Anal. Chem. 1994,66, 4068-4073
Determination of Planar PCBs by Combining On-Line SFE-HPLC and GC-ECD or GC/MS Hege Rebecka Johansen,’*t Georg Becher,t and Tyge Greibrokk* Department of Environmental Medicine, National Institute of Public Health, 0462 Oslo, Norway, and Department of Chemistry, University of Oslo, P.O. Box 1033, Blindern, 0315 Oslo, Norway
A method for determining non-ortho- and mono-orthosubstituted polychlorinated biphenyls (PCBs) in biological samples has been developed. High selectivity was obtained by on-line coupling of supercritical fluid extraction (SFE) and high-performance liquid chromatography (HPLC). Gas chromatography (GC) with electron capture detection or highresolution mass spectrometry was utilized for quantitative determinations. Group separation of different PCB congeners was achieved on a (2-( 1-pyrenyl)ethyl)dimethylsilylated silica column. Compared with off-line HPLC, the on-line coupling to SFE resulted only in a minor reduction in column efficiency. The average recoveriesof the planar non-ortho-substituted PCB congeners 77, 126, and 169 from crab hepatopancreas were 71-101%, with the highest recovery for PCB-169. For human blood serum and milk, the recoveries of the three congeners ranged from 35 to 57% (serum) and from 76 to 87% (milk), with the lowest recovery for PCB-169. The relative standard deviations of the complete analyses were determined to be 516%. The recoveries of the on-line SFE-HPLC technique were compared to those of conventional solvent extraction and off-line HPLC. For crab hepatopancreas, the two methods gave approximately the same result, but for blood serum, slightly higher recovery of the native non-ortho-substituted PCBs was obtained using SFE-HPLC. The present method demonstrates the high speed and selectivity which were obtained by on-line SFE-HPLC as a sample preparation technique prior to GC analyses for the determination of a group of highly toxic PCBs, usually found in very low concentrations compared to the bulk of the PCB congeners. Polychlorinated biphenyls (PCBs) with chlorine in the 3 , 4, and 5 positions (non-ortho-substituted) have the ability to assume coplanar configurations, similar to some of the highly toxic polychlorinated dibenzo-p-dioxins (PCDD) and dibenzofurans (PCDFs). With increasing number of ortho chlorine substituents, the PCBs are less likely to assume a coplanar configuration due to the increasing steric hindrance in the 2 and 6 positions. The congeners 3,3’,4,4’-tetrachlorobiphenyl (PCB-77), 3,3’,4,4’,5-pentachlorobiphenyl(PCB126),and 3,3’,4,4’,5,5’-hexachlorobiphenyl(PCB- 169), which are substituted in both para and at least two meta positions, are approximate isostereomers of 2,3,7,8-tetrachlorodibenzop-dioxin (TCDD), the most toxic dioxin congener, and induce effects similar to those caused by PCDDs and PCDFs.’ PCBs which have the ability to assume configurations with both
