptable range from the sensorial point of view. Sodium bicarbonate could indeed limit the formation of acrylamide if compared to other raising agents as ammonium hydrogen carbonate. This was demonstrated during experiments conducted on biscuits, in which also the addition of tartaric and citric acids was tested in order to reduce acrylamide content. The authors showed that a reduction of about 70% of acrylamide content was achieved when sodium bicarbonate was applied [31].

Taking into account the previously mentioned mycotoxin mitigation strategy, baking time/temperature play an important role in order to achieve a parallel significant DON mitigation; considering experiments carried out within the most severe time/temperature baking conditions, the greatest effect was observed with the baking step being performed at 200 ◦C for 8 min. In particular, an increase in time during the baking phase, in an acceptable technological range, can effectively reduce DON content in the final product [3].

Notably, an acrylamide reduction ranged from 77% to 100% was achieved in the finished product when baking was conducted at 180 ◦C for 5 min, though still remaining in the sensorial satisfactory range.

Overall, data collected within this study allow to proper set the baking time and temperature for controlling mycotoxin and acrylamide content in the finished product, as suggested by the response Contour Plot (Figure 4). The increase of baking parameters (which also goes in the direction of the obvious industrial requirement to speed-up the process, in order to increase the productivity as much as possible) within a range of mycotoxin mitigation (up to 20% of reduction) affects in a restricted manner the final acrylamide content (difference of the values of acrylamide expressed as predicted increase, named "delta acrylamide" in the figure caption), without implications on the organoleptic properties and consumer safety.

As a major outcome of this study, this allows one to design proper synergistic mitigation strategies for multiple contaminants along the food production chain.

**Figure 4.** Contour Plot: delta-acrylamide (predicted increase) values of the cocoa biscuit-making experiments; Temperature vs. cooking time.

#### **3. Materials and Methods**

#### *3.1. Chemicals*

Methanol and formic acid (p.a.), both HPLC gradient grade, were obtained from BDH VWR International Ltd. (Poole, UK). Acetonitrile was purchased from J.T. Baker (Deventer, The Netherlands) and ammonium acetate (MS grade) and glacial acetic acid (p.a.) were obtained from Sigma-Aldrich (Vienna, Austria). Standard acrylamide solution was purchased from Sigma-Aldrich (Milan, Italy). Acrylamide internal standard (13C3-acrylamide, 1 mg/mL in methanol) was obtained from Cambridge Isotope Laboratories, Inc. (Andover, MA, USA). Deionized water was used for all procedures. Water was purified successively by reverse osmosis and a Milli-Q plus system from Millipore (Molsheim, France). Deoxynivalenol standard was obtained from RomerLabs®Inc. (Tulln, Austria). OASIS® HLB 3 cc (60 mg) extraction cartridges were purchased from Waters (Manchester, UK). Glass vials with septum screw caps were purchased from Phenomenex (Torrance, CA, USA). Centrifugal filter units (Ultrafree MC 0.22 mm, diameter 10 mm) were obtained from Millipore (Billerica, MA, USA).

#### *3.2. Biscuit-Making Production in Pilot-Plant Scale*

Pilot-plant scale experiments were performed according to a previously published study [3]. Briefly, three batches of wheat bran naturally infected with Fusarium spp. were analysed with a focus on DON, selected and mixed with a blank wheat flour for the wholegrain biscuit production. Concerning cocoa biscuit trials, the same three mix flours were employed and three different batches of cocoa powder were selected.

Different doughs were prepared in order to obtain a final dough of about 1000 ± 30 g; regarding wholegrain biscuits, the ingredients were wheat flour (60%), bran (7%), cream of tartar, glucose syrup, and salt. Eggs, margarine, and dextrose were added depending on the value reported in recipes obtained from the experimental design (Table 3). Sodium bicarbonate was added in order to reach the appropriate pH value. Water amount ranged from 1.6% to 5%, depending on the technological requirements.


**Table 3.** Experimental data set for screening variables effects on acrylamide levels within the wholegrain biscuit-making process steps: full-factorial central composite design.

<sup>1</sup> Data expressed as mean value ± standard deviation.

In order to produce cocoa biscuits, 45% of wheat flour, 7% of bran, 4% of cocoa powder, margarine, glucose syrup, and salt were employed. Milk and dextrose amount were indicated in the model generated by Design of Experiment (Table 4). Sodium bicarbonate was added in order to reach the

appropriate pH value. The optimal amount of water to be added to each dough sample was established on the basis of internal technological knowledge.


**Table 4.** Experimental data set for screening variables effects on acrylamide levels within the cocoa biscuit-making process steps: full-factorial central composite design.

<sup>1</sup> Data expressed as mean value ± standard deviation.

The process for wholegrain and cocoa biscuit production consisted essentially of the following steps: creaming, dough preparation, and baking step. Firstly, wheat flour was mixed with all solid powder ingredients using a test planetary kneader for 2 min. Dextrose and margarine were mixed separately by using another test planetary kneader for 3 min (creaming step). At a later stage, cream and powders were mixed together for 3 min. Dough was shaped and rounded pieces of about 4 cm diameter (approximately 10 g) were obtained from dough and rested for 10 min at room temperature. Baking step was performed in a pilot-scale dynamic oven (Tagliavini, Parma, Italy). The overall process is summarized in Figure 1. Nineteen different tests for each process were performed (Tables 3 and 4, respectively).

Acrylamide content was examined in mix powders, before and after baking process. Before the acrylamide content analysis, samples were stored at −20 ◦C. Each sample was extracted and analyzed in duplicate.

#### *3.3. Moisture Content Determination*

The moisture contents of mix flours, doughs, and baked products were measured by takinga5g ground sample and heating it in a thermostatic oven at 105 ◦C for 6 h. All the results were compared on a dry matter (d.m.) basis.

#### *3.4. Experimental Design and Statistical Evaluation*

Design of Experiments (DoE) is used in many industrial issues, in the development and optimization of manufacturing processes, making a set of experiments representative with regards to a given question. DoE is a series of tests in which purposeful changes are made to the input variables of a system or process and the effects on response variables are measured: the analyst is interested in studying the synergistic effects of some interventions (the "treatments") to optimize the final process [28].

