**Reduced Toxicity of Trichothecenes, Isotrichodermol, and Deoxynivalenol, by Transgenic Expression of the** *Tri101* **3-***O***-Acetyltransferase Gene in Cultured Mammalian FM3A Cells**

**Nozomu Tanaka 1, Ryo Takushima 2, Akira Tanaka 2, Ayaki Okada 1, Kosuke Matsui 3, Kazuyuki Maeda 3, Shunichi Aikawa 4, Makoto Kimura <sup>3</sup> and Naoko Takahashi-Ando 1,2,4,\***


Received: 14 September 2019; Accepted: 7 November 2019; Published: 10 November 2019

**Abstract:** In trichothecene-producing fusaria, isotrichodermol (ITDol) is the first intermediate with a trichothecene skeleton. In the biosynthetic pathway of trichothecene, a 3-*O*-acetyltransferase, encoded by *Tri101*, acetylates ITDol to a less-toxic intermediate, isotrichodermin (ITD). Although trichothecene resistance has been conferred to microbes and plants transformed with *Tri101*, there are no reports of resistance in cultured mammalian cells. In this study, we found that a 3-*O*-acetyl group of trichothecenes is liable to hydrolysis by esterases in fetal bovine serum and FM3A cells. We transfected the cells with *Tri101* under the control of the MMTV-LTR promoter and obtained a cell line G3 with the highest level of C-3 acetylase activity. While the wild-type FM3A cells hardly grew in the medium containing 0.40 μM ITDol, many G3 cells survived at this concentration. The IC50 values of ITDol and ITD in G3 cells were 1.0 and 9.6 μM, respectively, which were higher than the values of 0.23 and 3.0 μM in the wild-type FM3A cells. A similar, but more modest, tendency was observed in deoxynivalenol and 3-acetyldeoxynivalenol. Our findings indicate that the expression of *Tri101* conferred trichothecene resistance in cultured mammalian cells.

**Keywords:** trichothecene; biosynthetic pathway; acetyltransferase; deacetylase; deoxynivalenol; 3-acetyldeoxynivalenol; isotrichodermol; isotrichodermin

**Key Contribution:** This is the first study to demonstrate that transfection of *Tri101* encoding trichothecene 3-*O*-acetyltransferase confers resistance to trichothecenes in mammalian cells. As the C-3 acetyl group of trichothecenes is liable to be digested by esterases present in cell culture, the toxicities of 3-*O*-acetyltrichothecenes may have been overestimated in previous studies.

#### **1. Introduction**

Trichothecenes are a group of mycotoxins produced by several filamentous fungi including *Fusarium*. They have a trichothecene skeleton of 12,13-epoxytrichothec-9-ene in common, but their side chain modification greatly varies, resulting in a large difference in their toxicity [1,2]. They exert toxicity mainly through the inhibition of protein synthesis in eukaryotes [3]. Some fusaria are notorious fungi, known to cause *Fusarium* head blight in important crops.

Based on the biosynthetic pathways, trichothecenes are largely divided into two groups: t-type trichothecenes, with a modifying group at C-3 position, and d-type trichothecenes, without a modifying group [4,5]. In the early biosynthetic pathway of trichothecenes, trichodiene is first synthesized through the cyclization of farnesyl pyrophosphate by Tri5p [6]. Trichodiene is then oxygenated to isotrichotriol, followed by spontaneous cyclization to produce the following t-type trichothecenes: *Fusarium graminearum* produces deoxynivalenol (DON), nivalenol, and their acetylated forms; *Fusarium sporotrichioides* produces T-2 toxin, neosalaniol, and diacetoxyscirpenol. On the other hand, in non-fusaria, including *Myrothecium*, *Trichoderma*, *Trichothecium, Stachybotrys,* and *Spicellum* [7], trichodiene is oxygenated to isotrichodiol followed by spontaneous cyclization to produce d-type trichothecenes. There is another classification method of dividing trichothecenes into four types (A–D) based on their chemical structures [8]. Type B group is distinguished from type A group by the presence of a ketone at C-8, and either type is synthesized by both t-type and d-type trichothecene producers. Type C group, which contains a 7,8-epoxide, and type D group, which contains a macrocyclic ring between C-4 and C-15, are exclusively produced by d-type trichothecene producers.

In fusaria, the first trichothecene produced is isotrichodermol (ITDol), which is immediately acetylated at the C-3 position by 3-*O*-acetyltransferase, Tri101p, into isotrichodermin (ITD) [5,9]. It has been reported that a 3-acetylated trichothecene is generally less toxic than its corresponding 3-hydroxy form [10–13]. Hence, it was suggested that Tri101p confers self-resistance to trichothecene produced in fusaria; indeed, *Schizosaccharomyces pombe* transformed with the *Tri101* gene has been found to be more resistant to T-2 toxin than the wild type (WT) [5]. So far, *Tri101* has been isolated from fusaria, including *F. graminearum* and *F. sporotrichioides*, and their enzyme kinetics have been evaluated extensively [14].

Trichothecenes are phytotoxins and may play a role as a virulence factor that contributes toxin producers to infect host plants [15–17]. Therefore, in order to combat *Fusarium* head blight, researchers have made extensive efforts to examine the effect of the transgenic expression of *Tri101* on trichothecene resistance and *Fusarium* infection in host cereals [18]. These studies have unequivocally proved that transgenic tobacco and rice showed improved trichothecene resistance [19,20]. In wheat, moderate tolerance to *Fusarium* infection was observed in a field trial [21]. On the contrary, in barley, deoxynivalenol may not be a virulence factor, and no effect on infection was observed in a field trial [22]. Thus, the gene manipulation of *Tri101* has been effective to confer tolerance to trichothecenes in some host plants, although the effect might be limited.

In contrast to the case with microbes and plants, the effects of trichothecene acetylation at the C-3 position and the transfection of *Tri101* into cultured mammalian cells are complicated to understand. Although there is approximately a 100-fold difference in terms of the in vitro inhibition of protein synthesis in rabbit reticulocytes between 3-acetylated trichothecenes and their corresponding 3-hydroxy forms, this difference was reduced to tenfold in terms of the in vivo inhibition of glycoprotein synthesis in BGK-21 cells [5]. Although it has been reported that a 3-acetyl trichothecene has lower toxicity than its corresponding 3-hydroxy form [2], there is a conflicting report of 3-acetyl T-2 toxin being as toxic as T-2 toxin in human cell cultures [23], and, so far, no reports have shown that improved resistance to trichothecenes is conferred by *Tri101* transfection in cultured mammalian cells. Moreover, it has been reported that the acetyl group at the C-3 position of trichothecene is easily removed in animal systems [24]. In this way, the toxicity of a C-3 acetyl trichothecene has tended to be inaccurately estimated by measuring the toxicity of a mixture of C-3 acetyl and C-3 hydroxy trichothecenes. Therefore, in this study, we attempted to maintain the 3-*O*-acetyl group attached to the trichothecene skeleton by the transgenic expression of *Tri101* in murine FM3A cells, resulting in a more accurate evaluation of the toxicity of 3-acetyl trichothecenes. We also examined whether *Tri101* transfection into mammalian cultured cells improves their resistance to 3-hydorxytrichothecenes.

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

#### *2.1. Deacetylation of 3-Acetyldeoxynivalenol (3-ADON) and ITD*

Assuming that the acetyl group of 3-acetyltrichothecenes was cleaved to produce more-toxic 3-hydroxytrichothecenes in cytotoxicity and animal studies, we first verified the extent of the deacetylation of two trichothecenes, 3-ADON, which is an acetylated form of the most common trichothecene DON, and, ITD, the first acetylated trichothecene produced by Tri101p in trichothecene biosynthesis.

First, 3-ADON or ITD was added to H2O, 125 mM Tris-HCl buffer (pH 6.5), and RPMI medium (without any additive), and the solutions were incubated in a CO2 incubator for 48 h. Neither DON nor ITDol was detected in H2O containing its corresponding 3-acetyltrichothecene, but some deacetylated forms were detected in both the buffer and the RPMI medium after incubation (Supplementary Table S1), which suggests that non-enzymatic deacetylation of these 3-acetylated trichothecenes occurred in them. Next, the actual culture medium for FM3A cells was examined. RPMI medium was supplemented with antibiotics, sodium pyruvate, β-mercaptoethanol, and deactivated fetal bovine serum (FBS). Considering the possibility that FBS contained contaminating esterases, we prepared the culture medium supplemented with non-boiled FBS (designated as N-medium) and boiled FBS (designated as B-medium). When 3-ADON or ITD was added to the media, the deacetylation rates of these trichothecenes were much higher in N-medium than B-medium, with ITD being deacetylated more efficiently than 3-ADON. While up to 96% of ITD was deacetylated to ITDol in the N-medium, only 4.6% of ITD was deacetylated in the B-medium. These results suggest that FBS contained broad-substrate-specificity esterases that could be deactivated by boiling, and ITD was more efficiently deacetylated than 3-ADON by these esterases.

FM3A cells were confirmed to grow in B-medium as normally as in N-medium; thus, the cells were incubated in B-medium containing 3-ADON or ITD in order to examine the stability of 3-*O*-acetyl group of these trichothecenes in a cell culture environment. As shown in Supplementary Table S1, 4.9% and 10.3% of 3-ADON was deacetylated to DON in 3 mL of cell culture medium containing 1 <sup>×</sup> 10<sup>5</sup> cells and 6 <sup>×</sup> 105 cells, respectively, while 17.0% and 39.1% ITD was deacetylated to ITDol, respectively. These results suggest that FM3A cells themselves contained esterases acting on C-3 of the trichothecene skeleton.

#### *2.2. Transfection of FM3A Cells and Screening Cell Lines with High Expression of Tri101*

First, we confirmed that the growth of FM3A WT cells was completely suppressed in the N-medium containing 30 μg/mL of blasticidin S and in the B-medium containing 0.1 μg/mL (0.40 μM) of ITDol. The growth of FM3A transfected with an empty vector was also completely suppressed in the B-medium containing 0.40 μM of ITDol. Thus, just after transfection of the vector carrying *Tri101*, the transformants were screened in N-medium with 60 μg/mL of blasticidin S and, subsequently, with increasing concentrations of the antibiotic (up to 250 μg/mL). Next, we transferred the screened cells into B-medium with 0.40 μM of ITDol, followed by B-medium with 0.80 μM of ITDol, in order to screen the transformants with a high Tri101p activity. Over 100 cell lines that survived the screening process were then cloned in B-medium containing 0.16 μM of ITDol by limiting dilution. Crude enzyme was prepared from each clone and the Tri101p activity toward ITDol or DON was initially evaluated by TLC, followed by HPLC. We selected one cell line, designated as G3, which showed the highest Tri101p activity. In the absence of the drugs, the growth rate of G3 was comparable to that of the WT cells.

#### *2.3. Acetylase and Deacetylase Activities of Crude Cell Extracts from WT and G3 Cells*

Next, we measured the in vitro acetylase activities of WT and the transformant G3 cells by HPLC. We added 169 μM DON or 200 μM ITDol at a final concentration to the reaction mixture, in order to measure the approximate Vmax, as these concentrations were much higher than the *K*<sup>m</sup> values for DON and ITDol, 11.7 and 10.2 μM, respectively [14]. The crude enzymes from the WT cells showed no

acetylase activities toward either substrate, with or without induction by dexamethasone (DEX; an inducer of transgene expression) (Figure 1, right). In contrast, the crude enzymes from the G3 cells showed acetylase activities toward both DON and ITDol. As expected, significantly higher acetylase activities were observed in the crude enzymes from G3 cells pretreated with DEX than in those without DEX (*p* < 0.05). This result is consistent with the previous observation that DEX resulted in a twoto fivefold increase in the levels of expression of a luciferase gene from the mouse mammary tumor virus-long terminal repeat (MMTV-LTR) promoter in FM3A cells [25]. All the crude enzyme prepared here showed deacetylase activities toward both 3-ADON and ITD (Figure 1, left).

