Advertisement

Chromatographia

, Volume 82, Issue 1, pp 235–250 | Cite as

Possibilities and Limitations of Isocratic Fast Liquid Chromatography-Tandem Mass Spectrometry Analysis of Pesticide Residues in Fruits and Vegetables

  • Steven J. LehotayEmail author
Original
  • 235 Downloads
Part of the following topical collections:
  1. 50th Anniversary Commemorative Issue

Abstract

Currently, state-of-the-art analytical methods for multiclass, multiresidue monitoring of pesticides in foods use ultrahigh performance liquid chromatography-tandem mass spectrometry (UHPLC-MS/MS) for analysis of LC-amenable analytes. UHPLC-MS/MS for > 100 pesticides typically takes 10 min per injection using gradient elution in the reversed-phase mode, plus typically 3–5 min for re-equilibration between injections. Isocratic mobile-phase conditions eliminate need for the re-equilibration, and can greatly speed analysis time. In this study, fast isocratic LC-MS/MS was evaluated using a C18 analytical column of 3 cm × 4.6 mm i.d. with 3 µm particles and a mobile phase consisting of 10 mM ammonium formate at pH 3 in 47.5/47.5/5 (v/v/v) acetonitrile/methanol/water. Flow rate was 0.4 mL/min and injection volume was 50 µL. Sample preparation entailed a formate-buffered QuEChERS method for fruit and vegetable samples, typically yielding extract of 1.27 g/mL sample equivalent in 94/6 acetonitrile/water without additional cleanup. The analysis time was 2.6 min covering 88 diverse pesticide analytes each with three ion transitions (dwell times of 5 ms and 5 ms interscan delays). Validation experiments involving fortification of water, pear, tomato, cucumber, eggplant, and cilantro at 10 and 100 ng/g (n = 10 for each matrix and level) showed that the method achieved acceptable quantification with 70–120% recoveries and ≤ 25% RSD for 32–62 (36–70%) of the analytes depending on the matrix. Using regulatory identification criteria, only 6 false positives occurred above 10 ng/g among 4400 analyte/matrix/sample combinations, but false negatives varied depending on the pesticide/matrix pair, with results improving significantly for analytes with retention times > 1.3 min. This study demonstrated the feasibility and limits of isocratic LC-MS/MS for rapid screening of common commodities monitored for pesticide residues.

Graphical Abstract

Keywords

Isocratic liquid chromatography-tandem mass spectrometry (LC-MS/MS) High-throughput analysis QuEChERS sample preparation Qualitative identification Pesticide residues Fruits and vegetables 

Introduction

Since the establishment of legislation to regulate pesticide use in agriculture, the goal of pesticide residue monitoring in foods has been to meet regulatory data quality objectives with the highest sample throughput and lowest cost possible. Ideally, a reasonably high percentage of sample lots would be tested to ensure food safety and protect the environment from misuse of pesticides. However, lab and technological capabilities have limited the number of samples that can be reasonably analyzed within the budgets provided to monitoring programs. For example, US Food and Drug Administration (FDA) monitors pesticide residues in < 0.1% of imported food shipments, and even fewer domestically [1].

Over the years, the number of pesticide analytes has grown dramatically, which has made the challenge of multiclass, multiresidue monitoring using a single analytical method even more difficult. The trend of increasingly lower maximum residue limits (MRLs), known as tolerances in the US, has also made the analytical chemist’s work more challenging. Fortunately, instrumental technologies and sample preparation techniques have improved dramatically to increase the chances of meeting the pesticide monitoring goals [2, 3, 4]. Due to extensive global food trade and continuing incidents of newsworthy chemical residue violations in food, more funding both from government and industry has been devoted to residue monitoring. Importers/exporters stand to lose millions of dollars if shipments are found to contain violative levels of pesticide residues. Currently, food trade is a ≈$3.45 trillion business [5], and growth of food monitoring labs has increased > 10% per year for several years to currently exceed $3 billion/year [6].

Multiclass, multiresidue monitoring of pesticides in foods commonly entail sample preparation by the “quick, easy, cheap, effective, rugged, and safe” (QuEChERS) approach followed by parallel analysis using a combination of gas chromatography-mass spectrometry (GC-MS) and ultrahigh performance liquid chromatography-mass spectrometry (LC-MS) [7, 8, 9, 10]. Triple quadruple tandem (MS/MS) for targeted analytes is more frequently conducted presently, but new instruments using high resolution MS are growing in usage [11, 12, 13, 14]. However, analysis time from the injection of one final extract to the next typically takes > 15 min in UHPLC and > 30 min in GC even when using advanced MS-based detection tools. In the case of GC-amenable analytes, low-pressure (vacuum-outlet) GC-MS(/MS) has been extensively demonstrated to speed analytical cycle times to < 15 min to double sample throughput [15]. This provides similar time savings as micro-bore GC analysis [16], but with greater robustness.

In the case of LC-amenable analytes, one way to further increase sample throughput is to employ isocratic conditions. Not only are isocratic conditions usually designed to shorten analyte retention times, the technique eliminates the mobile-phase re-equilibration time needed between injections with gradient elution. Depending on specific parameters, the common disadvantages of isocratic LC designed for fast analyses include the higher potential for ghost peaks from previous injections and the co-elution of many analytes and matrix co-extractives. Both problems can lead to large and variable matrix effects (MEs) in electrospray ionization (ESI) MS. Furthermore, data acquisition rate limits the number of co-eluting analytes that can be monitored at the same time.

Flow-injection analysis constitutes an even faster way to conduct analyses [17, 18, 19, 20], but the problem of chemical co-elutions and MEs are even more severe. So far, only a limited number of analytes and matrices have been tested by flow-injection MS analysis. Isocratic fast LC-MS/MS serves as a good intermediary approach that still provides a degree of separation in only a 1–2 min longer analysis time, which could also expand the number of analytes targeted using MS/MS detection.

The aim of this study was to test the challenge of conducting < 3 min isocratic LC-MS/MS analysis of nearly 100 representative pesticide residues added as low as 10 ng/g in common fruits and vegetables. QuEChERS sample preparation was to be used to maximize sample throughput, and analyses were conducted with and without cleanup of the extracts. Qualitative and quantitative method performance would be assessed by regulatory validation protocols to find the limits in the approach using an older instrument.

Experimental

Reagents

The acetonitrile (MeCN) and methanol (MeOH) solvents used in the study were HPLC-grade obtained from Fisher (Fair Lawn, NJ; USA). Water was 18.2 MΩ-cm quality from a Barnstead (Dubuque,IO; USA) E-Pure system. The 88 analytes included in this study and internal standard (IS) are listed in Table 1. High purity standards (typically > 99%) were obtained from the US Environmental Protection Agency (EPA) National Pesticide Standard Repository (Fort Meade, MD; USA), Chemservice (West Chester, PA; USA), and Dr. Ehrenstorfer (Augsburg, Germany). Individual pesticide stock solutions of 1–2 mg/mL were prepared typically in MeCN, which were used in the preparation of a 10 ng/µL working standard mixture in MeCN. Atrazine-d5 for use as an IS was obtained from C/D/N Isotopes (Pointe-Claire, QC; Canada), and a 10 ng/µL solution in MeCN was prepared as a working standard. Other pesticide and IS solutions were prepared as needed from the working standards for use as spiking solutions and calibration standards. All solutions were contained in amber glass vials and stored at − 20 °C when not in use.

Table 1

LC-MS/MS conditions in the analysis of the 88 analytes plus internal standard

No.

