In vitro screening of per- and polyfluorinated substances (PFAS) for interference with seven thyroid hormone system targets across nine assays

Abstract

The US Environmental Protection Agency is evaluating the ecological and toxicological effects of per- and polyfluorinated chemicals. A number of perfluorinated chemicals have been shown to impact the thyroid axis in vivo suggesting that the thyroid hormone system is a target of these chemicals. The objective of this study was to evaluate the activity of 136 perfluorinated chemicals at seven key molecular initiating events (MIE) within the thyroid axis using nine in vitro assays. The potential MIE targets investigated are Human Iodothyronine Deiodinase 1 (hDIO1), Human Iodothyronine Deiodinase 2 (hDIO2), Human Iodothyronine Deiodinase 3 (hDIO3), Xenopus Iodothyronine Deiodinase (xDIO3); Human Iodotyrosine Deiodinase (hIYD), Xenopus Iodotyrosine Deiodinase (xIYD), Human Thyroid Peroxidase (hTPO); and the serum binding proteins Human Transthyretin (hTTR) and Human Thyroxine Binding Globulin (hTBG). Of the 136 PFAS chemicals tested, 85 had sufficient activity to produce a half-maximal effect concentration (EC50) in at least one of the nine assays. In general, most of these PFAS chemicals did not have strong potency towards the seven MIEs examined, apart from transthyretin binding, for which several PFAS had potency similar to the respective model inhibitor. These data sets identify potentially active PFAS chemicals to prioritize for further testing in orthogonal in vitro assays and at higher levels of biological organization to evaluate their capacity for altering the thyroid hormone system and causing potential adverse health and ecological effects.

1. Introduction

The Organization for Economic Cooperation and Development has identified 4730 Chemical Abstracts Service registry numbers (CASRN) related to per- and polyfluoroalkyl substances (PFAS; OECD, 2018). The United States Environmental Protection Agency (USEPA) PFAS Action Plan (USEPA, 2020) indicates that there are >600 PFAS chemicals on the active Toxic Substances Control Act (TSCA) inventory. The PFAS Action Plan prescribes toxicity and effects research for seven of the most common chemicals but does not address in detail generation of toxicity data for the several hundred other PFAS of potential concern. To address this knowledge gap, the USEPA, in collaboration with the National Toxicology Program, has identified nearly 150 PFAS chemicals for initial testing screens. These chemicals, selected in two sets of 75 PFAS, were based on Agency priorities, exposure/occurrence considerations, availability of animal or in vitro toxicity data, and ability to procure and solubilize samples in dimethyl sulfoxide (DMSO) as a prerequisite for testing (Patlewicz et al., 2019, Patlewicz et al., 2022; https://www.epa.gov/chemical-research/pfas-chemical-lists-and-tiered-testing-methods-descriptions).

Some PFAS are known to interact with the thyroid axis in a broad range of species including, potentially, humans (Coperchini et al., 2021). However, the number of chemicals evaluated for actual effects on the thyroid hormone system is quite limited. Given the complexity within the hypothalamic-pituitary-thyroid axis (HPT axis) there are numerous potential targets/molecular initiating events (MIE) for these compounds. The HPT axis is a system of tissues and feedback mechanisms that function to control the synthesis, release and processing of the thyroid hormone and is highly conserved across vertebrate species. Briefly, the hypothalamus releases thyrotropin-releasing hormone (TRH) into pituitary portal circulation leading to release of thyrotropin stimulating hormone (TSH) from the pituitary. TSH travels via the blood to the thyroid gland and activates processes necessary for synthesis and release of 3,5,3′,5-tetraiodothyronine (T4). Upon activation of the TSH receptor, thyroid follicular cells take up iodide via the sodium iodide symporter (NIS) where it is acted upon by thyroid peroxidase (TPO) to facilitate iodination of tyrosines on thyroglobulin resulting in the formation of monoiodotyrosine (MIT) and diiodotyrosine (DIT). TPO further functions to facilitate the coupling of DITs, or DIT with MIT, to form T4 and to a lesser extent 3,5,3′-triiodothyronine (T3), respectively. In this process iodine on DITs and MITs that do not become part of the hormone is recycled by iodotyrosine deiodinase (IYD) helping maintain sufficient iodine levels in the gland. Upon release from the gland, T4 travels in the blood bound to thyroid hormone distributor proteins transthyretin (TTR) or thyroxine binding globulin (TBG). Once T4 reaches the tissue it is converted to the active form of the hormone, T3, via deiodination catalyzed by Types 1 (DIO1) and 2 deiodinase (DIO2). A third deiodinase, Type 3 deiodinase (DIO3), is also present in the tissues and is responsible for further removal of iodine and formation of less active forms of the hormone. This group of potential MIEs does not represent all aspects of the thyroid hormone system. Rather, they represent aspects of thyroid hormone economy, synthesis, activation, inactivation, and distribution.

Many of the chemicals in this screening test set are data poor with little known about their interaction or potential interference with thyroid hormone system homeostasis. In vitro assays have been developed that can be applied to evaluate whether chemicals have the potential to disrupt the thyroid hormone system at a number of different points within the axis. These include assays to address key points for thyroid hormone synthesis (Paul et al., 2014; Wang et al., 2019; Olker et al., 2021), thyroid hormone activation and inactivation (Renko et al., 2015; Hornung et al., 2018; Olker et al., 2019), and hormone binding and transport (Ren et al., 2016), among others (Murk et al., 2013). The assays used here cover seven MIEs within the network of thyroid Adverse Outcome Pathways (AOP) which are involved in thyroid hormone economy (Noyes et al., 2019). The seven MIEs link to 16 known or putative pathways in the AOP wiki (https://aopwiki.org/aops, Society for the Advancement of AOPs, 2020). This builds on the AOP network of Noyes et al. (2019) to link data derived from in vitro assays that measure chemical interactions with thyroid molecular targets to downstream events and adverse outcomes traditionally derived from in vivo testing. The in vitro screening presented here addresses bioactivity at MIEs within the thyroid axis AOP network, which includes potentially sensitive endpoints of mammalian neurological development and amphibian metamorphosis. Accordingly, the chemicals with activity in one or more of these initial screens could be considered for testing at higher levels of biological organization.

