Documente Academic
Documente Profesional
Documente Cultură
Food Chemistry
journal homepage: www.elsevier.com/locate/foodchem
Analytical Methods
a r t i c l e
i n f o
Article history:
Received 9 February 2011
Received in revised form 3 November 2011
Accepted 14 April 2012
Available online 21 April 2012
Keywords:
Infant milk formula
Pesticides
Microwave assisted extraction
LCMS/MS
a b s t r a c t
An extraction method consisting of both microwave-assisted extraction (MAE) and solid phase extraction
(SPE) was developed to extract pesticides of different polarities in infant milk formula. An experimental
central face composite design was established to evaluate the effects of several MAE variables simultaneously such as temperature and pH of extractant. With Derringer desirability function, all pesticides were
optimally extracted at 102 C within 20 min with 4.56 mL of extractant consisting of 0.1% of water in
methanol at pH 12. Extraction efciency was subsequently analysed with LCMS/MS. The extraction
recoveries ranging from 72% to 111% were obtained for all pesticides at fortication level of 5100 lg kg1
with relative standard deviations (RSDs) of less than 8%. Compared with Soxhlet/SPE, the MAE/SPE method
utilised less solvent and higher recoveries were achieved with a shorter extraction time.
2012 Elsevier Ltd. All rights reserved.
1. Introduction
Pesticides, which are also classied as persistent organic pollutants (POPs), are widely utilised for crop protection and pest control. The extensive application of pesticides is required to increase
the agricultural productivity such that the global food demand
can be met. Atrazine, for example, is one of the most frequently
used pesticides, for the pre- and post-emergence of grasses, weeds
and crops in United States. Other pesticides such as cyromazine are
utilised for y control in crop protection.
With the widespread usage of pesticides, residual amounts of
pesticides are being detected in both water and food cycles, posing
danger to human health. Epidemiologic studies were conducted to
evaluate the toxicity of pesticides. For example, US Environmental
Protection Agency (EPA) evaluated the chronic toxicity of atrazine
and its metabolites in Sprague Dawley rats to postulate the possible effects of persistent exposure to pesticides on humans (EPA,
1999). Atrazine was proposed to be classied as a likely human
carcinogen because of the elevated incidence of mammary tumours observed in female Sprague Dawley rats. In recent study,
the association between adverse birth outcomes and urinary biomarkers of prenatal atrazine exposure was assessed (Chevrier
et al., 2011). It was observed that the presence of quantiable levels of atrazine or a specic atrazine metabolite was associated with
fetal growth restriction and small head circumference for sex and
gestational age.
Corresponding author. Fax: +65 67791691.
E-mail address: chmlifys@nus.edu.sg (S.F.Y. Li).
0308-8146/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.foodchem.2012.04.076
Thus, the detection of pesticides at trace level is of vast importance to safeguard the wellbeing of the people. Analytical techniques have been developed to quantify the pesticides at trace
levels. One of the emerging areas of interest is the monitoring of
pesticides in infant products such as infant milk formula which is
the main source of food for infants. Pesticides can contaminate
the milk via three pathways: (i) the consumption of animal feed
or pastures, that have been sprayed with pesticides, by dairy cattle,
(ii) the spraying of pesticides on animals for ectoparasitic control
and (iii) the usage of pesticides in factory farms (Mezcua et al.,
2007). The concentration of pesticides present in milk is dependent
on its stability, the amount and time of exposure as well as the
metabolic rate of the animals. The triazine pesticides, for instance,
undergo degradation via three main reactions: dealkylation, oxidation and dechlorination in the environment. (Hapeman, Bilboulian,
Anderson, & Torrents, 1998) For example, cyromazine undergoes
dealklyation to form melamine which may result in renal diseases
with chronic consumption (Sun, Wang, Liu, Qiao, & Liang, 2009).
