Quantification of aflatoxin and ochratoxin contamination in animal milk using UHPLC-MS/SRM method: a small-scale study

Abstract

Mycotoxin contamination in animal milk is an emerging concern around the globe. Here we developed and validated an ultrahigh-performance liquid chromatography and mass spectrometry-selected reaction monitoring (UHPLC/MS-SRM) method to quantify low concentrations of aflatoxins (AFs) and ochratoxins (OTs) in routinely consumed animal milk samples collected from southern India. Stable isotope dilution methodology was applied to quantify AFB₁, AFB₂, AFG₁, AFG₂, AFM₁, AFM₂ and OTA, OTB in n = 38 different milk samples, using 1 mL of milk. Bioanalytical parameters including method accuracy, precision, recovery, regression analysis and stability were assessed. Dynamic ranges for quantification were between 15.6–1000 pg/mL for AFB₁, AFB₂, AFG₁, and OTA; 7.8–500 pg/mL for AFM₁, AFM₂ and OTB; 78.6–5000 pg/mL for AFG₂. Method accuracy ranged between 80–120%, with ± 15% precision. Recoveries for spiked standards were > 88% in water and 75% in milk, with limits of quantification (LOQ) ranging between 31.3 pg/mL for AFB₁, AFB₂, AFG₁ and OTA, 15.6 pg/mL for AFM₁, AFM₂ and OTB and 156 pg/mL for AFG₂. R² values for regression analyses ranged between 0.9991–0.9999. AFB₂ [mean: 38 pg/mL (0.038 µg/kg)] was quantified in goat milk, AFM₁ was quantified in cow, goat, pasteurized milk [mean: 331 pg/mL (0.331 µg/kg), 406 pg/mL (0.406 µg/kg), 164 pg/mL (0.164 µg/kg)]. Additionally, 90% of cow, goat and pasteurized milk samples were above European Union (EU) limits of 50 pg/mL (0.05 µg/kg) and 40% of cow and goat milk samples were above the Food Safety Standards Authority of India (FSSAI) limit of 500 pg/mL (0.5 µg/kg). AFM₂ was also quantified in cow, goat, and pasteurized milk samples [mean: 249 pg/mL (0.249 µg/kg), 375 pg/mL (0.375 µg/kg), 81 pg/mL (0.081 µg/kg)]. Our dynamic ranges for quantification are lower than other published methods, with need for a smaller volume of milk. This validated method can be applied for routine quantification of mycotoxins in milk.

Supplementary information

The online version contains supplementary material available at (10.1007/s13197-021-04986-w).

Introduction

Mycotoxin contamination in food items is an issue of growing concern around the globe, especially in countries like India, due to the emerging effects of climate change (Bhat et al. 2010). These toxins account for millions of dollars worth of losses in agricultural products and result in adverse impacts on human and animal health (Alshannaq and Yu 2017). The Food and Agriculture Organization estimates that 25% of the global food system is impacted by mycotoxin contamination. Several factors contribute to the presence or production of mycotoxins in foods and feed, including storage, environmental and ecological conditions (Hussein and Brasel 2001). In case of animal source foods, contamination happens primarily due to fungal growth in animal feed. Human exposure to mycotoxins occurs via intake of contaminated agricultural products such as cereals, corn and indirectly via the consumption of animal source foods such as milk and eggs (Flores-Flores et al. 2015). Tackling fungal growth that leads to mycotoxin/secondary metabolite contamination in food and feed is challenging due to their presence in a wide range of crops (Munkvold 2003; CAST 2003).

Secondary fungal metabolites produced by Aspergillus, Penicillium and Fusarium genera, that are known to contaminate crops and animal feed lead to a wide range of adverse health effects (Ismaiel and Papenbrock 2015). Aflatoxins (AFs), which are produced by Aspergillus parasiticus and Aspergillus flavus are primarily found in hot and humid climates (Giorni et al. 2007; Passone et al. 2010). Human exposure to AFs occurs mainly via consumption of staple crops such as maize and groundnuts, which have been impacted by A.flavus and/or A.parasiticus, in addition to foods derived from animals. AFB₁ is the most toxic of the AFs and chronic exposure is known to cause liver cancer and several other adverse health outcomes in humans. It has been categorized as a Group 1 carcinogen by the International Agency for Research on Cancer (IARC) (Ostry et al. 2017). AFB₁ is metabolized into an epoxide form that results in adduct formation with DNA and albumin through lysine modification (Sabbioni and Turesky 2017; Woo et al. 2011). Hepatic hydroxylation of AFB₁ by CYP1A2 forms AFM₁, which is excreted into the milk of lactating animals and is considered an important marker of long-term AFB₁ exposure (Sabbioni and Turesky2017; Woo et al. 2011). AFM1 is tenfold less toxic than AFB₁ and although previously classified as a group 2B carcinogen (probable carcinogen) by the IARC, has been re-classified to a group 1 carcinogen more recently (Ostry et al. 2017; IARC 2012).

