Fluorine Mass Balance and Suspect Screening in Marine Mammals from the Northern Hemisphere

There is increasing evidence that the ∼20 routinely monitored perfluoroalkyl and polyfluoroalkyl substances (PFASs) account for only a fraction of extractable organofluorine (EOF) occurring in the environment. To assess whether PFAS exposure is being underestimated in marine mammals from the Northern Hemisphere, we performed a fluorine mass balance on liver tissues from 11 different species using a combination of targeted PFAS analysis, EOF and total fluorine determination, and suspect screening. Samples were obtained from the east coast United States (US), west and east coast of Greenland, Iceland, and Sweden from 2000 to 2017. Of the 36 target PFASs, perfluorooctane sulfonate (PFOS) dominated in all but one Icelandic and three US samples, where the 7:3 fluorotelomer carboxylic acid (7:3 FTCA) was prevalent. This is the first report of 7:3 FTCA in polar bears (∼1000 ng/g, ww) and cetaceans (<6–190 ng/g, ww). In 18 out of 25 samples, EOF was not significantly greater than fluorine concentrations derived from sum target PFASs. For the remaining 7 samples (mostly from the US east coast), 30–75% of the EOF was unidentified. Suspect screening revealed an additional 37 PFASs (not included in the targeted analysis) bringing the total to 63 detected PFASs from 12 different classes. Overall, these results highlight the importance of a multiplatform approach for accurately characterizing PFAS exposure in marine mammals.


S3
Two clean-up steps were evaluated for their potential to remove inorganic fluorine and recovery of target analytes: 1) a solid phase extraction (SPE)-based clean-up, and 2) an EnviCarb-based clean-up. The SPE extraction method was based on Miyake et al. 2

and the EnviCarb extraction on
Powley et al. 1 Fish muscle samples were spiked with 250 ng PFOS (~162 ng F) and 500 ng NaF.
Method blanks showed high concentrations for the SPE clean-up with high variation, while the method blanks for the EnviCarb clean-up step were rather low and consistent. The unspiked samples showed similar concentrations for both methods, however EnviCarb showed a bit higher deviation. PFOS and NaF recoveries were calculated according to the following formula: (%) = Measured spike (ng F) − Measured no spike (ng F) Spiked concentration (ng F) × 100% PFOS recovery was high for both methods, i.e. 96% and 92% for SPE and EnviCarb, respectively.
Both extraction methods aim to remove inorganic fluorine, such as NaF, and get as low recovery as possible. NaF recovery was 12.5% and -0.2% for SPE and EnviCarb, respectively. Only the EnviCarb method was able to remove the inorganic fluorine effectively. After this extraction method comparison, EnviCarb was found to be the best suitable to use for analysis of the real samples, since this approach resulted in lower method blanks and more efficient removal of inorganic fluorine and was therefore considered the most suitable clean-up method.

Total and extractable organofluorine
A similar extraction procedure was applied to the liver samples prior to analysis with the CIC.
Since the CIC measures the total fluorine concentration, no internal standards were added, also no NH4OAc was added in the end. The final extracts (~ 1 ml) were split into two parts, in order to have a replicate of each sample. Also, since the sample boats have limited sample space, the final split extracts were concentrated to ~200 μl under a stream of nitrogen.

Targeted analysis
The system was operated in negative ion electrospray ionization (ESI-) mode. The source and desolvation temperatures were set at 150 ℃ and 350 ℃, respectively. The desolvation and cone gas flows (nitrogen) were set at 650 L/h and 150 L/h, respectively. The capillary voltage was set at 1.0 kV. Qualification and quantification were carried out using MassLynx 4.1 (Waters).

S4
Quantification was performed using internal standards via a 9-point calibration curve ranging from 0.008 to 150 ng/ml (linear, 1/x weighting). Precursor and product ions are presented in Table S7.
Analytes lacking an analogous labeled standard were quantified using the IS with the closest retention time and the data quality was defined as semi-quantitative (semiQ). Branched isomers were quantified using the calibration curve of the linear isomer.

Total and extractable organofluorine analysis
Measurements of total fluorine (TF) and extractable organofluorine (EOF) were carried out using a Thermo-Mitsubishi CIC using previously described methods. 3,4 Briefly, extracts (~200 µl for samples and 100 µl for standards) were placed in a ceramic sample boat containing glass wool (for fluid dispersion), while neat liver material (~100 mg) was weighed directly into the sample boat.
The samples were combusted slowly in a horizontal furnace (HF-210, Mitsubishi) at 1100 ℃ under a flow of oxygen (400 ml/min), argon (200 ml/min), and argon mixed with water vapor (100 ml/min) for approximately 5 minutes. Combustion gases were absorbed in MilliQ water during the entire length of the combustion process using a gas absorber unit (GA-210, Mitsubishi).
A 200 µl aliquot of the absorption solution was subsequently injected onto an ion chromatograph (Dionex Integrion HPIC, Thermo Fisher Scientific) equipped with an anion exchange column (2 × 50 mm guard column (Dionex IonPac AS19-4µm) and 2 × 250 mm analytical column (Dionex IonPac AS19-4µm) operated at 30 ℃. Chromatographic separation was achieved by running a gradient of aqueous hydroxide mobile phase ramping from 8 mM to 100 mM at a flow rate of 0.25 ml/min (Table S8), and fluoride was detected using a conductivity detector.
Quantification was carried out using a standard calibration curve prepared at 0.05 to 100 µg F/ml.
The calibration curve showed very good linearity with R 2 >0.98. The mean fluoride concentration in the method blanks was subtracted from the samples. The method detection limit (MDL) was defined as the mean concentration plus three times the standard deviation in the method blanks.

