**Emerging Fusarium Mycotoxins Fusaproliferin, Beauvericin, Enniatins, and Moniliformin in Serbian Maize**

#### **Igor Jaji´c 1, Tatjana Dudaš 1,\*, Saša Krstovi´c 1, Rudolf Krska 2,3, Michael Sulyok 2, Ferenc Bagi 1, Zagorka Savi´c 1, Darko Guljaš <sup>1</sup> and Aleksandra Stankov <sup>1</sup>**


Received: 30 March 2019; Accepted: 17 June 2019; Published: 19 June 2019

**Abstract:** Emerging mycotoxins such as moniliformin (MON), enniatins (ENs), beauvericin (BEA), and fusaproliferin (FUS) may contaminate maize and negatively influence the yield and quality of grain. The aim of this study was to determine the content of emerging *Fusarium* mycotoxins in Serbian maize from the 2016, 2017, and 2018 harvests. A total of 190 samples from commercial maize production operations in Serbia were analyzed for the presence of MON, ENs, BEA, and FUS using liquid chromatography-tandem mass spectrometry (LC-MS/MS). The obtained results were interpreted together with weather data from each year. MON, BEA, and FUS were major contaminants, while other emerging mycotoxins were not detected or were found in fewer samples (<20%). Overall contamination was highest in 2016 when MON and BEA were found in 50–80% of samples. In 2017 and 2018, high levels of MON, FUS, and BEA were detected in regions with high precipitation and warm weather during the silking phase of maize (July and the beginning of August), when the plants are most susceptible to *Fusarium* infections. Since environmental conditions in Serbia are favorable for the occurrence of mycotoxigenic fungi, monitoring *Fusarium* toxins is essential for the production of safe food and feed.

**Keywords:** emerging mycotoxins; *Fusarium*; LC-MS/MS; maize; Serbia

**Key Contribution:** Up to present, little effort has been put forth in terms of quality control regarding emerging mycotoxins in Serbian maize. This article gives the first insight into the current situation of emerging mycotoxin contaminations in Serbian maize and recommends further monitoring studies necessary for legislative purposes since climate conditions are favorable for the development of mycotoxin-producing fungi that pose a threat to food and food quality.

#### **1. Introduction**

Mycotoxins have become one of the most important food contaminants in modern society. They are toxic secondary metabolites that are usually produced by *Aspergillus*, *Penicillium*, and *Fusarium* fungi in favorable environmental conditions. Among these species, *Fusarium* are the most prevalent mycotoxin-producing fungi in the northern temperate regions, and mainly Central and SoutheasternEurope [1]. *Fusarium* molds are known as producers of several mycotoxins, including

both "traditional" as trichothecenes, zearalenone, and the fumonisins [2] and "emerging toxins" as moniliformin (MON), enniatins (ENs), beauvericin (BEA), and fusaproliferin (FUS) [3]. Additionally, Kovalsky et al. [3] found a co-occurrence of "traditional" toxins with EN, MON, and BEA. The main producers of emerging *Fusarium* mycotoxins in cereals are *Fusarium avenacum*, *Fusarium verticillioides (moniliforme)*, *Fusarium proliferatum,* and *Fusarium subglutinans* [1].

MON is a sodium or potassium salt of 1-hydroxycyclobut-1-ene-3,4-dione [4]. It was discovered by Cole et al. [5] while screening for toxigenic products of *F. verticillioides* isolated from southern leaf blight-damaged maize seed. BEA and ENs belong to a group of cyclodepsipeptidesthat may have antibiotic, insecticidal, and cytotoxic effects [6–9]. FUS is a toxic sesterterpene originally isolated by Ritieni et al. [10] from *F. proliferatum* in autoclaved maize cultures.

BEA is known as a cholesterol acyltransferase inhibitor [11] and it is toxic to several human cell lines [12,13]. Additionally, BEA can induce apoptosis and DNA fragmentation [14]. FUS is a sesterterpene identified from maize cultures of *F. proliferatum* isolated from maize [15]. FUS is toxic to *Artemiasalina*, to the lepidopteran cell line SF-9 and to the human nonneoplastic B-lymphocyte cell line IARC/LCL 171 [16]. FUS can induce teratogenic effects in chicken embryos [17]. Kriek et al. [18] found that ducklings and rats fed diets containing MON led to muscular weakness, respiratory distress, cyanosis, coma, and death.

There is limited data on the toxicity and occurrence of "emerging" mycotoxins. These mycotoxins are neither routinely determined nor legislatively regulated. Their presence has been reported in cereals from several countries [19–23]. In a recent EFSA report [24], an opinion on the presence of ENNs and BEA in food and feed was made, but the lack of relevant toxicity data did not a risk assessment. Currently, maximum levels for emerging *Fusarium* mycotoxins are not been regulated.

Maize is one of the most susceptible cereals to the presence of *Fusarium* molds. Infection of maize may lead to grain size and protein decreasing as well as harming germination. The final result is a decrease in yield and feed quality. Additionally, a consequential mycotoxin production is another highly problematic outcome of *Fusarium* infection.

In Serbia, arable land covers approximately 75.5% of utilized agricultural land. In the structure of sown arable land areas, cereals comprised 67.9%, industrial crops comprised 15.7%, vegetables comprised 2.6%, and fodder crops comprised 9.1% in 2016 [25]. However, cereals were grown on 1,763,575 ha in 2016, which is lower compared to 2015 (1,782,010 ha), and 2014 (1,819,188 ha). In 2016, maize was harvested from 1,010,097 ha, with a total production of 73,767,371 t. The average yield in 2016 was 7.3 t/ha, which was higher than in 2015 (5.4 t/ha), and slightly lower in comparison with 2014 (7.5 t/ha) [25]. When compared to 2016, the total production of maize decreased by 45.5% in 2017, while the average yield was only 4.0 t/ha [26]. In 2018, expected production of maize was 6,965,000 t, which was 73.3% higher than in 2017, with an average yield ofapproximately 7.6 t/ha [27]. High maize production in 2018 positioned Serbia among the top ten maize exporters [28] and among the top twenty maize producers in the world [29].

In considering these facts, the aim of this study was to determine the current state of the level of emerging *Fusarium* mycotoxins in Serbian maize. Additionally, an effort was made to relate the obtained results with the weather conditions recorded during the trial period.

#### **2. Results**

#### *2.1. Occurrence of Emerging Toxins in Maize Samples*

Maize samples collected during the 2016 harvest were analyzed and the results are shown in Table 1. MON and BEA had the highest presence among emerging mycotoxins (>80%), except in the West-Backa region (50%). Other emerging mycotoxins were not detected at all or were found in fewer samples (<20%). Overall, maize samples from the Middle-Banat region were the most contaminatedfor all investigated emerging mycotoxins. MON, BEA, and FUS were present in all regions. Mean levels of MON ranged from 189.97 μg/kg (West-Backa) to 920.10 μg/kg (Srem). BEA mean levels were between 6.82 μg/kg (West-Backa) and 34.79 μg/kg (Srem). FUS levels were the highest among all tested mycotoxins. They ranged from 328.50 μg/kg in South-Backa to 12,272.00 μg/kg in a sample from the West-Backa region. ENs were found in all regions except Srem, with the highest mean levels in samples originating from Middle-Banat.


**Table 1.** Occurrence of emerging toxins in maize samples collected in the Republic of Serbia in 2016.

The results of maize samples collected during 2017 are summarized in Table 2. ENs were not found in any of the four regions. MON, BEA, and FUS were found in all regions, except for FUS, which was not foundin the sample from North-Backa. All the samples from the South-Banat region were contaminated with MON and BEA. The highest mean levels of MON (499.00 μg/kg) and FUS (3415.88 μg/kg) were recorded in the West-Backa region, while the highest mean level of BEA (12.26 μg/kg) was recorded in the South-Banat region.

**Table 2.** Occurrence of emerging toxins in maize samples collected in the Republic of Serbia in 2017.



**Table 2.** *Cont.*

In 2018, MON was found in all three regions (South-Backa, North-Backa, and South-Banat), BEA and FUS were not found in samples from the South-Banat region and ENs were not present in any sample (Table 3). The mean levels of all tested mycotoxins were highest in the South-Backa region (MON 199.32 μg/kg, BEA 4.89 μg/kg and FUS 5793.79 μg/kg).

**Table 3.** Occurrence of emerging toxins in maize samples collected in the Republic of Serbia in 2018.


#### *2.2. Climate Conditions*

Reports from the Republic Hydrometeorological Service of Serbia [30] showed that the vegetation period of 2016 (April–September) in the territory of Serbia was warmer with somewhat higher precipitation than the long-term average. The deviation of mean daily temperatures during the vegetation period showed positive values (0.8 ◦C to 1.6 ◦C). The standardized precipitation index (SPI-3), determined for the summer period from 1 June to 31 August, showed normal humidity conditions for most of the territory of Vojvodina. However, in some parts of Vojvodina, moderate to extremely humid conditions were recorded (Figure 1B). Such conditions were registered in the Middle-Banat and West-Backa regions, and some parts of the South-Backa and South-Banat regions. In some production areas, strong winds and hail storms were recorded and certainly contributed to the damage of the grains and the occurrence of fungi on crops. The moisture in deeper soil layers in the middle of June was significantly reduced in Vojvodina as a result of a weaker inflow of precipitation in these areas.

If the weather conditions were observed in more detail, warm but unstable weather prevailed during the transition from May to June, and the agrometeorological conditions allowed the intensive development of maize. The trend of variable but warm weather continued in June. In the middle of the month, due to the influx of very hot air, the temperatures were considerably above the average for this period of the year. Maximum daily temperatures reached 36 ◦C on some days [30]. Thermal conditions were favorable for the intensive development of spring agricultural crops. By the end of the first decade of August, the weather was mostly dry and stable, but since the beginning of the second decade of August, the air temperatures moved around and were below average values. Maximum air temperatures were up to 28 ◦C, while the minimum morning temperatures were significantly below the average values for this time of the year [30]. Significant precipitation, mostly rain showers, was

recorded on the territory of the entire country. During July and August, 2 to 3 times more rain was registered in the territory of Serbia compared to the average quantities.

**Figure 1.** Regions of sample origin: **1**—West-Backa, **2**—South-Backa, **3**—Srem, **4**—Middle-Banat, **5**—South-Banat, **6**—North-Backa. (**A**) Humidity conditions in Serbia based on the Standardized Precipitation Index (SPI-3) determined for the summer period from 1 June to 31 August in 2016 (**B**) Reproduced from Agrometeorološki uslovi u proizvodnoj 2015/2016 godini, 2017, Republic Hydrometeorological Service of Serbia [30], 2017 (**C**) Reproduced from Agrometeorološki uslovi u proizvodnoj 2016/2017 godini, 2018, Republic Hydrometeorological Service of Serbia [31], and 2018 (**D**) Reproduced from Agrometeorološki uslovi u proizvodnoj 2017/2018, 2019, Republic Hydrometeorological Service of Serbia [32].

Detailed data on precipitation and temperature obtained from Metos®automatic weather stations (Metos®, Pessl Instruments, Weiz, Austria) in the observed regions in 2016 were compared to the multiannual average for 1981–2010 [33].Precipitation data (Figure 2) show that the Middle-Banat region, which overall was the most contaminated with emerging *Fusarium* toxins, and Srem, which showed the highest values of toxins, had precipitation values higher than the average in June, but they were noticeably lower in July. Average daily air temperature data (Figure 2) show that the temperature was around or slightly above the long-term average until August, when the temperature decreased. Average daily air temperatures were similar in all regions.

Reports from the Republic Hydrometeorological Service of Serbia [31] showed that the vegetation period of 2017 (April–September) was warmer and dryer than the multiannual average. The mean daily temperatures were 0.9–1.7 ◦C higher than the average, while precipitation was 20% lower than the average. SPI-3 showed extreme drought in the South-Banat region, while other observed regions were affected by high to moderate drought (Figure 1C).

The vegetation period started with unusually cold weather in April, but the weather conditions quickly normalized and became optimal for plant development during May [31]. During June, mean daily temperatures were higher than the average [31]. Maximum daily temperatures reached over 35 ◦C on some days, especially during the first and the last decade of the month. Hot weather continued during July. In most regions, precipitation was below the multiannual average, especially in the South-Banat region, where it was 50% lower than the average. The temperatures at the beginning of August were extremely high (38–42 ◦C) and higher than the average during the whole month [31]. Most regions were affected with drought, except the South-Banat region, which had 50% more rainfall than the average.

**Figure 2.** Deviation of total rainfall amount (columns) and average daily air temperature (lines) from the multiannual average (1981–2010) in 2016.

Precipitation and average daily air temperature data obtained from Metos®automatic weather stations (Metos®, Pessl Instruments, Weiz, Austria) in the observed regions during 2017 were compared to the multiannual average from 1981–2010 [33]. Precipitation data (Figure 3) showed that all regions had lower precipitation than the average during June. This trend continued through July for all regions except West-Backa, where high occurrences of MON and FUS were observed. The deviation of average daily air temperature (Figure 3) shows that temperature in April was lower than the multiannualaverage in all regions. During May, it was slightly above the multiannualaverage in all regions except South-Banat. In June the temperature was 1.4–2.4 ◦C higher than the average. Temperatures continued to be above the average until September.

**Figure 3.** Deviation of total rainfall amount (columns) and average daily air temperature (lines) from multiannual average (1981–2010) in 2017.

According to reports from the Republic Hydrometeorological Service of Serbia [32], the vegetation period in 2018 (April–September) was 1.8–2.6 ◦C warmer than the multiannual average. SPI-3 showed normal to moderately humid weather conditions in all observed regions (Figure 1D).

The vegetation period started with unusually warm weather in April when the mean daily temperature was 4–5 ◦C above the multiannualaverage [32]. During May, weather conditions were optimal for plant growth. In the beginning of June, maximum daily temperatures were high (28–34 ◦C), but temperature decreased in the last decade of the month when maximum daily temperatures were in the range from 19 ◦C to 24 ◦C [32]. Frequent rain showers were recorded during June and continued through July. Warm weather with temperatures above multiannualaverage continued through August and September [32].

The deviation of precipitation and average daily air temperature data collected from Metos®automatic weather stations (Metos®, Pessl Instruments, Weiz, Austria) during 2018 from the multiannual average of 1981–2010 [33] was observed. Precipitation data (Figure 4) showed that precipitation was below average during the whole vegetation period in the North-Backa region. Precipitation was lower than the average in all three regions in April, May, August, and September. In June, precipitation was above average in the South-Backa and South-Banat regions. South-Backa was the only region where high precipitation continued in July.Average daily air temperature data (Figure 4) show that the temperature was above the long-term average in all regions during April and May. Temperature decreased during June and in July, reaching the multiannualaverage value in the South-Backa region, while the North-Backa and South-Banat regionswere below average. In August and September, the temperature increased again above the multiannual average in all observed regions.

**Figure 4.** Deviation of total rainfall amount (columns) and average daily air temperature (lines) from the multiannual average (1981–2010) in 2018.

#### *2.3. Statistical Analysis*

The Kruskal–Wallis test (95% confidence level) found significant differences in MON levels deriving from different years and different regions. Additionally, MON levels significantly differed when compared in terms of average seasonal temperature and precipitation. For FUS, significant differences were found between years, average seasonal temperatures and precipitation, but not between regions. Statistically significant differences in BEA levels were found between years and regions. Furthermore, the Spearman correlation determined a slightly moderate negative linear correlation between average seasonal temperatures and mycotoxin contamination levels (*r* = −0.41 for MON, *r* = −0.5 for BEA, and *r* = −0.45 for FUS). For monthly weather data, only temperatures in May showed a moderate negative linear correlation (*r* = −0.5 for MON, r = −0.58 for BEA, and r = −0.48 for FUS), while precipitation in May showed a moderate positive linear correlation with contamination levels of observed mycotoxins (*r* = 0.51 for MON, *r* = 0.59 for BEA, and *r* = 0.49 for FUS). Stepwise regression showed (with 95% confidence level) that air temperatures from May to August had statistically significant differences in MON levels. For FUS, temperatures and precipitation in any month, as well as average values for the whole growing season, did not show statistically significant differences among contamination levels. May temperatures had a statistically significant difference in BEA levels. The adjusted R-squared values were between 0.101 and 0.2278 and showed that these models cannot be used as prediction models; however, they gave insight about statistically significant variables that influence mycotoxin contamination.

#### **3. Discussion**

Sutton [34] explained that for maize, *Fusarium* infection of the ear most frequently takes place through the tip of the ear when the fungi penetrate through the silk during maize flowering. Very humid weather during the period from silking to ripening enables ear contamination [35]. The ear is the most susceptible to contamination at the beginning of silking, while the susceptibility decreases with silk aging [36,37]. The silking period in the climatic region of Serbia takes place during July and the first half of August.

According to the Republic Hydrometeorological Service of Serbia [30], most of the critical period for *Fusarium* infection of maize (July–August) in 2016 was characterized as dry and stable weather. However, since the beginning of the second decade of August, air temperatures ranged around and below average values [30]. Maximum air temperatures were up to a maximum of 28 ◦C, while the minimum morning temperatures were significantly below the average values for this time of the year. However, moderate to extremely humid conditions occurred during the summer of 2016 in the Middle-Banat and West-Backa regions and some parts of the South-Backa and South-Banat regions, which may have led to *Fusarium* fungi growth and the consequent production of mycotoxins. Moreover, cool, cloudy and humid weather during July and August did not favor agricultural crops and such conditions probably caused plant stress and higher susceptibility to *Fusarium* infection.

In 2017, precipitation was lower than the multiannual average during summer months in all regions except South-Banat, where high levels of BEA were recorded. High precipitation in these two regions during silking (July and beginning of August), when the maize is the most susceptible to *Fusarium* infections were favorable for fungal development, may have led to high levels of emerging toxins in samples from these regions.

High precipitation during June 2018 in the South-Backa and South-Banat regions and during July 2018 in South-Backa resulted in South-Backa having the highest mean levels of MON, BEA, and FUS. Warm weather during July, together with high humidity in the South-Backa region enabled *Fusarium* infection of the ears, which may be related to the high mycotoxin contamination of samples from this region.

Emerging fusariotoxins were mostly investigated in Mediterranean countries. Juan et al. [22] analyzed 93 samples of organic cereals and organic cereal products from several local markets in Italy for the presence of different mycotoxins, including BEA, EN A, EN A1, EN B, EN B1. The authors found that levels of some emerging *Fusarium* mycotoxins ranged as follows: BEA 6.7–41 μg/kg, EN A 7.2–29.8 μg/kg, EN A1 5.3–64.3 μg/kg, EN B 5.5–102 μg/kg and EN B1 5.5–33.1 μg/kg. Among the commodities, the occurrence was the highest in wheat samples. Serrano et al. [21], investigated fusariotoxins' occurrence in the Mediterranean area. They found that BEA was present in 2 of 14 maize samples and in 1 of 22 maize-based products. Obtained levels were 2.1 and 73.9 μg/kg for maize and 5.2 μg/kg for maize-based products, respectively. A high divergence among detected BEA levels was also found in this study (0.03–136 μg/kg). Later, Serrano et al. compared levels of emerging

fusariotoxins between organic and conventional pasta in Spain [38]. They found that organic pasta was more contaminated than the conventional type. Contamination levels were 0.10–20.96 μg/kg for BEA, and 0.05–8.02 μg/kg for FUS, while ENs levels were 0.25–979.56 μg/kg. Remarkably high levels of emerging mycotoxins in raw cereals were found in Morocco by Zinedine et al. [39]. EN A1 was predominant among ENs with a presence in 39% of samples and levels ranging from 14 to 445 mg/kg. BEA was found in 26.5% of samples, with levels ranging from 1 to 59 mg/kg, while FUS was present in 7.8% of samples (levels from 0.6 to 2 mg/kg). Regarding maize samples, 42% contained ENs with mean levels of 207 mg/kg (EN A1), 54 mg/kg (EN B), 8 mg/kg (EN B1), while EN A was not detected. A similar situation was observed in cereals from the Spanish market [40]. The authors reported very a high presence of ENs (73.4%), wherein EN A1 was the most frequent with the highest levels (33.36–814.42 mg/kg). BEA was found in 32.8% of samples in the range of 0.51–11.78 mg/kg, and FUS levels were between 1.01–6.63 mg/kg with the presence in 7.8% of samples. In maize samples, the presence of ENs was 89%, BEA was found in 21% and FUS was in only one sample; on the other hand, ENs were found only during 2016 in this study, but not in any samples from 2017 and 2018, while BEA and FUS were detected every year. The highest mean level in maize was obtained for EN A1 of 813.01 mg/kg, while in this survey the highest mean level of EN A1 was only 9.30 μg/kg. In another study in Morocco on maize-based breakfast cereals, Mahnine et al. [41] obtained mean levels of 113 mg/kg for EN A1 and 20.1 mg/kg for EN B1, respectively, while EN A, EN B, FUS, and BEA were below LOQ in all samples. Tunisian cereals were highly contaminated as well. Maize-based cereals only contained EN A1 and ENB1 with mean levels of 113 mg/kg and 20.1 mg/kg. Oueslati et al. [20] obtained the presence of ENs in 96% of samples, where once more EN A1 was predominant (92.1%). Mean values were the highest in the case of EN A1 (up to 480 mg/kg). Only 3 samples of maize were analyzed. Two samples were positive, one with EN A1 (29.6 mg/kg) and another contained EN B1 (17.0 mg/kg). Notably, none of these authors analyzed results along with the weather conditions. Emerging fusariotoxins were also studied in rice. In Morocco, considerable contamination with ENs and BEA was revealed, but not with FUS [19]. In rice samples from Iran, a significant presence was found only in the case of BEA (40%), but in very low amounts [23].

