Evaluation of membrane filtration for cleanup in multi-residue pesticide analysis of spinach

Food Control 79 (2017) 134e142

Contents lists available at ScienceDirect

Food Control journal homepage: www.elsevier.com/locate/foodcont

Evaluation of membrane filtration for cleanup in multi-residue pesticide analysis of spinach Jangho Hong**, Ayato Kawashima*, Minami Okamoto, Noriaki Hamada Department of Environmental Science for Industry, Ehime University, 3-5-7 Tarumi, Matsuyama, Ehime, 790-8566, Japan

a r t i c l e i n f o

a b s t r a c t

Article history: Received 8 November 2016 Received in revised form 22 February 2017 Accepted 28 March 2017 Available online 29 March 2017

This study reports a new membrane filtration-based cleanup method for the analysis of pesticides. Recovery and cleanup by membrane filtration using 11 different membranes, classified by their molecular weight cut-off, pore size, and material, were examined. Three different eluent mixtures were also examined. The results indicated that membranes with a 0.1-mm pore size were the most effective among those tested. In particular, hydrophobic polyvinylidene difluoride and hydrophobic polytetrafluoroethylene gave better recoveries and cleanup than other membranes. Results from GC chromatograms and matrix effects showed that membrane filtration afforded better cleanup than the modified QuEChERS method. Furthermore, over 90% of the 89 pesticides tested had acceptable recoveries using these two membranes, according to an acceptable recovery range of 70e120%. Therefore, this technique has potential as an effective cleanup method. © 2017 Elsevier Ltd. All rights reserved.

Keywords: Membrane filtration Pesticide analysis Effective cleanup method Marix effects Recovery rates

1. Introduction Currently, hundreds of pesticides are in widespread use in agricultural fields globally. Residues of these pesticides affect agricultural products, especially fruits and vegetables. Due to consumer awareness of potentially hazardous pesticide residues in agricultural products, international trade issues, regulatory requirements, and other factors, agricultural products are monitored for pesticide residues. To meet the demands of consumers, farmers, regulators, and others, analytical methods for pesticide residues in complex matrices are continually being improved. “Quick, easy, cheap, effective, rugged, and safe” (QuEChERS) methods have evolved from the original version into multi-laboratory validated methods using acetate buffering (AOAC Official Method 2007.01) or citrate buffering (CEN Standard Method EN 15662) (Koesukwiwat, Lehotay, & Leepipatpiboon, 2010). These and other versions of QuEChERS have been adopted worldwide because of their benefi cial features (Anastassiades, Lehotay, Stajnbaher, & Schenck, 2003; , & Lehotay, 2003; Nguyen, Yu, Lee, & Lee, Anastassiades, Mastovska ndez-Borges, Cabrera Cabrera, Rodríguez-Delgado, 2008; Herna

* Corresponding author. ** Corresponding author. E-mail addresses: [email protected] (J. Hong), [email protected]. ac.jp (A. Kawashima). http://dx.doi.org/10.1016/j.foodcont.2017.03.045 0956-7135/© 2017 Elsevier Ltd. All rights reserved.

ndez-Sua rez, & Gala n Saúco, 2009; Húskova , Matisova , Herna  , & Svorc, Hrouzkova 2009; Słowik-Borowiec, Szpyrka, & Walorczyk, 2015). In routine analytical applications, sample throughput is an important issue to consider when selecting an analytical method. Recently, multi-class and multi-residue pesticides analysis methods in fruits, vegetables, and other commodities have been commonly applied worldwide to the regulation of agricultural product safety, international trade, toxicological risk assessment, research investigations, and a lot of other purposes. Among them, the QuEChERS approach to pesticide analysis in agricultural products provides rapid sample preparation (high sample throughput) (Lehotay, Koesukwiwat, van der Kamp, Mol, & Leepipatpiboon, 2011). Despite the many advantages and demonstrated feasibility of QuEChERS, the major challenge encountered in the analysis of pesticide residues in agricultural products is the presence of pigments, lipids, and fatty acid compounds that might be co-extracted with the pesticides. These compounds can create massive GC interference and become a burden to the GC column and detector (Kwon, Lehotay, & Geis-Asteggiante, 2012). In addition, they can also co-elute with pesticides and cause inconsistent pesticide recoveries (Chai & Elie, 2013). Membrane filtration is an efficient method for the removal of interference materials (Acero, Benitez, Real, & Garcia, 2009), and is widely used in separation, cleanup, and concentration (Ahn et al., 1999; Wang & Chung, 2005; Yang, F. J., Yang, D. L., Zhang, & Jian, 1993). The removal of pesticides by membranes has been

J. Hong et al. / Food Control 79 (2017) 134e142

135

reported (Acero et al., 2009). Although studies have investigated the analysis of pesticides using membranes (Han, Sapozhnikova, & Lehotay, 2014; Hatkeyama et al., 2006; Hong et al., 2016; Plakas, Karabelas, Wintgens, & Melin, 2006), there are few reports of the removal of interference materials by membrane filtration during pesticide analysis using GC/MS. In this study, we investigated the application of membrane filtration as an enhanced cleanup method in pesticide analysis. In this experiment, the recovery and cleanup by membrane filtration in pesticide analysis were examined. We used 11 different types of membrane, classified by molecular weight cut-off (MWCO), pore size, and material. Spinach was used in this study as a representative agricultural product because it is a typical leafy vegetable with a highly pigmented matrix.

