Development and validation of two multiresidue liquid

Development and validation of two multiresidue liquid chromatography tandem mass spectrometry methods based on a versatile extraction procedure for isolating non-steroidal anti-inflammatory drugs from bovine milk and muscle tissue Alessandra Gentili, Fulvia Caretti, Simona Bellante, Lucia Mainero Rocca, et al. Analytical and Bioanalytical Chemistry ISSN 1618-2642 Anal Bioanal Chem DOI 10.1007/s00216-012-6231-0

1 23

Your article is protected by copyright and all rights are held exclusively by SpringerVerlag. This e-offprint is for personal use only and shall not be self-archived in electronic repositories. If you wish to self-archive your work, please use the accepted author’s version for posting to your own website or your institution’s repository. You may further deposit the accepted author’s version on a funder’s repository at a funder’s request, provided it is not made publicly available until 12 months after publication.

1 23

Author's personal copy Anal Bioanal Chem DOI 10.1007/s00216-012-6231-0

ORIGINAL PAPER

Development and validation of two multiresidue liquid chromatography tandem mass spectrometry methods based on a versatile extraction procedure for isolating non-steroidal anti-inflammatory drugs from bovine milk and muscle tissue Alessandra Gentili & Fulvia Caretti & Simona Bellante & Lucia Mainero Rocca & Roberta Curini & Alessandro Venditti

Received: 12 March 2012 / Revised: 11 June 2012 / Accepted: 26 June 2012 # Springer-Verlag 2012

Abstract The main difficulties in analysing non-steroidal anti-inflammatory drugs (NSAIDs) in food and biological samples are due to the tight non-covalent interactions established with matrix proteins and the amount of occurring fatty material. The present paper describes an effective Electronic supplementary material The online version of this article (doi:10.1007/s00216-012-6231-0) contains supplementary material, which is available to authorized users. A. Gentili : F. Caretti (*) : S. Bellante : R. Curini : A. Venditti Department of Chemistry, Faculty of Mathematical, Physical and Natural Sciences, “Sapienza” University of Rome, P.le A. Moro no. 5, P.O. Box 34, Posta 62, 00185 Rome, Italy e-mail: [email protected] F. Caretti e-mail: [email protected] S. Bellante e-mail: [email protected] R. Curini e-mail: [email protected] A. Venditti e-mail: [email protected] L. Mainero Rocca Italian Workers’ Compensation Authority (INAIL)—ISPESL Area, Department of Occupational Hygiene, Laboratory of Chemical Agents, Via Fontana Candida 1, 00040 Monte Porzio Catone Rome, Italy L. Mainero Rocca e-mail: [email protected]

extraction procedure able to isolate fifteen NSAIDs (acetaminophen, salicylic acid, ibuprofen, diclofenac, flunixin and its metabolite 5-hydroxy-flunixin, nimesulide, phenylbutazone, meclofenamic acid, tolfenamic acid, meloxicam, carprofen, ketoprofen, naproxen and etodolac) from bovine milk and muscle tissue through two succeeding steps: (a) deproteinisation/extraction with organic solvent, essential to lower the medium dielectric constant and, therefore, to release the analytes from matrix; (b) SPE clean-up on OASIS cartridges. Lipids were easily removed during lowtemperature centrifugations. The advantages of the developed procedure pertain to the efficient removal of the fat substances (very low matrix effect and high recovery yields) and its versatility, since it can be applied both to milk and muscle with few adjustments due to the diversity of the two matrices. Ion-pairing reversed-phase chromatography combined with the negative electrospray detection was able to achieve low detection capabilities (CCβs) for all analytes and, in particular, for diclofenac whose Maximum Residue Limit (MRL) in milk is 0.1 μg kg−1. The methods were validated according to the guidelines of the Commission Decision 2002/657/EC and then applied for a small monitoring study. A number of samples showed traces of salicylic acid (SA), but its occurrence was not ascribed to a misuse of drugs (aspirin, salicylic acid) since SA, accumulating in plants in response to a pathogen attack, may be introduced into the food chain. Keywords NSAIDs . LC-tandem MS . Bovine milk . Bovine muscle tissue

Author's personal copy A. Gentili et al.

Introduction

Table 1 Maximum residue limits (MRLs) [6] and recommended concentrations (RCs) [7] for NSAIDs in bovine milk and muscle tissue

Since the 1970s, the use of non-steroidal anti-inflammatory drugs (NSAIDs) in breeding practices has progressively increased because of their efficacy in relieving pain and inflammation without the severe immunosuppressive and metabolic side effects of corticosteroids [1]. Nowadays, NSAIDs are recommended to treat mastitis in lactating cows, for long-term analgesia, and to control inflammation associated with arthritis, spondylitis and laminitis [2]; they have also been used in conjunction with antibiotics for the cure of acute bovine respiratory disease [1]. NSAIDs are considered “peripheral analgesics” because they can block the biosynthesis of prostaglandins (PGs) by inhibiting cyclooxygenases [1–3]. PGs generated by COX-1 isoform are constantly present in many tissues to impart a variety of physiologic effects such as the protection of gastrointestinal (GI) mucosa and hemostasis [2]. On the other hand, the inducible PGs are released only intermittently by COX-2 isoform to mediate the inflammation [2]. Therefore, the pharmacologic effect of NSAIDs is as much effective and safe as their action is limited to the COX-2 isoform [2]. On the basis of their inhibitory selectivity, NSAIDs are classified in COX non-selective inhibitors and COX-2 selective inhibitors [1–3]. All of them have the potential for adverse side effects (hepatotoxicity, hematopoietic and renal troubles, allergic reactions), but the group of COX nonselective inhibitors is the most toxic [1–3]. The main adverse effects relate to chemical irritation of the mucosa and the GI ulceration resulting from the inhibition of the constitutive PGs [2]. For these reasons, the incidental introduction of NSAID residues into the human food chain should be avoided and reliable analytical methods for their simultaneous determination in animal food products should be developed. Up to now, the US Food and Drug Administration has not approved the use of NSAIDs in animal husbandry with the only exception of the flunixin meglumine [4]. On the contrary, in the European Union, these drugs have been regulated since 1990 [5] through different laws, recently reorganised within the Commission regulation (EU) no. 37/2010 [6] where, in relation to the toxicity of their residue, some of them are licensed whereas others have a fixed MRL. For a few unauthorised NSAIDs, the Community Reference Laboratories have established Recommended Concentrations (RCs) in milk and meat [7]. Finally, no regulation has been set for acetaminophen, nimesulide, etodolac and meclofenamic acid. In this respect, Table 1 summarises MRLs and RCs for NSAIDs in bovine milk and muscle tissue. Advantages and disadvantages of the different chromatographic and electrophoretic techniques used for

