Environ. Sci. Technol. 2009, 43, 7902–7908
Branched Perfluorooctane Sulfonate Isomer Quantification and Characterization in Blood Serum Samples by HPLC/ESI-MS(/MS) N I C O L E R I D D E L L , * ,† GILLES ARSENAULT,† JONATHAN P. BENSKIN,‡ BROCK CHITTIM,† JONATHAN W. MARTIN,‡ ALAN MCALEES,† AND ROBERT MCCRINDLE§ Wellington Laboratories Inc., Guelph, Ontario, N1G 3M5, Canada, Department of Laboratory Medicine and Pathology, University of Alberta, Edmonton, Alberta, T6G 2G3, Canada, and Department of Chemistry, University of Guelph, Guelph, Ontario, N1G 2W1, Canada
Received April 28, 2009. Revised manuscript received August 7, 2009. Accepted August 31, 2009.
Perfluorooctane sulfonate (PFOS) is a global contaminant and is currently among the most prominent contaminants in human blood and wildlife samples. Although “total PFOS” (∑PFOS) analytical methods continue to be the most commonly used for quantification, recent analytical method developments have made it possible to resolve the various isomers of PFOS by HPLC-MS/MS. Characterized technical PFOS standards (i.e., containing a mixture of PFOS isomers) are now available that enable isomer specific quantification of PFOS, however the advantages of such an analysis have not yet been examined systematically. Herein, PFOS isomers have been individually quantified for the first time in real samples and the results are compared to a traditional ∑PFOS method; the influence of analytical standards and isomer specific electrospray and MS/ MS behavior were also investigated. The two human serum standard reference materials chosen for analysis contained dramatically different PFOS isomer profiles (∼30-50% total branched isomers) emphasizing that isomer patterns should not be ignored and may provide useful information on exposure sources (i.e., direct exposure to PFOS vs indirect exposure from PFOS-precursors). Depending on the sample and the particular MS/MS transition chosen for ∑PFOS analysis (i.e., 499f80 or 499f99), ∑PFOS concentrations may be over- or underestimated compared to the isomer specific analysis. Differences in the extent of in-source fragmentation and MS/MS dissociation contributed to the systematic analytical bias. It was also shown that ∑PFOS data are prone to interlaboratory variation due to various choices of PFOS standards and instrumental conditions used. In the future, for either ∑PFOS or isomer specific PFOS analyses, we suggest
* Corresponding author phone: +1 519 822 2436; fax: +1 519 822 2849; e-mail:
[email protected]. † Wellington Laboratories Inc. ‡ University of Alberta. § University of Guelph. 7902
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 20, 2009
that accuracy can be maximized and interlaboratory discrepancies minimized by using a common chemically pure technical PFOS standard characterized by 19F NMR.
Introduction Interest in the widespread distribution of perfluoroalkyl chemicals, including perfluorooctanesulfonate (PFOS, C8F17SO3-), in the environment (1-6) and their accumulation in humans has been steadily increasing (7-15). Quantification of PFOS in human plasma, blood serum, and whole blood is not trivial. A recent study addressed the biases associated with matrix effects in LC/MS-MS quantification methods (15). However, the difficulties associated with analyzing for PFOS in various matrices is compounded by the fact that its production from linear alkyl precursors using electrochemical fluorination is not a clean process but, instead, gives complex mixtures. For example, the commercial PFOS formerly produced by 3 M was a mixture of ca. 70% linear and ca. 30% branched isomers as measured by 19F NMR spectroscopy (16-18). Interest in the presence of the branched isomers of PFOS in the environment, particularly in biological samples, has also been growing due to issues relating to the quantification of total PFOS (19, 20) as well as questions about exposure sources and differences in bioaccumulation potential (19) and toxicity (21) among the various isomers. Unfortunately, although qualitative standards for many of the PFOS branched isomers are commercially available, quantitative standards are not. The generation of accurate quantitative data for PFOS is important for addressing environmental fate and toxicity questions, and ultimately for enabling accurate risk assessments by chemical regulators. Deviations from design values in round-robin interlaboratory studies demonstrate some of the difficulties associated with the analysis of PFOS in biological samples (22). The most common analytical method used for the analysis of PFOS is HPLC-MS/MS (23, 24). Despite the fact that multiple PFOS isomers are thought to be present in any environmental sample, traditional HPLC methods often avoid their separation so that the entire broadened peak, or unresolved group of peaks, may be easily integrated. This will be referred to henceforth as a “∑PFOS” method. It has already been acknowledged that ∑PFOS methods can lead to a systematic quantification bias of unknown proportions, since the various branched isomers are expected to have different ionization and fragmentation efficiencies during electrospray ionization (23, 25) and distinct collision induced dissociation (CID) patterns in MS/MS (25, 26). However, the extent of this bias has not previously been quantified, and quantitative isomer specific PFOS analysis has not yet been demonstrated as an alternative for any sample. In the current work, isomer specific quantification of PFOS is demonstrated in human serum for the first time and the cumulative impact of isomer specific response factors and various standards is discussed by comparison with a ∑PFOS quantification method.