(Waltham, MA, USA) was used for pesticide measurement. Aliquots (2 mL) from 1 mL of recovery sample and 500 mL of blank sample were injected into the GC system respectively. The oven program started at 50  C (held for 1 min), which was ramped at 30  C min
1 to 125  C, and 5  C min
1 to 200  C. Finally, the temperature was ramped at 10  C min
1 to 300  C and it was held for 11.5 min. MS detection was performed in 70 eV of electron ionization mode. Calibration standards were prepared in acetone at 50 ng mL
1, 100 ng mL
1, 200 ng mL
1, and 400 ng mL
1 A 1.5 m, 0.25 mm guard column obtained from Sigma-Aldrich (St. Louis, MO, USA) was coupled the DB-5MS capillary column (30 m, 0.25 mm i.d., 0.25 mm) obtained from Agilent Technologies (Santa Clara, CA, USA).

2. Materials and methods

2.4. Method performance

2.1. Chemicals and reagents

Calibration curves were used to determine pesticide concentrations, and were constructed for each pesticide using four different concentrations of the pesticide standard solution in the range of 50e400 ng g
1. The coefficients of determination (R2) exceeded 0.99 in all pesticides, except for allethrin, isoxathionoxon, phenothrin, fenbuconazole, flumioxazin and tolfenpyrad, which were little low between 0.92 and 0.98. The limit of quantitation (LOQ) was calculated as ten times the standard deviation. The LOQ obtained from pesticides were in the range 3e30 ng g
1.

All pesticide standards were of high purity. Pesticide standard solution 31 (for GC analysis) including 85 representative of pesticides was obtained from Kanto Chemical (Tokyo, Japan). Pesticide standard solution 31, which is name of the product, contains pesticides among the product of pesticide standard solution for GC analysis. It also contains various properties of pesticides such as log Kow (from
0.22e6.01), hydrophilic, hydrophobic, low MW, high MW, carbamate, organophosphate, organochloride and so on. An internal standard solution (containing phenanthrene-d10, anthracene-d10 and 9-bromoanthracene) was obtained from Wako Pure Chemical Industries (Tokyo, Japan). They were stored at
20  C. The standards were diluted by acetone for experiment and stock standard solutions were stored at 4  C. All organic solvents were pesticide grade and obtained from Wako (Tokyo, Japan). InertSep GC/NH2 (500 mg/500 mg/6 mL) cartridges which composed of graphitized carbon black (GCB) and aminopropyl (NH2) sorbent were obtained from GL Sciences (Tokyo, Japan). Trisodium citrate dihydrate, disodium hydrogen citrate sesquihydrate, sodium chloride, and magnesium sulfate were obtained from Kanto Chemical (Tokyo, Japan). 2.2. Membranes Seven different polyethersulfone (PES) membranes were used in this study: 0.1-mm-pore size was obtained from AS ONE Corporation (Osaka, Japan); a pore size of 0.05 mm, and MWCOs of 150 kDa and 50 kDa were obtained from Daisen Membrane Systems (Tokyo, Japan); MWCOs of 10 kDa and 5 kDa were obtained from Koch Membrane Systems (Wilmington, USA); and an MWCO of 1 kDa was obtained from Daisen Membranes Systems (Tokyo, Japan). Five different types of membrane with a pore size of 0.1 mm were also used in this study: PES membrane was obtained from AS ONE Corporation (Osaka, Japan); two types of polyvinylidene difluoride (PVDF) membrane, one hydrophobic and one hydrophilic, were obtained from Merck Millipore (Osaka, Japan); and two types of polytetrafluoroethylene (PTFE), one hydrophobic and one hydrophilic, were obtained from Flon Industry (Tokyo, Japan). Total 11 membranes were evaluated (PES membrane was duplicated). 2.3. Instruments and GC/MS analytical conditions HP4750 Stirred Cell obtained from Sterlitech (Kent, WA, USA) was used for membrane-filtration apparatus and a SepPak elution pump obtained from Waters (Milford, MA, USA) was used for sample elution from solid column. For centrifugation, Kubota 5922 from Kubota (Osaka, Japan) was used. GC (TRACE GC Ultra) coupled with a MS (Polaris Q) obtained from Thermo Fisher Scientific

