Towards the development of a standardized method for extraction and analysis of PFAS in biological tissues

Adam D. Point a, Thomas M. Holsen *b, Sujan Fernando c, Philip K. Hopke *d and Bernard S. Crimmins be
aInstitute for a Sustainable Environment, Clarkson University, Potsdam, NY, USA
bCivil and Environmental Engineering, Clarkson University, Potsdam, NY, USA. E-mail: tholsen@clarkson.edu
cCenter for Air Resources Engineering and Science, Clarkson University, Potsdam, NY, USA
dDepartment of Public Health Sciences, University of Rochester Medical Center, Rochester, NY, USA. E-mail: phopke@clarkson.edu
eAEACS, LLC, New Kensington, PA, USA

Received 1st September 2019 , Accepted 10th September 2019

First published on 16th September 2019


The stability and surface activity of per- and polyfluoroalkyl substances (PFAS) led to their extensive and diverse commercial and industrial use. However, these properties are accompanied by generally high aqueous solubility and toxic effects, resulting in growing scientific and public concern. Given the surge in PFAS analyses in recent years, recurrent incidents of analytical inconsistencies reported in the literature, and the demand for robust, standardized analytical methods, an effort to condense and quantitatively define/describe PFAS analytical issues was necessary. Thus, this work had two objectives. A review of PFAS analytical literature identified and compiled analytical challenges with specific attention to interlaboratory studies and information related to low analyte recoveries. Data from a series of systematic experiments quantitatively helped to identify those differences in the analytical methodologies that caused inconsistencies and may compromise PFAS analytical results. The adsorptive properties of these unique compounds were identified as the primary source of analytical irreproducibility, particularly for analytes with a perfluoroalkyl chain length exceeding nine carbons. More specifically, solvent composition (water/methanol ratio) was found to be the primary contributor to adsorptive analyte losses with a significant increase in adsorption occurring when water content of the analyzed solution was raised beyond 25%. The container material used also affected adsorptive losses for solutions with elevated water fractions. Contrary to literature reports, polypropylene containers generally resulted in greater analyte losses compared to amber borosilicate glass containers. These findings led to modifications of sample preparation procedures to provide a more robust analytical methodology. This methodology was subsequently utilized in the analysis of 300 biological tissue samples. A detailed procedure and discussion of its performance are provided. These findings and the connections made to the cited literature will provide the framework for a standardized method for PFAS extraction and analysis of biological tissues.



Water impact

Per- and polyfluorinated substances (PFAS) in the environment are of considerable concern for human and environmental health. However, there are difficulties in the analysis of PFAS in environmental matrices including water and biota that have produced inconsistent results. We have reviewed prior studies of analytical processes for PFAS, experimentally identified issues that produced inconsistent results, and suggest approaches that alleviate these issues.

Introduction

In this work, polyfluoroalkyl substances or PFAS are defined as any species within the family of aliphatic chemicals that contains at least one perfluoroalkyl (CF2) group [nomenclature taken from Buck et al.].1 A subset of PFAS includes the perfluoroalkyl sulfonic and carboxylic acids (PFSA and PFCA, respectively) that are the focus of the current study. Perfluorooctanoic acid (PFOA) and perfluorooctane sulfonate (PFOS) are the most studied PFAS. They were originally synthesized due to their unique stability and surface tension lowering properties. These attributes enabled their use in an assortment of applications and products including metal plating, inks, varnishes, waxes, lubricants, surface coatings for leather, paper, and textiles, and firefighting foams for the control of hydrocarbon fuel fires. PFOA and other related PFCAs also served as stabilizing agents during fluoropolymer production including polytetrafluoroethylene (PTFE) and polyvinylidene fluoride (PVDF).2,3

The widespread distribution,4 environmental persistence,5 and potential toxic effects6,7 led to the phase out of perfluorooctane sulfonyl fluoride (PFOSF) that was used to make PFOS and its precursors.8 A voluntary stewardship agreement between the United States Environmental Protection Agency (USEPA) and eight leading global producers aimed to eliminate emissions and product content of PFOA and related chemicals by 2015.9 However, despite these efforts to limit production and release in the United States, Western Europe, and Japan, the release of these compounds into the environment is expected to continue due to an increase in production in countries like India, China, Poland, and Russia10 as well as degradation of waste stocks (leaching from stored and/or disposed products).11 Thus, long-term monitoring will be needed to determine the environmental distribution and bioaccumulation of PFAS in the environment. However, at the core of these studies are having reliable analytical methods with well-characterized measurement uncertainties. These uncertainties have not been adequately documented and/or understood in previous studies.

Standardized methods such as EPA 537.1, ASTM D7979-17, and ASTM D7968-17a have been recently developed to provide a “validated” baseline for PFAS analysis. However, these methods primarily describe the instrumental analysis of aqueous matrices, with only one method (ASTM D7968-17a) addressing extraction from a solid (sand/soil) matrix. Furthermore, the scope section of method ASTM D7968-17a acknowledges the limitations of the extraction method described and suggests that a more exhaustive procedure should be used for difficult matrices.12 This leaves the individual researcher the responsibility of validating a suitable extraction and cleanup method if the matrix of interest is outside the scope of the standardized protocols. Advances in analytical technologies have made the instrumental reproducibility relatively trivial compared to workflows that require extraction and sample purification steps. Therefore, understanding the measurement uncertainty of the entire workflow (extraction, clean-up and analysis) is vital to ensuring data comparability for trends analyses13 and health/environmental risk assessments for the growing list of PFAS contaminated sites.14

As PFAS analyses became more common and various analytical challenges were encountered, a series of interlaboratory studies (ILSs) were performed to help develop robust methods. van Leeuwen et al. organized two PFAS ILSs in 2006 (ref. 15) and 2009.16 A variety of extraction and clean-up methods were employed by the participants to prepare samples for analysis generally by LC coupled to a tandem quadrupole mass spectrometer (MS/MS). Early ILSs revealed interlaboratory precision was quite poor. The reports of these ILSs documented moderate precision improvements although participants in both studies struggled with inefficient analyte extraction from distributed matrices, particularly for analytes with carbon chains ≥10 carbons. In the second ILS published in 2009,17 participants demonstrated greater precision when determining PFAS in human serum compared to the distributed study standard, suggesting that matrix cleanliness is not necessarily a predictor of method performance. Further precision improvements were reported for PFAS in human blood serum in an ILS published in 2010.18

