Unreported Aflatoxins and Hydroxylate Metabolites in Artisanal Oaxaca Cheese from Veracruz, Mexico

Aflatoxins (AFs) are toxic secondary metabolites of the fungi Aspergillus flavus, A. parasiticus and A. nomius. The fungi produce these AFs in cereals, oilseeds and spices. AFs have damaging effects on all organisms, including humans, and their symptoms can be classified as acute (vomiting, hemorrhage and death) or chronic (immunodepression, Reye syndrome, Kwashiorkor, teratogenesis, hepatitis, cirrhosis, and various cancers). The common AFs (AFB1, AFB2, AFG1, AFG2) are metabolized in the liver or by microbes that produce hydroxylates (AFM1, AFM2, AFP1) and aflatoxicol (AFL), which makes them soluble in water. This means that AFs can be excreted in fluids such as milk or urine, and AFs are not destroyed in the process of making cheese.


Introduction
Cheese is an economically important commodity worldwide. In Mexico, 76,696 tons of Oaxaca-type cheese were produced in 2005, with a value of 2,700 million Mexican pesos [1,2]. Most of the information about aflatoxins (AFs) in cheese is related to industrial production and sale through formal commercialized channels. However, most of the Oaxaca-type cheese consumed in Mexico is handmade artisanally, and there have been no reports about AFs in cheese and the quantities sold in Mexico.
The State of Veracruz is the sixth largest producer of milk in Mexico [3], and 53% of the total milk produced in the state is used without pasteurization to produce artisanal cheeses [4], which are sold in large cities such as the Port of Veracruz. One of the main artisanal cheeses produced in this region is Oaxaca-type cheese, which is made in the same way throughout the country. The process begins with warming milk to temperatures between 18°C and 25°C; the milk is then heated to 38°C, and rennet is added (9-12 mL for 100 L of milk). The milk is then acidified with acetic acid at pH 5.5, and upon curdling the curds are cut into 2-cm squares. The curds are left to rest for 25 min and are later shredded by hand and left to acidify for 20 min. The whey is drained, and the curds without whey are melted when mixed with hot water (60°C). The product is stretched by hand to form threads of 3 cm to 6 cm wide and then cooled with water (18°C). The cheese is left to drain, and salt is added (11 g to 50 g salt per 1 kg of threads). Finally, ball hanks are formed with the cheese threads [5].
Aflatoxins (AFs) are toxic secondary metabolites that chemically correspond to bis-dihydro-furanecoumarins and have well-known physicochemical properties [6]. AFs are mainly produced by the fungi Aspergillus flavus, A. parasiticus and A. nomius [7]. The common AFs found in cereals, which are present in balanced cow feed, are aflatoxin B 1 (AFB 1 ) and aflatoxin B 2 (AFB 2 ), which have blue fluorescence, and aflatoxin G 1 (AFG 1 ) and aflatoxin G 2 (AFG 2 ), which have green fluorescence [8].
The acute symptoms of AFs include vomiting, miscarriage, hemorrhage, diarrhea and death, and chronic symptoms include immunosuppression, fetal malformation, hepatitis B and C, cirrhosis, and carcinoma of the liver [9,10], cervix [11], colorectal system [12], breast [12] and pancreas [12]. AFs are considered potent carcinogens, and the International Organization for the Research on Cancer (IARC) classified them as Grade I in humans [13].
AFs can be present in balanced cattle feed [14] in countries with tropical weather, where high humidity and warm temperatures in storage warehouses facilitate fungal growth, as well as in agricultural Page 2 of 9 and unregulated local markets [15]. Cheese is an important source of nutrients for humans, and it is frequently contaminated with AFs [16]. AFs are present in the cereals and oilseeds that are used as ingredients in feed and silage for cattle [17]. They are common in countries where storage, harvest and climate conditions are suitable for fungal growth and in countries without food regulation legislation [17,18].
When dairy cattle consume fodder contaminated with AFB 1 or AFB 2 , these toxins are rapidly absorbed. The animal's liver and the microbial metabolism mitigates the damaging effects of the toxins by introducing a hydroxyl (OH -) into the AF molecule, forming hydroxylate metabolites that allow the toxins to be dissolved in water and expelled from the body via urine or milk. In this way, AFB 1 , AFB 2 , AFG 1 and AFG 2 from balanced feed of ruminants or from fodder are biotransformed into AFM 1 , AFM 2 , AFP 1 and aflatoxicol (AFL) [19], and these less toxic but still carcinogenic hydroxylate metabolites are secreted in cow's milk ( Figure 1).

