The treatment of surface and groundwater impacted with aqueous film-forming foam (AFFF) constitutes a major challenge for the United States as regulatory limits for per- and polyfluoroalkyl substances (PFAS) are instituted at State and Federal levels. In order to tackle this issue in a sustainable and cost-effective manner, effective removal technologies are needed that don’t require expensive disposal or energy-intensive regeneration. A major challenge is identifying materials that can effectively remove both long-chain and short-chain PFAS from the environment, given the wide variation in PFAS physicochemical properties (e.g., water solubility, hydrophobicity). It is well known that both short- and long-chain PFAS are highly bioavailable and, in a number of cases, bioaccumulate due in large part to their interactions with proteins. Proteins, therefore, suggest an attractive solution: by taking advantage of multiple types of affinities and interactions, protein-based sorbents could be tuned to capture a variety of PFAS, and the sorbents themselves would be biodegradable.

This project is being conducted in three phases. During Phase I, a framework for evaluating PFAS-protein binding as a potentially exploitable mechanism for PFAS removal from impacted water was established using equilibrium dialysis and several model proteins: serum albumin, liver fatty acid binding protein (L-FABP), and the peroxisome proliferator activated nuclear receptors, (PPAR)-α, -𝛾, and -δ. These proteins were selected based on their roles in bioaccumulation and toxicity of PFAS, thus suggesting strong interactions. Strong equilibrium binding affinities were measured for several of these proteins with long- and short-chain PFAS, but further work was needed to understand the kinetics of these interactions and how and whether they could be applied under more realistic water treatment conditions.

During Phase II of this project, the use of proteins as sorbents for PFAS removal was further evaluated through a combination of equilibrium and kinetic (time-resolved) experiments using equilibrium dialysis and isothermal titration calorimetry (ITC). These experiments showed highly variable results and low binding to all proteins tested. Three major challenges were identified: (i) the reliability and comparability of methods for measuring PFAS-protein binding affinities; (ii) the cost and durability of proteins selected; and (iii) the scalability of the approach to a practical water treatment application. The project then pivoted to identifying proteins that would be more effective to test—those from plant or bacterial origins, and low-molecular weight proteins that could be easily produced. A promising new method was also identified for evaluating binding affinity and kinetics simultaneously: surface plasmon resonance (SPR).

The overall goal of Phase III of this project is to develop a robust and practical treatment approach for AFFF-impacted water using biobased sorbents, utilizing a novel protein encapsulation approach, and leveraging lessons learned the initial studies. This will be accomplished through three complementary objectives:

  • Objective 1: Tiered in silico – in vitro sorbent screening. Effective sorbents for both long- and short-chain PFAS will be identified through tiered screening by leveraging computational and high-throughput in vitro tools via a multi-objective framework.
  • Objective 2: Production and entrapment of target proteins. The project will overcome challenges in cost-effectively producing protein sorbents using E. coli and yeast expression systems, and by entrapping proteins in a readily degradable yet protective structure of porous polymeric silica to enhance their durability.
  • Objective 3: Testing under realistic use conditions. Scalability of the identified sorbents will be addressed via flow-through system with multiple entrapped proteins under realistic conditions, including impacted water from field sites and experiments in the presence of natural organic matter (NOM). The encapsulated protein sorbents will be benchmarked against traditional granular activated carbon and ion exchange sorbents.


                                                                                                 Phase I Summary

Technical Approach

Phase I of this project initially focused on evaluating mammalian proteins associated with PFAS bioaccumulation and toxicity. Candidate protein sorbents were identified from existing literature based on interactions with PFAS and with analogous ligands such as fatty acids and acidic drugs. Protein-specific binding affinities for a set of eight selected perfluorinated alkyl acids — PFOS, PFNA, PFOA, PFHxS, PFHxA, PFHpA, PFBS, and PFBA —were then predicted using a multi-step molecular modeling framework. Finally, the most promising proteins — liver and intestinal fatty acid binding proteins, L-FABP and I-FABP, and the PPAR-α, -𝛾, and -δ — were experimentally evaluated for PFAS adsorption through a series of batch tests at the bench scale to determine the equilibrium dissociation constant, KD, and the equilibrium binding capacity, qmax. Equilibrium dialysis experiments were performed under ideal matrix conditions with single PFAS-protein pairs, under varied pH, ionic strength and temperature conditions, as well under “realistic” conditions using AFFF-impacted groundwater and single proteins.

During Phase II of this study, in parallel with time-resolved equilibrium dialysis testing, pilot experiments were conducted to evaluate the use of ITC for protein binding studies. Two proteins and three PFAS were evaluated: PFOS and PFBS with Bovine serum albumin (BSA), and PFOA with L-FABP. For the BSA experiments, 2.5 mM solutions of PFBS and PFOS were made using pH 7.4 phosphate buffered saline (PBS) buffer. When time-resolved dialysis experiments did not yield binding kinetics, SPR method development work replaced the originally proposed small column experiments to determine the binding kinetics of PFAS with selected proteins. Method development focused around minimizing non-specific binding between the PFAS and the SPR sensor surface, maximizing coverage of the sensor chip by L-FABP, and determining the appropriate concentrations of PFAS, L-FABP and regeneration to release binding. L-FABP immobilization was performed at concentrations of 3.33 and 5 µM, to ensure enough protein for binding. An injection at the highest PFOS concentration (1 mM) was then performed to confirm the activity of the protein. Multiple injections of PFOS at increasing concentrations from 300 μM to 1 mM were used to evaluate the dissociation constant, KD. Finally, smaller proteins from plant and microbial origins were identified by searching the Protein Data Bank for proteins with short-chain fatty acids as endogenous ligands. Identified proteins were then screened for predicted binding with short-chain PFAS via molecular docking simulations using Autodock Vina.

