Evaluating Solid Phase Adsorption Toxin Tracking (SPATT) for Passive Monitoring of Per- and Polyfluoroalkyl Substances (PFAS) with Ion Mobility Spectrometry-Mass Spectrometry (IMS-MS)

Abstract

Detection and monitoring of per- and polyfluoroalkyl substances (PFAS) in aquatic environments has become an increasingly higher priority of regulatory agencies as public concern for human intake of these chemicals continues to grow. While many methods utilize active sampling strategies (“grab samples”) for precise PFAS quantitation, here we evaluate the efficacy of low-cost passive sampling devices (Solid Phase Adsorption Toxin Tracking, or SPATTs) for spatial and temporal PFAS assessment of aquatic systems. For this study, passive samplers were initially deployed in North Carolina along the Cape Fear River during the summer and fall of 2016 and 2017. These were originally intended for the detection of microcystins and monitoring potentially harmful algal blooms, though this period also coincided with occurrences of PFAS discharge from a local fluorochemical manufacturer into the river. Additional samplers were then deployed in 2022 to evaluate changes in PFAS fingerprint and abundances. Assessment of PFAS showed legacy compounds were observed across almost all sampling sites over all 3 years (PFHxS, PFOS, PFHxA, etc.), while emerging replacement PFAS (e.g., Nafion byproducts) were predominantly localized downstream from the manufacturer. Furthermore, samplers deployed downstream from the manufacturer in 2022 noted sharp decreases in observed signal for replacement PFAS in comparison to samplers deployed in 2016 and 2017, indicating mitigation and remediation efforts in the area were able to reduce localized fluorochemical contamination.

Graphical abstract

1. Introduction

The global proliferation of per- and polyfluoroalkyl substances (PFAS) is of increasing environmental concern due to the chemical persistence of these molecules in addition to recent associations of PFAS exposure with adverse health outcomes.(Blake and Fenton, 2020; De Silva et al., 2021) The carbon-fluorine bonds that comprise PFAS provide chemical resilience and promote their utility as durable surfactants in a wide range of household and industrial applications including non-stick cookware, water repellent clothing, and firefighting foams. However, these physiochemical properties also render PFAS resistant to degradation and promote bioaccumulation throughout the food web.(Houde et al., 2008; Suja et al., 2009) The C8 Science Panel and the National Health and Nutrition Examination Survey (NHANES) correlated PFAS exposure with increased levels of cholesterol, hypertension during pregnancy, and various cancers.(Barry et al., 2013; Darrow Lyndsey et al., 2013; Lewis et al., 2015) Findings from these panels and other studies(Kotlarz et al.) have spurred a wave of regulatory action from governing bodies to gradually phase out production of so called “legacy” PFAS, generally described as fully fluorinated structures comprised of repeating CF₂ units greater than or equal to 6 carbon atoms (C6).(Brase et al., 2021) These measures have resulted in decreased production for some compounds, however, they also spurred efforts to develop chemically similar replacements such as shorter chain alternatives, inclusion of ether linkages, and heteroatom substitutions (e.g. chlorine replacing fluorine).(Brase et al., 2021; Renner, 2006) Collectively these replacement compounds are often referred to as “emerging” PFAS, many of which are relatively uncharacterized from a toxicological perspective compared to legacy compounds.(Brase et al., 2021)

