Human Biomonitoring of Mycotoxins in Blood, Plasma and Serum in Recent Years: A Review

This manuscript reviews the state-of-the-art regarding human biological monitoring (HBM) of mycotoxins in plasma, serum and blood samples. After a comprehensive and systematic literature review, with a focus on the last five years, several aspects were analyzed and summarized: (a) the biomarkers analyzed and their encountered levels, (b) the analytical methodologies developed and (c) the relationship between biomarker levels and some illnesses. In the literature reviewed, aflatoxin B1-lysine (AFB1-lys) and ochratoxin A (OTA) in plasma and serum were the most widely studied mycotoxin biomarkers for HBM. Regarding analytical methodologies, a clear increase in the development of methods for the simultaneous determination of multiple mycotoxins has been observed. For this purpose, the use of liquid chromatography (LC) methodologies, especially when coupled with tandem mass spectrometry (MS/MS) or high resolution mass spectrometry (HRMS) has grown. A high percentage of the samples analyzed for OTA or aflatoxin B1 (mostly as AFB1-lys) in the reviewed papers were positive, demonstrating human exposure to mycotoxins. This review confirms the importance of mycotoxin human biomonitoring and highlights the important challenges that should be faced, such as the inclusion of other mycotoxins in HBM programs, the need to increase knowledge of mycotoxin metabolism and toxicokinetics, and the need for reference materials and new methodologies for treating samples. In addition, guidelines are required for analytical method validation, as well as equations to establish the relationship between human fluid levels and mycotoxin intake.

The designation of mycotoxins includes a group of highly heterogeneous compounds, in terms of chemical structure and toxicological properties [2], with a low molecular mass [3]. The classification of mycotoxins is a complex task since they have diverse chemical structures and biosynthetic origins and, also, they are produced by a great variety of fungal species. Moreover, it should be noted that the same mycotoxin can be produced by several fungal species; for example, ochratoxin A (OTA) can be       Until now, only a few biomarkers have been validated for estimating mycotoxin exposure, namely, DON-glucuronides in urine, adduct of AFB1 with albumin (AFB1-alb) and AFB1-lys in human plasma, and AFB1-N7-guanine in urine [43]. Moreover, there is a shortage of HBM studies and biomarker definitions with respect to some mycotoxins, e.g., T-2 toxin (T-2), HT-2 toxin (HT-2), sterigmatocystin (STER) and nivalenol (NIV).
On the other hand, HBM requires sufficiently sensitive and validated analytical methods. There are three approaches to determining biomarkers of mycotoxins in biological matrices: direct, indirect and non-targeted analysis. Direct analysis uses standardized analytical methods that have been properly optimized and validated. However, this approach can be applied to parent compounds only, because just a few modified mycotoxins are available as reference substances. To overcome this problem, indirect determination can be used. In this case, modified mycotoxins are transformed into their free forms, which can then be analyzed using routine methods. Hydrolysis, reduction and other specific reactions can be used. Until now, direct and indirect methods based on liquid chromatography tandem mass spectrometry (LC-MS/MS) have been chosen to identify and quantify unmodified and modified mycotoxins [28,45].
While the potential of tandem mass spectrometry for the quantitative determination of multiple mycotoxins has been widely documented, there is currently growing interest in evaluating new mass spectrometer (MS) detection approaches, based mainly on high-resolution MS (HRMS), for a more complex task, i.e., the determination of unknown mycotoxin derivatives. In this case, non-targeted analysis is required and LC-HRMS is usually employed [43]. As an example, MS techniques based on high-resolution Orbitrap MS have advanced to the point that they now include the non-targeted analysis of fungal metabolites [48,49].
Furthermore, HBM requires easily accessible biological matrices such as urine, serum, plasma and breast milk. Urine analysis presents some advantages because sampling is non-invasive and collection is easy and it contains biomarkers of almost all mycotoxins; therefore, it is very often the matrix of choice for estimating human exposure [50,51]. Nevertheless, urine biomarkers reflect day-to-day variations in mycotoxin intake, so samples must be taken at different times over a 24-h period. In addition, there is variability in the volume of collected urine at different times and among individuals; this variability leads to changes in the concentration of excreted compounds in the samples. Several methods can be applied in order to reduce this variability [52]. Among them, the normalization of the mycotoxin levels for creatinine concentration (µg mycotoxin L -1 urine/g creatinine L −1 urine) is one of the most employed [30,53]. However, this procedure also has some drawbacks, since it is not clear if the mycotoxin/creatinine ratio can be used for interindividual comparison. Creatinine secretion can vary among different people related to muscle mass, sex, age, season, diet, etc. [52]. Clearly, breast milk can be used to monitor only lactating women; nonetheless, it is an excellent information source for exposure in breastfed babies. Serum and plasma matrices at least have the advantage of requiring less sensitive methods because they contain higher levels of compounds [48]. In addition, while urinary excretion normally indicates recent mycotoxin intake, plasma and serum measurements indicate long-term exposure [40]. However, they are limited in that they require invasive collection methods and medical professionals.
Another important aspect is that the level of a biomarker in the same individual usually varies over time. Coronel et al. (2010) [24] reviewed this aspect in relation to OTA and suggested that determining this mycotoxin in plasma would be useful if it were used to characterize populations instead of individuals. Toxicokinetic studies of different biomarkers in humans would be highly beneficial, but they are greatly limited for obvious ethical reasons.