Among the wholegrain and cocoa biscuit-making parameters, several conditions were varied during the experiments: DON contamination level on wheat bran; dextrose, margarine, egg and milk content (as percentage in recipes); pH value (as sodium bicarbonate content) and baking time and temperature. Each selected treatment was varied within a range defined according to the technological requirements to obtain an organoleptically appreciable product for consumers: the central experimental values, indicated in Table 1, represent the optimal combination of ingredients/recipe and operative conditions that permit to achieve the most appropriate finished product.

Experimental data were then analysed by a multi-variate analysis approach based on the partial least-squares (PLS) technique, using a dedicated statistical package (MODDE software, version 9.1, 2012; Umetrics, Umea, Sweden).

#### *3.5. Sample Extraction and Instrumental Conditions—Deoxynivalenol*

Concerning Deoxynivalenol, samples were extracted according to a previously published procedure [3,25] with slight modifications. Briefly, a total of 10.00 g of flour or dough or biscuit sample were extracted with 100 mL of an acetonitrile/water (84:16, *v*/*v*) mixture by homogenization at a medium-to-high speed for 2 min using a mixer (Oster, New York, USA). The extract was allowed to settle for 15 min. Afterwards, 5 mL were poured into a 10 mL vial, and evaporated to dryness under a nitrogen stream. The extract was reconstituted with 100 mL of 13C-DON internal standard solution (100 ng/mL in methanol) and 900 mL of water. Each extraction cartridge column was activated using 2 mL of methanol, and 2 mL of methanol:water (10:90, *v*/*v*). The sample extract was then slowly passed through the OASIS® HLB 3 cc (60 mg) (Waters, Manchester, UK) column using a vacuum chamber system. A solution of methanol:water (20:80, *v*/*v*) was used for washing, followed by elution with 1 mL of methanol. The eluate was evaporated under a gentle stream of nitrogen, and the residue was dissolved in 200 mL of eluent A (methanol:water, 20:80 *v*/*v*, 0.5% acetic acid, and 1 mM ammonium acetate) prior to UHPLCMS/MS analysis. Ultrahigh-performance liquid chromatography (UHPLC) was performed using a Dionex Ultimate® 3000 LC systems (Thermo Fisher Scientific Inc., Waltham, MA, USA) and a Kinetex Biphenyl column (2.6 mm; 100 × 2.10 mm; Phenomenex). The flow rate of the mobile phase was 400 mL/min, and the injection volume was 20 mL. The column oven was set to 30 ◦C. A linear binary gradient composed of (A) water (0.5% acetic acid, 1 mM ammonium acetate) and (B) methanol (0.5% acetic acid, 1 mM ammonium acetate) was employed. The gradient was as follows: 0–4 min to 40% B; 4–20 min to 80% B; 20–22 min, isocratic step 80% B; finally, a re-equilibration step at 10% B (the initial value) was performed for another 3 min, bringing the total analysis time to 25 min. Before UHPLC-MS/MS analysis, all samples were filtered through centrifugal filter units for clarification. ESI-MS/MS was carried out by a Q-Exactive (Thermo Fisher Scientific Inc., Waltham, MA, USA) mass spectrometer. Experiments were performed in full MS data scan for quantification and data-dependent scan with the following settings: the capillary temperature was set to 300 ◦C; the sheath gas and auxiliary gas flow rates were set to 40 and 10 units, respectively; the spray voltage was set to 3500 kV; the S-lens RF level was set to 55 V. All equipment control and data processing were performed by Excalibur software (Thermo Fisher Scientific Inc., Waltham, MA, USA). Deoxynivalenol measurements in all the samples were performed using isotopically labeled standard and calibration vs. matrix-matched standards.

#### *3.6. Sample Extraction and Instrumental Conditions—Acrylamide*

Sample extraction for acrylamide was performed according to a previously published procedure [29] with slight modifications. Briefly, samples were finely ground in a blender to homogeneity before extraction. 1 g of sample was weighed into a polypropylene graduated conical tube and different volumes of a 300 μl ml-1 internal standard solution (13C3-labeled acrylamide in 0.1% (*v*/*v*) formic acid) followed by 10 mL 0.1% (*v*/*v*) formic acid were added on the base of the acrylamide concentrations supposed to be present in the samples. After mixing for 10 min on a vortexer the extract was centrifuged at 1,0000 rpm for 5 min. A 3-mL portion of clarified solution was removed avoiding to collect top oil layer when present and filtered through a 0.45 μm nylon syringe filter (Phenomenex, Torrance, CA, USA) before injection into the HPLC-MS/MS system (injection volume 10 μL). LC-ESI-MS/MS in positive ion mode analysis was achieved using a Surveyor LC quaternary pump separation system (Thermo Fisher Scientific Inc.) coupled to a linear ion trap LXQ mass spectrometer (Thermo Fisher Scientific Inc., Waltham, MA, USA).

Chromatographic separation was performed using a Synergi Hydro-RP (150 × 2.0 mm) 4 μm analytical column (Phenomenex, Torrance, CA, USA). Elution was carried out at a flow rate of 0.2 mL/min, in isocratic conditions, at 30 ◦C using as mobile phase a mixture of 98.9% water, 1% methanol and a.1% formic acid (*v*/*v*/*v*). In these conditions, the retention time of acrylamide was about 4 min. A time programmed valve was used to discard the eluate from the column for the first 2.5 min in order to eliminate the compounds with retention times shorter than acrylamide. At 8 min the column flow was again diverted and the mobile phase changed to 100% methanol in order to clean the column from strongly retained compounds within a total run time of 10 min. MS/MS conditions were set as follows: capillary temperature was set to 160 ◦C; the sheath gas was set to 35 units; the spray voltage was kept at 4500 V; capillary voltage was kept at 9 V. All parts of the equipment and data processing were performed by the computer software Xcalibur (Thermo Fisher Scientific Inc.). MS/MS analysis was carried out by selecting the ions at *m*/*z* 72 and *m*/*z* 75 as precursor ions for acrylamide and 13C3-acrylamide respectively.