To streamline PFAS detection and quantitation, monitoring agencies are increasing efforts to establish unified sampling and testing methodologies. The predominant active sampling methods collect a known volume of water in the field (also known as “grab samples”, e.g. 250 mL of tap water for EPA Method 537.1),(Shoemaker and Tettenhorst, 2018) and concentrate analytes via solid-phase extraction prior to analysis by liquid chromatography coupled with mass spectrometry (LC-MS).(Chow et al., 2021; González-Barreiro et al., 2006) While collecting grab samples is highly informative with regard to quantitative assessment due to the defined sample volume, the data output is highly dependent upon the time and location of the instantaneous sampling period.(Coes et al., 2014) Passive sampling methods present a complementary approach that provides a “signal average” of chemical flux at the sampling location over the duration of the deployment period.(Coes et al., 2014) This signal averaging can help detect sharp instances of high/low chemical concentration events such as spills or short manufacturing outfalls that may be missed with discrete grab samples. In addition, passive samplers are temporally integrative through continual adsorption of target species over the duration of the sampling period (assuming equilibrium has not been achieved) and hence these methods can function as a pre-concentration mechanism to increase chemical abundance prior to analysis, which is of particular importance for relatively low-concentration pollutants such as PFAS.(Coes et al., 2014; Pétré et al., 2022) However, passive sampling methods also possess their own analytical challenges. For instance, quantitative assessment of aquatic passive samplers is extremely challenging due to several experimental factors including measurement and consistency of water flow through the sampler, pH variation, analyte degradation due to environmental factors such as temperature fluctuations or UV exposure, and uncharacterized equilibrium kinetics for analyte interactions with the sampling medium.(Alvarez, 2010; Kamali et al., 2022) For many stationary phases the sampling rate (RS) and sample-water partition coefficient (K_sw) measurements pertaining to the interaction of the resin with PFAS are uncharacterized, and as such pose a significant hurdle towards quantitative assessment. Several passive sampling devices commonly utilized for environmental profiling include the polar organic chemical integrative sampler (POCIS)(Alvarez et al., 2004) and the semipermeable membrane device (SPMD)(Petty et al., 2000; Vrana et al., 2005), with several studies devoted specifically to PFAS profiling utilizing a variety of stationary phase resins.(Kaserzon et al., 2012; Kaserzon et al., 2013; Lai et al., 2019; Li et al., 2016) A similar passive sampling device termed “solid phase adsorption toxin tracking” (SPATT) was developed in 2004 by MacKenzie et. al.(MacKenzie et al., 2004) to monitor algal toxins associated harmful algal blooms (HABs). SPATTs quickly garnered popularity in HAB monitoring studies due to their propensity to pre-concentrate toxins prior to analysis, low cost and ease of use.(Anderson et al., 2023; Howard et al., 2021; Pizarro et al., 2013; Roué et al., 2018; Zendong et al., 2016) One such study was performed by our research team to evaluate microcystin concentrations at four sampling locations along the Cape Fear River basin in North Carolina during the summer and fall months of 2016 and 2017 (Figure 1).(Kirkwood-Donelson et al.; Kotlarz et al.; Wiltsie et al., 2018) The sampling area includes sites upstream and downstream of a local fluorochemical manufacturer, with previous reports indicating PFAS discharge occurred during the sampling period.(Hopkins et al., 2018; Sun et al., 2016) The manufacturer implemented emission controls to mitigate these outcomes, which has been documented in previous reports.(Hopkins et al., 2018; Sun et al., 2016) Noting these observations, the objective of this study was to address the following questions:

1. Is the stationary phase used in SPATTs able to bind PFAS?

2. If so, did the SPATTs deployed in 2016 and 2017 adsorb PFAS and what is their chemical profile?

3. Are there any spatial or temporal differences between the sampling sites across both years?

4. Noting several remediation efforts pertaining to the fluorochemical releases enacted since 2016 and 2017,(Kotlarz et al.; Zhou et al., 2022) would the PFAS profile differ half a decade later?

Fig. 1.

Spatial map of the Cape Fear River basin in North Carolina. Samples were collected at 5 total sites along the river, with red text indicating samples obtained during the summer and fall months of 2016 and 2017, and purple representing SPATTs redeployed during a similar timeframe in 2022.

2. Experimental

2.1. SPATT Construction and Deployment

SPATTs were constructed utilizing the “embroidery hoop” method following the framework set forth by Rundberget et. al.(Rundberget et al., 2009) and refined by Howard et. al.(Howard et al., 2018) Briefly, each sampling protocol consists of packing 3 grams of Diaion HP20 resin into Nitex bolt cloth and encasing the resin and cloth in a rubber embroidery hoop (Figure S1).(MacKenzie et al., 2004; Zendong et al., 2016) The HP20 resin (a styrene-divinylbenzene stationary phase, chemical structure illustrated in Figure S2) was originally selected due its strong binding efficiency for microcystins and other algal toxins as noted in previous studies.(Roué et al., 2018; Zhao et al., 2013) Samplers were conditioned with methanol to activate the resin, washed with water, and subsequently deployed for the duration of the sampling period, typically 2 weeks. The initial deployment of SPATTs occurred during 2016 and 2017 at four sites denoted in red font in Figure 1. Sampling locations for Jordan Lake, Lillington and Fayetteville are located upstream from a local fluorochemical manufacturer, while Kings Bluff is located downstream of the plant. Following an initial pilot study of SPATT profiling from a subset of the 2016 samplers, additional SPATTs were redeployed in 2022 at Kings Bluff as well as at a new location upstream of the manufacturer, at Buckhorn Dam (purple font in Figure 1). Precise sampling periods and geographical coordinates for each sampler are provided in Table S1.

