Large declines in organofluorine contamination indicated by subarctic marine mammal tissues

Significance

Detrimental human and wildlife exposures to per- and polyfluoroalkyl substances (PFAS) have prompted shifts in chemical manufacturing toward novel compounds that are difficult to detect. Here, we show that bulk organofluorine concentrations in liver tissues from a subarctic marine mammal peaked in 2011 and declined by more than 60% by 2023. Four legacy PFAS accounted for ~75% of the detected organofluorine across all years. Significant temporal increases were consistently observed for only one novel PFAS. Results contrast past human serum measurements that contained a large fraction of unexplained organofluorine. Together, these results reinforce the success of regulatory actions for these global contaminants but suggest newer PFAS production may be predominantly accumulating in terrestrial and nearshore environments rather than the open ocean.

Abstract

The ocean is thought to be the terminal sink for per- and polyfluoroalkyl substances (PFAS), persistent organofluorine chemicals used widely in modern commerce for decades. Industry and stewardship programs phased out the most abundantly produced legacy PFAS in the early 2000s due to toxicity concerns. However, they have since been replaced by shorter carbon chain and “novel” chemistries, and past work hypothesized likely increases in these replacement PFAS that were not previously quantifiable. To address this gap, we measured bulk extractable organofluorine (EOF) in archived liver and muscle tissues from pelagic Subarctic pilot whales over the last several decades. Results show EOF concentrations peaked in 2011 and declined by over 60% by 2023. Among a broad suite of targeted and suspect PFAS measured using high-resolution mass spectrometry, only one was consistently increasing through 2023. Tissue concentrations of four main legacy PFAS that accounted for over 75% of EOF were all decreasing by 2023. The timing of peak concentrations depended primarily on whether they were transported to the subarctic by ocean circulation or atmospheric deposition, with the latter declining much faster. Oceanic transport and bioaccumulation modeling suggests that decadal-scale lags between production and food web bioaccumulation are primarily driven by marine transport processes. Large declines in tissue concentrations in this study reinforce the effectiveness of phase-outs in chemical production. However, other work showing stable or increasing EOF in human serum suggests many emerging PFAS with more neutral physicochemical properties may be preferentially accumulating in terrestrial and nearshore environments compared to legacy PFAS.

The ocean is thought to be the terminal sink for per- and polyfluoroalkyl substances (PFAS) produced widely in modern commerce since the 1950s (1). PFAS are not known to fully degrade in the environment, and their stability and surfactant properties have led to inclusion in diverse products such as fire-fighting foams, textiles, food packaging, electronics, and personal care products (2). In early decades, global PFAS production was concentrated in North America and Europe, making the Arctic and Subarctic particularly vulnerable regions for global pollution (3). Adverse effects of PFAS exposure on humans and wildlife have prompted global concern about these chemicals (4, 5), and seafood remains a primary source of dietary PFAS exposure for populations not directly impacted by local contamination (6, 7). Since the early 2000s, voluntary efforts and stewardship programs with North American and European industry, together with global regulatory conventions, have shifted PFAS production regionally away from chemicals with longer perfluorinated carbon chains toward shorter-chain and “replacement” PFAS chemistries (8–10). However, seawater PFAS concentration data are limited and implications of production shifts for temporal trends in marine environments are poorly understood (11).

Only a few dozen PFAS have analytical standards needed for measurement, but some government databases suggest the chemical family contains tens of thousands or more individual compounds (12). Bulk organofluorine measurements, such as extractable organofluorine (EOF), allow quantification of most organofluorine compounds in environmental samples (13, 14). Past work has shown stable or increasing EOF concentrations in human serum, despite declining concentrations of legacy PFAS, suggesting increasing exposure to novel or replacement PFAS (15, 16). Comparable EOF trends in wildlife are presently unavailable. Existing single time-point measurements of EOF in marine mammals suggest a wide range of contributions from unidentified PFAS (0 to 70%), but it remains unclear how this contribution varies with time and distance from contamination sources (17–19).

Marine mammals biomagnify PFAS to detectable levels, making them valuable sentinels of marine pollution and global change (20). North Atlantic pilot whales (Globicephala melas) integrate bioavailable chemical exposures horizontally across the eastern subarctic North Atlantic Ocean and vertically across their prey foraging depth (predominantly squid, cod, and blue whiting) between 400 to 700 m (21). Prior work by Dassuncao et al. reported an overall decline in 15 PFAS measured in North Atlantic pilot whale muscle between 1999 and 2013 (22). However, this decline was predominantly driven by a single PFAS transported atmospherically to the ocean, while concentrations of other legacy PFAS continued to increase. Many shorter-chain and “replacement” PFAS used in commerce were not detectable in that study due to limitations of analytical methods used at the time. Key questions therefore remained about how marine ecosystems respond to production shifts across PFAS classes, including the extent and variability in temporal lags between production and bioaccumulation.

