Determination of Aflatoxins M1, M2, B1, B2, G1, G2 and Ochratoxin A in Infant Formulas from Brazil Using a Modified QuEChERS Method and UHPLC-MS/MS

The aim of this study was to determine the contamination levels by aflatoxins M1, M2, G1, G2, B1 and B2 and Ochratoxin A in 123 infant formula powder samples from the metropolitan region of Rio de Janeiro, Brazil. A sensitive method using a modified Quick Easy Cheap Effective Rugged and Safe (QuEChERS) method and ultra-high performance liquid chromatography-tandem mass spectrometry (UHPLC-MS/MS) was validated for application in sample analyses. Ochratoxin A and aflatoxins B1, B2, G1 and G2 were not detected in the analyzed samples. Aflatoxin M1 was detected in 18 (14.6%) of the analyzed samples and quantified in eight (6.5%), at concentration levels ranging from 0.016 to 0.057 µg kg−1 and an average concentration of 0.031 µg kg−1. The aflatoxin M1 concentrations found in three of the analyzed samples (0.040, 0.044 and 0.057 µg kg−1) exceed the limit established by European Union regulations (0.025 µg kg−1). The estimate daily intake (EDI) of aflatoxin M1 for infants up to 12 months ranged from 0.39 to 1.53 ng kg−1 body weight day−1. The contamination results demonstrate the importance of carrying out monitoring studies to support aflatoxin M1 regulations for infant formulas in Brazil.


Introduction
Mycotoxins comprise a diverse group of toxic compounds produced as secondary metabolites of certain fungi and are found as contaminants in food and animal feeds worldwide. Aflatoxin B1 and naturally occurring aflatoxin mixtures, which include aflatoxins B1, B2, G1 and G2, are classified by the International Agency for Research in Cancer as human carcinogens, while aflatoxin M1 and ochratoxin A are classified as possible human carcinogens (IARC 1993). Aflatoxins M1 and M2 are hydroxylated metabolites eliminated into the milk of lactating animals that consume aflatoxins B1 and B2 through contaminated feed (Peraica et al. 1999;Fink-Gremmels 2008).
Exclusive breast-feeding is considered the ideal nutritional practice for infants up to six months, with the gradual introduction of complementary foods recommended after this age (WHO 2001;Brasil 2008). Despite the unquestionable nutritional benefits of this practice, complementary foods, including infant formulas, may expose infants to chemical contaminants, including mycotoxins. Infants are considered more susceptible than adults to chemical contaminants and the safety of foods consumed by this population is a priority public health issue (Sherif et al. 2009;Mielech et al. 2021).
In Brazil, many studies indicate the high occurrence of aflatoxin M1 in milk (Oliveira et al. 2006;Shundo et al. 2009), and the presence of aflatoxin B1 in this matrix has also been recently demonstrated (Scaglioni et al. 2014). However, scarce studies have investigated the presence of mycotoxins in infant formulas in Brazil, such as Iha et al. (2013), Ishikawa et al. (2016) and Tonon et al. (2018) that assessed these contaminants in seven, 16 and 26 infant formula samples, respectively.
The so-called QuEChERS-Quick Easy Cheap Effective Rugged and Safe-method, originally used to analyze pesticides in fruits and vegetables (Anastassiades et al. 2003), has been more recently extended to determine mycotoxins in food (Desmarchelier et al. 2010(Desmarchelier et al. , 2014Romero-Gonzalez et al. 2011;Tamura et al. 2011). To the best of our knowledge, the QuEChERS extraction method has not yet been applied in the determination of mycotoxins in infant formula.
In this context, the aim of the present study was to investigate the presence of aflatoxins M2, M1, G2, G1, B2, and B1 and ochratoxin A in several brands of infant formulas sold in the markets of Rio de Janeiro, Brazil. The samples were analyzed using a modified QuEChERS method and ultra-high performance liquid chromatography tandem mass spectrometry (UHPLC-MS/MS).

