Environ. Sci. Technol. 2002, 36, 1681-1685
Quantitative Characterization of Trace Levels of PFOS and PFOA in the Tennessee River K . J . H A N S E N , * ,† H . O . J O H N S O N , † J. S. ELDRIDGE,† J. L. BUTENHOFF,‡ AND L. A. DICK† 3M Environmental Laboratory and 3M Medical Department, Building 2-3E-09, P.O. Box 33331, St. Paul, Minnesota 55133-3331
Although there is evidence of widespread distribution of organic fluorochemicals such as perfluorooctane sulfonate and perfluorooctanoate, in the environment, the versatility of these compounds in industrial and commercial applications complicates characterization of pathways into the environment. A solid-phase extraction method coupled with HPLC-negative-ion electrospray tandem mass spectrometry was developed to quantitatively measure trace levels of organic fluorochemicals in drinking water and surface water. Using this method, certain fluorochemicals can be quantitatively measured in water samples down to 25 ppt, a level well below calculated drinking water advisory levels. To assess fluorochemical distribution in a localized geography and to ascertain whether fluorochemical manufacturing facilities contribute to environmental levels of fluorochemicals, 40 water samples were collected on an 80-mi stretch of the Tennessee River, near a fluorochemical manufacturing site in Decatur, AL. Low levels (ppt) of perfluorooctane sulfonate were determined throughout the stretch of river sampled. Concentrations of the measured fluorochemicals increased downstream of the fluorochemical manufacturing facility, indicating that effluent from manufacturing is one likely source of organic fluorochemicals into the river.
Introduction Organic fluorochemical compounds are used in a wide variety of industrial and commercial applications and processes; they are used as polymer additives, lubricants, fire retardants and suppressants, pesticides, and surfactants (1). Recent studies have characterized trace levels of certain fluorochemical compounds in the serum of nonoccupationally exposed humans and in some wildlife tissue samples; this information is accessible, among other places, on the public docket for the EPA Office of Pollution, Prevention and Toxic Substances (2). Specifically, in a limited set of human sera samples purchased from chemical supply companies, perfluorooctane sulfonate (PFOS, C8F17SO3-), perfluorooctanoate (PFOA, C7F13COO-), and perfluorooctanesulfonyl amide (PFOSA, C8F17SO2NH2) were determined to be present at >1.0-75 µg/L (3). These general population sera levels are below the serum levels measured previously in occupationally * Corresponding author e-mail:
[email protected]; phone: (651)733-2062; fax: (651)778-6176. † 3M Environmental Laboratory. ‡ 3M Medical Department. 10.1021/es010780r CCC: $22.00 Published on Web 03/12/2002
2002 American Chemical Society
exposed workers and are well below the serum PFOS concentration (approximately 100 µg/mL) associated with the earliest measurable clinical response, a decrease in total serum cholesterol, in cynomologus monkeys (2, 4-6). PFOS has also been detected in tissues collected from wildlife (2, 7). The discovery of organic fluorochemicals in human serum and in the environment has led to the initiation of studies to characterize the effects and the extent of fluorochemical presence in the environment (8). These studies have been challenged by the lack of sufficiently sensitive and compoundspecific analytical methods. Current methods of analysis for organic fluorochemicals include combustion techniques and high-performance liquid chromatography with UV detection, both of which are nonspecific tools, and nuclear magnetic resonance, gas chromatography (GC)-mass spectrometry, and GC-electron capture detector, all of which lack the selectivity or sensitivity needed for “natural” water analysis of these organic fluorochemicals. Additionally, the GC techniques require derivitization prior to analysis, and the derivitization methods have not been successfully applied to PFOS analysis (9-14). Recently, an HPLC-MS/MS method was utilized to characterize levels of specific fluorochemicals downstream of an airport where fluorochemical-containing fire-fighting foam was used (15). The purpose of this field study was to evaluate whether a newly developed method for trace level analysis of fluorochemicals could be applied to determine if fluorochemical manufacturing facilities may be a source of PFOS and PFOA in the environment.