The area of the chromatographic peaks of the extracted ion at *m*/*z* 55, due to the transition 72 > 55, and at *m*/*z* 58, due to the transition 75 > 58 were used for the quantitative analysis. The quantitative analysis was carried out with the method of the internal standard. The relative response factor of acrylamide with respect to 13C3-acrylamide was calculated daily by analysing a standard solution.

#### **4. Conclusions and Outlook**

Since precursors of acrylamide are present in the dough, modifications in recipe formulation and time-temperature control during baking process could be actually used to reduce acrylamide content in biscuits.

In the present study, the influence exerted by modifying ingredients and industrial conditions on acrylamide levels within a parallel mycotoxin mitigation strategy was investigated.

The obtained processing models showed a good fitting, robustness, and prediction capability, suggesting the most significant parameters. These can concretely contribute to the reduction of acrylamide levels in the final food product.

Acrylamide formation is evidently baking time- and temperature-dependent, therefore prolongation of heat treatments results in higher contents of acrylamide; however, when such parameters are moved within the optimal range for DON mitigation, the actual increase affects, in a limited way, the final acrylamide content without significant implications on the organoleptic properties.

In conclusion, the present report demonstrates the effectiveness of a careful design of process parameters for the mitigation of multiple contaminants in the final product, thus remaining within the consumer's sensorial acceptance.

**Author Contributions:** Conceptualization, M.S. and C.D.; Data Elaboration and Validation, M.S., S.G., C.D. and M.C.; Pilot Plant Trials, S.G.; Formal analysis, S.G. and M.C.; Writing—original draft, S.G. and M.S.; Writing—review & editing, M.S., C.D. and M.C.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors would like to thank Claudio Dall'Aglio (Barilla Pilot Plants) for his collaboration in the pilot plant trials, Nadia Morbarigazzi (Barilla Research & Development) for her suggestions and fruitful discussions and Dante Catellani (Barilla Food Research Labs) for his time and availability.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Determination of Deoxynivalenol Biomarkers in Italian Urine Samples**

#### **Barbara De Santis 1,\*, Francesca Debegnach 1, Brunella Miano 2, Giorgio Moretti 3, Elisa Sonego 1, Antonio Chiaretti 4, Danilo Buonsenso <sup>4</sup> and Carlo Brera <sup>1</sup>**


Received: 31 May 2019; Accepted: 18 July 2019; Published: 25 July 2019

**Abstract:** Deoxynivalenol (DON) is a mycotoxin mainly produced by *Fusarium graminearum* that can contaminate cereals and cereal-based foodstuff. Urinary DON levels can be used as biomarker for exposure assessment purposes. This study assessed urinary DON concentrations in Italian volunteers recruited by age group, namely children, adolescents, adults, and the elderly. In addition, vulnerable groups, namely vegetarians and pregnant women, were included in the study. To determine the urinary DON, its glucuronide and de-epoxydated (DOM-1) forms, an indirect analytical approach was used, measuring free DON and total DON (as sum of free and glucuronides forms), before and after enzymatic treatment, respectively. Morning urine samples were collected on two consecutive days, from six different population groups, namely children, adolescent, adults, elderly, vegetarians and pregnant women. Total DON was measured in the 76% of the collected samples with the maximum incidences in children and adolescent age group. Urine samples from children and adolescent also showed the highest total DON levels, up to 17.0 ng/mgcreat. Pregnant women had the lowest positive samples per category (40% for day 1 and 43% for day 2, respectively), low mean levels of total DON (down to 2.84 ng/mgcreat) and median equal to 0 ng/mgcreat. Estimation of DON dietary intake reveals that 7.5% of the total population exceeds the TDI of 1 μg/kg bw/day set for DON, with children showing 40% of individuals surpassing this value (male, day 2).

**Keywords:** mycotoxins; deoxynivalenol; children; adolescents; pregnant women; vegetarians; biomonitoring

**Key Contribution:** This study showed that total DON was measured in 76% of the collected samples, with the maximum incidences in children and adolescent age groups. The outputs of the statistical model associating food variables and urinary DON showed that an increase of total cereal consumption significantly associated with total DON in urine. In particular, the increase of pasta consumption affects the urinary DON content, confirming the relevant role of pasta in the Italian diet.

#### **1. Introduction**

Mycotoxins are natural food and feed contaminants, mainly produced by filamentous fungi of genera *Aspergillus*, *Penicillium*, *Fusarium* and *Alternaria* [1]. Mycotoxin production is promoted by either environmental and agronomic factors, while strong weather variability is a key factor for fungal infection, fungal colonization, and mycotoxin loads [2,3]. Among mycotoxins, trichothecenes represent the main group of *Fusarium* toxins commonly found in cereal grains. Deoxynivalenol (DON) is one of the most widely diffused natural-occurring trichothecene, is a sesquiterpenoid polar organic compound, belonging to the type B trichothecenes since it contains carbonyl group in C-8. *Fusarium graminearum* and *culmorum*, two important causing agents of the Fusarium Head Blight (FHB), are the most important producers of DON [4]. Acting as a virulence factor for cereal infection (namely in wheat, maize, barley, oat and rye), DON jeopardizes cereal grain quantity and quality from agriculture and health perspective [3] being persistent in cereal food derived products.

