Utilizing a Rapid Multi-Plug Filtration Cleanup Method for 72 Pesticide Residues in Grape Wines Followed by Detection with Gas Chromatography Tandem Mass Spectrometry

A convenient and fast multi-residue method for the efficient identification and quantification of 72 pesticides belonging to different chemical classes in red and white grape wines has been developed. The analysis was based on gas chromatography tandem quadrupole mass spectrometric determination (GC–MS/MS). The optimization strategy involved the selection of the amount of multi-walled carbon nanotubes (MWCNTs) and the number of cleanup procedure cycles for multi-plug filtration cleanup (m-PFC) to achieve ideal recoveries and reduce the sample matrix compounds in the final extracts. The optimized procedure obtained consistent recoveries between 70.2 and 108.8% (70.2 and 108.8% for white wine, and 72.3 and 108.4% for red wine), with relative standard deviations (RSDs) that were generally lower than 9.2% at the three spiking levels of 0.01, 0.05 and 0.1 mg/kg. The linearity was studied in the range between 0.002 and 0.1 mg/kg using pesticide standards prepared both in pure solvent and in the presence of the matrix, showing coefficients of determination (R2) higher than 0.9495 for all the pesticides. To improve accuracy, matrix-matched calibration curves were used for calculating the quantification results. Finally, the method was used successfully for detecting pesticide residues in commercial grape wines.


Introduction
Grape wine is one of the most commonly consumed alcoholic beverages in the world. In addition to its distinctive flavor, the moderate consumption of wine is correlated with reduced risks of both mortality and morbidity from human cardiovascular disease [1] and oxidative damage [2]. The worldwide consumption of wine is increasing steadily and has reached up to 240 Mhl per year, according to records collected by the International Organisation of Vine and Wine over recent years [3].

Chemicals and Materials
The initial sample preparation was identical to that used for the QuEChERS method  The main objective of this work was to develop a fast, sensitive and reliable analytical m-PFC method. In this work, 72 pesticides with different chemical structures in wine were determined using GC-MS/MS. This method was successfully applied for market survey samples.

Chemicals and Materials
The initial sample preparation was identical to that used for the QuEChERS method [38]. Standard compounds were provided by the Institute of the Control of Agrochemicals, Ministry of Agriculture, China. The purity of the standard pesticides was 95-99%. Stock solutions of 10 mg/L of the pesticide mixture were prepared in acetonitrile and stored at −20 • C. The working solutions were prepared daily. HPLC-grade acetonitrile was obtained from Fisher Chemicals (Fair Lawn, NJ, USA). Analytical-reagent-grade anhydrous sodium chloride (NaCl) and magnesium sulfate (MgSO4) were obtained from Sinopharm Chemical Reagent (Beijing, China). Tianjin Bonna-Agela Technologies (Tianjin, China) provided MWCNTs with different average external diameters and PSA. The MWCNTs were dried for 2 h at 120 • C to remove the absorbed water and then kept in desiccators for storage.

Apparatus and Conditions
Centrifugation was performed with two different instruments: an Anke TDL-40B centrifuge equipped with a bucket rotor (4 × 100 mL) (Shanghai, China) and a SIGMA 3K15 microcentrifuge equipped with angular rotor (24 × 2.0 mL) (Sigma Laborzentrifugen GmbH, Osterode am Harz, Germany). A QL-901 Vortex (Kylin-bell Lab Instruments, Jiangsu, China) was used for preparing the samples. A Meiling BCD-245W refrigerator freezer (Beijing, China) was used to control the temperature of the samples.
Determinations were performed using an Agilent 7000 A triple-quadrupole mass spectrometer interfaced to an Agilent 7890 A GC. An Agilent Technologies analytical capillary column (HP-5MS; 30 m × 250 µm × 0.25 µm film thickness) was used for GC separation, with helium (99.9999%) as the carrier gas at a constant flow rate of 1.2 mL/min. The column temperature was initially set at 50 • C and held for 1 min, then increased to 130 • C (and held for 1 min) at a rate of 30 • C/min, then increased to 250 • C at a rate of 5 • C/min, and finally increased to 290 • C at a rate of 10 • C/min, which was held for 5 min. The temperature of the injector port was 250 • C, and a volume of 1 µL was injected in splitless mode. The total running time was 38 min.
The mass spectrometer was operated in electron ionization mode (70 eV). The default instrument settings of a collision gas flow for N 2 of 1.5 mL/min and He of 2.25 mL/min, and a quadrupole temperature of 150 • C, were used in all the MS/MS experiments. The detector voltage was automatically set by the instrument after automated MS/MS tuning, which was typically 1250 V. A full autotune of the mass spectrometer, using the default parameters of the instrument, was performed before each sequence. Agilent MassHunter was used for instrument control and data acquisition/processing. For the final multiple reaction monitoring (MRM) acquisition method, two ion transitions at the experimentally optimized collision energy (CE) were monitored for each analyte. Both pairs of the MRM transitions were used for confirmation analysis, and the most sensitive transitions were selected for quantification analysis to obtain better separation efficiency. Table 1 summarizes the optimized MS/MS conditions for the individual analytes and their typical retention times (RT).

