Investigation of matrix effects by crops and their compensation methods in a multiresidue analysis using LC-MS/MS

Abstract

This study evaluated the matrix effect (ME) in a multiresidue analysis with the modified official Japanese method for agricultural products using liquid chromatography-tandem mass spectrometry (LC-MS/MS). In ME comparisons between the vegetable and fruit samples, it was found that more numerous analytes exhibited ion suppression in the vegetable samples than in the fruit samples, and substantial ion enhancement was not observed in most of the analyte–sample combinations. The ME could significantly vary, even within the same commodity, and it was suggested that sampling has greater influence than measurement when there is a wide ME variability. Dilution, the internal standard calibration method and the matrix-matched calibration method are practical countermeasures against MEs, but certain limitations in their applications should be considered. Moreover, a novel cleanup procedure suitable for hydrophilic neonicotinoid pesticides that minimized the usage of expensive internal standard solutions was suggested.

Introduction

Pesticide plays a crucial role in stable crop production, which improves production efficiency, and reduces labor requirements in agriculture. Pesticide residue tolerance are set for each pesticide–commodity combination. Agricultural products containing pesticide residues that exceed the maximum residue limit (MRL) or uniform limit are prohibited from sale, shipment, and import. When pesticides are used in accordance with approved direction of order, the agricultural product rarely contains pesticide residues exceeding the established limits. However, unintentional contamination of pesticides may occur through processes such as spray drift or the persistence of pesticide residues in the soil.^1,2) Therefore, the analysis of pesticide residues in agricultural products prior to shipment is important to ensure food safety.

Multiresidue analysis is an appropriate and efficient analytical procedure to precisely understand the pesticide exposure on agricultural products as this simultaneously targets several analytes in a single analysis. The Ministry of Health, Labor, and Welfare (MHLW) has informed the multiresidue analysis, which combines a cleanup procedure using liquid-liquid partitioning and solid-phase extraction and LC-MS(/MS) analysis (hereafter referred to as the official Japanese method).³⁾ Although this cleanup procedure has been validated for several target analytes in representative food samples, matrix removal can be inefficient depending on the specific pesticide–commodity combination. In this case, a problem called the matrix effect (ME) could compromise the results of the analysis because sample-derived matrices could enhance and suppress the ionization efficiency of the target analytes in the analyzers, resulting in changes in the peak shape and area of the target analytes.^4,5) In LC-MS/MS analysis, it has been reported that the matrices change the surface tension of the droplet, inhibit or accelerate the migration of analytes in the droplet, or compete with the analytes during ionization.^6,7) However, it has not been confirmed which pesticide–commodity combination is more susceptible to ME problems. Sulyok et al.⁸⁾ suggested that legumes and oilseeds, which contain high fat, starch, and/or protein might be the reason for the low recovery of analytes in legumes, oilseeds, and forage crops in comparison to grains due to ion suppression. Other studies have shown that bioflavonoids in citrus fruits may induce MEs.^9,10) However, much remains unknown about ME patterns in relation to most of the pesticide–commodity combinations in the multiresidue analysis using the official Japanese method. Some methods for calculating the ME have been used and reported in previous studies.¹¹⁾ One of the major methods is to calculate the ME using the peak area of the analyte in reagent-only and matrix-matched solutions at a specific concentration.^12,13) Another is to calculate the ME using the slope of the calibration curve of the analyte in the two solutions.^14,15) According to the criteria explained by Mol et al.,¹⁶⁾ substantial signal suppression or enhancement (ME<−20% and ME>20%, respectively) interfered with accurate determination via an LC-MS/MS analysis. However, ME measurements are currently not specified and are often non-unified among the guidelines or previous reports.¹⁷⁾ When the ME is not negligible in a pesticide residue analysis, reduction and/or countermeasures against ME should be discussed to satisfy the analytical criteria. Unfortunately, reduction or countermeasures against MEs are not always specified in the guidelines and are not unified among the guidelines.¹⁷⁾ Dilution is been considered a simple and effective way to reduce or eliminate MEs.^13,18) The internal and matrix-matched calibration methods have been proposed as countermeasures for ME.¹¹⁾ While these approaches are effective in compensating for ME, several challenges remain. For instance, the use of stable isotope-labeled (SIL) internal standards (IS) in internal calibration is associated with high analytical costs, and the applicability of blank samples in matrix-matched calibration requires further evaluation. Thus, it is important to improve these issues for the establishment and dissemination of a precise and efficient simultaneous analysis of pesticide residues in agricultural products.

This review presents selected findings on ME and their compensation methods in multiresidue analysis, organized into three chapters. In Chapter 1, the ME on the target analytes in the vegetable and fruit samples using the modified official Japanese method was preliminary determined. In Chapter 2, the sampling and measurement variance of the ME on the target analytes was evaluated. The relationship between ME and the physicochemical properties of the target analytes was analyzed. In Chapter 3, a novel analytical method for seven neonicotinoid pesticides in the vegetable samples was proposed. The analytical cost of using the internal standard calibration method was reduced by minimizing the SIL-IS waste.

Chapter 1: ME between vegetable and fruit samples in LC-MS/MS analysis¹⁹⁾

Materials and methods

Chemicals and reagents: 97 analytes of various octanol-water partition coefficients (log P_OW) were selected as the target analytes from the standard solutions (Table 1). The log P_OW was estimated by KOWWIN (v1.69) using EPI Suite (v 4.11) developed by the United States Environmental Protection Agency and Syracuse Research Corporation. Of the 97 target analytes, 82 pesticides were selected from the four standard solutions (WQ series) purchased from Fujifilm Wako Pure Chemical Corporation (Wako; Osaka, Japan), and 15 ISs were selected from the three standard solutions (PL series) purchased from Hayashi Pure Chemical Industry (Osaka, Japan) (Table 1). Pesticide analysis-grade acetonitrile, methanol, and toluene purchased from Wako were used for the analysis.

Table 1. Target analytes in Chapter 1.

