**Mycotoxins in Feed and Food Chain Present Status and Future Concerns**

Printed Edition of the Special Issue Published in *Toxins* Filippo Rossi Edited by

www.mdpi.com/journal/toxins

## **Mycotoxins in Feed and Food Chain**

## **Mycotoxins in Feed and Food Chain Present Status and Future Concerns**

Editor

**Filippo Rossi**

MDPI • Basel • Beijing • Wuhan • Barcelona • Belgrade • Manchester • Tokyo • Cluj • Tianjin

*Editor* Filippo Rossi Catholic University Italy

*Editorial Office* MDPI St. Alban-Anlage 66 4052 Basel, Switzerland

This is a reprint of articles from the Special Issue published online in the open access journal *Toxins* (ISSN 2072-6651) (available at: https://www.mdpi.com/journal/toxins/special issues/mycotoxins feed food chain).

For citation purposes, cite each article independently as indicated on the article page online and as indicated below:

LastName, A.A.; LastName, B.B.; LastName, C.C. Article Title. *Journal Name* **Year**, *Article Number*, Page Range.

**ISBN 978-3-03936-874-7 (Hbk) ISBN 978-3-03936-875-4 (PDF)**

c 2020 by the authors. Articles in this book are Open Access and distributed under the Creative Commons Attribution (CC BY) license, which allows users to download, copy and build upon published articles, as long as the author and publisher are properly credited, which ensures maximum dissemination and a wider impact of our publications.

The book as a whole is distributed by MDPI under the terms and conditions of the Creative Commons license CC BY-NC-ND.

## **Contents**



Reprinted from: *Toxins* **2019**, *11*, 640, doi:10.3390/toxins11110640 .................. **185**

## **About the Editor**

**Filippo Rossi** was born in Piacenza (1963) and graduated in Agricultural Sciences in 1989 (Catholic University, of Piacenza, Italy). In 1992, the Italian Ministry of Research awarded him with a Ph.D. in Molecular Biotechnology. Since 1993, he has been working at the Department of Food Science and Nutrition of the Faculty of Agricultural Sciences in Piacenza, where he is in charge of the Human Nutrition course in the Food Science and Technology degree. His research interests cover the role of food in the prevention of noncommunicable disease, particularly Mediterranean Diet and breast cancer, mycotoxins and liver cancer, dairy foods and hypertension, and starch digestion and obesity. Since 2004 (EU fundend project "Safe Food"), he has been involved in research regarding the prevention of mycotoxins contamination in the feed and food chain.

### *Editorial* **A Long Road to Safer Food**

#### **Filippo Rossi**

Section of Food Science and Nutrition, Department of Animal Sciences, Food and Nutrition, Faculty of Agricultural, Food and Environmental Sciences, Catholic University of Sacred Heart, Via Emilia Parmense 84, 29122 Piacenza, Italy; filippo.rossi@unicatt.it

Received: 17 June 2020; Accepted: 10 July 2020; Published: 14 July 2020

As a side effect of food production, mycotoxins have always accompanied humanity, even if the danger posed by these molecules has only recently been understood and new research has begun to identify and study ways to reduce their presence in food.

This Special Issue of *Toxins* includes papers on new findings concerning well-known mycotoxins, results of studies regarding emerging mycotoxins, such as alternaria and botryodiplodin, and new techniques to reduce mycotoxin contamination in processed cereals.

Reliable data on the presence of mycotoxins in food is very important in the toxicological evaluation of the risk associated with these toxic fungal compounds. Two papers cover this subject: Quevedo-Garza et al. [1] analyze Mexican infant formula food for aflatoxin M1 and Zentai et al. [2] determine the fumonisins in Hungarian maize-based food.

*Fusarium* spp., together with *Aspergillus* spp., are the most relevant fungi genus responsible for mycotoxin production. Researchers have focused their attention on cereals, while neglecting other crops. A paper from a Chinese group reports on the identification of the *Fusarium* species causing sweet pepper fruit rot and on the kinds of mycotoxins produced by these microorganisms [3].

A new toxic molecule produced by a fungal parasite of soybean is the focus of two papers from Abbas et al. who investigate the production of botryodiplodin [4] and its toxicity [5], while another contribution [6] considers secalonic acids, which are the main ergot ergochromes in overall ergot toxicity.

We have observed not only the appearance of new or emerging mycotoxins, but also of new foods, such as insects, that can also be contaminated by mycotoxins. On this topic, a paper in this Special Issue studies the metabolism of aflatoxin B1 in the larvae of the black soldier fly (*Hermetia illucens*) [7].

The reduction of mycotoxin contamination can be obtained by intervening during the cultivation or storage of products. Research carried out by Giorni et al. [8] tested the efficacy of the fungicide azoxystrobin on fungal parasites of rice and obtained a strong reduction (−67%) of sterigmatocystin while deoxynivalenol remained unaffected.

A clear reduction in *Fusarium*-produced toxins can be observed in the paper of Brodal et al. [9], by sieving oat grains and removing broken kernels, which are more contaminated than intact ones.

The last research article of the Special Issue describes an analytical method for the detection of 19 mycotoxins and three phytoestrogens in fish feed and fish meat [10]. The reduction of the risk posed to human health by mycotoxins requires the development and validation of reliable methods to monitor mycotoxins in feed and food.

The three reviews included in the Special Issue cover as many topics. Issues related to the use of lactic acid bacteria as aflatoxin binders in developing countries are discussed in the review of Ahlberg et al [11]. Kamle et al. [12] summarize the effect of fumonisin on human health and the strategies to reduce the level of this toxin in food. A group of emerging mycotoxins, those produced by *Alternaria*, is the focus of Crudo et al [13], who analyze "the most relevant data concerning the occurrence and toxicity of mycotoxins produced by *Alternaria* spp., ( ... .) alone or in combination with other mycotoxins and bioactive food constituents".

In conclusion, all the contributions to this Special Issue expand our current knowledge and, as Guest Editor, I am happy and proud to present this issue to the community of scientists involved in research on mycotoxins.

All research and review articles proposing novelties and overviews, respectively, were successfully and carefully selected for this Special Issue after rigorous revision by the expert peer reviewers. As the Guest Editor, I would like to express my deep appreciation to all the selfless and fair reviewers.

**Acknowledgments:** The editor would like to thank all the authors who contributed to this Special Issue and the reviewers for their evaluation work. The editor is also grateful to the MDPI management team for their valuable support.

**Conflicts of Interest:** The author declares no conflict of interest.

#### **References**


© 2020 by the author. 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* **Aflatoxin M1 Determination in Infant Formulae Distributed in Monterrey, Mexico**

**Patricia A. Quevedo-Garza 1, Genaro G. Amador-Espejo 2, Rogelio Salas-García 1, Esteban G. Ramos-Peña <sup>1</sup> and Antonio-José Trujillo 3,\***


Received: 27 December 2019; Accepted: 31 January 2020; Published: 4 February 2020

**Abstract:** The occurrence of aflatoxin M1 (AFM1) in infant formulae commercialized in the metropolitan area of Monterrey (Nuevo León, Mexico) was determined by using immunoaffinity column clean-up followed by HPLC determination with fluorimetric detection. For this, 55 infant formula powders were classified in two groups, starter (49 samples) and follow-on (6 samples) formulae. Eleven of the evaluated samples (20%) presented values above the permissible limit set by the European Union for infant formulae (25 ng/L), ranging from 40 to 450 ng/L. The estimated daily intake (EDI) for AFM1 was determined employing the average body weight (bw) of the groups of age in the ranges of 0–6 and 6–12 months, and 1–2 years. The results evidenced high intake values, ranging from 1.56 to 14 ng/kg bw/day, depending on the group. Finally, with the EDI value, the carcinogenic risk index was determined, presenting a high risk for all the evaluated groups. Based on these results, it is a necessary extra effort by the regulatory agencies to reduce the AFM1 presence in infant formulae consumed in Mexico.

**Keywords:** AMF1; infant formulae; estimated daily intake; carcinogenic risk index; Monterrey (Mexico)

**Key Contribution:** Aflatoxin B1 can be metabolized by mammals to aflatoxin M1 (AFM1), a form that retains potent carcinogenicity and which can be excreted into milk. There is scarce information on the occurrence of AFM1 in milk and dairy products, and no data are available in Mexico concerning infant formulae contamination by this mycotoxin. The results of the present study further demonstrate the potential risk for the infant population associated with the AFM1 presence in the infant formulae marketed in Monterrey (Mexico).

#### **1. Introduction**

Aflatoxins' presence in food products is one of the major health concerns of the regulatory agencies around the world. These toxins include around 20 metabolites produced by molds such as *Aspergillus flavus* and *A. parasiticus*, which is the most important of the aflatoxin B1 (AFB1) and is normally found in foods, especially those having high carbohydrate and/or fat contents [1]. Its occurrence has been reported in numerous food and feedstuff, including cereals and cereal-derived products [2].

Cattle feed with contaminated crops of AFB1 may lead to the formation of a hydroxylated metabolite named aflatoxin M1 (AFM1), which is excreted in the milk of lactating animals and whose name is due to the source detected [3]. Numerous researchers have reported a linear relationship of about 0.3–6.2% between the amount of AFM1 detected in milk and AFB1 in feed consumed by the animals [4]. Nevertheless, the percent of AFM1 excreted depends on various factors, including concentration of AFB1 in feed, milk yield, stage of lactation and breed [5].

Even though AFM1, the main monohydroxylated derivate of AFB1, presents less carcinogenic and mutagenic activity than AFB1, it exhibits a high level of genotoxic activity and certainly represents a health risk because of its elevated possibility of accumulation and binding to DNA [6]. Based on this, different health agencies such as the World Health Organization and the International Agency for Research on Cancer (IARC) have published articles in which AFM1 is a strong genotoxic and hepatotoxic agent [7]. Therefore, AFM1 has been evaluated as a possible human carcinogen agent, and although until 2002 it was classified in the 2B Group, with a tolerable daily intake (TDI) of 2 ng/kg bw [8], based on numerous scientific evidence that demonstrated carcinogenic and other (teratogenic, genotoxic and immunosuppressive) effects, it was reclassified into the first group [7].

Hence, the elimination of risk sources represents a major assignment for government agencies and food processors, not only for the contaminated products directly consumed by humans but also in feeding cattle that consume contaminated crops, whose products can reach the human being. Government regulations around the world concerning AFM1 limits differ from one other. The lowest AFM1 concentration was approved by the European Union (EU) and the Codex Alimentarius, fixing a maximum admissible level of 50 ng/L in fluid milk and dried or processed milk products [9,10]. On the contrary, higher AFM1 concentrations (500 ng/L) are permitted in the United States of America (USA) and some Latin American countries (such as Mexico and the MERCOSUR agreement), and China allows a maximum limit of 62.5 ng/L [11–14]. However, because of the higher susceptibility of infants to AFM1, the EU and the Codex Alimentarius fixed the maximum admissible level of 25 ng/L for infant formulae, follow-on formulae and dietary foods for medical purposes intended specifically for infants [9].

Another major problem concerning the presence of AFM1 in milk is the different dairy products it's included in (e.g., liquid milk, yogurt, cheese, milk powder, ice cream, regular cream, among others) and the fact that the aflatoxin cannot be eliminated by regular heat treatments such as pasteurization or ultra-high temperature processing [15]. Besides, one of the most important products manufactured from milk are the infant formulae, in which there is significant risk of AFM1 intoxication because small amounts of this toxin in the product may represent an important portion of aflatoxin intake [16].

Despite the danger associated to the AFM1 presence in milk, only a few articles are available regarding the presence of this toxin in milk and dairy products in Mexico [17,18], and no studies have been published regarding its presence in infant formulae or intake assessment for AFM1 in the country. Based on this, the aim of this study was to evaluate the AFM1 occurrence in infant formulae and to estimate the exposure of infant milk consumers to AFM1 by means of a sampling of the infant formulae brands distributed in Monterrey (Nuevo León, Mexico).

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

#### *2.1. Occurrence of AFM1 in Infant Formulae*

Table 1 shows the results obtained from the analyzed samples, with 20% of them being positive for the toxin in a range of 40 to 450 ng/L, and an average AFM1 concentration of 40 ± 99 ng/L for all analyzed samples, which is higher than the limit established for AFM1 in infant formulae by the Codex Alimentarius (25 ng/L) [11]. Nevertheless, when the infant formulae were evaluated separately (starter and follow-on groups), it can be observed that the AFM1 values increased from one group to another. In the starter formulae, the percentage of samples exceeding the AFM1 limit was 14%, remarkably lower than the percentage of samples above the limit in the follow-on formulae (67%). Furthermore,

the media in the starter formulae (20 ± 67 ng/L) was below the EU or Codex Alimentarius AFM1 limit (25 ng/L), compared to the follow-on formulae, with an average (180 ± 185 ng/L) exceeding the AFM1 limit. Although the AFM1 levels in starter formulae were significantly (*p* < 0.05) lower than those in follow-on formulae, it is important to notice the small number of samples evaluated in the follow-on formulae, compared to the infant formula evaluated in the starter group.


**Table 1.** Aflatoxin M1 presence in infant formulae.

\* Value in parentheses indicates the samples percentage above the limit set by the Codex Alimentarius (25 ng/L) with respect to the total. a,b Different online letters indicate significant mean differences among the different types of infant formulae (*p* < 0.05).

Regarding legislation about AFM1 limits in infant formulae, most of the countries do not have an established limit, which is the case of most of the Latin-American countries (including Mexico), which tends to apply the limit established by the Codex Alimentarius or the EU regulation (25 ng/mL) [11,19].

The occurrence of AFM1 in infant formulae varies in different countries. Gomez-Arranz and Navarro-Blasco [20] evaluated the presence of AFM1 in infant formulae in Spain, testing 69 samples and detecting the presence of AFM1 in 26% of them. In this case, all the detected samples were below the EU established limit. More recently, Akhtar et al. [21] determined the AFM1 presence in infant formulae in Pakistan, evaluating 13 samples, in which 53.84% of the samples were positive to the toxin presence and 30.76% exceeded the EU limit. Kanungo and Bhand [22] evaluated the AFM1 presence in infant formulae in India, determining that in 72 evaluated samples, all of them were above the EU permitted limit (25 ng/kg) and 75% of the samples exceeded the USA and Indian Food regulation limit (500 ng/kg). Er et al. [4] published a study evaluating the AFM1 presence in infant formula in Turkey, evaluating 84 samples with only one sample positive for the toxin. In this sense, Li et al. [14] detected the presence of AFM1 in powder base for infant formulae in China, evaluating a total of 1207 samples, with 56 samples being positive for the toxin without passing the Chinese limit (62.5 ng/kg). Awaisheh et al. [23] determined the AFM1 content in infant formulae (120 samples; 48 starter and 72 follow-on formulae) distributed in Jordan, with 58 positive samples for the toxin presence, with a media of 69 and 84 ng/kg for the starter and follow-on formulae, respectively.

#### *2.2. Infant Formulae Daily Intake by Age Group*

The present study is the first evaluation of the daily intake by Mexican minors, based on average consumption and body weight (Table 2). The Mexican Standard NOM-031-SSA2-1999 [24] classifies infants in two groups of infant formulae consumption: i) minor lactating (0–12 months), and ii) major lactating (one to two years). The consumption in these groups is starter and follow-on formulae for the first and the second year, respectively.

Based on the occurrence of AFM1 in infant formulae and the body weight of infants, the estimated daily intake (EDI) for AFM1 was in a range of 1.56 to 14 ng/kg bw per day, which represents the values estimated for one year-old infants when they are fed with starter or with follow-on formulae, respectively. However, when major lactating groups gain weight and reduce the follow-on formula intake (i.e., two years old), the EDI is reduced up to 4.28 ng/kg bw/day. Awaisheh et al. [23] have evaluated the infant formulae consumed in Jordan, presenting an EDI of 1.57 and 1.55 ng/kg bw/day for infants aged six and 12 months, respectively. On the other hand, Ismail et al. [25] reported an EDI value of 4.1 ng/kg bw/day for children aged one to three years in Pakistan. It is considerable the work developed by the food agencies seeking to reduce the presence of AFM1 in milk and infant

formulae. In this sense, Oliveira et al. [26] published an article evaluating the presence of AFM1 in infant formulae in Brazil with a daily intake of 22 ng/kg bw/day. In contrast, almost 20 years later, Ishikawa et al. [27] determined the AFM1 presence in infant formulae in the same country, presenting an important reduction in EDI values (0.078–0.306 ng/kg bw/day). Likewise, lower EDI values than the present study were detected in infant formulae consumed in Spain (n = 69) (0.02–0.13 ng/kg bw/day) [4]. Further, Ruangwises et al. [28] evaluated AFM1 presence in milk powder distributed in Thailand (90 samples) showing EDI values of 0.16 ng/kg bw/day in milk consumed by infants up to three years.

**Table 2.** Estimated aflatoxin M1 daily intake by average body weight and carcinogenic risk index (CRI) in children population based on the ENSANUT (2012).


\* According to the Kuiper-Goodman equation [8].

Comparing the results of AFM1 occurrence in infant formulae and in breast milk in Mexico, the results are quite similar. Thereby, Cantú-Cornelio et al. [29] evaluated the presence of AFM1 in breast milk of nursing mothers in central Mexico (112 samples), with an EDI value of 2.35 ng/kg bw/day, comparable results to the values obtained in the present study. These results show the importance of evaluating the presence of AFB1 in different products consumed by nursing mothers in order to reduce the toxin that may be transformed into AFM1 and reach infants by breast milk.

Table 2 also presents the result of the carcinogenic risk index (CRI) for the evaluated population. At this day, up to our knowledge, no CRI study evaluating the infant population of Mexico has been published. The AFM1 ingestion obtained in this study was greater than the TDI value (2 ng/kg bw/day) calculated by Kuiper-Goodman [8] dividing the TD50 by the safety factor 5000, indicating that there is a potential high risk for liver cancer due to the consumption of infant formulae in Mexican consumers groups studied.

#### **3. Conclusions**

The results of the current study have shown a high presence of AFM1 in infant formulae distributed in the Monterrey (Mexico) metropolitan area. From fifty-five samples evaluated, 20% exhibited a toxin content above the EU and Codex Alimentarius limit (25 ng/L), presenting a range of 40–450 ng/L. Further, in classifying the samples by the type of infant formulae and infant age for consumption (starter formula for minor infants up to one year, and follow-on formula for major infants between one and two years), different levels of AFM1 were obtained (20 ng/L for starter and 180 ng/L for follow-on formulae). Besides, based on the average body weight of the evaluated groups, the EDI value was calculated, with values in the range of 1.56–14 ng/kg bw/day. Finally, with the EDI data, the CRI was determined, obtaining a result of risk in all the evaluated groups. Based on these results, an important effort should be carried out by the regulatory agencies and milk producers in order to reduce AFM1 levels in milk in general, and, in particular, in batches that will be employed for infant formulae elaboration because of the high cancer risk associated with AFM1 presence and the infant consumers' vulnerability.

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

#### *4.1. Sample Collection*

Fifty-five infant formula samples from drug stores and supermarkets sold in Monterrey (Nuevo León, Mexico) were obtained. From these, 49 were starter formulae (0–12 months) and 6 were follow-on formulae (1–2 years). Among the starter formulae, 6 were pre-term formulae (formulated for prematurely born, regurgitation episodes by immature esophageal sphincter, or low birth weight infants), 11 were hypoallergenic formulae (specialized formula based on casein, whey or soy protein hydrolysates) and 9 were lactose free formulae (designed for lactose intolerant infants based on lactose hydrolysis by β-galactosidase or formulated from soy protein isolates).

All formulae were supplied as powder milks. Infant formula containers (cans or bags) were stored in dark at room temperature until analyses were performed.

#### *4.2. Sample Preparation*

Powder-based formula samples were suspended in deionized warm water according to the manufacturer instructions. The method used for sample preparation and AFM1 determination was that specified by the method ISO 14,501 [30]. Suspended infant formula samples were centrifuged at 4200 × *g* for 15 min to separate and remove the milk fat. Aliquots of skimmed milk (50 mL) were filtered (Whatman no. 4 filter paper) and slowly passed (1–2 drops/s) through an immunoaffinity column (AflaM1 HPLC, VICAM, Milford, MA, USA) fitted on a vacuum manifold, and washed twice with 10 mL of distilled water. Thereafter, the AFM1 was eluted with 4 mL of acetonitrile, allowing a time contact of at least 60 s. The eluate was collected in amber vials, the solvent was evaporated in a water bath at 40 ◦C with nitrogen, and the residue reconstituted in water:acetonitrile (67:33) and filtered by Millipore filters (0.45 μm) in amber vials.

#### *4.3. HPLC Analysis*

The HPLC analysis was carried out in a Varian HPLC model 9012 (Agilent Technologies, Santa Clara, CA, USA) connected with a fluorescence detector Varian ProStar (Agilent Technologies Santa, Clara, CA, USA). The separation column was a Phenomenex C18 with 4.5 × 250 mm and 5 μm of particle size (Phenomenex, Torrance, CA, USA). Water and acetonitrile mixture were used as a mobile phase in a proportion of 67:33 (v/v), at a flow rate of 1 mL/min, and an injection volume of 100 μL. Fluorometric detection was achieved at 360 nm excitation and 440 nm emission wavelength.

To assess the performance of the analytical method, linearity, limits of detection (LOD) and quantification (LOQ), recovery and precision (repeatability) were studied. Linearity was evaluated using standard calibration curves that were constructed by plotting the peak area versus the analyte concentration. The calibration curves were established using eight levels of concentrations from LOQ to 100 times LOQ. The regression curve obtained was y = 287.78 x + 75.10 giving appropriate value for the linearity (R2 = 0.998). LOD (2 ng/L) and LOQ (5 ng/L) were calculated as the sample blank value plus 3 and 10 times its standard deviation, respectively. In order to determine the recovery, reconstituted milk was added with 3 levels of AFM1 concentrations (50, 100 and 200 ng/L). The obtained values of recovery were between 83% and 104%. The precision (15.18%) was calculated as repeatability by means of triplicates in each of the levels analyzed in the recovery assay.

#### *4.4. Determination of AFM1 Exposure in the Population*

The determination of the exposure level or estimated daily intake (EDI) in the population of Monterrey to the AFM1 due to the consumption of infant formulae was carried out by combining data on the average daily consumption of milk by groups of age, with the average concentration of AFM1 found in this work, as well as the average body weight (bw) of the population by age groups. For this, Equation (1) was applied:

$$\text{Estimated AFM daily intake} \left(\frac{\text{ng}}{\text{kg bw}} / \text{day}\right) = \frac{\text{Milk intake (L)} \times \text{AFM1} \left(\frac{\text{ng}}{\text{L}}\right)}{\text{Body weight (kg)}} \tag{1}$$

where: Milk intake is the average amount of milk that the infant population ingests daily, expressed in liters. AFM1 is the average concentration of AFM1 contained in the analyzed samples, expressed in ng/L. Body weight is the bw average in the population by age groups in kilograms.

The data corresponding to the daily milk consumption by age groups was obtained from the National Survey of Health and Nutrition of Mexico (ENSANUT) [31], in the section corresponding to Nuevo León State.

In order to obtain the daily intake of AFM1 in the infant population, it was necessary to separate the population, as indicated by Mexican Standard NOM-031-SSA2-1999 [24] in: (1) minor lactating (newborn up to 6 months), at this stage of the infant's life, their diet is only based on breast milk or infant formulae for initiation; (2) minor lactating (from 6 to 12 months), at this stage, ablactation occurs, and the starter infant formulae and dairy infant formulae containing cereals and honey are taken as the infant diet at this stage of life; (3) major lactating (from 12 to 24 months), at this stage the dairy intake is determined by the follow-on formulae and those containing cereals and honey. From the ENSANUT [31] survey, the average weights of the infant population (minor and major lactating) were obtained.

Likewise, the CRI was estimated based on the proposal of Kuiper-Goodman [8], which estimates the TDI of AFM1 by dividing the TD50 (threshold dose by body weight; 10,380 ng/kg bw per day for AFM1) by the safety factor 5000, to give an estimated value of 2 ng/kg bw per day. A CRI of AFM1 higher than 2 ng/kg bw indicates liver cancer risk to consumers [8,32].

#### *4.5. Statistical Analysis*

All infant formulae were analyzed in duplicates. Collected data were statistically evaluated using the nonparametric Wilcoxon rank sum test with continuity correction of R Core Team (Vienna, Austria) [33]. AFM1 concentrations were expressed as mean ± standard deviation in order to show the occurrence of the toxin in infant formulae.

**Author Contributions:** Research concept, design and supervision: P.A.Q.-G. and A.-J.T.; HPLC method validation: P.A.Q.-G. and R.S.-G.; AFM1 analysis in infant formulae: P.A.Q.-G. and E.G.R.-P.; writing and correcting of the manuscript: P.A.Q.-G.; A.-J.T.; R.S.-G.; E.G.R.-P.; G.G.A.-E. All authors have read and agreed to the published version of the manuscript.

**Funding:** The authors appreciate the funding provided by the Universidad Autónoma de Nuevo León, Monterrey, México, through the PAICYT – UANL program, and by the CIRTTA of Universitat Autònoma de Barcelona, Bellaterra, Spain, for this study.

**Acknowledgments:** The authors are very grateful to Jesús Piedrafita (UAB, Spain) for his support in statistical analysis.

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

#### **Abbreviations**


#### **References**

1. Prandini, A.; Tansini, G.; Sigolo, S.; Filippi, L.; Laporta, M.; Piva, G. On the occurrence of aflatoxin M1 in milk and dairy products. *Food Chem. Toxicol.* **2009**, *47*, 984–991. [CrossRef] [PubMed]


© 2020 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 and Risk Assessment of Fumonisin B1 and B2 Mycotoxins in Maize-Based Food Products in Hungary**

#### **Andrea Zentai 1, Mária Szeitzné-Szabó 2, Gábor Mihucz 2, Nóra Szeli 2, András Szabó 2,3,\* and Melinda Kovács 2,3**


Received: 24 October 2019; Accepted: 3 December 2019; Published: 5 December 2019

**Abstract:** Fumonisins are toxic secondary metabolites produced mainly by *Fusarium verticillioides* and *Fusarium proliferatum.* Their toxicity was evaluated, and health-based guidance values established on the basis of both Joint FAO/WHO Expert Committee on Food Additives (JECFA) and European Food Safety Authority (EFSA) recommendations. This study presents the results of fumonisin analyses in different maize- and rice-based food products in Hungary and the potential health risk arising from their dietary intake. In total, 326 samples were measured in 2017 and 2018 to determine fumonisins B1 and B2 levels. Three-day dietary record data were collected from 4992 consumers, in 2009. For each food category, the average concentration values were multiplied by the relevant individual consumption data, and the results were compared to the reference values. With respect to the maximum limits, one maize flour, two maize grits, and two samples of other maize-based, snack-like products had total fumonisin content minimally exceeding the EU regulatory limit. The mean daily intake for all maize-product consumers was 0.045–0.120 μg/kg bw/day. The high intake (95 percentile) ranged between 0.182 and 0.396 μg/kg bw/day, well below the 1 μg/kg bw/day tolerable daily intake (TDI) established by EFSA. While the intake calculations resulted in comforting results, maize-based products may indeed be contaminated by fumonisins. Therefore, frequent monitoring of fumonisins' levels and evaluation of their intakes using the best available data are recommended.

**Keywords:** fumonisin; human exposure; maize products

**Key Contribution:** Fumonisin intake of Hungarian consumers from maize-based products is below the health-based guidance values. Maize-based products may be contaminated by fumonisins

#### **1. Introduction**

Fumonisins are secondary metabolites produced mainly by *Fusarium verticillioides* and *Fusarium proliferatum* [1]. Maize and maize-based products are most commonly contaminated by fumonisins, but fumonisins can be detected in several other cereal grains, such as rice, wheat, barley, rye, and oat [2,3]. More than 15 fumonisin homologues have been described, including fumonisin A, B, C, and P, and, among them, fumonisin B1 (FB1), FB2, and FB3 are the most frequent naturally occurring fumonisins [1,4]. FB1 typically accounts for 70%–80% of the total fumonisin produced, while FB2 usually makes up 15%–25% and FB3 3%–8% when cultured on maize, rice, or in liquid medium [1].

Among fumonisins, FB1 is the most toxic compound and has been shown to promote tumour growth in rats as well as equine leukoencephalomalacia [5] and porcine pulmonary oedema [6]. It was classified by the International Agency for Research on Cancer (IARC) in Group 2B (possibly carcinogenic in humans) [7]. FB1 also causes chronic liver and kidney toxicity when administered in repeated doses to rodents.

Fumonisin B toxins, as structural analogues of sphingoid bases, inhibit ceramide synthases, causing the disruption of the sphingolipid metabolism and leading to sphinganine (and sphingosine) accumulation in cells and tissues [8]. Toxicity studies have mainly focused on the effects of FB1, but FB2–4 appear to have similar toxicological profiles. Acute toxicity is not relevant for fumonisins.

The Scientific Committee on Food (SCF) as well as the European Food Safety Authority (EFSA) in Europe and the Joint FAO/WHO Expert Committee on Food Additives (JECFA) evaluated the dietary risk of fumonisin intakes [9–15].

The SCF established a tolerable daily intake (TDI) of 2 μg/kg bw/day for FB1 in 2000, based on an overall level of no observed adverse effect (NOAEL) of 0.2 mg/kg bw for liver and kidney in rodents [9]. This TDI was expanded into group TDI in relation to the total amounts of fumonisin B1, B2, and B3, alone or in combination [10]. JECFA published a risk assessment on FB1, FB2, FB3 in 2001 [11]. The assessment was essentially based on FB1 data, and the other toxins were considered as having similar toxicological profiles. A group provisional maximum tolerable daily intake (PMTDI) of 2 μg/kg bw/day per day was allocated based on a NOAEL of 0.2 mg FB1/kg bw per day for renal toxicity in a subchronic and a chronic rat study [11]. The PMTDI established by JECFA was retained in 2011 and in 2016 as well [12,13].

EFSA discussed food safety issues of mycotoxins, including fumonisins, in several documents. The chemical structure of mycotoxins can be altered by the defense reaction of plants, rendering them extractable conjugated and/or non-extractable bound mycotoxins or mycotoxin metabolites. Since these modified toxins are usually not detected during the analysis of mycotoxins, they are commonly termed "masked" or "bound". EFSA issued a scientific opinion in 2014 regarding certain modified mycotoxins in food and feed [14].

More recently, EFSA published a scientific opinion on the appropriateness to set a group health-based guidance value for fumonisins and their modified forms in 2018 [15]. For the establishment of the TDI, the benchmark dose lower confidence limit (BMDL10) of 0.1 mg/kg bw per day for induction of megalocytic hepatocytes in mice was used. Taking into account an uncertainty factor (UF) of 100 for intra- and interspecies variability, the TDI was established at 1.0 μg FB1/kg bw per day. FB2, FB3, and FB4 were included in the TDI, based on structural similarity and the limited available data indicating similar mode of action (MoA) and toxic potencies.

In Europe, Commission Regulation (EC) No 1881/2006, setting the maximum levels for certain contaminants in foodstuffs, established the maximum limits for fumonisins (sum of B1 and B2) in different commodities, including unprocessed maize, maize intended for direct human consumption, maize-based foods for direct human consumption, maize-based breakfast cereals, maize-based snacks, processed maize-based foods and baby foods for infants and young children, different milling fractions of maize, and other maize milling products not used for direct human consumption. Table 1 presents the specified maximum limits by commodities.

This article presents the results of fumonisin analyses in different maize- and rice-based food products in Hungary and, consequently, the potential health risk arising from their dietary intake.


**Table 1.** Maximum limits (μg/kg) for fumonisins established by Commission Regulation (EC) No 1881/2006.

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

#### *2.1. An Overview of the Measured Fumonisin Content*

Altogether, 326 samples were measured for fumonisins B1 and B2 mycotoxins levels. The types of samples were from the food categories of maize flour, maize grits, corn flakes, canned maize, other maize-based, snack-like products, white and brown rice, and other rice-based products. The limit of detection (LOD) and the limit of quantification (LOQ) for FB1 were 0.031 and 0.093 mg/kg, while those for FB2 were 0.051 and 0.154 mg/kg.

In total, 70 and 256 samples were analyzed in 2017 and 2018, respectively, and were considered together in our assessment.

We measured 64 maize flour samples, of which 33 (51.6%) had detectable FB1 content, and 6 (9.4%) had detectable FB2 content. The highest FB1 value was 1.46 mg/kg. The average FB1 and FB2 concentrations were 0.17–0.20 mg/kg for FB1 and 0.05–0.10 mg/kg for FB2. In no instance was FB2 detected if FB1 was undetected. Only in six cases, both FB1 and FB2 were detected at a measurable level (above LOQ), while FB2 was never detected alone.

Then, 62 maize grits were analyzed; 26 samples (41.9%) presented detectable FB1, and 4 (6.5%) detectable FB2. The highest concentrations found were 1.96 mg/kg for FB1 and 0.58 mg/kg for FB2. The average FB1 content was 0.13–0.16 mg/kg, while the average FB2 content was 0.03–0.08 mg/kg. Four samples contained both FB1 and FB2 above the LOQ.

Altogether, 8 of the 64 corn flakes samples (12.5%) had measurable FB1 content, whereas FB2 was not detectable in any of them. The average fumonisin B1 content ranged between 0.03 and 0.07 mg/kg, and the highest measured value was 0.46 mg/kg.

Only one of the 18 canned maize samples contained measurable FB1, but none of them contained FB2. The relevant FB1 concentration was 0.20 mg/kg.

Fumonisin B1 was measured in 20% of the other maize-based, snack-like products (17 of the 85 samples), and FB2 in only 2 samples. The average FB1 content ranged between 0.07 and 0.10 mg/kg, with a maximum content of 1.1 mg/kg.

Regarding white rice and brown rice samples and other rice-based products, FB1 and FB2 contents were in all cases below the LOQ. These commodities were therefore not included in our further risk assessment. The most important parameters of the analysis results for each food categories are summarized in Tables 2 and 3.

**Table 2.** Classification of the samples analyzed in this study in relation to fumonisins' limit of detection (LOD), limit of quantification (LOQ) \*, and the regulatory limit.


\* LOD and LOQ for FB1: 0.031 and 0.093 mg/kg, LOD and LOQ for FB2: 0.051 and 0.154 mg/kg. Nr: number of samples.



\* Method of mean calculation: Results below the LOD and between LOD and LOQ were taken into account in two ways. In the lower-bound (LB) scenario, 0 and LOD were inserted for values below LOD and between LOD and LOQ, respectively. In the upper-bound (UB) scenario, LOD and LOQ were inserted for values below LOD and between LOD and LOQ, respectively. \*\* There were only two measured values for the concerned food category and mycotoxin; na: not applicable. Nr: number of samples.

