Journal of Chromatography B, 991 (2015) 68–78

Contents lists available at ScienceDirect

Journal of Chromatography B journal homepage: www.elsevier.com/locate/chromb

Development of LC–MS/MS methodology for the detection/determination and confirmation of chloramphenicol, chloramphenicol 3-O-␤-d-glucuronide, florfenicol, florfenicol amine and thiamphenicol residues in bovine, equine and porcine liver Rick W. Fedeniuk ∗ , Massey Mizuno, Connie Neiser, Collin O’Byrne Centre for Veterinary Drug Residues, Canadian Food Inspection Agency, Saskatoon Laboratory, 116 Veterinary Road, Saskatoon, SK, Canada S7N 2R3

a r t i c l e

i n f o

Article history: Received 25 November 2014 Accepted 3 April 2015 Available online 10 April 2015 Keywords: Amphenicols Chloramphenicol-glucuronide Matrix effects QuEChERS Recovery

a b s t r a c t A method for the detection and confirmation of organic solvent extractable residues of the neutral, acidic, and basic analytes of the amphenicol class veterinary drugs and selected metabolites was developed and validated. Using a modified QuEChERS extraction with SPE cleanup and LC–MS/MS analysis, limits of detection and confirmation for the different analytes in bovine, equine, and porcine liver ranged from 0.1 ng/g for chloramphenicol to 1 ng/g for florfenicol amine. Tissue homogenization with an ammonium formate/EDTA solution and subsequent analyte partitioning against 7:3 acetonitrile:isopropanol solution and mixed-mode strong-cation exchange solid-phase extraction cartridge cleanup allowed for the extraction of all compounds from tissues with mean recoveries ranging from 50% (chloramphenicol 3-O␤-d-glucuronide) to 90% (thiamphenicol). Matrix effects ranged from greater than 85% suppression for florfenicol amine to 70% matrix enhancement for chloramphenicol 3-O-␤-d-glucuronide. Quantitation and confirmation were accomplished using commercially available penta-deuterated chloramphenicol as internal standard and multiple reaction monitoring (MRM) of two or three transitions per target analyte. Method accuracy was greater than 15% for all compounds except the glucuronide metabolite. Intra-lab method repeatability estimates ranged from 73% RSD for chloramphenicol 3-O-␤-d-glucuronide to 14% RSD for chloramphenicol. Only chloramphenicol 3-O-␤-d-glucuronide and florfenicol amine at the low end of their calibration ranges (0.25 and 1 ng/g, respectively) did not meet AOAC recommended HorRatr guidelines for intra-lab repeatabilities. Preliminary tests show that the method’s extraction protocol can be used to recover analytes of the ␤-agonists, corticosteroids, fluoroquinolones, sulfonamides, and tetracycline drug classes from the same matrices. Requirements for use in national chemical monitoring programs as a detection/confirmatory (florfenicol amine and chloramphenicol 3-O-␤-d-glucuronide) and determinative/confirmatory (chloramphenicol, florfenicol, thiamphenicol) analytical methodology are met. Crown Copyright © 2015 Published by Elsevier B.V. All rights reserved.

1. Introduction Chloramphenicol (CAP), florfenicol (FF), and thiamphenicol (TAP) (Fig. 1) are antimicrobial drugs due to their ability to inhibit protein synthesis in susceptible bacteria [1]. In Canada, the only drug of this class that is allowed for use in food animals is FF [2]. TAP is not regulated for use in food animals, and CAP is specifically banned for use in animals intended for food use, both by

∗ Corresponding author. Tel.: +1 306 385 7831. E-mail addresses: [email protected] (R.W. Fedeniuk), [email protected] (M. Mizuno), [email protected] (C. Neiser), [email protected] (C. O’Byrne). http://dx.doi.org/10.1016/j.jchromb.2015.04.009 1570-0232/Crown Copyright © 2015 Published by Elsevier B.V. All rights reserved.

Canada [3] as well as by the Codex Alimentarius [4]. The ban is due to CAP’s rare but known potential for causing aplastic anemia in susceptible humans, for which there is no currently characterized dose–response effect. Guidelines for the minimum required performance limits (MRPL) for CAP analytical methodologies [5] suggest a limit of detection (LOD) and confirmation (LOC) of 0.3 ng/g. However, there are no guidelines regarding MRPLs for CAP metabolites. CAP undergoes rapid metabolism [6]. A review by Wongvatchi et al. [1], as well as studies done by Korsrud et al. [7], Lynas et al. [8], and Mehdizadeb et al. [9] indicated that liver is a viable tissue for monitoring chloramphenicol residues, as well as being the preferred tissue for monitoring florfenicol and other residues. However, other tissues such as kidney are also viable tissues

R.W. Fedeniuk et al. / J. Chromatogr. B 991 (2015) 68–78

Fig. 1. Structures of chloramphenicol (A), chloramphenicol 3-O-␤-d-glucuronide (B), florfenicol (C), florfenicol amine (D) and thiamphenicol (E).

for monitoring chloramphenicol residues. Though metabolite profiles differ by species, treatment, and other factors, one of CAP’s common metabolites are the glucuronide conjugates, with the chloramphenicol 3-O-␤-d-glucuronide (CAP-GLUC) being the major conjugate [8,10]. Glucuronidation facilitates excretion and makes a drug much more polar, dramatically changing the extraction characteristics [11]. CAP has been shown to occur naturally in environment. Wongvatchi et al. [1] provided analyses of scenarios and rationale up to and including “worst-case” scenarios for ingestion of naturally occurring CAP and its impact on analytical results. It is shown that a “worst-case” scenario can potentially lead to occurrence of chloramphenicol residues in tissue. However, given that CAP does not have a known dose-response effect on occurrence of potentially negative effects upon human ingestion, and is therefore banned for use in animals intended for food, the source of CAP residues may be considered irrelevant, though a potential point of contention between producers and regulatory authorities. The QuEChERS method is a well-known methodology for the extraction of several classes of drugs, including pesticides and veterinary drugs, from different matrices. In brief, it uses liquid–liquid partitioning of acetonitrile salted out from aqueous phase via magnesium sulfate. However, it is known that extraction of very polar analytes, as well as those that are metal chelators, presents difficulties for the unmodified QuEChERS method [11,12]. A Canadian Food Inspection Agency funded project was carried out to develop and validate analytical methodology with improved detection capability for CAP in animal tissues. Project objectives included an improvement in the LOD for the parent compound, as well as expanding the capability for monitoring metabolites. It was decided that the method’s goals were to: (1) develop and validate a method with a CAP LOD and LOC below 0.3 ng/g. (2) Include detection and confirmation for CAP’s metabolite CAP-GLUC without the use of enzymatic digestion. (3) Maintain or improve method LODs and LOCs for FF and TAP relative to the lab’s current methodology. (4) Include determination of the FF metabolite, florfenicol amine (FFA). The diverse physicochemical properties of the target analytes (acidic, basic, neutral, variable polarities), required that the method be relatively non-selective but minimize matrix effects to address quantitation requirements. Subsequently, the method’s capabilities for extraction of analytes of the drug classes ␤-agonists, ␤-lactams, corticosteroids, fluoroquinolones, growth-promoters (steroid-based), macrolides, sulfonamides and tetracyclines, from the same matrices were assessed.