In 2012, Reiner et al.19 report generally high precision among seven ILS participants for PFOS in fish fillet tissue with %RSDs for PFOS below 7%, leading to the establishment of reference values for two NIST Standard Reference Materials (SRMs 1946 and 1947). However, poor agreement among the submitted data for PFAS in less frequently analyzed abiotic matrices (analyte-specific %RSDs 10–165%) distributed for an ILS published in 2015 (ref. 20) contrasted with the generally increasing trend in precision over time demonstrated by previously summarized ILSs. Generally, increasing native standard quality (purity and isomer composition), broadening the array of mass-labeled surrogates, and heightening sample cleanup effectiveness were identified as the most significant reasons for the improved interlaboratory precision over time. However, variability among ILS participants' analytical methods (e.g. surrogates used, extraction and/or cleanup methods implemented) generally inhibited deterministic reasoning for the discrepancies among the submitted data that further emphasizes the need for standardized methods and quantitative assessments describing how alterations to method parameters affect end results. A detailed discussion of the cited literature is provided in section S1 of the ESI including a comparison of cited ILSs in Table S1.

van Leeuwen et al.15 highlighted that early PFAS analytical methods were hampered by limited availability of mass-labeled standards. Application of such standards particularly in isotope dilution methods, can correct for both analyte loss during sample preparation and ionization effects during MS detection.16 Irreversible adsorption of dissolved PFAS to glass surfaces is generally suggested as a challenge in the PFAS literature.21–25 This adsorption was found particularly problematic for longer chain PFAAs, and especially when the sample solvent contained elevated water-to-organic solvent ratios. Applying a narrow range of mass-labeled surrogates across a broad range of target analytes with varying degrees of adsorption potential could have resulted in the poor method performance noted in the summarized ILSs. However, the origin of these assertions regarding PFAS adsorption to glass surfaces is unclear because these statements lack citations and/or supporting data. These publications most often suggest polypropylene (PP) as an alternative material to eliminate this undesired adsorption.

During method validation for alkaline digestion of biological samples, So et al.26 examined PFAA recovery as a function of potassium hydroxide (KOH) concentration in both water and methanol. Although not identified as adsorptive losses, given constant KOH concentration, recoveries for PFAAs containing a perfluorinated carbon chain ≥9 were consistently lower for aqueous solutions. More specifically, recoveries for perfluoroundecanoic acid (PFUnA) and perfluorododecanoic acid (PFDoA) were in the range of 2.5–54% for KOH concentrations spanning 0.01–2.0 N (recoveries decreased with increasing KOH concentration). Sepulvado et al.27 found similar results while testing their aqueous sample preparation method. For these tests, target analytes and corresponding surrogates were spiked into 50 mL PP tubes, and subsequently transferred to autosampler vials for analysis. Although analyte-specific recoveries were not provided, PFUnA and PFDoA (the longest chain analytes tested) recoveries fell to the 38–52% range.

Taniyasu et al.28 used a weak anion exchange (WAX) SPE cleanup method, which produced 80–100% recoveries of target PFAS for cartridge spike experiments, but the recoveries for 14 and 16 carbon PFCAs dropped to approximately 50% when processing spiked Milli-Q™ water. Similar recoveries of 75% and 50% were noted for 14 and 16 carbon PFCAs during testing of their basic (KOH) methanol extraction method, which included a 1[thin space (1/6-em)]:[thin space (1/6-em)]100 dilution of extraction solvent with Milli-Q™ water prior to SPE cleanup. Despite recovering 10–30% of this lost mass upon rinsing the sample tube with methanol, some mass remained unaccounted for suggesting irreversible adsorption to the polypropylene containers.

Kelly et al.29 reported mass-labeled surrogate recoveries of 30–120%, but details regarding which species demonstrated low recoveries were not provided. Shoemaker et al.24 reported recoveries in the 75–85% range for PFOS and PFCAs with ≥11 carbons during SPE concentration of spiked deionized water. The majority of the lost mass was recovered upon rinsing the sample bottle with methanol during SPE cartridge elution. De Silva et al.30 also employed the WAX SPE cleanup method, and experienced general target PFAS recoveries of >86%, but 13 and 14 carbon PFCA recoveries were only 71% and 45%, respectively. The variable recoveries among analytes described above support the application of a complete array of mass-labeled surrogates during quantitation.16

Preliminary testing in our laboratory was performed to validate the initially selected sample extraction, cleanup, and analysis methodology. However, low and inconsistent surrogate recoveries during this validation process warranted investigation to identify the probable cause(s) including ionization suppression in the mass spectrometer. Some of the relevant issues are discussed in section S2 of the ESI. Once these issues were resolved, the work focused on the extraction and cleanup steps. Based on a literature review, the use of glass autosampler vials during the instrumental analysis of samples was the first point of concern, followed by the relatively high sample diluent water/organic solvent fraction (2/3). A series of experiments were performed to identify the sources of the recovery inconsistencies. The first tests focused on the instrumental analysis process to discern how the sample vessel material and solvent composition choices affected adsorptive analyte losses. The second round of tests optimized the SPE sample cleanup method. These two sets of experiments were then used to provide the basis for a sample processing methodology as part of the Standard Operating Procedure (SOP 10.4) for extraction, cleanup, and analysis of biological tissues presented in the ESI.

Experimental

Materials

The analytical standards were purchased from Wellington Laboratories, Inc. (Guelph, Ontario) as catalogue numbers PFAC-MXB, MPFAC-MXA, M2PFTeDA, M2PFHxDA, and M5PFHxA. Table S2 describes all of the native and mass-labeled species used. In general, native and mass-labeled standards included PFCAs and PFSAs with perfluorinated chain lengths ranging from three to fifteen. All purchased stock solutions are supplied in amber glass ampules with methanol as the solvent, and all working solutions were prepared in amber borosilicate glass (ABG) vials using methanol as the diluent. Optima LC-MS grade solvents (water, methanol, and acetonitrile) and ammonium hydroxide solution (ACROS Organics™ brand 28–30% ACS reagent grade, part number 423305000) were purchased from Fisher Scientific along with Evergreen brand 15 mL PP centrifuge tubes (part number 05-558-33C). SPE cleanup was performed using Oasis WAX cartridges (3 cc, 60 mg sorbent, 30 μm particle size, Waters Corp. part number 186002490). ABG and PP autosampler vials were purchased from Agilent Technologies, Inc. (part number 5182-0716) and Waters Corp. (part number 186005219), respectively. Formic acid (98–100%) was obtained from Sigma Aldrich (part number 27001-500ML-R).

