Determination of glyphosate and AMPA in oat products for the selection of candidate reference materials

Abstract

The use of reference materials (RMs) is critical for validating and testing the accuracy of analytical protocols. The National Institute of Standards and Technology (NIST) is in initial stages of developing a glyphosate in oats RM. The first aim of this study was to optimize and validate a robust method for the extraction and analysis of glyphosate and aminomethylphosphonic acid (AMPA). The optimized method was used to screen thirteen commercially available oat products to identify candidate RMs. Glyphosate was detected in all samples, with the highest glyphosate mass fraction of 1100 ng/g; lower levels were measured in grains from organic agriculture. AMPA was quantified in nine samples up to 40 ng/g. The findings of this study led to the identification of candidate RMs, with “high” and “low” glyphosate levels. A preliminary stability study determined that glyphosate is stable in oat material at room temperature for six months.

1. Introduction

The National Institute of Standards and Technology (NIST) produces reference materials (RMs) and certified reference materials (CRMs) to support comparable and accurate analytical measurements and to promote method harmonization across laboratories. RMs are materials that are “sufficiently homogenous and stable with respect to one or more specified properties, which has been established to be fit for its intended use in a measurement process” (ISO, 2015). Standard Reference Materials (SRMs), a subset of RMs, are CRMs issued by NIST. CRMs are “characterized by a metrologically valid procedure for one or more specified properties, accompanied by an RM certificate that provides the value of the specified property, its associated uncertainty, and a statement of metrological traceability” (ISO, 2015). NIST RMs and SRMs can be used to support overall quality assurance, to verify the accuracy of specific measurements, and to support the development of new measurement methods for various areas including food safety (Barber et al., 2019; NIST, 2010; Phillips et al., 2019; Wise et al., 2012; Wise & Phillips, 2019).

There are several food-based matrix SRMs currently available from NIST to support nutritional content measurements. To expand into the food safety space, NIST is developing RMs that include only non-certified values in order to provide a wider range of products to customers more rapidly. NIST matrix SRMs comprised of certified values can take years to develop, especially if multiple orthogonal analytical methods are required to account for all analytical method bias and to fully characterize the material. In contrast, RMs comprised of reference values, or non-certified values, are still a high quality estimate of the true value, but not all sources of bias may have been investigated or accounted for by NIST (Beauchamp et al., 2020). The primary goal here is to rapidly develop and release RMs to meet immediate quality assurance needs of the measurement community (e.g., method harmonization of emerging contaminates) where no CRMs are currently available. By first releasing materials as RMs, NIST can then further engage with stakeholders to determine if changes are needed to the materials (matrix, analyte levels, etc.) or if there is a need to upgrade non-certified assigned values to certified values. In this model, NIST RMs can evolve in order to provide stakeholders with fit-for-purpose materials to meet both current and foreseeable needs.

When designing new natural matrix RMs, identification of appropriate measurands, measurand levels, and matrices that will be useful to the measurement community is critical. To develop RMs that contain endogenous measurands, multiple starting materials are often evaluated and blended to obtain an ultimate material that mimics an evaluation sample of manufacturing processes and/or obtain desired measurand levels for regulation limits and analytical testing of unknown materials. Screening candidate RMs is an important process in order to decide which materials advance into further RM development and ultimately production. Analytical screening methods often need to be sensitive, especially if low mass fraction levels are desired, or if blank level or low-level materials are used for blending with a higher mass fraction material.

Routine screening of foodstuffs for pesticide residues and other potential hazardous contaminants is important to ensure food safety. The availability of RMs is important in the quality control and assurance for these routine measurements (Lehotay & Chen, 2018). Furthermore, the availability of incurred or endogenous analytes in “real world” samples is critical in determining extraction efficiency. Several stakeholders have requested NIST to develop RMs for pesticides residues in foods, to include glyphosate. Glyphosate is of interest as the dominant pesticide used worldwide; in 2014, 125,384 tons of glyphosate were applied in the U.S. (Benbrook, 2016). A new practice emerged in the mid-2000s called “harvest-aid” (a late-season application of glyphosate for use as a desiccant on crops such as wheat, barley, and oats a few days before harvest in order to dry the crops for easier harvesting) which led to an increase of glyphosate use (Benbrook, 2016). Oats were selected as an initial matrix for an RM because glyphosate is heavily used on the crop and because oatmeal and oat-based cereals are very popular foods in Northern America.

The exposure risks of glyphosate on human health are currently being debated. Glyphosate is known to rapidly convert to aminomethylphosphonic acid (AMPA), its primary metabolite, from biotic or abiotic degradation in soils, which may also be toxic to humans even at low mass fractions (Grandcoin et al., 2017). Recently the non-profit organization Environmental Working Group (EWG) highlighted a health concern by reporting the presence of glyphosate up to 2,837 μg/kg in oatmeal and oat-based cereals, including snack bars (EWG, 2018). A monitoring of the Swiss market showed the presence of glyphosate and AMPA in diverse foodstuffs, with the main contributors for dietary exposure being cereal and pulses (Zoller et al., 2018). In breakfast cereals, levels from <0.001 μg/g to 0.291 μg/g of glyphosate and from <0.0025 μg/g to 0.010 μg/g of AMPA were found. For a study of the Canadian retail market between 2015 and 2017, 310 oat samples were extracted and glyphosate levels were between 0.006 μg/g and 3.1 μg/g (Kolakowski et al., 2020). All values reported in these studies were well below the glyphosate tolerance (maximum amount of a pesticide allowed to remain in a food) set by U.S. Environmental Protection Agency (EPA) which is 30 mg/kg for oats (Electronic Code of Federal Regulations, 2019).