Standard Solution Preparations
An ochratoxin A stock solution (40 µg mL −1 ) was prepared in a toluene-acetic acid (99:1 v/v) mixture; stock solutions of each aflatoxin (10 µg mL −1 ) were prepared in acetonitrile. The concentrations of the stock solutions were determined by UV spectrophotometry (Horwitz and Latimer 2005). More diluted individual solutions (200 ng mL −1 ) of mycotoxins other than ochratoxin A were prepared by dilution of the 10 µg mL −1 stock solutions with acetonitrile. When preparing a 200 ng mL −1 solution of ochratoxin A, an aliquot of a 40 µg mL −1 ochratoxin A stock solution was evaporated and an appropriate volume of acetonitrile was added to obtain a 200 ng mL −1 stock solution. All stock solutions were stored at -20 ºC. Aliquots of the aflatoxin and ochratoxin A solutions (200 ng mL −1 ) were combined and the volume was adjusted with acetonitrile to obtain an intermediate multi-mycotoxin solution at a concentration of 20 ng mL −1 . These solutions were prepared weekly and stored at -18 ºC.

Infant Formula Samples
One hundred and twenty-three infant formula powder samples from five different producers, representing thirty-one different brands, were acquired randomly in supermarkets and drugstores by the metropolitan region of Rio de Janeiro Public Health Surveillance System, in RJ, Brazil. Ninetytwo samples were acquired from May 2011 to August of 2013 and thirty-one samples from May 2018 to July of 2020. The samples were classified according to the label information supplied by the producers as follow-up (n = 52), starter (n = 49), special (n = 17), and toddler (n = 5) formulas. Aliquots (50 g) of the commercial samples were transferred to plastic containers and stored at -20 ºC until analysis. The samples were analyzed within six months after the end of each sampling.

Sample Preparations
For sample preparations, the method reported by Sartori et al. (2015) with some modifications was applied. To increase the sensitivity of the selected method, the sample extract aliquot volume (acetonitrile phase) evaporated to dryness (10 mL) was twice the aliquot used in the original method (5 mL). Briefly, ultrapure water (15 mL) was added to 1.5 g of each sample, weighed in a 50 mL centrifuge tube. The tube was then vigorously shaken until complete dissolution of the sample. Hexane (10 mL) and 15 mL of acetonitrile containing 1% (v/v) acetic acid were added, and the contents of the tube were shaken for 30 s. Subsequently, magnesium sulphate (6 g) and sodium chloride (1.5 g) were added and the tube was immediately shaken on a vortex (IKA Works) for 1 min and then centrifuged (HIMAC CF 7D2, Hitachi) at 3,000 rpm for 7 min. A 10 mL aliquot of the acetonitrile phase was evaporated to dryness under a gentle flow of nitrogen at 45 ºC (Turbo-Vac LV) and the residue was then dissolved in 1 mL of a methanol/water (1:1, v/v) solution. Finally, the obtained solution was filtered through a 0.22 μm filter before injection.

UHPLC-MS/MS Analysis
Liquid chromatography was performed on an ACQUITY UPLC™ system (Waters). A BEH C18 column (100 mm × 2.1 mm i.d., 1.7 μm particle size) was used as the stationary phase. The column temperature was maintained at 35ºC. Methanol (Phase B) and an aqueous ammonium formate 5 mM/acetic acid 1% solution (Phase A) were used as mobile phases. Mobile phase B was increased from 10 to 100% in 4 min and was then held constant for 1.5 min.
The system was then re-equilibrated for 2 min in the initial mobile phase composition. The flow rate was set at 0.3 mL min −1 . The injection volume was of 10 μL in full loop mode.
The detection was performed using a tandem quadrupole mass spectrometer (Waters, Quattro Premier™ XE) equipped with an electrospray ionization (ESI) source operated in the positive mode. The source parameters were capillary voltage 3.5 kV, extractor voltage 3 V, rf lens 0.1 V, multiplier 750 V, desolvation temperature of 350 ºC, source temperature of 120 °C. Nitrogen was used as the cone and desolvation gas at a flow of 50 L h −1 and 750 L h −1 , respectively. Argon was used as the collision gas at a pressure of 4 × 10 -3 mbar. The two ion transitions selected for each mycotoxin and the acquisition conditions are displayed in Table 1. The mycotoxins were allocated into three acquisition time windows as follows: aflatoxin M2 (1); other aflatoxin assessments (2) and ochratoxin A (3). For ochratoxin A and aflatoxin M2 the dwell time was of 200 ms and for all other aflatoxins, 30 ms. Interchannel delay and interscan delay were both 5 ms.

