Mycotoxins: Biotransformation and Bioavailability Assessment Using Caco-2 Cell Monolayer

The determination of mycotoxins content in food is not sufficient for the prediction of their potential in vivo cytotoxicity because it does not reflect their bioavailability and mutual interactions within complex matrices, which may significantly alter the toxic effects. Moreover, many mycotoxins undergo biotransformation and metabolization during the intestinal absorption process. Biotransformation is predominantly the conversion of mycotoxins meditated by cytochrome P450 and other enzymes. This should transform the toxins to nontoxic metabolites but it may possibly result in unexpectedly high toxicity. Therefore, the verification of biotransformation and bioavailability provides valuable information to correctly interpret occurrence data and biomonitoring results. Among all of the methods available, the in vitro models using monolayer formed by epithelial cells from the human colon (Caco-2 cell) have been extensively used for evaluating the permeability, bioavailability, intestinal transport, and metabolism of toxic and biologically active compounds. Here, the strengths and limitations of both in vivo and in vitro techniques used to determine bioavailability are reviewed, along with current detailed data about biotransformation of mycotoxins. Furthermore, the molecular mechanism of mycotoxin effects is also discussed regarding the disorder of intestinal barrier integrity induced by mycotoxins.

In addition, Penicillium species are known to produce mycophenolic acid (MPA) [9] and patulin (PAT) [10,11]. Aflatoxins, including aflatoxin B1 (AFB1), aflatoxin B2 (AFB2), aflatoxin M1 (AFM1), aflatoxin M2 (AFM2), aflatoxin G1 (AFG1), and aflatoxin G2 (AFG2), are the most studied group of mycotoxins produced by Aspergillus flavus [12,13]. Alternaria fungi contaminate a wide variety of food items, such as cereals, fruits, wheat, barley, and sorghum, producing several toxins, with alternariol  Ingestion of contaminated food is considered as a major route for exposure to many mycotoxins [51]. Upon ingestion, mycotoxins may induce local toxicity or cross the intestinal barrier to enter the bloodstream and reach target organs [2]. Nevertheless, to achieve any effect in a specific tissue or organ, the mycotoxins must be available in effective concentration at certain location, which refers to the compound's tendency to be extracted from the food matrix, and they must then be absorbed from the gut via the intestinal cells [67]. The term bioaccessibility refers to the fraction of a mycotoxin liberated from a food matrix that passes unmodified through complex biochemical reactions related to the gastrointestinal digestion and thus becomes available for absorption in the small intestine [68,69]. Bioaccessibility can be considered as an indicator for the maximal absorption of the toxin, which can be used for realistic worst-case risk assessment of the toxin in a consumer product [70]. In fact, foodborne mycotoxins can be degraded or modified by metabolic processes of the human body, and only a fraction of the initial content can pass the intestinal membrane to enter the bloodstream [71]. In this sense, bioavailability is defined as the portion of ingested contaminant in food that reaches the systemic circulation [72].
To determine the bioavailability of mycotoxins, different in vitro models or in vivo experiments have been carried out. In vivo experiments would be the best way to evaluate the efficacy of binding capacities [73]. However, to avoid the ethically questionable use of animals in the in vivo experiments, the in vitro models have been used instead. The bioavailability studies carried out in animals are complex, expensive, and lengthy, while the in vitro experiments can be simple, rapid, and cost-effective [72]. The advantages and disadvantages of each procedure are summarized in Table 3. Most of the in vitro studies of the gut were done with human colon tumorigenic cell lines Caco-2, T84, TC7, and HT-29 [74]. The brief description of the expression of transporters, enzymes, and other relevant proteins of available cell lines used for the in vitro biotransformation and bioavailability of drugs and xenobiotics is stated in Table 4. Among commercially available cell lines, Caco-2 cells have been widely used to study absorption, metabolism, and bioavailability of drugs and xenobiotics [2,74]. This model is generally suitable for screening drug and nutrient compounds due to a good in vitro-in vivo correlation [75]. Table 3. Advantages and disadvantages of in vivo and in vitro models in the evaluation of bioavailability.

Models Advantages Disadvantages
In vitro models

Simulation of gastrointestinal transformation
Similar to the physiological processes in the human body Suitable for high-throughput format Ability of testing a specific mechanisms of action Focus on small number of components Validation with reference material No hormonal and nervous control Lack of feedback mechanisms Absence of mucosal cell activity Deficiency of complexity of peristaltic movements, and involvement of the local immune system Homeostatic mechanisms are not present Difficult to achieve the anaerobic assay conditions  IPEC-J2 Neonatal pig small intestine CYP1A1, 1A2, 3A29 P-gp, MRP1, BCRP [86,87] P-glycoprotein (P-gp), multidrug resistance protein (MRP), breast cancer resistance protein (BCRP), uridinediphosphoglucuronosyl transferase (UGT), sulfotransferase (SULT), N-acetyltransferase (NAT), glutathione-S-transferase (GST), and cytochrome P (CYP).
This review mainly focuses on the biotransformation of mycotoxins via the expression regulation of some critical enzymes and the currently available data regarding the in vitro study of the bioavailability of mycotoxins using the Caco-2 monolayer. Furthermore, the usefulness and limitations of this model are also discussed.