To the best of our knowledge, besides the studies listed above, no additional wide studies in the scientific literature have focused on oat contamination by glyphosate. Overall, only a few studies are available for glyphosate and AMPA contamination in food (Ehling & Reddy, 2015; Xu et al., 2019). Moreover, glyphosate measurement was only implemented in the U.S. Food and Drug Administration (FDA) pesticide monitoring program in 2016 and the preliminary testing focused on soybeans, corn, milk, and eggs. Glyphosate was not targeted before by the U.S. FDA because the analysis requires the use of a separate analytical protocol compared to many pesticide residues that are currently being monitored (U.S. FDA, 2018). Results from the U.S. FDA 2017 report showed that over a two-year period (Federal Year 2016 and 2017), glyphosate was quantified in 59.5 % of corn and soy grain samples but no violative glyphosate levels were found in the tested samples.

The possible presence of glyphosate in oatmeal and oat-based cereals, which are widely consumed products, and the associated potential risks highlight the need to develop oat-based RMs for glyphosate. Only a limited number of food-based RMs are available for the analysis of pesticide residues and, to the best of our knowledge, no glyphosate RM or CRM is available for oatmeal or any other food matrices. However, quality control materials that are samples from past proficiency studies (FAPAS, 2019; Proof/ACS, 2019) with only short term stability data are available. CRMs and RMs are preferable over quality control materials because CRMs and RMs are typically available over longer periods of time due to long term stability assessment and larger production lots. In the case of this study, it was decided to first develop a RM instead of a CRM to expedite the release of the material.

Determination of glyphosate and AMPA is challenging due to their high polarity. Derivatization of glyphosate and AMPA is often required to ensure good sensitivity. Fluorenylmethylchloroformate (FMOC-Cl), leading to the formation of FMOC derivates, is the most common derivatization reagent (Demonte et al., 2018). For initial screening of candidate reference materials in this work, a protocol with derivatization using FMOC-Cl and solid-phase extraction (SPE) cleanup (Liao et al., 2018) was used to ensure ultra-trace levels of glyphosate in complex food matrices can be achieved. However, with analytical chemistry advances, glyphosate and AMPA can now be directly determined, without derivatization, with the use of hydrophilic interaction chromatography (HILIC) technology (Chen et al., 2013; López et al., 2019), and suppressed ion chromatography (Adams et al., 2017; Granby et al., 2003, Rajski et al., 2018). Multiple analytical methods are important for value assignment of RMs, and the use of a method without derivatization may likely be incorporated in future work.

The first aim of this study was to optimize and validate a robust method for the extraction and the analysis of glyphosate and AMPA in oat products including oatmeal, oat-based cereals and oat flour. As incurred materials are preferred for an RM, the optimized method was used to screen commercially-available oat products to identify candidate RMs. A total of thirteen products were obtained, including oatmeal and oat-based cereals, slightly to highly processed and oat flour from conventional and organic agricultural practices. This initial screening will help identify candidate reference materials with a wide range of glyphosate and AMPA mass fractions to proceed with RM development. Lastly, a preliminary stability study was conducted to evaluate the feasibility of a glyphosate in oat RM.

2. Experimental

2.1. Materials

2.1.1. Solvents and chemicals

All solvents were high-performance liquid chromatography (HPLC) grade and suitable for trace analysis. Acetonitrile (ACN), dichloromethane (DCM), methanol (MeOH) and water (H₂O) were purchased from Fisher Scientific (Hampton, NH).

For optimization of the liquid chromatography with tandem mass spectrometry (LC-MS/MS) method, glyphosate-FMOC (Dr Ehrenstorfer, 95.8 % purity) and AMPA-FMOC (Dr Ehrenstorfer, 99.89 % purity) were purchased from LGC Standards, (Manchester, NH, USA). Glyphosate (1000 μg/ml in H₂O, 99 % purity) and AMPA (100 μg/ml in H₂O, 99 % purity) solutions were purchased from RESTEK (State College, PA, USA). Glyphosate-1,2-¹³C₂,¹⁵N (100 μg/g in H₂O, Dr Ehrenstorfer, 97 % combined purity) and AMPA-¹³C,¹⁵N (100 μg/mL in H₂O, Dr Ehrenstorfer, 97 % combined purity) solutions were purchased from LGC Standards (Manchester, NH, USA).

Ammonium acetate (for molecular biology, ≥ 98 %), formic acid (puriss. p.a., ACS reagent, reag. Ph. Eur., ≥ 98 %) and hydrochloric acid (ACS reagent, 37 %) were purchased from Sigma Aldrich.

2.1.2. Solution preparation

Stock solutions of glyphosate-FMOC and AMPA-FMOC (25 μg/g, 5 μg/g, 0.5 μg/g and 0.1 μg/g) were prepared gravimetrically in MeOH and stored at −18 °C. For LC-MS/MS analysis, stock solutions were diluted in H₂O gravimetrically to obtain 1000 ng/g, 250 ng/g, 20 ng/g and 5 ng/g, respectively.