We hypothesized that EOF concentrations in pilot whales have been increasing in response to growing production of emerging and replacement PFAS, mirroring trends observed in humans (15, 16). To test this hypothesis, we measured concentrations of EOF, targeted and suspect PFAS in archived liver (n = 38) and muscle (n = 133) tissues from Subarctic North Atlantic pilot whales over several decades. We compared measured temporal trends in whales to modeled trends in seawater forced by data on chemical releases. This work provides insights into the fate of PFAS in the global environment and the effectiveness of past regulatory measures for reducing seafood PFAS exposures and the fate of PFAS in the global environment.

Results and Discussion

More than 60% Decline in Liver EOF Concentrations Since 2011.

We analyzed temporal trends in EOF concentrations in North Atlantic pilot whale liver (n = 38) between 2001–2023 using a Bayesian change point trend analysis to best represent uncertainty in available data (Methods). EOF concentrations peaked around 2011 [change point 95% credible interval: 2007–2017; Fig. 1A and SI Appendix, Table S1]. This represents a lag of approximately one decade after the phase out of legacy long-chain PFAS, which began in the early 2000s (23, 24). By 2023, EOF concentrations had declined by more than 60% of their 2011 peak (Fig. 1A), indicating that most biologically available organofluorine compounds in the remote North Atlantic have declined over the last decade.

Fig. 1.

Extractable organofluorine (EOF) concentrations in North Atlantic pilot whale liver (Globicephala melas). (A) shows the timeseries of liver EOF concentrations (blue circles represent individual samples) between 2001 and 2023. The solid line and light blue shading show the expected mean (posterior expected value) and 95% credible interval of Bayesian change-point analyses (Methods). (B) compares pilot whale liver EOF and summed concentrations of targeted PFAS as fluorine equivalents. Each point represents an individual sample. Error bars show the SD of EOF and targeted PFAS measurements calculated from replicate analyses. The solid line shows the weighted orthogonal regression with 95% confidence interval (gray shading). The 1:1 line is shown by the dashed line.

Relatively small fractions of unidentified EOF across years indicate that temporal declines predominantly reflect decreasing concentrations of targeted PFAS (i.e., PFAS that can be individually quantified using standard analytical methods). The sum of targeted PFAS analyzed in liver EOF extracts accounted for 90 to 280 ng F/g or 84 ± 2% (range 72 to 100%) of EOF measured in samples across years (Fig. 1B and SI Appendix, Fig. S1). Samples collected in 2001, 2006, 2015, and 2023 had a small but significant fraction of EOF (11 to 28%) that was not explained by the targeted PFAS. For liver samples from 2010 and 2019, targeted PFAS concentrations explained all the EOF detected (SI Appendix, Fig. S1). SI Appendix, section 1.2 contains additional information on targeted PFAS detected in EOF extracts.

Together, these results suggest that in pelagic marine ecosystems, production and use of novel PFAS has not yet offset the impacts of declines in legacy PFAS releases. Although EOF trends have not been previously measured in wildlife, comparisons with single time-point studies are consistent with this finding (SI Appendix, section 1.1). One caveat is that EOF measurements do not capture ultrashort-chain organofluorine compounds such as trifluoroacetic acid (TFA), which are known to be increasing globally but are thought to have a low propensity for bioaccumulation in aquatic ecosystems (25).

Four Legacy PFAS in Muscle Account for Most Targeted PFAS.

Between 1994 and 2023, the sum of targeted PFAS in pilot whale muscle (n = 133) ranged from 11 to 31 ng/g (Fig. 2). The composition of PFAS measured in muscle tissues in 2023 was lower in magnitude but similar in composition to that reported by Dassuncao et al. for samples collected between 2011–2013 (22). Even in 2023, the targeted PFAS composition in muscle tissues was dominated by four legacy PFAS (Fig. 2) with active production and environmental releases in North America and Europe that peaked in the 1990s and early 2000s (8, 10, 23, 24). These included i) 20% as perfluorooctane sulfonate (PFOS), a perfluorosulfonic acid with eight perfluorinated carbons (C8 PFSA); ii) 45% as perfluoroctane sulfonamide (FOSA), the C8 compound in the sulfonamide (FASA) precursor class that is transformed into PFSA; and iii) 15% as two long-chain (C10, C12) legacy perfluorocarboxylic acids (PFCA), which were also largely phased out of production (10, 23). PFOS and the C10 and C12 PFCA are terminal PFAS that are not known to degrade under natural conditions (26), while FASA are precursors to PFSA that are transformed by microbes or in vivo into terminal PFSA with the same carbon chain length.

Fig. 2.