Pesticide

MRM start (min)

tR (min)

MRM stop (min)

CV (V)

Precursor Ion (m/z)

CE ion 1 (eV)

Quant. ion 1 (m/z)

CE ion 2 (eV)

Qual. ion 2 (m/z)

CE ion 3 (eV)

Qual. ion 3 (m/z)

1

Acephate

0.851

0.96

1.35

16

183.8

8

142.8

18

124.9

28

94.7d

2

Acetamiprid

0.851

1.07

1.35

32

222.9

24

125.9

12

55.9

36

90.0

3

Cyromazine

0.851

1.00

1.35

32

166.9

18

84.9

18

60.0

16

124.9

4

Flumetsulam

0.851

0.98

1.35

32

325.9

24

128.9

54

108.8

18

261.9

5

Methamidophos

0.851

0.94

1.35

22

141.9

12

93.9

12

109.8

20

78.9

6

Aldicarb sulfone

0.861

0.94

1.36

22

222.8

8

147.9

18

85.9e

12

75.9e

7

Aldicarb sulfoxide

0.861

0.96

1.36

16

206.8

6

131.9

14

88.9

10

104.9

8

Methomyl

0.861

1.08

1.36

12

162.8

10

105.9

12

87.9

4

121.9

9

Monocrotophos

0.861

0.98

1.36

20

223.8

8

192.8

14

126.9

12

97.9

10

Omethoate

0.861

0.92

1.36

24

213.9

12

182.8

16

154.9

20

124.8

11

Oxamyl

0.861

1.02

1.36

12

236.9c

10

71.9

8

89.9

16

71.9f

12

Oxydemeton-methyl

0.861

1.05

1.36

20

246.8

14

168.8

26

108.8

12

104.9

13

Pymetrozine

0.861

1.07

1.36

28

217.9

18

104.9

32

78.6

26

97.8

14

Thiamethoxam

0.861

1.00

1.36

20

291.9

12

210.8

22

180.8

26

131.9

15

3-Hydroxycarbofuran

0.871

1.08

1.37

20

237.9

12

162.9

10

180.9

20

135.0

16

Clothianidin

0.871

1.06

1.37

18

250.0

12

168.8

14

131.8

26

110.1

17

Dicrotophos

0.871

1.05

1.37

20

237.9

12

111.9

22

71.9

8

192.9

18

Dimethoate

0.871

1.06

1.37

16

229.8

8

198.8

20

124.9

14

170.8

19

Imidacloprid

0.871

1.07

1.37

24

256.0

22

174.9

18

208.8

16

84.0

20

Phosmet-oxon

0.891

1.12

1.39

20

301.9

12

159.9

52

76.9

36

132.9

21

Aldicarb

0.901

1.12

1.40

8

207.9c

8

116.0

14

88.9

14

70.0

22

Atrazine-desethyl

0.901

1.10

1.40

32

187.8

16

145.9

24

78.9

24

103.9

23

Carbendazim

0.911

1.09

1.41

26

191.8

16

159.9

30

132.0

34

105.0

24

Carbofuran

0.911

1.14

1.41

22

221.9

10

164.9

20

122.9

28

55.0

25

Chlorsulfuron

0.911

1.07

1.41

26

357.9

18

166.9

16

141.0

44

110.8

26

Flutriafol

0.911

1.11

1.41

26

302.0

16

70.0

26

122.9

50

94.9

27

Parathion-oxon

0.921

1.13

1.42

22

275.9

14

219.8

32

93.9

24

173.8

28

Propoxur

0.921

1.12

1.42

16

209.9

14

110.9

8

167.9

24

92.9

29

Norflurazon

0.931

1.15

1.43

48

303.8

34

160.0

42

88.0

38

139.8

30

Fluxapyroxad

0.941

1.18

1.44

26

382.0

20

341.9

22

313.9

22

158.9

31

Imazethapyr

0.941

1.09

1.76

34

290.0

30

68.9

30

85.9

20

244.8

32

Thiabendazole

0.941

1.14

1.84

40

201.8

24

174.8

30

131.0

30

91.9

33

Carbaryl

0.951

1.15

1.45

18

201.9

14

144.9

30

126.9

24

116.9

34

Imazaquin

0.981

1.13

1.48

38

312.0

20

266.6

24

85.9

26

198.5

35

Fenobucarb

1.001

1.23

1.50

20

207.9

14

94.9

8

151.8

44

90.9

36

Atrazine

1.001

1.23

1.50

34

215.9

14

173.9

26

95.9

24

103.9

37

Atrazine-d5 (int std)