NSAID (acronym)

Bovine milk MRL or RC (μg kg−1)

Bovine muscle MRL or RC (μg kg−1)

Acetaminophen (ACF) Salicylic acid (SA)

Not regulated No MRL requireda

Not regulated

Acetylsalicylic acid (ASA)

Unauthorised (no RC)a Unauthorised (RC010)

Naproxen (NAP)

No MRL required No MRL required Unauthorised (RC010)

Ketoprofen (KPF)

No MRL required

No MRL required

Meloxicam (MLX) Phenylbutazone (PBZ)

MRL015 Unauthorised (RC05)

MRL020 Unauthorised (RC05)

Nimesulide (NIM) Ibuprofen (IBP)

Not regulated Unauthorised (RC010)

Not regulated Unauthorised (RC010)

Carprofen (CPF) Flunixin (FLX) 5-Hydroxy-flunixinc (5OH-FLX) Etodolac (ETO)

No MRL required – MRL040

MRL0500b MRL020 –

Not regulated

Not regulated

Diclofenac (DCF) Meclofenamic acid (MCL) Tolfenamic acid (TLF)

MRL00.1 Not regulated MRL050

MRL05 Not regulated MRL050

a

According to Article 14(7) of Regulation (EC) No 470/2009, acetylsalicylic acid and sodium salicylate cannot be used in animals producing milk or eggs for human consumption

b Marker residue expressed as sum of carprofen and carprofen glucuronide conjugate c

Marker residue in bovine milk

NSAIDs separation have already been described in detail [8, 9], since the former interest of the scientific community in these pharmaceuticals has been aroused by their involvement as emerging contaminants of the aquatic environment. In the food safety field, several methods using liquid chromatography coupled with mass-spectrometry detection (LC-MS) to analyse NSAIDs in animal food products have been published [10–22], but very few are the multi-analyte ones validated as confirmation methods [15–21] especially for the meat matrix [18]. This paper describes the development and validation of two multiresidue methods, based on LC-tandem MS, to confirm 15 NSAIDs (Fig. 1) in bovine milk and muscle tissue. All analytes selected for this study belong to the group of COX non-selective inhibitors and are the most used in veterinary medicine. Primary objectives of this work were: (a) to develop an extraction procedure able to overcome the strong interaction between NSAIDs and food

Author's personal copy Development and validation of two multiresidue liquid

Fig. 1 Names, structures and monoisotopic masses of the NSAIDs selected for this study

proteins, and to be adapted for the two bovine matrices; (b) to remove the large amount of fatty material otherwise responsible for low extractive efficiencies. A secondary aim was to optimise the LC-MS conditions in order to improve as much as possible the electrospray (ESI) response of diclofenac whose MRL was established at very low level in milk. All results obtained during this study will be presented and discussed along with those related to the method validation and the analysis of different milk and meat samples.

Experimental Materials, reagents and standard solutions Acetaminophen (ACF), acetaminophen-d3 [N-(4-hydroxyphenyl)acetamide-2,2,2-d3)], acetylsalicylic acid (ASA), meclofenamic acid (MCL), salicylic acid (SA), salicylic acid-d6 (2-hydroxybenzoic acid-d6), tolfenamic acid (TLF), carprofen (CPF), carprofen-d3 (6-chloro-α-methyl-d3-9Hcarbazole-2-acetic acid), ketoprofen (KPF), ketoprofen-d3

Author's personal copy A. Gentili et al.