Ochratoxins are produced by fungi of genera Aspergillus and Penicillium, commonly found in foodstuffs and feed (Zhu et al. 2017). OTA has been categorized as a group 2B carcinogen by the IARC (Ostry et al. 2017). Dietary exposure to OTA represents an important concern to public health and has been associated with several diseases in humans and animals (Heussner et al. 2015; Pfohl-Leszkowicz. 2009; Stoev et al. 2013). Several naturally occurring conjugates and metabolites of OTs have been identified to date, however the main forms include OTA, OTB, OTC and OTα (Malir et al. 2016). OTB is a non-chlorinated form of OTA and OTC is an ethyl ester of OTA.

A variety of methods exist for detection of mycotoxins in different food matrices and animal feeds including HPLC based fluorescence detection (HPLC-FD) (Kumar et al. 2020; Dhanshetty et al. 2019; Oulkar et al., 2018) and enzyme linked immunosorbent assays (ELISA) (Pestka et al. 1981). These methods generally lack in sensitivity and often times require a higher sample volume for use in the assay. Due to limitations in these methods, liquid chromatography-mass spectrometry based quantification using stable isotope labelled internal standards is the current gold standard in the field, for detection and quantification of lower levels of multiple-mycotoxins (pg to ng ranges) in food matrices.

Several published papers have described different kinds of mass spectrometry-based methods for analysis of mycotoxins, but have either not used labelled internal standards (ISTDs) or quantified AFM₂ (Mao et al. 2018; Zhang et al. 2018). We therefore developed and validated a multi-analyte method to quantify six aflatoxins (AFB₁, AFB₂, AFG₁, AFG₂, AFM₁ and AFM₂) and two ochratoxins (OTA and OTB) from a variety of milk samples (cow, goat, buffalo and pasteurized milk) by using labelled AFB₁-D₃ and OTA-D_5, as internal standards.

Experimental

Materials

Mycotoxin standards for AFB₁, AFB₂, AFG₁ and AFG₂ were purchased from Sigma-Aldrich (Bangalore, India). Standards for AFM₁, AFM₂, OTA, OTB, AFB₁-D3 and OTA-D5 were procured from Toronto Research Chemicals (Toronto, Canada). The purity of all of the analytes and deuterated internal standards was ≥ 98%. High purity MS grade solvents (water and acetonitrile) were purchased from Honeywell (Bangalore, India). Formic acid and ammonium acetate were obtained from Sigma-Aldrich (Bangalore, India). AFB₁, AFB₂, AFG₁ and AFG₂ were purchased as 1 mg; AFM₁ as 50 µg; AFM₂ as 100 µg; OTA/OTB as 2 mg powder forms. AFB₁-D3 and OTA-D5 ISTDs were purchased as 0.25 mg and 0.5 mg powder forms, respectively. These STDs and ISTDs were dissolved in acetonitrile to make 100, 10 µg/mL stocks and stored at −80 °C. Centrifugal membrane filters (0.45 μm, PVDF) were obtained from Thermo-Fisher Scientific (Bangalore, India). Animal milk (from cow, goat and buffalo) and pasteurized cow milk sample collection and storage conditions are presented elsewhere (Shetty et al. 2020).

Standard stock preparation

All mycotoxin STDs and two ISTDs working stock solutions and serial dilutions were prepared in acetonitrile. Working stock concentrations ranged from 1 μg/mL for AFB₁, AFB₂, AFG₁, OTA and OTB, 0.5 μg/mL for AFM₁ and AFM₂ and 10 μg/mL for AFG₂. For both AFB₁-D3 and OTA-D5 the working stock concentrations were 1 μg/mL each.