Since the presence of pesticides in food can also lead to acute
neurologic toxicity and chronic neuro-development impairment,
regulatory bodies such as US Food and Drug Administration (FDA,
1996) have set the maximum residue levels (MRLs) to monitor the
concentration of pesticides in a variety of food. Based on the Commission Directive 2006/125/EC (EU, 2006), the European committee
has regulated that not more than 10 lg kg1 of pesticides should be
present in infant food. In addition, Commission Regulation No.
1135/2009 (EU, 2009) has also stated that the concentration of
melamine should not exceed 2.5 mg kg1 in infant milk formula.
The MRL of pesticides are usually signicantly lower in infant food
2474
compared to that for adults as the infants body weight are lighter.
Moreover, infants consume the same type of food such as infant
milk formula repetitively. Hence, they are more prone to the adverse effects of pesticides in food than the adults.
In order to monitor pesticides at trace level in complex food
matrices, sample preparation involving both extraction and
clean-up procedure is essential. Liquidliquid extraction and Soxhlet extraction are established as the extraction method of choice to
analyse pesticides in food and environmental samples (Andreu &
Pico, 2004; Srogi, 2006). Despite their good extraction efciencies,
these methods have several drawbacks which include the usage of
large volume of organic solvents, time consuming procedures and
difcultly in automating such method. Recently, new extraction
techniques such as supercritical uid extraction (SFE), pressurised
liquid extraction (PLE) and microwave-assisted extraction (MAE)
which utilised elevated pressure and temperature have been reported for the extraction of pesticides. However, the experimental
set-ups of both PLE and SFE are highly costly. Moreover, organic
modiers like methanol, which are added to increase the polarity
of the carbon dioxide extractant, may also introduce additional
matrix materials that may be co-extracted in the SFE. This will reduce the selectivity of the extraction. On the other hand, for PLE,
poor extraction efciencies were observed for wet samples using
non-polar extraction solvent (Schantz, 2006).
Unlike traditional heating in which heat is transferred from the
container to the samples, MAE involves the use of microwave energy to heat the polar solvents and samples directly via ionic conduction and dipole rotation. Consequently, MAE has a greater
extraction efciency compared to traditional heating. MAE also offers several advantages which include low solvent consumption
and high sample throughput. Besides, MAE has demonstrated
excellent capability in extracting POPs such as lipophilic pesticides
and polycyclic aromatic hydrocarbons from complex matrices
(Wang, Meng, Lu, Liu, & Tao, 2007).
Since numerous publications have focused on the effects of several variables such as the type of solvent, temperature, time and pH
on MAE efciency (Stout, daCunha, & Allardice, 1996), experimental
design is applied in this study to examine the effects of these variables on MAE efciency. The traditional optimisation approach of
examining one variable at a time, involve the investigation of one
response by changing one variable while maintaining all the other
factors at a constant level. It is commonly utilised to determine
the best extraction condition, for a range of compounds despite
their difference in polarities. Its major disadvantage is that it does
not include the additive or synergistic interaction among the variables (Bianchi & Careri, 2008). As a consequence, this technique
does not portray the complete effects of the parameters on the response. In addition, more experiments are usually conducted based
on trial and error or through numerous ne tuning of the variables
to achieve good extraction recoveries for most of the compounds of
interest. This results in an increase in the analysis time and use of
reagents. With the application of experiment design, a more systematic approach can be adopted. The minimum and maximum value of variables is xed from the start of the experiment, taking into
account of the limitation of the instruments or extraction methods.
The number of experiments required is dependent on the designs
chosen such as Box-Behnken, full factorial design and Doehlent.
These symmetrical designs differ from one another with respect
to their selection of experimental points, number of levels for variables, and number of runs and blocks. Unlike the traditional approach, these experiments involve the optimisation of several
extraction variables simultaneously. In addition, the optimal MAE
condition for the pesticides of different polarities can be achieved
with the aid of the Derringer desirability function.
After extracting the pesticides with optimal MAE condition,
clean-up procedures will be carried out to remove the co-extracted
2475
"
#1=n
n
Y
pi
D
di
i1
where di is the partial desirability function of each response obtained from the transformation of the individual response of each
experiment, and Pi is the weight of the individual function, which
Pn
is normalised such that
i pi 1. For this study, the minimum
and the target value for the extraction recoveries were set at 80
110% respectively while the weight was xed as 1.