Biotransformation of Mycotoxins
Mycotoxins biotransformation is defined as all the complex modifications which alter the structure of mycotoxins by chemical reactions within the body [88]. Biotransformation is often referred to detoxification, but biotransformation enzymes can also convert certain chemicals into highly toxic Toxins 2020, 12, 628 6 of 36 metabolites ( Figure 1) in a process known as bioactivation [89]. Biotransformation of mycotoxins involves two distinct stages, namely phase I and phase II. The biotransformation process allows metabolites created during phase I to enter conjugation processes (phase II), but in some cases, the substances may be eliminated directly after phase I [90]. In phase I, the mycotoxin could be oxidized, reduced, or hydrolyzed based on their chemical structure [90]. The enzymes involved in detoxification belong to the cytochrome P (CYP) superfamily. The CYP superfamily contains the enzymes involved in oxidative metabolism, such as monooxygenases, prostaglandin synthases, amine oxidases and alcohol dehydrogenases; and reductive metabolism mainly governed by epoxide hydrolases, and aldehyde or ketone reductases [91]. CYP450 enzymes play an important role in the oxidative and reductive metabolism of many endogenous or exogenous chemical compounds [34], including most mycotoxins (Table 5). In mammals, CYPs are present in the endoplasmic reticulum and mitochondria of most cells [89]. Among CYPs, CYP3A with an average content from 50-70% of total enteric CYPs is the major subfamily expressed in the human small intestine [92].

Mycotoxins
Remaining mycotoxins In phase I, the mycotoxin could be oxidized, reduced, or hydrolyzed based on their chemical structure [90]. The enzymes involved in detoxification belong to the cytochrome P (CYP) superfamily. The CYP superfamily contains the enzymes involved in oxidative metabolism, such as monooxygenases, prostaglandin synthases, amine oxidases and alcohol dehydrogenases; and reductive metabolism mainly governed by epoxide hydrolases, and aldehyde or ketone reductases [91]. CYP450 enzymes play an important role in the oxidative and reductive metabolism of many endogenous or exogenous chemical compounds [34], including most mycotoxins (Table 5). In mammals, CYPs are present in the endoplasmic reticulum and mitochondria of most cells [89]. Among CYPs, CYP3A with an average content from 50-70% of total enteric CYPs is the major subfamily expressed in the human small intestine [92].  NEO, 3 -OH-T-2,  3 -OH-HT-2, T-2 triol, T-2 tetraol, and some C12,13-deepoxy products  Phase II reactions are known as conjugation reactions, which usually refer to covalent binding of endogenous hydrophilic substances such as glucuronic acid and sulfate. The reactions provide more hydrophilic compounds, which are quickly eliminated. In general, phase II reactions decrease the toxicity [89]. Uridine 5 -diphospho-glucuronosyltransferase (UDP-glucuronosyltransferase-UGT) and glutathione S-transferase (GST) enzymes play an important role in the phase II metabolism [89,91].
Although the liver is the main detoxification organ, extrahepatic tissues in the gastrointestinal tract (GI tract), kidney, and bladder also show metabolic activity. The GI tract is a first physical barrier for mycotoxins but it also influences the biotransformation process and bioavailability of mycotoxins in other ways. Microorganisms from guts have been reported to exhibit the capacity for degrading mycotoxins [131][132][133][134]. Additionally, P-glycoprotein (P-gp) and multidrug resistance protein (MRP), members of the ATP-binding cassette (ABC) superfamily of transport proteins, are able to pump mycotoxins out of the intestinal cells, leading to limit bioavailability of the substrates [71,135]. Both CYP450 and P-gp in the gut play a crucial role in defense mechanisms against mycotoxins that reach the intestinal mucosa [92].
Previous biotransformation studies mainly focused on AFB1, OTA, trichothecenes (T-2 and DON), ZEA, and FBs. Recently, emerging Fusarium and Alternaria mycotoxins have gained more interest [46], although in vivo metabolization data are still limited. The biotransformation products of mycotoxins are summarized in Table 3. These studies revealed that mycotoxins can induce the expression of CYP450 enzymes in animal and human cell lines.

Biotransformation of Ochratoxin A
In animals and humans, OTA can be metabolized by both phase I and phase II enzymes to many different products in the liver, kidney, and intestine ( Figure 3). Poor biotransformation and slow elimination of metabolites contribute to the toxicity, carcinogenicity, and organ specificity of OTA [139,141]. In the gut, ochratoxin α (OTα), a major metabolite and is formed by carboxypeptidases, which cleave the peptide bond in OTA [34]. Other types of major metabolites of OTA are 4-hydroxyochratoxin A (4-OH-OTA) and 10-hydroxyochratoxin A (10-OH-OTA) have been identified from the urine of rats and are also produced by human, pigs, goat, chicken, rat, and rabbit liver microsomes or human bronchial epithelial cells in vitro [142][143][144]. Most of the metabolites of OTA, such as OTα, OTB, 4-OH-OTA, and 10-OH-OTA, are less toxic than the original compound [129,139]. However, opening the lactone ring under alkaline conditions (called the lactone-opened OTA), found in rodents, leads to more toxic metabolites than OTA itself [126]. These phase I-type reactions probably relate to the action of the CYP450 enzyme family, including CYP1A1, CYP1A2, CYP3A1, CYP3A2, CYP3A4, CYP3A5, CYP2B6, and CYP2C9 [124][125][126]. Phase II biotransformation mainly occurs in the liver with conjugation of OTA with sulfate, glucuronide, hexose/pentose, and glutathione [127][128][129].