Sample Preparation
A QuEChERS-based approach was adapted to isolate the 72 analytes in the wine samples. The samples were obtained from a local supermarket and homogenized with a blender for 1 min at room temperature. For the determination of the recovery, the homogenized samples (10.0 ± 0.1 g) were spiked by adding the standard stock solutions at three concentrations: 0.01, 0.05 and 0.1 mg/kg. The spiked samples were set aside for 30 min before extraction.
An amount (10.0 ± 0.1 g) of each wine sample was weighed into a 50 mL centrifuge tube and 10 mL of acetonitrile was added. The resulting solution was shaken using a vortex for 1 min; then, 1 g of sodium chloride and 4 g of anhydrous magnesium sulfate were added. The tube was cooled immediately to room temperature in an ice-water bath. The centrifuge tube was shaken vigorously for 1 min to prevent salt agglomeration before centrifugation at 3800 rpm for 5 min. The 1 mL supernatant was used for further m-PFC.

m-PFC Procedures
The m-PFC procedure involved the following steps (shown in Figure 1): 1 mL of the supernatant was introduced into a 2.0 mL centrifuge tube. The sorbents (including 150 mg of anhydrous MgSO 4 ) in the column were adopted from the optimized d-SPE sorbents. As shown in Figure 1, the syringe needle was kept under the surface of the extract; then, the syringe piston was pulled and pushed to let the extracts pass through the sorbents for the purpose of cleaning. Finally, the layer was filtered through a 0.22 µm filter membrane. The extract was placed into a GC vial for chromatographic analysis.

Method Performance
The analytical method was validated according to the following parameters: the linearity, limit of quantification (LOQ), limit of detection (LOD), precision and accuracy. The test of linearity used matrix-matched calibration by analyzing samples of red wine and white wine. The precision and accuracy experiments were carried out in five replicates, each at three fortification levels (0.01, 0.05 and 0.1 mg/kg) for the sample matrix. According to SANTE/12682/2019 [65], the LOQs were determined as the concentrations of analyte giving a signal-to-noise ratio (S/N) ≥ 3 and analyte peaks from both product ions in the extracted ion chromatograms must fully overlap. Ion ratio from sample extracts should be within ±30% (relative) of average of calibration standards from the same sequence.

Analysis of Grape Wine Samples
Grape wine samples, including fifty red grape wines and twenty white grape wines, were purchased at supermarkets in Beijing and they belonged to several vintages between 2011 and 2018. Seventy samples were produced in wineries from different countries: China (42), France (7), Spain (5), Portugal (5), Italy (4), Australia (4) and the USA (3). The alcoholic strength ranged from 10% to 15% (Alc/vol). Bottled wines were stored in their original packaging at 5 • C.