Instrumentation

Ultra-high-performance liquid chromatography – tandem mass spectrometry (UPLC-MSMS) analyses were performed utilizing a Vanquish™ Flex Binary UHPLC System coupled with a TSQ Altis™ Triple Quadrupole Mass Spectrometer (Thermo Scientific™, Waltham, MA). System background was separated from injected target analytes using an isolator column (Waters Corp., part number 186004476) installed between the mobile phase mixer and the sample loop. Target analytes within the sample aliquots were separated using an Acquity UPLC® HSS T3 analytical column (2.1 × 100 mm, 1.8 μm, Waters Corp., part number 186003539). Details of the UPLC solvent gradient and monitored ions are supplied in Tables S3 and S4, respectively. Mass spectral data were acquired using Xcalibur version 4.2.28.14 and processed using TraceFinder™ 4.1 – EFS software (Thermo Scientific™, Waltham, MA) and Microsoft Excel (final data organization, calculations, etc.). Statistical analyses were performed using SigmaPlot Version 12.0 (Systat Software, Inc.).

Surface adsorption

Fig. 1 presents the study design for surface adsorption testing. Sample container materials (ABG and PP) and solvent compositions (75[thin space (1/6-em)]:[thin space (1/6-em)]25 and 25[thin space (1/6-em)]:[thin space (1/6-em)]75 water/methanol) were studied conjunctively by analyzing PFAA native standard solutions containing all of the native analytes listed in Table S2. To account for differences in volumetric capacities between the ABG and PP autosampler vials (2 mL and 0.7 mL, respectively), solutions in ABG and PP vials contained a total volume of 1.9 and 0.7 mL, respectively. All PFAA solutions contained 0.7 ng mL−1 of each native analyte, were prepared from the same stock solution, and then analyzed based on analyte-specific peak areas normalized using a single 13C-labeled injection standard (M5PFHxA). A concentration of 0.7 ng mL−1 was chosen because it is in the middle of the instrumentation's linear range and is within the concentration range typically encountered for environmental PFAS samples. For each solution, the required water volume was added last followed by 3 inversions by hand and 2 minutes of bath sonication immediately prior to the first injection. This procedure minimized possible variability in mixing time and intensity and kept the elapsed time between the water addition and injection minimal and as consistent as possible. Triplicate injections (10 μL) provided replicate determinations of the measured peak areas and the opportunity to evaluate adsorption behavior temporally relative to perfluorinated chain length, vial material, and solvent composition.
image file: c9ew00765b-f1.tif
Fig. 1 Methods flowchart for autosampler vial adsorption testing.

The twelve solutions were injected in sequential order that was then repeated thrice. After completion, all of the vials were recapped and equilibrated at rest overnight. On the following day, all vials were injected once more in the same sequential order as the previous day. Each vial was then mixed by vortexing (2 min, 700 RPM) and injected once more, then mixed in a bath sonicator (2 min) and injected once more. All solutions were then sonicated (2 min) and discarded, and each vial was rinsed with 0.5 mL of clean methanol followed by inversion ten times by hand and sonication (2 min). A single injection of each vial rinse was then made to measure any adsorbed analyte recoverable from the vial walls.

SPE optimization

All of the SPE optimization tests utilized a 2 ng spike of PFAA mass-labeled surrogate mixture (Table S2). The SPE cleanup was performed using a WAX SPE column as described by Taniyasu et al.28 and Chu and Letcher31 with some modifications. The extraction solvent was 2% formic acid in acetonitrile. For simplicity, the word “sample” will be used to refer to a solution or vessel containing target analytes. The flow rate was set to approximately 1 drop per second (1–2 mL per minute). In the following description, “drained under vacuum” is used to describe removing the hold-up volume from the SPE cartridge sorbent bed using 2–5 in Hg vacuum until solvent sputters from the tip of the solvent drain needle. “Drained” implies allowing the loaded solvent to drain at 1 drop per second until dripping stops and the hold-up volume is retained.

The SPE cartridges were connected to a glass vacuum manifold and preconditioned with 3 mL of methanol containing 1% ammonium hydroxide (28% ammonium hydroxide solution diluted 1[thin space (1/6-em)]:[thin space (1/6-em)]99 with water) and then 3 mL of methanol, which were each completely drained from the sorbent under vacuum, followed by 3 mL of water, keeping the meniscus above the sorbent. Three different sample loading procedures (discussed in detail below) were used to identify sources of variability in analyte recovery. After sample loading, the cartridge was washed with 1 mL of 2% aqueous formic acid to remove matrix components, and allowed to drain under vacuum. Two 1 mL water wash volumes were then processed through the cartridge followed by complete drying under vacuum (20 in Hg). Drying was terminated based on visual observation of a uniform lightening in the color of the sorbent bed.

Cartridges were then eluted with 1 mL of methanol (drained and discarded) followed by 1 mL of methanol rinse (drained and discarded) and 3 mL of methanol containing 1% ammonium hydroxide that was drained into a 15 mL PP centrifuge tube. This eluate was then evaporated to 0.5 mL in a heated water bath (50 °C) under a gentle stream of ultra-high purity nitrogen and subsequently diluted to 1 mL with methanol. A volume of 500 μL was then transferred to a 700 μL PP autosampler vial followed by addition of 175 μL of water. This volume was mixed by inversion by hand followed by bath sonication for 5 minutes to ensure adequate mixing.

For the first sample loading procedures, a pipettor was used to spike the standard solution (2 ng of each analyte) directly into approximately 0.1 mL of undrained conditioning water. This technique was used to gauge the method's maximum attainable recovery because it was the simplest loading method used and did not involve any secondary sample containers capable of adsorbing any analyte.