2. Methods

A set of PFAS chemicals prioritized for testing by the US EPA were screened using nine in vitro assays representing seven key molecular initiating events (MIE) within the thyroid axis. These include six enzyme inhibition assays (hDIO1, hDIO2, hDIO3, xDIO3, hIYD, and xIYD) previously described and used in screening of ToxCast chemicals (Hornung et al., 2018; Olker et al., 2019; Mayasich et al., 2021; Olker et al., 2021; Olker et al., 2022). These assays measure the release of iodide from the hormone substrate via a colorimetric endpoint using the Sandell-Kolthoff (SK) reaction. The other three assays (hTPO, hTTR, and hTBG) are fluorescence-based assays developed following previously published assays, with data reported here being the first large set of chemicals screened with these assays.

Each assay included positive and negative (solvent) control compounds; these are listed along with the substrate for each respective assay in Supplemental Table 1. DMSO was the negative (solvent) control in all assays, with concentration in final assay conditions based on the percent that did not produce a response or assay interference; this was 1% DMSO the enzyme inhibition assays and 0.5% DMSO in the hTTR and hTBG assays. The PFAS test chemicals were screened through these nine assays using a tiered screening approach with initial testing at a single high concentration followed by a subset of chemicals tested in concentration response. Responses for each chemical were normalized to % activity based on the positive and solvent control compounds for each assay. For the enzyme inhibition assays (hDIO1, hDIO2, hDIO3, xDIO3, hIYD, xIYD, hTPO), uninhibited enzyme response or 100% of control (in DMSO or lowest concentration of positive control, depending on the assay) represented 0% chemical activity and fully inhibited enzyme (in the high concentration of the positive control) represented 100% chemical activity. For the fluorescence-based hTTR and hTBG assays, 0% chemical activity represented 0% of T4 displacement of ANSA (100% of control, in lowest concentration of the positive control) and 100% chemical activity was 100% of T4 displaceable ANSA (in high concentration of positive control). More details are described below for each assay below.

Chemicals that produced activity of 20% or greater were considered ‘active’ in these assays. This cutoff was selected to have a consistent threshold across assays that is also supported by the background variability in each assay [calculated as three times the median absolute deviation (MAD) in the solvent controls (DMSO) across all replicates of all tested plates in each assay]. Chemicals producing activity of 50% or greater were moved forward for testing in concentration response. This response level was selected to focus on chemicals with greater separation from the background variability and for which an EC50 could likely be calculated when tested in full concentration response mode, making it possible to compare chemical potency within an assay. Data were processed and normalized using assay-specific automated pipelines in R (version 3.6.1; R Core Team, 2019), with single-point and concentration response data then submitted to the ToxCast pipeline for analysis and inclusion in the invitroDBv3.5 database (https://epa.figshare.com/articles/ToxCast_Database_invitroDB_/6062623).

2.1. PFAS test chemicals and approach

The test chemicals were procured commercially by Evotec (US) Inc. Branford, CT under contract to the US EPA. Chemicals were solubilized in 100% dimethyl sulfoxide (DMSO) at stock concentrations ranging from 5 to 30 mM. Plates were received with chemical identities masked; when single-point screening was completed, identities and actual plated concentrations were provided. Chemical source plates consisted of 165 samples from the two initial PFAS lists of 75 compounds each sourced as PFAS set 1 (EPAPFAS75S1) and set 2 (EPAPFAS75S2) (https://www.epa.gov/chemical-research/pfas-chemical-lists-and-tiered-testing-methods-descriptions) plus an additional 15 PFAS provided to replace chemicals on the initial lists that had known solubility or quality issues. As described in Patlewicz et al., 2019 and Patlewicz et al., 2022, these PFAS were selected from a larger PFAS screening library with a category-based approach to represent a diversity of structures, combined with physicochemical indicators of suitability for tiered toxicity testing (i.e., solubility in DMSO, logKow, vapor pressure). Evaluation of potential toxicity of these sets of PFAS is underway utilizing in vitro screening assays for multiple endpoints including developmental and neurotoxicity, endocrine disruption, immunotoxicity, and general toxicity (e.g., Houck et al., 2021), with a parallel analytical quality assurance (QA) effort to evaluate chemical presence and quality (Smeltz et al., 2023). Briefly, this QA evaluated PFAS stock solution quality and stability using mass spectrometry approaches to determine if the chemical of interest was present in each solubilized stock including confirmation of molecular weight, limited impurities/degradants, and an adequate analyte response, with concentration verification for 25 analytes for which certified standards were available. The 136 compounds reported here are those that passed QA check for stability in DMSO. PFAS excluded from the results reported here include those for which the target chemical could not be detected or there was evidence of degradation that could be due to DMSO (e.g., Perfluoro-2-methyl-3-oxahexanoic acid, Perfluamine, 1-Bromopentadecafluoroheptane), and the molecular weight was incorrect or multiple isomers were detected (e. g., 3H-Perfluoro-2,2,4,4-tetrahydroxypentane, Perfluorocyclohexanecar bonyl fluoride, 3H,3H-Perfluoro-2,4-hexanedione).

Single-Point Screen.

Chemicals were initially tested in the assays at a single high concentration dependent on the concentration received in the stock chemical plates. For hDIO1, hDIO2 hDIO3, XDIO3, hIYD, xIYD, and hTPO the single-point high test concentrations ranged from 50 μM to 300 μM which were achieved by a 1:100 dilution of initial chemical stock into the reaction mixture. Because 1% DMSO caused interference in the hTTR and hTBG assays initial stock samples were diluted 1:200 into the final reaction mixture to achieve 0.5% DMSO and resultant single-point high concentrations ranging from 25 μM to 150 μM. This single high concentration of each chemical was tested on three individual assay plates on the same day, for n = 3 replicate data points. Median values for the single-point screen were calculated with these 3 replicates.

Concentration response testing.

Those chemicals that had activity equal to or >50% in the single-point screen were moved forward to concentration response testing to establish a 50% effect concentration (EC50). For the hDIO1, hDIO2, hDIO3, xDIO3, hIYD, and xIYD assays, chemicals were evaluated using a 7-point concentration response curve with 200 μM being the highest test concentration (Supplemental Table 2). hTBG and hTTR were tested using a 12-point concentration response curve with 150 μM being the highest test concentration (Supplemental Table 2). hTPO was tested using an 8-point concentration response curve with 300 μM being the highest test concentration (Supplemental Table 2). These multiple-point concentration response curves were tested for each chemical on three separate assay plates on the same day for n = 3 replicate data points for each concentration of each chemical. As described in the Data Analysis section below, the concentration response data was analyzed with methods commonly applied to in vitro screening results using all replicates for each concentration of a chemical for the dose-response curve fitting and calculation of EC50.

