Article pubs.acs.org/JAFC

Development of an Ultraperformance Liquid Chromatography− Tandem Mass Spectrometry Method for the Analysis of Perfluorinated Compounds in Fish and Fatty Food Stephen W. C. Chung* and Chi Ho Lam Food Research Laboratory, Food and Environmental Hygiene Department, 4/F Public Health Laboratory Centre, 382 Nam Cheong Street, Hong Kong SAR, China S Supporting Information *

ABSTRACT: This paper reports the development of a method for the quantitative analysis of perfluorinated compounds (PFCs), including C6−C14 perfluorinated carboxylic acids (PFCAs) and C4−C12 perfluorinated sulfonates (PFSAs), in fish and fatty foods by ultraperformance liquid chromatography−electrospray ionization−tandem mass spectrometry (UPLC-ESI-MS/ MS) in which the UPLC was equipped a PFC Analysis Kit to eliminate background contamination. Rapid baseline separation was achieved for 17 PFCs within 12 min, and PFCs were well-resolved from potential interferences from taurodeoxycholic acid and branched isomers of PFCs. The method was validated according to Commission Regulation 2002/657/EC of the European Commission with matrices including salmon, beef, egg, and butter. Average spiked recoveries, measured at concentration levels of 0.02 (method limit of quantification (MLOQ)), 0.2, and 2 μg/kg, were in the range of 68−117% with relative standard deviations below 20%. Matrix effects were evaluated and found to be correctable by internal standardization, especially for short- and longchained PFCs. Trueness was verified against two certified reference materials. The method has also been successfully applied to the analysis of more than 200 food samples of a risk assessment study. KEYWORDS: perfluorinated compounds, PFCAs, PFSAs, fish, fatty food, UPLC-MS/MS

■

The Netherlands.22 Although eight labeled standards were used, the correction of hydrophilic short-chained and hydrophobic long-chained PFCs for a range of complex and heterogeneous matrices was still found to inefficient. Besides, phase separation occurred for fatty fish samples. According to the European Food Safety Authority (EFSA) report,23 most occurrence data of PFOS were reported without information on the isomer types analyzed: linear, branched, or both types. Owing to differences in bioaccumulation potentials,24 linear and branched isomers should be analyzed separately for better estimation of their dietary exposure. The developed ultraperformance liquid chromatography−mass spectrometry (UPLC-MS) method resolved linear PFOS from 10 different branched isomers and well-known interference taurodeoxycholic acid (TDCA). Besides, this method can quantify nine PFCAs and eight PFSAs within 12 min. Validation was performed with fish, meat, egg, and butter according to the requirement of the Commission Decision 2002/657/EC of the European Commission (EC), and the results fulfilled the recommendations in Commission Recommendation 2010/161/EU on monitoring of perfluoroalkylated substances in food. Furthermore, method accuracy and extraction efficiency were verified by two fish certified reference materials obtained from the National Institute of Standards and Technology.

INTRODUCTION As perfluorooctanesulfonate (PFOS) and perfluoroctanoic acid (PFOA) are persistent and can bioaccumulate in animals, it is not surprising that various studies reported the general public is exposed to these two chemicals.1,2 Food and water were found to be the main sources of exposure. Studies conducted by various authorities around the world, including Canada,3 China,4 the United States,5 the United Kingdom,6 Germany,7 The Netherlands,8 Spain,9 Sweden,10 and the European Union,11 found that PFOS and PFOA can present in certain foods up to 20 μg/kg. Some of these reports stated that fish (marine and fresh water) contain higher levels of these two PFCs. Studies also found some fatty foods including beef, offal (e.g., liver and kidney), butter, etc., can have PFOS/PFOA. Owing to the lack of native and labeled reference standards, early studies of PFCs in fish were concentrated in PFOS, PFOA, and PFCAs.12−16 More standards are available recently, including perfluoropentanesulfonate, perfluorononanesulfonate, and perfluorododecanesulfonate. Ion-pair extraction13,15,17 and solid−liquid extraction12,14,18,19 were commonly used. Both approaches coextract other matrix constituents, including lipids, and affect the mass spectrometric analysis (signal suppression).20 Alkaline digestion can reduce the interference from fatty matrix, but is time-consuming.7,18 Besides, the former extraction method may have the problem of loss of analytes in the solvent exchange process,13,17 whereas the latter would show low recoveries for long- and short-chain PFCs.19 In 2010, Ballesteros-Gomez et al. 21 introduced the THF−water extraction method for the analysis of 14 PFCs in food and applied it to the dietary intake study of PFOS and PFOA in © 2014 American Chemical Society