UHPLC-MS

A Sciex QTRAP 6500 (Sciex Singapore) mass spectrometer was used for mycotoxin analysis, which is equipped with a turbo V ion source for effective ionisation. The mass spectrometer is coupled to an Agilent 1290 infinity II UHPLC system (Agilent Technologies India Pvt. Ltd., India), and equipped with a column oven (set at 40 °C), auto-sampler with a thermo-controller (set at 10 °C). The system uses a flow through needle mode after injection and is furnished with a needle wash system (with 50% methanol) before injection to ensure zero percent carry over problems. The mobile phase solvent A was water (10 mM Ammonium Acetate, 0.1% Formic Acid) and Solvent B was acetonitrile (0.1% Formic Acid). In the UHPLC system, a C-18 column (2.1 × 100 mm, 1.8 μm, Agilent, Inc) was used for separation of mycotoxins. An optimised gradient to get maximum separation (0 to 3 min-10%B, 3 to 15 min- 10%B to 80% B, 15.1 to 17 min-100% B, 17.1 to 22 min-10%B) at 200 μL/min flowrate was used for analysis. The injection volume (10 μL) was kept constant throughout the analysis. Operating conditions were: Spray Voltage- 5500 V, Curtain Gas- 30 psi, Temperature- 500 °C, Gas 1- 30 psi, Gas 2- 50 psi, Entrance Potential (EP)- 10 V. Other parameters including De-clustering Potential (DP), Exit Potential (CXP), and Collision Energy (CE) were standardised based on the compound and incorporated into the method. Scan time was 50 millisec per transition with positive ion polarity. To obtain the details of tandem mass spectrometry (MS/MS) scans, a syringe pump was used to infuse STDs with flow of 10 μL/min. For each metabolite, the precursor ion was monitored and collision induced dissociation was used to generate product ions. The product ions were obtained by scanning the quadrupole-3 from m/z 50 to 500 with a cycle time of 1 s. This was followed by optimization of DP, CXP and CE for each of the intense product ions as shown in Table 1.

Table 1.

Selected reaction monitoring table

Analyte	Parent ion (m/z)	Product ion (m/z)	Retention time (min)	De-clustering potential (volts)	Collision energy (CE)	CXP (volts)	Equation	R2
AFB1	312.93	241.08	10.4	161	53	30	y = 0.0117x + 0.0138	0.9995
AFB2	314.95	287.08	9.9	176	39	26	y = 0.0039x − 0.001	0.9995
AFG1	328.92	243.05	9.9	161	39	22	y = 0.0076x − 0.0015	0.9999
AFG2	330.96	313.13	9.4	131	35	32	y = 0.0018x + 0.0047	0.9998
AFM1	328.96	273.04	8.8	171	33	38	y = 0.0092x + 0.0009	0.9998
AFM2	330.92	273.05	8.4	146	33	32	y = 0.0038x − 0.0003	0.9999
AFB1-D3	315.95	241.09	10.3	186	33	28
OTA	403.94	239.02	12.7	61	21	20	y = 0.0099x − 0.0056	0.9963
OTB	370.06	205.06	11.5	76	27	22	y = 0.0231x − 0.0102	0.9935
OTA-D5	409.02	239.05	12.7	16	33	32

Calibration curves

Calibration curves were set in the range of 15.6–1000 pg/mL for AFB₁, AFB₂ and OTA. For AFM₁, AFM₂ and OTB, the range was set to 7.8–500 pg/mL and for AFG₂ the range was higher at 78–5000 pg/mL. ISTD mix (AFB₁-D3 and OTA-D5) was spiked at a concentration of 1000 pg/mL to each standard. To extract mycotoxins from 1 mL of milk samples, 2 mL of cold acetonitrile with 2% formic acid was used. Similarly, to construct the standard curves, both STDs and ISTDs were spiked into 3 mL of water:acetonitrile (2% FA) (1:2) solution and processed similarly. Briefly, the STDs and ISTDs were extracted by separating the acetonitrile layer using a concentrated salt solution [300 μL of concentrated ammonium acetate (10 g/10 mL)]. The mixture was sonicated in a water bath sonicator for 10 min; it was subsequently removed and kept on ice for 5 min. It was then centrifuged for 5 min, (3800 × g) to separate the acetonitrile layer. Approximately, 1.5 mL of the acetonitrile from the supernatant was transferred to a 2 mL Eppendorf tube and dried (2.5 h in speed vacuum). The dried extract was then reconstituted with 100 μL of 50% acetonitrile (vortexed and centrifuged for 5 min at 15,000 × g) and 80 μL was transferred to HPLC vials. Next, 10 μL was injected to analyse mycotoxins using the UHPLC-MS/SRM method. STDs were prepared to construct a seven-point calibration curve on a daily basis and analysed along with five replicates of quality control (QCs) samples as shown in Table 2.

Table 2.