Emerging fusariotoxins were also studied in some non-Mediterranean European countries. Goertz et al. [42] investigated the contamination of different maize hybrids in Germany during 2006 and 2007. BEA was found in 52% of samples from 2006 and in 33% of samples from 2007. Mean levels were 390 μg/kg and 240 μg/kg, respectively. MON was detected in 45% and 43% of samples, respectively, with mean values of 280 μg/kg and 110 μg/kg, respectively. Among EN, only EN B was investigated. Although its presence was relatively high (41% and 30%, respectively) the levels were the lowest (mean of 70 μg/kg and 160 μg/kg, respectively) among investigated emerging fusariotoxins. The authors explained that moderate temperatures and frequent precipitation recorded during early growth stages in 2007 were favorable for *Fusarium* growth. This is in accordance with the results of correlation analysis in this study, which showed a moderately negative correlation between May temperatures and toxin contamination, together with a moderately positive correlation between May precipitation and toxin contamination. However, a higher mycotoxin presence and higher contamination levels in Germany were found in samples from 2006, which they associated with maize exposure to drought stress in July and September 2006. In Norway, Uhlig et al. [43] investigated MON occurrence in Norwegian grain (oats, barley, and wheat) during a three-year period (2000–2002). MON was found in 46% of samples and the obtained levels were between 43 and 950 μg/kg. The authors noted that the highest prevalence of MON was found in the 2002 season (67%), along with the highest concentration. In Poland, Chelkowski et al. [44] detected ENs and BEA in 18 out of 27 maize samples (levels of 0.8–46.0 mg/kg). Unfortunately, in both studies, the weather conditions were not discussed.

To the best of our knowledge, studies on emerging *Fusarium* toxins in Serbia have not been done to date. On the other hand, some studies have occurred in surrounding countries. In Romania, Stanciu et al. [45] found that ENs were the most frequent (73%) mycotoxins in both wheat and wheat flour, while EN B was detected the most (71%). The highest observed concentration was 407 μg/kg in

wheat samples. Mean values were 19 μg/kg in wheat flour and 128 μg/kg in wheat. In neighboring Croatia, Jurjevi´c et al. [46] investigated BEA presence in 209 maize samples originating from the 1996 and 1997 growing seasons. The authors found that 17.4% of samples from 1996 contained BEA at the mean level of 393 μg/kg and maximum concentration of 1864 μg/kg. In samples from 1994, only one of 104 samples contained BEA.

Based on the obtained results and available published data, the results from this study are in accordance with those found in Croatia, Italy, and Germany, while results from Poland, Spain, Morocco, and Tunisia are one order of magnitude higher. Unfortunately, studies from Romania, Norway, and Iran did not include maize or maize-based products and therefore a valid comparison cannot be made.

#### **4. Conclusions**

The main source of emerging *Fusarium* mycotoxins are cereals that are used in food and feed production, and they may thus pose a potential risk for human and animal health. Since environmental conditions in Serbia are favorable for the occurrence of mycotoxigenic fungi, monitoring of "traditional" but also "emerging" *Fusarium* toxins is essential for producing safe food and feed. The results indicated that most attention should be paid to fusaproliferin (FUS) and moniliformin (MON). Additionally, monitoring studies for emerging *Fusarium* mycotoxins are necessary for legislative purposes, because in the near future appropriate maximum contamination levels should be set for several mycotoxins by relevant authorities [38].

#### **5. Materials and Methods**

#### *5.1. Samples*

In total, 190 representative samples from commercial fields in Serbia were analyzed. Samples were collected during harvest in the northern Serbian province of Vojvodina, which is the country's most important agricultural area, over three years: 73 samples from 28 localities in 2016, 72 samples from 12 localities in 2017 and 45 samples from 13 localities in 2018. Localities were clustered based on their administrative area into 6 regions: West-Backa, South-Backa, Srem, Middle-Banat, South-Banat, and North-Backa (Figure 1A).

Each sample was transported to the laboratory immediately after sampling and stored in a freezer at −20 ◦C until analysis. Prior to analysis, the samples were allowed to reach room temperature. All samples were milled on a laboratory mill so that >93% passed through a sieve with a pore diameter of 0.8 mm and a portion was taken for analysis.

#### *5.2. Extraction and Mycotoxin Analysis in Maize Samples*

Five grams of each milled sample were extracted using a 20 mL extraction solvent (acetonitrile–water–acetic acid (VWR, Vienna, Austria), 79:20:1, *v*/*v*/*v*) followed by a 1 + 1 dilution using acetonitrile–water–acetic acid (VWR, Vienna, Austria) (20:79:1, *v*/*v*/*v*) and a direct injection of 5 μL diluted extract.

Liquid chromatography-tandem mass spectrometry (LC-MS/MS) screening of target fungal metabolites was performed at theInstitute of Bioanalytics and Agro-Metabolomics, Department of Agrobiotechnology (IFA-Tulln), University of Natural Resources and Life Sciences, Vienna, with a QTrap 5500 LC-MS/MS System (Applied Biosystems, Foster City, CA, USA) equipped with a TurboIon Spray electrospray ionization (ESI) source and a 1290 Series HPLC System (Agilent, Waldbronn, Germany). Chromatographic separation was performed at 25 ◦C on a Gemini® C18-column, 150 <sup>×</sup> 4.6 mm i.d., 5 <sup>μ</sup><sup>m</sup> particle size, equipped with a C18 4 × 3 mm i.d. security guard cartridge (all from Phenomenex, Torrance, CA, USA). The chromatographic method, chromatographic and mass spectrometric parameters, as well as the method validation data, are described by Malachova et al. [47]. Electrospray ionization-tandem mass spectrometry (ESI-MS/MS) was performed in the time-scheduled multiple reaction monitoring (MRM) mode both in positive and negative polarities in two separate chromatographic runs per sample

by scanning two fragmentation reactions per analyte. The MRM detection window of each analyte was set to its expected retention time of ±27 sec and ±48 sec in the positive and the negative mode, respectively. Confirmation of a positive analyte identification was obtained by the acquisition of two MRMs per analyte (excepting moniliformin, which exhibits only one fragment ion), which yielded 4.0 identification points according to commission decision 2002/657/EC. In addition, the LC retention time and the intensity ratio of the two MRM transitionswas in accordance with the related values of an authentic standard within 0.1 min and 30% rel., respectively.

Quantification was based on an external calibration using a serial dilution of a multianalyte stock solution, and results were corrected for apparent recoveries. The accuracy of the method is verified on a continuous basis by regular participation in proficiency testing schemes [47,48] organized by BIPEA (Gennevilliers, France). Based on the submitted results, a general expanded measurement uncertainty of 50% has been determined [49]. In the case of the 175 results already submitted for maize and maize-based feed, 168 results were in the satisfactory range (*z*-score between −2 and 2).

#### *5.3. Statistical Analysis*

Statistical analysis (Supplementary Materials) was performed using the computing environment *R* (*R* Core Team, Vienna, Austria) [50] on the data from regions where samples were collected during all three years of research (South-Backa and South-Banat). The Shapiro–Wilk normality test was used to check the distribution of the data. Since the data were not normally distributed, nonparametric tests were used for further analysis. The Kruskal–Wallis test was used to check whether the mean ranks of the mycotoxin levels were the same in all groups. Spearman's correlation was used to determine the correlation between climate conditions and mycotoxin contamination levels. Furthermore, Stepwise regression with backward steps was used to obtain the optimal model and significant months in terms of temperature and precipitation values that influence mycotoxin levels.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6651/11/6/357/s1, Figure S1. Spearman Correlation Results (*r*).

**Author Contributions:** Conceptualization, I.J. and F.B.; Data curation, T.D., S.K., Z.S., and A.S.; Formal analysis, T.D., S.K., and D.G.; Funding acquisition, I.J., R.K., and F.B.; Investigation, R.K., M.S., Z.S., D.G., and A.S.; Methodology, R.K., M.S., and D.G.; Project administration, I.J. and F.B.; Resources, I.J. and F.B.; Software, T.D. and S.K.; Supervision, I.J., R.K., and F.B.; Validation, I.J., R.K. and M.S.; Visualization, T.D. and S.K.; Writing—originaldraft, T.D. and S.K.; Writing—review and editing, I.J., R.K., F.B., and Z.S.

**Funding:** This research was funded by MyToolBox (EU's Horizon 2020 research and innovation programme, agreement No 678012), research project of the Ministry of Education, Science and Technological Development, Republic of Serbia (project III 46005) and Provincial Secretariat for Higher Education and Scientific Research of the Autonomous Province of Vojvodina through the project "Application of novel and conventional processes for removal of most common contaminants, mycotoxins and salmonella, in order to produce safe animal feed inthe territory of AP Vojvodina", Project No. 114-451-2505/2016-01.

**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/).

#### **Occurrence of the Ochratoxin A Degradation Product 2**- **R-Ochratoxin A in Co**ff**ee and Other Food: An Update**

**Franziska Sueck 1, Vanessa Hemp 1, Jonas Specht 1, Olga Torres 2,3, Benedikt Cramer 1,\* and Hans-Ulrich Humpf 1,\***


Received: 18 April 2019; Accepted: 3 June 2019; Published: 8 June 2019

**Abstract:** Food raw materials can contain the mycotoxin ochratoxin A (OTA). Thermal processing of these materials may result in decreased OTA levels but also in the formation of the thermal isomerization product 2- R-ochratoxin A (2- R-OTA). So far, only 2- R-OTA levels reported from 15 coffee samples in 2008 are known, which is little when compared to the importance of coffee as a food and trading good. Herein, we present results from a set of model experiments studying the effect of temperatures between 120 ◦C and 270 ◦C on the isomerization of OTA to 2- R-OTA. It is shown that isomerization of OTA starts at temperatures as low as 120 ◦C. At 210 ◦C and above, the formation of 25% 2- R-OTA is observed in less than one minute. Furthermore, 51 coffee samples from France, Germany, and Guatemala were analyzed by HPLC-MS/MS for the presence of OTA and 2- R-OTA. OTA was quantified in 96% of the samples, while 2- R-OTA was quantifiable in 35% of the samples. The highest OTA and 2- R-OTA levels of 28.4 μg/kg and 3.9 μg/kg, respectively, were detected in coffee from Guatemala. The OTA:2- R-OTA ratio in the samples ranged between 2.5:1 and 10:1 and was on average 5.5:1. Besides coffee, 2- R-OTA was also for the first time detected in a bread sample and malt coffee powder.

**Keywords:** Ochratoxin A; 2- R-ochratoxin A; 14(R)-ochratoxin A; coffee; degradation; processing; roasting; modified mycotoxins; masked mycotoxins

**Key Contribution:** 2- R-OTA occurs in most of the tested coffee samples in levels up to 3.9 μg/kg. Other food might also contain significant amounts of 2- R-OTA as the isomerization of OTA to 2- R-OTA starts at 120 ◦C.

#### **1. Introduction**

The mycotoxin ochratoxin A (OTA, Figure 1) can be found in a broad spectrum of food raw materials infested with fungi of the genera *Aspergillus* and *Penicillium* as well as in food products derived from these commodities [1]. Considering exposure within the European Union, cereals and cereal products, such as pasta, bread, and beer, are the most relevant OTA sources due to the high consumption rates of these food items. In Germany, for instance, these products are on average responsible for about 67% of the OTA intake in adults. Besides this, approximately 12% of the total OTA exposure occurs from coffee drinking and a further 6% from cocoa, 6% from meat, and 5% from wine [1].

**Figure 1.** Structures of ochratoxin A (OTA), its thermal isomerization product 2- R-ochratoxin A (2- R-OTA), and the degradation products 2- -decarboxy-ochratoxin A (DC-OTA) and ochratoxin α-amide (OTamide).

Most food items undergo food processing, such as milling, baking, roasting, frying, or fermentation. All of these steps can have an impact on the mycotoxin burden of the product, for instance by physical removal or chemical modification [2,3]. To what extent mycotoxins are modified during these processing steps depends on the parameters but also on the chemical structure of the compound. Chemical modifications can result in the formation of mycotoxin conjugates, such as the deoxynivalenol (DON) glucosides formed during the fermentation of dough [4–6], binding to food-matrix components as shown for fumonisins [7], or the formation of degradation products, such as the norDON series [8].

Depending on the commodity, OTA undergoes different types of processing, such as baking, extrusion cooking, roasting, and fermentation. The reduction of OTA levels during these processes has been found to range between no impact and almost a 100% decrease [3,9–12]. A strong OTA decrease was observed during coffee roasting, where up to a 90% reduction was reported in most of the studies [13–17]. On the other hand, in some other coffee-roasting experiments, only a slight reduction of less than 12% was noted [18,19]. Several efforts have been made to identify the chemical reaction that leads to lower OTA levels in roasted coffee. Bittner et al. (2013) showed that OTA binds to polysaccharides of the coffee bean during roasting [20]. The high temperatures present during coffee roasting also lead to a decarboxylation to form decarboxy-ochratoxin A (DC-OTA) as well as to racemization of the phenylalanine moiety to yield 2- R-ochratoxin A (2- R-OTA, previously reported as 14(R)-OTA) and to the formation of ochratoxin α-amide (OTamide) as shown in Figure 1 [20–22]. A quantitative analysis of 15 coffee samples from the German market indicated that DC-OTA is only formed in minor amounts, while the concentration of 2- R-OTA was up to 0.63 μg/kg. A ratio of OTA and 2- R-OTA lower than 4:1 was determined [21]. OTamide was not analyzed in this study.

OTA has been shown to be nephrotoxic, hepatotoxic, teratogenic, and immunotoxic in various species and was classified by the International Agency for Research on Cancer (IARC) as possibly carcinogenic to humans (group 2B) in 1993 [23,24]. The mode of action is still not clear and controversially discussed [25,26]. In comparison to OTA, only very little information on the toxicity of the degradation product 2- R-OTA is available: 2- R-OTA shows in a Cell Counting Kit-8 (CCK-8) assay a 10-fold lower cytotoxic effect on IHKE-cells compared to OTA [21,27]. Several studies have shown that OTA is detectable in blood of humans and animals [28], with human plasma concentrations ranging roughly between 0.05 ng/mL and 5 ng/mL [29,30]. The thermal isomerization product 2- R-OTA is also of relevance, as it was recently found in humans in comparable levels as OTA [31].

Little is known on the kinetics of the isomerization of OTA to 2- R-OTA, and the minimum temperature required for the thermal conversion of OTA to 2- R-OTA has not been elucidated [21]. Depending on these data, bakery products and breakfast cereals might also be potential sources of 2- R-OTA exposure.

Herein, we report the analysis of 51 coffee samples from Germany, France, and Guatemala to extend the database on 2- R-OTA levels in food. Furthermore, the kinetics of the isomerization of OTA to 2- R-OTA were studied in a model system covering temperatures between 120 and 270 ◦C. As isomerization of OTA to 2- R-OTA has already been observed at 120 ◦C, further thermally treated food samples were analyzed on the occurrence of OTA and 2- R-OTA.

#### **2. Results**

#### *2.1. Degradation of OTA in Model Heating Experiments*

The isomerization of OTA to 2- R-OTA as well as the formation of other degradation products were studied in model heating experiments. To that end, OTA was heated at temperatures of 120, 150, 180, 210, 240, and 260 ◦C without solvent for 1 to 30 min, respectively. The observed degradation curves of OTA are shown in Figure 2. Subsequent analysis by HPLC-FLD for the known degradation products 2- R-OTA and DC-OTA resulted in the curves shown in Figure 3. At the lowest tested temperature of 120 ◦C, OTA remained almost stable over the entire heating period with only 3% 2- R-OTA formed within 30 min. After the same time, but heating at 150 ◦C, approximately 20% of OTA was converted to 2- R-OTA. At 180 ◦C, a fast racemization of OTA towards 2- R-OTA was observed with an equilibrium between both compounds reached after approximately 20 min. Above this temperature, racemization of OTA towards 2- R-OTA was achieved after 1 to 5 min of heating, followed by further degradation of both diastereomers. After 30 min at 210, 240, and 260 ◦C, only 80%, 35%, and 20%, respectively, of the sum of OTA and 2- R-OTA were detectable. Screening for DC-OTA and OTamide revealed only small quantities of less than 1% (data not shown), and were not further considered.

**Figure 2.** Degradation curves of ochratoxin A (OTA) at temperatures between 120 and 260 ◦C for 1–30 min.

**Figure 3.** Formation of 2- R-ochratoxin A (2- R-OTA) during heating of OTA at temperatures between 120 and 260 ◦C for 1–30 min.

#### *2.2. Co*ff*ee Powder: Current Situation of OTA and 2*- *R-OTA Content*

The published dataset on 2- R-OTA contamination in food and the OTA:2- R-OTA ratio is limited to data from the analysis of 15 coffee samples in 2008 [21]. Therefore, in order to extend the database and to evaluate the current situation, we analyzed in total a set of 51 coffee samples from France, Germany, and Guatemala. Among the 14 commercially available roasted coffee samples from the French and German market, five were categorized as espresso powder and nine samples as coffee powder packages. Three of the coffee samples were labelled as organically grown and two as decaffeinated. The results are presented in Figure 4. In all coffee samples except two from Guatemala, OTA was quantifiable in a range between 0.26 and 28.4 μg/kg with a mean contamination of 1.98 μg/kg. No sample from Europe, but three samples from Guatemala (Figure 4, coffee samples 1-3) showed OTA levels above the legal limit of 5 μg/kg OTA set by the European Union [32]. Among these three, two samples exceeded the limit by factors of 5 and 4, respectively.

**Figure 4.** OTA and 2- R-OTA contents of the coffee powder samples from Germany (red), Guatemala (black), and France (blue). <sup>a</sup> decaffeinated, <sup>b</sup> espresso, <sup>c</sup> organically grown. The standard deviation is given for duplicate analysis.

In two decaffeinated coffee samples (Figure 4, samples 11 and 24), OTA concentrations of 1.30 and 0.88 μg/kg were detected. Interestingly, the three analyzed organic coffee powders, including one espresso sample (Figure 4, samples 32, 34, and 42), with OTA concentrations of 0.64, 0.62, and 0.38 μg/kg were among the coffee samples with the lowest detected OTA levels. The four other analyzed espresso samples (Figure 4, samples 12, 18, 26, and 38) contained 1.27, 1.09, 0.79, and 0.51 μg/kg OTA, which was comparable to the OTA contents of the other coffee samples from Europe.

The isomer 2- R-OTA was quantifiable in 18 of the 51 coffee samples (35%) with a mean concentration of 0.27 μg/kg. The highest 2- R-OTA level of 3.9 μg/kg was determined in the coffee sample from Guatemala containing 28 μg/kg OTA (Figure 4, sample 1). Among the European coffee samples, an espresso coffee powder (Figure 3, sample 12) with a 2- R-OTA concentration of 0.52 μg/kg and an OTA level of 1.27 μg/kg showed the highest 2- R-OTA contamination and, furthermore, the highest OTA:2- R-OTA ratio of 2.5:1. However, there was no indication that espresso coffee production favors 2- R-OTA formation, as the other analyzed espresso coffee samples contained no, or a low amount of, 2- R-OTA and showed OTA:2- R-OTA ratios between 4:1 and 10:1. In both decaffeinated coffee samples (Figure 3, samples 11 and 24), the 2- R-OTA concentration was below the limit of quantification (LOQ). Taking all coffee samples into account that were positive for 2- R-OTA, a OTA:2- R-OTA ratio between 10:1 and 2.5:1 with an average OTA:2- R-OTA-ratio of 5.5:1 was observed.