69

The unique features of this method are the modification of the QuEChERS extraction protocol to extract polar, ionized analytes, and the simultaneous detection and confirmation of CAP and its predominant metabolite CAP-GLUC without the use of enzymatic digestion. Additionally, estimates of accuracy, matrix effects, recovery and corresponding repeatability via variance component analysis allow for determination of method parameters that could be targeted for development if further improvements were required. Finally, the utility of the strategy of optimizing a method for a small but chemically diverse subset of target compounds for extraction of a much broader array of analytes is assessed. It is noted that dextramycin, the mirror image isomer of CAP, would not be distinguishable from chloramphenicol using the analytical methodology described within the manuscript. However, Berendsen et al. [13] reported that dextramycin is unlikely to occur either naturally as the result of production by plant fungal species nor is it likely to be present in pharmaceutical preparations. As well, given that dextramycin is reported to have substantially less antimicrobial activity compared to CAP, it is likely that a producer would not use such a substance for treatment, nor would residues occur through ingestion of natural sources of the compound. 2. Materials and methods 2.1. Reagents Unless stated otherwise, reagents were equivalent to reagent ACS grade or better and all organic solvents were distilled in glass. Type 2 deionized and distilled water was used throughout. Ammonium formate, oxalic acid, sodium chloride and sodium hydroxide were from Fisher Scientific (Toronto, ON, Canada). Ethylenediaminetetraacetic acid disodium salt dihydrate (EDTA) was from Sigma–Aldrich (Oakville, ON, Canada). Acetic acid, ammonium hydroxide (28–30%), formic acid, hydrochloric acid (Omnitrace) and sodium sulphate (Tracepure) were from EMD (VWR International, Toronto, ON, Canada). Acetonitrile, 95% nhexane, isopropanol (HPLC grade), methanol, and methyl tert-butyl ether (MTBE) were from Caledon (VWR International). Chloramphenicol (CAP) was purchased from MP Biomedicals, Inc. (Irvine, CA, USA). Chloramphenicol 3-O-␤-d-glucuronide (CAPGLUC) was purchased from Cerilliant Corporation (Round Rock, TX, USA). Florfenicol (FF) and thiamphenicol (TAP) were purchased from Sigma–Aldrich. Florfenicol amine (FFA) was a gift from Schering-Plough Animal Health (Kirkland, PQ, Canada). The deuterated internal standard chloramphenicol-D5 (CAP-D5) was purchased from Cambridge Isotope Laboratories, Inc. (Andover, MA, USA). Analytes from the drug classes ␤-agonists, ␤-lactams, corticosteroids, fluoroquinolones, growth-promoters (steroid-based), macrolides, sulfonamides and tetracyclines, were from multiple suppliers. 2.2. Materials Iris MCX SPE cartridges (6 mL, 200 mg, 25–35 ␮m) were from Canadian Life Science (Peterborough, ON, Canada). Oasis MCX SPE cartridges (6 mL, 150 mg, 30 ␮m) were from Waters (Mississauga, ON, Canada). PTFE syringe filters (0.2 ␮m, 13 mm) were from Pall Corporation (Ann Arbor, MI, USA). Tissues were obtained from Canadian Food Inspection Agency inspected abattoirs. 2.3. LC–MS/MS conditions A Waters 2695 LC module with a column heater and refrigerated autosampler in tandem with a Quattro Ultima mass spectrometer (Waters) was used. An ACE C18-AR column (50 mm × 2.1 mm, 3 ␮m packing) with an ACE C18-AR 2.1 mm guard cartridge

70

R.W. Fedeniuk et al. / J. Chromatogr. B 991 (2015) 68–78

were from Canadian Life Science. An Xbridge BEH C18 column (50 mm × 2.1 mm, 3.5 ␮m packing,) with an Xbridge BEH C18 Sentry 2.1 mm guard cartridge were from Waters. Column temperature was set to 40 ±1 ◦ C. Following method development, a two component mobile phase using the ACE column was chosen: A – aqueous 0.1% acetic acid, and B – 10:90 isopropanol:methanol. An acetonitrile component (C) was included at the end of each injection gradient to aid in removal of strongly retained compounds. Autosampler temperature was set to 5 ± 1 ◦ C, injection volume was 10 ␮L, and precolumn volume was 500 ␮L. A column flow diverter delivered effluent to the mass spectrometer from 0.9 to 6.0 min. Linear changes between steps in mobile phase composition were done. The mobile phase started with 0.3 mL/min of 100:0:0 A:B:C, changing to 70:30:0 at 0.2 min, held until 2.0 min, then to 50:50:0 at 6.5 min. This was followed by 10 column volumes of 10:45:45, then 15 column volumes of 100:0:0. Single component 1 ␮g/mL analyte in methanol solutions were infused into the mass spectrometer at 5 ␮L/min with an 80:20 A:B mobile phase flow rate of 0.3 mL/min. Settings were tuned to give the maximum response of the monitored MRM ion-transitions. Nominal global settings were as follows: ionization mode electrospray ionization ESI+ve and ESI−ve; capillary, 1 kV; hex 1, 27 V; hex 2, 0.3 V; aperture, 0.1 V; source temperature, 120 ◦ C; desolvation gas temperature, 450 ◦ C; cone gas flow (nitrogen), 60 L/h; desolvation gas flow (nitrogen), 1000 L/h; LM1, LM2, HM1 and HM2 resolution, 13.5 arbitrary units (approximately unit mass resolution); ion energy 1, 0.5 V; ion energy 2, 2.0 V; multiplier, 750 V; collision cell pressure, argon, ca. 2.5 × 10−3 mbar. Analyte specific mass spectrometer settings are given in Table 1. ESI+ve was used exclusively during the first 2 min of each injection for data collection, followed by exclusive use of ESI−ve. Note that as indicated in the Waters Quattro LCZ, Ultima, and Ultima PT performance maintenance protocol [14] instrument tolerances will result in nominal mass settings differing from theoretical exact masses. For the Quattro Ultima, the tolerance is 0.5 Da. Peak integration was preceded by data smoothing using two data point averaging and two passes. Integration parameters were adjusted to obtain valley-to-valley peak integration. The LC–MS/MS conditions used for assessment of the method’s extraction capabilities for analytes from the drug classes ␤agonists (10), ␤-lactams (9), corticosteroids (9), fluoroquinolones (4), growth-promoters (steroid-based – 8), macrolides (2), NSAIDS (9), sulfonamides (8) and tetracyclines (4) were those used in the lab’s corresponding validated methodologies for those drug classes.

2.4. Method development Factorial or adaptive one-variable-at-a-time (OVAT) experimental designs were used when appropriate for method development/optimization as suggested by Frey et al. [15]. 2.5. Final amphenicol extraction procedure Five grams of liver were homogenized with 5 mL of 0.5 M ammonium formate, 0.2 M disodium EDTA, 0.15 M sodium hydroxide, and 3 mL isopropanol. The mixture was shaken for 5 min, then 7 mL of acetonitrile added, and shaken for another 5 min. Four grams of sodium sulphate and 1 g of sodium chloride were added, shaken 5 min, centrifuged, and the organic layer poured into a new set of tubes containing one gram of sodium sulfate. Five mL of 7:3 acetonitrile:isopropanol were added to the tissue plugs and shaken 5 min, centrifuged and the organic layer added to the first extracts. The extracts were washed with 5 mL of hexane, and 2 mL of MTBE added. After shaking and centrifugation, the upper organic layer was transferred to a new set of tubes and evaporated under N2 at 60 ◦ C. The extract was reconstituted with 10 mL of aqueous 0.5 M ammonium formate, 0.4 M hydrochloric acid, 0.1 M oxalic acid, washed twice with 5 mL of hexane, then applied to Oasis MCX SPE cartridges preconditioned with methanolic 3% ammonium hydroxide, aqueous 5% ammonium hydroxide with 20% methanol, then aqueous 1% formic acid. The extracts were applied at a rate less than 1 drop/3–5 s. The cartridges were rinsed with 1 mL consecutive rinses of aqueous 1% formic acid and aqueous 1% formic acid with 10% methanol. The cartridges were dried under vacuum, and then rinsed with 2 mL MTBE. The analytes were eluted with 3 mL of methanolic 3% ammonium hydroxide, evaporated under N2 at 60 ◦ C, then reconstituted with 100 ␮L of 1:1 acetonitrile:methanol, 900 ␮L of 0.1% formic acid, and filtered through 0.2 ␮m PTFE syringe filters. 2.6. Method characterization/validation 2.6.1. Characterization: calibration ranges/LOD/LOQ/matrix effects/recovery/relative response factors For CAP, a target LOD of 0.2 ng/g was selected. For CAP-GLUC, a target LOD of 0.5 ng/g was selected, which is approximately equivalent to 0.3 ng/g CAP. Target LODs for FF and TAP were 0.5 ng/g. For FFA, method development activities indicated a target LOD of 2.0 ng/g.