2.2. hDIO1, hDIO2, hDIO3, xDIO3

The methods for hDIO1, hDIO2, hDIO3 and xDIO3 have been described in detail previously (Hornung et al., 2018; Olker et al., 2019; Mayasich et al., 2021). The model positive control compounds and substrates for the respective assays are summarized in Supplemental Table 1. Briefly, the assays determine chemical inhibition of the enzyme activity as measured by deiodinase-liberated iodide with the Sandell-Kolthoff (SK) reaction in a 96-well plate format. Test chemicals, HEK293 cell lysate with expressed hDIO1, hDIO2, DIO3 or xDIO3, and assay substrate (Supplemental Table 1) were combined in appropriate buffer. The assay plates were sealed, mixed, and incubated for 3 h at 37 °C. Following the incubation, samples were transferred to a 96-well, 2-ml filtration plate containing Dowex 50WX2 (Biotage USA, Charlotte, North Carolina) and the free iodide was eluted into a 96-well collection plate with application of 100 μl of 10% acetic acid. The free iodide was determined by the Sandell-Kolthoff (SK) reaction in which iodide catalyzes the reduction of cerium IV (Ce⁺⁴) which is yellow-colored to the colorless cerium III (Ce⁺³) in presence of arsenic (As⁺³), with the rate of this reaction dependent on the concentration of free iodide (Sandell and Kolthoff, 1937). In all four deiodinase assays, free iodide was detected using the SK reaction following methods previously described (Hornung et al., 2018).

Test chemical responses were normalized on a plate-wise basis, with the DMSO solvent control representing the uninhibited enzyme response (100% of control = 0% chemical activity) and the high concentration of the relevant model inhibitor representing fully inhibited enzyme (0% of control = 100% chemical activity). The model compound was included on each assay plate with graded concentrations to generate a concentration-response curve as a positive control. For hDIO1, propylthiouracil (PTU) was the model inhibitor for the assay and was included at 0.001, 0.033, 0.33, 1, 3.3, 10, 33, 1000 μM. Xanthohumol (XTH) was used at the model inhibitor for hDIO2, hDIO3, and xDIO3 assays, with concentrations of 0.0002, 0.002, 0.02, 0.2, 2.0, 20.0, and 200.0 μM (hDIO2) or 0.000183, 0.001827, 0.01827, 0.1827, 1.827, 18.27, and 182.7 μM (hDIO3, xDIO3). The net rate of change in the colorimetric readout from the SK reaction was calculated by subtracting the background rate of change, and then the solvent control reactions were used to normalize all data to percent of control. Results are reported as percent activity (calculated as 100 minus the percent of control).

2.3. hIYD and xIYD assays

The methods for hIYD and xIYD have been described in detail previously (Olker et al., 2021; Olker et al., 2022). The control compounds and substrate for the respective assay are summarized in Supplemental Table 1. Briefly, the assay measures deiodinase-liberated iodide with the Sandell-Kolthoff (SK) reaction in a 96-well plate format. Test chemicals, HEK293 cell lysate with expressed hIYD or Xenopus laevis liver microsomes and substrate were combined in appropriate buffer. The assay was initiated by adding of the reducing agent NADPH in 1% NaHCO₃. The assay plate was sealed, mixed, and incubated for 3 h at 37 °C. Following the incubation, samples were transferred to a 96-well Dowex 50WX2 (Biotage USA, Charlotte, North Carolina) and the free iodide was eluted into a 96-well collection plate with application of 100 μl of 10% acetic acid. The remaining steps through the SK reaction and assay-specific data processing are the same as those previously described (Olker et al., 2021, Olker et al., 2022).

Test chemical responses were normalized on a plate-wise basis, with the solvent controls (DMSO, NaOH) representing the uninhibited enzyme response (100% of control = 0% chemical activity) and the high concentration of the model inhibitor 3-nitro-L-tyrosine (MNT) representing fully inhibited enzyme (0% of control = 100% chemical activity). The model inhibitor, MNT was included on each assay plate with graded concentrations from 0.0005 to 200 μM (0.0005, 0.005, 0.025, 0.075, 0.16, 1.0, and 200 μM) to generate a concentration-response curve as a positive control. The net rate of change in the colorimetric readout from the SK reaction was calculated by subtracting the background rate, and then the solvent control reactions were used to normalize all data to percent of control. Results are reported as percent activity for a chemical (calculated as 100 minus the percent of control).

2.4. TPO assay

Methods for the Amplex Ultra-Red (AUR) hTPO inhibition assay is based on the work of Paul et al. (2014) and Friedman et al. (2016), except that enzyme was produced in HEK293 cell cultures using an adenovirus vector. Vector AdHM4CMVTPO was a generous gift of Dr. Y. Nagayama (Nagasaki Univ. School of Biomedical Sciences, Japan) (Guo et al., 2003). Briefly, virus was propagated in HEK293 cells under standard conditions (Graham and Prevec, 1991, Hitt et al., 1994). TPO was generated by infecting cells at a multiplicity of infection of >5, and incubating 48 h, at which time cytopathology was extensive, and cells were spontaneously detaching. Cells were scraped, pelleted, and suspended in one tenth volume of growth medium with 7.5% DMSO and frozen.

For chemical screening assays the hTPO cell lysate was thawed, tested for activity using methimazole (MMI) as the model inhibitor, and diluted to achieve approximately 20,000 fluorescence units using the Amplex Ultra Red (AUR) protocol (Invitrogen, Waltham, MA USA). All other reagents were from Sigma (St. Louis, MO USA).

On the day of the test, enzyme was thawed, vortexed gently, and 49 μl added to each well of a 96-well black round bottom plate (Costar, Corning, NY USA). Next 1 μl of test chemical in DMSO was added to each well using a Rainin Liquidator-20 (Rainin, Mettler-Toledo, Oakland, CA USA), followed by 25 μl AUR and 25 μl H₂O₂ to start the reaction. Final reaction volume was 100 μl containing approximately 5 μg hTPO cell lysate, 25 μM AUR, and 300 μM H₂O₂ in 0.2 M HEPES buffer. After AUR addition, all operations were done in low light. Plates were sealed, covered with aluminum foil, and incubated for 30 min at 37 °C. Post incubation, fluorescence was read in a BioTek Synergy Neo-2 plate reader (BioTek, Winooski, VT USA) at 544 nM excitation and 590 nM emission wavelength.

