ˇ ajka, Masˇtovsk, Lehotay, Hajsˇlov
C
1048
ˇ ajka1
Tomsˇ C
Katerˇina Masˇtovsk2
Steven J. Lehotay2
Jana Hajsˇlov1
Use of automated direct sample introduction with
analyte protectants in the GC–MS analysis of
pesticide residues
1
Institute of Chemical Technology,
Prague, Department of Food
Chemistry and Analysis,
Technick 3, 166 28 Prague 6,
Czech Republic
2
USDA, Agriculture Research
Service, Eastern Regional
Research Center, 600 East
Mermaid Lane, Wyndmoor, PA
19038, USA
Automated large-volume direct sample introduction, or difficult matrix introduction
(DMI), was investigated in the determination of 44 pesticide residues possessing a
wide range of physico-chemical properties (volatility, polarity, pKa) in fruit-based baby
food by means of gas chromatography – mass spectrometry (GC – MS) with a quadrupole mass analyzer. DMI has advantages over traditional injection because large
volumes (up to 30 lL) of potentially dirty sample extracts can be injected into the
GC – MS, but nonvolatile matrix components that would normally contaminate the
inlet are removed after every injection. The extra matrix and glass surfaces involved
in DMI, however, make the system more prone to the matrix-induced chromatographic enhancement effect, which adversely affects quantification of several pesticides.
To overcome this problem, matrix-matched calibration standards and/or the use of
analyte protectants were applied in the DMI approach, and the analysis of extracts
was also compared before and after undergoing clean-up by dispersive solid-phase
extraction. For best quantification, clean-up was still needed, and the combination of
matrix-matching with analyte protectants gave the most reproducible results. Depending on the application, however, the addition of analyte protectants (a mixture of 3ethoxy-1,2-propanediol, L-gulonic acid 3-lactone, and D-sorbitol) to sample extracts
and calibration standards in solvent (non-matrix matched), gave satisfactory quantification for most of the 44 pesticides tested. The lowest calibration levels for 34 of the
44 pesticides were d10 ng/g, which meets the standard required by the European
Union Baby Food Directive (2003/13/EC).
Key Words: Pesticide residue analysis; Direct sample introduction; Difficult matrix introduction;
Large volume injection; Analyte protectants; Gas chromatography – mass spectrometry; Food
analysis; Baby food;
Received: January 28, 2005; revised: March 10, 2005; accepted: March 13, 2005
DOI 10.1002/jssc.200500050
1 Introduction
Analysis of multiple pesticide residues in food is often a
time-consuming, labor-intensive, and expensive process
due to the complexity of the many analytes and matrices
involved. The use of organic solvents, such as acetone,
acetonitrile (MeCN), or ethyl acetate (EtAc), for extraction
provides high pesticide recoveries over a wide polarity
range, but further cleanup is nearly always required
before gas chromatographic (GC) analysis. Common
cleanup techniques, such as solid-phase extraction
(SPE), liquid-liquid partitioning, and/or gel permeation
chromatography (GPC), increase the overall cost of the
method, extend the analysis time and require additional
labor [1]. However, cleanup of the extracts is integral for
Correspondence: Steven J. Lehotay, USDA, Agriculture Research Service, Eastern Regional Research Center, 600 East
Mermaid Lane, Wyndmoor, PA 19038, USA.
Phone: +1 215 233 6433. Fax: +1 215 233 6642.
E-mail: [email protected] (S.J. Lehotay).
J. Sep. Sci. 2005, 28, 1048 – 1060
www.jss-journal.de
improving the ruggedness and reliability of the GC system [2]. Non-volatile matrix components build up in the
injector liner and capillary column, reducing performance
of the system until maintenance is performed [3].
Recently, Anastassiades et al. introduced the so-called
“quick, easy, cheap, effective, rugged, and safe” (QuEChERS) method for pesticide residue analysis. This
method uses MeCN for extraction of the sample and
simultaneous liquid-liquid partitioning resulting on adding
anhydrous MgSO4 and NaCl. After centrifugation, a portion of the extract (typically 1 mL) is transferred to a tube
containing primary secondary amine (PSA) sorbent and
anhydrous MgSO4. After brief mixing and centrifugation
steps, the extract is ready for GC analysis [4]. Unfortunately, degradation of base-sensitive pesticides (e. g. captan, folpet, dichlofluanid, chlorothalonil) occurred using
this approach. Lehotay et al. modified the method slightly
to improve results of those problematic pesticides through
the use of buffering [5]. The buffered QuEChERS method
i
2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Use of automated direct sample introduction with analyte protectants
The use of mass spectrometry (MS) in GC analysis permits the quantification and identification of a wide range of
GC-amenable pesticides in complex extracts [6]. Modern
GC – MS instruments can achieve detection limits that
approach those possible with traditional selective detectors [7], and the use of large-volume injection (LVI) can
further help to reduce detection limits if matrix interferences are not the limiting source of noise [2]. However,
cleanup of complex extracts becomes even more important in LVI to avoid contamination of the injection liner,
capillary column, and MS source with non-volatile matrix
components. Any non-volatile materials, such as salts,
carbohydrates, proteins, and lipids, will remain in or near
the injector and/or slowly migrate through the GC column.
These may not directly interfere in MS detection but they
can impair the performance of GC analysis and lead to frequent system maintenance.
An elegant way to overcome the problems caused by the
analysis of dirty samples is direct sample introduction
(DSI) [8 – 11], which has been commercially automated in
a technique called difficult matrix introduction (DMI) [12 –
16]. The DSI approach involves adding up to 30 lL of the
extract to a microvial that is placed in the GC liner. The
solvent is evaporated and vented at a relatively low temperature. After that, the injector is ballistically heated to
volatilize the GC-amenable compounds, which are then
focused at the front of the relatively cold analytical column.