1.001

1.23

1.50

32

220.9

18

179.0

26

100.9

n/a

n/a

38

Coumaphos-oxon

1.001

1.17

1.50

28

346.7

18

290.7

10

318.7

30

210.8

39

Diflubenzuron

1.001

1.24

1.50

22

310.9

14

157.9

30

140.9

44

112.9

40

Diuron

1.001

1.20

1.50

28

232.8

16

72.0

16

46.0

26

159.8

41

Malathion

1.001

1.21

1.50

18

330.9

12

126.9

22

98.9

6

284.8

42

Metalaxyl

1.001

1.21

1.50

22

280.0

14

219.9

10

247.9

18

192.0

43

Phosmet

1.001

1.15

1.50

20

317.8

12

159.9

50

76.9

38

132.9

44

Carboxin

1.011

1.16

1.51

22

235.8

14

142.9

22

86.8

30

92.9

45

Flubendiamide

1.011

1.17

1.51

12

683.0

12

407.7

30

273.8

66

147.0

46

Tetraconazole

1.011

1.17

1.51

34

371.9

26

158.8

22

70.0

78

88.9

47

Azinphos-methyl

1.031

1.16

1.53

16

317.8

8

159.9

16

131.9

36

76.9

48

Carfentrazone-ethyl

1.051

1.22

1.55

34

411.9

16

365.7

12

383.6

20

345.7

49

Fluopyram

1.051

1.19

1.55

36

397.0

28

172.9

20

207.9

54

145.0

50

Imazalil

1.051

1.28

1.59

30

296.9

22

158.8

20

68.9

16

200.8

51

Methoxyfenozide

1.051

1.21

1.55

12

369.1

18

149.0

6

312.9

42

90.9

52

Methiocarb

1.071

1.22

1.57

20

225.9

8

168.9

18

121.0

34

106.9

53

Novaluron

1.071

1.22

1.57

30

493.0

20

157.9

50

140.9

80

112.9

54

Triadimefon

1.071

1.22

1.57

24

293.9

20

69.0

16

196.9

12

224.9

55

Triazophos

1.071

1.22

1.57

26

314.0

18

161.9

32

119.0

32

96.9

56

Azinphos-ethyl

1.081

1.25

1.58

14

345.9

18

131.9

10

159.9

24

104.2

57

Linuron

1.081

1.24

1.58

26

248.9

18

159.9

16

181.8

34

132.9

58

Promecarb

1.081

1.23

1.58

20

208.0

8

151.0

16

108.9

26

90.9

59

Indoxacarb

1.091

1.25

1.59

28

528.0

14

292.8

16

248.8

26

149.9

60

Penthiopyrad

1.091

1.26

1.59

26

360.0

14

275.9

30

176.8

20

255.9

61

Propazine

1.091

1.26

1.59

36

229.9

16

187.9

24

145.9

32

103.9

62

Bitertanol

1.101

1.29

1.60

16

338.0

16

98.9

10

268.9

10

69.9

63

Chlorpyrifos-oxon

1.101

1.30

1.60

26

333.7

14

277.6

30

197.7

12

305.7

64

Fenarimol

1.101

1.28

1.60

38

330.9

28

80.9

22

267.9

36

138.9

65

Tebuconazole

1.101

1.31

1.60

30

308.0

20

70.0

30

124.8

30

150.9

66

Tebufenozide

1.101

1.27

1.60

10

353.2

20

133.0

6

296.9

44

105.0

67

Terbuthylazine

1.111

1.30

1.61

26

230.0

16

173.8

30

103.9

26

95.9

68

Trifloxystrobin

1.111

1.32

1.61

22

409.0

16

185.9

12

205.9

72

94.9

69

Ametryn

1.151

1.28

1.65

30

228.0

18

185.9

26

95.9

24

90.9

70

Pyraclostrobin

1.151

1.32

1.65

20

387.9

12

193.9

24

163.0

30

148.9

71

Pyrimethanil

1.151

1.32

1.75

46

199.9

24

107.0

24

82.0

36

76.9

72

Ethoprop

1.161

1.35

1.76

24

242.8

14

172.8

20

130.8

28

96.9

73

Temephos

1.161

1.36

1.66

32

466.9

36

124.8

16

404.7

22

418.7

74

Prometryn

1.171

1.37

1.77

38

241.9

22

157.9

18

199.9

30

84.9

75

Ethoxyquin

1.191

1.37

1.79

24

217.9

22

147.9

28

174.0

20

189.7

76

Penconazole

1.201

1.38

1.80

28

283.9

16

70.0

28

158.8

16

172.8

77

Prochloraz

1.201

1.38

1.80

18

375.8

10

307.8

24

70.0

22

85.0

78

Propiconazole

1.201

1.38

1.80

32

341.9

20

69.0

26

158.8

18

204.8

79

Spinosyn Aa

1.251

1.44

1.85

34

732.5

32

142.0

70

97.9

60

98.7

80

Triflumizole

1.251

1.43

1.85

16

345.9

10

277.8

16

73.0

20

55.0

81

Quizalofop-ethyl

1.301

1.46

1.90

34

373.0

18

298.8

28

90.9

26

271.0

82

Thiobencarb

1.301

1.46

1.90

18

257.9

16

124.9

12

99.9

46

89.0

83

Spinosyn Da

1.351

1.52

2.00

32

746.4

28

142.0

70

97.9

58

98.7

84

Fluroxypyr esterb

1.451

1.62

2.05

30

254.8

14

208.8

20

180.8

28

160.8

85

Hexythiazox

1.501

1.65

2.50

20

352.9

16

227.8

24

167.9

14

270.8

86

Pyridaben

1.651

1.88

2.50

20

364.7

26

147.0

12

308.8

44

132.0

87

Fenazaquin

1.851

2.04

2.50

30

307.1

24

57.0

16

161.0

20

146.9

88

Tribufos

1.851

2.10

2.50

28

314.9

14

168.8

22

112.8

12

224.9

89

Methoprene

2.001

2.27

2.60

12

311.1

6

279.0

12

191.0

24

94.9

MRM multiple reaction monitoring, tR retention time, CV collision voltage, CE collision energy

aSpinosad consists of 85% spinosyn A and 15% spinosyn D

bFluroxypyr-1-methylheptyl ester

cAmmonium adduct

dAmmonium adduct precursor of m/z 200.8, CV = 10 V

eAmmonium adduct precursor of m/z 240.0, CV = 12 V

fSodium adduct precursor of m/z 241.9, CV = 26 V

Formic acid (HCO2H) used in the extraction solution was 88% from GFS Chemicals (Columbus, OH: USA), and MS-grade (98%) HCO2H from Fluka (St. Louis, MO; USA) was used in the mobile phase. The ammonium formate (HCO2NH4) salt used in extraction was reagent grade from Sigma-Aldrich (St. Louis, MO; USA), and in the LC mobile phase, it was diluted from 10 M high-purity MS grade solution from Sigma. In sample cleanup experiments, the anhydrous magnesium sulfate (anh. MgSO4), primary secondary amine (PSA), and octadecylsilane (C18) reagents were from UCT (Bristol, PA; USA) and Biotage (Uppsala, Sweden). The pear, tomato, cucumber, eggplant, cilantro, clementine, raspberry, and Brussels sprout samples were all labeled organically-grown and purchased from local food stores.

Instrumentation and Analytical Conditions

A Waters (Milford, MA; USA) Acquity UPLC coupled to a TQD triple quadrupole MS/MS system was used for analysis. Instrument control and data acquisition were conducted through Waters MassLynx software. In the final method, the LC was operated under isocratic conditions with a 3 cm × 4.6 mm i.d. Phenomenex (Torrance, CA; USA) Prodigy C18 (3 µm particles, 100 Å pores) column plus 0.4 cm × 3.0 mm Phenomenex Security C8 guard/pre-column kept at 40 °C. Flow rate was 0.4 mL/min and injection volume was 50 µL. The mobile phase was 10 mM HCO2NH4 (pH 3) in 47.5/47.5/5 (v/v/v) MeCN/MeOH/H2O. A 200 mM aqueous HCO2NH4 solution was first prepared and adjusted to pH 3.0 (± 0.1) with HCO2H using a calibrated pH meter before 50 mL was added to 475 mL each of MeCN and MeOH.

Table 1 lists the final MS/MS conditions for each analyte (and the IS). Dwell times were the instrument minimum of 5 ms in all multiple reaction monitoring (MRM) ion transitions, and inter-dwell times were also 5 ms. Electrospray positive (ESI+) was used for ionization in all cases at 120 °C source and 450 °C desolvation temperatures, 3 kV capillary and 3 V extractor voltages, and 50 L/hr cone and 1000 L/h desolvation nitrogen gas flows. Ultrahigh purity argon was used as the collision gas. The divert valve sent the mobile phase flow to waste prior to 0.85 min and after 2.3 min in the method. The chromatographic system was flushed to waste with 1/1 (v/v) MeOH/MeCN and the spray cones were interchanged and cleaned between the sample sequences of about 80 injections.

Nuclear magnetic resonance (NMR) was used to determine water content in the different extracts from different sample preparation and cleanup procedures. A Varian (Palo Alto, CA; USA) Unity Plus 400 MHz instrument was used in the same procedure and conditions as reported previously [21]. In brief, 550 µL sample plus 100 µL CD3CN (Cambridge Isotopes, Woburn, MA; USA) were combined into NMR tubes, and integrated areas of the hydrogen peak were measured. Calibration standards of different v/v concentrations of H2O in MeCN were analyzed along with the samples to generate quadratic calibration curves of high precision.

Sample Processing and Preparation

Approximately 1 kg each of fruit and vegetable samples (pear, tomato, cucumber, eggplant, cilantro, raspberry, clementine, and Brussels sprout) were cut with a knife into ≈ 1.5 cm3 chunks and placed into a covered container in a -20 °C freezer for > 2 h. Each frozen sample was comminuted with pellets of dry ice in cold stainless steel bowls that had been just removed from the freezer. A Robotcoupe (Ridgeland, MS; USA) RSI 2Y1 was used to generate a fine powdery sample for later extraction. The procedure was done quickly to minimize the condensation of moisture from the lab atmosphere, and the samples were placed into jars in the freezer. The jars were only loosely sealed for > 1 h to give time for the dry ice to sublime and then tightly sealed until the samples were thawed for extraction.

The formate-buffered QuEChERS version as reported previously [22] was used for sample preparation of the blank and spiked fruit and vegetable samples. In the protocol, 10 g thawed sample was weighed into 50 mL polypropylene centrifuge tubes and 10 mL of MeCN containing 4.4% HCO2H (5%, v/v, of 88% pure reagent) and 5 g HCO2NH4 was added. Reagent blank consisted of 8.7 mL H2O to approximate the amount of moisture of fruits and vegetables. A dispenser was used to speed the solvent addition, and the analysts worked quickly to minimize the absorption of water from the atmosphere to the hygroscopic salt. The sample tubes were tightly sealed and placed in trays on a Glas-Col (Terre Haute, IN; USA) platform shaker (50 samples per batch), were shaken at 80% setting and full pulsation for 5 min. Then, the sample tubes were centrifuged for 3 min at 3711 rcf. No cleanup was conducted in the final method, and 1 mL extract was transferred to autosampler vials. To prepare matrix-matched (MM) and reagent-only (RO) calibration standards, 10 µL each of 10 ng/µL atrazine-d5 and appropriate pesticide mix concentrations in MeCN were added to yield 0, 5, 10, 20, 50, and 100 ng/g concentrations. MeCN/water (1 mL of 94/6, v/v, solution) was used in RO standards in place of matrix extracts to determine MEs.

