Developing Methods for Assessing Trophic Magnification of Perfluoroalkyl Substances within an Urban Terrestrial Avian Food Web

We investigated the trophic magnification potential of perfluoroalkyl substances (PFAS) in a terrestrial food web by using a chemical activity-based approach, which involved normalizing concentrations of PFAS in biota to their relative biochemical composition in order to provide a thermodynamically accurate basis for comparing concentrations of PFAS in biota. Samples of hawk eggs, songbird tissues, and invertebrates were collected and analyzed for concentrations of 18 perfluoroalkyl acids (PFAAs) and for polar lipid, neutral lipid, total protein, albumin, and water content. Estimated mass fractions of PFCA C8–C11 and PFSA C4–C8 predominantly occurred in albumin within biota samples from the food web with smaller estimated fractions in polar lipids > structural proteins > neutral lipids and insignificant amounts in water. Estimated mass fractions of longer-chained PFAS (i.e., C12–C16) mainly occurred in polar lipids with smaller estimated fractions in albumin > structural proteins > neutral lipids > and water. Chemical activity-based TMFs indicated that PFNA, PFDA, PFUdA, PFDoA, PFTrDA, PFTeDA, PFOS, and PFDS biomagnified in the food web; PFOA, PFHxDA, and PFHxS did not appear to biomagnify; and PFBS biodiluted. Chemical activity-based TMFs for PFCA C8–C11 and PFSA C4–C8 were in good agreement with corresponding TMFs derived with concentrations normalized to only total protein in biota, suggesting that concentrations normalized to total protein may be appropriate proxies of chemical activity-based TMFs for PFAS, which predominantly partition to albumin. Similarly, TMFs derived with concentrations normalized to albumin may be suitable proxies of chemical activity-based TMFs for longer-chained PFAS, which predominantly partition to polar lipids.


Fugacity and Chemical Activity
Fugacity and chemical activity are complementary concepts developed by G. N. Lewis in 1901 and used in the field of thermodynamics to characterize a chemical's capacity for transport and transformation and predict its environmental fate. 1,2 Defining the BMF and TMF in both fugacity and chemical activity formats is advantageous as it enables some substances, such as those with a significant presence in the gas phase, to be treated with the fugacity approach while other substances, which do not readily partition to the gas phase but have a considerable presence in water, are more easily treated with the chemical activity approach. 3,4 Normalising measured concentrations to act as proxies of fugacity or chemical activity does not require the selection of a specific standard reference state, which is difficult to determine for PFAS as PFAS generally occur in ionic form, since it cancels out in a fugacity or chemical activity ratio. Thus, when referring to chemical activity we are referring to apparent chemical activity rather than chemical activity to avoid specifying a standard reference phase for PFAS.

Biomagnification of Lipid Soluble Substances
Since lipid-soluble substances, such as polychlorinated biphenyls and other legacy POPs, preferentially partition into lipids in organisms, BMFs of lipid-soluble substances are usually derived from lipid normalised concentrations. [5][6][7][8][9][10][11] Thus, equation 1 is modified as: where and represent the sorptive capacity of the chemical in the organism; represents the sorptive capacity of lipids; the fraction of total lipid content; and , and , are the lipid normalised concentrations of the chemical (e.g., mol/g lipid) in the predator and prey. Lipid normalisation is usually a standard practice when investigating biomagnification (e.g., OECD 305 guidelines 12 ) and trophic magnification of lipid-soluble substances in organisms and is generally based on results from aquatic organisms and food-webs. [13][14][15] Accordingly, the TMF of lipid-soluble substances is derived from a linear regression of lipid normalised concentrations in organisms ( , ) as: This normalization approach inherently assumes that (i) the sorptive capacity of the organism is chiefly represented by lipids and directly proportional to the fraction of lipid content in the organism and (ii) that the sorptive capacity of the substance in lipids for all organisms is equivalent. However, these assumptions are not always applicable. For instance, if lipid content within organisms is low, then nonlipid organic matrices, such as proteins, will become the main site of chemical bioaccumulation. 8 In such cases, lipid normalisation can overestimate the fugacity and concentration in organisms with low lipid content and subsequently underestimate trophic magnification since many lower trophic level organisms have low lipid content.

Soil
Within 5 m of an invertebrate or earthworm collection site, a soil sample (diameter approximately 10 cm with a depth of approximately 5 cm) was dug up with a metal trowel or a bulb planter and placed into a 250 ml chemically rinsed jar then frozen at -20°C. The trowel or bulb planter were cleaned with a 10% ethanol solution prior to digging at each subsampling station.