Next, 1.5 mL of spiked extraction solvent was diluted to 10 mL with water in a 15 mL PP centrifuge tube, vortexed (2 min, 500 RPM), and loaded to a preconditioned cartridge. After loading was complete, the empty 15 mL tube was rinsed with 0.5 mL of methanol in order to solubilize any recoverable adsorbed analyte. The cartridge eluate was concentrated to 0.5 mL and analyzed in addition to the tube rinse.

Finally, to avoid potential analyte adsorption to the PP tube when diluting samples with water, a third test involved the dilution of the sample with water within the head volume of the cartridge. To accomplish this, 2.6 mL of water was loaded on the head of the conditioned cartridge followed by 0.5 mL of sample. To mix the sample aliquot with the water within the head volume of the cartridge, a pipette tip (containing sample) was inserted into the previously loaded water and approximately half of the aliquot was dispensed from the pipettor as close to the SPE sorbent as possible without overflowing the top of the cartridge. The pipette tip was then withdrawn upward toward the surface and the remainder of the sample aliquot was dispensed.

Results & discussion

Surface adsorption

The standard solutions in the vials containing 25% water yielded the greatest peak areas with lowest variability for all of the individual analytes during the two-day adsorption assessment. Peak areas for each analyte were normalized to the mean peak area from the first injection (per batch) of the solutions containing 25% water to account for differences in response factors among analytes. These results are summarized in Fig. 2.
image file: c9ew00765b-f2.tif
Fig. 2 Mean (n = 3) normalized peak areas vs. PFAA perfluorinated chain length for each vial material and solvent composition tested. For each material type, plotted values for each compound are normalized to the mean peak area for that compound in the first injections from the solutions containing 25% water. Bars represent sequential injections from day one (Inj. 1–3), day two (Inj. 4), day 2 after vortexing for 2 min (vortex), and day 2 after sonicating for 2 min (Sonicate). Embedded letters signify groups likely to be different based on multiple comparison testing (Tukey, α = 0.05) following one-way ANOVA testing (α = 0.05) on individual perfluorinated chain lengths for each material/solvent composition combination (each set of six bars represents one ANOVA/multiple comparison result). Error bars represent ± st. dev. * indicates PFSA.

Analytes with perfluorinated chain lengths less than seven were excluded from Fig. 2 because one-way ANOVA testing suggested that no differences existed among the six sequential injections (see Fig. S1). In general, the peak areas for analytes in 25% water solutions were very similar, regardless of vial material. For 25% water solutions, the small losses in peak areas over time for long chain analytes were mostly recovered after vortexing and/or sonicating the solutions. However, for 75% water solutions, the declines in the long chain analyte peak areas were much more prominent and permanent.

Declining relative peak areas over time for longer chain (perfluorinated chain length ≥10) PFAAs in solutions containing 75% water agreed with the published recovery data summarized in the introduction. Insufficient PFAA solubility may provide an explanation for the observed analyte loss from solution. PFAA solubility values are scarce, stemming from measurement difficulties resulting from their tendency to form micelles at elevated concentrations.32 The critical micelle concentration (CMC), or the concentration at which half of the molecules in solution are associated with micelles, has been suggested as a potential predictor of surfactant hydrophobicity.33 Martin et al.34 used data from Kissa35 to develop a linear regression relationship between PFAA CMCs and perfluorinated chain lengths from 4 to 10. Applying this relationship to a perfluorinated chain length of 13 corresponding to perfluorotetradecanoic acid (PFTeA), a CMC of 5 mg L−1 was estimated. Extending the relationship from Martin et al.34 may not be practical due to evidence suggesting that changes in solvation occur due to changes in molecular geometry beyond a perfluorinated chain length of 9.36 However, this relationship predicts that the PFTeA CMC is orders of magnitude higher than typical ng L−1 levels encountered for PFAAs in most environmental analyses. An additional consideration is the impact of solution purity on PFAA solubility. Ammonium perfluorooctanoate (APFO) solubility (as inferred by CMC) has been shown to decrease in the presence of co-dissolved salts. However, the decrease is only a factor of ∼4 as the salt concentration increases from 0.0 to 0.2 M.37 Thus, micelle formation or insufficient solubility leading to analyte precipitation from solution are unlikely to be significant factors at the concentrations and conditions typically encountered in environmental samples, including the analyses performed here. However, the solubility discussion above pertains to aqueous solutions. PFAS solubility in organic solvents are presumably higher than aqueous solubility, but literature is lacking in this area. Therefore, the interaction between molecules of more hydrophobic PFAS and the container walls is more likely.

Additional evidence for increasing PFAA hydrophobicity with increasing carbon chain length is provided by these analytes' reverse phase LC retention times (Table S4), critical micelle concentrations,35 and carbon normalized soil–water distribution coefficients.27,38 The minimum normalized mean peak areas observed for both vial types occurred at a perfluorinated chain length of 13 (PFTeA) in solutions containing 75% water. They were 0.35 and 0.13 for ABG and PP vials, respectively. Hydrophobicity may also explain the lower relative peak areas for the PFSA with a perfluorinated chain length of 10 (perfluorodecane sulfonate (PFDS)) relative to the equivalent PFCA (PFUnA) for solutions containing 75% water in PP vials. The reverse phase LC retention times (Table S4) of these two analytes suggest PFUnA exhibits greater hydrophobicity. However, the opposite is suggested by the CMC trend described by Martin et al.34 and the soil–water partitioning data.27,38 Involving more complex partitioning processes including protein binding, prior research on fish (rainbow trout, Oncorhynchus mykiss),34,39 plankton (Daphnia magna),40,41 and green mussels (Perna viridis)42 has consistently found PFSAs to bioconcentrate/bioaccumulate to a greater extent than PFCAs with equivalent fluorinated chain lengths.