In separate experiments to evaluate cleanup using dispersive solid-phase extraction (d-SPE) with different sorbents, 10 mL of initial extracts (combined replicates) were added to 15 mL polypropylene centrifuge tubes containing different sorbents: 1.5 g anh. MgSO4 only; 1.5 g anh. MgSO4 plus 500 mg each of PSA and C18; and 500 mg each PSA and C18. To measure the amount of co-extractives gravimetrically, 5 mL replicates of extracts were evaporated to dryness at 40 °C under nitrogen gas flow followed by 110 °C in an oven for 1 h in pre-weighed glass test tubes. The weight differences were measured on an analytical balance to assess cleanup.

Two approaches were conducted to estimate MEs in LC-MS/MS in the different sample preparation methods: (1) slopes of the calibration curves were compared to yield the %difference of the MM from the RO standards; and (2) the MRMs of several pesticides infused post-column into the ion source were monitored as blank matrix extracts were injected using the final LC-MS/MS conditions.

Results and Discussion

Development of LC-MS/MS Conditions

A pre-condition in this study was to keep the analytical cycle time < 3 min from injection from one sample to the next. Another defined limit was that the MS/MS instrument used in the study had a minimum ion transition dwell time and inter-dwell delay of 5 ms each per MRM, or total of 10 ms. Furthermore, at least 2 MRMs are needed per analyte for qualitative identification purposes based on their ion ratios [10, 23, 24], which would mean that at least 20 ms were required per analyte. However, this rapid sample preparation and analysis approach did not provide much selectivity in terms of cleanup or chromatographic separation, and thus relied predominantly on the selectivity provided by the MS/MS analysis. Even in traditional UHPLC-MS/MS analysis, monitoring of at least 3 MRMs helps with analyte identifications due to typical variations in measured ion ratios and likelihood of co-eluting interferences [24]. Since each ion ratio consists of 2 MRMs, the collection of 3 MRMs triples the number of ion ratios per analyte (ions 2 vs. 1, 3 vs. 1, and 3 vs. 2) that can be used to make improved identifications compared to when only 2 MRMs are acquired. Although the additional ion ratios help reduce the chances of false positives and negatives, this also extended the minimum data collection time per analyte to 30 ms in this study. Feasibility of the concept could still be demonstrated, but use of a faster and more sensitive instrument would have been helpful in this study.

The number of data points needed to detect a chromatographic peak is another limiting factor in the rapid analysis. Studies have shown that a minimum of 5–6 points across peaks yield a reasonable precision with RSD < 5%, and other factors limited RSD to ≈ 3% when ≥ 6 points formed the peak [25]. Thus, at least 6 points per peak were sought in this method, and the 30 ms data collection time limited the number of co-eluting targeted analytes possible in the method depending on peak widths.

A series of calculations with respect to data acquisition times, chromatographic peak widths, points per peak, MRMs/analyte, and number of simultaneous analytes was conducted using Excel, and many plots were generated. Figure 1a presents the plot that is most pertinent to the limits in this study. Due to the 30 ms data collection time per analyte, the only way to maximize the number of analytes monitored was to broaden chromatographic peak widths, which unfortunately also reduced separation efficiency and selectivity of analysis. Thus, to cover ≈ 100 co-eluting analytes to meet typical pesticide residue monitoring norms, the peak widths had to be ≈ 20 s, as used in the calculations shown in Fig. 1a (which is valid for FI-MS/MS, too).

Fig. 1

a Plot of the number of data acquisition points per co-eluting 20 s analyte peak using 10 ms per ion transition (MRM) depending on the number of MRMs per analyte, and b actual number of data acquisition points along the chromatogram per 16–21 s peak in the final method using 3 MRMs each of 10 ms for the 88 analytes plus internal standard

A list of 219 pesticides was provided by the EPA with prioritization for enforcement monitoring and risk assessment purposes, as previously described [26]. Among these, 130 were amenable for analysis by both LC and GC, 52 were only GC-amenable (mostly organochlorines), and 37 were LC-amenable only. A choice could have been made in this study to only monitor for those 37 targeted analytes by rapid LC-MS/MS and leave the remaining 182 for GC-MS(/MS) only, but overlapping analyses by orthogonally-selective methods provides even better quality of results and identification within the overall monitoring scope [23]. The larger goal was to test feasibility of this rapid approach to meet real-world needs, thus as many of the 130 LC- and GC-amenable pesticides were included in this isocratic LC-MS/MS method as possible within the given constraints.

Actually, FI-MS/MS was initially evaluated in this study for < 1 min analysis, but the MS/MS instrument used was not sensitive enough, especially when final extracts were diluted in an attempt to overcome ion suppression due to MEs [27]. Dilution of final extracts by factors > 100-fold normally overcome the MEs [28], but this also requires the most sensitive instrumentation available. Thus, a compromise was made to instead evaluate < 3 min isocratic LC-MS/MS analysis using the older instrument available for this study.

To achieve < 3 min analysis while yielding ≈ 20 s peak widths, a short and wide analytical LC column with large sample loading capacity had to be utilized. The guard and analytical column pair listed in Experimental served this purpose, and 50 µL injection led to 16–21 s peak widths within the 3 min limit. Fluroxypyr-1-methylheptyl ester was the last LC-only pesticide to elute from the column, but the insect growth regulator, methoprene, is better analyzed by LC-MS/MS than GC-MS/MS due its low mass base peak of m/z 73 when using electron ionization [16]. Thus, the retention time (tR) plus remaining peak of methoprene dictated the time of analysis in this method.

A series of experiments was conducted to assess tR of methoprene depending on isocratic mobile phase conditions, leading to the final choice given in Experimental with methoprene tR of 2.3 min and total analysis time of 2.6 min. The inclusion of an equal volume mixture of MeOH and MeCN as the organic component in the mobile phase has been shown to be beneficial in multiresidue pesticide analysis by LC-MS/MS [10, 26], and the inclusion of 10 mM HCO2NH4 (at pH 3) provided an excess source of \({\text{NH}}_{4}^{+}\) to help in the analysis of those pesticides that form adducts. The 5% water content of the mobile phase was also designed to closely match the water content of the injected final extracts, which was another consideration when choosing the final conditions. Slightly more water content in the mobile phase would have better dispersed the analytes along a 3 min chromatogram, but ghost peaks from more strongly retained matrix components would have been worse. Song et al. found that 30% water in the mobile phase led to > 10 min chromatograms for common pesticides in reversed phase LC [29], which does not increase sample throughput or analytical performance compared with gradient elution in UHPLC-MS/MS.

Table 1 lists the ultimate set of 88 analytes (consisting of 87 pesticides due to spinosad being a mixture of two spinosyns), plus the IS (atrazine-d5), listed in order of MRM start times in the final method. Measured peak widths proportionally grew from 16 s to 21 s from earlier to later eluting analytes. Table 1 lists the start and stop times of the 3 MRM ion transitions for each analyte, which were devised to incorporate the chromatographic peaks via calculations based on their tR and peak widths. Figure 1b further displays the number of points per peak along the chromatogram depending on the segments given in Table 1. Considering the ≈ 0.3 min peak widths, analytes that elute at tR ≈1.3 min are detected with 5–8 data points for each ion transition, and analytes with shorter and longer tR have > 10 points per peak, albeit the points are not always equidistant across the peaks.