Considering these results in light of the current maximum limits, one maize flour, two maize grits, and two samples of the other maize-based, snack-like products (mexicorn and a maize wafer) had total fumonisin contents minimally exceeding the regulatory limit (the sum was calculated according to the upper-bound (UB) scenario in case of a non-detectable value of FB2).

Our results were also compared with fumonisin contents measured and published in the previous decades in Hungary. Fazekas et al. [16] measured considerably high fumonisin concentrations in maize collected during storage and harvesting in 1993 and 1994. Of the moldy maize samples collected in the period of storage, 70.8% contained fumonisin B1 (0.05–19.8 mg/kg; average concentration: 2.6 mg/kg). Fumonisin B1 content measured in maize ears more or less affected by molds (affected

sample), collected in the period of harvesting, ranged between 0.095 and 52.4 mg/kg, with an average content of 6.64 mg/kg in 70% of the samples. Of the "average samples", 30% were contaminated with fumonisin B1 (0.06–5.1 mg/kg; average: 1.52 mg/kg). Fumonisin concentrations were determined by high-performance liquid chromatography methods.

Tóth et al. [17] investigated *Aspergillus* and *Penicillium* species and their mycotoxins in maize in Hungary in two consecutive years after harvest. Mycotoxin concentrations were measured with HPLC–MS technique. Fumonisins (B1 + B2) were observed in quantities exceeding the EU limit in some samples collected in different regions (4.66 mg/kg; 10.15 mg/kg; 5.13 mg/kg; 7.55 mg/kg) in 2010.

The IARC report cites contamination data in maize for Europe, including Hungary. Fumonisin B1 was detected in 248 out of 714 maize samples, at a concentration range of 0.007–250 mg/kg [7]. Similarly, the WHO series of Environmental Health Criteria dealt with fumonisin B1 in 2000 [18]. The report specifically cites the results of the Hungarian authors Fazekas et al. [19], measuring 0.05–75.10 mg/kg fumonisin B1 in 56 out of 92 maize samples.

Comparing our results with those of the above reports, fumonisin contamination in Hungary in recent years seems to be lower than that measured in previous decades. However, our measurements focused on processed food products (targeting the end consumer), which obviously have lower fumonisin contents than unprocessed maize samples.

#### *2.2. Correlation between FB1 and FB2 Levels*

FB2 content was always lower than FB1 content in our samples and was detected only in those samples also containing FB1. The relationship between fumonisin B1 and B2 contents was further analyzed, to understand whether a possible correlation coefficient could be set up.

The commodity groups of at least one sample containing measurable quantities of FB1 and FB2 together were maize flour (six samples), maize grits (four samples), and the other maize-based snacks (two samples). The correlation coefficient calculated for the maize flour commodity group based on the numerical concentrations was 0.95, indicating a strong correlation.

Taking into account all 35 samples where, beside FB1, FB2 was also detected but not measurable (i.e., between LOD and LOQ), the correlation coefficients were 0.79 and 0.77 in the lower-bound (LB) and UB scenarios, respectively. Considering only the pooled maize flour and maize grits samples (26 samples), the correlation coefficient values were 0.86 and 0.82 in the LB and UB scenarios, respectively.

These results suggest a possible correlation between the levels of fumonisins B1 and B2; however, a higher number of samples with measured fumonisin B1 and B2 concentrations would be necessary to draw further conclusions.

#### *2.3. Risk Assessment*

The resulting intake values—both mean and high percentile—were well below the reference values established by EFSA and JECFA. Table 4 presents the calculated population mean and 95 percentile intakes for the five commodity groups (maize flour, maize grits, corn flakes, canned maize, and other maize-based, snack-like products) concerned.

**Table 4.** Calculated mean and 95 percentile (P) for fumonisin intakes (μg/kg bw/day). The percentage of European Food Safety Authority (EFSA) tolerable daily intake (TDI) is included in brackets.


The mean daily intake for all maize-product consumers based on the LB and UB scenarios was 0.045–0.120 μg/kg bw/day. In addition, the high intake (95 percentile) ranged between 0.182 and 0.396 μg/kg bw/day, well below 1 μg/kg bw/day.

Regarding children (aged 0–18 years), the mean intake was 0.056–0.167 μg/kg bw/day, and the high intake (95 percentile) was 0.244–0.537 μg/kg bw/day.

Figure 1 presents the relative and cumulative frequencies of the resulting distributions of total fumonisin intakes for both total consumer population and children. The figure shows that most intakes cumulated below 0.5 μg/kg bw/day.

**Figure 1.** Relative and cumulative frequencies of total fumonisin intakes derived from maize-based products.

The results were compared to those of the exposure assessment conducted by EFSA in 2014 on the occasion of a derogation request for the maximum levels of several mycotoxins, including fumonisins [20]. On the basis of French contamination data of 2013, the mean exposure levels in children groups ranged between 0.17 and 1.52 μg/kg bw/day in the LB scenario and between 0.47 and 2.11 μg/kg bw/day in the UB scenario. The high (95 percentile) exposure levels ranged between 0.54 and 3.44 μg/kg bw/day and between 1.09 and 4.39 μg/kg bw/day in the LB and UB scenarios, respectively. In adult groups, the mean exposure levels were between 0.03 and 0.81 μg/kg bw/day in the LB scenario and between 0.15 and 1.19 μg/kg bw/day in the UB scenario. The 95th percentile, however, ranged between 0.08 and 1.76 μg/kg bw/day in the LB scenario and between 0.31 and 2.30 μg/kg bw/day in the UB scenario.

Our present results are in the same range or—especially in the case of children—considerably lower than reported results (Table 5).

Although the estimated mean and high intakes remained below both the JECFA and the EFSA reference values in all scenarios, it is worth noting that the maximum and some high values (over the 95 percentile) exceeded the 1 μg/kg bw TDI set by EFSA in 2018. In the case of all consumers, these high values amounted to 0.97% of the population, whereas in the case of children, they amounted to 2.36%. The maximum estimated intake value was 1.81 μg/kg bw. These specific high values were predominantly children's intake values, derived mainly from the consumption of canned and sweet maize and other maize-based snack-like products.

Considering that these intake results are based on the actually registered consumptions, representing only 4.8% of the total population and 7.6% of children consumers, the consequent health risk is probably negligible.


**Table 5.** Summary of estimated intakes (μg/kg bw/day) in comparison with EFSA estimations and health-based guidance values.

\* All (adult + children) consumers included in our calculations. JECFA: Joint FAO/WHO Expert Committee on Food Additives, PMTDI: provisional maximum tolerable daily intake.

#### *2.4. Commodity Contributions*

The contributions of different commodities to the summed intake estimated from all maize-based foods are presented in Table 6.

**Table 6.** Contribution of different commodities to total fumonisin intake from maize-based products.


\* Data are shown, but conclusions cannot be made due to extremely low registered consumption.

Considering the LB scenarios, maize-based, snack-like products contributed the most to the fumonisin intake of the total (all consumers) population (43.3%), followed by maize flour (29.1%) and corn flakes (15.1%). In the case of children, the main contributors in the LB scenario were, similarly, maize-based, snack-like products (60.4%), corn flakes (20.2%), and canned maize (11.2%) (see Figure 2).

**Figure 2.** Contribution of food commodities to total fumonisin intake from maize-based products.

#### *2.5. Uncertainty Considerations*

It should be mentioned that this assessment focused only on the intake of fumonisins B1 and B2 from five different maize-based commodity types. Other types and the modified or masked forms of fumonisins were not analyzed. Total fumonisin intake of the population could be somewhat higher, if all relevant (including also non-maize-based) commodity types were considered. However, given that maize is the focal commodity in relation to fumonisin contamination, the contribution of other food products to total fumonisin intake is considered low.

The effect of household food processing on fumonisin content (relevant only for maize flour and grit) was not taken into account in our calculations. While the change of fumonisin content as a result of processing operations was studied by several authors [21–25], and heating was reported to lead to some losses of the toxin, the results from different studies are variable [13]. Our approach might have led to a slight overestimation of exposure, taking into account that the effect of heating would lower the calculated intakes; however, this would not change our conclusions, considering that our results do not indicate serious health concern.

The fact that we took into account only those consumption days for which actual consumptions were registered also adds uncertainty. Given that maize-based commodities are non-staple commodities in Hungary, consumed only occasionally by the majority of the population, averaging the occasionally registered consumption values would be misleading. Similarly, including the zero-consumptions in our assessment would "dilute" the results.

However, it needs to be mentioned that current trends indicate an increase in gluten-free foods consumption, which is not strictly linked to the number of consumers intolerant to gluten. Regular consumers striving for healthy diets may as well choose maize-based foods. These facts highlight the importance of focusing more attention on these kinds of food products, considering that they also tend to be the focal commodities most highly contaminated with fumonisins.

As the consumption data were collected in 2009, certain changes might have occurred since then. In the case newer/more recent consumption data are published, repeating these evaluations would be of great value. In this regard, the consumption of different maize-based products could be studied in more detail. In our calculations, we linked the concentration data of an aggregated "maize-based, snack-like products" group to the consumption of an aggregated maize-based products group, including popped maize or extruded corn flakes. These calculations, however, could be refined by separately studying the consumptions of these specific products. Our measurement results indicate a relatively high contamination rate in this kind of commodity category.

#### **3. Conclusions**

Our calculations based on recent fumonisin analyses in maize-based foods and consumption data from a Hungarian survey produced comforting results. The calculated fumonisin intakes of the total population and of children consumers were well below the reference values established by JECFA and EFSA. The values were also in the same range or lower than the European exposure rates estimated by EFSA in 2014.

However, the recent trend of increasing the consumption of alternative, "healthy" foods, including maize-based commodities, needs to be monitored. Our results suggest that maize-based products may indeed be contaminated by fumonisins. Therefore, monitoring of fumonisins' levels and the frequent re-evaluation of their dietary intakes with the best available data are recommended.

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

#### *4.1. Sampling*

Maize-based products were purchased from the Hungarian market in three metropolitan regions, i.e., Kaposvár (n = 276), Budapest (n = 29), and other cities, e.g., Debrecen, Keszthely, Székesfehérvár (n = 21). Commercial products were collected from supermarkets, retail shops, and pharmacies. A total

amount of 326 samples purchased in 15 months (from August 2017 to November 2018) included maize flour (64), maize grits (62), corn flakes (64), canned maize (18), and other maize-based, snack-like products (85, extruded corn bread, tortilla chips, popcorn, nacho, maize chips, etc.). Beside these, 16 white and 10 brown rice and 7 rice-based products were also sampled. All information about the samples (i.e., producer, distributor, country of origin) was obtained from the products' labels and recorded. Samples were randomly selected, collecting as many as possible leader and minor brands available on the market.

#### *4.2. Laboratory Analysis*

#### 4.2.1. Chemicals

Fumonisin B1 (FB1) and B2 (FB2) were purchased from Merck-Sigma Aldrich (St. Louis, LO, USA). HPLC–MS-grade acetonitrile and water were purchased from Carl Roth GmbH (Karlsruhe, Germany), HPLC–MS-grade acetic acid was purchased from Merck (Darmstadt, Germany).

#### 4.2.2. Sample Preparation

Dry solid samples were ground using an ETA® Vital Blend II blender (ETA a.s., Praha, Czech Republic). Then, 5 g of sample was vortexed for 1 min with 20 mL of acetonitrile/water (50:50) on a VELP ZX-3 desktop vortex (Velp, Usmate, Italy) and 0.1% acetic acid and extracted for 60 min at 420 rotations/min speed on a horizontal desktop shaker (Edmund Bühler SM30A model, Bodelshausen, Germany). The supernatant of the extracted sample was centrifuged for 10 min at 14,000 rpm, and 4 ◦C. Aliquots of 10 μL internal standard solutions (13C-FB1, 6 μg/mL) were added to 970 μL aliquots of the supernatant of the centrifuged sample. The mixture was analyzed with LC–MS.

#### 4.2.3. High-Performance Liquid Chromatography

Liquid chromatography and mass spectrometry (LC–MS) analysis were performed with a Shimadzu Prominence UFLC separation system equipped with an LC–MS-2020 single quadrupole (ultra-fast) liquid chromatographer–mass spectrometer (Shimadzu, Kyoto, Japan) with electrospray source. Optimized mass spectra were obtained with an interface voltage of 4.5 kV and a detector voltage of 1.05 kV in negative mode and 1.25 kV in positive mode. Samples were analyzed on a Phenomenex Kinetex 2.6 μm XB-C18 100 Å column (100 mm × 2.1 mm, Phenomenex, Torrance, CA, USA). The column temperature was set to 40 ◦C; the flow rate was 0.3 mL/minute. Gradient elution was performed using LC–MS-grade water (Carl Roth GmbH, Karlsruhe, Germany) (eluent A) and acetonitrile (Carl Roth GmbH, Karlsruhe, Germany) (eluent B), both acidified with 0.1% acetic acid (Merck, Darmstadt, Germany). Then, 5 μL of each samples were analyzed with the gradient: (0 min) 5% B, (3 min) 60% B, (8 min) 95% B, followed by a holding time of 3 min at 95% eluent B and 2.5 min column re-equilibration with eluent 5% B. FB1 (diluted from 10 mg/L) standard solutions were used as references. MS parameters: source block temperature 90 ◦C; desolvation temperature 250 ◦C; heat block temperature 200 ◦C; drying gas flow 15.0 L/minute. Detection was performed using selected ion-monitoring (SIM) mode.

Detection (LOD) and quantification (LOQ) limits were 31 and 93 μg/kg for FB1 and 51 and 154 μg/kg for FB2.

For the calculation of LOD and LOQ, nine calibration points (0.1 μg/kg; 0.5 μg/kg; 1 μg/kg; 5 μg/kg; 10 μg/kg; 50 μg/kg; 100 μg/kg; 500 μg/kg; 1000 μg/kg) were measured, and the LOD and LOQ were calculated using the STHIBAYX function in Microsoft® Excel (Version 2013, Microsoft Corporation, Redmond, WA, USA). The slope of the calibration curve was determined using the nine calibration points.

*LOD* <sup>=</sup> (Peak area 1, Peak area 2, ... .; Concentration 1, Concentration 2, ... .)· 3,3333 Slope of calibration curve *LOQ* <sup>=</sup> (Peak area 1, Peak area 2, ... .; Concentration 1, Concentration 2, ... .)· <sup>10</sup> Slope of calibration curve

Ms Excel 2010 was used for the evaluation of the results.

#### *4.3. Analysis of the Measurements and Correlation between FB1 and FB2 Concentrations*

Main descriptive statistics (mean, maximum, 95th percentile) of the measured fumonisin contamination of the analyzed commodities were used. The measurement results were also characterized regarding the number of non-detected/not measurable values and the samples with fumonisin content exceeding the regulatory limit.

To take into account the uncertainty derived from the non-detected (<LOD) and detected but not measurable values (<LOQ), two scenarios were considered. First, to account for the worst-case option, assuming the highest possible concentration of these non-numerical values, LOD was inserted for values <LOD, and LOQ was inserted for values <LOQ, for both fumonisin B1 and B2 results. This was termed the upper-bound scenario. To illustrate with numbers, 0.031 mg/kg and 0.051 mg/kg were substituted for values of FB1 and FB2 <LOD, respectively, and 0.093 mg/kg and 0.154 mg/kg were substituted for values of FB1 and FB2 <LOQ, respectively.

To account for an optimistic scenario, assuming the lowest possible concentration, values <LOD were replaced with 0, and values <LOQ were replaced with the relevant LOD. This scenario was termed the lower-bound scenario. To illustrate with numbers, 0 was inserted for values of both fumonisins <LOD, and 0.031 mg/kg and 0.051 mg/kg were inserted for values of FB1 and FB2 <LOQ, respectively. Obviously, in the case of values >LOQ, the measured numerical values were used directly in all scenarios.

The possible correlation between fumonisin B1 and B2 contents in the samples was also analyzed, calculating the correlation coefficients. Besides considering only the corresponding numerical values of FB1 and FB2, we also analyzed a larger sample set, including those samples for which a numerical FB1 value was accompanied by a detected but not measurable (i.e., between LOD and LOQ) FB2 result. Lower- and upper-bound scenarios were calculated for these sample results as well.

#### *4.4. Food Consumption Data*

Consumption data were obtained from a survey carried out jointly by the Hungarian Food Safety Office (HFSO) and the Hungarian Central Statistical Office in 2009. Three-day dietary record data were collected from 4992 consumers, providing overall 14,976 daily food consumption data, including those of 934 children (aged below 18).

Relevant consumptions of maize products were recorded specifically for maize flour, maize grits, corn flakes, sweet maize, canned maize, frozen maize, extruded corn flakes, popped maize (with and without oil), and cheese-flavored popped maize. These products, and consequently their consumptions, were linked to the analyzed products, in order to perform intake calculations based on the concentration and consumption data of these specific commodities. Table 7 presents the commodities analyzed in relation to those consumed.


**Table 7.** Linking of analyzed values to consumed maize commodity categories by commodity name.

The maize-based, snack-like products measured mainly consisted of different types of snacks produced from maize, including nacho, tacoshells, corn flips, tortilla chips, extruded maize snack, etc. Although they had different compositions, they were dealt with in one aggregated commodity group called maize-based, snack-like products, as their compositions were not specified in the consumption data.

The effect of processing was not taken into account for two reasons. First, the effect of milling was not relevant, as the analytical measurements and consumptions were both recorded for milled maize products, enabling a direct linkage between them. On the other hand, the effect of heating was relevant for maize flour and maize grits; however, further studies would be necessary to conclude on the quantitative effect of heating, based on the literature.

We considered only the consumption days for which consumption of the selected foods was reported. Given that maize-based products are not consumed daily, including those individuals who did not report any consumption of these foods would unrealistically dilute our data. The main statistical parameters of the consumption data are summarized in Table 8.

#### *4.5. Risk Assessment Approach*

Risk assessment was performed by semi-probabilistic means, by considering the consumption values as a distribution, since there were exact individual food consumption data available. The concentration values of FB1 and FB2 were summed in each sample and considered accordingly in further calculations.

For each food commodity category, the average concentration was calculated for both the LB and the UB scenarios. These values were then multiplied by the relevant individual consumption data one-by-one, resulting in the relevant calculated fumonisin intake values for each individual consumption of each commodity.

The daily individual intakes calculated from each commodity category were then summed for each individual, resulting in the summed individual daily fumonisin intake from all the selected foods. The resulting distribution of individual total daily fumonisin intakes could then be further studied on a population level. Average and high (95 percentile) values were calculated to determine the fumonisin intake of average and high consumers. These calculations were also applied for the children population of consumers. Finally, the resulting values were compared to the reference values established by JECFA [11] and EFSA [15].

To estimate the commodity contributions to the summed intake estimated from the analyzed commodities, the population average intake from each commodity was calculated separately and then compared to the average summed intake, resulting in the proportion of contribution of each commodity.



**Author Contributions:** Study design, M.S.-S., M.K. and A.Z.; Methodology, A.Z. and M.S.-S.; Analytics, G.M., N.S. and A.S.; Formal Analysis, A.Z.; Resources, M.K.; Original Draft Manuscript Preparation, A.Z.; Review & Editing, A.Z., M.S.-S. and A.S.; Supervision, M.K. and M.S.-S.

**Funding:** This research was funded by the project GINOP-2.3.2.-15-2016-00046 and the EFOP-3.6.3.-VEKOP-16-2017-00005 programs.

**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*

## **Species Composition and Toxigenic Potential of** *Fusarium* **Isolates Causing Fruit Rot of Sweet Pepper in China**

**Jianhua Wang 1, Shuangxia Wang 2, Zhiyong Zhao 1, Shanhai Lin 1,3, François Van Hove <sup>4</sup> and Aibo Wu 2,\***


Received: 28 October 2019; Accepted: 21 November 2019; Published: 24 November 2019

**Abstract:** Apart from causing serious yield losses, various kinds of mycotoxins may be accumulated in plant tissues infected by *Fusarium* strains. Fusarium mycotoxin contamination is one of the most important concerns in the food safety field nowadays. However, limited information on the causal agents, etiology, and mycotoxin production of this disease is available on pepper in China. This research was conducted to identify the *Fusarium* species causing pepper fruit rot and analyze their toxigenic potential in China. Forty-two *Fusarium* strains obtained from diseased pepper from six provinces were identified as *F. equiseti* (27 strains), *F. solani* (10 strains), *F. fujikuroi* (five strains). This is the first report of *F. equiseti*, *F. solani* and *F. fujikuroi* associated with pepper fruit rot in China, which revealed that the population structure of *Fusarium* species in this study was quite different from those surveyed in other countries, such as Canada and Belgium. The mycotoxin production capabilities were assessed using a well-established liquid chromatography mass spectrometry method. Out of the thirty-six target mycotoxins, fumonisins B1 and B2, fusaric acid, beauvericin, moniliformin, and nivalenol were detected in pepper tissues. Furthermore, some mycotoxins were found in non-colonized parts of sweet pepper fruit, implying migration from colonized to non-colonized parts of pepper tissues, which implied the risk of mycotoxin contamination in non-infected parts of food products.

**Keywords:** *Fusarium* species; mycotoxin; toxigenic profile; mycotoxin migration; sweet pepper; fungal disease

**Key Contribution:** *Fusarium* species on sweet pepper in China is different from those in Canada and Belgium, and *F. equiseti*, *F. solani*, and *F. fujikuroi* were first reported causing pepper fruit rot in China. Toxigenic potential of *Fusairum* strains were analyzed in inoculated pepper fruit and diffusions of FA (fusaric acid), FB1 (fumonisin B1), FB2 (fumonisin B2), and MON (moniliformin) from lesions into the surrounding sound tissues were observed. This is the first report about the migration of FA and MON in sweet peppers.

#### **1. Introduction**

Sweet pepper or bell pepper (*Capsicum annuum* L.) is highly appreciated in the fresh vegetable markets worldwide due to its unique taste, aromas, and the multiple culinary uses. It represents one of the most important vegetables for the high content of phytochemicals, such as ascorbic acid and soluble phenols [1,2], having potential positive effects on human health. For example, ascorbic acid is an essential dietary nutrient in the human body with its vital biological function as an antioxidant. Sweet pepper is an economically important vegetable crop and widely used for direct consumption and manufacturing of sauce worldwide, where the global production reached 34.6 million tons of fresh fruit and 3.5 million tons of dried pods [3]. China is the largest producer and exporter of sweet pepper in the world, with a total fresh pepper production per year of almost 23 million tons. Sweet pepper is often grown in commercial greenhouses, which is favorable for the growth and survival of phytopathogenic fungi and the infection of pepper plants [4]. Thus, the disease poses a serious limitation to pepper cultivation, resulting in yield reduction or complete crop loss, as reported previously in the literature [4–6].

Several fungal diseases have caused economic losses in sweet pepper production in Canada, United Kingdom, and Belgium [4,7,8]. *Fusarium* infection on the stem- and blossom-end of pepper fruit caused by *F. solani* was first reported in Ontario and British Columbia in 1991 and caused approximately 5% fruit-yield loss [8–10]. Since the 1990s, this rot disease has been an increasing problem in pepper production in both Europe and North America [7,11,12]. More seriously, a severe outbreak of this disease resulting in a 50% yield loss in a greenhouse was reported in 1990 in Ontario [13].

In addition to external fruit rot, internal rot of sweet pepper fruit is also a big problem in pepper production [14]. In contrast to the external pepper fruit rot, the fruit is infected internally by a fungus. Unless severely infected and rotten, most infected fruits are difficult to cull before delivery to the market as the symptoms are not readily visible [4,8,10]. A comprehensive histopathology analysis performed by Yang et al. [10] indicated that internal fruit rot of greenhouse sweet pepper caused by *F. lactis* was initiated through the infection of the stigma and style during anthesis. Symptomless seed infection may contribute to disease spread, and air and insects also play an important role as intermediates in fungal spores spreading [15]. Based on the surveys to date, *F. lactis* is the principal causal agent of internal fruit rot of pepper, although the closely related *Fusarium* species belonging to the *Fusarium fujikuroi* species complex (FFSC), such as *F. proliferatum*, *F. subglutinans,* and *F. verticillioides*, have also been implicated in this disease [4,6,8].

Apart from causing significant yield losses, *Fusarium* species can produce fumonisins (FBs), trichothecenes (TCs), zearalenone (ZEN), and other mycotoxins in infected plant tissues which are harmful to consumers. For example, B-series fumonisins (Figure 1), which are mainly produced by *Fusarium* species from FFSC, are the most frequently detected mycotoxins in maize, and are involved in animal and human diseases by interfering with sphingolipid metabolism. Moreover, fumonisins have been associated epidemiologically with esophageal cancer in humans in some regions of the world [16,17]. On the basis of available toxicological evidence, the International Agency for Research on Cancer (IARC) has classified fumonisins as possibly carcinogenic (Group 2B) [18]. Mycotoxin contamination in grains, vegetables, and fruit poses a serious threat to food safety.

Mycotoxins can diffuse into tissues surrounding the pathogen-infected site. The water content of fresh sweet pepper is usually higher than 90%, which contributes to the dissolution of polar compounds produced by pathogens in pepper tissues. Fumonisins, fusaric acid (FA), and moniliformin (MON) are the most common mycotoxin contaminants produced by *Fusarium* species, and migration of some of these toxins has been reported in pepper and fruit previously [19,20]. A migration study of beauvericin (BEA) and fumonisins was reported by Monbaliu et al. [19], and the results indicated the migration of fumonisins into healthy parts of the sweet pepper, while beauvericin was not detected in the tissue surrounding the lesion. Mycotoxin contaminations originating from mycotoxin-producing *Fusarium* species in pepper should be considered as an importantly sensitive food safety concern. Therefore, it is of great necessity to assess the types and levels of mycotoxin contamination in pepper and its derived products [4].

**Figure 1.** Structural formulas of B-series fumonisins.

Effective management of mycotoxin contamination in sweet pepper relies on the control of the fungal infection and requires a better understanding of *Fusarium* biology and epidemiology. To our best knowledge, there is little information on the causal agent, etiology, toxigenic potential, and geographic distribution of the *Fusarium* species involved in pepper disease in China. The objectives of this current work were: (i) To isolate and identify the causal organism from *Fusarium* genus on sweet pepper in China; (ii) to assess the potential mycotoxin profiling of the fungal pathogens; (iii) to estimate the migration behavior of mycotoxins in sweet pepper.

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

#### *2.1. Isolation and Identification of Fusarium Isolates*

Sweet pepper fruits were sampled from Hainan, Heilongjiang, Hunan, Shanghai, Shandong, and Zhejiang provinces in China. Following isolation from diseased pepper tissues, a total of forty-two single-spore isolates were obtained and identified as belonging to *Fusarium* genus by morphology. Two of theses strains were isolated from Hainan, one strain was isolated from Heilongjiang, five strains were isolated from Hunan, twenty-one strains were isolated from Shanghai, three strains were isolated from Shandong, and ten strains were isolated from Zhejiang (Table 1).

The purified isolates were first identified to species with morphological characteristics [21], and this was subsequently confirmed by nucleotide sequences analysis of the translation elongation factor 1-α (*TEF*-1α) genes with partial of representative strains. Of the forty-two strains, twenty-seven were identified as *F. equiseti*, ten as *F. solani*, and five as *F. fujikuroi* (Table 1). Previously, a new disease on pepper caused by *F. concentricum* was reported by our group, and the strain MUCL54697 was isolated from Hunan province, China [22].

Eleven representative strains belonging to different species were selected for *TEF-1*α gene sequencing to further confirm the morphological identification results as described before [23–25]. Obtained sequences were subjected to alignment analysis using the network service tool BLASTn of the National Center for Biotechnology Information (NCBI) database. The sequence analysis of the portion of the *TEF-1*α genes of representative strains confirmed that all the strains belonged to *Fusarium* genus. The results indicated that nucleotide sequences of three strains (Q12002, Q12003 and Q12005) showed the highest identity (>99%) with the sequence of *F. equiseti*, three (Q12029, Q12030 and Q12034) showed the highest identity (>99%) with the sequence of *F. solani,* and five (Q12038-Q12042) showed the highest identity (>99%) with the sequence of *F. fujikuroi* in the NCBI database. The molecular identification results were all identical with the morphological results. The *TEF-1*α gene sequences generated in this study were deposited in GenBank, NCBI (http://www.ncbi.nlm.nih.gov/genbank/), under accession numbers KF208617–KF208627, which is presented in Table 1.


**Table 1.** *Fusarium* strains isolated in this study.

Among the forty-two *Fusarium* strains, all the ten *F. solani* (23.81%) strains were isolated from Shanghai, and the five *F. fujikuroi* (11.90%) strains were isolated from Hainan (one strain), Hunan (two strains), and Zhejiang (two strains). *F. equiseti* was the only species isolated from all six sampled provinces in China, and accounting for 64.29% of all *Fusarium* strains isolated (Table 1). According to our survey, *F. equiseti* was the predominant pathogen of sweet pepper fruit rot in China.

*Fusarium* strains were isolated from both external and internal rotten pepper fruits in this study. Of the forty-two strains, thirty-four were isolated from external rotten pepper fruits and eight strains were isolated from internal rotten pepper fruits (Table 1). For the twenty-seven *F. equiseti* strains, twenty-two were associated with external rot disease of pepper fruits and five (Q12002, Q12005, Q12008, Q12018, and Q12022) were isolated from internal rotten pepper fruits. For the ten *F. solani* and five *F. fujikuroi*, two (Q12034 and Q12035) and one strain (Q12041) were isolated from internal rotten pepper fruit, respectively. As such, *F. equiseti*, *F. solani*, and *F. fujikuroi* were the causal agents of external fruit rot of pepper in China, with *F. equiseti* being the predominant, while *F. equiseti*, *F. solani*, and *F. fujikuroi* were associated with internal fruit rot of pepper, with *F. equiseti* being the predominant. To our best knowledge, this is the first report of *F. fujikuroi*, *F. equiseti,* and *F. solani* associated with external and internal pepper fruit rot in China.

*F. solani* was reported as the predominant causal agent of external fruit rot of pepper in Europe and North America [7–13], which is different from China. However, it worth noting that *F. solani* is a common pathogen causing pepper fruit rot worldwide. *F. solani*, *F. lactis*, *F. proliferatum*, *F. subglutinans*, and *F. verticillioides* were reported to cause external or internal pepper fruit rot in Belgium, Canada, and United Kingdom, with *F. lactis* being the principal one [4,6,8]. In this study, *F. equiseti*, *F. solani*, and *F. fujikuroi* were found to be associated with internal pepper fruit rot, and with *F. equiseti* being the predominant. In light of the above, it is obvious that the population structure of *Fusarium* species associated with pepper fruit rot (external or internal) in China is quite different from those in surveys from Canada and Belgium [4,8,10]. Although the underlying factors for species distribution are unknown, climatic conditions (such as the annual temperature weather and humidity), hosts and their rotation, and adaptive evolution have been reported to influence the distribution of *Fusarium* species [14,26–28]. In view of the population genetic diversity and dispersal difference of the *Fusarium* pathogens, procedures for effective management of these pathogens on pepper are urgently needed.

#### *2.2. Molecular Phylogenetics*

Phylogenetic analyses were conducted on partial sequences of *TEF-1*α genes. Figure 2 shows the phylogenetic tree constructed with MEGA 5.10 [29]. As several *Fusarium* species have been reported to be the causal agents of pepper fruit rot, in addition to the nucleotide sequences obtained in this study (Table 1), corresponding sequences available in GenBank for the strains belonging to *F. concentricum*, *F. equiseti*, *F. fujikuroi*, *F. lactis*, *F. proliferatum*, *F. solani*, *F. subglutinans*, and *F. verticillioides* were retrieved and served as references.

**Figure 2.** Phylogenetic tree inferred from alignments of *TEF-1*α sequences of *Fusarium* species by the Neighbor-Joining method with program MEGA 5.10. The numbers beside branches are the percentages of congruent clusters in 1000 bootstrap trials. Bootstrap values higher than 75% are shown.

As shown in Figure 2, bootstrap analyses of the *TEF-1*α gene partial sequences clearly separated the thirty-six *Fusarium* isolates into three major clades with a bootstrap value of 100%; the *F. equiseti* clade (seven isolates), the *F. solani* clade (eight isolates), and the third clade that was composed of the remaining twenty-one isolates belonging to *F. concentricum*, *F. fujikuroi*, *F. lactis*, *F. proliferatum*, *F. subglutinans*, and *F. verticillioides*. The six *Fusarium* species mentioned above in the third clade are all from the *F. fujikuroi* species complex (FFSC) [30], and they are relatively closer species among the species analyzed.

It is obvious that six distinct subclades were formed with >97% bootstrap support in the third clade which separated the twenty-one isolates to *F. concentricum* (three isolates), *F. fujikuroi* (ten isolates), *F. lactis* (two isolates), *F. proliferatum* (two isolates), *F. subglutinans* (two isolates), and *F. verticillioides* (two isolates) (Figure 2). Thus, the *TEF-1*α sequences can efficiently differentiate these *Fusarium* species. The *TEF-1*α sequence BLASTn and phylogenetic analysis results strongly supported the identification results of *Fusarium* strains isolated.

#### *2.3. Multi-Mycotoxin Analysis in Pepper Fruit*

For large-scale screening of mycotoxin production capabilities in pepper by the sampled *Fusarium* isolates, the well-established multi-component LC-MS/MS method was used for scanning of the 36 mycotoxins reported in agricultural products. In this multi-mycotoxin analysis, two *F. equiseti* isolates (Q12002 and Q12004), three *F. solani* isolates (Q12029, Q12034 and Q12037), four *F. fujikuroi* isolates (Q12038, Q12039, Q12040 and Q12041), and the *F. concentricum* strain MUCL54697 [22] were selected to do the inoculation experiment and assess their mycotoxin production in different sites of sweet pepper fruits (Figure 3).

**Figure 3.** Inoculated pepper with the indication of inoculation site (red dot) and healthy site (black oval) for mycotoxin analysis.

Mycotoxin detection results from lesions (Figure 3, red dot) were used to do the mycotoxin profile analysis, and the results from the healthy parts (Figure 3, black oval) were used for the migration study. Mycotoxin profiles of the ten strains selected were summarized in Table 2. The multiple mycotoxin analysis resulted in the detection of BEA, FA, fumonisin B1 (FB1), fumonisin B2 (FB2), MON, and nivalenol (NIV) in pepper tissues (Figures 1 and 4). In one out of the ten isolates inoculated, none of the thirty-six mycotoxins investigated was detected. FA was the most frequently detected metabolite which was produced by nine isolates in concentrations that greatly varied from 41.44 to 10,662.36 μg/kg. BEA was produced by eight isolates with average concentrations varying between 5.20 and 1019.60 μg/kg. Both FB1 and FB2 were produced by the same four isolates ranging from 43.64 to 39,326.60 μg/kg and 26.96 to 3734.16 μg/kg, respectively. MON was produced by five isolates in greatly varying amounts (35.80–2439.48 μg/kg), while NIV was produced by only one isolate at a concentration of 184.16 μg/kg.