Validation of the Analytical Method
Method validation was performed by evaluating the following analytical performance parameters: selectivity, matrixeffect, linearity, trueness (recovery), precision (intra-day and inter-day repeatability), limit of detection (LOD) and limit of quantification (LOQ).
The selectivity of the method was evaluated by analyzing blank matrix samples of whole powdered milk and different infant formulas types. Mycotoxin identification was performed by comparing the analyte retention time in the samples with those of the standard solutions. Confirmation was performed by comparing the signal intensity ratios of the two ion transitions of each analyte in the sample with those of the standard solutions.
To investigate matrix effects, whole powdered milk and infant formulas from different producers available in the local market (starter milk-based infant formula, n = 4; follow-up milk-based infant formula, n = 3; lactose-free infant formula, n = 2) were selected. Calibration curves for the target compounds in matrix extracts (matrix-matched calibration) and in methanol/water (1:1 v/v) were prepared at four concentration levels, ranging from 0.1 to 1.5 ng mL −1 ; and each solution was then analyzed in triplicate. The calibration curve slopes were compared by an analysis of covariance (ANCOVA), considering a significance level of 5% (García-Campaña et al. 1997). Prior to performing the ANCOVA, the homogeneity of the residual variances of all the calibration curves was checked using a modified Levene test (Brown and Forsythe 1974).
The linearity of all calibration curves used in routine analyses was evaluated using matrix-matched calibration curves at four concentration levels, between 0.1 and 1.5 ng mL −1 and the solutions were analyzed in triplicate. The ordinary least squares method was applied for the elaboration of the  Souza and Junqueira (2005). As assumptions for regression analysis, the homocedasticity, independency and normality of the regression residuals were verified. Initially, the outliers were successively investigated by the Jacknife standardized residuals test (Belsley et al. 1980), until non-detection or a maximum exclusion of 22.2% in the number of original results (Horwitz 1995). The homocedasticity of residuals was verified by a modified Levene test (Brown and Forsythe 1974). The independency of residuals was verified by the Durbin-Watson statistic (Durbin and Watson 1951). The normality of residuals was verified by Ryan-Joiner test (Ryan and Joiner 1976). The regression significance and the lackof-fit were performed by an analysis of variance (ANOVA) (Draper and Smith 1998). Trueness and intra-day precision were evaluated by recovery studies using follow-up infant formula spiked with the investigated mycotoxins at three concentration levels (0.05, 0.1 and 0.15 µg kg −1 ), with four replicates for each level. Intermediate precision (inter day precision) assessments were performed by the analysis of four spiked samples (0.05 µg kg −1 ) analyzed on different days.
Infant formulas spiked with all the target mycotoxins at 0.05 µg kg −1 were used to calculate the limits of detection (LOD) and limits of quantification (LOQ), considering signalto-noise ratios of 3 and 10, respectively.

Estimated Daily Intake (EDI) of Aflatoxin M1
The EDI for aflatoxin M1 for infants was calculated based on the approach applied by Ishikawa et al. (2016), considering the exclusive use of infant formulas. Reconstituted infant formula consumption of 590, 642, 560 and 452 g, for infant ages one week, one month, six months and 12 months were considered, respectively (Brasil 2005). The average weights considered for male and female were, respectively, 3.3 and 3.2 kg for infant ages one week, 4.5 and 4.2 kg for one month, 7.9 and 7.3 kg for six months, and 9.6 and 8.9 kg for twelve months (WHO 2006). The International Programme on Chemical Safety/ Global Environment Monitoring System criteria (replacing all non-detectable results to LOD and all non-quantified results to LOQ), used when more than 80% of the samples present results are lower than the LOD, was adopted to calculate the average aflatoxin M1 concentration (0.0083 µg kg −1 ) in reconstituted infant formulas (IPCS/GEMS 1995).