The column then undergoes normal temperature programming to separate the analytes and cool to initial conditions, at which time the microvial is removed and discarded along with the non-volatile matrix components that
it contains. In this way, only those compounds with the
volatility range of the analytes enter the column, and the
GC should remain a pristine system after every analysis.
In the commercial DMI approach, the entire liner along
with the microvial is replaced after each injection. The
main advantages of DSI include: (i) reduced demands for
the GC system maintenance (contrary to other injection
techniques, contamination by non-volatile matrices does
not occur), and (ii) reduced need for sample cleanup.
However, without cleanup, the many (semi-)volatile matrix
components introduced from the sample into the injector
may still influence the quantitative aspects of the injection
process and/or interfere in the detection. Thus, MS or MS/
MS detection is preferred for the accurate quantification of
residues and reduction of chemical noise [10 – 14, 16].
Another important issue in the quantitative GC analysis of
pesticides in food is the “matrix-induced chromatographic
response enhancement effect” [17]. This effect is noted
J. Sep. Sci. 2005, 28, 1048 – 1060
www.jss-journal.de
by improved chromatographic peak intensity and shape of
affected compounds when they are injected in the presence of a complex matrix. The matrix fills active sites in
the liner and column, which reduces interaction of the analytes on these sites, and leads to enhanced analyte peaks.
Susceptible analytes (typically those with strong hydrogen
bonding potential) give poor peaks with low response in
the absence of matrix during the injection step. In theory,
extensive cleanup of the extracts could work to eliminate
the matrix components that cause this effect, but it is
untenable in practice due to the wide polarity range of the
analytes in multiclass, multiresidue methods and complexity of the matrices [18]. Moreover, the matrix-induced
enhancement effect gives larger and higher peaks, thus it
would be better to take advantage of this phenomenon
rather than eliminate it.
Ways to compensate for matrix effects in GC include:
(i) method of standard addition; (ii) use of isotopically
labeled internal standards; (iii) use of matrix-matched
standards; and (iv) use of analyte protectants. The latter
approach offers the most practical and convenient solution to the problems associated with calibration in routine
GC analysis of pesticide residues in diverse food samples. Essentially, analyte protectants are compounds that
strongly interact with active sites in the GC system, thus
decreasing degradation and/or adsorption of co-injected
analytes [4, 19, 20]. Therefore, the application of those
compounds can minimize losses of susceptible analytes,
thereby significantly improving their peak shapes and lowering detection limits. The analyte protectants are added
to sample extracts and matrix-free standards alike to
induce response enhancement in both instances, resulting in maximization and equalization of the matrix-induced
response enhancement effect. Various compounds have
been evaluated as analyte protectants [19], and a mixture
of 3-ethoxy-1,2-propanediol, L-gulonic acid c-lactone, and
D-sorbitol (in MeCN extracts) was found to most effectively cover a wide volatility range of GC-amenable pesticides [20].
A main goal of this study was to explore LV-DMI – GC –
MS applied to the analysis of 44 representative pesticides
in non-cleaned and cleaned buffered QuEChERS extracts
of foods. Additionally, quantification of the pesticides
using calibration with matrix-matched and solvent standards, both containing analyte protectants, was evaluated
and compared.
2 Experimental
2.1 Chemicals and materials
Pesticide reference standards, all 95% or higher purity,
were obtained from the National Pesticide Repository of
the US Environmental Protection Agency (Fort Meade,
MD, USA), Dr. Ehrenstorfer GmbH (Augsburg, Germany),
i
2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Original Paper
involves the extraction of the sample with MeCN containing 1% acetic acid (HAc) and simultaneous liquid-liquid
partitioning formed by adding sodium acetate (NaAc)
instead of NaCl along with the MgSO4.
1049
1050
ˇ ajka, Masˇtovsk, Lehotay, Hajsˇlov
C
Ultra Scientific (North Kingstown, RI, USA), and Chemservice (West Chester, PA, USA). Composite stock standard solution (20 lg/mL) of 44 pesticides (acephate, atrazine, azinphos-methyl, azoxystrobin, bifenthrin, bromophos, captan, carbaryl, chlorothalonil, chlorpyrifos, coumaphos, k-cyhalothrin, cyprodinil, p,p9-DDD, p,p9-DDE,
p,p9-DDT, deltamethrin, diazinon, dichlofluanid, dichlorvos, dimethoate, endosulfan sulfate, ethion, fenvalerate,
folpet, hexachlorobenzene (HCB), heptachlor, imazalil,
lindane, methamidophos, methiocarb, mevinphos,
omethoate, permethrin, o-phenylphenol, phosalone, pirimicarb, pirimiphos-methyl, procymidone, tebuconazole,
thiabendazole, tolylfluanid, trifluralin, and vinclozolin) was
prepared in MeCN containing 0.1% HAc (addition of 0.1%
HAc prevents degradation of base sensitive analytes in
MeCN [21]). A composite stock internal standard solution
(40 lg/mL) containing d6-a-HCH and triphenylphosphate
(TPP) was also prepared in MeCN containing 0.1% HAc.
Acetone, MeCN, methanol (MeOH), and toluene were
high purity grade solvents for pesticide residue analysis
from Burdick & Jackson (Muckegon, MI, USA), and the
glacial HAc was HPLC grade from Fisher Scientific (Fair
Lawn, NJ, USA). Anhydrous MgSO4 and NaAc N 3H2O
were obtained from Aldrich (Milwaukee, WI, USA), and
PSA (primary secondary amine) sorbent was from Varian
(Harbor City, CA, USA). The MgSO4 was heated for 5 h at
5008C in a muffle furnace to remove any residual water
and phthalates. For silanization, 5%-dimethyldichlorosilane (DMDCS) solution in toluene was obtained from
Supelco (Bellefont, PA, USA).