Method validation of aflatoxins and ochratoxins

Concentration in pg/ml	LOQ	LQC	MQC	HQC	LOQ	LQC	MQC	HQC
	AFB1				AFB2
	31.30	62.50	250.00	800.00	31.30	62.50	250.00	800.00
Inter-day mean (n = 3)	31.08	61.77	253.13	809.53	30.05	61.39	249.13	807.58
%CV (n = 3)	5.91	5.52	3.23	3.34	5.73	5.75	5.04	0.90
Accuracy (%)	99.30	98.84	101.25	101.19	96.00	98.23	99.65	100.95
	AFG1				AFG2
Concentration in pg/ml	31.30	62.50	250.00	800.00	156.00	313.00	1250.00	4000.00
Inter-day mean (n = 3)	31.53	63.74	253.27	793.57	1558.60	315.87	1244.67	4019.50
%CV (n = 3)	2.76	1.15	0.66	2.86	0.67	3.71	4.78	0.36
Accuracy (%)	100.75	101.98	101.31	99.20	101.67	100.92	99.57	100.49
	AFM1				AFM2
Concentration in pg/ml	15.60	31.30	125.00	400.00	15.60	31.30	125.00	400.00
Inter-day_mean (n = 3)	15.52	31.03	123.93	396.73	15.40	30.69	125.13	400.80
%CV (n = 3)	3.80	3.67	4.06	4.12	3.90	5.25	3.47	2.05
Accuracy (%)	99.49	99.15	99.15	99.18	98.72	98.04	100.11	100.20
	OTA				OTB
Concentration in pg/ml	31.30	62.50	250.00	800.00	15.60	31.30	125.00	400.00
Inter-day mean (n = 3)	31.46	61.80	253.20	815.55	15.36	31.27	127.73	409.23
%CV (n = 3)	3.87	2.16	3.09	1.41	5.88	2.65	1.49	0.94
Accuracy (%)	100.51	98.88	101.28	101.94	98.46	99.91	102.19	102.31

Method validation

The most important bio-analytical parameters for validation were conducted using criteria established by the United States Food and Drug Administration (US FDA) (Guidelines for Bioanalytical Method Validation 2018). Calibration linearity was studied by spiking two internal standards (one for each group of mycotoxins: AFB₁-D3 for AFs and OTA-D5 for OTs) to calibration solution at seven concentrations for mycotoxins. Previous studies have demonstrated the use of a single chemically similar ISTD for quantification of compounds in complex matrices, without a compromise in assay performance (Wieling 2002).

Integrated peak areas of the selected SRM transitions were used to build the STD curves. The highest intense product ion from each mycotoxin was used to build the STD curve. Curves were fitted by an equal weighted regression analysis using the quantification software Analyst (version 1.6.3). Precision and accuracy were evaluated using four concentration points (Limit of Quantification-LOQ; Lower Quality Control-LQC; Middle Quality Control-MQC and High-Quality Control-HQC). Accuracies of 85–115% and precisions of ± 15% were considered acceptable for LQC, MQC and HQC samples. Accuracies of 80–120% and precisions of ± 15% were considered acceptable for the LOQ as recommended. Five replicates for each point were analysed to determine the intra and inter-day accuracy and precision. This process was repeated over 3 days in order to determine the inter-day accuracy and precision using freshly prepared calibration curves. Accuracy was determined by the recovery of QCs, and precision was expressed as the coefficient of variation (CV) of the determination of the QCs. Inter-day accuracy and precision were calculated similarly for 15 replicates of each concentration point pooled from the three validation runs.

Recovery, stability and matrix effects

To check recovery from the milk matrix, triplicate QCs (LQC, MQC, HQC) and ISTDs were spiked to milk: acetonitrile (1:2) and water: acetonitrile (1:2) solvent system. As a control, both milk and water alone without the STDs were used. The STD curve was constructed in the same way to calculate the recovery from both milk and water matrices. The acetonitrile phase separation was done in the same way as mentioned above. The solvent was evaporated in a speed vacuum and reconstituted in 100 μL of 50% acetonitrile and 10 μL was injected for the analysis. Freeze thaw stability was checked by spiking QCs in triplicate alongside ISTDs to milk, which was allowed to freeze in the −80 °C for 20 min. We repeated the freeze thaw cycle thrice and calculated the concentration of the spiked toxins by using the STD curve. The stability in the auto-sampler vial was checked after 24 h for the highest standard.

To check for matrix effects, 12 aliquots of control pooled milk samples were processed in the same way as described above and the supernatant was then dried completely. Three samples were used as controls and in the remaining nine samples, triplicate QCs were spiked (n = 3 of each, LQC, MQC and HQC) on top of the processed milk matrix and reconstituted to 100 μL of 50% acetonitrile and 10 μL was injected for the analysis. As a control, QCs were spiked directly to reconstitution solution (100 μL of 50% acetonitrile) and 10 μL was injected for the analysis. These two datasets were compared to calculate matrix effects. We also examined matrix effects by comparing the slope of a calibration curve for standard solutions with matrix matched standard solutions in two experiments, where standards were spiked to the solvent system and milk matrix both pre-and post-extraction. Here, a lower slope for matrix matched standard solution suggests ion-suppression while a higher slope indicates ion enhancement (Zhou et al. 2017).