Glyphosate and AMPA solutions were mixed and diluted in H₂O gravimetrically to obtain a 50 μg/g glyphosate and 5 μg/g AMPA mixture (20-fold dilution of the purchased standards), and a 6 μg/g glyphosate and 0.6 μg/g AMPA mixture (166-fold dilution of the purchased standards).

Individual solutions of glyphosate and AMPA were mixed and diluted in H₂O to obtain a glyphosate 10 μg/g solution and an AMPA 2 μg/g solution. These solutions were mixed and diluted to obtain a mixture of glyphosate and AMPA at 100 ng/g each.

Glyphosate-1,2-¹³C₂,¹⁵N and AMPA-¹³C,¹⁵N solutions were mixed and gravimetrically diluted in H₂O to obtain a 10 μg/g glyphosate-1,2-¹³C₂,¹⁵N and 1 μg/g AMPA-¹³C,¹⁵N mixture. Individual solutions of glyphosate-1,2-¹³C₂,¹⁵N (100 μg/g in water) and AMPA-¹³C,¹⁵N (100 μg/g) were gravimetrically diluted in H₂O to obtain glyphosate-1,2-¹³C₂,¹⁵N 10 μg/g and AMPA-¹³C,¹⁵N 2 μg/g individual solutions. These solutions were mixed and gravimetrically diluted to obtain a mixture of glyphosate and AMPA at 400 ng/g. Solutions prepared in H₂O were stored at 6 °C.

2.1.3. Reagents preparation

FMOC reagent (20 mg/mL ACN) was prepared by diluting 5 g of FMOC chloride (HPLC derivatization, ≥ 99.0 %, Sigma Aldrich, St. Louis, MO, USA) in 250 mL of ACN. Borate buffer reagent (50 mmol) was prepared by diluting 4.8 g of sodium tetraborate decahydrate (ACS reagent, ≥ 99.5 %, Sigma Aldrich St. Louis, MO, USA) in 250 mL of H₂O. Ammonium hydroxide (NH₄OH) 2 % solution was prepared by diluting 5 mL of NH₄OH solution (28 % to 30.0 % NH₃ basis, Sigma Aldrich St. Louis, MO, USA) in 68 mL of H₂O.

2.2. Sample collection and preparation

Nine commercial breakfast cereals, including oatmeal and oat-based cereals, slightly to highly processed, from both conventional and organic agricultural practices, were purchased from local stores in the Gaithersburg, Maryland, area (Table 1). Conv_Oatmeal_1 and Conv_Oatmeal_1b samples were the same product and brand but from different lots. Before analysis, the breakfast cereals were ground into a fine powder with a Vitamix food processor. Every sample was extracted and analyzed in triplicate.

Table 1.

Sample codes and characteristics.

Code	Agriculture	Sample classification	Description
Org_Oatmeal_1	Organic	oatmeal	Instant oatmeal (whole grain rolled oats)
Org_Oatmeal_2	Organic	oatmeal	Instant oatmeal (whole grain rolled oats)
Conv_Oatmeal_1	Conventional	oatmeal	Instant oatmeal (whole grain rolled oats)
Conv_Oatmeal_1b	Conventional	oatmeal	Instant oatmeal (whole grain rolled oats)
Conv_Oatmeal_2	Conventional	oatmeal	Instant oatmeal (whole grain rolled oats) Addition of sugar and flavors
Conv_Oatmeal_3	Conventional	oatmeal	Instant oatmeal (whole grain rolled oats) Addition of sugar, colors, vitamins and flavors
Org_Oat-based_1	Organic	oat-based cereal	Oatmeal rings (whole grain oats) Addition of sugar, rice and honey
Conv_Oat-based_1	Conventional	oat-based cereal	Oatmeal rings (whole grain oats) Addition of corn starch, sugar and vitamins
Conv_Oat-based_2	Conventional	oat-based cereal	Oatmeal squares (whole grain oat flour, whole wheat flour) Addition of sugar, vitamins and flavors
Org_Oat-flour_1	Organic	oat flour	Whole grain oat flour
Conv_Oat-flour_1	Conventional	oat flour	Whole grain oat flour
Conv_Oat-flour_2	Conventional	oat flour	Whole grain oat flour
Conv_Oat-flour_3	Conventional	oat flour	Whole grain oat flour

Four commercial oat flours from both conventional and organic agricultural practices were purchased online. No sample preparation was performed before extraction and samples were extracted in duplicate.

2.3. Preliminary stability study

A preliminary stability study was performed to evaluate potential glyphosate or AMPA degradation in oat materials. An oatmeal sample from conventional agriculture, similar to Conv_Oatmeal_1 and Conv_Oatmeal_1b but obtained from a separate lot, was ground into a fine powder and extracted nine times to determine glyphosate and AMPA mass fractions at the beginning of the experiment (T0). The remaining ground oatmeal were sealed in four polyethylene bags that were then sealed in Mylar bags. Two bags were stored at room temperature (22 °C–23 °C) and two bags were stored in an oven set at 40 °C. Samples stored at room temperature and at 40 °C were extracted in triplicate after 16 weeks of storage and after 28 weeks of storage.

