Persistence of PFOA Pollution at a PTFE Production Site and Occurrence of Replacement PFASs in English Freshwaters Revealed by Sentinel Species, the Eurasian Otter (Lutra lutra)

Concentrations of 33 PFASs were determined in 20 Eurasian otters, sampled 2015–2019, along a transect away from a factory, which used PFOA in PTFE manufacture. Despite cessation of usage in 2012, PFOA concentrations remained high near the factory (>298 μg/kg ww <20 km from factory) and declined with increasing distance (<57 μg/kg ww >150 km away). Long-chain legacy PFASs dominated the Σ33PFAS profile, particularly PFOS, PFOA, PFDA, and PFNA. Replacement compounds, PFECHS, F-53B, PFBSA, PFBS, PFHpA, and 8:2 FTS, were detected in ≥19 otters, this being the first report of PFBSA and PFECHS in the species. Concentrations of replacement PFASs were generally lower than legacy compounds (max: 70.3 μg/kg ww and 4,640 μg/kg ww, respectively). Our study underscores the utility of otters as sentinels for evaluating mitigation success and highlights the value of continued monitoring to provide insights into the longevity of spatial associations with historic sources. Lower concentrations of replacement, than legacy, PFASs likely reflect their lower bioaccumulation potential, and more recent introduction. Continued PFAS use will inevitably lead to increased environmental and human exposure if not controlled. Further research is needed on fate, toxicity, and bioaccumulation of replacement compounds.


Methods
• Table S1: Otter sample selection breakdown • Further details on analytical determination • Table S2: List of determinands and analytical standards • Table S3: Variables relevant to testing association of PFAS concentrations in otters to distance from PTFE manufacturing site.• Table S4: Principal Component Analysis (PCA) results for PFASs which were detected in ≥70% of otters.
For quality assurance purposes, a blank and reference material sample (NIST 1946 [Lake Superior fish tissue] and NMCAG-RM1 spiked mussel tissue) were analysed with every 10 samples.The blank was a method (lab) blank.A chemical interference in the quantitative ion channel has been detected for PFBA in human placenta samples [1].The samples in the present study were spiked with labelled PFBA and concentrations recovery corrected, therefore any interference should be affecting native and labelled at the same level, so final concentrations are therefore corrected for any losses or interferences.The same has been carried out for PFPeA and PFHxA.Validations in biota samples, reference materials and interlabs have not shown problems with any of these chemicals, nor problems with recoveries of the labelled chemicals.This could suggest problems may be sample related more than uniquely method or chemical related.

Table S2: List of determinands and analytical standards
Details include the PFAS name, CAS number, abbreviation, carbon number (C n ), whether the PFAS is a legacy (L) or replacement (R) compound, analytical standard used, and limit of quantification (LOQ) measured in µg/kg wet weight achieved during this study.Nomenclature follows that of Buck et al., 2011.Compounds denoted with a * are new to the suite PFASs analysed by Cefas, the results for these compounds can be considered quantitative as they are from a fully calibrated internal standard method, however, caution should be taken as they have not been subject to a full method validation like the original suite.In our previous study [2], percentage of arable land in 10km radii of otters was retained (nonsignificant, variable importance: 0.39) via multi-model inference to explain PFOA concentration, as well as distance to the PTFE manufacturing site and wastewater treatment works load (WWTW) within a 10km radius.Consequently, we initially included both WWTW and arable land in the PFOA model for the current study, but for this, sample model validation steps showed a high collinearity [positive association] between arable land and distance from PTFE manufacturing site based on variance inflation factor, calculated using corvif function in the Car package [3].Percentage arable land was therefore not included in the starting model for the current analysis.First four principal components (PCs) shown, which explained 81.4% of the variation in the dataset.Calculated in R version 4.1.2(R Core Team 2021).Important loadings are highlighted (positive = yellow, negative = green).Cut off for important loadings was 0.20412 (sum of the squares of all loadings for each PC divided by number of variables; therefore, variables with a larger loading than this value contribute more than one variable's worth of information to the PC and are regarded as 'important').PFASs are arranged into classes, carbon number (C n ), whether the PFAS is a legacy (L) or replacement (R) compound are provided.Short chain compounds are highlighted blue.Compounds and class are denoted by their abbreviation, see  Biplot shows all compounds loading negatively on PC1, except PFOA and PFBA.Shorter chain compounds loading heavily positively (>0.20412) on PC2 with the majority of the other compounds loading negatively on PC2.

•
Figure S2: Model predicted decline in PFOA concentration with distance from the factory producing PTFE for the previous data set (2007-2009, blue) and this study's data set (2015-2019, red).• Figure S3: Model predicted decline in PFOA concentration with distance from the factory producing PTFE, without the 3 most southerly located otters.• Figure S4: Model predicted change in PC1 (largely representative of long chain PFASs which loaded negatively) and PC2 (largely representative of short chain PFASs which loaded positively, and PFOA, L-PFOS and PFOSA which loaded negatively) with distance from the factory and year.• Figure S5: Principal component analysis (PCA) plots with sex (top) and age (bottom) colour coded and ellipses shown.
Detail of data collection n / range Sex Determined at post-mortem examination (PM) by Cardiff University Otter Project (CUOP).Male n = 9 Female n = 11 Age Determined at PM by CUOP using following categories: -Female sub-adult = weight ≥2.1kg, immature uterus, no sign of previous/ current reproduction.-Female adult = mature uterus and/or sign of previous/ current reproduction.-Male sub-adult = weight ≥3kg, baculum <60mm.-Male adult = baculum ≥ 60mm.tip.Measured at PM by CUOP.Range: 885-1180 mm Normal distribution Body condition Scaled Mass Index (SMI) estimate of body condition (using Peig and Green, 2009).Calculated from length and weight measurements taken during PM by CUOP.Range: 5.307-8.771Normal distribution

Figure S2 :
Figure S2: Model predicted decline in PFOA concentration with distance from the factory producing PTFE for the previous data set (2007-2009, blue) and this study's data set (2015-2019, red).Blue dots show the raw data for the 2007-09 dataset, the solid blue is the model predicted concentrations.Red dots show the raw data for the 2015-19 dataset, the solid red line is the model predicted concentrations and the orange line shows the predicted concentrations forecast beyond the dataset.Dotted lines show standard error.Other variables retained in the top models of the respective datasets are controlled in the model predictions, see Table S5 for this study and O'Rourke et al. 2022 for previous dataset for details.Note; the datasets were not modelled together due to the differing sample selection (a spread of otters from across England and Wales for 2007-09 data, and a transect east from the factory for 2015-19 data -there was insufficient overlap between the datasets to enable robust comparison within the same model).

Table S4 : Principal Component Analysis (PCA) results for PFASs which were detected in ≥70% of otters.
Table S2 for full compound and class names.

Table S5 : GLM results for PFASs modelled.
Standardised estimates presented for averaged model.'Estimate' is the estimated coefficient, and SE is the standard error around that estimate.Models for PFOA and PFNA used raw data with Gamma family and log link function.Models for PFBA and PC1 used raw data with gaussian family and identity link function.All variables in the starting model (factory distance, year, latitude and WWTW load) were included as a continuous variables.