Mycotoxin Determination in Human Blood, Plasma and Serum
Due to the complexity of blood, plasma and serum samples, matrix components might interfere in analyte retention, in addition to reducing purification, recovery and method sensitivity and producing matrix effects when MS detectors are used. For these reasons, all the papers reviewed included Toxins 2020, 12, 147 9 of 44 the different steps designed to extract analytes and cleanup matrix components. The different methodologies used for preparing samples and determining mycotoxins in plasma, serum and blood samples over the last five years are shown in Table 2 and Figure 3. Toxins 2020, 12, x FOR PEER REVIEW 9 of 43

Mycotoxin Determination in Human Blood, Plasma and Serum
Due to the complexity of blood, plasma and serum samples, matrix components might interfere in analyte retention, in addition to reducing purification, recovery and method sensitivity and producing matrix effects when MS detectors are used. For these reasons, all the papers reviewed included the different steps designed to extract analytes and cleanup matrix components. The different methodologies used for preparing samples and determining mycotoxins in plasma, serum and blood samples over the last five years are shown in Table 2 and Figure 3. The most widely used mycotoxin-related extraction, cleanup and enrichment procedure was solid-phase extraction (SPE), based on the retention of analytes on a fixed support in a cartridge. SPE was used in 54.3% of the reviewed papers; most of them (76%) used it as a purification method in samples previously digested with a mixture of proteinases (Pronase ® ) to detect adducts of aflatoxin (AF-adducts). SPE can be used to purify and preconcentrate analytes. Liquid-liquid extraction (LLE) with different organic solvents was also widely used. Acetonitrile (ACN) [23,[54][55][56], chloroform [57], chloroform/isopropanol [55], ethylacetate (EtOAc) [48,58], diethyl ether [58] and mixtures of these with water, such as H2O/acetone/ACN [59][60][61][62][63], have proven to be the most efficient solvents for mycotoxin LLE extraction. An acid solution can help the extraction process by disrupting interactions between the toxins and sample constituents, such as proteins. Thus, ACN with formic or acetic acids [64,65] has also been used to extract mycotoxins. Therefore, the main sample preparation procedures described in the reviewed literature from the last five years are consistent with those included in the review by Escrivá et al. (2017) [31], although SPE took over from LLE as the most common technique. The use of Captiva ® EMR-Lipid cartridges has been described as a novel procedure in order to reduce matrix effects by eliminating phospholipids from plasma during sample preparation [66]. Apart from this one, no innovative improvements to the sample preparation process have been reported.
Quick, easy, cheap, effective, rugged, and safe (QuEChERS) (based on extraction with ACN), and immunoaffinity columns (IAC) (a variation of SPE in which the fixed support contains specific antibodies for mycotoxin retention) were not widely used ( Figure 3). The QuEChERS procedure is based on the extraction of analytes from an aqueous matrix using a polar organic solvent (MeOH or ACN). ACN and MeOH are miscible with water; therefore, the addition of salts to the aqueous phase is needed in order to achieve the separation of the two aqueous-organic phases. Also, salt addition favors the distribution of mycotoxins to the organic solvent. However, when very polar mycotoxins (such as PAT or FBs) are in the samples, high levels of organic solvent during the extraction procedure (that contributes to less polar mycotoxin extraction) lead to low recovery values, probably because they remain in the discarded aqueous phase during the extraction [67]. In these cases, a high percentage of water is needed [53] and less polar mycotoxins can be non-extracted. In the case of The most widely used mycotoxin-related extraction, cleanup and enrichment procedure was solid-phase extraction (SPE), based on the retention of analytes on a fixed support in a cartridge. SPE was used in 54.3% of the reviewed papers; most of them (76%) used it as a purification method in samples previously digested with a mixture of proteinases (Pronase ® ) to detect adducts of aflatoxin (AF-adducts). SPE can be used to purify and preconcentrate analytes. Liquid-liquid extraction (LLE) with different organic solvents was also widely used. Acetonitrile (ACN) [23,[54][55][56], chloroform [57], chloroform/isopropanol [55], ethylacetate (EtOAc) [48,58], diethyl ether [58] and mixtures of these with water, such as H 2 O/acetone/ACN [59][60][61][62][63], have proven to be the most efficient solvents for mycotoxin LLE extraction. An acid solution can help the extraction process by disrupting interactions between the toxins and sample constituents, such as proteins. Thus, ACN with formic or acetic acids [64,65] has also been used to extract mycotoxins. Therefore, the main sample preparation procedures described in the reviewed literature from the last five years are consistent with those included in the review by Escrivá et al. (2017) [31], although SPE took over from LLE as the most common technique. The use of Captiva ® EMR-Lipid cartridges has been described as a novel procedure in order to reduce matrix effects by eliminating phospholipids from plasma during sample preparation [66]. Apart from this one, no innovative improvements to the sample preparation process have been reported.
Quick, easy, cheap, effective, rugged, and safe (QuEChERS) (based on extraction with ACN), and immunoaffinity columns (IAC) (a variation of SPE in which the fixed support contains specific antibodies for mycotoxin retention) were not widely used ( Figure 3). The QuEChERS procedure is based on the extraction of analytes from an aqueous matrix using a polar organic solvent (MeOH or ACN). ACN and MeOH are miscible with water; therefore, the addition of salts to the aqueous phase is needed in order to achieve the separation of the two aqueous-organic phases. Also, salt addition favors the distribution of mycotoxins to the organic solvent. However, when very polar mycotoxins (such as PAT or FBs) are in the samples, high levels of organic solvent during the extraction procedure (that contributes to less polar mycotoxin extraction) lead to low recovery values, probably because they remain in the discarded aqueous phase during the extraction [67]. In these cases, a high percentage of water is needed [53] and less polar mycotoxins can be non-extracted. In the case of IAC, this was due to its specificity. This characteristic represents its main advantage for its use in single analysis; nevertheless, it is a drawback in multi-mycotoxin analysis, because it prevents the simultaneous retention of several compounds and their related metabolites.
In addition, some authors have added an enzymatic de-conjugation method to evaluate phase II metabolites. Plasma or serum samples were incubated with β-glucuronidase, or β-glucuronidase/sulfatase at 37 • C overnight, then the selected extraction procedure was applied [55,64,68,69].
With respect to the detection and quantification of mycotoxin in blood, plasma and serum samples, several techniques were applied ( Figure 3). Most of the papers referenced focused on determining one mycotoxin or some structurally related mycotoxins (78% of the studies reviewed). In other cases, the objective was to determine multiple mycotoxins simultaneously.