Test chemical responses were normalized to the graded concentrations of MMI from 0.0001 to 1000 μM (0.0001, 0.001, 0.01, 0.1, 1, 10, 100, and 1000 μM) to generate a concentration-response curve as a positive control, with uninhibited enzyme response (0.0001 μM MMI) set to 100% of control and fully inhibited enzyme (1000 μM MMI) set to 0% of control. An aliquot of enzyme was heat-inactivated in a water bath at 90 °C for 2 h to check for background fluorescence. Prior to normalizing based on the positive control, inactivated enzyme background fluorescence was averaged and then subtracted from each well’s total fluorescence. Results are reported as percent activity for a chemical, which was calculated as 100 minus the percent of control.

2.5. hTTR and hTBG binding assay

The methods for using hTTR in PFAS screening were developed based on Montaño et al. (2012). hTTR and all other reagents were sourced from Sigma (St. Louis, MO USA) and used in a fluorescence binding assay using the fluorescent ligand 8-anilino-1-naphthalenesulfonic acid ammonium salt (ANSA; Sigma chemicals) in black round bottom 96-well plates (Costar, Corning NY USA). Each well contained 200 μl reaction volume, with 0.5 μM TTR and 1.2 μM ANSA in 0.1 M phosphate buffer at pH 7.5. The native ligand, T4, was added to the plate in graded concentrations from 0.0149 to 4 μM (0.0149, 0.0332, 0.0738, 0.164, 0.3645, 0.81, 1.8, and 4 μM) to generate a concentration-response curve as a positive control. DMSO as chemical carrier was used at final concentration of 0.5%. Reaction plates were mixed, sealed, shaken for 2 min, and incubated in the dark for 2 h at 4 °C. Post incubation plates were read in a BioTek Synergy Neo-2 plate reader at an excitation wavelength of 380 nm and emission at 475 nM. Chemicals were screened at a single target concentration of 150 μM unless limited by solubility.

hTBG (TBG; Sigma St. Louis, MO USA) was also screened for the ability to bind PFAS. The methods were identical to those for hTTR above, with the exception that each well contained 0.0625 μM TBG and 0.6 μM ANSA, and the T4 reference curve was from 1.36 to 364.5 nM (1.36, 3.03, 6.73, 14.95, 33.22, 73.81, 164.03, 364.5 nM).

Processing and normalization of the hTTR and hTBG data used assay-specific automated pipelines to normalize data, calculate plate diagnostics, and assign assay-specific flags. First, the mean value from wells with hTTR only or TBG only was used to subtract the background fluorescence in each respective assay. Then, test chemical responses were normalized to the concentration-response curve of the model chemical T4, with high concentration T4 (4 μM or 364 nM, for TTR and TBG, respectively) representing 100% of ASNA displacement (0% of control) and low concentration T4 (0.015 μM or 1.36 nM for TTR and TBG, respectively) representing 0% of ANSA displacement (100% of control). Results are reported as percent activity for a chemical, which was calculated as 100 minus the percent of control.

2.6. hTTR native PAGE

Native polyacrylamide gel electrophoresis (PAGE) was performed to test whether PFAS chemicals caused disassociation of the hTTR tetramer in the screening assay. Eighty-nine chemicals that had >50% activity in single-point screening were tested in the native PAGE assay. Reactions were set up the same as the high-throughput screening assay. A reaction without test chemical served as a negative control. hTTR treated with 10 mM sodium dodecyl-sulfate (SDS) was used as a positive control for tetramer disassociation. Reactions were incubated at 4 °C for two hours. Following incubation, 10 μL of reaction mix were added to 10 μl sample buffer. The 20 μl sample was loaded into a 12% Criterion TGX precast midi protein gel (Bio-Rad, Hercules, CA), and gel electrophoresis was performed in TRIS running buffer using the Criterion vertical gel electrophoresis cell (Bio-Rad) for 55 min at 200 V. Native Mark unstained protein standard (Invitrogen, Waltham, MA) was run on each gel. Following electrophoresis, gels were washed three times in reverse osmosis (RO) water for 5 min. Gels were stained using Bio-safe Coomassie stain (Bio-Rad) for 60 min and washed again for 30 min.

2.7. Data analysis

Following the assay-specific data processing and normalization, single-point and concentration response data were submitted to the ToxCast pipeline (tcpl) for analysis and inclusion in the invitroDBv3.5 database (https://epa.figshare.com/articles/ToxCast_Database_invitroDB_/6062623). For this analysis, percent activity was used for the response value, with 20% as the threshold cutoff for both single-point and concentration response. For the concentration response data, all replicates for each concentration of a chemical were included to fit dose-response curves based on three models (constant, constrained Hill, and constrained gain-loss model) with the best model identified based on the lowest Akaike Information Criterion (AIC) value. For each assay, the single-point median % activity of each chemical is reported here. The concentration response results are displayed as % of control (where 0% control = 100% activity), with chemical concentration (log-10 scale) on the x-axis and percent of control on the y-axis, and absolute EC50s calculated from the tcpl results from fit parameters of the best model.

Five chemicals produced responses indicative of assay interference in one or more of the assays (Supplemental Table 3); these chemicals were identified based on comparison to responses produced by the controls and excluded from further analysis. Four iodinated PFAS were identified as interfering with the DIO and IYD assays, based immediate changes in absorbance which suggested presence of free iodide in the sample. Four PFAS were identified as interfering with the TPO assay based on fluorescence value of over 30,000 (compared to DMSO controls with value mean = 22,400 and median = 22,000). Chemical interference with the TTR and TBG assays was identified as those chemicals that produced a response >115% of the median DMSO response, which suggests potential auto-fluorescence. No chemicals were found to interfere with the hTTR assay and only one chemical was found to interfere with the hTBG assay (Supplemental Table 3). Further investigation of assay interference for each of these chemicals was beyond the scope of this study.

3. Results

3.1. hDIO1, hDIO2, hDIO3 and xDIO3

In the initial single-concentration screening, 28 of the test chemicals caused inhibition of the hDIO1 with an activity >20% (Table 1; Supplemental File 1). Nineteen of these chemicals showed inhibition ≥50% and were moved on for concentration response testing, which resulted in 19 chemicals for which EC50 values could be calculated (Table 2). The most potent inhibitor of hDIO1 activity was methyl perfluoro(3-(1-ethenyloxypropan-2-yloxy)propanoate) which had an EC50 of 0.37 μM compared to an EC50 of 0.52 μM for PTU, the model inhibitor (Table 2). The next most potent inhibitor, 1H,1H,6H,6H-perfluorohexane-1,6-diol diacrylate, had an EC50 of 5.71 μM. The inhibition curves for the five most active compounds are presented in Fig. 1a.