Compounds used as analyte protectants were 95% or better purity obtained from Sigma (St. Louis, MO, USA) and
Fluka (Buchs, Germany). A composite stock solution of 3ethoxy-1,2-propanediol (CAS: 1874-62-0), D-sorbitol
(CAS: 50-70-4), L-gulonic acid c-lactone (CAS: 1128-230) at concentration 40, 4, and 4 mg/mL, respectively, was
prepared in 7/3 H2O/MeCN for our experiments.
A sample of apple-based baby food was used for blanks,
fortified samples for recovery assays, and matrix-matched
standards for calibration in experiments.
2.2 Sample preparation
Matrix extracts were prepared according to the buffered
QuEChERS method [5] which entailed the following
steps: (1) weigh 15 g of thoroughly homogenized sample
into a 50 mL fluoroethylenepropylene (FEP) centrifugation tube; (2) add 15 mL 1% HAc in MeCN (v/v); (3) add
6 g anh. MgSO4 and 2.5 g NaAc N 3H2O; (4) shake vigorously for 1 min by hand; and (5) centrifuge the tube at
3450 rcf (relative centrifugal force) for 1 min. In experiments involving cleanup, dispersive-SPE entailed:
(6) transfer 1 mL of extract to a mini-centrifuge tube containing 50 mg PSA + 150 mg anh. MgSO4; (7) mix the
J. Sep. Sci. 2005, 28, 1048 – 1060
www.jss-journal.de
extract with the sorbent/dessicant for 20 s; and (8) centrifuge the tube at 3450 rcf for 1 min.
For the determination of recovery and repeatability of fortified pesticides in apple-based baby food (method validation), the procedure was similar, except the method was
scaled to 10 g sample weight and fortification was done
with appropriate volume to achieve 50 ng/g spikes and
100 ng/g internal standards before addition of the extraction solvent.
The matrix-matched standards were prepared by addition
of 20 lL of appropriate standards (125, 250, 500, 1250,
2500, 5000, and 10000 ng/mL pesticide mixtures each
containing 5000 ng/mL internal standards) in MeCN containing 0.1% HAc to 980 lL of blank extract. The standards in solvent were prepared by addition of 20 lL of
appropriate standards to 980 lL of MeCN containing
0.1% HAc. Concentrations of pesticides in both matrixmatched and solvent-based standards were as follows:
2.5, 5, 10, 25, 50, 100, and 200 ng/mL (which also correspond to ng/g in the sample). In the case of analyte protectants, 25 lL of the analyte protectant mixture was added
before the analysis to 1 mL of matrix-matched standards,
1 mL of standards in solvent, and 1 mL of final extracts for
determination of recovery and repeatability. The matrixmatched standards as well as extracts of fortified samples
were prepared from non-cleaned and cleaned buffered
QuEChERS extract.
For observation of the influence of water content during
DMI injection, 50 ng/mL pesticide solutions were prepared
in MeCN containing 0.1% HAc with different amount of
water (0%, 1%, 3%, 5%, 10%, and 15%). Before the analysis, 25 lL of the analyte protectant mixture was added to
1 mL of these solutions.
2.3 Silanization
Crimp-top direct thermal desorption (DTD) liners for
microvial and taper (part no. D100013; ATAS GL International, Veldhoven, The Netherlands), microvials (1.9 mm
ID, 2.5 mm OD, 15 mm long; part no. SPV1000; Scientific
Instrument Services, Ringoes, NJ, USA) and glass needle
guides (part no. D100008; ATAS GL International) were
washed in MeCN to remove glass particles, which were
present especially in the microvials. After drying at 2008C
for 2 h, the parts were inserted in 5% DMDCS in toluene.
After 45 min, the parts were washed in toluene and
inserted for 10 min in MeOH followed by drying at 2008C
for 30 min. To test silanization effects, a standard solution
at pesticide concentrations of 25 ng/mL in MeCN containing 0.1% HAc with and without addition of analyte protectants was used. The experiments were carried out using
(i) washed liner – needle guide – microvial; (ii) washed
needle guide – microvial and washed and silanized liner;
(iii) washed and silanized liner – needle guide – microvial.
i
2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Use of automated direct sample introduction with analyte protectants
1051
Table 1. Optimized conditions for the injection and GC analysis of the pesticides.
LV-DMI – GC – MS
GC – MS
Injector temperature program
1008C (3.5 min)
7 K/s to 2808C (held to the end of
GC – MS method)
2508C (constant temperature)
Vent time and flow
3.5 min and 50 mL/min
N/A
Carrier gas flow (He)
1 mL/min
1 mL/min
Injection volume
10 lL
1 lL
Splitless period
3.5 min
1.5 min
Oven temperature program
808C (3.5 min),
808C (1.5 min),
30 K/min to 1808C,
30 K/min to 1808C,
10 K/min to 2308C,
10 K/min to 2308C,
45 K/min to 3008C (9.61 min)
45 K/min to 3008C (8.61 min)
31.5 min
25 min
– Pre-injection steps
1 min
1 min
– Venting
3.5 min
0 min
– GC – MS analysis
23 min
20 min
– Cooling of GC oven
4 min
4 min
Total cycle time
2.4 LV-DMI – GC – MS conditions
3 Results and discussion
For the LV-DMI – GC – MS system, a Combi-PAL autosampler (CTC Analytics, Zwingen, Switzerland) with the
DTD/DMI accessory (ATAS GL International) was used in
combination with an Optic 3 programmable injector
(ATAS GL International). A Hewlett-Packard (Agilent, Little Falls, DE, USA) Model 5890 Series II Plus GC coupled
to a 5972 mass-selective detector was used for analysis.