No.	Analyte	Est. log Pow	STD	No.	Analyte	Est. log Pow	STD	No.	Analyte	Est. log Pow	STD
1-01	Nitenpyram	−0.66	WQ4	1-34	Furametpyr	2.36	WQ4, WQ5	1-67	Tebuconazole	3.7	WQ4, WQ5
1-02	Nitenpyram-d3	−0.66	PL-VII	1-35	Thiabendazole-13C6	2.47	PL-II	1-68	Isoxathion	3.73	WQ1-2
1-03	Dinotefuran	−0.64	WQ4	1-36	Azoxystrobin	2.5	WQ3	1-69	Pyraclofos	3.77	WQ5
1-04	Dinotefuran-d3	−0.64	PL-VII	1-37	Atrazine	2.61	WQ1-2	1-70	Diazinon	3.81	WQ1-2
1-05	Pymetrozine	−0.18	WQ4	1-38	Terbucarb	2.74	WQ1-2	1-71	Anilofos	3.81	WQ1-2
1-06	Thiamethoxam	−0.13	WQ4, WQ5	1-39	Fenobucarb	2.78	WQ1-2	1-72	Imazalil-d5	3.82	PL-II
1-07	Thiamethoxam-d3	−0.13	PL-VII	1-40	Simetryn	2.8	WQ1-2	1-73	Flusulfamide	3.83	WQ4
1-08	Imidacloprid	0.57	WQ4	1-41	Pyriminobac-methyl	2.84	WQ4, WQ5	1-74	Dimethametryn	3.9	WQ1-2
1-09	Imidacloprid-d4	0.57	PL-VII	1-42	Isoprothiolane	2.88	WQ1-2	1-75	Cadusafos	3.9	WQ5
1-10	Methomyl	0.6	WQ3	1-43	Cyproconazole	2.9	WQ5	1-76	Fipronil	4	WQ3
1-11	Clothianidin	0.7	WQ4	1-44	Boscalid	2.96	WQ4, WQ5	1-77	Cyprodinil	4	WQ4, WQ5
1-12	Clothianidin-d3	0.7	PL-VII	1-45	Ametryn	2.98	WQ4, WQ5	1-78	Dimepiperate	4.02	WQ1-2
1-13	Dimethoate	0.78	WQ1-2	1-46	Propanil	3.07	WQ4, WQ5	1-79	Piperophos	4.04	WQ1-2
1-14	Dimethoate-d6	0.78	PL-V	1-47	Simeconazole	3.1	WQ4, WQ5	1-80	Pretilachlor	4.08	WQ1-2
1-15	Acetamiprid	0.80	WQ4, WQ5	1-48	Metolachlor	3.13	WQ5	1-81	Thifluzamide	4.10	WQ4, WQ5
1-16	Acetamiprid-d3	0.8	PL-VII	1-49	Metolachlor-d6	3.13	PL-II	1-82	Isofenphos	4.12	WQ1-2
1-17	Thiacloprid	1.26	WQ4, WQ5	1-50	Pyridaphenthion	3.2	WQ1-2	1-83	Pirimiphos-methyl	4.2	WQ4, WQ5
1-18	Thiacloprid-d4	1.26	PL-VII	1-51	Cafenstrole	3.21	WQ1-2	1-84	Tebufenozide	4.25	WQ4
1-19	Triflumizole	1.4	WQ4, WQ5	1-52	Mefenacet	3.23	WQ1-2	1-85	Buprofezin	4.3	WQ1-2
1-20	Triflumizole-13C3	1.4	PL-V	1-53	Iprobenfos	3.34	WQ1-2	1-86	Difenoconazole	4.3	WQ4, WQ5
1-21	Propoxur	1.52	WQ5	1-54	Iprobenfos-d7	3.34	PL-II	1-87	Difenoconazole-d6	4.3	PL-V
1-22	Pyroquilon	1.57	WQ1-2	1-55	Napropamide	3.36	WQ1-2	1-88	Phosalone	4.38	WQ5
1-23	Metalaxyl	1.65	WQ1-2	1-56	Thiobencarb	3.4	WQ1-2	1-89	Esprocarb	4.6	WQ1-2
1-24	Fosthiazate	1.68	WQ5	1-57	Propyzamide	3.43	WQ1-2	1-90	Butamifos	4.62	WQ1-2
1-25	Fosthiazate-d5	1.68	PL-II	1-58	Bromobutide	3.47	WQ1-2	1-91	Dithiopyr	4.75	WQ1-2
1-26	Metribuzin	1.7	WQ4, WQ5	1-59	Edifenphos	3.48	WQ1-2	1-92	Clomeprop	4.8	WQ4
1-27	Bromacil	2.11	WQ4, WQ5	1-60	Prometryn	3.51	WQ4, WQ5	1-93	Pencycuron	4.82	WQ1-2
1-28	Quinoclamine	2.12	WQ5	1-61	Thenylchlor	3.53	WQ1-2	1-94	Chlorpyrifos	4.96	WQ1-2
1-29	Simazine	2.18	WQ1-2	1-62	Tetraconazole	3.56	WQ4, WQ5	1-95	Pyributicarb	5.18	WQ1-2
1-30	Methidathion	2.2	WQ1-2	1-63	Mepronil	3.66	WQ1-2	1-96	Pyriproxyfen	5.37	WQ1-2
1-31	Cyanazine	2.22	WQ4, WQ5	1-64	Propaphos	3.67	WQ5	1-97	Etofenprox	7.05	WQ1-2
1-32	Isoprocarb	2.31	WQ1-2	1-65	Pyrazoxyfen	3.69	WQ5
1-33	Metominostrobin	2.32	WQ4, WQ5	1-66	Flutolanil	3.7	WQ1-2

The target analytes are listed according to the order of estimated log P_OW (Est. log P_OW). Other abbreviations are as follows: Standard solution (STD), 66 Pesticides Mixture Standard Solution WQ-1-2 (WQ1-2), 63 Pesticides Mixture Standard Solution WQ-4 (WQ-4), 48 Pesticides Mixture Standard Solution WQ-5 (WQ-5), PL Pesticides surrogate mix II (PL-II), PL Pesticides surrogate mix V (PL-V), PL Pesticides surrogate mix VII (PL-VII).

Komatsuna (Japanese mustard spinach), mistuba (Japanese honewort), and young burdock were selected as the vegetable samples, whereas fig, grape, and satsuma (satsuma mandarin) were selected as the fruit samples. The edible portions of each food sample were prepared prior to homogenization: the roots of komatsuna and mitsuba were removed, while the entire young burdock was used. For satsuma, the exocarp and stalk end were removed, while only the stalk was removed from figs and grapes.

Clean-up procedure: the official Japanese method (multiresidue method I for agricultural chemicals by LC/MS (agricultural products))³⁾ was selected and modified. A 20.0 g homogenized sample was mixed with 100 mL of acetonitrile and was agitated for 30 min. The mixture was then suction-filtered, and the residue was rinsed with approximately 20 mL of acetonitrile. The combined extracts were brought to a final volume of 200 mL with acetonitrile. A mixture of 80 mL aliquot of the extraction solution, 10 g of sodium chloride and 20 mL of 0.5 M phosphate buffer (pH 7) was agitated for 10 min in a separation funnel. After phase separation, the acetonitrile layer was dried over anhydrous sodium sulfate and filtered. Moreover, the filtrate was concentrated and dried using a rotary evaporator (40°C) and a gentle stream of nitrogen gas. The residue was redissolved in acetonitrile/toluene (3 : 1 v/v) and passed through a graphite carbon/aminopropylsilanized silica gel-layered cartridge (InertSep® GC/NH₂; GL Science, Tokyo, Japan) and preconditioned with 10 mL of acetonitrile/toluene (3 : 1 v/v). An eluate obtained by filtering a 20 mL mixture of the residue and acetonitrile/toluene (3 : 1 v/v) mixture was concentrated to dryness, and the residue was reconstituted in methanol.