Materials and Methods Chemicals, Standards, and Serum Samples. The major branched isomers of PFOS were previously synthesized, isolated, and characterized as sulfonamide derivatives (17), and these were converted to the respective sulfonates (Figure 1) using proprietary methods at Wellington Laboratories (Guelph, ON) and are available as qualitative standards. HPLC grade methanol (MeOH) and water were purchased from 10.1021/es901261v CCC: $40.75
2009 American Chemical Society
Published on Web 09/17/2009
FIGURE 1. Structures of the prevalent PFOS isomers in technical mixtures and in samples. Caledon (Georgetown, ON). Reference standards of linear PFOS [13C4-MPFOS (MPFOS) and LPFOS] and a mixture of linear and branched isomers of PFOS (brPFOSK) were obtained from Wellington Laboratories (Guelph, ON). A sample of technical PFOS was purchased from TCI America (Portland, OR). Freeze-dried human serum samples A and B representing 10.0 and 10.7 mL, respectively, of reconstituted human serum were acquired through participation in the 2006 Worldwide PFC intercalibration study (22). Human Serum Sample A (NIST SRM 1589a) was a freeze-dried standard reference material (SRM) collected by NIST consisting of 50 pooled blood samples of 5 mL from donors who consumed fish caught around the Great Lakes. Human Serum Sample B (NIST SRM 1957) was a freeze-dried SRM also collected by NIST. This sample was prepared from a serum pool of 200 L across the United States. Detailed information on both SRMs and certificates of analysis can be found at www.nist.gov. LC/ESI-MS and LC/ESI-MS/MS. Separation, identification, and determination of the response factors of the PFOS isomers were completed using HPLC/ESI-MS and HPLC/ ESI-MS/MS experiments on a Waters Acquity Ultra Performance LC (Waters, Milford, MA) interfaced to a Micromass Quattro micro atmospheric pressure ionization (API) mass spectrometer (Micromass, Manchester, UK). Chromatography was performed on an Acquity UPLC BEH Shield RP18 column (1.7 µm, 2.1 × 100 mm) with gradient elution. The gradient started at 46% 80:20 MeOH/ACN and 54% water (both containing 10 mM ammonium acetate) at a flow rate of 350 µL/min. The program was ramped to 49% 80:20 MeOH/ ACN by 6 min, held for 17 min, then ramped again to 90% 80:20 MeOH/ACN over 0.50 min, held for 1 min, and returned to initial conditions. The mass spectrometer was set up in negative-ion electrospray mode with the following conditions: capillary voltage 2.00 kV, source temperature 110 °C, cone gas 60 L/h, desolvation gas flow 750 L/Hr, desolvation gas temperature 350 °C, cone voltage 60 V, collision energy 40 eV, and gas cell pressure ∼3.5e-3 mbar. The MS tune parameters for the individual branched isomers were optimized, however essentially no difference in their response factors was observed. Reduction in the cone voltage did not reduce the in-source fragmentation observed for isomers 2, 9, 10, and 11. Optimization of MS/MS tune parameters for isomer specific transitions using the individual isomers also did not result in a significant increase in response for any of the isomers (see Supporting Information (SI) for further details and Table S1). Serum samples A and B were extracted in triplicate and analyzed using an isomer specific analytical method (19). Briefly, 1 mL of serum was diluted with 0.1 M formic acid containing MPFOS as the internal standard and sonicated for 20 min. A 200 mg Oasis HLB cartridge (Waters) was conditioned with HPLC grade methanol (6 mL) and 0.1 M formic acid (6 mL). The samples were loaded onto the cartridge and slowly passed through under vacuum. The cartridges were rinsed with 15 mL of 0.1 M formic acid, 6 mL of 50% 0.1 M formic acid/50% methanol, and 1 mL of 1%
