Journal of Chromatography A, 1549 (2018) 31–38

Contents lists available at ScienceDirect

Journal of Chromatography A
journal homepage: [Link]/locate/chroma

Validation and application of analytical method for glyphosate and

glufosinate in foods by liquid chromatography-tandem mass
spectrometry
Yang Liao a,c , Jean-Marie Berthion a , Isabelle Colet a , Mathilde Merlo b ,
Alexandre Nougadère b , Renwei Hu a,∗
a
INOVALYS – Le Mans, Official Laboratory of Analysis, 72018, France
b
French Agency for Food, Environmental and Occupational Health & Safety(ANSES), Risk Assessment Department, Maisons-Alfort, 94701, France
c
Sichuan Normal University, College of Chemistry and Materials Science, Chengdu, 610068, China,

a r t i c l e i n f o a b s t r a c t

Article history: A reliable and sensitive method was developed for simultaneous determination of glyphosate and glufosi-
Received 22 November 2017 nate in various food products by liquid chromatography-tandem mass spectrometry. Based on extraction,
Received in revised form 3 March 2018 derivatization with 9-fluorenylmethylchloroformate and purification on solid phase extraction column,
Accepted 17 March 2018
quantification was done by using isotopic-labeled analytes as internal standard and calibration in matrix.
Available online 20 March 2018
Good selectivity and sensitivity were achieved with a limit of quantification of 5 ␮g/kg. The recoveries of
these two pesticides ranged from 91% to 114% with inter-day and relative standard deviation of 3.8–6.1%
Keywords:
in five matrices of cereal group spiked at 5, 10, and 20 ␮g/kg. An accuracy profile was performed for
Glyphosate
Glufosinate
method validation, demonstrating the accuracy and precision of the method for the studied food groups.
Food analysis The verification results in expanded food groups indicated extensive applicability for the analysis of
Accuracy profile glyphosate and glufosinate. Finally, the developed method was applied to analyze 136 food samples
Validation including milk-based baby foods from the French Agency for Food, Environmental and Occupational
LC–MS/MS Health & Safety. Glyphosate residues were detected in two breakfast cereal samples (6.0 and 34 ␮g/kg).
Glufosinate residues were found in a sample of boiled potatoes (9.8 ␮g/kg). No residues were detected in
the other samples, including milk-based baby foods with limits of detection ranging from 1 to 2 ␮g/kg.
The method has been applied for routine national monitoring of glyphosate and glufosinate in various
foods.
© 2018 Elsevier B.V. All rights reserved.

1. Introduction the analysis of these two pesticides in different kinds of food,

including baby foods.
Glyphosate [N-(phosphonomethyl) glycine] and glufos- Glyphosate and glufosinate are belonging to the amino acid
inate [DL-homoalanine-4-yl(methyl) phosphonic acid] are class of pesticides. These two compounds are challenging to
non-selective, broad spectrum herbicides and widely used for analyze at low concentration level due to their amphotericity,
weed control in agriculture, forestry, and urban applications [1,2]. highly polarity, non-volatility, and lack of chromophores or fluo-
They are considered to be non-persistent and safe in environ- rophores [8]. Until now, many methods have been applied for the
ment for their rapidly degradation in water and soil. However, analysis of glyphosate and glufosinate, mainly including gas chro-
glyphosate and glufosinate have attracted a high level of public matography coupled with mass spectrometry (GC–MS) [9], or gas
and scientific interest, because they are frequently detected in chromatography tandems mass spectrometry (GC–MS/MS) [10],
plants and soil, even in foods and human urine [3–7]. In order to liquid chromatography coupled with fluorescence (LC-FLD) [11],
better estimate human dietary exposure, it is necessary to expand ultraviolet (LC-UV) [12] and mass spectrometry (LC–MS) [13], liq-
uid chromatography-tandem mass spectrometry (LC–MS/MS) [14],
and so on. To reduce method complexity and time consumption,
the capillary electrophoresis and hydrophobic interaction chro-
matography, inductively coupled plasma MS (ICP-MS) were used
∗ Corresponding author. to analyze glyphosate without prior derivatization achieving sepa-
E-mail address: [Link]@[Link] (R. Hu).

[Link]
0021-9673/© 2018 Elsevier B.V. All rights reserved.
32 Y. Liao et al. / J. Chromatogr. A 1549 (2018) 31–38