LC-MS/MS conditions: the LCMS-8050® system in the ESI mode (Shimadzu Corp., Kyoto, Japan) was used for analysis. Supplemental Table S1 presents a list of the analytes, retention time (t_R), molecular formulas, precursor ion, product ion, Q1 and Q3 pre bias, collision energy (CE). The mobile phases were as follows: solvent A (2 mmol/L ammonium formate and 0.002% formic acid in ultrapure water) and solvent B (2 mmol/L ammonium formate and 0.002% formic acid in methanol). Analyte separation was conducted on a Kinetex® Biphenyl column (2.6 µm, 100×2.1 mm²; Phenomenex, California, USA) using a binary gradient mode, wherein the A : B ratio varied as follows: 0 min, 97 : 3; 1 min, 90 : 10; 3 min, 45 : 55; 10.5 min, 0 : 100; 12 min, 0 : 100; 12.01 min, 97 : 3; and 15 min, 97 : 3. Supplemental Table S2 shows the list of the other LC-MS/MS conditions. The limits of determination and quantification were 0.005 µg/kg and 0.01 µg/kg, respectively.

ME evaluation: The ME was calculated using the following equation, as described in earlier studies^12,13):

$$\text{ME}=\left(\frac{\begin{array}{l}\text{peak area of standard in }\\ \text{cleaned-up extract}\end{array}}{\begin{array}{l}\text{peak area of standards in }\\ \text{methanol}\end{array}}-\text{1}\right)\times \text{100\%}$$ (1)

The target analytes were spiked into the final solutions at 10 ng/mL. The pesticides contained in more than one standard solution were spiked at a concentration of 20 ng/mL (Table 1). The target analytes detected in non-spiked (blank) samples and/or those with defective peak shapes were excluded from the ME evaluation. The final solutions of two sample concentration levels were prepared (4 g sample/mL and diluted 0.8 g sample/mL), and each analysis was repeated three times.

Results and discussion

Figure 1 shows the number of target analytes showing ion suppression (ME<−10%), ion enhancement (ME>10%), and negligible ME (−10%≥ME≥10%). Ion suppression was shown in 44%, 45%, and 40% of the target analytes in komatsuna, mitsuba, and young burdock, respectively. In contrast, for fig, grape, and satsuma, only 11%, 2%, and 5% of the target analytes showed ion suppression, respectively, while ion enhancement was rarely observed for almost all analytes in both the vegetable and fruit samples. Supplemental Table S3 shows the list of ME in the 2 mL final solution derived from the vegetable and fruit samples. Quinoclamine and esprocarb exhibited substantial ion suppression across all the vegetable samples. Aside from these, no clear pattern was noted among the target analytes showing substantial ion suppression in the vegetable samples. Several studies have shown that the magnitude of the ME varies significantly depending on the pesticide–sample combinations.^14,20) Fernández-Alba et al.²⁰⁾ assessed the differences in ME among the three commodity groups, high water content, high acid and water content, and difficult or unique commodities, using analytical methods in accordance with EU SANTE/11945/2015. It was concluded that the number, distribution, and abundance of natural components varied greatly between matrices, even within the same commodity group. Our results are consistent with those previous studies, which suggested that the ME could be altered by the combination of the sample matrix and the physicochemical properties of the target analytes. However, our results suggested that ion suppression could be more problematic in the vegetable samples than in the fruit samples in the multiresidue analysis using the modified official Japanese method.

Fig. 1. The number of the target analytes were visualized according to the magnitude of ME: substantial ion suppression (ME<−20%), minor ion suppression (−20%≤ME<−10%), negligible ME (−10%≤ME≤10%), minor ion enhancement (20%≥ME>10%), substantial ion enhancement (ME>20%).

The results of this study exhibited the need for effective countermeasures against ME to achieve reliable analytical results. To assess the effectiveness of dilution, the MEs on target analytes were compared between the 2 mL and 10 mL final solution volumes (4 g sample/mL and diluted 0.8 g sample/mL, respectively). The results revealed that a fivefold dilution to a final volume of 10 mL was sufficient to eliminate substantial MEs for most analytes. In the 10 mL final solutions, substantial ion suppression was observed for only two analytes in fig, one analyte each in young burdock and satsuma, and none in komatsuna, mitsuba, and grape (Supplemental Table S4). In contrast, in the 2 mL final solutions, substantial ion suppression was observed in 16 analytes in komatsuna and mitsuba, 12 in young burdock, 4 in fig, 2 in satsuma, and none in grape. These findings indicated that dilution is effective, particularly in vegetable samples, for ME reduction. Previous studies have suggested that citrus fruits tend to exhibit substantial ion suppression and bioflavonoids such as hesperidin in citrus fruits, which could cause MEs.^9,10) Conversely, the ion suppression was rarely observed in satsuma in the 2 mL final solution in this study, which is inconsistent with the previous studies. This difference may be attributed to the exocarp of satsuma, which is rich in flavonoids,²¹⁾ that was removed in this study. However, the sample, analytical methods, and apparatus were not identical in these studies and further experiments should be conducted to support this probability.

Chapter 2: the ME between the vegetable and fruit-vegetable samples in the LC-MS/MS analysis^19,22)

Experiment 1: the evaluation of analyte properties related to the ME in the vegetable samples

Materials and methods

Reagents and samples: In Chapter 1, the ME of some pesticide–commodity combinations was not able to be calculated because the pesticides were detected from the blank samples. Therefore, 25 SIL-ISs of various log P_OW were selected as the target analytes in this study (Table 2). Hexazinone-d₆ purchased from CDN Isotopes, Inc. (Quebec, Canada) and Triphenyl phosphate (TPP) of Dr.Efrenstorfer™ purchased from LGC Limited (Teddington, UK) were considered as syringe spikes. Their corresponding pesticide standards were purchased from Wako. Acetonitrile and other reagents used is as mentioned in Chapter 1.

Table 2. Target analytes in Chapter 2.

No.	Analyte	Est. log Pow	STD
2-01	Nitenpyram-d3	−0.66	Hayashi Pure Chemicala)
2-02	Dinotefuran-d3	−0.64	Hayashi Pure Chemicala)
2-03	Thiamethoxam-d3	−0.13	Hayashi Pure Chemicala)
2-04	Acephate-d3	−0.9	Honeywellb)
2-05	Imidacloprid-d4	0.57	Hayashi Pure Chemicala)
2-06	Methomyl-d3	0.60	CDN Isotopesc)
2-07	Clothianidin-d3	0.64	Hayashi Pure Chemicala)
2-08	Acetamiprid-d3	0.80	Hayashi Pure Chemicala)
2-09	Thiacloprid-d4	1.26	Hayashi Pure Chemicala)
2-10	Propoxur-d3	1.52	CDN Isotopesc)
2-11	Cyanazine-d5	2.22	CDN Isotopesc)
2-12	Atrazine-d5	2.61	CDN Isotopesc)
2-13	Diuron-d6	2.67	CDN Isotopesc)
2-14	Triadimefon-d4	2.94	CDN Isotopesc)
2-15	Boscalid-13C6	2.96	Alsachimd)
2-16	Propyzamide-d3	3.43	CDN Isotopesc)
2-17	Fluxapyroxad-13C6	3.47	Alsachimd)
2-18	Prometryn-d4	3.51	Honeywellb)
2-19	Mepronil-13C, d3	3.66	Alsachimd)
2-20	Hexaconazole-d7	3.66	Sigma-Aldriche)
2-21	Tebuconazole-d3	3.70	Sigma-Aldriche)
2-22	Propiconazole-d3	4.13	Sigma-Aldriche)
2-23	Pirimiphos-methyl-d3	4.20	Honeywellb)
2-24	Fluopyram-d4	4.78	Sigma-Aldriche)
2-25	Flufenoxuron-d3	5.97	CDN Isotopesc)

^a) PL pesticides surrogate mix VII of Hayashi Pure Chemical Inc., ^b) Honeywell International Inc., ^c) CDN Isotopes Inc., ^d) Alsachim of Shimadzu Corp., ^e) Sigma-Aldrich® of Merck KGaA.