2.4. FMOC-Cl derivatization and SPE protocol

The extraction procedure was adapted from Liao et al. (2018), and the optimized protocol is described here. The ground cereal sample (1 g) was weighed in a 50 mL polypropylene centrifuge tube and an aliquot of the internal standard solution was gravimetrically added (20 μL to 50 μL of a solution containing 10 μg/g glyphosate-1,2-¹³C₂,¹⁵N and 1 μg/g AMPA-¹³C,¹⁵N prepared in H₂O). A 20 mL aliquot of a 50/50 mix of acidified H₂O (0.1 % formic acid) and MeOH, previously prepared and stored at −18 °C overnight, was added to the tube. The tube was mechanically rotated for 20 min at 10 rad/s (100 rpm) for uniform mixing and centrifuged for 30 min at 440 rad/s (4200 rpm) at 4 °C (Allegra X-14 R centrifuge, Beckman Coulter, Indianapolis, IN, USA). After centrifuging, 4 mL of the sample was transferred into a 15 mL polypropylene centrifuge tube containing 4 mL of the borate buffer reagent. After homogenization, 3 mL of the FMOC reagent was added. Samples were vigorously shaken and left 2 h in the dark during the derivatization reaction. Then, the pH of the samples was adjusted to 1.5 with addition of approximately 3 Pasteur pipettes drops of hydrochloric acid. Both MeOH and ACN were evaporated under a nitrogen (N₂) stream at 45 °C (approximately 1 h) and the tubes were centrifuged for 15 min at 419 rad/s (4000 rpm) to separate the excess sediment of FMOC solid. Prior to SPE, Oasis hydrophilic-lipophilic balance (HLB) SPE cartridges (3 cc, 60 mg, Waters, Milford, MA, USA) were conditioned with 3 mL of MeOH and 3 mL of acidified H₂O (0.1 % formic acid). The supernatant was loaded onto the SPE cartridges and the cartridges were rinsed with 3 mL of the acidified water and 3 mL of DCM. Compounds were eluted with 1.5 mL of a 70/30 MeOH/2 % NH₄OH solution mix in 2 mL glass vials. The eluate was evaporated under N₂ stream at 45 °C to 250 μL and H₂O was added to bring a final volume of 500 μL. Extracts were homogenized and filtered through a Whatman Mini-UniPrep syringeless filter with 0.45 μm polyvinylidene fluoride (PVDF) membranes (Sigma Aldrich St. Louis, MO, USA). Extracts were stored at 6 °C until LC-MS/MS analysis (no more than 7 d).

2.5. LC-MS/MS analysis

The analyses were performed on an Agilent (Santa Clara, CA, USA) 1290 Infinity LC coupled to an Agilent 6490A triple quadrupole mass spectrometer. The instrument was computer controlled using commercial software (MassHunter, Agilent). The separation was carried out on a C18 column with trifunctionally bonded ethylene bridged hybrid (BEH) particle (Acquity UPLC BEH C18 1.7 μm, Waters) with the following characteristics: 50 mm × 2.1 mm i.d., and 1.7 μm particle size. A guard column (Acquity UPLC BEH C18 1.7 μm, Waters, 5 mm × 2.1 mm i.d.) and a prefilter were added. Mobile phase A consisted of H₂O with 5 mmol/L ammonium acetate adjusted to pH 9 with the addition of ammonium hydroxide, and mobile phase B was ACN. The percentage of ACN in the chromatographic gradient was 5 % at 0 min, 5 % at 1 min, 35 % at 6 min, 95 % at 6.1 min, 95 % at 7.3 min, 5 % at 7.5 min, and 5 % at 9 min, with a total flow rate of 0.4 mL/min. Column temperature was set at 50 °C and injection volume was 20 μL. The MS analyses were carried out in negative polarity mode with a gas temperature of 180 °C, gas flow of 20 L/min, nebulizer pressure of 2.7 × 10⁵ Pa (40 psi), sheath gas temperature of 225 °C, sheath gas flow of 11 L/min, and capillary voltage of 3000 V.

Transitions were optimized using Agilent Optimizer by injecting 1 μg/g individual solutions of glyphosate-FMOC and AMPA-FMOC without a column. As glyphosate-1,2-¹³C₂,¹⁵N-FMOC and AMPA-¹³C,¹⁵N-FMOC are not commercially available, optimization of those transitions was not possible using the Agilent Optimizer. Transitions were deduced according to internal standard mass and glyphosate-FMOC and AMPA-FMOC transitions. Two transitions (unit resolution) were selected, one for quantification, the other for confirmation. Selected transitions and MS/MS parameters are given in the Supplementary Material (Table S1). This method was validated (to include linearity, matrix effects, accuracy, precision and limits of quantification) and used for screening all samples in Table 1 with the exception of the oat flour (Org_Oat-flour 1, Conv_Oat-flour 1, Conv_0at-flour 2, and Conv_Oat-flour 3).

The method was transferred to another instrument for the screening of oat flour, and for stability measurements. The second instrument used was an Agilent (Santa Clara, CA, USA) 1290 Infinity LC coupled to an Agilent 6460 triple quadrupole mass spectrometer. The same column and mobile phases were used with a modified solvent gradient. The percentage of ACN in the chromatographic gradient was 5 % at 0 min, 5 % at 0.5 min, 21 % at 7 min, 98 % at 7.1 min, 98 % at 8.5 min, 5 % at 8.6 min, and 5 % at 9.5 min, with a total flow rate of 0.4 mL/min. The MS analyses for this second method were also carried out in negative polarity mode with a gas temperature of 180 °C, gas flow of 11 L/min, nebulizer pressure of 2.7 × 10⁵ Pa (40 psi), sheath gas temperature of 225 °C, sheath gas flow of 11 L/min, and capillary voltage of 3000 V. The same transitions were used in both instruments. The accuracy of the transferred method using an Agilent 6460 was confirmed through matrix-matched spiked samples.

