Environ. Sci. Technol. 2004, 38, 3698-3704
Automated Solid-Phase Extraction and Measurement of Perfluorinated Organic Acids and Amides in Human Serum and Milk ZSUZSANNA KUKLENYIK, JOHN ADAM REICH, JASON S. TULLY, LARRY L. NEEDHAM, AND ANTONIA M. CALAFAT* Division of Laboratory Sciences, National Center for Environmental Health, Centers for Disease Control and Prevention, Atlanta, Georgia 30341
Organic fluorochemicals are used in multiple commercial applications including surfactants, lubricants, paints, polishes, food packaging, and fire-retarding foams. Recent scientific findings suggest that several perfluorochemicals (PFCs), a group of organic fluorochemicals, are ubiquitous contaminants in humans and animals worldwide. Furthermore, concern has increased about the toxicity of these compounds. Therefore, monitoring human exposure to PFCs is important.We have developed a high-throughput method for measuring trace levels of 13 PFCs (2 perfluorosulfonates, 8 perfluorocarboxylates, and 3 perfluorosulfonamides) in serum and milk using an automated solidphase extraction (SPE) cleanup followed by high-performance liquid chromatography-tandem mass spectrometry. The method is sensitive, with limits of detection between 0.1 and 1 ng in 1 mL of serum or milk, is not labor intensive, involves minimal manual sample preparation, and uses a commercially available automated SPE system. Our method is suitable for large epidemiologic studies to assess exposure to PFCs. We measured the serum levels of these 13 PFCs in 20 adults nonoccupationally exposed to these compounds. Nine of the PFCs were detected in at least 75% of the subjects. Perfluorooctanesulfonate (PFOS), perfluorohexanesulfonate (PFHxS), 2-(N-methylperfluorooctanesulfonamido)acetate (Me-PFOSA-AcOH), perfluorooctanoate (PFOA), and perfluorononanoate (PFNA) were found in all of the samples. The concentration order and measured levels of PFOS, PFOA, Me-PFOSA-AcOH, and PFHxS compared well with human serum levels previously reported. Although no human data are available for the perfluorocarboxylates (except PFOA), the high frequency of detection of PFNA and other carboxylates in our study suggests that human exposure to long-alkyl-chain perfluorocarboxylates may be widespread. We also found PFOS in the serum and milk of rats administered PFOS by gavage, but not in the milk of rats not dosed with PFOS. Furthermore, we did not detect most PFCs in two human milk samples. These findings suggest that PFCs may not be as prevalent in human milk as they are in serum. Additional studies are needed to determine whether environmental exposure to PFCs can result in PFCs partitioning * Corresponding author phone: (770)488-7891; fax: (770)488-4609; e-mail:
[email protected]. 3698 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 13, 2004
into milk. Large epidemiological studies to determine the levels of PFCs among the U.S. general population are warranted.
Introduction Perfluorochemicals (PFCs) have been used since the 1950s in a wide variety of industrial and consumer products, including protective coatings for fabrics and carpet, paper coatings, insecticide formulations, and surfactants. In 1999, the United States Environmental Protection Agency (U.S. EPA) began an investigation of PFCs after receiving data indicating that perfluorooctanesulfonate (PFOS), a widely used PFC, was persistent, bioaccumulative, and found in human populations (1-5) and in wildlife (6-13) around the world. Furthermore, animal data suggested potential developmental, reproductive, and systemic toxicity (14-16). In May 2000, 3M, the sole manufacturer of PFOS in the United States and the principal manufacturer worldwide, announced it was discontinuing its perfluorooctanyl chemistries, including PFOS. Shortly thereafter, in June 2000, the EPA identified possible related concerns with respect to perfluorooctanoic acid (PFOA) and fluorinated telomers. PFOA is used primarily in the production of fluoropolymers, such as poly(tetrafluoroethylene) (PTFE, Teflon) and poly(vinylidine fluoride), and fluoroelastomers. These polymers have numerous uses in many industrial and consumer products, including soil-, stain-, grease-, and water-resistant coatings on textiles and carpet, personal care products, and nonstick coatings on cookware, and uses in the automotive, mechanical, aerospace, chemical, electrical, medical, and