ELISA is a routine screening tool for the rapid monitoring of mycotoxins that offers high sensitivity, affordability and ease of analysis [70]. A wide range of commercial ELISA-based kits is available for different matrices, including human plasma, which makes it a highly versatile methodology. However, the potential for cross-reactivity with metabolites of target compounds or matrix components can give rise to overestimated values [71]. For this reason, AOAC International has not approved any ELISA method [72] and positive results must be confirmed (e.g., through chromatographic methods). Another shortcoming of ELISA is that it relies on specific antibodies for each mycotoxin; therefore, it is not appropriate for multi-mycotoxin determination. In fact, ELISA was applied in just 15% of all studies reviewed [73][74][75][76][77][78][79].
Liquid chromatography (LC), performed on reversed-phase columns, has become the main tool for determining mycotoxins in human fluids, including human blood, plasma and serum, and was used in 86% of all publications referenced. Advances in high-sensitivity detectors, in LC pump design and in column-packing materials have led to better limits of detection and improved chromatographic performance [80]. The introduction of Ultra-LC (UHPLC), characterized by uniform column packing material with a particle size of less than 2-3 µm and new pumps and detectors, has represented an improvement, given that it achieves shorter run times, reduces solvent consumption and improves chromatographic resolution and efficiency. UHPLC was reported in 42.8% of the chromatographic methods summarized in this review [48,53,58,64,65,68,69,[81][82][83].
The use of MS detectors in mycotoxin determination provides a valuable and confirmatory technique. The literature on mycotoxin determination in biological fluids is scarce, but chromatographic systems coupled with tandem MS are a reference technique for this purpose and were used in 58% of all articles referenced. Although the matrix effect on the MS signal is significant, instrumentation is expensive, and it requires high technical personnel training. MS detectors have key advantages, including high selectivity and sensitivity and the possibility of structural elucidation. Although several applications of LC-MS and LC-MS/MS have been developed for single-mycotoxin analysis, this methodology offers the possibility of simultaneously detecting multiple mycotoxins in a single run. Therefore, an increasing number of articles focused on biomonitoring several mycotoxins in human plasma, serum and blood samples, with a tendency towards LC-MS/MS ( Figure 4). The main ionization sources employed were electrospray ionization (ESI) and atmospheric pressure chemical ionization (APCI). Regarding the mass analyzers, the triple quadrupole (QqQ) and the quadrupole-ion trap (QTrap) were the most commonly used (57% and 32%, respectively).  Recent advances in high-resolution mass spectrometry (HRMS) [48,49,81] and the use of ulti-mycotoxin biomarker databases have improved the identification and validation of iomarkers of exposure and revealed new mycotoxin metabolites [43]. These studies have ontributed to a more in-depth knowledge of the biomarkers of exposure in humans.
Other approaches have also been investigated during the last few years for developing new ethodologies, based on biosensors, in order to determine mycotoxins in plasma, serum or blood. A iosensor is "any measuring device containing a biologically derived sensing (so-called iorecognition) element intimately associated with a sensor element (physicochemical transducer)" 93]. These devices have the advantage of being portable, sensitive and not as expensive as raditional approaches [93]. Recently, excellent reviews have been published regarding the evelopment and the applicability of biosensors for mycotoxin determination [94][95][96][97][98]  Recent advances in high-resolution mass spectrometry (HRMS) [48,49,81] and the use of multi-mycotoxin biomarker databases have improved the identification and validation of biomarkers of exposure and revealed new mycotoxin metabolites [43]. These studies have contributed to a more in-depth knowledge of the biomarkers of exposure in humans.
Other approaches have also been investigated during the last few years for developing new methodologies, based on biosensors, in order to determine mycotoxins in plasma, serum or blood. A biosensor is "any measuring device containing a biologically derived sensing (so-called biorecognition) element intimately associated with a sensor element (physicochemical transducer)" [93]. These devices have the advantage of being portable, sensitive and not as expensive as traditional approaches [93]. Recently, excellent reviews have been published regarding the development and the applicability of biosensors for mycotoxin determination [94][95][96][97][98]. Very few biosensor methodologies focusing on the analysis of mycotoxins in biological fluids have been published in the last 5 years. Moreover, all of them are limited to single-mycotoxin determination. Abnous et al. (2017) [99] obtained a good linear relationship in the range of 7-500 pg/mL, achieving a limit of detection (LOD) of 2.8 pg/mL and recovery values from 95.4% to 108.1% when serum samples were spiked with AFB1. Also, these authors demonstrated that their methodology was selective as regards to other possible interfering mycotoxins, such as AFM1  [103] determined ZEA levels in plasma and urine samples collected from healthy volunteers using electrochemical immunosensors. ACN acidified with 1% formic acid was added to the samples to precipitate proteins. After vortexing and centrifuging, the supernatant was dried (N 2 ) at 30 • C. Finally, the residue was dissolved in phosphate buffer solution (PBS) at pH 6.5 before the analysis. The recovery rates were from 90.2% to 107.4% for plasma samples and 90.3% to 106.3% for urine samples, and a LOD of 0.005 ng/mL was indicated for both matrices. This work was the only one that achieved the determination of the mycotoxin in real samples (two positive plasma samples and two positive urine samples). In addition, the obtained results were compared with those obtained using UHPLC-MS/MS and a good correlation for the mycotoxin levels by the two methodologies was found. Nevertheless, and despite the advantages of biosensors and their foreseeable interest for mycotoxin determination in the future, nowadays, biosensors are still under development and research [97,98] and new advances should be made in order to enhance their specificity and sensitivity [98].
There are some difficulties in the development and validation of analytical methods for HBM. First, standards of analytes are needed to carry out the method validation. They do not exist for some mycotoxins and obtaining them for the majority of metabolites is a complex task. Moreover, they are very expensive when available for purchase. In addition, when MS/MS detectors are used, it is often difficult to obtain matrix-free samples for the preparation of calibrators; in these cases, the use of labeled compounds can help. However, these labeled compounds are expensive and, more importantly, can have different retention times, recoveries and matrix effects than those of the parent compounds, and thus these factors should be taken into account during method validation [104,105]. Also, matrix reference materials with a known concentration of biomarkers are not at researchers' disposal for method validation.