2.4. Extraction procedures
2.4.1. Microwave extraction
A Biotage Initiator microwave 2.5 (Uppsala, Sweden) was utilised for the extraction of pesticides from infant milk formula. During the optimisation process, xed concentration of pesticides
Table 1
Range of MAE parameters selected for optimisation for different solvent system.
Variables
Temperature (C)
Time (min)
Volume of extraction solvent (mL)
pH
Percentage of water in organic solvent (%)
Methanol/water mixture
Acetonitrile/water mixture
Selected level
Selected level
60, 90,120
10,20,30
2, 3.5, 5
2, 7, 12
0, 15, 30
102
20
4.56
12
0.1
60, 90,120
10,20,30
2, 3.5, 5
2, 7, 12
20, 25, 30
61
12
4.92
12
20.1
2476
the linear relationship between concentration and area. The standards of higher concentration which deviated from linearity were
removed, determining the working range of the calibration curve.
The limit of detection (LOD) and limit of quantication (LOQ) were
calculated based on signal-to-noise ratio of 3 and 10, respectively.
Method accuracy and precision were evaluated by recovery studies
using fortied samples at three concentration levels (5, 25, and
100 mg kg1).
3. Results and discussion
3.1. LCMS/MS optimisation
In order to achieve good sensitivity, 1 mg kg1 of individual pesticide standard was prepared in water for MS optimisation through
ow injection analysis. The [M + H]+ precursor ions and daughter
ions were determined with the use of full scan and MS/MS spectrum. Subsequently, the fragmentator (20200 V), collision energy
(540 V) and dwell time (50200 s) were optimised in order to obtain high signal to noise and sufcient data points during multiple
reactions monitoring (MRM). The quantiers are assigned to the
most abundant daughter ions employed for quantication whereas
the qualiers are usually the next most abundant daughter ions for
conrmatory purposes.
For the quantication of pesticides in infant milk formula, the
chromatographic separation was rst optimised. It has been reported that highly polar molecules, specically melamine and its
analogues, had insufcient retention on traditional octadecylsilane
(C18) column. However, longer retention time of approximately
13 min was observed for melamine when two C18 columns were
used in series (Ehling, Tefera, & Ho, 2007). Alternatively, ion pairing
agent such as sodium n-heptanesulfate was added to the mobile
phase to improve the retention of melamine on C18 column (Sun,
Wang, Ai, Liang, & Wu, 2010). Nevertheless, the use of non-volatile
ion pairing agent as aforementioned is not recommended for LC
MS/MS system as the presence of these ion pairing agents might
contaminate the electrospray source leading to ion suppression
(Hernandez, Sancho, & Pozo, 2002).
To solve this problem, FDA proposed the use of hydrophilic
interaction liquid chromatography (HILIC) (FDA, 2008). This technique utilises hydrophilic stationary phase such as silica and relatively hydrophobic mobile phases consisting of a mixture of 540%
water or a volatile buffer and acetonitrile. The hydrophilic stationary phase was able to retain the highly polar compounds, such as
melamine and cyromazine with a retention time of approximately
5.8 and 6.3 min respectively (Fig. 1). Conversely, the more hydrophobic pesticides eluted earlier in the order of prometryn, ametryn, prometon, simetryn, atrazine and atraton.
3.2. MAE optimisation
3.2.1. Selection and optimisation of experimental design
A central face composite design which is a type of central composite design, in which the axial points are placed on the face centres
of the cube, was chosen. Each variable has only three levels instead
of ve when compared to the central composite design. Thus, the total number of experiments required (N) = 2(k 1)+2k + C0 where k
represented the number of factors and C0 referred to the total number of centre points (Ferreira et al., 2007). During optimisation, the
ve variables including extraction temperature, time, volume of
extraction solvent, pH, composition of water in solvent and the
nature of solvent for the MAE were investigated. With this design,
3 centre runs were conducted to produce a reasonable stability of
prediction variance. Hence, a total of 29 randomised experiments
were executed.