2.2. PFAS Extraction Methods

Given that each SPATT could only be extracted once, the initial priority of this study was to devise a testing scenario to ascertain whether the HP20 resin used in SPATTs could adsorb PFAS. It was theorized that the hydrophobic nature of styrene-divinylbenzene resin would be sufficient to bind the hydrophobic CF₂ portions of most PFAS. To test this hypothesis, a “mock” deployment trial was conducted wherein 2 L of LC-MS grade water were added in a 3.87 L polypropylene jug and spiked with ¹²C PFAS (natural isotopic abundance, not heavy labeled) at a sum concentration of ~40,000 ng/L to mimic concentrations observed in the Cape Fear River during the original deployment period.(Hopkins et al., 2018) A visual schematic of this workflow is provided in Figure S3 and Figure 2. A SPATT sampler was submerged in the spiked water and a stir bar was added to mimic river flow with moderate agitation set to 200 revolutions per minute. Following a 24-hour incubation period, the SPATT was retrieved, and the resin was separated from the Nitex cloth before subsequent transfer to a weigh boat for airdrying overnight. The dry resin was transferred to a pre-weighed 15 mL Falcon tube weighed to obtain the mass of resin collected, and 7.875 mL of methanol and 125 μL of heavy labeled ¹³C PFAS extraction mixture were added (MPFAC-C-ES, Wellington Labs, diluted 1:10 in methanol). The resulting slurry was vortexed, and 2 mL of water were added along with subsequent vortexing prior to sonication for one hour. The extract was then syringe filtered (Whatman glass microfiber filters, 0.45 μm with polypropylene housing) and speed vacuum concentrated to dryness. Samples were reconstituted in 200 μL of 60/40 water/methanol with 2 mM ammonium acetate, centrifuged, and 180 μL of extract was transferred to LC vials for analysis. Results illustrated in Figure S3 showed that ¹²C PFAS from the spiked water were adsorbed by the HP20 resin and were successfully extracted using the described methods. This workflow was carried forward to analyze SPATTs deployed along the Cape Fear River for comparative analysis as visualized in Figure 2. The only modification to the above workflow for the deployed samplers was a reduction in internal standard volume (75 μL as opposed to 125 μL) due to detector saturation observed in the initial trial.

Fig. 2.

Illustrative depiction of SPATT passive sampler deployement, extraction, and analysis methods utilized in this study.

2.3. LC-IMS-MS Analysis

SPATT extracts were profiled for PFAS content using a combinatorial separation methodology combining liquid chromatography, ion mobility spectrometry, and mass spectrometry (LC-IMS-MS) separations as illustrated in Figure 2. These analyses were performed on an Agilent 1290 binary LC system in conjunction with an Agilent 6560 IM-QTOF platform which has been characterized previously.(Dodds et al., 2020; Dodds et al., 2017; May et al., 2014; Stow et al., 2017) In brief, extracts were sampled using a 2 μL injection volume and chromatographically separated on an Agilent Zorbax RRHD Eclipse Plus C18 column (2.1 x 50 mm, 1.8 μm I.D., PN#959757-902) with a matching guard column (PN# 821725-901). Mobile phase A was comprised of water and mobile phase B was comprised of 95/5 methanol/water; both solvents were buffered with 5 mM ammonium acetate. All solvents were Optima LC-MS grade purchased through Fisher Scientific (Hampton, NH). Following chromatographic elution, analytes were ionized using the Agilent Jet Stream source operated in negative polarity and mobility separated in the uniform-field drift tube operated in multiplexed mode prior to mass analysis.(Causon et al., 2019) The instrument was mass calibrated using the Agilent Tune Mix for mass measurement accuracy within 10 ppm prior to sample analysis and the mobility dimension was calibrated using the same solution as their CCS values have been reported previously.(Stow et al., 2017) A primary objective of this study was to characterize the qualitative composition of adsorbed PFAS to SPATTs during this period of chemical outflow. As such, each sample was injected in duplicate to first acquire non-target MS1 data for suspect screening and peak area comparisons and later profiled via the alternating frames MS₁-MS/MS fragmentation scheme developed by Kirkwood-Donelson et. al. for structural characterization as necessary.(Kirkwood-Donelson et al.) For extended LC-IMS-MS parameters, see Figure S4.