Though validation is mandatory [106], there are no guidelines for the validation of mycotoxin quantification methods in human body fluids. Some authors [53,64,65] referred to EU Commission Decision No. 2002/657/EC [107], which mentions the rules for the analytical methods to be used in the testing of official samples (residues in live animals and animal products), or the EU Commission Regulation (EC) No. 401/2006 [108] which refers to methods for analyzing mycotoxins in foodstuffs. Other authors [48,64] validated their methodologies in accordance with guidelines on bioanalytical method validation issued by the Food and Drug Administration (FDA) [109] or the European Medicines Agency [110] which refer to drug development. Validation requires aspects such as criteria for determining the different validation parameters and controlling the analysis, along with the required limits of detection or quantification values for mycotoxin biomarkers in biological fluids.

Mycotoxin Biomarkers in Human Blood, Plasma and Serum
In the last five years, mycotoxin HBM in blood, plasma and serum has been carried out primarily through the analysis of parent compounds, e.g., OTA, ZEA, AFB1, aflatoxin B2 (AFB2), aflatoxin G1 (AFG1) and aflatoxin G2 (AFG2). In some cases, HBM was carried out by determining adducts. For instance, the measurement of AFB1-albumin adducts in plasma or serum, which are formed with the lysine amino acid of albumin, is used as a biomarker of AFB1. In the papers reviewed, AFB1-lys was the most commonly analyzed biomarker (29.9%); this was followed by OTA (21.6%), in both singleand multi-mycotoxin analyses. Table 3 provides a record of the biomarkers of mycotoxins in plasma and serum in the papers reviewed. These data are summarized in Figure 5.
Regarding AFs, the biomonitoring of plasma and serum is carried out analyzing AFB1 or by determining AFB1-albumin adducts. Albumin adducts are chosen primarily because they present a half-life of around 2-3 months and, hence, the presence of these adducts in plasma and serum samples indicates long-term and chronic exposure to AFB1. In addition, AFB1-albumin adducts are stable in serum samples stored at −80 °C for over 25 years, and can therefore be re-analyzed years later [120].
Regarding AFs, the biomonitoring of plasma and serum is carried out analyzing AFB1 or by determining AFB1-albumin adducts. Albumin adducts are chosen primarily because they present a half-life of around 2-3 months and, hence, the presence of these adducts in plasma and serum samples indicates long-term and chronic exposure to AFB1. In addition, AFB1-albumin adducts are stable in serum samples stored at −80 • C for over 25 years, and can therefore be re-analyzed years later [120].
Moderate-to-high levels of AFs in plasma and serum were reported globally, depending on the sampling country and the LOD of the methods used (Table 3). Cross-sectional surveys were conducted among the US [124] and Kenyan population [125]. The US study reported much lower AF levels than those of developing countries, where aflatoxin exposure has become a public health problem.
In recent years, several studies have been carried out to identify a link between plasma or serum AF levels and certain diseases, including hepatocellular carcinoma (HCC), chronic cirrhosis, chronic hepatitis B, colorectal cancer, liver cancer and human immunodeficiency virus [57,73,74,112,126]. Koshiol et al. (2017) [87] found a relationship between aflatoxin exposure and gallbladder cancer (GBC) based on levels of AFB1-lys in plasma samples from Shanghai. The AFB1-lys adduct was detected in 32% of the patients with this pathology (5.4 pg/mg albumin) and 15% of the control group (1.2 pg/mg albumin). Díaz-León et al. (2019) [81] reported exposure to AFB1-lys in a high number of the samples analyzed from healthy women, and observed a high correlation between AFB1 exposure and some markers of renal injury. In addition, Jolly et al. (2015) [126] conducted a study in Kumasi (Ghana) in which they detected high levels of AFB1-lys in the blood of study participants that varied according to season. The authors concluded that stored food was most probably the source of AFB1 exposure. In Taiwan, two case-control studies were performed in order to investigate the risk associated with AFB1 in HCC patients with different forms of hepatitis (B or C). These studies reported that high exposure to AFB1 increases the risk of HCC in populations with risk factors for cirrhosis, such as alcohol consumers and/or hepatitis patients [73,74].
The effect of aflatoxin exposure in vulnerable groups, such as pregnant women and children, is of particular concern [127]. Hernandez-Vargas et al. (2015) [75] established an association between AFB1 exposure in pregnant women and DNA methylation, which may have an impact on the health of their children. Based on the possible association between infant exposure and poor growth effects and on previous findings that have demonstrated that AFs cross the placental barrier [128,129], Lauer et al. (2018) [90] studied the association between maternal aflatoxin exposure during pregnancy and adverse birth outcomes. AFB1-lys was detected in 100% of serum samples from the expectant mothers, with values ranging from 0.71 to 95.60 pg/mg albumin. Elevated levels of AFB1-lys were associated with lower infant birth weight, lower weight for age and smaller head circumference for age in infants at birth. Moreover, the metabolism of AFs to epoxides may increase during pregnancy. Groopman et al. (2014) [129] concluded that the fetus has the capacity to metabolize aflatoxin to levels comparable to those of the mother and, consequently, pregnancy may pose a high risk of aflatoxin exposure for pregnant women and their fetuses.
A number of studies have been carried out to evaluate the presence of AFs in plasma or serum from children with severe acute malnutrition (SAM) in Kenya [92], Nigeria [49], Tanzania [113,117] Uganda [88,90], Mexico [81,91] and other countries [75,[77][78][79]83,87,114,115]. This is a particularly relevant topic due to the rates of childhood growth disorders in these countries. McMillan et al. (2018) [49] improved the accuracy of AFB1-lys quantification by means of a sample treatment that involved protein precipitation prior to Pronase ® digestion, followed by SPE cleanup; they also used 13 C internal standards and LC-MS/MS. The authors used this methodology to evaluate the presence of AFB1-lys in Nigerian children suffering from SAM. AFB1-lys levels in all children (control and SAM) ranged from 0.2-59.2 pg/mg albumin and were significantly higher in children with SAM (4.3 pg/mg albumin) compared to controls (0.8 pg/mg albumin) (p-value: 0.0083). Leroy et al. (2018) [91] studied the association between serum AFB1-lys levels and height or height-for-age difference at several ages of children that were exposed to a milk-based multiple micronutrient-fortified food in Mexico. Low aflatoxin exposure was associated with greater child linear growth. In contrast, Chen et al. (2018) [113] studied plasma samples from children at 24 months of age (n = 60) in Tanzania and analyzed them for AFB1-lys. Seventy-two percent of the children had detectable levels of AFB1-lys, with a mean level of 5.1 pg/mg albumin; however, no association was observed between aflatoxin exposure and growth impairment. Another study linked aflatoxin exposure to chronic hepatomegaly, which could represent a health risk for children [130].