Aflatoxins in cheese
AFs bind to proteins such as milk casein via a hydrophobic interaction [20]. Therefore, AFM 1 is present in cheese when contaminated milk is used. AFM 1 distributes in a 40% to 60% ratio between curd and whey, depending on the cheese-making method [21]. AFM 1 can withstand temperatures up to 320°C before decomposing and is resistant to thermal treatments, such as pasteurization, ultrapasteurization and acidification, that are used during the processing of cheeses [22,23]. Although AFM 1 is less toxic than AFB 1 , it is still carcinogenic, and it is frequently reported in dairy products. Nonetheless, other AFs certainly contribute to the risk associated with AFs due to the high consumption of dairy products. Therefore, many countries have established regulations for the maximum tolerable levels of AFs in milk and dairy products [24]. AFM 1 , AFM 2 , AFB 1 and AFL have been reported in Mexican milk [25,26]. Oaxaca-type cheese has economic, cultural and alimentary importance; therefore, the detection of AFs in artisanal cheeses commercialized in Veracruz is necessary. The purpose of this research was to find other AFs (AFB 1 , AFB 2 , AFG 1 , AFG 2 ) and the hydroxylate metabolites AFM 1 and AFM 2 for comparison as well as AFP 1 and AFL, which have not been reported, and to discuss their importance as carcinogens.

Sampling
The study consisted of 30 samples of 750 g of Oaxaca-type artisanal handmade cheese purchased in groceries and markets in the City of Veracruz. The cheese samples originated from 5 townships (Acajete, Medellin, Tlalixcoyan, Soledad and Veracruz) of the State of Veracruz, Mexico ( Figure 2). These townships practice dual-purpose cattle raising for both meat production and small artisanal cheese making with no sanitary legislation. A Matlab algorithm was used to randomly select the places from which the samples were purchased in the City of Veracruz. The cheeses were refrigerated immediately after sampling and were subjected to a drying process for a period of less than 12 hours. Samples of Oaxaca-type cheeses were purchased in March 2016, which is in the dry season when the cows are fed nutritionally balanced feed; during the rest of the year, cows typically eat grass.
Each cheese sample was manually unthreaded, the cheese samples were placed in a tray drier, and the dry samples were stored frozen until AF extraction and chemical analysis were performed.

Chemical extraction method for Aflatoxins
The R-Biopharm [27] method has been recommended for use

Page 3 of 9
with Total aflatoxin Easi-Extract Immunoaffinity Columns (IAC) (R-Biopharm Rhône Ltd., Glasgow, Scotland, UK). This method was performed according to the following protocol.
First, 15 g samples of dry, ground Oaxaca-type cheese were blended (Waring ETL laboratory blender 7010S model WF 2211214, Torrington, CT, USA) for 2 minutes at high speed with a mixture of 100 mL of MeOH/water (80:20 v/v) and 2 g NaCl to clarify the extract. The mixture was centrifuged at 4500 rpm for 15 min, and an amount of supernatant equivalent to 1 g of sample was dissolved in phosphatebuffered saline (PBS) at pH 7.4 at a proportion of 1:4(v/v) and homogenized for 1 minute in a vortex. Before adding the samples, each IAC was equilibrated with 20 mL of PBS at pH 7.4 applied at a flux of 5 mL/min. The buffered sample was passed through the IAC, and AFs were eluted using 1.5 mL of HPLC-grade MeOH followed by 1.5 mL of distilled water with reflux. The eluate was dried at 40°C in an oven (F135A Novatech Model, México City, Mexico) and then derivatized.

Derivatization
Derivatization is a process to increase the AF fluorescence [28,29] of AF standards to make calibration curves and to quantify the AFs in cheese samples Figure 3. The derivatization reaction with trifluoroacetic acid is the transformation of AFB 1 and AFG 1 that are less fluorescents, in their hemiacetals B 2a abd AFG 2a that are highly fluorescent. AFB 2 and AFG 2 are not affected by this reaction due to their saturated structure [28].
Eight dry AF standards (AFB 1 , AFB 2 , AFG 1 , AFG 2 , AFM 1 , AFM 2 , AFP 1 and AFL from Sigma-Aldrich; St. Louis, MO, USA) that were used to determine the AFs' linearity and percentage of recovery validation were dissolved in 200 µL of HPLC-grade acetonitrile (ACN), and 800 µL of derivatization solution was added. The derivation solution was prepared with 5 mL of trifluoroacetic acid (Sigma-Aldrich, St. Louis, MO, USA), 2.5 mL glacial acetic acid (Merck, Naucalpan, Estado de Mexico, Mexico) and 17.5 mL deionized distilled water and then vortexed (Vortex G-560, Bohemia, NY, USA) for 30 seconds. The vials containing the dry eluates were heated in a vapor bath at 65°C for 10 min. The derivatized samples were cooled to room temperature, and triplicate 60-µL samples were analyzed by HPLC with fluorescence (HPLC-FL).