2.4. Data Analysis

Each sampler extract was profiled by LC-IMS-MS in conjunction with a method blank (methanol carried through the extraction workflow) as well as a clean SPATT sampler which was not deployed to act as an extraction blank (referred to hereafter as the “SPATT blank”). The resulting .d datafiles were demultiplexed using PNNL Preprocessor(Bilbao et al., 2022) (v 2022.02.18) and imported into Skyline(MacLean et al., 2010; MacLean et al., 2018) for suspect screening analysis using a target list of ~100 PFAS as detailed in Table S2. Utilizing complementary separation mechanisms of retention time, mobility, and mass filtering for LC-IMS-MS data performed in Skyline is highly effective for PFAS analysis and has been described in detail previously.(Foster et al., 2022; Kirkwood et al., 2022a; Kirkwood et al., 2022b; MacLean et al., 2018) Briefly, chromatographic windows were established for each PFAS in the target list, and only those molecules detected in the samples and not in the method blank or SPATT blank (PFBA, GenX, TFMS, and PFMBA) were carried forward through subsequent analysis. Peak areas were extracted in Skyline given an instrumental mass resolving power of 20,000 (typically corresponding to an isolation window of ~0.02 m/z bin width) with a mobility resolving power of 40. For a visual illustration of the mass and mobility filtering afforded by Skyline, see Figure S5. In total, 78 PFAS were detected containing both legacy and emerging compounds from PFAS sub-classes including sulfonic acids (PFSAs), carboxylic acids (PFCAs), ether carboxylic acids (PFECAs), Nafion byproducts, and several annotated compounds described in Kirkwood-Donelson et. al.(Dodds et al., 2020; Hopkins et al., 2018; Kirkwood-Donelson et al.) Notably, very short chain PFAS such as PFBA, TFA and similar polar analytes less than ~200 m/z were not observed in SPATTs, further work is required and perhaps do not possess sufficient non-polar character to adsorb to the HP20 resin. In the previous non-targeted study a subset of SPATTs were analyzed via the same NTA workflow focused on discovery of novel PFAS and characterizing their molecular structure.(Kirkwood-Donelson et al.) These molecules were included in the suspect screening list to evaluate these identifications and assess their correlations to known PFAS over the duration of the sampling period. Skyline results were uploaded to PanoramaWeb for public access and are available in Tables S3, S4, and at https://panoramaweb.org/spattanalysis.url. Peak areas were exported via the document grid as a .csv file and normalized against the weight of resin extracted for each sampler and the number of days deployed. An additional normalization strategy was considered which ignored the number of days deployed (normalizing only against weight of resin extracted) and the resulting figures are provided in the Supporting Information (Figures S6–S8). Clustering results and corresponding PFAS trends were consistent using both normalization methods, and as such the normalization approach accounting for variations in deployment period was used for subsequent analysis. Normalized data was formatted for import into MetaboAnalyst for subsequent log10 transformation, statistical analysis, and data visualization.(Chong et al., 2018; Pang et al., 2021) Sample classifications were defined by deployment location and year apart from Jordan Lake as only 1 sample was collected during 2016 and 4 from 2017, and hence were grouped as one collective.