Sample Extraction and Salt-out Partitioning

Nonvolatile salts such as NaCl and MgSO4, pose problems in MS analysis due to their build-up on surfaces after evaporation of the solvent droplets in ESI. Volatile salts such as HCO2NH4, however, remain a gas at typical ion source operating temperature and are pumped away. Normally in LC-MS, the highly polar salts injected in the sample, independent of their volatility, pass unimpeded through reversed-phase LC columns, which allows them to be diverted to waste prior to the tR of the first analyte to reach the MS detector. When using FI-MS, however, any co-extracted salts in the injected extracts will be introduced into the MS ion source along with all analytes present because no chromatography is being done to separate them. Thus, only volatile salts should be used in FI-MS, and this is also the case for fast isocratic LC-MS in which the first-to-elute analytes overlap with unretained salts.

Although only a small amount of salts are co-extracted in the final QuEChERS approach using NaCl and MgSO4 [30], the concern is avoided by using a volatile salt in QuEChERS. Nanita et al. used moderately volatile NH4Cl in QuEChERS for a limited number of pesticide analytes in FI-MS/MS [19], and González-Curbelo et al. found that the more volatile HCO2NH4 worked well to salt-out a wide range of pesticide analytes for both LC- and GC-MS/MS [22]. Han et al. also demonstrated the use of HCO2NH4 in QuEChERS for hundreds of pesticides and environmental contaminants in foods [10].

Another option that avoids salts altogether involves the separation of the MeCN extract from the water in the sample via cold-partitioning [31]. However, this approach takes longer due to refrigeration time, and the partitioning of pesticide analytes into the upper phase has yet to be sufficiently evaluated. Thus, the extensively validated formate-buffered QuEChERS method [22] was employed in this study.

A common mistake in QuEChERS by analysts is to assume that the initial extract after partitioning from the aqueous phase equals the volume of MeCN added to the sample. This leads to a bias in the result if g/mL equivalent sample in final extracts is not correct or an IS is not employed properly to account for the volumetric difference. When using anh. MgSO4 in the post-extraction d-SPE cleanup step [21], the reduction of water in the MeCN extract leads to ≈ 1 g/mL equivalent sample in the final QuEChERS extracts. This is indicated in Table 2, listing results from an experiment using NMR, which showed that the separated extracts contain ≈ 3% water after the addition of anh. MgSO4 in d-SPE. Moreover, the finding that the initial fruit and vegetable extracts contained ≈ 6% water, which was reduced to 5% in the autosampler vial by addition of MeCN to compensate for the preparation of calibration standards, was one of the reasons that the isocratic mobile phase was prepared to contain 5% water.

Table 2

%Water in the original samples, initial formate-buffered QuEChERS extracts, and in the final extracts using different sorbents in d-SPE cleanup (150 mg anh. MgSO4 and/or 50 mg each PSA and C18 per mL initial extract)

Sample

%Water in sample

%Water in initial extract

%Water after d-SPE with PSA+ C18 only

%Water after d-SPE with anh. MgSO4 only

%Water after d-SPE with anh. MgSO4 + PSA + C18

Reagent blank

87

6.3

6.0

2.4

3.3

Raspberry

86

6.0

5.7

2.7

3.5

Brussels sprout

82

5.8

5.5

2.6

3.0

Clementine

86

6.0

5.7

2.9

3.3

If the extract is not dried, however, then the volume of the upper phase after the partitioning step must be known to calculate the equivalent sample concentration in the extract. Figure 2 presents this situation using the extraction conditions in this study with respect to moisture of the extracted sample. For water content from 5 to 8 mL (50–80% in the 10 g sample), the upper phase averaged ≈ 8.3 mL volume when extracting with 10 mL extraction solvent, which led to a ≈ 1.2 g/mL equivalent sample in those cases (or 20% concentration “boost”). Nanita et al. also used this concentration effect to lower detection limits in FI-MS/MS analysis of pesticides in water [19]. From 80 to 100% sample water content, the upper layer volume started to decrease significantly as shown in Fig. 2. Thus, fruits and vegetables with ≈ 90% moisture yielded ≈ 1.27 g/mL equivalent sample, and this had to be taken into account when determining analyte concentrations and recoveries. Even when using internal standards, the equivalent sample amounts need to be known when preparing calibration standards or converting analyte concentrations from ng/mL in solutions to ng/g in samples. If this is not done properly, 20–30% positive bias will occur in the results for typical watery food samples.

Fig. 2

Volume of initial MeCN extract in the formate-buffered QuEChERS method depending on the %water in the extracted sample (left axis, blue line) and the associated concentration factor (recovery compensation) of the extract (right axis, orange line)

Cleanup and Matrix Effects

The most critical limitation in this approach relates to MEs in ESI-MS analysis, especially due to ion suppression [27, 28]. Studies of MEs in UHPLC-MS/MS of QuEChERS extracts have shown that cleanup is not necessary except for the most complex fruits and vegetables [32], but even in those cases, the cleanup steps are not very effective. The components that cause the most ion suppression tend to be highly polar, which elute early in the chromatograms at highly aqueous mobile phase conditions. These components do not typically co-elute with pesticide analytes to pose problems in traditional LC-MS, but in this rapid LC-MS/MS approach, they create a serious challenge in the analysis.

Several experiments were done to assess MEs in this study, as presented in Figs. 3, 4, 5. First of all, gravimetric analysis of extracts before and after d-SPE cleanup, as shown in Fig. 3, found that significant nonvolatile material remains in final extracts even after cleanup steps using different sorbents. In some cases, the reagents used in the d-SPE step introduced more mass than matrix components that were removed. It is likely that small particles were transferred from the extracts to the weighed test tubes in those instances, but this also indicates a known drawback of d-SPE in that care is needed when making transfers of the final extracts to the autosampler vials. Furthermore, d-SPE is designed for rapid “just enough” cleanup, which is sufficient in traditional analysis, but it is not as thorough at removing matrix components as SPE using cartridges [32]. An estimation is provided in the bottom half of Fig. 3 showing the mass of nonvolatile material that is made in each 50 µg injection, which ranges from 17 to 66 µg. Usually, < 20 µL of final extract is injected even when using this older instrument in UHPLC-MS/MS, and 100-fold less equivalent sample are needed for analysis using state-of-the-art instruments, but the loss of sensitivity due to severe MEs, as shown in Figs. 4 and 5, required a larger injection volume to better meet analyte identification criteria at 10 ng/g.

Fig. 3

Co-extracted material in the initial and final extracts of different matrices using the formate-buffered QuEChERS method without and with d-SPE cleanup using different sorbents. The associated amounts of nonvolatile material injected per final extract appears below each matrix type with respect to the cleanup sorbents used

Fig. 4

Matrix effects of propiconazole infused throughout the chromatograms for a reagent blank, b raspberry, c Brussels sprout, and d clementine after cleanup or not using d-SPE with different sorbents

Fig. 5

Matrix effects on atrazine-d5 (internal standard) in raspberry final extracts to which 12 or 144 pesticide analytes were added at different concentrations as in matrix-matched calibration

The traces shown in Fig. 4 indicate the degree of MEs depending on cleanup throughout the chromatogram in the method. The observed pulses in the traces originate from the step-wise pulsation of analyte flow (propiconazole in this case) by the post-column infusion pump. Although data smoothing avoids the appearance of the pulses in the traces, it also broadens the effect, and the data was not smoothed in Fig. 4 to allow better comparison of MEs with tR of the analytes in Table 1. Comparison of the results shown in Figs. 3 and 4 give an idea of what type of components are being removed (or added) during the initial extraction and d-SPE cleanup using different sorbents. The unretained components begin to appear at 0.85 min, and typically induce the most ion suppression at 1.0–1.2 min.