HBM is also useful for assessing possible occupational exposure to AFs [86]. Saad-Hussein et al. (2016) [76] observed higher levels of AFB1-albumin in plasma from bakers compared to milling workers, and in both groups of workers compared to the control group. The authors attributed this to the exposure to different AFs in the workplace. Ferri et al. (2017) [86] studied levels of AFB1, AFB2, AFG1, AFG2 and aflatoxin M1 (AFM1), the most prevalent metabolite of AFB1, in mill workers exposed to contaminated dust. These authors combined a cleanup step that used a specific IAC and detection with an LC-FLD method to achieve good recoveries (84%-111%) for all AFs, except for aflatoxicol (AFOH) (60%). While there was no quantifiable presence of free AFs, AFB1-adducts were detected after treatment with Pronase ® .
All of these studies reinforced the idea that aflatoxin exposure is an important public health concern [81], and that HBM is a useful tool for studying the implication of AFs in different diseases [131]. In addition, they corroborate the premise that AF-adducts are the main biomarkers of AFs in plasma or serum.
With respect to OTA, direct detection in human plasma was widely performed. Because OTA binds to plasma proteins rapidly and with a high affinity, it constitutes a good biomarker of exposure due to its persistence in blood [17,24]. A good strategy for investigating human OTA exposure involves analyzing OTA levels in plasma or serum and converting internal OTA levels to estimated daily intake (EDI) values by means of the Klaassen equation (Equation (1)) [85,132]. The methodologies for OTA single-detection may include LLE or SPE of plasma or serum samples and analysis by LC-FLD [54,55,85,118] [62] studied the presence of OTA and 2´R-OTA (an OTA isomer that appears during the roasting process) in the blood of coffee drinkers. The authors reported the long-term persistence of the OTA isomer in blood and higher levels in coffee drinkers; consequently, 2´R-OTA could represent a good biomarker of OTA ingestion in these cases.
The relation between OTA plasma levels and specific disease risk markers, such as body mass index [133,134], kidney disease and inflammation (C-reactive protein) [119], was investigated. Di Giuseppe et al. (2012) [119] observed a positive association between OTA intake, C-reactive protein and cardiovascular risk. Despite the fact that OTA was detected in nearly all plasma samples analyzed in the studies reviewed, the value of 500 ng/L, which was related to the onset of kidney diseases, was exceeded in very few. Prati et al. (2016) [118] carried out a study to correlate OTA serum levels and liver function in two groups (one with and the other without chronic liver disease). The authors concluded that there was a relationship between OTA and C-reactive protein that confirms its inflammatory effect. However, the authors recommended further studies on this topic be carried out to clarify whether OTA is a risk factor for HCC or cirrhosis. In particular, OTA could pose a risk to people suffering from liver disease due to its capacity to induce DNA damage. In addition, OTA has a fairly long half-life in human blood and accumulates in the kidneys [135].  [54] carried out studies on plasma samples to assess OTA exposure in humans. OTA and OTα were determined through validated LC-FLD approaches that involved cleanup with LLE, with or without the cleavage of conjugates. OTA was detected in 100% and 48% of the samples, at levels of 0.72 and 0.14 ng/mL, respectively. The biomarker-based intake estimates (1.44 and 0.29 ng/kg.bw for both studies, respectively) fell well below the health-based guidance value (HBGV) (14 ng/kg.bw) [136]. In the study of Malir et al. (2019) [54], the authors also explored if OTA may contribute to kidney diseases such as renal tumors. However, there were non-significant differences between the OTA levels in Czech plasma samples from healthy volunteers and those from renal tumor patients. Woo et al. (2016) [85] analyzed OTA in serum samples from pregnant women and extrapolated the values obtained to the fetus. The authors considered EDI values for the fetus to be double those of the expectant mother. Although fetal exposure to OTA can be considered a high risk, limited information is available on such exposure and the associated toxicity problems.
In the case of ZEA, data suggesting a significant effect of exposure of ZEA metabolites on human health outcomes are scarce. Some studies have evaluated the relation between the presence of ZEA and its metabolites in biological samples with changes in estrogenic activity and adverse effects in the reproductive system. In a previous study, no differences between ZEA levels in plasma from control and cancer patients (cervical and breast cancer) was found [137]. Recently, Mauro et al. (2018) [68] studied the presence of ZEA and its metabolites in serum samples by performing enzymatic treatment of samples with glucuronidase, followed by purification through SPE columns and analysis by LC-MS/MS. The authors applied this methodology to samples from overweight or obese women in the USA. ZEA was detected in all obese women and exposure was associated with food intake (especially meat) and body mass index. The alteration of ZEA levels in the case of obese women, and consequently its potential implication for health, should be considered. Because there is evidence on endocrine disruption, particularly during fetal development, Fleck et al. (2016) [69] in USA, using a similar methodology, only detected ZEA in 1/11 serum samples from pregnant women, selected because of the presence of ZEA in the urine of the same individuals. Moreover, this positive serum sample had a level of ZEA near the limit of quantification (LOQ). In addition, a possible association with autism spectrum disorder (ASD) and ZEA exposure has been studied in Italy by De Santis et al. (2019) [58]. No ZEA values over the LOQ were obtained in serum samples from autistic children or from healthy controls.
CIT and its metabolite dihydrocitrinon (DH-CIT) in plasma have been described as good biomarkers of this mycotoxin in HBM [56]. Due to its nephrotoxicity, CIT may be a contributing factor to the high frequency of renal tumors in some countries [54] [23] to measure CIT and DH-CIT by means of an ACN protein precipitation, followed by centrifugation and analysis by LC-MS/MS, have been applied in studies on human plasma [23,55]. CIT was detected in almost all samples analyzed. However, due to the short half-life of this mycotoxin in human blood (about 9 h) and urine (6.7 h), and because CIT is extensively converted into DH-CIT (with a half-life of 8.9 h in urine), it does not accumulate in the organism and is considered to be of low concern in terms of human exposure, with the exception of some regions, including certain African countries, where further studies are required [56]. More biomarker-based analyses are needed to assess human health risks related to CIT exposure.