Composition of targeted PFAS concentrations in pilot whale (G. melas) muscle between 1994 and 2023. Note that the sum of targeted PFAS (and subsequent composition) was estimated from the sum of the geometric means for individual PFAS. Perfluorooctane sulfonamide (FOSA, dark green), perfluorooctane sulfonate (PFOS, dark blue), and the C10 and C12 perfluorocarboxylates (C10/C12 PFCA, dark orange) make up >75% of total measured PFAS across all years and are displayed by PFAS class. Other compounds in each PFAS class are shown separately, including perfluoroalkyl sulfonamides (FASA, light green) other than FOSA, perfluorosulfonates (PFSA, light blue) other than PFOS, and perfluorocarboxylates other than the C10/C12 chain lengths (PFCA, light orange). Composition for each sampling year is based on sample mean concentration as a fraction of the summed targeted PFAS. Data prior to 1994 are not shown because FOSA was not measured for these years. No data on the C5 FASA (included in “Other FASA”) were available between 2015–2020 and 2023. Measurements below detection are imputed as the method detection limited divided by the square root of two (MDL/$\sqrt{2}$).

It is striking that PFOS and FOSA continue to make up most of the pilot whale tissue PFAS concentrations more than two decades after North American and European production phase-outs that began in the mid-1990s to early 2000s, which contributed to rapid declines in environmental releases and exposures in near-source environments (8, 24, 27) (Fig. 2). Past work shows FASA, including FOSA, generally exhibit 1 to 2 orders of magnitude greater bioaccumulation than terminal PFAS of the same chain length. Many of these terminal PFAS have been identified as human health concerns (9, 28).

Pilot whales lack the enzyme present in most wildlife needed to biotransform some PFAS precursors into terminal PFAS (29), enabling clearer insights into the individual contributions of both terminal and precursor PFAS to total exposure. Shorter chain length FASA (C4–C6) have not been commonly monitored in wildlife, and we detected them in all samples analyzed. The C4-C6 FASA accounted for <10% of the total targeted PFAS in muscle (Fig. 2). However, detection of short chain FASA in this remote ecosystem highlights their long-range transport potential, in addition to their high bioaccumulation potential.

Monotonic FASA Trends Indicate Rapid Atmospheric Transport.

Fig. 3 shows contrasting temporal trends among the four main targeted PFAS present in pilot whale tissues. FOSA (C8 FASA) accounted for the greatest fraction of the sum of targeted PFAS (Fig. 2). FOSA exhibits the earliest decline among targeted PFAS detected in both liver and muscle and appears to follow a monotonic decrease across all study years (Fig. 3 A and B). Terminal PFAS like PFOS have a low pKa (<1) and are present mainly as the deprotonated acids in seawater, meaning some behave similarly to tracers for seawater circulation (30). By contrast, physicochemical properties for FASA more closely resemble neutral persistent organic pollutants (31, 32). FASA are volatile and have important atmospheric transport pathways in Arctic ecosystems (5, 33). Atmospheric transport results in relatively rapid coupling between declines in emissions and seawater concentrations in the North Atlantic compared to PFAS with only oceanic transport pathways (5, 24). The importance of atmospheric inputs to the ocean for FOSA thus likely explains its monotonic trends since the early 2000s.

Fig. 3.

Temporal changes in concentrations of selected PFAS in North Atlantic pilot whales (G. melas). Panels show contrasting temporal trends among the C8 FASA (A and B), C10 PFCA (C and D), C8 PFSA (E and F), and C4 FASA (G and H). Liver and muscle samples include 16 paired samples collected between 2015 and 2023, but otherwise reflect different individual whales. The solid line and colored shading show the expected mean (posterior expected value) and 95% CI of Bayesian change-point analyses (Methods). All trends are statistically significant, with 95% CI that do not include zero. Change point years are marked with vertical lines, showing the posterior mean and 95% CI. One sample below detection is shown as an open circle and has been replaced by the method detection limit divided by the square root of two (MDL/$\sqrt{2}$).

In contrast, tissue concentrations of the shortest chain length FASA appear to be monotonically increasing. Between 2001 to 2023, concentrations of the C4 FASA increased by approximately ~7% per year (Fig. 3 C and D), similar to rates previously observed for long-chain FASA and PFCA (22). The C5 FASA exhibited no clear trend in the liver (2001 to 2023) but showed increasing concentrations over the longer sampling period for muscle tissues collected between 1994 to 2023 (SI Appendix, Fig. S2). No other short-chain PFAS were detected in this study, and no other detected PFAS showed increasing concentrations. These unique patterns likely reflect a combination of ongoing commercial production of short-chain FASA, as well as their rapid atmospheric transport to the pelagic North Atlantic Ocean.

Other studies have reported increases in short-chain FASA in coastal beluga whales, including monotonically increasing trends for C4 FASA but no clear trends for C5 FASA from 2000–2017 (34). Limited information is available on production trends for the C5 FASA, but C4 FASA was produced as a replacement for PFOS in the early 2000s (35). Recent studies indicate that C4 FASA remains widely used, particularly in industries such as semiconductor manufacturing, and its production may continue to grow (35).

Terminal PFAA Show Decadal-Scale Lags Following Production Shifts.