Table 1.

Number of active PFAS compounds identified for each assay from the set of 136 PFAS compounds tested.

Molecular Initiating Event	≥ 20% activity at highest test concentration	≥ 50% activity at highest test concentration	Number with EC50
Human Deiodinase 1 (hDIO1)	28	19	19
Human Deiodinase 2 (hDIO2)	34	18	15
Human Deiodinase 3 (hDIO3)	23	13	11
Xenopus Deiodinase 3 (xDIO3)	24	17	12
Human Iodotyrosine Deiodinase (hIYD)	26	13	11
Xenopus Iodothyronine Deiodinase (xIYD)	19	8	6
Human Thyroid Peroxidase (hTPO)	15	4	3
Human Transthyretin (hTTR)	102	82	76
Human Thyroid Binding Globulin (hTBG)	37	9	7

Table 2.

Compounds active in hDIO1, hDIO2, and hDIO3 assays.

		EC50 (μM)
Name	CASRN	hDIO1	hDIO2	hDIO3	xDIO3
Propylthiouracil (PTU)a	51-52-5	0.52	ntc	nt	nt
Xanthohumolb	6754-58-1	nt	0.221	0.157	0.157
Methyl perfluoro(3-(1-ethenyloxypropan-2-yloxy) propanoate)	63863-43-4	0.37	31.8	3.36	11.1
1H,1H,6H,6H-Perfluorohexane-1,6-diol diacrylate	2264-01-9	5.71	11.2	43.6	15.3
1H,1H,9H-Perfluorononyl acrylate	4180-26-1	9.66	18.4		108.4
1H,1H,5H,5H-Perfluoro-1,5-pentanediol diacrylate	678-95-5	12.4	30.0	104.2	28.4
Perfluorotetradecanoic acid	376-06-7	15.8		29.2
Perfluoroundecanoic acid	2058-94-8	16.9	43.4	24.7
Perfluorotridecanoic acid	72629-94-8	24.7	364	35.4
Perfluoro-3,6,9-trioxatridecanoic acid	330562-41-9	25.6	43.7	22.6	70.0
Perfluorooctanesulfonamide	754-91-6	33.5
1H,1H,5H-Perfluoropentyl methacrylate	355-93-1	42.9	42.2	26.8	32.4
Perfluorohexanesulfonamide	41997-13-1	43.2
Perfluorodecanoic acid	335-76-2	51.1	72.1	70.1	126.4
3,3-Bis(trifluoromethyl)-2-propenoic acid	1763-28-6	70.6
11-H-Perfluoroundecanoic acid	1765-48-6	83.2	51.8	116.6	172.7
((Perfluorooctyl)ethyl)phosphonic acid	8022063-9	169.6
Potassium perfluorooctanesulfonate	2795-39-3	174.0	78.3		145.1
Perfluorooctanesulfonic acid	1763-23-1	184.3	107.5		132.5
9-Chloro-perfluorononanoic acid	865-79-2	218.9	128.7
Perfluorononanoyl chloride	52447-23-1	219.6	331.9
N-Ethylperfluorooctanesulfonamide	4151-50-2		178.8
2,2,3,3-Tetrafluoropropyl acrylate	7383-71-3		154.3
Perfluorooctanesulfonyl fluoride	307-35-7			215.1
2-(Perfluorobutyl)ethyl acrylate	52591-27-2				70.3

^aModel positive control compound for inhibition of DIO1 activity.

^bModel positive control compound for inhibition of DIO2 and DIO3 activity.

^cnt = not tested.

Fig. 1.

PFAS chemical inhibition of deiodinase enzymes. 1a = hDIO1, 1b = hDIO2, 1c = hDIO3, 1d = xDIO3. Solid circular symbols (●) show inhibition by the model inhibitor for each enzyme. Error bars represent standard error of three replicate runs.

In the initial single-point screening, 34 of the test chemicals caused inhibition of hDIO2 with an activity >20% (Table 1; Supplemental File 1). Eighteen of these chemicals showed inhibition ≥50% and were evaluated in concentration response testing, resulting in 15 for which EC50 values could be calculated (Table 2). H,1H,6H,6H-perfluorohexane-1,6-diol diacrylate showed the greatest potencey with an EC50 of 11.2 uM. This PFAS was also the second most potent inhibitor of hDIO1. The inhibition curves for the five most potent compounds are presented in Fig. 1b.

The initial single-point screening with hDIO3 resulted in 23 of the test chemicals inhibiting the enzyme by >20% (Table 1; Supplemental File 1). Thirteen of these chemicals showed inhibition ≥50% and were moved on to concentration response testing, with 11 for which EC50s values could be calculated. (Table 2). As with hDIO1 the most potent inhibitor of hDIO3 was methyl perfluoro(3-(1-ethenyloxypropan-2- yloxy)propanoate) with an EC50 of 3.36 uM. The inhibition curves for the five most potent compounds are presented in Fig. 1c.

For the xDIO3, the initial single-point screening identified 24 of the test chemicals that inhibited the enzyme activity by >20% (Table 1, Supplemental File 1). Seventeen of these chemicals showed inhibition ≥50% and were subsequently tested using the concentration-response protocol, resulting in 12 chemicals for which EC50s values could be calculated (Table 2). There were seven common PFAS for which EC50s were calculated for both hDIO3 and xDIO3. As with hDIO1 and hDIO3 the most potent inhibitor of xDIO3 was methyl perfluoro(3-(1-ethenyloxypropan-2-yloxy)propanoate) with an EC50 of 11.1 uM. The inhibition curves for the five most potent compounds are presented in Fig. 1d.

3.2. hIYD and xIYD

In the initial single-point screening 26 of the test chemicals caused inhibition of hIYD by >20% of control activity (Table 1; Supplemental File 1). Thirteen chemicals showed inhibition ≥50% and were moved forward for concentration response testing, resulting in 11 for which EC50 values could be calculated. In general, this group of compounds were not very potent relative to the model inhibitor 3-nitro-L-tyrosine (MNT). The two most potent PFAS had similar potency to each other but were far less potent than MNT. MNT had an EC50 of 0.52 μM, while the two most potent PFAS, perfluoroundecanoic acid and perfluorooctanesulfonic acid, had EC50s of 73.0 μM and 77.5 μM, respectively. The inhibition curves for the five most potent compounds are presented in Fig. 2a.

Fig. 2.

PFAS chemical inhibition of deiodinase enzymes. 2a = hIYD, 2b = xIYD. Solid circular symbols (●) show inhibition by the model inhibitor for each enzyme. Error bars represent standard error of three replicate runs.