**Figure 1.** Acetylase and deacetylase activities of crude enzymes from FM3A cells. Wild-type (WT) and the transformant (G3) cells were pretreated with or without DEX (dexamethasone) and the crude enzyme was prepared from the harvested cells. The enzyme reaction was carried out in 200 μL reaction mixture, with (**right**) or without (**left**) 1 mM acetyl CoA, and 10 μg of a trichothecene as a substrate; 148 μM 3-ADON (3-acetyldeoxynivalenol), 171 μM ITD (isotrichodermin), 169 μM DON (deoxynivalenol), or 200 μM ITDol (isotrichodermol) at a final concentration. The trichothecene denoted at the root of the arrows represents the substrate, while the trichothecene at the tip of the arrow is the product. The percentage reaction rate (%) represents the initial molar ratio of the product over the added substrate per protein (1.0 mg/mL) in the reaction mixture. The values represent the average ± standard deviation (SD) (*n* = 3).

#### *2.4. Acquired Trichothecene Resistance in the Cells Transfected with Tri101*

After the WT and G3 cells were seeded in the B-medium, the vehicle or ITDol was added to the culture medium, and the cell numbers were counted on day 3, 5, or 7 (Figure 2). Both WT and G3 cells grew normally until day 3 in the medium containing the vehicle, after which they underwent apoptosis. In contrast, in B-medium containing 0.40 μM ITDol, almost all the WT cells underwent apoptosis by day 3, but G3 cells continued to grow until day 5, after which the cell numbers decreased. In B-medium containing 0.80 μM ITDol, the G3 cells slowly continued to grow until day 5–7. In the medium containing a higher concentration of ITDol, even the G3 cells struggled to grow. Thus, it was concluded that G3 cells had acquired resistance to ITDol. This is the first study reporting that the transfection of *Tri101* into mammalian cells confers resistance to trichothecenes.

**Figure 2.** Growth of WT and the transformant G3 cells in B-medium containing ITDol. After pretreatment with DEX, the cells were seeded in 24-well plates, and ITDol solution was added to the cells. On day 3, 5, and 7, the cells were harvested and diluted with trypan blue solution, and the number of the live cells were counted. The values represent the average ± SD (*n* = 3). The numbers of live cells of G3 which showed statistical differences from those of their corresponding WT are marked with asterisks (\* *p* < 0.05; \*\* *p* < 0.01; \*\*\* *p* < 0.001).

#### *2.5. Cytotoxicity Evaluation of Each Trichothecene in WT and G3 Cells*

We performed a water-soluble tetrazolium salts (WST) assay to evaluate the cytotoxicity of DON, 3-ADON, ITDol, and ITD more accurately, based on the assumption that the cytotoxicity of 3-acetyltrichothecenes had been overestimated in previous studies. First, the half maximal inhibitory concentration (IC50) values of ITDol and ITD in WT cells were determined using N-medium or B-medium (Supplementary Figure S1). The observed IC50 values of ITDol and ITD were close when WT cells were incubated in N-medium (0.19 ± 0.01 and 0.26 ± 0.01 μM, respectively), however, a large difference was evident when assayed in B-medium (0.20 ± 0.01 and 2.14 ± 0.06 μM, respectively). The higher sensitivity of cultured mammalian cells to ITD in N-medium seems to be attributed to the contaminating esterases that act on ITD (Supplementary Table S1), resulting in the formation of much more toxic ITDol. This observation strongly suggests that without an elaboration to keep the 3-acetyl group attached to the ITDol ring, the toxicity of 3-acetyltrichothecene tends to be overestimated. Although the apparent IC50 value of 2.14 μM observed for ITD in B-medium could be closer to the actual IC50 value of intact ITD, this value may in fact be lower than the actual value, as an acetyl group may not remain attached to the C-3 position of the trichothecene skeleton due to the presence of intrinsic esterases hydrolyzing the 3-*O*-acetyl group in cultured mammalian cells (Figure 1, chemical reactions). Thus, we attempted to measure the concentration of ITDol possibly produced from ITD in the cell culture medium in this assay; however, ITDol was not detected in our HPLC analysis (limit of detection; 1.2 μM). As the IC50 value of FM3A cells against ITDol was as low as 0.22 μM (Figure 3),

ITDol levels below the detection limit possibly affected cell growth. The higher toxicity of ITD may be explained by the C-3 deacetylation of ITD during the assay.

Next, in order to evaluate cytotoxicity of 3-ADON and ITD more accurately, we determined their IC50 values using G3 cells expressing *Tri101*. Figure 3A shows the dose–response cytotoxicity curves of DON and 3-ADON using WT and G3 cells. In WT cells, the difference in the IC50 values between DON and 3-ADON was around 12-fold. In G3 cells, all the IC50 values were increased significantly compared to their corresponding values in WT cells. The IC50 value in G3 cells pretreated with DEX was slightly higher than that without DEX, and the values were significantly different in DON, but not in 3-ADON. No significant difference was observed in the IC50 value of these trichothecenes in WT cells with and without DEX pretreatment. These results indicate that the transformant G3 cells were more resistant to both DON and 3-ADON than WT cells due to the expression of *Tri101* in G3 cells. As it is likely that the more accurate IC50 of 3-ADON was that of G3 cells treated with DEX (11.68 μM), this value was significantly higher than that of WT cells (7.09 μM without DEX, 6.45 μM with DEX) suggesting that cytotoxicity of 3-ADON might be overestimated in previous studies. Here, the IC50 of 3-ADON was at least 21-fold higher than that of DON (0.56 μM) obtained in WT.

The effect of the transgenic expression of *Tri101* in G3 was more obvious in the IC50 values of ITDol and ITD than in those of DON and 3-ADON. Figure 3B presents the result of the WST assay of ITDol and ITD in WT and G3 cells. In WT cells, the difference of the IC50 values between ITDol and ITD was 13–16 fold. In G3 cells, without DEX induction, the IC50 values of ITDol and ITD were almost doubled compared to those in the WT cells. Moreover, with DEX induction, the IC50 values of ITDol and ITD were three- to fourfold compared with those in the WT cells. This indicates that DEX induced *Tri101* expression, which resulted in the increased acetylation of ITDol to ITD. Thus, the IC50 values of both ITDol and ITD were almost doubled in G3 cells treated with DEX compared to those without DEX, and the values were significantly higher in both ITDol and ITD by DEX pretreatment. In WT cells, no significant increase was observed in the IC50 value of these trichothecenes by DEX pretreatment. Similar to DON and 3-ADON, a more accurate IC50 of ITD was likely to be that of G3 cells treated with DEX (9.58 μM), thus, this value was significantly higher than that of WT cells (2.78 μM without DEX, 3.59 μM with DEX), suggesting that cytotoxicity of ITD in previous studies might be overestimated. It represented an at least 44-fold increase of IC50 value of ITD (9.58 μM), which is supposedly to be more accurate, compared to that of ITDol obtained in WT (0.21–0.22 μM).

In this study, the level of *Tri101* expression was found to affect the IC50 values of trichothecenes in mammalian transformants. However, the *Tri101* expression level may not be sufficient to fully compete with the C-3 deacetylation activity of strong endogenous esterases. Nevertheless, this study clearly showed that the 3-acetyl group of 3-*O*-acetyltrichothecenes could be hydrolyzed enzymatically in mammalian cells and that their cytotoxicities were overestimated in other studies. The IC50 of trichothecenes were much lower than the *Km* of Tri101p against them, making it difficult to completely acetylate the C-3 of the trichothecenes added to the culture.

This is the first report to unambiguously demonstrate the acquired trichothecene resistance of cultured mammalian cells transformed with *Tri101*. These results strongly support the theory that C-3 acetylation blocks the toxicity of t-type trichothecenes, which serves as a self-defense mechanism of the producing organisms [5]. In contrast to fungi and plants, the strong C-3 deacetylase activity of mammalian cells activates 3-acetyltrichothecenes, the toxicity of which has been overestimated in previous studies.

**Figure 3.** The dose–response cytotoxicity curves of the trichothecenes. (**A**) DON or 3-ADON and (**B**) ITDol or ITD was used as a toxin. Cytotoxicity assay of trichothecenes on FM3A cells was carried out using Cell Counting Kit 8 (CCK-8) reagent in 96-well plates. Cells were pretreated with or without 50 μM DEX. One microliter of a toxin or vehicle was added to 99 μL of cell culture, which was seeded (5 <sup>×</sup> 104/mL) one day before. After two days of incubation with a toxin or vehicle, a WST assay was performed. Growth inhibition (%) was calculated as follows: 100 × {(OD450 of vehicle control − OD450 of background) − (OD450 of trichothecene added − OD450 of background)}/(OD450 of vehicle control − OD450 of background). The IC50 values represent the average ± standard error.

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

#### *3.1. Production and Purification of Trichothecenes*

Each trichothecene was obtained from rice medium [26] or rice flour liquid medium [27] inoculated with *F. graminearum*: *Gibberella zeae* JCM 9873 for DON [28] and *F. graminearum* DSM 4528 for 3-acetyldeoxynivalenol (3-ADON) [29]. Isotrichodermin was obtained from the culture medium of *F. graminearum* MAFF 111233 *Fgtri11* disruptant (*Fgtri11-* , NBRC 113181) [27], while ITDol was obtained by the deacetylation of ITD in 2.8% ammonium solution. For purification, the ethyl acetate extract was applied to Purif-Rp2 equipped with Purif-Pack SI 30 μm SIZE (Shoko Scientific, Kanagawa, Japan), and the fraction containing a target trichothecene was concentrated. The concentrate was dissolved in ethanol and applied to preparative HPLC (UV detection at 254 nm for DON and 3-ADON, and at

195 nm for ITD and ITDol) equipped with a C18 column (Pegasil ODS SP100 10 ϕ × 250 mm, Senshu Scientific Co., Ltd., Tokyo, Japan). The concentration of each purified trichothecene was measured by HPLC equipped with a C18 column (Pegasil ODS SP100 4.6 ϕ × 250 mm) based on the standard curve previously obtained. The purity of each trichothecene used for the assays was confirmed to be >99%.

#### *3.2. Maintenance of Cultured Cells*

FM3A WT cells which were originally established from mammary carcinoma in C3H/He mouse were purchased from RIKEN BRC (Tsukuba, Japan). Cells were grown in RPMI1640 medium (Nacalai Tesque Co., Inc. Kyoto, Japan) with 10% FBS (Biowest, Nuaillé, France), penicillin/streptomycin (1000 units/mL each), 1 mM sodium pyruvate, and 50 μM 2-mercaptoethanol, denoted "N (non-boiled)-medium." In order to deactivate any contaminating esterases in FBS, we prepared the same medium with FBS boiled at 100 ◦C for 5 min beforehand, denoted "B (boiled)-medium." Cultured cells were incubated in a CO2 incubator (5% CO2, 37 ◦C).