ration from matrix component [15–17]. In most of these methods, (from SIGMA). SPE-C18 column was obtained from Waters and
LC–MS/MS is one of the most commonly used methods in many 9-fluorenylmethylchloroformate (FMOC) from SIGMA. Water was
laboratories for its accuracy and lower the limit of detection (LOD). purified by the Milli-Q water purification system (Millipore).
Though glyphosate can also be directly determined by Standard stock solutions of glyphosate and glufosinate
LC–MS/MS in foods as the Europe reference method “Quick Polar (1 mg/mL) were prepared in the Milli-Q water and stocked in the
Pesticides (QuPPe) method”, its sensitivity and selectivity are not plastic bottles at −20 ◦ C. The intermediate solutions (100 ␮g/mL
satisfactory at the level 10 ␮g/kg for baby food [18,19]. As we all and 10 ␮g/mL) were prepared from standard stock solutions.
know, the fruits and vegetables are rich in sugar and pigments, and The composite working solutions (1 ␮g/mL and 0.1 ␮g/mL) were
milk, cheese and butter to contain a lot of protein and fat. The com- diluted with Milli-Q water in the plastic bottles at 4 ◦ C. FMOC solu-
plexity of these matrices greatly leads to interference in the target tion was 20 mg/mL in acetonitrile. Disodium tetraborate was used
compound. To gain good sensitivity and selectivity, derivatization for the borate buffer (0.05 mmol/mL). The 100 mL eluent solvent
and purification by using solid phase extraction (SPE) have been was prepared as following: 10 mL NH4 OH (20%), 40 mL the Milli-Q
used as current popular pretreatment methods. The most common water and 50 mL methanol were added into a glass volumetric flask
approach to analysis of glyphosate and glufosinate always consists of 100 mL. A mixture of the two internal standards covered all the
of derivatization with 9-fluorenylmethyl-chloroformate(FMOC), sample including calibration standard, matrix blank and samples
and the column of C18 was used for purification and concentration at the concentration of 1 ␮g/mL.
and subsequent detection with LC–MS/MS [20,21].
As mentioned, most methods are generally suitable for only one 2.2. Sample preparation
matrix or one kind of matrix such as water, plants, cereals and
soil matrices [22–24]. It is being challenged to expand the scope 136 food samples came from ANSES. According to the European
of food samples and complex matrices and low concentration lev- Union SANTE guideline [27], all the samples were divided into 3
els stand in the way. The ideal method needs to be suitable for a groups for analysis (shown in Table 1). The samples with high water
wide range of matrices, and easy to carry out for most laboratories, content (≥60%) accounted for 21% of the samples analyzed (mainly
while being robust, selective, accurate and precise. In addition, the vegetables, fruits and drinks), and the ones with low water content
French Agency for Food, Environmental and Occupational Health such as biscuits or bread, for 13% of the samples. The products from
& Safety (ANSES) launched the first French Infant Total Diet Study animal origin, including milk-based baby foods represented 66% of
(iTDS), in order to estimate the dietary exposure and dietary risk the samples.
to infants and young children below 3 years old in 2012. This pro- Some samples need to be chopped and homogenized in a mixer
gram demanded to determine glyphosate and glufosinate residues (model Retsch-GM200) according to the documentation “Commis-
in various food samples, which are including plant products with sion Directive 2002/63/EC” [28]. All samples were stored at −20 ◦ C
different water, protein and sugar contents, and animal product until analysis.
such as meat, cheese and milk. Although the maximum residue
limits (MRLs) for two herbicides in food commodities have been 2.3. Extraction procedure
established in European Union (10–20 ␮g/kg), the limit of quantifi-
cation (LOQ) for baby food is required lower (5 ␮g/kg). Hence, it An accurately weighed homogenized sample was put into a
was imperious to develop a determination method for glyphosate 50 mL plastic centrifuge tube. A mixture of solvents including Milli-
and glufosinate applicable to a wide range food groups with high Q water, acidified water, methanol and dichloromethane (shown
sensitivity and selectivity. in Table 2) were added into the tube. The tube was rotated
This study describes a laboratory validation of a LC–MS/MS mechanically for uniform mixing for 20 min and then centrifuged
method for the determination of low concentrations of at 3000 rpm for 20 min at 4 ◦ C or −15 ◦ C (Rotanta 460R, Table 2).
glyphosate and glufosinate in various food products. The The aqueous extract of sample was transferred into a 20 mL plastic
extraction with a solvent mixture including acidified water, tube for derivatization. The DCM organic phase was removed.
methanol and dichloromethane, derivatization with 9- For each analysis series, a calibration curve (5–10–25–50 ␮g/kg)
fluorenylmethylchloroformate (FMOC) and purification conditions in the same matrix group, a reagent blank, a matrix blank (free of
by using solid phase extraction (SPE) column have been optimized. analytes like glyphosate and glufosinate) and a spiked sample at
An isotopic method has been used for the quantification of two 5 ␮g/kg were done.
compounds by LC–MS/MS. The whole method has been validated
and applied to 136 food samples provided by ANSES [25,26], such 2.4. Derivatization and purification procedure
as milk-based baby foods but also “conventional” food products
not specifically intended to infants and young children such as The derivatization steps were performed as follows: 4 mL of the
bread, biscuits, banana, apple, orange, potato, carrot, milk, cheese, aqueous extract of sample was pipetted into a glass tube of 20 mL
butter, juice, and so on. together with 50 ␮L of the labeled internal standard (1 ␮g/mL), 4
mL of the borate buffer (0.05 mmol/mL, pH 9) and 4 mL of FMOC
reagent (20 mg/mL in acetonitrile) and the tube was swirled and left
2. Material and methods overnight at room temperature. After that, 3 drops of hydrochloric
acid (HCl, 37%) was added into the tube and pH value was adjusted
2.1. Chemicals and reagents to 1.5.
A SPE-C18 cartridge (60 mg) was used to purify and concentrate
Glyphosate (purity 98%) and glufosinate (purity 97.5%) were the sample extract after derivatization. The cartridge was condi-
provided by Cluzeau Info Labo (CIL, France) as certified analyti- tioned with methanol (3 mL) and then water solution (3 mL, with
cal standards. Glyphosate (1,2-13 C, 15 N) and glufosinate (D3), used 0.1% formic acid). The total of the derivatized acidified extract was
as internal standards, were also obtained from CIL. The solvents added into the cartridge and the cartridge was washed by acidified
used were LC-grade acetonitrile (purchased from CARLO ERBA), water (3 mL, with 0.1% formic acid) and dichloromethane (3 mL).
formic acid, methanol (MeOH) and dichloromethane (from VWR). The eluent (1 mL) was added and collected to a vial of 2 mL. The
Other solvents and reagents were analytical grade such as dis- eluate was evaporated to dryness at 50 ◦ C under the protection
odium tetraborate decahydrate and ammonium acetate (NH4 Ac) of nitrogen. The residue was dissolved with 200 ␮L of a solvent
 Y. Liao et al. / J. Chromatogr. A 1549 (2018) 31–38 33