Validation of the extraction method
The validation of the analytical method and the analysis of the 30 Oaxacatype cheese samples were performed using known parameters [30].

Linearity of the system (Calibration curves)
Solutions with different concentrations of AFs were prepared from a stock concentration of 1000 ng AFM. The 0.25 mg AFM standards were diluted with benzene:acetonitrile (98:2 v/v), following a previously reported methodology [31], so that the pure AFs did not decompose.
a. The spectrophotometer (Genesys 10 UV Thermo Electron Corporation. Madison, WI, USA) was calibrated before the experiments to measure the absorbance of the AFM standard solutions at 357 nm.
b. Different formulas [31] were applied to calculate 1000 ng stock solutions of each AF concentration: c. Twelve concentrations (0.01, 0.05, 0.1, 0.5, 1, 2, 4, 8, 16, 32, 64 and 128 ng) of the 8 different AFs were independently created from the 1000 ng stock solution. These standard dilutions were then used to plot the analytic signal (area below the curve of each chromatographic peak) against the AF concentrations. The curve equation and statistical parameters were obtained. The slope value (b 1 ), ordinate to origin (b o ), determination coefficient (R 2 ), confidence interval for the slope to origin (IC(β)), variation coefficient percentage (% CV), standard deviation (SD), and the LOD and LOQ were calculated using Excel 2003.

LOD and LOQ
The LOD of the equipment was established in relation to the noise in the chromatogram. The LOD is equal to the AF concentration that gives a signal three times greater than the noise. The LOQ equals the AF concentration that is 10 times greater than the noise [32].

Recovery percentages
The recovery percentage is a measure of the accuracy of the method and expresses the proximity between the theoretical and experimental values. The arithmetic average, standard deviation, percentage of variation coefficient and confidence interval were calculated. To obtain accurate measurements, the AFs of the samples of dried, ground Oaxaca-type cheese, in 1 g aliquots, diluted in PBS (1:4 v/v), were individually spiked with three different concentrations (5, 20 and 40 μg kg -1 ) of the eight individual AF standards (AFB 1 , AFB 2 , AFG 1 , AFG 2 , AFM 1 , AFM 2 , AFP 1 ) and AFL; one aliquot without spiked AF was used as the control, which gave the basal contamination level. The samples were individually processed using the R-Biopharm extraction method [27]. AFs were purified and concentrated using an IAC, derivatized, and quantified by HPLC-FL, and the percentage of recovery for each AF was obtained. After the derivatization mixture was cooled to room temperature, triplicates of 60 μL of each sample were injected into the HPLC-FL method.

HPLC-FL quantitation
The HPLC-FL chromatographic system was an Agilent Series 1200 HPLC (Agilent Technologies, Inc., Santa Clara CA, USA) and consisted of an isocratic pump (Model G1310A); a fluorescence detector (Model G1310A Series DE62957044, Agilent Technologies, Inc., USA), which was set to an excitation wavelength of 360 nm and to two emission wavelength maxima of 425 nm for AFB 1 , AFB 2 , AFM 1 , AFM 2 , and AFL and 450 nm for AFG 1 , AFG 2 and AFP 1 ; and an autosampler (G1329A Series DE64761666). The chromatographic column was a VDS Optilab VDSpher 100 C18-E 5 µm 250 mm × 4.6 mm that was maintained at room temperature (22°C) with a mobile phase of water:ACN:methanol (65:15:20 v/v/v) that was degassed for 30 min under vacuum and added at a flux of 1.0 mL/min. The injection volume was 60 μL.
In the chromatograms the same letters mean that the samples were statistically the same, and were analyzed in a variance analysis with Tukey Test (P<0.05). AFB 1 and AFL were the most abundant AFs. The AFB 1 was not a product of the cow metabolism but rather indicated that maize starch had been added during the manufacturing of the cheese. This addition is legal [33] and accepted for fresh cheeses, but it is not good practice from the point of view of AF contamination. The amount of AFB 1 was very high, and it is the most carcinogenic AF. AFB 1 transforms itself into AFL, which is the interconverting hydroxylate form [34]. These 16. Variance analyses with the Tukey test at 95% were performed in triplicate, considering each cheese as an experimental unit. The graphs showing the data from the Tukey test and the standard deviations were produced using Kaleida Graph version 3.5. Kruskal-Wallis statistical analysis was used to find significance and differences of AFs. The Wilcoxon Rank Sums test was performed to find differences for each pair of samples and Bonferroni corrections of the samples.

