Understanding the Electrospray Ionization Response Factors of Per- and Poly-Fluoroalkyl Substances (PFAS)

Abstract

Per- and polyfluoroalkyl substances (PFAS) are used extensively in commercial products. Their unusual solubility properties make them an ideal class of compounds for various applications. However, these same properties have led to significant contamination and bioaccumulation given their persistence in the environment. Development of analytical techniques to detect and quantify these compounds must take into account the potential for these properties to perturb these measurements, specifically the potential to bias the electrospray ionization (ESI) process. Direct injection ESI analysis of 23 different PFAS species revealed that hydrophobicity and PFAS class can predict the ESI overall response factors. In this study, a method for predicting the behavior of individual PFAS compounds, including relative retention order in chromatography, is presented which is simply based on the number of fluorine atoms in the molecule as well as the class of the compound (e.g., perfluroalkylcarboxylic acids) versus computational estimations (e.g., non-polar surface area and logP).

INTRODUCTION

Per- and polyfluoroalkyl substances (PFAS) have been produced in the US since their initial discovery at DuPont in 1938.[1] These compounds are characterized by long fluorocarbon chains with fluorine atoms at sites normally occupied by hydrogen atoms in alkyl chains. This substitution of F for H in the alkyl chain yields a unique chemistry that has been capitalized on to manufacture many consumer products. The C-F bond is considered one of the strongest bonds in organic chemistry due in large part to fluorine’s electronegativity, which is the highest of all of the elements and twice that of hydrogen at 3.98.[2] The low polarizability of fluorine also contributes to very low surface energies for fluorocarbon chains and results in low intermolecular cohesion forces and thus reduced viscosity and unusually high volatility (often much greater than their hydrocarbon counterparts despite their higher molecular weights [3]). In addition to this atomic substitution, modern PFAS are also typically noted for containing a hydrophilic functional group which, together with the fluorine substitution, bestows a unique solubility property on these compounds, making them both hydrophobic and lipophobic (hence their value to consumer product applications).

Although these unusual solubility properties can make studying these compounds quite difficult, much work has been performed to understand these compounds from a chemical perspective. This research [4–8] has predominantly focused on the chemical state of these compounds as they exist in the environment in order to predict their transport properties and fate, although others have investigated ways to better predict their inherent properties (such as octanol/water partition coefficients, logP). Work has been done to improve these computational predictions, with one report noting a strong correlation between logP and the number of carbon atoms. [9]

Although there has been extensive use of mass spectrometry (MS) to detect and quantitate these compounds via negative ion mode electrospray ionization (ESI), there are no detailed discussions regarding how these compounds’ unique properties may impact the analytical process. In ESI, analytes are ionized as charged solvent droplets containing analyte material to undergo numerous fission and evaporation events prior to mass analysis. [10–14] These electrospray droplets are composed of charges (in the form of anionic hydroxyl groups in negative ion mode) spaced roughly evenly across the droplet’s surface.[15, 16] Moreover, hydrophobic or surface-active molecules (like PFAS) are at the surface of the droplet while more hydrophilic molecules are in the bulk “solution” and are therefore less likely to be ionized via droplet surface charges thus leading to diminished ESI ion abundances. [17–19] Molecules with unusual solubility or solvent partition properties (i.e., those that are highly surface active), have the potential to exhibit unexpected suppression, ionization, and/or saturation effects in the ESI process. For example, the presence of a high number of surface-active molecules on the surface of a droplet could alter the colligative properties of said droplet and reduce surface tension, such that the droplet is more likely to subdivide earlier than usual and promote a higher signal abundance of these same molecules (Figure 1). In this study, we sought to understand the roles that PFAS class and hydrophobicity have on ESI response factors for compounds that either co-elute or elute adjacent to one another.

Figure 1:

A theoretical depiction of the surface interaction of perfluorobutanoic acid (PFBA, strong acid) and perfluoropentanoic acid (PFPeA, also a strong acid) using space filling models (van der Waals radii). At higher concentrations of PFAS the droplet surface would quickly saturate, thus the highly ionizable PFAS present on the surface no longer accurately reflects the bulk PFAS concentration due in part to the lack of attractive forces between these molecules but also due to potential competition with other PFAS that may be present.