Validation of the Analytical Method
In the selectivity studies, absence of interference signals eluting at the same retention time of the selected mycotoxins was verified for all the evaluated matrices. Figure 1 displays a chromatogram of a follow-up formula spiked with the target compounds at 0.15 µg kg −1 (concentration of infant formulas reconstituted with water). Confirmation was performed by comparing the signal intensity ratios of the two ion transitions of each analyte in the sample with those of the standard solutions (Table 1).
The matrix effects results obtained for the target compounds in different matrices are displayed in Table 2, where positive values indicate signal enhancement and negative values indicate signal suppression. The homogeneity of the residual variances of all the calibration curves was confirmed (p-values > 0.05). For ochratoxin A and aflatoxin M1, no significant differences between the slopes of the calibration curves for all evaluated matrices and the slope of the calibration curve in solvent was observed (p-values > 0.05). An absence of matrix effects for these compounds in all studied matrices was, thus, observed. However, significant differences between the slopes of the calibration curves prepared in solvent and matrices for at least two of the target aflatoxins in each studied matrix were observed (p-values < 0.05). A strong suppression of the analytical signal was observed for the target mycotoxins in soy-based infant formula matrix. Therefore, the analytical method was considered inadequate for the determination of target mycotoxins in soy-based infant formula samples.
In the linearity studies, the normality of the residuals and the independency of the residuals for all the calibration curves were confirmed (p values > 0.05). A high regression significance (p values < 0.001) and non-significant lack of fit (p values > 0.05) were found, attesting the linearity of the calibration curves. The resulting correlation coefficients for all calibration curves were always higher than 0.99.
The trueness and precision results are displayed in Table 3, where the trueness is expressed by the recovery values and precision by the relative standard deviation (RSD). The recovery values ranged from 75 to 111%, with RSD lower than 12% for all investigated mycotoxins. The RSD for intermediate precision (inter day) was also lower than 12%. The results were evaluated according to the European Commission Decision 2002/657/EC (EC 2002). Hence, the method showed good trueness and precision for the evaluated concentration levels.
The summary results for the LOD and LOQ are displayed in Table 3. The sensitivity of the validated method was considered adequate for routine analysis. The LOQ of the method for aflatoxins B1, B2, G1 and G2 were much lower than the maximum permissible Brazilian limit (1 µg kg −1 ) for total aflatoxins (sum of the aflatoxins B1, B2, G1 and G2) in infant formulas. The LOQ for aflatoxin M1 (0.016 µg kg −1 ) was lower than the restriction limit set by the EU regulation (0.025 µg kg −1 ) for infant formulas (EC 2006) and comparable to the 0.010 and 0.0125 µg kg −1 values reported respectively by Zhang et al. (2013) and Tonon et al. (2018), who also employed the LC-MS/MS technique.