Biomonitoring data on other mycotoxins are very scarce and further research is required to gain more insight into their influence on human diseases.
Furthermore, humans suffer from exposure to multiple mycotoxins due to the fact that the human diet is varied and a number of fungi may be present in one raw material; therefore, the inevitable co-occurrence of different mycotoxins and their metabolites in human plasma and serum requires the development of a new approach to mycotoxin HBM.

Multi-Biomarker Studies
Co-contamination is particularly significant due to potential additive, antagonistic and/or synergistic toxic interactions [18,138]. Controlling this simultaneous presence in the human body requires the development of analytical methods to analyze multiple mycotoxins in biological fluids [139]. In fact, this is the current trend in analytical method development. As described above, chromatographic systems coupled with tandem MS have become the reference technique in this research and were used in 90% of all articles reviewed that included multi-mycotoxin analysis. As with single-mycotoxin monitoring, the main ionization sources employed were ESI and APCI. Regarding mass analyzers in multi-biomarker studies, QqQ and QTrap were the most commonly used, at 50% and 30%, respectively.
To date, only a few methods have been developed for multi-analyte detection of mycotoxins in plasma and serum samples [48,53,[58][59][60]64,65,82,111], and 90% of these methods have been developed in the last three years. Only five (50%) of these studies proposed methods for detecting more than 10 mycotoxins simultaneously. Additionally, some authors indicated that the LOQs obtained in these studies were not low enough for HBM studies [48], although no guidelines are available on this topic, as explained above (Section 2.2.).
The most common extraction procedure among the multi-detection methods reviewed was LLE (or SPE after an enzymatic procedure). This is due to the need for non-selective extraction procedures. Because sample pretreatment is critical, some authors assayed different preparatory methodologies to overcome the limitations presented by commonly used methodologies. For example, Osteresch et al. (2017) [59] proved that greater centrifugal force produces higher signal intensity and good limits of quantification in the lower pg/mL range for all 27 mycotoxins analyzed in serum samples. However, they reported high matrix effects. Slobodchikova and Vuckovic (2018) [48] used LC-HRMS (Orbitrap) to study different sample preparation procedures in order to reduce matrix effects, obtain a method for the multi-detection of 17 mycotoxins in human plasma and monitor the presence of unknown mycotoxins and metabolites. They selected three-step LLE, with which they obtained good recovery data. The low-cost method minimizes matrix effects, but is not suitable for OTA, FB1 or FB2 quantification. Arce-López et al. (2020) [66] recently developed and validated a LC-MS/MS methodology for the simultaneous analysis of 19 mycotoxins in plasma. Sample deproteinization and cleanup were performed in a single step by means of Captiva ® EMR-Lipid cartridges and several samples were processed simultaneously. Good recovery and matrix effect values were obtained due to the elimination of phospholipids from plasma during sample preparation and the use of matrix-matched calibration curves. LOD values ranged from 0.04 to 2.7 ng/mL (except for NIV) and mean recovery values from 68.8% to 97.6% (RDS ≤ 15%). The matrix effects were not significant for most of the mycotoxins and ranged from 75.4% and 109.3% (RDS ≤ 15%).
Some of the methods developed for multi-mycotoxin determination were designed to find a correlation between mycotoxin levels in plasma or serum and diseases, countries, seasons, gender, age and health statuses of donors (Table 3). In this study, OTA was the most abundant mycotoxin (27.7%). The EDI of OTA (2.4 ng/kg.bw/day) was lower than the tolerable daily intake (TDI) value, and the authors suggested that the potential health risk associated with this mycotoxin was low. Although the incidence and concentration of mycotoxins in males and females differed slightly, differences in mean concentrations between the two groups were not significant for all mycotoxins. More than 60% of the participants were exposed to one or more mycotoxin [65]. Ouhibi et al. (2020) [53] developed a method for detecting PAT, for the first time, and CIT in biological fluids to assess human exposure. Positive mycotoxin values were detected in 26% (PAT) and 36% (CIT) of the plasma samples from Tunisian participants, but no significant differences were observed between the control and colorectal cancer patients. In addition, higher CIT levels were detected in this study compared to previously published studies. Cao et al. (2018) [64] developed a multi-mycotoxin method for analyzing 11 carcinogenic mycotoxins in plasma from patients with HCC, including enzymatic de-conjugation with β-glucuronidase. Moreover, despite the fact that the authors obtained high recovery values with MeOH/acetic acid, they ended up opting for deproteinization with ACN/acetic acid, since lower matrix effects were observed. Sixty samples (30 control and 30 HCC patients) were collected for the study. AFB1 and STER were the most prevalent mycotoxins and were detected more frequently in patients with chronic liver disease (33% and 40%, respectively) than in control patients (13%). The authors suggested that it would be useful to study the possible influence of STER on HCC, since levels were higher in ill people than in controls. AFB2 was detected in 12% of the samples. AFG1, AFG2, AFM1, CIT and OTA were detectable only at the LOD. No PAT levels were found in the samples.
In a cross-sectional study, De Santis et al. (2017) [82] determined mycotoxins in serum samples from children with autism and healthy controls. After the digestion of samples with Pronase ® , purification with QuEChERS and an UHPLC-MS/MS analysis, the recoveries obtained were below 63%, except for AFB1 (82%). LOD ranged from 0.01 ng/mL for AFB1 to 11 ng/mL for gliotoxin (GLIO). In 2019, De Santis et al. [58] determined GLIO and OTA levels by means of UHPLC equipped with an FLD detector. Optimized recovery values for GLIO and OTA were obtained (63% and 75%). These studies sought to test the association between mycotoxins and ASD, since many gastrointestinal, inflammatory and neurological symptoms induced by mycotoxin exposure are similar to those often associated with ASD. Interestingly, these authors reported that children with autism have significantly lower levels of OTA in plasma when compared to their siblings and other healthy children. This could be explained by the altered OTA biotransformation pathway. Moreover, GLIO values in children were obtained for the first time.