MATERIALS AND METHODS

All experiments were performed on an Orbitrap Exploris 240 (Thermo Scientific, Bremen, Germany) incorporating a Thermo Scientific Vanquish LC system (Germering, Germany). Predicted logP and non-polar surface area (NPSA) values were generated from Chem3D version 20.0.0.41 (part of the ChemDraw Suite). LogP and NPSA values were estimated by using the CambridgeSoft Molecular Topology function in the Property Picker. NPSA values were calculated by subtracting the estimated polar surface area from the Connolly Accessible Area. 1H,1H,2H,2H-perfluorohexane sulfonate (4:2 FTS), 1H,1H,2H,2H-perfluorooctane sulfonate (6:2 FTS), 1H,1H,2H,2H-perfluorodecane sulfonate (8:2 FTS), 1H,1H,2H,2H-perfluorododecane sulfonate (10:2 FTS), N-ethyl perfluorooctane sulfonamido acetate (NEtFOSAA), N-methyl perfluorooctane sulfonamido acetate (NMeFOSAA), perfluorooctanesulfonamide (PFOSA), Nmethylperfluorooctanesulfonamide (MeFOSA), Perfluorobutane sulfonate (PFBS), perfluorohexanesulfonate (PFHxS), Perfluorooctane sulfonate (PFOS), perfluorobutyric acid (PFBA), perfluoropentanoic acid (PFPeA), perfluorohexanoic acid (PFHxA), perfluoroheptanoic acid (PFHpA), perfluorooctanoic acid (PFOA), perfluorononanoic acid (PFNA), perfluorodecanoic acid (PFDA), perfluoroundecanoic acid (PFUdA), perfluorododecanoic acid (PFDoA), perfluorotetradecanoic acid (PFTeDA), perfluorohexadecanoic acid (PFHxDA), and tetrafluoro-2-(heptafluoropropoxy)propanoic acid (HFPO-DA, Gen-X) were acquired from Cambridge Isotope Laboratories (Tewksbury, MA).

Chromatography and Direct Injection Method

A mixture of all 23 compounds, each at a concentration of 5,000 ng/L, were separated on a Phenomenex F5 Column (Torrance, CA) to determine the relative retention order of the PFAS compounds. A gradient (solvent A: water with 5% methanol, 0.1% formic acid, solvent B: methanol with 5% water, 0.1% formic acid) was run at 500 μL/min as follows: 0 min: 20% B, 7 min: 99% B, 8 min: 99%B, 8.01 min: 20% B, 10 min: 20% B. Binary mixtures (concentration of each compound = 2500 ng/L) were made up, based on the retention order, and were directly injected using an isocratic method run at 50:50 solvent A:B (solvent A: water with 5% methanol, 0.1% formic acid, solvent B = methanol with 5% water, 0.1% formic acid) at 500 μL/min. The mass spectrometer was run using a capillary voltage of (−)1500V, a sheath gas of 50 (arbitrary units), an aux gas of 12 (arbitrary units), a sweep gas of 0.5 (arbitrary units), and ion transfer tube temperature of 240 °C, and a vaporizer temperature of 300 °C. All peak areas were assessed using MS1 extracted ion chromatograms using exact masses of analytes. Data were integrated and tabulated using Skyline software [20] (version 20.2.1.454) developed by the MacCoss Lab at University of Washington.

RESULTS AND DISCUSSION

PFAS species are chemically diverse and exhibit both hydrophobic and lipophobic properties; this leads to them having some peculiar behavior in the ESI process. In general these compounds are expected to preferentially reside on the droplet surface, with the hydrophilic functional group partially dissolved in the droplet and the more hydrophobic fluorocarbon chains extending beyond the boundaries of the droplet. For example, Figure 1 shows a depiction of perflurorobutanoic acid (PFBA) which has limited intermolecular interactions competing for droplet surface space with perfluoropentanoic acid (PFPeA). The large dipole moment of the C-F bonds in these two molecules would (at the very least) result in a more dissociative relationship with the fluorocarbon chains from neighboring PFAS compounds and saturate the droplet surface at a relatively lower concentration. When multiple PFAS are present in the same droplet, there exists a potential for competitive occupation of this droplet surface, which would result in a surface occupancy (and therefore likely preferential downstream ionization) that is not indicative of the overall concentration of species present (Figure 1). In this theoretical depiction, even though PFBA and PFPeA are present in the initial droplet in equal proportions, the preferential surface occupation that PFPeA exerts means it is more likely to be ejected during Columbic fission. The greater hydrophobic properties of PFPeA (relative to PFBA) also mean it is more likely to be ejected from subsequent progeny droplets. Certain properties, such as logP and number of fluorines likely govern most of the ionization kinetics in this competitive scenario. This is a working model for the ESI of PFAS compounds.