2477
Fig. 1. The chromatograms from the analysis of the pesticides in milk matrices.
! b0
5
X
i1
bi xi
4 X
5
X
bij xi xj
i1 j1
2008). With the removal of some of the quadratic model coefcients as aforementioned, the R2 and the Q2 attained were greater
than 0.5, indicating good predictability. In addition, excellent
reproducibility was achieved based on the triplicate of the centre
points.
The normalised b coefcient values for Eq. (2) were displayed in
Table 2. The volume of extraction solvent and the pH of the extractant were observed to have more inuential effect on the extraction
recoveries for most compounds in this study at 95% condence level. The quadratic coefcient of volume of solvent pH has positive
effect to the extraction recoveries of most of the pesticides at 5%
signicance level. This indicated that the extraction recoveries increased further for most pesticides when the pH was basic and
with large volume of extraction solvent as compared to basic pH
with small volume of extractant.
The positive b coefcient values of the volume of extraction solvent indicated that increasing the volume of extraction solvent
would improve the extraction recoveries for most pesticides. In
addition, extraction temperature and time had a positive inuence
on the recoveries of most analytes except melamine and cyromazine. At high temperature and longer extraction time, the
2478
Fig. 2. Comparison of the extraction recoveries attained from MAE from different solvent mixture.
Table 2
Normalised b coefcient plots for the pesticides which were studied with the experimental design during MAE optimisation.
Prometryn
Constant
Temperature (temp)
Time
Volume of solvent (vol)
pH
% water
Temp Temp
Time Time
vol vol
pH pH
% water % water
Temp vol
Temp pH
Temp % water
Time % water
vol pH
vol % water
R2
Q2
Reproducibility
*
3.773
0.161
0.088
0.431*
0.457*
0.068
0.124
0.289
0.595*
0.102
0.310
0.072
0.144
0.113
0.143
0.228*
0.122
0.942
0.681
0.96
Ametryn
*
4.156
0.185*
0.082
0.345*
0.343*
0.092
0.150
0.247
0.600*
0.334*
0.323*
0.108
0.113
0.127*
0.164*
0.229*
0.079
0.96
0.73
1
Prometon
*
4.017
0.216*
0.064
0.326*
0.330*
0.142
0.104
0.271
0.761*
0.301
0.519*
0.138
0.120
0.115
0.188*
0.247*
0.047
0.926
0.602
0.99
Atrazine
*
3.441
0.084
0.124
0.245*
0.201
0.066
0.247
0.014
0.091
0.601
0.311
0.038
0.042
0.157
0.082
0.129
0.127
0.886
0.698
0.953
Simetryn
*
4.303
0.165*
0.069
0.521*
0.449*
0.114
0.114
0.275*
0.592*
0.113
0.385*
0.083
0.147
0.104
0.159*
0.247*
0.078
0.963
0.721
0.983
Atraton
*
4.355
0.236*
0.063
0.396*
0.318*
0.167
0.109
0.290
0.791*
0.260
0.570*
0.150*
0.133
0.073
0.191*
0.231*
0.038
0.939
0.599
0.987
Cyromazine
*
1.171
0.054
0.066
0.011*
0.356
0.146
0.248
0.267
0.895*
0.512
0.851*
0.035
0.088
0.096
0.107
0.223
0.028
0.661
0.506
1
Melamine
0.652*
0.023
0.092
0.456*
0.446*
0.188
0.165
0.250
0.566*
0.603*
0.640*
0.026
0.110
0.055
0.082
0.233*
0.082
0.866
0.564
1
hydrolysis of melamine and cyromazine might occur, hence, leading to lower extraction recoveries (Ono et al., 1998).