#### *3.3. Evaluation of Stability of 3-Acetyl Trichothecenes*

Twenty micrograms of 3-ADON (59.1 nmol) or ITD (68.4 nmol) in ethanol was added into a 6 cm dish containing 3 mL of water, 125 mM Tris-HCl buffer (pH 6.8), and RPMI1640 medium without any supplements. Each solution was incubated in a CO2 incubator (5% CO2, 37 ◦C) for 48 h. Similarly, 3 mL of the N-medium and B-medium containing 10 μg of 3-ADON (29.6 nmol) or ITD (34.2 nmol) was incubated in a CO2 incubator for 48 h. We also prepared FM3A WT cells (3.3 <sup>×</sup> 104/mL or <sup>2</sup> <sup>×</sup> 105/mL) in 3 mL B-medium supplemented with 10 μg of 3-ADON or ITD. After 48 h, each solution or medium was extracted with the same volume of ethyl acetate twice. The extract was dried up under an N2 stream and resuspended in 200 μL ethanol. These concentrating steps were necessary for detection of the mycotoxins. The filtered samples were subjected to HPLC.

#### *3.4. Plasmid Construction*

*Tri101* was excised from pCold-His-TRI101 [30] by double digestion with *Nde*I and *Eco*RI. The *Nde*I–*Eco*RI fragment was cloned into the corresponding sites of pColdIII-NFH (Supplementary Figure S2). The *Tri101* sequence, *N*-terminally fused with a FLAG-HA tag, was amplified by PCR from the resulting plasmid using primers FLAG-HA-Tri101F\_NheI (5 -AAAAGCTAGCATGGACTACAAGGACGACGAT-3 ) and FLAG-HA-Tri101R\_XhoI (5 -AAAACTCGAGCTAACCAACGTACTGCGCATA-3 ). After double digestionwith*Nhe*I and*Xho*I, the PCR productwasinserted between these sites downstream of aMMTV-LTR promoter in pMAM2BSD [25], yielding a DEX-inducible *Tri101* expression vector, pMAM2BSD\_FH\_Tri101.

#### *3.5. Transfection of FM3A Cells and Selection of Transformants*

The plasmid pMAM2BSD\_FH\_Tri101 or pMAM2BSD was transfected into FM3A cells using Lipofectamine® 2000 reagent (Thermo Fisher Scientific, Waltham, MA, USA) following the manufacturer's instructions. Plasmid DNA (5 μg) was diluted in 100 μL Opti-MEM medium without FBS. Five microliters of Lipofectamine® 2000 was diluted in 100 μL of Opti-MEM medium and left for 5 min. Diluted plasmid DNA and Lipofectamine were mixed gently and left for 20 min and added to FM3A cells (7~10 <sup>×</sup> 105/3 mL).

Transfected cells were incubated for one day before being transferred to N-medium containing 60 μg/mL of blasticidin S (Fujifilm Wako Pure Chemical Co., Osaka, Japan). During selection, the concentration of blasticidin S was increased to 250 μg/mL over one month. For the next selection step, the cells were transferred to B-medium containing 50 μM DEX and 0.1 μg/mL (0.4 μM) ITDol. The ITDol concentration was increased to 0.2 μg/mL (0.8 μM) over three weeks. The selected transformants (>100 cell lines) were cloned using limited dilution methods in B-medium containing 50 μM DEX and 0.04 μg/mL ITDol (0.16 μM). After limited dilution, both DEX and ITDol were removed from the medium. Among the survived transformants in this condition, the cell line whose

crude enzyme showed the highest 3-*O*-acetyltransferase activity was chosen and designated as G3. Measurement of activities of prepared crude enzymes was performed as described below.

#### *3.6. Preparation and Reaction of Crude Enzymes from Cultured Cells*

Wild-type FM3A cells and the selected transformants were seeded (10 mL aliquots of cell suspension at 1 <sup>×</sup> <sup>10</sup>4/mL). After a three-day incubation period, DEX (50 <sup>μ</sup>M at final concentration) was added for the induction of *Tri101* expression in the transformed cells, and for comparison, in WT cells. In parallel, the carrier solvent (10% dimethylsulfoxide [DMSO]; 0.1% DMSO at a final concentration) was added to these cells. Once the cells became semi-confluent, these were harvested and washed twice in 2 mL PBS. The washed cells were suspended in 500 μL of 125 mM Tris-HCl buffer (pH 6.8) and sonicated. After centrifugation (13,000 rpm at 4 ◦C for 5 min), the supernatant was subjected to a protein quantitation assay (PierceTM BCA protein assay kit, Thermo Fisher Scientific). The resulting crude enzyme fraction was assayed for modification activity against C-3 of the trichothecene skeleton.

To initiate the reaction, 100 μL of the crude enzyme (with a protein concentration around 1.0 mg/mL) was mixed with 100 μL of substrate solution, which contained a 3-hydroxy (169 μM DON or 200 μM ITDol) or 3-acetyl (148 μM 3-ADON or 171 μM ITD) trichothecene in 125 mM Tris-HCl buffer (pH 6.5), respectively, with or without 1 mM acetyl CoA. We used rTRI101 from *Escherichia coli* (BL21 (DE3)-pColdIII-*Tri101*) (1 mg/mL) [30] as a positive control and boiled rTRI101 as a negative control. Each enzyme assay was performed in triplicate. The mixture was incubated for 3 h at 37 ◦C and the reaction was terminated by adding 200 μL of ethyl acetate. Trichothecenes were extracted with ethyl acetate three times, concentrated under N2, resuspended in the appropriate amount of ethanol, and subjected to TLC and HPLC analysis.

#### *3.7. TLC and HPLC Analysis*

The trichothecenes in each sample were detected roughly by TLC and the concentration of each trichothecene was precisely measured by HPLC. The samples dissolved in ethanol were applied to TLC plates (Merck, Darmstadt, Germany) and developed with ethyl acetate/toluene (3:1). Trichothecenes on TLC plates were visualized using the 4-(p-nitrobenzyl)pyridine-tetraethylenepentamine method [31].

For HPLC analysis, the samples dissolved in ethanol were filtered, and 10 μL was applied to HPLC (LC-2000 plus; Jasco, Tokyo, Japan) loaded with a C18 reverse-phase column (4.6 ϕ × 250 mm, PEGASIL ODS SP100). Trichothecenes were eluted with a mobile phase of acetonitrile and water, according to the following steps: for DON and 3-ADON, 20% acetonitrile for 6 min, a linear gradient of 20%–50% acetonitrile for 9 min, 50% acetonitrile held for 6 min, then 20% acetonitrile held for 10 min; for ITDol and ITD, 20% acetonitrile for 6 min, a linear gradient of 20%–100% acetonitrile for 16 min, 100% acetonitrile held for 3 min, then 20% acetonitrile held for 10 min. Deoxynivalenol and 3-ADON were detected at 254 nm, while ITDol and ITD were detected at 195 nm. The molar concentration of each trichothecene was calculated from the peak area, based on the previously obtained standard curve.

#### *3.8. Toxicity Evaluation of Trichothecenes*

In order to evaluate the effect of *Tri101* gene transfection into FM3A cells, we counted the number of live WT and G3 cells, to which ITDol had been added beforehand. First, the cells were seeded at <sup>5</sup> <sup>×</sup> <sup>10</sup>4/mL in B-medium and were treated with 50 <sup>μ</sup>M DEX. After one day of incubation, 1 mL of the cells were transferred to a 24-well plate and ITDol (0, 0.4, 0.8, 1.6, 4.0, or 8.0 μM ITDol in 0.1% DMSO at a final concentration) was added and incubated for 3, 5, or 7 days in a CO2 incubator at 37 ◦C. The cells were carefully harvested and centrifuged (300× *g*, 5 min, 25 ◦C). The precipitated cells were then tapped gently, and the appropriate amount of medium was added. The cell suspension was diluted twofold with 0.5% trypan blue (Nacalai Tesque) and only the live cells were counted.

For the cytotoxicity evaluation, colorimetric cell viability assays using WST-8 were performed. Each trichothecene was prepared in 50% DMSO as follows: for DON, 0, 0.0034, 0.010, 0.034, 0.10, 0.34, 1.0, or 2.0 mM (0, 0.001, 0.003, 0.01, 0.03, 0.1, 0.3, or 0.6 mg/mL); for 3-ADON, 0, 0.030, 0.089, 0.30,

0.89, 3.0, 8.9, or 30 mM (0, 0.01, 0.03, 0.1, 0.3, 1.0, 3.0, or 10.0 mg/mL); for ITDol, 0, 0.004, 0.008, 0.016, 0.040, 0.080, 0.16, or 0.40 mM (0, 0.001, 0.002, 0.004, 0.01, 0.02, 0.04, or 0.10 mg/mL); for ITD, 0, 0.0068, 0.014, 0.034, 0.068, 0.14, 0.34, 0.68, 1.4, 3.4, or 6.8 mM (0, 0.002, 0.004, 0.01, 0.02, 0.04, 0.1, 0.2, 0.4, 1.0, or 2.0 mg/mL). Wild-type or G3 cells were suspended at a cell density of 5 <sup>×</sup> 104/mL in B-medium with or without 50 μM DEX and incubated for one day. Into each well of 96-well plates, 99 L of cell suspension was seeded and 1 μL of each trichothecene prepared above or 50% DMSO (vehicle) was added in triplicate. The cells were incubated for 48 h in a CO2 incubator. Into each well, 10 μL of Cell Counting Kit (CCK)-8 solution (Dojindo Molecular Technologies, Inc., Kumamoto, Japan) was added, and the microtiter plate was incubated for 3 h at 37 ◦C. OD450 was measured using a MultiskanTM FC plate reader (Thermo Fisher Scientific), and growth inhibition (%) caused by each trichothecene was calculated.

#### *3.9. Statistical Analysis*

Statistical analysis was performed using Student's *t*-test. Regarding the calculation and statistical analysis of IC50 values, log-logistic model was applied using R version 3.5.0 (R project for statistical computing).

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6651/11/11/654/s1, Table S1: The rate of deacetylation of 3-ADON and ITD under various conditions, Figure S1: The dose–response cytotoxicity curves of trichothecenes, Figure S2: Construction of pMAM2BSD\_FH\_Tri101.

**Author Contributions:** N.T., R.T., and A.T. performed the cell culture experiments and analyses. A.O., K.M. (Kosuke Matsui), and S.A. produced, purified and identified each trichothecene. K.M. (Kazuyuki Maeda) produced the plasmid. M.K. and N.T.-A. designed the experiments and discussed the draft manuscript. N.T.-A. wrote the paper.

**Funding:** This study was supported by the Tojuro Iijima Foundation for Food Science and Technology (H27-Kyodo-11, H25-Kojin-37), Grant-in-Aid for Scientific Research (KAKENHI 15K07459), and the Science Research Promotion Fund (the Promotion and Mutual Aid Corporation for Private School of Japan, 2018 and 2019).

**Acknowledgments:** We greatly acknowledge T. Sagawa, Y. Yashiro, and Y. Higuchi for their technical assistance, and T. Matsumura, a professional statistician, in WAKARA (Tokyo Japan) for his technical support with statistical analysis. We are also very thankful to D. Yamanaka and S. Takahashi from the University of Tokyo for their invaluable advice.