rings in the same plane have previously been termed coplanar PCBs,’ a term which is used in this paper, too, although knowledge of the actual planarity of different non-ortho congeners in solution and in the solid phase is scarce. Due to the low abundance of non-ortho-substituted PCBs, usually 3-5 orders of magnitude lower than the sum of PCBs, a separation from the other PCBs is required in order to determine this group of compounds. Solvent extraction, followed by purification of the extracted components and analyses by capillary gas chromatography coupled with electron capture detection or high-resolution mass spectrometry (GC-ECD or GC/MS), is the most common method for determination of PCBs in different matrices. Different analytical techniques for isolation and determination of non-ortho-substituted PCBs are described in the literature,2 most of which involve activated charcoal columns. By increasing the selectivity of each separation step, interferences from the sample will result in fewer problems in the complete analytical procedure. Supercritical fluid extraction (SFE) has been shown to provide a powerful alternative to traditional solvent extraction methods, particularly in extracting organochlorines from matrices containing high levelsof fate3-10Thus, SFE was selected for extraction due to high selectivity and also due to the possibility of transferring the entire extract on-line, prior to group separation of different groups of PCBs. The group separation of PCBs was obtained on a highly selective (2-(1-pyreny1)ethyl)dimethylsilylated silica column, first described by Haglund et al.,” to separate the non-ortho-chlorinated PCBs from the the mono-ortho- and multi-ortho-substituted congeners. On-line coupling of S F E to GCI2-l5 and to SFC16-19has been described, and recently some papers have also been (2) Creaser, C. S.;Krokos, F.; Startin, J. R. Chemosphere 1992.25, 1981-2008. (3) Nam, K. S.; Kapila, S.; Yanders, A. F.; Puri, R. K. Chemosphere 1990, 20, 873-880. (4) King, J. W. J . Chromatogr. Sci. 1989, 27, 355-363. ( 5 ) Murugaverl, B.; Voorhees, K. J. J . Microcolumn Sep. 1991, 3, 11-16. (6) Hopper, M. L.; King, J. W. J . Assoc. Off. Anal. Chem. 1991, 74, 661-666. (7) David, F.; Verschuere, M.; Sandra, P. Fresenius’ 2. Anal. Chem. 1992,344, 479-485. (8) France, J. E.; King, J. W.; Snyder, J. M. J . Agric. Food Chem. 1991, 42, 1871-1874. (9) Johansen, H. R.; Becher ,G.; Greibrokk, T. Fresenius’ 2. Anal. Chem. 1992, 344, 486490. (IO) Johansen, H. R.; Thorstensen, C.; Becher, G.; Greibrokk, T. J . High Resoluf. Chromatogr. 1993, 16, 148-152. (1 1) Haglund, P.; Asplund, L.; Jirnberg, U.; Jansson, B. Chemosphere 1990, 20, 887-894. (12) Raymer, J. H.; Velez, G. R. J . Chromatogr. Sci. 1991, 29, 468-415. (13) King, J. W.; France, J. E.;Snyder, J. M. Fresenius’Z. Anal. Chem. 1992,344.
474478.
+
National Institute of Public Health, Norway.
* University of Oslo.
(1) Safe, S. CRC Crit. Reu. Toxicol. Chem. 1990, 21, 51-88.
4060
Analytical Chemistry, Vol. 66, No. 22, November 15, 1994
(14) Nielen, M. W. F.; Sanderson, J. T.; Frei, R. W.; Brinkman, U. A. Th. J . Chromatogr. 1989, 474, 388-395. (15) Hawthorne, S . B.; Miller, D. J. J . Chromatogr. 1987, 403, 63-76.
0003-2700/94/0366-4068$04.50/0
0 1994 Amerlcan Chemical Society
concerned with the coupling of SFE to HPLC.2@24Only one of them dealt with P C B S ,and ~ ~ real samples were not included. Cleanup procedures of complex samples for PCB analyses are in most cases both time-consuming and associated with low recoveries. The aim of this report was to develop a simple and rapid on-line SFE-HPLC method for the quantitative extraction and cleanup of mono-ortho- and non-orthosubstituted PCBs prior to analyses by GC-ECD or GC/MS. The SFE-HPLC method was evaluated by using both standard solutions and biological samples with different fat content: blood serum, human milk, and crab hepatopancreas.