building and construction industries. In April 2003, the EPA released a preliminary risk assessment on PFOA that indicated potential human exposure to low levels of PFOA in the United States. (17). No clear association between human exposure to PFCs and adverse health effects has been established. However, on the basis of the following results from animal studies (15, 16, 18-20), a potential risk could exist for developmental and other adverse effects associated with exposures to PFCs in humans. Subchronic exposure to PFOS in rodents and primates results in adverse health effects, including reduction of body weight, liver hypertrophy, and decreased serum cholesterol and triglycerides (15, 16). Furthermore, exposure to PFOS during pregnancy in rats and mice results in maternal and developmental toxicity. Specifically, significant reductions of maternal weight gains and a marked decrease in circulating thyroid hormones in a dose-dependent manner were observed in pregnant rats and mice exposed to PFOS (20). The altered thyroid hormone metabolism, also detected in cynomolgus adult monkeys exposed to PFOS (15), is of potential concern because thyroid hormones regulate growth, metabolic rate, cardiac performance, and body temperature, and play a critical role in the normal development of the lung, inner ears, and nervous system (20). Perinatal hypothyroidism retards neurodevelopment and may have longlasting adverse effects on the development of intellectual and motor skills. Similarly, the alterations of thyroid hormones caused by PFOS exposure may interfere with cellular or functional maturation of target organs, such as the lung and the liver (19). The PFOS-exposed rat and mouse fetuses displayed a host of birth defects, including cleft palate, delayed ossification, anasarca (i.e., accumulation of serum in the connective tissue), and cardiac malformations (20). Furthermore, in utero exposure to PFOS in rats and mice 10.1021/es040332u Not subject to U.S. copyright. Publ. 2004 Am. Chem.Soc. Published on Web 05/28/2004
severely compromised postnatal survival in a dose-dependent manner (19). More importantly, the adverse developmental outcomes occurred at dose levels lower than those associated with teratologic outcomes (19). The medical surveillance of fluorochemical production workers in Decatur, AL, and Antwerp, Belgium, did not show substantial changes in lipid or hepatic clinical chemistry test results that were consistent with the known toxicologic effects of PFCs (1, 4). Similarly, no association between PFOA exposure and hepatic enzymes, lipoproteins, and cholesterol was found among 115 occupationally exposed workers (21). However, a recent retrospective cohort mortality study of the Decatur and Antwerp workers revealed an excess of bladder cancer: 3 deaths compared with 0.23 expected (22). In an earlier study, the relation between PFOA and mortality among 2788 male and 749 female workers employed during 1947-1983 at a plant that produced PFOA was examined (23). Ten years of employment in PFOA production jobs was associated with a 3.3-fold increase in prostate cancer mortality, compared with no employment in PFOA production jobs, although only six prostate cancer deaths occurred overall and four among the exposed workers. Because exposure to PFOS or PFOA in animals did not result in these particular toxicities, the results from these two retrospective cohort mortality studies must be interpreted cautiously. PFCs have been measured in biological (1-6, 8-12, 24) and environmental (25-28) matrixes by high-performance liquid chromatography-tandem mass spectrometry (HPLCMS/MS). Most of these methods used liquid-liquid extraction (LLE) to extract the PFCs from the biological matrix (1-4, 6, 8-12, 24). Methods using solid-phase extraction (SPE) instead of LLE have been reported for environmental (26, 27, 29) and biological (19, 20) samples. Although these methods have adequate sensitivity, the manual LLE or SPE steps are laborintensive and time-consuming. To address these limitations, we automated the extraction process and developed a sensitive method for measuring 13 PFCs, including PFOS and PFOA, in milk and serum by HPLC-MS/MS. We used this method to measure the levels of PFCs in the blood and milk of people nonoccupationally exposed to these compounds.