A study was conducted in 2018 to analyze exposure to AFB1 and other mycotoxins among workers at a waste-sorting plant [60]. AFB1, enniatin B (EnB) and OTA, as well as 2'R-OTA, were detected and quantified through a multi-mycotoxin LC-MS/MS approach. The authors concluded that this study confirmed co-exposure and different possible exposure routes.

Risk Characterization
Toxicokinetic data are very useful for human health risk assessment. Unfortunately, in the field of mycotoxins, these data are very scarce in humans. Related to half-life values, they have only been established for some of the toxins. CIT has a short half-life in human blood of 9.4 h [56], whereas, for OTA, a long half-life of 35.6 days has been described [135] due to its high binding to plasma proteins [140]. For ZEA, these data are still unknown [141], although Mukherjee et al. (2014) [142], using a physiologically-based toxicokinetic model, estimated a half-life for ZEA of 11.89 h in young girls. The half-life for AFB1 in four human volunteers was 64.4 h [143]; although it should be taken into account that authors did not discriminate between AFB1 and its metabolites or conjugates [144]. For AFB1-lys, and due to lysine stability in human serum, a half-life of 2-3 months was estimated by Mupunga et al. (2016) [145]. These authors also stated that AFM1 has a short half-life, though no value was indicated. No information about the human toxicokinetics of DON and its derivatives, T-2, HT-2, FUS-X, NIV [45], Ens and BEA [146] or diacetoxyscirpenol (DAS) [147] was found.
The health risks associated with mycotoxin exposure arise from their toxicity. To minimize the risk to human health, several international bodies, such as the European Food Safety Authority (EFSA) or the Joint FAO-WHO Expert Committee on Food Additives (JECFA), have carried out health risk assessments for mycotoxins (or group of mycotoxins) ( Table 4) and have established health-based guidance values (HBGV), such as the TDI. Due to the carcinogenic risk associated with some of the mycotoxins, the International Agency for Research on Cancer (IARC) has also evaluated and has classified some of them as: i) carcinogenic to humans (Group 1), based on sufficient human data; or ii) possibly carcinogenic to humans (Group 2B), based on sufficient experimental/animal data but limited human epidemiological information. Due to the lack of experimental data or epidemiological information at the time in which the IARC evaluation was carried out, many of the mycotoxins were also classified as Group 3 (not classifiable as to its carcinogenicity to humans). It should be mentioned that for compounds known to be genotoxic and carcinogenic, such as AFs or STER, the general assessment is that exposure from all sources should be as low as is reasonably achievable. Indeed, in these cases, EFSA did not consider it appropriate to establish HBGV and therefore proposed the margin of exposure (MOE) approach in their risk assessments.
Health risk can be evaluated by means of EDI through food consumption data/exposure data and/or biomarker levels in human plasma/serum. The exposure related-data obtained are then compared to the TDI established for the mycotoxin (or group) under study [22,148]. A strategy that combines both approaches can also be adopted [40]. In the papers reviewed, OTA and AFB1-lys were the most frequently detected biomarkers in plasma and serum samples. They reported positive levels of 64.9% and 76.9% of the samples analyzed for these mycotoxins, respectively. The samples analyzed were taken from documented patients with different diseases (SAM, HCC, GBC and ASD) and from control individuals. Therefore, although low levels were detected, it is possible to conclude that the world population in general is exposed to AFB1 and OTA. Assessing the risk posed by both mycotoxins is thus of great interest.
The correlation between OTA concentration in plasma, C p (ng/mL), and the estimated daily intake (EDI), expressed as ko, can be calculated by means of the Klaassen equation (equation 1) [85,132] where Cl renal is the daily renal clearance and A refers to OTA bioavailability: ko (ng/kg .bw/day) = Cl renal C P /A After considering several assumptions to express the equation based on human data (plasma clearance of 0.99 mL/kg of body weight/day and an estimated A value of 0.5), the Klaassen equation can be expressed as follows [33,118]: According to the studies included in this review, it should be noted that mean OTA concentrations in blood samples did not vary considerably (around 1 µg/L) across the world (Table 3). EDIs were calculated based on Equation (2) using the mean and maximum OTA levels observed in studies that addressed OTA detection. Based on the mean OTA concentration, the calculated EDI values ranged from 0.28 to 2.40 ng/kg.bw/day (Figure 6), which were considerably lower than the TDI for OTA (14 ng/kg.bw/day). Only one of the maximum values detected (9.18 µg/L) [71] led to an EDI value higher than the TDI established. Therefore, although positive OTA levels were observed in 64.9% of the total plasma and serum samples analyzed, the EDI values obtained did not exceed the risk value established for human health. ng/kg.bw/day). Only one of the maximum values detected (9.18 µg/L) [71] led to an EDI value higher than the TDI established. Therefore, although positive OTA levels were observed in 64.9% of the total plasma and serum samples analyzed, the EDI values obtained did not exceed the risk value established for human health. AFB1 is a known carcinogenic agent in humans (Group 1) and can lead to growth suppression, immune system modulation and malnutrition, even at low concentrations [27,161]. Intake of a small amount, such as 1 ng/kg.bw day, is considered dangerous and toxic for human health [22]. This is of particular concern in children, since lower values than those indicated for adults can have lethal consequences [162]. Therefore, the measurement of AFB1 and/or its metabolites in biological matrices is crucial for assessing potentially dangerous exposure to this toxin.
The correlation between AFB1-lys levels in plasma or serum and dietary exposure to AFB1 was investigated in the literature. It was calculated that around 1.4-2.3% of ingested AFB1 is covalently bound to albumin [37]. Moreover, a strong correlation coefficient of 0.80 between AFB1-lys levels in serum or plasma albumin and dietary exposure to AFB1 has been reported [163]. For AFs, however, no clear relationship between plasma concentration and EDI has been established [65], a fact that has made it impossible to calculate the EDI value [144].