**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*

## **Twenty-Eight Fungal Secondary Metabolites Detected in Pig Feed Samples: Their Occurrence, Relevance and Cytotoxic E**ff**ects In Vitro**

#### **Barbara Novak 1,\*, Valentina Rainer 1, Michael Sulyok 2, Dietmar Haltrich 3, Gerd Schatzmayr <sup>1</sup> and Elisabeth Mayer <sup>1</sup>**


Received: 27 August 2019; Accepted: 11 September 2019; Published: 14 September 2019

**Abstract:** Feed samples are frequently contaminated by a wide range of chemically diverse natural products, which can be determined using highly sensitive analytical techniques. Next to already well-investigated mycotoxins, unknown or unregulated fungal secondary metabolites have also been found, some of which at significant concentrations. In our study, 1141 pig feed samples were analyzed for more than 800 secondary fungal metabolites using the same LC-MS/MS method and ranked according to their prevalence. Effects on the viability of the 28 most relevant were tested on an intestinal porcine epithelial cell line (IPEC-J2). The most frequently occurring compounds were determined as being cyclo-(L-Pro-L-Tyr), moniliformin, and enniatin B, followed by enniatin B1, aurofusarin, culmorin, and enniatin A1. The main mycotoxins, deoxynivalenol and zearalenone, were found only at ranks 8 and 10. Regarding cytotoxicity, apicidin, gliotoxin, bikaverin, and beauvericin led to lower IC50 values, between 0.52 and 2.43 μM, compared to deoxynivalenol (IC50 = 2.55 μM). Significant cytotoxic effects were also seen for the group of enniatins, which occurred in up to 82.2% of the feed samples. Our study gives an overall insight into the amount of fungal secondary metabolites found in pig feed samples compared to their cytotoxic effects in vitro.

**Keywords:** *Fusarium*; *Aspergillus*; *Penicillium*; *Alternaria*; fungi; emerging mycotoxin; in vitro; IPEC-J2; occurrence data

**Key Contribution:** Less investigated fungal compounds were found more frequently and in higher concentrations in pig feed samples than regulated mycotoxins. Furthermore, some of them cause stronger cytotoxic effects than deoxynivalenol in vitro.

#### **1. Introduction**

Mycotoxins are described as secondary fungal metabolites produced by filamentous fungi in the field, and during storage and transportation under appropriate conditions. The main fungal species *Fusarium*, *Alternaria*, *Aspergillus*, and *Penicillium* possess a remarkable potential to produce a wide range of metabolites. Researchers assume that around 38% of the known 22,630 microbial products are of fungal origin; however, the exact number is the still subject of research [1].

The European Food Safety Authority (EFSA) set maximum allowed concentration levels or guidance values for some of these metabolites called "regulated mycotoxins" [2]. These include

aflatoxins, deoxynivalenol, HT-2/T-2 toxins, zearalenone, fumonisins, and ochratoxin A, which have been proven to cause detrimental effects in humans and animals. Some other fungal metabolites are so-called "emerging mycotoxins" and are claimed to be metabolites for which no guidance values exist yet. Scientific opinions and risk assessments have only been published for aflatoxin precursors (such as sterigmatocystin), ergot alkaloids, enniatins, beauvericin, and moniliformin so far [3–6]. However, for many other fungal metabolites, only a few, contradictory studies are available [7–9]. One reason for the underestimation of their potential threat might be the limited knowledge about their prevalence. Because of the continuous development of LC-MS/MS (liquid chromatography tandem mass spectroscopy) methods, a more sensitive detection of a wide range of secondary fungal metabolites in feed and food matrixes is now available [10]. Furthermore, studies indicate that climatic change might affect mycotoxin production patterns, leading to an increase in the emerging mycotoxin level [11,12]. In 2013, Streit et al. [13] published data on the *Fusarium* metabolites beauvericin, enniatins, aurofusarin, and moniliformin, which occur in 98%, 96%, 84%, and 76% of 83 analyzed feed samples, respectively. In the same study, the well-investigated mycotoxins deoxynivalenol (89%), zearalenone (87%), and fumonisins (22%) were additionally found in the analyzed samples. Besides, metabolites produced by *Alternaria* spp. were detected in many of the analyzed feed samples (e.g., alternariol, 80%; alternariol monomethyl ether, 82%; tenuazonic acid, 65%; and tentoxin, 80%). Although further publications revealed higher occurrences and concentrations of uncommon mycotoxins, they were limited to specific geographic regions and feed commodities [14–18].

The aim of this study was, first, to provide a complete picture of the effects of 28 chemically diverse fungal metabolites on the cell viability of an intestinal porcine epithelial cell line (IPEC-J2), and second, to compare the tested concentrations with the actual values found in pig feed samples worldwide. The cell line IPEC-J2 was chosen as an optimal test model since the intestine is the first target after ingesting mycotoxin-contaminated feed and most of the metabolites are absorbed in the jejunum [19]. The selection of fungal metabolites to be tested was mainly based on their relevance and occurrence in pig feed samples obtained from the BIOMIN mycotoxin survey program, conducted in the years 2014 to 2019. Here, the most relevant metabolites were *Fusarium*-derived compounds, such as enniatins, beauvericin, moniliformin, culmorin, fusaric acid, etc., but also many compounds produced by the genera *Alternaria*, *Penicillium*, and *Aspergillus* were evaluated. Cytotoxicity studies for selected metabolites are available for different cell types [9,20–23], but here we compared, for the first time, the in vitro effects of 28 metabolites taken from occurrence data of 1141 pig feed samples analyzed with the same analytical method [18].

#### **2. Results**

#### *2.1. Occurrence Data*

The challenge of collecting and analyzing manifold feed samples from all over the world is not only a logistical one, but is mainly found in trying to measure the samples with the same analytical method, limits of detection (LOD), and performance parameters at the same instrument [11] to achieve comparable results. Therefore, BIOMIN started a unique global mycotoxin survey including samples from 44 countries, obtained between February 2014 and February 2019, where 4978 feed samples were analyzed for more than 800 different metabolites using a multi-analyte LC-MS/MS-based method [24]. For our study, we selected 1141 out of the 4978 feed samples that were only intended for swine nutrition.

#### Secondary Fungal Metabolites in Finished Pig Feed Samples

Analyzed pig feed samples were taken as described in Section 5.3.2. and evaluated according to the method of Malachová et al. [24]. The list of analytes in this method has increased over recent years and covers now more than 800 metabolites. The most frequently occurring secondary fungal metabolites of 34 emerging and masked mycotoxins, as well as regulated mycotoxins in samples from 44 different countries, are summarized in Table 1. Twenty-eight metabolites, which are not regulated yet or are defined as an "emerging mycotoxin" [25], were selected to be tested for their effects on the viability of an intestinal porcine epithelial cell line (IPEC-J2). Gliotoxin and patulin were added to these tests because of their previously described harmful effects, making novel data beneficial for the scientific community [26,27]. The uncommon metabolites 15-hydroxy-culmorin, infectopyron, and asperglaucide could not be tested because authentic analytical standards were not commercially available. The prevalence (%), mean, median, and maximum concentrations (in μg/kg) for each compound are summarized in Table 1. In the following section, the focus lies only on the median and maximum concentrations for better legibility.

A ranking according to the prevalence of the main 34 secondary fungal metabolites as shown in Table 1 indicates that the cyclic dipeptide cyclo-(L-Pro-L-Tyr) was most frequently found with an occurrence of 87.6% and a maximum concentration of 34,910 μg/kg; followed by moniliformin (82.6%) and enniatin B (82.2%). Even though the median concentrations of the latter two were only 17 and 30 μg/kg, individual samples reached levels of up to 2053 μg/kg and 1514 μg/kg, respectively. In 82.1% of all pig feed samples, enniatin B1 had similar concentrations as enniatin B (max. 1846 μg/kg; median 34 μg/kg). Two other common enniatins, enniatin A1 and A, were found in 77.4% and 49.5% of all samples, albeit at lower concentrations compared to the B analogues (B and B1). Additionally, samples were vastly contaminated with the *Fusarium* metabolites aurofusarin and culmorin, which were identified in approximately 909 feed samples (79%), with median concentrations of 210 μg/kg and 118 μg/kg, which was comparable to the median concentration of deoxynivalenol (DON) with 193 μg/kg measured in 879 feed samples (77%). This contrasts with the maximum measured concentrations found for aurofusarin (85,360 μg/kg) and culmorin (157,114 μg/kg), while only 34,862 μg/kg for DON was detected. The second regulated mycotoxin on our list was zearalenone (ZEN), which was found at rank 10 with a prevalence of 73.3% and a median concentration of 18 μg/kg. Another uncommon metabolite, which was determined in 75.4% of all samples, was tryptophol, which was thus positioned before ZEN in the ranking according to prevalence. About 69% of the samples were positive for the hexadepsipeptide beauvericin (BEA) with a median concentration of only 6 μg/kg. Individual samples were contaminated with a maximum concentration of 413 μg/kg BEA.

Mainly unexpected compounds, such as emodin, brevianamid F, equisetin (EQUI), and cyclo-(L-Pro-L-Val), were found at the prevalence ranking positions of 13, 15, 16 and 34, respectively. These were nevertheless detected in about 70, 69, 64 and 62% of all feed samples, respectively, relating to 799, 787, 730 and 707 feed samples, respectively, out of a total of 1141. Compounds found to a lesser extent were *Alternaria* metabolites, such as tenuazonic acid (55.0%), alternariol (50.7%), alternariol monomethyl ether (40.3%), and tentoxin (37.3%). Another group of detected metabolites was produced by *Aspergillus* and/or *Penicillium* spp., but also by some *Fusarium* species, and these included 3-nitropropionic acid (3-NP), apicidin (API), kojic acid, bikaverin, fusaric acid, and mycophenolic acid. While 3-NP and API were found in 56.5% and 52.2% of samples, respectively, the others were detected in ≤33.7% feed samples.

Other regulated mycotoxins, such as fumonisin B1, ochratoxin A, and aflatoxin B1, did not occur frequently or in high concentrations in pig feed samples. These metabolites were determined in only 43% (median 82.7 μg/kg), 5% (median 2.6 μg/kg), and 3% (median 2.0 μg/kg) of all samples, respectively (data not shown). The mycotoxins gliotoxin and patulin, relevant due to their toxicity, scarcely occurred (0.2%) or could not be detected (<LOD) in these pig feed samples.


**Table 1.** Summary of 34 secondary fungal metabolites ranked according to their prevalence with mean, median, and maximum concentration, measured in 1141 finished feed samples for swine obtained from February 2014 to February 2019 using a multi-analyte LC-MS/MS-based method.

Concentrations in μg/kg. \* = not tested in vitro (lack of availability or known as regulated or masked mycotoxin). If not otherwise stated, a threshold (t) of >1.0 μg/kg or >LOD (limit of detection) was established.

As seen in Figure 1, 77.7% of the obtained samples had their origin in Europe, mainly from Germany (22.4%), Denmark (15.3%), and Austria (14.1%). Few samples came from other parts of the world, such as Central and South America (9%), Russia and Asian countries (5.6%), North America (3.5%), and Africa (2.4%). The analyzed feed samples obtained from Australia (0.6%) were marginal. Thirteen samples (1.1%) could not be assigned to a specific country. It can thus be stated that the picture of the occurrence primarily reflected the situation of Central Europe, since no major changes were seen in the ranking when samples from the other countries were excluded from the analysis.

**Figure 1.** Origin of 1141 pig feed samples obtained from different parts of the world.

#### *2.2. Cell Viability after 48 h Toxin Treatment*

The cell viability of intestinal porcine epithelial cells (IPEC-J2) was assessed after an incubation of 48 h with the respective metabolite in concentrations of up to 150 μM. Absolute and relative IC50 values were calculated. Deoxynivalenol (DON) (Figure 2A) was included in the test system as an internal standard to have comparable values representing a well-investigated mycotoxin. DON had already reduced the cell viability to 78.8% at 1.25 μM, resulting in absolute and relative IC50 values of 2.55 μM and 1.88 μM, respectively. The 28 tested fungal metabolites are listed according to their calculated absolute IC50 value, starting from the strongest (for 11 metabolites, relative and absolute IC50 value could be calculated) over moderate (for 5 metabolites, only a relative IC50 value could be calculated) to no cytotoxicity (for 12 metabolites, IC50 calculation was not possible). Apicidin was the metabolite that showed the strongest cytotoxic effect with an absolute IC50 value of 0.52 μM (relative 0.49 μM), followed by gliotoxin, bikaverin, beauvericin, and patulin. For all tested metabolites except enniatin B, the calculated relative IC50 value was lower than the absolute IC50 value (Figure 2).