Several characteristics are unique to the results for 75% water solutions within PP vials (Fig. 2). Generally, peak areas for analytes in these solutions exhibited the greatest variability. The peak areas for analytes with perfluorinated chain lengths between 7 and 11 increased initially during the three successive day one injections, then decreased, and then generally stabilized at near-initial levels on day two. This initial rise in peak area was not observed for short chain analytes (chain lengths <7) dissolved in the identical solvent composition in ABG vials. Additionally, peak areas for analytes with perfluorinated chain lengths exceeding 10 started at lower values and stabilized more quickly in 75% water solutions in PP vials compared to the same composition in ABG vials. It is hypothesized that the linear hydrocarbon structure of PP enhances the material's interaction with the hydrophobic tail of PFAA molecules relative to ABG. It is also hypothesized that the initial rise in short chain peak areas noted for analytes in PP vials may be caused by these compounds diffusing more quickly towards the vial walls where they weakly interact. Subsequently, they are displaced by the more strongly bound longer chain PFAA molecules. Similar competitive interaction of short and long chain PFAA homologs has been observed during bioconcentration in zebrafish (Danio rerio).43

Table 1 provides a concise summary of the results shown in Fig. 2. Table 1 summarizes likely differences in peak areas for individual analytes among the 4 vial material and solvent composition groups as identified by ANOVA and multiple comparison testing (n = 18) whereas Fig. 2 focuses on within-group differences for each tested vial material and solvent composition combination (n = 3). The majority of the differences in Table 1 occurred for long chain analytes between groups with different solvent compositions. For all instances where an analyte with a perfluorinated chain length <10 appears in Table 1, the 75% water solution had larger peak areas, while the opposite was true for analytes with perfluorinated chain length ≥10. Differences for short chain analytes were smaller than the long chain compound differences. No differences were found between ABG and PP vials containing 25% water solutions. However, probable differences emerged when comparing analyte peak areas between ABG and PP vials containing 75% water solutions, suggesting these differences were driven by only the vial material. For each of the three analytes in this category (PFDS, PFDoA, and PFTrA), peak areas were lower for PP vials than ABG vials. Although Fig. 2 shows an apparent difference in peak areas between ABG and PP vials for 75% water solutions for a perfluorinated chain length of 13 (PFTeA), the large variability among repeated injections for this analyte negated identification of a difference during statistical testing.

Table 1 Summary of Tukey multiple comparison test results (α = 0.05) following one1way ANOVA testing (α = 0.05) on normalized peak areas (values displayed in Fig. 2) for all vial material/solvent composition (% water) combinations. Column–row intersections correspond to individual comparisons. Gray-filled intersections highlight differences in solvent composition. Numbers indicate perfluorinated chain length with * indicating PFSAs. Bold red underlined text represents a likely difference after excluding all data for injection 1
image file: c9ew00765b-u1.tif


Based on Fig. 2, analytes with longer chain lengths are more likely to be present in the methanol vial rinsate on the second day of testing. The normalized mean PFTeA peak areas of the final injection from day two and the methanol rinsate from day two for all vial materials and solvent compositions are plotted in Fig. 3 supporting the assessment that peak area variability is due to both solvent composition and vial type/material. For the final injection, the mean PFTeA peak area for the PP vial with 75% water is less than half of the value for the ABG vial with an identical solvent composition. For both vial materials, total mean PFTeA peak area (final injection + rinse) for 75% water solutions was less than half of the corresponding value for 25% water solutions. Since all the test vials were spiked at the same time from the same stock solution, mixed well prior to the first and final two injections, and the analyte peak areas for the injections made at each time point had low variability (Fig. 2), the decrease in peak areas are likely due to mass loss to the vial walls rather than variability in spiked mass or solution heterogeneity. A subset of vials were rinsed with methanol a second time, but only negligible masses were recovered (data not shown), suggesting irreversible adsorption.


image file: c9ew00765b-f3.tif
Fig. 3 Normalized mean (n = 3) PFTeA peak areas for both the final injection after sonication on day two and the methanol rinse for each vial material and solvent composition. For each vial type, plotted values are normalized to the mean of the first injections (day one) from the solutions containing 25% water. Error bars represent std. dev. Negative and positive error bars correspond to final injection and methanol rinse, respectively.

The normalized mean PFTeA peak areas were 0.02 or less in the rinsate for both vial materials containing 25% water solutions, while the corresponding value for PP vials containing 75% water solutions (0.19) was larger than those found in the prior 4 injections from those vials (Fig. 3, 0.13–0.18). While vial contents was not completely removed upon decanting, only ≤2% of any analyte was retained in the vials with the 25% water solution. This result suggests that negligible adsorption occurred for all analytes dissolved in 25% water solutions. A solution containing 0% water was also tested (data not shown), which produced results nearly identical to 25% water, so no further testing of low water fractions was performed. Similarly, a solution containing 50% water was tested producing results very similar to that of 75% water. Therefore, 50% water composition was eliminated from further testing. These results suggest adsorption for long chain PFAAs decreases rapidly to negligible levels in the transition in solvent composition between 50% and 25% water.

Adsorption losses could raise method detection limits and introduce quantification error for those target analytes lacking dedicated mass-labeled analogs in the recovery standard. This effect is described in Fig. 4. The PFTeA peak areas are presented for each vial material and solvent composition combination relative to the PFDoA peak areas in the same test vial. PFDoA was chosen as the reference compound because it is the longest chain PFAA available in the commercial mass-labeled mixture purchased from Wellington Laboratories, Inc. (MPFAC-MXA). The mean peak area ratios for the solutions containing 25% water are approximately 1 with small variability for both vial materials, while the solutions containing 75% water yielded much more variable mean ratios that were larger than 1 due to greater adsorption of the more hydrophobic PFTeA. The elevated error in the peak area ratios for 75% water solutions, especially for PP vials, compared to those containing 25% water is attributed to the progressive adsorption of PFTeA. This result is illustrated in Fig. 2 by the decreasing peak areas during the six-injection sequence for PFTeA as compared to the relatively stable PFDoA values.


image file: c9ew00765b-f4.tif
Fig. 4 Box and whisker plot presenting PFDoA/PFTeA peak area ratios obtained for each vial material and solvent composition tested. Plotted data represent all injections prior to decanting and rinsing with methanol on day two (n = 18). Whiskers extend to 5th and 95th percentiles. Embedded letters signify groups likely to be different based on multiple comparison testing (Tukey, α = 0.05) following one-way ANOVA testing (α = 0.05).