The pH of the extractant is also one of the most inuential variables for MAE. With the b coefcient values tted to the Eq. (2), response contours plots could be generated to clearly visualise the
inuence of each variable on the extraction recoveries (Supplementary Fig. 1). Since the pKa of pesticides ranged from 2.35 to
5.27, the amine group on the pesticides would remain in its native
state at high pH, thereby, resulting in higher preference for solvent
mixture consisting of less percentage of water. Therefore, basic pH
condition was preferred for the extraction of most pesticides except atrazine. Since atrazine hydrolysed easily in acidic and basic
aquatic environment (Kamrin, 1997), higher extraction recovery
of atrazine could only be achieved at neutral pH with methanol
consisting of 222% of water.
The negative b coefcient values implied that the addition of
water to methanol resulted in poor recoveries for most of the pesticides. One possible explanation is due to the dissipation factor of
the solvent which comprised of dielectric loss and dielectric
constant. Methanol has low dielectric constant but higher dielectric loss compared to water. Thus, it absorbs microwave energy
poorly but dissipates microwave energy to thermal energy preferentially. Consequently, rapid heating can occur instantaneously,
resulting in higher extraction efciency for methanol as compared
to water (Eskilsson & Bjorklund, 2000).
3.2.4. Solid phase extraction
With the optimal MAE condition determined for methanol/
water solvent mixture, the MAE supernatant was subjected to
SPE for pre-concentration of pesticides and removal of interferences. The supernatant obtained from MAE was rst acidied with
hydrochloric acid to protonate the amine moieties and to enhance
its afnity for the anions on the sorbent. Oasis MCX sorbent which
consisted of sulphonate functional groups was selected instead of
octadecylsilane as it had more selectivity towards the protonated
amine moieties on the pesticides. The cartridge was pre-conditioned with water and methanol. After loading the sample, it was
washed with hydrochloric acid to further promote the binding
2479
between the pesticides and the negatively charged sorbent. Methanol was subsequently added to remove any hydrophobic interference. Basic eluent was added to deprotonate the ionised amine
group and reduce the pesticides afnity for the sorbent such that
high recoveries were achieved. Hence, 10 mL of 4% ammonium
hydroxide in methanol was added to ensure complete elution of
the pesticides from the sorbent.
Table 3
The extraction recoveries of three different concentrations of spiked pesticides in infant milk formula.
MAE/SPE
Soxhlet/SPE
5 lg kg1
Prometryn
Ametryn
Prometon
Atrazine
Simetryn
Atraton
Cyromazine
Melamine
25 lg kg1
100 lg kg1
25 lg kg1
Average n = 3
% RSD
Average n = 3
% RSD
Average n = 3
% RSD
Average n = 3
% RSD
88
90
88
73
91
91
78
72
5
5
2
5
8
3
3
4
98
95
87
90
92
95
76
87
4
2
3
7
0
5
7
6
101
101
111
85
101
106
94
91
4
4
4
3
4
4
6
7
75
80
66
57
76
82
62
35
2
2
4
2
2
1
7
13
Table 4
Calibration data, which included precisions and limit of detection and quantication.
Prometryn
Ametryn
Prometon
Atrazine
Simetryn
Atraton
Cyromazine
Melamine
Number of
standards
Mandel
test
Fcritical
(P = 99%)
r2
Linear
range (ppb)
Slope
Intercept
Intraday precision,
RSD% n = 3
Interday precision,
RSD% n = 6
LOD
(ppb)
LOQ
(ppb)
7
9
8
10
7
9
7
6
1.24
0.19
0.50
0.07
1.16
3.89
8.33
9.69
15.52
8.26
10.67
6.84
15.52
8.26
15.52
28.71
0.998
0.999
0.999
0.999
0.998
0.999
0.999
0.999
0.5250
0.1250
2.5600
0.5600
1.0250
0.5600
2.5500
7.5600
10,686
2424.3
12,477
1268.3
4068.4
17,726
548.3
120
3324.1
140.12
25,863
2486.9
2169.4
15,256
2387.9
527.58
0.26
0.37
0.93
0.44
0.28
0.30
0.17
0.82
6.73
2.20
12.52
0.66
1.23
11.95
5.08
0.56
0.14
0.12
0.69
1.23
0.28
0.55
2.53
1.61
0.46
0.41
2.30
4.11
0.93
1.85
8.42
5.36
2480