#### *2.3. Evaluation of Other Thermally Processed Food Materials as Sources of 2*- *R-OTA Exposure*

The data from the heating experiment with pure OTA suggest that, at a comparatively low temperature of 120 ◦C, a slow isomerization of OTA to 2- R-OTA can be observed. Such temperatures can be reached during different kinds of baking processes, and the conditions might be sufficient for the formation of significant quantities of 2- R-OTA. To that end, different thermal processed food samples were screened (Table 1). The choice of the analyzed food samples was based on thermal processing conditions and the overall contribution to OTA exposure [1]. Consequently, 30 samples of cocoa (8) and cereal products (22) were analyzed for OTA and 2- R-OTA. Details on the samples are given in Table 1.


**Table 1.** Results of a survey screening for the presence of OTA and 2- R-OTA in thermally processed food. (n: numbers of food samples; n.d.: not detectable).

Generally, OTA contamination of the analyzed food samples was low, resulting in only two samples with OTA levels above the LOQ and 13 samples above the limit of detection (LOD). However, despite the low OTA contamination, six food samples were found to be positive for 2- R-OTA. In particular, coffee surrogates, such as instant malt coffee powder and malt coffee powder, were all found to be positive for this compound. With an OTA concentration of 0.62 ± 0.04 μg/kg and a 2- R-OTA concentration of 0.22 ± 0.02 μg/kg, comparable racemization rates as for roasted coffee were observed. Additionally, one bread sample (pumpernickel) contained detectable amounts of 2- R-OTA. Pumpernickel is a long-term heated bread. To produce this type of bread, a loaf of rye is baked for 16–24 h at an oven temperature of approximately 110 ◦C.

#### **3. Discussion**

The thermal instability of the mycotoxin OTA has been reported for several food processing technologies. However only for roasted coffee have the degradation products 2- R-OTA and DC-OTA been described and quantified so far, and in a limited number of samples. In this study, it was shown in model heating experiments that a slow racemization of OTA to 2- R-OTA occurs at temperatures as low as 120 ◦C. Above these temperatures, a fast racemization of OTA was observed, with 25% 2- R-OTA being formed after 30 min at 150 ◦C, after 5 min at 180 ◦C, and after 1 min or less at 210 ◦C and above. Longer heating periods at temperatures of 240 ◦C and above resulted in a fast degradation of OTA and 2- R-OTA, which can be explained by other reactions, such as pyrolysis or polymerization. Compared with previously reported data, these results confirm the importance of 2- R-OTA as the main OTA degradation product. A slow conversion of OTA to 2- R-OTA occurring at temperatures as low as 120 ◦C has not yet been reported [21,22].

Thus, food samples, processed at temperatures far lower than those applied for coffee roasting, might be additional sources of 2- R-OTA exposure. However, to date, 2- R-OTA has only been detected

in blood samples from 34 coffee drinkers in concentrations between 0.021 and 0.411 ng/mL (mean: 0.21 ± 0.066 ng/mL) compared to 0.071–0.383 ng/mL (mean: 0.11 ± 0.093 ng/mL) for OTA. An average OTA:2- R-OTA ratio of 2:1 was determined and in some cases, the concentration of 2- R-OTA even exceeded that of OTA. No 2- R-OTA was detected in the set of 14 samples from non-coffee drinkers, suggesting that 2- R-OTA is predominantly present in roasted coffee. Nevertheless, the number of non-coffee drinkers participating in that study was low, and no correlation between the 2- R-OTA levels and overall coffee consumption was observed, making further sources of 2- R-OTA plausible [31,33].

Screening of a set of 51 coffee samples confirmed the role of roasted coffee as the key source of 2- R-OTA exposure, with 2- R-OTA levels of up to 3.9 μg/kg in highly contaminated coffee from Guatemala and up to 0.52 μg/kg in European coffee samples. No systematic differences between espresso coffee and other coffee with respect to the 2- R-OTA levels could be observed. For decaffeinated coffee, moderate OTA levels were detected, but no 2- R-OTA in concentrations above the LOQ. Although no information on the OTA content before roasting was available, the reported reductive effect of decaffeination on OTA levels seems to be limited [34]. These results are in good agreement with previously reported 2- R-OTA levels but also show that, in certain samples, such as one specific espresso powder, relatively high 2- R-OTA levels and OTA:2- R-OTA ratios as low as 2.5:1 can occur. In a previous study, for roasted coffee, OTA:2- R-OTA ratios between 10:1 and 4:1 (mean: 5:1) were determined from a set of 15 coffee samples. The variation of these ratios between different coffee samples also suggested a dependency of 2- R-OTA formation on the roasting process. Traditionally, green beans are roasted for between 8 and 20 min at temperatures between 160 and 240 ◦C using a drum roaster, a hot-air roaster, or a combination of both systems [35]. Oliveira et al. (2013) reported a correlation between roasting level and OTA degradation, while Castellanos-Onorio et al. (2011) observed for the two roasting techniques (drum roaster and hot-air roaster) a similar OTA reduction. In both cases, the formation or 2- R-OTA was unfortunately not studied [16,17]. In contrast to the traditional temperature regimes, to increase throughput, some companies have established a coffee-roasting process based on high temperatures of around 400 ◦C and roasting times of less than one minute [36].

To confirm the results from the model experiments, other thermally processed food samples, known to contribute to OTA exposure, were analyzed for the presence of 2- R-OTA, the most abundant thermal degradation product of OTA. Besides coffee, quantities of 2- R-OTA were detected in malt coffee as well as in traditionally baked rye bread (Pumpernickel). The latter is of special importance as it confirms the results from the model experiments, showing that OTA can be converted to 2- R-OTA at low temperatures between 100 and 120 ◦C. However, it has to be considered that this type of bread has a minimum baking time of 16 h, which makes a comparison with other bread and bakery products rather difficult. Other analyzed food samples contained no detectable amounts of 2- R-OTA but were also low for OTA.

The most relevant sources of OTA exposure are cereals, followed by coffee, cocoa, meat, and wine. For the cereal products biscuits and muesli, baking times are usually short and temperatures inside the product are mostly at 100 ◦C or below [37]. For roasted and expanded cereals, the situation is different; however, due to the low availability of industrially processed and naturally contaminated products, we were not able to prove whether a racemization of OTA during these processes may occur. Considering other typical food-processing procedures, wine can be excluded as potential source of 2- R-OTA due to low fermentation and processing temperatures. In the case of meat, temperatures above 100 ◦C are rarely reached in the inner part of the product (e.g., sausage). A different situation might be the manufacturing of canned meat, such as corned beef, where temperatures of up to 121 ◦C are applied for sterilization. Here, additional studies investigating the racemization of OTA in aqueous solutions are necessary. Thus, there is a need for more data on the occurrence of 2- R-OTA in these foods and on the toxic properties of this compound to allow for an adequate risk assessment.

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

Methanol (MeOH), acetonitrile (ACN), and toluene were obtained in gradient grade from Fisher Scientific (Schwerte, Germany). NaCl, formic acid, hexane, and Na2HCO3 were in pro analysi (p.a.) quality from Merck KGaA (Darmstadt, Germany). Potato dextrose agar (PDA), potato dextrose broth, and KH2PO4 were from Carl Roth (Karlsruhe, Germany), Xylene in p.a. quality was obtained from Honeywell (Seelze, Germany), NaHCO3 in p.a. quality was obtained from Grüssing (Filsum, Germany), and KCl in p.a. quality was obtained from VWR (Langenfeld, Germany). Purified water of ASTM type 1 quality was prepared with a Purelab Flex 2 system from Veolia Water Technologies (Celle, Germany).

#### *4.1. Biosynthesis of Standards*

OTA was isolated from cultures of *Aspergillus westerdijkiae* BFE 1115, provided by the Max Rubner Institute (Karlsruhe, Germany), which were activated for 2 days in liquid potato dextrose broth at room temperature and then incubated for two weeks at 27 ◦C on autoclaved durum wheat adjusted to a water content of 62.5% (*w*/*w*) and supplemented with 2.5% sodium chloride (*w*/*w*). OTA was extracted from the durum wheat cultures using tBME containing 0.5% formic acid, purified by liquid–liquid extraction with water at different pH levels and silica column chromatography with the solvent system toluene/tBME/formic acid (8/1.5/0.5 (*v*/*v*/*v*)). Finally, residual impurities were removed by crystallization from xylene/hexane (7/3 (*v*/*v*)). The purity of OTA was >99% as determined by HPLC-UV (220 nm) and NMR. 2- R-OTA, DC-OTA, OTα-amide, d5-OTA, and d5-2- R-OTA were prepared in-house as described elsewhere [21,22].

#### *4.2. Model Heating Experiments with OTA*

For the model heating experiments, a stock solution with a concentration of 274 μg/mL in acetonitrile was prepared. Aliquots of the solution (36.5 μL, 10 μg) were transferred into 1.5 mL vials and the solvent was evaporated to dryness under a stream of nitrogen at 40 ◦C. The dried thin film of OTA was heated for 1–30 min at temperatures of 120, 150, 180, 210, 240, and 260 ◦C, respectively. Subsequently, the samples were dissolved in 1 mL methanol/water/formic acid (63/37/0.15, (*v*/*v*/*v*)) and analyzed by HPLC-FLD for OTA and 2- R-OTA.

#### *4.3. Sample Collection*

European coffee samples were obtained from retail markets in France and Germany (and produced by industrial coffee roasting companies). The coffee samples originating from Guatemala were collected from small local markets and produced by local coffee roasters. Cocoa beans and nibs were provided as a gift from August Storck KG, Berlin, Germany. Other food samples reported in Table 1 were commercial products bought from retail markets in Germany.

#### *4.4. Sample Preparation*

Food samples were analyzed as described in literature with slight modifications [21,38]. To 5.00 g of homogenized ground coffee sample, 100 mL methanol/3% NaHCO3-solution (1/1, *v*/*v*) and 50 μL of a solution containing 100 ng/mL d5-OTA and 50 ng/mL d5-2- R-OTA in methanol were added. The mixture was extracted for three minutes using an Ultra-Turrax T25 mixer (IKA, Staufen, Germany) at a rotation speed of 9500 min−1. The obtained suspension was filtered through a 150 mm 3 HW folded filter (Sartorius-Stedim Biotech, Göttingen, Germany) and 5.00 mL of the filtrate were diluted with 45 mL phosphate-buffered saline (PBS) pH 7.4 (8 g NaCl, 1.2 g Na2HPO4, 0.2 g KCl, and 0.2 g KH2PO4 dissolved in 1 L H2O) before purification using an OchraTest WB (VICAM, available via Klaus Ruttmann, Hamburg, Germany) immunoaffinity column (IAC). After loading the IAC with the sample, the column was washed with 10 mL PBS and 10 mL water. OTA, 2- R-OTA, and DC-OTA were eluted with 2 mL methanol according to the protocol of the IAC manufacturer. The eluate was evaporated to dryness under a stream of nitrogen at 40 ◦C and reconstituted in 250 μL methanol/water/formic acid (60/40/0.1, *v*/*v*/*v*).

#### *4.5. Recovery Rate*

For the determination of the recovery rate, a coffee sample containing only traces of OTA and 2- R-OTA was fortified with three different concentrations of OTA (0.5, 2, and 8 μg/kg) and 2- R-OTA (0.4, 0.7, and 1.4 μg/kg) before analysis. To that end, aliquots of a standard solution of OTA and 2- R-OTA in acetonitrile were added to the homogenized sample and the solvent was allowed to evaporate for 2 h before extraction. The recovery rates were determined in duplicate and were 104.4 ± 4.0%, 102.9 ± 4.5%, and 103.9 ± 3.8% for 0.5, 2, and 8 μg/kg OTA and 106.0 ± 0.0%, 109.9 ± 11.2%, and 99.4 ± 3.8% for 0.4, 0.7, and 1.4 μg/kg 2- R-OTA, respectively.

#### *4.6. Calibration*

For HPLC-MS/MS experiments, an external seven-point calibration of OTA and 2- R-OTA with internal standards d5-OTA (1 ng/mL) and d5-2- R-OTA (0.5 ng/mL) in a concentration range from 0.1 to 10 ng/mL was used for quantification. The limit of detection (LOD) and the limit of quantification (LOQ) were determined by the signal-to-noise ratios 3 and 9, respectively. For OTA and 2- R-OTA, the LOQ was 0.1 ng/mL and the LOD was 0.03. For the HPLC-FLD experiments, a five-point calibration of OTA, 2- R-OTA (concentration range from 0.10 to 10.0 ng/mL), DC-OTA, and OTα-amide (concentration range from 0.05 to 2.00 ng/mL) was used for quantification. Samples exceeding the calibration range were diluted with methanol/water/formic acid (60/40/0.1, *v*/*v*/*v*) by an appropriate factor and reanalyzed. For all samples, calibration curves with correlation coefficient *r*<sup>2</sup> > 0.95 were calculated.

#### *4.7. HPLC-MS*/*MS*

An Agilent 1100 series HPLC (Agilent Technologies, Waldbronn, Germany) was coupled with an API 4000 QTRAP mass spectrometer (Sciex, Darmstadt, Germany) operated in electrospray ionization (ESI) positive mode with an ionization voltage of 5500 V. Data acquisition and quantification was carried out with the Analyst 1.6.2 software (Sciex). Chromatographic separation was achieved on a Nucleodur C18 Isis (150 × 2.0 mm; 5 μm) column with a 5 x 2 mm guard column of the same material (Macherey-Nagel, Düren, Germany) using a binary gradient at a flow rate of 0.3 mL/min. Methanol containing 0.1% formic acid was used as solvent A and water containing 0.1% formic acid as solvent B. The following linear gradient was used: 0 min 60% A, 1 min 60% A, 10 min 100% A, 10 min 100% A. The injection volume was 50 μL. The mass spectrometer was operated in selected reaction monitoring mode (SRM) with Gas 1 (nebulizer) set to 35 psi and Gas 2 (drying gas) set 350 ◦C and 45 psi. Nitrogen served as the Curtain gas (20 psi). The following transition reactions were monitored for 100 ms each (declustering potential (DP), collision energy voltage (CE), and collision cell exit potential (CXP) are given in brackets): OTA and 2- R-OTA: quantifier *m*/*z* 404.2→239.1 (DP 60 V, CE 33 V, CXP 14 V) and qualifier *m*/*z* 404.2→221.0 (DP 60 V, CE 50 V, CXP 15 V); d5-OTA and d5-2- R-OTA: quantifier *m*/*z* 409.2→239.1 (DP 60 V, CE 33 V, CXP 14 V), and quantifier *m*/*z* 409.2→221.0 (DP 60 V, CE 50 V, CXP 15 V).

#### *4.8. HPLC-FLD*

Samples of the degradation experiments were analyzed with HPLC Germany coupled to a fluorescence detector (X-LC, FP-2020 Plus, Jasco GmbH, Groß-Umstadt). Separation was achieved at 40 ◦C on a ReproSil-Pur C18-AQ (150 × 4.0 mm; 3 μm) column (Dr. Maisch GmbH, Ammerbuch, Germany) under isocratic conditions using a solvent mixture of methanol/water/formic acid (63/37/0.1, v/v/v) at a flow rate of 0.7 mL/min. The fluorescence detector was operated at an excitation wavelength of 330 nm and emission wavelength of 460 nm. The injection volume was 5 μL.

**Author Contributions:** Conceptualization, B.C., F.S. and H.-U.H.; investigation, F.S., O.T., V.H. and J.S.; resources, O.T. and H.-U.H.; writing—original draft preparation, F.S.; writing—review and editing, B.C. and H.-U.H.; supervision, B.C. and H.-U.H.; project administration, H.-U.H.; funding acquisition, B.C. and H.-U.H.

**Funding:** This research was funded by Deutsche Forschungsgemeinschaft (HU 730/10-2).

**Acknowledgments:** The authors thank Rolf Geisen for providing *A. westerdijkiae* BFE 1115.

**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* **The Influence of Processing Parameters on the Mitigation of Deoxynivalenol during Industrial Baking**

**David Stadler 1, Francesca Lambertini 2, Lydia Woelflingseder 3, Heidi Schwartz-Zimmermann 1, Doris Marko 3, Michele Suman 2, Franz Berthiller 1,\* and Rudolf Krska 1,4**


Received: 13 May 2019; Accepted: 30 May 2019; Published: 4 June 2019

**Abstract:** Deoxynivalenol (DON), a frequent contaminant of flour, can be partially degraded by baking. It is not clear: (i) How the choice of processing parameter (i.e., ingredients, leavening, and baking conditions) affects DON degradation and thus (ii) how much DON can be degraded during the large-scale industrial production of bakery products. Crackers, biscuits, and bread were produced from naturally contaminated flour using different processing conditions. DON degradation during baking was quantified with the most accurate analytical methodology available for this Fusarium toxin, which is based on liquid chromatography tandem mass spectrometry. Depending on the processing conditions, 0–21%, 4–16%, and 2–5% DON were degraded during the production of crackers, biscuits, and bread, respectively. A higher NaHCO3 concentration, baking time, and baking temperature caused higher DON degradation. NH4HCO3, yeast, vinegar, and sucrose concentration as well as leavening time did not enhance DON degradation. In vitro cell viability assays confirmed that the major degradation product isoDON is considerably less toxic than DON. This proves for the first time that large-scale industrial baking results in partial detoxification of DON, which can be enhanced by process management.

**Keywords:** mycotoxins; trichothecenes; thermal degradation; decontamination; mass spectrometry; food processing; detoxification; design of experiment; LC-MS/MS

**Key Contribution:** Deoxynivalenol (DON) can be partially detoxified during the industrial production of bakery products. The detoxification of DON can be enhanced by process management.

#### **1. Introduction**

Deoxynivalenol (DON) is along with aflatoxins, fumonisins, ochratoxin A, and zearalenone, one of the main mycotoxins of significance for human disease [1]. DON frequently contaminates cereals and cereal-based products [2]. The population of the European Union (EU) is frequently exposed to DON mainly due to the consumption of bread and other bakery wares [2]. Therefore, the European Commission (EC) has set maximum levels for DON in milling products and bread/bakery wares of 750 μg/kg and 500 μg/kg, respectively [3]. This legislation follows the rationale that it is feasible to reduce DON from flour during the production of bakery products by 33%. This assumption is supported by the recent report on the risk associated with DON in food by the European Food Safety Authority (EFSA) [2]. From 2007 to 2014, EFSA gathered occurrence data for the concentration of DON present in different food groups. For milling products, bread and rolls, and fine bakery wares, 4609, 2837, and 975 data points, respectively, were collected from the EU member states. The mean and 95th percentile concentrations of DON were 20–30% lower in bread and bakery wares (bread and rolls, fine bakery wares) than in flour (milling products).

During baking, DON was found to degrade mainly to isoDON, but also to norDON B and norDON C [4–6]. norDON B and norDON C were found to be considerably less cytotoxic than DON [5]. isoDON was found to have much lower inhibitory potency on the ribosome than DON, which is strongly indicative of its overall lower toxicity [6].

The reduction of the DON concentration during baking can be due to formulation (i.e., the dilution effect due to mixing contaminated flour with non-contaminated ingredients) or due to degradation of DON. The reduction due to formulation can be easily calculated from the percentage of flour of the finished bakery product, assuming no additional sources of DON. The extent of the DON degradation during baking, however, was reported to be very variable [2,7,8]. This high variation can be explained by errors in the calculation of the amount of DON that was degraded (e.g., by not considering the dilution with other ingredients) and by not performing baking trials with a suitable number of replicates. Recently, we identified all degradation products of DON, formed during baking of crackers, biscuits, and bread, using a stable isotope labelling approach combined with high resolution mass spectrometry analysis. After quantifying DON and all of its formed degradation products, we were able to provide the first comprehensive mass balance for those commodities [6].

In this current study, DON degradation during the production of crackers, biscuits, and bread from flour fortified with DON was determined to be 6%, 5%, and 2%, respectively, based on the increase of degradation products. As this study was aimed at the elucidation of the degradation products that are formed during baking, the experiments were carried out under standard baking conditions. As a step forward, the influence of the processing conditions should be evaluated. Furthermore, experiments with naturally contaminated flour need to be conducted in order to reproduce the actual industrial starting conditions in place during the large-scale production of bakery products.

The influence of different processing parameters on DON degradation during the production of crackers, biscuits, bread, and rusks was determined in previous studies [9–12]. Statistical evaluation of the DON concentration showed that baking temperature, baking time, and the concentration of the raising agent NaHCO3 influence the amount of DON in the final products. However, the statistical models as well as the determination of the DON degradation may have suffered from the high analytical variability of the determination of the DON concentration.