This analyte adsorption from solution results from the modification of the sample solvent composition (typically by dilution with water) resulting in an increased volume fraction of water. Such modifications most commonly occur prior to SPE cleanup of environmental sample extracts. The cleanup typically requires dilution of the extraction solvent to ≥80% water for adequate analyte retention on the SPE sorbent.28,29,31 Another applicable circumstance is when the sample solvent is diluted to ≥50% water content prior to instrumental analysis.30,44,45 The adsorption tests involved only a single sample-containing vessel prior to SPE while sample preparation procedures commonly employ several containers. The use of multiple containers per sample could compound the analyte losses during sample preparation. This recovery loss could be compensated for with the use of matrix matched calibration standards. However, this approach is laborious, and would still likely lead to some inconsistencies based on the results shown in Fig. 2 and 4.

SPE optimization

The results from the SPE method modification tests are summarized in Fig. 5. Dilution of sample with water prior to loading yielded low recoveries (≤51%) for analytes with perfluorinated chains greater than 10 carbons. Fig. 5 shows that significant mass was recovered in the tube rinse for longer chain analytes. As with the autosampler vial adsorption testing, some of this loss appears unrecoverable with methanol rinsing. Long chain analyte mean recoveries increased to at least 88% when the sample was diluted within the SPE cartridge except for the longest chain analyte (perfluorinated chain length of 15, 64%). This result was deemed acceptable given that an isotope dilution method is used for this analyte.
image file: c9ew00765b-f5.tif
Fig. 5 Mean (n = 3) percent recovery of 13C mass-labeled compounds obtained for each phase of WAX SPE cleanup method optimization. Bars correspond to the eluate and subsequent tube rinse for samples diluted within 15 mL PP tubes. Orange- and white-filled diamonds represent recoveries for within-cartridge dilutions and direct cartridge spikes, respectively. * denotes PFSA. Error bars represent ± st. dev.

Fig. 2 and 5 showed consistent trends in recovery with respect to perfluorinated chain length. In general, recovery decreased for analytes containing perfluorinated chains exceeding 8 to 9 carbons. Fig. 5 shows cartridge spike recoveries were essentially constant for all perfluorinated chain lengths. However, the other two sample loading procedures involving additional sample containers and larger sample-surface contact areas (15 mL tube and SPE cartridge walls) produced lower analyte recoveries. Comparison of the “eluate” and “dilute on SPE” series plotted in Fig. 5 as well as the results in Fig. 2 suggest that long chain PFAA recoveries decrease as the contact time and/or area between a container's surface and a sample consisting of high water content is increased. PFAA structural properties likely explain these observed tendencies.

Ellis et al.36 utilized fluorine-19 nuclear magnetic resonance (19F NMR) spectroscopy to probe the solution-phase conformational properties of C2–C14 PFCAs and their respective hydrogenated analogs. Their objective was to elucidate correlations between molecular structure and physical properties and thus, the environmental behavior of PFCAs. Patterns in both chemical shift and relaxation times for the fluorine nuclei at multiple, fixed locations within the perfluorinated chain supported the occurrence of a distinct change in molecular geometry, and therefore solvation of PFCAs with a perfluorinated chain of at least 9 carbons. The spin–spin relaxation time patterns for the terminal CF3 group were noticeably different among PFAAs with perfluorinated chain lengths of 1–5, 7–9, and 10–13, suggesting unique structural differences among these respective groups. Additionally, notable decreases in chemical shift and spin–lattice relaxation times associated with the terminal CF3 group for perfluorinated chain lengths of 10 and 11, respectively further support structural differentiation between short and long chain PFAAs. Generally, other trends in chemical shift and relaxation times associated with chain length were similar between fluorinated and hydrogenated analogs, and were attributed to changes in molecular size. These findings agree with the observations that the solid state structures of perfluorinated alkyl chains adopt a helical character above (CF2)8, with increasingly dominant helical nature to (CF2)12.46

Final method performance

Using insights from the above experiments, a final procedure for extraction, cleanup, and analysis of biological tissues was drafted (supplied in the ESI as SOP 10.4). Dilution of the sample extracts with water in the SPE cartridge during cleanup and decreasing the final water content of analyzed PFAS solutions from 66% to 25% during the UPLC-MS analysis increased surrogate recovery and improved precision. The addition of a two-minute soaking step during SPE cartridge elution was also found to increase surrogate recovery reproducibility. This optimized procedure has been applied to over 300 samples including whole organism homogenates of over 17 fish and macroinvertebrate species analyzed over a period spanning more than two years. These samples have been acquired as part of the United States Environmental Protection Agency's Great Lakes Fish Monitoring and Surveillance Program (GLFMSP) that has been studying bioaccumulative toxic substances in the Laurentian Great Lakes since the 1970s (https://www.epa.gov/great-lakes-monitoring/great-lakes-fish-monitoring-and-surveillance). The details regarding surrogate recoveries and precision based on duplicate sample extractions and replicate extractions of NIST SRM 1947 are summarized in Tables S5, S7, and S8. Mean recoveries of surrogates spiked to samples ranged from 72% to 96% (overall mean of 86%) with %RSDs ranging from 8% to 26% (overall mean of 15%). Likewise, mean recoveries of surrogates spiked to method blanks ranged from 60% to 94% (overall mean of 86%) with %RSDs ranging from 7% to 22% (overall mean of 11%). Mean % differences for individual analyte concentrations quantified in duplicate sample extractions ranged from 5% to 10.5% with minimum and maximum values of the 95% confidence intervals of 2.5% and 14.6%, respectively. Application of this procedure to NIST SRM 1947 yielded consistently high analyte concentrations relative to the reference and information values supplied in the certificate of analysis. Percent recoveries (quantified concentration normalized to corresponding certificate of analysis concentration) and %RSDs for the five certified PFAS were: PFNA (143%, 10%), PFOS (139%, 10%), perfluorodecanoic acid (PFDA, 116%, 11%), PFUnA (140%, 9%), and perfluorotridecanoic acid (PFTrA, 161%, 15%). While these analyte concentrations exceed those specified in the certificate of analysis for SRM 1947, the interlaboratory data lacked adequate precision to develop reference values and uncertainty thresholds for PFNA, PFDA, PFUnA, and PFTrA that were only assigned information values.19 Additionally, a true assessment of accuracy cannot be made without prior knowledge of an analyte's concentration in a given sample. So, it is possible that the values in the analysis certificate under predict the true analyte concentration due to incomplete extraction as previously identified by the ILSs.15,16