Sample preparation from different milk samples

The aliquoted milk samples (1 mL) were allowed to thaw on top of ice and transferred to 15 mL centrifuge tubes. Ice-cold acetonitrile (2 mL) with 2% formic acid was then added, and 10 μL of ISTDs were spiked on top. The sample was then vortexed for 1 min and kept on ice for 5 min to allow for complete protein precipitation. Following this, 300 μL of concentrated ammonium acetate (10 g/mL) solution was added, sonicated in a water bath sonicator (10 min), kept on ice (5 min), and then centrifuged (5 min, 3800 × g) to separate the acetonitrile layer. Approximately 1.5 mL of the acetonitrile from the supernatant was transferred to 2 mL Eppendorf tube and dried in a speed vacuum. The final reconstitution was done with 100 μL of 50% acetonitrile and filtered through a 0.45 μm PVDF membrane centrifugal filter (5 min, 15,000 × g). The top 80 μL from each sample was transferred to HPLC vials and 10 μL from each were injected to analyse mycotoxins in milk by the UHPLC-MS/SRM method. A seven-point calibration curve was constructed, and pooled cow milk samples and solvent alone controls were used throughout the study for quantification of mycotoxins.

Results

UHPLC-MS/MS analysis of mycotoxins

Infusion of 10 μg/mL solution of each mycotoxin in acetonitrile was conducted in the mass spectrometer through the syringe pump at a flow rate of 10 μL/min. The source temperature was kept at 0 °C and the other parameters including curtain gas-20 psi, gas 1–10 psi were set to check the full MS and the MS/MS details. Both AFs and OTs showed the M + H ion as major parent ion in the mass spectrum (Supplementary Fig. 1). The collision induced dissociation of the M + H ions showed the product ions of which the highest intense one was selected for quantification. AFB₁ & B₂, AFG₁ & G₂ and AFM₁ & M₂ showed similar patterns of product ions (Fig. 1). In case of OTA and OTB the highest intense product ion is similar with and without a chlorine moiety (m/z: 239 for OTA and 205 for OTB). The highest intense product ions were selected to make the SRM method in the MS (Table 1). AFB₁-D3 and OTA-D5 showed the highest intense product ions, which are similar to AFB₁ and OTA indicating the common unlabelled product ions.

Fig. 1.

MS/MS analysis of aflatoxins, ochratoxins and the corresponding internal standards

Method development

Pooled cow milk samples were initially checked for use as control matrix during method development. However, AFM₁ and AFM₂ were present in quantifiable amounts, which made it difficult to use milk as a control matrix for method validation. We therefore used the same amount of extraction solvent (1:2 of water and acetonitrile) with 2% FA as matrix for method development and validation. Briefly, both STDs and ISTDs were spiked to 3 mL of extraction solvent and the acetonitrile layer was separated, dried, and reconstituted in 50% acetonitrile and injected. Both AFs and OTs showed clear peaks in the UHPLC-MS/SRM chromatogram (Fig. 2). Standard curves were constructed using AFB₁-D3 as ISTDs for all AFs and OTA-D5 for both OTs in order to quantify milk samples. There is a 0.4 min difference in the retention time between AFB₁ to B₂, AFG₁ to G₂ and AFM₁ to M₂. In the case of OTs, there is a 1.3 min retention difference between OTA and OTB. Another peak in the AFB₂ channel, which is close to its retention time, is associated with AFB₁-D3 ISTD. The internal standard showed a major peak at 316 m/z and a minor peak at 315 m/z in the AFB₂ channel. Since there is a 0.4 min difference in retention time, the peak does not cause interference with quantification of AFB₂. Both AFs and OTs elute between 8 and 13 min in the gradient (30–70%B) in the analytical column. All mycotoxins were stable in the auto-sampler for at least 24 h at 10 °C. In order to assess recovery, QCs (LQC, MQC, and HQC) were spiked to both milk and water matrices, following the same procedure described above. Recovery was calculated based on the concentration of mycotoxins in milk and the amount spiked on top of milk. Under these conditions the recovery for all toxins are above 88% when spiked into water and above 75% when spiked into pooled milk (Supplementary Table 1). Recovery calculations were made based on controls using both water and milk alone. Matrix effects did not appear to be of concern and recoveries were above 85% for MQC and HQC. At the LQC level, both OTA and AFM₂ showed 30% reduction in concentration compared to the concentration spiked (Supplementary Table 2). We also examined matrix effects by comparing the slope of calibration curves for standard spiked to neat solution (solvent system) and matrix matched standard solution using two different pre-and-post extraction protocols. Results are presented in Supplementary Table 3. No differences were observed between slopes of calibration curves in these experiments, suggesting no significant signal suppression or enhancement due to the milk matrix. As noted by Panuwet et al. (2016), the use of isotopically labelled internal standards can minimize the impact of biological matrix effects on quantification, although there remains a lack of definitive consensus on management of such effects during bioanalytical method development and validation (Panuwet et al. 2016).