Initial single-point screening with xIYD found 19 of the test chemicals caused inhibition of xIYD by >20% of control activity (Table 1; Supplemental File 1). Eight of the chemicals produced inhibition ≥50% and were evaluated in concentration response testing, with six for which EC50 values could be calculated. This group of PFAS also showed weak activity relative to the model inhibitor MNT. The strongest inhibitor of xIYD activity was perfluoro-3,6,9-trioxatridecanoic acid with an EC50 of 91.9 μM compared to an EC50 of 0.52 μM for MNT. The next most potent PFAS inhibitor, perfluorooctanesulfonamide had an EC50 of 93.7 μM. The inhibition curves for the five most active compounds against xIYD are presented in Fig. 2b.

3.3. hTPO

Of the seven MIEs examined in these studies, hTPO was least sensitive to PFAS chemicals. In the initial single-point screening 15 caused inhibition of hTPO with >20% activity (Supplemental File 1). Four of the chemicals which produced inhibition ≥50% were moved on to concentration response testing, resulting in three for which EC50 values could be calculated. In general, this group of compounds were not very potent when compared to the model inhibitor MMI. The most potent inhibitor of hTPO activity was perfluorooctanoyl fluoride which had an EC50 of 76.6 μM compared to an EC50 of 0.14 μM for MMI.

3.4. Thyroid hormone distributor proteins

In the initial single-point screening 102 of the test chemicals competed with ANSA for binding to hTTR with >20% activity (Table 1; Supplemental File 1). Eighty-two of these chemicals showed ≥50% displacement of ANSA and were moved forward for concentration-response testing resulting in 76 chemical for which EC50 values could be calculated (Supplemental Table 4). T4 had an average EC50 of 0.209 μM for displacement of ANSA binding. Three of the test chemicals, perfluoroheptanesulfonic acid, (perfluorobutyryl)-2-thenoylmethane, and perfluorohexanesulfonic acid inhibited ANSA binding at concentrations lower than T4 with EC50s of 0.067 μM, 0.183 μM, 0.184 μM, respectively. The binding competition curves for the five most active compounds are presented in Fig. 3a.

Fig. 3.

PFAS chemical binding to human thyroid hormone distributor proteins transthyretin and thyroid binding globulin. 1a = hTTR, 1b = hTBG. Solid circular symbols (●) show binding by the native ligand T4. Error bars represent standard error of three replicate runs.

Given the high proportion of PFAS shown to have activity in the hTTR binding assay we had concerns that the chemicals may have a nonspecific impact on the hTTR tertiary structure resulting in ANSA displacement. To determine if the PFAS were causing disruption of the hTTR tetramer we used native PAGE to look for potential disassociation of the tetramer. We evaluated all chemicals that had EC50s determined. Reactions were set up as they were in the in vitro screening assay, and under these conditions, dissociation of the hTTR tetramer was not observed (Fig. 4).

Fig. 4.

Native Page gel of PFAS chemicals under assay conditions. Lane 1 Ladder, Lane 2 TTR, Lane 3 disassociated TTR (red box highlights degraded tetramer), Lane 4 TTR + ANSA+DMSO. Lanes 5–18 contain hTTR plus chemical. Lane 5: Perfluoroheptanesulfonic acid. Lane 6: Perfluorohexanesulfonic acid. Lane 7: (Perfluorobutyryl)-2-thenoylmethane. Lane 8: Perfluoro-3,6,9-trioxadecanoic acid. Lane 9: 9-Chloro-perfluorononanoic acid. Lane 10: Perfluoro-1-octanesulfonyl chloride. Lane 11: 8H-Perfluorooctanoic acid. Lane 12: Perfluoro-4-isopropoxybutanoic acid. Lane 13: Perfluorohexanesulfonamide. Lane 14: Perfluorooctanoic acid. Lane 15: Perfluoroheptanoic acid. Lane 16: 1H,1H,5H,5H-Perfluoro-1,5-pentanediol diacrylate. Lane 17: Potassium perfluorooctanoate. Lane 18: Potassium perfluorohexanesulfonate.

Unlike the high proportion of chemicals competitive with ANSA for binding to hTTR, only 37 PFAS in the initial single-point screening caused disruption of binding to human thyroxine binding globulin (hTBG) with >20% activity (Table 1; Supplemental File 1). Nine of the chemicals showed ≥50% displacement of ANSA of which seven produced curves from which EC50 values could be calculated. In general, this group of compounds showed weak activity relative to T4, which had an EC50 of 0.023 μM. The most potent binder, (perfluorobutyryl)-2- thenoylmethane, had an EC50 of 10.6 μM. The binding competition curves for the 5 most active compounds against hTBG are presented in Fig. 3b.

4. Discussion

The USEPA is evaluating the ecological and toxicological effects of perfluorinated chemicals. In support of that effort, we used nine separate in vitro assays to probe MIEs in the thyroid axis. Six of the assays (hDIO1, hDIO2, hDIO3, xDIO3 and hIYD and xIYD) have been previously used in our laboratory to screen chemical libraries (Hornung et al., 2018; Olker et al., 2019; Olker et al., 2021; Mayasich et al., 2021; Olker et al., 2022). Three of the assays, hTPO inhibition, hTBG binding and hTTR binding have been modified from methods published in the literature (Montaño et al., 2012, Paul et al., 2014). The chemicals tested consisted of 136 perfluorinated compounds selected by the USEPA for the initial assessment of perfluorinated alkyl substances (https://www.epa.gov/chemical-research/pfas-chemical-lists-and-tiered-testing-methods-descriptions). These experiments were designed to identify potential thyroid- active chemicals in vitro to prioritize for further testing in vivo at higher levels of biological organization. The chemicals identified here with the greatest binding to carrier proteins (lowest EC50 values) or the lowest EC50 values for enzyme inhibition could be considered along with other bioactivity data, toxicokinetic, and exposure information for screening in models of greater biological complexity.