Table 1 Analyte specific mass spectrometer settings. ESI mode

Analyte

Precursor ion (m/z)

Product ion (m/z)

Dwell time (s)

+ve

Florfenicol amine (FFA)

247.9 247.9 247.9

129.9a 197.0 230.0

0.18 0.18 0.18

5 5 5

21 21 10

1.1–2.0

−ve

Thiamphenicol (TAP)

353.9 353.9

184.8a 289.9

0.18 0.18

10 10

16 4

2.1–3.0

−ve

Florfenicol (FF)

356.0 356.0 356.0

118.7 184.9 336.0a

0.18 0.18 0.18

10 10 10

30 16 5

3.0–3.8

−ve

Chloramphenicol glucuronide (CAP-GLUC)

497.0 497.0

192.8a 397.1

0.30 0.30

30 30

14 14

3.8–4.7

−ve

Chloramphenicol (CAP)

320.9 320.9 320.9

151.8a 193.9 256.9

0.18 0.18 0.18

5 5 5

14 6 4

4.6–5.9

−ve

Chloramphenicol-D5 (CAP-D5)

325.9

156.8

0.15

5

14

4.6–5.9

a

Quantitation ion.

Cone voltage (V)

Collision energy (V)

Time window (min)

R.W. Fedeniuk et al. / J. Chromatogr. B 991 (2015) 68–78

Estimates of amphenicols’ relative response factors, accuracy, matrix effects, recovery, and repeatability at 1.0×, 5.0× and 20.0× the target LODs were obtained. Weighted linear regression (1/concentration) of five-gram samples spiked with analyte and 1 ng/g of the CAP-D5 internal standard using one of bovine, equine, and porcine livers in three separate experiments was done. Other drug classes were quantified using the lab’s corresponding validated LC–MS/MS procedures. Coefficients of determination (R2 ) were used to assess the suitability of the linear model for calibration. One sample from each of bovine, equine, and porcine liver tissue was prespiked in each experiment at the 1.0×, 5.0×, and 20.0× target LODs (n = 3 for each level) to provide estimates of overall method repeatability, within-experiment (between-species) repeatability, and among-experiment repeatability. Post-extraction spikes at the 1.0×, 5.0×, and 20.0× LODs were done in corresponding tissues and in chemical standards to estimate recovery and matrix effects. To estimate recoveries, the ratio of the pre-spike quantitation ion area counts to the corresponding tissue’s post-spike area counts, multiplied by 100, were calculated. To estimate matrix effects, the ratio of the area counts in the post-spiked extracted samples to the area counts in chemical standard solutions, multiplied by 100, then subtract 100, were calculated. Overall method precision at each concentration was calculated using a full model (i.e. all data) and separated into two components using a nested experimental design, assuming random effects. Within-experiment (which is between-species confounded with intra-day precision) was nested among-experiment (confounded with inter-day precision). The variance component for withinexperiment was the error term, whereas the variance component for among-experiment was the difference of the mean square for the full model from mean square error, divided by three. Variance components were reported as percent relative standard deviation (%RSD) and used to determine which component contributed the most to overall method precision. The analyte’s quantitation ion area counts in 5.0× and 20.0× samples, divided by the relative concentration, and then divided by the area counts in the corresponding matrix at the 1.0× spike, were used to estimate relative responses. A 2-sample t-test assuming unequal variances was used to determine if statistically significant differences (P < 0.05) existed between relative responses of the treatments. The LOD and LOQ for the analytes were estimated using a modification of the Miller and Miller method [16]; rather than use the variation about the predicted values at all calibration levels, only the variation about the predicted values for the three lowest calibration levels, 0.5×, 1.0× and 1.5× the target LOD from each of the 3 experiments was used, providing nine data points to estimate Sy/x . LOD and LOQ were defined as 3 × Sy/x and 10 × Sy/x , respectively.

2.6.2. Validation: accuracy/method repeatability/specificity The concentration ranges chosen for validation were from 0.5× to 5.0× the target LODs. The concentration ranges were: CAP – 0.1–1.0 ng/g; CAP-GLUC, FF, and TAP – 0.25–2.5 ng/g; FFA – 1.0–10.0 ng/g. The matrix-fortified calibration standards used were 0.5×, 1.0×, 1.5×, 2.0×, 3.0×, and 5.0× the target LODs. Validation experiments estimated accuracy and repeatability using triplicate blind spikes at three different concentration ranges from nine randomly selected tissues (three sources each of bovine, equine and porcine liver). The blind spike ranges were: low (1.0× − 2.0× target LOD), midlevel (2.0× − 3.0× target LOD) and high (3.0× − 5.0× target LOD). The experiments were performed on three separate days, using one of bovine, equine, and porcine liver matrix-fortified calibration curves in each experiment. The LOC was defined as the lowest concentration with all samples having at least one calculated ratio of the qualifier ion area counts/quantitation ion