2.5. Cleanup by membrane filtration Pesticide-free spinach was selected in this study which obtained from a local market (Matsuyama, Ehime, Japan). We use modified QuEChERS method based on Lebotay method (Lehotay et al., 2010) for initial extraction. The procedure was as follows: (1) Weighed 10 g of each a chopped spinach sample in to a 50-mL polypropylene centrifuge tubes; (2) Added acetonitrile (MeCN, 10 mL) into the tubes and it was vigorously shaken for 1 min by hand after all tubes was sealed. Poured trisodium citrate dehydrate (1 g), disodium hydrogen citrate sesquihydrate (0.5 g), sodium chloride (1 g), and magnesium sulfate (4 g) to the tube and shaken vigorously by hand for 1 min after all tubes was sealed; (3) The mixtures were centrifuged at 3500 rpm (2330 rcf) for 10 min to separate the organic phase (MeCN) from the aqueous and solid phases; (4) Transferred 5 mL of each MeCN supernatant into three tubes and three types of mixed solvent were added to each tube, respectively; (5) For yield a total volume of 12.5 mL, three types of analyte solution was mixed with extract. The 5:5 watereMeCN (v/v) analyte solution was made by adding 1.25 mL of MeCN and 6.25 mL of water, the 6:4 watereMeCN (v/v) analyte solution was made by adding 7.5 mL of water and the 7:3 watereMeCN (v/v) analyte solution was made by following method: the supernatant (5 mL) was first evaporated to 2 mL, then add 1.75 mL of MeCN and 8.75 mL (6) Before membrane filtration, first of all, membrane was installed in the membrane-filtration apparatus. 20 mL of mixed solvent (5:5, 6:4, or 7:3 watereMeCN (v/v)) was added on the membrane by pipette for pre-washing (0.5 MPa, 30  C, and 400 rpm). Then, 10 mL of the extract was added on the membrane for filtration. Finally, 5 mL of mixed solvent was added for rinse, total volume of filtered solution was 15 mL. At this point, cleanup effect was determined visually by the color of the extract. 2.6. Recovery with spiked standard solutions using modified QuEChERS method Three different 12.5-mL crude standard samples (5:5, 6:4, and 7:3 watereMeCN (v/v)), containing 125 mL of pesticide standard solution, were used as pesticide-contaminated samples, while

136

J. Hong et al. / Food Control 79 (2017) 134e142

three different 12.5-mL crude reference samples (5:5, 6:4, and 7:3 watereMeCN (v/v)), containing 125 mL of acetone without pesticide, were used to calculate recovery rates and matrix effects. Following procedure is based on the previous work in Fig. 1 (Hong et al., 2016). Briefly, for membrane filtration, a 10-mL aliquot was used as the analyte solution. After membrane filtration (refer to section 2.5) and dehydration by Na2SO4, the analytes were concentrated to approximately 50 mL using an evaporator (100 hPa and 40  C), followed by cleanup. The extract which eluted from GCB/NH2 cartridge by adding 20 mL of mixed solvent were concentrated using an evaporator (100 hPa and 40  C), and a recovery sample (from the crude standard sample) and reference sample (from the crude reference sample) were made by adding acetone to a total volume of 1 mL. To determine recoveries, 50 mL of the syringe spike was added into the recovery sample to a final volume of 1 mL, and 25 mL of the syringe spike was added into the blank sample to a final volume of 500 mL to check for spinachrelated interference. To determine matrix effects, 25 mL of the syringe spike and 50 mL of the standard solution were added into the matrix sample to a final volume of 500 mL. Finally, pesticides were identified and quantified by GC/MS.

dehydration by Na2SO4, the analytes were concentrated to approximately 50 mL using an evaporator (100 hPa and 40  C), followed by cleanup with a GC/NH2 cartridge. The following method was the same as in section 2.6. Briefly, after the cleanup procedure, the analytes were measured by GC/MS.

3. Results and discussion 3.1. Preliminary study to find optimal parameters for membrane filtration

2.7. Pesticide analysis in spinach by membrane filtration and modified QuEChERS method

Our previous study indicated that recovery increased as the MWCO increased and that cleanup improved when the eluent contained water (Hong et al., 2016). GK membrane with a 3 kDa MWCO using 1:1 watereMeCN (v/v) gave the best results, with 3 K being the highest MWCO in the previous study. Therefore, in this study, we chose membranes that were over 1 kDa and three different water-containing eluents (5:5, 6:4, and 7:3 watereMeCN (v/v)). Pesticide recovery was determined by calculating the recovery rate (%) of the spiked pesticide standard and the cleanup achieved by membrane filtration was determined by visual observation of the spinach extracts. Pesticide-free spinach was selected as the representative agricultural product because it is a typical leafy vegetable with a highly pigmented matrix.

Ten grams of spinach samples were prepared. One was used as crude standard sample which was fortified with pesticide standard mixture, and the other as a crude reference sample which was pesticide-free, to calculate recovery rates and matrix effects. Modified QuEChERS method was performed as detailed in section 2.5, but with 125 mL of pesticide standard solution mixed with 10 g of spinach. The supernatants (5 mL) were transferred into two tubes for preparing a recovery sample and a reference sample. The analyte solution was made by adding a suitable ratio of water to MeCN to a final volume of 12.5 mL (5:5, 6:4, and 7:3 watereMeCN (v/v)). After membrane filtration (refer to section 2.5) and

3.1.1. Effect of watereMeCN ratio in the eluent We used a hydrophobic PTFE membrane (pore size, 0.1 mm) in this experiment. The experimental procedure was as described in sections 2.5 and 2.6. Fig. 2 shows that the percentages of pesticides recovered efficiently from 89 pesticides (acceptable recoveries ranged from 70 to 120%) by membrane filtration using three different eluents, namely watereMeCN mixtures in ratios of 5:5, 6:4, and 7:3 (v/v). The recovery (%) of spiked pesticides is defined by following equation:

Fig. 1. Schematic diagram of the modified QuEChERS method.