NH4OH in water. The cartridge was then purged with air and the analytes were eluted with 6 mL of methanol (1% NH4OH) into 15-mL centrifuge tubes. The extracts were evaporated under N2 to 0.5 mL and transferred to methanol rinsed polypropylene microvials (Fisher), with polyethylene caps (Supelco). Analysis by HPLC-MS/MS involved 10-µL injections of extract onto a FluoroSep RP Octyl column (3 µ 100 Å, 15 cm × 2.1 mm, ES Industries, West Berlin, NJ). Gradient elution conditions were 200 µL/min, and starting conditions were 40% A (water adjusted to pH 4.0 with ammonium formate) and 60% B (100% methanol). Initial conditions were held for 0.3 min, ramped to 64% B by 1.9 min, increased to 66% B by 5.9 min, 70% B by 7.9 min, 78% B by 40 min, 88% B by 42 min, 100% B by 55 min, returning to initial conditions by 60 min, and allowing 30 min for equilibration. Mass spectral data were collected using a hybrid triple-quadrupole linear ion trap mass spectrometer (4000QTRAP, MDS Sciex, Concord, ON) equipped with an electrospray interface operating in the negative ion mode. The MS/MS tune parameters utilized for quantification were optimized for isomer specific transitions using the brPFOSK standard (Table S1). Chromatograms were recorded by selected reaction monitoring in MS/MS mode.
Results and Discussion Comparison of PFOS Quantification by Total and Isomer Specific Methods. Quantification biases were investigated in serum samples by first generating chromatograms using previously developed isomer specific methods (19). These were subsequently analyzed using three distinct standard curve quantification techniques: (i) ∑PFOS quantification using m/z 499f99 and m/z 499f80 transitions and summing the response of all isomers relative to the technical brPFOSK standard, (ii) quantification of linear and total branched isomers separately using m/z 499f99 and m/z 499f80 transitions relative to a LPFOS and brPFOSK standard, respectively, and (iii) the linear and each of the major perfluoromonomethyl PFOS isomers were quantified individually using separate calibration curves generated from the known isomer composition of brPFOSK (determined by 19 F NMR) and isomer specific transitions (Table S1 and Figure S1). For each quantification technique, six-point calibration curves (each curve the average of standards run before and after serum samples) were employed using concentrations ranging from 0.5 to 50 ng/mL total PFOS. All of these methods were applied to two human serum samples that many laboratories have in their possession following the second Worldwide Interlaboratory Study on PFCs. An interesting result was the striking difference in the branched isomer profiles of human serum samples A and B (Figure 2). The HPLC-MS/MS chromatograms of these samples qualitatively showed that each had a substantially different ratio of linear:branched isomers, and the relative profile of the branched isomers also varied. PFOS exposure sources are not well characterized, thus the current state of knowledge provides no explanation for these different isomer profiles. Although limited to only two samples, these findings illustrate that PFOS exposure sources can vary and, irrespecVOL. 43, NO. 20, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
7903
FIGURE 2. HPLC-MS/MS chromatograms produced using the Benskin et al. method (19) showing the PFOS isomer profiles in (a) serum A, (b) serum B, and (c) the brPFOSK standard on the FluoroSep RP Octyl column tracking m/z 499f80 (red), m/z 499f99 (green), and m/z 499f130 (blue). tive of improvements to accuracy, isomer specific analytical methods could be a useful tool for exposure tracking in biomonitoring studies. Nevertheless, the chromatograms of neither human serum sample looked exactly like the brPFOSK standard, thus some systematic analytical bias should be anticipated from traditional ∑PFOS methods. Quantification of ∑PFOS in the serum samples relative to brPFOSK, using either the m/z 499f80 or m/z 499f99 transition, produced lower concentrations than the reference values generated in the associated interlaboratory study (22) (Table S2). It