Biotransformation of Ochratoxin A
In animals and humans, OTA can be metabolized by both phase I and phase II enzymes to many different products in the liver, kidney, and intestine ( Figure 3). Poor biotransformation and slow elimination of metabolites contribute to the toxicity, carcinogenicity, and organ specificity of OTA [139,141]. In the gut, ochratoxin α (OTα), a major metabolite and is formed by carboxypeptidases, which cleave the peptide bond in OTA [34]. Other types of major metabolites of OTA are 4-hydroxy-ochratoxin A (4-OH-OTA) and 10hydroxyochratoxin A (10-OH-OTA) have been identified from the urine of rats and are also produced by human, pigs, goat, chicken, rat, and rabbit liver microsomes or human bronchial epithelial cells in vitro [142][143][144]. Most of the metabolites of OTA, such as OTα, OTB, 4-OH-OTA, and 10-OH-OTA, are less toxic than the original compound [129,139]. However, opening the lactone ring under alkaline conditions (called the lactoneopened OTA), found in rodents, leads to more toxic metabolites than OTA itself [126]. These phase I-type  reactions probably relate to the action of the CYP450 enzyme family, including CYP1A1, CYP1A2, CYP3A1, CYP3A2, CYP3A4, CYP3A5, CYP2B6, and CYP2C9 [124][125][126]. Phase II biotransformation mainly occurs in the liver with conjugation of OTA with sulfate, glucuronide, hexose/pentose, and glutathione [127][128][129].
Other DON-biotransformation products, including DON-glutathione conjugates and the products of glutathione degradation, such as DON-S-cysteinyl-glycine and DON-S-cysteine, have been reported in cereals. Thanks to intestinal microflora, DON could be metabolized in animals and humans but not deposited in the tissues [151,152].
porcine, rat, chicken, bovine, and human [34,102,[148][149][150]. Other DON-biotransformation products, including DON-glutathione conjugates and the products of glutathione degradation, such as DON-S-cysteinyl-glycine and DON-S-cysteine, have been reported in cereals. Thanks to intestinal microflora, DON could be metabolized in animals and humans but not deposited in the tissues [151,152].
Toxins 2020, 12, x FOR PEER REVIEW 13 of 37 I enzyme, contributing to the rapid metabolism of T-2 to HT-2 [100]. A recent study revealed that cholic acid supplementation promotes the T-2 metabolism through activation of the farnesoid X receptor, which was found to have significantly increased the expression of CYP3A37 [99]. In phase II, glucuronidation of T-2 toxin, HT-2 toxin, and further phase I metabolites essentially contribute to the metabolism and excretion. The transformation of T-2 to T-2-3-glucuronide and HT-2 to HT-2-3-glucuronide and HT-2-4-glucuronide occurs in liver microsomes of rats, mice, pigs and humans [155]. The activities of GSTs and sulfotransferases can be also attributed to the conjugation reaction as a response to T-2 exposure [100,101].

Biotransformation of Fumonisins
After oral ingestion, FB1 are excreted primarily in the feces, either in the intact form or converted into aminopentol (HFB1) and partially hydrolyzed FB1 (pHFB1) by the intestinal microbiota ( Figure 6) [116]. The supplementation with fumonisin carboxylesterase FumD results in the gastrointestinal degradation of FB1 and is considered as an effective strategy to detoxify FB1 in the digestive tract of turkeys and pigs [156]. The findings of Daud et al. [157] provided evidence that human fecal microbiota are capable of FB1 degradation, and LC-MS/MS fragmentation patterns indicated microbial biotransformation to hydrolyzed and partially hydrolyzed FB1 [157]. FB1 is not metabolized by CYPs. Moreover, it is a selective inhibitor of CYP2C11 and

Biotransformation of Fumonisins
After oral ingestion, FB1 are excreted primarily in the feces, either in the intact form or converted into aminopentol (HFB1) and partially hydrolyzed FB1 (pHFB1) by the intestinal microbiota ( Figure 6) [116]. The supplementation with fumonisin carboxylesterase FumD results in the gastrointestinal degradation of FB1 and is considered as an effective strategy to detoxify FB1 in the digestive tract of turkeys and pigs [156]. The findings of Daud et al. [157] provided evidence that human fecal microbiota are capable of FB1 degradation, and LC-MS/MS fragmentation patterns indicated microbial biotransformation to hydrolyzed and partially hydrolyzed FB1 [157]. FB1 is not metabolized by CYPs. Moreover, it is a selective inhibitor of CYP2C11 and CYP1A2, while the activities of CYP2A1:2A2, CYP2B1:2B2, CYP3A1:3A2, and CYP4A are not significantly affected. The significant inhibition of CYP2C11 might be related to suppressed protein kinase activity as a result of the inhibition of sphingolipid biosynthesis caused by FB1 [158][159][160]. FB1, HFB1, and pHFB1 can be acetylated to form N-acetylated fumonisins with fatty acid of various lengths, and N-acyl forms proved to be more toxic than the parent FB1 [161][162][163].
CYP1A2, while the activities of CYP2A1:2A2, CYP2B1:2B2, CYP3A1:3A2, and CYP4A are not significantly affected. The significant inhibition of CYP2C11 might be related to suppressed protein kinase activity as a result of the inhibition of sphingolipid biosynthesis caused by FB1 [158][159][160]. FB1, HFB1, and pHFB1 can be acetylated to form N-acetylated fumonisins with fatty acid of various lengths, and N-acyl forms proved to be more toxic than the parent FB1 [161][162][163].

Biotransformation of Zearalenone
ZEA is mainly biotransformed into α-zearalenol (α-ZEA), which shows the highest binding affinity to human and porcine estrogen receptors, whereas in broilers and rats, β-zearalenol (β-ZEA) with the low affinity to the receptor is predominantly produced [103][104][105]. ZEA upregulates mainly mRNA levels of CYP2B6, CYP3A4, CYP1A2 and CYP1A1, followed by CYP3A5 and CYP2C9, together with activation of their transcriptional regulators-aryl the hydrocarbon receptor (AhR), constitutive androstane receptor (CAR), and pregnane X receptor (PXR) [106]. It is well known that ZEA, α-ZEA, and β-ZEA are substrates of UGT, the enzyme responsible for the glucuronidation (Figure 7) [78,105,164,165]. However, the UGT was not only saturated but also inhibited by high concentration of ZEA [166]. Although zearalenone-14-glucoside (ZEA14Glc) has lower toxicity than ZEA due to inability to interact with estrogen receptors, the possible systemic hydrolysis and further activating metabolism of ZEA14Glc leads to ZEA-mediated toxicity [167]. Due to the adverse effect of ZEA on human and animal health, microorganisms have gained great interest in the modulation of ZEA adsorption and transformation [168,169]. Eukaryotic cells were able to biotransform ZEA to α-ZEA and β-ZEA, while prokaryotic cells only absorbed ZEA without any metabolization of this mycotoxin and sequestered ZEA by binding to the cell wall [170,171].