Deiodinases are a group of enzymes that are involved in the synthesis and catabolism of the thyroid hormones (Luongo et al., 2019). DIO1 and DIO2 are responsible for the deiodination of T4 to the active form of the hormone T3. Deiodinase 3 on the other hand is responsible for the deiodination of T3 to a much less active form of the hormone. The strongest inhibitor of DIO1 activity was methyl perfluoro(3-(1-ethenyloxypropan-2-yloxy)propanoate) with an EC50 of 0.37 μM compared to an EC50 of 0.52 μM for PTU, the model inhibitor. This compound was also the strongest inhibitor of DIO3 activity with an EC50 of 3.36 μM, and fourth strongest inhibitor of DIO2 with an EC50 of 31.7 μM (Table 2). This result would lead one to conclude that of all the chemicals tested, methyl perfluoro(3-(1-ethenyloxypropan-2-yloxy)propanoate) would be the highest priority candidate for in vivo testing to understand the significance of these in vitro results. Of the 19 compounds that inhibited DIO1, 15 of those also inhibited DIO2 and 10 showed inhibition of DIO3 (Table 2). Only two compounds were specific to DIO2 (2,2,3,3-tetrafluoropropyl acrylate and N-ethylperfluorooctanesulfo namide), while only one chemical was specific to DIO3, perfluorooctanesulfonyl fluoride. This is not surprising given the similarity in the thyroid hormone substrates that these enzymes act upon. This finding is consistent with previous reports of chemical activities on these enzymes in vitro (Hornung et al., 2018; Olker et al., 2019).

TPO is an enzyme critical for the synthesis of thyroid hormone within the thyroid gland and is known to be sensitive to chemical inhibition resulting in disruption of thyroid hormone synthesis leading to adverse effects on development in a range of species (Köhrle and Frädrich, 2021). Only six compounds of the 136 tested here inhibited enzyme activity by >50% of control. The most active chemical tested, perfluorooctanoyl fluoride, had an EC50 of 76.5 μM compared to the model TPO inhibitor, MMI whose EC50 was 0.14 μM. The low number of PFAS identified as potential actives, and low potency of the most active compound suggests that TPO is not a particularly sensitive target of this group of compounds.

IYD is an enzyme found in the thyroid gland, liver and kidney of vertebrates and is involved in the maintenance of iodine in living organisms. IYD’s function is to recycle iodine from monoiodo- and diiodotyrosine (MIT and DIT) within the thyroid gland, liver, and kidney to ensure maintenance of this limited essential trace element (Moreno and Visser, 2010). Thirteen chemicals showed inhibition of greater than of 50% of control activity with the most potent, perfluoroundecanoic acid, having an EC50 of 73.0 μM compared to the EC50 of 0.023 μM for MNT, the model inhibitor used here. As with thyroid peroxidase, IYD does not appear to be a sensitive target of the perfluorinated chemicals tested in these studies.

In most mammals there are three major thyroid hormone distributor proteins (THDPs), serum albumin, TBG, and TTR (for review see McLean et al., 2017, Janssen and Janssen, 2017, Rabah et al., 2019). Their roles are to bind thyroid hormones, thus controlling transport throughout the body and limiting the supply of available hormone to specific tissues. Disruption of thyroid hormone binding can have significant impact on thyroid homeostasis and development (Rabah et al., 2019). We examined PFAS disruption of binding to hTBG thyroxine and hTTR. In our screening studies, over half of the PFAS tested (80 of 136) bound to hTTR with a potency sufficient to generate EC50’s for completive displacement of ANSA binding (Table 1). Three compounds perfluoroheptanesulfonic acid, (Perfluorobutyryl)-2-thenoylmethane perfluorohexanesulfonic acid had greater affinity to bind than T4. Worth noting is perfluorohexanesulfonic acid that has previously been shown to impact the thyroid hormone system in vivo in rodent studies (Butenhoff et al., 2009a, 2009b, Ramhøj et al., 2018, Ramhøj et al., 2020). Investigators Ren et al. (2016) and Weiss et al. (2009) have tested several of the same PFAS and also found that they bind hTTR. Table 3 shows the EC50s obtained using human hTTR, and although there were differences in the methods used by each research team, the values are generally within 2- to 5-fold of each other. Ren et al., 2016 used a fluorescence polarization based competitive binding assay with a fluorescent T4 probe, whereas Weiss et al. (2009) used ¹²⁵I-T4 as the binding ligand. There are several other literature reports of PFAS binding to TTR. Perfluoroheptanesulfonic acid, the chemical with highest affinity for hTTR in our experiments, was found by Zhang et al. (2015) to have an EC50 of 0.64 μM compared to our value of 0.09 μM. Perfluorohexanesulfonic acid also had high affinity for hTTR in our assay, but there are no literature reports on binding of this to hTTR. The related salt form, potassium perfluorohexanesulfonate was found to bind to hTTR (Ren et al. (2016) and Weiss et al. (2009)) (Table 4). Following the top two strong hTTR binders, four other PFAS, (perfluorobutyryl)-2-thenoylmethane, perfluoro-3,6,9-trioxadecanoic acid, 9-chloro-perfluorononanoic acid, and perfluoro-1-octanesulfonyl chloride also had EC50 values similar to T4. To our knowledge, the observation that these chemicals interact with human hTTR in vitro may be the first such report. This is true for many of the other active chemicals reported here for which no other literature references related to hTTR binding were found. The in vivo biological significance of the data presented here is not yet clear. Further investigation using orthogonal in vitro assays and in vivo exposures may be necessary to determine the importance of these observations with TTR.

Table 3.

Comparison of PFAS binding to human TTR in this study and two literature sources.

Chemical Name	CASRN	EC50 (μM) this study	EC50 (μM) Ren et al., 2016.	EC50 (μM) Weiss et al., 2009.
Perfluoroheptanoic acid	375-85-9	0.40	1.128	1.565
Perfluorononanoic acid	375-95-1	0.41	1.977	2.737
Potassium perfluorooctanesulfonate	2795-39-3	0.43	0.13	0.94
Perfluorooctanoic acid	335-67-1	0.44	0.378	0.949
Potassium perfluorohexanesulfonate	3871-99-6	0.49	0.594	0.717
Potassium perfluorobutanesulfonate	29,420-49-3	0.71	13.33	nta
Perfluorodecanoic acid	335-76-2	0.71	1.623	8.954
Perfluorooctanesulfonamide	754-91-5	0.73	nt	6.124

^aNot tested = nt.

Table 4.

PFAS compounds screened in this study and reported in published literature to disrupt mammalian thyroid hormone system in vivo.