**Figure 2.** Viability (%) after 48 h of incubation with deoxynivalenol (**A**), apicidin (**B**), gliotoxin (**C**), bikaverin (**D**), beauvericin (**E**), and patulin (**F**) (tested at (0.156–20 μM)); except for apicidin (0.0049–5 μM)) of confluent intestinal porcine epithelial cells (IPEC-J2). Data represent mean ± standard deviation.

The next six fungal metabolites that showed a strong cytotoxic effect are presented in Figure 3. The four enniatins B, A, B1, and A1 showed similar absolute IC50 values of 3.25 μM, 3.40 μM, 3.67 μM, and 4.15 μM, respectively. Aurofusarin was less toxic, resulting in an absolute IC50 value of 11.86 μM. Although emodin was tested at higher concentrations, viability did not further decrease at concentrations >50 μM. Hence, IC50 values of 18.71 μM (absolute) and 13.09 μM (relative) were calculated.

**Figure 3.** Viability (%) after 48 h of incubation of enniatin B (**A**), enniatin A (**B**), enniatin B1 (**C**), enniatin A1 (**D**), aurofusarin (**E**), and emodin (**F**) (tested at (0.156–20 μM), except for emodin (0.625–150 μM)) of confluent intestinal porcine epithelial cells (IPEC-J2). Data represent mean ± standard deviation.

Assessment of the following five fungal metabolites resulted in moderate cytotoxicity on IPEC-J2, as relative, but no absolute, IC50 values could be calculated (Figure 4). Examination of the effects of equisetin was only possible up to a concentration of 20 μM, leading to a decreased viability of 64.0% (±7.6%). The other fungal metabolites seen in Figure 4 (B–E) were tested in a concentration range of 0.625 to 150 μM. The *Alternaria* metabolites tenuazonic acid (B) and alternariol (C) showed similar effects of reducing cell viability starting at 20 μM (76.2% and 76.3%), resulting in relative IC50 values of 20.88 μM and 20.26 μM. However, despite increased concentrations being used, viability remained around 50%. Rubrofusarin (D) reduced viability started at 50 μM (64.0 ± 8.7%) with a calculated relative IC50 value of 21.54 μM. Interestingly, mycophenolic acid (E) led to an abrupt decreased viability (26.1% ± 12.1%) at very low concentrations (0.156 to 1.25 μM), but no further loss in viability was determined at concentrations from 2.5 to 150 μM.

**Figure 4.** Viability (%) after 48 h of incubation with mycophenolic acid (**A**) (0.156–150 μM), equisetin (**B**) (0.156–20 μM), alternariol (**C**), tenuazonic acid (**D**), and rubrofusarin (**E**) ((0.156–150 μM) for C–E) of confluent intestinal porcine epithelial cells (IPEC-J2). Data represent mean ± standard deviation.

For the following twelve fungal secondary metabolites, shown in Figures 5 and 6, no reduced cell viability was seen, and hence, no IC50 calculation was possible.

**Figure 5.** Viability (%) after 48 h of incubation with culmorin (**A**), moniliformin (**B**), roquefortine C (**C**), tentoxin (**D**) ((0.156–20 μM) for A–D), alternariol monomethyl ether (**E**), and kojic acid (**F**) ((0.625–150 μM) for E–F) of confluent intestinal porcine epithelial cells (IPEC-J2). Data represent mean ± standard deviation.

**Figure 6.** Viability (%) after 48 h of incubation with cyclo-(L-Pro-L-Tyr) (**A**), cyclo-(L-Pro-L-Val) (**B**), tryptophol (**C**), 3-nitropropionic acid (**D**), and brevianamid F (**E**), and fusaric acid (**F**) ((0.625–150 μM) for A–F) of confluent intestinal porcine epithelial cells (IPEC-J2). Data represent mean ± standard deviation.

An overview of the obtained IC50 values (in μM and μg/kg) for the 16 secondary fungal metabolites with strong and moderate effects on viability is presented in Table 2, together with their median and maximum concentrations in μg/kg found in 1141 pig feed samples.


**Table 2.** List of absolute and relative IC50 values in μM (left columns) and in μg/kg (middle columns), compared to median and maximum concentration in μg/kg (right columns) ranked according their cytotoxicity.

nc = not calculable.

#### **3. Discussion**

Twenty-eight fungal metabolites plus DON were assessed for their effects on the viability of intestinal porcine epithelial cells (IPEC-J2) in comparison to their prevalence in 1149 analyzed pig feed samples. The strength of our study is that analytical measurements were performed by using one single LC-MS/MS multi-mycotoxin method, as well as the same cell line and viability assay, for all samples to achieve comparable data [24]. A general problem with global mycotoxin occurrence data is usually the difference in methodologies, which make a comparison of concentrations challenging due to different limits of detection (LODs), sample extractions, and performance parameters [11]. Research on toxicology and occurrence of emerging mycotoxins is still scarce, although it has been steadily rising during the past few decades. This can partly be explained by the fact that the sensitivity and potential of LC-MS-based methods have been improved by analyzing hundreds of metabolites simultaneously, as well as by lower detection limits [24]. Furthermore, tremendous climatic abnormalities in some parts of the world favor an increasing formation of uncommon fungal metabolites [12]. Even though not all the detected metabolites might be relevant regarding food and feed safety, the abundance and co-occurrence of those compounds in different feed matrixes might pose a certain risk to susceptible animals. IPEC-J2 provides an optimal in vitro model, as swine is considered as the most sensitive species regarding mycotoxicosis [28]. This non-transformed cell line is isolated out of jejunal epithelial cells, in which the absorption of nutrients and other compounds mainly takes place. Furthermore, it possesses strong morphological and functional similarity to intestinal epithelial cells in vivo [29,30] compared to cancer cell lines such as Caco-2 and HT-29 cells. For in vitro experiments, we decided to test concentrations of up to 150 μM or, in order not to exceed a solvent concentration of 1%, the highest possible test concentration. A solvent concentration of 1% did not negatively influence viability in our test system (data not shown). For our study, we have chosen to discuss the absolute IC50 value, as this value is representative for the concentration, where the half-maximal inhibitory concentration was calculated. As described by Sebaugh [31], when the response of more than two assay concentrations is above 50%, the calculation of the relative IC50 value would be ambiguous. Furthermore, the calculation of the relative IC50 value uses the top and bottom plateau, even when values do not reach 50% viability. Therefore, those IC50 values would result in false positive results. Mycophenolic acid (MPA) is a representative example, as viability was never lower than 66%; however, a relative IC50 value of 0.543 was calculated, making this toxin one of the most toxic in our ranking. The calculation for the

absolute IC50 value was not possible for this toxin, and therefore, the calculation of the absolute IC50 values was used, first, because it was more accurate, second, only with this value were we able to rank these fungal metabolites according to their toxicity, and third, for reflecting a realistic scenario. For the sake of completeness, both IC50 values are shown if calculation was possible. For comparison with the occurrence data, maximum and median concentrations are used, as the median concentration seems to be a more accurate measure of central tendency because of a generally skewed data distribution.

Deoxynivalenol (DON) was included in our study as a comparable internal reference toxin since it is well investigated and manifold deleterious effects of DON are known, as summarized in the review by Pestka [32]. For DON, an IC50 concentration of 2.55 μM (= 756 μg/kg) was calculated, whereas a median concentration of 193 μg/kg was found in swine feed. Although the detected concentration was lower than the determined IC50 value, it is known that in particular chronic low doses of DON lead to immune dysregulation, growth retardation, and impaired reproduction [32]. This is of increased importance because the bioavailability for DON after oral administration is 98.6% ± 23.6% in pigs [33]. This might also be true for other metabolites, but neither feeding trial nor bioavailability studies have been conducted for most of them. Therefore, the choice of in vitro concentrations is challenging. Stability and retention time in the gastrointestinal tract (GIT) have not been sufficiently researched. Thus, an accumulation of metabolites in the GIT by ingesting chronic low concentrations is very likely and we tried to test a broad concentration range up to 150 μM, if possible.

Apicidin (API), only discovered in the year 1996 and isolated from *Fusarium* spp., showed the strongest cytotoxicity in our test system (IC50 of 0.52 μM = 324 μg/kg). This is in accordance with a study on other cell lines, reporting an IC50 concentration of 0.16 to 3.8 μM on cell proliferation [34]. Furthermore, this compound possesses antiprotozoal activity [35] and causes toxic effects in rats leading to death at levels of 0.05 and 0.1% [36]. Although the measured median concentration in feed samples was rather low (8 μg/kg), API was found in more than half of the samples and individual samples reached concentrations of up to 1568 μg/kg. Streit et al. [13] detected a maximum concentration of 160 μg/kg in 66% positively tested samples, but additional occurrence studies are missing. The second-ranked cytotoxic compound was gliotoxin (GLIO) with an IC50 value of 0.64 μM (= 209 μg/kg). This *Aspergillus fumigatus* metabolite hardly occurs in swine feed, but was detected in corn silage used as cattle feedstuff [37]. GLIO has been reported to cause immunosuppressive, genotoxic, apoptotic, and cytotoxic effects determined in a rat intestinal cell line (IEC-6), in hamster ovary cells (CHO), and in mouse macrophages (RAW264.7) [38,39], and might pose a risk for other animal species that are frequently exposed to this mycotoxin. Literature about the effects and occurrence of our third-ranked toxic compound bikaverin (BIK) is scarce. It is reported as a red pigment produced by different *Fusarium* spp. with antibiotic and antibacterial properties [40,41]. A study from 1975 reported its cytotoxicity against three different cancer cell lines, leading to ED50 values of 0.5–4.2 μg/mL (1.31–10.99 μM) [42], which is comparable with our IC50 value of 1.86 μM. Furthermore, we found BIK in almost 30% of the feed samples with a maximum concentration of 1564 μg/kg, whereas another study reported contaminations of up to 51,130 μg/kg [13]. Our in vitro data give the first evidence of its cytotoxic effects on epithelial cells, and considering an IC50 value of only 1.86 μM (= 711 μg/kg), the high contamination level might be alarming. We would like to point out that for this toxin, viability did not further decrease below 50% after applying higher concentrations, which has also been observed for other metabolites such as enniatin B, emodin, and tenuazonic acid. More research has been carried out for the cyclic hexadepsipeptides beauvericin (BEA) and the group of enniatins (ENNs; A, A1, B, B1). Their cytotoxicity was demonstrated in a variety of cell culture models [7,9,43–46] and is attributed to their ionophoric properties. A recently published study from Fraeyman et al. [46] revealed that proliferating IPEC-J2 are more sensitive to BEA, but not to the ENNs. Differentiated cells seem to be more robust against their detrimental effects, which was already shown by Springler and Broekaert et al. [47,48]. According to Fraeyman's study, the overall cytotoxicity after 24 h of incubation was ranked as BEA > ENN A >> ENN A1 > ENN B1 >>> ENN B. This contrasts with our results that led to the following ranking: BEA > ENN B > ENN A > ENN B1 > ENN A1 in proliferating

IPEC-J2. This discrepancy could be due to different incubation time points (24 h vs. 48 h) and different measured endpoints (flow cytometry vs. Sulforhodamine B assay). Additionally, more data about their occurrence are available. We mostly found ENN B and B1 (≈82% positive); however, 77.4%, 68.7%, and 49.5% of the feed samples were also contaminated with ENN A, BEA, and ENN A1, respectively. The maximum detected concentrations varied between 1514, 1846, 307, 413 and 549 μg/kg, respectively. Considering that those compounds usually co-occur, the total amount can exceed a level of 4500 μg/kg or even 5543 μg/kg, as described in Kovalsky et al. [18]. A high prevalence of BEA and ENNs in cereal grains was already described in other peer-reviewed studies [14,15,18,49]. Additionally, a high oral bioavailability, particularly seen for ENN B1, has been reported [50,51], which turn them into a potential risk factor for exposed animals.