**Table 2.** Mycotoxins produced by selected *Fusarium* strains in sweet pepper samples.

<sup>1</sup> IS, Inoculation site; HS, Healthy site.

**Figure 4.** Structural formulas of moniliformin (**A**), nivalenol (**B**), fusariuc acid (**C**), and beauvericin (**D**).

As shown in Table 2, significantly different toxigenic profiles were observed among different species in inoculated pepper fruits. For the investigated *Fusarium* species, the most abundant of toxic metabolites were produced by *F. fujikuroi*, and all the four *F. fujikuroi* isolates can produce BEA, FA, FB1, FB2, and MON. With regard to the individual mycotoxin, relatively higher contents of FB1 were generated compared to FB2 by the same strain, which were consistent with the previously reported studies [31,32]. The ratios between the two fumonisins for individual *F. fujikuroi* strain were in the range of 1.05–1.83 in pepper lesions. Meanwhile, significant differences in fumonisin production capacities of *F. fujikuroi* strains were observed. For example, strain Q12040 produced FB1 and FB2 in concentrations of 43.64, 26.96 μg/kg, respectively, while large amounts of FB1 and FB2 were produced by isolate Q12041 (3926.60, 3734.16 μg/kg, respectively) which were about 90 and 139 times as higher than those produced by Q12040 under the same conditions. Note that similar or higher amounts of FA and MON were produced by the four *F. fujikuroi* strains when compared with FBs. For example, the amount of FA is higher than the total fumonisins (FB1 + FB2) produced by three out of four isolates. Strain Q12002 produced NIV, a kind of type B trichothecene, at a concentration of 184.16 μg/kg, while trichothecenes were not observed with the other strain Q12004. These two *F. equiseti* isolates showed a considerable intraspecies variation in profiles of trichothecene production, similar to results reported by Hestbjerg et al. [33]. Similarly, intraspecies variation in toxigenic profiles was also observed in

*F. solani*. Regarding the three *F. solani* strains, only FA was produced by strain Q12029, compared to BEA and FA (94.84, 78.98 μg/kg, respectively) that were produced by strain Q12034, while none of the thirty-six mycotoxins were detected with strain Q12037.

Large variation among strains, both in terms of their toxigenic profiles and the quantity of mycotoxins produced in pepper, was found in this study. Based on the mycotoxin profile results, it could be concluded that interspecies toxigenic profile variation appears to be a species-specific characteristic, while the intraspecies quantity variation appears to be a strain-specific characteristic.

The results of the migration study are summarized in Table 2. Among the mycotoxins detected in this study, diffusion phenomenon of FA, FB1, FB2, and MON from a moldy area to healthy tissues was observed in pepper fruit, while no detectable BEA and NIV were found in unaffected parts.

As shown in Table 2, FA was the most frequently detected mycotoxin in unaffected parts, with concentrations varying from 13.80 to 874.04 μg/kg. As a phytotoxin [34], the ratios of FA detected from lesions and healthy parts were in the range of 5.54–51.87. FB1 was detected in unaffected parts with concentrations varying from 41.08 to 183.04 μg/kg. The ratios of FB1 detected from lesions and healthy parts were in the range of 21.45–37.95. FB2 was detected in unaffected parts with concentrations varying from 24.30 to 144.92 μg/kg and the ratios of FB2 detected from lesions and healthy parts were similar to FB1 (25.77–37.58). MON was detected in unaffected parts with concentrations varying from 172.13 to 253.12 μg/kg, and the ratios of the compound from lesions and healthy parts were in the range of 7.61–12.59. Migration of FB1 and FB2 in sweet pepper was reported by Monbaliu et al. [19], and this is the first report about the diffusion of FA and MON in sweet pepper.

Mycotoxins can persist in infected plant tissues, and depending on physical or chemical properties (solubility, polarity, hydrophilicity, molecular weight, concentration, etc.) and tissue components, might also transfer from a rotten part of the plant tissues into the surrounding sound tissues, even in the absence of fungal growth. In this study, no NIV was detected in healthy pepper tissues maybe due to the low concentration, while BEA was not detected even at a high concentration (1019.60 μg/kg). Similar results were reported by Monbaliu et al. [19] that BEA can not be detected in surrounding tissues even with an extremely high concentration, 73,800 μg/kg, in pepper lesions. As shown in Figure 4, BEA is a cyclic hexadepsipeptide that contains three D-hydroxyisovaleryl and three N-methylphenylalanyl residues in an alternating sequence [35]. As an organic and non-polar compound, BEA is insoluble in the aqueous environment. Vegetable, and sweet peppers in particular, contain >90% water, which is probably why BEA was not detected in the surrounding tissues. These results demonstrated the possible risk of mycotoxin contamination in non-infected parts of food products, and some mycotoxins can diffuse into sound tissues. Since second-quality vegetables and fruits may be used to produce derivatives such as juices, jam, etc., further studies on migration behaviors and affecting factors of different mycotoxins in various vegetables and fruits are very important to establish suitable means of protecting consumers from exposure to toxic substances [20].

#### **3. Conclusions**

*Fusarium* species causing pepper fruit rot (external and internal) were analyzed in this study. Altogether, forty-two isolates belonging to *F. equiseti* (27 isolates), *F. solani* (10 isolates), and *F. fujikuroi* (five isolates) were identified with *F. equiseti* being the predominant species. To our best knowledge, this is the first report of *F. fujikuroi*, *F. equiseti* and *F. solani* associated with pepper fruit rot in China. Toxigenic profiles of ten pathogens were determined in sweet peppers, and six toxic metabolites (BEA, FA, FB1, FB2, MON, and NIV) were detected in total. Significantly different toxigenic profiles were observed among the three *Fusarium* species. Diffusions of FA, FB1, FB2, and MON from lesions into the surrounding sound tissues were observed in sweet peppers, and this is the first report about the migration of FA and MON in sweet peppers. Further studies on migration behaviors and affecting factors of different mycotoxins in various vegetables and fruit should be conducted.

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

#### *4.1. Isolation and Purification of Fusarium Strains*

During the growing season, pepper fruit samples were collected from different regions in China, including Hainan, Heilongjiang, Hunan, Shandong, Shanghai, and Zhejiang provinces (Table 1). Pepper fruits with visible symptoms (external or internal when cut open) were selected for pathogen isolation. Fungi were isolated using conventional methods as follows: Symptomatic tissues (3 × 3 cm) were surface-sterilized in 0.1% HgCl2 for 1 min, transferred into 70% ethanol for 30 s, then rinsed three times in sterilized distilled water, dried, and plated on 90 mm Petri dishes containing potato dextrose agar (PDA). After incubation for 3–5 days at 28 ◦C in the dark, colonies resembling morphologically to *Fusarium* were transferred onto new PDA. Plates were incubated at 28 ◦C in the dark until colonies developed, and then purified through serial transfers. No more than one strain per fruit was isolated. For each strain, a single spore culture was obtained by single-sporing as described before [36]. The pure cultures were used for the morphological and molecular characterization. Monoconidial strains were cryopreserved and maintained in tubes on PDA in the lab.

#### *4.2. Nucleotide Maniplation*

Mycelia plugs from 3-day-old PDA cultures were transferred to 50 mL of potato dextrose broth (PDB) medium and incubated with shaking (100 rpm) at 28 ◦C in the dark for 3 days. After incubation, mycelium was filtered through two-layered cheesecloth, washed with sterile water, then freeze-dried and ground to a fine powder using a TissueLyser II system (Qiagen Tissuelyser II, Retsch, Haan, Germany). Genomic DNA was extracted and purified using a Cetyl Trimethylammonium Bromide (CTAB) protocol as described before [37]. In brief, after homogenization, the ground power was suspended with 600 μL of CTAB lysis buffer, mixed well by shaking, and incubated at 65 ◦C for 1 h. After incubation, the solution was cooled at room temperature for 5 min, cellular debris were pelleted by centrifugation at 12,000 rpm for 10 min, and 500 μL of supernatant was transferred into a new tube. The supernatant was extracted with 500 μL of chloroform:isoamyl alcohol mixture (24:1, *v*/*v*). After centrifugation at 12,000 rpm for 10 min, 400 μL of aqueous phase was transferred to a fresh tube, then 400 μL of ice-cold isopropyl alcohol and 40 μL of sodium acetate (3 M, pH 5.2) were subsequently added to the samples. The tubes were mixed by gentle inversion. After incubation for 1 h at −20 ◦C, DNA was precipitated by centrifugation at 12,000 rpm for 10 min. The DNA sample was washed with 800 μL of pre-chilled 70% ethanol and air-dried before resuspension in 50 μL TE buffer (10 mM Tri-HCl, 0.1 mM EDTA, pH 8.0). DNA quantification and quality analysis were carried out by agarose gel electrophoresis with known DNA marker as standard.

#### *4.3. Fusarium Strain Identification*

The purified *Fusarium* strains were identified to species by morphological characteristics, and this was confirmed by *TEF-1*αgene sequence analysis of the representative strains. Methods for determining phenotypic characters and mycelial growth of *Fusarium* strains were from published protocols [21].

In order to verify the identity of *Fusarium* strains collected, DNA sequence comparisons were made for a subset of the strains using the *TEF-1*α gene, known as one of the most pertinent gene for determining the species rank in the *Fusarium* genus [23–25]. Portions of the *TEF-1*α gene were amplified with primer pair EF-1 (3 -ATGGGTAAGGA(A/G)GACAAGAC-5 ) and EF-2 (3 -GGA(G/A)GTACCAGT(G/C)ATCATGTT-5 ) in a thermal cycler. Polymerase chain reaction (PCR) was performed in a 50 μL reaction system afterwards [4], with minor modifications. PCR reaction mixtures contained 1× TransStar FastPfu Fly PCR SuperMix (TransGen Biotech, Beijing, China), 0.2 μM of each primer, and 50 ng of genomic DNA template. A negative control omitting the DNA template was used in every set of reactions. The thermal cycler (T100 Thermal Cycler, Bio-Rad, Hercules, CA, USA) conditions consisted of an initial denaturation step at 94 ◦C for 2 min, followed by 30 cycles of denaturation at 94 ◦C for 20 s, annealing at 58 ◦C for 20 s and extension at 72 ◦C for 30 s, then

a final extension of 72 ◦C for 5 min. Amplified products (50 μL) were separated by electrophoresis on 1.5% (*w*/*v*) agarose gels. Gels were stained with ethidium bromide and photographed under UV light in the Bio-Imaging system (Bio-Rad, Hercules, CA USA). Fragments were excised and extracted from the gel using the QIAquick gel extraction kit (QIAGEN, Hilden, Germany) according to the manufacturer's instructions. Purified amplicons were sequenced in both directions using an ABI3730XL DNA sequencer (Applied Biosystems, Foster City, CA, USA) for each strain, and contigs were assembled with Sequencher version 4.1 program (Gene Codes Corporation). The *TEF-1*α gene sequences generated in this study were subjected to similarity searches with the BLASTn network service of NCBI nucleotide database.

#### *4.4. Phylogenetic Analysis*

Portions of the *TEF-1*α gene sequences from eleven *Fusarium* strains were generated for phylogenetic analysis in this study (Table 1). All sequences were compared with sequences of *Fusarium* species available in the GenBank database through BLASTn searches for similar sequences.

Phylogenetic analyses were conducted using MEGA v. 5.10 [29] to characterize the genetic diversity and evolutionary relationships of the strains. *TEF-1*α sequences of twenty-five fungal strains belonging to *Fusarium* genus retrieved from the NCBI database were also used as references for constructing a phylogenetic tree. In total, thirty-six sequences were analyzed, including seven *F. equiseti* strains, ten *F. fujikuroi* strains, eight *F. solani* strains, three *F. concentricum* strains, two *F. lactisi* strains, two *F. proliferatum* strains, two *F. subglutinans* strains, and two *F. verticillioides* strains. The *TEF-1*α sequences of these *Fusarium* strains were all available in NCBI nucleotide database, and detailed information about their strain code, geographic origin, and host/substrate is listed in Table 3.



All sequences were aligned initially with ClustalX software [44] and the alignments manually edited. Phylogenetic analyses of the sequences were performed with MEGA5.1 for Neighbor-joining (NJ) analysis, and Kimura-2 parameter model and pairwise deletion option for gaps were used. The reliability of the tree topologies was evaluated using bootstrap support with 1000 pseudoreplicates of the data.

#### *4.5. Mycotoxin Production Analysis in Pepper Fruits via LC-MS*/*MS*

Fungal strains were initially grown on PDA in 90 mm diameter Petri dishes for 7 days at 28 ◦C in the dark, after which they were used to inoculate pepper fruits. Mature pepper fruits were inoculated with mycelium plug as described by Van Poucke et al. [4]. After incubation, the fruit tissues from different positions, including the inoculation site and healthy site, were collected, homogenized separately, and processed for mycotoxin detection analysis. Mycotoxin detection results from lesions were used to do the mycotoxin profile analysis, and the results from the healthy parts were used for the migration study.

For each sample, 2 g of the grounded material was extracted with 8 mL extraction solvent (acetonitrile:water = 84:16, *v*/*v*). Multi-component analysis of mycotoxins in the inoculated samples was performed as described previously [45,46]. In total, 36 mycotoxins were included in the multi-mycotoxin analyses: Aflatoxin B1 (AFB1), aflatoxin B2 (AFB2), aflatoxin G1 (AFG1), aflatoxin G2 (AFG2), aflatoxin M2 (AFM2), aflatoxin M1 (AFM1), beauvericin (BEA), citrinin (CIT), cyclopiazonic acid (CPA), deoxynivalenol (DON), 3-acetyldeoxynivalenol (3-ADON), 15-acetyldeoxynivalenol (15-ADON), deoxynivalenol-3-glucoside (DON-3G), deepoxydeoxynivalenol (Deep-DON), diacetoxyscirpenol (DAS), fusaric acid (FA), fumonisin B1 (FB1), fumonisin B2 (FB2), fusarenon-X (FUSX), gliotoxin, HT2 toxin (HT2), moniliformin (MON), neosolaniol (NEO), nivalenol (NIV), ochratoxin A (OTA), patulin (PAT), penitrem A (PenA), sterigmatocystin (SMC), T-2 toxin (T-2), verruculogen (VER), zearalenone (ZEN), α-zearalanol (α-ZOL), β-zearalanol (β-ZOL), α-zearalenol (α-ZAL), β-zearalenol (β-ZAL), and zearalanone (ZAN). HPLC or analytical grade of acetonitrile, hexane, and other chemical agents were purchased from Merck (Darmstadt, Germany). Deionized water purified by a Milli-Q water system (Millipore, Billerica, MA, USA) was used throughout the experiments.

**Author Contributions:** S.W. and J.W. collected the samples; Z.Z. performed the mycotoxin analysis; S.W., S.L. and J.W. performed the experiments and analyzed the data; F.V.H. provided some protocols for isolation; writing—original draft preparation, J.W.; writing—review and editing, A.W.; supervision, A.W.

**Funding:** This research was funded by the National Key Research and Development Program of China (2017YFC1600304), National Natural Science Foundation of China (31871896, 31401598, and 31602124), Shanghai Agriculture Commission Basic Research Programs (2014NO.7-3-7 and 2011NO.4-3).

**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*

## **Toxin Production in Soybean (***Glycine max* **L.) Plants with Charcoal Rot Disease and by** *Macrophomina phaseolina,* **the Fungus that Causes the Disease**

#### **Hamed K. Abbas 1,\*, Nacer Bellaloui 2, Cesare Accinelli 3, James R. Smith <sup>2</sup> and W. Thomas Shier 4,\***


Received: 13 September 2019; Accepted: 29 October 2019; Published: 6 November 2019

**Abstract:** Charcoal rot disease, caused by the fungus *Macrophomina phaseolina*, results in major economic losses in soybean production in southern USA. *M. phaseolina* has been proposed to use the toxin (-)-botryodiplodin in its root infection mechanism to create a necrotic zone in root tissue through which fungal hyphae can readily enter the plant. The majority (51.4%) of *M. phaseolina* isolates from plants with charcoal rot disease produced a wide range of (-)-botryodiplodin concentrations in a culture medium (0.14–6.11 μg/mL), 37.8% produced traces below the limit of quantification (0.01 μg/mL), and 10.8% produced no detectable (-)-botryodiplodin. Some culture media with traces or no (-)-botryodiplodin were nevertheless strongly phytotoxic in soybean leaf disc cultures, consistent with the production of another unidentified toxin(s). Widely ranging (-)-botryodiplodin levels (traces to 3.14 μg/g) were also observed in the roots, but not in the aerial parts, of soybean plants naturally infected with charcoal rot disease. This is the first report of (-)-botryodiplodin in plant tissues naturally infected with charcoal rot disease. No phaseolinone was detected in *M. phaseolina* culture media or naturally infected soybean tissues. These results are consistent with (-)-botryodiplodin playing a role in the pathology of some, but not all, *M. phaseolina* isolates from soybeans with charcoal rot disease in southern USA.

**Keywords:** fungi; mycotoxins; phaseolinone; LC/MS; soybean; charcoal rot disease; root infection mechanism

**Key Contribution:** This is the first report of a toxin being found in the tissues of soybean plants naturally infected in the field with charcoal rot disease, specifically finding (-)-botryodiplodin and not phaseolinone. This is also the first report of results consistent with some isolates of *M. phaseolina* using different toxins, other than (-)-botryodiplodin, to facilitate root infection in soybean.

#### **1. Introduction**

The fungus *Macrophomina phaseolina* (Tassi) Goidanich [1], also known by the teleomorph *Sclerotium bataticola* Taub. [2], is the cause of charcoal rot disease, and other named diseases, in soybeans and about 500 other crop and ornamental species in the United States and internationally [3–5]. Charcoal rot

disease, also known as summer wilt, dry weather wilt, or black root disease, results in crop yield loss and seed quality deterioration in soybeans and other crops [6–11]. Charcoal rot disease is more prevalent in heat- and drought-stressed conditions [12,13]. *M. phaseolina* can spread to adjacent plants with interdigitating roots through the soil, infecting the roots and spreading throughout the infected plant through the vascular system [14,15]. *M. phaseolina* forms black spore-like mycelial structures called microsclerotia, which allow the fungus to survive over winter. These microsclerotia are the grey and black dots in the stems and roots of soybean plants that give charcoal rot disease its name [16]. Common agricultural practices such as managing planting dates, fungicide applications, and biological control have been ineffective in controlling this disease [17–23]. Despite extensive efforts to control charcoal rot disease by developing resistant soybean genotypes [24–26], currently available genotypes are still not sufficiently resistant to prevent the disease in the field, although moderately resistant genotypes have been shown to have lower levels of *M. phaseolina* in plant tissues [27–29].

The mechanism used by *M. phaseolina* to infect plants with charcoal rot disease is not yet understood, in part because of the diversity in *M. phaseolina* isolates [30–33]. *M. phaseolina* has been reported to produce toxins, including (-)-botryodiplodin and phaseolinone [34–37]. It has been proposed that a toxin may play a role in an early step of the mechanism used by *M. phaseolina* to infect susceptible plants through the roots from the soil reservoir, where the fungus normally lives, particularly over the winter [7,36].

The objective of the present study is to investigate the involvement of toxins, particularly (-)-botryodiplodin, in the charcoal rot disease of soybeans. Soybeans are selected as the subject for these studies because charcoal rot disease causes major economic losses for soybean production in the midsouthern USA (Mississippi, Arkansas, and Louisiana) [10,11,38–40]. The role(s) of toxins in root infection is investigated in these studies by assessing the production of (-)-botryodiplodin, phaseolinone, and other toxins in cell-free culture filtrates of charcoal rot disease-causing *M. phaseolina* isolates and in roots and other tissues from soybean plants naturally infected with charcoal rot disease in the field. Studies on the culture filtrates of charcoal rot disease-causing *M. phaseolina* isolates resulted in the discovery that some, but not all, isolates produce (-)-botryodiplodin, but not phaseolinone, and some isolates that do not produce (-)-botryodiplodin do produce another as yet unknown toxin(s). Studies on toxins present in soybean plant tissues provided the first demonstration of a mycotoxin known to be produced by *M. phaseolina* in soybean plant tissues naturally infected with charcoal rot disease, specifically (-)-botryodiplodin, but not phaseolinone.

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

#### *2.1. Toxin Production in Culture by M. Phaseolina Isolates from Plants with Charcoal Rot Disease*

Toxin production in culture by *M. phaseolina* isolates from many USA sites and numerous types of plant sources were examined as the toxicity of cell-free culture medium filtrates in soybean leaf disc cultures from two soybean genotypes, DS97-84-1 and DT97-4290 (Table 1). Toxicity assessments with the two genotypes exhibited a similar rank order with no substantive difference between the two, whether assessed at 50% strength or at full strength. The same cell-free culture filtrates from *M. phaseolina* isolates were also assayed by LC/MS for levels of (-)-botryodiplodin, the toxin previously [35] found associated with culture filtrates of a *M. phaseolina* isolate from a soybean plant in Mississippi with charcoal rot disease. Observed concentrations of (-)-botryodiplodin ranged from not detectable to 6.11 μg/mL (Table 1). The majority of isolates (51.4% of isolates studied) produced quantifiable levels of (-)-botryodiplodin in culture filtrates, while 37.8% of isolates studied produced trace levels (i.e., above the limit of detection (1 <sup>×</sup> 10−<sup>5</sup> ng/μL), but less than the limit of quantitation (1 <sup>×</sup> 10−<sup>2</sup> ng/μL), and 10.8% of isolates studied produced no detectable level of (-)-botryodiplodin in culture filtrates.




 *Macrophomina phaseolina, Mp*; Mississippi, MS; Kentucky, KY; Arkansas, AR; Louisiana, LA; Dakota, SD; Tennessee, TN; Texas, TX; Dakota, ND; Minnesota,MN; Oklahoma, OK; Kansas, KS; Nebraska, NE; North Carolina, NC; Michigan, MI; Florida, FL; Arizona, AZ; Georgia, GA; light, l; dark, d. a Toxicity score measured in soybean leaf disccultures of two soybean genotypes: (i) DT97-4290, which is moderately resistant to charcoal rot disease and (ii) DS97-84-1, which is susceptible to charcoal rot disease. Toxicity wasassessed qualitatively according to the following symptom rating scale: healthy tissue < a little browning around the edges of the leaf disc, + < moderate browning around the edgesthe leaf disc, ++ < browning of the whole leaf disc, +++ < browning of the leaf disc with some photobleaching, ++++ < photobleaching of the whole leaf disc, +++++. b Color densitywas assessed qualitatively according to the following color density scale: whitish < light yellow < light tan < light grey < tan < beige or amber < dark tan < dark brown or dark grey black. c (-)-Botryodiplodin concentrations in culture medium filtrates were measured quantitatively by LC/MS.

 of <

Whether *M. phaseolina* isolates were from trees, soybeans, melons, or other plant sources, cell-free culture filtrates were toxic in soybean leaf disc cultures, and toxicity levels varied from not detectable to very toxic (Table 1). Culture filtrates from *M. phaseolina* isolates that contained high levels of (-)-botryodiplodin (>1 μg/mL) were all very toxic in soybean leaf disc cultures, resulting in maximal or near maximal toxicity with both DT97-4290 and DS97-84-1 soybean leaf discs at 100% and 50% strength. Culture filtrates that contained intermediate levels of (-)-botryodiplodin (0.2–1.0 μg/mL) were moderately toxic in soybean leaf disc cultures. However, some other *M. phaseolina* isolate culture filtrates that contained only trace levels or even no detectable (-)-botryodiplodin were highly toxic in soybean leaf disc cultures. This observation is consistent with some disease-inducing isolates of *M. phaseolina* producing one or more toxins other than (-)-botryodiplodin. This is the first report of results supporting the hypothesis that different isolates of *M. phaseolina* may use different toxins to facilitate root infection in soybeans. Further studies are needed to determine if any of those isolates use the other toxin(s) to facilitate root infection by a mechanism analogous to the one by which (-)-botryodiplodin might facilitate root infection. Some culture filtrates from the disease-inducing isolates of *M. phaseolina* contained very little toxicity in soybean leaf disc cultures, despite the isolate being able to cause charcoal rot disease. Explanations for this observation include the possible presence of a toxin-production regulatory mechanism that suppresses toxin production by those isolates under the culture conditions used in this study, or the possibility that charcoal rot disease in soybeans may be caused by a seed-borne *M. phaseolina* endophyte that would not need a root infection mechanism or any toxins associated with it [14]. The *M. phaseolina* isolate from which phaseolinone was originally isolated [34,37] was a seed-borne endophyte.

Also included in Table 1 is an assessment of the color of week-old cultures of *M. phaseolina*. Dunlap and Bruton [41] reported that a *M. phaseolina* isolate formed pigment in an infected muskmelon (*Cucumis melo*) and in liquid culture media containing glycine and some other amino acids. Some *M. phaseolina* isolates that cause charcoal rot disease in soybeans have been observed to form large numbers of black microschlerotia under the same culture conditions that induce the production of (-)-botryodiplodin [42]. In the data in Table 1, the rank order of pigment production, as assessed qualitatively according to the color density scale used, differed substantially from the rank order of toxicity in cell-free culture filtrates as assessed in soybean leaf disc cultures at either full strength or 50% dilution and from the relative amount of (-)-botryodiplodin present as measured by LC/MS. Thus, pigment production as assessed in this study appeared to be unrelated to toxin production, consistent with the previously identified correlations not being a general phenomenon when larger numbers of *M. phaseolina* isolates are examined.

#### *2.2. Analysis of Toxin Levels in Tissue Samples from Soybean Plants Naturally Infected with Charcoal Rot Disease*

If the hyphae of a *M. phaseolina* strain that causes charcoal rot disease use a toxin(s) to create a necrotic area in the root and thereby facilitate entry into soybean plant roots from a soil reservoir, those hyphae are expected to produce a toxin(s) at least from the time the fungus detects the root in the soil until the fungal hyphae inside the plant have detected that a stable infection has been established there. Fungi that spread from plant to plant through interdigitating roots, as *M. phaseolina* does in the charcoal rot disease of soybeans, may also secrete a toxin(s) inside the roots of fully infected plants in order to create a necrotic area within the root from which hyphae may exit the plant to spread to adjacent plants. Thus, soybean plants exhibiting the symptoms of charcoal rot disease may contain a chemically and metabolically stable toxin in tissues at a level detectable by standard analytical methods such as LC/MS. If the *M. phaseolina* strain causing charcoal rot disease in a soybean plant is a constitutive (continuous) producer of the toxin, comparable levels of the toxin may be expected in all the affected tissues of diseased plants. Therefore, naturally infected soybean plants exhibiting symptoms of charcoal rot disease were collected from different infected areas in commercial soybean production fields in Mississippi in the 2004 growing season. Control soybean plants not exhibiting symptoms of charcoal rot disease were also collected. Samples of roots, leaves, stem pulp, branches, twigs, and seeds were individually extracted and analyzed by LC/MS for levels of (-)-botryodiplodin, phaseolinone, phomenone, and gigantenone (Table 2). Only (-)-botryodiplodin was detected and only in the roots of soybean plants exhibiting symptoms of charcoal rot disease, not in other tissues of diseased plants and not in the roots or any other tissues of control soybean plants not exhibiting symptoms of charcoal rot disease. This is the first report of a toxin being found in infected plant tissues associated with charcoal rot disease in soybeans. This observation would be expected if *M. phaseolina* used (-)-botryodiplodin in its mechanism for (i) initial root infection and (ii) to exit heavily infected plants in order to spread to and infect adjacent plants. However, additional studies are needed to establish a role for (-)-botryodiplodin in either the initial root infection or the root exit mechanism. No phaseolinone, phomenone, or gigantenone was found in any tissue of soybean plants with charcoal rot disease in this study. As shown in Table 3, these observations are confirmed by a similar study conducted in 2007, in which root tissue was collected from naturally infected soybean plants from commercial production fields in Mississippi and Kentucky, USA, and analyzed by LC/MS for levels of (-)-botryodiplodin, phaseolinone, phomenone, and gigantenone. As observed in the first study (Table 2), only (-)-botryodiplodin was detected in diseased roots, not phaseolinone, phomenone, or gigantenone. Again, (-)-botryodiplodin levels varied from traces to 3.14 μg/g, that is, greater than a 1000-fold concentration range. The wide range of (-)-botryodiplodin levels in charcoal rot-diseased soybean roots (Tables 2 and 3) paralleled the wide range of (-)-botryodiplodin production levels in cell-free culture filtrates of *M. phaseolina* isolates from plants with charcoal rot disease. In both experimental systems, there were a substantial number of cases in which (-)-botryodiplodin production was too low for it to be a toxin that could play a role in the pathology caused by those *M. phaseolina* strains, whether by facilitating root infection or any other mechanism.


**Table 2.** Mycotoxin levels in root and other tissues of soybean plants collected from soybean fields in Mississippi in 2004.

aIdentification and quantification of toxins in samples by LC/MS were based on one standard due to the limited amount of standards available. b Soybean plants of the Saline cultivar with no detectable sign of charcoal rot disease were collected from commercial fields in Mississippi. Samples of the same six tissues were taken, pooled, and extracted in the same way as tissues from diseased plants and the extracts were assayed by LC/MS in the same manner. Extracts of all undiseased soybean tissues, including roots, contained no detectable (-)-botryodiplodin or other toxin.


**Table 3.** Toxins in the roots of soybean plants exhibiting charcoal rot disease properties collected from commercial soybean fields in Kentucky and Mississippi in 2007.

\* Soybean plants exhibiting symptoms of charcoal rot disease were collected in the indicated commercial field numbers in the indicated states, brought to the laboratory, tissues harvested and stored at −20 ◦C until assayed. Soybean root samples had symptoms of charcoal rot and were run by LC/MS. Determination and quantification of these mycotoxins was by LC/MS based on one standard because a limited amount of these standards were available.

#### **3. Conclusions**

A wide range of (-)-botryodiplodin levels were observed in both cell-free culture medium filtrates from *M. phaseolina* isolates from plants with charcoal rot disease and in the roots, but not in the aerial parts, of soybean plants naturally infected with charcoal rot disease. Cell-free culture medium filtrates from some *M. phaseolina* isolates from plants with charcoal rot disease were strongly phytotoxic, despite containing only traces or no (-)-botryodiplodin. No phaseolinone was detected in either cell-free culture medium filtrates from *M. phaseolina* isolates or in tissues from soybean plants naturally infected with charcoal rot disease. The results of this study are consistent with some, but not all, isolates of *M. phaseolina* associated with charcoal rot disease in soybean-producing (-)-botryodiplodin. Some isolates of *M. phaseolina* cultured from soybean plants with charcoal rot disease produce no detectable (-)-botryodiplodin in culture, but do produce other unknown toxins. Further research is needed to determine what role, if any, (-)-botryodiplodin and other toxins produced by *M. phaseolina* isolates play in the root infection mechanism of the charcoal rot disease of soybeans.

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

#### *4.1. Soybean Plant and Greenhouse Conditions*

The soybean genotype DT97-4290 [28] was selected as an example of a genotype that is moderately resistant to charcoal rot disease, and the soybean genotype DS97-84-1 [43] was selected as an example of a genotype that is susceptible to charcoal rot disease. Plants were germinated in trays with vermiculite, and the seedlings of each genotype were transplanted into six soil-filled 9.45 L pots, each containing four plants of the same genotype. During the growth period, the soil water potential of the plants was maintained at approximate field conditions, 15 to −20 kPa. Six pots, each containing four plants, were used for each genotype. The greenhouse temperature was set to 34 ◦C for the day cycle and 28 ◦C for the night cycle. Light intensity ranged from that of sunny to cloudy days. Plants were harvested during the vegetative stage.

#### *4.2. M. phaseolina Culture Sources*

The collection locations and plant hosts of the 37 cultures of *M. phaseolina* used in the study are presented in Table 1. Some *M. phaseolina* cultures were isolated from infected plant tissues in the Abbas laboratory in 2013 using the method of Mengistu et al. [4,25], while other cultures were provided by colleagues from their collections, particularly G.L. Sciumbato, Mississippi State University.

#### *4.3. Preparation of Cell-Free Culture Extracts*

Potato dextrose broth (PDB) was prepared by boiling 200 g of peeled potatoes, straining them through a cheesecloth, and adding 20 g of dextrose per liter of water. PDB (150 mL) was placed in 500 mL Erlenmeyer flasks, covered with cotton plugs, autoclaved for 15 min, and allowed to cool to room temperature. Upon cooling, each flask was inoculated with three to four plugs of *M. phaseolina* isolate and placed on an Innova 40 Benchtop Incubator Shaker (New Brunswick Scientific Co., Inc., Edison, NY, USA) for seven days at 128 rpm, 28 ◦C. The color change of each culture after one week of incubation was observed and recorded according to the following color density scale: whitish < light yellow < light tan < light grey < tan < beige or amber < dark tan < dark brown or dark grey < black.

After seven days of incubation, the culture medium was passed through Whatman No.1 filter paper into a plastic beaker. The filtrate was then filtered through an 0.45 μm membrane filter in a disposable filter unit (Nalgene Company, Rochester, NY, USA, Size 250 mL cellulose nitrate CN Filter Unit) using a laboratory vacuum to achieve a cell-free filtrate that was stored at −20 ◦C until used.

#### *4.4. Toxicity of Cell-Free Filtrates of M. phaseolina Culture Media in Soybean Leaf Disc Cultures*

The toxicity of *M. phaseolina* culture filtrates was assessed by rating the appearance of soybean leaf discs from two genotypes (DT97-4290, which is moderately resistant to charcoal rot, and DS97-84-1, which is susceptible) after four to five days in half (50%) and full strength (100%) cell-free culture filtrates, *M. phaseolina* isolates were grown on potato dextrose agar (PDA) for seven days at 28 ◦C. True mature leaves with no signs of damage were harvested from 3- to 4-week-old soybean plants, and 4 mm discs were cut from the leaves using a sterile cork borer (No.4). Three leaf discs were placed in each well of sterile 24-well tissue culture trays with low evaporative lids (Becton Dickinson and Company, Franklin Lakes, NJ, USA) containing 1.5 mL of culture filtrate in triplicate at two concentrations (50% and 100%). The trays were then incubated in a growth chamber at 25 ◦C under continuous light for 96 h. The discs were observed for signs of toxic effects after 24, 48, 72, and 96 h. Toxicity was assessed qualitatively according to the following symptom rating scale: healthy tissue < a little browning around the edges of the leaf disc, + < moderate browning around the edges of the leaf disc, ++ < browning of the whole leaf disc, +++ < browning of the leaf disc with some photobleaching, ++++ < photobleaching of the whole leaf disc, +++++.