			In vivo reference	Assays in which chemical was active (≥20% activity)
Name	Acronym	CASRN		hDIO1	hDIO2	hDIO3	hlYD	hTTR	hTBG	hTPO
Perfluorobutanesulfonic acid	PFBS	375-73-5	NTP, 2019	–a	–	–	–	Xb	–	–
Potassium perfluorobutanesulfonate	K-PFBS	29,420-49-3	Ladics et al., 2008	–	–	–	–	Xb	–	–
Potassium perfluorohexanesulfonate	K-PFHxS	3871-99-6	Ramhøj et al., 2020, Butenhoff et al., 2009a, b, Ramhøj et al., 2018	–	–	–	–	Xb	–	–
Perfluorooctanesulfonic acid	PFOS	1763-23-1	Dong et al., 2016	Xb	Xb	–	Xb	Xb	X	X
Potassium perfluorooctanesulfonate	PFOS-K	2795-39-3	Seacat et al., 2002, Chang et al., 2008, Thibodeaux et al., 2003, Lau et al., 2003, Yu et al., 2009, Yu et al., 2011, Chang et al., 2009, Martin et al., 2007, Curran et al., 2008, Luebker et al., 2005, Wang et al., 2011	Xb	Xb	–	Xb	Xb	X	X
Perfluorobutanoic acid	PFBA	375-22-4	Butenhoff et al., 2012	–	–	–	–	Xb	–	–
Perfluorohexanoic acid	PFHxA	307-24-4	NTP, 2019	–	–	–	–	Xb	–	–
Perfluorooctanoic acid	PFOA	335-67-1	NTP, 2019	–	X	–	X	Xb	X	–
Ammonium perfluorooctanoate	APFOA	3825-26-1	Martin et al., 2007, Butenhoff et al., 2002	–	–	–	X	Xb	X	–
Perfluorononanoic acid	PFNA	375-95-1	NTP, 2019	X	X	–	Xb	Xb	X	X
Perfluorodecanoic acid	PFDA	335-76-2	Gutshall et al., 1989, Langley and Pilcher, 1985, Van Rafelghem et al., 1987, NTP, 2019	Xb	Xb	Xb	Xb	Xb	X	–
Perfluorotetradecanoic acid	PFTEDA	376-06-7	Hirata-Koizumi., 2015	Xb	X	Xb	–	Xb	X	–
8:2 Fluorotelomer alcohol	8:2 FTOH	678-39-7	Ladics et al., 2008	–	X	–	–	–	–	–

Chemicals were deemed in vivo active if changes in hormone levels or thyroid gland histopathology were reported.

^aNot active (– -) in this screening study (<20% activity).

^b≥50% activity and EC50 could be calculated in concentration response testing.

Fewer of the PFAS tested bound to hTBG than to hTTR, and with much lower affinity. Marchesini et al. (2008) tested perfluorooctanesulfonate for binding to hTBG and found none. Ren et al. (2016) tested 16 PFAS for binding to hTBG and found only two (perfluorotridecanoic acid and perfluorotetradecanoic acid) that showed weak binding. The different binding patterns of chemicals to hTTR and hTBG might be due to differences in their T4 binding pocket, which is a hydrophobic channel in hTTR versus a surface pocket in TBG (Hamilton et al., 1993; Zhou et al., 2006). Recent studies in our laboratory testing the ToxCast Phase 1, ToxCast Phase 2 and e1K libraries also showed unexpectedly high hit rates for hTTR (unpublished data). These libraries consist of 1813 chemicals in total and we found that 61% were active at >20% in single concentration screening, with 35% inhibiting at >50%.

Of the 136 chemicals tested, a total of 85 had an EC50 value in at least one of the assays. The nine assays discussed here are by no means a comprehensive list of the possible MIEs within the thyroid axis which could be impacted by interaction with xenobiotic chemicals, although the assays reported here are those of highest priority (Noyes et al., 2019). As an initial step to understand the utility of this data set to inform potential impacts in animals, we conducted a search of the literature and found 13 PFAS for which there is strong evidence of impacts on the thyroid hormone system in mammalian studies (Table 4). Chemicals were deemed in vivo active if changes in hormone levels or thyroid gland histopathology were reported. All 13 in vivo active compounds showed activity in at least one of the assays reported here. Twelve of the in vivo actives showed in vitro activity towards hTTR, nine of which at sub-micromolar concentrations. Seven of the chemicals, potassium perfluorooctanesulfonate, perfluorononanoic acid, perfluorooctanesulfonic acid, Perfluorooctanoic acid, perfluorotetradec anoic acid, perfluorodecanoic acid, and 8:2 Fluorotelomer alcohol showed inhibition of a least one of the DIOs although the activity was relatively low compared to the model inhibitors. Seven of the in vivo actives also had weak interaction with hTBG. Thus, activity in the nine assays reported here is associated with known thyroid effects in vivo, but the nature and strength of the relationships are yet to be elucidated.

A well-studied chemical like PFOA has various documented in vivo effects in multiple species; rats (Cai et al., 1996), rainbowfish (Miranda et al., 2020), quail (Smits and Nain, 2013) and others. Yet in our tests PFOA only showed activity <50% in the hTTR binding assay. Another well studied chemical with known in vivo effects, K-PFOS, was active in six of the human based assays, hDIO1, hDIO2, xDIO3, hIYD, hTTR, and hTPO. Perfluoro-3,6,9-trioxatridecanoic acid has no literature on in vivo effects yet it was active in eight of the nine assays presented here (Supplemental File 1). This may suggest a greater likelihood that it could produce in vivo effects but given the examples of PFOA and PFOS above the number of assays activated may not, in itself, be particularly informative.

5. Conclusions

Multiple high throughput screening assays have met proof of concept for testing thyroid MIEs and can be used to screen more PFAS. They can also be applied to other chemical sets of interest to the USEPA or to other investigators. These experiments are not designed to evaluate perfluorinated chemical effects at concentrations found in the environment, but rather to prioritize chemicals for further testing at higher levels of biological organization. The chemicals identified here with the highest relative binding affinity to carrier proteins or the lowest EC50 values for enzyme inhibition can be considered for elevation to higher tier testing in mammalian and amphibian cell based and in vivo testing models. The fact that many of these PFAS were not active in these assays or activity was at relatively high concentrations compared to the positive control, suggests many of the PFAS may be of low potential for thyroid hormone system impacts. However, we also recognize that this suite of nine assays does not provide complete coverage of potential thyroid hormone system MIEs, and importantly, other MIEs such as enzymes and transporters related to liver metabolism and elimination of thyroid hormones are not covered in the suite of assays presented here. Further evaluation and analysis of thyroid hormone system impacts of PFAS is warranted to identify potential active chemicals and to develop a better understanding of the best use of in vitro screening assays for their predictive capacity for in vivo adverse effects.

This research was conducted under an animal care plan approved by the Institutional Animal Care and Use Committee of the EPA Office of Research and Development.

Data availability

Data will be made available on request.