Komatsuna and spinach were selected to represent the vegetable commodities, while tomato and aubergine were selected to represent the fruit-vegetable samples. Five varieties of each sample cultivated on different farmlands in Osaka were purchased from several local markets (cultivars unknown) on the same day. The edible part of the samples was homogenized and used for the analysis.

Clean-up procedure and LC-MS/MS condition: the clean-up procedure was performed according to a method mentioned in Chapter 1. The volume of the final solution was 2 mL in this chapter (4 g/mL of methanol). Supplemental Table S5 shows a list of the analytes, t_R, molecular formulas, precursor ion, product ion, Q1 and Q3 pre bias, and CE. A Kinetex® C₁₈ column (2.6 µm, 100×2.1 mm²; Phenomenex, California, USA) coupled with SecurityGuard™ ULTRA biphenyl guard cartridge (4×2.1 mm², Phenomenex, California, USA) was used for analyte separation in LC. Other conditions followed as mentioned in Chapter 1. The limits of determination and quantification were 0.00125 and 0.0025 mg/kg, respectively.

ME evaluation: understanding MEs over a wide range of concentrations is essential to achieve reliable and precise results from multiresidue analysis, because the concentration of pesticide residues is unknown prior to conducting the analysis. Therefore, the following equation was selected to evaluate the ME on target analytes in this chapter^14,15):

$$\text{ME}=\left[\left(\frac{\begin{array}{l}\text{slope of matrix-matched }\\ \text{calibration}\end{array}}{\begin{array}{l}\text{slope of reagent-only }\\ \text{calibration}\end{array}}\right)-\text{1}\right]\times \text{100\%}$$ (2)

The reagent-only and matrix-matched calibrations were measured using linear regression (5, 10, 25, 62.5, and 175 ng/mL) with weighing (weighing factor=1/x², x=concentrations). A weighted regression was applied to calculate the calibration curves since the residuals of the five concentrations were not equivalent (Supplemental Fig. S1).²³⁾ Each sequence began with the reagent-only 0 standard (reagent blank) followed by the matrix-matched 0 standards (sample blank) to check the presence of false detects. The order of sample injections in the sequences was from low to high concentrations according to a previous study.¹⁴⁾ This pattern was repeated for the reagents and samples. The ME evaluation was replicated four times for each sample. The grand mean of ME was calculated by dividing the sum of each mean ME by the number of sample varieties (h=5).²⁵⁾ A one-way ANOVA (The degree of freedom (df) within groups=15, df between groups=4, F critical value=3.045 (p=0.05)) was also performed to investigate significant differences in the mean ME between various samples.²⁵⁾ In the comparison of ME between SIL-IS and pesticides, both the SIL-ISs and their corresponding pesticides were spiked into the final solution for all analytes, except propyzamide and propiconazole. For these two combinations, the pesticides and SIL-ISs were spiked separately into the sample solution to avoid identification errors in the LC-MS/MS analysis. The ME of syringe spikes (ME_ss) was calculated by the equations shown in Supplemental data.

The evaluation of the sampling and measurement variances of the ME: the calculation of the measurement variance (σ₀²) and sampling variance (σ₁²) was performed according to the theory explained in Ref. 25. Supplemental data shoes the equations used for calculation.

The analysis of the physicochemical characteristics: the chemical formula of target analytes were searched from the pesticide abstract provided by the Food and Agricultural Materials Inspection Center (FAMIC)²⁴⁾ and PubChem provided by the National Library of Medicine.²⁵⁾ The physicochemical characteristics were estimated from their structures using ChemDraw software (ChemDraw Professional 19.0)²⁶⁾ and EPI Suite system (Supplemental Table S6). The estimated physicochemical parameters were used in this study because some experimentally obtained parameters were derived under differing environmental conditions such as temperature.

Recovery test: a pesticide standard solution was spiked into 20.0 g of the homogenized komatsuna at 0.0025, 0.01, and 0.04 mg/kg concentrations, respectively. A standard solution of SIL-IS was spiked into the final solution at 0.0025 mg/kg. The recovery was calculated using three methods: external, internal, and matrix-matched calibration methods. All calibration curves were investigated in the range of 5–175 ng/mL with weighting (weighing factor=1/x²). TPP and hexazinone-d₆ were added to the final solution as syringe spikes (100 ng/mL). Recovery tests were performed in five replicates. Recovery was calculated for all analytes using the external standard and matrix-matched calibration methods. For the internal calibration method, the recoveries of propyzamide and propiconazole were not analyzed because the discrimination of the pesticides from their SIL-IS was not possible using LC-MS/MS when these analytes were mixed in the final solution. This could be attributed to the fact that the molecular mass of the pesticides and SIL-ISs were nearly identical, since the pesticides contains two chlorine atoms but their SIL-ISs contains three d isotopes, resulting in natural isotopes of similar molecular mass.

Results and discussion

Repeatability of the ME evaluation

The suitability of TPP and hexazinone-d₆ as syringe spikes was evaluated and the LC-MS/MS performance was assessed throughout the experimental period. The repeatability of both TPP and hexazinone-d₆ did not exceed more than 20%,⁷⁾ showing that LC-MS/MS performance was good throughout the study (Supplemental Table S7). The ME_ss of TPP and hexazinone-d₆ was between −7% and 3%, which is sufficiently low to be ignored (Supplemental Table S7). Throughout the sequence, TPP and hexazinone-d₆ showed a slight downward trend from the low to high concentrations of the calibration (Supplemental Fig. S2). Lehotay et al. reported a similar downward trend in multiresidue analysis using the modified QuEChERS method and LC-MS/MS for apple, spinach, orange, and rice samples.¹⁴⁾ They suggested that this trend is presumably caused by the accumulation of co-injected matrix materials in the system or increasing analyte concentration in the calibration standards with sequence progression.¹⁴⁾ Since no accumulation of hexazinone-d₆ in LC-MS/MS was observed, hexazinone-d₆ was overall a better syringe spike than TPP in compensating for the slight downward trend in the sequence.