71

area counts lie within tolerances to ratios obtained from chemical standards as defined in EU document 2002/657/EC [17]. Accuracy at each blind spike range was estimated by summing the % deviation of the estimated concentration from known blind spike concentration over the three experiments The standard deviation of the % deviations for each analyte was used as an estimate of the %RSD for method repeatability. Specificity was evaluated by determining if statistically significant differences (P < 0.05) existed between the LODs and LOQs of the characterization and validation data sets. The characterization experiments contained 0.5 − 1.5× target concentrations of 39 additional drug analytes monitored as part of Canada’s National Chemical Residue Monitoring Program, and therefore have the highest probability of being interferences. The validation experiments contained only the amphenicol analytes. 2.6.3. Validation: ruggedness testing A 7-factor Plackett–Burman experimental design with spikes at the 0.5× the target LOD for each analyte was used. The factors and levels chosen were: (1) bovine versus porcine liver; (2) organic solvent/tissue/salts shaking time – 5 min versus 15 min; (3) organic solvent extract evaporation time – to dryness versus to dryness plus 30 min; (4) aqueous reconstitution solution – 0.5 M ammonium acetate, 0.4 M hydrochloric acid, 0.1 M oxalic acid versus 1% formic acid; (5) hexane washes – two versus three; (6) Oasis MCX (waters) versus IRIS MCX SPE; (7) SPE eluate evaporation time – to dryness versus to dryness plus 20 min. Determination of statistically significant differences (P < 0.05) between factor effects was evaluated using a modified 2-sample ttest between sample means with the error variance obtained from repeated analysis of the unmodified method [18,19]. 2.6.4. Validation: stability testing 2.6.4.1. Extracts. Six blank samples from each of bovine, equine and porcine liver were extracted and post-spiked at 2.0× target LOD, evaporated, and placed in −70 ◦ C storage. One sample from each species was reconstituted and placed in 5 ◦ C storage at times to represent storage of 14, 10, 7, 4, 2 and 0 days in a 5 ◦ C autosampler. All samples were run within the same batch, using linear regression analysis for each species with storage time (days) and response as independent and dependent factors, respectively. A slope statistically different from zero (df = 4, P < 0.05) was used as an indicator of extract instability over 14 days. 2.6.4.2. Storage time. Six blank samples from each of bovine, equine, and porcine liver were spiked at 2.0× target LOD concentrations and placed in −70 ◦ C storage. Samples were transferred to −20 ◦ C storage at times to represent storage of 0, 7, 14, 21, 28 and 35 days at −20 ◦ C. All samples were processed within the same batch and analyzed immediately, using linear regression analysis for each species with storage time (days) at −20 ◦ C as independent factor and response as the dependent factor. A slope statistically different from zero (df = 4, P < 0.05) was used as an indicator of analyte instability at −20 ◦ C over 35 days. 2.6.4.3. Freeze–thaw. Six blank samples from each of bovine, equine, and porcine liver were spiked with 2.0× target LOD concentrations and placed in −20 ◦ C storage. After day 1, two samples from each tissue type were thawed at room temperature for 4 h before being placed back in −20 ◦ C storage. After 7 days of storage, the day 1 samples and an additional two samples from each tissue type were thawed at room temperature for 4 h before being placed back into −20 ◦ C storage. On day 14, all samples were processed within the same batch, and analyzed immediately, using linear regression analysis for each species with the number of freeze/thaw cycles as independent factor and the response as the dependent factor.

72

R.W. Fedeniuk et al. / J. Chromatogr. B 991 (2015) 68–78

A slope statistically different from zero (df = 4, P < 0.05) was used as an indicator of analyte instability as a function of number of freeze/thaw cycles. 3. Results and discussion 3.1. Method development 3.1.1. Column and mobile phase selection, mass spectrometer parameters Electrospray positive gave the largest signal/noise for FFA, whereas electrospray negative gave the largest signal/noise for all the other amphenicols. Methanol-containing mobile phase was observed to result in larger ion area counts and s/n peaks for CAPGLUC and FFA than acetonitrile, and therefore was used for all further tests. Acidic (0.01% formic acid) and alkaline (0.1% ammonia) aqueous mobile phases using the ACE C-18AR and Xbridge columns, respectively, were compared. CAP and FF chemical standards showed no area count differences between the two phases. In matrix samples, FF signal suppression was 30% greater in acidic mobile phase. TAP had 50% greater area counts in the acidic mobile phase. CAP-GLUC had a lower capacity factor in the alkaline mobile phase (6 versus 10) but approximately twice the area counts in chemical standards. Matrix samples showed approximately 20% matrix enhancement and 20% suppression in acidic and alkaline mobile phases, respectively. CAP-GLUC peaks in alkaline mobile phase were asymmetrical and split, and retention times continually decreased within an injection sequence. In contrast, an acidic mobile phase resulted in symmetrical peak shape and no retention time shifts. FFA showed no substantial difference in area counts between the two mobile phases. The capacity factor in acidic mobile phase was less than in alkali mobile phase (1.8 versus 4.7), but matrix suppression (70%) was the same in both phases. Overall, the acidic mobile phase was advantageous relative to the use of the alkaline mobile phase. Cech and Enke [20] reported that ammonium hydroxide use in mass spectrometry results in unstable anion formation and irreproducible ion counts. Fuks et al. [21] reported that glucuronic acid binds to active metal sites in the column and solvent path, altering retention times and peak shapes. The observed results and published theories on the negative aspects of ammonium hydroxide use indicated that an acidic mobile phase was more suitable for further testing. Acetic acid increased signal strength of the parent amphenicols relative to formic acid. Wu et al. [22] speculated that acetic acid improves electrospray negative compound detection relative to formic acid due to acetic acid’s higher pKa and proton acceptor capability. The use of 0.1% acetic acid was integrated into the method. Ammonium formate addition to the mobile phase was tested to determine if matrix effects can be reduced in matrix samples, as observed by Choi et al. [23]. However, even at 0.02 mM, FFA’s signal counts were reduced by approximately 40% in both chemical standards and matrix samples. Other amphenicol’s were not affected. No further investigation was done. The addition of 10% isopropanol to methanol increased signal strength for CAP-GLUC by approximately 20%, and was integrated into the method. BadhuTawiah et al. [24] reported that isopropanol improves electrospray desolvation. 3.1.2. Tissue extraction and cleanup Estimates of amphenicol extraction into water-immiscible organic solvents (ethyl acetate, hexane, and MTBE) when partitioned 1:1 (v/v) against acidic (0.1 M HCl) or alkali (0.1 M NaOH) aqueous solutions were obtained. Hexane did not extract any of

the compounds. In both acidic and alkali solutions, CAP, FF and TAP extraction by ethyl acetate was about 50, 70 and 70% respectively. In acidic solution, CAP-GLUC and FFA extraction into ethyl acetate was 30 and 0% respectively, whereas in alkali solution, it was 0 and 40%, respectively. In both acidic and alkali solutions, CAP, FF and TAP recovery into MTBE was about 90, 70 and 20%, respectively. MTBE did not extract CAP-GLUC and FFA from either solution. The increased solubility of CAP-GLUC, FFA, and TAP in ethyl acetate relative to MTBE may be attributed to ethyl acetate’s higher hydrogen bonding capacity (7.2 MPa versus 5.0 MPa) which is calculated from Hansen solubility parameters [25]. The water immiscible organic solvents did not have universal utility for amphenicol extraction. The method of Van de Riet et al. [26], which uses acetone extraction for analysis of CAP, FF, FFA and TAP from aquatic species’ tissues was tested. Applying this method to liver, recoveries of 70, 60, 30 and 50% were obtained for CAP, FF, FFA and TAP. CAP-GLUC had 0% recovery. The addition of 1% formic acid in acetone increased CAPGLUC recovery to 60%, while FFA recovery was reduced to 6%, with all other compounds being 80% recovered. Acidification gelatinized the tissue, which complicated sample processing. Using 1% formic acid in acetonitrile as extractant, tissue precipitated readily from solution. CAP-GLUC and FFA recovery was 40 and 80% respectively. CAP, FF and TAP recoveries ranged from 50 to 70%. Visible amounts of matrix were present in the dried extracts. Geis-Astegiantte et al. [27] reported that bovine muscle acetonitrile extracts contained approximately 2.9% of the wet-matrix. Assuming animal tissue solids content of approximately 30% [28], 9% of the tissue solids are solubilized. Hall et al. [29] suggested that excessive matrix components negatively impact long term mass spectrometer performance. To reduce matrix content in acetonitrile extracts, three different SPE cleanup treatments were tested: Oasis HLB (polymer hydrophilic–lipophilic balanced), Oasis MCX (mixed-mode, HLB and strong cation-exchange), and Oasis MAX (mixed mode, HLB and strong-anion-exchange). The 150 mg, 30 ␮m particle size cartridge formats were used. Oasis HLB cartridges recovered more than 80% of CAP, FF, and TAP from acidic and alkali aqueous solutions. CAP-GLUC and FFA were recovered only when applied in solutions with pH adjusted to neutralize charges. Oasis MAX cartridges recovered more than 80% of all compounds from alkali aqueous solution but CAP-GLUC recovery decreased to 40% in alkali acetonitrile matrix extracts. It was speculated that this was due to competition for ion-exchange sites from tissue matrix components, since tissue buffering capacity is primarily due to phosphates and organic acids [30]. The Oasis MCX cartridges recovered more than 80% of all compounds in acidic (pH 0.99) from 0.1 to 10.0 ng/g for all amphenicols were obtained. CAP-GLUC and FFA recoveries were 50 and 80%, respectively. After evaporation, aqueous reconstitution, and hexane washing, the extracts no longer plugged SPE cartridges. SPE cartridge rinsing procedures after extract application affected results. Alkali (ammonium hydroxide) aqueous rinses less than 1 mL retained all amphenicols, and reduced matrix effects by 20–50%. A 2-mL MTBE wash step was also added to aid in removal of non-polar material, though the effect of this was not tested. Matyash et al. [38] showed that MTBE is effective for selective extraction of lipids and phospholipids from biological samples. An organic solvent concentration of 10% acetonitrile/10% methanol (v/v) in the final LC extract maximized analyte area counts, likely by reducing non-specific binding to glass walls [39] and disruption of micelles formed by endogenous emulsifier compounds such as bile salts [40]. However, this concentration distorted FFA peak shape, and so a final organic solvent content of 5% acetonitrile/5% methanol (v/v) was chosen. When applied to the additional non-target analytes using the corresponding LC–MS/MS conditions from the drug classes corresponding validated methodologies, the estimated recovery ranges were: ␤-agonists 40–80%; ␤-lactams 0–20%; corticosteroids 40–80%; fluoroquinolones 30–80%; steroid-based growth promoters 5–30%; macrolides 30–60%; NSAIDS 0–10%; sulfonamides 10–70%; and tetracyclines 5–30%. When the aqueous alkaline SPE wash step was omitted, total recoveries increased for: ␤-agonists 60–90%; corticosteroids 50–90%; fluoroquinolones 40–80%; sulfonamides 40–80%; and tetracyclines 30–60%. Because of improved recoveries for several of the drug classes, the alkaline SPE wash