The system was also equipped with a split/splitless injector and electronic pressure control. In the case of splitless
injection a double taper liner with internal volume of
800 lL (part no. 5181-3315, Agilent) was utilized. A Varian VF-5MS EZ-guard column (30 m60.25 mm6
0.25 lm) with integrated retention gap (5 m60.25 mm) at
the front and a 1 m of uncoated capillary at the back was
used in GC. The uncoated back section is very useful in
GC – MS because the heated MS transfer line is not controlled (cooled) by the column oven, and if air is introduced
into the column (as in DSI), then the stationary phase has
no hot zones where a reaction would occur with the oxygen in air. Optimized conditions of the splitless and DMI
methods are summarized in Table 1. The quadrupole was
operated in selected ion monitoring (SIM) mode detecting
2 – 3 ions for each analyte as listed in Table 2. The temperatures of the MSD interface, ion source, and quadrupole were 2808C, 1508C, and 2308C, respectively. The
electron multiplier voltage was set to +200 V above the
optimum after the standard instrument tuning procedure.
The responses (peak heights) of pesticides were normalized to the sum of peak heights for d6-a-HCH (m/z 224)
and TPP (m/z 326).
3.1 Optimization of LV-DMI – GC – MS conditions
J. Sep. Sci. 2005, 28, 1048 – 1060
www.jss-journal.de
Generally, optimization of DMI, especially in the case of a
multiresidue method, is a rather demanding task since
many instrument parameters are involved (initial and final
injector temperature, column flow, venting time and flow,
and transfer time). If DMI is also used as a LVI technique,
the process has to begin with solvent evaporation in the
injection port. Ideally, the solvent evaporation would be
conducted gently at a temperature below the boiling point
of the particular solvent. Venting of the solvent at low temperature is important not only to reduce solvent expansion, but also to minimize losses of highly volatile analytes [22]. Initial inlet temperatures of 508C and 608C were
reported when EtAc (boiling point 778C at 1 atm) was
used in DMI [12 – 14]. The boiling point of MeCN is 81.88C
at 1 atm, and under normal circumstances, 10 lL of
MeCN would take little time to evaporate with flowing gas.
However, the narrow opening of the DMI microvial and
higher pressure in the injection port makes the evaporation process take longer. Evaporation of 10 lL MeCN took
more than 5 min at 808C and 908C, which was not practical. Thus, the injector port temperature was set to 1008C
with 3.5 min evaporation time in the final method.
Although it is unlikely that transfer of the earliest eluting
analyte (dichlorvos) onto the GC column was 100% efficient, we obtained acceptably reproducible results for
dichlorvos and the other early-eluting pesticides. The final
injector temperature was set to 2808C (higher temperatures did not lead to larger peaks of the last eluting components, deltamethrin and azoxystrobin). The optimal trans-
i
2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ˇ ajka, Masˇtovsk, Lehotay, Hajsˇlov
C
1052
Table 2. Optimized conditions of the MS method used with LV-DMI – GC (ions used for quantification are given in bold characters).
Pesticide
Start time
[min]
Dichlorvos
6.50
tR
[min]
SIM ions
[m/z]
Dwell time
[ms]
7.1
185
109
220
7.3
141
94
95
8.2
127
192
109
Acephate
8.4
136
142
94
o-Phenylphenol
8.9
170
169
141
Methamidophos
Mevinphos
8.00
Omethoate
9.5
156
110
126
9.7
306
264
290
d6-a-HCH (IS)
10.2
224
222
187
HCB
10.3
284
286
249
Dimethoate
10.4
125
87
93
Trifluralin
9.60
Atrazine
10.5
200
215
173
10.7
179
199
304
Lindane
10.8
219
183
181
Chlorothalonil
11.1
266
264
268
Pirimicarb
11.2
166
238
72
Diazinon
Vinclozolin
10.64
11.6
285
212
198
Carbaryl
11.50
11.9
144
115
116
Heptachlor
11.9
272
274
100
Pirimiphos-methyl
Methiocarb
12.05
12.0
290
305
12.1
168
153
109
Dichlofluanid
12.2
123
167
224
Chlorpyrifos
12.3
314
197
258
12.6
331
329
125
Bromophos
12.52
Cyprodinil
12.8
224
225
210
Tolylfluanid
12.8
238
137
181
13.0
283
285
96
Captan
Procymidone
12.89
13.0
79
149
264
Thiabendazole
13.1
201
174
Folpet
Imazalil
13.27
13.1
260
262
297
13.4
215
217
173
p,p9-DDE
13.4
318
316
246
Ethion
13.7
231
153
384
p,p9-DDD
13.8
235
237
165
p,p9-DDT
14.00
Endosulfan sulfate
14.1
235
237
165
14.2
272
387
274
163
Tebuconazole
14.3
250
125
TPP (IS)
14.3
326
325
Bifenthrin
14.4
181
165
Cyhalothrin I
14.8
197
181
208
14.9
182
367
154
Cyhalothrin II
14.9
197
181
208
www.jss-journal.de
i
20
15
20
20
25
30
25
20
20
166
Phosalone
J. Sep. Sci. 2005, 28, 1048 – 1060
14.60
40
35
2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Use of automated direct sample introduction with analyte protectants
1053
Table 2. Continued.
Pesticide
Start time
[min]
tR
[min]
15.0
160
132
77
15.40
15.5
183
163
165
15.6
183
163
165
Azinphos-methyl
Permethrin I
Permethrin II
Coumaphos
Fenvalerate I
15.7
362
226
210
17.0
225
167
419
17.2
225
167
419
17.40
17.6
253
181
255
17.8
253
181
255
18.1
344
388
403
Deltamethrin II
Azoxystrobin
Dwell time
[ms]
16.40
Fenvalerate II
Deltamethrin I
SIM ions
[m/z]
17.98
50
100
100
100
compounds ( – OH); (iv) amino compounds (R – NH – );
(v) imidazoles, benzimidazoles ( – N2); and (vi) urea derivates ( – NH – CO – NH – ) [15, 19]. In addition to free silanol groups and metals potentially present on the glass surfaces (even those declared as “deactivated”), additional
active sites can originate from non-volatile co-extractives
deposited in the front part of the GC system from previous
injections of real-world samples.