#### *4.5. Toxin Standards for LC*/*MS Analyses*

The structures of toxins measured in this study are presented in Figure 1. (±)-Botryodiplodin was synthesized, as described in the accompanying manuscript [44], as a white powder with purity over 98%. A stock solution of (±)-botryodiplodin (1000 ng/μL) was prepared in chloroform. Working standards were prepared in the concentration range 1.0 <sup>×</sup> 10−<sup>5</sup> ng/μL to 40 ng/μL in ethyl acetate. Gigantenone and phomenone were gifts from Gary A. Strobel, Montana State University, Bozeman, MT. Phaseolinone was synthesized (Figure 2) from a sample of the phomenone (6.5 mg, 0.0246 mmole) dissolved in 1 mL chloroform and mixed with a 1.2 molar excess of m-chloroperoxybenzoic acid (Acros Organics, 0.029 mmole, 7.15 mg of 70% pure material) and pyridine (4.7 μL, 4.6 mg, 0.058 mmole) dissolved in 200 μL chloroform. The mixture was incubated for 1 h at −10 ◦C with stirring and then allowed to warm to room temperature overnight. The reaction mixture was diluted with ether, extracted twice with water, once with 1N HCl to remove pyridine, twice with saturated sodium bicarbonate-brine solution to remove product m-chlorobenzoic acid and unreacted m-chloroperoxybenzoic acid, dried over anhydrous sodium sulfate, and evaporated in vacuo. The product (7.5 mg) gave a single peak at *m*/*e* 281 (phaseolinone + H+) in LC/MS analysis under the conditions described below, with no detectable phomenone starting material at *m*/*e* 265. A single peak was observed in LC/MS for the phaseolinone preparation, even though the reaction conditions would be expected to produce a mixture of phaseolinone and epi-phaseolinone, presumably because the two forms were not resolved under the liquid chromatography conditions used.

**Figure 1.** Chemical structures of the toxins measured by LC/MS in *M. phaseolina* culture media and soybean root tissues.

**Figure 2.** The chemical reaction used in the semi-synthesis of the LC/MS standard phaseolinone from the natural toxin phomenone. MCPBA = meta-chloroperoxybenzoic acid.

#### *4.6. Preparation of Plant Tissue and M. phaseolina Culture Medium Extracts for LC*/*MS Analyses*

Soybean root and other tissue samples were cleaned of adherent earth, dried in an oven at 45 ◦C for two to three days, and ground to the consistency of flour using a Stain Laboratory Mill Grinder, Model M-2 (Fred Stein Laboratories, INC., Atchison, Kansas, USA). Ethyl acetate (10 g) was added to 50 g of each sample, shaken for 1 h, filtered through filter paper (Whatman No.1), and transferred to vials for analysis by LC/MS as described below. *M. phaseolina* culture medium cell-free filtrate samples were extracted with ethyl acetate in a 1:1, *v*:*v* ratio on a vortex mixer for 1 min and allowed to separate into two distinct layers. The ethyl acetate layer was transferred to vials for analysis by LC/MS.

#### *4.7. LC*/*MS Analysis*

LC/MS analyses of toxin samples obtained prior to 2007 were conducted on a Thermo Finnigan LCQ Advantage instrument coupled to a Thermo Finnigan Surveyor MS and a Thermo Finnigan Surveyor MS Pump (Thermo Electron Corporation, West Palm Beach, FL, USA). After 2007, a more advanced and upgraded LTQ XL Ion Trap Mass Spectrometer, Finnigan Surveyor Autosampler, and Finnigan Surveyor MS Pump (Thermo Scientific, West Palm Beach, FL, USA) were used. Analyses were carried out in positive scan mode at ambient temperature using a Waters Nova-Pak C18 column, a 10 μL partial loop injection, and mobile phases (A) 1% acetic acid in methanol, (B) water, and (C) methanol at a flow rate of 500 μL/min. The analysis occurred over 25 min using a gradient of 20% A and 80% B for 12 min, then 20% A, 5% B, and 75% C for 3 min, and then back to 20% A and 80% B for the duration of the 25 min. The analysis utilized the following scan events of a full scan from *m*/*e* 100 to 300. The confirmation of (-)-botryodiplodin used three masses: *m*/*e* 127, 145, and 109. The limit of

detection (LOD) was 1 <sup>×</sup> 10−<sup>5</sup> <sup>μ</sup>g/mL and the limit of quantitation (LOQ) was 0.01 <sup>μ</sup>g/mL. The LOQ was based on the regression of the standards used for analysis. The full scan run of phomenone, gigantenone, and phaseolinone was from *m*/*e* 100 to 500, and their confirmations were identified by using *m*/*e* 265, 265, and 281, respectively.

#### *4.8. Statistical Analysis*

The analysis of variance was performed on toxin concentration data using the PROC GLM procedure in SAS Version 9.22 (Cary, NC, USA, 2010). Means were separated by Fisher's Least Significant Difference test with *p* ≤ 0.05 level of significance.

**Author Contributions:** Conceptualization, H.K.A., N.B., C.A., J.R.S., W.T.S.; methodology, H.K.A., W.T.S.; validation, H.K.A.; investigation, H.K.A., W.T.S.; resources, H.K.A., W.T.S.; data curation, H.K.A., W.T.S.; writing—original draft preparation, H.K.A., W.T.S.; writing—review and editing, H.K.A., N.B., C.A., J.R.S., W.T.S.; visualization, W.T.S.; supervision, H.K.A.; project administration, H.K.A.; funding acquisition, H.K.A., W.T.S.

**Funding:** This research was funded by the Mississippi Soybean Promotion Board, grant numbers 34-2016 and 34-2017.

**Acknowledgments:** The authors are grateful to Alemah Butler, Bobbie J. Johnson (retired), Jeremy Kotowicz, and Vivek H. Khambhati for their technical assistance. Also, the authors are grateful for G.L. Sciumbato (retired) for supplying cultures and plant samples. Trade names are used in this publication solely for the purpose of providing specific information. The mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the United States Department of Agriculture.

**Conflicts of Interest:** All 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*

## **Phytotoxic Responses of Soybean (***Glycine max* **L.) to Botryodiplodin, a Toxin Produced by the Charcoal Rot Disease Fungus,** *Macrophomina phaseolina*

**Hamed K. Abbas 1,\*, Nacer Bellaloui 2, Alemah M. Butler 1, Justin L. Nelson 3, Mohamed Abou-Karam <sup>3</sup> and W. Thomas Shier 3,\***


Received: 13 September 2019; Accepted: 18 December 2019; Published: 1 January 2020

**Abstract:** Toxins have been proposed to facilitate fungal root infection by creating regions of readily-penetrated necrotic tissue when applied externally to intact roots. Isolates of the charcoal rot disease fungus, *Macrophomina phaseolina*, from soybean plants in Mississippi produced a phytotoxic toxin, (−)-botryodiplodin, but no detectable phaseolinone, a toxin previously proposed to play a role in the root infection mechanism. This study was undertaken to determine if (−)-botryodiplodin induces toxic responses of the types that could facilitate root infection. (±)-Botryodiplodin prepared by chemical synthesis caused phytotoxic effects identical to those observed with (−)-botryodiplodin preparations from *M. phaseolina* culture filtrates, consistent with fungus-induced phytotoxicity being due to (−)-botryodiplodin, not phaseolinone or other unknown impurities. Soybean leaf disc cultures of Saline cultivar were more susceptible to (±)-botryodiplodin phytotoxicity than were cultures of two charcoal rot-resistant genotypes, DS97-84-1 and DT97-4290. (±)-Botryodiplodin caused similar phytotoxicity in actively growing duckweed (*Lemna pausicostata*) plantlet cultures, but at much lower concentrations. In soybean seedlings growing in hydroponic culture, (±)-botryodiplodin added to culture medium inhibited lateral and tap root growth, and caused loss of root caps and normal root tip cellular structure. Thus, botryodiplodin applied externally to undisturbed soybean roots induced phytotoxic responses of types expected to facilitate fungal root infection.

**Keywords:** botryodiplodin; root infection mechanism; root toxicity; *Macrophomina phaseolina*; hydroponic culture

**Key Contribution:** Botryodiplodin was observed to be phytotoxic in cultured leaf discs from soybean genotypes susceptible or resistant to charcoal rot disease, but the phytotoxic response was greatest in susceptible genotypes. Botryodiplodin was shown to be phytotoxic when applied externally to intact *Lemna pausicostata* plantlets. Botryodiplodin treatment of undisturbed soybean seedling roots in hydroponic culture resulted in loss of root tips, creating a lesion of a type that may facilitate root infection.

#### **1. Introduction**

Charcoal rot is a plant disease caused by the fungus, *Macrophomina phaseolina* (Tassi) Goid [1], in over 500 commercially-important plant species ranging from ornamental plants to trees to major food

and fiber crops, including soybean (*Glycine max* L. (Merr.)). An example of the impact of charcoal rot disease on agriculture was provided by attempts to establish commercial natural rubber production with guayule (*Parthenium argentatum* Gray) in the arid southwest region of the US as an alternative to imported material from the rubber tree (*Hevea brasiliensis*) [2]. Guayule rubber production was only competitive when plants were grown close enough together that the roots interdigitated, under which conditions charcoal rot could spread from plant to plant destroying the crop [3]. Because charcoal rot is favored by hot, dry conditions [4], it is a climate-impacted plant disease that is predicted to be an increasingly important agronomic problem going forward, given that climate change is predicted to result in hotter, drier conditions in the majority of the world [5].

Research on (−)-botryodiplodin as a food contaminant has mainly focused on its production by the blue cheese fungus, *Penicillium roqueforti* [6,7]. (−)-Botryodiplodin production by *P. paneum* in bread and silage is also a concern [8,9]. Concerns about (−)-botryodiplodin as a possible contaminant in Roquefort cheese and other foods have led to extensive studies of its possible toxic effects in mammalian systems [10]. Because *M. phaseolina* is known to produce (−)-botryodiplodin and to be present in seeds as an endophyte, contamination of food items such as tofu and vegetable oil by (−)-botryodiplodin is a concern [10]. However, studies on foods and feeds impacted by charcoal rot disease have not been reported.

Although soybean cyst nematode is the major cause of soybean yield losses most years in most parts of the US, charcoal rot has traditionally been the most economically-important disease of soybean in the mid-southern region of the US (i.e., in Arkansas, Mississippi, and Louisiana) [11–15]. However, rising average temperatures and increased prevalence of drought have made the disease an increasingly important cause of yield losses during hot, dry growing seasons in all but northern parts of the US and other parts of the world [16]. Extensive studies have been carried out attempting to use selective breeding to develop soybean genotypes that are resistant to charcoal rot, but this approach has yielded only tolerant or moderately resistant genotypes [17,18]. Attempts to use various agronomic techniques to prevent the disease have also failed, so research on charcoal rot continues [17,19].

The mechanism used by *M. phaseolina* to infect soybean plants from the soil reservoir is poorly understood. *M. phaseolina* enters plants through the roots, then spreads through conductive tissues, reducing conduction volume, plant weight, and height, as well as reducing seed quality and quantity [16,20]. Inside plant tissues, *M. phaseolina* produces microsclerotia that appear as gray to black dots in and on stems and leaves and serve as reproductive structures that survive over winter in soil or as endophytes in the infested seed [21]. Fungi are widely believed to gain admission to plant roots from the soil by either (i) physical penetration of tissue; or (ii) secretion of toxins that kill plant tissue locally, creating a necrotic region through which fungal hyphae can easily propagate [19,22–24]. The mechanism(s) used by toxins to create localized necrosis in plant roots is not well understood, but two possible mechanisms are (i) secretion of hydrolytic enzymes or toxins that induce activation of endogenous hydrolytic enzymes; and (ii) secretion of toxins that specifically kill dividing meristematic cells near root tips, which creates necrotic tissue in a place that provides convenient access to the plant's vascular system through which the fungus can spread throughout the plant [19].

*M. phaseolina* has been reported to produce several mycotoxins that are candidates for toxin-mediated initiation of infection by generating a necrotic zone. These mycotoxins include phaseolinone [25], botryodiplodin [26], and patulin, because the *M. phaseolina* genome contains genes for its biosynthetic enzymes [27]. Siddiqui et al. (1979) [25] identified phaseolinone in culture extracts of pathogenic *M. phaseolina* isolated as an endophyte of mung bean. Dhar et al. (1982) [28] proposed the structure of the isolated toxin to be an epoxidized analog of a known phytotoxin, phomenone, which is part of an extensive family of phytotoxic eremophilane sesquiterpenoid (C-15) toxins produced by numerous plant pathogenic fungi [29]. Phaseolinone has been synthesized by Kitahara et al. (1991) [30] by conversion of another known eremophilane sesquiterpenoid toxin, phomenone. A series of 12 eremophilane analogs, including synthetic phaseolinone, were shown to be phytotoxic, producing either green islands on monocot leaves or necrotic lesions on dicots [29,31].

Ramezani et al. (2007) [32] and Abbas et al. (2019) [33] found no detectable phaseolinone in culture extracts of *M. phaseolina* isolated from infected soybean plants in the Mississippi Delta region of the southern USA. Bioassay-guided fractionation of the extracts led to the isolation of a different, known mycotoxin, botryodiplodin, which was first isolated by Sen Gupta et al. (1966) [26] from culture filtrates of *Botryodiplodia theobromae* Pat. (syn. *Lasiodiplodia theobromae* (Pat.) Griffon & Maubl), a cellulolytic fungus first isolated in 1944 from mildewed tent fabric in India, and subsequently shown to be a plant pathogen in many economically-important crops in the tropics and sub-tropics around the world [34].

The objectives of the present study were to investigate the identity of the phytotoxin produced by *M. phaseolina* isolates from Mississippi soybeans with charcoal rot disease as botryodiplodin, and to characterize some botryodiplodin root toxicity properties that could enable it to play a role in the initial stages of the soybean root infection mechanism of *M. phaseolina*.

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

#### *2.1. Synthesis of (*±*)-Botryodiplodin*

Chemically synthesized (±)-botryodiplodin exhibited potent phytotoxicity in each of a series of experimental systems, including *L. pausicostata* axenic cultures (Figure 1), soybean leaf discs in culture (Figures 2 and 3), and soybean seedling roots in hydroponic (Figure 4) and sand (Figure 5) culture. One explanation for the observation [32] that bioassay-guided fractionation of phytotoxicity produced by *M. phaseolina* isolates that cause charcoal rot disease in Mississippi soybeans yielded (-)-botryodiplodin, but no detectable phaseolinone, was that the toxin preparations contained a small percentage of either phaseolinone or another unknown, but the potent toxin that was responsible for the observed phytotoxicity. Phaseolinone was proposed by Siddiqui et al. (1979) [25] to mediate infection in charcoal rot disease based on its isolation from culture filtrates of an *M. phaseolina* endophyte from mung beans in India. When a phytotoxin is purified from nature, it is never 100% pure, so that it is always possible that the phytotoxicity may actually reside in a highly toxic impurity, rather than in the major component of the preparation. Chemical synthesis is one approach that can provide evidence that the major component of the preparation is the actual toxin. Chemical synthesis of a toxin is unlikely to produce the same impurities as found in material purified from nature. Even if the impurities in the two types of preparations are both toxic, they are unlikely to induce identical pathology in all toxicity tests. Therefore, identical phytotoxic properties are unlikely to be observed in synthetic and natural preparations of a toxin, if the activities of either are due to a highly active impurity. At least seven syntheses of botryodiplodin have been reported, since the initial success by McCurry & Abe (1973) [35]. None of these syntheses could conceivably produce phaseolinone or any other eremophilane sesquiterpenoid as a by-product. Although the method used in this study to synthesize (±)-botryodiplodin was selected because it involved only five steps using simple, standard chemistry and low cost reagents, it also could not conceivably produce phaseolinone or any other eremophilane sesquiterpenoid as a contaminant. Antibacterial activity was the first biological activity identified for (-)-botryodiplodin [26], and the easiest to assay. Chemically synthesized (±)-botryodiplodin was shown to exhibit antibacterial activity indistinguishable from that of (−)-botryodiplodin purified from *M. phaseolina* cultures [32] (data not shown). In addition, (±)-botryodiplodin induced phytotoxic responses indistinguishable from those induced by (−)-botryodiplodin, when compared in duckweed (*L. pausicostata*) plantlet cultures and soybean leaf discs in culture (see below). Identical activity of (±)-botryodiplodin and (−)-botryodiplodin is consistent with extensive studies on the mechanism of action of (−)-botryodiplodin by Moule et al. (1981a; 1981b; 1982) [36–38], which indicated that the toxin acts by chemical reactions in cell nuclei that covalently cross-link proteins to DNA, and not by interacting with a chiral binding site on any enzyme or receptor that might require an optically active form. Although (+)-botryodiplodin has been prepared by chemical synthesis [39], its biological activity, or the lack thereof, has not been reported by these investigators or others. More extensive structural

alterations of botryodiplodin in the form of epimers have been reported to be inactive in the case of 4-*epi*-botryodiplodin [40]. Félix et al. (2019) [41] observed that cytotoxicity of 3-*epi*-botryodiplodin measured in Vero monkey kidney cells and 3T3 mouse fibroblast cultures was 0–5% of the cytotoxicity of botryodiplodin. However, in a leaf puncture assay in young tomato plant leaves, 3-*epi*-botryodiplodin produced a much larger lesion with different morphology than botryodiplodin, but similar to the lesion produced by botryodiplodin acetate. Thus, the possibility that (+)-botryodiplodin might be an inactive diluent in the (±)-botryodiplodin preparations used in this study, cannot be rigorously excluded, but if it were inactive, all conclusions drawn would be the same, with reported (−)-botryodiplodin activities occurring at half the stated concentrations.

**Figure 1.** Inhibition of duckweed (*Lemna pausicostata*) plantlet growth in axenic cultures containing a range of concentrations of (±)-botryodiplodin in the culture medium. Duckweed growth was measured as percent inhibition of frond production ± SEM relative to controls not treated with toxin. Phytotoxicity was assessed at 24 h (-), 48 h (-), 72 h (), and 96 h (). The full toxic response was observed by 48 h (IC50 = 0.22 μg/mL); that is, the percent growth reduction at 48, 72, and 96 h were not significantly different from each other, but all were significantly greater than that at 24 hours, *p* < 0.05, multiple linear regression analysis.

**Figure 2.** Phytotoxicity rating scale used to determine the percent severity of (±)-botryodiplodin phytotoxicity on soybean leaf discs, in which 0% = healthy tissue; 10% = slight browning around the edges of the leaf disc; 25% = moderate browning around the edges of the leaf disc; 50% = browning around the edges of the leaf disc with slight bleaching; 75% = extensive browning of the leaf disc with bleaching; and 100% = complete bleaching of the leaf disc.

**Figure 3.** Phytotoxicity effects of (±)-botryodiplodin in cultured leaf discs from three different soybean genotypes, DT97-4290 (), which was released as a charcoal rot disease resistant genotype, Saline () and DS97-84-1 (-). The phytotoxic response is shown at (**A**) 24 h, (**B**) 48 h, (**C**) 72 h and (**D**) 96 h. The phytotoxicity rating scale is described in Figure 2. Saline was significantly (*p* < 0.05, multiple regression) more susceptible to the phytotoxic effects of (±)-botryodiplodin than DS97-84-1 and DT97-4290 at each time point. Results are the mean of three replicates.

**Figure 4.** Effects of various (±)-botryodiplodin concentrations (0 to 80 μg/mL) in hydroponic culture medium on soybean seedlings. A reduced number of lateral roots and discoloration occurred at all (±)-botryodiplodin concentrations tested.

**Figure 5.** Soybean seedling roots treated with a range of concentrations of (±)-botryodiplodin (10 to 300 μg/mL) in sand culture served as controls for unsupported soybean seedling roots in hydroponic culture. A reduced number of lateral roots and discoloration occurs at higher concentrations of (±)-botryodiplodin.

The simplest explanation for differences in the type of toxin produced in culture by endophytic *M. phaseolina* isolated from mung beans in India [25] and pathogenic *M. phaseolina* isolated from soybeans in Mississippi [32] is that the isolate studied by Siddiqui et al. (1979) [25] produced both phaseolinone and (−)-botryodiplodin, whereas only (−)-botryodiplodin was produced by the Mississippi isolates [10]. Production of multiple, structurally dissimilar mycotoxins by a single fungus has been well-documented [42], and there are numerous examples in the scientific literature of regional variations in mycotoxin production by the same species of fungus [43,44].

#### *2.2. Phytotoxicity of (*±*)-Botryodiplodin in Lemna Pausicostata (Duckweed) Cultures*

A series of studies were initiated to determine if botryodiplodin possesses properties useful for a mycotoxin to play a role in mediating root infection by *M. phaseolina* from a soil reservoir. Specifically, to be an effective mediator of root infection, a toxin must be able to kill undisturbed, actively growing root tissue in the absence of an insect, nematode, or other vector that physically damages root tissue. Root cells killed by the toxin should create a necrotic region, preferably one that

would provide fungal hyphae with facile access to the plant vascular system. A toxin-mediated fungal root infection mechanism should be able to facilitate tissue entry in the absence of fungal structures such as appressoria [45] that enable fungal cells to physically penetrate plant leaf tissue in the absence of a vector.

(±)-Botryodiplodin (0 to 64 μg/mL) dissolved in the culture medium of parallel axenic cultures of the aquatic plant of *Lemna pausicostata* (duckweed) induced a phytotoxic response in intact, growing plantlets floating on the surface of the culture medium over a 96-hour period (Figure 1) that was indistinguishable from the phytotoxic response to (−)-botryodiplodin prepared as described by Ramezani et al. (2007) [32]. Phytotoxicity was measured as percent growth reduction measured by the number of plantlet fronds produced relative to parallel control cultures not treated with (±)-botryodiplodin. Additional phytotoxicity occurred as a formation of necrotic tissue with light brownish color around the edges of the fronds and some bleaching progressing to 100% growth inhibition, 100% mortality, and complete bleaching. No detectable toxicity was observed at 24 hours, because growth was measured as frond number and more time than that was needed for a plantlet to generate a new frond under conditions used. However, the full extent of toxicity was observed at 48 hours with IC50 = 0.22 μg/mL. The dose-response curves at 72 hours (IC50 = 0.19 μg/mL) and 96 hours (IC50 = 0.18 μg/mL) were not significantly different from each other (Pearson's *r* = 0.993, *p* = 0.601, multiple linear regression analysis), or from that at 48 hours (Pearson's *r* = 0.994, *p* = 0.995 at 72 hr; r = 0.984, *p* = 0.874 at 96 hrs, multiple linear regression analysis).

Phytotoxicity of (±)-botryodiplodin was determined in leaf discs from charcoal rot tolerant and susceptible soybean genotypes. (±)-Botryodiplodin (0 to 320 μg/mL) in culture medium for 96 hours induced the same phytotoxic response in soybean leaf discs cut from mature leaves of three- to four-week old soybean seedlings as observed with (-)-botryodiplodin [32], specifically progressive browning (necrosis) around the edges of the leaf disc and bleaching (light-induced loss of chlorophyll) progressing to complete browning of the leaf disc and 100% bleaching. (±)-Botryodiplodin phytotoxicity was compared in leaf discs from the following three genotypes: DT97-4290, which was released as a charcoal rot disease resistant soybean genotype; and two others that are considered susceptible to charcoal rot disease, DS97-84-1 and Saline. The percent severity of phytotoxic responses was quantitated at 24, 48, 72, and 96 hours using the rating scale given in Figure 2. At each time period, Saline was significantly (*p* < 0.05, multiple regression) more susceptible to the phytotoxic effects of (±)-botryodiplodin than DS97-84-1 and DT97-4290. While DS97-84-1 was more susceptible to (±)-botryodiplodin than DT97-4290 at some times, the differences were not significant. At 24 hours (Figure 3A), phytotoxicity was observed only at the highest (±)-botryodiplodin concentrations with IC50 values of 320 μg/mL for each of Saline, DS97-84-1, and DT97-4290, respectively. At 48 hours, (Figure 3B) phytotoxicity was observed at lower (±)-botryodiplodin concentrations with IC50 values of 136 μg/mL for Saline and 272 μg/mL for DS97-84-1 and DT97-4290. At 72 hours (Figure 3C), substantial phytotoxicity was observed at progressively lower (±)-botryodiplodin concentrations with IC50 values of 59.5 μg/mL for Saline and 132 μg/mL for DS97-84-1 and DT97-4290. At 96 hours (Figure 3D), substantial phytotoxicity was observed at much lower (±)-botryodiplodin concentrations with IC50 values of 14.9, 38.5, and 42.9 μg/mL for Saline, DS97-84-1 and DT97-4290, respectively. The observation that the three soybean genotypes examined in the study exhibited susceptibility to the phytotoxic effects of (±)-botryodiplodin in the order Saline > DS97-84-1 > DT97-4290 is consistent with the charcoal rot tolerance reported for genotype DT97-4290 [18] resulting from a change expressed in multiple tissues, including leaf tissue. Given that the level of resistance expressed by genotype DT97-4290 is not sufficient to prevent charcoal rot disease and infection by *M. phaseolina* [18], subsequent studies focused on investigating root-specific responses believed to be associated with initial infection.

#### *2.3. Root Toxicity of (*±*)-Botryodiplodin in Soybean Seedlings*

Studies on root toxicity of (±)-botryodiplodin used soybean seedlings in hydroponic culture with the toxin being added to culture medium bathing only the roots. Soybean seedlings in hydroponic culture were treated for four days with a range of (±)-botryodiplodin concentrations (10 to 80 μg/mL) in the nutrient solution bathing the roots. Control seedlings produced abundant lateral roots during the hydroponic culture period. The addition of (±)-botryodiplodin to the nutrient solution reduced lateral root production even at 10 μg/mL, the lowest concentration tested in initial trials (Figure 4). Inhibition of root growth by (±)-botryodiplodin treatment was quantified by the dry weight relative to that of control plants exposed to 0 μg/mL (±)-botryodiplodin. There was significantly greater toxicity to lateral roots than to tap roots (*p* < 0.05, regression analysis) (Figure 6). (±)-Botryodiplodin exposure resulted in about an eight-fold reduction in lateral root growth, but only in about a two-fold reduction in tap root growth. There was significant reduction in tap root growth at the highest (±)-botryodiplodin concentrations tested (≥40 μg/mL), but the IC50 (23.5 μg/mL) was 5.6-fold higher than the IC50 for lateral roots (4.2 μg/mL), which exhibited significant (*p* < 0.05, Student's *t*-test) reduction in lateral root growth at ≥5 μg/mL (Figure 6). Thus, botryodiplodin caused toxicity to undisturbed soybean roots when applied externally, which is a property expected for a toxin capable of playing a role in facilitating fungal root infection from a soil reservoir (Figures 4 and 6).

**Figure 6.** Inhibitory effects of various (±)-botryodiplodin concentrations in hydroponic culture medium (0 to 80 μg/mL) on lateral soybean seedling root growth (—♦—) with inhibition at IC50 = 4.2 μg/mL (100% dry weight = 11.5 ± 2.1 mg), and on tap root growth (- - - - - - -) with inhibition at IC50 = 23.5 μg/mL (100% dry weight = 13.8 ± 2.0 mg). Root growth presented on the vertical axis was measured as dry weight of excised lateral or tap roots after (±)-botryodiplodin exposure for 96 h at room temperature in continuous light. Results are the mean of three replicates ± SEM. \* Significantly reduced soybean root growth at *p* < 0.05; \*\* significantly reduced soybean root growth at *p* < 0.01 (Student's *t*-test).

A set of experiments exposing soybean seedling roots to (±)-botryodiplodin in sand culture was conducted to eliminate the possibility that physical contact of soybean roots with solid soil particles might induce or maintain a protective layer on roots. Seedlings germinated in soil were transplanted to sand culture, acclimatized, and then the roots exposed to (±)-botryodiplodin (10, 100, and 300 μg/mL) dissolved in fresh culture medium (Figure 5). (±)-Botryodiplodin treatment resulted in greatly reduced lateral root production, particularly at the higher concentrations. There was no indication that the sand used in the sand culture system interfered with phytotoxicity either by adsorption of toxin on silica surfaces, or by interfering with conduct of the experiment either by preventing continuous visual monitoring of toxin-induced damage or causing root damage when washing sand away.

Pink to red discoloration of exposed roots occurred at the highest (±)-botryodiplodin concentration (300 μg/mL) in sand culture (Figure 5), and at the higher concentrations tested in liquid hydroponic culture (Figure 4), with the darkest coloration at the highest concentration (80 μg/mL). Formation of pigment by reaction of botryodiplodin with protein and other amines has been observed numerous times [26,35,39,46,47]. The (±)-botryodiplodin used in the present study has been shown to react with proteins, amino acids, and a wide variety of other amines to give red to yellow pigments [48]. Given that soybean seedlings have been reported to express proteins such as nutrient and water transporters on root surfaces [49], the pink to red pigment observed on soybean seedling roots treated with (±)-botryodiplodin in hydroponic culture (Figures 4 and 5) may have formed by a similar reaction with root surface proteins.

The production of abundant lateral roots by soybean seedlings under the stationary hydroponic conditions used in this study (Figures 4 and 6) presumably results from disruption of oxygen and ethylene exposure to roots, which has been shown in *Arabidopsis thaliana* to be genetically defined and environmentally regulated [50,51]. Soybean has similar ethylene receptors and associated regulatory gene products [52], which provides an explanation for the well-documented occurrence of lateral root production by soybean when soil becomes waterlogged [53,54]. In plant root growth, cell division occurs solely in meristematic regions near root caps, and root extension primarily results from subsequent cell elongation. Botryodiplodin has been shown to target DNA synthesis and dividing cells in a wide variety of biological systems, including bacteria [26,55], fungi [56], yeast [26], plants [32], and mammalian cells [35,37,57–59]. A phytotoxin such as (−)-botryodiplodin, which kills dividing cells, would be expected to target meristematic tissue near the tips of both tap and lateral roots. The higher reduction in lateral root growth (~eight-fold) than in tap root growth (~two-fold) by (±)-botryodiplodin (Figure 6) is consistent with the toxin acting on meristematic tissue, which makes up a larger percentage of total tissue weight in small lateral roots than it does in the larger tap root.

Soybean seedlings growing in hydroponic culture with roots exposed to (±)-botryodiplodin (15 μg/mL) in culture medium (Figure 7) resulted in the loss of the root cap and meristematic tissue without involvement of a vector or physical injury, and were consistent with the toxin targeting dividing cells in the meristem. Similar loss of the root cap and meristematic tissue occurred at the higher (±)-botryodiplodin concentrations tested (35 and 80 μg/mL). Additional studies are needed to determine how rapidly the root tip loss occurs at various (±)-botryodiplodin concentrations.

**Figure 7.** Light micrographs of root tips of soybean seedlings after four days in hydroponic culture in 10% Villagarcia medium in water with no (±)-botryodiplodin (left panel, 400×) or with (±)-botryodiplodin (15 μg/mL) (right panel, 200×).

Thus, (±)-botryodiplodin applied externally to undisturbed soybean roots induced phytotoxic responses of a type expected to facilitate fungal root infection. An example of a plausible root infection mechanism involving the observed responses of soybean root to (±)-botryodiplodin could involve hyphae of a fungus like *M. phaseolina* propagating outward from a plant-derived nutrient source through the soil in all directions until hyphae detect the presence of a root tip, stimulating release of (−)-botryodiplodin. The released (−)-botryodiplodin would be expected to cause loss of the root tip and exposure of the vascular system that should facilitate the propagation of fungal hyphae into the vascular system and subsequently throughout the plant [19]. However, additional studies will be needed to confirm that targeting of root tip meristematic cells is involved in the actual root infection mechanism used by *M. phaseolina* in charcoal rot disease of soybeans in the field.

#### **3. Conclusions**

The toxin, botryodiplodin, produced by *M. phaseolina*, the fungus that causes charcoal rot disease in many plant species, is phytotoxic in soybean leaf disc cultures and in actively growing *Lemna pausicostata* plantlet cultures. Botryodiplodin exposed to undisturbed roots of soybean seedlings in hydroponic culture results in a root tip destruction response that would facilitate fungal infection of the root.

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

#### *4.1. Preparation of (*±*)-Botryodiplodin*

(±)-Botryodiplodin was selected for use in these studies, because it is readily synthesized chemically in larger amounts than were available by fermentation [32]. The mechanism of action of botryodiplodin has been extensively studied by Moule et al. [36–38], who provided evidence for non-enzymatic (i.e., chemical) crosslinking of DNA to protein. There have been no reports of botryodiplodin binding specifically to a chiral binding site on any enzyme or receptor. A non-enzymatic mechanism of action for botryodiplodin would result in phytotoxicity of synthetic (±)-botryodiplodin being equivalent to that of fermentation-derived (−)-botryodiplodin. The (±)-botryodiplodin used in this study was synthesized by preparing α-methyl-α-angelicalactone, using a modification of the method of Helberger et al. (1949) [60], followed by its conversion to the final product using four steps that are included in the synthetic method developed by Mukaiyama et al. (1974) [61] (Figure 8). Briefly, α-methyllevulinic acid (**1**) (500 mg) (TCI America, Portland, OR, USA), was treated with phosphoric acid (1% wt/wt) and subjected to vacuum distillation at 120–130 ◦C and ~40 Torr to provide α-methyl-α-angelicalactone (**2**) in approximately 80% yield. The product was treated under argon with boron trifluoride etherate and formaldehyde generated in situ by thermal degradation of paraformaldehyde. The reaction was quenched with NaHCO3 aqueous solution and extracted into dichloromethane. The product, (±)-cis-α-methyl-β-acetyl-γ-butyrolactone (**3**), was purified by chromatography on silica gel in diethyl ether:hexane 4:1 and crystallized from hexane. The ketone group of **3** was blocked with ethanethiol in the presence of zinc chloride and the product **4** extracted into dichloromethane and purified by chromatography on silica gel in diethyl ether:hexane 4:1. Reduction of **4** with diisobutylaluminium hydride in tetrahydrofuran at −78 ◦C yielded the diethanethiol derivative of (±)-botryodiplodin (**5**), which was purified by chromatography on silica gel using a 5% to 20% diethyl ether:hexane gradient. Unblocking of lactol **5** in acetone containing 1% water, CuCl2 and CuO was accomplished at room temperature in 30–60 minutes. (±)-Botryodiplodin (**6**) was extracted from the reaction mixture into dichloromethane and purified by chromatography on silica gel using ether:hexane 4:1 followed by re-chromatography on silica gel using dichloromethane:methanol 20:1 to yield 116 mg (20.9% overall yield) at a purity of >98% based on thin layer chromatography and nuclear magnetic resonance spectroscopy. The (±)-botryodiplodin (**6**) exhibited 1H nuclear magnetic resonance spectroscopy values and thin layer chromatographic Rf values identical to those reported in the literature [39,62] and those obtained in this laboratory with (−)-botryodiplodin purified from cultures of *M. phaseolina* [32], except that (±)-botryodiplodin was not optically active.