European Regulation (EC) 1881/2006 [5] has set maximum levels of DON in unprocessed cereals and cereal foods products intended for direct human consumption in the range of 1750 to 200 μg/kg. The scaled down levels of the contamination from the milling products and the processed ones reflect a well-known reduction rate of about 30% because of the cleaning and processing [6,7] but it has also to be considered the DON degradation process that occurs during industrial baking [8]. The legislative provisions serve to control the levels in food. However, surveys carried out worldwide [2] and in Italy [9] confirmed DON occurrence in cereal samples like wheat and maize, including processed products, thereby suggesting an exposure assessment issue. The correlation between DON dietary exposure and DON presence in urine has already been reported and confirmed [10,11]; therefore, the biomonitoring of DON and DON metabolites in urine may constitute a valuable indicator of the dietary exposure. On the basis of the available information derived from the toxicity studies, a temporary tolerable daily intake (TDI) of 1 μg/kg body weight (bw)/day was established in 2002 [12]. On the basis of the more recent available scientific data, this TDI has been recently confirmed by the European Food Safety Authority (EFSA) as a group-TDI for the inclusion of 3-AcDON, 15-AcDON and DON-3-glucoside, as DON plant metabolites [4]. In regards to acute risk, the Joint Expert Committee on Food and Additives (JECFA) [13] concluded that DON is a probable factor for acute pathologies in humans and derived an Acute Reference Dose (ARfD) of 8 μg/kg bw. In 2017, an extensive review on the presence of DON and its metabolites in human urine was carried out [14]. Authors showed that it is ubiquitous and highlighted geographical differences. So far, the metabolic profile of DON in animals has already been established. DON is metabolized to 13-deepoxy (DOM-1) in animals and is excreted via the feces and urine [15,16]. In addition to DOM-1, other conjugated products may be present as excreted metabolites, namely DON-3-glucuronide (DON-3GlcA) and DON-15-glucuronide (DON-15GlcA), which seem to be the major metabolites in humans, along with their iso forms and DON-8,15-hemiketal-8-GlcAc [17], sulfate and sulfonate forms [18,19]. More recently, DON-7-glucuronide (DON-7GlcA) has been tentatively identified and introduced as a new metabolite [20–22] in humans.

The analysis of urinary glucuronides is crucial for the study of trichothecene biomarkers, because approximately 80% of DON excreted via urine is conjugated with glucuronic acid [4]. A number of biomonitoring studies of DON in urine is available, however, different analytical method approaches, different limits of quantification and different ways to express measured mycotoxin (total DON or free DON, not specified, creatinine corrected, etc.) may give difficulties in comparison results. When expressed in ng/mg of creatinine (ng/mgcreat), DON levels in urine ranged from 0.2 up to 903.7 ng/mgcreat [23–28], with a worst case reported for a group of pregnant women from Croatia [21].

In the present study, morning urine samples were collected over two consecutive days from 203 Italian volunteers; the participants were recruited by age group, namely children (3–9 years), adolescents (10–17 years), adults (18–64 years), and elderly (65 years or above). Vulnerable groups, namely vegetarians and pregnant women, were also included in the study. Each participant was asked to collect a first-morning urine sample on two consecutive days and to complete a Food Frequency Questionnaire (FFQ) reporting dietary habits over a 1-month period and a Food Diary (FD) with detailed information about food items consumed on the day preceding the collection of urine samples. Associations between food consumption and urinary DON were assessed using ordered logistic regression models.

To determine the urinary total DON expressed as sum of free and glucuronides forms, an indirect analytical approach, measuring free and total DON, before and after enzymatic treatment, respectively, was fully validated and applied to the collected urine samples [29,30].

This paper affords a distinctive data set detailing the DON exposure of Italian population in an urban setting, providing data on levels of DON in human urine samples collected, as analyzed by liquid chromatography coupled with mass spectrometry (LC-MS).

#### **2. Results and Discussion**

#### *2.1. Analytical Method*

The method used in this experimental study was previously published by Turner et al. [10] and applied with minor modifications. In particular, in order to reduce the amount of methanol needed for the elution of mycotoxins from the immunoaffinity column (IAC), three volumes were tested, 4 mL as recommended in the work of Turner et al., 2 mL and 1.5 mL. The performances were evaluated in terms of the recovery factor (*n* = 3), the obtained results were 102 ± 5%, 99 ± 5% and 56 ± 10%, for 4, 2 and 1.5 mL, respectively. The selected volume was 2 mL, for which a good recovery was achieved also permitting the reduction of the amount of organic solvent used and of the time needed for drying the sample in the subsequent analytical step. Since the IAC was designed for DON analysis, the IAC cross reactivity for DOM-1 was also tested (*n* = 3) with good results in terms of recovery (98 ± 1%). In regards to the determination step, a single injection for DON and DOM-1 was performed instead of the separated chromatographic run proposed in the reference paper [10].

#### *2.2. In-House Validation*

Before applying the method to the collected samples, an in-house validation was performed on 4 different contamination levels; the results obtained for the investigated performance parameters are reported in Table 1. The limits of detection and quantification (LoD and LoQ) values set during validation fit with the purpose of having an analytical method that is able to detect low amounts of DON and DOM-1 for exposure assessment purposes. Recovery factors are in the ranges 95–109% and 81–93% for DON and DOM-1, respectively, and the precision, evaluated in terms of Relative Standard Deviation of repeatability (RSDr), is always below 10%. Despite the absence of a specific official reference, the obtained values were considered satisfactory under the criteria of Regulation (EC) 401/2006 [31], which apply to any DON analysis on food. The average values estimated for method uncertainty ranged between 8 and 14% for DON and between 15 and 18% for DOM-1, being in both cases well below the 44%, which is the maximum acceptable value when the Horwitz approach is applied [31].


**Table 1.** Performance characteristics obtained during the in-house validation of the method.

#### *2.3. Description of Study Population*

In total, 406 first morning urine samples were collected from 203 volunteers belonging to six population groups, namely children, adolescents, adults, elderly, vegetarians and pregnant women. All samples were kept at −20 ◦C until analysis. The volunteers enrolled for the study are reported, with their anthropometric data, grouped by category, gender and age in Table 2.

