Aflatoxin M₁ and ochratoxin A in a human milk bank in the city of São Paulo, Brazil

Abstract

Because infants are more susceptible to the adverse effects of mycotoxins, this work was carried out to determine aflatoxin M₁ (AFM₁) and ochratoxin A (OA) in milk from the Human Milk Bank of the Southern Regional Hospital, São Paulo, Brazil. Analytical methods were first established and evaluated. The methods involved the extraction of AFM₁ with methanol and OA with 1% aqueous sodium bicarbonate solution and methanol, clean-up with immunoaffinity columns having antibodies specific for each mycotoxin and quantification by high performance liquid chromatography (HPLC) with fluorescence detection. The method established for AFM₁ had mean recovery percentages of 94, 77 and 82% and coefficients of variation of 17.5, 3.4 and 4.2% at 0.01, 0.03 and 0.05 ng ml⁻¹, respectively. For the OA method, the corresponding values were 84, 84 and 75% for recovery and 14.1, 3.7 and 4.0% for the coefficient of variation. The limit of quantification for both methods was 0.01 ng ml⁻¹. Of a total of 50 samples analysed, only one was contaminated with AFM₁, at 0.024 ng ml⁻¹, and two with OA, at 0.011 and 0.024 ng ml⁻¹. Although the incidence observed was low, it is recommended that the study be extended to other milk banks of the city of São Paulo.

Keywords:

• Aflatoxin M1
• ochratoxin A
• human milk
• analysis

Introduction

Aflatoxin M₁ (AFM₁) is the principal hydroxylated aflatoxin metabolite in the milk of dairy cows fed a diet contaminated with aflatoxin B₁ (AFB₁). It is also present in the milk of nursing mothers who consumed foodstuffs with AFB₁ (Neal et al. 1998). The fact that milk, the basic food of infants, can be contaminated with AFM₁, a highly toxic substance, has been a matter of considerable concern. There are few papers on the occurrence of AFM₁ in human milk in the world (Wild et al. 1987; El-Nezami et al. 1995; Jonsyn et al. 1995; Saad et al. 1995; Alla et al. 2002; Abdulrazzaq et al. 2003; Turconi et al. 2004). While the percentages of contaminated samples in France, Italy, Australia and Zimbabwe were 0 (0/42), 0.4% (1/231; 0.194 ng ml⁻¹), 15% (11/73; range, 0.028–1.031 ng ml⁻¹) and 11% (6/54; range 0.014–0.05 ng ml⁻¹), respectively, in Thailand and Sierra Leone AFM₁ was found in 45% (5/11; range, 0.039–1.736 ng ml⁻¹) and 31% (35/113; range, 0.2–99 ng ml⁻¹), respectively, of the samples. Egypt and the United Arab Emirates also had high incidence and levels, 55% (66/120; range, 0.02–2.09 ng ml⁻¹) and 91% (127/140; range, 0.053–3.4 ng ml⁻¹), respectively.

Ochratoxin A (OA) is a nephrotoxic and nephrocarcinogenic mycotoxin produced by Penicillium verrucosum in temperate or cold climates and a number of species of Aspergillus in warmer and tropical parts of the world (Pittet 1998). OA has been reported as a contaminant in human milk (Gareis et al. 1988; Breitholtz-Emanuelsson et al. 1993; Jonsyn et al. 1995; Micco et al. 1995; Miraglia et al. 1995, 1998; Zimmerli & Dick 1995; Apostolou et al. 1998; Skaug et al. 1998; Brera et al. 2001; Skaug et al. 2001; Alla et al. 2002; Turconi et al. 2004), the surveys showing marked differences in OA incidence and levels in the different countries.

The contamination percentage in Italy varied from 20–87% (22/111, 9/33, 294/465, 74/85, 198/231), the levels being generally low (range, 0.001–12.0 ng ml⁻¹). OA was also found at varying incidence but at low levels in Norway (33%; 38/115; range, 0.01–0.13 ng ml⁻¹ and 21%; 17/80; range, 0.01–0.182 ng ml⁻¹), Germany (11%; 4/36; range, 0.017–0.03 ng ml⁻¹), Australia (2%; 2/100; range, 3–3.6 ng ml⁻¹), Switzerland (10%; 4/40; range, 0.005–0.014 ng ml⁻¹) and Sweden (58%; 23/40; range, 0.01–0.04 ng ml⁻¹). Sierra Leone and Egypt had the highest levels of contamination (35%; 40/113; range, 0.2–337 ng ml⁻¹ and 36%; 43/120; range 5.07–45.01 ng ml⁻¹, respectively).

Children are considered to be more susceptible than adults to the effects of mycotoxins because of their lower body weight, higher metabolic rate, lower ability to detoxify and incomplete development of some organs and tissues, notably the central nervous system (Kuiper-Goodman 1989; Galvano et al. 1996).

The objectives of this study were:

1.	To standardize and validate analytical methods for the determination of AFM1 and OA in human milk.
2.	Determine AFM1 and OA in human milk samples from the Human Milk Bank of the Southern Regional Hospital in São Paulo, Brazil.

