High-Throughput Small-Scale Platform for Synthesis, Characterization, and Modeling of Per- and Polyfluoroalkyl Substances Analogs

Abstract

Per- and polyfluoroalkyl substances (PFAS) are a global challenge due to their exceptional thermal and chemical durability which leads to environmental persistence, bioaccumulation, and toxicity. Tackling this challenge is a complex endeavor as the ever-expanding number of emerging PFAS hinders their monitoring while current countermeasures remain limited. Thus, there is a need for rapid strategies that can transform PFAS into safer, degradable analogs or expand libraries for untargeted monitoring. Here, we describe the implementation of a high-throughput (1 Hz) desorption electrospray ionization mass spectrometry (HT-DESI-MS) platform for the chemical transformation of perfluorocarboxylic acids (PFCAs) via a data-driven workflow that led to 915 new PFCA analogs (89% success rate) and revealed reactivity trends. Tandem mass spectrometry (MS/MS) enabled online structural confirmation and diagnostic fragment identification, supporting standard-free LC-MS/MS analysis. Further integration with ion mobility spectrometry (IMS) provided drift time measurements correlating with molecular size and shape, adding a new dimension that can improve feature annotation in untargeted PFAS analysis. Complementary quantum mechanical calculations of dipole moment and HOMO− LUMO gap predicted polarity and electronic reactivity, guiding analog selection. Collectively, this workflow combines rapid synthesis, structural annotation, and multidimensional profiling, with potential to discover safer PFAS and enhance environmental monitoring.

Introduction

Per- and polyfluoroalkyl substances (PFAS) have been widely incorporated into industrial products since the 1950s as they possess exceptional thermal stability, chemical resistance, and surfactant properties. (1–3) However, these characteristics have contributed to their environmental persistence, bioaccumulation, and associated toxicity, prompting global concern from the early 2000s onward. (4–9) Multiple efforts are being pursued to tackle the crisis generated by these so-called “forever chemicals”. For instance, short-chain alternatives and structurally modified legacy PFAS have been developed as replacements, yet their environmental implications and toxicological profiles remain similar or even unknown. (10–15) Remediation strategies such as pollutant removal by adsorption (16,17) or decomposition (18–21) are still challenging due to the requirement of adsorbent disposal or incomplete defluorination and unknown toxicology of degradation products. On top of this, the broad-coverage monitoring of PFAS is still a challenge despite of the wide range of emerging analytical methodologies (22–35) that now detect sizable fractions of previously unidentified organofluorides, as many newly formed PFAS still remain uncharacterized. (30)

The ability to broadly diversify PFAS structures through chemical transformations is relevant to tackling this crisis as it provides opportunities to (i) develop safer analogs for industrial use, (36,37) (ii) introduce motifs that enhance chemical degradation and reduce bioaccumulation, (38,39) and (iii) enlarge the synthetic PFAS libraries that underpin untargeted analytical workflows (27–31) used to assess environmental transformations in soil or wastewater. (32) To fulfill these goals, a pipeline for PFAS diversification requires rapid reactions that allow for efficient chemical space exploration using minimal substrate amounts to avoid generating further contaminants, as well as computational assistance to guide the evaluation of high-value modifications toward specific physicochemical properties.