J. Hong et al. / Food Control 79 (2017) 134e142

Fig. 2. Percentage of 89 pesticides with 70e120% recoveries in the hydrophobic PTFE membrane according to the watereMeCN ratio (v/v) in the eluent. (Red bars represent standard deviations, n ¼ 3). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

137

factor of recovery. The percentages of pesticides recovered efficiently were 79% for 0.1 mm and 0.05 mm membranes, 83% for the 150 kDa membrane, and <70% for the other membranes. Therefore, these three membranes were used in the subsequent cleanup test. Cleanup was confirmed from the extract colors by visual observation after membrane filtration. The results showed that the color of each extract from the three different membranes was not substantially different (data not shown). In addition, considering the pressure and permeation velocity during membrane filtration, a larger pore size or higher MWCO membrane was more favorable. Volumetric flow rate was decreased when a size of membrane was too small and it caused by clogging of extracts during membrane filtration. But, clogging was remarkably decreased as size of membrane was increased. As membrane filtration was completed at a relatively low pressure (0.05 MPa) and fast volumetric flow rate (1.25 mL/s) using the 0.1 mm membrane compared with the others, the 0.1-mm membrane was determined to be the best for membrane filtration.

Recovery (%) ¼ (peak area of analytes in recovery sample / peak area of analytes in matrix sample)  100 (1)

3.2. Recovery and cleanup with five different membranes using the optimal parameters

The percentage of efficiently recovered pesticides was 96% for 5:5 watereMeCN, 98% for 6:4 watereMeCN, and 85% for 7:3 watereMeCN (all v/v). Pesticide recoveries using 5:5 and 6:4 watereMeCN (v/v) mixtures were higher than for 7:3 watereMeCN (v/v), with the highest recovery achieved using 6:4 watereMeCN (v/ v). These results were represented by the extract colors, which became lighter as the watereMeCN ratio increased (data not shown). However, despite 7:3 watereMeCN (v/v) producing a slightly lighter extract color compared with other solvents, the recovery was the lowest. This was due to the hydrophobic effect, which is the tendency of nonpolar substances to aggregate in aqueous solution and exclude water molecules. From these results, we chose 6:4 watereMeCN (v/v) as the best eluent.

Using the optimal parameters from the preliminary study, we investigated five different 0.1 mm membranes (PES, hydrophilic and hydrophobic PVDF, and hydrophilic and hydrophobic PTFE membranes) using 6:4 watereMeCN (v/v). The experimental procedure was the same as described in section 2.7. The recoveries are shown in Fig. 4. Over 80% of the pesticides from the 89 total pesticides were recovered by all membranes, except for PES. The PES membrane experienced shrinkage in the presence of the watercontaining solvent and resisted swelling, which might have caused fouling effects that can lead to a lower recovery. In contrast, about 90% of the pesticides were recovered using hydrophobic PVDF and hydrophobic PTFE membranes, with the latter showing the best recovery. For the PVDF and PTFE membrane filtrations, the hydrophobic membranes gave higher recoveries than the hydrophilic membranes, particularly for hydrophobic pesticides. Fig. 5 shows pesticide recoveries according to the octanol/water partition coefficient by 0.1-mm membrane filtration, and clearly demonstrates that hydrophobic pesticides were well-recovered through hydrophobic membranes. In particular, PTFE gave a higher recovery than other membranes, perhaps due to its solubility parameter. Fig. 6 shows the percentage of 89 pesticides above ±20% matrix effect in the membranes, of which hydrophobic PVDF had the lowest percentage. Hydrophobic PVDF and hydrophobic PTFE membrane filtration were less matrix-effected than hydrophilic PVDF and hydrophilic PTFE membrane filtration. In addition, extract color was slightly lighter in hydrophobic membranes compared in hydrophilic membranes (data not shown). This might be due to the hydrophobic effect. In addition, surface interactions, such as van der Waals' forces during mass transfer, might have removed interferences (Chen, Taylor, Mulford, & Norris, 2004). In the case of PES, shrinking during membrane filtration might have been the primary reason low matrix-effect. The results showed that the hydrophobic PVDF and hydrophobic PTFE filtration methods were best in terms of recovery and cleanup. Fig. 7 shows matrix effects of pesticides according to retention time (min) by three different membranes. Matrix effects relatively increased after 15e22 min and 25e30 min of retention time of GC analysis. Fatty acid derivatives (RT ¼ 15.8 min), phytol (RT ¼ 19.63 min), tocopherol (RT ¼ 28.14 min) and phylloquinone (RT ¼ 28.66 min) are shown in Fig. 8. These materials cause matrix effects during GC analysis. Thus, matrix effects could be increased unless these interferences were removed. But, matrix effects were decreased by hydrophobic PVDF filtration method because interferences were removed or decreased by this method.

3.1.2. Effect of membrane pore size and MWCO We investigated the effect of pore size and MWCO on recovery and cleanup. Seven different types of PES membranes, classified by their pore size and MWCO, were used with the 6:4 watereMeCN (v/ v) eluent. The membranes had pore sizes of 0.1 and 0.05 mm and MWCOs of 150, 50,10, 5 and 1 kDa. The recovery results, showing the percentage of pesticides recovered efficiently from 89 pesticides by membrane filtration (acceptable recovery range from 70 to 120%), are presented in Fig. 3. The results indicated that the recovery decreased when the MWCO was below 150 kDa. Moreover, the recovery decreased remarkably when the MWCO was below 5 kDa. These results demonstrated that size exclusion can be a dominant

Fig. 3. Percentage of 89 pesticides with 70e120% recoveries according to the MWCO and pore size of PES membranes with 6:4 watereMeCN (v/v). (Red bars represent standard deviations, n ¼ 3). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

138

J. Hong et al. / Food Control 79 (2017) 134e142

Fig. 4. Percentage of 89 pesticides with 70e120% recoveries according to membrane material with 6:4 watereMeCN (v/v). (Red bars represent standard deviations, n ¼ 3). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 5. Recovery of pesticides according to octanol/water partition coefficient by 0.1 mm membrane filtration.