We observed a similar trajectory to EOF concentrations (Fig. 1A) and a predictable lag based on the perfluorinated carbon chain length for the legacy perfluoroalkyl acids (PFAA, including PFSA and PFCA, e.g., C8 PFSA: PFOS and C10 PFCA: PFUnDA; Fig. 3 C and D and SI Appendix, Fig. S3 B and C). For example, PFUnDA peaked between 2013–2016 followed by a decline of 13 to 17% per year (Fig. 3 C and D and SI Appendix, Tables S2 and S3). PFOS similarly peaked in muscle and liver between 2017–2020 and declined sharply by 11 to 29% per year after this time period (Fig. 3 E and F). The earliest declines were exhibited by the compounds with the longest perfluorinated carbon chains, with differences in the change points predicted across the perfluorinated carbon chain lengths spanning approximately a decade. For example, the C13 PFCA began to decline around 2011–2014, while the C8 PFCA began to decline in 2020 (SI Appendix, Figs. S3 and S4). In contrast, the C6 and C7 PFSA did not show significant temporal changes across the study window (SI Appendix, Fig. S2 and Table S3).

Contrasting temporal trends among targeted PFAA likely reflect the combined effects of shifts in chemical production and chain-length-dependent patterns in oceanic fate and transport. For example, the lack of significant trend observed for the C6 PFSA likely reflects a combination of phase-outs by major manufacturers in the early 2000s as well as ongoing industrial production of C6 perfluoroalkane sulfonyl fluoride (PHxSF) and its derivatives, including C6 FASA and C6 PFSA, at a small number of manufacturers within the last decade (24, 36). Observed declines in the C8-C13 PFCA, PFOS, and the C10 PFSA reflect approximately 10 to 20-year time lags from their peak production in North America and Europe in the early 2000s (24). By contrast, in coastal ecosystems declining biological concentrations of the same legacy PFAS have been reported almost immediately following shifts in chemical production (34). It is therefore likely that the decadal-scale lag observed in the subarctic North Atlantic reflects the timescales required for oceanic transport of environmental releases and food web uptake (30). Additional comparisons of trends observed among marine ecosystems are discussed in SI Appendix, section 1.2.

Earlier peaks in concentrations for PFAS with the longest perfluorinated carbon chain lengths (Fig. 3 and SI Appendix, Figs. S2 and S3) may reflect relatively more efficient vertical scavenging by settling particles in seawater to depths most relevant for pilot whale foraging (400 to 700 m) compared to the shorter-chain-length PFAS. Using published literature data on organic carbon fluxes (37) and vertical transport velocities (38, 39) (SI Appendix, section 1.3), we estimated that vertical PFAS transport in the North Atlantic upper water column (top 1,000 m) by particle sinking in seawater is the dominant transport process (>50% of vertical flux) for PFCA with greater than nine perfluorinated carbons. Deep water formation (convective jets) transports surface waters containing PFAS throughout much of the Northeastern Atlantic to the mesopelagic and deep ocean leading to the detectable presence of both short- and long-chain PFAS throughout the vertical water column (1, 40). In addition, PFAS partitioning to suspended particles for the longest carbon chain-length compounds may accelerate vertical transport and uptake into the pilot whale food web (41,42, 43). Particle-associated vertical transport is therefore likely an important contributor to shifts in the decadal timescales of long-range oceanic transport for long-chain PFAS.

Oceanic Transport Drives Decadal-Scale Lags.

Fig. 4A shows historical chemical releases from Wang et al. for the targeted PFAS that account for >75% of the sum of targeted PFAS in pilot whale muscle and liver across all years (23, 24). These include FOSA (C8 FASA), PFOS (C8 PFSA), and the C10 and C12 PFCA (PFUnDA and PFTrDA). All four compounds show substantial declines in production in North America and Europe after the year 2000, with more sustained production for the C10/C12 PFCA between 2010 and 2020 compared to FOSA and PFOS (Fig. 4A). Peak production years across the four compounds occurred between the mid-1990s and early 2000s. Additional details on the production data shown in Fig. 4A are provided in the Methods.

Fig. 4.

Comparison among PFAS production trends, seawater PFOS concentrations, and EOF concentrations in the North Atlantic pilot whale liver. (A) shows annual production of key PFAS detected in pilot whale liver and muscle (Fig. 2), including the eight perfluorinated carbon chain (C8) perfluorosulfonic acid (PFSA): PFOS, the C8 perfluorosulfonamide (FASA): FOSA, the C10 perfluorocarboxylic acid (PFCA): PFUnDA, and the C12 PFCA: PFTrDA. PFAS release data are from Wang et al., for “country group I” between 1958 and 2027 and reflect geometric mean of the high and low “plausible” scenarios (SI Appendix, section 2.8) (23, 24). (B) shows modeled peak PFOS concentrations (ca. 2010) for an areal cross-section of the North Atlantic Ocean [adapted from Zhang et al. (30)], where the core distribution area of North Atlantic pilot whales is outlined by the white rectangle (44) (SI Appendix, Fig. S6). (C) shows the modeled trajectory in seawater PFOS concentrations between the surface and 720 m depth, which approximately the foraging depth of pilot whales (predominant prey: squid, cod, blue whiting) (21). The solid line shows the modeled median concentrations and gray shading shows 95% confidence interval. (D) compares the sum of PFAS production for >75% of the PFAS detected in pilot whale tissue to the temporal trajectory of EOF concentrations in pilot whale liver samples from this study.