Here we report the implementation of a workflow for the rapid generation of PFAS analogs through the combination of high-throughput (HT) experimentation with quantum mechanical calculations using perfluoroalkyl carboxylic acids (PFCA) as model PFAS. Within this pipeline, HT nanogram-scale synthesis is carried out via desorption electrospray ionization (DESI), (40–42) a unique ionization method which serves as a synthetic tool via accelerated microdroplet reactions. (43–53) This acceleration, often orders of magnitude faster than bulk-phase chemistry, (44) arises from the unique interfacial properties in microdroplets, including partial solvation, (50,51) strong electric fields, (52) and the presence of reactive species. (53) As a result, products are formed within milliseconds in DESI droplets and can be directly detected by mass spectrometry (MS) for combined synthesis and screening. HT-DESI-MS is then achieved through fast (~1 s per reaction) sampling of a high-density array of small-volume (50 nL) reaction mixtures. (54–56) This fully automated approach has already found broad application in drug discovery, both for label-free biological assays (57–61) as well as for rapid diversification of drug molecules. (62–64) The usefulness of the vast data sets obtained with this platform was demonstrated by successfully informing a liquid chromatography (LC) tandem mass spectrometry (MS/MS) methodology, which is the gold standard for PFAS analysis. (24–26) Additionally, we show that coupling DESI with ion mobility spectrometry (IMS), an emerging technique for PFAS analysis, (22,31) can provide complementary information for further characterization of the newly synthesized PFCA analogs. Finally, we demonstrate the integration of theoretical calculations of dipole moments (μ) and HOMO–LUMO gaps (E_H-L) as an efficient strategy for prioritizing promising analogues and guide cost-effective experimental coverage of the vast PFAS chemical space.

Experimental Section

HT-DESI-MS(/MS) Synthesis

The automated HT-DESI-MS platform combines a robotic liquid handling system and a mass spectrometer equipped with a DESI source, integrated via custom hardware and software. A detailed description of the platform can be found in the Supporting Information (SI) as well as in published manuscripts. (55,56) Briefly, the synthesis workflow involves the automated preparation of reaction mixture arrays (16 replicates per reaction, 50 nL each) onto a PTFE-coated glass slide that is then automatically transferred to the DESI stage where a charged solvent (methanol) spray is used to desorb the reaction mixtures, carrying them in secondary splashed microdroplets where accelerated reactions occur, thus providing on-the-fly the synthesis of analogs at a throughput of 1 s per reaction. Hits were determined using the signal-to-noise and conversion ratios (SNR and CR, respectively; see SI for details). Online structural confirmation of the products was conducted by revisiting the hit spots for HT-DESI-MS/MS analysis at a throughput of 10 s per reaction.

LC-MS(/MS) and DESI-IMS-MS

All the reaction mixtures were incubated (12 h), and eight reaction mixtures were pooled together by selecting nonisobaric products to increase the LC-MS analysis efficiency. LC-MS/MS was carried out in the multiple reaction monitoring (MRM) mode and was conducted by setting up the MRM transitions rapidly obtained via HT-DESI-MS/MS. Reaction mixtures were deposited (1 μL) onto a PTFE-coated glass slide and screened using DESI via a line scan using a traveling wave ion mobility spectrometry (TWIMS) coupled to MS.

Theoretical Calculations

The geometries of the PFCA derivatives were obtained through a two-step process. First, Open Babel (65) was used to generate 50 conformers for each compound and to select the lowest energy conformer using the Merck Molecular Force Field (MMFF94). (66) Next, the selected lowest energy conformer was optimized in the gas phase at the density functional level of theory using Gaussian16 (67) package. All the quantum mechanical calculations were carried out using B3LYP functional and 6–31G(d,p) basis sets.

Results and Discussion

Small-Scale Diversification of PFCAs via HT-DESI-MS(/MS)

The HT-DESI-MS platform was used to rapidly derivatize PFCAs via amide-forming coupling between carboxylic acids and amines (Figure 1A, Table S1). A stepwise data-driven screening workflow (63) was utilized so that only conditions or functionalization reagents found successful in an initial screen are used in a subsequent screening stage. The first screen involved four coupling reagents and 14 representative amines that were tested using perfluorooctanoic acid (PFOA) as model PFCA to identify the most effective activator (Figures S1–S2). EDC (1-ethyl-3-(3-(dimethylamino)propyl)carbodiimide hydrochloride) and diamino containing amines showed superior reactivity (Figure S2), therefore, eight diamines were selected and proceeded to the second screening stage for determination of optimal reaction conditions (Figure S3). Seven conditions, including two different forms of EDC, different orders of mixing reagents, and the use of additives, were screened. Mixing PFOA with HCl-EDC before adding the amine and omitting any additives (see Condition 1 in Figure S3) was found optimal. We expanded the screening (third stage) to C3–C12 PFCAs with 103 different amines spanning seven functional classes (Table S1, A1-A103; Figure 1B and Figures S4–S5).