Conclusions

Systematic experiments helped to elucidate that both solvent composition and container material affect dissolved PFAA adsorption to container walls, and can negatively affect multiple stages of sample preparation and PFAS analysis. Consistent, sufficient mixing of samples, application of isotope dilution methods whenever possible, and minimizing contact time and surface area between sample container walls and PFAA solutions containing ≥25% water can help minimize analyte loss. In scenarios involving dilution of PFAS solutions with water where water content will exceed 25% relative to organic solvent, use of borosilicate glass as opposed to polypropylene containers can increase long chain PFAA recoveries. This recommendation is especially true if rinsing the sample container walls to recover adsorbed analyte is impractical (e.g. transferring a sample aliquot for subsequent processing). The application of an optimized analytical procedure applying knowledge of PFAS behavior (i.e. adsorptive tendencies) can provide precise methods for PFAS analysis. Because of differences in analytical methods such as the increase of commercially available mass-labeled surrogates over time, caution is recommended when directly comparing quantitative PFAS data, especially when comparing pre- to post-2009 studies. The work presented here was not all encompassing in terms of PFAS methodology. Therefore, additional studies of this kind as well as transparent communication of analytical difficulties are recommended and encouraged in order to further develop robust standardized analytical methods for PFAS in multiple matrices in which these compounds are present.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

Funding for this work was provided by the Great Lakes National Program Office under the United States Environmental Protection Agency, Grant No. GL96594201 and 00E00454. This work is solely attributed to the authors and does not necessarily represent the views of the United States Environmental Protection Agency. We also thank all staff and students working at the Center for Air Resources Engineering & Science laboratory who helped in the progression of this work in any way.