This surface activity and lack of intermolecular forces is also possibly the reason why PFAS compounds have a much greater ionization efficiency compared to their hydrocarbon fatty acid counterparts. Fatty acids are difficult to ionize via ESI such that most work on them utilizes some form of derivatization.[21] As discussed earlier, this lower ionization efficiency is attributed to both their lower relative surface activity as well as the strong bonding of the head groups.

The chemical differences between hydrocarbon-fluorocarbon species are especially evident with regard to the octanol-water partition coefficient (P). The predicted logP values between hydrocarbons and their fluorocarbon counterparts differ at times by more than two units (Table 1). However, these values can be especially difficult for modern software to accurately predict and for this reason, a more universal indicator would be of value. Previously [9], researchers have noted a correlation between the number of carbons and logP values; however, this relationship is expected to be less accurate for some PFAS molecules that contain additional carbons (in the form of methylene, methyl or ethyl groups) without the augmentation of fluorocarbon properties (e.g., perfluoroalkyl sulfonamides, fluorotelomer carboxylic acids and fluorotelomer sulfonic acids). The number of fluorines correlates much better with the predicted logP than the number of carbon atoms (Figure 2). When a series of PFAS compounds (4:2 FTS, 6:2 FTS, 8:2 FTS, 10:2 FTS, NEtFOSAA, NMeFOSAA, PFOSA, MeFOSA, PFBS, PFHxS, PFOS, PFBA, PFPeA, PFHxA, PFHpA, PFOA, PFNA, PFDA, PFUdA, PFDoA, PFTeDA, PFHxDA, and HFPO-DA (also known as Gen-X) are plotted as a function of either number of fluorines or number of carbons versus their predicted logP, it can be seen that both properties yield correlated results. However, two compounds, in particular, do not correlate well when graphing the number of carbons against predicted logP values. These two compounds are the two perfluoroalkyl sulfonamides that were tested, NMeFOSAA (Figure 2, A) and NEtFOSAA (Figure 2, B). Additionally, the fluorotelomer sulfonic acids, 4:2 FTS, 6:2 FTS, 8:2 FTS, and 10:2 FTS (Figure 2, call-outs C, D, E, and F, respectively) poorly correlate with carbon number for PFAS compounds. These compounds contain hydrocarbon motifs that skew the carbon correlation with regards to solubility properties. Interestingly, the correlation when using non-polar surface area (NPSA) reverses and carbon number correlates better than fluorine number. Importantly, logP and NPSA are not experimentally measured values but predicted.

Table 1:

Comparison of logP values for matched hydrocarbon-fluorocarbon carboxylic acid molecules species as predicted by Chem3D software.

Hydrocarbon	logP	Fluorocarbon	logP
Butanoic acid	0.7231	PFBA	1.7075
Octanoic acid	2.9350	PFOA	4.7037
Dodecanoic acid	5.1684	PFDoA	7.6999

Figure 2:

A) Comparison of the correlation between the number of fluorines or number of carbons in a PFAS compound and the predicted logP. Arrows indicate those compounds that do not correlate as well for number of carbons vs. LogP (A- NMeFOSAA, B- NEtFOSAA, C- 4:2FTS, D- 6:2 FTS, E- 8:2 FTS, F- 10:2 FTS). B) Comparison of the correlation between number of fluorines or number of carbons in a PFAS compound versus non-polar surface area. In this case, number of carbon atoms is a better correlative than number of fluorine atoms.

Given that all PFAS species have a high surface activity (based on logP values) and they do not possess any electrostatic attractive forces due to the large dipole moment of the C-F bond, at higher concentrations it is predicted that saturation effects would quickly dominate this interaction to the extent that the highly ionizable PFAS present on the surface no longer accurately reflects the bulk PFAS concentration. In other words, saturation of the PFAS ion abundance (potentially resulting in a parabolic calibration curve) would occur long before other saturation events would occur, such as chromatography column overload or mass analyzer detector saturation. For this reason, when analyzing mixtures it is important to test for contribution to this saturation from co-eluting PFAS species.