EXPERIMENTAL SECTION Reagents and Chemicals. Supercritical fluid extractions were performed with carbon dioxide, SFC grade (Scott Speciality Gases, Plumsteadville, PA). All solvents were HPLC grade from Rathburn (Walkerburn, U.K.). The 13Clabeled PCB-77, PCB-126, and PCB 169 were from Cambridge Isotope Laboratories (Woburn, MA). The normal PCB standards were purchased from Cambridge Isotope Laboratories or from Amchro (Sulzbach, Germany). SFE System. The SFE system was composed of a SFC300 pump and an extraction unit SFE-30 (Carlo Erba, Milan, Italy). The extraction was performed in a thermostated chamber which was provided with a separately heated outlet to the on-line system, and the outlet of the restrictor was kept at 280 "C. Extraction cells used in this study included 0.8 and 1.7 cm3 (50 mm X 4.6 mm i.d. and 100 mm X 4.6 mm i.d., respectively) cells from Keystone Scientific (Bellefonte, PA). In addition to the filter in the extractor cell, a 0.5 pm filter was placed between the extraction cell and the restrictor to prevent plugging of the restrictor. Basic alumina from Merck, 70-230 mesh, was used as an adsorbent for the lipids. The alumina was activated at 190 "C overnight and stored in a screw-capped bottle. The alumina was placed at the outlet end of the extraction cell and in a separate cell after the extraction. HPLC System. The separation of the mono-ortho and non-ortho-substituted PCBs was performed with a Waters 6000A pump, a Valco 7060 switching valve, and two 150 mm X 4.6 mm Cosmosil 5-PYE columns with (2-(l-pyrenyl)ethy1)dimethylsilylated silica gel, 5 pm particles (Nacalai Tesque Inc., Kyoto, Japan), coupled in series. Hexane was used as a mobile phase at a flow rate of 0.7 mL/min. On-Line Coupling. The on-line system was accomplished by connecting a 1/16 T coupling (Valco ZT1 C, drilled up to 0.5 mm i.d.) to the linear restrictor, Figure 1. The restrictor, King, J. W. J. Chromatogr. Sci. 1990, 28, 9-14. Xie, Q. L.; Markides, K. E.; Lee, M. J. Chromatogr. Sri. 1989,27, 365-370. Levy, J. M.: Cavalier, R. A.; Bosch, T. N.; Rynaski, A. F.; Huhak, W. E. J. Chromatogr. Sci. 1989, 27, 341-346. . . Berg, B. E.; Hansen, E. M.; Gjerven, S.;Greibrokk, T. J. High Resolut. Chromatogr. 1993, 16, 358-363. (20) Unger, K. K.; Roumeliotis, P. J. Chromatogr. 1983, 282, 519-526. (21) Cortes, H. J.; Green, L. S.: Campbell, R. M. Anal. Chem. 1991,63, 27192724. (22) Liu, M. H.; Kapila, S.;Nam, K. S.; Elseewi, A. A. J. Chromatogr. 1993,639, 151-1 5 7 . (23) Stalling, D. L.; Saim, S.;Kuo,K. C.; Stunkel, J. J. J. Chromatogr. Sci. 1992, 30,
486-490.
(24) Bewadt, S.; Pelusio, F.; Montanarella, L.; Larsen, B.; Mapelli, G. Proceedings of the European Symposium on Analytical SFC/SFE; Riva del Garda, Italy; 1993; pp 114-120.
18
3 Figure 1. Schematic diagram of the on-line SFE-HPLC system. 1, Con gas reservoir; 2, SFE pump: 3, extractor oven; 4, extractor cell; 5, alumina column; 6, switching valve, V1; 7, restrictor;8, restrictorheater: 9, Vaico T; 10, switching valve, V2; 11, HPLC pump; 12, restrictor tip; 13, steel tubing;l4, union; 15, Switching valve, V3; 16, 200 pm i.d. fused silica capillary: 17, PYE columns; 18, UV detector.