Fig. 6. Percentage of 89 pesticides above ±20% matrix effect according to membrane materials with 6:4 watereMeCN (v/v). (Red bars represent standard deviations, n ¼ 3). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

J. Hong et al. / Food Control 79 (2017) 134e142

139

Fig. 7. Matrix effects of pesticides according to retention time (min) of GC analysis by three different membranes with 6:4 watereMeCN (v/v).

Fig. 8. GC/MS chromatograms for blank spinach samples using membrane filtration and modified QuEChERS methods.

3.3. Comparison and reproducibility study between membrane filtration and the modified QuEChERS method The experimental procedure used is described in section 2.7. From the above results, we used hydrophobic PVDF and hydrophobic PTFE membranes to examine the reproducibility of the membrane filtration method, based on the RSD (relative standard deviation) and comparisons of cleanup based on chromatograms. In addition, we also demonstrated the potential of membrane filtration for use as an efficient method by comparison with the modified QuEChERS method. Spinach was used in the test as a representative agricultural product. 3.3.1. Recovery Table 1 shows the recovery rates (%) and RSDs (%) of pesticides using the modified QuEChERS and membrane filtration methods (n ¼ 3). Recoveries of all pesticides by the modified QuEChERS

method were in range 84e117%, with except for 2 no detection (N.D.). RSD values (n ¼ 3) were below 20% for all pesticides. The percentage of 89 pesticides with 70e120% recoveries was 98% using the modified QuEChERS method. The recoveries of most pesticides by the hydrophobic PVDF membrane filtration method were in the range 71e97%, except for 2 N.D. In addition, seven pesticides had recoveries below 70%, namely tecnazene, benfluralin, quintozene, fenpropimorph, triadimefon, fthalide, and buprofezin. RSD values (n ¼ 3) were below 20% for all but six pesticides. The percentage of pesticides recovered efficiently was 90% for the hydrophobic PVDF filtration method. The recoveries of all pesticides using the hydrophobic PTFE filtration method were in the range 73e108%, except for 2 N.D. RSD values (n ¼ 3) were below 20% for all but four pesticides. The percentage of pesticides recovered efficiently was 98% for the hydrophobic PTFE filtration method. Using the hydrophobic PVDF filtration method, the recovery was slightly lower than that of other methods. This might have been due to adsorption

140

J. Hong et al. / Food Control 79 (2017) 134e142

Table 1 Recoveries and matrix effects of pesticides (200 ppb) on spinach by modified QuEChERS and membrane filtration methods (n ¼ 3) with 6:4 watereMeCN (v/v). Pesticides

XMC Tecnazene Propachlor Propoxur Benfluralin Monocrotophos Dicloran Carbofuran Quintozene Clomazone Simazine Atrazine Propyzamide Cyanophos Triallate Isazofos Iprobenfos Benoxacor Bromobutide Acetochlor Phosphamidon Chlorpyrifos-methyl Vinclozolin Metalaxyl Ametryn Prometryn Propanil Fenpropimorph Ethofumesate Chlorthal-dimethyl Triadimefon Quinoclamine Nitrothal-isopropyl Bromacil Bromophos-methyl Diphenamid Fthalide Allethrin-1 Allethrin-2 Allethrin-3 Allethrin-4 Dimethametryn Dimepiperate Methidathion Tetrachlorvinphos Alpha-Endosulfan(I) Fenothiocarb Napropamide Fenamiphos Flutriafol Profenofos DEF Oxadiazon E-Metominostrobin Isoprothiolane Isoxathion_oxon Buprofezin Flamprop-methyl Bupirimate Oxyfluofen Z-Metominostrobin Azaconazole Isoxathion Imazamethabenzmethylester Ethion Beta-Endosulfan(II) Fluacrypyrim Oxadixyl Benalaxyl Carfentrazone-ethyl Trifloxystrobin

Modified QuEChERS method

Hydrophobic PVDF method

Hydrophobic PTFE method

Recovery (%)

RSD (%)

Matrix effect (%)

RSD (%)

Recovery (%)

RSD (%)

Matrix effect (%)

RSD (%)

Recovery (%)

RSD (%)

Matrix effect (%)

RSD (%)

102 98 102 102 98 95 98 104 89 101 104 98 102 104 99 106 100 103 101 100 99 101 102 105 100 99 99 95 103 102 97 96 99 105 99 101 96 N.D N.D 99 114 99 101 105 93 110 101 98 97 102 95 93 101 99 100 89 100 101 96 97 97 96 96 94

7 18 9 7 6 13 7 12 9 8 7 11 10 10 7 3 9 5 6 7 9 8 9 7 12 11 8 6 7 9 4 10 7 14 8 10 13 N.D N.D 2 11 8 4 10 17 17 8 3 7 13 2 5 3 3 1 10 5 3 6 7 6 4 3 11