The recruitment of subjects was subdivided between Istituto Superiore di Sanità (ISS) and the hospital "Agostino Gemelli (UCSC)", with ISS responsible of the recruitment of adults, elderly and vegetarians and UCSC of children and adolescents by the clinics of Pediatric Unit and pregnant women from the clinics of Gynecology and Obstetrics outpatients. The selection of different groups of the population was specifically requested by EFSA while being considered as an asset for the study due to having different potential metabolic susceptibility, body weights, dietary habits, and consumption rates. Pediatric biological systems and detoxification processes might be widely different, causing an amplification of children's susceptibility to hazards that would have negative consequences [32]. It should be noted that the cereal-based products consumption intakes figures of children in Italy are comparable with those of adults, showing, in addition, that children are consumers of a higher variety of those food products than adults (data from the FFQ and food diary, not shown). These two aspects provide children with, at least, the same exposure risk to that of adults. As for pregnant women, due to a decreased immunocompetence as a consequence of hormone levels changes and their complex and multifactorial interplay [33], this population group is of special interest for its possible susceptibility to DON, which has a recognized impact on the immune response system. As for vegetarians, they were included to attempt to verify and to measure if a special diet, which could be enriched by cereal based products, is a factor that may affect exposure estimates. However, the literature shows that these subgroups were already enrolled as special populations in other DON

biomonitoring surveys. Piekkola et al., and Hepworth et al., [26,34] indicated the potential risk to mothers and their babies from DON exposure during pregnancy. Pestka highlighted the need to study the relationship between DON consumption and possible growth effects in susceptible populations such as children and vegetarians [35].

During the recruitment, UCSC initially encountered difficulties for pregnant women relating to a general concern from potential participants regarding the possible outputs of the study, abandonment of the study after initial recruitment, and incorrect collection of urine samples and compilation of questionnaires from participants. Therefore, ISS supported the UCSC hospital in its interactions with under-recruited subgroups (pregnant women and vegetarians). In regards to vegetarians, it was quite difficult to recruit male volunteers to the study since they constitute a low percentage of the vegetarian population.

The evaluation of the ethic aspects related to the study protocol and its approval by an Ethics Committee was requested and obtained before the starting date of the study. The UCSC ethical approval was granted by the Local Ethics Committee on 23 March 2014 (for the Gynecology Unit) and 7 April 2014 (for the Pediatric Unit). ISS ethical approval was granted by the ISS Ethic Committee on 18 February 2014. Informed consent was provided by the participants during their first visit. Approval code: Prot. PRE 84/14 and Prot. CE 14/413.

**Table 2.** Anthropometric data and number of individuals recruited in Italy grouped by category, gender and age.


<sup>a</sup> F, Female; M: Male; <sup>b</sup> N: number of subjects; <sup>c</sup> BMI, Body Mass Index.

#### *2.4. DON Biomarker Levels in Urine Samples*

For each collected urine sample three analyses were performed, one for free and one for total DON and DOM-1, the first and second corresponding to before and after enzymatic treatment, with the third for creatinine. The results obtained are reported in Table 3, namely mean values, median (P50) and interquartile (IQR) values are listed as non-adjusted and creatinine-adjusted total DON. In Table 3 the percentage of samples above LOD for total DON is also reported for each category. Moreover, in Table 3 the percentage contribution of free DON and DON-GlcA to total DON is reported.

The mean DON level in urine for the total population studied was 7.67 and 7.93 ng/mL, but differences arose when the different categories were separately taken into account. Among the selected group, pregnant women had the lowest positive samples per category (40% for day 1 and 43% for day 2, respectively), low mean levels of total DON (4.37 ng/mL and 2.70 ng/mL on day 1 and day 2, respectively), and a median equal to 0 ng/mL. Conversely, urine from children and adolescents showed the highest concentrations of total DON, up to 75.9 ng/mL. The vegetarian group, apart from females on day 1, showed median values very close to the mean values, indicating that the values are close to normal distribution.

The median values of total DON in the morning urine ranged from 0 ng/mL for pregnant women on day 1 and day 2 to 12.60 ng/mL for adults male on day 2. However, when the creatinine adjustment was taken into account the median values ranged from 0 ng/mgcreat for pregnant women (day 1 and day 2) to 13.0 ng/mgcreat for children male (day 2), while the median value for adults male on day 2 decremented to 6.84 ng/mgcreat. These results emphasize the importance of creatinine adjustment for the correct interpretation of results.

Regarding the DOM-1 content, only 6 samples out of the 406 analyzed urines (1.5%) had detectable levels of this toxin (ranges <LOD-2.6 ng DOM-1/mgcreat and <LOD-1.7 ng DOM-1/mgcreat, for day 1 and day 2, respectively), confirming that this is a minor route for DON in human metabolism.

In view of the ambiguous statistical differences among males and females' DON levels found in literature [11,30,36–38], DON ng/mgcreat content was checked between sexes in the general population excluding pregnant women. No statistical differences were highlighted for any of the age group (Mann-Whitney test). Conversely, a difference statistically significant was obtained comparing women adult group with pregnant women pointing out that pregnant women showed DON ng/mgcreat levels lower than adult females in general (*p*-value 0.025).

Comparing the results obtained within the EFSA study [39], the median concentrations of total DON adjusted for creatinine in morning urine in Italy and Norway were quite similar within similar population groups, whereas the median concentrations in the corresponding population groups in the UK were approximately 3-fold higher than in Italy and Norway. The Italian results are also in line with the values published by Solfrizzo et al. [30], which reported for total DON a median of 10.32 ng/mL.

Regarding the contribution of free DON and DON-GlcA to total DON, throughout all the population groups, the DON-GlcA represented 66% and 71% of the total DON for day 1 and day 2 respectively, confirming that the glucuronidation is an important route for DON excretion, as has already been reported in other studies [40].


*Toxins* **2019**, *11*, 441

 of free

**Table 3.**

Non-adjusted

 and

creatinine-adjusted

 total DON

concentrations

 in urine samples by day and sex in sampled population groups, and contribution

a F: Female; M: Male; in parenthesis, number of subjects.

#### *2.5. Regression Analysis Between Food Consumption and DON Level in Urine Samples*

The logistic regression model was applied to the dataset to assess the effect of the general variables on total DON concentration adjusted for creatinine. The output of the model refers to a unitary variation of the considered variable. With the aim to express the output of the food variable in a more comprehensive way, an increment of 10 grams of cereal food intake (i.e. total food, food category or item) was considered.