20 cm x 28 pm i.d.,which was made of a linear fused silica capillary (Polymicro Technologies, Phoenix, AZ) gave a flow of 133 mL/min of gaseous C02. The restrictor was pushed through the T and connected in a union (Valco ZU.l, Vici AG, Schenken, Switzerland) with a 0.75 mm opening. The cooling effect due to the decompression of the fluid helps the analytes to concentrate in a narrow band in the decompression zone, Le., inside a steel tubing with an i.d. of 0.5 mm and a length of 20 cm (see 13, Figure 1). Restrictor plugging due to the decrease in solubility was avoided as the temperature was held at 280 "C. The low i.d. of the restrictor prevented any significant amount of sample to be lost due to backdiffusion to the extraction cell. The extraction cell was placed in an oven, thermostated to 60 "C during extraction. A Valco six-port valve V2 (N6W FG 3 10, see 10, Figure l), was placed between the T piece and the HPLC mobile phase reservoir to avoid back-diffusion of COz to the mobile phase during extraction. Dissolved C02 in the hexane phase resulted in large peaks in front of the chromatogram. After the extraction was completed, valve V2 was switched to the flow mode to introduce the HPLC eluent. In addition, a switching valve, V3, Rheodyne Model 7000 (see 15, Figure l), was placed at the columns inlet to prevent C02 from entering the columns during extraction. During extraction, the decompressed COz gas was vented to the atmosphere through a 200 pm i.d. fused silica capillary (see 16, Figure 1). Before the extraction started, the valves V2 and V3 were placed in the locked position, Le., the C02 could not enter the mobile phase reservoir or the columns. All the extractions were performed in the dynamic mode. The extraction time was 40 min, and the pressure was 14.5 MPa. Optimal extraction time and density for extracting PCBs from similar samples have been tested in earlier st~dies.~JOWhen the extraction was completed, valve V2 was opened and the LC eluent was introduced. After 10 s, valve V3 was opened and the chromatography was started. GC System. G C analyses were performed with a PerkinElmer 8700 gas chromatograph equipped with an electron capture detector and an AS-8300 autosampler. Hydrogen was used as carrier gas with a linear velocity of 28 cm/s (at 100 "C). Argon/methane (5%) was used as makeupgas with Analytical Chemistry, Vol. 66, No. 22, November 15, 1994
4069
a flow rate of 50 mL/min. Injector temperature was 270 "C, and the detector was operated at 330 "C. A DB-5 capillary column, 50 m X 0.25 mm i.d., 0.25 pm film thickness (J&W Scientific, Folsom, CA), was temperature programmed from 60 "C (1 min hold) to 200 "C at 20 deg/min and then to 280 "C (10 min hold) at 2.0 deg/min. GC/HRMS. The GC/HRMS instrument consisted of a VG AutoSpec high-resolution mass spectrometer with Opus Quan software program and a Hewlett-Packard 5890 gas chromatograph with a DB-5 capillary column, 30 m x 0.25 mm i.d., 0.25 pm film thickness (J&W Scientific). The initial column temperature was 130 "C (2 min hold), and then the column was temperature programmed to 200 "C at 20 deg/ min, then to 212 "C at 0.6 deg/min, and then to 300 "C at 10 deg/min. Helium was used as carrier gas with a linear velocity of 25 cm/s (at 200 "C) . Injections were performed in the splitless mode using a H P 7673 A autosampler. The injector temperature was 280 "C. Selected ion detection was carried out in the electron impact mode with a MS ion source temperature of 245 "C, an electron energy of 53 eV, and a resolution of 6000. Dwell time for each ion was 80 ms. Perfluorokerosene (PFK) was used to provide suitable lock masses. The transfer line was held at 280 "C. Prior to analyses, the sample volume was reduced to about 20 pL, and n-nonane was added to 30 pL. To