Experimental Section Standards and Reagents. PFOS and PFOA were obtained from Fluka (Milwaukee, WI). HPLC-grade methanol was purchased from E. M. Science (Gibbstown, NJ), and ammonium acetate was received from Aldrich (Milwaukee, WI). Sep-Pak Cac 6 cm3 (1 g) C18 SPE (solid-phase extraction) cartridges from VWR (Milford, MA) were used for extraction. Field Campaign and Site. Sample collection took place November 14-16, 2000, starting nearly 35 mi upstream of the fluorochemical manufacturing facility discharge. The full length of the Tennessee River is approximately 650 mi long; it originates where the Holston and French Broad Rivers meet in Knox County, TN (16). The Tennessee River eventually joins the Ohio River in Kentucky before flowing into the Mississippi River. The total flow of the river measured at Wheeler Dam through the hydroturbans (no spillway releases occurred) on November 14, 2000, was 55 600 ft3/s; on November 15, the flow was measured at 49 200 ft3/s; and on November 16, it was 46 800 ft3/s. The average historical flow rate from 1940 to the present including spillway releases is 49 000 ft3/s (17). The field study encompassed approximately 80 mi of the Tennessee River in Alabama. At the first sample collection point furthest upstream (mile marker 337), the Tennessee River is approximately 0.4 km across. The river remains relatively narrow until just upstream of the city of Decatur, AL (mile marker 305). As the river approaches Wheeler Dam, it widens to approximately 1-2 km. The width remains within that range until the last sample collection point furthest downstream is reached, just upstream of Wilson Dam (mile marker 261). Each sample was collected in the center channel of the river. This stretch of the Tennessee River was chosen for sampling because discharge from a fluorochemical manufacturing facility enters the river midway through the sample VOL. 36, NO. 8, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1681
collection area. The city of Decatur, AL is located just upstream from the industrial discharge from the fluorochemical manufacturing facility at Baker’s Creek. Although there are several other sources of industrial discharge into Baker’s Creek and into the Tennessee River at other locations downstream of Decatur, there are no other obvious, potential sources of fluorochemicals into the river except perhaps the Decatur Wastewater Treatment Plant discharge that is located several miles upstream from the fluorochemical manufacturing facility. Fluorochemical and nonfluorochemical manufacturing processes are discharged together into Baker’s Creek at an average rate of approximately 16 million gal/day. Sample Collection. New 1-L Nalgene (Rochester, NY) polypropylene containers with narrow-mouth bottles and screw tops were used for collection of samples in the field. Teflon bottles and Teflon-lined caps were avoided throughout the experiment as interferences may be introduced to the samples of extracts. Glass sample collection bottles were also avoided as the target analytes may bind to the glass in aqueous solutions. The containers were rinsed with methanol and deionized water prior to use. Samples were collected over 3 days at approximately 2-mi intervals along the Tennessee River; a fluorochemical manufacturing facility is located on the river about halfway through the section of river included in the sample set. Samples were collected using a subsurface grab-sampling device that allowed the collection container to be opened and closed 15 cm below the surface. Deepwater sample locations from the center channel of the river were selected to minimize the amount of sediment in the samples. Each sample was sealed individually in a zip-lock plastic bag, stored on ice in a cooler, and refrigerated at 4 °C once they were returned to the lab. Forty samples of river water were collected. Along with each sample collection, several field parameters were logged; latitude and longitude by GPS, river depth, pH, conductance, dissolved oxygen, water temperature, air temperature, and general remarks about weather conditions. Sample Collection QC. In addition to the 40 water samples collected, a large number of QC samples were also collected. Five field blanks, defined as sample containers filled with deionized water prior to sampling, were prepared and exposed to the same conditions as the samples. One field blank was collected for each sample cooler used for sample storage and shipping. Six field duplicates, Tennessee River water samples collected in duplicate at six different locations and handled exactly as the 40 original samples, were also collected. For each sampling day, at least one field matrix spike (FMS) was prepared. The FMS consisted of a sample of Tennessee River water spiked with PFOS and PFOA at 1000 ng/L in the field. Two field spike control samples, deionized water spiked with the target analytes in the field, were also collected each day of sample collection. QC samples that were spiked in the field were maintained in a separate cooler to avoid any cross contamination. Solid-Phase Extraction. After the samples were