is notable that a pure linear standard (LPFOS) was provided for this interlaboratory study, however, its use was not made mandatory. Differences were also evident between ∑PFOS quantification with m/z 80 and m/z 99, and this depended on the sample. Specifically, the m/z 99 product ion produced a higher ∑PFOS concentration than m/z 80 for serum A (two-tailed student’s t test, p ) 0.006), whereas in serum B the m/z 99 product ion produced a lower concentration relative to m/z 80 (p ) 0.003, Table 1). It is important to note that the m/z 99 product ion is the most commonly used PFOS quantification ion due to endogenous interferences associated with the m/z 80 product ion (19), however these interferences are chromatographically separated by the perfluorooctyl stationary phase used here and thus do not contribute to the reported variation. The linear PFOS isomer was then quantified using both m/z 80 and m/z 99 product ions, and LPFOS and brPFOSK standards. The calibration curve for linear PFOS in the brPFOSK standard was produced by simply adjusting the calibrated mass of linear PFOS in the mixture to 78.8% of the total mass based on 19F NMR spectroscopy data (supplied with the standard). Regardless of the product ion (m/z 80 or m/z 99) chosen, or the standard used, the resulting concentrations of linear PFOS were indistinguishable in both serum A and B (Table S3). This result was anticipated, and it furthermore validates the 19F NMR spectral data provided 7904
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 20, 2009
with brPFOSK. For simultaneous quantification of all of the branched isomers (hereafter referred to as “total branched isomers”) using brPFOSK, a calibration curve was constructed by adjusting the concentration of the branched PFOS isomers in each calibration standard to 21.2% of the total concentration based on 19F NMR spectroscopy data (supplied with the standard). Quantification was achieved using both m/z 499f80 and 499f99 transitions (Table 1). Good agreement was observed for the data obtained using the m/z 80 and m/z 99 product ions for quantification of total branched isomers (determined to not be significantly different using a two-tailed student’s t test, p ) 0.32). Finally, individual perfluoromonomethyl isomers were quantified by building calibration curves for the individual isomers using isomer specific transitions (Figure S1 and Table 1) after adjusting for their composition in the brPFOSK standard. This can only be done for the PFOS isomers that have been characterized by 19F-NMR. Utilizing the brPFOSK standard, we were able to detect and quantify five perfluoromonomethyl isomers. The rank order of individual isomers differed between the two samples, but the general rank order was isomer 7 > 5 > 6∼4 > 2 (Table 1). The sum concentration of individual branched PFOS isomers is not a true total, because some minor isomers are not accounted for, which explains why the sum of all individual isomers was somewhat less than that estimated by the “total branched PFOS” analysis (Table 1). However, since the levels of perfluorodimethyl isomers observed in the serum samples were lower than those in the brPFOSK standard, any effect on the accuracy of the data resulting from the exclusion of the dimethyl branched isomers during isomer specific quantification is expected to be minimal; albeit this cannot necessarily be assumed for other human matrices or other organisms. It is also possible that quantification of total branched isomers by summing all branched peaks in the 499f80 transition may cause overreporting because all but 3 isomers have relative response factors greater than 100% in this transition (Table 2). Therefore, if laboratories undertake an isomer-specific analysis, there may still be a benefit to reporting results from an accompanying “total branched” analysis. Relative Response Factors of Branched PFOS Isomers. In an effort to