The evaluation of these mycotoxins in human milk has not been studied in this country.

Materials and methods

Samples

A total of 50 samples of human milk, 22 samples collected in the winter (2001–2002) and 28 samples in the summer (2002), were obtained from the Human Milk Bank of the Southern Regional Hospital in São Paulo, Brazil. The samples (about 60 ml each) were kept in glass bottles at −20°C until analysed. The mothers signed a written permission for the donation of the samples and the use of information about their diets obtained through a questionnaire.

Reagents and materials

HPLC solvents were of HPLC grade and other chemicals were of analytical reagent grade. The immunoaffinity clean-up columns, specific for AFM₁ and OA, were obtained from Rhône-Diagnostics Technologies Ltd. (Glasgow, UK). Immunoaffinity columns were kept under refrigeration at 2–8°C and brought to room temperature before analysis. Standards of AFM₁ and OA were purchased from Sigma Chemical Co. (St Louis, MO, USA). AFM₁ stock solution was prepared in benzene-acetonitrile (90:10 v/v) and for OA was prepared in toluene-acetic acid (99:1 v/v). The concentrations were 0.16 µg ml⁻¹ and 9.72 µg ml⁻¹, respectively, and were determined according to AOAC International (1995), checked with UV spectrofotometry in 350 nm for AFM₁ and 333 nm for OA, using ε = 5440 m² mol⁻¹ (OA). Working solution (0.02 µg ml⁻¹) for AFM₁ and OA were prepared together by appropriate dilution in methanol. The solutions were stored below −20°C and protected from light.

Determination of AFM₁

The samples were extracted and submitted to clean-up according to instructions from the producer of the immunoaffinity columns, with minor modifications. The analytical procedure is shown schematically in Figure 1. For AFM₁, an aliquot (25 ml) of the sample of human milk was diluted with 10 ml deionized water and centrifuged at 1100 × g for 15 min. The thin fat layer was removed, 35 ml of methanol were added, and the mixture was mechanically shaken for 10 min and centrifuged at 1100 × g for 15 min. The upper layer, after addition of 50 ml of deionized water, was passed through the immunoaffinity column for AFM₁ at a slow steady flow rate of ca 2–3 ml min⁻¹. The column was washed with 30 ml of deionized water and dried by passing air through it. AFM₁ was slowly eluted from the column with 2 ml acetonitrile:methanol (3:2, v/v). Acetonitrile:methanol was allowed to remain in contact with the column for at least 60 s to ensure complete removal of the bound toxin from the antibody. The eluate was brought to dryness under a gentle stream of N₂. The residue was dissolved in 500 µl of 1% acetic acid:acetonitrile:methanol (40:35:25, v/v) and a 20 µl aliquot was injected into the HPLC equipment.

Figure 1. Flow diagram of the analytical procedure.

Although the immunoaffinity column was expected to be highly specific, the identity of AFM₁ was confirmed by derivatization with trifluoroacetic acid (AOAC International 1995).

Determination of OA

An aliquot (25 ml) of the sample of human milk was mixed with 10 ml of 1% aqueous sodium bicarbonate solution and the mixture was shaken for 3 min and centrifuged at 1100 × g for 15 min. The thin fat layer was removed and 30 ml of methanol were added. After mechanical shaking for 10 min and centrifugation at 1100 × g for 15 min, the upper layer was taken, 50 ml of phosphate buffered saline (PBS) were added and the total volume passed through the immunoaffinity column for OA at a slow steady flow rate of ca 2–3 ml min⁻¹. The column was washed with 30 ml deionized water and dried by passing air through it. OA was slowly eluted from the column with 2 ml acetic acid: methanol (2:98, v/v). Acetic acid:methanol was allowed to be in contact with the column for at least 60 s to completely remove the bound toxin from the antibody. The eluate was brought to dryness with a gentle stream of N₂. The residue was dissolved in 500 µl of 1% acetic acid:acetonitrile:methanol (40:35:25, v/v) and an aliquot of 20 µl was injected into the HPLC chromatograph. The identity of OA was confirmed by derivatization with concentrated HCl according to Zimmerli and Dick (1995) and Skaug et al. (2001).

HPLC equipment and operating conditions

All analyses were carried out using a Shimadzu (Kyoto, Japan) SCL-10AVP LC system equipped with an on-line degasser, automatic injector, quaternary pump LC-10ADVP, a CTO-10ASVP column oven kept at 35°C, fluorescence detector model RF-10AXL and CLASS-VP 5.02. The analytical column was a LiChrosorb C18 column (250 × 4 mm, 10 µm, Merck, Germany) and the mobile phase consisted of 2% acetic acid:acetonitrile:methanol (40:35:25, v/v) pumped at a flow rate of 1 ml min⁻¹. For fluorescence detection, excitation was set at 360 nm and emission at 430 nm for AFM₁, and at 333 nm and 443 nm, respectively, for OA. The same mobile phase was used for determinations of AFM₁ and OA, but was carried out in separation injections.