Biotransformation of Zearalenone
ZEA is mainly biotransformed into α-zearalenol (α-ZEA), which shows the highest binding affinity to human and porcine estrogen receptors, whereas in broilers and rats, β-zearalenol (β-ZEA) with the low affinity to the receptor is predominantly produced [103][104][105]. ZEA upregulates mainly mRNA levels of CYP2B6, CYP3A4, CYP1A2 and CYP1A1, followed by CYP3A5 and CYP2C9, together with activation of their transcriptional regulators-aryl the hydrocarbon receptor (AhR), constitutive androstane receptor (CAR), and pregnane X receptor (PXR) [106]. It is well known that ZEA, α-ZEA, and β-ZEA are substrates of UGT, the enzyme responsible for the glucuronidation (Figure 7) [78,105,164,165]. However, the UGT was not only saturated but also inhibited by high concentration of ZEA [166]. Although zearalenone-14-glucoside (ZEA14Glc) has lower toxicity than ZEA due to inability to interact with estrogen receptors, the possible systemic hydrolysis and further activating metabolism of ZEA14Glc leads to ZEA-mediated toxicity [167]. Due to the adverse effect of ZEA on human and animal health, microorganisms have gained great interest in the modulation of ZEA adsorption and transformation [168,169]. Eukaryotic cells were able to biotransform ZEA to α-ZEA and β-ZEA, while prokaryotic cells only absorbed ZEA without any metabolization of this mycotoxin and sequestered ZEA by binding to the cell wall [170,171].

Biotransformation of Enniatins
For ENNs, the most information is currently available for ENN B and B1. In vitro and in vivo studies demonstrated that CYP3A4, CYP2C19, and CYP1A2 play the major role for ENN B metabolism in human microsome,s and CYP3A and CYP1A are also included in this process in rats and dogs [109]. The 12 biotransformation products were characterized after the incubation of ENN B with rat, dog, and human liver microsomes ( Figure 8): M1-M5 were monohydroxylated and M6 and M7 were N-demethylated, whereas M8-M12 were the result of multiple oxidations [110]. However, only eight metabolites could be detected in the case of chicken liver microsomes, particularly five hydroxylated (M1-M5) and three carboxylated (M9, M11 and M12) metabolites. Moreover, M4 and M13 were major metabolites in egg samples, while M11 and M13 were found in liver and serum samples collected after broilers and hens were given contaminated feed containing ENN B [111]. Similarly, ENN B1 is mainly metabolized by CYP3A4 [112]. In vitro incubation with minipig and slaughter swine liver microsomes resulted in the detection of ten ENN B1 metabolites (M2-M11) and M1 occurred only in the minipig assays, while six metabolites (M5-M8) were detected also in vivo [113]. Rumen microbiota also proved to be able to degrade ENN B up to 72% after 48 h of incubation [114]. Any sulfated or glucuronidated phase II metabolites of ENN B or ENN B1 were detected (Figure 9) [115].

Biotransformation of Enniatins
For ENNs, the most information is currently available for ENN B and B1. In vitro and in vivo studies demonstrated that CYP3A4, CYP2C19, and CYP1A2 play the major role for ENN B metabolism in human microsomes and CYP3A and CYP1A are also included in this process in rats and dogs [109]. The 12 biotransformation products were characterized after the incubation of ENN B with rat, dog, and human liver microsomes ( Figure 8): M1-M5 were monohydroxylated and M6 and M7 were N-demethylated, whereas M8-M12 were the result of multiple oxidations [110]. However, only eight metabolites could be detected in the case of chicken liver microsomes, particularly five hydroxylated (M1-M5) and three carboxylated (M9, M11 and M12) metabolites. Moreover, M4 and M13 were major metabolites in egg samples, while M11 and M13 were found in liver and serum samples collected after broilers and hens were given contaminated feed containing ENN B [111]. Similarly, ENN B1 is mainly metabolized by CYP3A4 [112]. In vitro incubation with minipig and slaughter swine liver microsomes resulted in the detection of ten ENN B1 metabolites (M2-M11) and M1 occurred only in the minipig assays, while six metabolites (M5-M8) were detected also in vivo [113]. Rumen microbiota also proved to be able to degrade ENN B up to 72% after 48 h of incubation [114]. Any sulfated or glucuronidated phase II metabolites of ENN B or ENN B1 were detected (Figure 9) [115].

Biotransformation of Beauvericin
Very few studies have been carried out on BEA ( Figure 10) in this regard. No BEA metabolites were detected in the mice feed with BEA in the study of Rodríguez-Carrasco et al. [107], suggesting a higher metabolic stability for BEA [107]. Mei et al. [108] reported that BEA is a potent inhibitor of diverse CYP450 enzymes, including CYP3A4/5 and CYP2C19 in human liver microsomes and CYP3A1/2 in rat liver microsomes [108].

Biotransformation of Beauvericin
Very few studies have been carried out on BEA ( Figure 10) in this regard. No BEA metabolites were detected in the mice feed with BEA in the study of Rodríguez-Carrasco et al. [107], suggesting a higher metabolic stability for BEA [107]. Mei et al. [108] reported that BEA is a potent inhibitor of diverse CYP450 enzymes, including CYP3A4/5 and CYP2C19 in human liver microsomes and CYP3A1/2 in rat liver microsomes [108].