Materials and Methods Chemicals. Perfluorooctanesulfonamide (PFOSA), 2-(Nethylperfluorooctanesulfonamido)acetic acid (Et-PFOSAAcOH), 2-(N-methylperfluorooctanesulfonamido)acetic acid (Me-PFOSA-AcOH), potassium perfluorohexanesulfonate (PFHxS), potassium perfluorooctanesulfonate, ammonium perfluorooctanoate, and 1H,1H,2H,2H-tetrahydroperfluorooctanesulfonate (THPFOS) were provided by 3M Co. (Saint Paul, MN). Perfluoropentanoic acid (PFPeA), perfluorohexanoic acid (PFHxA), perfluoroheptanoic acid (PFHpA), perfluorononanoic acid (PFNA), perfluorodecanoic acid (PFDeA), perfluoroundecanoic acid (PFUA), and perfluorododecanoic acid (PFDoA) were purchased from Oakwood Products (West Columbia, SC). [1,2-13C2]perfluorooctanoic acid ([13C2]PFOA) was provided by Dupont Co. (Wilmington, DE). HPLC-grade methanol, acetonitrile, and water were purchased from Caledon (Ontario, Canada), acetic acid (glacial) was purchased from Sigma-Aldrich (St. Louis, MO), formic acid (98% min, GR) was purchased from EM Science (Gibbstown, NJ), and ammonium hydroxide (30%) was purchased from J.T. Baker (Phillipsburg, NJ). Formic acid (0.1 M, 500 mL) was filtered through a 500 mg Oasis HLB cartridge (Waters Corp., Milford, MA) before use. All other chemicals and solvents were used without further purification. Preparation of Standard Solutions and Quality Control (QC) Materials. Standard stock solutions (10 mg/mL) were prepared by dissolving solid standards in methanol. Nine working standard solutions containing all analytes were
prepared by serial dilutions in methanol to final concentrations such that a 50 µL spike in 1 mL of serum or milk would cover a concentration range from 0.1 to 100 ng/mL. The THPFOS and [13C2]PFOA stock solutions (1 mg/mL) in methanol were prepared from the solid standard and were diluted to provide working solutions such that a 50 µL spike onto 1 mL of serum or milk provided an approximate final concentration of 10 ng/mL. All stock solutions and standards were aliquotted and stored in polypropylene vials at or below -20 °C. The QC materials for milk and serum analysis were prepared in bulk from pooled breast milk purchased from Mother’s Milk Bank (San Jose, CA) and from calf serum (Gibco, Grand Island, NY), respectively. The pools were mixed uniformly, divided into three subpools, dispensed into small aliquots (ca. 2-7 mL) into prerinsed polypropylene vials, and stored at -20 °C until use. One subpool was used as a blank QC and to prepare the calibration standards, and the other two were enriched with PFCs as needed to afford lowconcentration (QCL, ∼7 ng/mL) and high-concentration (QCH, ∼35 ng/mL) subpools. The QC pools were characterized to define the mean and the 95% and 99% control limits of PFC concentrations by a minimum of 30 repeated measurements in a 3 week period. QC materials reextracted and analyzed after the initial characterization showed that the PFCs remained stable at -20 °C for at least 3 months. Subjects. The blood samples were collected in July 2003 from 20 Atlanta residents with no documented exposure to PFCs. The study participants (50% men) were predominantly Caucasian (90%) with a mean (standard deviation) age of 42.3 (10.2) years. Informed consent was obtained from all participants. The samples were collected in the morning (before 9:30 a.m.), refrigerated (2-4 °C) immediately after collection, and shipped via courier service on dry ice to the Centers for Disease Control and Prevention’s National Center for Environmental Health (CDC/NCEH) laboratory on the day of collection. On arrival, the blood samples were centrifuged on an Allegra 6 centrifuge (Beckman Coulter Inc., Fullerton, CA) at 2500 rpm for 15 min. The serum was transferred to polypropylene vials and stored at -40 °C until analysis. No information from the human milk donors or about the sampling procedures was available. The milk samples, shipped frozen to the CDC, were stored at -40 °C in their original glass or plastic containers until analysis. Sample Preparation. Standards, QCs, and blanks were prepared and processed using the same procedure. Only polypropylene labware was used for the preparation and analysis of samples. All standard solutions were allowed to reach room temperature before they were spiked into the matrix. To 3 mL of 0.1 M formic acid, placed in an 8 mL polypropylene test tube and spiked with 50 µL of internal standard solution and 50 µL of standard solution (for standards only) was added 1 mL of serum or milk. This solution was vortex-mixed, sonicated for 20 min, and placed on the Zymark RapidTrace Station (Zymark Corp., Hopkinton, MA). Automated SPE. Before each extraction run, the extractor lines were purged with methanol and water. Next, a 60 mg/3 mL Oasis-HLB column (Waters) was conditioned with HPLCgrade methanol (2 mL) and 0.1 M formic acid (2 mL). Afterward, the sample was loaded onto the cartridge at 1 mL/min. For milk samples, the SPE cartridge was washed with 3 mL of 0.1 M formic acid and 2 mL of 50% 0.1 M formic acid/50% methanol. For serum samples, the cartridge was washed with 3 mL of 0.1 M formic acid, 6 mL of 50% 0.1 M formic acid/50% methanol, and 1 mL of 1% NH4OH/water. Then, the cartridge was vented with 5 mL of air at 20 mL/ min. The PFCs were eluted from the SPE column with 1 mL of 1% NH4OH/acetonitrile and collected in a 4 mL polyproVOL. 38, NO. 13, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3699
TABLE 1. Multiple-Reaction-Monitoring Analysis of PFCs Indicating the PFC Precursor and Product Negative-Ion Masses (m/z)a analyte
abbreviation
precursor ion (M - H)- (m/z)
product ion (m/z)
perfluorooctanesulfonamide 2-(N-methylperfluorooctanesulfonamido)acetic acid 2-(N-ethylperfluorooctanesulfonamido)acetic acid perfluorohexanesulfonic acid perfluorooctanesulfonic acid tetrahydroperfluorooctanesulfonic acid perfluoropentanoic acid perfluorohexanoic acid perfluoroheptanoic acid perfluorooctanoic acid perfluorononanoic acid pefluorodecanoic acid perfluoroundecanoic acid perflurododecanoic acid [13C2]perfluorooctanoic acid
PFOSA Me-PFOSA-AcOH Et-PFOSA-AcOH PFHxS PFOS THPFOS (internal std) PFPeA PFHxA PFHpA PFOA PFNA PFDeA PFUA PFDoA [13C2]PFOA (internal std)
498 570 584 399 499 427 263 313 363 413 463 513 563 613 415
78 512 526 80b/99 80b/99 80 219 269 319 369 419 469 519 569 370
a M represents the molecular ion; the product ions were (M - CO H)- for the carboxylic acids; (M - SO H)- for Me-PFOSA-AcOH and Et2 3 PFOSA-AcOH, (SO3)- for PFOS, PFHxS, and THPFOS, and (SNO2)- for PFOSA. b Quantitation ion. We monitored the m/z 99 ion (FSO3-) for PFOS and PFHxS to confirm the presence of these compounds.