the final extract was added 250 pg of PCB-189 to measure the recovery of the 13C-labeled analogs. Preparation of Blood Samples. Serum was stored at -20 "C until sample preparation. A 10 mL sample of blood gave approximately 5 mL of serum. C18silica gel, 0.5 g with porous 40 pm particles, (Analytichem Bonded, Harbour, CA), was filled in a 1.7 cm3 extraction cell from Keystone Scientific. The packing was rinsed with n-hexane to remove any lipophilic contamination, dried with air flow through the sorbent for 5 min, and conditioned with methanol (5 mL) and then water (5 mL). Internal standards, PCB-159 for the mono-orthosubstituted PCBs and 13Clabeled analogs for the non-orthosubstituted PCBs, were added to the serum and mixed with a whirlmixer before the serum was passed through the sorbent by vacuum. Basic alumina, about 0.5 g, was filled in the outlet end of the extraction cell mainly to adsorb any residual water. In addition, 0.5 g of basic alumina was placed in a separate cell coupled to the extraction cell to serve as a selective adsorbent for the lipids. Preparation of Human Milk. The human milk was prepared in the same way as the serum, but the proteins had to be precipitated with acetonitrile (1: 1 v/v) before the milk was added to the cl8 silica gel in order to avoid plugging of the pores. Preparation of Crab Samples. Internal standards, PCB159, and I3Clabeled analogs of the non-ortho-substituted PCBs were added to crab hepatopancreas before it was mixed and homogenized with anhydrous sodium sulfate (1:4 w/w) in a mortar. The dry, saltlike mixtures, about 1 g, were placed in the extraction chamber. Basic alumina was placed in the outlet end of the extraction cell and in a separate cell after the extraction chamber (1:l w/w). Glass wool was placed in each end of the extraction chamber in order to reduce any empty space in the cell. 4070
Analytical Chemistry, Vol. 66, No. 22, November 15, 1994
Table 1. Number of Theoretical Plates, N, In Both On-Line and OH-Line Modes N
PCB IUPAC no. 105 118 156 77 126 169
chlorination
on-line mode
Mono-Ortho-Substituted PCBs 2,3,3',4,4' 14 200 2,3',4,4',5 9 000 2,3,3',4,4',5 10 500 Non-Ortho-Substituted PCBs 3,3',4,4' 15 100 3,3',4,4',5 16 100 3,3',4,4',5,5' 16 300
off-line mode 18 000 13 900 16 500 18 200 18 100 19 100
Solvent Extraction. (i) Blood Serum. A method described by Papke et al.25for extraction of dioxins in blood samples was modified for the extraction of PCBs. Chem Elut CE 1010 columns from Analytichem were topped with 4 g of sodium chloride and 5 g of Hydromatrix (plankton marine diatomite from Analytichem). Thereafter, the adsorbents were washed with heptane/ 2-propanol (3:2) and methylene chloride and dried at 50 "C overnight. Internal standards were added to the blood serum before it was diluted with water and ethanol (1: 1:O. 12) and applied on the column. The extraction was performed by eluting with 80 mL of heptane/2-propanol (3:2). The eluate was reconcentrated with purified nitrogen, dissolved in 3 mL of cyclohexane, and treated twice with an equal volume of sulfuric acid (Scan Pure, 98.3%, from Chem Scan A/S, Elverum, Norway) to remove the major part of the lipids and other interfering organic compounds. The sulfuric acid phase was re-extracted with cyclohexane. The organic phases were combined and reduced to about 150 pL before fractionation on the PYE columns. (ii) Human Milk and Crab Hepatopancreas. The extractions of both crab hepatopancreas and human milk were performed according to a method described by Norheim et a1.26 The method included extraction with cyclohexane, acetone, and water after ultrasonic disintegration. The lipids were removed by treatment with sulfuric acid and gel permeation chromatography before the PCBs were fractionated on the PYE columns.