Table 1
Classification, group and food product of analyzed samples.

Classification Groups Food products (sample number)

Plant products with high 1. High water content (19 samples) Bananas (1), kiwis (1), peeled pears (1), peeled apples (1), broccolis (1),
water content (≥60%) carrots (1), zucchinis (1), cauliflowers (1), spinaches (1), beans (1),
mixed vegetables (1), boiled potatoes (1*), fried potatoes (1), peas (1),
leeks (1), pumpkins (1), mashed potatoes (1), vegetable soup (1),
tomatoes (1)
2. High acid content and highwater content (10 samples) Oranges (1), clementines/mandarins (1), fruit juices (5), drinks with
milk and fruits (1), diluted syrup extracted from fruits (1), sodas (1)

Plant products with low 3. High sugar and low water content (2 samples) Fruit compote except apples (1), applesauce (1)
water content (<60%) 4. High starch and/or protein content and low water and Fruit biscuits (1), natural or fruit cereals (1a ),brioche and brioche bread
fat content (6 samples) (1),chocolate bread (1*), chocolate mousse (1), chocolate cereals (1)
5. “Difficult or unique commodities”(10 samples) Cocoa powder (1), chocolate milk (1),chocolate spreads (1),Other
products (7)

Products of animal origin 6. Meat (muscle) and seafood (1 sample) Ravioli filled pasta (1)
7. Milk-based baby foods and other milk products (80 Infant formulaefor infants below 6 months (27), infant formulae (6–12
samples) months) (35), follow-on formulae (9), other milk products (6), yogurt
(3)
8. Fat from food of animal origin (8 samples) Cheese (5), cream (2), butter (1)
a
The residues of glyphosate or glufosinatewere found in samples.

Table 2
The extraction solvents for glyphosate and glufosinate.

Classification Sample (g) Acidified water (mL) Methanol (mL) Dichloromethane (mL) Centrifugation temperature (◦ C)

High water content 5 20 – 20 4

Low water content 2.5 10 10 10 −15
Products of animal origin 5 10 10 10 −15

mixture (acetonitrile, the mobile phase A for LC–MS/MS and ammo- FMOC and (404.1 → 136.1 for quantification, 404.1 → 208 for
nium acetate at 0.05 mmol/L in water, the mobile phase B; 10/90; qualification) of glufosinate-FMOC. 395.2 → 91.1 transition of
V/V). The final extract was then transferred into the conical vial for glyphosate (1,2-13 C, 15 N)-FMOC and 407.1 → 139.1 of glufosinate
LC–MS/MS analysis after the ultrasonic shock. (D3)-FMOC were used for internal standard quantification.