Figure 1.

HT-DESI-MS screening of PFCAs with a large set of amines. (A) Reaction scheme for the coupling reaction between PFCAs (PFCA3-PFCA12) with 103 amines (see Supporting Information Figure S4 and Table S1 for the full list of amines and their structures). The amine structures shown are related to the different groups as described in Figure S4. (B) Heatmap of conversion ratio (CR) for the screening results of PFCAs with amines. (C) Representative spectra of the reactions of different PFCAs with N,N-dimethylethylenediamine (A26). (D) representative spectra of the reactions of PFOA with (1R,2R)-(−)-1,2-diaminocyclohexane (A8), N-methylpropane-1,3-diamine (A18), (1-benzylpyrrolidin-3-yl)methanamine (A56), N,N′-dimethylethylenediamine (A79), 4-piperidinopiperidine (A93), 2-(4-Methyl-piperazin-1-yl)-ethylamine (A100), and 2-amino-N,N,N-trimethylethanaminium chloride hydrochloride (A102). R.A.: relative abundance.

In total, 1030 distinct transformations including 16,480 individual measurements, were carried out in less than 4.5 h. This campaign yielded 915 PFCA analogs (88.8% success rate) with picomole-level reagent consumption via accelerated reactions in DESI microdroplets, thus without the need for any lengthy incubation (Figure 1B and Figure S5). The screen also provided insights into broad reactivity trends. For example, longer chain PFCAs generally had better reactivity than shorter chain ones (Figure 1B, C), likely due to higher surface affinity (68) and hence stronger microdroplet interfacial effects. Primary amines (G1–G3 and G6 classes, e.g., A8, A18, A56, A100 in Figure 1D) out-performed secondary amines (G4–G5, e.g., A79) because of reduced steric hindrance, whereas charged amines (G7, e.g., A102) were largely inert. Note that despite the high speed and tiny sample size, HT-DESI-MS provides high-quality spectra (Figure 1C, D) and high-precision CR values (Figure S6).

To benchmark this DESI-based synthetic methodology, we carried out standard bulk reactions with overnight incubation for PFOA with 103 amines followed by LC-MS analysis (Figure 2A and Figure S7). Eight reaction mixtures were pooled together in a single LC-MS run to achieve more efficient analysis (Figure 2A, see SI for details). Overall, 99 out of 103 reaction products were detected (Figure S7), validating the reaction screening results by HT-DESI-MS, which provided all 103 hits (96% agreement). Importantly, this similar performance is obtained despite HT-DESI-MS being at least 300-fold faster (see SI, Section II) than conventional bulk reactions followed by LC-MS.

Figure 2.

(A) LC-MS full mass chromatogram of reaction mixture containing eight reactions (bottom). TIC of the LC-MS analysis and (top) EIC of the expected m/z of each product. (B) HT-DESI-MS/MSof reaction products of PFOA with 2,2-dimethylpropane-1,3-diamine (A7), N-isopropylethylenediamine (A17), 3-(dimethylamino)propylamine (A31), and N1,N3-dimethylpropane-1,3-diamine (A81). Note that the fragmented site forming the neutral loss (NL) is indicated in arrows. (C) LC-MRM (normalized for each species) guided by HT-DESI-MS/MS without need of database or authentic standards. The transitions of A7 (499→482), A17 (499→440), A31 (499→454), and A81 (499→468) were selected by the product ions observed in (B).