During Phase III of this study, the project team will conduct in silico and in vitro evaluation of protein-PFAS binding affinity with complementary approaches: molecular dynamics, SPR, equilibrium dialysis, and fluorescence displacement. In an advance from previous phases, proteins to be tested will be produced in-house, using a tiered system of expression in E. coli and P. pastoris. This multimethod approach to production and binding evaluation will aid in identifying the most effective and readily scalable candidates. Targeted analysis will be used to evaluate sorbent performance for 10 single PFAS (6 perfluoroalkyl carboxylic acids, PFCAs, and 4 perfluoroalkane sulfonic acids, PFSAs). We will also test sorbent performance using binary mixtures to evaluate potential competition between PFAS that are effectively removed when tested alone, and finally test the removal efficacy for realistic AFFF-derived environmental mixtures from groundwater sourced from three sites (Point Mugu, Cape Canaveral, and Barksdale). The silica-based entrapment and support system will be tested for its ability to improve the durability of the protein-based sorbent. The third year of Phase III will focus on scale-up and robustness under storage. An iterative process will be adopted whereby failure at any step will feed back to earlier screening, taking into account the relevant properties of the failed candidate to refine the screening approach.

This project combined molecular simulations with in vitro experiments to identify protein-based sorbents for PFAS and evaluated the PFAS-protein binding affinities.


 Phase I and II Results

The key findings from Phase I are summarized as follows and are available in the Phase I Final Report. During Phase I, the binding of PFAS to different proteins was found to be variable, and both chain length- and condition-specific, suggesting the ability to tune PFAS removal by the use of different proteins alone or in combination. Interestingly, the nuclear receptor PPAR-δ was found to bind strongly with both PFBA and PFHxS, while PPAR-α bound strongly to PFHxA (as well as the long-chain PFNA). This is the first report to show such a strong association of biomolecules with short-chain PFAS, raising both prospects for treatability and potential concern about toxic effects. Moreover, the binding capacities calculated from protein-PFAS data for some short-chain PFAS were substantially higher than with either granular or particulate activated carbon, illustrating the potential of our approach for targeting difficult-to-treat PFAS.

The key findings from Phase II are summarized as follows and are available in the Phase II Final Report. Key findings from Phase 2 included comparative evaluation of ITC and SPR as approaches to assess PFAS-protein binding. From ITC experiments, no binding was observed for PFBS with BSA, but a binding affinity was calculated for PFOS in good agreement with reported values from similar studies. However, this required high concentrations of protein. For L-FABP, it was determined that, for the protein quantities available for ready testing, a more sensitive micro-ITC system was required. This remains an option for future work. In initial testing of the SPR platform, no non-specific binding was found between PFOS and the SPR sensor chip, and subsequent immobilization experiments with L-FABP indicated that the immobilized protein was active, both positive indications. However, only minimal PFOS-LFABP binding was detected at concentrations of PFOS below 1 mM, indicating that SPR was also hitting sensitivity limitations and it was determined this was in large part due to the size difference between the immobilized protein and the PFAS tested. Thus, either a more sensitive SPR-based assay was needed, smaller proteins closer in size to the target PFAS could be evaluated, or the system could be “flipped” by tethering the PFAS to the sensor surface, which would require additional assay development. Alternatively, proteins from sources that could be more cheaply scaled up could provide a route where even weaker binding could be harnessed for PFAS removal by increasing the sorbent quantity. Molecular screening and modeling to identify such sources yielded three promising candidates for follow-on investigation: a phospholipase enzyme from rice, a fluoroacetyl co-enzyme A from Streptomyces cattleya, and a thermophilic esterase from Thermogutta terrifontis.


Results of Phase I confirmed the potential of protein-based sorbents for PFAS remediation by identifying, through complementary molecular modeling and in vitro experiments, proteins that associated strongly with both long- and short-chain PFAS. Of particular benefit was the ability to tune an adsorbent by incorporating multiple protein-based moieties and/or by use of strategic changes in feed composition (ionic strength, pH), to address a wide variety of PFAS structures. The combination of molecular modeling and batch testing formed the basis of a robust and powerful design framework for developing protein-based sorbents for PFAS water treatment.

Results of Phase II indicated a need to increase assay sensitivity for protein binding. While both ITC and SPR showed some limitations, in pivoting to method development we identified a promising new direction of research to directly measure PFAS-protein binding kinetics and simultaneously evaluate the stability, regenerability, and overall robustness of different protein candidates. Moreover, several nonhuman enzymes that have never been investigated for PFAS-binding activity were identified that may point to new avenues of discovery for short-chain sorbents. While PFBA continued to pose a challenge to identification of sorbents, the predicted binding affinities for PFBA were substantially higher for phospholipase A2 and for the thermophilic esterase than for the mammalian proteins that had been the focus of Phase I.

Phase III of this project will produce fundamental new knowledge about the interactions between diverse non-human proteins and PFAS, and provide valuable data on the treatability of highly mobile short-chain PFAS. Methods of protein production will be evaluated from micromolar to gram scales, giving insight into the scalability and potential commercialization of the protein entrapment approach. A robust sorbent made with degradable support would be of tremendous benefit for more sustainable treatment of AFFF-impacted sites. (Anticipated Phase III Completion - 2026)


Khazaee, M., E. Christie, W. Cheng, M. Michalsen, J. Field, and C. Ng. 2021. Perfluoroalkyl Acid Binding with Peroxisome Proliferator-Activated Receptors ⍺, γ, and δ and Fatty Acid Binding Proteins by Equilibrium Dialysis with a Comparison of Methods. Toxics, 9(3):45. doi.org/10.3390/toxics9030045.