Amount of the MWCNTs
After the analytes had been extracted using 10 mL of acetonitrile, followed by the partitioning of the analyte molecules in an organic solvent in the presence of a salt mixture (the salting-out effect), 1 mL of the acetonitrile phase was further cleaned by the m-PFC procedure. Zhao et al. [62] found that different amounts of MWCNT sorbents had a significant influence on the purification and recovery of the pesticide extracts. To evaluate the effect of this parameter, different amounts of MWCNT were investigated in the same procedure. The amount of sorbent material was progressively increased from 5 mg to 10, 15 and 20 mg. The experiment was performed using 1 mL of the acetonitrile extract at a spiking level of 0.1 mg/kg and it was then cleaned by the m-PFC method with different amounts of MWCNT. The recovery of most of the analytes increased with the amount of MWCNTs and the results were within an acceptable range: 70-120% for red wine. As shown in Figure 2, upon increasing the amount of MWCNTs from 5 to 10 mg, the recovery levels for epoxiconazole, profenofos, azoxystrobin and bifenthrin remained acceptable (70-103%). However, the recovery decreased to 33-69% when the amount of MWCNTs was increased to 15 and 20 mg. In addition, although better recovery was achieved with 5 mg of MWCNT materials, the performance was not as good as that with 10 mg, and there was more chromatography interference when 5 mg was used. The recovery was also acceptable with 10 mg of MWCNTs. Consequently, 10 mg (1 mL of the extract) was used as the optimum amount for m-PFC in further studies, since acceptable recovery and good cleanup performance was obtained with this amount.

Optimization of the m-PFC Procedure Cycle Times
In order to obtain the best recovery and cleanup performance, the cycles of pulling and pushing during the m-PFC procedure were optimized. The recovery was acceptable with one and two cycles of pulling and pushing, but the cleanup performance was not as good as that with three cycles and there was more chromatography interference for one or two pull-push cycles. In addition, four cycles were also tested, but there was no significant difference in cleanup performance from that with three cycles. As a result, three cycles of pulling and pushing were chosen for the optimized m-PFC procedure.

Optimization of the m-PFC Procedure Cycle Times
In order to obtain the best recovery and cleanup performance, the cycles of pulling and pushing during the m-PFC procedure were optimized. The recovery was acceptable with one and two cycles of pulling and pushing, but the cleanup performance was not as good as that with three cycles and there was more chromatography interference for one or two pull-push cycles. In addition, four cycles were also tested, but there was no significant difference in cleanup performance from that with three cycles. As a result, three cycles of pulling and pushing were chosen for the optimized m-PFC procedure. Figure 3 shows the purification effects of different cleanup cycles.

Optimization of the m-PFC Procedure Cycle Times
In order to obtain the best recovery and cleanup performance, the cycles of pulling and pushing during the m-PFC procedure were optimized. The recovery was acceptable with one and two cycles of pulling and pushing, but the cleanup performance was not as good as that with three cycles and there was more chromatography interference for one or two pull-push cycles. In addition, four cycles were also tested, but there was no significant difference in cleanup performance from that with three cycles. As a result, three cycles of pulling and pushing were chosen for the optimized m-PFC procedure. Figure 3 shows the purification effects of different cleanup cycles.

Linearity and Matrix Effects
Linearity was studied in the range of 0.002-0.1 mg/L for all the pesticides at five calibration levels (0.002, 0.005, 0.01, 0.02, 0.05 and 0.1 mg/L) by a matrix-matched standard calibration in blank extracts of red wine and white wine. Linear calibration graphs were constructed by plotting analyte concentrations versus the relative peak areas of the calibration standards. The linearity values, calculated as the determination coefficients (R 2 ) for each pesticide from the matrix-matched calibration (m-PFC cleanup) plots, are shown in Table 2. The quantitative results of the detection method greatly depend on its calibration. Both pure solvent-based as well as matrix-matched calibrations gave R 2 values better than 0.985. This was remarkable, considering the complexity of the matrices. The matrix effects (ME) were evaluated in terms of slope ratios: 100 × (1-slope acetonitrile/slope matrix) [62,66].
The matrix effects include enhancement or suppression effects, so the concentration results obtained can be erroneous, depending on the solvent calibration curves [18]. To examine the matrix effects, matrix-matched standards were compared with solvent standards. Table 2 summarizes the results. In our work, it was considered that, if the value was in the range of −10 to 10%, the matrix effect could be ignored; if the value was lower than −10% or higher than 10%, this showed a matrix-suppression or an enhancement effect, respectively [62,66]. The results show that, in red and white wine, 45 and 25 of the pesticides presented an enhancement effect (ME > 0), respectively, and the other 27 and 47 of the pesticides showed a suppression effect (ME < 0), respectively; 44 and 34 of the pesticides expressed distinct matrix-suppression and enhancement effects. Therefore, for more accurate results, validation experiments were performed for pesticide residue concentrations in non-compliant samples, calculated using matrix-matched calibration standards and excluding any influence produced by matrix effects, as recommended in SANTE/12682/2019 [65]. In order to overcome the adverse impact of matrix effects on the quantified results, we calibrated the sample results with matrix-matched standards to guarantee the correct quantification of the pesticide concentrations in real samples.