To assess the effect of the general variables, the mean values for the total DON content (ng/mgcreat) on day 1 and day 2 were calculated for each subject and then categorized in tertiles, resulting in only one dependent variable. In regards to the age groups, the odds ratio (OR) to have a higher level of mean DON adjusted for creatinine in adults compared to children is 0.13 (*p* = 0.000), confirming the critical scenario for the children age group. The OR to have a higher level of mean concentration of total DON adjusted for creatinine in pregnant women compared to non-pregnant women is 0.198 (*p* = 0.001), in accordance with the very low DON levels in urine for this category. The gender, BMI, physical activity and vegetarian variables were not significant when added into the model.

Regarding the food variables, the analyses were performed by assessing the effect on total DON adjusted for creatinine on day 1 and day 2 of (i) total cereal food intake; (ii) each food group intake; (iii) single food items. While no statistically significant association between total food intake and DON levels was observed for day 1, the increase of 10 g in the total cereal food intake raised the OR by about 2.6% for day 2 (*p* = 0.027). The effects of the food groups provided significant results for pasta and pasta-like products, while for day 1 and day 2, an increased intake of 10 g caused the OR to have a higher level of concentration of total DON adjusted for creatinine by about 4% (*p* = 0.047) and 5.5% (*p* = 0.052), respectively. Considering the food item variables, the increased intake of durum wheat pasta caused the OR to have a higher level of total DON adjusted for creatinine by about 6% (*p* = 0.023) and 7.9% (*p* = 0.008), respectively.

#### *2.6. Estimated DON Daily Intake*

Starting from the total DON concentration levels measured in the collected urine samples, the Estimated Dietary Intake (EDI) was calculated using the following formula [30]

$$EDI\_{DON} = C \times \frac{V}{bw} \times \frac{100}{E} \tag{1}$$

EDI (ng/kg bw/day);

C = total DON concentration in the analyzed urine samples (ng/mLurine);

V = mean 24 h human urine volume (1.0 mL per kg of bw per hour for adults [41]; 2.0 mL per kg of bw per hour for children [42]);

bw = body weight reported in the questionnaire;

E = urinary excretion rate of DON in 24 h, 72.3% [10]

The calculated individual exposure followed a non-normal distribution (Shapiro-Wilk test) hence a non-parametric approach has been used. In Table 4, the mean, median (P50), and 95th percentile (P95) are reported. The obtained EDI values were compared with the Tolerable Daily Intake (TDI) set by EFSA for DON at 1000 ng/kg bw [4], in the last column of Table 4, the number of individuals and the percentage exceeding the TDI are reported.


**Table 4.** Mean, median (P50) and 95th percentile (P95) of EDI (ng/kg bw/day) calculated for DON are reported based on the category, gender and age group, together with the number and percentage of individuals exceeding the TDI.

The EDI mean values for the total population of the study are quite low, representing around the 30% of the TDI reported by EFSA, and the percentage went down to 20% when the median was considered. The scenario was more differentiated when the single categories were considered. The highest EDI values were obtained for children and to a minor extent for the adolescent age group, with mean values ranging from 577 to 937, and from 296 to 427 ng/kg bw/day for children and adolescent, respectively. Also, the number of individuals exceeding the TDI was higher for this age group, with the highest percentage being the 40% of the male children category on day 2. It is important to note that the two highest estimated EDI's were obtained from female children on day 1 (5036 ng/kg bw/day) and male children on day 2 (3193 ng/kg bw/day), both values were below the ARfD (8000 ng/kg bw/day). As far as chronic exposure is concerned, all the values over the TDI should be duly considered as possible concern for public health as reported by EFSA, especially for vulnerable groups such as infants, toddlers and other children [4].

For a better visualization of the distribution of the obtained results, the calculated individual EDIs are reported in Figure 1. The depicted scenario for children and adolescent must be considered while taking into consideration body weights, homeostasis water balance (i.e., average food and beverage water input, urine feces, skin water output) and food consumption rates, especially for the children, that are not dissimilar when compared to the ones of the rest of the population, producing an unfavorable body weight/intake ratio.

The concentration values obtained in this study were also used by EFSA for exposure estimations. When comparing EFSA exposure estimates from biomarkers for adolescents, adult, elderly with corresponding mean averaged values obtained in this study (data not shown), exposure scenarios are confirmed. Discrepancies arise for children, and in particular, EFSA [4] obtained lower exposure estimates from biomarkers. These differences are due to different urine volumes considered for the

24 h, since EFSA used 0.5 L of urine for children during the 24 h, while in this study, calculations were made depending on the body weight of the subject [42], leading to a urine volume in the 24 h that ranged from 0.71 L/day to 2.49 L/day (an average of 1.25 L/day). The estimation made for the total Italian population in this paper, 329 and 343 ng/kg bw/day for day 1 and day 2 respectively, is also comparable with the probable daily intake (PDI) of 590 ng/kg bw/day reported by Solfrizzo [30], which was calculated considering a 50% excretion rate and 1.5 L for total urine in the 24 h for all of the recruited population.

**Figure 1.** Distribution of the calculated individual EDI around the median value (red small line) and compared to the DON TDI (red line).

#### **3. Conclusions**

The high incidence of DON in urine confirms its ubiquitous presence in cereal food products in Italy. However, the results of this study showed moderate mean levels of DON in urine samples of the studied cohort, despite differences that emerged when the different categories were considered, with children being the most susceptible group. The critical scenario depicted for children was also confirmed when the estimated daily intake was considered, showing the highest mean values for this age group (close to TDI and over in the case of P95), and the highest percentage of individuals exceeding the TDI, while for the total population the mean EDI represented around the 30% of the TDI. The higher exposure estimated for children and, to a minor extent for adolescents, can be explained by taking their unfavorable body weight/intake ratios into consideration.

In this study, the contribution of free DON and DON-GlcA to total urinary DON was also investigated and the results confirmed that DON-GlcA is the major DON metabolites for humans, on the other hand the very limited number of samples with DOM-1 above LOD confirms that this is a minor route for DON in the human metabolism.