We compared production curves to modeled trends in seawater PFOS forced by environmental release data (23, 24, 30) across the known foraging territory of the North Atlantic pilot whales included in this study (Fig. 4 B and C). Peak modeled seawater PFOS concentrations occurred ca. 2010 (Fig. 4B), predominantly reflecting the lag time needed for transport from the Western Atlantic Ocean in the North Atlantic current. Qualitatively, the modeled decline in seawater PFOS concentrations using a 3-D ocean transport model (MITgcm) is similar to the temporal trajectory measured in pilot whale tissues (Figs. 4C and 3 E and F) (30). Peak modeled PFOS concentrations in seawater occurred within a year of the peak in EOF, although the modeled seawater decline (3.2% per year) was shallower than the observed trend. The observed whale muscle and liver PFOS concentrations peaked between 2017–2020 and declined by 11 to 29% per year after this time period. Differences between modeled seawater concentrations and temporal trends observed in whales likely reflects uncertainty in chemical release data as well as potential lags between terrestrial releases and transport to coastal environments that are not captured by the model (Methods).

A decadal scale lag between chemical production and EOF concentrations in whale tissues are indicated by the comparison between cumulative production of the four major PFAS and temporal trends in EOF concentrations (Fig. 4D). Modeled seawater trajectories in the MITgcm are not available for other PFAS but, as discussed above, the long-chain PFCA (approximately C10 and longer) likely have accelerated vertical transport in the water column due to their higher propensity for partitioning to suspended particles (K_d ≥ 4.2) compared to PFOS (K_d = 3.7) (41, 45). This helps to explain an earlier decline in the C10/C12 PFCA compared to PFOS in pilot whale tissues. FOSA exhibits the earliest decline in pilot whale tissues among these four PFAS (Fig. 3 A and B) and most closely resembles chemical production due to its important atmospheric transport pathway (Fig. 4A). Overall, the observed time lags between PFAS production and bioaccumulation can be largely explained by environmental transport processes.

To characterize the role of trophic transfer and bioaccumulation on any additional lags between production and bioaccumulation, we parameterized an existing food-web bioaccumulation model to represent the pilot whale food web (46, 47). Modeled PFOS half-lives in juvenile males were only one year, while smaller, lower trophic level organisms such as squid and fish had even shorter half-lives, ranging from days to weeks. This is much faster than the observed decadal time lag between emissions and concentrations in pilot whales (Fig. 4d), emphasizing the rapid equilibrium established for bioavailable PFAS in seawater and aquatic organisms. Longer-chain PFAS generally exhibit longer biological half-lives, but we observed a shorter lag time between peak emissions and concentrations in whales, again suggesting this is not driven by food-web processes. We infer from these results that environmental transport processes are the main driver of the lag time between PFAS emissions and pilot whale tissue concentrations rather than food web bioaccumulation processes. This contrasts the timescales required for uptake, trophic transfer, and elimination of other persistent organic pollutants and methylmercury that can require a decade or more for changes in exposure to be fully reflected in top predators (48, 49).

Suspect PFAS Show Declining Temporal Trends.

Using high-resolution mass spectrometry (Methods), we identified 16 additional PFAS in pilot whale liver, based on a list of 50 compounds previously identified by nontargeted analysis of marine mammal tissues (17, 19, 34, 50) (SI Appendix, Table S4 and Fig. S5). The suspect PFAS identified were all long-chain compounds (C6–C15) from five distinct classes, including two classes already partially assessed by targeted PFAS analyses in this study (PFCA and n:3 fluorotelomer carboxylic acids, n = 7), as well as three additional classes of H- and Cl-substituted PFAA (n = 11) (SI Appendix, Table S9) (51, 52).

Bayesian change point analyses of the suspect PFAS peak areas revealed that most of the suspect PFAS exhibited declining peak area abundance in recently measured samples (SI Appendix, Fig. S5), while none showed statistically increasing peak areas in recent years. The observed trends follow monotonic or single change-point patterns, similar to those observed for targeted legacy PFAA and FASA, suggesting similar production patterns and transport pathways. Additional discussion of these patterns across classes and chain lengths can be found in SI Appendix, section 1.4.

Notably, none of the ether-linked or cyclic PFAS previously detected in marine mammals and in Arctic environments were detected in this study (3, 17, 19). Similarly, none of the fluorotelomer sulfones that were found to comprise a large fraction of EOF in East Greenland killer whale blubber were detected in preliminary measurements of three pilot whale blubber samples from 2023 measured in this study (SI Appendix, section 2) (50). Overall, these data suggest biologically available concentrations of most suspect PFAS identified in this study are decreasing and are consistent with trends observed for most of the targeted PFAS.