2.5. Analysis by LC–MS/MS

3. Results and discussions
2.5.1. Chromatographic conditions in liquid chromatography (LC)
Dionex UltiMate 3000 LC system was used and equipped 3.1. Pretreatment optimization
with a C18 column (Atlantic150 × 2.1 mm, 2.6 ␮m, Waters).
Acetonitrile was used as the mobile phase A and ammo- 3.1.1. Extraction procedure
nium acetate at 0.05 mmol/L in water (pH = 5.0), as the mobile Based on the polarity and water solubility of glufosinate and
phase B. Correspondingly, the percentage of acetonitrile was glyphosate, water, acidified water (0.1% formic acid), methanol
changed linearly as follows: 0 min, 5%; 2.5 min, 5%; 8.0 min, (MeOH), and dichloromethane (DCM) were used as extraction sol-
95%; 10.0 min, 95%, and 10.1 min, 5% with a total flow rate vent in the matrix samples at spiked level 100 ␮g/kg. When using
of 200 ␮L/min and the injection volume of 5 ␮L. The retention just water as extraction solvent, the recoveries were not high, even
times for glyphosate-FMOC/glyphosate (1,2-13 C,15 N)-FMOC and only 50% for glyphosate in the milk as Fig. 1 shown. The choice
glufosinate-FMOC/glufosinate (D3)-FMOC were respectively 7.0 of an acidified water (0.1% formic acid) as extraction solvent was
and 7.2 min. The total run time was 20.0 min. dictated by the very polar character of both glyphosate and glufos-
inate and their insolubility in most organic solvents. Although the
2.5.2. Mass spectrometry conditions recovery can be improved with acidified water (0.1% formic acid),
TSQ Quantiva mass spectrometer (Thermo Fisher Scientific) the recovery was not satisfied with milk sample and interferences
with an electrospray ionization (ESI) source was performed for MS persisted and affected the analyst. The main problem with this pro-
analysis. cedure is the presence of other water-soluble components in the
The mass spectrometer analyses were carried out in positive ion extract that will hamper the analyte. Therefore, two other solvents
mode and the ion source type was heated electrospray ionization were added to remove interferences: methanol for the removal of
(HESI). The selective reaction monitoring (SRM) mode was used. protein and starch precipitation (only for the matrix in medium and
The MS instrumental settings parameters were shown as below: high protein content such as cereals, meet. . .) and dichloromethane
spray voltage was 4000 V, the sheath gas and auxiliary gas pressures (DCM) (for all the matrix) for the removal of hydrophobic matrix
were 40 (Arbitrary) and 15 (Arbitrary) respectively, ion transfer components or non-ionic low to medium water solubility matrix
tube temperature was 300 ◦ C, and the vaporizer temperature was components such as lipids. A simultaneous process of extraction
250 ◦ C. The capillary temperature was 320 ◦ C and the collision gas and purification was used to minimize the sample handing time.
(Ar) pressure was 1.5 mTorr. The obtained data were handled by Adding dichloromethane and methanol for extraction improved the
Xcalibur software. recovery of glyphosate from 50% to 92% and glufosinate from 70%
[M+H]+ ions (m/z) were selected as the precursor ions and to 98%. After a series of experiments, the proportion between acidi-
two transitions were selected for each derivative (392.2 → 88.2 fied water (0.1% formic acid), methanol and dichloromethane were
for quantification, 392.2 → 214.1 for qualification) of glyphosate- chosen which is shown in Table 2.
34 Y. Liao et al. / J. Chromatogr. A 1549 (2018) 31–38

solvent mixture of the mobile phases of LC–MS/MS to reduce sol-

vent effect. Finally, the sample extract was concentrated 20 times
(from 4 mL to 0.2 mL) before the analysis on LC–MS/MS. Following
the above steps, the recovery of glyphosate could reach 92% and
one of glufosinate, 94% in the milk.
A combination of several steps as indicated below makes our
method more sensitive and selective: an efficient extraction with a
solvent mixture, the FMOC derivatization, a SPE-C18 cartridge for
purification and the concentration 20 times (from 4 mL to 0.2 mL)
for sample extract before instrument analysis and finally, the deter-
mination by LC–MS/MS with its high sensibility and selectivity.
Therefore, the combination of all these factors leads to the high
sensitivity and selectivity of our method.
In addition, a calibration curve in the same matrix group was
used to quantify glyphosate and glufosinate in the presence of the
labeled internal standard (glyphosate 13 C, 15 N and glufosinate D3)
to reduce the matrix effect.