Figure 1. Optimized conditions in the LV-DMI – GC – MS
method.
fer time (splitless period) was 3.5 min; a longer period
caused only broadening of analyte peaks. With respect to
GC conditions, the initial oven temperature was set to
808C to provide good analyte peak shapes and avoid oxidation of the stationary phase during DMI (when the liners
are exchanged, a small amount of air is also introduced).
Figure 1 illustrates the final conditions of the LV-DMI –
GC – MS method.
3.2 DMI system deactivation
Initial experiments involving DMI injections of solvent-only
solutions gave poor peak shapes and sensitivities for
many pesticides. This problem relates to degradation
and/or irreversible adsorption of susceptible analytes on
active sites in the injection liner, microvial, needle guide,
and column. Compounds containing the following functional groups in their molecule structures are typically
troublesome in this respect: (i) organophosphates
(–P=O); (ii) carbamates ( – O – CO – NH – ); (iii) hydroxy
J. Sep. Sci. 2005, 28, 1048 – 1060
www.jss-journal.de
DMI has the great advantage of preventing non-volatile
matrix components accumulating in the GC system
because they remain in the microvial and are removed
after each GC run. However, in comparison with other
injection techniques, DMI incorporates greater glass surface area in the inlet, which needs to be deactivated. Also,
the activity of the inlet may vary greatly throughout the GC
sequence because a new (different) liner, microvial, and
needle guide are introduced into the system each time,
potentially decreasing accuracy (both precision and trueness) of the results. Thus, an effective and reproducible
deactivation is essential in routine application of DMI for
GC analysis of pesticide residues and other susceptible
analytes.
We tested two deactivation approaches: (i) silanization of
glass surfaces with DMDCS to prevent the interaction with
silanol groups and (ii) addition of analyte protectants to
the injected samples. As mentioned in the Introduction,
analyte protectants strongly interact with the active sites,
thus their application can serve as a rather convenient
way of GC system (both inlet and column) deactivation
performed at each injection. As Figure 2 shows, silanization of particular glass parts (i. e. liner, needle guide, and
microvial) or of all components did not lead to observable
improvement in precision. On the other hand, the use of
analyte protectants resulted in significant improvement in
precision, regardless of whether in combination with silanization or not. Thus, we utilized a simple addition of analyte protectants instead of the time-consuming, laborintensive, and mainly ineffective silanization process.
i
2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
1054
ˇ ajka, Masˇtovsk, Lehotay, Hajsˇlov
C
Figure 2. Influence of silanization and use of analyte protectants on signal reproducibility (expressed as RSD, n = 3):
(A) and (D) washed liner – needle guide – microvial; (B) and
(D) washed needle guide – microvial and washed and silanized liner; (C) and (F) washed and silanized liner – needle
guide – microvial; (A) – (C) standard at concentration 25 ng/
mL in MeCN containing 0.1% HAc without analyte protectants, 10 lL injected using DMI; (D) – (F) standard at concentration 25 ng/mL in MeCN containing 0.1% HAc with analyte
protectants, 10 lL injected using DMI.
3.3 Influence of water content on analyte transfer
and peak focusing
The QuEChERS method uses MeCN for extraction, which
is miscible with water, and the salting out procedure using
anhydrous MgSO4 does not remove all of the water from
the MeCN extract. The water content in non-cleaned and
cleaned QuEChERS extract was found to be L14% and
L2%, respectively [4]. Thus, the injection of these rather
different extract solutions could impact transfer and peak
focusing of target analytes. To observe if this was the
case, 50 ng/mL pesticide standard solutions with analyte
protectants were injected, which contained different
amount of water (0%, 1%, 3%, 5%, 10%, and 15%). Significant influence of water content on results was
observed in the case of dichlorvos, where higher concentration of water resulted in about 40% decrease of its signal intensity (see Figure 3.A). On the other hand, peak
heights of most of the pesticides involved in our study
were not affected by the presence of up to 15% water in
MeCN, albeit variability increased slightly in injections
containing 10% and 15% water (see Figure 3.B).
3.4 Injection of baby food extracts with and
without clean-up
To evaluate the feasibility of the LV-DMI – GC – MS
approach for routine analysis of pesticide residues in real
J. Sep. Sci. 2005, 28, 1048 – 1060
www.jss-journal.de
Figure 3. Influence of water content in MeCN on the analyte
peak height for (A) Dichlorvos, and (B) p,p9-DDD, both at
50 ng/mL with analyte protectants added, 10 lL injected
using DMI (n = 3).
samples, a number of analyses were conducted for the
pesticides in MeCN containing 0.1% HAc (solvent standard with analyte protectants) and in buffered QuEChERS
extracts (extracts of fortified baby-food samples and
matrix-matched standards with analyte protectants) as
described in Experimental. Furthermore, the buffered
QuEChERS extracts were injected as non-cleaned (crude)
and cleaned extracts. Although one of the main advantages of the QuEChERS method is its high sample
throughput, we wanted to see if it can be increased even
further by avoiding the dispersive-SPE clean-up and analyzing the crude extracts satisfactorily using DMI. Of
course, the non-cleaned extracts contain more matrix compounds that could interfere with MS detection. Figure 4
shows how the clean-up step reduced the amount of nonvolatile components retained in the used microvials (only a
thin film of co-extractives were present after the analysis of
cleaned extracts). In the case of the crude extracts, the
non-volatile co-extractives on the wall of microvial are visible. We should also note that the syringe plunger often
became stuck when the non-cleaned extracts were repeatedly injected (presumably due to the higher amounts of
less volatile components). The injection of rather polar analyte protectants can also pose this problem, but the use of
acetone-water mixture (1:1, v/v) for syringe washing
serves as a good preventative measure [20].
i
2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Use of automated direct sample introduction with analyte protectants
1055
Table 3. Results of the validation study (mean recoveries and relative standard deviations, RSDs) obtained in cleaned and noncleaned buffered QuEChERS extracts of apple-based baby food samples spiked at 50 ng/g (n = 6). Matrix-matched standards
with analyte protectants were used for calibration, for which lowest calibration levels (LCLs) are given.