We hypothesized that DON degradation during industrial baking is influenced by the choice of processing parameter (i.e., ingredients, leavening, and baking conditions). Hence, the objectives of this study were to: (i) Produce bakery products (crackers, biscuits, and bread) from naturally contaminated flour using different processing conditions, (ii) determine DON degradation by LC-MS/MS based quantification of the DON degradation products and iii) develop a statistical model which shows the influence of the processing parameters on the DON degradation. Due to the comprehensive baking trials in combination with the highly sensitive LC-MS/MS analysis and the consideration of all formed degradation products, this study presents the most accurate report on the effect of industrial baking on DON degradation to date. Moreover, a first characterization of the cytotoxicity of isoDON, compared to DON, is presented.

#### **2. Results**

To study the main processing factors affecting the mitigation of DON during industrial baking, crackers, biscuits, and bread were chosen as representative commodities of the main bakery categories

produced with very different technologies. A Design of Experiments (DoE) approach was applied to set up the experimental trials. In agreement with results previously obtained by our research group [9,10,12], the main technological parameters monitored were: (i) Baking conditions (i.e., time and temperature), (ii) pH modifying agents (NaHCO3, NH4HCO3, and other minor ingredients) and (iii) leavening conditions (time, temperature, presence, and type of leavening agents). Each factor was modified within a range according to technological feasibility. Some conditions stressed the process over the organoleptic acceptability of the final product. These processing conditions were chosen to promote significant changes of the DON concentration in order to understand the impact of the applied processing conditions. In the present work, all the food commodities were produced at a pilot plant level. As a starting point of the DoE approach, the same processing parameters as those on common industrial processing lines were chosen. Figure 1 exemplifies the influence of the baking conditions on the final appearance of biscuits, bread and crackers.

**Figure 1.** Display of the influence of baking conditions on the appearance of biscuits (top), bread (middle) and crackers (bottom).

#### *2.1. Incluence of Processing Parameters on DON Degradation during the Production of Bakery Products*

For all three baking commodities, the results of the experimental trials were evaluated and presented in two sub-chapters. First, the sum of the DON degradation products, which is the measure for DON degradation, was shown as a bar graph. The individual experiments were listed according to the processing parameters that were found to be important for DON degradation based on a visual assessment. In this way, Figures 2, 4, and 6 show the DON degradation that was achieved under the specified processing conditions ranked from high to low DON degradation. These bar graphs give a first impression of the DON degradation and the influence of the processing parameter for the individual commodities. Second, to confirm and refine the conclusions of the visual assessment, a DoE model was built using the MODDE software tool. The outcome of this statistical analysis is

visualized by an effects plot which displays the change of the sum of the DON degradation products when a processing factor is varied from its lowest to its highest value (Figures 3, 5, and 7, top). The data set obtained from the DoE model was used to build a prediction model for the influence of the process parameter that led to a statistically significant increase of DON degradation. These predictive factor plots show how DON degradation can be enhanced by the design of the production process (Figures 3, 5, and 9, bottom). To allow for the easy comparability of the results between the three baking commodities, the same scaling was chosen for the individual graphs.

#### 2.1.1. Biscuits

Biscuits, from the technological point of view, represent a quite simple bakery preparation characterized by a baking step and the use of raising agents. In order to monitor the impact of these two factors on the DON content in the final products, a very basic recipe was chosen, containing only flour, oil, sugar, and water, with the addition of NaHCO3 and NH4HCO3 as leavening agents. Several batches of biscuits were prepared at different conditions according to the DoE inputs based on the Screening Interaction model. In the present study, the processing parameters NaHCO3 concentration, NH4HCO3 concentration, baking time, baking temperature, and sugar content were varied (Table S1). Due to the different production conditions, DON degradation varied between 4–16% (Figure 2).

**Figure 2.** Increase of the deoxynivalenol (DON) degradation products isoDON and norDONs A-C during the production of biscuits using different processing conditions. The increase was determined as a ratio of the molal concentration of the DON degradation products to the molal concentration of DON resulting from the natural contamination of the flour. The experimental trials were listed according to the NaHCO3 concentration, baking temperature, and baking time. Error bars represent the process standard deviation.

A higher DON degradation was observed in batches with a higher concentration of NaHCO3, baking temperature, and baking time. The variation of the processing parameter also led to a changed ratio of the degradation products with the main degradation products being norDON B and C, instead of isoDON, for experiments with a higher baking temperature and baking time. norDON A, which was not found to be a degradation product in our previous study [6], was found only at processing conditions that lead to higher DON degradation.

The DoE model confirmed that the main factors affecting DON degradation were: (i) The quantity of NaHCO3 used, (ii) the baking temperature, and (iii) the baking time (Figure 3, top). In addition, the statistical model indicates that a synergistic effect of baking temperature and time exists. Changes in NH4HCO3 and sucrose concentration did not significantly influence DON degradation products. The prediction model for the influence of the processing parameter on DON degradation revealed that by increasing the NaHCO3 concentration from 0.2 to 0.6%, DON degradation is increased by 5% (Figure 3, bottom). An increase in baking temperature and time is predicted to enhance DON degradation by about 2%.

**Figure 3.** Top: Effects plot, which shows the change of the sum of the DON degradation products when a processing factor is varied from its lowest to its highest value and all other factors are kept at their averages, which was obtained for the design of the experiment (DoE) data set in pilot-scale biscuit making experiments. Error bars represent the confidence interval corresponding to a 95% confidence level. "\*" between two processing factors indicates a synergistic effect, which cannot be explained solely by addition of degradation caused by the two parameters individually. Bottom: Predictive factor effect plots, which show the influence of the following processing parameters on the deoxynivalenol (DON) degradation during the biscuit production: (i) NaHCO3 (left), (ii) baking temperature (middle), and (iii) baking time (right). The dotted lines represent the confidence interval corresponding to a 95% confidence level.

#### 2.1.2. Bread

Bread is surely the most representative and most studied commodity of soft bakery products. Its production workflow implies dough fermentation before baking. Both fermentation and subsequent baking can be managed for time and temperature. The fermentation phase can also be modulated

for relative humidity conditions, yeast, bakery improvers, raising agents, and pH. As a consequence, the number of factors to investigate for assessing potential DON mitigation effects was quite high. Therefore, in comparison to biscuits and crackers, more pilot plant trials had to be carried out. Based on literature data and our experience, the following variables were taken into account for the DoE approach: Baking temperature, baking time, leavening time, yeast, sucrose, and cider vinegar concentration. Each factor was varied within a range defined according to technological feasibility (Table S4). Depending on the processing parameter, DON degradation varied from 2–5% (Figure 4).

process standard deviation.

**Figure 4.** Increase of the deoxynivalenol (DON) degradation products isoDON and norDONs B-C during the production of bread using different processing conditions. The increase was determined as a ratio of the molal concentration of the DON degradation products to the molal concentration of DON resulting from the natural contamination of the flour. The experimental trials were listed according to baking temperature, baking time, and sucrose concentration. Error bars represent the

During the bread production, the achievable DON degradation was found to be low compared to the biscuit and cracker production. Although the influence was small, we found that the baking time and temperature as well as the sucrose concentration impacted the DON concentrations. The DoE model confirmed that the main factors influencing DON degradation were: (i) Baking time, (ii) baking temperature, and (iii) sucrose concentration (Figure 5, top). Changes in cider vinegar and yeast concentration as well as the leavening time did not influence DON degradation. The prediction model confirmed that changes in the processing parameter have only a minor effect of approximately 2% on DON degradation (Figure 5, bottom).

**Figure 5.** Top: Effects plot, which shows the change of the sum of the DON degradation products when a processing factor is varied from its lowest to its highest value and all other factors are kept at their averages, which was obtained for the design of the experiment (DoE) data set in pilot-scale bread making experiments. Error bars represent the confidence interval corresponding to a 95% confidence level. Bottom: Predictive factor effect plots, which show the influence of the following processing parameters on the deoxynivalenol (DON) degradation during the biscuit production: (i) Baking time (left), (ii) baking temperature (middle), and (iii) sucrose concentration (right). The dotted lines represent the confidence interval corresponding to a 95% confidence level.

#### 2.1.3. Crackers

Crackers were chosen as being representative of a complex baking commodity involving different technological aspects and combining several factors which were demonstrated to affect DON mitigation in the previous experiments carried out on biscuits and bread. Among the cracker making parameters, the following conditions were considered as factors in the experimental design: Baking time, baking temperature, acidic mother content, and NaHCO3 concentration. For this trial, each factor was varied

within a range defined according to technological feasibility (Table S7). During the cracker production, the DON degradation varied from 0–21%, depending on the choice of processing parameters (Figure 6).

**Figure 6.** Increase of the deoxynivalenol (DON) degradation products isoDON and norDONs A-C during the production of crackers using different processing conditions. The increase was determined as a ratio of the molal concentration of the DON degradation products to the molal concentration of DON resulting from the natural contamination of the flour. The experimental trials were listed according to the NaHCO3 concentration, baking time, and baking temperature. Error bars represent the process standard deviation.

A high DON degradation was observed in experiments with a high NaHCO3 concentration, baking time, and baking temperature. Similar to the biscuit production, the variation of the processing parameter caused a change in the ratio of the degradation products. The DoE model revealed that NaHCO3 concentration and baking time were the main factors affecting DON degradation (Figure 7, top). In addition, the statistical model indicated that a synergistic effect of baking time and NaHCO3 concentration exists. The acidic mother concentration, leavening time, and baking temperature did not affect DON degradation. The prediction model revealed the potential DON reduction that can be achieved by process management (Figure 7, bottom). An increase of the NaHCO3 concentration from 0 to 1% is predicted to increase DON degradation by 10%. Similarly, the increase of baking time from 1 to 6 min is predicted to increase DON degradation by 10%.

**Figure 7.** Top: Effects plot, which shows the change of the sum of the DON degradation products when a processing factor is varied from its lowest to its highest value and all other factors are kept at their averages, which was obtained for the design of the experiment (DoE) data set in pilot-scale bread making experiments. Error bars represent the confidence interval corresponding to a 95% confidence level. "\*" between two processing factors indicates a synergistic effect, which cannot be explained solely by addition of degradation caused by the two parameters individually. Bottom: Predictive factor effect plots, which show the influence of the following processing parameters on the deoxynivalenol (DON) degradation during the biscuit production: (i) NaHCO3 concentration (left) and (ii) baking time (right). The dotted lines represent the confidence interval corresponding to a 95% confidence level.

#### *2.2. Cytotoxic E*ff*ects of isoDON*

The cytotoxic potential of isoDON in direct comparison to DON was evaluated in two different human colon cell lines (Figure 8).

**Figure 8.** Cell viability of human colorectal adenocarcinoma cells (HT-29) (left) and non-tumorigenic human colon epithelial cells (HCEC) (right) after 24 h of incubation with different concentrations of deoxynivalenol (DON) and isoDON. The concentration at which cell viability was reduced by 30% (IC30) was calculated from a dose response curve fitted to the individual data points (×).

Usually, the half maximal inhibitory concentration (IC50) is used to compare the potency of two substances. As the highest tested isoDON concentrations did not reduce cell viability by 50% under all tested conditions, IC30 values where cell viability was reduced by 30% were used instead. For the human colorectal adenocarcinoma cell line HT-29 the following IC30 values were determined: 1.9 μM (DON) and 91 μM (isoDON), whereas for the non-tumorigenic human colon epithelial cells (HCEC) the following IC30 values were identified: 0.4 μM (DON) and 53 μM (isoDON). In both cell viability assays, isoDON was much less potent (factor 48 and 133 for HT-29 and HCEC, respectively) in reducing cellular viability compared to DON.

#### **3. Discussion**

#### *3.1. Influence of Di*ff*erent Processing Parameters on DON Degradation in Bakery Products*

#### 3.1.1. The pH Value of the Dough

As DON has been reported to be unstable in alkaline solutions, the pH of the dough is clearly an important processing factor to consider when designing a production process that is optimized for a high DON degradation [13]. For the production of biscuits and crackers, the pH of the dough was regulated primarily by chemical raising agents. We found that the type of the chemical raising agent as well as its concentration was crucial regarding DON degradation. Whereas the use of NaHCO3 led to higher DON degradation, the use of NH4HCO3 did not influence DON concentration. This can be explained by the different pH values of the dough resulting from the different chemical nature of the raising agents. Gökmen et al. prepared cookies from two different doughs containing 0.7% NaHCO3 and NH4HCO3, respectively [14]. When NaHCO3 was used, the pH was initially 8.5 and increased slightly during baking to pH 9. In the absence of acidic compounds, NaHCO3 is thermally converted to Na2CO3, H2O and CO2 (Equation (1)). As CO3 <sup>2</sup><sup>−</sup> is a stronger base than HCO3 −, the thermal degradation of NaHCO3 leads to an increase of the ph. The use of NH4HCO3 as a leavening agent led to a decrease of the pH from 8 in the dough to 6–7 in the baked cookies. The decrease of pH can be explained by the degradation of NH4HCO3 to NH3, CO2 and H2O, which evaporate during baking (Equation (2)).

$$2\ \text{NaHCO}\_3 \rightarrow \text{Na}\_2\text{CO}\_3 + \text{H}\_2\text{O} + \text{CO}\_2\tag{1}$$

$$\text{NH}\_4\text{HCO}\_3 \rightarrow \text{NH}\_3 + \text{H}\_2\text{O} + \text{CO}\_2 \tag{2}$$

Besides chemical raising agents, further ingredients were shown to affect the pH of the dough. High pH values due to the presence of NaHCO3 were only observed when sucrose was used as sugar. At a higher pH, sucrose was not hydrolyzed and thereby did not cause a change in pH [14]. When glucose was used, it was partially hydrolyzed to fructose resulting in a decrease of the pH from 8.5 in the dough to 6–7 in the baked cookies. This can be rationalized by the report of Feng et al. who calculated the pKa values for glucose and fructose [15]. Fructose was shown to be considerably more acidic (pKa 12) compared to glucose (pKa 14). In the presence of an acid, NaHCO3 was converted to CO2 which caused a decrease of the pH (Equation (3)).

$$\text{NaHCO}\_3 + \text{H}^+ \rightarrow \text{Na}^+ + \text{H}\_2\text{O} + \text{CO}\_2 \tag{3}$$

For the biscuit production, our results confirm the outcome of a previous study, where the NaHCO3 concentration was the main factor that enhanced DON degradation [9]. For the cracker production, our results contradict a previous report in which baking temperature and time were found to be of higher importance compared to the NaHCO3 concentration. However, in the previous study, fewer experimental trials were carried out and the accuracy of the analytical methodology was lower.

#### 3.1.2. Baking Conditions

For all three commodities, the baking conditions (i.e., temperature and/or time) influenced DON degradation. However, for the production of biscuits and crackers, baking conditions were less important compared to the pH of the dough. For the production of crackers and bread, baking time was found to be more important than baking temperature. The contrary was observed for biscuit production. The reason for this observation might relate to differences in moisture content and surface to volume ratio of the individual baking commodities.

#### 3.1.3. Ratio of the DON Degradation Products

For the production of biscuits and crackers, the ratio of the DON degradation products shifted from isoDON to norDON B and C in the experiments that led to high DON degradation. This supports the proposed mechanism, that isoDON is further converted to norDON B and C [13]. Some baking conditions led to the formation of norDON A. Most likely, the reason why norDON A was not detected in a previous study was a higher limit of detection (LOD) and limit of quantification (LOQ) for norDON A compared to isoDON, norDON B and C in the analytical method used [6]. Interestingly, the formation of norDON A was not observed in the experiments with the highest DON degradation (i.e., biscuits: N4, N16; crackers: N11, N12, N15, N16) in total. This might have been due to an initial formation of isoDON, which was further degraded into norDON A and finally to norDON B or norDON C.

#### *3.2. Toxicity of the DON Degradation Products*

For a comprehensive assessment of the health risks of eating bakery products produced from DON contaminated flour, the toxicity of all degradation products has to be determined. So far, norDONs A–C were shown to be considerably less cytotoxic than DON [5]. In comparison to DON, isoDON was found in in vitro translation assays to have a considerably lower potency (factor 94 and 60 in wheat and rabbit ribosomes, respectively) to inhibit translation, known to be the major mechanism of toxicity of trichothecenes [6]. As the 60S subunit binding site of ribosomes is the main molecular target of DON, the results of the translation assay were indicative of considerably lower cytotoxicity [16,17]. However, various aspects, such as different cellular absorption, changes in distribution, and especially differences in fitting into the pockets of the ribosomal A-site due to changes in polarity might lead to significant differences in cytotoxicity of isoDON and DON. Therefore, the cytotoxic effects of isoDON in comparison to DON were evaluated.

#### Cytotoxic Effects of isoDON

Performing cell viability assays in two different human colon cell lines, we found that isoDON induced significantly less cytotoxicity compared to DON in HCEC and HT-29 cells. In Figure 9, the structures of DON and isoDON, showing structural differences at the carbon atoms C-7 to C-10 of the trichothecene backbone, are displayed.

**Figure 9.** Structure of iso-deoxynivalenol (isoDON) (left) and DON (right). The C-7 to C-10 atoms of the trichothecene backbone are highlighted.

Our findings are supported by structure activity relationships which have been conducted for various trichothecenes [18–20]. The absence of a double bond at the C-9 and C-10 position, which is the case for isoDON, led to a significant decrease in both inhibition of protein biosynthesis and cytotoxicity. Although the toxicity of DON is often ascribed to the epoxide moiety, we could provide additional proof that a double bond between C-9 and C-10 is an essential structural feature for ribosomal inhibition and thus, for trichothecene-induced toxicity.

#### **4. Conclusions**

Depending on the processing conditions, 0–21%, 4–16%, and 2–5% of DON were degraded during the production of crackers, biscuits, and bread, respectively. For biscuits and crackers, DON degradation can be increased by 10–20% by process management. For bread, process management can lead to a minor increase in DON degradation of about 3%. DON degradation was enhanced by a high NaHCO3 concentration, baking time, and temperature. The processing parameters NH4HCO3, yeast, cider vinegar, acidic mother content, and sucrose concentration as well as leavening time did not influence DON concentration.

To put our findings in context to the results reported in recent literature, a graphical summary is presented in Figure 10.

**Figure 10.** Comparison between the deoxynivalenol (DON) degradation reported in recent literature studies [10–12,21–29] and the DON degradation values determined in this study under different processing conditions. Each dot represents one baking experiment.

Compared to previous studies found in the literature, we achieved a much narrower range of DON degradation, which was most pronounced for bread. The reason for that is most likely that we applied the currently most accurate analytical methodology to determine DON degradation [6].

To gain further knowledge about the toxicity of DON degradation products, the cytotoxicity of isoDON was evaluated by performing in vitro cell viability assays. isoDON was found to be considerably less cytotoxic (factor 48 and 133 for HT-29 and HCEC cells, respectively) compared to DON. As all DON degradation products are considerably less cytotoxic than DON, we conclude that DON degradation during baking results in a lower toxicity. Thus, for the first time, we have presented proof that baking under real industrial conditions can lead to a partial detoxification of DON.

#### **5. Materials and Methods**

#### *5.1. Chemicals and Reagents*

Acetonitrile (ACN, gradient grade) was purchased from VWR International GmbH (Vienna, Austria). Acetic acid (LC–MS gradient grade) was obtained from Sigma Aldrich (Vienna, Austria). In all experiments, ultra-pure water (purified by a Purelab Ultra system ELGA LabWater, Celle, Germany) was used. Liquid calibrant solutions of DOM-1, DON-3-Glc, DON, and U-[13C15]-DON were supplied by Romer Labs GmbH (Tulln, Austria). Reference standards of the DON degradation products isoDON and norDONs A, B and C were synthesized according to published procedures [5,6,21,30].

#### *5.2. Preparation of Bakery Products*

The food commodities used in the present study were produced from naturally contaminated flour according to the procedure previously described by our group by applying a suitable scaling up factor of the starting ingredients that fits with the production under pilot plant facilities [6].

Biscuits: The final dough (3000 g, 14% moisture content (MC)) was used to form the individual biscuits (approximately 11 g each) which consisted of 56 wt % flour, 40 wt % fat and sugar, and 4 wt % water.

Bread: The final dough (8000 g, 14% MC) was used to form the individual loafs of 450 g each (40% MC), that reached a final weight of 400 g and consisted of 65 wt % flour, yeast, oil, and salt and 35 wt % water after baking.

Crackers were shaped (approximately 10 g each) from the final dough (4127 g, 26% MC) and they consisted of 97 wt % flour, yeast, and malt extract and 3 wt % water.

#### *5.3. Design of Experiment (DoE) Setup and Statistical Evaluation*

Design of Experiments (DoE) was used to determine the influence of processing parameters on the degradation of DON and the increase of its degradation products during the baking of crackers, biscuits, and bread from naturally contaminated flour. The screening interaction model was chosen and processing parameters that we hypothesized to influence DON degradation were used as model quantitative factors. Table 1 provides a list of the processing parameters that were investigated in our study. For more detailed information on the processing parameters used in the individual experiments, the reader is referred to Tables S1, S4, and S7.