Fig. 2.

UHPLC-MS/SRM chromatogram of Standards and Internal Standards in HQC level. The arrow indicates the peak for AFB₂ standard

Calibration curve and limit of quantification

Calibration curves for quantification of each of the mycotoxins were linear over a seven-point standard curve range spanning 64-fold, with regression correlation coefficients ranging between 0.9991 and 0.9999 (Supplementary Fig. 2). Standard curves for AFB₁, B₂, G₁ and OTA ranged between 15.6 pg/mL to 1 ng/mL; for AFM₁, M₂ and OTB ranged between 7.8 and 500 pg/mL and for AFG₂ ranged between 78 pg/mL to 5 ng/mL. Regression analyses were conducted for calibration curves with the mean (± SD) for the slope and intercept on three different days (Table 1). The LOQs were defined as the lowest analyte concentration that could be quantified with an accuracy of 80–120% and precision of ± 15% for replicates (n = 5) on three different days, with a corresponding signal to noise ratio > 10 (Table 2). The ranges for each mycotoxin were selected based on detected response in the mass spectrometer.

Assay accuracy, precision and stability

Overall, excellent accuracy and precision were obtained for the analysis of all mycotoxins (Table 2). Inter-day accuracy (n = 3) for the LQC (62.5 pg/mL for AFB₁, AFB₂, AFG₁ and OTA; 31.3 pg/mL for AFM₁, AFM₂ and OTB; 313 pg/mL for AFG₂) for all toxins ranged from 98 to 102%. For MQC (250 pg/mL for AFB₁, AFB₂, AFG₁ and OTA; 125 pg/mL for AFM₁, AFM₂ and OTB; 1250 pg/mL for AFG₂) from 99 to 102%. For HQC (800 pg/mL for AFB₁, AFB₂, AFG₁ and OTA; 400 pg/mL for AFM₁, AFM₂ and OTB; 4000 pg/mL for AFG₂) from 99 to 102%. Inter-day precisions (n = 3) were in the range of 1–6% (for LQC), 1–5% (for MQC) and 1–4% (for HQC). The typical STD curves for all AFs and OTs are shown in the supplementary Fig. 2. Lines were constructed using ratios (Analyte/ISTD) against the concentrations and passed close to the point of origin with regression ranges between 0.999 to 1. Details of the method development data are tabulated in Table 2. The precision and accuracy data for the analysis of the LOQ, LQC, MQC, and HQC samples that were re-analysed after 24 h standing in the auto sampler gave essentially identical data to that obtained from the original analyses.