3.2. Method validation

Fig. 1. Influence of extraction solvent on the recoveries for glyphosate and glufosi-
nate. Validation of the developed method was performed as described
in the following documents: European Union SANTE guideline [27]
3.1.2. Derivatization conditions and French Standard NF V03-110 [29]. The calibration curves were
Because of high polarity, retention in reversed phase columns obtained by the linear analysis (Y = aX + b and linear 1/X) at matrix-
and sensitivity in ESI/MS is either low in the case of underivatized matched calibration, which was conducted at four levels ranging
glyphosate and glufosinate. Since the work of Moye and Boning between 5, 10, 20 and 100 ␮g/kg with two replicates for five days.
[30], who were the first to use FMOC for glyphosate derivation, The accuracy was evaluated by the recoveries (%) and the precision
many procedures have been reported in the literatures [20,21]. was expressed by RSDR (%) of the spiked samples. The applica-
After derivatization, analytes’ polarity is significantly reduced tion fields, the limit of detection (LOD), the limit of quantification
enabling better retention on the reversed phase and thereby bet- (LOQ), and the accuracy and precision of the method were discussed
ter separation from matrix components. In addition, the ionization respectively as below.
efficiency of analytes in ESI source was enhanced by its hydropho-
bic parts in molecules and limit of detection (LOD) was improved.
3.2.1. Choice of the food groups and food categories
Therefore, the FMOC was used as derivatizing reagent in this
Glyphosate and glufosinate herbicides are used at pre- and
method. In terms of reaction time, date ranged from 5 min [21] to
post-emergence weed control, and their residues are theoretically
overnight [24]. The effect of different derivatization reaction times
mainly expected in cereals, which were divided into the commod-
and temperatures was studied. With the increase of reaction time,
ity groups with high starch and/or protein content and low water
the recovery rate increased significantly. The maximum reaction
and fat content shown in Annex A of SANTE guideline [27]. Con-
yield was observed after 4 h (shown in Fig. 2). In practice, 12 h was
sequently, maize, rice, wheat and barley were selected as matrix
chosen to be reaction time. An influence of the reaction tempera-
blank to validate the developed method. Considering the diversity
ture was observed as Fig. 2b. Higher temperature reaction tended
of food products to be tested, the verification of method in differ-
to decrease the recovery of product. It is the same as reported by
ent matrices spiking samples was also done, as shown in method
Druart [11]. Therefore, the derivatization was completed at room
application. The linearity, the limit of detection (LOD), the limit of
temperature.
quantification (LOQ), and the accuracy and precision of the method
The effect of FMOC concentration was not investigated in our
were discussed respectively as below.
study but was fixed at 20 mg/mL. The use of FMOC in excess was
to compensate for high reactivity with water leading to the forma-
tion of FMOC-OH products and matrix components-primary and 3.2.2. Linearity, LOQs and LODs
secondary amines. The linear range was determined by two replicate analyses
of calibration standards (matrix-matched calibration), which was
3.1.3. SPE purification conducted at four levels ranging between 5, 10, 20 and 100 ␮g/kg
To remove the interferences, excessive FMOC and concentrate with internal standard mixture for five days. The calibration curves
the analytes to increase sensitivity, a SPE-C18 cartridge was chosen were obtained by the linear analysis (Y = aX + b with 1/X). In five
for purification and concentration. Firstly, methanol and acidified matrices (2 maize samples, 1 rice, 1 wheat and 1 barley) the coef-
water were used to condition the cartridge. Once the derivatization ficient of determination (R2 ) for the standard curves of glyphosate
reaction took place overnight, hydrochloric acid was added to stop and glufosinate were greater than 0.99 as attachment material A1.
the reaction, by low-ering the pH. Then the derivatized acidified In theory, the LODs and LOQs are obtained by the ratio of the
extract was added into the C18 cartridge. Acidified water and DCM signal-to-noise (S/N) of 3 and 10 from the lowest concentration
were chosen for rinsing to remove the excess FMOC and some polar levels of the spiked samples. However, the results of reproducibility
interference. In order to carefully check for possible analyte losses and repeatability were not satisfying in practical application using
during dichloromethane washing, the solution was also analyzed the value of 10 S/N as LOQ. In this study, although the target LOQ
from washing step with water and dichloromethane and the result for glyphosate and glufosinate was 10 ␮g/kg, the LOQs for the two
revealed no indication of analyte losses. The eluent at pH 9 was pesticides were chosen at 5 ␮g/kg for French Infant Total Diet Study.
selected to elute the analytes from the cartridge. The eluate was The experiment was carried out for duplicate in the five matrices
then evaporated to dryness and the residue was dissolved with a during five days. We used three parameters (accuracy, fidelity and
 Y. Liao et al. / J. Chromatogr. A 1549 (2018) 31–38 35

Fig. 2. Influence of derivatization time and temperature on the recoveries for glyphosate and glufosinate.

Table 3
Accuracy and precision of the proposed method in studied matrices.

Matrix, day [Ca ] Recoveries (%) Avg.R(%) RSDR (%)

Maize, day1 Maize, day2 Rice, day3 Wheat, day4 Barley, day5

glyphosate 5 100 94 106 100 114 105 109 100 108 103 104 5.60
10 97 102 109 107 99 98 112 109 101 107 104 6.38
20 99 103 100 95 101 108 100 93 106 99 100 4.53

Glufosinate 5 100 98 107 108 109 106 106 103 96 92 103 5.88
10 107 98 95 94 99 103 108 108 102 105 102 5.12
20 91 96 101 103 98 99 98 94 100 100 98 3.82
a
Which is spiked level (␮g/kg).

homogeneity) to test LOQ. The accuracy, fidelity and homogeneity LOQ) and NF V03-110 (repeatability: n ≥ 2, reproducibility: n ≥ 3,
were calculated as following: level: n ≥ 3, level of concentration: no request).
First, the accuracy was evaluated by absolute mean bias, rel-
CS −CM ative mean bias (%) and average recovery (%). The precision was
(1) Accuracy criterion (≤10%):| √ |, where Cs was the chosen
SD× 10
expressed by standard deviations (SD) of repeatability, inter-series
value; CM , mean value and SD, standard deviation
& intermediate fidelity and relative standard deviation (RSDR %).
(2) Fidelity criterion (≤ 20%, relative standard deviation, RSD):
The recoveries for glyphosate ranged from 93% to 114% in inter-
SD/CM x 100%
day studies and the average recoveries (%) for three levels (5, 10
(3) Homogeneity: the homogeneity of data was tested by SHAPIRO-
and 20 ␮g/kg) were 104, 104 and 100 respectively. Accordingly,
WILK statistic.
the recoveries for glufosinate ranged from 91% to 109% in inter-
day studies and the average recoveries (%) for three levels (5, 10
The values obtained of accuracy were 2.2% and 1.3% for and 20 ␮g/kg) were 103, 102 and 98 respectively in Table 3. The
glyphosate and glufosinate respectively. Two values obtained of RSDR % values for glyphosate and glufosinate ranged from 3.8 to
fidelity for glyphosate and glufosinate were both 5.7% and they 6.1 in inter-day, which reached the evaluation requirements for
reached the requirements of the precision. According to the test of
SHAPIRO-WILK, the calculated results demonstrated the data fol-