To focus the attention on the increase of DON degradation products during baking, multivariate data analysis was performed on the concentration (μg/kg) obtained by LC/MS-MS analysis in the finished products of the sum of DON degradation products. DoE set up and data elaboration were carried out by using MODDE software 11 (Umetrics, Umea, Sweden).


**Table 1.** Processing parameters that were investigated for their impact on the degradation of deoxynivalenol during the production of bakery products from naturally contaminated flour under pilot plant conditions.

#### *5.4. Analysis of the Samples*

To determine DON degradation during the industrial baking process, the concentrations of DON and its degradation products (i.e., isoDON, norDON B, and norDON C) were analyzed in naturally contaminated flour and the products made thereof. Additionally, the concentration of the suspected DON degradation and conversion products DON-3-glucoside, DOM-1 and norDON A was monitored. Sample preparation and analysis are discussed in detail elsewhere [6]. In brief, 5.00 ± 0.01 g of ground and homogenized sample was extracted with 20.0 mL of ACN:H2O, 84:16, *v*:*v*. A total of 200 μL of sample extract were dried down and reconstituted in 100 μL of water. Subsequently, the concentrated sample extracts were centrifuged at 10 ◦C and 4500 rpm for 10 min. Four μL aliquots of the concentrated sample extract were injected together with 0.4 μL of internal standard into the HPLC system. Each sample was worked-up once and injected in duplicate. The analytical method used for the analysis was based on LC-MS/MS and was validated for recovery, repeatability, LOD, and LOQ. All results were corrected for the recovery factor determined in the validation study [6]. The concentration of the analytes that was found in naturally contaminated flour and in the finished bakery products can be found in Table S2 and S3 (biscuits), Table S5 and S6 (bread), and Table S8 and S9 (crackers) of the Supplementary Material, respectively.

#### *5.5. Calculation of the DON Degradation*

The sum of the DON degradation products was shown to be a much more accurate measure of the DON degradation than the change of the DON content itself [6]. Therefore, the DON degradation was calculated based on the increase of its degradation products isoDON and norDONs A–C. As the flour used for the production of the bakery products was naturally contaminated with the degradation products at different concentration levels, the increase of the degradation products was calculated separately before being summed up to obtain the final DON degradation.

#### *5.6. Calculation of the Increase of the DON Degradation Products Due to Baking*

For the production of bakery products from naturally contaminated flour, the increase of the degradation products was calculated according to:

$$\text{dilution factor} = \frac{m\_{flow}}{m\_{\text{bakery product}}}$$

$$b\_{\text{analyte,dilution of nat.out. flow}} = b\_{\text{analyte, flourr}} \ast \text{dilution factor}$$

$$\text{Change due to baking} = \frac{b\_{\text{analyte, bakary product}} - b\_{\text{analyte, lilution of nat.out. flow}}}{b\_{\text{DON, dilution of nat.out. flow}}}$$

First, the dilution factor, which describes the mass ratio of naturally contaminated flour present in the finished bakery product, was determined. The dilution factor was 0.97 (crackers), 0.56 (biscuits), and 0.65 (bread), respectively. Second, the theoretical molal concentration b (mol/kg) of the analyte resulting from the dilution of the flour by the addition of other ingredients (*banalyte*,*dilution o f nat*.*cont*. *flour*) was determined. Finally, the molal concentration of an analyte resulting from the natural contamination of the flour was subtracted from the molal concentration of the analyte in the bakery product (*banalyte*, *bakery product*) and expressed as a fraction of the molal concentration of DON resulting from the natural contamination of the flour. The result was the fraction of DON that is converted to DON degradation products during the production of bakery products.

#### *5.7. Estimation of the Process Standard Deviation*

In each DoE, three center points were set, which means that three experiments were carried out under identical processing conditions: Experiment number 17, 18 and 19 (crackers), 17, 18 and 19 (biscuits) and 33, 34, 35 (bread), respectively. The process standard deviation was calculated for the sum of the change of the degradation products of the three replicates which were produced using the same processing parameter. The process standard deviation was 0.9%, 1.4%, and 0.8% for crackers, biscuits, and bread, respectively, and thus considerably higher than the analytical standard deviation for the determination of the sum of the DON degradation products which was in the range of 0.2–0.4% [6].

#### *5.8. Human Cell Culture*

Human colorectal adenocarcinoma cells, HT-29, were purchased from the German Collection of Microorganisms and Cell Cultures (DSMZ, Braunschweig, Germany). For cell culture Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 10% (*v*/*v*) heat inactivated fetal calf serum and 50 U/mL penicillin and 50 μg/mL streptomycin was used. The non-tumorigenic human colon epithelial cell line, HCEC [31], was kindly provided by Prof. Jerry W. Shay (UT Southwestern Medical Center, Dallas, TX, USA). HCEC cells were cultivated in DMEM (high glucose) supplemented with 2% (*v*/*v*) Medium 199 (10×), 2% (*v*/*v*) cosmic calf serum, 20 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), 50 μg/mL gentamicin, insulin-transferrin-selenium-G (10 μg/mL; 5.5 μg/mL; 6.7 μg/mL), recombinant human epidermal growth factor (18.66 ng/mL), and hydrocortisone (1 μg/mL). Both cell lines were cultivated in humidified incubators at 37 ◦C and 5% CO2. Media, supplements, and further material for cell culture were purchased from GIBCO Invitrogen (Karlsruhe, Germany), Lonza Group Ltd. (Basel, Switzerland), Sigma-Aldrich Chemie GmbH (Munich, Germany), Sarstedt AG & Co. (Nuembrecht, Germany), and Fisher Scientific GmbH (Vienna, Austria).

#### *5.9. Cytotoxicity of isoDON*

Per well, 1500 HCEC or 5500 HT-29 cells were seeded in a 200 μL culture medium in 96-well plates and allowed to grow for 48 h. Then, the culture medium was withdrawn and cells were treated with different concentrations of isoDON and DON in the respective cell culture medium for 24 h. As a solvent control, 1% water (LC–MS grade) was incubated and 1% (*v*/*v*) Triton X-100 served as a positive control. After the treatment period, the medium was withdrawn and cells were rinsed once with phosphate buffered saline (100 μL/well) and incubated with DMEM containing 10% (*v*/*v*) alamarBlue® reagent (Invitrogen™ Life Technologies, Karlsruhe, Germany) for 75 min. Subsequently, fluorescence intensity was measured (excitation: 530 nm; emission: 600 nm) on a Cytation 3 Imaging Multi Mode Reader (BioTek, Bad Friedrichshall, Germany). For quantification, blank-values were subtracted and measured data were referred to solvent control. Data are presented as a mean of at least five independent experiments, each performed in technical triplicates.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6651/11/6/317/s1, Table S1: Processing parameters used for the production of biscuits in different experiments (Exp. No.). Ingredients are given as the weight percentage of the dough; Table S2: Concentration of the analytes in the naturally contaminated flour, dilution factor by which the flour gets diluted in the final biscuits, and the resulting

concentration (assuming 0% degradation). <LOQ: concentration was below the limit of quantification (LOQ) of the analytical methodology. Table S3: Concentration of deoxynivalenol (DON) and related compounds in the biscuit samples. <LOQ: concentration was below the limit of quantification (LOQ) of the analytical methodology; Table S4: Processing parameters used for the production of bread in different experiments (Exp. No.). Ingredients are given as the weight percentage of the dough; Table S5: Concentration of the analytes in the naturally contaminated flour, dilution factor by which the flour gets diluted in the final bread, and the resulting concentration (assuming 0% degradation). n.d. (not detected): Concentration was below the limit of quantification of the analytical methodology; Table S6: Concentration of deoxynivalenol (DON) and related compounds in bread samples. n.d. (not detected): Concentration was below the limit of quantification of the analytical methodology; Table S7: Processing parameters used for the production of crackers in different experiments (Exp. No.). Ingredients are given as the weight percentage of the dough; Table S8: Concentration of the analytes in the naturally contaminated flour, dilution factor by which the flour gets diluted in the final crackers, and the resulting concentration (assuming 0% degradation). n.d. (not detected): Concentration was below the limit of quantification of the analytical methodology; Table S9: Concentration of deoxynivalenol (DON) and related compounds in cracker samples. n.d. (not detected): Concentration was below the limit of quantification of the analytical methodology.

**Author Contributions:** Conceptualization, D.S., F.L., M.S., F.B. and R.K.; methodology, D.M., M.S. and R.K.; data curation, D.S., F.L. and L.W.; toxicity study, L.W and D.M.; visualization, D.S. and F.L.; writing-original draft preparation, D.S., F.L. and L.W.; writing—review and editing, all authors; supervision, D.M., M.S., F.B. and R.K.; project administration and funding acquisition, D.M., M.S. and R.K.

**Funding:** This project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No. 678012 for the MyToolBox project, also covering the APC. We also acknowledge financial support from the Austrian Science Fund (FWF) via the special research project Fusarium (F3718).

**Acknowledgments:** The authors would like to thank the colleagues for the pilot plant trials execution support: Dante Catellani, Marta Filomeno, Gennaro Cangiano, Sergio Cerasti and Umberto Tanzi.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **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* **Fate of Ergot Alkaloids during Laboratory Scale Durum Processing and Pasta Production**

#### **Sheryl A. Tittlemier \*, Dainna Drul, Mike Roscoe, Dave Turnock, Dale Taylor and Bin Xiao Fu**

Grain Research Laboratory, Canadian Grain Commission, 1404-303 Main Street, Winnipeg, MB R3C 3G8, Canada; dainna.drul@grainscanada.gc.ca (D.D.); mike.roscoe@grainscanada.gc.ca (M.R.); dave.turnock@grainscanada.gc.ca (D.T.); dale.taylor@grainscanada.gc.ca (D.T.); binxiao.fu@grainscanada.gc.ca (B.X.F.)

**\*** Correspondence: sheryl.tittlemier@grainscanada.gc.ca; Tel.: +1-204-984-3456

Received: 6 March 2019; Accepted: 29 March 2019; Published: 31 March 2019

**Abstract:** The fate of ergot alkaloids during the milling of durum and subsequent production and cooking of pasta was examined. Durum samples containing varying amounts of ergot sclerotia (0.01–0.1% by mass) were milled, and all milling product was analyzed for 10 ergot alkaloids using liquid chromatography with tandem mass spectrometry. Spaghetti was prepared from the semolina obtained during milling. Ergocristine, ergocristinine, and ergotamine were the predominant ergot alkaloids observed in the milling fractions and spaghetti. Approximately 84% of the total ergot alkaloid mass of the whole grain durum resided in the milling product fractions associated with the outer kernel layers (bran, shorts, feeds). No consistent loss of ergot alkaloids was observed during the production or cooking of spaghetti. However, changes in the ratio of *R*- to *S*-enantiomers occurred during the milling and cooking of spaghetti. Products containing bran, shorts, and feeds, as well as cooked spaghetti, contained a higher proportion of the less biologically active *S*-enantiomers. The results of this study emphasize the need to monitor *R*- and *S*-enantiomers, and to consider food and feed products, as opposed to whole grain, when assessing any exposure of consumers to ergot alkaloids.

**Keywords:** mycotoxin; milling; bran; semolina; cooking; dietary exposure

**Key Contribution:** No substantial losses of ergot alkaloids were associated with boiling pasta. However, cooking pasta resulted in epimerization and a shift towards higher concentrations of the less biologically active *S*-enantiomers.

#### **1. Introduction**

Ergot alkaloids (EAs) are a group of mycotoxins produced during the fungal infection of cereals. In Canada, the fungus *Claviceps purpurea* causes ergot infection of rye, barley, oats, wheat, and durum [1]. During *C. purpurea* infection, healthy kernels are replaced by dark-colored sclerotia that contain high concentrations of various EAs [2–4]. The incidence of ergot infection of durum grown in Canada is variable amongst growing years, but appears to be increasing since the early 2000s [5]. This increase suggests that the potential for EAs occurring in durum, and durum-based food products, is also increasing. A number of EAs have been measured in Canadian cereals [5], as well as those grown in western [6] and eastern Europe [7], and Australia [3].

The EAs most commonly associated with cereal grains are amide-like derivatives of lysergic acid [8]. These compounds contain a stereogenic center, and can exist in *R*- (indicated by an "ine" suffix in their names) and *S*- (indicated by an "inine" suffix in their names) enantiomeric forms. The enantiomers can undergo reversible epimerization, which can reach equilibrium. Epimerization has been reported to occur in solution under alkaline conditions and lengthy periods of storage [9].

Consumption of ergot alkaloids has been long known to cause ergotism in both humans and livestock [10,11]. Symptoms of ergotism include gangrene, gastrointestinal effects, reduced lactation, as well as effects on the central nervous system.

Current practices to avoid or minimize exposure to ergot alkaloids in food and feed involve grain handling and milling procedures to exclude ergot sclerotia from entering food and feed channels. Tolerances are also used to regulate the amount of ergot sclerotia in grain. For example, the Canadian maximum level for ergot sclerotia in durum is 0.02% on a mass basis for the higher quality grades of No.1 Canada Western Amber Durum (CWAD) and No.2 CWAD [12], and the Codex Alimentarius Commission maximum level of ergot sclerotia in wheat is 0.05% [13].

Currently, no jurisdictions have set maximum levels (MLs) for ergot alkaloids in grain, food, or feed. However, regulatory agencies are exploring the use of MLs as a tool to manage health risk from exposure to EAs. The Canadian Food Inspection Agency (CFIA) consulted stakeholders on proposed limits for a number of contaminants, including ergot alkaloids, in livestock feeds [14]. The complexity of EA exposure and hazard assessments due to processing effects on EA concentrations, toxicity of individual EAs, and synergistic or antagonistic effects in EA mixtures, as well as variable susceptibilities due to age, sex, and physiological state, coupled with limited published scientific information in these areas, led the CFIA to revisit setting MLs for EAs in the future [15].

There are limited published studies on the fate of ergot alkaloids in food or feed products during the processing of grain. Most of the studies focus on rye-based food and feed. Fajardo et al. [16] investigated the fate of six EAs in milling products of red spring wheat. Flour containing EAs was also used to prepare pasta, Asian noodles, and bread. Overall, EAs were concentrated in the milling products associated with reduction streams, and some losses were noted in cooked noodles. However, a limitation of this work is the inclusion of only *R*-enantiomers in the EA analyses. Merkel et al. [17] studied the fate of twelve EAs in cookies baked with rye flour and subjected to in vitro digestion. The authors reported degradation of EAs during baking, as well as a shift from *R*- to *S*-enantiomers for all EAs. A feeding study conducted by Dänicke [18] included an assessment of the fate of twelve EAs in rye-based feed for laying hens, which had been processed under heat and pressure. Substantial epimerization of EAs was noted after processing, with an increase in concentration of *S*-enantiomers for all EAs.

The objective of this work was to take advantage of ongoing pasta quality and functionality studies, and examine the fate of EAs in milled durum, as well as during the processing and cooking of spaghetti. This study used durum, a cereal grain important in the production of pasta. The fate of EAs during the processing of durum has not been well studied. This study also monitored the fate of four *R*/*S* enantiomeric pairs, as well as two additional EA *R*-enantiomers for which the *S*-enantiomer was not readily available at the time of the study. The outcomes of this study are valuable for use in exposure assessments to ensure that consumers' exposure to EAs in food, or feed, is accurately determined.

#### **2. Results**

#### *2.1. Ergot Alkaloids in Whole Grain and Milling Products*

Table 1 lists the EA concentrations measured in the comminuted whole grain and milling products for the six samples of CWAD containing varying amounts of ergot sclerotia. Concentrations are provided as the sum of 10 ergot alkaloid analyte concentrations. Milling yields from the two mill runs of 2 kg each are also listed in Table 1.

Total EA concentrations for whole grain and all milling products increased as the percentage of ergot sclerotia in the original durum sample increased. The correlation between the percentage of ergot sclerotia and total EA concentration was statistically significant for whole grain and all milling products (Pearson Product Moment Correlation, *p* < 0.0025), aside from bran (*p* < 0.111). The total EA concentration of bran increased from 1112 to 5695 μg/kg as the ergot sclerotia content increased from 0.01% to 0.1%, and reached a maximum of 5715 μg/kg at an ergot sclerotia content of 0.04%.


**Table 1.** Total ergot alkaloids concentration (μg/kg) in whole grain durum and milling products. Average ± standard deviation (2 × 2 kg mill runs for each of the 6 samples) milling yields are provided for the milling products.

EAs were observed in all fractions from the five samples that contained ergot sclerotia. Only feeds and the first semolina fraction from the CWAD sample with no visible ergot sclerotia contained measurable EAs. Total EA concentrations in these fractions were low, ranging from 12 to 16 μg/kg. Total EA concentrations from the other samples were highest in bran, shorts, and feeds and the lowest was in the first semolina fraction. The concentrations in this semolina fraction were approximately 20–40× lower than the maximum measured in bran, shorts, or feeds.

Ergocristine and ergocristinine, followed by ergotamine, were the most predominant EAs observed in whole grain and milling products. The mean percentage of total EAs on a molar (as opposed to mass-based concentration) basis for each EA in the five samples containing ergot sclerotia was determined; these are presented in Figure 1.

**Figure 1.** Mean ± standard deviation percentage of total moles of ergot alkaloids present in whole grain and milling products. Means and standard deviations were calculated from whole grain and milling products from the five durum samples with ergot sclerotia ranging from 0.01–0.1% by mass.

In bran, shorts, and feeds, ergocristinine was the predominant EA at 29–35% of total EAs. The molar percentage of ergocristinine in bran and shorts were significantly different from those of the semolina and flour fractions (*p* < 0.05). Ergocristine was predominant in whole grain and other milling products, ranging from 37–55% of total EAs. The mean molar percentage of ergocristine was significantly different between semolina 1 and bran, shorts, and feeds (*p* < 0.05).

No other significant differences were noted amongst the other milling products. Ergotamine ranged from 10–14% of total EAs in whole grain and all milling products.

#### *2.2. Ergot Alkaloids in Spaghetti*

Table 2 lists the mean recoveries of individual EAs from fortified boiled spaghetti and cooking water. Mean recoveries for all EAs ranged from 93% to 113% from cooking water. Mean recoveries from boiled spaghetti were assessed only for the *R* enantiomers in order to preserve standard material, and because no difference in recoveries was observed for cooking water and whole grain [5]. Mean recoveries of the *R* enantiomer EAs from fortified boiled spaghetti ranged from 98% to 143%.

**Table 2.** Mean ± standard deviation percent recovery of ergot alkaloids from cooking water and cooked spaghetti. Cooking water samples were fortified with ergot alkaloids to produce a concentration of 40 μg/kg for ergonovine, 200 μg/kg for the other *R*-enantiomers and 100 μg/kg for the *S*-enantiomers. Boiled spaghetti samples were fortified to produce a concentration of 40 μg/kg for ergonovine and 200 μg/kg for the other *R*-enantiomers; spaghetti was not fortified with *S*-enantiomers. Triplicate replicates of cooking water and boiled spaghetti were analyzed for *R*-isomers. Duplicate replicates of cooking water were analyzed for *S*-isomers.


Table 3 lists concentrations measured in freshly extruded spaghetti, cooked spaghetti, and cooking water from the six samples of CWAD containing varying amounts of ergot sclerotia. The moisture contents of the spaghetti are also provided. The concentrations provided in Table 3 are on a fresh weight basis; that is, they are not normalized to a specific moisture content. The moisture content of cooked spaghetti was approximately twice that of the freshly extracted spaghetti.


**Table 3.** Mean ± standard deviation total ergot alkaloids concentration (μg/kg) in freshly extruded spaghetti, cooked spaghetti, and remaining cooking water.

Similar to whole grain, semolina, and flour fractions, ergocristine, ergocristinine, and ergotamine were the most predominant EAs observed in freshly extruded spaghetti. The mean percentage and standard deviation of total EAs on a molar basis for these three EAs in the five samples containing ergot sclerotia were 42 ± 8% for ergocristine, 19 ± 3% for ergocristinine, and 15 ± 5% for ergotamine. For cooked spaghetti and cooking water, ergocristinine was the predominant EA at 55 ± 8% and 73 ± 28% of the total EA molar content, respectively.

#### **3. Discussion**

#### *3.1. Fate of Ergot Alkaloids in Milling Products and Spaghetti*

Most of the EAs associated with sclerotia were in the by-products of semolina milling, i.e., bran, shorts, and feeds from outer kernel layers. Higher amounts of EAs were associated with bran, shorts, feeds (Figure 2), which are milling products containing material from the outer kernel. These three milling products contained 84% of the total EA mass in the durum.