Received: January 2, 2014 Accepted: June 5, 2014 Published: June 5, 2014 5805

dx.doi.org/10.1021/jf502326h | J. Agric. Food Chem. 2014, 62, 5805−5811

Journal of Agricultural and Food Chemistry

■

Article

maximal recovery of the eluate. The eluate was evaporated to dryness (50 °C, N2) and reconstituted with 0.5 mL of MeCN/water 1:1 (v/v). Finally, the extract was filtered by syringe (Terumo, Leuven, Belgium) with a 0.2 μm regenerated cellulose syringe filter (Sartorius AG, Goettingen, Germany) into a PP LC vial with a PP screw cap (Waters) for UPLC-MS/MS analysis. For fat/oil samples, no defatting and freeze-drying were done. The rest of the sample preparation was the same as described above. Ultraperformance Liquid Chromatography−Mass Spectrometry. The chromatographic separation was carried out using a Waters Acquity UPLC system, which consisted of a sample manager, a column manager, and a binary solvent manager (Milford, MA, USA) equipped with a UPLC column with perfluoro-C8 stationary phase, Epic FO LB, 1.8 μm, 5 cm × 2.1 mm (ES Industries Chromega, West Berlin, NJ, USA). Column temperature was set at 35 °C. Gradient elution was made with MeCN and aqueous buffer with 1.3 mM ammonium formate and 0.7 mM formic acid at a constant flow rate of 0.45 mL/min. Initial mobile phase composition of 13% MeCN was held for 3 min. MeCN composition was linearly increased to 55% at 7.5 min and subsequently to 95% at 10 min. Washing with 95% MeCN was held from 2 min. MeCN composition was then reduced back to 13% and followed by a 3 min reconditioning of the column. The total run time was 15 min. Five microliters of extract was injected in this study. Besides, the UPLC system was installed with an Acquity PFCs analysis kit. Epic FO LB was used as an isolation column. The UPLC system was coupled to a Qtrap 5500 triple-quadrupole mass spectrometer (AB Sciex, Framingham, MA, USA) equipped with a TurboV ion source for analysis. The data system was installed with Analyst software (version 1.5.2). The tandem mass spectrometer was operated in negative electrospray ionization mode (−ve ESI-MS/MS). Scheduled multiple reaction monitoring algorithm mode was used to maximize the instrument performance. Ion spray voltage was set at −4 kV. Nitrogen was used as collision gas and set at medium. Curtain gases, GS1 and GS2, were set at 20, 50, and 50, respectively. Source temperature was set at 600 °C. Both quadrupole 1 (Q1) and quadrupole 3 (Q3) resolution were set as unit. The entrance potential was set as −10 V. The optimized settings for collision energy (CE) and the declustering (DP), entrance (EP), and collision cell exit potentials (CXP) were tested for each target analyte and their ISs by flow injection analysis resulting from the most abundant precursor ion and are summarized in the Supporting Information (Table S1), as well as the indicative retention times on the column. All PFCAs were monitored as their deprotonated ions ([M − H]−), whereas all PFSA anions were directly monitored ([M]−). Softwares Acquity UPLC Console (Waters) and Analyst (AB Sciex) were used to operate the UPLC and MS/MS, respectively. Software MultiQuant (AB Sciex) was used for data processing. Internal standardization was used for PFC quantification. Matching of native and labeled standards is summarized in Table S1 in the Supporting Information. Fifteen-point calibration curves with concentrations from 0.05 to 50 μg/L (which covered 3 orders of magnitude) using weighted linear regression were established for quantification. Two MRM transitions were monitored to fulfill the identification points and ion ratio requirement. The maximum permitted tolerances on the MRM ratio as well as relative retention time (RRT) deviation tolerances (2.5%) followed 2002/657/EC guidelines. The minimum signal-to-noise ratio for both quantification and confirmation MRM of a quantified signal was 3.

MATERIALS AND METHODS

Chemicals and Standards. LC-MS grade acetonitrile (MeCN) and methanol (MeOH) were purchased from Anaqua Chemicals Supply (Houston, TX, USA) and Fisher Scientific (Waltham, MA, USA), respectively. Water was purified through a Milli-Q synthesis system integral with LC-Pak polisher from Millipore (Billerica, MA, USA). Absolv. grade n-hexane was purchased from Tedia Co. (Fairfiled, OH, USA). Analytical reagent grade ammonia solution was purchased from VWR international (Poole, UK). HPLC grade tetrahydrofuran (THF), ammonium acetate, acetic acid, ammonium formate, and formic acid were purchased from Sigma-Aldrich (St. Louis, MO, USA). Native perfluorinated compounds (PFCs), including perfluoro-nhexanoic/heptanoic/octanoic/nonanoic/decanoic/undecanoic/dodecanoic/tridecanoic/tetradecanoic acid (PFHx/Hp/O/N/D/Ud/Do/ TrD/TeDA) and sodium perfluoro-1-butane/pentane/hexane/heptane/octane/nonane/decane/dodecanesulfonate (PFB/Pe/Hx/Hp/ O/N/D/DoS), were purchased from Wellington Laboratories (Ontario, Canada). Isotopically labeled standards perfluoro-n[1,2,3,4,6-13C5]hexanoic acid (13C5PFHxA), perfluoro-n-[1,2,3,4-13C4]heptanoic acid (13C 4 PFHpA), perfluoro-n-[ 13 C 8]octanoic acid (13C8PFOA), perfluoro-n-[13C9]nonanoic acid (13C9PFNA), perfluoro-n-[1,2,3,4,5,6-13C6]decanoic acid (13C6PFDA), perfluoro-n[1,2,3,4,5,6,7- 13 C 7]undecanoic acid ( 13 C 7PFUdA), perfluoro-n[1,2-13C2]dodecanoic acid (13C2PFDoA), perfluoro-n-[1,2-13C2]tetradecanoic acid (13C2PFTeDA), sodium perfluoro-1-[1,2,3-13C3]hexanesulfonate (Na13C3PFHxS), and sodium perfluoro-1-[13C8]octanesulfonate (13C8PFOS) were purchased from Wellington Laboratories (Ontario, Canada), and ammonium perfluoro-[18O2]butanesulfonate (18O2PFBS) was obtained from RTI International (Research Triangle Park, NC, USA). TDCA sodium salt was purchased from Steraloids Inc. (Newport, RI, USA), and perfluoro-1/-2/-3/-4/-5/-6-methylheptanesulfonate (P1/2/3/4/5/6MHpS), perfluoro-2/-3/-4/-5/-6-methylheptanoic acid (P2/3/4/5/6MHpA), perfluoro-4,4/-5,5/-3,5/-4,5-methylhexanesulfonate (P44/55/35/45DMHxS), and perfluoro-4,4/-5,5/-3,5/-4,5methylhexanoic acid (P44/55/35/45DMHxA) in methanol were purchased from Wellington Laboratories. Intermediate mixed standard solutions (10 mg/L) were prepared with methanol. Working standard solutions for calibration were freshly prepared with MeCN/water (5:5 v/v). All solutions were stored at −20 °C. Sample Preparation. A portion of 10 g of homogenized sample was weighed and placed into a 50 mL polypropylene centrifuge (PP) tube (Sarstedt, Nümbrecht, Germany). Isotopically labeled internal standards each of 1 ng were spiked. The sample was then defatted by vortexing with 20 mL of n-hexane for 10 min with a multipulse vortexer (Glas-Col, LLC, Terre Haute, IN, USA). The mixture was centrifuged by a centrifuge (Kendro, Newtown, CT, USA), and the nhexane layer was discarded. The defatted sample was then freeze-dried for 24 h to remove the sample moisture by a freeze-dryer (SP Industries, Warminster, PA, USA). Twenty milliliters of THF/water (75:25) solution was added for extraction. The mixture was vortexed for 10 min and was recentrifuged for 10 min at 8500 rpm (or 6770g) (Falcon 6/300, MSE, UK). Any residual oil layer was discarded. Ten milliliters of extract was pipetted into a 15 mL PP centrifuge tube (Sarstedt). The volume of extract was reduced to 6 mL under a slow stream of nitrogen at 50 °C and then made up to 15 mL with water before cleanup. The 15 mL sample solution was loaded onto a preconditioned weak anion exchange cartridge, Strata X-AW (6 mL, 200 mg) obtained from Phenomenex, at a rate of 1 drop/s. After loading, the cartridge was washed with 4 mL of 25 mM acetate buffer at pH 4 and 6 mL of a 50:50 (v/v) THF/MeCN mixture at a rate of 2 drop/s. Then, the SPE WAX cartridge was coupled to the ENVI-carb cartridge (6 mL, 250 mg), which obtained from Sigma-Aldrich, via a suitable polyethylene (PE) adaptor cap. The target analytes were eluted out with 6 mL of 0.1% ammonia in MeOH. The eluate passed through the graphitized carbon cartridge, too. Cartridges were dried under vacuum to ensure

■

RESULTS AND DISCUSSION Method Development. Despite more and more PFC standards becoming commercially available, PFCs being mainly present in fish and meat, linear and branched PFCs should be quantified separately, and improvement in high-throughput analytical procedure is requested. The key issue of this paper was to develop a new method with good resolution power of PFCs for rapid quantification in fatty foods. As PFPeS (C5), PFNS (C9), and PFDoS (C12) are commercially available, the scope of analysis increased to nine PFCAs (C6−C14) and eight 5806

dx.doi.org/10.1021/jf502326h | J. Agric. Food Chem. 2014, 62, 5805−5811

Journal of Agricultural and Food Chemistry

Article

Figure 1. Chromatograms of the two MRM transitions of (a) PFOA and their branched isomers of a mixed standard solution, (b) PFOS, their branched isomers, and TDCA of a mixed standard solution, and (c) PFOS and their branched isomers of a fish sample. The blue trace refers to 413/ 369 and 499/80 for PFOA and PFOS, respectively, whereas the red trace refers to 413/169 and 499/99, respectively.

PFSAs (C4−C12). SPE WAX conditions were slightly modified for this wider scope of PFCs. We attempted to include perfluorohexadecanoic acid and perfluorooctadecanoic acid as well but did not succeed because their recoveries were comparatively low and none of the internal standards was found to be suitable for correction of recovery. The run time was much shortened after the LC analysis migrated to a UPLC system. Seventeen PFCs were eluted within 12 min. It is worth mentioning that a small change in pH value of the buffering system affected PFC retentions. To ensure reproducibility, the mobile phase buffers were prepared by mixing of acid and salt solution instead of adjusting to the desired pH. Formic acid to ammonium formate ratios of 1:4, 1:2, and 1:1 in mobile phase were tested. PFBS was found to have no retention under buffer ratio 1:4, whereas PFTeDA had a strong retention, which led to undesirable peak boarding under buffer ratio 1:1. Although a systematic multivariate analysis of chromatographic conditions was not performed, the buffer system with an acid-to-salt ratio of 1:2 produced a suitable retention for both short- and long-chain PFCs. Background PFCs Contamination. The UPLC-MS/MS offered excellent sensitivity to the PFCs, and analytes at the nanograms per liter level could be easily detected. As such, stringent control on background contamination from reagents, labware, and instrument was essential for reliable quantification of any trace amount of PFCs detected in samples. LC-MS grade MeOH (LabScan or Fisher) was used in the entire method as significant residual PFOA was noted in HPLC grade MeOH. Using PP centrifuge tubes, non-PTFE SPE vacuum manifold accessories, PP LC vials, and septumless polyethylene caps were necessary to reduce background. An Acquity PFC Analysis Kit