Sample Analyses
In order to protect public health, regulations for the control of aflatoxin M1 contamination levels in milk and infant formula have been established in several countries (FAO 2004;EC 2006). The current regulation in Brazil, IN nº 88 / 2021, set the maximum limit of 1 µg kg −1 for total aflatoxins (sum of G2, G1, B1 and B2) in infant formula. The limit regulating the presence of aflatoxin M1 in infant formula, however, has not yet been established in Brazil (Brasil 2021). For evaluation of sample contamination, the reconstitution stated on the formula containers and the limits according to EU regulation were considered.
The incidence and concentrations of aflatoxin M1 in the analyzed samples are displayed in Table 4. Aflatoxin M1 was found in 27 (22.0%) of the analyzed samples. In Brazil, Table 2 Matrix effect evaluation for infant formulas and powdered milk a Obtained by analysis of covariance comparing the slopes of the calibration curves prepared for each matrix (different producers); b obtained by analysis of covariance comparing the slopes of the matched calibration matrix with the slopes of the calibration curve in solvent; c obtained by analysis of covariance comparing all the slopes of the calibration curves prepared in the matrix.  (Kim et al. 2000;Baydar et al. 2007;Alvito et al. 2010;Gómez-Arranz and Navarro-Blasco 2010;Meucci et al. 2010;Beltran et al. 2011;Zhang et al. 2013, Omar 2016and Akhtar et al. 2017. Aflatoxin M1 was quantified (≥ LOQ) in eight of the analyzed samples (6.5%), at concentration levels ranging from 0.016 to 0.056 µg kg −1 , averaging 0.031 µg kg −1 . Concerning the samples acquired from 2011 to 2013, aflatoxin M1 was quantified in five (5.5%) of the analyzed samples at concentration levels ranging from 0.021 to 0.044 µg kg −1 , averaging 0.026 µg kg −1 . With regard to samples acquired from 2018 to 2020, aflatoxin M1 was quantified in three (9.4%) of the analyzed samples at concentration levels ranging from 0.016 to 0.056 µg kg −1 , averaging 0.037 µg kg −1 .
The average aflatoxin M1 concentration observed in this study (0.031 µg kg −1 , calculated using concentrations ≥ LOQ), was two-fold higher than the value of 0.015 µg kg −1 reported by Tonon et al. (2018), and similar to the concentration of 0.024 µg kg −1 reported by Ishikawa et al. (2016) in Brazil. Compared to results for other countries, the mean value reported herein is tenfold higher that that reported by Gómez-Arranz and Navarro-Blasco (2010), of 0.0031 µg kg −1 , similar to the values reported respectively by Quevedo-Garza et al. (2020) and Akhtar et al. (2017), of 0.040 µg kg −1 and 0.037 µg kg −1 , and half the value of 0.06 µg kg −1 reported by both Kim et al. (2000) and Baydar et al. (2007).
The concentration levels (0.040, 0.044 and 0.057 µg kg −1 ) found for aflatoxin M1 in three (2.4%) of the analyzed samples exceeded the maximum permissible level set by EU regulations for infant formulas, of 0.025 µg kg −1 . In Table 3 Validation parameters of the method for infant formulas LOD, Limit of detection (µg kg −1 ); LOQ, Limit of Quantification (µg kg −1 ); Concentration of reconstituted infant formulas; Rec (%), recovery; a RSD (%) relative standard deviation (intra-day, n = 4); b RSD (%) relative standard deviation (inter-day, n = 4 days) 0.05 µg kg −1 0. other studies carried out in Brazil, this level was exceeded in three (18.8%) of the samples analyzed by Ishikawa et al. (2016), while all samples investigated by Iha et al. (2013) and Tonon et al. (2018) exceeded the established level. In studies carried out in other countries by Quevedo-Garza et al. (2020), Akhtar et al. (2017) and Omar et al. (2016), the maximum allowed level was exceeded respectively by 20.0, 30.8 and 85.0% of the analyzed samples, while none of the samples analyzed by Gómez-Arranz and Navarro-Blasco (2010) exceeded the permitted level. Figure 2 displays the chromatograms for one of the samples naturally contaminated by aflatoxin M1 (0.044 µg kg −1 ). The other investigated mycotoxins were not detected in the analyzed samples.

Estimated Daily Intake (EDI) of Aflatoxin M1
The results obtained in the EDI studies are displayed in the Table 5. The EDI values found for aflatoxin M1 in the present study (0.39 to 1.53 ng kg −1 bw day −1 ) were higher than the values reported by Ishikawa et al. (2016) for infant formula intake by infants up to 12 months old in Brazil (0.078 to 0.306 ng kg −1 bw day −1 ).
The EDI results obtained in our study were higher than the values of 0.14 ng kg −1 bw day −1 for infant formulas and 0.014 ng kg −1 bw day −1 for pre-term feeding previously reported in Spain (Gómez-Arranz and Navarro-Blasco 2010). In Mexico, the EDI for aflatoxin M1 reported by Quevedo-Garza et al. (2020) ranged from 1.56 to 14 ng kg −1 bw day −1 , representing the values estimated for one year-old infants when fed starter or follow-on formulas, respectively.

Conclusions
A suitable modified QuEChERS method for the analysis of aflatoxins M1, M2, B1, B2, G1, G2 and ochratoxin A in infant formulas was validated. The method is sensitive and fast, involving a simple simultaneous extraction and sample clean-up followed by concentration of the extract and analysis by UHPLC-MS/MS. As an advantage, the analytical method shows no matrix effect for aflatoxin M1 and ochratoxin A, not requiring the construction of calibration curves in the sample matrix for the determination of these mycotoxins during routine analyses.
The validated method was applied to the determination of the target mycotoxins in samples from the metropolitan region of Rio de Janeiro, RJ, Brazil, all but three of which were in accordance to the established EU regulation. The contamination results reported herein demonstrate the importance of conducting monitoring studies to support aflatoxin M1 regulations for infant formulas in Brazil.
Funding No funding was received for this study.
Data Availability All data generated or analyzed during this work are included in this published article.