**Figure 2.** Distribution of total ergot alkaloids mass amongst durum milling products. Pie slices represent the mean fraction of total ergot alkaloid mass across the five durum samples with ergot sclerotia ranging from 0.01–0.1% by mass.

The presence of the EAs predominantly in these milling fractions is consistent with the observation by Franzmann et al. [19] that the amounts of EAs in rye flour increased with a higher amount of bran in the flour. Franzmann et al.'s work attributed this EA distribution on rye kernels to the coincidental contact and abrasion between sound rye and ergot sclerotia during ordinary grain handling and movement.

Even though EAs will be transferred to sound kernels from ergot sclerotia during movement, the fate of ergot sclerotia during the milling process will be the most important factor affecting EA content of milling products because the concentration of EAs in ergot sclerotia are orders of magnitude greater than in sound grain. Total EA concentrations in rye ergot sclerotia were approximately 300× greater than concentrations in rye that had been mixed with ground ergot sclerotia and subject to cleaning [19]. Mean concentrations of the 10 EAs included in the current study ranged from 500 to 1000 mg/kg in ergot sclerotia obtained from infected durum plants [4].

The predominance of EAs in the bran, shorts, and feeds milling products fractions is also consistent with the path taken by ergot sclerotia through a wheat milling procedure reported by Farjado et al. [16]. The aforementioned Farjado et al. [16] milled wheat containing various amounts of ergot sclerotia. They noted that the sclerotia congregated in the reduction system because they did not flake when passing through the break rolls. This led to approximately 75% of EAs present in the whole grain wheat residing in bran and shorts milling fractions in their study.

The fate of the individual EAs was also examined over the milling process. Interestingly, there were differences in the EA profiles amongst the milling products associated with the outer kernel layers (bran, shorts, and feeds) and endosperm (semolina, flour), in addition to the variation in concentration.

These differences in EA profiles are illustrated in Figure 3. The ratio of *R*- to *S*-enantiomers was lower in milling products associated with the outer kernel layers as compared to whole grain, indicating a predominance of the *S*-enantiomers. The *R*-enantiomers were predominant in semolina and flour, and were present in these fractions to greater extent than in whole grain.

**Figure 3.** Difference in ratios of *R*-enantiomer to *S*-enantiomer concentrations between whole grain durum, milling products, and pasta matrices. The difference in ratios was calculated using mean *R*/*S* concentration ratios determined from the five groups containing 0.01–0.1% ergot sclerotia by mass.

The work in this study does not directly address or examine the cause of the differences in occurrence amongst individual EAs. However, past research has investigated or noted the epimerization of EAs. Epimerization of EAs is reported to be promoted by exposure to light [20], therefore EAs in the outer kernel layers may be subject to more light and subsequent epimerization than EAs in the inner kernel layers. Even though heat also appears to facilitate epimerization [17,18], it is unlikely that the milling process used in this study promoted epimerization in the bran, shorts, and feeds fractions, as temperatures generated during the roller milling of wheat are around 35 ◦C [21].

The cooking of spaghetti did not appear to considerably affect the presence of EAs. No substantial losses of EAs were consistently evident after the preparation of spaghetti by extrusion, nor after boiling the spaghetti. Figure 4 provides a comparison of the amount of EAs in freshly extruded spaghetti, cooked spaghetti, and the cooking water, to semolina. The comparison is on the basis of EA mass, therefore the impact from varying moisture content of the products is avoided. Across the five durum samples with varying ergot sclerotia contents, freshly extruded spaghetti and cooked spaghetti samples contained 77 ± 14% and 93 ± 18% of the total EA content observed in semolina, respectively. Cooking water contained a negligible amount of EAs (0.04 ± 0.02%) of the total semolina EA content.

Overall, the results observed do not indicate a consistent and extensive loss of EAs during processing. While the amounts of EAs observed in the freshly extruded spaghetti seem to suggest some loss, the amounts of EAs observed in cooked spaghetti do not demonstrate similar losses. The apparently lower EA amounts measured in the freshly extruded spaghetti may reflect differences in the ability of the analytical method to extract and/or measure EAs in this matrix.

Merkel et al. [17] reported small losses of 2–30% in cookies due to degradation of EAs during baking. Dänicke noted an average loss of 11% for EAs in heat treated rye [18], but the changes in concentrations across the five chicken diets examined, ranged from a loss of 26%, to an increase of 15%. This inconsistency suggests that the heterogeneous nature of ergot contamination of whole grain may have contributed to the apparent loss of EAs in the heat treated rye.

**Figure 4.** Amount of total ergot alkaloids in processed spaghetti and cooking water relative to semolina. The mean of duplicates is reported for freshly extruded spaghetti. Single analyses were performed for the other matrices.

Fajardo et al. [16] also reported losses of EAs during the cooking of Asian noodles and spaghetti made with wheat flour. However the analytical method used in that study only included the *R*-enantiomers as analytes. Any occurrence of *S*-enantiomers formed by epimerization during cooking would not be observed. Therefore changes in concentrations of EAs due to epimerization would appear instead as losses of EAs.

Even though no substantial losses of EAs were observed during the processing and cooking of spaghetti, the EA profile did change during cooking. As seen in Figure 3, the EA content of semolina and freshly extruded spaghetti is dominated by *R*-enantiomers, which changed to a predominance of *S*-enantiomers in cooked spaghetti and cooking water. The changes in the EA profile observed suggest that the heat of extrusion (45 ◦C) used to prepare the fresh spaghetti is not enough to promote epimerization, whereas the heat of boiling water can facilitate epimerization.

The enrichment of *S*-enantiomers after observed after cooking spaghetti is consistent with changes observed in other research. Merkel et al. reported the epimeric ratio shifted toward the *S*-enantiomer for all EAs in baked cookies [17]. Dänicke [18] also noted a consistent increase in the proportion of *S*-enantiomers (and concomitant decrease in *R*-enantiomers) in rye with varying ergot sclerotia content after heat treatment.

#### *3.2. Implications of Ergot Alkaloid Fate in Milling Products and Spaghetti*

The wider implications of this work relate to the distribution of EAs amongst the durum milling products and the epimerization of EAs observed. The association of EAs with the bran, shorts, and feeds fractions after durum milling will lower the exposure for populations consuming food products, such as pasta, made from semolina, as compared to whole grain durum. In turn, any incorporation of bran, shorts, and feeds fractions into animal feed will increase exposure of livestock, as compared to use of whole grain durum.

The epimerization observed from cooking and milling durum indicate that feed products and cooked pasta will contain a higher proportion of the less biologically active *S*-enantiomers. Overall, the deleterious health effects of EAs are associated with the ability of EAs to act as ligands for a variety of receptors. The ligand activity is reported to be greater for *R*-enantiomers, compared to *S*-enantiomers [18,20].

The fractionation of EAs amongst milling products and the epimerization observed during milling and cooking highlights the need for exposure assessments to consider concentrations in food products, or to use a processing factor, as opposed to using whole grain durum EA concentrations to estimate consumers' exposure. The results of this study also indicate that both *R*- and *S*-enantiomers should be monitored and assessed, in order to obtain an accurate view of consumers' exposure to EAs.

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

#### *4.1. Samples*

A No.1 Canada Western Amber Durum sample (CWAD, 30 kg) was cleaned and hand-picked to remove ergot bodies and any kernels with dark discoloration. The cleaned CWAD wheat that did not contain ergot sclerotia was divided into 6 × 5 kg sub-samples. Sclerotia hand-picked out of other naturally-infected CWAD samples were added back to five of the sub-samples in order to prepare grain with 0.01, 0.02, 0.03, 0.04, and 0.1% (m/m) ergot sclerotia. The sixth 5 kg sub-sample was kept at 0% ergot sclerotia. The 5 kg sub-samples were divided into 2 × 2 kg and 1 × 1 kg portions using a rotary sample divider (Materials Sampling Solutions, Southport, Australia). The 2 × 2 kg portions were milled as described below. The 1 kg whole grain portion was comminuted using a Retsch SR 300 rotor beater mill fitted with a 750 μm screen and coupled with a Retsch DR 100 vibratory feeder.

#### *4.2. Milling and Processing into Spaghetti*

Durum wheat samples were milled on a four stand Allis-Chalmers laboratory mill coupled with a small-scale semolina purifier previously described by Dexter et al. [22]. The mill flow consists of four corrugated break roll passages, five corrugated sizing roll passages, and 10 purification steps. The mill room was controlled at 21 ◦C and 60% relative humidity. Durum wheat samples were tempered to 16% moisture for 16 h before milling. Particles retained on a 425 μm screen after the four break passages were collected as bran. After each break passage, coarse material retained on a 630 μm screen was passed through sizing rolls. After sizing passages, flour was collected by combining materials with fine particles passing through the 180 μm sieve, and particles retained on a 700 μm screen were collected as shorts. The remaining milling streams were passed through a purifier. After each purification step, streams with particles passing through the 571 μm screens, but retained on the 183 μm screens, were collected as semolina. Particles retained on the 630 μm screens on purifiers 7–10 were collected as feed.

Following the method of Fu et al. [23], spaghetti was produced from semolina using a customized micro-extruder (Randcastle Extrusion Systems Inc., Cedar Grove, NJ, USA). For dough preparation, semolina was mixed with water in a high speed asymmetric centrifugal mixer (DAC 400 FVZ SpeedMixer, FlackTec, Landum, SC, USA) at constant water absorption of 31.5%. Vacuum was applied to eliminate the introduction of air bubbles, after which the dough crumbs were extruded through a four-hole Teflon coated spaghetti die (1.8 mm).

The fresh pasta was subsequently dried in a pilot pasta dryer (Bühler, Uzwil, Switzerland) coupled with a 325 min drying cycle and maximum temperature of 85 ◦C.

#### *4.3. Cooking and Moisture Content Determination*

Spaghetti (10 g) was added to 100 g of water brought to boil. The spaghetti was cooked uncovered for 10 min. After 10 min, cooked spaghetti was immediately removed from the cooking water with tongs, excess adhered water was left to evaporate for 1 min, and the mass of cooked spaghetti was determined. A 10 g portion of boiled spaghetti was then placed into a polyethylene sample bottle, which was closed and stored for approximately 2 h until extraction and EA analysis. The remainder of the boiled spaghetti was re-weighed, stored in sealed plastic bags, and frozen until moisture content determination. The remaining cooking water was cooled in a refrigerator and stored for approximately 4 h until extraction and EA analysis.

The moisture content of cooked spaghetti was determined gravimetrically. The remainders of the boiled spaghetti that were weighed immediately after cooking were placed on sieves and air dried for 24 h. The air dried samples were ground using a Brabender MLI-204 break mill, and then dried for 1 h in an oven held at 130 ◦C. The moisture content of the cooked spaghetti was calculated from the decrease between the post-cooking and air dried masses as well as the pre- and post-oven dried masses.

Moisture content of freshly extruded spaghetti was calculated based on the amount of water added and originally present in the semolina, minus 0.5% during to moisture loss during pasta processing.

#### *4.4. Determination of Ergot Alkaloids*

Samples were analyzed for 10 ergot alkaloids according to the liquid chromatography tandem mass spectrometry method described by Tittlemier et al. [5]. The 10 ergot alkaloid analytes are listed in Table 2. Four of the alkaloids were *S*-epimers of *R*-enantiomers. The two *S*-epimers that were not included in the method (ergonovinine and ergotaminine) were not easily obtained at the time of analysis. In order to correct for variations in final extract volume and injection volume, dihydroergotamine was used as an internal standard.

Milling products and spaghetti (10 g) were extracted with 50 mL 84:16 (*v*/*v*) acetonitrile/3.03 mM aqueous ammonium carbonate. The slurries of milling products were shaken on a flatbed shaker for 30 min, and spaghetti was comminuted in extraction solvent using a handheld laboratory homogenizer for 3 min at 12,000 rpm. After extraction, sample extracts were then centrifuged, and an aliquot of supernatant was diluted with 3.03 mM aqueous ammonium carbonate. Internal standard was added to all samples prior to analysis. Post-extraction fortified calibration standards were prepared in wheat matrix.

Water (1 g) was transferred from the cooled cooking water, combined with 1 mL acetonitrile, and vortex mixed for 10 s. The solution was left to sit for 2 min prior to a 1 mL aliquot being taken and combined with 1.5 mL of 3.03 mM aqueous ammonium carbonate. Internal standard was added, final sample extracts were then vortex mixed and filtered using PTFE filters prior to analysis.

Sample extracts were chromatographed on a C18 column followed by analysis in the positive electrospray ionization mode using multiple reaction monitoring. Two transitions were monitored for each analyte. In order to avoid epimerization of ergot alkaloids, the exposure of sample extracts to light was minimized by covering samples and extracts during processing, and by keeping the illumination option off on the liquid chromatograph sample manager.

Analytes were considered to be positively identified and quantitated if their retention times were within 0.1 min of the average retention time of the corresponding analyte in the external calibration standards; the peak had a signal-to-noise ratio greater than 9:1, and the ratio of qualification to quantitation ions was within acceptable tolerances [24]. Analyte peak areas were normalized to the dihydroergotamine peak area in samples during calculation of concentrations.

Blank wheat samples fortified with a solution of standards and an in-house rye reference material were analyzed with each batch of samples and were used to monitor the performance of the analytical method during the study.

#### *4.5. Evaluation of Method for Analysis of Cooking Water and Boiled Spaghetti*

Because water and boiled spaghetti were different matrices than the grain used in the initial validation of the EA analytical method [5], additional method evaluation was performed for these two matrices. Boiled commercially-available durum spaghetti, and the cooking water used to boil the spaghetti, were fortified with ergot alkaloids, and analyzed as described above.

The boiled spaghetti was fortified with the *R*-enantiomers only; fortification concentrations were 200 μg/kg for each alkaloid aside from ergonovine, which was fortified at 40 μg/kg. The boiled spaghetti was drained, and as much water adhering to the cooked pasta as possible was removed by shaking. The cooked spaghetti was then weighed within a minute after boiling to obtain a relevant cooked mass, and 10 g was transferred to a 250 mL centrifuge bottle, capped, and stored in darkness for 30 min to cool prior to fortification. After fortification, samples sat for 20 min in darkness at room

temperature prior to extraction. Extraction and analysis was performed as described above, using a ratio of 5:1 (*v*/*m*) extraction solvent to mass of boiled spaghetti.

The cooking water was fortified with both *R*- and *S*-enantiomers. After the cooked spaghetti was removed, the water was placed in a refrigerator for 30 min. After cooling, two portions of 10 mL were removed. One portion was fortified with *R*-enantiomers, and the other was fortified with *S*-enantiomers. Concentrations of ergot alkaloids in both portions were as mentioned above for boiled spaghetti. Three aliquots were taken from the *R*-enantiomers fortified 10 mL, and two aliquots were taken from the *S*-enantiomers fortified 10 mL. Extraction and analysis was performed as described above. All aliquots were analyzed for all ergot alkaloids in order to examine if epimerization had occurred during sample processing.

#### *4.6. Statistical Analyses*

Statistical analyses were performed using SigmaPlot 13.0 (Systat Software Inc., Chicago, IL, USA). The mean molar percentages of ergocristine and ergocristinine were compared amongst milling products using a Kruskal-Wallis One Way Analysis of Variance on Ranks. The Tukey Test was used to isolate milling products whose mean molar percentages differed from others. Statistical tests were limited to these two EAs because they were the predominant compounds in the milling products.

**Author Contributions:** Conceptualization, S.A.T.; Methodology, B.X.F and S.A.T.; Formal Analysis, S.A.T.; Investigation, D.D., M.R., D.T. (Dave Turnock), and D.T. (Dale Taylor); Resources, B.X.F. and S.A.T.; Writing-Original Draft Preparation, S.A.T.; Supervision, B.X.F. and S.A.T.; Project Administration, S.A.T.

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

**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/).

## *Review* **Mycotoxins during the Processes of Nixtamalization and Tortilla Production**

#### **Sara Schaarschmidt \* and Carsten Fauhl-Hassek**

German Federal Institute for Risk Assessment (BfR), Department Safety in the Food Chain, Max-Dohrn-Str. 8-10, D-10589 Berlin, Germany; carsten.fauhl-hassek@bfr.bund.de

**\*** Correspondence: sara.schaarschmidt@bfr.bund.de

Received: 28 March 2019; Accepted: 11 April 2019; Published: 16 April 2019

**Abstract:** Tortillas are a traditional staple food in Mesoamerican cuisine, which have also become popular on a global level, e.g., for wraps or as snacks (tortilla chips). Traditional tortilla production includes alkaline cooking (nixtamalization) of maize kernels. This article summarizes the current knowledge on mycotoxin changes during the nixtamalization of maize and tortilla production. Upon nixtamalization, mycotoxins can be affected in different ways. On the one hand, the toxins can be physically removed during steeping and washing. On the other hand, mycotoxins might be degraded, modified, or released/bound in the matrix by high pH and/or high temperature. This also applies to the subsequent baking of tortillas. Many studies have shown reduced mycotoxin levels in alkali-cooked maize and in tortillas. Most of the available data relate to aflatoxins and fumonisins. The reduction (and detoxification) of aflatoxins during nixtamalization might, however, be partially reversed in acidic conditions. The loss of fumonisin concentrations is to some extent accompanied by hydrolyzation and by lower toxicity. However, some studies have indicated the potential formation of toxicologically relevant modified forms and matrix-associated fumonisins. More data are required to assess the influence of alkaline cooking regarding such modified forms, as well as mycotoxins other than aflatoxins/fumonisins.

**Keywords:** aflatoxins; alkaline; hydrolyzed fumonisins; fumonisins; food processing; maize; masa; matrix-associated mycotoxins; modified mycotoxins; tortillas

**Key Contribution:** The paper provides a critical overview of the effect of masa and tortilla production on mycotoxin concentrations considering the potential degradation and transformation of mycotoxins and matrix–toxin interactions.

#### **1. Introduction**

Mycotoxins are secondary fungal metabolites that are produced in the field and/or during the storage of crops and raise health concerns for humans and animals due to their toxic potential. Typically, several mycotoxins occur in parallel in crops. They can be produced by different fungal species, but single species are also usually capable of producing a distinct set of toxins [1]. Aflatoxins are mainly produced by *Aspergillus* and *Pencillium* species, with aflatoxin B1 (AFB1) being the most toxic and carcinogenic. Aflatoxins—including AFB1, aflatoxin B2 (AFB2), aflatoxin G1 (AFG1), and aflatoxin G2 (AFG2)—are particularly common in maize and other crops produced in warmer climates and are a serious health threat in many regions worldwide (for more information see e.g., [2,3]). Fumonisins belong to a large group of toxins referred to as *Fusarium* toxins that are produced by several *Fusarium* species, such as *Fusarium verticillioidies*. Moreover, fumonisins can be produced by some species of the *Aspergillus niger* complex. Fumonisins B1, B2, and B3 (FB1, FB2, FB3) are frequently found in raw maize and can exhibit liver and kidney toxicity [4]. Other *Fusarium* toxins, which are often present in maize, include zearalenone (ZEN [5]) and trichothecenes, such as deoxynivalenol (DON) [6,7]. In addition to such 'traditional' mycotoxins, which have been assessed and monitored in relative depth, so called 'emerging' mycotoxins have been identified. Similar to traditional mycotoxins, emerging mycotoxins are directly produced by fungi. Although some have been known for several decades already, emerging mycotoxins are still, however, less investigated and understood. One example of emerging mycotoxins is moniliformin (MON), which is also produced by some *Fusarium* species [8].

In addition to the free mycotoxins produced by toxigenic fungi (free parent compounds), mycotoxins can be modified in their chemical structure by biological or chemical processes [9–11]. Further, mycotoxins can be connected to the matrix, either by being physically entrapped or by covalent binding to matrix molecules. A proposal for a harmonized terminology of modified and matrix-associated mycotoxins was provided by Rychlik et al. [11]. Such forms can raise analytical challenges, which is particularly true for matrix-associated mycotoxins. For those, special treatments of the matrix, after extraction of free mycotoxins, are required to be able to extract the bound forms. However, free modified forms are, similar to emerging mycotoxins, often not covered by routine analysis.