The automated platform also enables rapid structural confirmation via online MS/MS, a crucial capability in PFAS structural analysis. (26) Microdroplet reaction products from hit spots were subjected to MS/MS and the spectra obtained provided diagnostic neutral losses (NLs) rich in structural information. With this information, regioisomers arising from multivalent amines were distinguished (Figure S8). For instance, the MS/MS spectrum of the product from the reaction between PFOA and N-methylethylenediamine showed two different fragments: (i) NL 31 Da (−NH2CH3) indicating amide formation via the primary amine, and (ii) NL 17 Da (−NH3) consistent with a secondary-amine attack and a free primary amino group. Analog controls with ethylenediamine and N,N′-dimethylethylenediamine validated these assignments (Figure S8B). Diagnostic NLs also differentiated isomeric products as in the case of the reactions between PFOA and amines A31, A7, A17, and A81, which all lead to a product of m/z 499 (Figure 2B and Figure S9).

This information can also be used to guide LC-MRM analysis, which is typically used for PFAS detection and commonly depend on standards or databases, both of which can be limited. Four pooled reaction mixtures that could not be assigned by full-scan MS due to their isobaric products (Figure S9), were fully resolved and assigned once the transitions derived from HT-DESI-MS/MS were applied (Figure 2C). Additionally, as homologous PFCAs fragment similarly, MRM transitions defined for PFOA analogues can be transferred directly to all the PFCAs evaluated (Figure S10). For instance, the MS/MS spectrum of the PFOA analogs generated through reaction with N,N-dibutyl-1,3-propanediamine (DBPA, A35) showed two fragments (NL 129 Da and NL 157 Da) in HT-DESI-MS/MS. Implementing these same NLs for the other PFCAs as MRM transitions allows for the ten DBPA-derived products to be detected and assigned with high specificity without the need for any analytical standards (Figure S10). Thus, HT-DESI-MS/MS does not only provide online structural confirmation of the synthesized analogs but also gives immediately usable transitions for LC-MRM analysis, thus enabling standard-free retention time assignments for new molecules.

DESI-IMS-MS Evaluation of PFCA Derivatives

As a tool for synthesis and analysis, DESI can also be readily coupled to both IMS and MS, thereby introducing a gas-phase separation that precedes mass analysis and provides additional information on the shape or size of the ions. (69) When combined with MS, IMS enhances the resolution of complex mixtures making it particularly useful for nontargeted PFAS analysis. (22,27)

A total of 40 PFCA analogs, synthesized in situ via DESI, were selected for in depth characterization of their molecular properties via IMS and exact mass analysis (Table 1, Tables S2–S3 and Figures S11–S12). This set included all the DBPA-derived PFCAs together with 29 amine-derived PFOA analogues. The observed IMS drift times increased linearly with the PFCA chain length in the DBPA series (Figure S11), aligning with the expectation of larger collision cross sections corresponding to longer perfluoroalkyl chains. Importantly, for isomeric products, subtle structural differences generated measurable drift time shifts (Figure S12). For instance the drift time of the 1,3-bis(aminomethyl)cyclohexane (A10) derived product (3.17 ms) was found longer than that of the 2-(1-methylpiperidin-4-yl)ethanamine (A66) derived product (3.02 ms), reflecting the more extended geometry (i.e., larger cross-section) of the cyclohexyl core likely due to the rigid structure of the cycle and the presence of an additional substituent. Another interesting example is the (1-benzylpyrrolidin-3-yl)methanamine (A56) functionalized PFOA (m/z 587), which despite being heavier than the DBPA-derived PFOA analog (m/z 583) has a shorter drift time, illustrating how a compact conformation can offset the effect of the ion mass (Table 1). These observations demonstrate that coupling DESI-based microdroplet synthesis with IMS extends the experimental data set accessible via HT experiments from m/z and MS/MS fragmentation pattern to cross sections (which can be calculated with drift time calibration), an attribute that can be correlated with blood–brain-barrier permeability or incorporated into spectral libraries for nontargeted PFAS monitoring. (70)

Table 1.