**Figure 8.** Chemical synthesis of (±)-botryodiplodin.

#### *4.2. Assay of Antibacterial Activity of Botryodiplodin*

The biological activity level of (±)-botryodiplodin was confirmed using antibacterial activity, the type of activity used to guide the initial isolation of (−)-botryodiplodin by Sen Gupta et al. (1966) [26], and the most easily measured of its numerous reported biological activities, including antifungal [56], phytotoxic [25,32], anti-cancer [37,57], mutagenic [36], and antifertility [58] activities. Antibacterial activity was compared on samples of (±)-botryodiplodin, prepared as described above, and (−)-botryodiplodin purified as described by Ramezani et al. (2007) [32] from culture filtrates of *M. phaseolina* isolated from a soybean plant with charcoal rot disease in Mississippi. Antibacterial activity was measured by serial dilution from 20 to 0.1 μg/mL in Mueller-Hinton broth in triplicate in the wells of a 96-well tray using (±)-botryodiplodin and (−)-botryodiplodin samples sterilized by dissolution at 10 mg/mL in 95% ethanol. The wells were inoculated with an actively growing culture of *Bacillus subtilis*, strain 1a1, isolated in this laboratory from lawn soil and shown to be susceptible to all antibiotics in a 28-member panel except thiostrepton. Trays were cultured overnight at 37 ◦C and bacterial growth estimated as the OD at 600 nm in a plate reader (BioTek Instruments Synergy HT, Winooski, VT, USA).

#### *4.3. Plant Growth and Environmental Conditions*

Soybean genotypes DT97-4290 (moderately resistant to charcoal rot) [18], DS97-84-1 and Saline [63] (both of which are susceptible to charcoal rot) were grown in the greenhouse. Seeds were planted and germinated in flat trays of vermiculite, and similarly sized seedlings were transplanted into 9.45 L pots filled with a silt loam soil (24% sand, 54% silt, and 22% clay with 1.3% organic matter) pH 6.5, 17 cmol/kg cation exchange capacity. Plants were watered as needed to maintain soil water potential at field capacity, i.e., between –15 to –20 kPa. Four pots were used for each soybean genotype, and three plants were grown in each pot. Greenhouse conditions were about 34 ◦C ± 8 ◦C during the day and approximately 28 ◦C ± 6 ◦C at night with a photosynthetic photon flux density of about 850–2100 <sup>μ</sup>mol·m−<sup>2</sup> <sup>s</sup><sup>−</sup>1, as measured by Quantum Meter (Spectrum Technologies, Inc., Aurora, IL, USA). The range of light intensity reflects the range from a cloudy day (850 <sup>μ</sup>mol·m−<sup>2</sup> <sup>s</sup><sup>−</sup>1) to a sunny day (2100 <sup>μ</sup>mol·m−<sup>2</sup> <sup>s</sup><sup>−</sup>1). The source of lighting in the greenhouse was a mixture of natural and artificial lights. Plant leaves were harvested during the vegetative phase of growth.

#### *4.4. Phytotoxicity of (*±*)-Botryodiplodin in Soybean Leaf Discs*

Dose-response curves were obtained for phytotoxic responses to a range of (±)-botryodiplodin concentrations by triplicate cultures of three soybean leaf discs. Leaf discs cut from healthy leaflets from true mature leaves of 3- to 4-week-old plants of each of the three soybean types were used to determine the phytotoxicity of (±)-botryodiplodin. All leaves were harvested in the laboratory and three soybean leaf discs measuring 4-mm diameter were cut with a sterile cork borer (#4) and placed in sterile 24-well tissue culture plates with low evaporative lids (Becton Dickinson and Company, Franklin Lakes, NJ). (±)-Botryodiplodin solutions in water (1.5 mL) over a range of concentrations (0, 2.5, 5.0, 10, 20, 40, 80, 160, and 320 μg/mL) were added to the wells of plates in triplicate. Leaf discs were incubated in a growth chamber at 25 ◦C under continuous visible light for 96 h and examined for signs of phytotoxicity after 24, 48, 72, and 96 h using the following symptom rating scale: Healthy tissue, 0%; a narrow zone of brown (necrotic) tissue forming around the edges of the leaf disc, 10%; a substantial zone of brown tissue forming around the edges of the leaf disc, 25%; brown tissue throughout the leaf disc, 50%; brown tissue throughout the leaf disc with bleaching, 75%; complete bleaching of the leaf disc, 100% (Figure 2).

#### *4.5. Phytotoxicity of (*±*)-Botryodiplodin in Duckweed Plant Cultures*

Dose-response curves were obtained for phytotoxic responses to a range of (±)-botryodiplodin concentrations by triplicate cultures of three-frond duckweed plantlets. Cultures containing three duckweed (*Lemna pausicostata* Helgelm.) plantlets were used to bioassay phytotoxicity, as described by Tanaka et al. (1993) [64], with some modification. Briefly, three duckweed plantlets containing three fronds each were transferred from a laboratory maintenance culture with clean forceps to each well of a sterile 24-well tissue culture plate with a low evaporation lid. Aliquots (1.5 mL) of culture medium containing a range of (±)-botryodiplodin concentrations (0, 0.03, 0.06, 0.13, 0.25, 0.5, 1, 2, 4, 8, 16, 32, and 64 μg/mL) were added in triplicate to the wells of culture plates. Duckweed plants were subsequently incubated in a growth chamber at 25 ◦C under continuous light for 96 hr. Duckweed plantlets were observed for signs of phytotoxicity after 24, 48, 72, and 96 hr. Growth was measured as addition of fronds in treated cultures relative to control cultures not treated with (±)-botryodiplodin. No additional fronds being produced in a treated culture was scored as 100% inhibition of growth.

#### *4.6. Hydroponic Culture of Soybean Seedlings*

The effects of (±)-botryodiplodin on soybean root growth were investigated in soybean seedlings germinated from a commercial soybean seed variety assumed to be charcoal rot-susceptible (Kansas Soybean Commission, Topeka, KS, USA) in autoclaved soil and grown under continuous light to the cotyledon stage (VC, 4–7 cm). Seedling roots were washed free of soil particles and transplanted to hydroponic growth medium. Seedlings were grown under hydroponic conditions for four days before use in root toxicity assays in individual 16 × 100 mm glass tubes containing 5 mL of a mixture of 90% distilled water and 10% Villagarcia medium [65]. The Villagarcia medium used consisted of distilled water (999 mL) containing CaSO4.2H2O (690 mg), KH2PO4 (34 mg), KNO3 (200 mg), MgSO4.7H2O (61 mg), and 1 mL of 1000-fold concentrated microsolute nutrient solution containing FeSO4.7H2O (50 mg), KCl (14 mg), H3BO4 (5.7 mg), MnSO4.H2O (1.5 mg), ZnSO4.7H2O (2.6 mg), CuSO4.5H2O (0.45 mg), and (NH4)6Mo7O24 (2.1 mg). Seedlings were held in place by the tube walls and maintained with roots covered with medium added daily as needed. Seedlings placed under sand culture conditions were grown four days in 5 mL of washed, sterile sand, which was added after the seedling was placed in the tube and kept soaked with 10% (*v*/*v*) Villagarcia medium in water added daily as needed.

#### *4.7. Root Toxicity of (*±*)-Botryodiplodin in Soybean Seedlings in Hydroponic Culture*

Dose-response curves were obtained for phytotoxic responses to a range of (±)-botryodiplodin concentrations by the roots of groups of three soybean seedlings cultured individually in hydroponic medium. Soybean seedlings were grown in continuous light for four days at room temperature in 5 mL of hydroponic growth medium consisting of 10% Villagarcia medium and 90% water in individual 16 × 100 mm glass tubes, using the walls of the glass tubes to hold the seedlings upright. The medium was withdrawn from seedling cultures with a Pasteur pipet, replaced by fresh medium containing a range of (±)-botryodiplodin concentrations in triplicate in three individual glass tubes (0 to 300 μg/mL in an initial range-finding assay, and 0 to 80 μg/mL in subsequent studies), and cultured for an additional four days at room temperature in continuous light. Root growth was quantified by removing seedlings from the culture tubes and washing the roots with a stream of deionized water from a wash bottle. Roots were excised at the stem line with a scalpel. Lateral roots were cut from the tap roots and the two root types dried separately overnight under vacuum in a desiccator over Drierite desiccant at room temperature. Lateral and tap roots were weighed separately on a sensitive balance (Mettler Toledo UMX2 Ultra-Microbalance, Mettler-Toledo International, Columbus, Ohio), and the dry weights of triplicate samples plotted as mean ±standard error versus (±)-botryodiplodin concentration.

#### *4.8. Light Micrographs of Soybean Seedling Roots Exposed to (*±*)-Botryodiplodin in Hydroponic Culture*

Soybean (commercial variety) seedlings were established in hydroponic culture as described above, then transplanted to individual new 16 × 100 mm glass tube containing 5 ml of (±)-botryodiplodin (0, 15, 35, and 80 μg/ml) in 10% Villagarcia medium and 90% water. Seedlings were incubated at room temperature for 4 days with continuous light, at which time control seedlings in 0 μg/ml (±)-botryodiplodin had abundant lateral roots. Seedlings in 15 μg/ml (±)-botryodiplodin had substantially reduced numbers of lateral roots and seedlings in 35 and 80 μg/ml (±)-botryodiplodin had stunted roots stained pink. An Xacto knife was used to cut the roots off at slightly above where the root begins. The excised roots were placed in labeled glass scintillation vials filled to the top with Karnovsky's fixative [66]. Root samples were embedded in resin and thick-sectioned on an Ultracut UCT microtome (Leica, Buffalo Grove, IL, USA) using a diamond knife. Sections were collected on glass slides, stained with toluidine blue, and imaged using bright-field light microscopy on an Eclipse 90i (Nikon Inc., Melville, NY, USA) with a D2-Fi2 color camera running Nikon Elements software.

#### *4.9. Data Analysis*

Phytotoxic responses were quantified as IC50 values (the concentration of (±)-botryodiplodin that causes 50% of the maximal toxic response) determined graphically by interpolation on plots of toxic response versus log (±)-botryodiplodin concentration prepared using the graphing package included in Microsoft Excel 2010. Statistical analyses (correlation analysis, multiple linear regression analysis, Student's *t*-test) were conducted using the statistical package included in Microsoft Excel 2010. *p* ≤ 0.05 was considered significant.

**Author Contributions:** Conceptualization, H.K.A. and W.T.S.; Formal analysis, A.M.B. and W.T.S.; Funding acquisition, H.K.A.; Investigation, H.K.A., N.B., A.M.B., J.L.N., M.A.-K. and W.T.S.; Project administration, H.K.A.; Writing—original draft, H.K.A.; Writing—review & editing, N.B., A.M.B., J.L.N., M.A.-K. and W.T.S. All authors have read and agree to the published version of the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported in part by the Mississippi Soybean Promotion Board and the Mississippi State University Special Research Initiatives grants program.

**Acknowledgments:** The authors are grateful to Bobbie J. Johnson (retired), Jeremy Kotowicz, and Vivek H. Khambhati for their technical assistance. Trade names are used in this publication solely for the purpose of providing specific information. Mention of a trade name, propriety product, or specific equipment does not constitute a guarantee or warranty by the USDA-ARS and does not imply approval of the named product to exclusion of other similar products.

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

#### **References**


© 2020 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* **Ergochromes: Heretofore Neglected Side of Ergot Toxicity**

**Miroslav Flieger 1, Eva Stod ˚ulková 1, Stephen A. Wyka 2, Jan Cern ˇ ý 3, Valéria Grobárová 3, Kamila Píchová 1, Petr Novák 1, Petr Man 1, Marek Kuzma 1, Ladislav Cvak 4, Kirk D. Broders <sup>2</sup> and Miroslav Kolaˇrík 1,\***


Received: 25 June 2019; Accepted: 23 July 2019; Published: 25 July 2019

**Abstract:** Ergot, fungal genus *Claviceps*, are worldwide distributed grass pathogens known for their production of toxic ergot alkaloids (EAs) and the great agricultural impact they have on both cereal crop and farm animal production. EAs are traditionally considered as the only factor responsible for ergot toxicity. Using broad sampling covering 13 ergot species infecting wild or agricultural grasses (including cereals) across Europe, USA, New Zealand, and South Africa we showed that the content of ergochrome pigments were comparable to the content of EAs in sclerotia. While secalonic acids A–C (SAs), the main ergot ergochromes (ECs), are well known toxins, our study is the first to address the question about their contribution to overall ergot toxicity. Based on our and published data, the importance of SAs in acute intoxication seems to be negligible, but the effect of chronic exposure needs to be evaluated. Nevertheless, they have biological activities at doses corresponding to quantities found in natural conditions. Our study highlights the need for a re-evaluation of ergot toxicity mechanisms and further studies of SAs' impact on livestock production and food safety.

**Keywords:** mycotoxins; ergot alkaloids; ergochromes; secalonic acid; food safety; cereals; tetrahydroxanthones; *Claviceps*

**Key Contribution:** Ergot alkaloids (EAs) were considered as the only factor responsible for the ergot toxicity. We showed—using the robust sample size—that ergochromes are similarly abundant as ergot alkaloids and possess high toxicity to human cells. Thus, their importance for human and animal health should be further investigated.

#### **1. Introduction**

Ergot, the genus *Claviceps* (Ascomycota: Hypocreales) includes obligate plant parasitic fungi that develop in the ovary of grasses (including cereals), sedges, and rushes and form sclerotia containing toxins. The famous rye ergot, *Claviceps purpurea*, is a member of the section *Claviceps* which is specified by the production of highly toxic ergopeptines [1]. Recently it has been shown, that *C. purpurea sensu lato* (*s. l.*) is a complex of four cryptic species with different host grass spectra. While common land grasses are often infected by both *C. purpura sensu stricto* (*s. s.*) and *C. humidiphila*, cereal crops seem to be infected by just *C. purpurea s. s.* In the Palearctic region, these two species are the most important from an agricultural point of view [2,3]. Furthermore, recent findings suggest further cryptic diversification among North American *C. purpurea s. l.* specimens and their impacts on agricultural remains uncertain [4].

Ergot poisoning causes ergotism in humans and livestock. Most of the research related with ergotism has been focused on the ergot alkaloids (EAs) as these are among the most important natural pharmaceuticals and toxins in human history [5,6]. There is still some pharmaceutical research being conducted on EAs, however, research into their toxic effects on human health have relatively diminished. Humans are no longer at risk of ergotism in most of the world due to advanced seed cleaning and food screening for the presence of EAs. However, there has been a resurgence of research interested in the toxicoses of livestock or wild animals in recent years which has brought to light the substantial challenge of elucidating alkaloid-induced effects of animal responses to exposure [7,8]. Currently, the EU Scientific Panel on Contaminants in the Food Chain of the European Food Safety Authority (EFSA) recommended 12 priority alkaloids for monitoring in food and feed, all of which are grouped in the ergopeptines produced by *C. purpurea s. l*.

A recent review by Klotz [9] detailed the collective knowledge on ergotism and fescue toxicoses of livestock. He noted the complexity of this area of research as EA toxicity is affected by changes in alkaloid concentrations, proportions, and availability as well as individual's genetic predispositions, prior exposures, and ambient environments. This led to the overall conclusion that the impacts of EAs on livestock, especially convulsive (neurological and abortogenic) symptoms are not caused by the sole action of a single toxin, but rather the combined impact and synergistic action of multiple EAs derived from *Claviceps* and *Epichloë* species [9,10]. While most of the available data of EAs effects on animals both address the symptoms and define the problem, there is still a lack of research on other fungal metabolites and their potential harmful effects on livestock and humans.

In *C. purpurea s. l.*, the average EAs content in sclerotia varies between studies and ranges from 0.01–1.3 mg/g [11–16] to 2.88–7.26 mg/g [17], with individual values rarely reaching 5–10 mg/g (d/w) [12,16,18]. In addition to EAs, *C. purpurea* produces many other secondary metabolites. Most have been identified and inspected for toxicity while others have still eluded proper examination, with no research on their effects on livestock [10]. In addition, some of these other metabolites are produced in greater quantity than the heavily researched EAs. Ergot sclerotia can contain 1–2% of pigments, predominantly yellow biphenyl pigments called ergochromes (ECs) [19], which typically reach 5 mg/g of the sclerotia dry weight [20]. The main ECs of ergot are secalonic acids A–C (SAA, SAB, SAC), whereas related ergoflavin, ergochrysin A, B and chrysergonic acid are produced in negligible amounts. Other minor pigments, such as anthraquinone derivates endocrocin and clavorubin, are present [21]. Secalonic acids D–F were described from various moulds and lichens [22,23]. Secalonic acids exhibit various biological activities with the best studied secalonic acid D (SAD) showing mutagenic, teratogenetic, and cytotoxic activity [24–26]. Strong biotoxic activity against animals, plants, or microbial cells was also documented in secalonic acid A [27,28], F [29], and G [30].

Surprisingly, across the distribution of ergot species, the quantity and environmental role of ECs are still unknown, despite their proved activity against mammalian cells. There is currently no knowledge on the effect of ECs on livestock or how synergistic actions of ECs with other ECs or EAs affect their toxicity on livestock or humans. Therefore, we pose the following question: Could ECs represent an important and so-far neglected part of ergot toxicity? For that purpose, we quantified ECs content across a large set of ergot sclerotia and present the basic cytotoxicity assays on human cell lines.

#### **2. Results**

#### *2.1. Ergot Alkaloids and Ergochromes Content*

Three major SAA, SAB, and SAC and two minor ECs, endocrocin, and ergochrysin were identified in sclerotia. The average content of all SAs and EAs across all 111 samples was 4.08 mg/g (SD 4.39) and 3.58 mg/g (SD 2.46), respectively (Table 1, Figures 1 and 2, Table S1). These values were statistically not different (paired *t*-test, *p* = 0.3) and moderately, but significantly correlated (Figure 3, Pearsson

coefficient 0.46, two tailed *t*-test, *p* < 0.001). The SAs content and proportion differed substantially between species and between European and North American (NA) populations of *C. purpurea s.s. Claviceps arundinis* and *C. humidiphila* contained significantly more SAs than EAs and had the highest content of all SAs among analyzed species. The dominance of SAs over EAs was also found in *C. capensis*, *C. macroura*, *C. monticola*, *C. pazoutovae*, and *C. fimbristylidis* (not tested for significance due to the low sample size). The opposite ratio was found in *C. purpurea* (all sample set), *C. spartinae* (significant different in both cases), *C. cyperi*, and *C. nigricans* (not tested for significance). European *C. purpurea s. s.* population had similar content of SAs and EAs, whereas the NA population had significantly more EAs than SAs (Table 1, Figure 2, Table S1). Both populations differed significantly in SAs (Mann-Whitney test, *p* < 0.005), but not in EAs content. The spectrum of SAs was shown to have an obvious chemotaxonomic value which separates European *C. purpuera* and *C. spartinae* (SAA as a dominant and often single secalonic acid), from the NA *C. purpurea* (SAC dominant, followed by SAB and SAA) and *C. arundinis* (SAB dominant, followed by SAC) and *C. humidiphila* (SAB dominant, SAC minor) (Figure 4, Table S1).

**Table 1.** Concentration summary of ergochromes and ergot alkaloids in *Claviceps* spp. sclerotia. Content in dry sclerotia is expressed as minimum-average (standard deviation)-maximum. The category of other species includes: *Claviceps* sp. 1, sp. 2, *C. capensis*, *C. cyperi*, *C. fimbristylidis*, *C. macroura*, *C. monticola*, *C. nigricans*, and *C. pazoutovae* (Table S1).


secalonic acid A

secalonic acid C

**Figure 1.** Structures of the main ergochromes from this study, secalonic acid A–C.

**Figure 2.** Box plot graph summarizing the total content of secalonic acids and ergot alkaloids in the dry sclerotia. The category of other species includes: *Claviceps* sp. 1, sp. 2, *C. capensis*, *C. cyperi*, *C. fimbristylidis*, *C. macroura*, *C. monticola*, *C. nigricans*, and *C. pazoutovae* (Table S1). \*\*\* *p* < 0.001, \*\* *p* < 0.01, \* *p* < 0.05, ns—not significant, in the Wilcoxon Signed Rank test.

**Figure 3.** Linear regression of total secalonic acid (SAA, SAB, SAC) and ergot alkaloid content in sclerotia.

**Figure 4.** Principal component analysis showing relatedness of samples based on content of SAA, SAB, and SAC. PCA (Principal Component Analysis) axis 1 and 2 explained 76.3% and 16.1%, respectively.

#### *2.2. Biological Activity In Vitro*

Studied compounds were tested using cancer-derived Jurkat (Figure 5) and HeLa cell lines along with primary skin-derived fibroblasts (Figure 6). Toxicity tests on Jurkat cells (24 h exposure, Figure 5) showed that endocrocin (anthraquinone) did not cause apoptosis or cell death even in extremely high concentrations (LD50 705.5 μg/mL). The toxicity of ergoxanthine (LD50 142.0 μg/mL) and ergochrysin (LD50 118.8 μg/mL) was lower in comparison to SAs and EAs. Toxicity after 24 h exposure to SAA, SAB, SAC, and EA began at the lowest tested concentration (0.75 μg/mL), with 12 (EAs) and 25 (SAs) μg/mL resulting in 50%, and 50 (EAs) and 100 (SAs) μg/mL resulting in 100% of dead or apoptotic cells in all variants. The fraction of apoptotic and dead cells only slightly differed between these compounds, with EAs (LD50 13.5 μg/mL) showing the highest toxicity, followed by SAA (LD50 19.8 μg/mL), SAB (LD50 35.9 μg/mL), and SAC (LD50 36.5 μg/mL). For SAA + EAs (LD50 12.8 μg/mL) and SAB + SAC + EAs (LD50 15.4 μg/mL) neutral to synergistic effects on the number of dead cells could be observed.

Strong effects including changes in morphology and cell contact was observed in human primary fibroblasts and HeLa cells. At concentrations above 25 μg/mL cells incubated with SAA, SAB, SAC, and EA stopped dividing and showed loose stress fibers which started to detach. Striking effects on cellular morphology could be detected using Mitotracker probe (Figure 6). In the case of secalonic acid, concentrations above 12 μg/mL elicited rounded mitochondria with extremely high positivity in greater than 50% of fibroblasts and 100% of HeLa cells, indicating modulation of mitochondrial function, namely proton gradient value. In contrast, specific effects of ergotamine to both HeLa (100% affected cells above 50 μg/mL) and fibroblasts (25 μg/mL) caused the formation of swollen vacuolar structures. We can estimate that these swollen structures (not acidic—negative for Lysotracker probe) could stand for more than half of the cellular volume. Representative data for SAA and ergotamine treatments (various concentrations and simultaneous detection of actin cytoskeleton, mitochondria, and nuclei) are shown as Figure S1.

**Figure 5.** Live, Apoptotic, and dead cell events in Jurkat cells after 24 h incubation. Representative toxicity values for endocrocin, ergochromes, ergotamine, and the combinations of SAs and ergotamine. Combinations reflect the mixtures of SAs and EAs found in the real sclerotia. Cells are shown as events detected by flow cytometry and expressed in percentages (total amount of cells = 100% of events).

**Figure 6.** Mitochondrial architecture and Mitotracker positivity in HeLa cells and primary skin-derived fibroblasts. Cell cultures grown on glass cover slips were treated for 24 h with 12 μg/mL secalonic acid A or ergotamine and in vivo incubated with MitoTracker® Red CMXRos. Magnification 20×.

#### **3. Discussion**

#### *3.1. Ergochrome Quantity and Distribution Across the Species*

Ergochrome production in ergot fungi seems to be limited to the section *Claviceps* [1]. This section, represented by *C. purpurea s. l.*, is the most widely distributed section of the genus *Claviceps* and infects the largest number of host plants [1,2]. Due to its cosmopolitan distribution with over 400 potential hosts, including many economically important crops such as rye, wheat, triticale, and barley, the EFSA has selected *C. purpurea s. l.* to be the focus of chemical analysis for food safety concerns [7]. Among the members of *C. purpurea s. l.*, only *C. purpurea s. s.* and *C. humidiphila* infect cereals or cultivated and wild forage grasses and thus have an agricultural importance. *Claviceps cyperi*, infecting *Cyperus* spp. is distributed only in South Africa and is a known causal agent of ergotism in cattle [31]. *Claviceps arundinis* (mostly on *Phragmites*, *Mollinia*, or *Leymus arenarius*) and *C. spartinae* (*Spartina*, *Distichlis*) can be very common in their particular habitats, but their grass hosts are not significantly grazed by animals [2,3,32,33]. Populations of others, i.e., Paleartic *C. nigricans* (*Eleocharis, Scirpus*) or South African *Claviceps capensis* (*Ehrharta*), *C. fimbristylidis* (*Fimbristylis*), *C*. *pazoutovae* (*Ehrharta* and *Stipa*), *C. macroura* (*Cenchrus*), or *C. monticola* (*Brachypodium*) are rare in abundance and their hosts have low palatability [34,35].

Three major combinations of SAs in sclerotium were found and their mutual ratios have taxonomic value. Franck [36] analyzed three *C. purpurea* sclerotia from rye and found SAA as the dominant SA, followed by SAB and SAC, which is concordant with our data for SAs in European *C. purpurea s. s.* Concerning the total content of ECs, the only reliable publication reports 5 mg/g [20] which fully corresponds with our results. Besides SAs, other pigments were also found in several samples (*C. arundinis*, *C. humidiphila*, and *C. spartinae*) but in negligible quantity. The observed quantity of endocrocin (maximum 205 μg/g) corresponds with Franck [20] which reports maximal values of 40 μg/g (Table 1, Table S1).

Based on our data, the total EAs and SAs content in sclerotia was significantly correlated and the amount of toxic compounds in a single sclerotium is thus cumulative. This correlation also shows that measuring EAs can be used to some extent as a proxy for SAs abundance. It is already known that the pigment content of sclerotia is proportional to its alkaloid content and thus a pigments quantification (namely clavorubrin) can provide a method for measuring the samples' toxicity [37,38].

Both toxin groups are metabolically independent, but their production in sclerotia is contemporary and seems to be regulated by the same stimuli (phosphate level) [39], which can explain the observed correlation. Fungal pigments typically have a light protection role, and this ability is also expected in the case of ECs [39]. Contrary to this, EAs are light sensitive [40] and the correlation between the contents of EAs and SAs could also just be a consequence of light induced degradation of EAs in the sclerotia with primary lower contents of pigments (SAs).

#### *3.2. Ecological Role of SAs*

From an ecological point of view, SAs can contribute to ergot toxicity (see below) and are thus involved in the protective mutualism known in *C. purpurea s. l.* [41]. SAs can play an important role in the protection of the sclerotia against light or in antibiosis (resistance to microbial attacks). In particular, SAB showed activity against *Bacillus megaterium*, *Escherichia coli*, and *Microbotryum violaceum* [42], and SAA showed antibiotic activity against *Bacillus subtilis*, *Piricularia oryzae* [27], *Microccocus luteus* (MIC 4–8 μg/mL), and *Enterococcus faecalis* (MIC 32 μg/mL) [43]. SAA is also a highly potent non-host specific phytotoxine acting as a possible virulence factor of *Pyrenophora terrestris* [28]. Interestingly, the whole section *Claviceps*, which is unique due to the presence of ECs, also has an extraordinary broad host range [1]. This suggests that ECs could potentially play some role in the virulence cycle, but are surely not essential for it, as was shown in infections test with mutant *C. purpurea* strains with blocked ECs synthesis [39].

#### *3.3. Ergochromes Toxicity and Significance*

The main aim of this study was to determine the quantity of ECs produced by *C. purpurea* and other members of the section *Claviceps* to help assess their potential role in ergotism pathogenicity. The effort to control the toxicosis of forage and human food has generally been successful; however, there has been a resurgence of research interested in the toxicosis of livestock due to increased ergot abundance in recent years [44] and continual reports of ergotism on livestock.

While many researchers continue to focus on compounds that possess the tetracyclic ergoline ring of ergot alkaloids, we showed that so far neglected SAs are equally abundant as alkaloids and have similar toxicity to cell cultures. While our data shows the essential first step, further research into the toxicity of SAs is needed to elucidate their real impact on human and animal health. Understanding the connection of SAs to ergotism might help explain some of the unexplained aspects of egotism. Klotz [9] published a thorough review of the inconsistent nature and occurrence of ergotism in livestock. While his review only covered the toxicosis of *Claviceps*-derived ergotamine and ergocristine and *Epichloë*-derived ergovaline, he was still able to determine that the impacts of alkaloids on livestock are not caused by the sole action of a single toxin but rather the combined impact and synergistic action of multiple alkaloids. This was evident as many researchers were generally unsuccessful in replicating the complete effects of ergotism by introducing individual or even combinations of multiple alkaloids to livestock. For example, individual applications of ergovaline and ergotamine and combined applications of ergocornine, ergocryptine, and ergocristine were unable to produce gangrenous ergotism or fescue foot in all of the exposed livestock [45–48]. Similar inconclusive results were also observed for other aspects of ergotism such as fat necrosis and male-specific effects [9]. Such inconsistencies might be the results of concentration levels, alkaloid proportions, isomeric forms, accumulation, as well as individual's genetic predispositions, prior exposures, and ambient environments.

Therefore, SAs might represent a missing piece in the larger picture of ergotism on livestock. A few experiments on living animals were conducted with SAs. The major compound of ergot sclerotia, SAA, is lethal to mice at the peritoneal injected doses 50–100 mg/kg [27]. In another study, SAA caused edema and inflammation in rats after the intraperitoneal application of 12.5–50 mg/kg [49]. This toxicity is comparable to well recognized toxin SAD which LD50 for mice in intraperitoneal injection was reported as 42 mg/kg [50] or in a range of 26.5–51.7 mg/kg [24]. Teratogenic effects were observed in rats injected with SAD and teratogenic activity started at doses of 5 mg/kg body weight [25,51]. Data from humans are missing, but it is known that SAD produces cleft palate as the only malformation in fetal mice which is of potential relevance to human health [52]. Furthermore, SAs have various biological effects at non-toxic doses. SAA at the concentrations 0.15–0.75 mg/kg injected peritoneally to mice affect the metabolism of dopaminergic neurons [53,54].

In cases of cell line cultures, the toxicity to cancer cells is much higher than for normal cells. SAA is toxic against cultured mouse (IC50 = 0.5 μg/mL, [55]) or human leukemia cells (ED50 = 3.5 μg/mL, [43]). These values roughly correspond to the LD50 ranging between 20–37 μg/mL found in SAA-SAC for Jurkat cells in our study. The IC50 for murine melanoma cells was 1.8 μg/mL for SAA and 0.18 μg/mL for SAD, but to affect healthy keratinocytes a 3.5× (SAA) or 13× (SAD) higher concentration was needed [56]. SAD is also cytotoxic against various carcinoma cell lines at very low concentrations (IC50 = 0.05–0.76 ug/mL) and its mechanism of function is the best studied among all ECs [57].

Thus, the question is whether SAs of ergot origin can negatively affect animal health. The only available data are for SAD, and optical antipode of SAA, which have very similar toxicity to animal models and potentially have the same mode of action [58]. Whereas, at the intraperitoneal application, the LD50 for SAD and SAA ranges between 25–37 mg/kg (applied one times) while there is an 11-fold lower sensitivity when introduced by an oral route in mice [59]. Thus, an LD50 of around 400 mg/kg body weight can be close to toxicity values found in field conditions. Scenarios accounting for repeated applications of sublethal doses increase the toxicity by approximately three times as toxic effects of SAD were found to be cumulative in a five-day feeding experiment, which indicates that the toxin can be accumulated within the organism [59]. The question, what is the exposition of livestock to SAs under

natural conditions, remains unanswered. In EAs the lowest save doses (no-observed-adverse-effect levels) are 0.22–0.60 mg/kg body weight per day [7]. Concerning the fact that SAs and EAs have very similar concentrations in ergot, we can expect that animals encounter similar doses. Based on our above review, the lethal toxicity of SAs is more than 125 times lower than in EAs and their importance in the acute intoxication can play a role in case of highly contaminated fodder only. Nevertheless, SAs have various biological activities in much lower doses (i.e., 0.15 mg/kg injected peritoneally, [53,54]) corresponding to expositions in natural conditions.

#### *3.4. Mode of Action*

SAs exhibit various bioactivities, generally cytotoxic and cytostatic. SAD and SAF exhibit antitumor activity in low micromolar ranges, including induction of cell cycle arrest in G1 or induction of apoptosis [60,61]. The molecular mechanism behind the antitumor selectivity and toxicity remain putative. Opposite to cytotoxic effects, SAA applications in a Parkinson's disease mouse model protected against dopaminergic neuron death [53]. The best studied SA, SAD, exhibited various modes of cytotoxicity towards multidrug resistance cells due to induction of ABCG2 degradation via calpain-1 activation [57]. Our observations agree with this general view that particular bioactivities are cell type dependent (in our case leukemic cell line, adenocarcinoma, and primary human fibroblasts) in terms of effective concentrations and particular phenotypes (proliferative block, loss of stress fibres, proportion of apoptotic cells). The hallmark physiological and morphological changes in all cell types treated with SAs was transformation of the typical mitochondrial network to isolated rounded mitochondria with extremely high positivity for Mitotracker probe, indicating interference with mitochondrial function. For example, electron-transport chain functions and proton gradient levels. This is in agreement with previously published results that SAD uncoupled the oxidative phosphorylation in isolated rat mitochondria [62]. Another indication, that mitochondria could be the cellular target for SAs is based on the observation that SAA protects dopaminergic neurons from 1-methyl-4-phenylpyridinium induced cell death via the mitochondrial apoptotic pathway [54].

#### **4. Conclusions**

Despite the fact that ergochromes do not play a significant role in acute toxicity of ergot, the complexity of the secondary metabolites produced by *Claviceps* species including secalonic acids points to a more complex fungus-grazing animal interaction. Grazing animals consuming ergotised grasses are constantly exposed to doses of secondary metabolite mixtures that can have a profound biological effect on their physiology or development. This publication is focused on the major secondary metabolite family, SAs, which can offer the tempting explanation to the complex ecological interaction between the herbivore and the fungus. Historically, the major body of research was focused on the ergot alkaloids. The observation that *Claviceps* secondary metabolome is much more diverse than anticipated, providing new research direction. Future research needs to examine metabolism, absorption, and excretion of SAs to determine their potential in short term versus long term toxicosis. For example, the described effect of SAs on mitochondria (temping molecular mechanism behind the bioactivity) indicating high interference with function needs detailed studies to be understood completely. These research directions are essential to understand ergotism itself in its complexity and to help determine whether forage and human food contaminated with *Claviceps purpurea* should be monitored for the control of ECs content, as proposed in this publication.