Biotransformation of Beauvericin
Very few studies have been carried out on BEA ( Figure 10) in this regard. No BEA metabolites were detected in the mice feed with BEA in the study of Rodríguez-Carrasco et al. [107], suggesting a higher metabolic stability for BEA [107]. Mei et al. [108] reported that BEA is a potent inhibitor of diverse CYP450 enzymes, including CYP3A4/5 and CYP2C19 in human liver microsomes and CYP3A1/2 in rat liver microsomes [108].

Biotransformation of Beauvericin
Very few studies have been carried out on BEA ( Figure 10) in this regard. No BEA metabolites were detected in the mice feed with BEA in the study of Rodríguez-Carrasco et al. [107], suggesting a higher metabolic stability for BEA [107]. Mei et al. [108] reported that BEA is a potent inhibitor of diverse CYP450 enzymes, including CYP3A4/5 and CYP2C19 in human liver microsomes and CYP3A1/2 in rat liver microsomes [108].

Biotransformation of Alternaria Mycotoxins
AOH and AME form the metabolites hydroxylated at C-2, C-4, and C-8 by activation of the CYP1A1 enzyme ( Figure 11) [172,173]. AOH and AME activate the AhR pathway, which induces CYP1A1 expression [117,118]. AOH is known for its genotoxicity [118]. However, the phase I metabolites, 4-OH-AOH and 4-OH-AME, had minor effect compared to AOH or AME in topoisomerase inhibition and DNA strand-breaking effects [174]. Phase II metabolism includes conjugation with glucuronic acid and sulfate [119]. AME and AOH were enzymatically glycosylated using whole-cell biotransformation system, producing highly effective rates of 58% AOH-3-glucoside, 5% AOH-9-glucoside, and 24% AME-3-glucoside [120]. However, human gut microbiota was not capable of metabolizing AOH, AME, and ALT [175]. The conversion of ATX-II, significantly more genotoxic than AOH, to ATX-I by de-epoxidation in Caco-2 cells did not showed an adequate detoxification but an attenuation of genotoxicity [176]. The metabolic pathway of AOH, AME and other Alternaria mycotoxins, such as TEN, TeA, ALT and ATXs, are summarized in Figure 11.

Biotransformation of Alternaria Mycotoxins
AOH and AME form the metabolites hydroxylated at C-2, C-4, and C-8 by activation of the CYP1A1 enzyme ( Figure 11) [172,173]. AOH and AME activate the AhR pathway, which induces CYP1A1 expression [117,118]. AOH is known for its genotoxicity [118]. However, the phase I metabolites, 4-OH-AOH and 4-OH-AME, had minor effect compared to AOH or AME in topoisomerase inhibition and DNA strand-breaking effects [174]. Phase II metabolism includes conjugation with glucuronic acid and sulfate [119]. AME and AOH were enzymatically glycosylated using whole-cell biotransformation system, producing highly effective rates of 58% AOH-3-glucoside, 5% AOH-9-glucoside, and 24% AME-3-glucoside [120]. However, human gut microbiota was not capable of metabolizing AOH, AME, and ALT [175]. The conversion of ATX-II, significantly more genotoxic than AOH, to ATX-I by de-epoxidation in Caco-2 cells did not showed an adequate detoxification but an attenuation of genotoxicity [176]. The metabolic pathway of AOH, AME and other Alternaria mycotoxins, such as TEN, TeA, ALT and ATXs, are summarized in Figure 11.

Biotransformation of Patulin
PAT induces the upregulation of PXR and AhR accompanied by the enhancement of CYP1A1, CYP1A2, CYP2B6, CYP2C9, CYP 3A4, and CYP3A5 expression [130]. Moreover, PAT reacts with intracellular glutathione in gastrointestinal mucosa cells [182,183]. The extracellular enzymes of Lactobacillus casei YZU01 induced by PAT mainly degrades PAT, and the cell wall of this bacteria can also absorb a small amount of PAT [184]. Similarly, the degradation of PAT was observed by Saccharomyces cerevisiae during cider fermentation into E-ascladiol and Z-ascladiol (Figure 12), which are not toxic to human [185]. The biotransformation of PAT in humans and animals is not well understood and remains to be established.

Biotransformation of Patulin
PAT induces the upregulation of PXR and AhR accompanied by the enhancement of CYP1A1, CYP1A2, CYP2B6, CYP2C9, CYP 3A4, and CYP3A5 expression [130]. Moreover, PAT reacts with intracellular glutathione in gastrointestinal mucosa cells [182,183]. The extracellular enzymes of Lactobacillus casei YZU01 induced by PAT mainly degrades PAT, and the cell wall of this bacteria can also absorb a small amount of PAT [184]. Similarly, the degradation of PAT was observed by Saccharomyces cerevisiae during cider fermentation into E-ascladiol and Z-ascladiol (Figure 12), which are not toxic to human [185]. The biotransformation of PAT in humans and animals is not well understood and remains to be established.

Assessment of Bioavailability of Mycotoxins Using Caco-2 Cell Monolayer
The Caco-2 cell line is the most common and extensively used in vitro model to study the intestinal absorption of mycotoxins via the intestinal membrane enterocytes [2,10,187,188]. It was originally derived from a heterogeneous human epithelial colorectal adenocarcinoma cells established by Fogh and coworkers in 1977 [189]. The Caco-2 cells have the ability to spontaneously differentiate into a monolayer of cells, expressing many properties typical of absorptive enterocytes with a brush border layer, tight junctions, and efflux and uptake transporters as found in the small intestine [190][191][192]. Moreover, several phenolic compounds (e.g., kaempferol) are able to regulate the MAPK pathway, which is beneficial to the barrier functions [193]. Kaempferol treatment showed significant an increase in claudin 3, claudin 4, and occluden [194]. On the other hand, several mycotoxins-deoxynivalenol, zearalenone, fumonisin B1, T-2 toxin, aflatoxin M1, and ochratoxin A-have a deleterious effect on tight junctions of claudin 3, claudin 4, claudin 7, and occluden [195][196][197][198].
The Caco-2 cells have been shown to be a suitable model for biotransformation study because they express various phase-I hydroxylation and phase-II conjugation enzymes, and transport proteins of the ATP-Binding