the precision. The uncertainty (I) included  (RMB)
relative mean bias

for accuracy and RSD for precision I = (RMB)2 + RSD2, , and it
lowed a normal distribution. Consequently, it was satisfying with
the homogeneity of data. Thus, the chosen LOQs of glyphosate and ranged from 9% to 16% for glyphosate and 9% to 13% for glufosinate.
glufosinate at 5 ␮g/kg were suitable in the selected matrices. The Second, a tolerance interval of expectation is constructed which
LODs of glyphosate and glufosinate were 1.7 ␮g/kg correspondingly contains a probability ␤ % (80% chosen). To use an accuracy profile to
(LOD = 3/10 LOQ). In addition, it is observed that the values of the validate a method, the two decision criteria (limits of acceptability
ratio S/N for glyphosate and glufosinate were higher than 1900 and and probability ␤) was established as follows. The result was shown
20000 respectively at 5 ␮g/kg (far more than 10 S/N), which shows in attachment material A3 and attachment material A4.
that in practice a lower LOQ can be achieved for glufosinate. The
results were shown in attachment material A2.
(1) Limits of acceptability. They serve to translate the practical
objectives of the users and delimit an interval around the refer-
3.2.3. Accuracy and precision ence value. In most cases, these limits are regulated or derived
In the validation plan, we used an accuracy profile to validate a from regulation. In this study, the recovery (between 70 and
method. The details as follows: 120%) for accuracy and RSDR limits (≤20%) were used according
The number (n) of series, repetition and levels for the validation to the guidance document SANTE.
plan were chosen (repeatability: n = 2, reproducibility: n = 5, level: (2) Probability ␤. It represents the proportion of future average
n = 3, level of concentration: LOQ, 2 LOQ and 4 LOQ) according to results within the tolerance intervals. The value chosen for ␤
the guidance document SANTE (repeatability: n ≥ 1, reproducibil- depends largely on the application field. It is obvious that the
ity: n ≥ 5, level: n ≥ 2, level of concentration: LOQ and 2 LOQ or 10 smaller ␤, eg. 70%, the more likely the method is to produce
36 Y. Liao et al. / J. Chromatogr. A 1549 (2018) 31–38

Fig. 3. Accuracy profile of glyphosate and glufosinate for the proposed method.

Table 4
The recoveries of glyphosate and glufosinate in expanded matrices.

Matrices [Ca ] Recoveries (%) Avg.R(%) RSDR (%)

Banana Potato Milk Cheese Yogurt

glyphosate 5 94 106 102 94 96 96 108 90 96 82 97 7.9

10 103 96 98 103 97 88 100 105 102 89 98 6.1
Glufosinate 5 104 100 90 112 96 108 104 97 106 102 104 7.3
10 97 99 99 95 83 88 116 92 96 101 95 6.0
a
Which is spiked level (␮g/kg).

results that do not correspond to the announced specifications. shown in Fig. 4. The 4 peaks are respectively shown in the chro-
In NF V03-110 this probability ␤ has been set at 80%, at least. matogram for each matrix: the quantitative peak of glyphosate (top
left) and one of its internal standard (bottom left), and one of glufos-
Based on the accuracy, precision, uncertainty and a tolerance inate (top right) and one of its internal standard (bottom right). The
interval of expectation, the accuracy profile of the method was last chromatogram (bottom right) was the blank matrix. From the
calculated and the results were shown in Fig. 3. All different figure, it displayed little interference in the matrices. Therefore, it
combinations complied with the acceptable method performance indicated that the two pesticides in these different matrices can be
parameters suggested by SANTE guide line. The experiment results determined, and that the developed method was sensitive enough
indicated that the developed method was in the satisfying range to quantify glyphosate and glufosinate in foods.
expected for analysis.
3.3.3. Analytical results
3.3. Application To ensure the quality of the analytical results, a reagent blank, a
matrix blank and two spiked samples at 5 ␮g/kg (QC) for two com-
3.3.1. Application fields pounds and at 2 ␮g/kg for glufosinate (as requested by ANSES) were
This developed method has been certified by the French certifi- analyzed in each kind of matrix and each series experiment, for the
cation organization (COFRAC), and has been applied to the analysis 136 food samples analyzed by this method. Finally, among these
of 136 food samples provided by ANSES in January 2013, and cur- samples, only 3 samples contained quantified levels of glyphosate
rently for routine national monitoring. The milk-based baby foods or glufosinate residues. Glyphosate residues were quantified at
were mainly provided “as consumed” (liquid) by ANSES and ana- 6.0 and 34 ␮g/kg in two breakfast cereal samples respectively.
lyzed as such by the laboratory. The food samples were divided into Glufosinate residues were quantified at 9.8 ␮g/kg in a sample of
three groups as shown in Table 1. boiled potatoes. No residues were detected in the other samples
including milk-based baby foods, with LOD ranging from 1 ␮g/kg
3.3.2. Verification in expanded matrices for glufosinate to 2 ␮g/kg for glyphosate as requested by ANSES.
Considering the variety of food products to be tested (Table 1), For these samples, in practice, better analytical limits were ver-
the verification of the method in expanded new matrices spiking ified for each analytical batch and the results were satisfied. In
samples was also conducted according to SANTE guideline. The order to better monitor the dietary exposure of infants and young
plant products with low water content (<60%) have been validated. children to these two compounds and their related residues (e.g.
So according to three kinds of actual samples, the different matrices trimethyl-sulfonium, AMPA, acetyl-glyphosate and acetyl-AMPA),
were selected for verification including fruit (banana) and veg- further analytical developments and most frequent analysis should
etable (potato), milk, yogurt and cheese. The average recoveries (%) be undertaken in the framework of monitoring programs, espe-
and RSDR (%) were calculated by the spiked samples including five cially for cereals and potato products.
kinds of matrix in two replicates at two levels (5 and 10 ␮g/kg). The
recoveries of glyphosate and glufosinate ranged from 82% to 116% 4. Conclusion
(between 70 and 120%) and the RSDR , from 6.0 to 7.9% (≤20%).
The result was displayed in Table 4. So, the method was verified A reliable and sensitive method was developed to detect and
in these application fields with acceptable accuracy and precision. quantify glyphosate and glufosinate residues in food by LC–MS/MS.
Meanwhile, the chromatograms at the level (5 ␮g/kg) in 5 differ- High accuracy and precision could be achieved by off-line deriva-
ent spiked matrices (banana, potato, cheese, milk and yogurt) and a tization and purification. Good validation parameters for linearity,
blank matrix (free of analytes like glyphosate and glufosinate) were recovery, LOD and LOQ were acquired and the verification results
 Y. Liao et al. / J. Chromatogr. A 1549 (2018) 31–38 37

