Identifying PFAS in Soil Samples

There are more than 15,000 per- and polyfluoroalkyl substances (PFAS), according to a chemicals database organized by the United States Environmental Protection Agency (EPA). PFAS are associated with numerous health risks, and analytical scientists around the world are investigating better ways to detect these synthetic chemicals in the environment and food.

Jonathan Zweigle, of the University of Tübingen, is one of these scientists. Zweigle is using non-targeted screening based on high-resolution mass spectrometry (HRMS) to detect PFAS in soil samples. HRMS allows scientists to determine the elemental and isotopic composition of a sample with high accuracy. LCGC International sat down with Zweigle to discuss his research and the latest in PFAS identification and analysis.

Why is non-targeted screening (NTS) based on HRMS your preferred method for analyzing PFAS in a variety of samples?

In our research, soil samples from different PFAS-contaminated sites were selected, which have a high sample complexity. Typically, at these soil contamination sites PFAS occur that cannot be determined by targeted analysis because there are usually no authentic reference standards available. Therefore, the use of HRMS detection is of high relevance to cover all ionizable PFAS by the measurement method. This allows a much wider analytical window and allows identification of previously unknown PFAS. This also applies to other sample matrices.

What results did this approach yield regarding highly fluorinated PFAS, and even PFAS that were fluorinated at slightly lower levels?

The mass to carbon ratio (m/C), where the carbon number C can be estimated from the HRMS raw data, allows the discrimination of molecules with a high content of heavy heteroatoms compared to hydrogen. While most natural compounds which are coextracted from soils have a rather low m/C, PFAS with a certain fluorine content have a much higher m/C than most non-PFAS compounds. The higher the fluorine content, the easier the separation which is further improved by applying the mass defect to carbon ratio (MD/C) as a second dimension. Depending on the sample matrix, we observed that also PFAS with a considerable H-content can be prioritized from a large fraction of other compounds. However, for compounds with very low fluorine atoms compared to H, (for example, only one CF3-group in a typical organic molecule) this approach does not work since these compounds are at the same location as compounds without fluorine.

What other PFAS-specific techniques did you use in conjunction with the non-targeted HRMS screening?

In our PFAS-NTS workflows we use a combination of several techniques for prioritization and identification. After feature detection and blank filtering, the MD/C-m/C approach is applied to the MS1 data as an initial filtering step to remove a large fraction of unlikely features (for example, 90%). The next step is the application of Kendrick mass defect (KMD) analysis to detect homologous series of suspected repeating units (CF2, CF2O). We further apply a prioritization to the MS2 data where all fragmentation spectra are searched for certain mass differences (∆CF2, ∆HF) and PFAS-diagnostic fragments. Finally, a suspect screening with different databases is applied to the prioritized features. Both accurate mass and isotope patterns are used to match the features with suspect lists. Spectra without a database hit are interpreted manually for structure elucidation.

Can you discuss, in general, the characterization of the 80+ PFAS you found in contaminated soils in southwest Germany? What were the concentrations of some of these chemicals?

We investigated two contaminated sites in Germany. At the site in mid Germany more than 70 PFAS were identified which could also be semi-quantified. The total estimated concentrations reached approximately 30 µg/g soil while the class of SF5-perfluorosulfonic acids made up the largest fraction. At this site mainly perfluorinated compounds and a low fraction of polyfluorinated precursors were identified. In contrast, in southwestern Germany, larger polyfluorinated precursors such a fluorotelomer alkyl phosphate esters (PAPs) and fluorotelomer mercapto alkyl phosphate esters (FTMAPs) act as a long-term source of perfluoroalkyl acids (PFAAs). Concentrations of PFAS were not assessed in this study.

How does your research add to previous discoveries in this subject area, and what will you be studying next regarding this topic?

Several PFAS were already known in these soils, however, these analyses are often restricted to a limited number of target analytes. Our research showed that a much larger number of PFAS with relevant concentration fractions are present at these soils which were overlooked by target approaches. In the future, my research will focus on complementary chromatography to cover an even wider polarity range of analytes which will be combined with semi-quantification approaches. Semi-quantification is urgently needed in NTS to provide the missing concentration values which are necessary for robust risk assessment and a wider use of NTS to assess environmental pollution.

Further Reading

1. Zweigle, J.; Bugsel, B.; Zwiener, C. Efficient PFAS prioritization in non-target HRMS data: systematic evaluation of the novel MD/C-m/C approach. Anal. Bioanal. Chem. 2023, 415 (10), 1791–1801. DOI: 10.1007/s00216-023-04601-1
2. Zweigle, J.; Bugsel, B.; Zwiener, C. FindPFΔS: Non-Target Screening for PFAS - Comprehensive Data Mining for MS2 Fragment Mass Differences. Anal. Chem. 2022, 94 (30), 10788–10796. DOI: 10.1021/acs.analchem.2c01521
3. Zweigle, J.; Bugsel, B.; Röhler, K.; Haluska, A.A.; Zwiener, C. PFAS-Contaminated Soil Site in Germany: Nontarget Screening before and after Direct TOP Assay by Kendrick Mass Defect and FindPFΔS. Environ. Sci. Technol. 2023, 57 (16), 6647–6655. DOI: 10.1021/acs.est.2c07969
4. Zweigle, J.; Bugsel, B.; Fabregat-Palau, J.; Zwiener, C. PFΔScreen - an open-source tool for automated PFAS feature prioritization in non-target HRMS data. Anal. Bioanal. Chem. 2024, 416 (2), 349–362. DOI: 10.1007/s00216-023-05070-2