RESULTS AND DISCUSSION Band Focusing and Trapping Efficiency. To test the efficiency of the trapping system, six PCB congeners were spiked on a glass fiber filter, and the quantitative results after SFE-HPLC were compared with those of the same solution injected directly into the HPLC system. Compared with direct injection of the components, 7-12% lower peak areas were obtained with the on-line method. The chromatographic peak shapes after SFE-HPLC compared favorably with those obtained after direct injection of the components into the HPLC. Compared to regular HPLC, the on-line coupling of the SFE part to the HPLC system resulted in 12-36% loss in the number of theoretical plates, depending on the identity of the congeners (Table 1). The largest reduction was obtained for the mono-orthosubstituted PCBs, which are eluted closer to any traces of M.;Lis, Z. A.; Scheunert, K. Chemosphere 1989, 19,941948. ( 2 6 ) Norheim, G.;Skaare, J . U.;Wiig, 0. Enuiron. P o l h i . 1992, 77, 51-57. ( 2 5 ) Papke, 0.; Ball,
Table 2. Recoveries of 40 ng Samples of PCBs Extracted from a Glass Fiber Filter with SFE and Fractionated On-Llne with HPLC'
IUPAC no. recovery,% SD. %b
Table 3. Recoveries of Spiked Non-Ortho-Substituted PCBs Extracted wlth SFE and Fractionated On-Line with HPLC'
crab hepatopancreasb
105
118
156
159
77
126
169
PCB congener
67 5
68 9
69 13
68 7
79 3
88 4
99 7
77 126 169
blood samplesC
recovery
SD
recovery
(%)
(%)
(%)
(a)
71 89 101
10 5 7
57 49 35
10 16 15
SD
human milkd recovery
SD
(a)
(%)
82 87 76
11 14 13
Analyzed on GC-ECD or GC-MS. The experimental conditions are given in the text. Five replicates. Seven replicates. Five replicates.
geners are less retained than the tetrachlorinated (77), pentachlorinated ( 126), and hexachlorinated (169) non-orthosubstituted congeners. The multi-ortho-substituted PCB 1105 congeners are eluted in the front of the chromatogram, but as they usually are present in a much higher level, they can be quantitated directly on GC-ECD. The peak shape may be improved by using additional external cooling to focus the analytes, but as the obtained result was satisfactory for this separation, no such attempts were done. The retention of the PCBs on the PYE columns can be reduced by the presence of lipids." The recovery of PCB- 1 18 and -77, which are the earliest eluting congeners in each fractions, can be reduced if they are partly eluted in the wrong fraction. The cutting points for the fractions were based on 0 8 16 MIN. the retention times of the standards. Samples with large differences in the fat content, from about 0.5% in blood serum Figure 2. On-line SFE-HPLC chromatogram of six PCB congeners: the monoortho-substituted 105 (2,3,3',4,4'), 118 (2,3',4,4',5), and 156 to about 15% in crab hepatopancreas, were therefore examined (2,3,3',4,4',5) and the nonortho-substituted 77 (3,3',4,4'), 126 to validate the method. An automation of the collection steps (3,3'4,4',5), and 169 (3,3',4,4',5,5') congeners extracted from a glass will probably lead to an increase in the recovery of PCB-77, fiber filter. The extraction and chromatographic conditions are given in the text. as this congener elutes first in the coplanar fraction. Blood Samples. The recoveries of PCB-77 and -126 were about 60% and 50% respectively, which is in agreement with C02 and are thus expected to be more influenced than the results obtained in an earlier report using SFE-GC.'O non-ortho-substituted PCBs. However, the recovery for PCB-169 was low, about 35%, the An impactor interface has been e m p l ~ y e d to~ trap ~~~,~~ standard deviation was quite high (Table 3). Recoveries the analytes. When the collector (0.5 mm i.d. X 2 cm steel for PCB-77, -126, and -169 in blood plasma were reported to tubing) was packed with silica or Florid, the band broadening be about 70% by Asplund et al.,*'using the same PYEcolumns became significant. However, the cooling effect obtained due for