equilibrated to room temperature, the sample bottle was inverted to thoroughly mix the sample. A total of 40 mL of the sample was transferred to a 50-mL polypropylene centrifuge tube (VWR; Milford, MA); an appropriate aliquot of the matrix spike solution prepared in methanol was added to the lab control matrix spike samples at this step. The SPE cartridge was conditioned by passing 10 mL of methanol through the cartridge followed by 5 mL of Type 1 water (approximately 2 drops/s). The column was not allowed to go dry at any time. The analytical sample was loaded onto the SPE cartridge, and the aqueous eluate was collected at a rate of about 1 drop/s and discarded. The cartridge was then washed with 5 mL of 40% methanol in water; this wash was also discarded. The analytes were eluted with 100% 1682
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 8, 2002
methanol so that exactly 5 mL was collected into a 15-mL polypropylene centrifuge tube. A portion of the methanol eluate was transferred to an HPLC vial (Kimble/Kontes; Vineland, NJ) for analysis. HPLC-Electrospray Tandem Mass Spectrometry. Prior to detection, analytes were separated using an Agilent Technologies 1100 series HPLC system modified with low dead-volume tubing. A 10-µL sample of each extract was injected onto a 50 × 2 mm (5 µm) Keystone (Bellefonte, PA) Betasil C8 column. The mobile phase consisted of (A) 2 mM ammonium acetate and (B) methanol. At a flow rate of 300 µL/min, the separation proceeded from initial conditions of 40% B to 100% B at 7.5 min before returning to 40% B. The column temperature was maintained at 35 °C, and the total run time was 14 min. The PFOS and PFOA standard material are mixtures of linear and branched isomers (approximately 80% linear), purified from other homologues using prescale HPLC. To maximize analyte response, the isomers were not separated chromatographically but are represented as a single peak. Isomerically separated standards for PFOS and PFOA are not available. It is assumed that the response factor for branched and linear isomers are equivalent and that the standard mixture is representative of that identified in the samples. For quantitative determination, the HPLC was interfaced to a Micromass Ultima atmospheric pressure ionization tandem mass spectrometer (Beverly, MA) in the negativeion electrospray mode. Instrumental parameters, such as cone voltage, were optimized to 60 and 15 V to transmit the [M - K]- (m/z ) 499) and [M - NH4]- (m/z ) 413) ions for PFOS and PFOA, respectively. For each analyte, the energy of the collision gas (argon) was optimized for quantitation based on single-product ion transition noted here: PFOS (499 > 99), PFOA (413 > 169). For PFOS, the m/z ) 99 transition corresponds to the FSO3- fragment; the m/z ) 169 transition for PFOA corresponds to the C3F7- fragment. Although the 499 > 80 transition (m/z ) 80 corresponds to the SO3- fragment) gives a stronger signal for PFOS, occasionally interferences are observed from this transition and the 499 > 99 transition was preferred. In our experience, optimal collision gas energies vary greatly from system to system; in our system 45 eV was used to generate the fragment for PFOS and 30 eV was used to generate the fragment for PFOA. The desolvation gas temperature was 250 °C, and the desolvation gas was set at 650 L/h. The cone gas was set at 50 L/h, and the source block was heated to 150 °C. A pressure of 3 × 10-3 mbar was maintained in the collision cell. A dwell time of 0.4 s was used to monitor each transition. Extraction QC and Characterization. To monitor potential contamination during extraction, nine method blanks consisting of bottled drinking water were extracted along with the samples. Six river water samples were split: one split was analyzed as per the method, the second split was centrifuged and decanted prior to extraction according to the method. Two eight-point extracted curves spanning concentrations from 10 to 5000 ng/L and six mid-level calibration checks were prepared along with the samples. An extracted curve is a calibration curve prepared by spiking samples of bottled water with a specific amount of each target analyte followed by extraction of the water sample. The extracted curves used for quantitation, consisting of at least five active points and ranging from approximately 10 pg/mL to 10 ng/mL in bottled drinking water (corresponding to a final extract concentration of 0.080-80 ng/mL), were plotted using a 1/X-weighted linear fit. The curve was not forced through zero, and all active curve points were required to be within 30% of the theoretical value. Low- or high-level points that were outside the area of linearity or that did not otherwise meet stated criteria were deactivated. Mid-level extracted calibration curve
TABLE 1. Accuracy of SPE Extraction Method Determined by Evaluating Extracted Samples of Drinking Water vs a Standard Curve Extracted from Bottled Water
50 ng/L 100 ng/L 500 ng/L a
PFOS % recovery ( SD
PFOA % recovery ( SD
101 ( 29a 112 ( 16a 96 ( 11
91 ( 11 97 ( 8 83 ( 18
One sample result eliminated as an outlier using the Grubb’s Test.