Implications for Future Contamination.

We hypothesized that recent global increases in novel PFAS production not captured by targeted analysis might lead to an increase in EOF concentrations in biological tissues (22). By contrast, our results show declining trends since the early-to-mid 2010s for EOF, summed targeted PFAS, and some suspect PFAS in North Atlantic long-finned pilot whales over the past three decades. FOSA and PFOS have remained the dominant PFAS in pilot whale tissues since the onset of chemical production. Less than 30% of EOF was unidentified, and preliminary suspect screening analyses did not identify any novel or short-chain PFAS with increasing trends. These findings emphasize the success of voluntary and regulatory efforts that phased out production of many particularly bioaccumulative PFAS in Europe and North America in the early 2000s.

Among the replacement and suspect PFAS measured in this study, only short-chain (C4-C5) FASA were observed to increase continuously over the study period between 1987 and 2023. This reinforces concerns about the bioaccumulation potential of short-chain FASA (28). A prior study of coastal beluga whales from the St. Lawrence Estuary in North America reported increases in short-chain FASA, short-chain PFAA, and H-PFCA through 2017, but in this study, we find only short-chain FASA were both detectable and increasing in North Atlantic pilot whales by 2023. The C4 FASA has also been detected in other remote marine environments, highlighting its long-range atmospheric transport potential (3, 19). Past work shows a high bioaccumulation potential for short-chain FASA that is comparable to PFOS (28). Furthermore, recent work has found that sulfonamides, including the C4 FASA, are similarly or more potent developmental toxicants than PFOS and long-chain PFCA (53, 54). Increasing short-chain FASA accumulation in remote food webs underscores the need for additional monitoring in wildlife and associated human and animal toxicity assessments across FASA chain lengths.

Despite the increasing presence of short-chain FASA, overall accumulation of shorter-chain or novel “replacement” PFAS in North Atlantic pilot whales has remained low. The observed decline of EOF in these whales, in contrast to stable or increasing trends in humans, suggests fundamental differences in exposure pathways. Humans typically experience continuous, direct exposure to replacement PFAS through consumer products, leading to more rapid and sustained accumulation, even for PFAS that are less bioaccumulative. In pelagic marine environments, environmental transport processes slow down and dilute wildlife exposures to replacement PFAS, particularly for shorter-chain alternatives. For example, diffuse exposure to short-chain PFAS in marine environments may result in low accumulation levels that are presently below detection. Transformation of neutral short-chain PFAS precursors to less bioaccumulative acids during transport could also contribute to lower overall bioaccumulation in wildlife compared to humans (28).

Greater retention of novel PFAS in terrestrial environments can also contribute to low levels of accumulation in marine mammals. This hypothesis is supported by previous evidence for limited novel PFAS accumulation in both subarctic and coastal cetaceans, as indicated by semiquantitative and nontargeted analyses identifying FASA and legacy FTCA as the predominant contributors to unknown EOF (17, 19). Legacy PFAA are mainly strong acids, facilitating transport in surface waters and groundwater. In contrast, more neutral novel PFAS precursors can be more strongly sorbed to sediment and soils (55, 56). Some FASAmay persist in terrestrial soil and groundwater reservoirs for centuries due to their strong retention in soils and slow microbial biotransformation (57, 58). As PFAS production has shifted away from PFAA, a greater number of novel PFAS chemistries could be retained near contaminated sites, a hypothesis that requires further investigation. Centurial-scale retention of alternative PFAS in terrestrial ecosystems would substantially delay the response of marine ecosystems to shifts in chemical production.

These processes could also lead to long-term persistence and greater cumulative exposures in terrestrial ecosystems, potentially surpassing that of legacy PFAS. Recent evidence suggests unexpected trophic magnification of short-chain (< C6) PFAA in terrestrial wildlife. This may reflect short-chain precursor contributions to elevated PFAS accumulation in the upper trophic levels of these ecosystems (59). Together, this underscores the need for expanded monitoring of terrestrial ecosystems using a comprehensive suite of PFAS analytical tools to better capture the organofluorine mass budget.

Methods

Pilot Whale Tissue Sample Collection.

Staff at the Faroese Environment Agency and the Faroe Marine Research Institute routinely collect and archive tissue samples from North Atlantic long-finned pilot whales (G. melas). Tagged tracer data for pilot whales in the Faroe Islands show they have high site-fidelity within a constrained geographic area in the North Atlantic (SI Appendix, Fig. S6). Stomach contents of pilot whales indicate they have a simple diet that predominantly consists of European flying squid (Todarodes sagittatus) and blue whiting (Micromesistius poutassou) (21, 44).