3. Results and discussion

The goal of this study was to evaluate whether PFAS bound to the SPATTs and if so, assess PFAS changes at various locations. To perform these analyses, MetaboAnalyst was utilized for statistical interrogation of SPATT data using several methods. First, sample grouping was assessed using principal component analysis (PCA) as illustrated in Figure 3A. Sample grouping was primarily observed based on deployment location relative to the chemical manufacturer (upstream or downstream) and by year. For the 2016/2017 samples, SPATTs deployed downstream from the fluorochemical plant at Kings Bluff are spatially distinct in the PCA plot from sites located upstream (Fayetteville, Lillington, and Jordan Lake). Kings Bluff samples collected in 2016 are the most spatially isolated from upstream samples during the same deployment period, however as time proceeds the Kings Bluff samples collected in 2017 cluster closer to the upstream samples. The redeployed samples collected from Kings Bluff in 2022 clustered near samples collected upstream from the manufacturer, indicating less PFAS discharge from the manufacturer into the river. These observations are consistent with the hierarchical clustering heatmap in Figure S9. Noting consistency for samples collected in each year per location site, samples from each group were averaged (typically 5-7 samples per site for each year) in the hierarchical cluster for ease of visualization as illustrated in Figure 3B. PFAS profiles extracted from the SPATTs clustered from a sample-based perspective (horizontal tree) into two primary groups: (1) the 2016/2017 Kings Bluff samples and (2) upstream samples and 2022 Kings Bluff redeploys. From a compound-driven perspective (vertical tree in Figure 3B), detected PFAS profiles clustered primarily into emerging PFAS (principally Nafion byproducts and other ether-linked species) and legacy compounds. For legacy PFAS and other well-characterized compounds, molecules from distinct PFAS subclasses also tended to cluster together, such as fluorotelomers and polyfluorinated phosphates. Clustering tendencies were also noted for short and long chain PFSAs and PFCAs as highlighted in the brace callouts (e.g., PFBS, PFPeS, and PFHxS, all PFSA molecules comprised of C4-C6, located in close spatial proximity via clustering). These observations were conserved in the correlation diagrams in Figures S10 and S11. Pertaining to emerging PFAS and replacement compounds, structurally similar molecules also tended to group together. For example, Nafion byproducts 2, 4, and 6 cluster closely along with NVHOS and the recently detected compounds 2 and 3 as noted by Kirkwood- Donelson et al. in Figure 3B.(Kirkwood-Donelson et al.) Although not all features noted previously by Kirkwood-Donelson et al. were able to be definitively annotated with distinct chemical structures, the clustering and correlation methods utilized here indicate which compounds fluctuate in correlation with fully annotated PFAS. For example, while MS/MS spectra collected in the previous work was unable to definitively assign a molecular structure for a signal annotated as “Unknown 3,” (chemical formula, C₉H₂F₁₆O₆S), the hierarchical clustering heatmap and corresponding correlation diagrams (Figure S10 and S11) detailed in the current work illustrate that the relative abundance of this analyte fluctuates similarly to that of Compound 8 and Nafion byproduct 2 across sample groups (red asterisks in the hierarchical clustering of Figure 3B). The molecular structures of these two compounds share a large degree of structural similarity and the similar chemical formula noted for these two analytes and Unknown 3 (C₈HF₁₅O₆S, and C₇H₂F₁₄O₅S), which may suggest a common synthetic pathway (Figure 3C).

Fig. 3.

A) Principal component analysis (PCA) of SPATTs deployed in the Cape Fear river from 2016/2017 as well as 2022 at the 5 locations illustrated in Figure 1. B) Sample-averaged hierarchical clustering heatmap of SPATTs deployed in this study. Callouts are provided to illustrate sampling locations upstream and downstream of the fluorochemical manufacturer, and clustering tendencies for legacy and emerging compounds. C) Highlighted structures for Nafion byproduct 2, Compound 8, and Unknown 3 which grouped during clustering and correlation analysis and respresentative structures when avaliable.