Biota
Cooper's hawk eggs were chosen to represent the apex predator because eggs represent a maternal concentration of contaminants at a given point in time and are frequently used as a matrix for environmental contamination monitoring ( Figure S3). [21][22][23] The hawks' main target prey species were American Robins (Turdus migratorius), European Starlings (Sturnus vulgaris), and House Sparrows (Passer domesticus). 24 We supplemented our targeted songbird prey species with samples of 12 other known Cooper's Hawk prey species, [24][25][26][27][28][29]  Association, Burnaby, BC. The lower levels of the food web were represented by terrestrial invertebrates (such as beetles, earthworms, and sowbugs) commonly eaten by birds ( Figure S3). [30][31][32][33][34][35] Sample sizes of each species monitored in the current study are listed in Table S1.

Sample Preparation
Air PUF disks used for air sampling were precleaned and shipped as previously reported in Schuster, et al. 36 . PUF disks were precleaned using an Accelerated Solvent Extractor ( Table S1. Cooper's hawk eggs were homogenized by whisking the yolk and albumen together. Songbirds were defeathered (i.e., plucked) and large keratinized or boney tissues (e.g., beaks, wings, legs, and feet) were removed prior to further processing. All frozen or semi-thawed biota samples (songbirds and invertebrates) were processed by cutting tissues into small pieces and homogenizing them with a ball-mill (Retsch TM MM400 Mixer Mill, Fisher Scientific). Prior to and after homogenization, samples were stored at -40 °C.

S13
Chemical Analysis Table S2. Perfluoroalkyl acids (PFAA) and corresponding internal standards that were included in the chemical analysis. Physicochemical properties were obtained from the CompTox Chemicals Dashboard for neutral molecules. 37,38 Values are experimental averages (in black) or predicted averages (in red) from the Open (  Passive air sampling chemical analysis has been previously published in Shoeib, et al. 40 and Rauert,et al. 41 . Briefly, the extraction method consisted of spiking the PUF disks with surrogates of 0.5 ng of labelled PFAA (Table S2). PUFs were then extracted using ASE (ASE 350, Dionex Corporation, Sunnyvale, CA, USA) with acetonitrile (3 cycles) to extract the PFAAs. Extracted fractions were volume reduced to 0.5 mL using TurboVap (Biotage, Charlotte, NC, USA) and nitrogen blowdown. Each sample fraction was further purified with activated carbon columns containing 100 mg of ENVI-Carb (100-400 mesh, Supelco, St. Louis, MO) and eluted using 4 mL of acetonitrile before final solvent exchange to methanol and volume reduction to 0.5 mL. Prior to instrumental analysis, each sample was spiked with 0.5 ng of 13 C8 PFOA and 13 C8 PFOS for use as injection standards. Injection standards were used to quantify the surrogates and calculate surrogate recoveries. The separation and detection of PFAS was performed using ultra performance liquid chromatography tandem mass spectrometry (UPLC-MS/MS), following previously published methods in Shoeib, et al. 40 .
Air concentrations (pg/m3 or ng/m3) of target analytes were derived from the mass of the chemical collected on the PUF divided by an effective air sampling volume. This air volume was determined with the GAPS template 42 , using the average temperature during the deployment period and an estimated PUFair partition coefficient from Ahrens, et al. 43 . The effective air volume was calculated as the number of days the PUF was deployed (i.e., 90) multiplied by a sampling rate of 4 m 3 /day. 44 Only one ion transition is monitored for PFBA and PFPeA, so there is a greater level of uncertainty in their reported concentrations compared to the other PFAS. 41 Table S3. PFAA chemicals with names, abbreviations, associated surrogate/injection standard, and quantifier transition used in processing and analysis of PUF disks. The column temperature was held at 50°C while the sample was maintained at 20 °C. The PFAAs were S16 detected by negative electrospray ionization (ESI-) in multiple reaction monitoring scanning mode (MRM). The relative response of a given analyte to its mass-labelled internal standard is used to calculate the concentration of the analyte, which corrects for any experimental losses of both the analyte and its internal standard.

Standards and Chemicals
All the PFSA/PFCA standards (including isotopically enriched internal standards in

Air
During the installation of the air samples, three field blanks were deployed and collected (i.e., opened, placed in the metal housing, removed, and sealed as a negative control) at three sites: Inter River Park, North Vancouver; Heather St. and W. 20th Ave, Vancouver -West.; and Ladner Harbour Park, Delta. During the air sample processing, two procedural sample blanks were run with separate sets of 4 -5 samples each (Table S4). Each sample was spiked with surrogate standards. The limit of detection (LOD) for each contaminant in the air was determined as 2xSD of the mean concentration of the field and procedural blanks (Table S5). The method detection limit (MDL) for each contaminant in the air was determined as the mean concentration of the field and procedural blanks + 2×SD (Table S5). Data below the MDL were reported as non-detect (ND). Air samples were recovery and blank corrected using the MDL of the field and procedural blank concentrations.