Food processing, in general, is capable of affecting mycotoxins present in the raw materials. A reduction in mycotoxin concentrations might be caused by fractionation or (partial) degradation of the toxins, leading also to lower toxicity. However, often, lower mycotoxin levels (also) involve modification or binding mechanisms. In such cases, the resulting structures might still harbor unknown toxicity or might be (re)converted into a toxic form. Further, concentrations of free toxins can even increase during food processing by releasing mycotoxins from matrix components (if matrix-associated mycotoxins are present in the raw material and depending on the processing conditions). Tortillas are a traditional staple food for the Mesoamerican population and are increasing in popularity throughout the world [12], which also applies to related products, such as tostadas, tortilla chips, and maize chips. Their unique flavor is caused by an alkaline cooking of maize—a special processing procedure called nixtamalization. This process can cause several physicochemical changes in maize kernels and is capable of affecting mycotoxins. This review presents the current knowledge on mycotoxin changes during the process of tortilla production with a particular focus on the nixtamalization step. In doing so, changes in the concentration of free parent forms are considered, as well as their potential modification and the putative toxicological impacts.

#### **2. Processes Involved in Nixtamalization and Tortilla Production**

Nixtamalization describes an ancient food processing procedure developed and applied by indigenous Mesoamerican (e.g., Aztec and Mayan) civilizations [13], which is still used nowadays. It represents alkaline cooking of maize kernels. Traditionally, this is done using lime, which mainly consists of Ca(OH)2. Classic nixtamalization also makes use of wood ash. In traditional nixtamalization (Figure 1), maize kernels are cooked in lime water followed by steeping at room temperature, which typically takes place overnight. During nixtamalization, the elevated pH and high temperature facilitate the softening of the endosperm and the release of the pericarp. After steeping, the cooking and steeping liquid, the so-called nejayote, is removed. The alkaline-cooked kernels (nixtamal) are then washed with water to remove excessive lime, as well as (part of) the loosened pericarp. The aleurone layer, i.e., the outermost layer of the endosperm that is rich in protein and vitamin B1–3, stays attached to the starchy endosperm. The aleurone layer also contributes to a reduction of protein and starch losses during cooking, steeping, and washing [14]. After washing, the nixtamal is stone-ground upon the addition of water to form a maize dough called masa. Small portions of masa are formed into balls that are flattened into thin discs. The so-formed tortillas are baked on a hot plate or in an oven. By frying, tortillas and tortilla strips can be further processed into tostadas and tortilla chips (or similar products). Additionally, masa can also be used to prepare maize chips. With respect to storage, masa can be dried and later remoistened for further processing. Moreover, a dry masa flour can be produced from low-moisture nixtamal by fine grinding under dry conditions. In this process, due to the low moisture, no release of starch granules from the protein matrix occurs compared with fresh masa

production, in which nixtamal is ground at a high moisture level. Dry masa flour is often used in commercial tortilla production.

**Figure 1.** Scheme of typical steps and (by)products in traditional nixtamalization and tortilla production. The industrial production of tortillas often makes use of dry masa flour, which is made from dried nixtamal by fine grinding (not shown).

Alkaline cooking of maize causes several physical, as well as (bio)chemical, changes [14,15]. Some of those are associated with the enhanced nutritional value of the grain and are of particular importance in diets mainly relying on maize. The improved bioaccessibility of calcium and niacin (vitamin B3) are likely the most important of these changes. Thanks to the latter, pellagra—a niacin-deficiency disease typically related to maize- and sorghum-based nutrition—is not common in tortilla-eating countries. Moreover, nixtamalization can improve protein digestibility and can positively affect the protein quality of maize by partially adjusting the leucine-to-isoleucine disproportion [16]. In tortillas, few amino acids were found to be reduced—including leucine, which can act as an antagonist of isoleucine. Nonetheless, maize protein is, in general, deficient in lysine and tryptophan but relatively rich in the sulfur-containing amino acids methionine and cystine. Nixtamalization furthermore enhances the content of resistant starch, which is accompanied by a lower glycemic index [17,18]—a factor that is nowadays of special interest. Furthermore, traditional tortilla production is capable of lowering mycotoxin contaminations, as outlined below, which is of particular significance for a healthy cereal-based diet. Lime-cooked maize and products thereof are, moreover, characterized by a unique flavor, which contributes to the increasing popularity of such foods on a global level [12,19].

#### **3. Aflatoxins during Nixtamalization and Tortilla Production**

#### *3.1. Impact on Aflatoxin Concentrations by Traditional Nixtamalization*

Because aflatoxins, particularly AFB1, are a severe health threat and are often found in maize produced in warmer climates, strategies for reducing aflatoxin contaminations are of special importance. Interestingly, traditional nixtamalization is in several studies described to have a high potential for

lowering aflatoxin concentrations (for details, see Table 1). The nejayote, the main waste byproduct, which typically also contains a solid fraction that mainly consists of maize tip cap, pericarp, and germ, can in return show a certain accumulation of aflatoxins.

The loss in aflatoxins and potential transfer to the lime water is also dependent on the type of aflatoxin [20,21]. In a study by Ulloa-Sosa et al. [21], AFB1 + AFB2 were less reduced (by around 40%–50% in nixtamal and masa) compared with AFG1 + AFG2 (by around 75%). The total aflatoxin concentration in nixtamal and masa was approximately reduced by 60%–65%. Remarkably, whereas most of the AFB1 + AFB2 were detected in the nejayote, no AFG1 or AFG2 could be detected in this matrix, pointing to a degradation or transformation into undetectable form(s) upon exposure to alkaline pH. Tortillas showed approximate reductions in total aflatoxins, AFB1 + AFB2, and AFG1 + AFG2 levels of 70%, 60%, and 80%, respectively. Additionally, de Arriola et al. [20] found AFG1 and AFG2 to be somewhat more reduced during nixtamalization than AFB1 and AFB2 (average reduction of AFB1, AFB2, AFG1, and AFG2 in masa approximately 93%, 90%, 98%, and 97%, respectively). Both studies were performed with fungal-inoculated kernels. Abbas et al. [22] analyzed the impact of the entire tortilla production process (covering nixtamalization using a 2% Ca(OH)2 solution) on natural aflatoxin contaminations. Here, the AFB1 content was on average reduced by 40%, and the AFB2 content was reduced by only 28%. AFG1 and AFG2 were not detected in unprocessed maize (or in tortillas).

Enhancing the concentrations of Ca(OH)2 for cooking and steeping would not necessarily cause a more pronounced aflatoxin reduction in tortillas [20,23]. Lime concentrations of >2% are even described to produce tortillas with organoleptic characteristics unsuitable for human consumption [20]. De Arriola et al. [20] found average reductions of total aflatoxin concentrations in masa and tortillas of 94% and 95%, respectively, at 0.6% lime without significant difference compared with the use of 1.87% lime. The experiments were done on highly aflatoxin-contaminated maize obtained by fungal inoculation. Moreover, the cooking procedure (i.e., cooking in an open kettle versus pressure cooking in an autoclave) made no significant difference regarding the change in total aflatoxin concentration [20]. When comparing the impact of five different nixtamalization processes on aflatoxin concentrations (see Table 1), Price et al. [23] found cooking, prolonged steeping, and washing of nixtamal to facilitate aflatoxin reductions in nixtamal, masa, and tortillas. Here, aflatoxin levels were reduced by approximately 50%–70% in tortillas compared with the naturally contaminated unprocessed maize (initial aflatoxin level of around 140 μg/kg). When using AFB1-spiked kernels (100 μg/kg) in previous experiments, the authors mentioned that the toxin was almost not detected after tortilla production [23].

Based on their experimental data, Moreno-Pedraza et al. [24] proposed a (traditional) process for nixtamalization and tortilla production, which is supposed to completely reduce AFB1 contaminations. The key steps are as follows: (i) cooking kernels in 1% lime at 90 ◦C for 45 min; (ii) further steeping in the alkaline solution at 25 ◦C for 18 h; (iii) removal of the nejayote followed by only one washing step to remove the pericarp but to preserve the alkaline conditions (approximately pH 10) of the nixtamal; (iv) resting the masa, which has been obtained by grinding the washed nixtamal and still features the high pH value at room temperature for 40 min; and (v) flattening and cooking of the tortillas. The authors found that the nixtamalization step, the resting of the alkaline masa for ≥30 min, and/or the tortilla baking completely or almost completely eliminated AFB1 levels (either low natural contamination or a 115–125 μg/kg spike in the masa). In this study, the alkaline treatment lead to the formation of at least two unidentified degradation or transformation products: one of 301.25 Da (molecular formula: C17H16O15) and another of 325.33 Da (molecular formula: C17H18O5) [24].

To follow the fate of aflatoxin during traditional nixtamalization, radio-labeled AFB1 was used [25]. Natural AFB1 contamination (at low and high levels) was found to be lowered by 97%–100% after traditional nixtamalization using a chromatographic method. When radio-labeled AFB1 was spiked in unprocessed maize, the loss in radioactivity amounted to 84% in the masa. The remaining radioactivity was detected in the nejayote and the washing liquids (with decreasing levels in later washings). Whereas the aflatoxin retained in the masa appeared intact (detectable by an antibody-based method), aflatoxin in the liquid waste fractions was undetectable by means of chromatography [25].

#### *3.2. Aflatoxin Reductions by Alternative Nixtamalization Processes*

The abovementioned studies point to a high potential of traditional nixtamalization to lower aflatoxin concentrations. However, traditional nixtamalization is a time-consuming process characterized by a relatively high input of water and energy. Moreover, the nejayote is considered to be a highly problematic byproduct due to its excessive pH, its high content of organic and insoluble matter, and other factors [26], and it is usually disposed in landfills, not utilized [27]. Thus, to reduce water and energy inputs and the amount of wastewater, alternative processing procedures were considered. Such methods were not only tested regarding organoleptic and nutritional characteristics, but also in view of aflatoxin reductions as outlined below.

One promising alternative method for masa and tortilla production might be the extrusion of maize meal upon the addition of lime. At a lime concentration of 0.2%, based on maize meal, organoleptic properties were found to be comparable to traditionally produced masa/tortillas (using a 0.33% lime solution, which represented 1% lime based on maize meal). Total protein and lysine contents were similar (or only slightly lowered), but the tryptophan loss was much lower in extruded masa compared with traditional masa [28]. A higher nutritional value, accompanied by an elevated weight gain and protein efficiency ratio in a rat feeding trial, was, in addition, shown for tortillas produced with maize meal extruded at 0.15%–0.25% lime (relative to maize meal mass) compared with traditional tortillas produced with masa prepared with 2% lime based on kernel mass (= 0.67% lime solution) [29]. Elias-Orozco et al. [30] evaluated the alkaline extrusion process regarding aflatoxin reductions in naturally contaminated maize. Astonishing, in the highly contaminated batch, they found a high level of aflatoxin M1 (AFM1) in the raw maize, which was only around 20% less than the AFB1 level. It had been previously described that *Aspergillus* spp. are capable of producing AFM1 (and aflatoxin M2); however, usually relatively low amounts of AFM1, compared with AFB1 or total aflatoxins, are found in maize or in culture media [31–34]. Furthermore, they detected AFB1-dihydrodiol in the raw maize, which is formed via enzymatic oxidation of AFB1 followed by non-enzymatic hydrolysis. Tortillas produced after extrusion of maize meal lacking any lime showed reductions in AFB1, AFB1-dihydrodiol, and AFM1 levels of approximately 46%, 54%, and 20%, respectively. When lime was added at 0.3% relative to maize meal mass, reductions increased to approximately 74%, 70%, and 52%, respectively. At 0.5% lime, they amounted to 83%–89%. Moreover, the authors tested extrusion upon the addition of 0.75%–3% H2O2, alone or in combination with 0.3% lime. Similar to lime, adding H2O2 to the extrusion process can increase aflatoxin reductions in tortillas. However, the combination of lime and H2O2 showed no or minor benefits compared with lime alone, except regarding AFM1 reduction. Here, reduction was enhanced from approximately 52% (0.3% lime, only) to 61%–73% (0.3% lime + 0.75%–3% H2O2 based on maize meal). To compare, traditional nixtamalization of the highly contaminated maize resulted in AFB1, AFB1-dihydrodiol, and AFM1 reductions of 92%–94%. Similar results on the effect of extrusion treatments and traditional nixtamalization were found for a maize batch contaminated with AFB1 at a lower level; other aflatoxins were here not detected [30] (see Table 1 for details).

Pérez-Flores et al. [26] tested the use of a microwave for the nixtamalization process. For that, maize grits (obtained from fungal-inoculated kernels) were cooked in a minimized amount of lime water (0.5% Ca(OH)2) in a microwave (for details, see Table 1). After steeping (3 h), no water removal or washing, which could have caused fractionation of mycotoxins, was indicated by the authors. The so-produced masa had 36%–82% lower aflatoxin concentrations than the maize kernels, with higher reduction at higher initial contamination levels. In tortillas, aflatoxin levels were lowered by 68%–84% of the initial amount. Tortillas produced by such microwave nixtamalization showed comparable physicochemical (moisture, pH, color) and technological properties (puffing, rollability, weight loss) as described for traditionally produced tortillas.

Minimization of water and energy input was also tested with a so-called ecological nixtamalization process [35]. Here, maize meal was mixed with a minimum amount of hot (92 ◦C) 0.375% lime solution for only 10 min. After steeping (2 h), the nixtamal was ground into masa without any water removal or washing steps. With such a process, AFB1 + AFB2 levels were lowered by 25%–40%, 13%–25%, and 61%–78% in the nixtamal, masa, and tortillas, respectively. Higher percentage reductions were, however, detected at lower initial contamination levels. Although applied to milled maize, the ecological nixtamalization was overall less effective in reducing aflatoxins compared with a traditional nixtamalization process applied to kernels of the same (fungal-inoculated) batch. The tested traditional nixtamalization covered a higher concentration of the lime water (1% lime), longer incubation times (70 min cooking, 12 h steeping), removal of the nejayote, and washing of the nixtamal (which typically also removes the loosened pericarp). Here, aflatoxin reductions amounted to 83%–92%, 87%–89%, and 90%–92% in nixtamal, masa, and tortillas [35].

Torres et al. [36] compared a traditional process (that included the cooking of kernels in lime water) with a commercial one. In the latter, whole maize kernels were mixed with lime and boiling water without further cooking (similar as described to the aforementioned ecological nixtamalization of maize meal [35]). In both processes tested by Torres et al. [36], the nejayote was removed after a 14 h steeping, and the nixtamal was washed twice. In the commercial process, the pH levels of the nixtamal, masa, and tortillas were lower compared with those in the traditional one (5.8–5.9 compared to 6.7–6.8). Moisture content was also slightly lower. Further, the loss of solids was reduced (4.1% in the commercial versus 6.8% in the traditional process). These factors likely contributed to the lower efficiency of aflatoxin reduction: The commercial processing reduced the level of total aflatoxins in tortillas by 30%, whereas the traditional tortilla production was more efficient (52% aflatoxin loss). Maize chips and tortilla chips showed aflatoxin reductions of 71% when using traditionally produced masa and 79%–85% upon use of masa produced with the tested commercial process. However, maize at different initial aflatoxin concentrations was applied to the two processes, which might also have affected the aflatoxin reduction efficiency. In this study, samples were acidified upon extraction (before filtration of suspended samples) to cause a reconversion of potential transiently transformed aflatoxins [36] (see below).

#### *3.3. Potential Reconversion of Modified Aflatoxins*

In general, besides the leaching of aflatoxins into liquid fractions, alkaline conditions can cause the opening of the lactone ring of aflatoxins (including AFB1), resulting in a loss of fluorescence of the molecules and thus a loss of fluorescence-based detection. Further, a strongly reduced toxicity and mutagenicity after cleavage of the lactone ring was described [37]. In nixtamalized maize (products), the lowered aflatoxin concentrations were found to be accompanied by lower mutagenicity and oxidative stress in vitro [23,38]. Vázquez-Durán et al. [38] showed that for extracts of raw maize, a more pronounced lipid peroxidation in kidney Vero cells occurred than for extracts of tortillas, which were produced from the raw maize by a microwave nixtamalization process (as described by Pérez-Flores et al. [26]) and had a 84% lower aflatoxin level. Further, no mutagenic toxicity was detected in the tortillas in the Ames test, but it was present in the unprocessed maize. Similarly, Price and Jorgensen [23] observed a reduced mutagenic potential for masa and tortilla samples compared with raw maize when testing different nixtamalization processes (although the number of revertants in the Ames test did not always correlate with the detected aflatoxin level regarding the extent of reductions).

The modification of aflatoxins during alkaline treatment is not necessarily permanent, however, and might be reversed upon exposure to acidic conditions, as present in monogastric digestive systems. Price and Jorgensen addressed this issue by acidifying the suspended samples in the course of aflatoxin extraction (original pH around 11; acidified: 5–6), mimicking acidification in the human stomach. In fact, in doing so, part of the undetectable modified aflatoxin(s) was reconverted into fluorescent form(s). After acidification, the total aflatoxin reduction in the tortillas amounted to only 20%–46%,

instead of 48%–73% when lacking such a step. For masa, the reduction in fluorescent aflatoxin(s) was approximately 14%–56% without and 4%–29% with acidification. Additionally, Méndez-Albores et al. [39] showed that the reduction in aflatoxin concentrations by nixtamalization is partly reversible. After acidification, aflatoxins became, to some extent (approximately 5% of the concentration in raw maize), detectable in the dried nejayote, which originally had a pH of 12. When the extracts of the samples were acidified (initial pH of samples: 8.2–8.3), aflatoxin concentrations were around 57% and 34% higher compared with those in the alkaline extracts of masa and tortilla, respectively. However, compared with the raw maize, the aflatoxin levels were still very low with reductions of 78% in masa and 91% in tortillas (for the alkaline extracts, reductions amounted to 86% and 93%, respectively). Pérez-Flores et al. [26] found that only very low amounts of aflatoxins in the extracts of masa and tortillas were recovered by an acidification step. Here, the pH of masa and tortillas was again around 8.2 prior acidification (and adjusted to 3). Different from Price and Jorgensen [23], in the two latter studies [26,39], not the suspended samples, but the sample extracts were acidified. However, when treated with weak bases and during ammoniation, AFB1 was found to interact with matrix macromolecules, including non-protein fractions [40,41]. Hence, it is tempting to speculate that the extraction efficiency of the modified aflatoxin(s) is dependent on the pH of the matrix and that it is higher under acidic conditions. If this is true, matrix-associated aflatoxins potentially present in tortillas could be also released in the stomach.

In the study by Price and Jorgensen [23], acidification of the samples was found to be further capable of restoring mutagenicity in the Ames test. The mutagenic effects for the tortilla samples were even somewhat higher than for raw maize, which, however, contradicted the reduced aflatoxin concentrations that were observed in the acidified tortilla samples [23]. This might indicate the formation of additional mutagenic form(s) during tortilla production, which could also explain the rather low correlation between mutagenicity and aflatoxin concentration determined by the authors. In general, the efficiency and persistency of aflatoxin transformation/detoxification by elevated pH is dependent on several factors. Positive effects of temperature, time, and kernel moisture on AFB1 reduction under alkaline conditions were shown for ammoniation at atmospheric pressure [42] and under elevated pressure [43]. Differences in aflatoxin reductions depending on the processing procedures were also found for nixtamalization and tortilla production (Table 1). Further, initial contamination levels and type of contamination (contamination in the field, post-harvest contamination, spiking with pure standard) might affect mycotoxin reduction efficiencies. Such factors should be considered when assessing the aflatoxin loss in view of toxicity in alkaline-processed maize and products thereof, such as tortillas.





**Table 1.** *Cont.*


*Toxins* **2019**, *11*, 227

**Table 1.** *Cont.*



analyzing extracts before acidification.

batches are given and separated by this symbol. AFB1: aflatoxin B1; AFB1-diol: aflatoxin B1 dihydrodiol; AF: aflatoxin(s) not further specified in the cited study but likely total aflatoxins

B1 + B2 + G1 + G2; total AFs: aflatoxins B1 + B2 + G1 + G2. accum.:

 ~: Approximate

 values that were calculated for this overview by using the data provided in the cited literature. |: Here, individual data of two

accumulation;

 LOD: limit of detection; n.d.: not detected; O/N: overnight; orig.: original.