[2-(3-benzoylphenyl)propionic acid (methyl-d3), diclofenac (DCF), diclofenac-13C6 {[2-(2,6-dichlorophenylamino)phenyl13C6] acetic acid}, etodolac (ETO), phenylbutazone (PBZ), phenylbutazone-13C12 (4-butyl-1,2-diphenyl-13C12-pyrazolidine3,5-dione), flunixin (FLX), flunixin-d3 {2-[2-methyld3-3-(trifluoromethyl)phenylamino]pyridine-3-carboxylic acid}, ibuprofen (IBP), ibuprofen-d3 [α-methyl-d3-4-(isobutyl)phenylacetic acid], meloxicam (MLX), meloxicam-d3 {4-hydroxy-2-(methyl-d3)-1,1-dioxo-benzo[e]-1,2-thiazine-3carboxylic acid (5-methyl-2-thiazolyl)amide}, naproxen (NAP) and nimesulide (NIM) were purchased from AldrichFluka-Sigma Chemical (Milan, Italy). 5-Hydroxy-flunixin (5OHFLX) was kindly given by Dr. B. Neri (Istituto Zooprofilattico Sperimentale of Lazio and Tuscany, Rome, Italy). Etodolac-d3, {1,8-(diethyl-d3)-1,3,4,9-tetrahydropyrano[3,4b]indole-1-acetic acid)}, naproxen-d3 (6-methoxy-α-methyld3-2-naphthaleneacetic acid), nimesulide-d5 [N-(4-nitro-2(phenoxy-d5)phenyl)methanesulfonamide], meclofenamic acid-d4 {2-[(2,6-dichloro-3-methylphenyl)amino]benzoic acid-d 4 } and tolfenamic acid-d 4 {2-[(3-chloro-2methylphenyl)amino]benzoic acid-d4} were synthesised in our laboratories (see Electronic Supplementary Material, Schemes S1, S2, S3 and S4). Purity was ≥98 % for all standards and internal standards. Individual stock solutions of the standards and internal standards were prepared in methanol at 1 μg μL−1. Composite working solutions of standards and internal standards were made by mixing the above solutions and diluting with methanol in order to obtain concentrations suitable for the several studies and experiments. When unused, stock solutions were stored at −18 °C, while working solutions, standards and internal standards at 4 °C. For LC, distilled water was further purified by passing it through the Milli-Q Plus apparatus (Millipore, Bedford, MA USA). Acetonitrile, methanol, acetone and hexane (RS Plus grade for HPLC) were obtained from Carlo Erba (Milano, Italy). Trifluoroacetic acid (TFA) was ReagentPlus®, formic acid and dibutylamine (DBA) were puriss. p.a. (AldrichFluka-Sigma Chemical). For extraction studies, OASIS HLB SPE cartridges of 6 mL capacity (500 mg of sorbent) were supplied by Waters (Waters, Milano, Italy). PTFE filters (0.45 μm) were purchased from Alltech (Deerfield, IL, USA). Chemical synthesis of deuterated internal standards and data analysis The synthesis methods and the data of naproxen-d3, meclofenamic acid-d4, tolfenamic acid-d4, nimesulide-d5 and etodolac-d3, used as internal standards, are illustrated in Schemes S1, S2, S3 and S4 (see Electronic Supplementary Material).

The structure of the target compounds were confirmed by NMR analysis and high-resolution MS. The 1H-13C NMR spectral data of the synthetic internal standards were measured on a Varian Mercury-300, from Varian Inc., Palo Alto, CA, USA, (300 MHz for 1H and 75 MHz for 13C). All compounds were dissolved in deuterated chloroform (CDCl3) or methanol (CD3OD) purchased from AldrichFluka-Sigma Chemical. Chemical shifts are given in ppm (δ) using tetramethylsilane (TMS) as the internal standard. Mass spectra were registered using a Q-TOFMICRO spectrometer (Micromass, Waters, Manchester, UK) equipped with an ESI source, in the negative and/or positive ion mode. Data were analysed using the MassLynx software developed by Waters. Column chromatography was carried out on silica gel 60 F254 (Merck). Unless otherwise noted, all reagents were purchased from commercial suppliers and used as received. Food samples All food samples were purchased from randomly chosen supermarkets in Rome. The pasteurised, homogenised whole milk samples were stored at 4 °C, while the fresh raw meat samples were chopped into small pieces and stored at −18 °C. The samples used for the method development and validation were previously determined to be free of the drugs considered. Extraction procedure The established aliquot of food sample was spiked with known variable amounts of the analytes and their corresponding internal standards; a 30-min period was allowed for equilibration at room temperature. Thereafter, a two-step extraction procedure, based on deproteinisation/ extraction with organic solvent and clean-up via SPE, was applied to extract NSAIDs from bovine milk and muscle tissue. Deproteinisation/extraction Milk An aliquot (5 mL) of milk was transferred into a 50-mL screw-capped polyethylene centrifuge tube, and 10 mL of acetonitrile was added. The mixture was first vortexed for 2 min, then placed in an ultrasound bath (model Starsonic 18 from Liarre, Bologna, Italy) for 10 min, and finally centrifuged at 7,000 rpm for 10 min at 0 °C (model PK131R from A.L.C. International, Cologno Monzese, Milan, Italy). To remove the organic solvent, the supernatant fraction was concentrated in a water bath at 30 °C under a gentle nitrogen stream, till a final volume of 5 mL. The extract was centrifuged for 10 min at 7,000 rpm at 0 °C.

Author's personal copy Development and validation of two multiresidue liquid

Muscle tissue A 5-g sample of muscle tissue was put in a beaker containing 200 μL of methanol, homogenised by means of an Ultra-Turrax T8 homogeniser (IKA-Werke GmbH & Co. KG, Staufen, Germany) and then placed in a 50-mL centrifuge tube. After that, 10 mL of acetonitrile was added and the mixture was vortexed, ultrasounded, and centrifuged just as described for milk. The supernatant was then transferred to a new tube and the extraction process was repeated with 10 mL of acetone. The two organic extracts were pooled, concentrated to 5 mL in a water bath thermostated at 30 °C, and finally centrifuged at 7,000 rpm for 10 min at 0 °C. Clean-up OASIS HLB cartridge was fixed to a vacuum flask so as liquids were forced to pass through it by the aid of a waterjet pump. After preconditioning with 6 mL methanol and 15 mL Milli-Q water, the aqueous extracts from the previous step were diluted to 150 mL with Milli-Q water and loaded onto the cartridge. Then, the cartridge was dried under vacuum for 10 min and washed with 6 mL of hexane to remove fat substances. The analyte elution was accomplished with 5 mL of methanol and 15 mL of acetone. The eluate was collected into a glass vial with a conical bottom, concentrated to a volume of 500 μL at 30 °C under a gentle stream of nitrogen, and adjusted to the final value of 1 mL with methanol. Finally, the extract was filtered on PTFE filters (0.45 μm) and injected (20 μL) into the LC-MS/MS system.