Pesticide
Cleaned extracts
Non-cleaned extracts
Recovery [%]
RSD [%]
LCL [ng/g]
Recovery [%]
RSD [%]
LCL [ng/g]
Acephate
–
–
100
–
–
>200
Atrazine
99
6
5
94
4
5
Azinphos-methyl
–
–
100
–
–
200
Azoxystrobin
98
7
2.5
106
15
2.5
Bifenthrin
98
4
2.5
92
10
2.5
Bromophos
101
3
2.5
96
8
2.5
Captan
–
–
100
–
–
>200
Carbaryl
93
12
25
–
–
200
Chlorothalonil
82
19
2.5
83
23
10
Chlorpyrifos
96
3
2.5
97
5
5
Coumaphos
100
11
2.5
100
15
10
k-Cyhalothrin I+II
95
5
2.5
103
5
2.5
Cyprodinil
97
1
2.5
99
4
2.5
p,p9-DDD
91
6
2.5
106
11
2.5
p,p9-DDE
97
3
2.5
93
4
2.5
p,p9-DDT
104
17
2.5
95
7
2.5
Deltamethrin I+II
93
6
10
101
2
10
Diazinon
101
4
2.5
96
6
10
Dichlofluanid
86
14
10
84
20
25
Dichlorvos
101
17
5
97
8
10
Dimethoate
99
14
25
–
–
200
Endosulfan sulfate
97
15
2.5
97
18
10
Ethion
102
8
2.5
98
10
2.5
Fenvalerate I+II
96
6
5
104
5
5
Folpet
98
11
25
–
–
100
HCB
98
7
2.5
94
8
2.5
Heptachlor
99
5
2.5
95
4
5
Imazalil
104
4
2.5
95
6
5
Lindane
94
5
5
96
5
5
Methamidophos
87
11
25
–
–
>200
Methiocarb
–
–
Interference
–
–
Interference
Mevinphos
99
9
5
91
11
10
Omethoate
–
–
100
–
–
200
Permethrin I+II
99
5
2.5
108
5
2.5
o-Phenylphenol
99
8
2.5
93
6
2.5
Phosalone
97
20
10
94
29
25
Pirimicarb
97
5
5
96
4
2.5
Pirimiphos-methyl
100
6
2.5
98
6
2.5
Procymidone
95
6
5
98
6
10
Tebuconazole
95
2
2.5
96
4
5
Interference
Thiabendazole
–
–
Interference
–
–
Totylfluanid
86
15
10
92
15
50
Trifluralin
100
10
2.5
93
5
2.5
Vinclozolin
96
2
5
100
4
10
J. Sep. Sci. 2005, 28, 1048 – 1060
www.jss-journal.de
i
2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
1056
ˇ ajka, Masˇtovsk, Lehotay, Hajsˇlov
C
Table 3 compares lowest calibration levels (LCLs)
obtained for the tested pesticides in cleaned and noncleaned buffered QuEChERS extracts. The analysis of
cleaned extracts yielded better detectability for many analytes (dichlorvos, mevinphos, omethoate, dimethoate, diazinon, chlorothalonil, vinclozolin, carbaryl, heptachlor,
dichlofluanid, chlorpyrifos, tolylfluanid, procymidone, captan, folpet, imazalil, endosulfan sulfate, tebuconazole,
phosalone, and coumaphos). Moreover, methamidophos,
acephate, and azinphos-methyl could not be determined in
the crude extracts even at the highest tested concentration
level of 200 ng/g. A co-elution of methiocarb and thiabendazole with matrix compounds precluded quantification of
these analytes in both the crude and cleaned extracts.
Figure 4. DMI liner after 10 lL injection of: (1) cleaned babyfood extract; (2) non-cleaned baby-food extract; and
(3) detail of microvials used for introduction of non-cleaned
extracts.
In the case of k-cyhalothrin and deltamethrin, formation of
second peaks (isomers) with the same MS spectra was
observed. The extent of conversion of both pesticides to
their isomers was sample dependent, increasing in order:
pure solvent (MeCN with 0.1% HAc) a cleaned QuEChERS extract a non-cleaned QuEChERS extract (see
Figure 5). The isomerization process of k-cyhalothrin and
Figure 5. Chromatograms of 10 lL DMI injections of 100 ng/mL (A) k-cyhalothrin (m/z 197) and (B) deltamethrin (m/z 253) and
their isomers (denoted with asterisks) obtained in: (1) MeCN with 0.1% HAc; (2) cleaned buffered QuEChERS extract; (3) noncleaned buffered QuEChERS extract. Analyte protectants were added to all solutions.
J. Sep. Sci. 2005, 28, 1048 – 1060
www.jss-journal.de
i
2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Use of automated direct sample introduction with analyte protectants
deltamethrin in MeCN (and acetone) was described previously [21], demonstrating that formation of the second
peak depends on the activity of the GC system and presence of different co-injected matrix components. To
avoid quantification errors, peak heights of both isomers
were summed for quantification purposes.
3.5 Method validation
The combination of the buffered QuEChERS sample preparation method and optimized LV-DMI – GC – MS was
evaluated in a validation study, involving analysis of 6
replicates of apple-based baby food spiked at 50 ng/g by
the tested 44 pesticides. Table 3 provides mean recoveries, relative standard deviations (RSDs), and LCLs
obtained by the analysis of cleaned and non-cleaned buffered QuEChERS extracts. The best overall results were
obtained by analysis of the cleaned extracts, in which
case, 38 pesticides were satisfactorily quantified (only
acephate, omethoate, methiocarb, captan, thiabendazole, and azinphos-methyl could not be determined at the
spiking level of 50 ng/g). Using matrix-matched standards
(with analyte protectants) for calibration, average recoveries of the determined pesticides were between 82 and
104% with RSDs d 20%. In crude extracts, four additional
pesticides (carbaryl, dimethoate, folpet, and methamidophos) could not be quantified and the recoveries of the 34
determined pesticides were between 83 and 108% with
RSDs d 20%, except for chlorothalonil and phosalone
which gave slightly less reproducible results (RSDs 23%
and 29%, respectively). Thus, with the exception of chlorothalonil and phosalone in crude extracts, the validation
results for the determined pesticides in both extracts fully
met the EU criteria for quantitative methods set for the
pesticide concentration range of 10 – 100 ng/g, requiring
repeatability (RSDs) d 20% and mean recovery within the
range of 70 – 120% [23]. For the majority of the analytes,
the LCLs were d10 ng/g (mostly 2.5 ng/g, see Table 3).