Soil
To check for contamination of analytes from solvents or the extraction process, one procedural blank, one spiked matrix, and one duplicate was prepared and analysed for the batch of 12 soil samples.
Initial calibration of the LC-MS/MS instrument is performed by the analysis of five or more calibration S17 solutions and the lowest calibration standard is used as the reporting limit (RL ; Table S6). A mid-level calibration standard is analysed to verify the initial calibration and injected after every 12 hours or after every 10 client samples (whichever comes first). Sample specific detection limits (SDL) are determined by converting the area equivalent to 3 times the estimated chromatographic noise height to a concentration in the same manner that target peak responses are converted to final concentrations (Table S6). The SDL accounts for any effect of matrix on the detection system and for recovery achieved through the analytical work-up. For some samples, the percent recoveries for some surrogate compounds did not meet the method criteria limits. But as the isotope dilution method of quantification produces data that is recovery corrected, these variances from method criteria were deemed to not affect the quantification of the target analytes. Percent surrogate recoveries are used as general method performance indicator only; percent recoveries for PFAS in the spiked matrix ranged from 73 -98 % (Table S6). Soil samples were blank corrected with the procedural blank.

Biota
To check for contamination of analytes from solvents or the extraction process, a blank sample was prepared and analysed for each batch of 10 -11 samples (n = 8; Table S4). In-house reference materials, The method limit of quantification (MLOQ) was measured by performing replicate analyses of blank matrix samples (n=8) spiked with analytes at a concentration of 3 -5 times the estimated detection limit and calculating the standard deviation: MLOQ = t × SD (Table S5). Where t = Student's value for a 99 % confidence interval and a standard deviation with n-1 degree of freedom, and SD = standard deviation of S18 the replicate control. In this experiment t (n-1, 0.99) = 2.998. MLOQ were determined for samples of hawk eggs, songbirds, and invertebrates (Table S5).
The method limit of detection (MLOD) was determined separately for samples of hawk eggs, birds, and invertebrates (Table S5). MLODs were defined as the concentration that would give a signal to noise ratio (peak to peak) of 3, which means there is no detected peak (or a signal to noise ratio [peak to peak] lower than 3) of the target compound found in the respective ion extracted chromatogram within its retention time window. PFAS concentrations reported are corrected for background contamination by subtracting their respective method blank concentration values.    Next, 20 μL of each standard was transferred in duplicate into a 96 well plate. Then, 80 μL of a prepared reaction mix (85 μL assay buffer, 1 μL enzyme mix, 1 μL PLD enzyme, and 1 μL dye reagent) was added to each well; and the well plate was wrapped in aluminum foil and incubated at room temperature for 30 mins. Finally, the absorbance of the samples and standards was measured at 570 nm in a microplate reader (SpectraMax® M2e, Molecular Devices, LLC, San Jose, CA). The fraction of non-polar lipids in each sample was then derived by subtracting the fraction of polar lipids from the fraction of total lipids.

Protein Composition
To determine the fractions of total protein, albumin, and structural protein in the samples, we measured the amount of total protein in each sample with a Bradford assay method, 46  Reagent (500-0006, Bio-Rad Laboratories Ltd., Montreal, QC) in deionized water was transferred to each standard and sample well; and the well plate was wrapped in aluminum foil and incubated at room temperature for 5-10 mins. Finally, the absorbance of the samples and standards was measured at 595 nm in a microplate reader (SpectraMax® M2e, Molecular Devices, LLC, San Jose, CA). Organic matter (e.g., fiber and carbohydrates) in each sample was estimated as the dry weight (i.e., 100water content %) minus the total protein and total lipid content.