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

#### *5.1. Specimens Analyzed*

The analyzed sclerotia (*n* = 111) covers material from four continents. All 13 agriculturally and environmentally important ergot species producing ergochromes, were collected from the wild as well as cultivated grasses and cereals; i.e., *C. purpurea s.s*. (collections from Europe, North America, New Zealand), *C. humidiphila* (Europe), *C. arundinis*(Europe), *C. nigricans*(Europe), *C. spartinae*(Europe), *C. cyperi*, *C. fimbristylidis*, *C. macroura*, *C. monticola*, *C. pazoutovae*, (all from South Africa), and two undescribed *Claviceps* sp. (USA, New Zealand). Materials originated from previous studies [2,34,63] or were collected during the course of this study (Table 1, Table S1). Sclerotia were identified using the ITS rDNA sequence barcode. DNA from sclerotia was isolated using the fast NaOH protocol [64] for European samples or using the PowerPlant Pro DNA Isolation Kit (MoBio/Qiagen) for US/NZ samples. PCR and sequencing of the ITS barcode was performed according to Pažoutova et al. [2]. DNA and chemical analyses were done from the same sclerotium in the case of larger sclerotia. In cases of very small sclerotia, chemical analysis was performed from the whole sclerotium and DNA based identification was completed using a sclerotium collected from the same or adjacent grass spiclet.

#### *5.2. Sample Preparation, Extraction, and HPLC Analyses*

Pulverized sclerotia (1–10 mg) were mixed with extraction mixture dichloromethane and concentrated ammonia (500:1, *v*:*v*, 0.5–2.0 mL) and gently stirred for 1 h. Supernatant was separated by centrifugation and kept in the freezer until use. The same HPLC instrumentation and method as published earlier [1] was used for the analysis of EAs and ECs in sclerotia of all *C. purpurea s.s.* analyzed.

#### *5.3. Ergot Alkaloids and Ergochromes Identification and Quantification*

#### 5.3.1. General Workflow

Identification of EAs was based on retention time which was compared with standard compounds and UV–VIS spectra of individual compounds. Secalonic acids A, B, and C, were isolated and identified using UV–VIS spectra, FTMS, NMR, and optical rotation. The isolated compounds were used as chromatographic standards for quantification of their content in individual sclerotia. Further isolated ECs were determined as endocrocin and ergochrysin by the same procedures as used for the SAs. In both cases the data obtained were in agreement with previously published data [20].

#### 5.3.2. FTMS

Samples were measured using 15T solariX FTMS equipped with an ESI/MALDI ion source and ParaCell (Bruker Daltonics, Billerica MA). The analysis was performed using an electrospray ionization (ESI) in a positive ion mode as described in Flieger et al. [65] with the following differences: The collision energy was kept at −15.5 V, the mass range for MS data acquisition started at *m*/*z* 150 a.m.u., resulting in a resolution of 250,000 at *m*/*z* 400. The detailed FTMS data for SAA-SAC are presented in Figure S2.

#### 5.3.3. NMR

NMR spectra were measured on a Bruker Avance III 600 MHz spectrometer (600.23 MHz for 1H, 150.93 MHz for 13C) in CD3CN (20 and 30 ◦C). Residual signals of solvent were used as an internal standard (at 20 ◦C: δ<sup>H</sup> 1.941, δC 1.41, at 30 ◦C: δ<sup>H</sup> 1.936, δ<sup>C</sup> 1.35). NMR experiments: 1H NMR, 13C NMR, COSY, 1H-13C HSQC, 1H-13C HMBC, and ROESY were performed as described in Stod ˚ulková et al. [66]. Detailed NMR data of the identified compounds are provided in Table S2 (13C NMR) and Table S3 ( 1H NMR).

#### 5.3.4. Quantification of Ergot Alkaloids and Ergochromes

A standard solution of quantified compounds, i.e., ergotamine and ergochrysin, were prepared in methanol and SAA in acetone at final concentrations of 31.75, 62.5, 125, 250, 500, and 1000 μg mL−1. The calibration graphs were constructed by plotting the integrated peak areas of individual compounds versus concentration. The following linear regression equations and correlation coefficients were obtained: Ergotamine; y = 5298.3x, *R*<sup>2</sup> = 0.9994, UV = 315nm; secalonic acid A; y = 14537x + 6452, *<sup>R</sup>*<sup>2</sup> <sup>=</sup> 0.9995, UV <sup>=</sup> 315nm; ergochrysin; y <sup>=</sup> 6205.7x <sup>−</sup> 2015, *<sup>R</sup>*<sup>2</sup> <sup>=</sup> 0.9991, UV <sup>=</sup> 315nm.

#### *5.4. Biological Activity Testing*

For toxicity studies, immortalized T-lymphocyte-derived cancer cell line Jurkat was cultivated in 96-well plates at a density of 2 <sup>×</sup> 105 cells in a final volume of 300 <sup>μ</sup>L of RPMI1640 (LONZA, USA). Cells were treated with different concentrations of SAs, ergoxanthin, endocrocin, ergotamine, and a combination of SAA and ergotamine (1:1) and SAB, SAC, and ergotamine (0.5:0.5:1) (Figure 5). Cells cultured in RPMI1640 and in RPMI1640 with only DMSO were only used as negative controls. After the incubation (24 h), cells were washed in PBS containing 0.02% gelatine and 0.01% sodium azide (Sigma-Aldrich, St. Louis, MO, USA). Hoechst 33258-stained cells were analyzed with the FACS LSRII instrument (BD Biosciences, San Jose, CA, USA) and FlowJo 10.5.3 software (Tree Star, Ashland, OR, USA).

For fluorescent microscopy adenocarcinoma cell line HeLa and primary human skin-derived fibroblasts were used. Cells were cultivated in DMEM medium with 10% FCS (Gibco, Invitrogen, Grand Island, NY, USA) and seeded on glass cover slips (up to density 50%) in 24-well plates. Cells treated with different concentrations of tested compounds were cultured in DMEM, wells supplemented with only DMSO were used as negative controls. After 24 h incubation, cells were incubated (10 min, Lysotracker® Red, or MitoTracker® Red CMXRos (Molecular Probes-Invitrogen, Carlsbad, CA, USA) and fixed (3.7% paraformaldehyde in PBS, 20 min, RT), permeabilized (0.1% Triton X-100 in PBS), blocked (1% BSA in PBS), and stained with Phalloidin-Alexa Fluor®488. All fluorescent dyes were from Molecular Probes (Invitrogen Carlsbad, CA, USA). Morphological observations were performed using Olympus IX71microscope equipped with DP70 camera 20× objective. Nuclei were stained and specimens mounted using Fluoroshield DAPI (Sigma Aldrich).

#### *5.5. Statistical Analysis*

Data were visualized on PCA (Principal Component Analysis). The normality of the data was tested using Chi-squared test and the correlation between SAs and EAs production was done using the linear Pearson test. A non-parametric Wilcoxon signed rank test was used to test the null hypothesis of no difference in the EAs and ECs concentrations within the particular species or population. Non parametric Mann-Whitney test or parametric paired *t*-test was used to compare EAs and SAs content between populations or across all samples. These statistical analyses were done using the PAST 3.25 software [66]. The LC50 values of tested compounds were calculated using probit analysis [67,68] using Microsoft Excell® Professional Plus 2013 software (Microsoft Corp., Redmond, WA, USA).

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6651/11/8/439/s1, Table S1: List of analyzed ergot samples and the content of ergochromes and ergot alkaloids, Table S2: 13C NMR data for secalonic acid A–C, Table S3: 1H NMR data for secalonic acid A–C, Figure S1: Cell cultures grown on glass cover slips were treated for 24 h with secalonic acid A or ergotamine and in vivo incubated with MitoTracker® Red CMXRos, fixed, permeabilized, and labeled with Phalloidin-Alexa Fluor®488, Figure S2: FTMS data for secalonic acid A–C.

**Author Contributions:** Conceptualization, M.K. (Miroslav Kolaˇrík) and M.F.; methodology, M.F., J.C., and M.K. ˇ (Miroslav Kolaˇrík); resources, M.K. (Miroslav Kolaˇrík), S.A.W., K.D.B., J.C., and M.F.; investigation, E.S., K.P., ˇ S.A.W., P.N., M.K. (Marek Kuzma), P.M., and V.G.; formal analysis, M.F., M.K. (Miroslav Kolaˇrík), and J.C.; ˇ validation, M.F., M.K., and J.C.; writing—original draft preparation, M.K. and S.A.W.; writing—review & editing, ˇ M.K., S.A.W., M.F., and J.C.; funding acquisition, M.K. (Miroslav Kolaˇ ˇ rík), K.D.B.

**Funding:** This research was funded by the Czech Science Foundation (GACR), grant number 13-00788S and by ˇ European Regional Development Funds, grant number CZ.1.05/1.1.00/02.0109 BIOCEV. Stephen A. Wyka and Kirk D. Broders were supported by American Malting Barley Assoc. Grant No. 17037621.

**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* **Aflatoxin B1 Conversion by Black Soldier Fly (***Hermetia illucens***) Larval Enzyme Extracts**

**Nathan Meijer 1,\*, Geert Stoopen 1, H.J. van der Fels-Klerx 1,\*, Joop J.A. van Loon 2, John Carney 3,4 and Guido Bosch <sup>5</sup>**


Received: 24 July 2019; Accepted: 10 September 2019; Published: 12 September 2019

**Abstract:** The larvae of the black soldier fly (*Hermetia illucens* L., BSFL) have received increased industrial interest as a novel protein source for food and feed. Previous research has found that insects, including BSFL, are capable of metabolically converting aflatoxin B1 (AFB1), but recovery of total AFB1 is less than 20% when accounting for its conversion to most known metabolites. The aim of this study was to examine the conversion of AFB1 by S9 extracts of BSFL reared on substrates with or without AFB1. Liver S9 of Aroclor-induced rats was used as a reference. To investigate whether cytochrome P450 enzymes are involved in the conversion of AFB1, the inhibitor piperonyl butoxide (PBO) was tested in a number of treatments. The results showed that approximately 60% of AFB1 was converted to aflatoxicol and aflatoxin P1. The remaining 40% of AFB1 was not converted. Cytochrome P450s were indeed responsible for metabolic conversion of AFB1 into AFP1, and a cytoplasmic reductase was most likely responsible for conversion of AFB1 into aflatoxicol.

**Keywords:** aflatoxin; mycotoxin; black soldier fly; BSFL; *Hermetia illucens*; S9 fraction; cytochrome P450; metabolic conversion; enzyme induction

**Key Contribution:** The S9 fraction of black soldier fly larvae (*Hermetia illucens* L., BSFL) contains cytochrome P450s that metabolically convert aflatoxin B1 (AFB1) into AFP1, and a cytoplasmic reductase is responsible for conversion of AFB1 into aflatoxicol.

#### **1. Introduction**

Aflatoxins are a group of mycotoxins that are primarily produced by the molds *Aspergillus flavus* and *Aspergillus parasiticus*. The four major aflatoxins are B1, B2, G1, and G2, which can be found in various food products such as peanuts and maize [1]. Aflatoxins are carcinogenic to humans (IARC Group 1) and a major economic and health problem globally, but especially in sub-Saharan Africa, Latin America, and Asia, since people and animals are exposed to levels that substantially elevate mortality and morbidity. Aflatoxin B1 (AFB1) has generated the most concern due to its toxicity and high contamination levels in food and feed commodities in certain areas such as Africa [2,3]. AFB1 is converted by animals and humans into a variety of metabolites, such as aflatoxin M1, Q1, P1, and aflatoxicol (AFL) [1,4]. AFB1 is a "procarcinogen" in the sense that hepatic microsomal cytochrome

P450 (CYP450) enzymes convert AFB1 to AFB1-8,9-epoxide (AFBO), which has reactive and electrophilic properties that underlie the toxicity of the compound [5].

Although prevention of contamination of crops by aflatoxigenic molds is paramount, a variety of decontamination strategies have been developed. Postharvest detoxification methods for AFB1 include physical (heat and irradiation), chemical (acidification, ammoniation, and ozonation), and biological (whole organism or extracts thereof and enzymatic) treatments [6–9]. Although degradation levels of AFB1 are generally high for enzymatic treatments, treatment times are also high (up to several days), and there is uncertainty regarding the degradation products formed [6]. Since metabolites in treated products may still be toxic, determination of degradation products is a principal requirement for assessing the safety and efficacy of enzymatic detoxification treatments. Detoxification mechanisms are generally classified into three phases: "(I) introduction of reactive and polar groups into substrates through oxidation, reduction, or hydrolysis; (II) conjugation of metabolites with other compounds to create more polar or more easily excretable molecules; and (III) transport and elimination of compounds" [10]. Cytochrome P450 monooxygenase enzymes play a major role in the bioactivation of AFB1 in phase I metabolism [1,11]. These enzymes can be found in almost all (aerobic) organisms, but different P450 isoforms are species specific [12,13]. Some compounds may act as inhibitors of certain P450s. The best-known example of such an inhibitor is piperonyl butoxide (PBO) [13].

Insects have developed physiological and metabolic strategies to cope with potential toxic compounds, such as mycotoxins. Earlier work on the fruit fly *Drosophila melanogaster* Meigen (Diptera: Drosophilidae) [14–17] and on yellow mealworm (*Tenebrio molitor* L.; Coleoptera: Tenebrionidae; YMW) [18] has shown that some insects are capable of metabolizing AFB1. More recently, Bosch et al. (2017) [19] found that both black soldier fly larvae (*Hermetia illucens* L., BSFL) and YMW have a high AFB1 tolerance and that the toxin did not accumulate in these species. Moreover, the amount of AFB1 lost (from substrates to insects) varied from 83% to 95% for BSFL and 89% to 96% for YMW. However, the YMW formed AFM1 (present in the excreta) and AFB1 was detected in YMW when provided with feed containing 0.023 mg/kg of AFB1 or more. The concentration decreased when the YMW were starved before harvesting, which resulted in the larvae emptying their guts. This suggested that the gut contents contributed significantly to the measured AFB1 levels in the YMW. Camenzuli et al. (2018) [20] subsequently assessed the effects of a variety of mycotoxins, including AFB1, on BSFL and lesser mealworm (*Alphitobius diaperinus* Panzer; Coleoptera: Tenebrionidae). The mycotoxin metabolites AFL, aflatoxin P1, Q1, and M1 were taken into consideration in the chemical and bioaccumulation analyses. Mass balance calculations for BSFL suggested recovery of total AFB1 of less than 20%. Of the other analyzed metabolites, only AFL was detected at 0.2% of the overall mass balance; aflatoxin Q1, P1, and M1 were not detected (<0.001 mg/kg for larvae, <0.005 mg/kg for residual material (spiked feed and gut clean)).

In the corn earworm (*Helicoverpa zea* L.; Lepidoptera: Noctuidae), the toxicity of AFB1 depends on the CYP-mediated metabolic bioactivation [21]. Niu et al. (2008) [22] reported that dietary phytochemicals (i.e., xanthotoxin, coumarin, or indole-3-carbinol) induced midgut enzymes including CYP321A1 that can degrade AFB1 into mainly AFP1 and, to a lesser extent, an undefined metabolite. Feeding AFB1 without the phytochemical did not increase CYP321A1 transcripts and resulted in reduced growth and development, confirming that phytochemicals induced CYP enzymes that detoxify AFB1 [23]. Incubation of AFB1 with a homogenate of the larvae of the navel orangeworm (*Amyelois transitella* Walker; Lepidoptera: Pyralidae) resulted in the formation of mainly AFL and, to a lesser extent, aflatoxin B2a and AFM1 [24]. This was in line with findings in testes of the fruit fly using a similar in vitro approach [17]. CYP6AB11 from navel orangeworm did not metabolize AFB1 [25]. Importantly, the in vitro study of Lee and Campbell (2000) [24] reported that PBO did not impact AFL formation by navel orangeworm, which suggested that this metabolite was formed by cytosolic NADPH-dependent reductase. Incubation of AFB1 with a homogenate of larvae of the codling moth (*Cydia pomonella* L.; Lepidoptera: Tortricidae) did not result in the metabolites AFL, AFB2a, and AFM1, which may relate either to absence of the metabolic system, different metabolic

pathways, or that the system was not activated in the larvae, as these were not exposed to AFB1 before the study [24]. In honey bees (*Apis mellifera* L.; Hymenoptera: Apidae), there are also indications of P450-mediated metabolic detoxification of AFB1 [26].

In summary, BSFL have high tolerance to AFB1, and when AFB1 is provided in the feed, most of it cannot be recovered in the larvae and residual material. It is not clear whether and, if so, to what extent AFB1 is metabolically converted. As an alternative to live animals, an enzyme extract can be prepared to assess the potential for metabolic conversion of the species in vitro. In this manner, individual or several metabolic conversion pathways can be isolated and identified. The aim of this study was to examine the conversion of AFB1 by S9 extracts of BSFL reared on a substrate with AFB1. The S9 enzyme fraction contains both the membrane-bound as well as the soluble enzymes [27]. Liver S9 of Aroclor-induced rats was used as a reference. To investigate whether cytochrome P450 enzymes specifically are involved in the conversion of AFB1, PBO was tested in a number of treatments. We conclude that cytochrome P450s were indeed responsible for metabolic conversion of AFB1 into AFP1, and that a cytoplasmic reductase was most likely responsible for conversion of AFB1 into AFL.

#### **2. Results**

#### *2.1. E*ff*ects of AFB1 in Feed on Larval Development*

Live BSFL were subjected to two treatments, each applied in triplicate: one treatment in which the feed was spiked with AFB1 to a concentration of 0.5 mg/kg, and one control treatment without AFB1 added to the feed. Per replicate, 100 larvae less than 24 h old were provided with the feed and harvested after 9 days. Survival after these 9 days was high for both the control (average: 99.0) and the AFB1 treatment (average: 97.3) (*p* = 0.007). Average total biomass obtained was, respectively, 15.2 and 15.0 g (*p* = 0.685).

#### *2.2. AFB1 Conversion by S9 Fractions*

Table S1 shows the molar concentrations of AFB1 and the analyzed metabolites after incubation for all replicates. The results from the treatment with AFB1 but without S9 (−S9 + AFB1, t = 2 h) show that only AFB1 was found at the same concentration as what was spiked, and that no metabolites were formed. In the treatment with S9 but without AFB1 (+S9 − AFB1, t = 2 h), no AFB1 or metabolites were detected. This indicates that the AFB1 that was present in the larval feed was not converted into the analyzed metabolites by the larvae and did not accumulate.

Figure 1 shows the average molar concentrations (nM) of AFB1 and the analyzed metabolites for the three types of S9 fractions (rat, BSFL-control, and BSFL-AFB1) at two different points in time after addition of AFB1: after directly (t = 0 h) halting enzymatic activity (+S9 + AFB1, t = 0 h) and after incubation for2h(+S9 + AFB1, t = 2 h). For all three S9 fractions at t = 0 h, only AFB1 was present. The AFB1 concentration at t = 0 h was half of the concentration that was spiked at the start due to the addition of 100 μL of acetonitrile to the 100 μL mixture of Regensys A buffer, NADPH, AFB1, and S9. The results of the BSFL-control and BSFL-AFB1 S9 fractions that were incubated for 2 h show that part of the AFB1 was converted into AFP1 (23.44 nM, *p* = 0.847) and AFL (21.32 nM, *p* = 0.824). The total molar concentrations (AFB1 + AFP1 + AFL) of these two treatments were equal to the total molar concentration of AFB1 in the t = 0 h treatments (BSFL-control: *p* = 0.275; BSFL-AFB1: *p* = 0.211). This indicates that no metabolic conversion occurred other than the type that was observed (i.e., AFB1 into AFP1 and AFL).

**Figure 1.** Molar concentrations (nM) of aflatoxin B1 (AFB1) and metabolites (AFM1, AFP1, and aflatoxicol (AFL)) for incubation of AFB1 with S9 fractions from rat liver, black soldier fly larvae (*Hermetia illucens* L., BSFL)-control, and BSFL-AFB1 after directly halting enzymatic activity (t = 0 h) and after 2 h of incubation. Significance of differences is indicated in the figure with \* (*p* ≤ 0.05) or NS (not significant, *p* > 0.05).

The results of the rat S9 treatments that were incubated for 2 h show that AFM1 (29.34 nM) and, to a lesser extent, AFP1 (2.59 nM) had formed. The amount of AFB1 that was recovered after incubation from the treatment with the rat S9 fraction (3.15 nM) was less than what was recovered from the BSFL treatments (BSFL-AFB1: 37.23 nM; BSFL-control: 30.57 nM). In addition, the total molar concentration of AFB1 and analyzed metabolites for the rat S9 treatment after incubation for 2 h (35.17 nM) was less than the total AFB1 molar concentration for the rat S9 treatment at t = 0 h (82.29 nM). This indicates that some of the spiked AFB1 was converted by the rat S9 into different metabolites than those that have been analyzed.

#### *2.3. E*ff*ect of PBO on AFB1 Conversion by S9 Fractions*

Figure 2 shows the average molar concentrations (nM) of AFB1 and the analyzed metabolites for the two types of S9 fractions (rat and BSFL-AFB1) after incubation with AFB1 for 2 h. One treatment contained an S9 fraction (rat or BSFL-AFB1) and AFB1 (+S9 + AFB1, t = 2 h); the second also contained dimethyl sulfoxide (DMSO, (+S9 + AFB1 + DMSO, t = 2 h); and in the third, PBO (dissolved in DMSO) was added to the S9 fractions and AFB1 (+S9 + AFB1 + DMSO + PBO, t = 2h). For the AFB1 larvae S9 treatments, the differences between the treatment containing DMSO and the treatment without further additives were not significant for each included metabolite (AFB1 (*p* = 0.296), AFL (*p* = 0.758), AFP1 (*p* = 0.491)). This indicates that the DMSO in which the PBO was dissolved did not affect the conversion of the BSFL-AFB1 S9 fraction. Compared with the BSFL treatment without additional additives, the AFP1 concentration in the PBO treatment was reduced (*p* = 0.002), while the AFL (*p* = 0.001) and AFB1 (*p* = 0.004) concentrations were elevated. Comparing the rat treatment with PBO to the treatment without additives shows that the conversion into AFP1 was completely halted and the conversion into AFM1 was reduced (7.54 nM). The AFB1 molar concentration was higher in the PBO treatment than in the treatment without additives, but the total molar concentration of the analyzed metabolites in the PBO treatment was equal to the total AFB1 molar concentration at t = 0h(+S9 + AFB1, t = 0 h; *p* = 0.129). This indicates that the PBO halted the conversion of AFB1 by rat S9 into different metabolites than those that have been analyzed.

**Figure 2.** Molar concentrations (nM) of AFB1 and metabolites (AFM1, AFP1, and AFL) for incubation of AFB1 with S9 fractions from rat liver and BSFL-AFB1, and with dimethyl sulfoxide (DMSO) or DMSO + cytochrome P450 inhibitor piperonyl butoxide (PBO) added, after 2 h of incubation. Significance of differences is indicated in the figure with \* (*p* ≤ 0.05) or NS (not significant, *p* > 0.05).

#### **3. Discussion and Conclusions**

Body weight and survival of control larvae and larvae exposed to AFB1 were similar. We therefore conclude that the BSFL were unaffected by the addition of AFB1 to their feed, which is in line with the findings of Bosch et al. (2017) [19] and Camenzuli et al. (2018) [20].

The study showed that S9 preparations of BSFL converted approximately 60% of the AFB1 to AFL and AFP1. The remaining 40% of AFB1 was not converted into the analyzed metabolites. The amounts of AFL and AFP1 were more or less equal, and there was no difference in activity of S9 prepared from larvae grown on substrates with or without AFB1. This suggests that the enzymes involved in the biotransformation of AFB1 are part of constitutive detoxification systems of the BSFL. Activation of the system in the larvae via pre-exposure—as hypothesized by Lee and Campbell (2000) [24], discussed above—is therefore not required for the system's functioning.

The addition of cytochrome P450 inhibitor PBO partially inhibited the formation of AFP1 by BSFL S9 extracts, indicating that a P450 enzyme is involved in the conversion from AFB1 into AFP1. Conversion to AFL by the BSFL S9 fraction was not inhibited when PBO was added, indicating that it is not catalyzed by P450 enzymes. The total recovery of AFB1 and metabolites in the BSFL PBO treatment exceeded the total molar concentration of metabolites in the treatment without additives at approximately 122% (*p* = 0.001), but this was within the range of 2 \* SD.

Since AFB1 is converted to AFL by a cytosolic NADPH-dependent reductase [24,28,29], we therefore propose that this conversion to AFL by BSFL occurs via the same pathway. Figure 3 shows selected metabolic conversion pathways known for AFB1. The black arrows denote metabolic pathways that have been found to be active in BSFL S9 fractions in this study; the grey arrows denote known pathways in other species.

**Figure 3.** Selected metabolic conversion pathways known for AFB1 (adapted from Lee and Campbell, 2000 [24] and Dohnal et al., 2014 [5]).

The reaction from AFB1 to AFL is reversible. The cofactor for the reduction of AFB1 is NADPH, which was added to the AFB1 at the start of the trials, together with the Regensys A regenerating system. The cofactor for the dehydrogenation of AFL yielding AFB1 is NADP, which accumulates when the regeneration of NADPH stops [28]. It cannot be ruled out that BSFL possess this microsomal dehydrogenase, which would revert the reaction and increase the level of AFB1 again, thereby negating detoxification. This reversion could, for instance, occur in case of an incubation time longer than 2 h or in the absence of an NADPH regenerating system. AFM1 was not formed by the BSFL S9 fraction in this study. The latter conversion is catalyzed by the cytochrome P450 enzyme CYP1A2 [5]. The absence of the formation of AFM1 in the BSFL treatment (in this study as well as in Camenzuli et al., 2018 [20]) and its presence in the treatment with rat S9 suggests the absence of this enzyme in BSFL. However, the enzyme may also have been deactivated during preparation of the BSFL S9 fraction.

Compared with the conversion of AFB1 by live BSFL, as studied by Camenzuli et al. (2018) [20], there are a few major differences in how the AFB1 was metabolized by the S9 fraction observed in this study. Firstly, no aflatoxin P1, Q1, and M1 could be recovered by Camenzuli et al. (2018) [20], and the amount of AFL was negligible (0.2% of mass balance). In the current study, however, approximately equal proportions of AFL and AFP1 were recovered. Moreover, while less than 20% of AFB1 could be recovered in the mass balance of Camenzuli et al. (2018) [20], 100% could be recovered in this study. It is unclear what the exact reasons are for these discrepancies, but the following hypotheses may be considered. Since live larval cells are expected to contain a wider variety of cofactors (other than NADPH, as used in conjunction with the S9 fraction in this study), a larger number of enzymes may be activated. It is possible that enzymes in live larvae first convert the AFB1 into AFL and AFP1, which, in turn, are precursors for other compounds. These may, for instance, be reactive metabolites that bind to other proteins. A second option is that the conversion of AFB1 into AFL and AFP1 in the S9 fraction is accelerated due to the absence of other cofactors that would catalyze different metabolic pathways. More research on the exact pathways of AFB1 conversion by live BSFL is recommended in order to identify and quantify degradation products so that the efficacy and safety of reared larvae can be assessed. This could, for instance, be achieved by performing the analyses described in this manuscript with inhibitors of specific cytochrome P450 and/or NADPH-dependent reductase enzymes.

In conclusion, BSFL S9 fractions converted AFB1 into AFP1 and AFL. Furthermore, exposing BSFL to AFB1 did not impact the conversion capacity, suggesting that the enzymes involved are part of a general metabolic system. No other analyzed metabolites were formed. Cytochrome P450s were responsible for metabolic conversion of AFB1 into AFP1. A cytoplasmic reductase was most likely responsible for conversion of AFB1 into AFL.

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

The overall methodology for the different treatments is shown schematically in Figure 4. Firstly, BSFL were reared on feed that had been spiked with AFB1 (0.5 mg/kg). A second batch of larvae was reared on noncontaminated feed as a control. From these two batches of larvae, separate S9 fractions were prepared, and a commercial rat S9 fraction was used for comparison. These S9 fractions were incubated with AFB1 for 2 h (+S9 + AFB1, t = 2 h). In addition, an incubation was included in which PBO dissolved in DMSO was added to the mixture of the S9 fraction and AFB1 (+S9 + AFB1 + DMSO + PBO, t = 2 h). Four control treatments were used in this study. Firstly, acetonitrile was directly added to an S9 and AFB1 mixture at t = 0 h in order to halt enzymatic activity (+S9 + AFB1, t = 0 h). Secondly, one mixture was prepared excluding S9 fractions (−S9 + AFB1, t = 2 h), and a third control treatment excluded AFB1 (+S9 − AFB1, t = 2 h). Finally, a solvent control treatment containing DMSO was used (+S9 + AFB1 + DMSO, t = 2 h).

Differences between treatments were tested for significance by multiple one-way ANOVA tests (α = 0.05) using the Analysis ToolPak add-in for Microsoft Excel 2016 MSO 32-bit (version 16.0.4849.1000, Microsoft Corporation, Redmond, WA, United States of America, 2016). This was done by comparing the molar concentrations of individual (AFB1, AFM1, AFP1, and AFL) and total metabolites between treatments.

**Figure 4.** Schematic representation of treatments (control treatments within dotted lines; red letters indicate the difference between that treatment and the + S9 + AFB1, t = 2 h treatment).

#### *4.1. Larvae Treatment*

A standard dry wheat-based mash feed (layer meal based) was spiked with AFB1 (*A. flavus*, 99.6% purity, Sigma-Aldrich, Saint Louis, MO, USA) by an external laboratory (Ducares B.V., Utrecht, The Netherlands) to reach a concentration of 0.5 mg/kg of feed. This was the highest concentration used by Bosch et al. (2017) [19], which had no effect on the mortality and growth of the larvae. The feed used was the same batch that was used by Camenzuli et al. (2018) [20], which had been prepared by Ducares B.V., Utrecht, The Netherlands. A sample of the nonspiked feed was used as a control.

Per dietary treatment, three plastic boxes (17.8 × 11.4 × 6.5 cm) were prepared for each replicate. Each box contained 18 g (±0.1 g) of feed, which was manually mixed with 25 mL (±0.1 mL) of tap water. One hundred larvae less than 24 h old and originating from the BSF colony maintained at the Laboratory of Entomology (Wageningen University, Wageningen, The Netherlands) were added to the box. The box was then closed with a perforated lid. All boxes were kept in a climate cabinet (27 ± 1 ◦C and 88% ± 1% relative humidity) for 9 days. After 9 days, the larvae were collected and counted. The larvae were cleaned by rinsing with lukewarm tap water, dried with paper, and snap-frozen in liquid nitrogen. Larvae were stored at −80 ◦C until further analyses.

#### *4.2. Preparation of S9 Fraction*

Frozen larvae were ground to a fine powder with a precooled mortar and pestle under addition of liquid nitrogen. The frozen powder was transferred to a precooled polypropylene tube (50 mL Greiner, VWR, Amsterdam, The Netherlands) and 1.5 mL of ice cold buffer (1.15% KCl in 50 mM Tris/HCl pH 7.4) was added per gram of powder (7 g of powder + 10.5 mL of buffer). The sample was mixed thoroughly by tapping on the bench to bring the powder in contact with the buffer. Care was taken that the powder did not thaw before it was mixed with the extraction buffer. After obtaining a homogenous suspension, the material was further extracted by gently inverting the tubes 100 times. The suspensions were centrifuged in a precooled rotor for 25 min at 8960 rcf and 4 ◦C. The supernatants were collected, pooled, and mixed. Then, 500 μL aliquots were snap-frozen in liquid nitrogen and stored at −80 ◦C.

The protein concentration was determined using the DC Protein assay (Bio-Rad Laboratories, Veenendaal, The Netherlands) according to the manufacturer's protocol and bovine serum albumin (BSA) was used as a standard. The protein content of the insect S9 fractions was on average 36 mg/mL. The protein content of the rat liver S9 was 38 mg/mL (data provided by the manufacturer).

#### *4.3. S9 Incubations with AFB1*

Samples were prepared on ice and contained 1× Regensys A buffer (Moltox, Boone, USA), 5 mM NADPH, 50 ng/mL AFB1 (50 μg/kg), and 2.5 mg/mL S9 protein in a final volume of 100 μL. NADPH was prepared freshly in Regensys A buffer. AFB1 was dissolved in DMSO and dilutions in Regensys buffer were prepared prior to the incubations. The final concentration of DMSO in the assay was 0.03%. The reactions were started by addition of S9 to the mixture and transferring the tubes to 37 ◦C in an Eppendorf thermomixer. Most samples were incubated for 2 h. t = 0 samples were prepared by adding 100 μL of cold acetonitrile prior to addition of S9.

To study the role of cytochrome P450 enzymes in the conversion of AFB1, 1 mM of PBO (or 3% DMSO as solvent control) was included in the S9 mixes. Samples were incubated for 2 h and the reactions were stopped by addition of 100 μL of ice cold acetonitrile. Samples were vortexed thoroughly, put on ice for 5 min, and finally stored at −80 ◦C.

#### *4.4. Chemicals*

Regensys A buffer and rat liver S9 (Aroclor-induced rats; lyophilized S9 preparation) were purchased from Trinova Biochem (Gießen, Germany). Regensys A buffer consists of 100 mM phosphate buffer pH 7.4, 33 mM KCl, 8 mM MgCl2, and 5 mM glucose-6-phosphate (NADPH regeneration system). NADPH, AFB1, AFM1, DMSO-HybriMax, and PBO were purchased from Sigma-Aldrich (Zwijndrecht, The Netherlands); AFP1 from TRC (Toronto Research Chemicals, North York, Canada); and AFL from Enzo Life Sciences BVBA, (Brussels, Belgium). Potassium chloride was obtained from Merck (Amsterdam, The Netherlands) and Tris-buffer from Fisher Scientific (Landsmeer, The Netherlands).

#### *4.5. LCMS Analyses*

Analyses were performed in largely the same way as in Camenzuli et al. (2018) [20]. Samples were defrosted, vortexed, and centrifuged for 5 min, 14,000 rpm at room temperature. From the supernatant, 190 μL was transferred to an LCMS vial and 10 μL of 13C-labeled internal standard solution was added. Samples were mixed and 5 μL was analyzed with an LC-MS/MS-based method for the analysis of mycotoxins in feed and food materials. The accredited scope of this method was extended in order to also quantify the AFB1 and its metabolites in larvae and residual material (excreta and residual feed) of BSFL.