73

step was removed from the method. Recovery and matrix effects for analytes from the ␤-agonists, corticosteroids, fluoroquinolones, sulfonamides, and the tetracycline drug classes were estimated using the labs’s corresponding validated instrumental methodologies for those drug classes. Representative quantitation ion chromatograms for the amphenicols in blank and 0.5× spiked bovine liver are shown in Fig. 2. 3.2. Method characterization CAP and CAP-GLUC relative responses were not affected by the presence of matrix over a 1–20× LOD concentration range (Table 2). FFA chemical standards showed no significant differences in relative responses, but significantly decreased in matrix-matched standards. In contrast to FFA, FF and TAP chemical standards showed significant decreases in relative response, but no significant effects in matrix-matched samples. Zacarias et al. [41] reported a decrease in phospholipid’s ESI negative ionization relative response with increasing concentration. For all amphenicols, matrix relative response at 1× was not significantly different (P < 0.05) from that at 5×. FFA showed substantial matrix suppression (85%) whereas CAPGLUC showed substantial matrix enhancement (70%) (Table 3). It was speculated that co-eluted materials were effective proton acceptors, thereby decreasing FFA ionization but increasing CAPGLUC ionization. The majority of the matrix variance was due to within-experiment (between-species confounded with intra-day) effects. Though matrix effects were substantial for all the compounds, overall method % RSDs for internal standard corrected results were within HorRatr guidelines for intra-lab repeatability [42] (Table 4). The HorRatr is defined as: RSDr = C −0.15 where RSDr is the relative standard deviation, and C is the mass fraction of the analyte in the matrix. Estimated intra-lab repeatability should be within the range of 0.5× to 2.0× of the guideline. The HorRatr is distinguished from the HorRatR which is the traditional Horwitz equation characterizing inter-laboratory reproducibility. The HorRatr by definition is one-half that of the HorRatR . Extraction of the additional tested compounds and quantitation using the instrumental conditions of the drug classes corresponding validated methodologies generally exhibited good linearity with matrix-fortified calibration curves, with most coefficients of determination exceeding 0.98 (Table 5). Repeatability estimates for the majority of the compounds were within HorRatr guidelines. Thus, extraction of compounds from the same test matrices beyond the initially tested chemically diverse subset of compounds (the amphenicols) was feasible. The lack of success with certain drug classes (␤-lactams, steroid-based growth-promoters, NSAIDs) is speculated to be due to the subset amphenicol compounds not having comparable extraction physicochemical properties. 3.3. Method validation 3.3.1. Accuracy/LOC/LOD/LOQ/specificity CAP, FF and TAP accuracy ranged showed good accuracy and repeatability (Table 6). The FFA quantitation ion transition showed a nearby eluting peak that was baseline resolved. In addition to being baseline resolved, the peak had substantially different ion ratios from FFA, and therefore it cannot be mistaken for FFA. Within-expt (intra-day) variance was generally the largest contributor to overall method variance. CAP-GLUC and FFA repeatability statistics at the 0.5× target LOD did not meet HorRatr guidelines for within-lab repeatability, but did meet guidelines at the 5.0×

74

R.W. Fedeniuk et al. / J. Chromatogr. B 991 (2015) 68–78

Fig. 2. Full-scale quantitation ion chromatograms for florfenicol amine, thiamphenicol, florfenicol, chloramphenicol 3-O-␤-d-glucuronide, chloramphenicol, and chloramphenicol-D5 in blank and 0.5× target LOD spikes in bovine liver.

level. Given that both CAP-GLUC and FFA have different chemical properties from that of the internal standard’s (CAP-D5) chemical properties, it is probable that the internal standard did not suitably correct for the analytes’ recovery and matrix effects. The estimated LODs for CAP, CAP-GLUC, FF, FFA, and TAP obtained from the validation experiment data were 0.03, 0.24, 0.11, 0.83, and 0.12 ng/g, respectively. Corresponding LOQs were 0.11, 0.81, 0.38, 2.75, and 0.41 ng/g, respectively. Corresponding LOCs were, 0.1, 0.5, 0.25, 1.0, and 0.25 ng/g, respectively. CAP’s LOD (0.03 ng/g) and LOC (0.1 ng/g) are below the suggested MRPL

guidelines, whereas CAP’s common metabolite CAP-GLUC’s LOD (0.24 ng/g) and LOC (0.5 ng/g) allow for detection and confirmation of the equivalent of 0.3 ng/g CAP without the need for timeconsuming enzymatic digestion procedures. The estimated LODs for CAP, CAP-GLUC, FF, FFA, and TAP from the characterization experiment data were 0.08, 0.27, 0.16, 0.92, and 0.12 ng/g, respectively. Corresponding LOQs were 0.26, 0.89, 0.55, 3.05, and 0.41 ng/g, respectively. The samples for the characterization dataset contained 39 additional compounds from the ␤-agonist, corticosteroid, fluoroquinolone, sulfonamide, and

R.W. Fedeniuk et al. / J. Chromatogr. B 991 (2015) 68–78

75

Table 2 Average analyte relative responses in chemical standards and post-spiked matrix samples. Analyte

Sig. Diff?a

Concentration 1×

5×

CAP Postspike Chem std Sig Diff?a

1 1

0.92 0.93 No

CAP-GLUC Postspike Chem std Sig Diff?a

1 1

FF Postspike Chem std Sig Diff?a

Sig. Diff?a

Concentration 5×

20×

No No

0.92 0.93 No

0.75 0.77 No

1.06 1.11 No

No No

1.06 1.11 No

1 1

0.89 0.71 No

No Yes

FFA Postspike Chem std Sig Diff?a

1 1

0.85 1.02 Yes

TAP Postspike Chem std Sig Diff?a

1 1

0.88 0.85 No

a

Sig. Diff?a

Concentration 1×

20×

No No

1 1

0.75 0.77 No

No No

1.14 1.38 No

No No

1 1

1.14 1.38 No

No No

0.89 0.71 No

0.78 0.40 Yes

No Yes

1 1

0.78 0.40 Yes

Yes Yes

No No

0.85 1.02 Yes

0.70 1.02 Yes

No No

1 1

0.70 1.02 Yes

No No

No Yes

0.88 0.85 No

0.82 0.65 No

No Yes

1 1

0.82 0.65 No

No Yes

Significant difference assessed using a 2-sample t-test assuming unequal variances, n = 3, P < 0.05.