Validation parameters
The Limit of Detection (LOD) in ng g -1 and recovery percentage *Decreasing order of abundance of AFs in Oaxaca-type cheese:1) AFL, 2) AFB 1 , 3) AFM 1 , 4) AFG 2 , 5) AFM 2 , 6) AFP 1 , 7) AFG 1 and 8) AFB 2 .    two toxic metabolites are more frequent in Oaxaca-type cheese than are AFM 1 and AFM 2 . Other AFs, such as AFP 1 and AFG 1 , were present in trace amounts [19], and AFG 2 appeared more frequently. AFL can be formed through the enzymatic or synthetic reduction of AFB 1 , and Page 6 of 9 it has high toxicity and carcinogenicity [34]. Although the toxicity of AFL is 18 times lower than that of AFB 1 , both molecular structures have similar potency to form an exo-epoxide analogue that can bind to DNA [34]. AFL interconverts with AFB 1 , has electrochemical properties like those of AFB 1 , and these compounds have been experimentally demonstrated to have high carcinogenicity and toxicity. AFM 1 contamination was in third place in Oaxaca-type cheese samples (Table 2), consistent with the results obtained for other kinds of cheese, such as cream cheese [35], white pickled cheese [36], sheep curd [37], Grana Padano cheese [38], parmesan [39], Turkish kashar cheese [40], and Serbian hard cheese [41]. AFM 2 contamination has been less frequently reported. There have been several studies [42] on carryover from cows fed AFB 1 -contaminated rations to AFM 1 in milk. The degree of toxicity and carcinogenicity of AFs is in the following order: B 1 >G 1 >B 2 >G 2 .
We performed a Kruskal-Wallis analysis to find differences in the concentrations of AFB 1 , AFB 2 , AFG 1 , AFG 2 , AFP 1 and AFL among the 30 samples, as shown in Table 3. There were statistically significant differences among the samples for AFB 1 and AFP 1 . For AFG 2 , the differences were not statistically significant at 5%, but if a different significance value is considered (10%, for example), the differences among the samples may be significant. Table 4 presents the Wilcoxon Rank Sums test to find the differences for each pair of samples for AFB 1 . We found that 21 of the 25 samples had concentrations that were not significantly different from zero, but when the Bonferroni correction was applied, these samples differed from the remaining four samples, 18 to 21.
For AFP 1 , we found that 13 of the 15 samples had concentrations that were not significantly different from zero. When the Bonferroni correction was applied, these samples differed from the remaining two (samples 16 and 10). It is important to note that most samples had at least one replicate different from zero. Samples 2 and 5 had two replicates different from zero, and samples 10 and 16 had three replicates different from zero. Table 5 shows the statistics for AFP 1 in the 15 samples that had values different from zero.