pylene tube, and the needle was rinsed with water. The automated SPE of 100 samples was completed in ∼4 h. The SPE eluate was evaporated to ∼100 µL under a stream of dry nitrogen (UHP grade) in a Turbovap evaporator (Zymark) at 55 °C. A total of 200 µL of 90% 20 mM acetic acid/10% methanol was added to the evaporated extract, and the extract was transferred into a polypropylene autosampler vial (with a polypropylene cap) and centrifuged. HPLC. A total of 12 µL of the reconstituted serum or milk extract was injected using an Agilent 1100 HPLC system (Agilent Technologies, Wilmington, DE) operating at a 300 µL/min flow rate with 20 mM ammonium acetate (pH 4) in water and methanol as mobile phase A and mobile phase B, respectively. The analytes were separated from other extracted components using a Betasil C8 column (3 × 50 mm, 5 µm; ThermoHypersil-Keystone, Bellefonte, PA), preceded by a Betasil C8 precolumn (3 × 10 mm), heated at 40 °C. The HPLC gradient program (14 min) was as follows: started at 60% mobile phase B, next, the mobile phase organic content was increased in 0.5 min to 80% mobile phase B and kept for 9 min, and then the mobile phase organic content was decreased in 0.5 min to 60% mobile phase B and kept for 3 min to equilibrate the column. MS/MS. Negative-ion TurboIonSpray (TIS), a variant of electrospray, was used to convert liquid-phase ions into gasphase ions on an API 4000 triple-quadrupole mass spectrometer (Applied Biosystems, Foster City, CA). The TIS settings were curtain gas (N2) 20 arbitrary units (au), heated TIS gas (zero air) 35 au, heated TIS gas temperature 400 °C, and ion spray voltage 4500 V. Ionization parameters and collision cell parameters were optimized individually for each analyte (Table 1). Unit resolution was used for both Q1 and Q3 quadrupoles. Data Analysis. Data acquisition and analysis for all of the samples, blanks, standards, and QC materials were performed using the Analyst software of the API 4000. Each ion of interest in the chromatogram was automatically selected and integrated. The peak integrations were corrected manually if necessary. We used the peak area ratio of each analyte to THPFOS or [13C2]PFOA (i.e., response factor, RF) for quantification. THPFOS was used for sulfonamides and sulfonates, and [13C2]PFOA was used for the carboxylates. Nine standard analyte concentrations encompassing the entire linear range of the method were used to construct two daily calibration curves in calf serum or milk of RF versus standard amount. The full set of nine standards was injected twice: before the samples (e.g., QCs, blanks, unknowns), and after all samples, to monitor sensitivity changes. The average calibration curve weighted by the reciprocal of the standard amount (1/x) was 3700
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 13, 2004
used for quantification. The levels of PFCs in the milk used for the preparation of the calibration standards were below the limit of detection (LOD). The native amounts of PFCs in the calf serum used for the preparation of calibration standards and QCs were calculated from an average calibration curve obtained from spiked standards extracted from water. The calf serum contained small amounts (between 4 may cause protein precipitation on the HPLC column or in the LC/MS interface during the analysis of serum extracts (36, 37). Therefore, to ensure the ruggedness of the HPLC method, we used an eluent buffered at pH 4 and a maximum 80% methanol content during the HPLC gradient program. Coelution of PFOA and THPFOS did not affect the quantitation of PFOA. PFOA concentrations obtained in the presence or absence of THPFOS using [13C2]PFOA as internal standard were unchanged. Spiked serum and breast milk were analyzed repeatedly to determine the LOD, accuracy, and precision of the method. The LOD was calculated as 3S0, where S0 is the standard deviation as the concentration approaches zero (38). S0 was determined from five repeated measurements of low-level 3702