Statistical analysis confirmed that age significantly affected urinary DON concentration (the difference between children and adults). Considering the outputs of the statistical model for food variables, increasing total cereal consumption was significantly associated with total DON in urine and in particular the increase of pasta consumption affected the urinary DON content more than the increased intake of other studied food items, confirming the relevant role of pasta in the Italian diet.

The results obtained in this study underpin the need for other studies in order to collect data for providing a more comprehensive exposure assessment of the Italian population. Moreover, support for the biomarker approach in exposure assessment is represented by the availability of either validated analytical methods and harmonized references for those critical figures (such as urine outputs or excretion rates), which sensibly influence the exposure scenarios.

#### **4. Materials and Methods**

#### *4.1. Analytical Method*

#### 4.1.1. Chemicals and Reagents

Methanol HPLC grade (Carlo Erba Reagenti, Cornaredo, MI, Italy) and ultrapure water (Millipore, Burlington, MA, USA) were used for the LC-MS analysis.

In order to perform the planned validation study and the analytical work on collected urine samples, the standard of DON (product number: D0156, 1mg), U-[13C15]-DON (99.0% 13C) and DOM-1 (product number: 34135, 2 mL) were purchased from Sigma (Saint Louis, MI, USA). The enzyme β-glucuronidase (Type IX-A from *E. coli*; product number: G7396—2MU) was purchased from Sigma too. The IAC DONtest WBTM (product number: G1066) were purchased from Vicam (Milford, MA, USA).

#### 4.1.2. Sample Preparation

Urinary DON and metabolites concentrations were measured using a two-step process. Stored urine samples were allowed to thaw, and were then centrifuged (1790 g, +4 ◦C, 15 min). For each participant, two aliquots (1 mL) were prepared by mixing 13C-DON internal standard solution, to provide a final concentration of 20 ng/mL.

Aliquot 1 was used to determine total DON concentrations, defined as the sum of glucuronide metabolites and free DON. To measure DON-glucuronides and free DON, each sample was adjusted to pH 6.8 with drop wise addition of KOH or HCl and digested using β-glucuronidase solution (23,000 U/mL, 250 μL) in a shaking water bath at 37 ◦C for 18 h, ensuring a gentle mixing. After this pre-treatment, the samples were diluted to a final 4 mL with phosphate buffered saline (PBS, pH 7.4). The diluted urine sample was passed through a wide bore DON IAC using a VisiprepTM vacuum manifold (Sigma-Aldrich, Saint Louis, MI, USA). DON was eluted from columns with methanol (2 mL) and extracts were dried at 40 ◦C under a gentle stream of nitrogen and reconstituted in 90% methanol (250 μL) before LC-MS analysis. DOM-1 was quantified on the same aliquot and analyzed for DON-GlcA. Aliquot 2 was used to assess free DON using the aforementioned procedure, but without any β-glucuronidase treatment.

Urine creatinine was assessed by the enzymatic method described by Mazzachi et al. [43].

#### 4.1.3. LC-MS Determination

Chromatographic separation of DON and DOM-1 was performed using UHPLC (Ultra-High-Performance Liquid Chromatography) with Waters RP Acquity BEH C18 column (100 × 2.1 mm, 1.7 μm, Milford, MA, USA) kept at 30 ◦C, and a mobile phase sequence of 10 minutes duration, starting with 20% MeOH, changing to wash of 75% MeOH after 4.50 min and reverting to 20% MeOH after 8 min (flow rate 0.350 mL/min with a volume injection of 10 μL). DON elutes at 2.5 minutes under these conditions, while DOM-1 elutes at 4.5 min.

The mass spectrometric analysis was carried out with a Quattro Premier XE (Waters Milford, MA, USA) in SIR (Selective Ion Recording) acquisition mode with an ESI (ElectroSpray Ionization) interface. The analysis was performed in positive ion-mode. The following mass spectrometer conditions were optimized by direct infusion of DON and DOM-1 standard solutions: capillary voltage 4.0 kV, desolvation gas flow rate 500 L/h at 350 ◦C, source temperature 110 ◦C, cone voltage 40 V. The monitored masses for DON were 319.2 and 334.2 m/z corresponding to [DON-Na]<sup>+</sup> and [ 13C-DON-Na]<sup>+</sup> respectively, and 303.2 m/z for [DOM-1-Na]+.

For quantification purposes, calibration curves with a labeled internal standard for DON, and an external standard approach for DOM-1 were used. The curves covered the range 2–250 ng/mL, corresponding to 0.50–62.5 μg/Lurine for DON, and 2–200 ng/mL corresponding to 0.50–50 μg/Lurine for DOM-1. An acceptability criterion of *R*<sup>2</sup> > 0.995 was applied during routine analysis.

DON-GlcA values were estimated indirectly by subtracting free DON values from total DON values for each analyzed urine sample.

#### *4.2. In-House Validation*

The in-house validation was performed in accordance with the Eurachem guideline [44]. Selectivity and specificity were guaranteed by the clean-up step with the IAC containing specific antibodies for the selected mycotoxins. LoD and LoQ were identified by the injection of diluted standard solutions, the requirements were a S/N = 3 for LOD and a S/N = 10 for LoQ. The LoQ was included in the validated contamination levels. The validation was performed on 4 different contamination levels for DON and DOM-1 by repeated analyses on spiked urine samples. Trueness was evaluated in terms of recovery factors, while precision was estimated in terms of Relative Standard Deviation of repeatability (RSDr) calculated on repeated analyses for each contamination level. Expanded uncertainty was also estimated by a metrological approach in accordance with Eurachem guideline [45]. The combined standard uncertainty was calculated by a metrological approach by summing up the standard uncertainty contributions from repeatability (Type A), recovery (Type A), pipetting volumes (Type B), and calibration (Type A). By applying a coverage factor of 2, the expanded uncertainty accounted for the 95% confidence interval.