Apparent Chemical Activity
We obtained measured distribution coefficients for PFAS from Allendorf, et al. 16 , Droge 18 ,and Bischel,et al. 17 (Table S7). We used the distribution coefficients measured by Allendorf, et al. 16 for our chemical activity calculations and modelling instead of those from Bischel, et al. 17 and Droge 18 because (i) the distribution coefficients reported in these two studies were determined at 21 -25C, which are not environmentally relevant body temperatures for birds [47][48][49] or preferred optimal temperatures for terrestrial invertebrates like earthworms, [50][51][52] and (ii) since Allendorf, et al. 16 reported distribution coefficients for different tissue types this maintains consistency across experimental conditions.
For PFAS that did not have measured tissue-water distribution coefficients, we predicted the respective distribution coefficients from linear regressions using all the measured values from Allendorf, et al. 16 , Droge 18 ,and Bischel,et al. 17 and the molar volume of PFAS (Table S8). Molar volume was chosen rather than total surface area or carbon chain length to predict distribution coefficients since several quantitative structure property relationships based on molar volume of PFAS were often statistically better than those based on total surface area and can help distinguish between PFAS with the same number of attached fluorines and/or carbon chain length 53 . Also, carbon chain length, molar volume, molar weight, and the number of attached fluorine all appeared to have similar linear relationships with the measured distribution coefficients ( Figure S4). We derived separate linear regressions for PFCA and PFSA, which were all statistically significant (p < 0.01) and demonstrated reasonably good fits ( Figure S5 and S6) apart from the storage lipid-water distribution coefficient (DSLW) for PFSA (p > 0.05). The relationship for DSLW of PFSA had a very large confidence interval likely due to the small sample size in the model, which adds greater uncertainty to subsequent analyses relying on these predictive values.
S23 Table S7. Measured tissue-water distribution coefficients of PFAS obtained from Allendorf, et al. 16

VM MW
Allendorf, et al. 16 Bischel, et al. 17 16 Allendorf,et al. 16 r 2 = 0.96,p < 0.001;r 2 = 0.97,p < 0.001. b Predicted with experimental values from Allendorf, et al. 16 and Bischel,et al. 17 using equations: PFCA Log DALBW = 0.0525 × VM -0.000102×VM r 2 = 0.63,p < 0.001;PFSA Log DALBW = 0.00935 × VM + 2.26,r 2 = 0.46,p = 0.038. c Predicted with experimental values from Allendorf, et al. 16   As substances like PFAS are typically solids at environmental temperatures, the activity coefficient must be adjusted by a fugacity ratio (F) in order to visualise the solid chemical behaving as a semi-cooled liquid. 1, 3 F (unitless) is a function of the enthalpy of fusion and the melting point of the chemical. By applying Walden's rule, which states that the enthalpy of fusion at the melting point is roughly 56.5 J/mol K, F can be estimated with the melting point ( ; K) and the environmental or organism body temperature (T; K) as 1, 3 : ln ≅ −6.79 ( − 1) Table S9 were obtained from the CompTox Chemicals Dashboard 37 and represent the median value if the range of predicted or experimental values was large (i.e., SD greater than 2) or the average value if the range of values was small. Experimental values were prioritised. Tissue solubilities (mol/L) for PFAS in Table S8 were calculated as the product of the water  Table S9. Tissue solubilities (mol/L) for PFAS were converted to a mass basis assuming a density of 1.0 kg/L for water, 1.36 kg/L for albumin, 1.097 kg/L for polar lipids, 0.95 kg/L for neutral lipids, and 1.36 kg/L for structural protein. Solubility of PFAS within organisms was subsequently determined as:

Melting points and water solubilities in
where represents the fraction of albumin in the sample; the fraction of neutral or storage lipids; the fraction of polar or membrane lipids; the fraction of structural protein; the fraction of water content; and , , , , and represent the solubilities of albumin, neutral lipids, polar lipids, structural proteins, and water, respectively.

Biochemical Composition of Biota
Based on the scatterplots in the pairplot of the variables ( Figure S7), we constructed individual linear regressions between TP and the mass fraction (%) of each tissue phase with particular interest in albumin and polar lipid as they appeared to have stronger correlations with TP. Absolute correlations between neutral lipid and total lipid, albumin and total protein, and structural protein and total protein were likely very high due to collinearity and from being constituents of total lipid or total protein ( Figure   S4).

S28
Chemical mass of PFAS (ng) in 1 g of each tissue phase was determined by assuming that the chemical activities in each tissue phase were equivalent and represented by the activity in the organism as:

= ×
Relative fractions of chemical mass of PFAS (%) in each tissue phase were subsequently estimated from the total mass of PFAS measured in the sample as:
However, Ahrens, et al. 54 measured PFAS in air samples from Toronto using SIP-PAS and PUF-PAS and found that concentrations of PFSA from the two sampling techniques were generally within a factor of 2,

S32
indicating good agreement between the techniques and that both methods are suitable for measuring PFAS in air.

Mass Distribution of PFAS
Since log DALBW values from Allendorf,et al. 16 differ by almost an order of magnitude from log DALBW values from Bischel,et al. 17 for some PFAS, notably PFDA and PFUdA (Table S7) (Table S14). S43 Table S17. Trophic magnification factors of PFAS from terrestrial and aquatic food-webs determined with wet weight concentrations (TMFW), dry weight concentrations (TMFD), and concentrations normalised to total protein content (TMFP).