Two MRM transitions were included for each metabolite in the MS/MS method. Details on this and additional MS/MS settings can be found in Tables S2 and S3 of the Supplementary Materials. Each metabolite was identified by its retention time and the peak area ratio between two transitions: the quantifier and the qualifier. Quantification was performed by bracketed calibration (an interval of not more than 10 injections) on the peak area of the quantifier (qn) of calibration solutions in solvent. Concentrations of AFB1 and metabolites were corrected for matrix effects with the use of their respective 13C-isotope-labeled standards (AFB1 and AFM1) or by means of matrix-matched calibration standards (AFP1 and AFL).

The limit of quantification (LOQ) was defined as the lowest calibrated level which complied with the required QC parameters as mentioned in SANTE/11945/2015. Metabolite-specific LOQs can be found in Table S4 in the Supplementary Materials.

The LC-MS/MS system consisted of an injection and pump system from Waters (Waters, Milford, MA) and an AB Sciex QTRAP 6500 triple quad system equipped with an electrospray ionization (ESI) source operated in positive mode (AB Sciex, Nieuwerkerk a/d IJssel, The Netherlands). For LC separation, a 100 × 2.1 mm ID, 3 μm Restek Ultra Aqueous C18 column (Interscience, Breda, The Netherlands) was used. Details on the LC-MS/MS settings can be found in Table S5 of the Supplementary Materials. The LC eluent gradients were 1 min isocratic at 100% A, followed by a linear gradient to 100% B in 4 min. For complete elution of all matrix coextractants from the column, the final composition at 100% B was kept for 2 min. In 30 s, the initial conditions were restored and then equilibrated for 2 min prior to the next injection.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6651/11/9/532/s1. Table S1. Molar concentrations (nmol/L) of aflatoxin B1 (AFB1) and analyzed metabolites (aflatoxicol (AFL), aflatoxin P1 (AFP1), and aflatoxin M1 (AFM1)) after incubation. Results of individual replicates. Table S2. MS/MS parameters. Table S3. MS/MS transitions. Table S4. LOQs of analyzed compounds. Table S5. LC-MS/MS parameters.

**Author Contributions:** Conceptualization, G.S., H.J.v.d.F.-K., J.J.A.v.L., J.C., and G.B.; Formal analysis, G.S.; Methodology, J.C. and G.B.; Resources, J.J.A.v.L.; Writing—original draft, N.M.; Writing—review and editing, H.J.v.d.F.-K., J.J.A. v.L., and G.B.

**Funding:** This study was part of the strategic research program of Wageningen UR "Customized Nutrition" and financed by Mars, Inc. (grant no. 1277360201) and Wageningen UR.

**Acknowledgments:** We want to thank Ruud van Dam (Wageningen Food Safety Research, WFSR) for performing the LCMS analyses, and Lonneke van der Geest (WFSR) for contributing to the design of the experiments.

**Conflicts of Interest:** This study was part of the strategic research program of Wageningen UR and financed by Wageningen UR, an industrial partner, and the Dutch Ministry of Agriculture, Nature, and Food Quality. 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. The authors declare no conflicts 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* **E**ffi**cacy of Azoxystrobin on Mycotoxins and Related Fungi in Italian Paddy Rice**

#### **Paola Giorni 1,\*, Umberto Rolla 2, Marco Romani 2, Annalisa Mulazzi <sup>3</sup> and Terenzio Bertuzzi <sup>3</sup>**


Received: 14 May 2019; Accepted: 28 May 2019; Published: 30 May 2019

**Abstract:** AbstractThe efficacy of azoxystrobin was evaluated in the presence of mycotoxigenic fungi and relative mycotoxins in Italian paddy rice during the growing season in the field. Three experimental fields were considered and the applied experimental design was a strip plot with three replicates; rice samples were collected at four different growing stages. The efficacy of the fungicide treatment on rice fungal population was demonstrated with around 20% less total fungal incidence in sprayed samples compared to untreated ones; the same decrease was noted also in *Fusarium* spp. species but not in *Aspergillus versicolor*. Of the mycotoxins considered, ochratoxin A (OTA) and aflatoxins (AFBs) were never detected, deoxynivalenol (DON) was found in 46% of samples at levels always lower than 100 μg/kg, while sterigmatocystin (STC) occurred in all the paddy rice samples collected after flowering, with a maximum value of 15.5 μg/kg. Treatment with azoxystrobin was not effective in reducing DON contamination, but it had an important and significant effect on STC content, showing a decrease of 67% in the sprayed samples.

**Keywords:** rice; mycotoxins; sterigmatocystin; STC; deoxynivalenol; DON; growing season; azoxystrobin; fungicide

**Key Contribution:** This study takes into consideration, for the first time, the effect of fungicide treatment on mycotoxigenic fungi population in Italian paddy rice and on the presence of mycotoxins during the growing season with particular interest to fungicide influence on sterigmatocystin, a relevant emerging mycotoxin.

#### **1. Introduction**

Rice (*Oryza sativa* L.) is the staple food for almost half of the world's population [1]; it is cultivated mainly in Asian regions and China is the largest producer. The main rice producer in Europe is Italy, accounting for around 50% of total European production. Rice cultivation is principally in Northern Italy (Piedmont and Lombardy) and is destined for several food uses, baby foods included.

Different diseases can affect rice and, in particular, fungi can be particularly dangerous for plant and grain health [2] during both the growing season and post-harvest [3]. Panicle blast caused by *Pyricularia grisea* [4] and brown spot caused by *Bipolaris oryzae* [5] are the most dangerous diseases for Italian rice crops, occurring frequently and causing production and economic losses [4]. Nowadays, particular attention must also be paid to fungal species which can produce, in favorable environmental and substrate conditions, various mycotoxins that can impact human health. The presence of

mycotoxigenic fungi on paddy rice has been indicated in several reports; in particular, *Fusarium* spp., responsible for trichothecenes (expecially deoxynivalenol (DON)) and fumonisins (FBs) [6,7], *Aspergillus flavus*, able to produce aflatoxins (AFs) [8], and *Penicillium* spp. for ochratoxin A (OTA) and citrinin (CIT) [9]. Recently, sterigmatocystin (STC) produced by *Aspergillus versicolor* was also found in rice and resulted the most common mycotoxin in Italian rice [10]. The European Commission fixed strict limits for mycotoxins in cereals; in particular, for rice destined for human consumption limits are present for AFs (2.0 μg/kg), OTA (3.0 μg/kg), and DON (750 μg/kg), making mycotoxin containment very important for product exchanges (EU Regulation 165/2010; EC Regulation 1881/2006; EU Regulation 1006/2015).

Among possible strategies to control mycotoxigenic fungi development in the field and, consequently, mycotoxin production, fungicides can be used. Negative effects have been reported in some cases, such as the reduction of beneficial microorganisms for plant growth due to acidification of the soil [11] or the possible selection of fungicide-resistant fungal strains [12]. Moreover, each toxigenic fungal species responds differently to fungicides because several factors can contribute to their reaction; in particular, weather, active ingredients, plant development stage, and cultivar resistance can play a role [13,14] and can act as stressors in the production of mycotoxins [15]. For example, it has been found that triazole applications can reduce both *F. graminearum* and DON occurrence [16,17], especially if the treatments are carried out before fungal infection [11]. However, in some cases the use of fungicide can increase mycotoxin content; Dors et al. [15] reported that tebuconazole was able to act as an elicitor of stress for mycotoxigenic fungi and, consequently, enhanced the presence of several mycotoxins.

There are few fungicides allowed by Italian Regulations for rice; of these, azoxystrobin is an active ingredient belonging to the strobilurin chemical group and it is one of the most used on Italian rice because of its demonstrated efficacy on several crops and its major role in reducing *Pyricularia grisea* [18] and *Bipolaris oryzae* infections in rice [19] and Fusarium Head Blight (FHB) in wheat [20]. However, its possible effect on mycotoxins is still uncertain since in some studies the use of azoxystrobin in wheat could result in an increase in DON content up to 42% [21,22]; for this reason, this possible effect needs to be evaluated also in rice in order to assist farmers in their selection of fungicides.

The aim of this study was to define the efficacy of azoxystrobin on mycotoxigenic fungal species present in paddy rice during the growing season from flowering to over ripening (1 June–30 September) and determine its possible effect on the production of their relative mycotoxins.

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

#### *2.1. E*ffi*cacy of Azoxystrobin on Mycotoxigenic Fungi*

The highest fungal incidence was found at the full ripening stage with more than 70% of the rice kernels infected (Table 1). Fungi seem to increase their incidence throughout the growing season up to ripening, then they significantly decrease if left in field for an additional 14 days obtaining around a 10% reduction for total fungi incidence (Table 1). The same level of reduction was not observed for mycotoxigenic species that reach their maximum incidence at harvest time (full ripening) and maintain their presence even in the case of over ripening. Both for *Fusarium* spp. and *A. versicolor*, the only mycotoxigenic species resulting with a significant presence in field, no differences were found between the full ripening and over-ripening stages (Table 1).

The same was found in a previous study on paddy rice [10] with the only exception of *Fusarium* spp. that seemed to decrease in over ripening in accordance with the total fungi trend; probably, different meteorological conditions registered after full ripening, in particular the almost total absence of rain observed in the area in year 2018, could have influenced *Fusarium* spp. vitality.

As expected, different rice varieties showed different levels of fungal contamination; in particular, Terra CL showed the highest fungal content while CL26 the lowest (Table 1). *Fusarium* spp. exhibited the same trend while *A. versicolor* presence resulted always very low and with no significant differences between rice varieties. However, interestingly, *A. versicolor*, differently from other fungal species, showed a higher incidence in CL26, which was the rice variety least contaminated by other fungal species, and a lower incidence in Terra CL and CL15 varieties which were the most contaminated by other fungal species (Table 1). This was probably due to varying fungal abilities to compete in extreme environmental conditions; the year 2018, in fact, was notable in Italian rice cultivation areas for an almost total absence of rain (total rainfall was only 157.4 mm in the period 1 June–30 September) and extreme temperatures (up to 36 ◦C). This was undoubtedly favorable for xerophilic species, such as *A. versicolor* [23], in particular on rice varieties where fungal incidence and, as a consequence, fungal competition were lower.

The efficacy of the fungicide treatment on rice fungal population was demonstrated with around 20% less total fungal incidence in sprayed samples than the untreated ones (Table 1). The same decrease was noted also in the *Fusarium* spp. species but not in *A. versicolor* which was unchanged (Table 1). The effect of strobilurins against fungi is well documented, they appear able to enhance rice plant defenses against pathogen attacks [19,24], shown also in wheat against mycotoxigenic *Fusarium* species like *F. graminearum* [20]. The incidence of *A. versicolor* was too low to obtain a significant reduction in treatment with azoxystrobin, although a reduction of 5% was observed.

**Table 1.** Analysis of variance (ANOVA) of fungal incidence and contamination of sterigmatocystin (STC) and deoxynivalenol (DON) at different sampling times in different rice varieties sprayed or unsprayed with fungicides formulated with azoxystrobyn (250 g/L) in three different experimental fields. Data refer to mean data; all experiments were conducted with three replicates.


Different letters mean significant differences according to Tukey Test; n.s.: not significative; \*: *p* ≤ 0.05; \*\*: *p* ≤ 0.01.

Differences in fungal contamination were found between the three different experimental fields; in particular, experimental field C was the most contaminated with also the highest *Fusarium* spp. incidence. No significant differences were found between experimental fields in the presence of *A. versicolor*.

#### *2.2. E*ffi*cacy of Azoxystrobin on Mycotoxin Production*

Among the considered mycotoxins considered, OTA and AFs were never detected; DON was found in 46% of the samples at levels always lower than 100 μg/kg, while STC occurred in almost all the paddy rice samples, showing a maximum value of 15.5 μg/kg. These data partially accord with a previous study, carried out on rice samples collected in the same area, that found DON and AFs only sporadically and in low amounts, while STC was always detected, appearing as crucial in rice contamination [10].

Regarding mycotoxin accumulation, it is important to note that both DON and STC follow the same trend of their producing fungi. In particular, DON was highest at full ripening and remained constant up to over ripening as happened for *Fusarium* spp. fungi, while STC was highest at full ripening and significantly decreased in over-ripening, as happened to the presence of *A. versicolor* (Table 1). A similar result for STC was found in a previous research, even if this decrease was not so intensive [10]. This could be due to environmental and substrate conditions that probably reduce fungal ability to produce STC while they have no effect on DON production. Significant differences in mycotoxin contamination were found between rice varieties with Sole CL resulting one of the most contaminated by DON and the one with the highest STC content (Table 1).

The highest DON contamination was found in the rice variety CL15 while the lowest was in the rice variety Centauro (Table 1); none of the rice varieties considered in the study showed a DON contamination above the limits fixed by the European Commission of 1250 μg/Kg for paddy rice (EU Regulation 1881/2006). These results seem to confirm the findings of a previous study where DON was found only in low amounts [10] suggesting that this mycotoxin could be considered a minor risk for Italian paddy rice.

The treatment with azoxystrobin was inefficient in reducing DON contamination; contrarily, DON increased in sprayed samples even if the results were not statistically different (Figure 1). The fungicide had an important and significant effect on STC content, showing a decrease of 67% in the sprayed samples (Table 1). These data partially agree with previous findings on wheat, where the use of strobilurin obtained a good reduction of FHB, but with an uncertain impact on DON reduction [20,25]. The results obtained in STC reduction are very promising because this mycotoxin seems to be the most dangerous for Italian rice production and treatment with azoxystrobin, one of the active ingredients allowed by the Italian government on rice, could be useful for reducing STC contamination in field during the growing season.

Significant differences in mycotoxin contamination were observed between experimental fields (*p* ≤ 0.01); in particular, experimental field C was the most contaminated having the highest incidence of total fungi. Moreover, the same experimental field showed the highest STC content (1.9 μg/Kg vs 0.7–1.6 μg/Kg) and the highest DON content (40 μg/Kg vs 20–22 μg/Kg) (Table 1). Differences in mycotoxins contamination between experimental fields were expected since many variables can contribute to their presence such as susceptibility of rice variety, preceding crop, tillage, and pest presence [26,27]. However, even if we tried to keep the differences in agronomic management minimal between experimental fields, environmental factors, such as relative humidity and temperature, can always play a relevant and unpredictable role in both fungal contamination and mycotoxin occurrence.

Considering treatment with azoxystrobin on single rice varieties, important reductions were observed. In particular, Sole showed the highest STC reduction being, respectively, of 62% and 77% in experimental field B and C (Figure 1). The lowest reduction (22%) in STC obtained in sprayed samples was found in the rice variety CLXL745 (Figure 1). Differences in azoxystrobin efficacy between rice varieties were probably due to their different susceptibility to fungicide, as already found in other studies with other active ingredients [28] and on other crops [29,30].

**Figure 1.** Mean sterigmatocystin (STC) and deoxynivalenol (DON) contamination in the different experimental fields considered in the trial in case of unsprayed and sprayed paddy rice varieties.

#### **3. Conclusions**

At first, this study confirmed that *Fusarium* spp. and *A. versicolor* were the most frequently mycotoxigenic species found on Italian paddy rice during the growing season. As a consequence, DON and STC can occur in paddy rice samples at different plant growing stages. Treatment with azoxystrobin as an active ingredient seems to be efficient in reducing total fungi and *Fusarium* spp. incidence and STC contamination, while they have no effect on *A. versicolor* presence and DON level. Rice varieties played an important role for their different and proven susceptibility to fungal diseases and fungicide efficacy.

STC, an emerging mycotoxin that is not routinely checked, has been confirmed as a relevant mycotoxin in paddy rice, confirming a previous survey that collected rice samples of different origin [31]; the contamination level of this mycotoxin could be considered for future EU legislative regulation.

The results obtained showed that the use of azoxystrobin could help farmers to develop a potential method for STC containment in conditions particularly conducive for fungal development and mycotoxin production, being necessarily cautious in their use due to uncertain reduction of the presence of DON.

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

#### *4.1. Field Samples*

Sampling of rice was conducted at four different growing stages (flowering (BBCH 69), early dough (BBCH 83), ripening (BBCH 89), and over-ripening (BBCH 92, 15 days post-ripening) in 2018 in three experimental fields located close to Mortara (PV) in Lombardy, the main Italian rice production region. Nine rice varieties, both long B and round grain were considered (Table 1). In experimental field A were cultivated four long B grain rice varieties (CLXL 745, CL26, Sirio CL, Mare CL) while five common round grain rice varieties (Sole CL, Selenio, Centauro, Terra CL, CL15) were cultivated in experimental fields B and C. Soil texture and sowing period varied, while the meteorological conditions can be considered similar because of the proximity of the fields (all within 10 km).

#### *4.2. Fungicide Treatment*

The applied experimental design was a strip plot with three replicates; each strip (6 m × 100 m), containing rice plants of 1 variety with a sowing density of 150 kg/ha of seeds, was subdivided into three plots considered as replicates. Working at field level, it was not possible to apply a randomized plot design.

In each experimental field, two assays were considered for each rice variety: "unsprayed strip" used as control, where no fungicide was sprayed, and "sprayed strips" where treatment with fungicide was carried out. The distance between "unsprayed" and "sprayed" strips was 20 meters, which was considered sufficient to prevent problems linked to spray drift and possible contamination between the two considered assays.

Commercial formulations with azoxystrobin in the same concentration (250 g/L) were distributed in open field, on the whole width of plots chosen as "sprayed", using a backpack sprayer (mod. SP 126, Oleo-Mac Bagnolo in Piano, Reggio Emilia, Italy) calibrated to spread 250 L/ha of solution. In all the sprayed experimental assays, only 1 fungicide treatment was scheduled when rice plants were at the stage of panicle emergence (BBCH 51-55). We decided to do the fungicide treatment at this plant growing stage because this time is normally chosen also for the scheduled paddy rice treatment against the main pathogens such as *Pyricularia grisea* and *Bipolaris oryzae*.

For each rice variety and experimental field, rice plants were collected from each plot with an X-shape design, then the plants were shelled and the grains obtained considered as representative. For each plot, representing a replicate, 500 g of grains were randomly chosen as sample. Samples were used for mycological analyses and then dried, milled using a cyclone hammer mill (1 mm sieve, Pulverisette, Fritsch GmbH, Idar-Oberstein, Germany), homogenized and kept at 4 ◦C until chemical analysis.

#### *4.3. Monitoring of Mycotoxigenic Fungi*

Fifty kernels were randomly selected from each sample, surface disinfected in 1% sodium hypochlorite for 2 min and in 90% ethyl alcohol for 2 min and then transferred onto Petri dishes containing potato dextrose agar (PDA, Biolife, Milano, Italy). The Petri dishes were incubated at 25 ◦C (12 h light photoperiod) and after 5–7 days the incidence of kernels infected by fungi was quantified. *Fusarium* spp. and *Penicillium* spp. isolates were identified at Genus level thanks to observations with binocular microscope (40×); only *Aspergillus* spp. isolates were identified at species level observing their morphological characteristics with magnification between 100× and 400× according to Raper and Fennell [32].

#### *4.4. Monitoring of Mycotoxins*

The analyses were carried out using the following methods: AFs were determined by HPLC-FLD (liquid chromatography with fluorimeter detector) as reported by Bertuzzi et al. [33]; OTA by HPLC-FLD [34], DON by GC-MS (gas chromatography coupled to mass spectrometer) [35], STC by LC-MS/MS (liquid chromatography coupled to mass spectrometer) [36]. The analyses were recently described in the work of Bertuzzi et al. [10].

#### *4.5. Data Analysis*

The data were transformed before statistical analysis; in particular, fungal incidence was arcsine transformed and mycotoxin content was ln transformed [37]. Analysis of variance (ANOVA) was calculated using the generalized linear model (GLM) procedure of the statistical package IBM SPSS Statistics 21 (IBM Corp., Armonk, NY, USA) while significant differences were highlighted using the Tukey test (*p* ≤ 0.05) for mean separation.

**Author Contributions:** The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript. T.B., M.R., and P.G. designed the research. P.G. and A.M. performed the experiment. T.B. and P.G. analyzed the data and wrote the manuscript. U.R., M.R., P.G., and A.M. assisted with the experiment. T.B. and P.G. supervised the research and edited and approved the final manuscript.

**Funding:** This work supported by Lombardy region, PSR 2014-2020 program, project BABYRICE.

**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* **Removal of Small Kernels Reduces the Content of** *Fusarium* **Mycotoxins in Oat Grain**

#### **Guro Brodal \*, Heidi Udnes Aamot, Marit Almvik and Ingerd Skow Hofgaard**

Norwegian Institute of Bioeconomy Research (NIBIO), P.O.Box 115, N-1431 Ås, Norway; heidi.udnes.aamot@nibio.no (H.U.A.); marit.almvik@nibio.no (M.A.); ingerd.hofgaard@nibio.no (I.S.H.)

**\*** Correspondence: guro.brodal@nibio.no

Received: 30 April 2020; Accepted: 22 May 2020; Published: 23 May 2020

**Abstract:** Cereal grain contaminated by *Fusarium* mycotoxins is undesirable in food and feed because of the harmful health effects of the mycotoxins in humans and animals. Reduction of mycotoxin content in grain by cleaning and size sorting has mainly been studied in wheat. We investigated whether the removal of small kernels by size sorting could be a method to reduce the content of mycotoxins in oat grain. Samples from 24 Norwegian mycotoxin-contaminated grain lots (14 from 2015 and 10 from 2018) were sorted by a laboratory sieve (sieve size 2.2 mm) into large and small kernel fractions and, in addition to unsorted grain samples, analyzed with LC-MS-MS for quantification of 10 mycotoxins. By removing the small kernel fraction (on average 15% and 21% of the weight of the samples from the two years, respectively), the mean concentrations of HT-2+T-2 toxins were reduced by 56% (from 745 to 328 μg/kg) in the 2015 samples and by 32% (from 178 to 121 μg/kg) in the 2018 samples. Deoxynivalenol (DON) was reduced by 24% (from 191 to 145 μg/kg) in the 2018 samples, and enniatin B (EnnB) by 44% (from 1059 to 594 μg/kg) in the 2015 samples. Despite low levels, our analyses showed a trend towards reduced content of DON, ADON, NIV, EnnA, EnnA1, EnnB1 and BEA after removing the small kernel fraction in samples from 2015. For several of the mycotoxins, the concentrations were considerably higher in the small kernel fraction compared to unsorted grain. Our results demonstrate that the level of mycotoxins in unprocessed oat grain can be reduced by removing small kernels. We assume that our study is the first report on the effect of size sorting on the content of enniatins (Enns), NIV and BEA in oat grains.

**Keywords:** T-2 toxin; HT-2 toxin; deoxynivalenol (DON); enniatin B (EnnB); size sorting; unprocessed cereals

**Key Contribution:** Removing small kernels can reduce the mycotoxin content of oat grain lots, and thereby improve the grain quality and increase the number of lots that can be accepted as safe for food and feed.

#### **1. Introduction**

Several species of the fungal genera *Fusarium* are common pathogens of small grain cereals. *Fusarium* spp. infect and cause damage to the head and grain of cereals, especially under moist conditions. The disease, known as Fusarium Head Blight (FHB), is one of the most important diseases in wheat (*Triticum aestivum*), oats (*Avena sativa*) and barley (*Hordeum vulgare*). During development and maturation of infected heads, *Fusarium* species can produce several mycotoxins which can lead to severe contamination of grain [1]. Mycotoxin-contaminated grains do not necessarily show disease symptoms which makes them difficult to identify. Consumption of grain and grain-based products containing *Fusarium* mycotoxins can cause many harmful health effects in humans and animals, and *Fusarium* toxins are therefore one of the most important quality and safety risks of cereal grain for food and feed [2–4]. In addition, workers at grain elevators and mills may be exposed to mycotoxins

by inhalation and skin permeation of grain dust during grain processing [5]. To reduce the risk, the European Union (EU) has set maximum levels for some mycotoxins in cereal grain and cereal-based food products for human consumption, and has recommended guidance values for its content in animal feed [6,7].

The mycotoxin deoxynivalenol (DON) is common and frequently occurs in oat grains [8–11]. Moreover, the HT-2 toxin and T-2 toxin are often found more frequently in oats than in other cereal species, and sometimes at high concentrations [9,10,12–15]. HT-2 and T-2 toxins are closely related (HT-2 is the deacetylated form of T-2), often occur together [12] and their occurrence concentrations are often considered together as a sum of HT-2 and T-2. Throughout this study we use the denomination HT2+T2 for the sum of these toxins. In addition to DON and HT2+T2 toxins, zearalenone (ZEA) and several unregulated mycotoxins such as nivalenol (NIV), enniatin A (EnnA), enniatin A1 (EnnA1), enniatin B (EnnB), enniatin B1 (EnnB1) and beauvericin (BEA) often occur in oat grains [8–11]. In Norway, 3-acetyl-deoxynivalenol (3-ADON) is the dominating acetylated chemotype, although 15-acetyl-deoxynivalenol (15-ADON) has been detected [16].

Besides the importance of oats as a raw material for animal feed concentrate (compound feed), oats grown for human consumption have increased during the last few years due to their beneficial nutritive properties [17]. On the other hand, it has been reported that consumers with a high relative intake of cereals compared to their body weight have a HT2+T2 and a DON exposure that may exceed the Tolerable Daily Intake (TDI) [18]. HT2+T2 toxins are considerably more toxic than DON (TDI 0.02 and 1.0 μg/kg body weight/day, respectively) [19]. The maximum levels for DON and ZEA in unprocessed oat grain for food are 1750 and 100 μg/kg, respectively [6]. However, no regulated maximum levels have been set so far for HT2+T2 in cereals and cereal products, but the European Commission has recommended an indicative level of 1000 μg/kg for HT2+T2 in unprocessed oats [20]. When concentrations of HT2+T2 are detected above this level, the EU member states should perform investigations to identify factors resulting in these levels and investigate the effect of feed and food processing on the presence of these toxins.

In oats, the largest proportion of the mycotoxins is located in the hulls. Thus, de-hulling, i.e., removing hulls (glumes and husk) from the kernels before further processing into oat flakes and other products is an efficient method to reduce the mycotoxin content of oats. Commercial processing of oats has been reported to reduce the content of DON, T2 and HT2 by 80%–95%, with the major loss occurring during de-hulling [21–25]. However, de-hulling is part of the processing of cereal grain. On the other hand, cleaning and size-sorting of raw grain, normally performed as a first step to remove dust, weed seeds, chaff/straw pieces and small, lightweight and damaged kernels before further processing, is accepted according to European legislation, to be carried out on unprocessed grains [6,20]. Several studies have shown that cleaning and sorting of cereal grain can reduce the content of mycotoxins, although variable effects have been reported. Most data are available for reduction of DON in wheat, but also other *Fusarium* toxins, e.g., HT2+T2, NIV and ZEA have been analyzed. The effects of cleaning and sorting by various methods and procedures on the reduction of mycotoxins in wheat, as well as in a few studies in barley and oats, have been reported to vary from no reduction to up to more than an 80% reduction and have even been reported to increase levels in a few cases [3,26–28]. The effect of removing the small grain fractions by size sorting, i.e., after separating the kernels on sieves according to kernel size, varies depending on the sieve sizes used. Despite different degrees of mycotoxin reduction reported by the cleaning and sorting of grains, overall results indicate that these operations may efficiently reduce the mycotoxin levels in highly contaminated cereals before further use/processing.

Data on the effects of removing small grain kernels on the mycotoxin content in oats are limited. The aim of the present study was therefore to investigate to what extent the removal of small kernels can contribute to the reduction of the mycotoxin content in oat grains and thereby improve the quality of the remaining grain. Our hypothesis was that it is feasible/achievable to reduce the mycotoxin content in raw oat grain by size sorting/cleaning out the smaller kernels. Grain from 24 Norwegian oat

grain lots (14 from 2015 and 10 from 2018) were sorted into a large and a small kernel fractions by passing the grain through a 2.2 mm laboratory sieve. The large and small kernel samples, in addition to samples of unsorted grain, were analyzed for content of HT2 and T2 (reported together as HT2+T2), DON, ADON (3- and 15-acetyl-deoxynivalenol analyzed together), NIV, EnnA, EnnA1, EnnB, EnnB1, BEA and ZEA. The mycotoxin concentrations in the two grain fractions were compared with the concentrations in the unsorted grain.

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

#### *2.1. Mycotoxin Content in the Oat Grain Lots*

A visual summary of the concentrations of the ten mycotoxins detected in the 24 oat grain lots (unsorted grain) is shown in Figure 1A,B. Most grain lots contained all the tested mycotoxins. Moderate to high mycotoxin levels in unsorted grain samples were detected for HT2+T2 toxins and EnnB in 2015, and DON in 2018 (Figure 2). Similar contrasting occurrences of HT2+T2 vs. DON in oats have been reported in other studies [9,29,30]. The mycotoxin levels in samples of unsorted grain of the remaining mycotoxins, i.e., ADON, NIV, EnnA, EnnA1, EnnB1, BEA and ZEA were generally low in samples from both years. ZEA was only detected in 2015 samples. Our results are in line with other reports on occurrences of mycotoxins in Norwegian oat grains [9,12,25,31].

**Figure 1.** *Cont.*

**Figure 1.** (**A**) Median, interquartile range, range and outlier (\*) for the content (μg/kg) of the mycotoxins HT2+T2 and EnnB (Enniatin B) in 14 oat grain lots from 2015 and 10 oat grain lots from 2018 (unsorted samples). (**B**) Median, interquartile range, and outliers (\*) for the content (μg/kg) of the mycotoxins DON (deoxynivalenol), ADON (3- and 15-acetyl-deoxynivalenol), NIV (nivalenol), EnnA (enniatin A), EnnA1 (enniatin A1), EnnB1 (enniatin B1), BEA (beauvericin) and ZEA (zearalenone) in 14 oat grain lots from 2015 and 10 oat grain lots from 2018 (unsorted samples).

**Figure 2.** Concentration levels (μg/kg) of HT2+T2 toxins (**A**), deoxynivalenol (**B**) and Enniatin B (**C**) in 24 unsorted oat grain lots and in large and small kernel fractions after size sorting on sieve size 2.2 mm. Lot 1–14 from 2015, lot 15–24 from 2018. Note the different values on the concentration level axes.

#### *2.2. Mycotoxin Content in Unsorted and in Large Kernel Fraction of Oats*

#### 2.2.1. HT2+T2

All grain lots contained HT2+T2 toxins (Figure 2A). The levels were considerably higher in the grain from 2015 (lot 1–14) than in the grain from 2018 (lot 15–24). The HT2+T2 concentrations in unsorted grain varied from 486 to 1368 μg/kg (mean 745 μg/kg) among the samples from 2015, and from 92 to 282 μg/kg (mean 178 μg/kg) among the samples from 2018 (Table 1). After sorting, we detected significantly lower HT2+T2 in the large kernel fractions than in the unsorted grain. The concentrations in the large kernel fraction varied from 197 to 522 μg/kg (mean 328 μg/kg) among the samples from 2015, and from 70 to 187 μg/kg (mean 121 μg/kg) among the samples from 2018. In 2015, this corresponds to an average reduction in HT2+T2 concentration of 56% (varying from 24% in lot 10 to 76% in lot 5) in the large kernels compared to the unsorted grain. In 2018, the average reduction in HT2+T2 concentration was 32% (varying from only 2% in lot 15 to 66% in lot 23) (Table 1). The average weight reduction after removal of the small grain fraction was 15% and 21% of the weight of the samples from the two years respectively (Table 2). We did not observe any relationship between the percentage weight

reduction and percentage of HT2+T2 reduction (*R*<sup>2</sup> = 0.00, *p* = 0.919), or between weight reduction and HT2+T2 levels in unsorted samples (*R*<sup>2</sup> = 0.04, *p* = 0.328) when calculated for all 24 grain lots together. However, a significant relationship between percentage weight reduction and HT2+T2 levels was observed (*R*<sup>2</sup> = 0.47, *p* = 0.010) for samples from 2015, by omitting the highest contaminated seed lot (No. 2). Only a few studies on the effects of removing small grain kernels on HT2+T2 in oats by size sorting have been found. A Swedish study reported markedly reduced concentrations of HT2+T2 (not quantified) after removing the kernels that passed through a sieve size of 2 mm [32]. A study in Finland, also using a sieve size of 2 mm obtained around a 30%–35% reduction [33], which agrees with our results from 2018 samples and is somewhat lower than what we obtained from the 2015 samples. It was interesting to observe that the effect of removing small kernels was considerably higher in the oats with relatively high HT2+T2 levels (mean 745 μg/kg, 2015 samples) than in oats with lower levels (mean 178 μg/kg, 2018 samples). The extent of toxin reduction increased with toxin levels, and linear regression showed that this relationship was significant (*p* = 0.000, *R*<sup>2</sup> = 0.44, Figure 3). Moreover, a few other studies have reported the highest reduction of HT2+T2 in the highest contaminated oats by size sorting and de-hulling [21,32].

**Figure 3.** Percentage reduction of HT2+T2 toxins in oat grain samples from 2015 and 2018 in large kernel fraction (after removing small kernel fraction by sieve size = 2.2 mm) vs. concentration level in grain lots (unsorted grain).




*p*-valuepaired largecompared *p*-values 0.05/2representsignificant different from the unsorted fraction. 2 Percentage change in toxin level in large or small grain fractions compared to the unsorted fraction. Reductions in toxin level comparedto the unsorted fraction are shown as negative values and increase as positive values. 3 n.a.—not analysed due to no significant difference and/or low toxin levels. 4 n.d.—not detected.


**Table 2.** Origin (municipality, field number) and cultivars of oat grain lots from 2015 and 2018, and weight proportions (%) of small and large kernel fractions after size sorting (sieve size 2.2 mm).

<sup>1</sup> Based on fraction weights of all samples.

As in our study, large differences between samples on the effect on HT2+T2 levels in oats by cleaning/sieving (laboratory-scale grain cleaner, sieve size 1.75 × 20 mm) were reported in a German study [22]. They observed reductions in the range from 0% to 100%, which is even more inconsistent than our data. In a study of barley, the content of HT2 was reduced by 68% and T2 by 81% on average, after around 13% of the sample was removed by using a 2.5 mm sieve [34]. In durum wheat, 54% reduction in the HT2+T2 concentration was observed after a vigorous cleaning procedure (aspiration and two sieves: 5 × 15 mm and 2 × 19 mm) [35]. Concentration levels of HT2+T2 in oats have been reported to be higher in rachis and glumes than in kernels [36]. As small kernels contain a higher proportion of glumes and pericarp than large kernels, removing small kernels will contribute to a reduced mycotoxin content. Our results and other reported data imply that cleaning and size sorting can be useful methods to reduce the concentrations of HT2+T2 in unprocessed grain of oats and other small grain cereals, although the effect will vary among grain lots.