We subsampled liver and muscle tissues from available juvenile male whales (individuals < 494 cm in length) to limit variability due to calving and age (21), following the same approach for sample selection as Dassuncao et al. (22). We followed all ethical protocols and international conventions for species protection (National Marine Fisheries Service Permit No. 26689). Frozen tissues (−20 °C) were subsampled, shipped frozen, and transferred to similar storage conditions at Harvard University prior to analysis.

This study leverages the data compiled by Dassuncao et al. for muscle tissues between 1986–2013 (n = 87) and includes additional liver and muscle samples analyzed for this work (22). We obtained muscle subsamples from 46 whales collected between 2015 and 2023, resulting in a total of 133 muscle samples. Liver samples were collected from 38 pilot whales between 2001 and 2023. The Supporting Information (SI Appendix, Table S5) contains additional information on sample collection (sampling date, location, and agency). Paired liver and muscle samples were obtained from 16 whales, and all other samples are from different individuals.

Tissue Sample Extraction.

Tissue samples were extracted following previously published methods (60). Frozen tissue samples were homogenized using an OMNI International TH homogenizer and 1 to 1.5 g of the wet weight tissue was subsampled for extraction. Extractions followed similar procedures for targeted PFAS and EOF but muscle samples intended for targeted analysis were initially spiked prior to extraction with an isotopically labeled PFAS mixture as an internal standard for quantification (EIS) at concentrations ranging from 1,500 to 1,800 ng/L (13, 60). Tissues were extracted with acetonitrile and bead blending, followed by offline weak anion exchange (WAX) solid phase extraction (SPE). The eluent was further reduced to near-dryness and reconstituted in methanol as the final extract (60).

For liver and blubber samples extracted for organofluorine analysis, similar methods were used, but SPE-WAX extractions included an additional clean-up step using 0.1% NH4OH in Milli-Q to remove inorganic fluorine. An aliquot of the final extract was spiked with internal standard after extraction and before quantification of individual PFAS compounds. The remaining extract was directly measured for EOF. Additional details of sample extraction methods are available in the SI Appendix, section 2.3.

EOF Analysis.

We analyzed EOF in the liver rather than muscle because liver tissues have higher concentrations and therefore more reliably detectable levels. We additionally measured EOF in a small number of blubber samples to test for further evidence of elevated blubber EOF, which was previously observed in killer whales (18, 50). All samples were analyzed by combustion ion chromatography (CIC) using a Metrohm CIC with combustion unit from Analytick Jena (Jena, Germany), 920 Absorber Module, and 930 Compact IC Flex ion chromatograph from Metrohm (Herisau, Switzerland) at Harvard University. Instrumental analyses followed previously described protocols (13, 60). Method detection limits (MDLs) were calculated as the average plus three times the SD of procedural blanks, adjusted by the dilution factor (61). We evaluated removal of inorganic fluorine using sodium fluoride spikes (98.1 to 100% removal) and organofluorine recovery using PFAS mixture spikes (91.2 to 95.0% recovery) in procedural and matrix samples. Details on blanks, sample replicates, and spike recoveries are provided in the SI Appendix, section 2.4.

Targeted PFAS Analysis.

We analyzed muscle and liver tissue samples included in this study for a suite of 28 targeted PFAS (SI Appendix, Table S7). Prior work by Dassuncao et al. included 15 of these targeted PFAS for muscle samples, including data from 1986–2013 for 6 PFAS (C8 perfluoroalkyl sulfonic acid (PFSA); C8-10, C12-13 perfluoroalkyl carboxylic acids (PFCA) and from 1994–2013 for 9 PFAS [C4, C6-7, C10 PFSA; C5-C7 and C11 PFCA; C8 perfluoroalkyl sulfonamide (FASA)] (22). We selected a subset of 24 of the 87 historical muscle samples that were previously analyzed between 1994 and 2013 to obtain comparable data for the additional 13 targeted PFAS included in this work (C5 and C9 PFSA; C3-C4 PFCA; C4-C6 and C10 FASA; ammonium 4,8-dioxa-3H-perfluorononanoic acid: ADONA; C4, C6 n:2 fluorotelomer sulfonic acid (FTSA); C3, C5 n:3 fluorotelomer carboxylic acid (FTCA).

Samples were analyzed at Harvard University using a Vanquish Flex ultrahigh-performance liquid chromatograph (ThermoFisher, U.S.) coupled with a quadrupole Orbitrap Exploris 120 MS (ThermoFisher, U.S.) in ESI-mode (UHPLC-HRMS). Data were acquired using full scan MS1 mode with resolution of 60,000 and a scan range of 200 to 800 m/z and data-dependent analysis mode with resolution of 30,000 for MS2 confirmation (SI Appendix, Table S8). For PFAS without matched internal standards, the standard closest in retention time was used for quantification.