Evaluation of the methods

A pool of samples of human milk was analysed repeatedly to verify if AFM₁ and OA were detected (uncontaminated samples). To evaluate recoveries, the methods were applied to the analysis of uncontaminated samples spiked with 0.01, 0.03 and 0.05 ng ml⁻¹ of AFM₁ and OA. For both mycotoxins, the recovery test was carried out five times at the concentration of 0.01 ng ml⁻¹ and four times at 0.03 and 0.05 ng ml⁻¹.

The limit of quantification was determined by the addition of decreasing amounts of the mycotoxin standards to an uncontaminated sample.

Standard curves for the two mycotoxins, bracketing the expected levels in the samples, were constructed with seven points (from 0.5–1.0 ng ml⁻¹), six injections being made for each point. The curves passed through the origin and demonstrated linearity within the concentration range, the coefficients of correlation being 0.9997 for AFM₁ and 0.9983 for OA and the coefficients of variation being 1.68 and 3.20%, respectively.

Results and discussion

The method established for AFM₁ had mean recovery percentages of 94; 77 and 82% with coefficients of variation of 17.5; 3.4 and 4.2% at concentrations of 0.01, 0.03 and 0.05 ng ml⁻¹, respectively (see Table I). The method for OA had mean recovery percentages of 84, 84 and 75% and coefficients of variation of 14.1; 3.7 and 4.0% at concentrations 0.01, 0.03 and 0.05 ng ml⁻¹, respectively (Table I). Therefore the limit of quantification for both methods was 0.01 ng ml⁻¹ defined as the minimum level at which the analyte can be quantified with acceptable accuracy and precision (Garfield et al. 2000).

Table I. Recovery of AFM₁ and OA added to samples of human milk.

Quantity added (ng ml^−1)	Mean recovery (%)^a	Coefficient of variation (%)^a
Aflatoxin M1
0.01	94.0	17.5
0.03	77.0	3.4
0.05	81.5	4.2
Ochratoxin A
0.01	83.9	14.1
0.03	84.5	3.7
0.05	75.0	4.0
^aMeans of five determinations at 0.01 ng ml^−1 and four determinations at 0.03 and 0.05 ng ml^−1.

Of a total of 50 samples analysed, only one was contaminated with AFM₁, at 0.02 ng ml⁻¹, and two with OA, at 0.01 and 0.02 ng ml⁻¹ (Table II). Chromatograms of AFM₁ and OA contaminated samples are shown in Figure 2.

Figure 2. HPLC-FD chromatograms of (a) sample spiked with standards of AFM₁ and OA at 0.5 ng ml⁻¹; (b) sample of human milk contaminated with AFM₁ at 0.02 ng ml⁻¹ and (c) sample of human milk contaminated with OA at 0.02 ng ml⁻¹.

Table II. Incidence of AFM₁ and OA in Milk Bank.

Season/number of days after delivery	Number of samples analysed	Number of positive samples	Mycotoxin	Concentration^a (ng ml^−1)
Winter
1–7	12	1	AFM1	0.02
7–15	2	0	–	ND^b
>15	5	0	–	ND
Not specified	3	0	–	ND
Summer
1–7	14	0	–	ND
7–15	2	1	OA	0.02
>15	6	0	–	ND
Not specified	6	1	OA	0.01
^aLimit of quantification: 0.01 ng ml^−1; ^bND: not detect.

The contamination levels obtained in the present study of AFM₁ and OA were lower than Abdulrazzaq et al. (2003) and Micco et al. (1995). In our evaluation a body weight of 4 kg assumed as a basis for the exposure assessment of the babies was below the estimated Tolerable Daily Intake (TDI). With a TDI of 5 ng kg⁻¹ bw/day suggested by the Scientific Committee for Food at European level (European Commission 1998), the maximum tolerable OA daily intake for babies should be 20 ng.

The levels found in our study by comparing the legal limits for OA in milk and baby food that exist in some countries e.g., Italy, France and Switzerland, were not significant, however in one sample, the level of AFM₁ was above or equal to the legal limit (FAO 2004).

We expected a difference between summer and winter in the samples of human milk because in our country in winter the eating of peanuts and peanut products is greater than in summer indicating the possibility of contamination from aflatoxins. The samples of human milk contaminated with OA were collected from two mothers who consumed rice bran, beer and coffee in the summer.

Although the incidence of AFM₁ and OA in the human milk samples from the Milk Bank of the Southern Regional Hospital was low, it is recommended that the study be extended to other milk banks of the city of São Paulo. The fact that three samples were contaminated indicated the possibility of contamination, exposing infants to these two mycotoxins.

There was an intention to correlate the incidence of the mycotoxins to the mothers’ diets, with information obtained through a questionnaire. The low incidence observed prevented such a correlation being made.

Acknowledgements

The authors would like to thank the employees of the Human Milk Bank of the Southern Regional Hospital, city of São Paulo, for their collaboration in collecting the samples of human milk and for filling out the questionnaires to obtain information about the mothers’ diets.