#### 2.2.2. DON

DON was detected in all grain lots (Figure 2B), however its concentration levels were higher in the grain from 2018 (lot 15–24) than in the grain from 2015 (lot 1–14). The concentration levels varied from 100 to 309 μg/kg (mean 191 μg/kg) in unsorted samples from 2018 and from quantification limit (LOQ = 1 μg/kg) to 153 μg/kg (mean 46 μg/kg) in samples from 2015 (Table 1). For the grain harvested in 2018, removing the small kernels resulted in significantly lower DON concentrations in the large kernels (varying from 89 μg/kg to 249 μg/kg, mean 145 μg/kg) compared to the unsorted grain. On average, we detected 24% less DON in the large kernels in 2018 samples, however, the reduction varied from only 3% (lot 17) to up to 35% (lot 22). Despite low DON levels in the 2015 samples,

we observed a trend towards a lower mean concentration in large kernel fractions (30 μg/kg) compared to unsorted grain (46 μg/kg) (Table 1). However, the effect of removing the small kernels varied from a reduction to an increase in DON content, and no significant difference in DON levels was detected between the grain fractions.

Limited data exist on the size-sorting effects on DON in oats. A Finnish study observed a 30%–40% reduction in DON concentrations using a 2 mm sieve [33], which is somewhat higher than the 24% reduction we obtained in our 2018 samples. Size sorting of barley in the same study resulted in around a 50% lower content of DON in large grain. Another study in barley reported an 80% reduction of DON after removing the small kernels by using a sieve size of 2.5 mm, however, the effect differed between cultivars [34]. Several studies in wheat, using different sieve sizes, aspiration and cleaning technologies resulting in large variations in the amounts of by-products (e.g., waste, screenings, offals, dockage, pellets etc.) have reported from no and up more than 80% reduction in DON concentrations by cleaning and sorting, but also increased levels have been observed in a few cases [26–28]. For industrial cleaning, an expected reduction rate of 20% has been suggested for DON in wheat [37], which is near the 24% reduction we detected in oats by removing the small kernel fraction in the 2018 materials. Our result supports the previous finding that removing small kernels can reduce the concentration of DON in oats, however as with other small grain cereal species, the effect is likely to vary among grain lots.

#### 2.2.3. Enniatins (Enns) and BEA

The prevalence and concentration levels of Enns and BEA in our grain lots were in accordance with previous studies from Nordic countries where these toxins have been reported as common contaminants of cereals occurring generally at low concentration levels. However, they occur occasionally at high levels, and often EnnB is the most common [8–10,25,31].

All grain lots contained EnnB (Figure 2C). The concentration levels in unsorted samples were considerably higher in most grain lots from 2015 than in grain from 2018, ranging from 92 to 5356 μg/kg (mean 1059 μg/kg), and from 8 to 25 μg/kg (mean 15 μg/kg), for the two years respectively. After removing the small kernel fraction, the EnnB concentrations in large kernel samples from 2015 was significantly lower than in unsorted grain and varied between 48 and 3064 μg/kg (mean 594 μg/kg) (Table 1). On average, this represents 44% less EnnB content than in unsorted grain. However, the reduction varied between 5% (lot 14) and 63% (lot 6), and in lot 2 we recorded an increase of 2%. For samples from 2018, no reduction in the mean EnnB concentrations was detected after size sorting. EnnB1 was detected in all grain lots. In samples from 2015, the concentrations ranged from 6 to 452 μg/kg (mean 120 μg/kg), and a trend towards lower EnnB1 concentrations in large grain (mean 62 μg/kg) compared to unsorted was observed (Table 1). The EnnB1 concentrations in samples from 2018 ranged from 9 to 38 μg/kg (mean 18 μg/kg), with no difference in the content between large and unsorted grain. EnnA and EnnA1 were also detected in all unsorted grain samples, however, the levels were low, especially in samples from 2018 (Table 1). In samples from 2015, a trend towards a lower content of EnnA and EnnA1 in the large grains compared to unsorted grain was observed. BEA occurred at very low levels in all unsorted grain lots, however, a trend towards a reduction was observed after removing the small kernel fraction in samples from 2015 (Table 1). We assume that our study is the first report on the effect of size sorting on the content of Enns and BEA in cereal grains, as we did not find any published data on this. However, a study of Enns in milling fractions of wheat reported that approximately 40% remained in the final wheat flour, compared to whole grain [38].

Enns and BEA have shown cytotoxic, genotoxic and immunomodulating effects, as well as toxic effects on reproductive systems [39,40]. These toxins have been reported to accumulate in animal tissues and eggs [41]. In 2014, the European Food Safety Authority (EFSA) concluded that acute exposure to Enns and BEA do not indicate concern for human health, but chronic exposure might be of concern [42]. However, due to a lack of relevant in vivo toxicity data, a human risk assessment could not be performed. So far, maximum levels for content of these mycotoxins in food and feed have not been established and at present there is no regulatory requirement to consider or reduce the contamination of Enns and BEA in cereal grains. Our results indicate than EnnB can be substantially reduced in oats by removing small kernels, although no consistent effect in relation to concentration levels was found.

#### 2.2.4. NIV, ADON and ZEA

NIV was detected in all but two grain lots from 2015 and in all 10 lots from 2018, however, the overall concentration levels were low (Figure 1B). After removing the small kernel fraction, NIV was detected in all large kernel samples from 2018, but only in 8 of the 14 large kernel samples from 2015. Despite the low levels, a trend towards lower content of NIV in the large grains (mean 11 μg/kg) compared to unsorted grains (mean 23 μg/kg), was observed in samples from 2015 (Table 1). No published data on the effect of sorting and cleaning on the NIV content in oats have been found, however, what is almost an elimination (> 98%) of NIV during oat processing has been reported [21]. In barley, 94% less NIV were reported after removing the small kernels using sieve size 2.5 mm [43]. In wheat, from below a 10% to around an 80% reduction in NIV by cleaning and size sorting has been reported [28]. Based on our limited results and the literature on the effect on NIV in other cereal species, we assume that removal of small kernels can reduce the NIV content in oat grain.

In this study 3- and 15-ADON were analyzed together as ADON. Although 15-ADON is detected in Norway, 3-ADON is the most common chemotype [16]. ADON was detected in all 24 grain lots. However, the overall concentration levels were low (Figure 1B), ranging from 1 (= LOQ) to 35 μg/kg (mean 8 μg/kg) in the samples from 2015, and from 1 to 4 μg/kg (mean 2.2 μg/kg) in the samples from 2018 (Table 1). After removing the small kernel fraction, ADON was still detected in all large kernel samples from 2018 (mean 1.6 μg/kg), and the level was significantly lower than in the unsorted samples. In 2015, ADON could not be detected in five samples after removal of small kernels. Although the average level (5 μg/kg) in the large kernel fraction was lower than in the unsorted grain, this was not statistically significant. A recent Norwegian study on the distribution of mycotoxins in oat grain reported around a 90% reduction of 3-ADON by de-hulling of grain containing relatively high levels of 3-ADON (mean 485 μg/kg) [25]. Based on that study, together with our limited data, it is reasonable to believe that the content of ADON can be reduced by removing small oat kernels.

ZEA was detected at very low levels (close to LOQ = 1 μg/kg) in 12 of the 14 unsorted oat grain samples from 2015, and no ZEA was detected in the 2018 samples (Table 1). Studies in wheat have found ZEA to be mainly concentrated in the outer tissues of the grain, however, the reduction of ZEA content by cleaning and processing has been reported to vary from a few to up to around 40% [43–45]. As no data have been found on the effect of size sorting on the level of ZEA in oats, and because we only detected very low levels in our study, it is not possible to conclude on the effect of size sorting on ZEA in oats.

#### *2.3. Mycotoxin Content in the Small Kernel Fraction*

Since most mycotoxins are mainly concentrated in the small kernel fractions, in the hulls and in the outer tissues of the grains [21,24,26,44] cleaning, size sorting, de-hulling and further processing will increase the mycotoxin concentrations in the by-products (e.g., screenings, offals, dockage, pellets, bran etc.) [3] In our study, the mean concentration of HT2+T2 in the small kernel fraction was 272% higher for the 2015 samples, and 187% higher in the 2018 samples, compared to the unsorted grain (Table 1). The increase in single samples varied from below 100% to up to 840%, which is the same magnitude that has been measured for HT2+T2 in oat by-products from other size sorting and cleaning studies [21,24,32]. The DON concentrations in our samples were generally low, and despite a mean increase of 71% (ranged from 42% to 145%) in the small grain fraction in samples from 2018, the levels were still moderate (Table 1). A high accumulation (ten-fold) of DON in the offals after cleaning was reported in oats in a Polish study [46]. In our study, the mean concentration of EnnB in the small kernel fraction was 98% and 120% higher in samples from 2015 and 2018, respectively, compared to the unsorted grain (Table 1). However, the concentrations in the small kernel samples varied considerably from a reduction of 60% to an increase of 568%. Higher EnnB1 concentrations were found in the small grain fraction compared to unsorted grain in samples from 2018, representing an increase of 89% (Table 1), whereas in samples from 2015, a trend towards higher mean EnnB1 concentrations in small grains compared to unsorted grains was observed. Except for a study reporting a considerably higher content of EnnB (200% and 375% in shorts and bran respectively) and EnnB1 (around 240% and 300% in shorts and bran respectively) after milling of wheat, no other data have been found on the distribution of Enns in cereal grain fractions [38]. Despite low ADON levels, a significant increase was detected in the small kernel fraction compared to unsorted grain for 2018 samples (Table 1). The by-product fractions are commonly used as raw materials for animal feed. It is important for feed producers to be aware of the risk of extensive increases in the mycotoxin content in by-products. To manage the mycotoxin risk and to decide what to be done with potentially contaminated by-products, proper sampling and analysis are necessary. Based on the contamination level, an evaluation of the economic value of the by-products and the carry-over potential of the mycotoxins, a decision on the inclusion level of feed ingredients can be made [3]. In our study, the highest measured HT2+T2 concentration in the small kernels fraction was 6427 μg/kg (lot No. 2) and 4889 μg/kg (lot No. 14) which would have been of concern if it had been used in feed production.

#### *2.4. Conditions Influencing on the E*ff*ect of Grain Size Sorting*

We observed variable effects of size sorting on the content of HT2+T2, DON and EnnB among the grain lots (Figure 2). One reason for this variation can be the diversity in cultivars (Table 2) which is likely to differ in resistance to *Fusarium* infections and mycotoxin development. In addition, our grain lots originated from two different years and partly from different locations. The degree of *Fusarium* mycelium growth into the kernels and mycotoxin development varies between *Fusarium* species and are in addition to host plant resistance against infection, influenced by cultivation conditions and local weather during the susceptible stages of the host plant and during grain-filling and maturation stages [36,47,48]. This can result in a different distribution of the different mycotoxins in kernels and therefore likely contribute to the different effects of size sorting among the samples in our study. Moreover, cultivar differences in phenological kernel traits itself, such as kernel size probably also contributed to the variation in size-sorting effects. Grain materials for this study were not selected to allow for an examination of the influence of cultivar or location on the size-sorting effect. One important reason for the variation between different studies in the effect of cleaning and size sorting on the mycotoxin content in grain is in the use of different sorting and cleaning methods/technologies, e.g., different sieve sizes and machinery settings, and some studies have included a pre-cleaning step without or with aspiration with varying fan speeds to remove dust, broken grain and other debris. By removing some of the "waste products" prior to sorting, less effects will be obtained by further cleaning and size sorting. This diversity in method will also result in large differences in volume and weight proportions removed in the pre-processing stages e.g., [21,34,44].

#### *2.5. Grain Weight Reduction by Size Sorting and Mass Balance Calculations*

By passing the raw grain samples through a laboratory sieve (sieve size 2.2 mm) and removing the small kernel fraction, the grain weight was in average reduced by 15% and 21% for samples from 2015 and 2018, respectively (Table 2), i.e., a larger proportion passed through the sieve in the samples from 2018 than in the samples from 2015, indicating generally smaller kernels in 2018 than in 2015. However, the weight reduction varied between 6% and 33% among the 14 samples from 2015, and between 9% and 32% among the 10 samples from 2018. No data was found in the literature on the weight reduction by cleaning or size sorting in oats, however, dehulling has been reported to remove around 30% to 40% of the whole grain weight [21,25]. In a study of barley, the weight reduction varied between 6% and 25% among 15 samples of different cultivars when grain passed a sieve size of 2.5 mm [34].

Mass balance calculation of mycotoxin concentrations in unsorted grain and in the sum of the size fractions is an important quality control tool [44]. Mass balance calculated for amounts of HT2+T2 (2015, 2018), DON (2018) and EnnB (2015) in the two fractions from the mycotoxin concentrations and the weight of each fraction, and the sum was compared to the mycotoxin content in the unsorted sample (Table 3). For HT2+T2, the recovery in the sum of the two fractions compared to unsorted grain ranged between 59% and 120% (mean 87%) for 2015 samples and between 75% and 138% (mean 105%) for 2018 samples. For DON, the recovery in the sum of the two fractions ranged between 85% and 117% (mean 96%). The recovery of EnnB in the sum of the two fractions ranged between 51% and 188% (mean 77%). Regression analysis of the mycotoxin amounts in the unsorted grain and the sum amounts in the two fractions (Figure 4) indicated rather a good relationship for DON (*R*<sup>2</sup> = 0.93, 2018 samples) and EnnB (*R*<sup>2</sup> = 0.90, 2015 samples), however, some variation in the recovery among the samples were observed for HT2+T2 in samples from both years (*R*<sup>2</sup> = 0.60 and *R*<sup>2</sup> = 0.64, for 2015 and 2018 samples respectively). This indicates that the analysis was somewhat inaccurate.

**Table 3.** Comparison between mycotoxin concentrations (μg/kg) in unsorted grain and the mycotoxin mass balance calculated from the weighted sum of mycotoxins in the large and small kernel fractions and percentage recovery in calculated compared to measured amounts.


**Figure 4.** Regression analysis of the sum of toxin in large and small kernel fractions compared to the unsorted fraction for (**A**) HT2+T2 in 2015 (*n* = 14); (**B**) HT2+T2 in 2018 (*n* = 10); (**C**) deoxynivalenol (DON) in 2018 (*n* = 10); and (**D**) enniatin B (EnnB) in 2015 (*n* = 14).

Reasons for some of the discrepancies between the sum of the two fractions and the amounts in the unsorted sample may be due to inaccuracies in sampling, including sample preparation before analysis, the drawing of a small ground sample (5 g) for analysis, and the recovery of the analysis method itself. Sampling is a major source of error in monitoring mycotoxins in cereal grains. It is difficult to obtain homogenous samples partly due to an uneven distribution of mycotoxins within a grain lot [49]. If a sample is not representative it can cause an over- or under- estimation of mycotoxin contamination, and this might have contributed to the variable effects on the mycotoxin content we obtained by size sorting. The type of grinder as well as the method for dividing can influence on the heterogeneity of mycotoxins in a sample. The grinder and sieve size (1 mm) used in our study gave what was perhaps a higher heterogeneity due to a higher particle size than what is optimal [50]. Moreover, small samples with low mycotoxin levels can cause considerable measurement uncertainty.

#### **3. Conclusions**

Our study showed that by removing the small kernel fraction from the grain, representing on average 15% and 21% of the weight of the samples from the two years respectively, the content of *Fusarium* mycotoxins was considerably reduced. The most notable effects were seen on the concentrations of T2+HT2 toxins, which were reduced by 56% and 32% on average for samples from 2015 and 2018 respectively, and EnnB, which was reduced by 44% in grain lots from 2015. We also observed a clear reduction, on average 24%, in the DON concentrations in the 2018 samples. Moreover, despite low levels, our analyses showed a trend towards reduced content of DON, ADON, NIV, EnnA, EnnA1, EnnB1 and BEA after removing the small kernel fraction in samples from 2015.

For HT2+T2, the reduction obtained by sorting increased with the mycotoxin levels of unsorted grain lots. Ours and other studies experienced variable effects on the mycotoxin content by size sorting, however, removing the small kernel fraction in oats can be a useful method to reduce the mycotoxin contamination. Grain lots are still defined as unprocessed after cleaning and size sorting. Thus, by performing these operations, the grain industry may safely utilize a higher number of unprocessed oat grain lots for further processing in the food and feed chain. Knowledge about the different content of mycotoxins in various grain size fractions increases the possibility of better utilization of oat grains for food and feed and can help to identify grain lots at risk. Because of variable effects, it is also important to analyze the mycotoxin content after size sorting. For several of the mycotoxins, the concentrations were considerably higher in the small kernel fraction compared to unsorted grain. We assume that our study is the first report on the effect of size sorting on the content of Enns, NIV and BEA in oat grains.

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

#### *4.1. Oat Grain Materials*

Samples (approximately 0.5 kg, 14% moisture) from 24 oat grain lots (several cultivars) were obtained from various field experiments in southeast Norway (Table 2). Fourteen lots were harvested in 2015 at four different field locations, and ten were harvested in 2018 at one location. The grain lots were chosen based on preliminary tests showing that they contained HT-2 and T-2 toxins, which were the mycotoxins we were most interested in in this study. A sample of approximately 300g of harvested grain (raw material) of each lot was obtained by deviding on a riffle divider (Rationell Kornservice AS, Esbjerg, Denmark). After slight air cleaning (blowing) at "low speed" to remove dust, weed seeds and trash/straw pieces, each sample was further divided into sub-samples of approximately 100 g (unsorted sample) and 200 g. Each of the 200 g samples was size sorted into a large and small kernel fraction by passing the grain through a laboratory scale grain screening machine (in-house made, sieve size = 2.2 mm) at Kimen Seed Laboratory. The weight of unsorted, large and small kernel samples from each grain lot was recorded. Materials of the two size fractions, in addition to the unsorted sample, were ground on a high-speed rotor mill (ZM 200, Retsch, Haan, Germany) fitted with a 1 mm sieve and stored at −20 ◦C until analyses.

#### *4.2. Mycotoxin Analyses*

All grain samples were analyzed for the content of eleven different mycotoxins by using LC-MS/MS. The sample preparation was done according to the procedure published by Klötzer and Lauber [51] except that only 5 g aliquot of each sample was extracted with 20 mL mixture of acetonitrile and water (80:20 v/v). The analyses of the mycotoxins detected as cations (HT-2, T-2, Enns, BEA and ZEA) in grain samples harvested in 2015 were carried out using a Waters Ultima Pt MS/MS-detector, whereas the analyses of the mycotoxins detected as anions (DON, NIV, sum of 3-acetyl-DON and 15-acetyl-DON) were performed with a Thermo high resolution accurate mass (HRAM) Q-Exactive Orbitrap instrument. The mycotoxin analysis of grain from 2018 was carried out using HRAM Q-Exactive Orbitrap exclusively, using electrospray polarity switching in order to detect all the ionized mycotoxins in one run (Table 4). The toxins were separated on a Thermo Accucore aQ (100 × 2.1 mm i.d., 2.6 μm) column. A linear mobile phase gradient was used, starting with 100% water in 5 mM ammonium acetate reaching 100% methanol in 5 mM ammonium acetate after 9 min. The total run time was 18 min. The injection volume was 5 μL (Waters instrument) or 1 μL (Thermo instrument), the flow 0.3 mL/min and the column temperature was 30 ◦C. In the negative mode, mycotoxins were detected as acetate-adducts [M+CH3COO]<sup>−</sup> and in the positive mode mycotoxins were detected as ammonium adducts [M+NH4] <sup>+</sup> or hydrogen adducts [M+H]+. The identification criteria were retention time (RT) matched to reference standard, precursor ion accurate m/z mass within 5 ppm accuracy and the presence of at least one targeted product ion with accurate mass within 5 ppm accuracy and produced by fragmentation of the precursor ion. An in-house library of product ion spectra (MS2) for the mycotoxins aided in the identification. Quantification was based on the peak height of the precursor ions. Reference standards of the mycotoxins were purchased from Merck, Darmstadt, Germany. Calibration standards were prepared in the range of 1–1000 μg/kg. Limit of quantification (LOQ) was 1–10 μg/kg. The recovery of HT-2 and T-2 was confirmed to 100% using a certified oat reference material. Recovery of the other toxins was determined from spiked control samples that were prepared with each batch of samples. Recovery was 100% for DON, 3+15-Acetyl-DON and EnnB, 60%–70% for NIV, ZEA, EnnA, EnnA1 and EnnB1, and 45% for BEA. Our method reported the correct levels of HT-2, T-2, DON and ZEA (z-scores lower than 0.35) in oat meal in a proficiency test in 2019 [52].


**Table 4.** Parameters for the high resolution accurate mass (HRAM) detection of the analytes including retention time, precursor ion, adducts type of precursor and one of the product ions.

#### *4.3. Data Analyses*

The mean, minimum and maximum toxin concentration was calculated for each toxin for both years using Minitab 18. Percentage reductions of toxins in the large grain fractions and percentage increases in the small grain fractions were calculated compared to concentrations in the unsorted grain. Percentage weight reduction was calculated from the weight of the unsorted sample (= 100%) and the weight of the large and small kernel fractions (sum = 100%). The mean toxin levels in the small or the large grain fractions were compared to the mean toxin level in the unsorted grain fraction using a paired *t*-test in Minitab 18. The confidence level was adjusted according to the Bonferroni method to

obtain a simultaneous confidence level of 95%, and the differences between the means were considered as significant when *p*-values ≤ 0.05/2. The following relationships were analyzed by linear regression in Minitab 18: (i) the percentage of HT2+T2 reduction vs. the concentration level in the grain lots (unsorted grain), (ii) the sum of toxin concentration (HT2+T2 2015, 2018; DON 2018; EnnB 2015) in the small and the large grain fraction vs. the toxin concentration in the grain lots (unsorted grain), (iii) the percentage weight reduction vs. the percentage HT2+T2 reduction, and (iiii) the percentage of weight reduction vs. the HT2+T2 level in the grain lots. The relationships in (i), (iii) and (iiii) were studied across both years.

**Author Contributions:** G.B. designed the experiments, supervised the research and wrote the major parts of the manuscript; H.U.A. contributed to the data calculations and to the writing of the manuscript; M.A. analyzed the mycotoxins and contributed to the writing of the manuscript; I.S.H. was project leader, contributed to the data calculations and the writing of the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by The Agriculture and Food Industry Research Funds—FFL/JA (The Research Council of Norway grant number 254751/E50), Graminor AS, Lantmännen Corporate R&D, Felleskjøpet Agri SA, Felleskjøpet Rogaland Agder, Fiskå Mølle Moss AS, Strand Unikorn AS, Norgesmøllene AS and Kimen Seed Laboratory.

**Acknowledgments:** Grain lots were provided from field experiments at NIBIO and the Norwegian Agricultural Extension Service. We thank Kari Wahltoft at Kimen Seed Laboratory for cleaning/size sorting and weighing the samples, Ely Gauslaa and Chloè Grieu at NIBIO for grinding samples, Børge Holen for the development of the mycotoxin extraction method and analysis method on Waters LC-MS/MS, Jan Kristian Larsen for mycotoxin extraction of oat grain samples and Torfinn Torp for advice on statistical calculations.

**Conflicts of Interest:** The authors declare no conflict of interests related to this research. The funders had no role in the design of the study; in the collection, analysis, or interpretation of data; or in the writing of the manuscript and agreed to publish the results.

#### **References**


© 2020 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*

## **Development and Validation of a Liquid Chromatography High-Resolution Mass Spectrometry Method for the Simultaneous Determination of Mycotoxins and Phytoestrogens in Plant-Based Fish Feed and Exposed Fish**

**Amritha Johny 1,\*, Christiane Kruse Fæste 1, André S. Bogevik 2, Gerd Marit Berge 3, Jorge M.O. Fernandes <sup>4</sup> and Lada Ivanova <sup>5</sup>**


Received: 11 March 2019; Accepted: 11 April 2019; Published: 13 April 2019

**Abstract:** New protein sources in fish feed require the assessment of the carry-over potential of contaminants and anti-nutrients from feed ingredients into the fish, and the assessment of possible health risks for consumers. Presently, plant materials including wheat and legumes make up the largest part of aquafeeds, so evaluation of the transfer capabilities of typical toxic metabolites from plant-infesting fungi and of vegetable phytoestrogens into fish products is of great importance. With the aim of facilitating surveillance of relevant mycotoxins and isoflavones, we have developed and validated a multi-analyte LC-HRMS/MS method that can be used to ensure compliance to set maximum levels in feed and fish. The method performance characteristics were determined, showing high specificity for all 25 targeted analytes, which included 19 mycotoxins and three isoflavones and their corresponding aglycons with sufficient to excellent sensitivities and uniform analytical linearity in different matrices. Depending on the availability of matching stable isotope-labelled derivates or similar-structure homologues, calibration curves were generated either by using internal standards or by matrix-matched external standards. Precision and recovery data were in the accepted range, although they varied between the different analytes. This new method was considered as fit-for-purpose and applied for the analysis of customised fish feed containing wheat gluten, soy, or pea protein concentrate as well as salmon and zebrafish fed on diets with these ingredients for a period of up to eight weeks. Only mycotoxin enniatin B, at a level near the limit of detection, and low levels of isoflavones were detected in the feed, demonstrating the effectiveness of maximum level recommendations and modern feed processing technologies in the Norwegian aquaculture industry. Consequently, carry-over into fish muscle was not observed, confirming that fillets from plant-fed salmon were safe for human consumption.

**Keywords:** Atlantic salmon; zebrafish; liquid chromatography high-resolution mass spectrometry; mycotoxins; phytoestrogens; plant-based feed

**Key Contribution:** A multi-analyte LC-HRMS/MS method for 25 targeted mycotoxins and phytoestrogens was developed and validated in feed and fish matrices. Mycotoxins above the respective LOD were not detected in feed and dietary exposed fish, whereas phytoestrogens were found in soy and pea protein-based diets but carry-over into fish was not observed.

#### **1. Introduction**

Global fish production reached more than 171 million tonnes by 2016, of which 88% were directly used for human consumption and 12% (20 million tonnes) were used for the production of fishmeal and fish oil in aquaculture [1]. Fish and fishery products are an important source of essential nutrients in the human diet, and demand is growing in line with the increasing world population [2]. Aquaculture is the fastest-growing food industry and the intensification of the production depends on the utilisation of other resources for aquafeeds than fishmeal, for which exploitation is reaching an unsustainable level. Therefore, agricultural crops, mainly legumes, cereal grains and oilseeds, have been introduced in steadily increasing amounts into fish feeds, completely or partially replacing marine protein sources [3].

Plant protein sources mainly include soy, pea, lupine, alfalfa, wheat, corn, rape seeds, sunflower seeds, cotton seeds, sesame seeds, mustard oil cake, and white leadtree leaves [4]. Moreover, proteins from insects, microalgae, krill and single-cell proteins have been explored as replacements for fishmeal, but plant proteins are by far the most used ingredients in feed in aquaculture. The considerable changes in the diet composition of farmed fish include ingredients with physicochemical properties that potentially could lead to challenges regarding fish health and welfare, and product quality [5]. However, new processing technologies for plant protein extraction of undesirable components such as fertilisers, pesticides, persistent organic pollutants and heavy metals have allowed the transition from marine to agricultural sources [6]. The growth performance of plant-fed fish has been found to be adequate in short feeding studies [7], but concern about potential negative health effects from natural toxins and anti-nutritional factors including phytoestrogens remains [4,8]. Some anti-nutritional factors are considerably resistant against heat and digestion and have the potential for carry-over into the food chain. Several studies have shown that bioactive compounds may affect physiological functions in animals and humans including negative effects on intestinal health [9]; however, information for fish is limited [4]. The potential transfer of undesirable substances from new sources of aquafeeds might thus lead to potential health risks for consumers of fish products [10]. The assessment of transmissibility requires analytical methods that can be reliably applied for the detection of relevant natural contaminants in agricultural crops, and the considerable prevalence of mycotoxins and phytoestrogens makes them priority target analytes. However, only a few recent studies have surveyed mycotoxin levels in fish feed or farmed fish [11–16], and phytoestrogens are even less investigated [17,18].

There is a risk of mycotoxicosis in farmed fish due to the presence of mycotoxins in plant feed ingredients, but information on effects in fish is limited [11,19]. Mycotoxins comprise a large variety of secondary metabolites produced by fungi such as *Fusarium* spp., *Aspergillus* spp., *Alternaria* spp. and *Pencillium* spp. that infect agricultural crops both in the field and during storage, depending on their preferred growth conditions [20]. The presence of mycotoxins in practically all feed- and foodstuffs worldwide, although at different levels, is critical for nutritional security and safety, and important for animal and human health and welfare [21]. In moderate climate zones, major mycotoxin classes associated with *Fusarium* crop infections are trichothecenes, zearalenones and enniatins. The most important trichothecenes (polycyclic sesquiterpenoids) are A-type HT-2 toxin (HT-2) and T-2 toxin (T-2) and B-type deoxynivalenol (DON), including the acetylated and glucosidated derivatives 3-acetyl-deoxynivalneol (3-ADON), 15-acetyldeoxynivalenol (15-ADON) and deoxynivalenol-3-glucoside (DON-3G), as well as nivalenol (NIV). Furthermore, the mycoestrogen zearalenone (ZEN) shows considerable occurrence and toxicity. The ionophoric enniatins (ENN) B, B1, A, and A1 are detectable in almost all grain samples and considered an emerging threat [22]. In contrast, toxicity caused by ergot alkaloids such as ergosine, ergonovine, ergotamine, ergocristin, ergocornine and α-ergocryptine in *Claviceps purpurea*-infected cereals has been known as ergotism for centuries. Ergot contamination is a sporadic issue but appears to have increased in recent years. The storage mycotoxin of main concern in Nordic countries is ochratoxin A (OTA), a pentaketidic isocoumarin produced by *Penicillium* or *Aspergillus* sp. In contrast, aflatoxins and fumonisins normally do not occur in Norwegian feed commodities [23]. The European Commission has recommended maximum levels for important mycotoxins in different feed commodities [24]. Fish ingredients and composite fish feed are not specifically mentioned but the guidance levels for DON (5 mg/kg); ZEN (2 mg/kg) and OTA (0.25 mg/kg) also apply to aquaculture. Additionally, an indicative value for the sum of T-2 and HT-2 (250 μg/kg) in compound feed is provided by the EU Commission recommendation [25]. Comparable values have not been established for NIV, enniatins or ergot alkaloids because of the limited occurrence and toxicity data.

Phytoestrogens are plant-derived polyphenolic non-steroidal compounds with structural and functional similarity to animal oestrogens, which can bind to oestrogen receptors and activate oestrogen receptor-dependent pathways in mammals and fish [26]. Thus, they have the potential to disrupt the endocrine system by competing with endogenous hormones. Phytoestrogens can be broadly differentiated into isoflavones, coumestans and lignans, depending on the alkylation pattern in the basic isoflavone molecule structure [27]. Legumes, especially soy, are rich in isoflavones, which occur in plants mainly in glucosidated form, whereas the unconjugated molecules are prevalent after uptake. Important representatives of this substance class are the glucosides daidzin, genistin, glycitin and their respective free counterpart's daidzein, genistein and glycitein [28]. They are also potential substrates for metabolic glucuronidation or sulphatation reactions in the liver and kidneys due to the hydroxyl groups in the molecule and could be excreted as conjugates [29]. Processed soy protein concentrates have an increased aglycon content, which results in improved phytoestrogen absorption from the diet [30]. Exposure of fish to phytoestrogens in feed has been shown to cause reproductive effects and to affect growth and metabolism [31], but the levels in the edible tissue of soy-fed fish and potential human exposure have not been investigated so far.

The assessment of possible health risks from the consumption of fish fed with plant-derived feed requires the development of appropriate analytical methods for the detection of transferred contaminants and bioactive compounds. Mycotoxins are usually analysed by liquid chromatography tandem mass spectrometry (LC-MS/MS) with different multi-toxin methods and in various matrices such as bulk cereals, flour, nuts, food products and hay bales [32–40]. Advanced sampling schemes and extraction protocols have been developed, resulting in improved homogeneity and recovery so that method validation can be performed [41]. Sample preparation often includes single-step solvent extraction using acidic acetonitrile/water mixtures, followed by solid-phase extraction (SPE) or immunoaffinity purification [39]. Matrix effects can be controlled by using matrix-matched calibration and isotope-labelled internal standards (ISTD), which are available for trichothecenes but not for enniatins and ergot alkaloids [32,33,36–38,40]. Notably, fewer LC-MS/MS methods have been described for ergot alkaloids than for *Fusarium* toxins, focussing on rye, feed and seeds as typical matrices [34,37]. In contrast, phytoestrogens are mostly measured in physiological samples including human and animal plasma, milk and urine in connection with monitoring of dietary exposure [42,43]. The LC-MS/MS methods developed for the detection of phytoestrogens in soy and food items use methanol-water extraction and reversed-phase (RP) chromatography [44,45].

Earlier studies have measured several mycotoxins in feed ingredients, aquafeeds and fish fillets [11,13,14,16,46] but ergot alkaloids were not among the analytes. In addition, we have found one report of the occurrence of phytoestrogens in foods of animal origin, including a few fish samples [47]. Considering the potential consumer health risk resulting from the extensive introduction of agricultural crops into fish feed and contaminant carry-over, analytical methods for the reliable detection of natural toxins and bioactive compounds are required. The present study was thus intended to fill this gap

by developing a multiplexed LC-MS/MS method for the simultaneous quantification of 25 relevant feed-borne mycotoxins and phytoestrogens in feed and fish.

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

#### *2.1. Fish Feed with Fixed Contents of Wheat Gluten, Soy Protein or Pea Protein*

Finished feed has to comply with national and international legislation regarding maximum contents of certain contaminants including some mycotoxins [24,25]. In the present study, the fish diets were prepared in a fully equipped feed technology research facility based on materials that are commonly used in Norwegian aquaculture. Since the focus of the fish experiments was the potential transfer of natural contaminants from feed into fish, and not digestibility or feed utilisation, the composition was balanced with regard to plant-based ingredients (Table 1). Constant levels of 15% or 30% wheat gluten, soy protein concentrate or pea protein concentrate were achieved by adjusting the amount of fishmeal, which resulted in slight differences in the total crude protein and total lipid contents between the diets (Table 1). By keeping the ratio of plant-derived ingredients constant, comparability of the analytical results for the targeted metabolites was ensured.

**Table 1.** Composition of customised salmon and zebrafish feed (FM, fish meal; SPC, soy protein concentrate; PPC, pea protein concentrate).