Different chromatographic profiles in the MEs occur depending on the different reagents and commodities. Brussels sprout (Fig. 4c) caused the greatest extent of ion suppression, consistently reaching − 95% even after d-SPE cleanup, and extending all the way to the end of the chromatogram. Raspberry (Fig. 4b) induced the least MEs, which were estimated by triangulated area beneath the 0% baseline to be ≈ 44% less than Brussels sprout. Clementine (Fig. 4d) caused a curious double dip suppression effect, which makes sense considering the type of ion suppression profiles observed in similar experiments using LC-MS/MS of QuEChERS extracts of citrus fruits [33]. The co-extracted components causing the second dip at 1.3 min co-elute with many pesticide analytes even when using UHPLC-MS/MS with high separation efficiencies.

Unfortunately, chemical reagents used in sample preparation (Fig. 4a) are substantial contributors underneath matrix effect profiles observed for the injected fruit and vegetable extracts. For example, MgSO4 + PSA + C18 in d-SPE of the reagent blank yielded ≈ 85% of the estimated triangulated area compared with the raspberry extract undergoing similar treatment. Meanwhile, the reagent blank without cleanup gave ≈ 1/3 less overall matrix effect by area. Higher purity extraction reagents would have probably reduced the MEs, but cost of implementation in this method would have been prohibitive. Despite that MgSO4 + PSA + C18 increased MEs in reagent blanks, the traces for the commodities in Fig. 4a–c demonstrate removal of highly polar matrix components that induce greater ion suppression at 0.85–1.0 min. Otherwise, very little difference can be observed in the traces between 1.0 and 2.6 min for each commodity after undergoing d-SPE cleanup. Although the experiments demonstrated that d-SPE cleanup helped reduce the amount of co-extracted matrix components from produce extracts and the MEs overall, the extent of cleanup was too small to justify the additional step in this type of rapid analysis.

Another commonly unappreciated issue in quantitative analysis by ESI-MS is the additional MEs caused by analytes with each other in multiresidue methods. The mechanism of ion suppression works the same whether the cause is due to co-eluting matrix components or analytes at high concentrations [27]. Method validation studies usually involve spiking all analytes at the same concentration levels, and calibration standards are prepared in the same way, both of which induce MEs especially when all analytes are at high concentrations. However, real samples rarely contain more than a few pesticides, and concentrations are typically less than 100 ng/g [1]. High level calibration standards in which many analytes co-elute can cause suppression and may lead to a high bias in the result when only a single analyte is present in the sample. This situation is exacerbated in the rapid isocratic LC-MS/MS approach because more analytes co-elute.

An experiment was conducted using the final method in which MEs on the IS (atrazine-d5) at 100 ng/g equivalent concentration were compared for different MM calibration standards in raspberry extracts. The results in Fig. 5 show how the blank matrix induced − 59% signal suppression, and the addition of 144 pesticides at 1, 10, and 100 ng/g equivalent concentrations to the raspberry extract caused additional (not relative) suppression of − 4%, − 18%, and − 23%, respectively. Even the presence of just 12 pesticides at 100 ng/g caused − 17% more matrix effect. If normalization of results to the IS does not compensate for these MEs for all analytes [33], then a bias in the results is introduced even in recovery experiments in which spiked pesticides occur in the final extracts at lower concentrations than some of the high level mixed pesticide standards used for calibration.

Validation Results

The final method was investigated using a similar regulatory validation protocol as previously [10, 24, 26]. The quality of method performance with respect to both qualitative identification and quantitative determination of the analytes in different matrices were assessed at 0 (blank), 10, and 100 ng/g spiking levels (n = 10 each). Six-point each of MM and RO calibration standards were prepared and analyzed at the beginning and end of each sequence. Peak areas of the quant. ion (see Table 1) were normalized to atrazine-d5 in the same injection, and the MM standards were used to generate a linear forced-zero best-fit equation for quantification. Table 3 provides the results of the 3-day experiment involving water and pear (Day 1), tomato and cucumber (Day 2), and eggplant and cilantro (Day 3). The complexity of the matrices and co-extractives increased as the experiment progressed.

Table 3

%Recoveries (%RSD) and number of identifications out of 10 spikes each at *10 ng/g and 100 ng/g in fast isocratic LC-MS/MS analysis of the analytes (listed in 4 groups of 22 according to tR) in different matrices with normalization to the internal standard (atrazine-d5)

Text in italics indicates recoveries (R) < 70% or > 120% and RSD > 25%. Bold text indicates > 10% false negatives, and boxed results meet validation criteria. The number of analytes meeting validation criteria for the given column for each group of 22 appear in the group rows

*n = 10 (max) at 10 ng/g n = 10 (max) at 100 ng/g

1Spinosad consists of 85% spinosyn A and 15% spinosyn D; 2 fluroxypyr-1-methylheptyl ester

Internal standard at 100 ng/g without normalization, n = 3

Figure 6 displays examples of ion chromatograms for 4 of the analytes at 10 ng/g in tomato. These organophosphorus pesticide metabolites are not as commonly monitored as their parent forms in multiresidue analysis, but they usually remain part of the regulatory definition due to their toxicity. They are quite polar and are better analyzed by LC than GC methods. Only parathion-oxon could not be identified in this example, which construed one of its 4 false negatives at this level in tomato, as indicated in Table 3.

Fig. 6

Fast isocratic LC-MS/MS analysis of example pesticides spiked at 10 ng/g in tomato, including a false negative for parathion-oxon

In terms of identification, the MS/MS regulatory criteria established by the USDA and FDA were used as in previous studies [10, 23, 24]. Namely, an analyte is said to be identified if: (1) the tR falls within ± 0.1 min of the reference tR for the analyte; (2) one ion ratio is within ± 10% or two ion ratios are within ± 20% of the reference ion ratios (absolute differences, not relative); and (3) the determined analyte concentration must be greater than the regulatory action level (or 10 ng/g desired reporting level in this case). The tR was not a factor in this study because MRM segments rather trimly covered the expected peak widths, and extraneous peaks beyond ± 0.1 min were incomplete and could not be integrated by the instrument software. Therefore, ion ratio differences became the most important factor in identification. Reference ion ratios were set contemporaneously in each sequence from the averages of 50–100 ng/g equivalent RO calibration standards using integrated peak areas for each MRM ion transition.

An Excel spreadsheet was developed to automatically make identifications of the analytes based on the ion ratios meeting the given criteria above. In an evaluation of the resulting rates of false positives and negatives, different identification threshold settings were compared. An “ion ratio point” was granted for each ion ratio that met the ± 10% criterion or when two ion ratios met the ± 20% criterion. A maximum of 6 points was awarded when all three ion ratios fell within ± 10% of the reference ion ratios, which also met the ± 20% criteria for all three pairs of ion ratios. If the tR and reporting level criteria are also met, regulatory identification of an analyte by traditional GC- or LC-MS/MS analysis only requires 1 ion ratio point to be attained. However, if these normal criteria yield too many false positive identifications by the rapid LC-MS/MS method, then excessive unnecessary confirmatory analyses defeats the purpose of this streamlined high-throughput concept.

Without considering determined concentration as a factor, use of an identification threshold of 1 ion ratio point led to a rate of 1.4–3.1% “false positives” for the analytes in the 6 matrices with n = 880 (10 replicates and 88 analytes in each matrix). False identifications ranged from 0.1 to 1.7%, 0.0–1.4%, and 0.0–1.1% when using 2, 3, and 4-point thresholds, respectively, independent of concentration. However, nearly all of these were < 1 ng/g, and use of a 1-point threshold merely led to 6 instances with concentration ≥ 10 ng/g: parathion-oxon at 26 ng/g in one out of 10 cucumber blanks; diflubenzuron at 28 and 107 ng/g in an eggplant and cilantro blank, respectively; atrazine-desethyl at 18 ng/g in a cilantro blank; and azinphos-methyl at 16 and 19 ng/g in two cilantro blanks. Using a 2-point threshold eliminated all of these “false positives” except one of the azinphos-methyl identifications, which disappeared when using a threshold of 3 ion ratio points.