Liquid chromatography and mass spectrometry The HPLC system consisted of a Series 200 micro-HPLC, a Series 200 autosampler (equipped with a 20-μL loop) and a degassing device, all from Perkin Elmer (Norwalk, CT, USA). The chromatographic separation of analytes was achieved on an XTerra-MS C18 column (150 × 4.6 mm I.D., 5 μm particle size), protected by a guard column of the same type (20×3.9 mm I.D., 5 μm particle size) from Waters (Milano, Italy). Analytes were identified and quantified by an API 3000TM triple-quadrupole mass spectrometer (AB Sciex, Foster City, CA, USA), equipped with a TurboIonSpray (TIS) source operating in negative ionisation mode. Air was produced by a compressor (Jun-Air 4000-40M, Bromsgrove, UK) and it was used both as nebuliser (2 L min−1) and drying (8 Lmin−1) gas. Nitrogen was produced by a generator (Parker-Balston model 75A74, Haverhill, MA, USA) connected to the air compressor; it was utilised as curtain (1.5 Lmin−1) and collision (4 mTorr) gas. A

temperature of 200 °C was set to heat the drying gas and a voltage of −4,500 V was applied to the capillary. Mass axis calibration of mass-resolving quadrupoles Q1 and Q3 was attained by the infusion of a polypropylene glycol solution at 10 μL min−1. Unit mass resolution was assured by maintaining a full width at half maximum of approximately 0.7±0.1 Da. For the identification of every analyte, two Selected Reaction Monitoring (SRM) transitions were chosen after acquiring and studying the corresponding product ion scan spectra; instrumental parameters of the TIS source and the analysers were optimised for each SRM ion current, infusing the corresponding standard solution with a syringe pump (concentration at 1 ng μL−1 and flow rate of 10 μL min−1). In separating NSAIDs, phase A was acetonitrile–methanol (50:50, v/v) and phase B water, both of them containing 0.2 mmol L−1 DBA as ion pairing agent. The mobile phase gradient profile was as follows (t in min): t0, A00 %; t1, A0 0 %; t2, A020 %; t20, A075 %; t20, A0100 %; t25, A0100 %. The flow rate of the LC eluent was 1 mL min−1 and it was split so that 10 % of the column effluent was diverted to the TIS source by means of a T-junction. The most important LC-MS parameters for the acquisition and identification of fifteen NSAIDs are summarised in Table 2. The chromatographic run was divided into two acquisition periods in order to improve instrumental sensitivity for acetaminophen and salicylic acid, increasing their corresponding dwell time. Labelled internal standards and related LC-MS parameters used for NSAIDs quantitation are listed in Table 3. Figure 2 shows a representative LC-ESI(-)-SRM chromatogram of a meat extract spiked with the selected analytes.

Method validation The developed methods were validated in accordance with the Commission Decision 2002/657/EC [23] and its implementation [24] for a quantitative confirmation method. The validation study of NSAIDs in bovine milk and muscle tissue was carried out at three fortification levels chosen around a validation level (VL). Concerning this, the analytes selected for this study were subdivided in two groups: – –

NSAIDs with an established MRL (MRL as validation level) Unauthorised NSAIDs with and without RC, permitted NSAIDs, and unregulated NSAIDs (CCβ as validation level)

For every analyte, the method performance was assessed in terms of selectivity, identification, recovery, precision,

Author's personal copy A. Gentili et al. Table 2 LC-MS/MS parameters needed to the identification of NSAIDs, selected in this study, in bovine milk and meat extracts

Periods

ANALYTES

Retention time (min) Average (RSD)

SRM transitionsa

Ion ratiob Average (RSD)

Tolerance ion ratios 2002/657/EC (%)

I period (0–6.5 min)

Acetaminophen

4.78 (0.6)

150/118 150/107

0.35 (12)

±50

Salicylic acidc

5.22 (0.4)

137/65 137/93

35 (5.3)

±25

Naproxen

7.73 (0.4)

229/170 229/185

95 (4.0)

±20

8.17 (0.3)

253/197 253/209

9.0 (7.5)

±50

253/185 253/197

2.0 (8.0)

II period (6.5 – 18.5 min)

Ketoprofend

Milk Meat

a

In the first line is reported the least intense SRM transition (qualifier) and in the second line the most intense one (quantifier)

Meloxicam

8.58 (0.3)

350/113 350/286

39 (9.5)

±25

Phenylbutazone

9.04 (0.3)

307/92 307/279

57 (8.0)

±20

b

The relative abundance between the two SRM transitions is calculated as ratio of Qualifier Intensity/Quantifier Intensity

Ibuprofen

10.22 (0.3)

205/161 205/159

60 (8.3)

±20

Nimesulide

10.28 (0.3)

307/198 307/229

15 (4.0)

±30

c

Owing to the quick degradation of acetylsalicylic acid, ion currents of salicylic acid account for both drugs

5-Hydroxyflunixin

10.44 (0.2)

311/160 311/227

24 (10)

±25

Carprofen

10.98 (0.3)

272/260 272/228

10 (13)

±50

d Owing to the occurrence of a compound interfering the m/z 253/209 SRM transition, the confirmation of ketoprofen in muscle tissue was based on the m/z 253/185 ion current