The AFB1 adduct concentrations (pg/mg albumin) reported by the papers reviewed were as follows: mean values ranged from 0.8 to 31.2, and maximum values ranged from 0.10 to 211. The worst scenario was observed in studies carried out in developing countries. The lowest values were those found in developed countries and in some control samples in Nigeria [49] and China [87].
To improve risk assessment, further studies should be carried out to monitor rarely studied toxins such as T-2 and HT-2, as well as other forms in which mycotoxins are found, such as glucuronide conjugates [22].

Conclusions
Mycotoxins are recognized as toxic compounds of great concern in the context of human health and the global economy. HBM of mycotoxin biomarkers is considered a good approach to obtain data that could help determine human exposure, assess risks and identify relationships between Figure 6. Comparison between estimated daily intake (EDI) of mean and maximum values and TDI for OTA exposure. Each dot corresponds to the EDI mean (blue) or EDI max (red) for the different exposure studies retrieved for OTA. A total of 15 studies were evaluated. The black line represents the most recent TDI established for OTA [136].
AFB1 is a known carcinogenic agent in humans (Group 1) and can lead to growth suppression, immune system modulation and malnutrition, even at low concentrations [27,161]. Intake of a small amount, such as 1 ng/kg.bw day, is considered dangerous and toxic for human health [22]. This is of particular concern in children, since lower values than those indicated for adults can have lethal consequences [162]. Therefore, the measurement of AFB1 and/or its metabolites in biological matrices is crucial for assessing potentially dangerous exposure to this toxin.
The correlation between AFB1-lys levels in plasma or serum and dietary exposure to AFB1 was investigated in the literature. It was calculated that around 1.4-2.3% of ingested AFB1 is covalently bound to albumin [37]. Moreover, a strong correlation coefficient of 0.80 between AFB1-lys levels in serum or plasma albumin and dietary exposure to AFB1 has been reported [163]. For AFs, however, no clear relationship between plasma concentration and EDI has been established [65], a fact that has made it impossible to calculate the EDI value [144].
The AFB1 adduct concentrations (pg/mg albumin) reported by the papers reviewed were as follows: mean values ranged from 0.8 to 31.2, and maximum values ranged from 0.10 to 211. The worst scenario was observed in studies carried out in developing countries. The lowest values were those found in developed countries and in some control samples in Nigeria [49] and China [87].
To improve risk assessment, further studies should be carried out to monitor rarely studied toxins such as T-2 and HT-2, as well as other forms in which mycotoxins are found, such as glucuronide conjugates [22].

Conclusions
Mycotoxins are recognized as toxic compounds of great concern in the context of human health and the global economy. HBM of mycotoxin biomarkers is considered a good approach to obtain data that could help determine human exposure, assess risks and identify relationships between diseases and mycotoxins. This creates new challenges in the field of mycotoxin research. In the present review, some of these challenges, together with other related aspects, have been identified.
Good biomarkers for each mycotoxin of interest should be described, including the ones for the so-called modified mycotoxins. Consequently, studies on the metabolism and toxicokinetics of mycotoxins are of vital importance. In addition, validated analytical methods should be developed. The current analytical trend is to simultaneously detect multiple mycotoxins in a single run with a view to saving time and reducing costs. The LC methodology coupled to several detectors, especially MS/MS and HRMS, has proven to be a useful analytical technique for multi-mycotoxin biomonitoring. The development of these methods requires adequate and, if possible, affordable standards and reference materials. New methodologies for sample treatment that reduce matrix effects are needed. Finally, guidelines for the validation of analytical methods should also be developed.
AFB1-lysine and OTA in plasma and serum have been the most widely studied biomarkers in recent years. For AFB1-lys detection, most articles proposed digestion with Pronase ® and a purification step by means of SPE before analysis with LC coupled to FLD, MS/MS or HRMS detectors. AFs such as AFB1, AFB2, AFG1, AFG2 and AFM1, CIT and ZEA have also been analyzed, but to a lesser extent. Mycotoxins such as T-2 and HT-2 were not studied in the papers reviewed.
A high percentage of the samples analyzed for OTA and AFB1 (mostly as AFB1-lys) presented some level of these mycotoxins; it can therefore be concluded that the general population is exposed to them. In the case of OTA, the EDI values calculated, based on either the mean or maximum values, were, in all cases, lower than the TDI value defined for this mycotoxin. For the other mycotoxins, equations that relate biomarkers of exposure and concentration in biological fluids or tissues need to be determined. This aspect is crucial for improving knowledge and the interpretation of the results obtained through HBM.
Several authors have searched for a relationship between diseases and mycotoxin levels in plasma or serum. Some relationships have been detected, but others remain unclear. It is not even clear whether the presence of mycotoxins in biological samples is the cause (or a contributor, along with other factors) of a disease, or, on the contrary, whether its presence is the result of the metabolism pathway alteration produced by the illness itself, which could increase mycotoxin levels in ill people.
Based on all of the above, it is possible to conclude that new avenues are emerging, and much more research is required on the interesting and important topic of mycotoxin HBM.

Materials and Methods
A systematic review strategy was carried out based on PRISMA Statement [164]. For this purpose, the PubMed and Web of Science databases were used. First, two general searches were performed with the aim of obtaining a general overview of the subject, as shown in Figure 7 (bold). Additional searches were then carried out with more specific medical subject heading (MeSH) terms: "biomonitoring" OR "exposure" OR "review" OR "disease" OR "detection" OR "analytical methods" OR "presence" (added to the first general search); and "risk assessment" (added to the second search). A total of 2388 articles were obtained with this search. Inclusion criteria were: full-text was available, papers addressed the research topic, papers described analytical methodologies, articles were written in English and articles focused on human biological fluids, especially plasma, serum and blood. Articles that did not meet these criteria were excluded, as well as duplicated records. Other sources were also used, such as the reference section of the papers reviewed, the IARC, the European Commission and the EFSA. In total, 164 articles were evaluated (Figure 7).   Funding: This work was supported by the Spanish "Ministerio de Economia, Industria y Competitividad, Agencia Estatal de Investigación" (AGL2017-85732-R) (MINECO/AEI/FEDER, UE).

Conflicts of Interest:
The authors declare that there are no conflicts of interest.