#### **4. Fumonisins during Nixtamalization and Tortilla Production**

#### *4.1. General Impact on Fumonisin Concentrations and Fumonisin Hydrolyzation*

Fumonisins are very water-soluble mycotoxins, which can thus leach into the liquid fraction during cooking and steeping procedures. Furthermore, an alkaline treatment can result in a hydrolysis of the *O*-acyl bonds of fumonisins, leading to the formation of hydrolyzed fumonisins. Sydenham et al. [44] found, upon steeping of maize kernels and maize meal in 0.1 M Ca(OH)2 (at room temperature, under continuous stirring), a reduction in FB1 concentrations and an accumulation of fully hydrolyzed FB1 (HFB1; also referred to as aminopentol). For maize meal naturally contaminated with FB1, almost all of the mycotoxin was lost. Here, around 78%–89% of the FB1 was converted into HFB1, with 68%–72% being transferred into the steeping liquid and 11%–17% remaining in the alkali-treated maize meal. The latter contained only up to 9% of the FB1 level of untreated maize meal. In total, around 11%–25% of the FB1 was retained as FB1 or HFB1 in the maize meal. When treating whole kernels (also naturally contaminated) in the same manner, the reduction in FB1 concentration amounted to 76%–99%. After treatment, kernels were manually sorted by the extent of pericarp loss. Kernels with fully removed pericarp showed almost no FB1 left, and only approximately 4% of the FB1 was detected to be present as HFB1. In kernels with partly removed pericarp, approximately 7% of the initial FB1 was present as HFB1, and 24% remained in the parent form [44]. Accordingly, the removal of the nejayote and of maize pericarp would contribute to fumonisin reduction in nixtamalized maize.

Pilot-scale processing of naturally contaminated maize simulating commercial nixtamalization and tortilla (chips) production showed significant reductions in concentrations of FB1 and FB2 [45,46]. Although having a similar pattern, the extent of fumonisin reduction varied in both studies among individual runs, independent of initial concentration. Voss et al. [45] used different maize batches for five runs. Dombrink-Kurtzman et al. [46] examined the same maize batch in two runs but found nonetheless strong variations, particularly for FB2 (for details, see Table 2). FB1 reduction in nixtamal was in both studies accompanied by an accumulation of HFB1 in the steeping and washing liquids. Voss et al. [45] detected, besides HFB1, some partially hydrolyzed FB1 (PHFB1) in the raw maize. However, this compound did not accumulate in the nejayote but, if present, decreased over time. The decrease in FB1 and PHFB1 was accompanied by an increase in the fully hydrolyzed form. In general, hydrolyzation particularly takes place in the nejayote, which typically has a pH of ≥11. Palencia et al. [47] found the molar ratio of HFB1 to FB1 to be 21 in the nejayote but around 1 in wash water, masa, and tortillas. The overall transfer of fumonisins to the nejayote amounted in the study of Voss et al. [45] to approximately 45% of the total initial amount (on a molar basis) of FB1, PHFB1, and HFB1. Additional amounts were detected in the washing water. Dombrink-Kurtzman et al. [46] described the liquid fractions to contain on average of 76% of the initial FB1: 72.5% of FB1 was converted into HFB1 and 3.5% remained as FB1. The study also indicated the potential for further lowering of FB1 and FB2 levels during masa/tortilla production, in case they were somewhat less reduced in the nixtamal. FB1 and FB2 reductions in tortillas amounted in both runs to 88%–92% and 71%–91% compared with the levels in unprocessed maize, respectively. However, in one of the runs, the nixtamal showed reductions of around 75% for FB1 and only 20%–30% for FB2 [46] (Table 2).

The potential impact of tortilla baking on fumonisin reductions also became obvious in a study that tested a microwave nixtamalization process using maize grits. Here, total fumonisin levels were not significantly lowered in masa. This was in accordance with the tested processing procedure, because no removal of nejayote or washing of nixtamal was indicated by the authors. Different from masa production, the baking of tortilla lowered fumonisins by approximately one half. The reduction during the heat treatment was likely facilitated by the high pH of masa and tortillas (i.e., around 8.1–8.3). The tested physicochemical and technological characteristics were similar to those described for traditionally produced tortillas [48].

The fate of fumonisins was also investigated during commercial processing into tortilla (chips) [49,50]. Scudamore et al. [49] analyzed industrial tortilla chip production in United Kingdom (UK) plants, involving mixing of a maize flour dough followed by sheeting, cutting, baking, and frying. Alkaline conditions were, however, not indicated. However, because nixtamalization contributes to the typical flavor of tortillas and tortilla chips, we assume that dry masa flour or a similar ingredient was involved in the commercial production process. When analyzing 11 runs (that comprised two different compositions of maize flour mixtures), FB1 + FB2 were lowered by 32%–78% on the product 'as is' basis (average: 59%). Because the moisture content of the chips is usually more or less comparable to that of dry maize ingredients, a similar fumonisin reduction would apply when related to dry weight. Commercial tortilla production in Texas was studied by De La Campa et al. [50]. Here, the reduction in FB1 levels was high overall and ranged from 80% to 100% in masa and from 83% to 100% in tortillas. Production conditions strongly differed between the four processing plants, regarding, for example, lime concentration and cooking time.

In experimental studies, De La Campa et al. [50] further investigated the impact of these factors at different initial FB1 levels using fungal-inoculated maize. In doing so, they found a positive impact of lime concentration (when testing lime solutions of around 0.25%–1.6%) on FB1 reduction. This effect was independent of the initial FB1 concentration, which also had a significant effect on FB1 half-life. Regarding boiling time (15 versus 60 min), the authors mentioned that this factor had no apparent effect, but data were not shown [50]. Additionally, De Girolamo et al. [51] described the low effect of cooking time on the hydrolyzation of fumonisins, when comparing cooking times of 15, 30, and 60 min. In this study, which tested lime solutions with concentrations of around 0.33% and 1.67%, nixtamalization lowered mean FB1 + FB2 levels in masa by 26%–48%. Interestingly, the same process lacking lime resulted in a somehow stronger FB1 + FB2 reduction. Here, PHFB (PHFB1 + PHFB2) levels in masa were also lowered, but the loss was not accompanied by the formation of HFBs (HFB1 + HFB2). Reductions are likely solely caused by the leaching of fumonisins into the liquid fractions. By contrast, the use of lime provoked the formation of (partially) hydrolyzed forms of FB1 and FB2. Hydrolyzation was again more pronounced at higher lime concentration [51].

#### *4.2. Potential Further Transformations of Fumonisins*

On the one hand, De Girolamo et al. [51] discovered that alkaline cooking can somehow facilitate the release of bound fumonisins. In their study, the total mass of FB1 + FB2, PHFBs, and HFBs recovered after nixtamalization (also including the liquid waste fractions) exceeded the initial mass by around 50%–80%. In the course of nixtamalization, (part of) the released matrix-associated fumonisins were suggested to be hydrolyzed. Although the water-cooked maize showed a higher reduction in (free) FB1 + FB2 than the alkali-cooked maize, matrix-associated fumonisins would still be present. Moreover, bound fumonisins in food (and feed) products can, in general, increase health concerns, because free toxins might be released during digestion. Promoting the release from the matrix followed by hydrolyzation of fumonisins could contribute to a detoxification by nixtamalization. Different from FB1 and FB2, no (liver) cancer-promoting activity or weight loss was found for HFB1 and HFB2 in rats [52]. In contrast, in vitro tests showed a higher toxicity on primary rat hepatocytes in this study. Hence, it was concluded that the hydrolyzed fumonisins are not adsorbed from the gut [52]. A lower or lacking hepatic, intestinal, and neural toxicity of the hydrolyzed form compared with the parent compound was also shown in pigs and mice [53,54], although an impact on sphingolipid metabolisms in vivo was demonstrated at a high dose of HFB1 [54]. Inhibition of ceramide synthase and disruption of sphingolipid metabolism is the critical biochemical effect underlying fumonisin cytotoxicity.

On the other hand, hydrolyzation might also favor interaction with other compounds, including matrix macromolecules. Interestingly, Park et al. [55] were able to detect matrix-associated fumonisins in some retail tortilla chip samples. However, when analyzing retail samples, it cannot be excluded that the forms were already present in the raw material. To address this question, Burns [56] investigated a nixtamalization process by applying the detection method developed by Park et al. [55]. In doing so, a significant increase in protein-bound and other matrix-associated FB1 during nixtamalization was demonstrated. When maize kernels were processed in the same manner, but lacking lime, no significant change was observed in the concentration of total matrix-associated FB1 [56]. A reduction in recoverable (H)FB1 was described for experimentally produced and extruded masa flour [57]. However, here, the underlying mechanisms (degradation, binding, or modification to undetected free forms) remained unknown.

*N*-(carboxymethyl)-FB1 was previously shown to be formed under alkaline conditions at elevated temperatures by using pure FB1 incubated overnight with D-glucose [58] and also when heating HFB1 with D-glucose [59]. Interaction with glucose during extrusion cooking of maize resulted in a strong reduction in fumonisin-induced toxicity in rats [60]. However, when analyzing nixtamalization and tortilla chip production mimicking commercial processing, no indications were given for a (relevant) formation and accumulation of fumonisin–sugar adducts, namely *N*-(carboxymethyl)-FB1 and *N*-(1-deoxy-D-fructos-1-yl)-FB1 [45]. By contrast, Park et al. [61] could detect *N*-fatty acyl fumonisins in a tortilla chip sample (in 1 out of 38 retail samples), indicating a potential formation of those modified fumonisins in alkali-treated and fried maize products. In vitro studies implicate a high toxicity of several *N*-fatty acyl fumonisins. In addition, such modified forms can be more rapidly taken up and accumulated in human/animal cells than FB1 (for an overview, see [4]). However, further studies are required to obtain more information on the toxicity of *N*-acetylated fumonisins and their occurrence in foods. The same is true for the interaction of fumonisins with other molecules that potentially takes place during nixtamalization and the possible contribution to fumonisin-related toxicity.

Using bioassays and feeding trials, several studies indicate a reduced toxicity of FB1-contaminated maize raw material after being processed by nixtamalization [47,62–64]. To analyze (potential) kidney damage, in addition to histological analysis, sphinganine can be used as a biomarker for fumonisin-induced ceramide synthase inhibition. Palencia et al. [47] detected reduced accumulation of sphinganine in cell lines treated with extracts of tortillas compared with those treated with extracts of raw maize. This was in conjunction with lowered FB1 levels. Here, the sum of FB1 and HFB1 in tortillas (on molar basis) was half of the initial FB1 level detected in the raw maize. The toxic potential was found to be lowered by 60% for extracts of tortillas compared with extracts of unprocessed maize [47]. Similarly, in feeding trials on rats, kidney sphinganine and sphingosine concentrations were not increased or less increased in rats fed a diet containing nixtamalized maize (meal) compared with those fed non-nixtamalized maize (meal) [62,63]. In both studies, nixtamalization was performed with a 1.2% lime solution. Voss et al. [63] used raw maize with three different FB1 contamination levels. Rats that ate a diet containing nixtamalized maize showed no or only week symptoms of nephropathy. This was much different from when the diet contained uncooked maize. Burns et al. [62] additionally included a mock-nixtamalization control (i.e., cooking of maize meal without lime). Similar to the findings of De Girolamo et al. [51], this procedure also lowered the FB1 level, but much less hydrolyzation took place compared with cooking with lime (for details, see Table 2). Both nixtamalization and mock-nixtamalization strongly reduced kidney damage, as well as renal toxicity (evaluated by number of apoptotic tubule cells), compared with uncooked maize meal [62]. Due to the clearly reduced toxic effects caused by nixtamalized maize (products), a significant formation of matrix-associated FB1 or unknown free fumonisin forms that contributed to toxicity was not indicated in these two studies. This differed from a former rat feeding trial performed by Hendrich et al. [64] using highly FB1-contaminated maize (obtained by fungal inoculation). Here, although nixtamalization was able to lower toxicity in some cases, a more pronounced effect would be expected in view of the high loss in FB1 (approximately 98%–100%). When considering the molecular weights, the formed HFB1 amounted to approximately 60%–72% of the initial FB1. In this study, it became further obvious that the nutritional status was capable of impacting toxicological effects caused by fumonisins present in the non-nixtamalized and nixtamalized maize. Hence, more research is needed regarding the potential formation and occurrence of so far undetected and/or unknown toxic fumonisin form(s) in alkali-cooked maize.




*Toxins* **2019**, *11*, 227

**Table 2.** *Cont.*



142



*Toxins* **2019**, *11*, 227

corrected for change in moisture. d: Here, it is not clear if the change in the mycotoxin concentration

basis. ~: Approximate separated by this symbol. FB1/2: fumonisin B1/2; HFB1/2: hydrolyzed fumonisin B1/2; PHFB1/2: partially hydrolyzed FB1/2. accum.: accumulation;

n.s.: not significant; O/N: overnight; orig.: original.

 values that were calculated for this overview by using the data provided in the cited literature. |: Here, individual data of two production runs are given and

 is corrected for change in moisture. e: As equivalent to parent form. f: On molar

 n.a.: not analyzed; n.d.: not detected;

#### **5. Other Mycotoxins during Nixtamalization and Tortilla Production**

Little data were found on mycotoxins other than aflatoxins and fumonisins during nixtamalization and tortilla production (Table 3). Abbas et al. [65] experimentally produced tortillas (including traditional nixtamalization with 2% Ca(OH)2) using two batches of maize naturally contaminated with the *Fusarium* toxins ZEN and DON, as well as maize spiked with the purified toxins (by injection into the embryos). Here, no difference in the percentage reduction of mycotoxin was obvious depending on the type of contamination. For ZEN, the initial levels that were present as *trans*-ZEN were lowered by 59%–100%. For the two maize samples with the highest concentrations (one spiked and one naturally contaminated one), some ZEN (<0.4% of the total amount) could be detected in the nejayote. Further, some isomerization from *trans*-ZEN to *cis*-ZEN took place for these maize samples. However, most of the ZEN was degraded into undetectable form(s), and it was supposed that the alkaline treatment attacked the lactone ring of ZEN. Whether this transformation would be stable under acidic conditions was not addressed [65]. For DON, reductions amounted to 72%–82%. The naturally contaminated maize batches contained, in addition, the acetylated form 15-acetyl-DON, which was completely destroyed in tortillas. Neither DON nor 15-acetyl-DON could be detected in the nejayote.

The potential to lower ZEN and DON by alkaline steeping of maize was also shown when using 0.1 M sodium carbonate. Here, steeping of raw maize kernels at 22 ◦C for 24 h lowered ZEN and DON by around 45% and 70%, respectively. An extended steeping over 72 h reduced the concentrations by 88% and 95%, respectively [66]. The baking and frying steps in commercial tortilla chip production were analyzed by Scudamore et al. [49] regarding mycotoxin changes on a product 'as is' basis. In the (probably alkaline) maize flour mixture(s) used to prepare the dough for tortillas, ZEN was present at low levels only, and the change during processing was very variable. However, if the initial ZEN level was higher than 13 μg/kg, the reduction amounted to 35%–64%. If the initial level was below 9 μg/kg, the detected change ranged from a 7% reduction to a 116% increase. This was probably caused by difficulties in representative sampling of industrial processes. DON levels in the tortilla chips were on average lowered by 32%, with the highest reductions at the highest initial levels. The sensitivity of DON towards food production processes that involve alkaline additives was also observed in the production of bakery wares and during the cooking of noodles (for an overview, see [67]).

The reduction of the emerging mycotoxin MON during tortilla production was studied by Pineda-Valdes et al. [68]. In pilot-scale experiments, MON was reduced by 97% after cooking of maize kernels in a 0.25% lime solution. After steeping or further processing, MON could not be detected anymore. When determining laboratory-scale processing using fungal-inoculated maize with an around 10-fold higher initial concentration, MON was lowered by 54% during cooking. After steeping and washing, MON reduction accounted for 64% and 69%, respectively. In masa and tortillas, the loss was around 70%. MON was not detected in any of the liquid fractions [68], although it is characterized by low molecular size and high water solubility [8]. Thus, MON might have been either modified into undetected form(s) during the 20 min alkaline cooking step or was degraded due to the action of high temperature and/or high pH. In a former study, Pineda-Valdes and Bullerman [69] demonstrated an affection of MON at elevated pH and temperature. Heating to 100 ◦C in an aqueous environment with pH 10 for 60 min lowered MON by around one half. However, after 20 min of cooking, the MON loss amounted to less than 20%. Therefore, nixtamalization of MON-contaminated maize showed a relatively high efficiency in reducing the concentration of this emerging mycotoxin, probably by a pH of >10 of the lime water.


*Toxins* **2019**, *11*, 227

#### **6. Conclusions and Outlook**

Nixtamalization and tortilla baking can affect mycotoxins in different ways, including physical and chemical action: (1) Water-soluble mycotoxins can leach into the liquid fractions during cooking, steeping, and washing. (2) Mycotoxins present in the pericarp, tip cap, and germ are removed when these tissues are (partly) separated by thorough washing of the nixtamal. (3) Action of high pH and elevated temperature during cooking and baking can result in degradation, modification, and/or binding or release of mycotoxins.

For traditional nixtamalization, a high potential to lower free parent forms of mycotoxins is described (see also Figure 2). Aflatoxin concentrations of raw maize were found to be lowered by around 15%–85% and 20%–100% in the nixtamal and masa, respectively. Tortillas mostly showed aflatoxin reductions of 50%–100%. For FB1, the reduction mainly amounted to around 75%–100% in nixtamal, masa, and tortillas. For ZEN, DON, and MON, reductions of around 60%–100%, 70%–80%, and 70%–100% are described. However, only very limited data is available regarding maize mycotoxins other than aflatoxins and fumonisins. More data on such toxins would help to evaluate the benefits of alkali-processed maize in more detail.

Besides reduction in the free parent forms, modification of mycotoxins can occur, and interaction with matrix compounds can be altered. To analyze such processes during nixtamalization, intense efforts have already been undertaken to establish and optimize appropriate detection methods. Although it must be noted that, when analyzing food matrices, which can harbor several challenges, analytical recovery must in general be taken into account, and data should be corrected accordingly (which was often not done or at least not mentioned for the data presented in the current review). Careful conclusions on the reduction factors of mycotoxins must certainly also take a reasonable contribution of variability and uncertainty into account. Furthermore, the stability of the present forms and their bioavailability need to be considered when analyzing toxicological impacts. In addition, precise knowledge on the critical parameters in nixtamalization and tortilla production is important to optimize production procedures to furthermore reduce potential health risks to the consumers, e.g., by reconversion of aflatoxins after consumption.

In general, further research is needed to evaluate possible modifications and matrix–mycotoxin interactions during nixtamalization, as well as the occurrence and potential toxicity of the formed structures in the final food items. In doing so, a possible reconversion and/or release of parent forms in the gastrointestinal tract, as well as by activity of the gut microflora, need to be considered. Reliable analytical data would be the basis for precise understanding of the processes and the factors in mycotoxin reduction. Moreover, knowledge on the fate of mycotoxins and their toxicity is required to evaluate possible utilization strategies for the nejayote.

**Figure 2.** Graphical overview of mycotoxin changes during nixtamalization and tortilla production. (**a**) Free parent mycotoxins. (**b**) Sum of free parent form(s) and detected modified/ matrix-associated form(s). Columns indicate (approximate) ranges; lines represent (approximate) mean values of changes in mycotoxin concentrations from raw maize material to tortillas or intermediate products. The numbers on the x-axis refer to studies described in the literature, which are listed in Tables 1–3. More than one number can refer to the same reference if different process conditions or technologies were compared. For details on the studies (including references, processing parameters, and mycotoxin content in raw maize material), see Tables 1–3. \*: Only mixing of kernels with hot lime water, without further cooking. \*\*: Microwave cooking. \*\*\*: Extrusion cooking. Total AFs: aflatoxins B1 + B2 + G1 + G2; Total AFs (acid): aflatoxins B1 + B2 + G1 + G2 detected in acidified samples/extracts; AFB1/2: aflatoxin B1/2; AFB1 + 2 (acid): aflatoxin B1 + B2 detected in acidified samples/extracts; AFB1-diol: aflatoxin B1 dihydrodiol; AFG1/2: aflatoxin G1/2; AFM1: aflatoxin M1; FB1/2/3: fumonisin B1/2/3; (P)HFBs: partly + fully hydrolyzed fumonisins B1 + B2; (P)HFB1: partly + fully hydrolyzed fumonisin B1; ZEN: zearalenone; DON: deoxynivalenol; 15-ac-DON: 15-acetyl-deoxynivalenol; MON: moniliformin.

**Funding:** This work was performed as part of the MyToolBox project (www.mytoolbox.eu), which has received funding from the European Union's Horizon 2020 research and innovation program under grant agreement no. 678012. The article reflects only the authors' view, and the European Commission is not responsible for any use that may be made of the information it contains.

**Acknowledgments:** We appreciate the work of all colleagues on this topic and apologize to those whose studies were not cited here.

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

#### **References**

1. Pitt, J.; Wild, C.; Baan, R.; Gelderblom, W.; Miller, J.; Riley, R.; Wu, F. *Improving Public Health through Mycotoxin Control. Chapter 1: Fungi Producing Significant Mycotoxins*; International Agency for Research on Cancer, WHO Press: Geneva, Switzerland, 2012; pp. 1–30.


© 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/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Toxins* Editorial Office E-mail: toxins@mdpi.com www.mdpi.com/journal/toxins

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18