separation of the non-ortho-substituted PCBs from the to the expansion of the supercritical fluid was sufficient to other PCBs after traditional extraction methods. focus the PCBs inside the 20 cm empty steel tubing, with a The differences in recoveries appear not to be related to loss of not more than 7-12%. the use of SFE-HPLC, since extraction from glass fibers gave UV detectors are not adequate for measuring recoveries good yields (Table 2) and since conventional extraction (of of low levels of PCBs. Thus, 40 ng of seven mono-ortho- and native PCBs) gave even lower yields (see Validation of the non-ortho-substituted congeners, respectively, were extracted Method, below). Furton and Lin28reported a lower recovery from a glass fiber filter with SFE, on-line fractionated with for the highly chlorinated PCB species from a c18 sorbent at HPLC and quantitated after GC-ECD analyses. The results 13.6 MPa and 60 OC. They observed, however, low recoveries are given in Table 2. Due to the additional steps, Le., collection, (10-3076) of all the tested PCB congeners. Lower recovery evaporation of the mobile phase, and injection on the GC, for the higher chlorinated PCBs correlated with increasing lower recoveries were obtained compared with the results retention of biphenyls with increasing chlorine content.29The obtained comparing off-line and on-line HPLC. However, reported recoveries of PCBs from cl8 sorbents seem to be the reproducibility is good and the use of IT-labeled analogs very dependent of the extraction conditions and on the as internal standards for the coplanar PCBs can adjust for differences in the cl8 material. Lohleit et a1.30andAlexandrou analyte losses. Figure 2 shows a chromatogram of six PCB congeners, (27) Asplund, L.; Svensson, B.-G.; Nilsson, A.; Eriksson, U.;Jansson, B.; Jensen, mono-ortho- and non-ortho-substituted, extracted from a glass S.; Widequist, U.;Skerfving, S. Ph.D. Thesis, University of Lund, 1993. fiber filter with SFE, trapped in the steel tubing, and separated (28) Furton, K. G., Lin, Q.J . Chromatogr. Sri. 1993, 31, 201-206. (29) Aitzetmflller, K. J . Chromatogr. 1975, 107, 411-415. on-line with HPLC. The pentachlorinated (105 and 118) (30) Lohleit, M.;Hillmann, R.; BBchmann, K. Fresenius' J . Anal. Chem. 1991, and the hexachlorinated (1 56) mono-ortho-substituted con339,470-474. I
Analytical Chemistry, Vol. 66, No. 22, November 15, 1994
4071
Table 4. Quantltatlon of Mono- and NonQrtho-Substituted PCBs In Crab Hepatopancreas and Human Blood Serum, SFE vs Solvent Extraction. crab heptaopancreas human blood serum
PCB congeners
S FE
solvent extraction
SFE
solvent extraction
105b 118b 156b 77c 126c 169c
6.1 ng/g 17.2 ng/g ndd 179.1 pg/g 50.1 pg/g 16.3 pg/g
5.2 ng/g 19.0 ng/g ndd 189.8 pg/g 54.7 pg/g 14.0 pg/g
0.3 ng/g 1.1 ng/g 0.7 ng/g 3.9 pg/g 7.4 pg/g 1.8 pg/g
0.2 ng/g 0.8 ng/g 0.5 ng/g 2.8 pg/g 5.7 pg/g 1.2 pg/g
The samples were quantitated using PCB-159 and 3C-labeled analogues of PCBs 77, 126, and 169 as internal standards. The results are given on wet weight basis. Analyzed on GC-ECD. Analyzed on GC-MS. Not determined.
lmll
on
"I "I
126
13..sj
I69
I E A -
f.
Figure 3. GC-ECD chromatogram of 0.3 g of crab hepatopancreas spiked with 6 ng of the nonsrtho-substituted PCBs after on -line SFEHPLC extraction and fractionation.The extraction and chromatographic conditions are given in the text. (A) Fraction 1, Le., the multisrthosubstituted PCBs, after separation on the PYE columns. The extract was dissloved In 300 HL, and 2 FL was injected into the GC. (8)Fraction 2, i a , the mono-ortho-substituted PCBs. PCB-156 was below the detection limit. (C) Fraction 3, Le., the non-ortho-substituted PCBs, in which 40 pg of each PCB congener was injected into the GC.
et have reported quantitative recoveries of PCBs from CIS sorbent at 25 MPa/70 OC and 40 MPa, respectively. (31) Alexandrou, N.; Lawrence, M. J.; Pawliszyn, J. Anal. G e m . 1992,64, 301311.