MDLs were calculated as either the lowest detectable calibration concentration value multiplied by the dilution factor, or when method blanks had detectable concentrations, the average detectable procedural blank concentration plus three times the SD of the detectable blanks (17). Only values > MDL are reported (SI Appendix, Table S9). Matrix and procedural spikes were analyzed alongside direct measurement of the native spike that was not subjected to any extraction procedures. Spike recoveries are reported in comparison to the native spike measurement rather than the nominal spike concentration. Spike recoveries were within 100 +/- 30% for all analytes except some FASA, which were slightly higher (SI Appendix, Table S12). Details on instrument parameters, blanks, sample duplicates, standard reference materials, and procedural and matrix spike recoveries are provided in SI Appendix, section 2.5.

Suspect Screening Analysis.

We focused suspect screening analyses in pilot whale liver tissues on 50 PFAS previously identified using nontargeted and suspect screening analyses of marine mammal tissues (17, 19, 34, 50). Other PFAS within the same homologous series as previously identified compounds were identified by Kendrick mass defect. Temporal trends were evaluated for identified compounds based on detected peak area. Confidence levels were assigned based on the Schymanski and Charbonnet scales (51, 52). Additional details about sample analysis and suspect identification are provided in SI Appendix, section 2.6.

Statistical Analyses.

We compared liver EOF concentrations to the sum of targeted PFAS measurements measured (without EIS added prior to extractions) to assess the proportion of unidentified organofluorine in each sample. Compounds with detection frequencies below 50% were not included in the targeted PFAS fluorine equivalent totals. This has little influence on results because the sum of the 13 compounds in this category, when detected, made up <1% of the total targeted PFAS. We compared measured EOF concentrations across samples and the sum of targeted PFAS (as fluorine-equivalents) using linear regression and used the mean and 95% CI of the slope to estimate the fraction of unidentified organofluorine as a portion of the EOF in each year and across all years.

We omitted targeted PFAS with detection frequencies below 75% from summary statistics and temporal trend analyses (SI Appendix, Table S13). For compounds with >75% detection frequencies, we imputed values below the MDL using simple imputation (MDL/√2). Outliers were removed prior to statistical analyses based on iterative Grubb’s tests across 5-year moving windows (p < 0.05). All statistical analyses were performed in R version 4.2.2 and RStudio version 2021.09.2.

We analyzed temporal changes in tissue concentrations using multiple change point regression models with Bayesian inference (R package “mcp”) (62). We log-transformed concentration data and fit piecewise linear regression models that force adjacent linear segments to meet at a change point. We chose this statistical approach to better characterize uncertainty in the timing of shifts in concentrations from increasing to decreasing following changes in chemical production. Baseline models were run for all trend analyses using Gaussian likelihoods, uniform priors for the change point parameter across all possible change points, and broad uniform priors for the model intercept and slope parameters to exclude extreme estimates (SI Appendix, section 2.7 and Table S14). For each linear segment of the best-selected models, we considered a trend to be significantly increasing or decreasing if the 95% CI of the slope parameters did not include zero. Here, the 95% CI, more precisely referred to as a credible interval, is calculated as the highest posterior density interval, or the shortest interval that contains 95% of the posterior probability mass.

Because we do not have strong prior expectations about the number of change points in our data, we account for uncertainty in best fit model structure by comparing models with zero or one change point for all trend analyses. In general, only models for which all parameter estimates had converged (Gelman–Rubin statistic < 1.15) were included in the model comparisons. To evaluate the best model structure, we used leave-one-out cross validation to characterize and compare model fits, following an analysis analogous to a t test (63, 64) (R package “loo” version 2.8.0). Additional details about model selection are presented in SI Appendix, section 2.7.

Ocean and Bioaccumulation Modeling.

Seawater C8 PFSA (PFOS) concentrations were simulated in the North Atlantic Ocean using the MITgcm general circulation model, following an adaptation of the model developed by Zhang et al. (30). Dissolved and particle bound PFOS transport were simulated following continental discharges from 1958 to 2027. We estimated total annual emissions by calculating the geometric mean of the high and low “plausible” scenarios presented for country group I in Wang et al. (24). We estimated continental discharges using wastewater discharge data, following previous findings that wastewater discharges accounted for the majority of C8 PFSA releases to the environment (30, 65). We revised previous methods for the spatial and temporal distribution of estimated continental discharge using newly available wastewater discharge data from the global HydroWASTE database (66), PFAS measurements in wastewater discharge (67), and riverine discharge data from the global HydroSHEDS database (68). We introduced an additional 5-y time lag to published emissions trends using a simple box model, which improved model fit to coastal measurements made after 2012. We approximated releases by distributing the total mass of estimated continental releases proportionally based on annual wastewater discharge volume. Additional model details are provided in SI Appendix, section 2.8.

We adapted an existing mechanistic food web bioaccumulation model to simulate expected temporal trends of the C8 PFSA in the North Atlantic pilot whale food web using modeled ocean concentrations (46, 47). We parameterized a time-dependent model to predict annual whole-body tissue concentrations over time for five trophic levels: phytoplankton, zooplankton, fish, squid, and pilot whale (SI Appendix, section 2.9) (22, 46). Half-lives were calculated using modeled elimination rates for whales between 5 and 15 y of age.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information.