Property Table for Products of PFOA Derivatives with Selected Amines. The full list can be found in Tables S2–S3.

Index	Name	Measured m/z	Error (ppm)	RT (min)	DT (ms)	EH-L (eV)	μ (Debye)
A2	1,3-Diaminopropane	471.0548	−1.3	12.25	2.55	5.95	3.41
A18	N-Methylpropane-1,3-diamine	485.0708	−0.4	12.43	2.58	6.64	5.67
A13	1,3-Diamino-2-propanol	487.0500	−0.6	11.91	2.67	6.99	6.16
A25	N-(2-Hydroxyethyl)-1,3-propanediamine	515.0817	0.2	12.24	2.86	6.10	1.63
A24	2-(2-Aminoethylamino)-ethanol	501.0654	−1.2	12.11	2.77	5.78	4.60
A23	3-(Benzylamino)propylamine	561.1013	−1.8	13.87	3.36	5.78	3.12
A35	3-(Dibutylamino)propylamine	583.1811	0.9	15.06	3.70	5.51	4.64
A10	1,3-Bis(aminomethyl)cyclohexane	539.1133	−8.7	13.50	3.19	5.94	3.02
A66	2-(1-Methylpiperidin-4-yl)ethanamine	539.1173	−6.3	13.11	3.02	5.10	3.44
A65	(1-Benzyl-1,2,3,6-tetrahydropyridin-4-yl)methanamine	579.1450	−7.4	13.85	3.47	5.46	4.13
A68	1-Benzylpiperidin-4-amine	587.1131	−8.3	14.15	3.52	5.73	3.78
A56	(1-Benzylpyrrolidin-3-yl)methanamine	587.1173	−1.2	14.02	3.51	5.37	3.55

Note that for the unmodified PFOA, E_H-L is 7.33 eV, and μ is 2.07 D. RT stands for retention time, and DT stands for drift time.

E_H-L and μ Calculations

To complement the experimental descriptors, electronic properties such as E_H-L and μ were computed for the selected set of 40 derivatives using Gaussian16 (Tables S2–S3 and Figure S13). These properties provide essential information about the polarity as well as the chemical and photochemical reactivity of the newly formed PFCA analogs. For the DBPA derivatives, μ increased markedly across the PFCA series whereas E_H-L modestly decreased (Table S2). Further insights into the effect of structural differences on these electronic properties were obtained by comparing products of different amines with PFOA (Table 1 and Table S3). In general, the derivatized PFOAs exhibited higher μ and reduced E_H-L than unmodified PFOA, however, minor structural modifications produced disproportionate electronic effects across the reaction products. For example, similar amine-derived products, such as A2, A13, and A18, which differ only by a methyl or hydroxyl group, displayed widely spaced μ values. Even when E_H-L values were found nearly identical, a single −CH2– insertion (cf. A24 and A25) or a positional isomer change (cf. A21 and A22) altered μ appreciably. Note that the fact that the PFOA derivatives showed lower E_H-L, especially those generated with amines containing aromatic groups (A56, A65, and A68), indicates greater chemical reactivity or easier electron transfer. Thus, these simulations have the potential to support the rational modification of PFAS toward specific physicochemical properties.

In summary, this study demonstrates that HT-DESI-MS/MS, together with LC-MRM, IMS, and computational modeling, can operate within a unified workflow that links rapid synthesis with multidimensional PFAS characterization. As shown, this combined approach, which can be expanded to diverse chemical space, (63) has the potential to accelerate the discovery of new analogs with specific properties, facilitate the construction of databases (25) for untargeted PFAS analysis of environmental samples, or even lead to the development and testing of remediation strategies such as precipitation (71) or photosensitization by chemical transformation. It is also worth noting that considering microdroplets, viz. aerosols, are of major importance in the environment, the accelerated reaction methodology presented here provides a system suitable for investigating environmentally relevant PFAS chemistry. (72,73)