Table 3 Mean and repeatability (expressed as %RSD) of recoveries and matrix effects for chloramphenicol (CAP), chloramphenicol-3-O-␤-D-glucuronide (CAP-GLUC), florfenicol (FF), florfenicol amine (FFA), and thiamphenicol (TAP) at the 1X and 5X (recovery only) target concentrations in liver (bovine, equine and porcine). Analyte

Recovery (1×)

CAP CAP-GLUC FF FFA TAP

Recovery (5×)

Matrix effect (average 1× and 5×)

Mean

RSD (%)

Mean

RSD (%)

Mean

%RSD

83 70 81 105 89

19 45 11 7.6 16

65 50 79 86 91

27 34 19 39 20

−49 70 −66 −85 −66

49 33 19 8.3 10

Table 4 Mean estimated% accuracy for the amphenicol analytes at the 1× and 5× concentrations in bovine, equine and porcine liver from nominal values, and within and amongexperiment variance (expressed as %RSD) obtained during method characterization. Analyte

CAP CAP-GLUC FF FFA TAP

Accuracy (%)

Overall %RSD

Within-expt (between-species, intra-day precision) %RSD

Among-expt (inter-day precision) %RSD

1×

5×

1×

5×

1×

5×

1×

5×

4.7 −13 −2.9 7.6 −6.1

2.0 −12 −1.4 3.3 4.7

14 24 10 35 13

7.5 32 15 33 16

9.7 24 8.0 33 11

2.7 25 15 32 15

9.7 19 5.7 9.0 7.2

7.0 19 0.0 8.7 4.8

tetracycline drug classes, whereas the validation data did not. These compounds are monitored in Canada’s National Chemical Residue Monitoring Program, and therefore are possible/likely interferences. LOD and LOQ estimates from the validation and characterization datasets were not statistically significantly different from each other (P < 0.05), indicating that the method is specific for the target analytes in the presence of probable interferences.

The accuracy, repeatability, ruggedness and specificity statistics support the method’s suitability for meeting the suggested MRPL criteria for detection and confirmation of CAP residues (parent and glucuronide), for quantitative determination and confirmation of FF and TAP residues at concentrations above 0.4 ng/g, and for detection and confirmation of FFA residues at concentrations greater than 1 ng/g.

3.3.2. Ruggedness CAP and FF were not affected by the chosen factors. CAP-GLUC showed a significant decrease of approximately 15% with excessive organic solvent extract drying and a significant increase of about 15% with excessive SPE eluate drying. FFA was affected by changes in several factors; tissue source, aqueous reconstitution solution of the organic solvent extract, number of hexane washes, and SPE cartridge source. TAP was significantly affected by the tissue source.

3.3.3. Stability-extracts CAP and TAP showed no statistically significant changes. CAP-GLUC showed a significant decrease in bovine extracts of approximately 1%/day over 14 days. FF showed a significant increase in bovine extracts of approximately 2%/day over 14 days. FFA showed a significant increase in equine extracts of approximately 3%/day over 14 days. The statistics suggest that analysis should occur as soon as practically possible after preparation.

76

R.W. Fedeniuk et al. / J. Chromatogr. B 991 (2015) 68–78

Table 5 Summary statistics for application of the optimized amphenicols analytical methodology to extraction of analytes from additional drug classes.a Analyte

Analytical range (ng/g)

Accuracy (1×)

Recovery (1×)

R2

Matrix effect (1×)

Mean

RSD (%)

Mean

RSD (%)

Mean

RSD (%)

0.10–4.0 0.10–4.0 0.10–4.0 0.10–4.0 0.10–4.0 0.10–4.0 0.10–4.0 0.50–20 0.10–4.0 0.50–20 0.50–20 0.10–4.0 0.50–20

27 3.0 11 2.3 8.5 9.5 26 2.0 −1.8 9.8 18 7.7 7.7

49 34 35 39 23 50 53 5.4 56 32 39 23 32

66 64 68 67 72 62 62 67 73 68 68 74 65

26 12 9.9 29 15 26 31 27 14 11 14 13 11

−42 −19 −41 −36 −22 −34 −47 −47 −16 −6 −17 −34 26

65 90 46 96 102 113 48 38 165 392 102 57 99

0.96 0.99 0.98 0.97 0.98 0.97 0.99 0.99 0.99 0.99 0.98 0.98 0.98

Corticosteroids Beclomethasone Betamethasone Dexamethasone 20-Dihydroprednisolone 20-Dihydroprednisone Flumethasone Methylprednisolone Prednisolone Prednisone Triamcinolone acetonide

0.50–20 0.50–20 0.50–20 0.50–20 0.50–20 0.50–20 0.50–20 0.50–20 0.50–20 0.50–20

3.9 0.9 21 2.8 −3.3 1.2 20 16 −36 9.2

13 34 33 42 16 24 24 55b 54b 26

38 48 46 73 76 47 47 58 67 43

44 55 54 41 38 55 54 32 28 57

−36 −71 −64 −62 −55 −40 −61 −69 −52 −36

12 28 14 14 15 22 14 12 10 11

0.99 0.98 0.99 0.99 0.99 0.99 0.99 0.98 0.98 0.99

Fluoroquinolones Ciprofloxacin Danofloxacin Enrofloxacin Sarafloxacin

0.50–20 0.50–20 0.50–20 0.50–20

11 0.5 0.7 28

61b 27 20 26

69 64 88 67

20 39 15 16

−24 23 −16 −37

215 184 284 118

0.96 0.93 0.97 0.95

␤-Agonists Brombuterol Cimaterol Clenbuterol Clenpenterol Hydoxymethylclenbuterol Isoxsuprine Mabuterol Ractopamine Ritrodine Salbutamol Terbutaline Tulobuterol Zilpaterol

Sulfonamides Sulfachloropyridazine Sulfadiazine Sulfadimethoxine Sulfadoxine Sulfaethoxypyridazine Sulfamethazine Sulfaquinoxaline Sulfathiazole

5.0–200 5.0–200 5.0–200 5.0–200 5.0–200 5.0–200 5.0–200 5.0–200

11 8.5 9.6 5.6 7.0 8.3 18 9.8

23 35b 25 22 23 18 36b 30

55 54 50 61 59 65 31 60

15 13 21 14 17 8.5 39 14

−38 −42 −27 −27 −27 −23 −30 −54

23 17 33 40 35 41 48 26

0.99 0.99 0.99 0.99 0.99 0.99 0.99 0.99

Tetracyclines Chlortetracycline Doxycycline Oxytetracycline Tetracycline

25–1000 25–1000 25–1000 25–1000

−0.9 −0.8 2.8 14

23 19 18 14

33 41 35 28

17 24 16 30

−50 −25 −36 −26

19 52 27 64

0.99 0.99 0.99 0.91

a b

Recovery and matrix effect estimates were obtained used the LC–MS/MS conditions from lab’s corresponding validated methodologies for those drug classes. %RSD exceeds HORRATr guidelines.