Carprofen glucuronidee

–

448/113 448/272

–

–

Flunixin

Meat

11.14 (0.2)

296/197 286/212

30 (5.4)

±25

Diclofenacf

Milk

11.34 (0.5)

296/252 294/250

75 (14)

±20

294/214 294/250

5.5 (4.5)

±50

Meat

e

Theoretical SRM transitions of carprofen glucuronate

f

Owing to the low MRL value in milk the [M37Cl–H]−/[M37Cl– H–CO2]− was selected as qualifier SRM transition

Milk

Meclofenamic acid

12.42 (0.4)

294/214 294/258

34 (6.0)

±25

Tolfenamic acid

12.70 (0.3)

262/218 260/216

25 (3.5)

±25

linearity and analytical limits. Some experiments were combined as much as possible in order to minimise the workload. The most intense SRM transition (quantifier transition) was used for the quantitative analysis, while the least intense one (qualifier transition) for the estimation of the detection limit (CCα) and the detection capability (CCβ) so that a S/N ratio greater than 3 could always be respected. The MS data acquisition and processing system was the “Analyst” software from Applied Biosystem. Selectivity The presence of potentially interfering substances was investigated by analysing twenty blanks for each food matrix. Moreover, since the method is intended to quantify more than one analyte, increasing amounts of every compound and internal standard were injected individually and the two SRM transitions of each analyte were monitored for intrusive peaks.

Identification The criteria for the analyte identification in bovine milk and muscle tissue are reported in Table 2. The four identification points (IPs) earned by selecting two SRM transitions were enough both to confirm NSAIDs with an established MRL and the unauthorised ones. During the validation study, the relative standard deviation of the retention time was within ±2.5 % and the ion ratio between the two SRM transitions was inside the tolerances recommended by the Commission Decision 2002/657/EC (see Table 2) for all analytes. Recovery and precision Blank samples were fortified at: – –

0.5, 1.0 and 1.5 times the VL level with drugs having an established MRL 1.0, 1.5 and 2.0 times the CCβ with all of other drugs

Author's personal copy Development and validation of two multiresidue liquid Table 3 Labelled internal standards and related LC-MS/MS parameters used for NSAIDs quantitation Periods

I period (0–6.5 min)

Internal standard

SRM transitions

Acetaminophen-d3

4.77 (0.5) 153/107

Salicylic acid-d6

5.20 (0.4) 143/97

II period (6.5–18.5 min) Naproxen-d3

7.71 (0.5) 232/188

Ketoprofen-d3

8.16 (0.3) 256/212

Meloxicam-d3

8.57 (0.3) 353/289

Phenylbutazone-13 C12

a

Retention time (min) Average (RSD)

9.05 (0.4) 319/291

Ibuprofen-d3

10.20 (0.3) 208/164

Nimesulide-d5

10.26 (0.4) 312/234

Carprofen-d3

10.97 (0.3) 275/231

Flunixin-d3a

11.12 (0.3) 298/254

Etodolac-d3

11.17 (0.3) 289/215

Diclofenac-13C6

11.33 (0.4) 300/256

Meclofenamic acid-d4 Tolfenamic acid-d4

12.43 (0.5) 298/262

For NSAIDs with a fixed MRL, the CCα values were determined by analysing 20 blank samples fortified at the corresponding permitted limit (CMRL); CMRL plus 1.64 times the standard deviation of the within-laboratory reproducibility was taken as the decision limit (α05 %). For the other NSAIDs, CCα was calculated as the amount of analyte producing a signal, on average, three times as high as the noise of the baseline in a chromatogram of a blank material; S/N ratio was valued as the mean of the 20 representative blank samples analysed by the developed method. For both groups of drugs, CCβ was estimated by analysing 20 blanks fortified at the previously estimated decision limit; CCα plus 1.64 times the corresponding standard deviation was used to yield the detection capability (β05 %).

Results and discussion

12.68 (0.5) 264/220

Optimization of the LC conditions and MS detection

Also used for the 5-hydroxy-flunixin quantitation in milk

Six replicates of each fortification level were analysed on 1 day. The procedure was repeated on two additional days. The data from the above-described experiments were also used to calculate the recovery and method precision (repeatability and within-laboratory reproducibility), the latter expressed as relative standard deviation (RSD). Criteria for acceptable values at the different fortification levels were adopted from the 2002/657/EC document. Linearity Matrix-matched calibration curves were constructed for each analyte on each validation day. For this purpose, bovine milk/meat blank samples were spiked at 0.0, 0.5, 1.0, 1.5, 2.0, 2.5 and 3.0 times the VL level. The calibration curves were calculated by linear regression, using the quantifier transition and plotting the relative peak area (area of analyte/area of internal standard) as a function of fortification level. Squared linear regression coefficients (r2) greater than 0.99 were considered acceptable. Decision limit (CCα) and detection capability (CCβ) Using the qualifier transition for the CCα and CCβ estimation of each analyte, it was always possible to observe both ion currents with their characteristic ion ratio (Table 2) according to the tolerances established by the Commission Decision 2002/657/EC.