Notes and references

  1. R. C. Buck, J. Franklin, U. Berger, J. M. Conder, I. T. Cousins, P. de Voogt, A. A. Jensen, K. Kannan, S. A. Mabury and S. P. van Leeuwen, Integr. Environ. Assess. Manage., 2011, 7, 513–541 CrossRef CAS PubMed.
  2. A. G. Paul, K. C. Jones and A. J. Sweetman, Environ. Sci. Technol., 2009, 43, 386–392 CrossRef CAS PubMed.
  3. K. Prevedouros, I. T. Cousins, R. C. Buck and S. H. Korzeniowski, Environ. Sci. Technol., 2006, 40, 32–44 CrossRef CAS PubMed.
  4. J. P. Giesy and K. Kannan, Environ. Sci. Technol., 2001, 35, 1339–1342 CrossRef CAS PubMed.
  5. M. Filipovic, A. Woldegiorgis, K. Norström, M. Bibi, M. Lindberg and A.-H. Österås, Chemosphere, 2015, 129, 39–45 CrossRef CAS PubMed.
  6. U.S. Environmental Protection Agency Office of Water (4304T) Health and Ecological Criteria Division, Drinking water health advisory for perfluorooctane sulfonate (PFOS) (EPA Document Number: 822-R-16-004), Washington, DC, 2016.
  7. U.S. Environmental Protection Agency Office of Water (4304T) Health and Ecological Criteria Division, Drinking water health advisory for perfluorooctanoic acid (PFOA) (EPA Document Number: 822-R-16-005), Washington, DC, 2016.
  8. W. A. Weppner, Re: Phase-Out Plan for POSF-based Products. USEPA Administrative Record AR226-0600. Available from: www.regulations.gov, as document EPA-HQ-OPPT-2002-0051-0006, 2000.
  9. United States Environmental Protection Agency, Document number 20460. 2010/15 PFOA Stewardship Program Commitments. Available from: https://www.epa.gov/assessing-and-managing-chemicals-under-tsca/letter-inviting-participation-pfoa-stewardship-program, Washington, DC, 2006.
  10. Z. Wang, I. T. Cousins, M. Scheringer, R. C. Buck and K. Hungerbühler, Environ. Int., 2014, 70, 62–75 CrossRef CAS PubMed.
  11. Z. Wang, J. M. Boucher, M. Scheringer, I. T. Cousins and K. Hungerbühler, Environ. Sci. Technol., 2017, 51, 4482–4493 CrossRef CAS PubMed.
  12. ASTM D7968-17a, Standard Test Method for Determination of Polyfluorinated Compounds in Soil by Liquid Chromatography Tandem Mass Spectrometry (LC/MS/MS), ASTM International, West Conshohocken, PA, 2017.
  13. S. B. Gewurtz, A. O. De Silva, S. M. Backus, D. J. McGoldrick, M. J. Keir, J. Small, L. Melymuk and D. C. G. Muir, Environ. Sci. Technol., 2012, 46, 5842–5850 CrossRef CAS PubMed.
  14. T. A. Bruton and A. Blum, Environ. Health, 2017, 16, 120 CrossRef PubMed.
  15. S. P. van Leeuwen, A. Kärrman, B. van Bavel, J. de Boer and G. Lindström, Environ. Sci. Technol., 2006, 40, 7854–7860 CrossRef CAS PubMed.
  16. S. P. J. van Leeuwen, C. P. Swart, I. van der Veen and J. de Boer, J. Chromatogr. A, 2009, 1216, 401–409 CrossRef CAS PubMed.
  17. G. Lindström, A. Kärrman and B. van Bavel, J. Chromatogr. A, 2009, 1216, 394–400 CrossRef PubMed.
  18. J. M. Keller, A. M. Calafat, K. Kato, M. E. Ellefson, W. K. Reagen, M. Strynar, S. O'Connell, C. M. Butt, S. A. Mabury, J. Small, D. C. G. Muir, S. D. Leigh and M. M. Schantz, Anal. Bioanal. Chem., 2010, 397, 439–451 CrossRef CAS PubMed.
  19. J. L. Reiner, S. G. O'Connell, C. M. Butt, S. A. Mabury, J. M. Small, A. O. De Silva, D. C. G. Muir, A. D. Delinsky, M. J. Strynar, A. B. Lindstrom, W. K. Reagen, M. Malinsky, S. Schäfer, C. J. A. F. Kwadijk, M. M. Schantz and J. M. Keller, Anal. Bioanal. Chem., 2012, 404, 2683–2692 CrossRef CAS PubMed.
  20. J. L. Reiner, A. C. Blaine, C. P. Higgins, C. Huset, T. M. Jenkins, C. J. A. F. Kwadijk, C. C. Lange, D. C. G. Muir, W. K. Reagen, C. Rich, J. M. Small, M. J. Strynar, J. W. Washington, H. Yoo and J. M. Keller, Anal. Bioanal. Chem., 2015, 407, 2975–2983 CrossRef CAS PubMed.
  21. J. W. Martin, K. Kannan, U. Berger, P. De Voogt, J. Field, J. Franklin, J. P. Giesy, T. Harner, D. C. G. Muir, B. Scott, M. Kaiser, U. Järnberg, K. C. Jones, S. A. Mabury, M. Simcik, C. Sottani, B. Van Bavel, A. Kärrman, G. Lindström and S. Van Leeuwen, Environ. Sci. Technol., 2004, 38, 249A–255A Search PubMed.
  22. M. K. So, S. Taniyasu, N. Yamashita, J. P. Giesy, J. Zheng, Z. Fang, S. H. Im and P. K. S. Lam, Environ. Sci. Technol., 2004, 38, 4056–4063 CrossRef CAS PubMed.
  23. N. Yamashita, K. Kannan, S. Taniyasu, Y. Horii, T. Okazawa, G. Petrick and T. Gamo, Environ. Sci. Technol., 2004, 38, 5522–5528 CrossRef CAS PubMed.
  24. J. A. Shoemaker, B. Boutin and P. Grimmett, J. Chromatogr. Sci., 2009, 47, 3–11 CAS.
  25. L. Ahrens, K. Vorkamp, P. Lepom, N. Bersuder, N. Theobalk, R. Ebinghaus, R. Bossi, J. L. Barber and E. McGovern, ICES Tech. Mar. Environ. Sci., 2010, 1–16 Search PubMed.
  26. M. K. So, S. Taniyasu, P. K. S. Lam, G. J. Zheng, J. P. Giesy and N. Yamashita, Arch. Environ. Contam. Toxicol., 2006, 50, 240–248 CrossRef CAS PubMed.
  27. J. G. Sepulvado, A. C. Blaine, L. S. Hundal and C. P. Higgins, Environ. Sci. Technol., 2011, 45, 8106–8112 CrossRef CAS PubMed.
  28. S. Taniyasu, K. Kannan, M. K. So, A. Gulkowska, E. Sinclair, T. Okazawa and N. Yamashita, J. Chromatogr. A, 2005, 1093, 89–97 CrossRef CAS PubMed.
  29. B. C. Kelly, M. G. Ikonomou, J. D. Blair, B. Surridge, D. Hoover, R. Grace and F. A. P. C. Gobas, Environ. Sci. Technol., 2009, 43, 4037–4043 CrossRef CAS PubMed.
  30. A. O. De Silva, C. Spencer, B. F. Scott, S. Backus and D. C. G. Muir, Environ. Sci. Technol., 2011, 45, 8060–8066 CrossRef CAS PubMed.
  31. S. Chu and R. J. Letcher, J. Chromatogr. A, 2008, 1215, 92–99 CrossRef CAS PubMed.
  32. E. A. Simister, E. M. Lee, J. Ren Lu, R. K. Thomas, R. H. Ottewill, A. R. Rennie and J. Penfold, J. Chem. Soc., Faraday Trans., 1992, 88, 3033–3041 RSC.
  33. J. Tolls and D. T. H. M. Sijm, Environ. Toxicol. Chem., 1995, 14, 1675–1685 CrossRef CAS.
  34. J. W. Martin, S. A. Mabury, K. R. Solomon and D. C. G. Muir, Environ. Toxicol. Chem., 2003, 22, 196–204 CrossRef CAS PubMed.
  35. E. Kissa, Fluorinated surfactants: Synthesis properties applications, Marcel Dekker Inc., New York, 1994 Search PubMed.
  36. D. A. Ellis, K. A. Denkenberger, T. E. Burrow and S. A. Mabury, J. Phys. Chem. A, 2004, 108, 10099–10106 CrossRef CAS.
  37. N. Downes, G. A. Ottewill and R. H. Ottewill, Colloids Surf., A, 1995, 102, 203–211 CrossRef CAS.
  38. C. P. Higgins and R. G. Luthy, Environ. Sci. Technol., 2006, 40, 7251–7256 CrossRef CAS PubMed.
  39. J. W. Martin, S. A. Mabury, K. R. Solomon and D. C. G. Muir, Environ. Toxicol. Chem., 2003, 22, 189–195 CrossRef CAS.
  40. Z. Dai, X. Xia, J. Guo and X. Jiang, Chemosphere, 2013, 90, 1589–1596 CrossRef CAS PubMed.
  41. X. Xia, A. H. Rabearisoa, X. Jiang and Z. Dai, Environ. Sci. Technol., 2013, 47, 10955–10963 CrossRef CAS PubMed.
  42. C. Liu, K. Y. H. Gin, V. W. C. Chang, B. P. L. Goh and M. Reinhard, Environ. Sci. Technol., 2011, 45, 9758–9764 CrossRef CAS PubMed.
  43. W. Wen, X. Xia, D. Hu, D. Zhou, H. Wang, Y. Zhai and H. Lin, Environ. Sci. Technol., 2017, 51, 12358–12368 CrossRef CAS PubMed.
  44. V. I. Furdui, N. L. Stock, D. A. Ellis, C. M. Butt, D. M. Whittle, P. W. Crozier, E. J. Reiner, D. C. G. Muir and S. A. Mabury, Environ. Sci. Technol., 2007, 41, 1554–1559 CrossRef CAS PubMed.
  45. B. F. Scott, C. Spencer and E. Lopez, Water Qual. Res. J. Can., 2009, 44, 263–277 CrossRef CAS.
  46. J. Wang and C. K. Ober, Liq. Cryst., 1999, 26, 637–648 CrossRef CAS.

Footnote

Electronic supplementary information (ESI) available. See DOI: 10.1039/c9ew00765b

This journal is © The Royal Society of Chemistry 2019