In an effort to build our understanding of which species would “co-elute” we separated out the 23 PFAS compounds using reversed-phase chromatography. Figure 3 shows the overlaid extracted ion chromatograms with each compound at an identical concentration (5,000 ng/L). Inspection of these data demonstrate remarkably variable ion abundances spanning nearly an order of magnitude for PFAS species across five chemical classes (Table 2). Note that there is no “elution window” for each PFAS class – they are interlaced throughout the entire retention time (vide infra). The relative retention order of these 23 compounds drove the generation of a series of binary mixtures (compound 1 and 2, compound 2 and 3, etc.). These binary mixtures were directly injected into the mass spectrometer in the absence of an analytical column in order to encourage competitive ionization.

Figure 3:

Overlaid extracted ion chromatograms for the 24 PFAS species used in this study. The color-coded abbreviated names denotes the PFAS chemical class as shown in Table 2. All species are detected at the [M-H]¹⁻ except for Gen-X which is detected as both [M-H]¹⁻ and its decarboxylated form.

Table 2:

The classes of PFAS species used in this study with both a color code and a letter code for each class along with the generic structure of these compounds which reveals their chemical diversity.

Figure 4 shows three examples of the negative ion mode direct injection analysis (from the total of 22 binary mixtures that were analyzed). For each sample injection consecutively eluting PFAS species (as indicated by the reversed-phase LC chromatogram) were grouped together to form equal concentration (5,000 ng/L) binary mixtures (i.e., the first eluting peak was mixed with the second eluting peak, and the second eluting peak was mixed with the third eluting peak). Thus, the compounds most likely to co-elute in a reversed-phase LC-MS method were run together to encourage and probe a competitive ionization environment by removing chromatographic separation (samples were directly injected). Importantly, in the top trace, 6:2 FTS has 13 F atoms while PFOA has 15 F atoms. Based on logP values alone, PFOA would be expected to have a higher response factor. In the middle trace, PFOA has two fewer fluorines and is in a less competitive Group compared to PFOS resulting in PFOS having a peak area roughly twice that of PFOA. In the bottom trace, both PFDA and FOSA have the same number of F atoms but very different response factors. This suggests that not only is the number of F atoms (which is strongly correlated to NPSA and logP) an important predictor of ESI response, but the chemical class also clearly plays a role.

Figure 4:

Three representative chronograms for three different binary mixtures directly injected into the mass spectrometer in the absence of a chromatographic column. The top trace shows a nearly identical response factor. Despite the fact that Group A is predicted to preferentially reside on the droplet surface more than Group C, PFOA and 6:2 FTS have similar peak areas, likely due to the greater number of fluorines on PFOA, which cause it to compete equally with PFOA for surface occupancy. The middle trace and bottom trace show examples where one molecule has a response factor approximately half that of the other due either to Group identity and/or number of fluorines.

Table 3 shows a summary of the results with injection pair, chemical class, number of fluorine atoms and logP for each component in the binary mixture along with the peak area ratio (ESI response factor). Each row represents a sample injection that was performed. Each sample injection contained two PFAS molecules, chosen based on their elution order from a reversed-phase LC column (as detailed earlier). Each column lists the stated property for both analytes in order of their LC elution order. The final column shows the calculated Peak Area Ratio where the peak area of the first compound listed was divided by peak area of the second compound listed in the row.

Table 3:

Data from direct injections of binary mixtures of PFAS compounds.