Since most of NSAIDs are best responsive to negative ESI ionisation, the literature mainly describes methods based on ion-suppression RP chromatography with negative ESI detection [17–20], while a dual polarity detection has been adopted in studies involving some NSAIDs that give a more intense signal in positive ion mode [13, 15, 21, 22]. In both cases, the response of NSAIDs detected in negative ionisation is suppressed by the presence of (formic or acetic) acid in the mobile phases, whereas the advantage of a slightly higher response of NSAIDs detected in positive ionisation is reduced switching polarity because of the longer pause time between scans (as verified during this study for flunixin, 5hydroxy-flunixin, meclofenamic acid, tolfenamic acid, meloxicam and acetaminophen). A step-by-step approach was applied to optimise the LCMS conditions in order to achieve analytical limits as low as possible. The main difficulty in separating the selected NSAIDs was related to their different pKa values ranging from 3 to 9.5. Peak-tailing was observed for meloxicam and salicylic acid working in ion suppression RP chromatography (2.5–5 mM formic acid). Using ammonium formate or acetate as ion pairing agent (1–5 mM), the extent of peaktailing was decreasing, the S/N ratio in negative ion mode was improving, but the chromatographic selectivity was getting worse. Sharp, symmetrical peaks and very high S/ N ratios were instead obtained for all analytes with (0.1, 0.2 and 0.4 mM) DBA, but a more selective organic modifier was required for separating diclofenac and meclofenamic acid, isomers sharing one SRM transition. The best compromise in terms of chromatographic resolution and S/N ratio was attained using a mixture 50:50 v/v methanol:acetonitrile

Author's personal copy A. Gentili et al.

Fig. 2 SRM-chromatogram (quantifier transitions) of a bovine meat extract spiked with analytes at their validation level (MRL and CCβ)

as phase B and a 0.2-mM DBA concentration in both mobile phases. Under these conditions, the ion-pairing RP chromatography coupled to the negative ESI ionisation has proved to be more effective than the ion-suppression RP chromatography with the dual-polarity detection. Few other considerations on the behaviour of some analytes, during ESI ionisation and collision induced decomposition (CID), deserve to be reported. It has been verified that aspirin

(acetylsalicylic acid, ASA) is prone to deacetylation both in solvent (water and methanol) and in the ESI source, even at low declustering potential values (see Fig. 3). For these reasons, the two SRM transitions (m/z 137/93 and m/z 137/65) belonging to salicylic acid and shared by ASA can be used to monitor an incidental occurrence of both drugs in bovine milk and meat samples. In source degradation has been observed for carprofen as well: its Q1 full scan spectrum has shown the

Author's personal copy Development and validation of two multiresidue liquid Fig. 3 Owing to the quick deacetylation of acetylsalicylic acid (ASA) in ESI source (panel A: degradation kinetics of ASA at t0) and in solvent (panel B: degradation kinetics of ASA at t06 h), the SRM transitions (m/ z 137/65; m/z 137/93) of salicylic acid are also shared by ASA

pseudomolecular anion at m/z 272 and two fragment ions at m/ z 228 and 260; the latter is also the m/z value of the tolfenamic acid pseudomolecular ion which, under (CID) conditions, generates the product ion at m/z 216. Consequently, the m/z 260/216 SRM transition is shared by both drugs (see the extracted ion current profile of tolfenamic acid in Fig. 2). The poor fragmentation of tolfenamic acid constitutes a problem in achieving the number of identification points needed to its confirmation in real samples. The presence of a chlorine atom in its molecule has made possible to use [M37Cl–H]−/[M37Cl–H–CO2]− as the qualifier transition. The same SRM transition has also been selected to obtain a detection limit as low as possible for diclofenac

because of the very low MRL established in bovine milk (see Table 2). Unlike what has been reported by other researchers, presumably because of different instrumentations, we have found that the CID fragmentation of ketoprofen produces two product ions (m/z 197 and 209) instead of only one (m/z 209). The structural isomery between nimesulide and phenylbutazone has not represented a problem; they are well separated chromatographically and, characterised by different fragment ions. Theoretical SRM transitions of carprofen glucuronate (see Table 2), whose standard is commercially unavailable, were supposed after a fragmentation study of model compounds (estriol-3-glucuronide, estradiol-3-glucuronide,

Author's personal copy A. Gentili et al.

estrone-3-glucuronide, estriol-16-glucuronide, estradiol-16glucuronide). The product ion scan spectra of estrogen conjugates have always shown two fragment ions: the free estrogen and a product of the glucuronic acid at m/z 113. On this basis, an analogous behaviour has also been assumed for carprofen glucuronide. Optimisation of the extraction procedure The extraction procedure was optimised evaluating the absolute recoveries of analytes. To this end, the internal standards were introduced adopting the volumetric method [25, 26], i.e. adding the ISs at the end of the extraction procedure. Absolute recovery of each analyte was assessed by normalising its profile to the peak area of the IS, and comparing this ratio to that obtained by injecting an extract from a related blank sample to which the analytes and were added post-extraction. Deproteinisation/extraction step

having an increasing eluotropic strength. In fact, the use of acetonitrile allowed yields greater than 71 % for most analytes with the exception of meclofenamic acid (50 %) and tolfenamic acid (43 %). Mixtures (20 mL) composed by acetonitrile–ethyl acetate (1:1, v/v) and acetonitrile:acetone (1:1, v/v) improved recovery of these two NSAIDS but worsened that of more polar analytes; in addition, the eluate appeared turbid, especially using ethyl acetate. The sequential extraction with 10 mL of acetonitrile and 10 mL of acetone provided recoveries ranging between 79 % and 100 % for all NSAIDs. Since carprofen and carprofen glucuronide are the marker residues in bovine muscle tissue [6], acid [15] and enzymatic [18] digestions were also tested as preliminary step to deproteinisation/extraction with organic solvents. Nevertheless, the hydrolysis made the whole procedure longer and sensitivity and analytical limits worse: in fact, a large amount of interfering compounds freed from matrix was responsible for a severe suppression of the ESI signal, observed in spite of a final SPE clean-up step.