2.6. Glyphosate and AMPA quantification

Glyphosate-1,2-¹³C₂,¹⁵N and AMPA-¹³C,¹⁵N, isotopically labelled internal standards (IL-IS), were used to quantify glyphosate and AMPA, respectively using response factors (Rf). The Rf were calculated as follows:

$$\text{Rf}=({\text{m}}_{\text{x}}\times {\text{A}}_{\text{IS}})/({\text{A}}_{\text{x}}\times {\text{m}}_{\text{IS}}).$$

where m_x and m_IS are the mass fractions of the analyte and its internal standard, respectively, and A_x and A_IS are the areas of the analyte and its internal standard, respectively. The amount of IL-IS in samples was adjusted to fit the same range of expected concentration in the sample.

2.7. Method performance and validation

Performances were validated according to European Commission guidance document on analytical quality control and method validation procedures for pesticide residues and analysis in food and feed (European Commission, 2018). This includes linearity, matrix effects, accuracy, precision and limit of quantification.

Linearity was evaluated using matrix-matched samples by spiking a blank material at five different mass fraction levels: 5 ng/g, 100 ng/g, 400 ng/g, 1000 ng/g and 2000 ng/g for glyphosate, 5 ng/g, 10 ng/g, 40 ng/g, 100 ng/g and 200 ng/g for AMPA. Linear regression and squared correlation coefficient (r²) were calculated by plotting compound peak areas versus its concentration (n = 5).

Matrix effects (ME) were evaluated by comparing the slope of response curves prepared by the addition of standards in matrix samples (matrix-matched) and by the addition of standard directly in solvent. Samples were prepared at 5 ng/g, 100 ng/g, 400 ng/g, 1000 ng/g and 2000 ng/g for glyphosate and 5 ng/g, 10 ng/g, 40 ng/g, 100 ng/g and 200 ng/g for AMPA. ME was calculated according to the following equation:

$$\text{ME}=\left[\right(\text{matrix-matched response slope}-\text{solvent response slope})/(\text{solvent response slope}\left)\right]\times 100$$

Method accuracy and precision were evaluated with a recovery study. Matrix-matched spiked samples were prepared at 5 ng/g, 100 ng/g, 400 ng/g, 1000 ng/g and 2000 ng/g for glyphosate and 5 ng/g, 10 ng/g, 40 ng/g, 100 ng/g and 200 ng/g for AMPA (n = 5 to 15) and quantified using IL-IS. Accuracy was determined by the average recovery for each spike level tested and precision was determined with relative standard deviations (RSD) calculation for each spike level tested. The LOQ is then the lowest spike level meeting performance requirements with recovery between 70 % and 120 % and repeatability RSD ≤ 20 % (European Commission, 2018).

The accuracy of the method, including quantification using isotope dilution, was verified through a standard addition experiment. A sample consisting of conventional agriculture rolled oats (similar to Conv_Oatmeal_1 and Conv_Oatmeal_1b but obtained from a separate lot), with glyphosate and AMPA mass fractions that were previously estimated at 800 ng/g and 40 ng/g, respectively, was aliquoted into four samples and spiked with increasing amounts of glyphosate and AMPA: no spike (0 ng), 378 ng, 750 ng and 1131 ng of glyphosate, and no spike (0 ng), 39 ng, 77 ng and 116 ng of AMPA. In each sample, 500 ng of glyphosate-1,2-¹³C₂,¹⁵N-FMOC and 50 ng of AMPA-¹³C,¹⁵N-FMOC were added. Glyphosate and AMPA responses were normalized by the response of their labelled internal standards. Linear regressions were performed (glyphosate or AMPA normalized areas versus glyphosate or AMPA added amount) to determine the slope and the intercept. The amount of glyphosate and AMPA were determined by the absolute value of the x-intercept. Linear regression statistics were performed by OriginPro 2017 Software (Northampton, MA, USA).

Method reproducibility over multiple days was evaluated by the extraction of two different materials (ground oatmeal and oat flour from conventional agriculture) four times, at one-week intervals. Samples were extracted in triplicate.

2.8. Quality assurance

Precautions were taken to minimize laboratory background contamination. Before all LC-MS/MS analyses, instrumental performances (peak shapes, sensitivity) were checked. Water blanks were injected before each sample to avoid memory effects. For each extraction batch, method performance (LOQ and recovery) were checked with the extraction of spiked samples. Compound identification was verified for every batch using retention times and the ion ratio between quantification and confirmation of multiple reaction monitoring (MRM) transitions.

Potential sources of artificially introduced contamination during the extraction process were evaluated by processing and analyzing protocol blanks in an identical fashion to the cereal samples. Procedural blanks were prepared by the addition of IS in the extraction solvent and were subjected to the entire extraction protocol.