Page 7 of 9
Bonferroni correction, we found that all samples with observations different than zero were equal and significantly different than zero.
AFL is produced in several fungi: Aspergillus flavus, A. parasiticus, A. niger, Eurotium herbariorum, Rhizopus spp. and other non aflatoxicogenic A. flavus [43]. Reduction of the 1-keto group of AFB 1 produces AFL [44]. AFL is equally carcinogenic as AFB 1 , so its formation is not a significant detoxification mechanism [45,46]. AFL has approximately 70% the mutagenicity of AFB 1 [47], and it has two forms, A(Ro) and B, both of which are produced from the biological reduction of AFB 1 and mainly by Tetrahymena pyriformis, Dactylium dendroides and Rhizopus spp. AFL A is 18 times less toxic than AFB 1 in the duckling biliary hyperplasia assay, and the biological activity of AFL B is unknown [48,49]. AFL is the major metabolite of AFB 1 in many plants and animals, and it has been detected in milk [26,50], fermented dairy products [51], cereals and nuts [52], eggs [53], blood [54,55], human brain [56], the sera and liver of humans with kwashiorkor and marasmic kwashiorkor in Ghana and Nigeria [57][58][59][60], human urine [61], urine of heroin addicts [58], a breast-fed infant with neonatal hepatitis [62], the muscle of broiler chickens fed with contaminated diets [63], and poultry fed chronic low doses of mycotoxins, with the liver having the highest levels [64]. AFB 1 , AFM 1 and AFL accumulate in the tissues and urine of calves [65,66]. AFL-DNA adducts that were produced in vivo were identical to those produced by AFB 1 and had similar molecular dosimetry responses and toxicity to the target organ [67]. Regarding DNA adduction and hepatocarcinogenicity in rainbow trout, the tumorigenic potencies were AFB 1 =1.00, AFL=0.936, AFM 1 =0.086. AFL is a more potent toxin than AFM 1 , which can reconvert with AFM 1 , becoming AFL M 1 [68]. AFL-induced hepatocellular carcinomas in rats and fish have a lower tumor incidence than those induced by AFB 1 [69]. There is an interconversion of AFB 1 and AFL, mediated by intracellular enzymes in rat blood [70]; guinea pigs [71]; sharks, which reconvert 30% of AFL to AFB 1 [72][73][74]; and cultured human epidermal cells [75]. AFL converts into AFB 1 , which is the most carcinogenic and toxic of all AFs [76]. AFL is oxidized readily back to AFB 1 , so it can serve as a 'reservoir' for AFB 1 in vivo, thereby prolonging the effective lifetime in the body [77]. If pH has a role in the interconversion AFB 1 -AFL, it could act in the normal human digestion of milk, where pepsin lowers the pH. The isomerization of AFL to AFB 1 was observed in culture media with a low culture pH [76].
The genotoxicity of AFM 1 has been demonstrated by in vitro and in vivo experiments. The carcinogenic potency of AFM 1 is 2% to 10% weaker than that of AFB 1 [78]. Therefore, the Food Safety Commission of Japan in 2013 [78] concluded that the AFB 1 that is present in animal feed is extremely unlikely to affect the health of humans who have consumed contaminated milk or other livestock products. However, AF and the hydroxylate metabolites are also genotoxic carcinogens and are more likely to be found in livestock products, so AFB 1 contamination in feed and AFM 1 contamination in milk need to be reduced as much as possible. In particular, attention should be paid to the fact that the intake of milk per 1 kg of body weight is higher in infants than in other age groups [78].
Risk assessment parameters for AFB 1 , AFL and AFM 1 have been compared [79]. The virtually safe dose for AFL was 1.7 times higher than that for AFB 1 .
The incidence of hepatocellular carcinoma in rats and fish dosed with AFL was lower than that in animals treated with AFB 1 at the same dosage. AFL in milk might still be a health hazard, particularly for infants whose staple diet is milk-based. AFM 1 was not the most abundant AF, and the risk increases when the AFL contamination in milk is added. AFM 1 is possibly carcinogenic to humans and was classified as Group 2B by the IARC (1997) [69]. AFM 1 has been found in Mexican milk [25], so its presence in cheese was not unexpected, where most AFs were metabolized to AFL. Autumn milk was significantly more contaminated with AFL (p<0.0002). AFB 1 had no significant correlation with season, and it is not clear if the presence of vegetable oil helped to decrease the AFL contamination [26]. AFB 1 was generally present in milk at trace levels (0.05 mg L -1 to 0.42 mg L -1 ) in 5.2% of the 290 samples [26] and is not considered a health risk, but cheese had more concentrated amounts (0.04 to 49.2 ng g -1 ), with an average of 11.2 ng g -1 in the 30 samples and can be considered a health risk.
The hydroxylates AFP 1 and AFL are not accepted as toxicologically important in many countries. Polish and European Union legislations (Commission Regulation No. 152/98) agree that all food should be free from AF [80]. Oltipraz was shown to reduce AFB 1 adduct biomarkers [81] and inhibit AFM 1 production by bovine hepatocytes [82], so it can be used to lower the risk related to cheese consumption. It is necessary to balance the availability of milk in relation to the health risk, not only for cancer but also for other diseases, such as immune suppression, hepatitis and cirrhosis. This fact makes mycotoxin regulation difficult and very incomplete.
AFs are recurrent and occasionally unavoidable contaminants of milk, cereals and oilseeds, and their thermal stability rules out both pasteurization and ultrapasteurization as effective control methods. The best control strategy is to keep raw materials and feed under obligatory mycotoxin regulation. In the case of cheese, it is recommended not to add maize flour during the manufacturing process.

Conclusion
Although the legislation regarding maximum tolerance levels has attempted to decrease the level of AFM 1 contamination in cheeses and although there is no direct evidence of human toxicity resulting from the consumption of cheese contaminated with AFs, the problem of ingesting AFB 1 and AFL is still present in fresh cheeses, such as the artisanal Oaxaca cheese.