ME evaluation in the vegetable and fruit-vegetable samples

The similarity of MEs on 25 pesticides and their SIL-ISs was assessed using an identical blank komatsuna sample, to evaluate whether ME assessment using SIL-ISs can serve as an alternative to direct ME assessment on the pesticides. The average difference in the MEs of the 25 pesticides and those of the SIL-IS was 2%. The maximum difference in the ME of boscalid and boscalid-d₆ was only 8%. A two-tailed t-test (t=2.31 (p=0.05), n=5) was used to determine the significance.²⁵⁾ The results indicated no significant difference in the MEs on the 18 pesticides and SIL-ISs. The exceptions were thiamethoxam, thiacloprid, diuron, boscalid, triadimefon, prometryn, and their SIL-ISs. For these pesticide–commodity combinations, an ME evaluation was conducted as reference. Overall, the MEs on SIL-ISs were assessed to estimate the ME patterns of the pesticides in the samples. Before the ME evaluation, the linearity of the calibration curves of the target analytes was also assessed to determine the LC-MS/MS performance in obtaining reliable MEs. The results showed that the deviations ranged between −9% and 8%, which were considered satisfactory for all pesticide–commodity combinations according to the guidelines (not exceeding ±20%).²⁷⁾ The r-squared (r²) values of the calibration curves were greater than 0.989 for all the vegetable samples. Satisfactory results were obtained in both the reagent-only and matrix-matched calibration curves.

The results of the ME evaluation on the target analytes in four kinds of vegetable samples revealed that the MEs were mostly not consistent across analytes in the same type of vegetable (Fig. 2). Substantial ion suppression (ME<−20%) was observed for dinotefuran-d₃, clothianidin-d₃, and thiacloprid-d₄ for komatsuna, acetamiprid-d₃ and flufenoxuron-d₃ for spinach, and nitenpyram-d₃ for tomato, in at least one result of the four replicates (Supplemental Table S8). Conversely, the grand mean of the ME which was considered substantial was only thiacloprid-d₄–komatsuna combinations. Moreover, the results of the one-tailed F-test indicated that 20 out of 25 analytes in komatsuna and 17, 17, and 12 out of 25 analytes in spinach, tomato, and aubergine showed a significant difference in the ME between various samples (Supplemental Table S8). Lehotay et al.¹⁴⁾ suggested that the MEs were generally consistent across most analytes when comparing different types of apples (e.g., Golden Delicious, Braeburn, Pink Lady, and Fuji), spinach (e.g., navel, juicy, moro, and biologique), and rice (e.g., brown, red, and white) in the multiresidue analysis using the modified QuEChERS method. The standard deviation of MEs among these varieties in the LC-MS/MS analysis was typically below 10%, except for certain combinations involving specific pesticides and orange samples. Based on these findings, it was concluded that a single variety can reasonably represent other types within the same matrix in LC-MS/MS analysis. Besil et al. also showed that the MEs were different among citrus species but similar between varieties in orange and mandarin.⁹⁾ Conversely, Kruve et al. showed indicated differences in the ME among apple varieties for some pesticides treated with QuEChERS method and the RSD values were as high as 34% for some pesticides.²⁸⁾ Since significant differences in the MEs were observed between varieties of samples in the majority of the analyte–sample combinations in our study, it was suggested that the MEs could vary due to sampling and/or measurement differences in the multiresidue analysis using the modified official Japanese method.

Fig. 2. Box plot diagram of the matrix effect (ME) on 25 stable isotope-labelled internal standards for four types of vegetables when the peak area is compensated for by hexazinone-d₆. Five samples of each type were analyzed, and each sample was measured four times. In the boxplot, the box’s bottom end represents the lower quartile of the data, the bold line in the middle represents the median of the data, and the box’s top end represents the upper quartile of the data. The lower and upper whisker represents the minimum and maximum data, respectively. The outliers are shown in white dots.

Sampling and measurement variances of ME

The estimated variances of measurement and sampling were assessed to understand ME variation. The estimated standard deviation (ESD) of measurement was within the range of 0.6–3% for komatsuna, 0.4–2% for spinach, 0.5–3% for aubergine, and 0.6–4% for tomato, respectively (Fig. 3). The results indicated that the measurement variance was low (≤4%) when the samples were undergoing the same clean-up procedure by the same analyst using the same instrument. Conversely, the ESD of sampling was within the range of 0–9% for komatsuna, 0–13% for spinach, 0–5% for aubergine, and 0–8% for tomato, indicating a slightly wider range than the ESD of measuring. The ESD of sampling for acetamiprid-d₃ and flufenoxuron-d₃ in spinach was 13%. The wide variability in the ME for dinotefuran-d₃ was shown in the previous section (Fig. 2). The sampling variances were greater than the measurement variances for all cases involving dinotefuran-d_3. These results suggested that while both sampling and measurement contribute to ME variability, sampling tends to have a greater influence on a wide variability. The consistency of ME was suggested when the standard deviation values were <10% in a previous study.¹⁴⁾ Our results suggest the consistency of ME between the same types of vegetable samples, except for acetamiprid-d₃ and flufenoxuron-d₃ in spinach. However, the ME could significantly vary among the same kinds of samples by sampling and measurement (Supplemental Table S8). These results suggest that the similarity of ME between the test samples and matrix-matched samples should be considered for the assessment of the applicability of the matrix-matching method for ME compensation.

Fig. 3. Estimated standard deviation (ESD) of the matrix effect on 25 stable isotope-labelled internal standards for four types of vegetables when the peak area is compensated for by hexazinone-d₆. σ₀ and σ₁ represent the square root value of the measurement and sampling variances, respectively. Five samples of each type were analyzed, and each sample was measured four times.

ME and physicochemical characteristics of the analytes

Figure 4 shows the scattered diagram of the relationship between the mean ME and the log P_OW. When the log P_OW was more than 1.52, the mean values of the absolute ME were 2%, 1%, 1%, and 1% in the komatsuna, spinach, tomato, and aubergine samples, respectively. Similarly, the ESD of the ME on the analytes were 1%, 3%, 1%, and 1% in the komatsuna, spinach, tomato, and aubergine samples, respectively. However, substantial ion suppressions in at least one of the replicates in the five varieties of each sample were observed in the analytes with a log P_OW below 1.52. These results suggest that the analytes with a high log P_OW rarely exhibit substantial ME in the modified official Japanese method and the ME might be severe and altered in the analytes with a low log P_OW. The analytes with a high estimated log P_OW and some properties related along with a high log P_OW such as long t_R tend to exhibit a small magnitude and ESD of the ME (Supplemental Fig. S3). Since not all of the analytes with a low log P_OW exhibited an unstable ME among the same types of samples, further research is needed to better understand the properties related to ME. However, the results revealed no clear correlation between the single physicochemical property of the target analytes and the MEs in any of the samples, indicating that multiple properties should be discussed to understand their relationship.

Fig. 4. Relationship between mean matrix effect (ME) on the five samples of komatsuna and the estimated octanol-water coefficient (log P_OW). Each sample was analyzed four times. The error bar represents the standard deviation.