#### *4.3. Study Design*

The dataset in the present analysis represents a subset of data collected for a larger study entitled "Experimental study of deoxynivalenol biomarkers in urine" conducted for the European Food Safety Authority (EFSA) GP/EFSA/CONTAM/2013/04 [39]. In brief, this study explored the occurrence of DON and its metabolites in urine from different population groups (children, adolescents, adults, elderly, and pregnant women; total *n* = 635) in three European countries (UK, Italy, and Norway) and the relationships between urinary DON levels and its metabolites and self-reported dietary intake of cereal-based food items.

#### Recruitment of Participants and Urine Sample Collection

A target sample size of at least 200 individuals was established. The population groups included in the study were divided according to the age groups used within the EFSA Comprehensive European Food Consumption Database [46]. A relatively higher number of potentially more susceptible population groups such as children, adolescents, vegetarian and pregnant women was planned to be included. The planned sampled population included six different subgroups for recruitment, i.e., children (aged 3–9, 20%), adolescents (aged 10–17, 20%), adults (aged 18–64, 10%), elderly (aged above 65, 15%), vegetarians (15%) and pregnant women (20%).

Exclusion criteria included subjects not being able to give informed consent or complete the questionnaire, individuals affected by acute pathologies and any chronic illness (chronic renal, hepatic or cardiac problems, cancer), with chronic gastrointestinal conditions (e.g., celiac disease), gluten sensitivity or eating disorders, such as food allergies and those subjects recently on a weight loss diet, depression and psychosis, or hospitalized subjects within 3 months of admission. Inclusion criteria at the specialized recruitment centers (hospitals, clinics, institutions) required only healthy people, either not being on any medication or stable medication (for more than three months) that did not affect appetite (such as oral steroid use). As far as vegetarians, only people following the diet for >1 year and above the age of 18 years were recruited.

Collection equipment and instructions on how to collect and store a first morning urine sample were provided to participants. On two separate days, prior to urine sample collection, participants were required to complete a food diary, reporting food items consumed throughout those days. Participants provided a first morning urine sample the following morning. The urine samples (kept frozen at home) were returned to the clinical trial units after the second day of collection.

#### *4.4. Regression Analysis Between Food Consumption and DON Levels in Urine Samples*

#### 4.4.1. Food Frequency Questionnaire and Food Diary

Since a rich cereal-based diet plays a key role in the DON exposure, a semi-quantitative Food Frequency Questionnaire (FFQ) was designed. Type, frequency and quantity of the food consumed were obtained by this FFQ which has been prepared based on a validated questionnaire used in a Spanish study targeted to pregnant women [47]. The adopted FFQ included information about portion size and usual food frequency intake, with a recall period up to one month. In order to make the compilation of the questionnaire easier, photographic examples of portion sizes were also provided to participants. The food list in the adopted FFQ included specific food categories containing all food sources susceptible to DON contamination such as wheat, maize and barley products with emphasis on breads (whole meal, white, soft grain, other), breakfast cereals (high-fiber and other), pasta, pizza, fruit pies, biscuits, buns/cakes and beer. The reported food items reflected the specific Italian food habits. Therefore, in order to get information on the dietary habits of each population group, the FFQ was designed for gathering information on the amount and type of cereals and cereal-based products commonly more frequently consumed by the volunteers in the month prior the study and to capture and verify the most common food sources of DON in the diet. In order to capture such food categories, the EFSA's Food Classification System FoodEx2 database was used [46].

Beside the FFQ, a Food Diary (FD) was also prepared with the aim of collecting information on the intake of the same food items as those included in the FFQ, but referred to the two days immediately before the urine sampling.

A database collecting all the data related to the enrolled subjects concerning individual information (age, gender, BMI, and other information) and food intake was prepared.

#### 4.4.2. Statistical Analysis

Statistical analyses were carried out using STATA/SE 12.0. For descriptive statistics measures (e.g. mean, median) a substitution method was applied. Values below LOD were substituted with 0. Shapiro-Wilk test was used for testing of normality distribution. Two sample Wilcoxon rank sum test (Mann-Whitney test) was used to compare within categories (e.g., Male/Female or Vegetarian Y/N).

To assess the effect of all the observed variables on the total DON content (expressed as ng/mgcreat) an ordered logistic regression model was used. As variables, the model used a selection of the information gathered from the questionnaires including general variables such as age, gender and BMI, and food variables, namely the different food items reported on the FFQ and the FD. The model was run matching independent variables with the dependent variable. The variables that were shown to be significant were matched together to assess how much of the observed phenomenon is explained by the variables considered. When variables were significant, also interactions were considered. In order to apply the ordered logistic regression model, a categorization in tertiles (33th, 66th, 100th) of the dependent variable, namely the total DON content (ng/mgcreat) on day 1 and day 2 was performed.

The odds ratio (OR) is used to quantify the extent of the association of the examined variable on the total DON (ng/mgcreat) values.

A variable is considered significant when the null hypothesis of the relative coefficient of the variable is null with a significance level set to 0.95 (1-alpha). For this purpose, the *p* value was used. When the *p* value is higher than 0.05, the examined variable is considered to not be significant, conversely if the *p* value is lower than 0.05 the examined variable is considered to be significant.

**Author Contributions:** Conceptualization, B.D.S. and C.B.; Data curation, B.D.S., F.D. and G.M.; Formal analysis, F.D., B.M., responsible for the biological sampling (urine samples collection), A.C. and D.B.; Validation, B.D.S., F.D., B.M. and E.S.; Writing—original draft, B.D.S., F.D. and E.S.; Writing—review & editing, C.B.

**Funding:** The data presented in this study is extracted from a larger study funded by the European Food Safety Authority (EFSA) call GP/EFSA/CONTAM/2013/04 into the "Experimental Study of Deoxynivalenol Biomarkers in Urine". Sole responsibility lies with the author and the Authority is not responsible for any use that may be made of the information contained therein.

**Acknowledgments:** Marco Finocchietti, Gabriele Moracci, Maria Cristina Barea Toscan and Giuliana Verrone are acknowledged for their technical assistance.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*