The same evaluation was done with respect to false negatives in the different matrices spiked at 10 and 100 ng/g, and Fig. 7 shows the results for tomato and cucumber (note that these results also involve sample preparation, not just the analysis step). The rates of false negatives increase in an almost linear relationship as the identification threshold grows from 1 to 4 points. Since very few false positives occur when using the normal regulatory MS-based identification criteria in this method, and false negatives increase significantly as more stringent criteria are set, the decision was made to keep the regulatory identification criteria unchanged by setting a 1-point ion ratio threshold in this rapid LC-MS/MS method. Figure 8 exhibits the rates of true identifications using these criteria in the validation study for the different commodities, and Table 3 presents the results for the individual analyte/matrix pairs.

Fig. 7

Rates of false negatives when setting different ion ratio thresholds for analyte identifications for the 88 analytes spiked into tomato and cucumber at different levels, n = 880 (10 for each analyte/matrix and level)

Fig. 8

Rates of correct identifications in the final method for the 88 analytes spiked (or not) into the different matrices at different levels, n = 880 (10 for each analyte/matrix and level)

Bold text is used in Table 3 to designate when the rate of false negatives was > 10% for the given pesticide, level, and matrix, and italics denotes when recoveries were < 70% or > 120% and RSDs > 25%. The order of analytes is the same in Tables 1 and 3, which closely relates to tR. To better evaluate the results both qualitatively and quantitatively, consideration should be given to points per peak shown in Fig. 1b and MEs provided in Fig. 4. In Table 3, the 88 analytes are divided into 4 groups of 22: Group 1 consists of analytes with tR ≈ 0.9–1.1 min (acephate to atrazine-desethyl); Group 2 covers tR ≈ 1.1–1.2 min (carbendazim to flubendiamide); Group 3 has tR ≈ 1.2–1.3 min (tetraconazole to terbuthylazine); and Group 4 with tR ≈ 1.3–2.3 min incorporates trifloxystrobin to methoprene. The numbers of analytes that meet the validation criteria signified in normal font in each column is listed per group. The results that meet all criteria are boxed within the table.

A clear relationship occurred in improved quality of results as the tR increased, which can be discerned from the compiled results by group in Table 3. Only ethoxyquin did not meet all validation criteria in multiple matrices among Group 4 analytes, whereas in Group 1, only dicrotophos, dimethoate, and phosmet-oxon gave acceptably quantitative and qualitative results in at least two of the food commodities. Of course, the results also worsened as the matrices became more complex, and this approach was shown to be more useful for simple fruits and vegetables. However, direct chemical matrix interferences were not much of a problem in this approach, as evidenced by the few false positives listed above and merely 3 “interf.” shown in Table 3 (atrazine in cilantro and penconazole in eggplant and cilantro). The number of data points per chromatographic peak were also not a limiting concern in the approach because many Group 2 pesticides determined from peaks with 5–7 points gave similar RSDs as Group 3 and 4 analytes with > 10 points per peak. Thus, ion suppression due to MEs were a major source of difficulties in the method, and one path to achieve better results for a wider polarity range of pesticides in the method would be to reduce polar co-extractives by using higher purity reagents and/or more effective cleanup. Independent of sample preparation, use of a more sensitive MS detector would surely have allowed dilution of extracts to yield better results, and greater selectivity and faster data acquisition of modern instruments would also help.

From a broad perspective, all quantitative and qualitative validation criteria were met in all matrices and levels for only 7 pesticides: triazophos, tebuconazole, pyraclostrobin, prometryn, prochloraz, propiconazole, and triflumizole. From another perspective, it is surprising that the “quick and dirty” approach with an older instrument met the criteria for any pesticides. Even though only 13 pesticides met all qualitative and quantitative validation criteria in cilantro, 56 pesticides in the complex commodity could be still be satisfactorily identified at 100 ng/g, which is typically lower than set regulatory limits albeit default limits are 10 ng/g for unregistered pesticides. The results improved significantly for the less complex matrices, with 28, 42, 47, 50, and 49 pesticides meeting all validation criteria for eggplant, cucumber, tomato, pear, and water, respectively. In all, 22 pesticides on the target list were not acceptably extracted, determined, and/or identified by the method in at least one matrix, but only novaluron did not meet the < 10% false negative rate at 100 ng/g in any matrix. However, it was still correctly identified 70–80% of the time in 100 ng/g spikes in all but the eggplant and cilantro, which is better than 100% false negatives if it is not monitored.

Conclusion

Although this high-throughput approach does not match the method performance capabilities of current (U)HPLC-MS/MS methods, it does acceptably meet both quantitative and qualitative regulatory validation criteria for many pesticides in common commodities. Feasibility for screening by MS/MS identification at regulatory levels of concern was demonstrated, and key factors for improvements have been pinpointed: (1) use of modern MS instruments would allow dilution of extracts to reduce MEs while also leading to faster data acquisition for narrower peaks, inclusion of more pesticides, and greater LC and MS selectivity; (2) use of higher purity reagents and/or salt-free (cold) extract phase partitioning would reduce the amount of ion suppression in ESI-MS; (3) use of automated and/or online SPE or other forms of efficient and effective cleanup [10, 26, 30, 32] would also reduce MEs; and (4) use of < 5 min rather than < 3 min analyses via different isocratic LC mobile and stationary phase conditions would still triple sample throughput compared to current UHPLC gradient elution methods.

In the newly developed method, the sample throughput was 23 analyses/hour, and a single analyst could also extract a batch of at least 20 pre-comminuted samples within an hour. Summation integration of chromatographic peaks and automatic identification of targeted analytes without human review also enables high throughput data management [26, 32], albeit the particular instrument software used in this study did not (yet) possess that feature. Sample comminution remains the current limiting factor in being able to conduct higher throughput in real-world, start-to-finish multiresidue monitoring for pesticides in food commodities (and environmental solids) [2], but this isocratic fast LC-MS/MS concept is an important step toward the overall goal.

Notes

Acknowledgements

The author thanks Robyn Moten for technical assistance in the laboratory and Gary Strahan for NMR analysis of the extracts to determine water content.

Disclaimer

Mention of brand or firm names does not constitute an endorsement by the U.S. Department of Agriculture above others of a similar nature not mentioned.

Compliance with Ethical Standards

Conflict of interest

The author declares no conflict of interest.