Analysis of the SPATT data through both PCA and hierarchical clustering suggests that PFAS contamination, particularly for emerging compounds such as Nafion byproducts, has decreased substantially from mid- 2017 and subsequently to 2022; the manufacturer started to collect process wastewater in the second half of June 2017. This information can also be observed in a bar chart focusing on two representative compounds from both the legacy and emerging species. Figure 4 displays this information for PFOS (A) and Nafion byproduct 2 (B), wherein both peak areas were normalized per gram of resin collected per day deployed. While the abundance of PFOS does fluctuate over time, it is consistently observed both upstream and downstream of the fluorochemical plant across all sampling periods. Downstream abundances were lower than upstream abundances for the same time periods because river discharge at the downstream location was 1.4-2.7 times that of the upstream location (see Table S5). Similar trends were noted for other legacy compounds such as PFHxS, PFHxA, PFNS, and 6:2 FTS, among others. Legacy PFAS such as PFOS and PFOA have been produced since the 1940s; even though they are no longer produced or used in the U.S., the data highlight that they continue to occur in the Cape Fear River, likely as a result of inputs from contaminated groundwater, landfill leachate, and or precursor degradation. In contrast, widescale production of Nafion began as recently as the mid-2000s.(Kim et al., 2015) Nafion byproduct 2 is predominantly observed downstream of the manufacturer in 2016 and 2017 samples, with a marked decrease in abundance noted for the 2022 redeploys. Average river discharge during the 2022 redeploys was similar in 2017 (1,530-1,630 ft³/s) and 2022 (1,320-1,760 ft³/s) suggesting a similar dilution factor in both years. Similar results were noted for Hydro-EVE, PFO4DA, and other Nafion byproducts. The fluorochemical plant is a known manufacturer of Nafion, and hence these observations were expected and are consistent with previous targeted (suspect screening) studies of the Cape Fear River.(Dodds et al., 2020; Hopkins et al., 2018; Kotlarz et al.; Zhou et al., 2022) The fluorochemical manufacturer has initiated several steps as a response to the North Carolina Department of Environmental Quality (NC DEQ) including reduction of process wastewater discharge since 2017,(Kotlarz et al.) installation of a thermal oxidizer in 2019,(Zhou et al., 2022) and construction of a retaining wall to prevent PFAS transport via groundwater to the river.(2023) In addition, in 2017 the manufacturer began exporting ~26,500 gallons of processing water per day to Texas for deep well injection as reported in an email to the Environmental Protection Agency and NC DEQ.(Garon, 2017; Hopkins et al., 2018) Taking these events into account, observations noted in this work suggest the Cape Fear river contains less PFAS specifically for emerging compounds in comparison to half a decade prior consistent with similar studies in the Cape Fear River.(Hopkins et al., 2018; Pétré et al., 2022)

Figure 4.

Normalized abundances for candidate molecules representing both legacy and emerging PFAS, here PFOS and Nafion byproduct 2, A and B, respectively. Labels are included to denote sampling sites upstream (Lillington and Buckhorn Dam) and downstream (Kings Bluff) of the nearby flurochemical manufacture along with corresponding sampling periods.

4. Conclusion

Overall, the results collected in this work demonstrate several principles. First, the HP20 resin commonly utilized in SPATTs can act as an effective stationary phase for passive sampling of PFAS in aquatic environments. This observation was noted in both the mock deployment trial and with samples obtained from the Cape Fear River during 2016 and 2017, a period which coincided with known PFAS discharge in the area. Samplers were redeployed in 2022 to assess the river’s relative contamination profile 5 years later, and the results collected herein suggest PFAS contamination in the river has been reduced for emerging compounds, though legacy PFAS remain persistent. The observed decrease in compounds associated with local fluorochemical manufacturing is consistent with previous targeted studies conducted in the Cape Fear River during this time.(Hopkins et al., 2018; Kotlarz et al.) We acknowledge that the HP20 resin utilized herein is predominately non-polar, and as such may struggle to absorb short chain polar PFAS (e.g. TFA, PFBA, etc.) and should be investigated further in subsequent studies with alternative resins (e.g. bare silica) given the breadth of research devoted to highly mobile polar PFAS.(Ateia et al., 2019; Zheng et al., 2023) While the results of this work are by no means quantitative, collectively they illustrate the unique propensity of passive sampling methods to preconcentrate low abundance species prior to analytical characterization for spatial profiling of pollutant monitoring. While grab sampling remains the gold standard in quantitative aquatic assessments, passive sampling devices such as those demonstrated herein can serve as a complementary sampling strategy for contaminant profiling. In addition, statistical approaches including clustering and correlation methods were used to associate unannotated MS signals with profiles of fully characterized contaminants, potentially providing an avenue for sample-driven data association of chemical origin for current unknowns and future environmental targets.

Highlights:

• Passive samplers provided spatial and temporal profiling of fluorinated chemicals

• Emerging contaminants were primarily observed downstream from fluorochemical plant

• Fluorochemical abundances have decreased over the past 5 years

Data availability

Skyline results were uploaded to PanoramaWeb for public access and are available in Tables S3, S4, and at https://panoramaweb.org/spattanalysis.url.