Recovery test using the external, internal, and matrix-matched calibration methods

Since the appropriate dilution factor to ignore the ME influence was unknown before the analysis for the other samples, the countermeasures against ME other than dilution should be discussed for a multiresidue analysis. The pesticide recovery based on external, internal, and matrix-matched calibration was discussed. In the recovery test, compensation for the peak area of the target analyte by hexazinone-d₆ was not performed because the number of samples in the sequence was less than that in the ME evaluation and the RSD of hexazinone-d₆ throughout the sequence was only 2%. Table 3 shows the recoveries of the 25 pesticides from the blank komatsuna samples. In the external calibration method, the recoveries of acephate, nitenpyram, methomyl, and thiacloprid were lower than 70% for at least one concentration level, which did not satisfy the MHLW-criteria²⁹⁾ (recovery range: 70–120%). The recoveries of methomyl were 63%, 69%, and 66% at 0.0025 ng/kg, 0.01 ng/kg, and 0.04 ng/kg spiking levels, respectively, and none of them satisfied the criteria. The RSD values for the analyte recovery were below 15% for all analytes which were satisfactory (RSD<25% at 0.0025 ng/kg and RSD<15% at 0.01 and 0.04 ng/kg). The application of the internal and matrix-matched calibration methods improved the recovery of methomyl (Fig. 5). The recoveries of methomyl using the matrix-matched calibration method improved to 93%, 96%, and 91% at 0.0025 ng/kg, 0.01 ng/kg, and 0.04 ng/kg spiking levels, respectively. The recoveries of methomyl using the internal calibration method were 82%, 85%, and 86% at 0.0025 ng/kg, 0.01 ng/kg, and 0.04 ng/kg spiking levels, respectively. The average recovery of the target analytes among the calibration methods was compared using Tukey’s multiple comparison test (T critical value=4.60 and 6.98 (p=0.05 and 0.01, respectively), n=5). The application of the internal and matrix-matched calibration methods significantly improved the recovery of methomyl compared with the external calibration method. However, a significant difference was observed between the recoveries calibrated using the internal and matrix-matched methods for methomyl at 0.0025 and 0.01 mg/kg. Although no significant differences were observed between the MEs on methomyl and those on methomyl-d₃ in blank komatsuna (Table 3), the use of a matrix-matched calibration method with a blank sample identical to the test sample may contribute to more precise residue analysis. If an identical blank sample is not available, caution should be exercised. Based on the results obtained (Supplemental Table S8), the similarity of MEs between the test and matrix-matched samples should be evaluated to avoid under- or over-compensation of the peak area. This study also suggests that, in most cases, the use of the internal calibration method entailing the addition of low concentrations of SIL-ISs (0.0025 mg/kg) into the samples delivers improved recoveries compared with those when an external calibration method entailing low to high concentrations is used. The unsatisfactory recoveries of acephate and nitenpyram were not attributable to ME, as they were not improved by changing the calibration method. Since these pesticides are not target analytes of the official Japanese method, improvements in the cleanup procedure may be necessary to enable better analysis of acephate and nitenpyram in vegetable samples.

Table 3. Average recovery and relative standard deviation (RSD) of the pesticides from blank komatsuna sample in five replicates (%). The bold values show the recovery or RSD that did not satisfied the Ministry of Health, Labour, and Welfare-criteria (recovery: 70–120%, RSD%<25% for 0.0025 mg/kg concentration level and RSD>15% in 0.01 and 0.04 mg/kg concentration level). The pesticides are listed according to the estimated log P_OW.

No.	Pesticide	Average recovery (RSD) (%)
		External calibration			Matrix-matched calibration			Internal calibration
		0.0025 mg/kg	0.01 mg/kg	0.04 mg/kg	0.0025 mg/kg	0.01 mg/kg	0.04 mg/kg	0.0025 mg/kg	0.01 mg/kg	0.04 mg/kg	SIL-IS
2-01′	Nitenpyram	43 (8)	51 (6)	45 (12)	51 (7)	55 (6)	47 (11)	48 (10)	48 (7)	45 (17)	Nitenpyram-d3
2-02′	Dinotefuran	85 (4)	80 (4)	71 (2)	92 (4)	90 (4)	81 (2)	91 (4)	88 (3)	85 (5)	Dinotefuran-d3
2-03′	Thiamethoxam	85 (6)	90 (3)	82 (3)	95 (6)	94 (3)	85 (3)	82 (11)	89 (6)	93 (7)	Thiamethoxam-d3
2-04′	Acephate	0 (—)	0 (—)	0 (—)	0 (—)	0 (—)	0 (—)	0 (—)	0 (—)	0 (—)	Acephate-d3
2-05′	Imidacloprid	86 (3)	86 (3)	81 (2)	94 (3)	94 (3)	90 (2)	92 (4)	90 (3)	89 (4)	Imidacloprid-d4
2-06′	Methomyl	63 (5)	69 (2)	66 (2)	93 (4)	96 (2)	91 (2)	82 (4)	85 (4)	86 (5)	Methomyl-d3
2-07′	Clothianidin	86 (11)	84 (4)	77 (3)	99 (11)	96 (4)	88 (3)	96 (5)	85 (7)	90 (6)	Clothianidin-d3
2-08′	Acetamiprid	93 (1)	91 (1)	86 (1)	100 (1)	98 (1)	92 (1)	93 (5)	90 (3)	95 (5)	Acetamiprid-d3
2-09′	Thiacloprid	72 (10)	70 (2)	68 (1)	89 (11)	94 (2)	92 (1)	88 (7)	84 (2)	85 (2)	Thiacloprid-d4
2-10′	Propoxur	86 (3)	88 (2)	86 (3)	96 (3)	94 (2)	91 (3)	89 (4)	89 (2)	92 (5)	Propoxur-d3
2-11′	Cyanazine	93 (3)	94 (2)	92 (3)	97 (3)	97 (2)	95 (3)	91 (3)	95 (3)	95 (4)	Cyanazine-d5
2-12′	Atrazine	92 (1)	95 (2)	90 (1)	100 (1)	100 (2)	95 (1)	92 (2)	93 (2)	94 (2)	Atrazine-d5
2-13′	Diuron	91 (3)	90 (2)	85 (1)	98 (3)	96 (2)	91 (1)	91 (3)	90 (2)	90 (1)	Diuron-d6
2-14′	Triadimefon	102 (3)	99 (3)	96 (2)	105 (3)	100 (3)	97 (2)	98 (6)	96 (4)	96 (6)	Triadimefon-d4
2-15′	Boscalid	97 (13)	94 (4)	93 (3)	95 (13)	95 (5)	95 (3)	86 (11)	88 (8)	101 (16)	Boscalid-13C6
2-16′	Propyzamide	97 (7)	99 (4)	95 (2)	97 (8)	100 (4)	96 (2)	NA	NA	NA	Propyzamide-d3
2-17′	Fluxapyroxad	97 (2)	96 (1)	92 (1)	99 (2)	98 (1)	94 (1)	94 (3)	91 (3)	89 (2)	Fluxapyroxad-13C6
2-18′	Prometryn	94 (4)	94 (2)	94 (1)	98 (4)	97 (2)	97 (1)	93 (4)	91 (2)	95 (2)	Prometryn-d4
2-19′	Mepronil	98 (1)	98 (3)	92 (2)	100 (1)	101 (3)	95 (2)	93 (3)	90 (6)	92 (10)	Mepronil-13C, d3
2-20′	Hexaconazole	98 (2)	95 (2)	93 (2)	97 (2)	97 (2)	95 (2)	91 (5)	89 (4)	89 (5)	Hexaconazole-d7
2-21′	Tebuconazole	106 (13)	98 (7)	96 (2)	100 (14)	96 (7)	96 (2)	105 (18)	95 (10)	88 (13)	Tebuconazole-d3
2-22′	Propiconazole	98 (3)	99 (2)	98 (2)	95 (3)	99 (2)	98 (2)	NA	NA	NA	Propiconazole-d3
2-23′	Pirimiphos-methyl	96 (2)	95 (2)	96 (3)	97 (2)	96 (2)	97 (3)	93 (4)	92 (3)	96 (3)	Pirimiphos-methyl-d3
2-24′	Fluopyram	101 (2)	97 (2)	94 (1)	100 (2)	100 (2)	98 (1)	97 (2)	92 (3)	95 (2)	Fluopyram-d4
2-25′	Flufenoxuron	101 (2)	101 (2)	102 (2)	98 (2)	98 (2)	100 (2)	93 (2)	94 (3)	96 (1)	Flufenoxuron-d3