44 12 26 26 18 54 53 32 12 24 24 16 20 25 1 17 22 6 2 3 18
7
6 21 23 10 25 13 2
9 11 24 13 19
2 4 5 N.D N.D 19 3 17
14
19 27
25
19 2 8 10 7
2
18
2
4 0
15
13
11
13
6
12
15 56

17 12 16 13 19 18 20 15 18 15 13 14 15 15 10 10 15 14 13 14 15 13 13 10 12 13 12 10 9 11 10 16 12 9 12 11 11 N.D N.D 14 15 13 18 22 21 18 34 15 22 7 13 17 13 15 13 11 15 13 13 15 15 12 10 6

90 57 88 87 67 89 81 88 62 87 85 87 89 88 71 96 84 85 82 85 83 80 85 89 87 84 88 39 92 84 51 85 84 84 76 86 68 N.D N.D 92 71 84 85 96 95 80 82 87 82 85 82 75 83 89 88 81 23 88 79 76 89 89 83 91

3 2 1 4 9 1 5 4 4 1 2 2 0 2 4 8 5 6 8 3 6 3 3 5 5 4 2 32 5 7 35 4 4 6 3 5 7 N.D N.D 8 11 5 0 5 5 5 9 0 4 4 3 9 2 2 2 5 25 4 2 14 1 3 1 37

32 10 16 17 7 32 39 27 9 22 32 23 24 23 6 15 8 5
2
1 6
5
3 28 25 13 23 4 5
5 6 19 7 18
1 6 9 N.D N.D 8 5 7
15
16 6
11
16 1
13 2
4
3
9
6
7 11
13
3
10
15
6
8
10 75

12 17 3 8 3 26 5 4 6 3 3 6 2 1 5 8 2 1 5 5 4 3 3 6 4 5 8 4 5 6 2 5 6 4 4 3 6 N.D N.D 4 2 2 15 148 26 5 34 11 19 11 13 8 3 10 8 5 4 12 6 19 8 11 8 76

99 86 96 98 85 96 99 99 73 97 97 97 100 101 82 93 95 96 95 95 96 90 91 97 97 93 96 82 97 92 93 92 93 90 85 94 88 N.D N.D 97 93 95 97 95 101 85 84 97 93 99 96 80 94 98 99 97 96 96 96 87 99 94 95 92

7 31 11 11 13 10 7 7 12 6 5 5 6 9 12 9 8 11 10 13 12 8 7 7 10 7 7 10 10 7 22 10 10 7 9 7 14 N.D N.D 10 13 12 10 2 12 9 17 7 10 9 9 9 6 8 9 6 8 8 5 11 9 7 4 10

80 7 24 33 4 83 61 24 2 24 36 15 20 40 0 5 10 5
5 3 14
9
8 21 23 8 52 3
3
13 4 44 2 14
11 3 5 N.D N.D 2 4 5
12
7 16
16
20 10 0 14 5
3
14
6
3 34
10
2
8
15
4
5
14 55

22 3 3 5 10 11 9 4 7 1 3 5 8 7 10 11 12 6 11 8 11 11 10 7 5 9 9 10 13 14 11 5 12 13 12 9 7 N.D N.D 7 8 12 8 61 7 7 18 2 14 4 5 8 5 12 6 10 5 4 3 20 6 5 7 23

99 94 101 97 98 97 101

3 7 2 5 16 5 5


7
28
1
10
6
8
10

16 4 17 13 19 15 17

80 87 85 85 83 84 87

4 8 4 4 15 3 6


12
18
8
11
5
8
9

8 6 8 13 18 17 8

93 100 98 98 109 98 96

10 11 7 7 8 11 10


11
19
3
5
23
5
2

10 6 12 5 19 9 8

J. Hong et al. / Food Control 79 (2017) 134e142

141

Table 1 (continued ) Pesticides

Quinoxyfen Propargite Norflurazon Diclofop_methyl Hexazinone Pyridaphenthion Bromopropylate Piperophos Methoxychlor Phosmet Phenothrin-1 Phenothrin-2 Tetradifon Pyrazophos Fenbuconazole Flumioxazin Flumiclorac-pentyl Tolfenpyrad

Modified QuEChERS method

Hydrophobic PVDF method

Hydrophobic PTFE method

Recovery (%)

RSD (%)

Matrix effect (%)

RSD (%)

Recovery (%)

RSD (%)

Matrix effect (%)

RSD (%)

Recovery (%)

RSD (%)

Matrix effect (%)

RSD (%)