Assessment of Bioavailability of Mycotoxins Using Caco-2 Cell Monolayer
The Caco-2 cell line is the most common and extensively used in vitro model to study the intestinal absorption of mycotoxins via the intestinal membrane enterocytes [2,10,187,188]. It was originally derived from a heterogeneous human epithelial colorectal adenocarcinoma cells established by Fogh and coworkers in 1977 [189]. The Caco-2 cells have the ability to spontaneously differentiate into a monolayer of cells, expressing many properties typical of absorptive enterocytes with a brush border layer, tight junctions, and efflux and uptake transporters as found in the small intestine [190][191][192]. Moreover, several phenolic compounds (e.g., kaempferol) are able to regulate the MAPK pathway, which is beneficial to the barrier functions [193]. Kaempferol treatment showed significant an increase in claudin 3, claudin 4, and occluden [194]. On the other hand, several mycotoxins-deoxynivalenol, zearalenone, fumonisin B1, T-2 toxin, aflatoxin M1, and ochratoxin A-have a deleterious effect on tight junctions of claudin 3, claudin 4, claudin 7, and occluden [195][196][197][198].
The Caco-2 cells have been shown to be a suitable model for biotransformation study because they express various phase-I hydroxylation and phase-II conjugation enzymes, and transport proteins of the ATP-Binding Cassette (ABC) superfamily [166]. Furthermore, a good correlation has been found for data on oral absorption in humans and the results in the Caco-2 model [199].
To closer mimic the intestinal barrier in vivo, Caco-2 cells were seeded on permeable membranes to form a confluent monolayer with a well-defined tight junction for approximately 21 days post-seeding [78]. The integrity of the Caco-2 monolayer was monitored by measuring the transepithelial electrical resistance (TEER), or by examining the permeability of paracellular markers, such as mannitol, inulin, Dextran, PEG 4000, Lucifer yellow, and phenol red [191,200]. Studies that have investigated the bioavailability of mycotoxins by Caco-2 cells are listed in Table 6. The results of these studies show that mycotoxins are transported through Caco-2 monolayer in different efficiencies. ZEA was substrates for ABCC1, ABCC2 and metabolites into αand β-zearalenol and glucuronides. DON, NIV, ZEA ENNs, and BEA cross easily the cell barrier. DON is efficiently transported through the intestinal barrier possibly either by passive/facilitated diffusion [202] or by paracellular passage through intercellular tight junctions [207]. All of the apparent permeability (P app ) values greater than 1 × 10 −6 cm/s suggest that these mycotoxins were absorbed efficiently [208]. P app values for DON have been reported by many researchers. Sergent et al. [207] reported an average P app value of 5.02 × 10 −6 cm/s for absorption (apical (AP)-basolateral (BL) compartment) and excretion (BL-AP direction) [207]. In other study, absorption and excretion P app values ranged 1.23-2.06 × 10 −6 and 2.68-2.8 × 10 −6 cm/s, respectively [202]. Finally, P app value of 3.3 × 10 −6 and 2.8 × 10 −6 cm/s for absorption and excretion, respectively, were determined in study of Kodota et al. [209]. A faster bidirectional transport of DON in the mixture comparing to pure DON was observed, suggesting that the presence of other mycotoxins including AFB1, FB1, and OTA may promote intestinal transport of DON [210]. For NIV, transcellular transport probably occurred by passive diffusion in the absorptive direction, and P app values were also higher than 10 −6 cm/s [203]. The P app values obtained with a concentration of 20 µM ZEA in the apical compartment and an incubation time of 1 h were 10.47 ± 4.7 × 10 −6 cm/s [211]. About 30% of initial ZEA crossed the cell monolayer after 3 h of exposure, and 40% of ZEA was absorbed by the intestinal after 8 h [78]. ZEA presented higher bioavailability than its metabolites, α-ZEA, ranging from 10% to 36% (0-4 h; 30 µM) [72]. Unlike DON-3-glucoside (neither absorbed or cleaved by Caco-2 cells), ZEA-14Glc and ZEA-16Glc could cross the cell barrier and be absorbed by Caco-2 cells, resulting in further cleavage and the subsequent release of their parent deglycosylated forms [212]. BEA bioavailability was variable from 50% to 54% [213]. Higher duodenal bioavailability compared to colonic bioavailability of ENNs was observed. Particularly, the duodenal bioavailability of ENNs ranged from 58% to 77% for ENN A, from 69% to 70% for ENN A1, from 65% to 67% for ENN B, and from 62% to 65% for ENN B1. Colonic bioavailability ranged from 17% to 33% for ENN A, from 41% to 50% for ENN A1, from 48% to 55% for ENN B, and from 52% to 57% for ENN B1 [67]. In contrast, FB1 was not absorbed by Caco-2 cells [214].
Berger et al. [215] showed that OTA was absorbed by the human intestinal mucosa by passive diffusion of the undissociated form of OTA and it was not appreciably metabolized by Caco-2 cells [215]. DON and NIV were not significantly metabolized or accumulated in Caco-2 cells as well [71,202,203,207,216,217]. Therefore, upon ingestion, these mycotoxins can be absorbed from the gut via intestine cells, then entered into the systemic circulation and thus transported to the whole body. Nevertheless, the intestinal absorption of OTA would be limited thanks to the presence of the MRP2 [215] and breast cancer resistance protein (BCRP) [204]. An efflux of AFB1 was also associated with BCRP [218], and DON was a substrate for both P-gp and MRP2 [202]. P-gp has been shown to be involved in the efflux of FB1 [214], and NIV interacted with P-gp and MRP2 [203]. Several studies showed that DON transport was unaffected by the transporter [207,209]. However, DON uptake and efflux are carrier-mediated processes, and P-gp and organic anion-transporting peptides may be the major efflux/uptake transporters for DON in Caco-2 cells, respectively [219]. The stepwise c-Jun-N-terminal kinase-Akt-nuclear factor kappa-light-chain-enhancer of activated B cells (JNK-Akt-NF-κB) pathway elaborates upon P-gp induction following DON exposure in mammalian cells and provides a self-protection mechanism to resist exogenous toxic compounds such as DON and T-2 [220]. These dissimilarities may be consequences of differences in exposure conditions to the toxin. Particularly, transport experiments were performed in pH gradient, and the acidification of the apical compartment may increase the fraction of the uncharged molecules facilitating diffusion across the cell membrane and intracellular accumulation [221]. Furthermore, differences in the culture medium, passage number, and time in culture before splitting may lead to significant differences in ABC transporter expression and functionality [222].