References

  1. 1.
    US Government Accountability Office (GAO). Food Safety FDA and USDA should strengthen pesticide residue monitoring programs and further disclose monitoring limitations. GAO-15-38. 2014, pp 105Google Scholar
  2. 2.
    Lehotay SJ, Chen Y (2018) Hits and misses in research trends to monitor contaminants in foods. Anal Bioanal Chem 410:5331–5351CrossRefGoogle Scholar
  3. 3.
    Stachniuk A, Fornal E (2016) Liquid chromatography-mass spectrometry in the analysis of pesticide residues in food. Food Anal Methods 9:1654–1665CrossRefGoogle Scholar
  4. 4.
    González-Curbelo M, Socas-Rodríguez B, Herrera-Herrera AV, González-Sálamo J, Hernández-Borges J, Rodríguez-Delgado M (2015) Evolution and applications of the QuEChERS method. Trends Anal Chem 71:169–185CrossRefGoogle Scholar
  5. 5.
    World Trade Organization (2018), International Trade Statistics 2015. www.wto.org/english/res_e/statis_e/its2015_e/its2015_e.pdf. Accessed July 2018
  6. 6.
    Strategic Consulting (2014) Food Contract Lab Report. www.strategic-consult.com/product/food-contract-lab-report. Accessed July 2018
  7. 7.
    Ferrer C, Lozano A, Uclés S, Valverde A, Fernández-Alba AR (2017) European Union proficiency tests for pesticide residues in fruit and vegetables from 2009 to 2016: overview of the results and main achievements. Food Control 82:101–113CrossRefGoogle Scholar
  8. 8.
    Villaverde JJ, Sevilla-Morán B, López-Goti C, Alonso-Prados JL, Sandín-España P (2016) Trends in the analysis of pesticide residues to fulfil the European regulation (EC) No. 1107/2009. Trends Anal Chem 80:568–580CrossRefGoogle Scholar
  9. 9.
    Raina-Fulton R (2015) New trends in pesticide residue analysis in cereals, nutraceuticals, baby foods, and related processed consumer products. J AOAC Int 98:1163–1170CrossRefGoogle Scholar
  10. 10.
    Han L, Sapozhnikova Y, Lehotay SJ (2016) Method validation for 243 pesticides and environmental contaminants in meats and poultry by tandem mass spectrometry coupled to low-pressure gas chromatography and ultrahigh-performance liquid chromatography. Food Control 66:270–282CrossRefGoogle Scholar
  11. 11.
    Senyuva HZ, Gokmen V, Sarikaya EA (2015) Future perspectives in Orbitrap™-high-resolution mass spectrometry in food analysis: a review. Food Addit Contam A 32:1568–1606CrossRefGoogle Scholar
  12. 12.
    Wang J, Chow W, Chang J, Wong JW (2017) Development and validation of a qualitative method for target screening of 448 pesticide residues in fruits and vegetables using UHPLC/ESI Q-Orbitrap based on data-independent acquisition and compound database. J Agric Food Chem 65:473–493CrossRefGoogle Scholar
  13. 13.
    Uclés S, Uclés A, Lozano A, Martínez-Bueno MJ, Fernández-Alba AR (2017) Shifting the paradigm in gas chromatography mass spectrometry pesticide analysis using high resolution accurate mass spectrometry. J Chromatogr A 1501:107–116CrossRefGoogle Scholar
  14. 14.
    Mol HGJ, Tienstra M, Zomer P (2016) Evaluation of gas chromatography - electron ionization - full scan high resolution Orbitrap mass spectrometry for pesticide residue analysis. Anal Chim Acta 935:161–172CrossRefGoogle Scholar
  15. 15.
    Sapozhnikova Y, Lehotay SJ (2015) Review of recent developments and applications in low-pressure (vacuum outlet) gas chromatography. Anal Chim Acta 899:13–22CrossRefGoogle Scholar
  16. 16.
    Lee J, Kim L, Shin Y, Lee J, Lee J, Kim E, Moon JK, Kim JH (2017) Rapid and simultaneous analysis of 360 pesticides in brown rice, spinach, orange, and potato using microbore GC-MS/MS. J Agric Food Chem 65:3387–3395CrossRefGoogle Scholar
  17. 17.
    Nanita SC, Kaldon LG (2016) Emerging flow injection mass spectrometry methods for high-throughput quantitative analysis. Anal Bioanal Chem 408:23–33CrossRefGoogle Scholar
  18. 18.
    Nanita SC (2011) High-throughput chemical residue analysis by fast extraction and dilution flow injection mass spectrometry. Analyst 136:285–287CrossRefGoogle Scholar
  19. 19.
    Nanita SC, Kaldon LG, Bailey DL (2015) Ammonium salting out extraction with analyte preconcentration for sub-part per billion quantitative analysis in surface, ground and drinking water by flow injection tandem mass spectrometry. Anal Methods 7:2300–2312CrossRefGoogle Scholar
  20. 20.
    Mol HGJ, van Dam RCJ (2014) Rapid detection of pesticides not amenable to multi-residue methods by flow injection-tandem mass spectrometry. Anal Bioanal Chem 406:6817–6825CrossRefGoogle Scholar
  21. 21.
    Anastassiades M, Lehotay SJ, Štajnbaher D, Schenck FJ (2003) Fast and easy multiresidue method employing acetonitrile extraction/partitioning and “dispersive solid-phase extraction” for the determination of pesticide residues in produce. J AOAC Int 86:412–431Google Scholar
  22. 22.
    González-Curbelo M, Lehotay SJ, Hernández-Borges J, Rodríguez-Delgado M (2014) Use of ammonium formate in QuEChERS for high-throughput analysis of pesticides in food by fast, low-pressure gas chromatography and liquid chromatography tandem mass spectrometry. J Chromatogr A 1358:75–84CrossRefGoogle Scholar
  23. 23.
    Lehotay SJ, Sapozhnikova Y, Mol HGJ (2015) Current issues involving screening and identification of chemical contaminants in foods by mass spectrometry. Trends Anal Chem 69:62–75CrossRefGoogle Scholar
  24. 24.
    Lehotay SJ, Mastovska K, Lightfield AR, Nuñez A, Dutko T, Ng C, Bluhm L (2013) Rapid analysis of aminoglycoside antibiotics in bovine tissues using disposable pipette extraction and ultrahigh performance liquid chromatography - tandem mass spectrometry. J Chromatogr A 1313:103–112CrossRefGoogle Scholar
  25. 25.
    Dallüge J, Vreuls RJJ, van Iperen DJ, van Rijn M, Brinkman UATh (2002) Resistively heated gas chromatography coupled to quadrupole mass spectrometry. J Sep Sci 25:608:614Google Scholar
  26. 26.
    Sapozhnikova Y (2018) Development and validation of a semi-automated high-throughput analytical method for 265 pesticides and environmental contaminants in meats and poultry. J Chromatogr A.  https://doi.org/10.1016/j.chroma.2018.08.025 Google Scholar
  27. 27.
    Stahnke H, Kittlaus S, Kempe G, Hemmerling C, Alder L (2012) The influence of electrospray ion source design on matrix effects. J Mass Spectrom 47:875–884CrossRefGoogle Scholar
  28. 28.
    Stahnke H, Kittlaus S, Kempe G, Alder L (2012) Reduction of matrix effects in liquid chromatography-electrospray ionization-mass spectrometry by dilution of the sample extracts: how much dilution is needed? Anal Chem 84:1474–1482CrossRefGoogle Scholar
  29. 29.
    Song S, Zhu K, Han L, Sapozhnikova Y, Zhang Z, Yao W (2018) Residue analysis of 60 pesticides in red swamp crayfish using QuEChERS with high-performance liquid chromatography-tandem mass spectrometry. J Agric Food Chem 66:5031–5038CrossRefGoogle Scholar
  30. 30.
    Han L, Matarrita J, Sapozhnikova Y, Lehotay SJ (2016) Evaluation of a recent product to remove lipids and other matrix co-extractives in the analysis of pesticide residues and environmental contaminants in foods. J Chromatogr A 1449:17–29CrossRefGoogle Scholar
  31. 31.
    Shao G, Agar J, Giese RW (2017) Cold-induced aqueous acetonitrile phase separation: a salt-free way to begin quick, easy, cheap, effective, rugged, safe. J Chromatogr A 1506:128–133CrossRefGoogle Scholar
  32. 32.
    Lehotay SJ, Han L, Sapozhnikova Y (2016) Automated mini-column solid-phase extraction cleanup for high-throughput analysis of chemical contaminants in foods by low-pressure gas chromatography—tandem mass spectrometry. Chromatographia 79:1113–1130CrossRefGoogle Scholar
  33. 33.
    Kwon H, Lehotay SJ, Geis-Asteggiante L (2012) Variability of matrix effects in liquid and gas chromatography—mass spectrometry analysis of pesticide residues after QuEChERS sample preparation of different food crops. J Chromatogr A 1270:235–245CrossRefGoogle Scholar

Copyright information

© This is a U.S. government work and not under copyright protection in the U.S.; foreign copyright protection may apply 2018

Authors and Affiliations

  1. 1.US Department of Agriculture, Agricultural Research ServiceEastern Regional Research CenterWyndmoorUSA

Personalised recommendations