understand the above analytical biases, the relative response factors of all PFOS isomers under electrospray and MS/MS conditions were investigated using a more rapid UPLC-MS(/MS) method. The preparation and purification of 10 different branched PFOS isomers (Figure 1) permitted the relative response factor of each to be studied, the determination of their elution order on a UPLC BEH Shield RP18 column (Figure 3), and hence the assignment of the signals in various other technical PFOS mixtures. The elution order differed substantially from that previously observed using a perfluorophenyl (25) or the perfluorooctyl (19) stationary phase used above. For example, on the UPLC Shield RP18 column the elution order was (in order of increasing retention time) 9 < 10 < 8 < 11 < 6 < 5 < 4 < 7 < 3 < 2 < 1, whereas on the perfluorophenyl column the order was 8-11 < 4 < 5 < 2 < 3 < 6 < 7 < 1, and on perfluorooctyl it was 8-11 < 2 < 4 < 3 < 5 < 6 < 7 < 1. This highlights the importance of carefully identifying each isomer peak using isolated standards or CID patterns when undertaking an isomer-specific analysis, rather than relying on elution orders reported in the literature. This UPLC method is more rapid than existing isomer specific methods, however in our experience it had limited applicability for biological samples since it did not adequately separate endogenous steroid interferences (11, 19). However, this UPLC method may be very useful for abiotic samples such as water, air, or dust. It is also important to point out that, even with this highly resolved method, not all branched isomers were baseline separated from each other or the linear isomer. Specifically,
TABLE 1. Quantification of PFOS (ng/mL) in Human Serum A and B by the Traditional ∑PFOS Method, a Total Branched Method, and an Isomer Specific Method (All Methods Used the brPFOSK Technical Standard; All Data Were Collected in Selected Reaction Monitoring Mode Using m/z 499 as the Precursor Ion (m/z Values in the Table Are Product Ions)) ∑PFOS and total branched PFOS ∑PFOS
a
m/z
∑PFOS
99
b
isomer specific analysis
total branched PFOS
80
99
80
linear 80
7
6
5
4
2
80
130
330
130
419
sum of branched isomers
sum of all isomers
mean 95% CI
3.72 (0.53
2.85 (0.82
1.15 (0.85
0.92 (0.22
1.8 (0.7
Serum A 0.29 0.07 (0.02 (0.04
0.19 (0.09
0.12 (0.03
0.06 (0.05
0.7 (0.13
2.5
mean 95% CI
18.1 (4.20
21.1 (4.89
9.6 (2.31
9.5 (2.54
10.6 (2.86
Serum B 3.58 1.06 (0.80 (0.18
1.44 (0.31
0.82 (0.12
0.39 (0.14
7.3 (1.41
17.9
a Quantification by summing the areas of the m/z 80 product ions for all of the peaks. areas of the m/z 99 product ions for all of the peaks.
b
Quantification by summing the
TABLE 2. MS and MS/MS Relative Response Factors (%) for the Various PFOS Isomers Relative to Linear PFOS (Isomer 1) (Uncertainty of All Measurements is ±16%) isomer
1
2
3
4
5
6
7
8
MS relative response factor for m/z 499 using electrospray ionization 100 40 106 80 101 109 91 84 3.0 m/z 499 f m/z 99 m/z 499 f m/z 80
100 100
77 117 0
In-source fragmentation (% ) 100 × A/B)a 6 4.6 1.7 1.7 1.8 MS/MS relative response factors 97 49 39 43 78 135 241 142
78 123
7.3 10 113
9
10
11
69
73
74
44
38
22
0 118
0 220
19 90
a “A” is the sum of the signals from the different channels set at m/z 80, 99, 130, 180, 230, 280, 330, 169, 219, 269, 419, and “B” is equal to “A” + signal at m/z 499.
isomer 2 coeluted with linear PFOS (Figure 3), however quantification by using their unique product ions in MS/MS (m/z 419 and 80, respectively) allowed for their resolution and accurate quantification. The relative response factors for the branched PFOS isomers under electrospray ionization were determined in separate injections by UPLC-MS using MPFOS [13C4-PFOS] as an internal standard under selected ion monitoring conditions optimized for linear PFOS (1). In the selected ion monitoring mode (m/z 499), isomers 2, 9, 10, and 11 showed a substantially lower response (