Intestinal absorption of AOH was more extensive and faster than AME. About 23-26% of the apically applied AOH reached the basolateral compartment, while only about 3-7% of the initial amount of AME in the apical chamber reached the basolateral side. In basolateral medium, several metabolites were also detected: Three AOH metabolites (3-O-sulfate, 3-, and 9-O-glucuronide) and AME-3-O-glucuronide [119]. Several authors have already shown the ability of Caco-2 cells to metabolize ZEA into αor β-ZEA, as well as into its glucuronidated and sulphated forms [78,166,211]. Videmann et al. [206] established that facilitated or active transport was involved in the transportation of ZEA and its metabolites. Particularly, they were substrates for ABCC1-3 transporters. ZEA and α-ZEA were mostly extruded by ABCC2 at the AP side and ABCC3 was able to transport β-ZEA at the BL side [206].
Treatment of Caco-2 cells with mycotoxins at reasonable concentrations must have no significant effect on cell viability, cell damage, and barrier integrity. FB1 at a concentration of up to 138 µM did not induce variation on cell viability and differentiation [214]. Similarly, ZEA concentration of up to 200 µM had no significant effect on cell viability and cell damage [78,206], and the integrity of the cell monolayers was preserved throughout the incubation with ZEA at a concentration of up to 40 µM, indicating that ZEA does not have detrimental effects on epithelial integrity in vitro [212]. Moreover, Caco-2 cells exposure to 5 µM of NIV showed neither a significant increase in the sucrose flux nor a significant decrease in TEER values [203]. DON also had no significant effect on Caco-2 cell viability at a concentration of up to 33 µM [202,209].
However, other studies reported that mycotoxins such as ZEA, DON, FB1, T-2, PAT, AFB1, and OTA decreased the TEER of intestinal epithelial cell lines in porcine as well as in human epithelium [10,195,196,198,[223][224][225][226][227]. A reduction in TEER can cause an increase in the paracellular permeability, changes in transcellular flux through altered plasma channels or pumps, and uncontrolled cell death within the monolayer [228]. Pfeiffer et al. [211] showed that 20 µM of ZEN was able to affect the apparent permeability coefficients of Caco-2 cells, leading to their quick absorption from the intestinal lumen into the portal blood [211]. Moreover, the important indicators of intestinal permeability are tight junction proteins, which are comprised of several multiprotein complexes, including transmembrane proteins (claudin, occludin, and junctional adhesion molecule) and cytoplasmic scaffolding protein and signaling proteins, including zonula occludens [229]. DON, ZEA, FB1, T-2, AFM1, and OTA have a deleterious effect on tight junctions of claudin 3, claudin 4, claudin 7, and occluden [195][196][197][198].
Tight junction structure and function can be regulated by signaling molecules involved in the mitogen-activated protein kinase-dependent (MAPK) pathways [230]. Therefore, the rapid activation of MAPK, ZEA, and DON decreased the expression of tight junction proteins, resulting in intestinal barrier impairments [134,197]. DON and other trichothecenes are known for their binding of the ribosomal peptidyltransferase, inhibition of protein synthesis, and rapid activation of MAPK via inducing two signal transduction pathways of a process named the ribotoxic stress response [227,[231][232][233]. The first pathway consists of the double-stranded RNA-activated protein kinase, leading to stimulation of JNK and p38 [25]. The second pathway involves hematopoietic cell kinase belonging to the Src tyrosine kinase family, which are upstream transducers of activation of MAPK. Among the primary MAPK subfamilies, such as p44/42 extracellular signal-regulated protein kinase (ERK), p38, and JNK [234], p44/42 ERK can be involved in intestinal epithelial cell morphology and in the structure of tight junctions. It was reported that the DON-induced activation of the p44/42 ERK signaling pathway inhibits the expression of claudin-4, which leads to reduces the barrier function of the intestine evaluated by TEER, paracellular permeability [197,227]. Treatment with 10 µM of DON also increased ERK, P38, JNK, and c-Jun phosphorylation levels by 2-fold, 30-fold, 61-fold, and 5-fold, respectively, and altered the gene expression levels of occludin, claudin-3, and the composition of tight junction proteins ( Figure 13) [235]. The activation of p44/42 MAPK was partially involved in the detrimental effects of the integrity of tight junction caused by AFM1 and OTA [224]. Toxins 2020, 12, x FOR PEER REVIEW 24 of 37 In addition to the tight junction, the maintenance of intestinal barrier-related paracellular secretions, such as cytokines and chemokines, are important as well. ZEA metabolites, α-and β-ZEA, can be beneficial to the intestine by decreasing the expression of both interleukin-8 (IL-8) and interleukin-10 in a dose-dependent manner. Its metabolites have a rather anti-inflammatory effect on the epithelial intestinal cells [225]. However, cytokines are related to the impairment of intestinal integrity when exposed to ZEA and FB1 [225,226,236]. Moreover, the correlation between permeability and IL-8 secretion induced by DON in the intestine was investigated by the authors of [209]. IL-8 was examined as a factor affecting intestinal barrier function, and the increased IL-8 secretion may be involved in the TEER decrease [237]. Similar results were reported by the authors of [238]. Consequently, exposure to certain mycotoxins, particularly DON, may cause damage to the intestinal integrity and lead to various chronic intestinal inflammatory diseases, such as inflammatory bowel disease [195]. In addition, the synergic effects of OTA and AFM1 that might exacerbate intestinal inflammation were also reported [239].