Analysis of mycotoxins from different milk samples

Mycotoxins from different milk samples (cow, goat, buffalo and pasteurized milk) were processed and analysed using the validated UHPLC-MS/SRM method. AFG₁, G₂ and OTB showed no peaks, AFB₁ and B₂ showed slight humps at the corresponding retention times and AFM₁, M₂ and OTA showed clear peaks in cow milk samples. The typical UHPLC-MS/SRM chromatogram of one of the cow milk samples is shown in Fig. 3. A small hump was seen in the AFB₁ and AFB₂ channels at the exact retention times in the chromatogram. A background peak was seen at 12.72 min in the AFB₁ channel, which was away from the AFB₁ peak. This validated method allows for detection of even trace levels (~ 10 pg/mL) of these toxins in animal milk samples. AFB₁ in all milk samples was below the LOQ level (31.3 pg/mL). One of the cow’s milk samples, one pasteurized milk and a few of the goat milk samples were above the first point (detection limit) of the standard curve (15.6 pg/mL). Buffalo milk samples were below the limit of detection for AFB₁. For AFB₂, both buffalo and pasteurized milk samples were below the first point of the STD curve (15.6 pg/mL). Most of the goat milk samples and one cow’s milk sample showed levels above the LOQ (31.3 pg/mL) for AFB₂. Mean levels of AFB₂ in goat milk samples were 38 pg/mL (0.038 μg/kg). In case of AFG_1, all milk samples were below the first point and LOQ except one goat milk sample, which showed ~ 33 pg/mL (~ 0.033 μg/kg). For AFG₂, all milk samples were below the LOQ except one cow’s milk sample, which showed ~ 160 pg/mL (~ 0.16 μg/kg). In case of AFM₁/AFM₂, with the exception of buffalo milk, all other samples showed quantifiable amounts. Seven of the fresh cow’s milk samples and all goat/commercial milk samples showed levels that were higher than the permissible limits for AFM₁, which is 50 pg/mL (according to the EU). Similar patterns were seen for AFM₂ in milk samples. Four cow milk samples and three goat milk samples showed concentrations above the allowable limits (500 pg/mL according to FSSAI and CODEX). Mean values for AFM₁ in cow milk and in goat milk samples were 331, 406 pg/mL (0.331, 0.406 μg/kg) and for AFM₂ in cow milk and in goat milk samples were 249, 375 pg/mL (0.249, 0.375 μg/kg). OTA concentrations were less than the LOQ value but were above the limit of detection (15.6 pg/mL) for cow and goat milk (22 pg/mL or 0.022 μg/kg). Interestingly, both buffalo and pasteurized milk samples showed levels of OTB ~ 12 pg/mL (0.012 μg/kg) that were higher when compared to cow and goat milk samples which showed ~ 8 pg/mL (0.008 μg/kg). Both were slightly above the limit of detection (7.8 pg/mL). A complete analysis of AFs and OTs from a variety of milk samples is shown in Fig. 4.

Fig. 3.

UHPLC-MS/SRM chromatogram of cow milk sample. The arrow indicates the peak for AFB₁, AFB₂ and OTA

Fig. 4.

Analysis of Mycotoxins in milk. Aflatoxins (AFB₁, AFB₂, AFG₁, AFG₂, AFM₁ and AFM₂), ochratoxins (OTA and OTB). (CM-cow milk, GM-goat milk, BM-buffalo milk, PM-Pasteurized milk)

Discussion

India is the leading producer of liquid milk, accounting for 17% of the world’s total milk production (Douphrate et al. 2013). Milk, particularly from dairy animals such as cow, goat and buffalo provide essential micro and macronutrients, which help children and adults meet their daily energy and nutrient needs. It is estimated that on average 1 glass of milk (250 mL) per day, is consumed by the general populace in India, either directly or indirectly in a variety of forms such as tea, coffee, curd, butter milk etc., (Kumar et al. 2014). Pasteurized or ultra-high temperature treated milk is consumed in urban areas across the country. In rural regions, however fresh milk is collected directly from animals and distributed to households on a daily basis. High levels of milk consumption make it imperative to assess contamination and adulteration, routinely. A recent National Milk Quality Survey (2018) conducted by the Food Safety and Standards Authority of India (FSSAI) collected samples from 29 states and 7 union territories across the country and found that 5.7% of milk samples were contaminated with AFM₁ (FSSAI. 2018). This study also used LC-MS/MS for quantification of AFM₁ in milk samples, however standard curve ranges were lower in our analysis, thus allowing for detection of pg/mL levels of exposure. Previous studies on animal milk have reported AFM₁ in samples from across the globe at levels in the range of 0.3–1718 pg/mL, thus higher method sensitivity can improve detection of even trace concentrations of exposures in complex matrices such as milk (Ketney et al. 2017).

During the method development process, we checked different sample preparation protocols including direct protein precipitation (using ethanol, methanol and acetonitrile), liquid–liquid extraction (LLE) with hexane after protein precipitation and reverse phase solid phase extraction (RP-SPE) after protein precipitation for mycotoxin extraction from milk. We observed poor recoveries under all the aforementioned conditions (data not shown) and therefore used ice-cold acidic acetonitrile for mycotoxin extraction. To separate the acetonitrile layer, concentrated ammonium acetate solution was used instead of sodium acetate, magnesium sulphate and sodium chloride. This is mainly to avoid issues with ionization in the mass spectrometer. Better extraction efficiency with minimal steps in sample preparation (protein precipitation, acetonitrile phase separation, drying the acetonitrile layer and analysis) may explain our ability to detect and quantify lower concentrations (pg/mL) of AFM₁/M₂ and AFB₂ in milk samples.