Fig. 4. Chromatograms (LC–MS/MS) in expanded matrices at the level (5 ␮g/kg).

demonstrated extensive applicability of the method for the analysis References

of glyphosate and glufosinate in various food products. The method
can be used not only for high water content samples, but also for [1] S.O. Duke, S.B. Powles, Mini-review glyphosate: a once-in-a-century
herbicide, Pest Manag. Sci. 64 (2008) 319–325.
samples with low one, high protein and products of animal origin. [2] K. Zoe, M.S. Jill, The rise of glyphosate and new opportunities for biosentinel
In conclusion, the developed method can be a dependable and sen- early-warning studies, Conserv. Biol. 31 (6) (2017) 1293–1300, [Link]
sitive one for the routine monitoring of glyphosate and glufosinate org/10.1111/cobi.12955.
[3] EFSA (European Food Safety Authority), The 2015 European Union report on
in food. Analytical perspectives should include the development of pesticide residues in food, EFSA J. 15 (4) (2017) 4791.
analytical methods to better monitor glyphosate related metabo- [4] T. Bøhn, M. Cuhra, T. Traavik, M. Sanden, J. Fagan, R. Primicerio, Compositional
lites in food, e.g. trimethyl-sulfonium, AMPA, acetyl-glyphosate and differences in soybeans on the market: glyphosate accumulates in Roundup
Ready GM soybeans, Food Chem. 153 (2014) 207–215.
acetyl-AMPA. [5] R. Mesnage, N. Defarge, J. Spiroux de Vendômois, G.E. Séralini, Potential toxic
effects of glyphosate and its commercial formulations below regulatory
limits, Food Chem. Toxicol. 84 (2015) 133–153.
Conflict of interest [6] M. Krüger, P. Schledorn, W. Schrödl, H.-W. Hoppe, W. Lutz, A.A. Shehata,
Detection of glyphosate residues in animals and humans, J. Environ. Anal.
Toxicol. 4 (2014) 2–5.
The authors declare no conflict of interest. [7] EFSA (European Food Safety Authority), Conclusion on the peer review of the
pesticide risk assessment of the active substance glyphosate, EFSA J. 13 (11)
(2015) 4302.
Acknowledgments [8] P. Thomas, J.B. Ignaz, B. Astrid, D.M. Markus, E.B. Marianne, Occurrence of the
herbicide glyphosate and its metabolite AMPA in surface waters in
Switzerland determined with on-line solid phase extraction LC-MS/MS,
This work was supported by ANSES. Environ. Sci. Pollut. Res. 24 (2) (2017) 1588–1596.
[9] M. Motojyuku, T. Saito, K. Akieda, H. Otsuka, I. Yamamoto, S. Inokuchi,
Determination of glyphosate glyphosate metabolites, and glufosinate in
Appendix A. Supplementary data human serum by gas chromatography −mass spectrometry, J. Chromatogr. B.
875 (2008) 509–514.
[10] A. Royer, S. Beguin, J.C. Tabet, S. Hulot, M.A. Reding, P.Y. Communal,
Supplementary data associated with this article can be found, Determination of glyphosate and aminomethylphosphonic acid residues in
in the online version, at [Link] water by gas chromatography with tandem mass spectrometry after
036.
38 Y. Liao et al. / J. Chromatogr. A 1549 (2018) 31–38