3. Results and discussion

3.1. Optimization of the extraction protocol: effect of reducing derivatization time and solvent evaporation prior to SPE

The extraction procedure was adapted from Liao et al. (2018) with some modifications to reduce derivatization time and to improve sensitivity. The derivatization reaction time was reduced from overnight to two hours as suggested by Demonte et al. (2018) in order to avoid a two-day experiment. No difference in derivatization was observed between a two hour and an overnight derivatization time; both conditions gave complete and reproducible reactions.

Another modification of the protocol was to fully evaporate MeOH and ACN prior to the SPE cleanup step. When the solvents were evaporated completely, glyphosate-FMOC and AMPA-FMOC signals increased by a factor of 25 (Figure S1, Supplementary Material) compared to when the solvents were not evaporated. The evaporation of the solvents also allowed for the excess FMOC-Cl to precipitate and prevented subsequent transfer to the SPE cartridges, avoiding FMOC-Cl interferences in the final extract.

3.2. Method validation: linearity, matrix effects, accuracy, precision, LOQ and reproducibility

An organic oatmeal was chosen for the blank material to prepare matrix-matched samples, however, detectable amounts of glyphosate (≈1 ng/g) and AMPA (≈1 ng/g) were found in this material. Guidance for method validation procedures for pesticide residues (European Commission 2018) recommend that spiking levels should be at least three times the level present in the blank material. Given the intrinsic background of glyphosate and AMPA present in the blank matrix, no matrix-matched spike samples under 5 ng/g were prepared during this study. Procedural blanks did not indicate the presence of glyphosate. Despite all the precautions taken, AMPA was quantified in every procedural blank, at levels corresponding to 0.5 ng/g. This result is not surprising as AMPA is not only the degradation product of glyphosate but can also result from the degradation of phosphonates that are commonly used in industrial and household applications (Grandcoin et al., 2017).

Linearity was evaluated by plotting compound peak area versus mass fraction of the matrix-matched calibrants. Determination coefficients (r²) were 0.9989 ± 0.0006 (s.d., n = 5) for glyphosate and 0.9987 ± 0.0013 (s.d., n = 5) for AMPA matrix-matched calibration curves, demonstrating that the method is linear and appropriate for glyphosate levels ranging from 5 ng/g to 2000 ng/g and for AMPA levels ranging from 5 ng/g to 200 ng/g.

Matrix effects (ME) were evaluated by comparing the slope of matrix-matched and solvent calibrations (Figure S2, Supplementary Material). For both glyphosate and AMPA, good linearity was achieved with r² greater than 0.9990 for matrix-matched calibration and greater than 0.9970 for solvent calibration. Strong positive matrix effects were calculated, 197 % for glyphosate and 50 % for AMPA, highlighting that the use of the isotopically labelled internal standards and matrix-matched calibration are required for the accurate quantification of glyphosate and AMPA in oat samples.

The results for the glyphosate and AMPA recovery study are reported in Figure 1. For this study, glyphosate and AMPA were quantified using IL-IS. For each spike level tested, accuracy and precision were validated as recoveries were all comprised between 92 % and 111 % (70 % to 120 % required) and RSD were all under 8 % (≤ 20 % required) (European Commission, 2018). The LOQ is then defined by the lowest level of validated spike: 5 ng/g for both glyphosate and AMPA. The signal-to-noise values for the lowest validated spikes for glyphosate and AMPA were around 500 and 150, respectively, suggesting that lower LOQ can be achieved. Because no lower matrix spiked samples can be prepared due to trace levels of glyphosate and AMPA in the matrix blank, the extraction of solvent spiked samples at 2.5 ng/g (n = 3) verified that lower detection limits can be achieved. For this experiment the average recovery of glyphosate was 102 % and the RSD was 6 % for the 2.5 ng/g solvent spiked sample. For the quantification of oat products in this work, the glyphosate LOQ was determined to be 5 ng/g (determined by the lowest level of validated spike). Measured levels between 1 ng/g and 5 ng/g are reported as “detected but not quantified”. Because AMPA levels in the protocol blanks were 0.5 ng/g, the AMPA LOQ was set to 10 times this level at 5 ng/g (in accordance with the lowest verified spiked level) and no results under that limit are reported.

Figure 1.

Glyphosate and AMPA recoveries obtained at different levels of spiked samples (number of replicates, RSD).

The accuracy of the quantification by isotope dilution was confirmed through a standard addition experiment. The standard addition plot flor glyphosate and AMPA is given in Figure S3 (Supplementary Material), where compound area (normalized by the area of the IS in the sample) was plotted versus the amount of compound added. Linearity for both glyphosate and AMPA was good with r² values of 0.9993 and 0.9950, respectively. Intercept with the abscissa gave the mass in the sample (1 g extracted). Standard errors of the intercept and the slope were used to estimate the standard error on the calculated mass fractions. Mass fractions determined by standard additions were 839 ng/g ± 9 ng/g and 39 ng/g ± 2 ng/g for glyphosate and AMPA, respectively. Comparatively, the mass fractions determined with the internal standard method in this sample were 824 ng/g ± 25 ng/g for glyphosate and 37 ng/g ± 2 ng/g for AMPA. The uncertainties were estimated using a 3 % variation for glyphosate and a 5 % variation for AMPA based on the variability at similar levels presented above in the recovery study (Figure 1). The mass fractions determined by standard addition and internal standard quantification agree for both glyphosate and AMPA, providing additional confirmation of the accuracy of the method.