4072
Analytical Chemistry, Vol. 66, No. 22, November 15, 1994
Quantitative recoveries are also reported from use of ODS to trap the PCBs after off-line SFE.32 Human Milk. Good recoveries were obtained for all of the non-ortho-substituted PCBs from human milk (Table 3). A slightly lower recovery was obtained for PCB-169, but it was considerably higher than that obtained for the blood samples. Crab Hepatopancreas. Figure 3 shows GC-ECD chromatograms of crab hepatopancreas spiked with 6 ng of the three planar PCBs after SFE extraction and on-line fractionation on HPLC. Assuming that the spiked analytes behave like the native analytes during extraction, the recoveries from crab hepatopancreas were very good (see Table 3). The good recovery of PCB-77, compared to extraction of standards from glass fiber filters and the peak profile on the GC-ECD chromatogram, indicates that very low levels of fat were extracted together with the PCBs. The recoveries of spiked PCBs from crab hepatopancreas, human milk, and blood serum correlate with the results obtained in an earlier study using on-line SFE-GC.l0 The low recovery for blood serum is not easily explained; however, lower recoveries for blood serum compared with human milk correlate with reports obtained using conventional method^.^^,^^ Recoveries reported from blood serum vary usually from 59 to 103%,33while the recoveries from human milk most often vary between 90 and Validation of the Method. Very little has been reported concerning quantitative on-line SFE-HPLC. One of the problems has been irregular baseline noise and variable detector response due to gas bubbles introduced into the HPLC solvent. Detector instability was no problem during this study, after a valve (10, V2 Figure 1) was introduced between the T and the HPLC mobile phase reservoir. To validate the recoveries obtained after SFE-HPLC extraction and fractionation, a few samples were compared with the results obtained using conventional extraction and cleanup methods. Samples of crab hepatopancreas, blood serum, and human milk were quantitated using both solvent extraction and SFE. The results are presented in Table 4. Compared with solvent extraction, SFE gave a slightly higher level of native PCBs (32) Bswadt, S.;Johansson, B.; Rovid, C.; Pelusio, F.; Larsen, B. R. Proceedings of the 2nd European Symposium on Analytical SFC/SFE; Riva del Garda, Italy; 1993; pp 121-128. (33) Luotamo, M. Ph. D.Thesis, University of Helsinki, 1991. (34) SkAre, J , U. , NVH/VI, Norway, personal communication.
from blood serum, for crab hepatopancreas, the two methods gave approximately the same results. For samples with high levels of lipids, SFE was a faster method compared with conventionalextraction methods, which usually require both treatment with sulfuric acid and gel permeation chromatography (GPC) to remove the lipids. SFEHPLC and concentration before the extract could be analyzed on GC-ECD or GC/MS took about 90 min. The large amounts of solvents required with liquid extraction is also avoided by using SFE, but solvent extraction has the advantage that simultaneous parallel extractions can be performed. However, as the removal of lipids with GPC and the
fractionation on the PYE columns have to be done separately after solvent extraction, the SFE method is still a much faster alternative.
ACKNOWLEDGMENT Financial support from the Norwegian Research Council for Science and Humanities is greately appreciated. Mr. 0. J. Rossland is gratefully acknowledged for performing the mass spectrometric analyses. Received for review May 2, 1994. Accepted August 22, 1994.@ ci Abstract
published in Advance ACS Abstracts. October 1, 1904.
AnalyticalChemistry, Vol. 66, No. 22, November 15, 1994
4073