Table 6 Mean estimated % accuracy of amphenicol blind spikes at the low, medium and high ends of their respective analytical ranges in bovine, equine, and porcine liver from nominal values, and within- and among-experiment variances (expressed as %relative standard deviation) obtained during method validation. Analyte

CAP CAP-GLUC FF FFA TAP

Accuracy (%)

Overall %RSD

Within-expt (between-species, intra-day precision) %RSD

Among-expt (inter-day precision) %RSD

Low

Med

High

Low

Med

High

Low

Med

High

Low

Med

High

2 42 8 16 12

−7.6 24 1.3 −12 −1.7

−5.3 11 5.8 −3.0 −1.2

9 73 25 52 24

14 57 17 27 19

13 34 15 32 22

9 51 19 46 20

11 29 17 19 19

9 30 15 32 22

0 52 17 23 13

8 48 0 19 0.4

10 17 4.3 0 0

3.3.4. Stability – tissue storage at −20 ◦ C and freeze/thaw cycles Storage of tissues at −20 ◦ C over 35 days showed no significant concentration changes for all analytes. FF showed a significant decrease of about 20% after 2 freeze–thaw cycles in bovine liver. CAP-GLUC showed a significant decrease of about 50% after 2 freeze–thaw cycles in bovine and porcine liver. CAP showed a significant increase of about 20% in bovine liver, and 10% in equine

liver. Given that CAP and FF are known to be readily metabolized, the concentration changes in metabolically active tissues were expected. The lack of residue concentration changes in tissues that remained frozen over 35 days, while residue concentrations changed when tissues were exposed to freeze–thaw cycles, suggests that sample preparation and analysis procedures should be done in a manner to prevent repeated freeze–thaw cycles.

R.W. Fedeniuk et al. / J. Chromatogr. B 991 (2015) 68–78

4. Conclusions The simultaneous determination and confirmation of residues of the banned compound CAP to 0.1 ng/g, and confirmation of the presence of its metabolite CAP-GLUC to 0.5 ng/g, allows for compliance with international guidelines for method minimum required performance limits while increasing the spectrum of detection to a common metabolite without the need for time-consuming enzymatic digestion steps. Method variance was predominantly due to variability between-species (intra-day) rather than variations in method application (inter-day). Utilization of Hansen solubility parameters for improving organic solvent extraction properties aided the rational development of a modified QuEChERS method for the determination and confirmation of ionized and polar analytes from liver. Utilizing a subset of chemically diverse analytes during method development allows for repeatable extraction of analytes with similar chemical properties. The implications are that the logistics of developing a multi-residue/multi-class analytical method may be simplified by choosing appropriate chemical property matched representatives, subsequently decreasing standards preparation and data processing requirements. Acknowledgements This project was financially supported by funding from the Canadian Food Inspection Agency. Suggestions from Ryan Denkert, Johanna Matus, and Kathleen Verity during method development, and instrumental analysis performed by Roger Munro for the additional tested drug classes is appreciated. The rest of the staff of CFIA Saskatoon is acknowledged for their work in maintaining laboratory operations. References [1] J. Wongtavatchai, J.G. McLean, F. Ramos, D. Arnold, WHO Food Additive Series: 53. Chloramphenicol, Joint Expert Committee on Food Additives, 2004, http://www.inchem.org/documents/jecfa/jecmono/v53je03.htm (accessed 26.08.14). [2] List of Maximum Residue Limits (MRLs) for Veterinary Drugs in Foods. Health Canada, Government of Canada. http://www.hc-sc.gc.ca/dhp-mps/vet/mrllmr/mrl-lmr versus new-nouveau-eng.php (accessed 07.03.14). [3] List of Banned Drugs. Section B.01.048. Health Canada, Government of Canada. http://www.hc-sc.gc.ca/dhp-mps/vet/banned drugs list interdit medicaments-eng.php (accessed 26.08.14). [4] Report of the twenty-first session of the Codex Committee on Residues of Veterinary Drugs in Foods. Minneapolis, USA, CL 2013/26-RVDF. Codex Alimentarius Commission, Codex Secretariat, Rome, Italy, 2013, August. [5] CRL Guidance Paper (7 December 2007). CRLs view on state of the art in analytical methods for national residue control plans. http:// www.bvl.bund.de/SharedDocs/Downloads/09 Untersuchungen/EURL Empfehlungen Konzentrationsauswahl Methodenvalierungen EN.pdf? blob=publicationFile&v=2 (accessed 03.07.14). [6] R.M. Parker, I.C. Shaw, Determination of chloramphenicol in tissues – problems with in vitro metabolism, Analyst 113 (1988) 1875–1876. [7] G.O. Korsrud, J.M. Naylor, J.D. MacNeil, W.D.G. Yates, Persistence of chloramphenicol residues in calf tissues, Can. J. Vet. Res. 51 (1987) 316–318. [8] L. Lymas, D. Currie, C.T. Elliot, J.D. McEvoy, S.A. Hewitt, Screening for chloramphenicol residues in the tissues and fluids of treated cattle by the four plate test, Charm II radioimmunoassay and Ridascreen CAP-Glucuronid enzyme immunoassay, Analyst 123 (1998) 2773–2777. [9] S. Mehdizadeh, H.R. Kazerani, A. Jamshidi, Screening of chloramphenicol residues in broiler chickens slaughtered in an industrial poultry abattoir in Mashad, Iran, Iran. J. Vet. Sci. Technol. 2 (2010) 25–32. [10] M. Chen, D. Howe, B. Leduc, S. Kerr, D.A. Williams, Identification and characterization of two chloramphenicol glucuronides from the in vitro glucuronidation of chloramphenicol in human liver microsomes, Xenobiotica 37 (2007) 954–971. [11] C. Pouech, M. Tournier, N. Quignot, A. Kiss, L. Wiest, F. Lafay, M.M. FlamentWaton, E. Lemazurier, C. Cren-Olivé, Multi-residue analysis of free and conjugated hormones and endocrine disruptors in rat testis by QuEChERSbased extraction and LC–MS/MS, Anal. Bioanal. Chem. 402 (2012) 2777–2788. [12] J. Wang, D. Leung, The challenges of developing a generic extraction procedure to analyze multi-class veterinary drug residues in milk and honey using ultra-high pressure liquid chromatography quadrupole time-of-flight mass spectrometry, Drug Test. Anal. 4 (2012) 103–111.