For patulin (PAT), a low IC50 value of 3.21 μM (= 495 μg/kg) was determined as well, but since not a single sample was contaminated with PAT, it does not seem to be relevant regarding porcine health. PAT is better known as a feed contaminant in fruits, especially apples and vegetables, and its toxicity was demonstrated in different animal species [52,53]. Other conclusions have to be drawn regarding the *Fusarium* metabolite aurofusarin (AURO), which led to an IC50 value of 11.86 μM (= 6766 μg/kg) and was detected in a median and maximum concentration of 210 μg/kg and 85,360 μg/kg, respectively. An interference of AURO with antioxidants and fatty acids in the eggs and embryos of quails has already been reported [54,55], as well as its negative effect on the growth performance in red tilapia [56] and cytotoxicity in mammalian cells [23,57,58]. Our results are comparable with a study from Jarolim et al. [23], reporting a significant decrease in viability of HT29 and HCEC-1CT cells starting at 5 μM AURO after 48 h. Even though the calculated IC50 value of 6766 μg/kg seems to be high in our approach, when considering the maximum found concentration of 85,360 μg/kg, a critical risk assessment is urgently required. The last metabolite where an IC50 calculation was possible was emodin (EMO), which is claimed to be a therapeutic agent of various diseases used in traditional Chinese medicine for centuries. As EMO is produced by a range of different plant families, and found ubiquitously in herbs, trees, and shrubs [59], a potential risk to animal health seems very unlikely.

Five out of the tested metabolites, mycophenolic acid (MPA), equisetin (EQUI), alternariol (ALT), tenuazonic acid (TeA), and rubrofusarin (RUB) led to a slight decrease in viability, and relative IC50 values could be determined. Despite increasing concentrations, a reduction of more than 50% was not found; therefore, absolute IC50 values could not be calculated. We have chosen the SRB assay to measure the protein content because this cell target was the most sensitive one in preliminary studies. However, it seems that for other compounds, a different cell target might be more suitable. As described in an study from 2017 [60], different endpoint analyses can lead to a different outcomes. Especially for MPA, a study about its cytotoxicity has been published where the mitochondrial activity of Caco-2 cells was only decreased by 45% at 780 μM MPA after 48 h of incubation, but no effect was seen in THP-1 monocytes [61]. These results vary from ours since we determined a constant decline of 20.0 to 33.6% viability already starting from 1.25 to 150 μM MPA by measuring the cellular protein content. Interestingly, similar results about ALT were published, in which the cell viability of HepG2 cells did not decrease in a concentration-dependent manner up until a concentration of 100 μM [62,63].

None of the other compounds, culmorin (CUL), moniliformin (MON), roquefortine C, tentoxin, alternariol monomethyl ether, kojic acid, cyclo-(L-Pro-L-Tyr), cyclo-(L-Pro-L-Val), tryptophol, 3-nitropropionic acid, brevianamid F, and fusaric acid (FA) showed negative effects in the tested concentration range on IPEC-J2. However, we would like to refer to four of them: although the metabolite culmorin (CUL) elucidated no cytotoxic effect in our tests, it was found in 79.7% of all samples, with remarkable concentrations of 157,114 μg/kg. Interestingly, its natural occurrence is always correlated to the occurrence of DON. A recent study showed that CUL is able to inhibit the glucuronidation of DON in human liver microsomes, and thus, its detoxification process [64,65]. These findings make the compound highly relevant regarding synergistic effects, not only for the detoxification of DON, but also for other toxins. MON was already described as a hazardous

contaminant for poultry, and therefore, a risk assessment has been recently published by EFSA [6,66]. Finally, although being non-cytotoxic in our experimental approach, cyclo-(L-Pro-L-Tyr) was found to be the most frequently detected compound in our survey program with 87.6% positive samples. Only a few publications described its antibacterial activity against two gram-negative bacteria, as well as its cytotoxic and genotoxic effects in lymphocytes and various cell lines to date [67–69]. However, due to its high occurrence, along with its high concentration, further studies are needed. Furthermore, we would like to stress that other metabolites, such as aflatoxin B1 and fumonisin B1, are also known for their non-cytotoxic effects in vitro and their detrimental effect in animals. Thus, a lack of cytotoxicity in vitro does not necessarily indicate a complete harmlessness.

#### **4. Conclusions**

Taken together, our study focused on the occurrence data and concentrations of 28 secondary fungal metabolites without regulatory guidelines and their effect on the viability of an intestinal porcine cell line. In the majority of the analyzed pig feed samples, low median concentrations (15 of 28 metabolites gave median concentrations of <20 μg/kg) were determined; however, some individual samples were contaminated with high concentration levels, which might be relevant for animal health. The maximum occurrence values exceeded the absolute IC50 concentrations for apicidin, bikaverin, and aurofusarin. Moreover, even if acute exposure to most of the metabolites is low, concerns regarding chronic exposure at lower levels are rising. An important factor that needs to be considered for further investigations comprise the absorption, distribution, metabolism, and excretion (ADME) of the substances and their ability to enter the target cell. Therefore, cytotoxicity studies provide a first overall picture of the relevance of the detected compounds and serve as a suitable alternative and prerequisite to animal testing. For a qualified risk assessment, however, reliable combination studies to investigate synergistic, additive, and antagonistic effects are needed due to the frequent co-occurrence of toxic compounds, especially with the regulated main mycotoxins. Finally, data from feeding trials in productive livestock with chronic exposure of those compounds have to be the logical target for the testing process within the next few years.

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

#### *5.1. Cell Culture*

The porcine jejunal intestinal epithelial cell line, IPEC-J2 (ACC 701; Leibnitz Institute DSMZ, German Collection of Microorganisms and Cell Cultures, Braunschweig, Germany), originated from a neonatal, unsuckled piglet. These non-transformed cells were continuously maintained in complete cultivation medium consisting of Dulbecco's modified eagle medium (DMEM/Ham's F12 (1:1), Biochrom AG, Berlin, Germany), supplemented with 5% fetal bovine serum, 1% insulin-transferrin-selenium, 5 ng/mL epidermal growth factor, 2.5 mM Glutamax (all GibcoTM, Life Technologies, Carlsbad, CA, USA), and 16 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (Sigma-Aldrich, St. Louis, MO, USA), and grown at 39 ◦C in a humidified atmosphere of 5% CO2. IPEC-J2 between passages 1–15 were routinely seeded at 1 <sup>×</sup> 106 cells/mL in 150 cm2 tissue culture flasks (Starlab, Hamburg, Germany) with 28 mL complete cultivation medium and subcultured upon confluence every 3–4 days. For assays, confluent cells were detached using Trypsin (0.25%)-ethylenediaminetetraacetic acid (EDTA, 0.5 mM, Sigma-Aldrich, St. Louis, MO, USA). Cell culture was regularly tested and found to be free of mycoplasma contamination via PCR (Venor®GeM Mycoplasma Detection Kit; Minvera Biolabs GmbH, Berlin, Germany).

#### *5.2. Material*

All tested chemicals are listed in Table 3 and were dissolved either in dimethylsulfoxid (DMSO, Sigma-Aldrich, St. Louis, MO, USA), ethanol (EtOH absolut, VWR International, Radnor, PA, USA), acetonitrile (ACN, Sigma-Aldrich, St. Louis, MO, USA), or distilled water.


**Table 3.** List of chemicals.

Adipogen Life Sciences, Switzerland; BioAustralis, Australia; Biopure, Austria; MedChem Express, Sweden; Santa Cruz, Germany; Sigma-Aldrich, USA.

#### *5.3. Method*

#### 5.3.1. Cell Viability Assay

For the measurement of cell viability, 3 <sup>×</sup> 104 cells/well were seeded into a flat-bottom, cell-culture-treated 96-well microplate (Eppendorf) in 200 μL cultivation media and grown for 24 hours at 39 ◦C and 5% CO2. After reaching confluence, IPEC-J2 were treated with a broad range of concentrations of all chemicals (listed in Table 3).

A sulforhodamine B (SRB) assay (Xenometrix, Allschwil, Switzerland) was used to determine cellular protein content and was performed according to the manual. Briefly, the supernatant was discarded and the cell layer was washed with 300 μL SRB I solution per well. Then, 100 μL SRB II fixing solution was added to each well and the plate was incubated for 1 h at 4 ◦C. After the incubation time, cells were washed three times with 200 μL distilled water. Cells were stained with 50 μL SRB III labelling solution/well for 15 min, and afterwards, cells were washed again two times with 400 μL SRB IV rinsing solution. Then, bound SRB was solubilized with 200 μL SRB V and after 15–60 min, absorbance was measured at 540 nm and a reference wavelength of 690 nm using a microplate reader (Synergy HT, Biotek, Bad Friedrichshall, Germany).

#### 5.3.2. LC-MS/MS Multi-Analyte Method

A total of 1141 samples from 75 countries were provided by the BIOMIN Mycotoxin Survey for measuring up to 800 analytes, including fungal and bacterial metabolites, pesticides, and veterinary drug residues using a LC-MS/MS multi-mycotoxin analysis method [24]. Samples were collected after instruction or by trained staff only from February 2014 until February 2019. For the present study, data from finished pig feed samples including 28 fungal metabolites were chosen for detailed analysis (see Table 1). The threshold *(t)* was set to be >1.0 μg/kg or the limit of detection (LOD), whichever was higher.

A minimum of 500 g of sample was submitted to the laboratory of the Institute of Bioanalytics and Agro-Metabolomics at the University of Natural Resources and Life Sciences Vienna (BOKU) in Tulln. After reception, samples were immediately milled, homogenized, and finally analyzed. Samples were extracted with a mixture of acetonitrile (ACN), water, and acetic acid (79:20:1, per volume) on a rotary shaker for 90 min. After centrifugation, the supernatant was transferred to a glass vial and diluted with a mixture of ACN, water, and acetic acid (20:79:1, per volume), and was injected into the LC-MS/MS system (electrospray ionisation and mass spectrometric detection employing a quadrupole mass filter). Quantification was done according to an external calibration using a multi-analyte stock solution.

The method was performed according to the guidelines from the Directorate General for Health and Consumer affairs of the European Commission, published in document No. 12495/2011 [70].

#### 5.3.3. Statistics and Evaluation

The half-maximal inhibitory concentrations (IC50) were calculated using GraphPad Prism (GraphPad Prism Version 7.03, San Diego, CA, USA). For calculation of the relative IC50 value, data was log-transformed and fitted to a four-parameter logistic equation:

$$Y = \text{Bottom} + (Top - \text{Bottom}) / (1 + 10((LogIC50 - X) \ge HillSlope))\tag{1}$$

The molar concentration of a substance that reduced viability to 50% between the top and the bottom plateau was calculated.