To assess mycotoxin contamination in milk, we collected fresh cow, goat and buffalo samples (n = 28). We also collected pasteurized cow milk samples (n = 10), to see if there are differences in mycotoxin levels due to the effects of ultra-high temperature processing. With the exception of three cow milk samples, all other animal milk samples (cow, goat and pasteurized) showed AFM₁ above 50 pg/mL, which exceed maximum allowable limits set by the EU. Four of the cow and goat milk samples exceeded the FSSAI and CODEX maximum allowable level (MAL) of 500 pg/mL. Interestingly, none of the buffalo milk samples showed AFM₁ contamination. Buffalo milk samples were acquired from a hilly and elevated part of southern India, where an improved grazing environment may explain lower levels of mycotoxin contamination in animal feed.

We found that AFM₂, a derivative of AFB_2, is also present at higher concentrations [mean of AFM₂ in cow milk: 249 pg/mL (0.249 μg/kg), goat milk: 375 pg/mL (0.375 μg/kg), buffalo milk: 11 pg/mL (0.011 μg/kg) and pasteurized milk: 81 pg/mL (0.081 μg/kg)]. Four of the cow milk samples and two of the goat milk samples exceed 500 pg/mL (0.5 μg/kg) limit for AFM₂. The biological implications of AFM₂ contamination on health have not been well characterized. Further research is needed to understand the synergistic or additive effects of AFM₁ and AFM₂ on health. Levels of AFB₁ and AFB₂ in milk were much lower when compared to AFM₁ and AFM₂, which suggest that AFB₁ and AFB₂ are metabolized and secreted into milk in the form of AFM₁/AFM₂. Levels of AFG₁ and AFG₂ are below LOD in most of the milk samples. Studies conducted globally show that approximately 9.8% of milk samples exceed maximum limits set for AFM₁ whereas other mycotoxins are found at lower concentrations in milk (Milicevic et al. 2015). There is a need for more systematic investigation of the levels of aflatoxins in milk used for regular human intake, particularly in countries with high dairy consumption, such as India. In case of ochratoxins, most of the milk samples had concentrations that were above the first point in the STD curve for OTA and OTB, but were below the LOQ. OTA has been previously reported in milk, infant formula and milk-based products (Malir et al. 2016).

Quantification of lower pg levels of mycotoxins in food matrices, using smaller volumes of matrix is a major challenge. LC–MS offers the best approach for quantification of metabolites, nutrients and other contaminants in food matrices. Due to higher method sensitivity, selectivity and specificity, UHPLC-MS/SRM is considered the gold standard method for quantification of biomolecules, when compared to UV, fluorescence and immuno-affinity-based methods (Becker-Algeri et al. 2016; Flores-Flores et al. 2017, 2018; Hashemi et al. 2016; Rastogi et al. 2004). Although the use of specific internal standards for quantification of individual analytes is ideal, cost and availability are an important concern, particularly in our context. For this reason, we used only two labelled mycotoxins (AFB₁-D3 and OTA-D5) as internal standards for each group. Other authors have similarly used a single internal standard to quantify mycotoxins of similar structure (Cervino et al. 2008).

The approximate cost per sample for this analysis ranges between INR 3000 to 5000, and relative ease of processing efficiency can reduce time and increase overall sample throughput. Furthermore, this method can be applied for detection of mycotoxins in human breast milk samples and for routine monitoring in animal milk meant for human consumption (Mehta et al. 2020). Overall, our findings are strengthened by the use of LC–MS/MS and stable isotope dilution methodology, to quantify lower concentrations of mycotoxins in smaller volumes of milk.

Conclusion

We developed and validated an UHPLC-MS/SRM method to quantify AFs and OTs in animal milk samples. As a preliminary study, we checked 38 raw milk samples, from a variety of animals. All samples except buffalo milk exceeded the MAL values for AFM₁ set by the EU (50 pg/mL) and four raw milk samples from cows and goats also exceeded MAL values set by FSSAI and CODEX (500 pg/mL). Similar trends were seen for AFM₂ across milk samples. Further investigation with larger sample sets is warranted to understand the true extent of contamination, in addition to studies of animal food and feed from these regions of the country. Findings from this investigation suggest a need for steps to be taken to control potential fungal contamination in animal feed and to mitigate mycotoxin contamination in milk, routinely used for human consumption.

Abbreviations

AF

Aflatoxin

OT

Ochratoxin

UHPLC/MS

Ultrahigh performance liquid chromatography and mass spectrometry

SRM

Selected reaction monitoring

IARC

International agency for research on cancer

LOD

Lowest detection limit

LOQ

Limit of quantification

LQC

Lower quality control

MQC

Middle quality control

HQC

Highest quality control

MAL

Maximum allowed level

DP

De-clustering potential

CXP

Exit potential

EP

Entrance potential

CE

Collision energy

Compliance with ethical standards

Conflicts of interest

The authors declare no conflicts of interest.