90 97 103 96 100 98 98 97 100 96 92 93 93 92 97 91 84 93

6 3 1 5 6 6 8 3 9 5 18 10 9 11 11 14 10 7


20
6 2
2 8
7
6 10
17
5 20 1
17 23 33 30 36 40

11 15 18 10 8 20 17 18 20 16 28 20 8 21 11 17 19 17

73 84 89 86 87 82 83 90 71 87 81 78 71 84 88 91 97 85

2 4 3 2 5 8 5 3 9 3 30 38 11 5 4 6 13 5


11
16 2
8 5
14
13
7
6
7
11
12
2 8 42 9
5 6

7 12 28 11 13 23 17 17 6 20 38 28 3 19 15 13 21 26

79 88 95 93 96 99 92 97 93 95 85 80 95 100 96 102 102 101

6 9 9 12 15 12 10 11 6 10 13 23 20 7 9 13 20 16


11
4 39 1 31 7
6 9
13 11
3
22
4 44 87 33 24 64

3 13 4 6 10 14 9 14 5 4 25 36 9 5 14 5 5 19

of pesticides onto the membrane by means of hydrophobic effects, solubility parameters, and more (Schmidt & Lutze, 2013; Spillman, 1995). However, results showed that over 90% of pesticides were efficiently recovered using all three methods. The results showed that these methods have higher recovery than previous work which represented that 80% of pesticides were recovered (Hong et al., 2016). 3.3.2. Matrix effects The conventional methods to estimate the effect of interferences are analysis of matrix effects and chromatograms. Matrix effects are influence on an analytical measurement caused by all other components of the sample being measured as the analyte. To evaluate matrix effects, a standard 200 mg kg
1 solution of was prepared. The matrix effects (%) were defined by the following equation. Matrix effects (%) ¼ ((peak area of analytes in matrix sample / peak area of analytes in standard solution)
1)  100 (2) For each compound, values between ±20% were considered that they did not show matrix effect, whereas values higher than þ20% indicated signal enhancement and values lower than
20% indicated signal suppression. Table 1 shows the matrix effects (%) and RSDs (%) of pesticides using the modified QuEChERS method and membrane filtration method (n ¼ 3). Interferences are known to be problematic in pesticide analysis using GC/MS, with matrixinduced signal enhancement occurring in GC when active surfaces in the system cause retention and/or degradation of analytes (Erney, Gillespie, & Gilvydis, 1993). Thus, cleanup is very important in pesticide analysis. Matrix effects for most pesticides in the modified QuEChERS method were obtained in the range within ±20%, with the exception of 23 pesticides. RSD values (n ¼ 3) were below 20% for all but seven pesticides. The percentage of matrixeffected pesticides (matrix effect were above ±20%) was 28% for the modified QuEChERS method. Matrix effects for most pesticides in the hydrophobic PVDF filtration method were obtained in the range within ±20%, with the exception of 14 pesticides. RSD values (n ¼ 3) were below 20% for all but 11 pesticides. The percentage of matrix-effected pesticides was 18% for the hydrophobic PVDF filtration method. Matrix effects for most pesticides in the hydrophobic PTFE filtration method were obtained in the range within ±20%, with the exception of 23 pesticides. RSD values (n ¼ 3) were

below 20% for all but six pesticides. The percentage of matrixeffected pesticides was 28% for the hydrophobic PTFE filtration method. These results indicated that the number of matrix-effected pesticides was lower using hydrophobic PVDF than the modified QuEChERS method, while that of the hydrophobic PTFE filtration method was the same as the modified QuEChERS method. 3.3.3. Chromatogram interferences The second method to estimate the effect of interference materials was to compare chromatograms. Fig. 8 shows GC/MS total ion chromatograms of blank spinach samples after analysis by the modified QuEChERS and membrane filtration methods. Although, some primary peaks from the spinach sample appeared in both chromatograms, fewer and less intense interfered peaks were shown by the membrane filtration methods. Furthermore, coextractives such as phytol (RT ¼ 19.63 min), tocopherol (RT ¼ 28.14 min), and phylloquinone (RT ¼ 28.66 min) were removed by the membrane filtration methods. It could be help to reduce a burden to the GC column and detector. The matrix effects and chromatograms showed that hydrophobic PVDF filtration methods were more effective at removing interference materials than the modified QuEChERS method. In addition, this results showed interference materials were remarkably removed compared with previous work (Hong et al., 2016). 4. Conclusions This study showed novel cleanup method for pesticide analysis in spinach. The key of this method was cleanup by UF membrane filtration. 3 types of solvent and 11 different types of membranes, classified by their MWCO, pore size, and material were investigated for finding efficient cleanup by UF membrane filtration. Primary interfered peaks were almost removed by this method. The results showed cleanup ability is higher our method than the modified QuEChERS method. With regard to matrix effect, our method (especially when hydrophobic PVDF was used) showed prominently reduced matrix effect compared with the modified QuEChERS method with no prominently different recoveries. This study shows efficient cleanup method by membrane filtration for pesticides analysis. Although, it has potential applications for pesticides analysis in various matrices, we have to improve this method for the determination of pesticide residues in routine analysis.