Thus for these pesticides, the presented method can be
used for control of the 10 ng/g maximum residue limit
recently established in the EU for pesticides in baby
food [24].
In terms of the retention time reproducibility, the RSDs
were a0.05% for cleaned buffered QuEChERS extracts
(except for the early eluting analytes methamidophos and
dichlorvos, which gave RSDs of 0.27% and 0.13%,
respectively) and less than 0.06% for pesticide residues in
non-cleaned extracts.
In addition to the calibration using matrix-matched standards, we also evaluated the possibility to quantify the
tested pesticides using matrix-free, solvent standards.
For effective deactivation of the GC system (see Section
3.2), analyte protectants were added to both types of calibration standards as well as fortified samples extracts.
Moreover, the use of a suitable combination of analyte
J. Sep. Sci. 2005, 28, 1048 – 1060
www.jss-journal.de
1057
protectants in solvent standards is a more convenient
approach to compensate for the matrix-induced response
enhancement effect than the laborious and time-consuming matrix-matching approach.
Masˇtovsk et al. [20] evaluated various combinations of
promising analyte protectants to overcome matrix effects
in analysis of GC-amenable pesticides in QuEChERS
extracts using hot splitless injection (1 lL) at 2508C. A
mixture of 3-ethoxy-1,2-propanediol, L-gulonic acid c-lactone, and D-sorbitol (added at 10, 1, and 1 mg/mL,
respectively, to both solvent standards and matrix
extracts) provided the best overall results in terms of
effective compensation for matrix-induced enhancement
and improved ruggedness in a long-term performance
test. For the 10 lL DMI injections in this study, we used
the same mixture of analyte protectants, but 10-fold lower
concentrations were added to achieve the same amount
of analyte protectants introduced into the GC system as in
the case of the 1 lL splitless injection (10, 1, and 1 lg
injected, respectively).
Figure 6 compares pesticide recoveries in cleaned
extracts obtained using matrix-matched and solvent standards, both with addition of analyte protectants. The latter
approach provided comparable results meeting the 70 –
120% recovery criterion for most of the determined pesticides. Slightly higher results were observed for p,p9-DDD
and cyprodinil (recoveries 125 and 126%, respectively),
whereas the recoveries of imazalil and azoxystrobin were
more overestimated, reaching values of 163 and 204%,
respectively. In the case of endosulfan sulfate, the recovery calculated vs. solvent standard was unexpectedly
lower than with the use of matrix-matching (64 vs. 97%).
Lower recoveries (48% – 68%) were also obtained for a
group of base-sensitive pesticides folpet, dichlofluanid,
tolylfluanid, and chlorothalonil (not shown in Figure 6), in
which case the presence of matrix components probably
led to partial degradation in both samples and matrixmatched standards. Thus, higher responses were
achieved in matrix-free, MeCN solutions, where the addition of 0.1% HAc effectively prevented degradation of
these analytes [21].
For comparison purposes, we also injected the matrixmatched standards in cleaned extracts and solvent standards, both with analyte protectants (at 10, 1, and 1 mg/
mL in this case), using 1 lL hot splitless injection. Very
similar responses were obtained for all determined pesticides in both calibration standard types, thus as observed
previously [20], the optimized mixture of 3-ethoxy-1,2-propanediol, L-gulonic acid c-lactone, and D-sorbitol was
highly effective for compensation of matrix-induced
response enhancement in hot splitless injection. Figure 7
shows overlays of chromatograms for three susceptible
pesticides o-phenylphenol, cyprodinil, and azoxystrobin
i
2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
1058
ˇ ajka, Masˇtovsk, Lehotay, Hajsˇlov
C
Figure 6. Recoveries of pesticides spiked at 50 ng/g in apple-based baby food samples, extracted using the buffered QuEChERS method with dispersive-SPE clean-up and analyzed by LV-DMI – GC – MS. Calibration was performed using (A) matrixmatched standards and (B) standards in pure solvent (MeCN with 0.1% HAc). Analyte protectants were added to calibration solutions as well as extracts of fortified samples. The error bars represent standard deviations (n = 6); dashed lines denote acceptable mean recovery range of 70 – 120%.
(all at 50 ng/mL) obtained by hot splitless and DMI injections in matrix-matched and solvent standards, both with
analyte protectants. Also, injections in solvent standards
without analyte protectants are shown for comparison,
demonstrating dramatic improvement in peak shapes and
intensities achieved with the use of analyte protectants.
As discussed above, cyprodinil gave slightly higher signals in matrix-matched vs. solvent standards in DMI. In
the case of azoxystrobin injected by DMI, the matrixinduced enhancement was more pronounced, whereas
sufficient compensation for this effect was observed when
using hot splitless injection. It should be noted, however,
that the analytes problematic in this respect in DMI (azoxystrobin, imazalil, and cyprodinil) are rather LC- than GCamenable pesticides, for which LC – MS generally gives
more reliable results than GC – MS in routine practice [5].
the best quantification for the most pesticides, but crude
QuEChERS extracts with analyte protectants added can
be analyzed directly in DMI to achieve acceptable quantitative results for most of the pesticides.