The intermediate reproducibility of the method over multiple days was evaluated by the extraction of two different materials (ground oatmeal, similar to Conv_Oatmeal_1 but from a different lot, and oat flour from conventional agriculture, similar to Conv_Oat-flour_1 but from a different lot) four times, at one-week intervals. Results are presented in Figure 2 and Figure 3. For the oatmeal sample, average mass fractions (calculated for 3 replicates) over the four weeks were 703 ng/g, 766 ng/g, 732 ng/g and 708 ng/g for glyphosate and were 30 ng/g, 30 ng/g, 31 ng/g and 30 ng/g for AMPA, respectively (Figure 2). Glyphosate and AMPA within-day variability were low, with a RSD for glyphosate < 1.7 % and a RSD for AMPA < 2.5 %. For glyphosate the average mass fractions of the four experiments was 723 ng/g with a between-day RSD of 4 %. For AMPA, the average of mass fractions of the four experiments was 30 ng/g with a between-day RSD of 2 %. For the oat flour sample, average mass fractions over the four weeks were 501 ng/g, 532 ng/g, 486 ng/g and 468 ng/g for glyphosate and were 18 ng/g, 19 ng/g, 17 ng/g and 17 ng/g for AMPA (Figure 3). Glyphosate within-day variability was low (RSD < 3 %) except for the second week experiment where the RSD was 10 %. This was linked with an unexpected high level (590 ng/g) measured in one replicate, which remains unexplained. AMPA within-day RSD was under 6 %. For glyphosate the average of mass fractions of the four experiments was 497 ng/g with a between-day RSD of 7 %. For AMPA, the average of mass fractions of the four experiments was 18 ng/g with a between-day RSD of 4 %. While the reproducibility results were better for the oatmeal sample compared to the oat flour, the variability values for both materials were acceptable as they were under the SANTE/11813/2017 guidance document criterion for within-laboratory reproducibility (20 %), showing good intermediate reproducibility of the method over the course of several weeks.

Figure 2.

Glyphosate and AMPA mass fractions (ng/g) measured in an oatmeal sample over four weeks. The within-day average ± s.d is provided for each day of extraction. The between-day average, s.d, and RSD over the four-week period are also provided.

Figure 3.

Glyphosate and AMPA mass fractions (ng/g) measured in an oat flour sample over four weeks. The within-day average ± s.d is provided for each day of extraction. The between-day average, s.d, and RSD over the four-week period are also provided.

The proposed optimized method meets the performance requirements to be applied for the extraction and analysis of real oatmeal and oat-based cereal samples.

3.3. Glyphosate and AMPA screening in oat products

The optimized method was used to screen thirteen oat products for glyphosate and AMPA to identify potential candidate RMs. The samples included oatmeal, slightly to highly processed oat-based breakfast cereals, and oat flour from conventional and organic agricultural practices. Graphic representation of the amount of glyphosate and AMPA in the samples are provided in Figure 4. Glyphosate was detected in all thirteen oat samples. The lowest levels were obtained in breakfast cereals with grains obtained from organic agricultural practices. Glyphosate was detected but not quantified in two organic cereal samples and was quantified in one organic cereal sample (26 ng/g) and in the organic oat flour (11 ng/g). In conventional oatmeal, glyphosate levels were 1100 ng/g, 837 ng/g, 62 ng/g and 277 ng/g, in oat-based cereals levels were 768 ng/g and 901 ng/g and in oat flour levels were 535 ng/g, 7 ng/g and 554 ng/g. These levels are in accordance with those reported by the non-profit organization EWG (EWG, 2018). The highest glyphosate mass fraction (1100 ng/g) was obtained for one of the conventional instant oatmeal samples (two lots of the same product were analyzed). All of these samples were under the tolerance set by the EPA for oats which is 30 ppm (30 μg/g) (Electronic Code of Federal Regulations, 2019). Oatmeal and oat-based cereal samples were extracted and analyzed in triplicate. RSDs for all samples were all < 10 %, showing the applicability of the method to real samples. Because oat flour samples were only extracted and analyzed in duplicate, no precision statements for these materials will be made.

Figure 4.

Glyphosate and AMPA mass fractions (ng/g) measured in the oatmeal, oat-based breakfast cereals and oat flour (NQ = <LOQ (5 ng/g), detected but not quantified indicates that glyphosate levels were between 1 ng/g and 5 ng/g). Error bars represents the standard deviation of n=3 replicates (oatmeal and oat-based samples) or n=2 replicates (oat flour samples).

In this study, AMPA was present at low levels and was not quantified in the organic samples tested. In conventional samples, AMPA was quantified at 40 ng/g, 6 ng/g and 19 ng/g in oatmeal, at 23 ng/g and 25 ng/g in oat-based cereals and at 20 ng/g, 15 ng/g and 25 ng/g in oat flour. AMPA levels are well below glyphosate levels, with a glyphosate/AMPA ratio between 10 and 35. Oatmeal and oat-based cereal samples were extracted and analyzed in triplicate. RSDs were all < 10 % for all samples, demonstrating the applicability of the method for AMPA determination in real samples. Because oat flour samples were only extracted and analyzed in duplicate, no precision statements for these materials will be made.

3.4. RM candidate selection

Candidate RMs must have mass fractions of the targeted compounds that are fit of purpose as quality control materials and in accordance with the current method validation needs of stakeholders. For pesticide residues in food, the tolerance or maximum residue limit (MRL) level is often a good value to use in preparation of an RM as laboratories must show good analytical performance at that level (European Commission, 2018). In the case of glyphosate, the tolerance set for oats is much higher than most pesticide residues. For example, the tolerance level set by the U.S. EPA is 30 μg/g and the MRL set by the European Union (EU) is 20 μg/g (Electronic Code of Federal Regulations, 2019; EU Pesticides Database - European Commission, 2019). Stakeholders, however, have expressed little interest for a RM for glyphosate at levels near the US and EU tolerance levels, and some stakeholders have expressed interest in RMs with levels between 50 ng/g and 500 ng/g for food product specification applications for evaluation of organic agriculture products. These requested glyphosate levels align with the glyphosate levels that were determined in commercial oat-based products investigated in this study. From initial discussions from stakeholders and from results in this study, producing a RM with a “high” and a “low” glyphosate level would be of interest to the measurement community. No specific levels have been requested by stakeholders for AMPA.

In addition to choosing appropriate mass fraction levels that will be fit for purpose, it is critical to produce a material that is homogenous. Therefore, a homogeneity study must be conducted before the material is released as a reference material. This entails a study on the final packaged production lot using stratified random sampling to ensure the material is homogenous across the entire production lot. A detailed homogeneity study is forthcoming because RM production is only at the material selection phase. However, a consideration of homogeneity during the early stages of RM development has been conducted to ensure that the material is processed appropriately (ground, mixed, etc.) before it is packaging and that a minimal yet fit-for-purpose sample size is achievable.

The sample protocol for screening oat products in this study included either grinding the material (oatmeal and oat-based cereals) or using a previously ground material (commercial oat flour) before extraction and analysis. The reproducibility data for both laboratory ground and previously ground samples produced acceptable RSDs (Figure 2 and Figure 3), conveying that the ground materials are likely more homogenous than cereals. In contrast, extracting intact rolled oats produces relatively large RSDs (> 45 %; data not shown). The material in this work was ground on an analytical scale using a blender. The oat material will either be milled during RM production or purchased pre-ground and mixed uniformly to ensure adequate homogeneity of the resulting product.

In this work, only one sample size (1 g) was investigated. Typical sample sizes found in the literature range from 1 g to 10 g (Adams et al., 2017; Chen et al., 2013; Granby et al., 2003; Liao et al., 2018; López et al., 2019; Rajski et al., 2018). Future work includes an in-depth homogeneity study across the entire production lot of the final packaged material together with a determination of a recommended minimal sample size.

Considering all those elements, Conv_Oatmeal_1, Conv_Oat-flour_1 and 3 could be selected as potential RM for the high level while Org_Oatmeal_1 and Org_Oat-flour_1 would be good candidates for a low-level glyphosate RM. Multiple samples can be well blended to achieve desired level RM materials.

3.5. Preliminary Stability study

A six-month preliminary stability study was conducted with a ground oatmeal sample (similar to Conv_Oatmeal_1) stored both at room temperature and at 40 °C. Results are presented in Figure 5. At T0, glyphosate initial mass fraction was 1027 ng/g. After 28 weeks of storage, glyphosate mass fractions were still in the reference variability range, even for the sample stored in the “extreme” condition (40 °C). These initial results suggest that glyphosate could be stable stored at room temperature. For AMPA however, initial mass fraction was 39 ng/g and mass fractions after 16 weeks of storage were under the reference variability range. However, AMPA appears stable between 16 and 28 weeks. More data are needed to conclude on AMPA stability.

Figure 5.

Glyphosate and AMPA mass fractions (ng/g) after 16 and 28 weeks of storage at room temperature and 40 °C. Error bars represents the standard deviation of n=9 replicates (T0) or n=3 replicates (16 and 28 weeks of storage samples).

To the best of our knowledge, there are no published data regarding glyphosate and AMPA stability in oat samples. However, a recent study reported that glyphosate was stable in wheat stored at room temperature or 4 °C over a 15 months period (Tittlemier et al., 2017).

4. Conclusions

A SPE-LC-MS/MS protocol for the analysis of glyphosate and AMPA in oats was optimized and successfully validated. This method was applied for the screening of commercially available oatmeal, oat-based breakfast cereals and oat flour in the aim of identifying potential candidate RMs. The initial findings of this study provide a useful survey of glyphosate and AMPA levels in a small subset of oat products. Over the thirteen samples tested, three appeared as potential candidates for glyphosate levels found in typical oat-based products, and two could be used for a low-level glyphosate in oats RM. A preliminary stability study showed that glyphosate was stable in oats over a 6 months period both at room temperature and 40 °C. This result is encouraging for the stability of the future reference material.

Future work includes packaging of candidate reference materials, a detailed homogeneity study, and a long-term stability of glyphosate and AMPA in such a material After the release of the RM, NIST will continue to engage with stakeholders to evaluate if the RMs need to be adapted, for example, adding/changing matrices, analytes, levels, and/or upgrading to an SRM if needed.

Highlights.

• A SPE-LC-MS/MS method was validated for the analysis of glyphosate and AMPA in oats

• Glyphosate was detected in all the samples and quantified up to 1100 ng/g

• AMPA was only quantified in 9 samples and mass fractions were lower (< 40 ng/g)

• Oat products were selected for potential candidate reference materials

• Glyphosate was stable in a ground oat material for six months