TABLE 2. Determination of Extraction Efficiency, LOQ, and LOD of SPE Method Determined by Evaluating Extracted Samples of Bottled Drinking Water vs a Standard Curve Prepared in Methanol PFOS % extraction efficiency ( SD
PFOA % extraction efficiency ( SD
50 ng/L 100 ng/L 500 ng/L
120 ( 2 127 ( 3 92 ( 1
112 ( 34 121 ( 3 109 ( 6
method LOQ method LOD
10-25 ng/L 5 ng/L
25-50 ng/L 25 ng/L
verification (CCV) samples were analyzed after every five samples to ensure the continued validity of the calibration curve. CCVs were required to be within 30% of the theoretical value.
Results and Discussion Characterization of the Method. Prior to sample analysis, a series of experiments was designed to characterize the analytical method. Bottled drinking water was used for all samples, and calibration QC was used for the characterization
experiments. Accuracy and reproducibility were determined versus an extracted curve by extracting eight samples of bottled drinking water spiked at three different levels across the linear range of the method, for a total of 24 analyses per analyte. The results for all analytes in each of the three concentration levels, along with the standard deviation for the analysis, are summarized in Table 1. Two results from the 48 analyses were obvious outliers (as determined by the Grubb’s Test) and were not included in the calculations summarized in Table 1. For all analytes, reproducibility was within 6%. Precision of the river water sample extraction, determined by triplicate extraction of a single river water sample, was within 10% for all analytes. Instrumental precision was determined by five replicate injections of a single drinking water extract spiked at 100 ng/L and was within 6% for all target analytes. The limit of detection was determined as per EPA Regulation 40 CFR part 136, Appendix B. For each analyte, seven low-level spikes were prepared and analyzed. On the basis of the standard deviation associated with the replicate analysis, a limit of detection was calculated. This calculated limit of detection was verified by analyzing a blank sample that was spiked at the calculated level and extracted. The limit of quantitation was determined on a run by run basis and was based on the standard curve generated with each individual data set. The limit of quantitation was defined as the lowest active point in the calibration curve. The low point was required to have a peak area higher than that of the blank and be back-calculated to quantitate within 30% of the spiked concentration. A summary of the LOD and the approximate LOQ for each analyte is included in Table 2. Extraction efficiency was determined by evaluating extracts of bottled drinking water prepared at three levels (50, 100, and 500 ng/L) versus an unextracted curve prepared in methanol. Three separate extracts were prepared at each level. PFOS recoveries ranged from 92 to 127%; PFOA
FIGURE 1. Results of the SRM analysis for each target analyte of (1) an extraction blank, (2) a field blank, and (3) a sample from the Tennessee River downstream of Baker’s Creek. The y-axis depicts relative ion abundance for each set of chromatograms. In the sample of river water (which corresponds to 100% ion abundance for comparison purposes), PFOS was measured to be approximately 112 ng/L; PFOA was measured at approximately 275 ng/L. VOL. 36, NO. 8, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1683
TABLE 3. Results of Analysis of Water Samples Collected on the Tennessee River, near Decatur, ALa
sample 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
approx mile marker 337 335 333 331 329 327 325 323 321 319 317 315 313 311 309 307 305 303 302 301 299 297 295 293 291 289 287 285 283 281 279 278.2 277 275 273 271 269 267 265 263 261 259.4
river depth (ft)
conductance (µΩ/cm)
14 184 36 183 20 184 20 184 28 183 38 186 39 184 40 185 20 184 37 184 26 184 22 185 21 185 20 185 19 186 24 184 22 184 33 185 26 184 24 184 23 186 25 185 23 188 26 187 20 187 21 186 18 185 28 183 29 178 32 180 46 182 Wheeler Dam 26 187 50 191 19 183 24 185 38 191 48 199 55 201 70 202 75 201 Wilson Dam
[PFOS] (ng/L)
[PFOA] (ng/L)
27.8 28.9 28.8 25.8 36.9 16.8 27.4 31.0 26.9 22.3 21.8 21.4 18.4 31.6 51.9 52.6 37.1 39.4 39.4 54.1 37.3 30.3 74.8 96.4 98.0 107 136 140 106 134 106