Although the Caco-2 cells model offers several advantages, such as the reproducibility of results, controlled environment, and in-depth mechanistic insight [240], some limits of Caco-2 for assessing the bioavailability were also reported [241]. The main disadvantages of these models are the lack of the regulatory processes of the complex mucosal barrier and inability to accurately calculate the fractional transport and flux rate through the static transport conditions [242]. Moreover, it has been shown that significant variation of the expression level of efflux transporters, such as BCRP, MRP2, and MDR1 in the Caco-2 cell monolayer in human small and large intestines, affect the results as well [243,244]. The Caco-2 cell monolayer is somewhat unsuccessful in simulating in vivo intestinal environment due to lack of expression of CYP3A4, which is responsible for the biotransformation of many compounds in the human epithelial cell [245]. Further drawbacks of these models include the incapability of simulating the changes of intestinal pH system, since it is performed at constant pH conditions. In addition, variations in TEER and permeability were also reported to be related to the source of Caco-2 cell and interlaboratory differences in protocol design [192]. In addition to the tight junction, the maintenance of intestinal barrier-related paracellular secretions, such as cytokines and chemokines, are important as well. ZEA metabolites, αand β-ZEA, can be beneficial to the intestine by decreasing the expression of both interleukin-8 (IL-8) and interleukin-10 in a dose-dependent manner. Its metabolites have a rather anti-inflammatory effect on the epithelial intestinal cells [225]. However, cytokines are related to the impairment of intestinal integrity when exposed to ZEA and FB1 [225,226,236]. Moreover, the correlation between permeability and IL-8 secretion induced by DON in the intestine was investigated by the authors of [209]. IL-8 was examined as a factor affecting intestinal barrier function, and the increased IL-8 secretion may be involved in the TEER decrease [237]. Similar results were reported by the authors of [238]. Consequently, exposure to certain mycotoxins, particularly DON, may cause damage to the intestinal integrity and lead to various chronic intestinal inflammatory diseases, such as inflammatory bowel disease [195]. In addition, the synergic effects of OTA and AFM1 that might exacerbate intestinal inflammation were also reported [239].
Although the Caco-2 cells model offers several advantages, such as the reproducibility of results, controlled environment, and in-depth mechanistic insight [240], some limits of Caco-2 for assessing the bioavailability were also reported [241]. The main disadvantages of these models are the lack of the regulatory processes of the complex mucosal barrier and inability to accurately calculate the fractional transport and flux rate through the static transport conditions [242]. Moreover, it has been shown that significant variation of the expression level of efflux transporters, such as BCRP, MRP2, and MDR1 in the Caco-2 cell monolayer in human small and large intestines, affect the results as well [243,244]. The Caco-2 cell monolayer is somewhat unsuccessful in simulating in vivo intestinal environment due to lack of expression of CYP3A4, which is responsible for the biotransformation of many compounds in the human epithelial cell [245]. Further drawbacks of these models include the incapability of simulating the changes of intestinal pH system, since it is performed at constant pH conditions. In addition, variations in TEER and permeability were also reported to be related to the source of Caco-2 cell and interlaboratory differences in protocol design [192].
To reduce the heterogeneity of the Caco-2 parental cell line and to improve the performance and the stability of this cellular model, some clonal derivative of Caco-2 cells have been established. The Caco-2/TC7 cell line, which was isolated from a late passage of the parental Caco-2 line, is suitable for intestinal absorption model due to a less heterogenic cellular population, resulting in better reproducibility of results [246]. The human intestinal HT-29 cell line is another cell line from colorectal origin with epithelial morphology and has a large proportion of mature goblet cells that can produce mucins. Therefore, the co-culture of Caco-2 and HT-29 with a ratio of 9:1 was used to provide a better representation of the intestinal tract [247]. In addition, the human colon carcinoma (HCT-116) and human colon adenocarcinoma (SW480) cells used in unraveling cancer-related mechanisms and the human duodenum adenocarcinoma (HuTu-80) cell line simulating duodenal cells are less popular [191]. More recently, a combination of in vitro digestion and Caco-2 absorption was used to simulate the physiological settings in the gastrointestinal tract and determine the bioaccessibility and bioavailability of the ZEA reaction products [72].

Conclusions
Scientific insights in the production of mycotoxins, their toxicities, biotransformation, and metabolism in different organisms have greatly contributed to a more detailed understanding of the chemical hazards in food. Mycotoxins can notably biotransform and detoxify in the liver, as well as in the digestive tract. The results obtained with Caco-2 monolayer are useful in the prediction of mycotoxins' intestinal permeability, transport mechanism, and gene regulation of transporters and enzymes in humans, and may help interpret properly data of mycotoxins' absorption for better comprehension of their possible adverse effects. Furthermore, the combined usage of in vitro digestion models with in vitro intestinal absorption models using Caco-2 cells may offer more complete picture during digestion in the intestinal tract. However, the correlation between in vitro Caco-2 data and in vivo situation necessitates further investigation.

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