Suspect screening
Suspect screening was carried out using a Dionex Ultimate 3000 liquid chromatograph coupled to a Q Exactive HF Orbitrap (Thermo Scientific), based on a previously described method. 5 The flow rate was held constant at 0.4 ml per minute throughout the run. The mobile phases and eluent program used for non-target/suspect screening were the same as those used for target analysis (i.e.

S5
by UPLC-MS/MS). The instrument was run in negative ion, full scan (200-1200 m/z) data dependent acquisition (DDA) MS/MS mode (50-1200 m/z). The resolution was set to 120 000 (15 000 for MS/MS), the automatic gain control (AGC) was set to 3e6, and other instrumental parameters are presented in Table S11. Briefly, CL = 5 is assigned when only the exact mass is known. CL = 4 is used when the unknown analyte ion can be assigned an unambiguous formula, but no structural information is available. CL = 3 represents tentative candidates whose possible structure can be proposed but lack sufficient information to assign an exact structure. CL = 2a represents probable structures by comparing to library spectra where spectrum-structure is unambiguous. CL = 2b can be assigned when no standard or literature information is available for confirmation and there is only diagnostic evidence. Finally, CL =1 represents confirmed structures, that match a reference standard with MS, MS/MS and RT.

Targeted analysis
Limits of quantification (LOQs) were determined by the lowest calibration concentration that showed a well-shaped peak with intensity >1e3 and signal-to-noise (S/N) >3. For compounds that were not present in the calibration standard, but that were detected in the samples (PFPeDA, PFHpS, and branched isomers), the LOQ from the corresponding standard was used. For compounds where method blank contamination was observed (PFBS, PFOS, and FOSAA), the LOQ was determined as the average of the quantified concentrations in the method blanks plus ten times the standard deviation. The compound-specific LOQs are listed in Table S6.
Method accuracy and precision for most substances was very good, with percent recoveries ranging from 73-130% and standard deviations ranging from 3-30% ( Figure S2). The exceptions were for PFHxDA, PFOcDA, 4:2 FTSA, and 8:2 FTSA, which showed very high recoveries (278%, 397%, 212%, and 227%, respectively), while HFPO-DA, 3:3 FTCA, 5:3 FTCA, and 7:3 FTCA showed very low recoveries (22%, 34%, 55%, and 53%, respectively). These deviating recoveries are likely due to matrix effects, which were not accounted for because of the absence of an exactly matching isotopically-labeled internal standard. Nevertheless, the targets with very high recoveries were included in the analysis, since their concentrations in the samples were so low (<1 ng/g, ww). The targets with low recoveries were also included in the analysis, albeit S6 measured concentrations may be underreported. Finally, the method was externally validated by analyzing a standard reference material (SRM) sample of NIST serum 1957. Results are presented in Table S9 and were generally in good agreement with certified values and prior measurements of this material by other researchers.

Total and extractable organofluorine
All boats were baked out prior to sample combustion to minimize background contamination. Each run started and ended with a calibration curve and after every 8-10 samples, a blank and a midlevel calibration standard were analyzed for quality control. The removal efficiency of inorganic fluoride was tested by spiking a range of known concentrations of NaF (0.25, 0.5, 0.75, 1, and 2 µg) into liver tissue followed by extraction ( Figure S3). Furthermore, recovery of organic fluoride was determined by spiking PFOS (0.08, 0.13, 0.25, 0.5, and 1 µg/ml) to liver tissue and performing the extraction ( Figure S4). CIC analysis of both the extracted liver residue and the EnviCarb used for clean-up showed that the inorganic fluoride remained in the extracted liver; in other words, it was not extracted during the initial acetonitrile extraction step. The obtained recovery for PFOS was used to correct the measured concentrations of EOF in real samples. In theory, since the recovery is different for each target analyte, the recovery should be determined for each individual compound. However, practically this would mean a large number of experiments and therefore only the recovery for the most abundant compound, PFOS, was assessed here.
S7 Figure S1. Diagram of the experimental design.
S8 Figure S2. Recovery ± standard deviation (%) of native compounds spiked in seal samples (n=4). Right panel shows severe overrecovery of four targets, attributable to matrix-induced ionization enhancement.      No MS/MS spectra were available. S19 Figure S13. EICs for unknowns (class 12) in pygmy sperm whale. MS/MS spectra for four compounds within class 12 with molecular formulas assigned to the most common fragments.     a Compounds that were present in the method blanks and for these the LOQ was determined alternatively by calculating the average contamination concentration plus ten times the standard deviation. b Compounds that were not present in the calibration curve, but that were present in the samples.  Table S8. Eluent program for the ion chromatography part of the CIC analysis.