No concerns were observed in this study with respect to
the theoretical aspects of the DMI approach. The two
major advantages entail the large-volume injection that
DMI provides and the removal of non-volatile matrix components after every injection that would normally contaminate the GC system in traditional injection techniques.
However, the commercial DMI device has practical concerns (e. g. potential for leaks, difficulty of decrimping the
caps, software issues) that should be eliminated, and
indeed, the manufacturer has introduced a revised model.
A competing manufacturer has also entered the market
with a similar approach, which should make the automated DSI possible in a number of routine applications.
4 Conclusions
In this study, the feasibility of LV-DMI – GC – MS was evaluated for 44 diverse pesticides in apple-based baby food.
The results indicated that the approach was very useful in
this application for nearly all pesticides evaluated. A key
aspect in the success of the approach was the use of analyte protectants to deactivate glass surfaces in the DMI
system. Determination limits were d10 ng/g for 34 of the
pesticides in the spiked baby food extracts, which is particularly important to meet the EU baby food directive. The
use of dispersive-SPE clean-up was needed to achieve
J. Sep. Sci. 2005, 28, 1048 – 1060
www.jss-journal.de
Acknowledgments
This work was supported in part by Research Grant Award
No. US-3500-03 from BARD, the United States-Israel Binational Agricultural Research and Development Fund.
Disclaimer
Mention of brand or firm name does not constitute an endorsement by the U.S. Department of Agriculture above
other of a similar nature not mentioned.
i
2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Use of automated direct sample introduction with analyte protectants
1059
Figure 7. Comparison of peak responses and shapes of 50 ng/mL o-phenylphenol (m/z 170), cyprodinil (m/z 224), and azoxystrobin (m/z 344) obtained in: (1) MeCN with 0.1% HAc; (2) MeCN with 0.1% HAc and addition of analyte protectants; (3) cleaned
extract with addition of analyte protectants. Samples injected using (A) 1 lL splitless injection at 2508C and (B) 10 lL DMI injection. In both cases a mixture of 3-ethoxy-1,2-propanediol, D-sorbitol, and L-gulonic acid c-lactone was used as analyte protectants (10 lg, 1 lg, and 1 lg injected).
References
[1] F.E. Ahmed, TrAC Trends Anal. Chem. 2001, 20, 649 – 661.
[2] K. Masˇtovsk, S.J. Lehotay, J. Hajsˇlov, J. Chromatogr. A
2004, 1054, 335 – 349.
[3] J. Zrostlkov, J. Hajsˇlov, M. Godula, K. Masˇtovsk, J.
Chromatogr. A 2001, 937, 73 – 86.
J. Sep. Sci. 2005, 28, 1048 – 1060
www.jss-journal.de
ˇ tajnbaher, F.J.
[4] M. Anastassiades, S.J. Lehotay, D. S
Schenck, J. AOAC Int. 2003, 86, 412 – 431.
[5] S.J. Lehotay, K. Masˇtovsk, A.R. Lightfield, J. AOAC Int.
2005, 88, 615 – 629.
[6] M. Careri, F. Bianchi, C. Corradini, J. Chromatogr. A 2002,
970, 3 – 64.
[7] H.-J. Stan, M. Linkerhgner, J. Chromatogr. A 1996, 727,
275 – 289.
i
2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
1060
ˇ ajka, Masˇtovsk, Lehotay, Hajsˇlov
C
[8] A. Amirav, S. Dagan, Eur. J. Mass Spectrom. 1997, 3, 105 –
111.
[9] H. Jing, A. Amirav, Anal. Chem. 1997, 69, 1426 – 1435.
[10] S.J. Lehotay, J. AOAC Int. 2000, 83, 680 – 697.
[11] S.J. Lehotay, A.R. Lightfield, J.A. Harman-Fetcho, D.J.
Donoghue, J. Agric. Food Chem. 2001, 49, 4589 – 4596.
[12] S. de Koning, G. Lach, M. Linkerhgner, R. Lscher, R.H.
Tablack, U.A.Th. Brinkman, J. Chromatogr. A 2003, 1008,
247 – 252.
[13] K. Patel, R.J. Fussell, D.M. Goodall, B.J. Keely, Analyst
2003, 128, 1228 – 1231.
[14] K. Patel, R.J. Fussell, D.M. Goodall, B.J. Keely, Food Addit.
Contam. 2004, 21, 658 – 669.
[15] J. Hajsˇlov, J. Zrostlkov, J. Chromatogr. A 2003, 1000,
181 – 197.
[16] K. Masˇtovsk, S.J. Lehotay, J. Chromatogr. A 2003, 1000,
153 – 180.
[17] D.R. Erney, A.M. Gillespie, D.M. Gilvydis, C.F. Poole, J.
Chromatogr. A 1993, 638, 57 – 63.
J. Sep. Sci. 2005, 28, 1048 – 1060
www.jss-journal.de
[18] F.J. Schenck, S.J. Lehotay, J. Chromatogr. A 2000, 868,
51 – 61.
[19] M. Anastassiades, K. Masˇtovsk, S.J. Lehotay, J. Chromatogr. A 2003, 1015, 163 – 184.
[20] K. Masˇtovsk, S.J. Lehotay, M. Anastassiades, Anal.
Chem., submitted.
[21] K. Masˇtovsk, S.J. Lehotay, J. Chromatogr. A 2004, 1040,
259 – 272.
[22] M. Godula, J. Hajsˇlov, K. Masˇtovsk, J. Krˇivnkov, J.
Sep. Sci. 2001, 24, 355 – 366.
[23] Quality control procedures for pesticide residues analysis –
guidelines for residues monitoring in the European Union,
3rd ed., European Commission document SANCO/10476/
2003, Brussels, 2004.
[24] Commission Directive 2003/13/EC of 10 February 2003
amending Directive 96/5/EC on processed cereal-based
foods and baby foods for infants and young children, Off. J.
Eur. Union 2003, L41, 33 – 36.
i
2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim