Shape-Selective Extraction of PCBs and Dioxins from Fish and Fish

Mar 8, 2007 - Erik Spinnel , Jerker Fick , Patrik L. Andersson and Peter Haglund .... Martin Hansen , Kristine A. Krogh , Bent Halling-Sørensen , Erl...
2 downloads 0 Views 532KB Size
Download PDF
Anal. Chem. 2007, 79, 2945-2951

Shape-Selective Extraction of PCBs and Dioxins from Fish and Fish Oil Using In-Cell Carbon Fractionation Pressurized Liquid Extraction P. Haglund,*,† S. Sporring,‡ K. Wiberg,† and E. Bjo 1 rklund‡

Department of Chemistry, Umeå University, S-901 87 Umeå, Sweden, and Department of Analytical Chemistry, Lund University, P.O. Box 124, S-221 00 Lund, Sweden

This paper describes a new shape-selective, pressurized liquid extraction (PLE) procedure for extracting polychlorinated dibenzo-p-dioxins and dibenzofurans (PCDD/Fs) and PCBs from food and feed samples with an integrated carbon fractionation step. Initially this was done using specially designed inserts for 34-mL cells, but subsequently, large solid cells (66 mL) were machined to increase the capacity and robustness of the system. Depending on the carbon load and extraction solvent strength, the non-ortho PCBs were recovered either with the bulk of the PCBs or with the PCDD/Fs. The former is preferable if PCDD/Fs are the targets. In most cases, however, data are required for all indicator PCBs, WHOPCBs, and PCDD/Fs. Therefore, further efforts focused on developing, optimizing, and validating a cost- and timeefficient PLE procedure that can extract these targets, separate non-ortho PCBs and PCDD/Fs from the bulk of the PCBs, allow gravimetric fat determinations, and requires a minimum of postextraction cleanup. The performance of the resulting procedure was assessed in experiments with a fish tissue reference material. The trueness of the WHO-PCB-TEQ, PCDD/F-TEQ, and totalTEQ data were - 8, - 5, and - 7%, respectively, and the corresponding CVs were 1.5, 0.5, and 1.3%; within the limits set by the European community for gas chromatography-high-resolution mass spectrometry methods for food and feed control. Since the Belgian dioxin crisis in 1999,1 there has been increasing demand for fast, reliable, and cost-efficient analytical methods to determine polychlorinated dibenzo-p-dioxins (PCDDs), polychlorinated dibenzofurans (PCDFs), and non- and mono-ortho chlorine-substituted PCBs (non- and mono-ortho PCBs or WHOPCBs) in food and feed samples. The traditional methods, such as open-column extraction and Soxhlet extraction, often require long extraction times and tedious cleanup procedures before extracts can be analyzed.2 Another drawback of the traditional methods is that they consume large amounts of very clean, costly, * To whom correspondence should be addressed. E-mail: peter.haglund@ chem.umu.se. Fax: +46-90-128133. † Umeå University. ‡ Lund University. (1) European Commission. Official J. Eur. Commission L 310 1999, 62-70. (2) Liem, D. TrAC, Trends Anal. Chem. 1999, 7, 499-507. 10.1021/ac0624501 CCC: $37.00 Published on Web 03/08/2007

© 2007 American Chemical Society

organic solvents. An alternative method, pressurized liquid extraction (PLE), has been shown to have considerable potential for the extraction of persistent organic pollutants in environmental samples3,4 as well as in food and feed samples.5 PLE dramatically reduces the extraction time and the amount of solvents required for the extractions. Recently, a relatively new PLE approach with in-cell cleanup was published by several authors,6-11 in which a fat retainer (acidic alumina, neutral alumina, basic alumina, Florisil, or silica impregnated with concentrated H2SO4) is incorporated into the extraction cell for fat-free extraction of PCBs from food or feed samples.6 We have previously optimized the extraction times and number of cycles required to efficiently extract PCBs from a lipid matrix using selective PLE.7 In that study, and most of the other cited studies, the Dionex ASE 200 system was used with 33-mL extraction cells. Later, Sporring et al.8,9 used a Dionex ASE 300 and larger cells (100 mL) to increase the sample capacity and allow the selective extraction of PCBs from lightly contaminated food and feed samples. These cells also hold sufficient material to allow analysis of PCDD/Fs in various food and feed samples.10,11 The main improvements, compared to the traditional methods, are the reductions in solvent consumption and the times required for extraction and cleanup. The extracts obtained only require carbon column fractionation and miniaturized multilayer silica column cleanup. However, the main advantage of the system, that all fat is retained in the extraction cell, may also be a drawback. If the results have to be presented on a lipid weight basis, a separate lipid weight determination is needed, so an additional extraction has to be done. An alternative system has been developed by Focant et al.,12 based on PLE extraction and on-line cleanup and fractionation (3) Bjo ¨rklund, E.; Nilsson, T.; Bøwadt, S. TrAC, Trends Anal. Chem. 2000, 19, 434-445. (4) Ramos, L.; Kristenson, E. M.; Brinkman, U. A. Th. J. Chromatogr., A 2002, 975, 3-29. (5) Carabias-Martinez, R.; Rodrigues-Gonzalo, E.; Revilla-Ruiz, P.; HernandezMendez, J. J. Chromatogr., A 2005, 1089, 1-17. (6) Bjo ¨rklund, E.; Mu ¨ ller, A.; von Holst, C. Anal. Chem. 2001, 73, 4050-53. (7) Mu ¨ ller, A.; Bjo¨rklund, E.; von Holst, C. J. Chromatogr., A 2001, 925, 197205. (8) Sporring, S.; Bjo ¨rklund, E. J. Chromatogr., A 2004, 1040, 55-161. (9) Sporring, S.; von Holst, C.; Bjo ¨rklund, E. Chromatographia 2006, 64, 553557. (10) Wiberg, K.; Sporring, S.; Haglund, P.; Bjo ¨rklund, E. J. Chromatogr., A 2006, 1138, 55-64. (11) Bernsmann, T.; Fu ¨ rst, P. Organohalogen Comp. 2004, 66, 157-161. (12) Focant, J. F.; Pirard, C.; De Pauw, E. Talanta 2004, 63, 1101-1113.

Analytical Chemistry, Vol. 79, No. 7, April 1, 2007 2945

using several columns (multilayer silica, basic alumina, and carbon). Column switching is used to direct the flow between the columns and to collect fractions containing WHO-PCBs and PCDD/Fs. Although fully automated, it still has drawbacks. It is costly, since it utilizes expensive prepacked columns, it consumes amounts of solvent similar to the traditional methods, and as described above for the selective PLE procedure,10 it is not possible to perform a lipid weight determination. In this paper, the development, optimization, and validation of a new shape-selective extraction method for PCDD/Fs and PCBs in food or feed samples is described. By using specially designed inserts for the PLE cells (34 mL), it was possible to integrate the carbon fractionation step into the extraction cell. Subsequently, larger cells (66 mL) were developed, to further improve the capacity and robustness of the shape-selective extraction procedure. Based on experience from open-column carbon fractionation,13,14 it was initially assumed that the target analytes could be separated into three fractions containing the bulk of the PCBs (bulk-PCBs), mono-ortho PCBs, and non-ortho PCBs and PCDD/ Fs, respectively. Later it was realized that matrix constituents (primarily lipids) compete effectively with mono-ortho PCBs for carbon adsorption sites and prevent class separation of bulk and mono-ortho PCBs. Therefore, further efforts focused on developing, optimizing, and validating a selective PLE procedure that can extract all target analytes, separate ortho-chlorine-substituted PCBs from non-ortho PCBs and PCDD/Fs, allow lipid weight determinations to be made, and require a minimum of postextraction cleanup. This paper reports the results of these efforts and presents a cost- and time-efficient strategy for extraction, cleanup, fractionation, and analysis of regulatory indicator PCBs (CBs 28, 52, 101, 138, 153, and 180; I-PCBs), WHO-PCBs, and PCDD/Fs present in biological samples. EXPERIMENTAL SECTION Samples. Two fish oils were used during the method development: a herring oil, originating from the EU project DIFFERENCE15 (fish oil 1); and a commercially available dietary supplement cod liver oil, Mo¨ller’s Tran (Peter Mo¨ller, Oslo, Norway) that was bought in a local store (fish oil 2). In addition, the inhouse reference material used at Umeå University, a homogenate of wild salmon tissue and sodium sulfate (1:5, w/w) (fish tissue), was used for method validation. Chemicals. Sodium sulfate (p.a. >99%), methanol, and dichloromethane (DCM) were purchased from Fluka Chemica, and the sodium sulfate was baked at 400 °C for 10 h prior to use. Acetone D (for analysis of dioxins), n-heptane (Pestanal grade), toluene (Pestanal grade), and Celite were obtained from Riedel-de Hae¨n. GF/A glass microfiber filters for covering cell caps came from Whatman International Ltd. (Maidstone, England). Activated AX21 carbon originated from Anderson Development. The PCDD/F solutions used for quantification originated from Wellington Laboratories (Guelph, ON, Canada), and the non-ortho and monoortho PCB solutions were from Cambridge Isotope Laboratories (13) Danielsson, C.; Wiberg, K.; Koryta´r, P.; Bergek, S.; Brinkman, U. A. Th.; Haglund, P. Chromatogr., A 2005, 1086, 61-70. (14) Smith, L. M.; Stalling, D. L.; Johnson, J. L. Anal. Chem. 1984, 56, 18301842. (15) van Loco, J.; van Leeuwen, S. P. J.; Roos, P.; Carbonelle, S.; de Boer, J.; Goeyens, L.; Beernaert, H. Talanta 2004, 63, 1169-1182.

2946 Analytical Chemistry, Vol. 79, No. 7, April 1, 2007

Figure 1. Packing of the PLE cells (left) and expected elution sequence when applying selective PLE. The second fraction was extracted with bisolvent mixtures of either n-heptane/dichloromethane or n-heptane/acetone.

(Andover, MA). The amounts and origin of the 13C-labeled internal standards and syringe spikes used have been published elsewhere.10 Equipment. (a) PLE Extraction and Sample Cleanup. All extractions were performed using an ASE 300 system (Dionex, Sunnyvale, CA) with an extraction temperature of 100 °C, a purge time of 90 s, a flush volume of 60%, and an extraction time of 5 min. The cells were packed, as illustrated in Figure 1, from the bottom up. When the carbon/Celite had been added, the cell was connected to a vacuum pump and the carbon was conditioned by drawing 4 mL of DCM/methanol/toluene (15:4:1, v/v/v), 1 mL of DCM/n-heptane (1:1, v/v), and 5 mL of n-heptane through it. Next, the sample was added. The herring oil and the cod liver oil were applied to Na2SO4 in the cell. The salmon tissue was added as a Na2SO4 homogenate. Internal standards were added on top of the samples, and the dead volume was filled with Na2SO4. Initially, especially designed inserts for the PLE cells (34 mL) were used to integrate the carbon fractionation step with the extraction. Subsequently, larger cells (66 mL) were developed, to further improve the capacity and robustness of the shapeselective extraction procedure. Table 1 presents a detailed description of the extraction sequences used in trials with both the inserts and cells. For the fractions where a fat determination was needed, a subsample was taken and the fat content was determined gravimetrically by weighing the residue after complete solvent removal. Finally, lipids and other polar residues were removed by passing the extracts through multilayer silica columns. Large columns (30mm i.d.) were used for the first (lipid-rich) fractions, intermediate columns (16-mm i.d.) for the middle fractions, and small columns (8-mm i.d.) for the last (PCDD/F-containing) fractions. Each of these columns was filled with 2 cm of KOH-impregnated silica (30%, w/w), 1 cm of silica, and 3 cm of concentrated H2SO4impregnated silica (40%, w/w). Prior to cleanup, the columns were rinsed with two bed volumes of n-heptane, and after applying the sample, they were eluted with two bed volumes of n-heptane. Finally, the solvent was evaporated to 1 mL, and each fraction

Table 1. Parameters for the PLE Extraction Experimentsa

trial

n

cell vol (mL)

1 2:1 2:2 2:3 2:4 3:1 3:2 4:1 4:2 4:3 5

1 1 1 1 1 3 1 2 2 1 3

34 34 34 34 34 66 66 66 66 66 66

fraction 1 2 × n-heptane 2 × n-heptane 2 × n-heptane 2 × n-heptane 2 × n-heptane 2 × n-heptane 2 × n-heptane 2 × n-heptane 2 × n-heptane 2 × n-heptane 2 × n-heptane

cycles × solvent fraction 2 2 × n-heptane/DCM (1:1) 1 × n-heptane/DCM (9:1) 1 × n-heptane/DCM (9:1) 2 × n-heptane/DCM (1:1) 2 × n-heptane/DCM (1:1) 1 × n-heptane/DCM (9:1) 1 × n-heptane/acetone (1:2.5) 1 × n-heptane/acetone (1:2.5) 1 × n-heptane/acetone (2.5:1) 1 × n-heptane/acetone (1:2.5) 1 × n-heptane/acetone (1:2.5)

fraction 3

carbon/Celite, % C/total mass (g)

sample

sample amt (g)

2 × toluene 2 × toluene 2 × toluene 2 × toluene 2 × toluene 4 × toluene 4 × toluene 4 × toluene 4 × toluene 4 × toluene 4 × toluene

7.9/0.5 7.9/0.5 25/0.5 25/0.5 50/0.5 25/1.5 25/1.5 25/2.4 25/2.4 25/1.5 25/2.4

fish oil 1 fish oil 2 fish oil 2 fish oil 2 fish oil 2 fish oil 2 fish oil 2 fish oil 2 fish oil 2 fish oil 2 fish tissue

3 3 3 3 3 3 3 3 3 3 15

a The experiments were performed in five series of trials. The numbers after the colons refer to experiments within the series (e.g., trial 2:4 is the fourth experiment in trial 2) and n is the number of replicates. In all cases, the extraction time was 5 min, the flush volume was 60%, and the extraction temperature was 100 °C.

was evaporated further into 40 µL of n-tetradecane, which was added as a keeper. (b) Instrumental Analysis. The instrumental analysis was carried out using two different gas chromatography-mass spectrometry (GC-HRMS) systems. The first was used during the method development and optimization phases (trials 1-3, Table 1). It consisted of a gas chromatograph (Hewlett-Packard 5890, Agilent Technologies, Palo Alto, CA) equipped with a 60-m DB 5 fused-silica column (i.d. 0.32 mm; 0.25-µm film; J&W Scientific, Folsom, CA) connected to a VG 70-250S (formerly VG Analytical) double-focusing mass spectrometer operating at a resolution of 8000. Sample aliquots were injected splitless (2 µL, split opened after 2 min) at an injector temperature of 280 °C. The oven temperature was kept at 190 (I-PCBs and mono-ortho PCBs) or 200 °C (all other target analytes) for 2 min after injection and then raised by 3 °C/min to 300 °C, where it was held for 3 min. Helium was used as carrier gas at a head pressure of 18 psi. The MS was operated in electron ionization selected ion monitoring mode, at an electron energy of 35 eV, and with both the transfer line and ion source temperature set at 250 °C. For each target compound, two of the most intense ions of the molecular ion isotope distribution cluster were monitored. The second GC-HRMS system was used during the method validation phase (trials 4 and 5). It consisted of a Hewlett-Packard 6890 equipped with a 60-m DB 5 fused-silica column (i.d. 0.25 mm; 0.25-µm film). For the mono-ortho PCB fraction, an additional column was used in order to resolve PCBs 128 and 167 (DB 5MS; 60 m; i.d. 0.32 mm; 0.25-µm film). The GC was connected to a Waters Autospec double-focusing mass spectrometer (Milford, MA) operating at a resolution of 10 000. The oven temperature was kept at 160 (I-PCBs and mono-ortho PCBs) or 200 °C (all other target analytes) for 2 min after injection and then raised by 3 °C/min to 300 °C, which was held for 3 min. All other instrumental conditions were as above, except that the helium carrier gas was set to a constant flow of 1.2 mL/min using the electronic pressure control. (c) Quantification and Toxic Equivalency (TEQ) Calculations. Authentic calibrants were used for all compounds except PCB 81, which was quantified using the relative response factor of PCB 77. The quantification was done by the isotope dilution method.

Figure 2. Distribution of 13C-labeled indicator PCBs (I), mono-ortho PCBs (mo), non-ortho PCBs (77, 126, 169), and tetra-octa-CDD/Fs (Te-O) that were added to fish oil 1 and detected in the three PLE fractions from trial 1. Fractions 1, 2, and 3 were eluted with n-heptane, n-heptane/dichloromethane, 1/1 (v/v), and toluene, respectively. The carbon trap contained 0.5 g of 7.9% AX-21 on Celite.

For each sample, a dioxin TEQ value was calculated using the TEQ concept, which was developed for dioxin risk assessment and is frequently used in food and feed risk assessments. The basic assumption underlying the concept is that the contributions to the total toxicity made by all toxic congeners are additive. The 2,3,7,8-TCDD is the most persistent and toxic PCDD/F congener and has been assigned a toxicity equivalency factor (TEF) of 1. All other 2,3,7,8-chlorinated PCDD/Fs and WHO-PCBs have been assigned TEFs relative to that of TCDD.16 The TEQ for a group of compounds is the sum of the products of the individual concentrations and TEFs. Throughout this report, lower-bound TEQs are presented; i.e., only values that were higher than the limits of detection (LOD) for the respective compounds were included in their calculations. RESULTS AND DISCUSSION The shape-selective PLE procedure developed in this study was based on a previously published method for open-column (16) van den Berg, M.; Birnbaum, L.; Bosveld, A. T. C.; Brunstro ¨m, B.; Cook, P.; Feeley, M.; Giesy, J. P.; Hanberg, A.; Hasegawa, R.; Kennedy, S. W.; Kubiak, T.; Larsen, J. C.; van Leeuwen, F. X. R.; Liem, A. K. D.; Nolt, C.; Peterson, R. E.; Poellinger, L.; Safe, S.; Schrenk, D.; Tillitt, D.; Tysklind, M.; Younes, M.; Waern, F.; Zacharewskr, T. Environ. Health Perspect. 1998, 106, 775-792.

Analytical Chemistry, Vol. 79, No. 7, April 1, 2007

2947

Figure 3. Distribution of 13C-labeled indicator PCBs (I), mono-ortho PCBs (mo), non-ortho PCBs (77, 126, 169), tetra-CDD/F (Te), and pentaocta-CDD/Fs (Pe-O), added to fish oil 2 and detected in the three PLE fractions, which were eluted with n-heptane, a bisolvent mixture, and toluene, respectively. The trial number, carbon trap, and solvent mixture compositions are given above each panel. Ac, acetone; DCM, dichloromethane.

fractionation of PCBs and PCDD/Fs,10,13 which is usually used to separate PCBs and PCDD/Fs in lipid-free extracts into three fractions containing the bulk of the PCBs (bulk-PCBs), monoortho PCBs, and non-ortho PCBs and PCDD/Fs, respectively. When lipids were present, and under the more aggressive elution conditions used in the PLE (Table 1, trial 1), all PCB components did however, as expected, elute earlier (Figure 2), and the monoortho and a large proportion of the non-ortho PCBs were found 2948 Analytical Chemistry, Vol. 79, No. 7, April 1, 2007

in fraction 1 (n-heptane). The residual non-ortho PCBs were found in fraction 2 (n-heptane/DCM, 1/1, v/v) and the PCDD/Fs in fraction 3 (toluene, back-flush). The bulk of the fat was recovered in the first fraction (>98.5%), and only 0.3% was found in the last (PCDD/F) fraction. This method was considered very suitable for rapidly screening biological samples. After shape-selective PLE, only a miniaturized multilayer silica column was required to remove residual coextractants (primarily lipids). This combination

Figure 4. Schematic diagram of the 66-mL cell and the insert (in a 34-mL cell) used in the study. The larger cell contains a longer, wider carbon column and thus allows larger samples to be extracted.

was therefore tested for PCDD/F-TEQ analysis of food (fish oil) and feed (fish meal) samples, with promising results. The TEQ levels, as determined using a chemically activated fluorescence gene expression (CAFLUX) bioassay, compared well with those determined by a traditional GC-HRMS method.17 In principle, the combination of shape-selective PLE and CAFLUX detection may also be used to determine PCB-TEQs, although this remains to be confirmed. If and when that has been demonstrated, the procedure could be used for large-scale screening of food and feed items for compliance with the EU maximum residue limits for both PCDD/F-TEQ and total-TEQ (sum of the WHO-TEQ and PCDD/F-TEQ).18 In that case, fractions 1 and 2 would be combined into a “PCB fraction” and the lipid content would be determined gravimetrically. Lipids and other coextractants would then be removed from the PCB fraction and PCDD/F fraction (fraction 3), respectively, using multilayer/miniaturized multilayer silica columns, and the PCB-TEQ and PCDD/F-TEQ would be determined using the bioassay. The purpose of the following rounds of experiments was to shift the non-ortho PCB away from the other PCBs, preferably to the last fraction, thereby allowing interference-free GC-HRMS

determination of the non-ortho PCBs. In most cases, it would then be possible to simultaneously determine the levels of all non-ortho PCB and 2,3,7,8-PCDD/F congeners since they are normally present at comparable levels in biological matrixes. The monoortho PCBs are present at much higher levels and could therefore be determined in the bulk-PCB fraction, provided that a suitable GC column is used. In round 2 (trials 2:1-2:4) the carbon content was increased from 7.9 to 25 or 50% and the strength of the extraction solvents was also varied (see Table 1). The results indicated that increasing the carbon content or decreasing the elution strength of the solvent mixture used for fraction 2 increases the retention of nonortho PCBs (Figure 3). However, the results were not conclusive. Stronger retention was observed with an n-heptane/DCM ratio of 9/1 (trial 2:1) than with the same solvents at a ratio of 1/1 (trial 1), i.e. with a solvent mixture of lower strength. Similarly, the retention increased when the carbon content was increased from 7.9 (trial 1) to 25% (trial 2:3). However, the proportions of the target analytes retained with an n-heptane/DCM ratio of 9/1 and a carbon content of 25% (trial 2:2) were very similar to those retained in trial 1. In the last experiment with the inserts (trial 2:4), the cell started to leak, indicating that the back pressure created with a 50:50 carbon to Celite mixture in the inserts we used was too high. It was concluded that the cells were insufficiently robust for routine work. New, solid cells were therefore developed, rather than inserts, with threading and dimensions identical to that of the original Dionex 300 66-mL cells. This simplified the packing and cleaning procedures and minimized the risk of leakage. The designs of the 34- and 66-mL cells used in the study are compared in Figure 4. As shown, the larger cells have a longer, wider section for the carbon trap, and a larger sample reservoir, which is of paramount importance when analyzing ultralow-level samples. The amount of carbon/Celite mixture was increased from 0.5 to 1.5 g, and the repeatability of extractions using this configuration was tested (trial 3-1). The results clearly showed that the repeatability was much better with the larger cells than with the inserts. The bulkPCBs and >99% of the mono-ortho PCBs were extracted with the n-heptane, while the non-ortho PCBs were distributed between all the fractions, but notably, most of PCBs 126 and 169 were found

Figure 5. Comparison of results obtained in trials 3.1 (left) and 4.3 (right). The graphs show the distribution of 13C-labeled indicator PCBs (I), mono-ortho PCBs (mo), non-ortho PCBs (77, 126, 169), tetra-CDD/F (Te), and penta-octa-CDD/Fs (Pe-O), added to fish oil 2 and detected in the three PLE fractions, which were eluted with n-heptane, n-heptane/dichloromethane, 9/1 (v/v), and toluene, respectively. The carbon trap was filled with 1.5 g of a 25:75 mixture (w/w) of carbon and Celite.

Analytical Chemistry, Vol. 79, No. 7, April 1, 2007

2949

Figure 6. Comparisons of indicator PCB (top panel), WHO-PCB (middle panel), and tetra-octa-CDD/F (Te-O; lower panel) levels (pg/g of fresh weight) obtained using the ASE-carbon and the traditional multistep procedures. The levels of HpF2 were below the LOD for both methods, and the levels of HpF1 and HpD were below the LOD for the PLE method.

in the toluene fraction (Figures 3 and 5, trial 3:1). Even better separation between non-ortho PCBs and the other PCBs was obtained when using n-heptane/acetone (1:2.5) for the second fraction (Figure 3, trial 3-2). It was concluded that this solvent mixture was superior to the previously used 1:1 (v/v) n-heptane/ DCM mixture since it provided better performance and a better work environment, at lower cost. (17) Nording, M.; Sporring, S.; Wiberg, K.; Bjo¨rklund, E.; Haglund, P. Anal. Bioanal. Chem. 2005, 381, 1472-1475. (18) The commission of the European communities; commission regulation 199/ 2006; 3 February 2006.

2950

Analytical Chemistry, Vol. 79, No. 7, April 1, 2007

In trial 4:3, an experiment equivalent to trial 3:1 was performed, to assess the reproducibility of the system. The results were very similar to those of trial 3:1 (Figure 5). Thus, the system seems to provide sufficiently repeatable and reproducible extractions for accurate isotope-dilution GC-HRMS determinations of WHOPCBs and PCDD/Fs. In addition, an attempt was made to shift the non-ortho PCBs even further into the toluene fraction. The large cells were filled with 2.4 g of carbon/Celite mixture, very (19) Nording, M.; Nichkova, M.; Spinnel, E.; Persson, Y.; Gee, S. J.; Hammock, B. D.; Haglund, P. Anal. Bioanal. Chem. 2006, 385, 357-366.

close to the maximum amount they can hold, and fraction 2 was eluted with n-heptane/acetone in the v/v proportions 1:2.5 (trial 4:1) or 2.5:1 (trial 4-2). Both experiments resulted in almost complete separation of non-ortho PCBs from the other PCBs. There was virtually no difference in the distributions of the target compound among the fractions obtained using these two solvent mixtures (Figure 3). However, the more polar mixture, n-heptane/ acetone (1:2.5), was preferred since it extracts lipids more efficiently. Thus, the experimental conditions of trial 4:1 were used in the final method, i.e., two cycles of n-heptane, one cycle of n-heptane/acetone (1:2.5), and four cycles of toluene. The performance of this method was evaluated in triplicate analyses (trial 5) of an in-house salmon tissue reference material. The results compared well with those obtained using the traditional method20,21 (Figure 6). There were no significant differences in the indicator-PCB or mono-ortho PCB levels determined by the two methods, as manifested in overlapping 95% confidence intervals. The same was true for all non-ortho PCBs and PCDD/ Fs, except CB126 and 2,3,7,8-TCDD, for which slightly higher levels (pg/g of fresh weight) were obtained using the traditional method (390 ( 4 vs 430 ( 26 for CB126 and 2.4 ( 0.01 vs 2.7 ( 0.1 for TCDD). However, these differences were primarily due to the exceptionally low variability in the selective PLE data. It was therefore concluded that the two methods produced data of very similar quality. The performance of the method was also compared with the European community (EC) requirements that analytical methods must meet to be used as confirmatory methods in the control of PCDD/F-TEQ and WHO-PCB-TEQ levels in food.22 According to these requirements, the trueness has to be within (20% and the precision, measured as the coefficient of variation (CV), has to be less than 15%. The trueness of the WHO-PCB-TEQ, PCDD/ F-TEQ, and total-TEQ values obtained from the salmon tissue analyses using the selective PLE method were -8, -5, and -7%, respectively, and the corresponding CVs were 1.5, 0.5, and 1.3%. The recoveries of most 13C-labeled internal standards were also acceptable, i.e., within the range 60-120%, although they were (20) Wiberg, K.; Bergman, A.; Olsson, M.; Roos, A.; Blomkvist, G.; Haglund, P. Environ. Toxicol. Chem. 2002, 21, 2542-2551. (21) Danielsson, C.; Wiberg, K.; Koryta´r, P.; Bergek, S.; Brinkman, U. A. Th.; Haglund, P. J. Chromatogr., A 2005, 1086, 61-70. (22) The commission of the European communities, commission directive 2002/ 69/EC, 26 July 2002.

lower for highly chlorinated congeners. The average recovery of the octa-CDD/Fs was only ∼40%, most likely due to strong sorption to the carbon trap. However, their overall contribution to the TEQ was much lower than 10%, so the data still comply with the requirements. If necessary, the recoveries of the highly chlorinated PCDD/Fs could be improved by adding more toluene extraction cycles. Thus, the new shape-selective PLE method seems to comply with the EC requirements. However, before the method could be used in routine work, further tests with additional samples would be needed, to assess its robustness and determine the rate of false negatives it yields, which has to be below 1% according to the EC directive. CONCLUSION The shape-selective PLE method developed in this study has proven its value for extracting fish and fish oil samples, and it seems reasonable to believe that all types of food and feed samples can be extracted using this method. Previously, sepiolitic clay has proven difficult to extract by a selective PLE method using n-heptane and in-cell cleanup with sulfuric acid-impregnated silica.10 However, the new shape-selective method, which includes use of toluene, is expected to efficiently extract planar organic compounds, such as WHO-PCBs and PCDD/Fs, from mineral surfaces. When a similar technique (using the ASE 100 and a smaller cell size) was applied to 10 different soil samples, it efficiently extracted PCDD/Fs from solid matrixes that differed widely in degree of pollution, soil type, and amounts of organic and soot carbon.19 Thus, even feed samples containing sepiolitic clay or other mineral filling agents are expected to be quantitatively extracted using the shape-selective PLE procedure. Work is currently in progress to verify this hypothesis. ACKNOWLEDGMENT This work was performed as parts of the DIAC and DIFFERENCE projects, both of which were funded by the European Commission (contracts G6RD-CT-2001-00572 and G6RD-CT-200100623, respectively).

Received for review December 28, 2006. Accepted February 5, 2007. AC0624501

Analytical Chemistry, Vol. 79, No. 7, April 1, 2007

2951