Recovery and Precision
The recovery and repeatability of the method were established to evaluate the method's performance. The repeatability and the accuracy of the method were tested by carrying out five consecutive extractions (n = 5) of spiked matrices at three concentration levels (0.01, 0.05 and 0.1 mg/kg). All the recovery values were determined from analyses of the 72 pesticides in the matrices. The values were calculated using matrix-matched calibration standards, as stated in Section 3.3.1. Table 3 shows detailed recovery and repeatability data for all the pesticides analyzed in wine matrices. The recovery rates of all the pesticides were in the range of 70.2-108.8% (between 70.2 and 108.8% for white wine, and between 72.3 and 106.0% for red wine). The relative standard deviations (RSDs) were below 8.3% for all the cases. All the recovery values and RSDs are in the acceptable range of SANTE/12682/2019 [65]. Table 3. Average recovery, RSD, LOD and LOQ after application of the m-PFC procedure, determined by GC-MS/MS in wine.

Limits of Quantitation and Limits of Detection
The described method was tested for the simultaneous extraction and determination of 72 analytes in wine matrices, which manifested varying LODs and LOQs. Since LODs and LOQs are matrix-dependent, it is recommended to perform matrix-matched calibrations for the quantitative analysis of unknown samples in complex matrices. Table 3 shows the LOD and LOQ values for the pesticides in wine under study. The LODs and LOQs ranged from 0.002 to 0.01 mg/kg and from 0.01 to 0.05 mg/kg, respectively.
In general, the validation data for all the analytes were in accordance with the EU guidelines (2019) SANTE/12682/2019 [65] for pesticide residue analysis, reflecting the good performance of the method. Comparison with other QuEChERS methods for determining pesticides in grape wines, the m-PFC method which our group has proposed, showed significant advantages in terms of the amount of time that is consumed, each sample took less than two minutes for m-PFC cleanup, so it would be time-saving during the processing of enormous samples. Regarding recoveries, RSDs, LOQs and the number of pesticides, the m-PFC method presented similar effects when compared to other methods (Table 4).

Method Application
The developed QuEChERS method with a m-PFC cleanup step was applied to real samples. Seventy reference samples (50 for red wine samples and 20 for white wine samples) from the supermarkets in Beijing were treated and analyzed by GC-MS/MS. Since MRLs have not yet been set for wine, according to SANTE/2020/12830 [66] and EC 657/2002 [67], it is accepted that the MRLs for wine are the same as those for wine grapes. Table 5 shows the detected concentrations of pesticides in the real samples from supermarkets in Beijing. Pesticide residues were detected in seven samples (10%). The most frequently detected pesticides were difenoconazole (2.9%), pyridaben (4.3%), carbosulfan (2.9%), pyr imethanil (1.4%), propyzamide (1.4%), simazine (4.3%) and atrazine (4.3%). To date, and with regard to grapevine products, MRLs have only been set for grapes, taking the transfer in the wine into account. The MRLs for wine are still not widely established [15][16][17]. Therefore, we referred to the maximum permitted residue levels set by EU2018/555 [68], The concentrations of pesticides in the selected wine samples did not exceed the permitted residue levels. The average concentration (X) and standard deviation (SD) of each compound were calculated considering the pesticide residues below the limit of quantification (LOQ) for the method as non-detected(ND); n the number of 3 times each sample had been analyzed.

Conclusions
An efficient and effective m-PFC multi-residue method was developed for the determination of 72 pesticides in wine by GC-MS/MS. The m-PFC method, which could be carried out without any solvent evaporation, vortexing or centrifugation procedure, proved to be a simple and rapid cleanup method. The method achieved high-quality results (good repeatability and recovery, and a wide analytical scope) and had several practical benefits (low cost, little labor, high sample throughput, hardly any waste and low labware equipment and space demands). The method was found to be very sensitive and gave a LOQ of <0.05 mg/kg for all the analytes. In conclusion, m-PFC could be used as a feasible, convenient and rapid high-throughput cleanup method for the analysis of analytes in wine.