For the calculation of the absolute IC50 value, data was log-transformed and following equation was used:

$$Y = \text{Bottom} + (Top - Bottom) / (1 + 10 \left( (LogIC50 - X) \ge HillSlope + log \left( \frac{Top - Bottom}{Fifty - Bottom} - 1 \right) \right) \tag{2}$$

Fifty = 50; Top = 100

The molar concentration of a substance that reduced viability to 50% of the maximum viability was calculated.

**Author Contributions:** B.N. conceived and designed the experiments; B.N. and V.R. performed the experiments; B.N., M.S., and E.M. analyzed the data; M.S. provided the analysis tool (LC-MS/MS method); B.N. wrote the paper; D.H., G.S., and E.M. reviewed the paper.

**Funding:** This research received funding from the Austrian Research Promotion Agency (Österreichische Forschungsförderungsgesellschaft FFG (grant numbers 853863 and 859603), as well as EFREtop (grant number 864743).

**Acknowledgments:** We would like to thank Philipp Fruhmann (University of Natural Resources and Life Sciences, Vienna) for the generous gift of the culmorin stock, which was produced according to Weber et al. [65].

**Conflicts of Interest:** B.N., V.R., G.S., and E.M. are employed by BIOMIN. However, this circumstance did not influence the design of the experimental studies or bias the presentation and interpretation of results. The other two authors, M.S. and D.H., 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* **BmTudor-sn Is a Binding Protein of Destruxin A in Silkworm Bm12 Cells**

#### **Jingjing Wang, Weina Hu and Qiongbo Hu \***

Key Laboratory of Bio-Pesticide Innovation and Application of Guangdong Province, College of Agriculture, South China Agricultural University, Guangzhou 510642, China; wangjingjing@stu.scau.edu.cn (J.W.); hwn688094@163.com (W.H.)

**\*** Correspondence: hqbscau@scau.edu.cn; Tel.: +86-20-8528-0308; Fax: +86-20-8528-0292

Received: 19 December 2018; Accepted: 23 January 2019; Published: 24 January 2019

**Abstract:** Destruxin A (DA), a hexa-cyclodepsipeptidic mycotoxin secreted by the entomopathogenic fungus *Metarhizium anisopliae*, was reported to have an insecticidal effect and anti-immunity activity. However, its molecular mechanism of action remains unclear. Previously, we isolated several potential DA-affinity (binding) proteins in the *Bombyx mori* Bm12 cell line. By docking score using MOE2015, we selected three proteins—BmTudor-sn, BmPiwi, and BmAGO2—for further validation. First, using Bio-Layer Interferometry in vitro, we found that BmTudor-sn had an affinity interaction with DA at 125, 250, and 500 μM, while BmPiwi and BmAGO2 had no interaction signal with DA. Second, we employed standard immunoblotting to verify that BmTudor-sn is susceptible to DA, but BmPiwi and BmAGO2 are not. Third, to verify these findings in vivo, we used a target engagement strategy based on shifts in protein thermal stability following ligand binding termed the cellular thermal shift assay and found no thermal stability shift in BmPiwi and BmAGO2, whereas a shift was found for BmTudor-sn. In addition, in BmTudor-sn knockdown Bm12 cells, we observed that cell viability increased under DA treatment. Furthermore, insect two-hybrid system results indicated that the key site involved in DA binding to BmTudor-sn was Leu704. In conclusion, in vivo and in vitro experimental evidence indicated that BmTudor-sn is a binding protein of DA in silkworm Bm12 cells at the 100 μM level, and the key site of this interaction is Leu704. Our results provide new perspectives to aid in elucidating the molecular mechanism of action of DA in insects and developing new biopesticide.

**Keywords:** Destruxin A; *Bombyx mori*; binding protein; BmTudor-sn; Bm12 cell

**Key Contribution:** BmTudor-sn is a Binding Protein of Destruxin A in Silkworm Bm12 cells by in vivo and in vitro experimental evidences, and the key site of this interaction is Leu704.

#### **1. Introduction**

Destruxins are cyclodepsipeptidic mycotoxins, and there are 39 analogues [1]. Destruxin A (DA, Figure 1A), the common analogue secreted by the entomopathogenic fungus *Metarhizium anisopliae*, has a strong insecticidal effect and anti-immunity activity, which includes breaking the balance between calcium and hydrogen ions and subsequently affecting the function of phagocytosis and encapsulation in hemocyte [2,3]. Because of the immunosuppression activity of DA in insects, the majority of studies have focused on the mechanism of action of the effect of DA on the immune-related pathway or stress reaction (response), such as changes to the transcriptome [4] and proteome [5] or immune regulation by microRNA [6] in *Plutella xylostella*. There have been various studies on the effects of DA, including its impact on transcriptome [7], proteome, transcription factor, and antibacterial peptide expression [8] in *Bombyx mori* and its influence on the Toll or Imd pathway in *Bemisia tabaci* [9]. However, for drug research or development, it is important to look for direct targets, most of which are proteins, in

appropriate tissues or cells. In order to clarify the molecular mechanism more directly, previously, we screened and isolated DA-binding proteins in ovary-derived Bm12 cells using a label-free small molecule drug target identification method called drug affinity responsive target stability (DARTS) [10]. The DARTS results indicated that DA greatly induced heat shock proteins (HSPs) in cultured cells, and we successfully demonstrated that DA binds to an HSP in vitro by molecular interaction validation [11]. However, evidence from only one approach is not sufficient, especially for studies only performed in vitro.

Here, we selected three proteins, BmTudor-sn, BmPiwi, and BmAGO2, from the DARTS analysis based on their molecular docking score and determined whether these proteins are DA binding proteins through a series of in vivo and in vitro experiments. In brief, BmPiwi and BmAGO2 belong to the Argonaute family [12], which plays a vital role in germ cell development and represent a core component of the RNA interference (RNAi) pathway and function in transcriptional regulation [13]. Meanwhile, BmTudor-sn, a multifunctional protein containing four staphylococcal nuclease domains and a Tudor domain, participates in cellular pathways involved in gene regulation, cell growth, and development and interacts with Argonaute proteins [14], and it is known as stress granule protein in *Bombyx mori* [15]. These three proteins are all critical components in the RNAi pathway [16]. Notably, few studies have associated DA with germ line-derived proteins that are involved in several important physiological processes. In addition, this study may provide novel insights to exploit new targets and pathways for the development of pesticides.

**Figure 1.** Binding pose of Destruxin A (DA) with BmTudor-sn, BmPiwi, and BmAGO2. DA is colored in cyan, and the surrounding residues in the binding pockets are colored in orange. The backbone of the receptor is depicted as spectrum ribbon. (**A**) The structure of DA. (**B**) The binding poses of DA with BmTudor-sn. (**C**) The binding mode of DA with BmPiwi. (**D**) The binding patterns of DA with BmAGO2.

#### **2. Results**

#### *2.1. Molecular Docking*

Faced with hundreds of isolated proteins from the DARTS analysis, molecular docking was a convenient and efficient method to screen for proteins that are of research value. Ultimately, BmTudor-sn, BmPiwi, and BmAGO2 protein were selected from hundreds of candidates because of their relatively high binding scores and because they play crucial roles in many physiological processes in *Bombyx mori*.

Through molecular docking, we obtained the binding modes of DA with the BmTudor-sn, BmPiwi, and BmAGO2 proteins. The estimated binding free energies indicated by GBVI/WSA dG scoring are listed in Table 1. A lower binding free energy suggests a higher binding affinity. The ligands and DA had docking scores that ranged from −9.1947 to −11.002 kcal/mol, suggesting a good binding affinity with these proteins. The binding modes of DA with BmTudor-sn, BmPiwi, and BmAGO2 are depicted in Figure 1. DA fits the pocket well in terms of the shape in the binding sites. For the BmTudor-sn docking pattern, the Arg343 side chain of the SN3 domain could have a hydrogen bond interaction with the carbonyl group of DA, which contains a number of carbonyl groups that might form hydrogen bonds with nearby basic amino acids. For the BmPiwi binding mode, the carbonyl in DA, which is regarded as a hydrogen bond donor, forms one hydrogen bond with the side chain of Cys648 in BmPiwi, which forms other hydrogen bonds with Gln695 and Gly651. At the bottom of the binding pocket, hydrogen bonds are formed between Cys614 and several residues around it. For the binding mode of DA with BmAGO2, one oxygen atom of the carbonyl near the pyrrolidine of DA, which is regarded as a hydrogen bond donor, forms one hydrogen bond with the side chain of Arg981 in BmAGO2 and forms another intramolecular hydrogen bond with a nitrogen atom. The other oxygen atom of the carbonyl near the pyrrolidine of DA, which is regarded as a hydrogen bond donor, forms one hydrogen bond with the side chain of Lys761 in BmAGO2.


**Table 1.** The docking score of DA against BmTudor-sn, BmPiwi, and BmAGO2.

*2.2. Assessing the Interaction of DA with BmTudor-sn, BmPiwi, and BmAGO2 by Bio-Layer Interferometry (BLI) In Vitro*

The determination of the affinity constant between heterologously expressed recombinant proteins with a small molecule through in vitro biophysical approaches is a generally accepted and frequently used method for target validation. In this study, we selected bio-layer interferometry [17] because it is label free, has high sensitivity, and can be performed in real-time to assess the binding affinity. Recombinant proteins were expressed and purified from eukaryotic expression in lepidoptera *Spodoptera frugiperda* 9 (Sf9) cells because of their similar tertiary structure with native proteins. The BLI analysis results indicated that DA interacts with BmTudor-sn at concentrations of 125 μm, 250 μm, and 500 <sup>μ</sup>m but not BmPiwi or BmAGO2 (Figure 2A,B). And the affinity constant KD is 5.87 × <sup>10</sup><sup>−</sup>4M. These results possibly contradict the above docking results, which indicated that DA formed hydrogen bonds with the three proteins.

**Figure 2.** Results of recombinant proteins interacting with DA using BLI. (**A**) Data analysis with software showing there are interactions between BmTudor-sn and DA. (**B**) Molecular interaction kinetic data of BmTudor-sn with DA. (**C**,**D**) Processed data indicated no interaction of BmPiwi or BmAGO2 with DA.

#### *2.3. Assessing the Interaction of DA with BmTudor-sn through Protein Stability and RNAi In Vivo*

DARTS is based on the principle that protein folding stability is more resistant to protease and heat treatments under conformational modifications caused by small molecules or other ligands. Notably, false-positive DARTS results will be obtained with the high expression of proteins caused by drug treatment and the stress response thus, it was necessary to confirm that BmTudor-sn was not completely hydrolyzed by proteolysis. The immunoblotting results (Figure 3A) clearly indicated that BmTudor-sn, but not BmPiwi and BmAGO2, was stable with DA treatment in a dosage-dependent manner under proteolysis, which suggested that BmTudor-sn is a DA binding protein. Additional evidence was obtained using a method that directly monitored target engagement based on the shift in protein thermal stability induced by a small molecule, which is termed the cellular thermal shift assay (CETSA) [18]. As depicted in Figure 3B, the thermal stability of BmTudor-sn increased following DA binding modification, as indicated by the binding protein solubility in the supernatant being positively correlated with the temperature gradient. These protein stability shift assay results provided sufficient evidence to demonstrate that BmTudor-sn is a binding protein of DA in Bm12 cells. Then, cytotoxicity and viability assays were performed following DA treatment with BmTudor-sn knockdown, and the results revealed that viability increased by approximately 20% (Figure 3C).