77

[13] B.J.A. Berendsen, T. Zuidema, J. de Jong, L.A.M. Stolker, M.W.F. Nielen, Discrimination of eight chloramphenicol isomers by liquid chromatography tandem mass spectrometry in order to investigate the natural occurrence of chloramphenicol, Anal. Chim. Acta 700 (2011) 78–85. [14] Waters Corporation, Quattro LCZ, Ultima, and Ultima PT Performance Maintenance Protocol. P/N 715000704, 2012, http://www.waters.com/waters/ (accessed keywordSearch.htm?cid=505548&q=715000704&locale=en CA 14.10.14). [15] D.D. Frey, F. Engelhardt, E.M. Greitzer, A role for one-factor-at-a-time experimentation in parameter design, Res. Eng. Des. 14 (2003) 65–74. [16] J.N. Miller, J.C. Miller, Chapter 5. Calibration methods: regression and correlation, in: J.N. Miller, J.C. Miller (Eds.), Statistics and Chemometrics for Analytical Chemistry, Pearson Education Limited, Essex, England, 2005, pp. 107–149. [17] Commission Decision of 12 August 2002 implementing Council Directive 96/23/EC concerning the performance of analytical methods and the interpretation of results. 2002/657/EC. http://eur-lex.europa.eu/LexUriServ/ LexUriServ.do?uri=OJ:L:2002:221:0008:0036:EN:PDF (accessed 05.03.14). [18] J.N. Miller, J.C. Miller, Chapter 4. The quality of analytical measurements, in: J.N. Miller, J.C. Miller (Eds.), Statistics and Chemometrics for Analytical Chemistry, Pearson Education Limited, Essex, England, 2005, pp. 74–106. [19] J.N. Miller, J.C. Miller, Chapter 7. Experimental design and optimization, in: J.N. Miller, J.C. Miller (Eds.), Statistics and Chemometrics for Analytical Chemistry, Pearson Education Limited, Essex, England, 2005, pp. 181–212. [20] N.B. Cech, C.G. Enke, Practical implications of some recent studies in electrospray ionization fundamentals, Mass Spectrom. Rev. 20 (2001) 362–387. [21] L. Fuks, D. Filipuk, W. Lewandowski, Lanthanide ions complexation by uronic acids, J. Mol. Struct. 563–564 (2001) 587–593. [22] Z. Wu, W. Gao, M.A. Phelps, D. Wu, D.D. Miller, J.T. Dalton, Favorable effects of weak acids on negative-ion electrospray mass spectrometry, Anal. Chem. 76 (2004) 839–847. [23] B.K. Choi, D.M. Hercules, A.I. Gusev, LC–MS/MS suppression effects in the analysis of pesticides in complex environmental matrices, Fresenius J. Anal. Chem. 369 (2001) 370–374. [24] A. Badhu-Tawiah, C. Bland, D.I. Campbell, R.G. Cooks, Non-aqueous spray solvents and solubility effects in desorption electrospray ionization, J. Am. Soc. Mass Spectrom. 21 (2010) 572–579. [25] Appendix A: Table A.1, in: C.M. Hansen (Ed.), Hansen Solubility Parameters: A User’s Handbook, 2nd ed., CRC Press, Inc., Boca Raton, FL, USA, 2007, pp. 345–483. [26] J.M. Van de Riet, R.A. Potter, M. Christie-Fougere, G. Burns, Simultaneous determination of residues of chloramphenicol, thiamphenicol, florfenicol, and florfenicol amine in farmed aquatic species by liquid chromatography/mass spectrometry, J. AOAC. Int. 86 (2003) 510–512. [27] L. Geis-Asteggiante, S.J. Lehotay, A.R. Lightfield, T. Dutko, C. Ng, L. Bluhm, Ruggedness testing and validation of a practical analytical method for >100 veterinary drug residues in bovine muscle by ultrahigh performance liquid chromatography–tandem mass spectrometry, J. Chromatogr. A 1258 (2012) 43–54. [28] H.O. Hultin, Chapter 12. Characteristics of muscle tissue, in: O.R. Fennema (Ed.), Food Chemistry, 2nd ed., Marcel Dekker, Inc., New York, NY, USA, 1985, pp. 725–789. [29] T.G. Hall, I. Smukste, K.R. Bresciano, Y. Wang, D. McKearn, R.E. Savage, Chapter 19. Identifying and overcoming matrix effects in drug discovery and development, in: J.K. Prasain (Ed.), Tandem Mass Spectrometry – Applications and Principles, InTech Europe, Rijeka, Croatia, 2012, http://dx.doi.org/10.5772/ 32108, Available from: http://www.intechopen.com/books/tandem-massspectrometry-applications-and-principles/identifying-and-overcomingmatrix-effects-in-drug-discovery-and-development [30] M. Kylä-Puhju, M. Ruusunen, R. Kivikari, E. Puolanne, The buffering capacity of porcine muscles, Meat Sci. 67 (2004) 587–593. [31] K.J. Bisceglia, A.L. Roberts, M.M. Schantz, K.A. Lippa, Quantification of drugs of abuse in municipal wastewater via SPE and direct injection liquid chromatography mass spectrometry, Anal. Bioanal. Chem. 398 (2010) 2701–2712. [32] E. Chambers, D.M. Wagrowski-Diehl, Z. Lu, J.R. Mazzeo, Systematic and comprehensive strategy for reducing matrix effects in LC/MS/MS analyses, J. Chromatogr. B 852 (2007) 22–34. [33] F.J. Schenck, P. Callery, P.M. Gannett, J. Daft, S.J. Lehotay, Comparison of magnesium sulfate and sodium sulfate for removal of water from pesticide extracts of foods, J. AOAC Int. 85 (2002) 1177–1180. [34] B.J.A. Berendsen, L.A.M. Stolker, M.W.F. Nielen, Selectivity in the sample preparation for the analysis of drug residues in products of animal origin using LC–MS, Trends Anal. Chem. 43 (2013) 229–239. [35] A.M.J. Crowell, M.J. Wall, A.A. Doucette, Maximizing recovery of watersoluble proteins through acetone precipitation, Anal. Chim. Acta 796 (2013) 48–54. ˜ E. Prabhakaran, J.M. Scholtz, Protein structure, stability and [36] C.N. Pace, S. Trevino, solubility in water and other solvents, Phil. Trans. R. Soc. Lond. B 359 (2004) 1225–1235. [37] J.T. Chin, S.L. Wheeler, A.M. Klibanov, Communication to the editor on protein solubility in organic solvents, Biotechnol. Bioeng. 44 (1994) 140–145. [38] V. Matyash, G. Liebisch, T.V. Kurzchalia, A. Shevchenko, D. Schwudke, Lipid extraction by methyl-tert-butyl ether for high-throughput lipidomics, J. Lipid Res. 49 (2008) 1137–1146.

78

R.W. Fedeniuk et al. / J. Chromatogr. B 991 (2015) 68–78

[39] M. Pawula, G. Hawthorne, G.T. Smith, H.M. Hill, Chapter 12. Best practice in biological sample collection, processing, and storage for LC–MS in bioanalysis of drugs, in: W. Li, J. Zhang, F.L.S. Tse (Eds.), Handbook of LC–MS Bioanalysis: Best Practices, Experimental Protocols, and Regulations, John Wiley and Sons, Inc., Hoboken, NJ, USA, 2013, pp. 139–164. [40] M.G. Khaledi, Review. Micelles as separation media in high-performance liquid chromatography and high-performance capillary electrophoresis: overview and perspective, J. Chromatogr. A 780 (1997) 3–40.

[41] Zacharias, D. Bolanowski, A. Bhatnagar, Comparative measurements of multicomponent phospholipid mixtures by electrospray spectroscopy: relating ion intensity to concentration, Anal. Biochem. 308 (2002) 152–159. [42] AOAC Guidelines for Single Laboratory Validation of Chemical Methods for Dietary Supplements and Botanicals, 2002, http://www.aoac.org/imis15 prod/ AOAC Docs/StandardsDevelopment/SLV Guidelines Dietary Supplements.pdf (accessed 07.03.14).