The abbreviations are as follows: Stable isotope-labeled internal standard (SIL-IS), and not available (NA)

Fig. 5. Recovery of methomyl from the blank komatsuna sample calibrated using external standard (std.), internal std., and matrix-matched calibration (cal.) methods in three concentration levels (0.0025, 0.01, and 0.04 mg/kg) (n=5). The differences in the average recovery of each pesticide was assessed using Tukey’s multiple comparison test (** p<0.01, * p<0.05).

Chapter 3: Simultaneous analysis of seven neonicotinoid pesticides using a novel method^19,30)

Materials and methods

Reagents and samples: seven neonicotinoid pesticides, namely, acetamiprid, clothianidin, dinotefuran, imidacloprid, nitenpyram, thiamethoxam, and thiacloprid were selected as the target analytes in this chapter. The PL pesticides surrogate mix VII including dinotefuran-d₃, nitenpyram-d₃, thiamethoxam-d₃, imidacloprid-d₄, clothianidin-d₃, acetamiprid-d₃, and thiacloprid-d₄, was purchased from Hayashi Pure Chemical Inc (Osaka, Japan). Acetonitrile and other reagents used is as mentioned in Chapter 1.

Brown rice, peanuts, and grapes were selected as the food samples. Before homogenization, the chaffs and shells of the brown rice and peanut samples were removed, respectively. Brown rice and peanuts were ground into a homogeneous powder using a mill (Panasonic Corporation, Osaka, Japan), while 10.0 g of the powder was soaked in 20 mL of ultrapure water prior to the analysis. The grape sample was thoroughly homogenized using a food processor (Panasonic), and 20.0 g was used for analysis.

Clean-up procedure and LC-MS/MS condition: the modified official Japanese method and proposed method were assessed.

The modified official Japanese method: the clean-up procedure was performed according to a method mentioned in Chapter 1. The acetonitrile extract was increased to 200 mL and a 10 mL aliquot of the extract solution was transferred for the clean-up procedures. The residue was reconstituted with 1.0 mL of methanol and analyzed using LC-MS/MS.

Proposed method: the 10 mL aliquot of the extract solution was obtained using the modified official Japanese method. The aliquot was concentrated by a rotary evaporator (Büchi, Flawil, Switzerland) at 40°C. The residue was dissolved in 5 mL of methanol/ultrapure water (2 : 8 v/v) and submitted to InertSep® Pharma FF (fast-flow) cartridge (500 mg/6 mL) (GL Sciences, Tokyo, Japan) which was pre-conditioned with 10 mL of methanol and a methanol/ultrapure water (2 : 8 v/v) solution, respectively. The condensed sample was percolated through a Pharma FF cartridge under a vacuum manifold and washed with 5 mL of methanol/ultrapure water (2 : 8 v/v). The cartridge was then dried with a stream of nitrogen gas for 30 min. Moreover, the dried cartridge was combined with InertSep® GC cartridge (500 mg/6 mL) (GL Sciences) pre-conditioned with 10 mL of methanol and loaded with 15 mL of methanol. After the Pharma FF cartridge was removed, the GC cartridge was then loaded with 5 mL of methanol. The collected eluent was condensed with an evaporator and a stream of nitrogen gas. The residue was re-dissolved in 1.0 mL of methanol.

Supplemental Table S9 shows a list of the analytes, t_R, molecular formulas, precursor ion, product ion, Q1 and Q3 pre bias, and CE. A Kinetex® Biphenly column (2.6 µm, 100×2.1 mm²; Phenomenex, California, USA) coupled with a SecurityGuard™ ULTRA biphenyl guard cartridge (4×2.1 mm², Phenomenex, California, USA) was used for analyte separation. Other conditions follows as mentioned in Chapter 1. The limit of determination and limit of quantification were 0.005 µg/kg and 0.01 µg/kg, respectively.

ME evaluation: the ME of the proposed method was measured using Eq. (1), as described in Chapter 1. The target analytes were spiked into the final solution at a concentration of 10 ng/mL. Each sample was analyzed in triplicate.

Recovery test: the target analytes were spiked into a homogenized sample at a concentration of 0.01 mg/kg. The residue was reconstituted in 1 mL of methanol (equivalent to 0.5 g/mL for cereal samples and 1 g/mL for grape samples). The calibration curve was generated without weighting. The modified official Japanese method was performed in triplicate, while the proposed method was conducted five times for each sample. For the internal standard calibration method, the calibration curves were constructed by plotting the peak area ratio (native/surrogate) versus the mass fraction ratios (native/surrogate). The recovery of the surrogates spiked at two different timings; they were evaluated before the extraction procedure and after obtaining the aliquot and compared using a two-tailed t-test (t=2.31 (p=0.05), n=5). The concentration of the surrogates in the final solution was 10 ng/mL for both cases.

Results and discussion

ME evaluation: the ME of the pesticide–sample combinations ranged from −27 to 37%, and the results exhibited significant differences in the matrix effect according to the food type (Fig. 6). Ion enhancement was observed in all the target analytes in the presence of brown rice and grape matrices. In contrast, ion suppression was shown in the peanut matrix, except in the case of imidacloprid. The results indicated that no substantial ME was observed in the brown rice, suggesting that ME might affect the results in the method validation. A similar ME was observed in the surrogates.

Fig. 6. Matrix effect (ME) on seven neonicotinoid pesticides in brown rice, peanut, and grape samples cleaned-up with the proposed method (n=4). The error bar represents the standard deviation.

Surrogate recovery: the recoveries and RSDs of the surrogates were comparable whether the compounds were spiked into the sample before extraction or after aliquot collection. When the surrogates were spiked prior to extraction, the recoveries ranged from 82 to 120%; when spiked into the aliquot, the recoveries ranged from 76 to 119% (Table 4). In all cases, the RSDs were below 8.4% across the surrogate–sample combinations. No significant differences were observed between the two spiking methods (p=0.05), suggesting that aliquot separation may not affect surrogate recovery. These results met the performance criteria as internal standard, as specified in the MHLW-guidelines (the recovery of ISs must be more than 40%).²⁹⁾ Furthermore, the findings suggest that post-aliquot spiking can effectively compensate for analyte loss during SPE clean-up and MEs in the LC-MS/MS analysis, which is comparable to the conventional method of pre-extraction spiking.