#	Injection Pair	Elution Order	Chemical Class	Fluorine Number	logP	Peak Area Ratio 1st/2nd
1	PFBA / PFPeA	1 – 2	C – C	7 – 9	1.7 – 2.5	0.5
2	PFPeA / PFBS	2 – 3	C – B	9 – 9	2.5 – 1.9	0.3
3	PFBS / 4:2 FTS	3 – 4	B – A	9 – 9	1.9 – 2.6	1.2
4	4:2 FTS / PFHxA	4 – 5	A – C	9 – 11	2.6 – 3.2	1.9
5	PFHxA / HFPO-DA (Gen-X)	5 – 6	C – E	11 – 11	3.2 – 2.8	6.3
6	HFPO-DA (Gen-X) / PFHpA	6 – 7	E – C	11 – 13	2.8 – 4.0	0.2
7	PFHpA / PFHxS	7 – 8	C – B	13 – 13	4.0 – 3.4	0.5
8	PFHxS / 6:2 FTS	8 – 9	B – A	13 – 13	3.4 – 4.1	1.1
9	6:2 FTS / PFOA	9 – 10	A – C	13 – 15	4.1 – 4.7	1.1
10	PFOA / PFOS	10 – 11	C – B	15 – 17	4.7 – 4.9	0.6
11	PFOS / PFNA	11 – 12	B – C	17 – 17	4.9 – 5.5	1.8
12	PFNA / 8:2 FTS	12 – 13	C – A	17 – 17	5.5 – 5.6	1.2
13	8:2 FTS / PFDA	13 – 14	A – C	17 – 17	5.6 – 6.2	0.9
14	PFDA / FOSA	14 – 15	C – D	17 – 17	6.2 – 4.9	2.0
15	FOSA / PFUdA	15 – 16	D – C	17 – 17	4.9 – 7.0	0.5
16	PFUdA / 10:2 FTS	16 – 17	C – A	17 – 17	7.0 – 7.1	1.6
17	10:2 FTS / PFDoA	17 – 18	A – C	17 – 19	7.1 – 7.7	0.8
18	PFDoA / NMeFOSAA	18 – 19	C – D	19 – 21	7.7 – 4.9	3.4
19	NMeFOSAA / MeFOSA	19 – 20	D – D	21 – 21	4.9 – 5.3	0.7
20	MeFOSA / NEtFOSAA	20 – 21	D – D	21 – 23	5.3 – 5.3	1.6
21	NEtFOSAA / PFTeDA	21 – 22	D – C	23 – 27	5.3 – 9.2	0.3
22	PFTeDA / PFHxDA	22 – 23	C – C	27 – 31	9.2 – 10.7	1.7

It is important to note that there is one outlier in these data: Gen-X is known to be more labile than the other compounds used in this study. The data in Table 3 reveals the following “rules”. If the binary mixture contains two PFAS species that are from the same class, the number of F’s determines the relative response factors due to the greater number of fluorine atoms increasing the surface activity of the molecule (for example see injection #1 “PFBA/PFPeA”, also depicted in Figure 1). Moreover, the number of fluorine atoms perfectly correlates with relative retention order in reversed-phase separations (Figure 3). If the binary mixture contains two PFAS species that are from different groups, the preferential ionization becomes an important factor with Group A and B > C > D > E. Thus, from a chemical perspective, Group E is likely not accurately ranked because of the low number of compounds run from this class (due to low commercial availability). However, the number of fluorines also factors into the relative response of co-eluting PFAS: if there is both a fluorotelomer sulfonic acid (Group A) with a carboxylic acid (Group C) present in the sample, but the carboxylic acid has a greater number of fluorine atoms, the response of the carboxylic acid is “rescued”, and the two species have the same overall ESI response factor (Figure 4 top panel). However, if you have the same number of fluorine atoms the ESI response factor will follow the relative ranking of the chemical groups (e.g., Figure 4 bottom panel). It is important to note that Group E is only represented by Gen-X and as stated above, this PFAS species is known to decarboxylate during the ionization and desolvation process.

Collectively, these data indicate that the number of fluorines is a simple predicator of the properties of individual PFAS species when one considers the chemical class. This simple predictor can aid researchers in various facets of PFAS research. For example, few stable isotope labeled internal standards are available for targeted quantification assays, (though over 5000 PFAS species have been documented) resulting in researchers having to choose the most appropriate surrogate internal standards. The use of matched stable-isotope internal standards, whenever possible, mitigates the adverse ramifications due to the ESI response factors. When a matched internal standard is not available, selection of a surrogate internal standard based solely on the number of fluorine atoms and chemical class provides a pairing that is commensurate in terms of logP and NPSA and experimentally with respect to chromatographic retention time (#F atoms) and will yield optimal quantification.

CONCLUSION

PFAS are highly utilized in consumer products due to their unusual chemical properties. However, these same properties can have profound effects on analytical methods designed to detect and quantitate them. Data presented herein demonstrates that each PFAS species has general rules for ionization that depends on compound class and number of fluorines. These general rules can aid researchers when designing analytical methods for these analytes. In particular, these data demonstrate that when choosing a surrogate internal standard, one should be chosen based on PFAS chemical class and number of fluorine atoms.