Milk SPE clean-up step It has been demonstrated that NSAIDs establish a tight interaction with the plasma proteins due to the hydrogen and ionic bonds, and that the value of drug–protein association constant is increasing with decreasing pH [27]. This means that acid deproteinisation is inadequate to extract NSAIDs from a biological/food sample: in fact, as verified in this study, the analyte recovery from two milk samples, spiked before and after acid deproteinisation, was under 50 % in the first case and greater than 90 % in the second one. Addition of neutral water-miscible organic solvents has been used successfully to precipitate proteins and to extract NSAIDs from milk and meat [11, 13, 16, 17, 19, 21, 22]. Its advantageous use is due to the decrease of the medium dielectric constant: protein molecules aggregate and precipitate, while pKa value of NSAIDs is rising steeply. This step was here optimised for both the food matrices, testing the efficiency of different solvents (acetonitrile, acetone, ethyl acetate), their mixtures and volumes. Moreover, performing centrifugation steps at 0 °C, the co-extracted fatty material was removed in large part and the extractive performances were improved. High recovery values of all analytes were achieved both with acetone and with acetonitrile, but the latter was preferred for two main reasons: (a) acetaminophen yield was 93 % (instead of 60 %) and (b) milk deproteinisation/extraction by means of acetone caused a significant occurrence of lipids in the supernatant. Muscle tissue Unlike from milk, deproteinisation/extraction of NSAIDs from meat required the sequential elution with solvents

A clean-up step, following the deproteinisation/extraction with solvent, was added to reduce matrix effect and to concentrate the final extract. Since OASIS-HLB cartridges have already been tested successfully for isolating NSAIDs from aqueous environmental samples [9, 10], we decided to prove their effectiveness also for food samples. In order to prepare the organic extract from the first step for its SPE loading, the supernatant was concentrated to 5 mL (in the case of milk this is roughly the volume corresponding to its aqueous component) and then diluted to a volume of 150 mL with Milli-Q water. The viscosity reduction, due to the dilution, achieved a better flow rate during the SPE loading. Other experiments were performed to establish pH of the aqueous extract before its SPE loading, dilution volumes, extractants and their volumes. After loading of sample, the washing of SPE cartridge with hexane (6 mL) was useful to remove the co-extracted fatty material still remaining. By using one 5-mL aliquot of methanol, followed by three 5mL aliquots of acetone, absolute recoveries exceeded 76 % with the exception of acetaminophen (63 %) and phenylbutazone (50 %), anyway an acceptable range for a multicompound quantitative method. In accordance with Dubreil-Chéneau et al. [20], we verified that the use of antioxidant agents such as ascorbic acid [15, 17–19] did not lead to significant improvements in recovery values. Besides oxidation, another difficulty in the analysis of phenylbutazone might be due to the concurrent equilibrium of the three forms: enol, keto and enolate [28]. Acidification of the 150 mL of aqueous extract with

Author's personal copy Development and validation of two multiresidue liquid

formic acid at pH 3.5, before its SPE loading, did not show considerable effect on extraction efficiency, so it was avoided to simplify the procedure. Analytical method validation A summary of the validation parameters is reported in Tables 4 and 5. During the assessment of method selectivity, no signal interfering with that of the interested compounds was observed within their retention window (tr±2.5 %). One exception was noticed for the quantifier SRM transition (m/z 253/209) of ketoprofen after extraction from bovine muscle tissue (see Fig. 4). In this case, the less intense but “cleaner” SRM transition m/z 253/185 was utilised for identification purposes (see Table 2). All recovery and precision experiments were performed using food samples which, by preliminary analyses, resulted analyte-free. The relative average recovery at the different fortification levels varied from 97 % to 102 %. Repeatability results were obtained at the three fortification levels under identical conditions by the same operator on a particular day. To estimate the within-laboratory reproducibility, the three analytical sessions were performed by different operators employing different batches of solvents, working solutions and food samples. The 2002/657/EC document states

that the precision of a quantitative method shall be as low as possible for mass fractions less than 100 μg kg−1; satisfactory results were accomplished here both for the analytes with a VL level lower than 100 μg kg−1 and for carprofen, whose MRL has been fixed at 500 μg kg−1. Quantitation of NSAIDs was performed by the external standard procedure. The analyte–area/internal standard–area ratio was plotted against the fortification level to generate linear calibration curves with a r2 greater than 0.9930 for bovine milk and than 0.9922 for muscle tissue. An effective correction of the matrix effect was achieved combining an exhaustive extraction procedure with an adequate internal standardisation. In order to assess the reliability of the estimated analytical methods, 20 blanks of bovine milk and meat were fortified with the calculated concentrations of CCα and CCβ. All NSAIDs were always identified at CCα in 50 % of all cases and at CCβ in 95 %, at least [24]. Moreover, in accordance with the CRL Guidance Paper [7], the figures generated for CCα (Tables 4 and 5) were lower than the recommended concentrations (Table 1) for naproxen, phenylbutazone, and ibuprofen both in bovine meat and milk. Application to real samples The feasibility of this method in measuring NSAIDs at trace levels in food matrices was checked analysing eight

Table 4 Overall summary of validation parameters related to the confirmatory method of NSAIDs in bovine milk NSAIDs

Recoverya (repeatibility)b (%)

Intra-laboratory reproducibilityb (%)

Linear regression parameters

with CCβ as VL

Spiked level

Spiked level

Slope

Acetaminophen Salicylic acid Naproxen Ketoprofen Phenylbutazone Nimesulide Ibuprofen Carprofen Etodolac Meclofenamic acid with MRL as VL Meloxicam 5-Hydroxy-flunixin Diclofenac Tolfenamic acid