exchange ion resin purification and derivatization, Application on vegetable [20] M.E. Báez, E. Fuentes, M.J.J. Espina. Espinoza, Determination of glyphosate and
matrixes, Anal. Chem. 72 (2000) 3826–3832. aminomethyl-phosphonic acid in aqueous soil matrices: a critical analysis of
[11] C. Druart, O. Delhomme, A. Vaufleury, E. Ntcho, M. Millet, Optimization of the 9-fluorenylmethyl chloroformate derivatization reaction and application
extraction procedure and chromatographic separation of glyphosate, to adsorption studies, J. Sep. Sci. 37 (2014) 3125–3132.
glufosinate and aminomethylphosphonic acid in soil, Anal. Bioanal. Chem. [21] T. Arkan, I. Molnár-Perl, The role of derivatization techniques in the analysis
399 (2011) 1725–1732. of glyphosate and aminomethyl-phosphonic acid by chromatography,
[12] I. Hanke, H. Singer, J. Hollender, Ultratrace-level determination of glyphosate, Microchem. J. 121 (2015) 99–106.
aminomethyl- phosphonic acid and glufosinate in natural waters by [22] W.C. Koskinen, L.J. Marek, K.E. Hall, Analysis of glyphosate and
solid-phase extraction followed by liquid chromatography–tandem mass aminomethylphosphonic acid in water, plant materials and soil, Pest Manag.
spectrometry: performance tuning of derivatization, enrichment and Sci. 72 (2016) 423–432.
detection, Anal. Bioanal. Chem. 391 (2008) 2265–2276. [23] Y.T. Han, L. Song, P.Y. Zhao, Y.J. Li, N. Zou, Y.H. Qin, X.S. Li, C.P. Pan, Residue
[13] A.M. Botero-Coy, M. Ibáñez, J.V. Sancho, F. Hernández, Improvements in the determination of glufosinate in plant origin foods using modified Quick Polar
analytical methodology for the residue determination of the herbicide Pesticides (QuPPe) method and liquid chromatography coupled with tandem
glyphosate in soils by liquid chromatography coupled to mass spectrometry, mass spectrometry, Food Chem. 197 (2016) 730–736.
J. Chromatogr. A 1292 (2013) 132–141. [24] P. Kaczyński, B. Łozowicka, Liquid chromatographic determination of
[14] L.J. Marek, W.C. Koskinen, Simplified analysis of glyphosate and glyphosate and aminomethylphosphonic acid residues in rapeseed with
aminomethylphosphonic acid in water, vegetation and soil by liquid MS/MS detection or derivatization/fluorescence detection, Open Chem. 13
chromatography-tandem mass spectrometry, Pest Manag. Sci. 70 (2014) (2015) 1011–1019.
1158–1164. [25] M. Hulin, N. Bemrah, A. Nougadère, J.L. Volatier, V. Sirot, J.C. Leblanc,
[15] H.Y. Chiu, Z.Y. Lin, H.L. Tu, C.W. Whang, Analysis of glyphosate and Assessment of infant exposure to food chemicals: the French total diet study
aminomethylphosphonic acid by capillary electrophoresis with design, Food Addit. Contam. Part A Chem. Anal. Control Expo. Risk Assess 31
electrochemiluminescence detection, J. Chromatogr. A. 1177 (2008) 195–198. (7) (2014) 1226–1239.
[16] Z.X. Guo, Q.T. Cai, Z.G. Yang, Ion chromatography/inductively coupled plasma [26] ANSES, Etude De l’alimentation Totale Infantile Tome 2 − Partie 4, Résultats
mass spectrometry for simultaneous determination of glyphosate glufosinate, Relatifs Aux résidus De Pesticides, Rapport d’expertise Collective, ANSES
fosamine and ethephon at nanogram levels in water, Rapid Commu. Mass Edition scientifique, Paris, 2016, pp. 378.
Spectrom. 21 (2007) 1606–1612. [27] European Commission Guidance document on analytical quality control and
[17] C.F.B. Coutinho, L.F.M. Coutinho, L.H. Mazo, S.L. Nixdorf, C.A.P. Camara, Rapid method validation procedures for pesticides residues analysis in food and
and direct determination of glyphosate and aminomethylphosphonic acid in feed. SANTE/11945/2015. rev. 0, 2015.
water using anion-exchange chromatography with coulometric detection, J. [28] Commission Directive 2002/63/EC of 11 July 2002, Community methods of
Chromatogr. A. 1208 (2008) 246–249. sampling for the official control of pesticide residues in and on products of
[18] M. Anastassiades, D.I. Kolberg, E. Eichhorn, A. Benkenstein, S. Lukačević, D. plant and animal origin and repealing Directive 79/700/EEC, J. Eur. Commun.
Mack, C. Wildgrube, I. Sigalov, D. Dörk, A. Barth, Quick Method for the Analysis 16 (7) (2002).
of Numerous Highly Polar Pesticides in Foods of Plant Origin via LC–MS/MS [29] Norme Française AFNOR, Analyse des produits agricoles et alimentaires −
Involving Simultaneous Extraction with Methanol (QuPPe-Method), 2015 Protocole de caractérisation en vue de la validation d’une méthode d’analyse
[Link] [Link]. quantitative par construction du profil d’exactitude. NF V03-110, May 2010.
[19] A.M. Botero-Coy, M. Ibáñez, J.V. Sancho, F. Hernández, Direct liquid [30] H.A. Moye, A.J. Boning Jr., A versatile fluorogenic labelling reagent for primary
chromatography-tandem mass spectrometry determination of underivatized and secondary amines: 9-Fluorenylmethyl Chloroformate, Anal. Lett. 12 (B1)
glyphosate in rice, maize and soybean, J. Chromatogr. A. 1313 (2013) 157–165. (1979) 25–35.