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 13, 2004
standards. Except for PFHxA, PFPeA, and PFDoA in serum, which had a lower SPE recovery and higher LODs than the other PFCs, the LODs were 0.1-0.6 ng/mL in serum and 0.1-1 ng/mL in milk (Table 2). These values, which reflect the good sensitivity of the method, are lower than the detection limits previously achieved in serum (2, 24) and environmental water (25, 26) samples. The calibration curves showed adequate linearity (i.e., correlation coefficients of 0.97-0.99). The interday variation of the calibration curve slopes, measured as the average coefficient of variation (CV), was 7-30%. The method accuracy was assessed by five replicate analysis of serum and milk spiked at three different concentrations (1, 5, and 25 ng/mL) and expressed as a percentage of the expected value. The intraday variability, reflected in the method accuracy, was very good (77-136%) for most analytes at all spike levels (Table 2). We determined the method precision by calculating the CVs of 30 repeated measurements of the QCL and QCH materials over a period of 3 weeks (Table 3). These CVs, which reflect the intraday and interday variability of the method, show very good precision (18% and 10%) for PFOA for which a labeled internal standard was available. For the rest of the analytes, the CVs
TABLE 3. Precision of Measurements of PFC Concentrations in Spiked QC Serum Pools (N ) 30)a QC low
QC high
analyte
mean
CV (%)
mean
CV (%)
PFOSA Me-PFOSA-AcOH Et-PFOSA-AcOH PFHxS PFOS PFOA PFNA PFDeA PFUA PFDoA
7.3 4.8 1.7 6.9 6.6 8.2 6.3 4.3 5.0 4.6
19 27 29 14 27 18 22 19 25 27
31.3 25.7 9.1 36.8 35.7 39.1 30 22.9 23.3 21.5
17 25 28 12 21 10 16 21 26 28
a Mean concentrations in nanograms per milliliter. CV (%) is the coefficient of variation.
were good (i.e., 12-29%) and adequate for quantification despite the lack of isotope-labeled internal standards. We tested the usefulness of our method by analyzing milk and serum of Sprague-Dawley rats administered PFOS by gavage for a cross-fostering study conducted at 3M several years ago (14). Specifically, we analyzed 10 archived milk and serum samples collected on lactation day 14 from two case and eight control rats during the 3M cross-fostering study. We found PFOS at parts-per-million levels in the serum (and milk) of the two treated animals: 196 µg/mL (100 µg/ mL) and 116 µg/mL (13.7 µg/mL). We found that the mean (and range) of PFOS concentrations in serum of the control rats was 80 ng/mL (1-335 ng/mL). We did not detect PFOS in the milk samples of the control animals. The serum PFOS levels determined using our method were comparable to those previously obtained by 3M scientists (39). No measurements of PFOS in milk were done at 3M. The presence of PFOS in the milk of the treated rats suggested that even though PFOA (and presumably all PFCs) is tightly bound to proteins in the plasma (40), PFCs may be incorporated into milk. However, we did not find detectable concentrations of most PFCs in two human milk samples, only PFPeA (1.56 ng/mL) in one of them and PFHxA (0.82 ng/mL) in the other. These findings suggest that PFCs may not be as prevalent in
FIGURE 2. Typical HPLC-MS/MS chromatogram of an extract of a 1 mL serum sample spiked with a 2.5 ng/mL standard mixture of PFCs, [13C2]PFOA, and THPFOS. human milk as they are in serum. However, because we only analyzed two human milk samples and information on the donors and collection procedures was not available, these data should be interpreted cautiously. Additional studies are
TABLE 4. Concentrations (ng/mL) of Selected PFCs Measured in Serum Samples Collected in Atlanta, GA, in July 2003 from 20 Adultsa sex
age
PFOSA
Me-PFOSA-AcOH
Et-PFOSA-AcOH
PFHxS
PFOS
PFOA
PFNA
PFDeA
PFUA
F F F F F F F F F F M M M M M M M M M M
54 31 36 52 67 49 31 36 23 36 39 36 42 40 54 50 46 49 39 36
0.7 0.3 0.2 0.2