Table 4. Recovery and relative standard deviation of neonicotinoid pesticides obtained by the proposed method.

No.	Analyte	Sample	External calibration		Internal calibration
			BEa)	AFb)	BEa)	AFb)	Surrogate
3-01′	Nitenpyram	Brown rice	91 (6.9)	87 (1.0)	103 (0.4)	103 (0.9)	Nitenpyram-d3
		Grape	81 (4.0)	83 (9.3)	103 (1.7)	103 (1.8)
		Peanuts	87 (4.0)	84 (2.6)	102 (0.9)	101 (1.8)
3-02′	Dinotefuran	Brown rice	99 (1.7)	101 (2.7)	109 (1.5)	108 (4.3)	Dinotefuran-d3
		Grape	99 (5.3)	97 (7.3)	101 (3.1)	103 (3.1)
		Peanuts	90 (2.8)	88 (2.3)	111 (2.5)	108 (4.7)
3-03′	Thiamethoxam	Brown rice	101 (2.4)	102 (0.9)	102 (2.0)	103 (1.0)	Thiamethoxam-d3
		Grape	110 (1.7)	103 (7.1)	103 (2.3)	103 (0.7)
		Peanuts	85 (5.6)	81 (2.8)	101 (2.1)	99 (1.4)
3-04′	Imidacloprid	Brown rice	119 (1.3)	118 (0.9)	107 (1.3)	106 (1.0)	Imidacloprid-d4
		Grape	119 (0.8)	111 (6.6)	103 (2.1)	104 (1.8)
		Peanuts	115 (3.8)	113 (2.8)	105 (2.6)	105 (2.4)
3-05′	Clothianidin	Brown rice	108 (1.2)	110 (2.3)	102 (2.0)	106 (5.2)	Clothianidin-d3
		Grape	119 (0.5)	115 (5.0)	99 (5.3)	108 (1.5)
		Peanuts	86 (5.1)	84 (2.5)	101 (5.0)	101 (2.6)
3-06′	Acetamiprid	Brown rice	100 (1.4)	100 (0.8)	103 (1.8)	103 (1.1)	Acetamiprid-d3
		Grape	103 (1.5)	95 (8.3)	105 (1.6)	103 (1.4)
		Peanuts	101 (1.2)	106 (1.4)	97 (1.8)	104 (1.0)
3-07′	Thiacloprid	Brown rice	95 (2.2)	96 (1.5)	96 (1.2)	97 (1.5)	Thiacloprid-d4
		Grape	96 (1.9)	91 (8.2)	99 (0.5)	100 (2.0)
		Peanuts	96 (1.7)	91 (2.7)	95 (0.9)	93 (1.2)

Surrogates addition: ^a) Before extraction, ^b) after fractionation.

Method validation: the recoveries and RSDs of neonicotinoid pesticides obtained by the proposed method calculated using the external calibration method were in the ranges of 91–119 and ≤6.9%, 81–119 and ≤5.3%, and 85–115 and ≤5.6% for brown rice, grapes, and peanuts, respectively (Table 4). The recoveries and RSDs met the criteria in the MHLW guidelines for the validation of the pesticide-residue analysis methods (70–120 and <20%, respectively)²⁹⁾ for all analyte–sample combinations and validated the applicability of the proposed method. The results were better than those of the modified official Japanese method wherein the recovery of hydrophilic dinotefuran and nitenpyram was not satisfactory (Supplemental Fig. S4). Moreover, the recoveries and RSDs of the proposed method were calculated using the internal standard calibration method. When the surrogates were added to the sample before extraction and used for compensation, the improved recoveries and RSDs were obtained in the ranges of 96–109 and ≤2.0%, 99–105 and ≤5.3%, and 95–111 and ≤5.0% for brown rice, grapes, and peanuts, respectively (Table 4). The improvement can be attributed to the compensation of ME as shown in Fig. 3. However, most of the surrogates were left unused in the diluted extract after the process of fractionation, resulting in the high analytical cost. Since the surrogate concentrations in the final solution did not show a significant difference when the surrogates were spiked into the sample before extraction or after aliquot collection, the mitigation of the usage of futile surrogates was considered. Moreover, the identical recovery rates of the neonicotinoid pesticides were obtained when ISs were added to the aliquots for samples (Table 4). Since the internal standard calibration method is commonly unavailable to compensate for extraction efficiency, the absence of surrogates in the extraction procedure would not be a problem in the compensation of ME. The results indicated that the spiking procedure could reduce the analytical cost for determining neonicotinoid pesticide residues in food samples, while maintaining accuracy comparable to that of the conventional internal standard calibration method.

Conclusion

In Chapter 1, the MEs induced by the vegetable and fruit samples in a multiresidue analysis using the modified official Japanese method were compared, and the results showed that more analytes were affected by ion suppression in the vegetable samples than in the fruit samples. In contrast, ion enhancement was hardly observed. The comparison of MEs in the final solution of different matrix concentrations (4 g sample/mL and 0.8 g sample/mL) exhibited that a fivefold dilution could overcome ME in most cases.

In Chapter 2, not only ME but also its sampling and measurement variances of 25 SIL-ISs were evaluated for the vegetable and fruit-vegetable samples. The ESD of measurement was 0.4–4% for all the analyte–sample combinations. The ESD of sampling was less than 10% for all the analyte–ample combinations except for spinach and the two analyte combinations. The results of the one-way ANOVA showed significant differences in the ME among samples for more than about half of the target–analyte combinations. The results revealed that the sampling variance could be a significant factor in the wide variation of ME among the same agricultural products. In the evaluation of ME and the physicochemical properties of pesticides, the average MEs of the analytes with a log P_OW greater than 1.52 were nearly 0% in all crops. Conversely, the log P_OW of analytes which exhibited substantial ion suppression tended to be lower than 1.52, suggesting that ME is likely to be a problem in the analysis of analytes with a low log P_OW in the vegetable samples. However, multiple properties other than log P_OW should be discussed to understand the ME patterns. In the recovery test, the internal and matrix-matched calibration methods resulted in better recovery compared with that of the external calibration method in most cases. Moreover, the addition of SIL-IS (0.0025 ng/mL) is effective for internal calibration at low concentrations and for ME compensation in a wide concentration range (0.01 and 0.04 ng/mL, respectively).

In Chapter 3, it was shown that SPE cleanup by InertSep® Pharma FF and GC can achieve excellent recoveries of neonicotinoid pesticides from different types of agricultural products. Furthermore, the addition of surrogate standards to an aliquot of the extract solution, rather than to the sample prior to extraction, could reduce analytical costs while ensuring excellent recovery.

The findings contributed to the understanding of the ME in the multiresidue analysis of agricultural products using LC-MS/MS and emphasized the importance of selecting a suitable countermeasure against ME.

Electronic supplementary materials

The online version of this article contains supplementary material, which is available at https://www.jstage.jst.go.jp/browse/jpestics/.