**Figure 3.** Protein stability shift assay by immunoblotting and CETSA results indicated that DA specifically binds to BmTudor-sn but not BmPiwi or BmAGO2. (**A**) Immunoblotting results demonstrating that BmTudor-sn stability was influenced by interacting with DA in a dosage-dependent manner. (**B**) CETSA showing that BmTudor-sn thermal stability was subject to thermal gradient treatment after adding DA. (**C**) BmTudor-sn knock down cells showed higher viability under DA treatment. Significant differences between columns are indicated by an \* (*p* < 0.05) according to *t* test.

#### *2.4. Screening Key Amino Acid Sites of Interaction between DA and BmTudor-sn Using an Insect Two-Hybrid (I2H) System*

The above experiments assessed and verified the interaction of DA with BmTudor-sn. However, the exact interaction sites were not clear. Here, we evaluated the binding site in vivo using the insect two-hybrid (I2H) protein-protein interaction method [19] (Figure 4). This method was used to assess key sites where DA disrupted the interaction of BmTudor-sn with BmAGO1 in the Spodoptera frugiperda 9 (Sf9) cell line. Six mutants were prepared according to optimal docking sites, and 0.02 and 0.2 μg/mL DA treatments were selected based on our previous results. The results clearly indicated that for the 0.02 and 0.2 μg/mL treatments, the signals for 2 mutants, Ser707Ala and leu704Ala, and 3 mutants, Ser701Ala, Leu704Ala, and Tyr708Ala, respectively, were each different from the wild type signal, which suggested that the binding mode is dependent on the DA dosage. Apparently, DA interacts with BmTudor-sn at both concentrations at Leu704. At the protein domain level, the differences in the mutants were both in the Tudor domain (Ser707, Leu704, Ser701, Tyr708) rather than the SN3 (Lys492) and SN4 (Lys582) domains. In addition, DA impeded the protein pair interaction at Ser707 and Tyr708 and promoted it at Leu704 and Ser701 based on the increase or decrease in the signal, respectively.

**Figure 4.** Key amino acid sites of interaction between DA and BmTudor-sn. (**A**) Schematic of the principle of the insect two-hybrid (I2H) system [19]. (**B**) Differences in the mutants with the 0.02 and 0.2 μg/mL DA treatments. Significant differences between columns are indicated by an \* (*p* < 0.05) according to DMRT. (**C**) Sketch of the domain structure of BmTudor-sn and the key amino acid sites of the Tudor domain. The text continues here.

#### **3. Discussion**

In drug research and development, it is important to clearly determine the mechanism of action of small molecules in cells, the most common target of which are binding proteins [20]. Previously, we attempted to elucidate the mechanism for DA using different approaches, such as analyzing phenotype differences and changes in the transcriptome<sup>7</sup> and proteome [21]. Unfortunately, because DA has no active group, it is not possible to bond affinity chromatography tags for use in chemical proteomics. With DARTS, we have the ability to bond DA with potential proteins. However, DARTS can return false-positive results when highly expressed proteins are not digested completely by protease [22]. Indeed, BmTudor-sn, BmPiwi, and BmAGO2 proteins were upregulated by DA 2-3-fold at the transcriptional level. Immunoblotting is typically used for verification [23], and the dosage-dependent blot protein is the binding protein. Based on this, we confirmed that BmTudor-sn is a binding protein of DA but not BmPiwi or BmAGO2.

DARTS analysis revealed dozens of potential binding proteins, and it was necessary to determine which were the most probable. Molecular docking is a reasonable, efficient, and convenient strategy to narrow down candidates among all proteins compared to other methods such as heterologous expression and kinetic analysis, which require verification one by one. However, it seems paradoxical in this study that proteins with the top docking scores were not the binding protein [24]. Actually, docking is a completely theoretical speculation, and other experiments are required to draw a conclusion. Here, BLI kinetic analysis in vitro and CETSA in vivo were performed to address these drawbacks, demonstrating that DA binds to BmTudor-sn rather than BmPiwi or BmAGO2. BLI is a recently developed, frequently used bio-molecular interaction analysis method based on optical interference signals and is a fast, real-time method that uses a small amount of sample. Meanwhile, CETSA is a small molecule target engagement strategy based on protein thermal stability shifts caused by ligand binding, and it has been successfully used to identify several target proteins in drug studies in recent years [25]. Moreover, in the BLI analysis, we found that DA bound to BmTudor-sn at the 100 μM level, and this result agreed with the DARTS experiment in which BmTudor-sn only appeared

in the 200 μg/mL treatment with a specific proteinase and not at the lower dosage and with other proteinases. Furthermore, the CETSA results demonstrated that BmTudor-sn showed thermal stability with the 200 μg/mL treatment and not at the lower dose. I2H was first used to study protein-protein interactions20, and with it, we screened the key amino acid sites of the protein-ligand complex using different mutants because of its convenience and sensitivity and because a purified protein was not required. I2H was also practical for comparing the drug inhibition rates of protein pairs.

DA has been previously reported to play a role in immune related pathways and in inducing the immunosuppression phenotype in insects [26]. The majority of studies have focused on these roles, while few have addressed the mechanism of action in cells. Therefore, previously, we assessed the ion concentration inside and outside the cell [2] and found that DA strongly induced the expression of heat shock proteins (HSPs) in cells and bound to one of them in vitro [11], and here, we showed that BmTudor-sn is a DA binding protein. Tudor-sn, a multifunctional protein, was reported to regulate downstream expression and slice RNAi under normal conditions and participate in the formation of stress granules and processing bodies under stress [15]. This may also explains why BmTudor-sn only appeared in the high concentration treatment in DARTS. Another question is whether DA binds to Tudor-sn in humans. DA is one of fungal toxins generated by the commercial biopesticide *M. anisopliae*, which also produces Destruxin B (DB), and DB was reported to have clear anti-cancer activity [27]. Interestingly, we previously showed that DB is more sensitive to V-ATPase than DA in silkworm hemocytes, and this ion channel is known to be the most probable target of DA. However, DA likely does not bind to Tudor-sn in humans because Tudor-sn has different functions in different species.

In conclusion, in vivo and in vitro experimental evidence indicated that BmTudor-sn is a binding protein of DA in silkworm Bm12 cells at the 100 μM level, and the key amino acid site of this interaction is Leu704.

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

#### *4.1. Cell Lines and Culture*

The silkworm Bm12 cell line was donated by Professor Cao Yang (College of Animal Science at South China Agricultural University) and cultured in TNM-FH culture medium (Hyclone, Pittsburgh, MA, USA) and 10% fetal bovine serum (Gibco, Waltham, MA, USA). The *Spodoptera frugiperda* 9 (Sf9) cell line was cultured in SFX culture medium (Hyclone) with 5% fetal bovine serum. Cells were cultured at 27 ◦C and maintained at over a period of 2–4 days. Cells in the logarithmic phase were used for the experiment.

#### *4.2. Destruxin A and Treatment*

Destruxin A (DA) was isolated and purified from the *Metarhizium anisopliae* var. anisopliae strain MaQ10 in our laboratory [28]. A DA stock solution of 10,000 μg/mL was made up from 1 mg of DA and 100 μL of dimethyl sulfoxide (DMSO, Sigma-Aldrich, Darmstadt, Germany). To begin treatment, the DA stock solution was added to a cell well at a final concentration of 200 μg/mL DA. The control group was only supplemented with 0.1% DMSO.

#### *4.3. Homology Modeling and Molecular Docking*

Homology modeling was conducted in MOE v2015.1001. The structure was determined by homology modeling of the target sequence. Template crystal structures were identified using NCBI BLAST and downloaded from the RCSB Protein Data Bank. The Protonate module of MOE v2015.1001 was used to calculate the protonation state at pH = 7. Ten independent intermediate models were built. These different homology models were obtained from the mutational selection of different loop candidates and side chain rotamers. Then, the intermediate model that scored the highest according to the GB/VI scoring function was chosen as the final model and subjected to further energy minimization using the AMBER12: EHT force field.

MOE Dock was used for molecular docking simulations. The 2D structure of the ligand was drawn in ChemBioDraw 2014 and converted to 3D in MOE v2015.1001 through energy minimization with the MMFF94x force field. The protein structures were constructed by homology modeling. The Site Finder module in MOE was used to predict the potential binding pockets. Then, the protonation state of the protein and the orientation of the hydrogens were optimized using LigX at a pH of 7 and temperature of 300 K. Prior to docking, the force field of AMBER12: EHT and the implicit solvation model of Reaction Field (R-field) were selected. The docking workflow followed the "induced fit" protocol, in which the side chains of the receptor pocket were allowed to move according to ligand conformations, with a constraint on their positions. The weight used for tethering side chain atoms to their original positions was 10. For each ligand, all docked poses were ranked by London dG scoring first, and a force field refinement was then carried out on the top 30 poses followed by a rescoring of GBVI/WSA dG.

#### *4.4. Bio-Layer Interferometry (BLI)*

All proteins were prepared by eukaryotic expression in the Sf9 cell line. These were tagged with His-tag and purified by nickel affinity chromatography. BLI analysis was performed on a ForteBio OctetQK System (K2, Pall Fortebio Corp, Menlo Park, CA, USA). Generally, the protein samples were coupled with a biosensor for immobilization. Serial not gradient dilutions of DA (500, 250, 125, 62.5, 31.25, 15.63, and 7.813 μM) were used for treatment. PBST buffer (0.05% Tween20, 5% DMSO) was used for the reference and dilution buffers. The working procedure was baseline for 60 s, association for 60 s, and dissociation for 60 s. Finally, the raw data were processed with Data Analysis Software (9.0, Pall Fortebio Corp, Menlo Park, CA, USA).

#### *4.5. Immunoblot and Cellular Thermal Shift Assay (CETSA)*

The Bm12 cell line was used to conduct the CETSA and immunoblot experiments. Cells were treated with DA. For immunoblotting, cell extracts from different DA treatments, collected after incubation with RIPA lysis buffer, were digested with proteinase K and then heated in water at 90 ◦C for 5 min. Samples were analyzed using SDS-PAGE and transferred to PVDF membranes. The membranes were then incubated in skim milk powder, followed by incubation with primary and HRP antibodies. ECL was added to the chemiluminescence reaction. For CETSA, Bm12 cells, treated with 200 μg/mL DA, were divided into 8 aliquots, heated at 37–58 ◦C, and lysed by a freeze-thaw cycle. The supernatants of the lysed cells were used for the western blot analysis as described above.

#### *4.6. RNAi and Viability and Toxicity Assessment*

SiRNAs were prepared by synthesis in vitro. The sequence of *BmTudor-sn* siRNA is 5 -CCAAAGGACCGCCAACAAUTT-3 and 5 -AUUGUUGGCGGUCCUUUGGTT-3 . SiRNA and FuGENE transfection reagent were each diluted in serum-free medium and then mixed. The mixture was added to Bm12 cells after the DA treatment. Viability and toxicity assessment were performed following the manufacturer's instructions.

#### *4.7. Insect Two Hybrid (I2H) System*

BmTudor-sn and mutants were cloned into the I2H vector pIE-AD, and BmAGO1 was cloned into the I2H vector pIE-DBD using the Gateway system. These vectors and the luciferase vector were co-transfected into the Sf9 cell line and treated with DA at 0.02 and 0.2 μg/mL. The luciferase activities in the cell extracts were determined using a Luciferase Reporter Assay System (Promega, Beijing, China) and Synergy™ H1 (BioTek, Winooski, VT, USA).

**Author Contributions:** J.W. designed and completed the experiments and wrote the paper. W.H. completed the isolation and characterization of proteins. Q.H. conceived and designed the experiments and revised the paper.

**Funding:** This research is supported by National Natural Science Foundation of China (31772184 and 31272057) and Guangdong Province Science and Technology Project (2016B020234005).

**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/).

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

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

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

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