142

J. Hong et al. / Food Control 79 (2017) 134e142

Therefore, in the future research, the cleanup method needs to be extended to a different range of pesticides and more test matrices such as grains, fruits, other vegetables and so on. Acknowledgement We are grateful anonymous reviewers for their valuable comments that greatly improve this manuscript. References Acero, J. L., Benitez, F. J., Real, F. J., & Garcia, C. (2009). Removal of phenyl-urea herbicides in natural waters by UF membranes: Permeate flux, analysis of resistances and rejection coefficients. Separation and Purification Technology, 65, 322e330. Ahn, K. H., Song, K.-S., Cha, H. Y., & Yeom, I. T. (1999). Removal of ions in nickel electroplating rinse water using low-pressure nanofiltration. Desalination, 122, 77e84.  Anastassiades, M., Lehotay, S. J., Stajnbaher, D., & Schenck, F. J. (2003). Fast and easy multiresidue method employing acetonitrile extraction/partitioning and ‘‘dispersive solid-phase extraction’’ for the determination of pesticide residues in produce. Journal of AOAC International, 86, 412e431. , K., & Lehotay, S. J. (2003). Evaluation of analyte Anastassiades, M., Mastovska protectants to improve gas chromatographic analysis of pesticides. Journal of Chromatography A, 1015, 163e184. Chai, L. K., & Elie, F. (2013). A rapid multi-residue method for pesticide residues determination in white and black pepper (Piper nigrum L.). Food Control, 32, 322e326. Chen, S. S., Taylor, J. S., Mulford, L. A., & Norris, C. D. (2004). Influence of molecular weight, molecular size, flux, and recovery for aromatic pesticide removal by nanofiltration membranes. Desalination, 160, 103e111. Erney, D. R., Gillespie, A. M., & Gilvydis, D. M. (1993). Explanation of the matrixinduced chromatographic response enhancement of organophosphorus pesticides during open tubular column gas chromatography with splitless or hot oncolumn injection and flame photometric detection. Journal of Chromatography A, 638, 57e63. Han, L., Sapozhnikova, Y., & Lehotay, S. J. (2014). Streamlined sample cleanup using combined dispersive solid-phase extraction and in-vial filtration for analysis of pesticides and environmental pollutants in shrimp. Analytica Chimica Acta, 827, 40e46. Hatkeyama, E., Kajita, H., Sugawara, T., Sasaki, A., Takahashi, S., et al. (2006). Simultaneous determination of pesticides in agricultural products by LC/MS/MS using clean-up with ultrafiltration. Food Hygiene and Safety Science, 47, 137e145.

ndez-Borges, J., Cabrera Cabrera, J., Rodríguez-Delgado, M. A., Herna ndezHerna n Saúco, V. (2009). Analysis of pesticide residues in bananas Su arez, E. M., & Gala harvested in the Canary Islands (Spain). Food Chemistry, 113, 313e319. Hong, J., Kawashima, A., Okamoto, M., Kanetsuki, K., Makino, T., & Hamada, N. (2016). Fundamental study of a novel membrane filtration cleanup method for pesticide analysis in agricultural products. Food Control, 64, 1e9.  , R., Matisova , E., Hrouzkov Húskova a, S., & Svorc, L. (2009). Analysis of pesticide residues by fast GC in combination with negative chemical ionization mass spectrometry. Journal of Chromatography A, 1216, 6326e6334. Koesukwiwat, U., Lehotay, S. J., & Leepipatpiboon, N. (2010). High throughput analysis of 150 pesticides in fruits and vegetables using QuEChERS and lowpressure gas chromatographyetime-of-flight mass spectrometry. Journal of Chromatography A, 1217, 6692e6703. Kwon, H., Lehotay, S. J., & Geis-Asteggiante, L. (2012). Variability of matrix effects in liquid and gas chromatographyemass spectrometry analysis of pesticide residues after QuEChERS sample preparation of different food crops. Journal of Chromatography A, 1270, 235e245. Lehotay, S. J., Koesukwiwat, U., van der Kamp, H., Mol, H. G., & Leepipatpiboon, N. (2011). Qualitative aspects in the analysis of pesticide residues in fruits and vegetables using fast, low-pressure gas chromatography time-of-flight mass spectrometry. Journal of Agricultural and Food Chemistry, 59, 7544e7556. Lehotay, S. J., Son, K. A., Kwon, H., Koesukwiwat, U., Fu, W., Mastovska, K., et al. (2010). Comparison of QuEChERS sample preparation methods for the analysis of pesticide residues in fruits and vegetables. Journal of Chromatography A, 1217, 2548e2560. Nguyen, T. D., Yu, J. E., Lee, D. M., & Lee, G. H. (2008). Multiresidue method for the determination of 107 pesticides in cabbage and radish using QuEChERS sample preparation method and gas chromatography mass spectrometry. Food Chemistry, 110, 207e213. Plakas, K. V., Karabelas, A. J., Wintgens, T., & Melin, T. (2006). A study of selected herbicides retention by nanofiltration membranesdthe role of organic fouling. Journal of Membrane Science, 284, 291e300. Schmidt, P., & Lutze, P. (2013). Characterisation of organic solvent nanofiltration membranes in multi-component mixtures: Phenomena-based modelling and membrane modelling maps. Journal of Membrane Science, 445, 183e199. Spillman, R. (1995). Economics of gas separation membrane processes. Membrane Separation Technology, 2, 589e668. Słowik-Borowiec, M., Szpyrka, E., & Walorczyk, S. (2015). Gas chromatographic determination of pesticide residues in white mustard. Food Chemistry, 173, 997e1005. Wang, K. Y., & Chung, T. S. (2005). The characterization of flat composite nanofiltration membranes and their applications in the separation of cephalexin. Journal of Membrane Science, 247, 37e50. Yang, F. J., Yang, D. L., Zhang, S. H., & Jian, X. G. (1993). Synthesis of polypiperazineamide thin-film membrane on PPESK hollow fiber UF membrane. Chinese Chemical Letters, 18, 966e968.