was installed into the Waters UPLC system; those parts such as stainless steel tubings/screws/ferrules/filters of the LC system that contained PTFE parts were replaced by PEEK. An Epic FO LB was connected in-line between the solvent mixer and the injection of the LC system so that the PFCs leached out from the system were retained in it. The same type of column was used for both isolation and analytical separation such that the separation of the leached out PFCs could be ensured. High back pressure was generated by the isolator column of the PFC Analysis Kit and analytical separation column (i.e., two UPLC columns in series). The problem was resolved by replacing MeOH with MeCN as organic mobile phase and setting the analytical column temperature at 35 °C. Trace amounts of PFHxA, PFHpA, and PFOA in method blank were controlled at ∼0.02 μg/L, which was less than the lowest calibration level at 0.05 μg/L in vial (i.e., 1/4 method limit of quantification (MLOQ)), whereas other PFC background contamination was much lower. Method Performance. Specificity. The method specificity, which referred to the discrimination between analytes and closely related substances, was mainly focused on eliminating interference from branched PFCs isomers and TDCA as the method applied to fish and meat. Branched (br-) PFOSs and PFOAs (dimethylhexane/ methylheptane isomers) can also be monitored by the same principle transition from 499 to 80 and from 413 to 369 as PFOS and PFOA, respectively, but with different responses ranging from 7 to 340%. A similar observation was also noted for their secondary transitions from 499 to 99 and from 413 to 169. A systematic trend cannot be identified, and the MRM ratio varied among isomers. Thus, coelution of any branched 5807

dx.doi.org/10.1021/jf502326h | J. Agric. Food Chem. 2014, 62, 5805−5811

Journal of Agricultural and Food Chemistry

Article

Figure 2. Average recovery and precision of sample spike of four matrices.

summarized in Figure 2, with details given in Table S2 in the Supporting Information. The average recoveries in general fulfilled the recommended 70−120% in Commission Recommendation 2010/161/EU. Slightly lower recoveries (>65%) were observed for PFTeDA. Poor recoveries (>50%) and precision (>20%) of PFTrDA and PFTeDA in beef were improved when the recently available 13C2PFTeA was used as internal standard for correction. Method accuracy and trueness were also evaluated by four replicate analyses of certified reference materials (CRMs), fish tissues SRM 1946 and SRM 1947. Only reference values of PFOS were available in the two CRMs, and information values of PFNA, PFDA, PFUnDA, and PFTrDA were available in SRM 1947. Results obtained were compared to these values and are given in Table 1. Good method performance was

isomers could affect quantification as well as confirmation of PFOA/PFOS. Standard solutions of 10 br-PFOSs and 9 brPFOAs at concentrations of 2 ng/mL were tested by the UPLC-MS/MS system and were baseline separated from their straight-chain isomer. Chromatograms of mixed standards of PFOAs and PFOSs are depicted in Figure 1, panels a and b, respectively, to show the separations achieved. From the real sample results, br-PFOS isomers were detected in almost all of the fish and seafood samples. Hence, good chromatographic separation was necessary for analyzing PFOS. An example is depicted in Figure 1c. In some cases, br-PFOS isomers could contribute to >30% of the signal, and overestimation occurred if br-PFOS isomer signals were not resolved. Further separation of branched isomers individually was not attempted. TDCA in matrices of animal origin has been reported to interfere with the 499/80 MRM transition of PFOS, especially for fish. A standard solution of TDCA in 20 ng/mL was analyzed to evaluate the resolution of the UPLC program. It was found to elute much earlier than PFOS (Figure 1b). No interference from TDCA was expected as its retention time fell outside the MRM monitoring window in our finalized method. Interference was noted for the MRM transition from 402 to 80 for 13C3 PFHxS in fish samples. Thus, a matrix interference free of MRM transition of from 402 to 99 was used for correction. Calibration. Linearity was checked through calibration curves, which were constructed by 15 concentration levels from 0.05 to 50 μg/L, equivalent to a working range from 0.005 to 5 μg/kg in sample, with coefficients of determination (R2) values >0.995. Individual residual deviations were