1.0 CCβ

1.5 CCβ

2.0 CCβ

99 (5) 100 (5) 100 (3)

99 (5) 100 (5) 100 (4)

100 (4) 100 (5) 100 (4)

100 (3) 100 (3) 98 (6) 98 (5) 101 (3) 100 (4) 99 (4) 100 (3) 100 (3) 100 (3) 100 (3) 100(3) 99 (4) 98 (4) Spiked level 0.5 MRL 1.0 MRL 100 (5) 100 (4) 98 (5) 98 (4) 100 (5) 100 (5) 97 (5) 99 (4)

101 (3) 99 (5) 100 (4) 100 (3) 99 (2) 100 (4) 99 (4) 1.5 MRL 100 (4) 100 (4) 100 (4) 99 (4)

1.0 CCβ 7 8 6 5 9 7 6 5 8 7 Spiked level 0.5 MRL 8 8 9 8

a

Relative recovery. The results are reported as the average of six replicates

b

Method precision was expressed as relative standard deviations

1.5 CCβ

r2

Analytical limits (μg kg−1) CCα

CCβ

2.0 CCβ

7 7 7

7 7 6

0.084 0.146 0.060

0.9975 0.9994 0.9931

84.2 4.56 1.24

196 11.2 2.76

5 8 6 6 5 6 6

4 8 6 5 5 5 6

0.064 0.033 0.538 0.134 0.003 0.158 0.032

0.9947 0.9986 0.9975 0.9930 0.9984 0.9935 0.9993

1.08 0.71 0.45 3.55 3.68 0.33 1.46

2.12 1.23 0.74 7.85 13.4 0.69 3.36

1.0 MRL 8 8 8 7

1.5 MRL 7 6 8 7

0.088 0.029 2.660 0.323

0.9986 0.9946 0.9994 0.9998

22.3 48.7 0.14 55.9

29.8 55.2 0.18 61.7

Author's personal copy A. Gentili et al. Table 5 Overall summary of validation parameters related to the confirmatory method of NSAIDs in bovine muscle tissue NSAIDs

Recoverya (repeatability)b (%)

Intra-laboratory reproducibilityb (%)

Linear regression parameters

with CCβ as VL

Spiked level

Spiked level

slope

CCβ

1.5 CCβ

2.0 CCβ

97 (6) 99 (5)

99 (6) 100 (5)

99 (5) 101 (4)

8 9

8 9

8 8

0.990 2.140

0.9963 0.9965

100 3.25

288 7.03

100 (3)

100 (3)

100 (4)

7

7

6

0.752

0.9978

1.52

2.80

99 (3) 97 (6)

100 (4) 97 (6)

100 (3) 98 (5)

6 9

5 9

5 9

0.102 0.199

0.9922 0.9987

1.12 0.92

2.67 1.84

Nimesulide

100 (3)

100 (3)

100 (3)

8

7

6

2.060

0.9992

0.30

0.68

Ibuprofen

100 (4)

100 (3)

101 (4)

5

5

5

0.941

0.9995

3.84

9.90

Etodolac

99 (5)

99 (4)

100 (4)

6

5

5

0.624

0.9993

0.20

0.47

Meclofenamic acid

98 (5)

99 (5)

99 (4)

9

8

8

0.203

0.9991

1.92

4.10

1.0 MRL 6 4 6

1.5 MRL 6 4 5

0.428 0.435 2.030

0.9946 0.9985 0.9991

5 8

5 8

0.963 1.000

0.9979 0.9975

Naproxen Ketoprofen Phenylbutazone

with MRL as VL Meloxicam Carprofenc Flunixin Diclofenac Tolfenamic acid a

Spiked level 0.5 MRL 1.0 MRL 101 (4) 100 (5) 99 (3) 102 (3) 98 (4) 99 (3) 98 (5) 97 (5)

99 (5) 101 (5)

1.5 MRL 98 (4) 97 (5) 100 (3) 99 (4) 100 (5)

1.5 CCβ

CCα

1.0 CCβ Acetaminophen Salicylic acid

1.0 CCβ

r2

Analytical limits (μg kg−1)

Spiked level 0.5 MRL 7 5 6 6 8

2.0 CCβ

31.6 652 38.8 8.05 70.3

48.3 796 50.2 12.5 90.6

Relative recovery. The results are reported as the average of six replicates

b

Method precision was expressed as relative standard deviations

c

Recovery of carprofen glucuronide could not be valued due to the unavailability of its standard

different brands of pasteurised fresh milk and eight different samples of bovine meat. Three replicates of each sample were analysed.

Most samples showed the occurrence of the salicylic acid at levels included between CCα and CCβ (samples suspected positive but without a statistical certainty), with the

Fig. 4 Chromatographic SRM profiles for ketoprofen extracted from bovine muscle tissue (panels A) and milk (panels B). The occurrence of an interfering compound in the meat extract induced to substitute the m/z 253/197 SRM transition for the m/z 253/209 one (see Table 2)

Author's personal copy Development and validation of two multiresidue liquid

exception of two meat samples where salicylic acid concentration was larger than CCβ (60±2 and 14.8±0.6 μg kg−1, respectively). Aspirin (ASA) is quickly metabolised to salicylate (SA) [1, 2] which, in turn, is rapidly distributed into extracellular fluids such as milk; therefore, ASA is not authorised in animals producing milk for human consumption. Nevertheless, the presence of salicylic acid in foods of animal origin cannot be simply ascribed to the use of ASA or SA because this simple phenolic metabolite is a key signalling molecule for activation of defence responses and resistance to various pathogens in many host plant species [29]. One sample of meat was suspected for diclofenac (CCα< CCx