Screening the ToxCast Chemical Libraries for Binding to Transthyretin

Abstract

Transthyretin (TTR) is one of the serum binding proteins responsible for transport of thyroid hormones (TH) to target tissue and for maintaining the balance of available TH. Chemical binding to TTR and subsequent displacement of TH has been identified as an endpoint in screening chemicals for potential disruption of the thyroid system. To address the lack of data regarding chemicals binding to TTR, we optimized an in vitro assay utilizing the fluorescent probe 8-anilino-1-napthalenesulfonic acid (ANSA) and the human protein TTR to screen over 1,500 chemicals from the U.S. EPA’s ToxCast ph1_v2, ph2, and e1k libraries utilizing a tiered approach. Testing of a single high concentration (target 100 μM) resulted in 888 chemicals with 20% or greater activity based on displacement of ANSA from TTR. Of these, 282 chemicals had activity of 85% or greater and were further tested in 12-point concentration-response with target concentrations ranging from 0.015–100μM. An EC50 was obtained for 276 of these 301 chemicals. To date, this is the largest set of chemicals screened for binding to TTR. Utilization of this assay is a significant contribution towards expanding the suite of in vitro assays used to identify chemicals with the potential to disrupt thyroid hormone homeostasis.

Introduction

Thousands of chemicals are developed each year adding to the thousands of chemicals already in use ^{1, 2}. It has been demonstrated that exposure to environmental contaminants can have adverse biological impacts; thus, it is necessary to characterize the risks of these chemicals ^{3, 4}. As the number of chemicals grows each year, it becomes increasingly prohibitive to test them for effects in vivo. As such, regulatory agencies are turning to new approach methodologies (NAMs), including in vitro high throughput screening, to assess a chemical’s potential to interact at a specific molecular target and to identify endocrine disrupting chemicals (EDCs) ^{5, 6}. The US Environmental Protection Agency’s (US EPA) Endocrine Disruptor Screening Program has focused on identifying EDCs that interfere with the estrogen, androgen, and thyroid systems ⁷. Much work has been done in validating in vitro assays to test chemical interactions at the estrogen and androgen receptors to prevent reproductive impacts ^8–10; however, progress in developing new methods for targets within the hypothalamus-pituitary-thyroid (HPT) axis has been slower due to complexity of the thyroid system and multiple targets for potential disruption ^{7, 11}.

The HPT axis is tightly regulated due to its importance in growth, development, and metabolism ^{12, 13}. There are several targets of potential interference along this HPT axis as well as in serum binding proteins and peripheral tissues ^{12, 14–16}. Further, a feedback system signaling when to turn on and off production of thyroid hormone (TH) is in place to maintain TH homeostasis ^{17, 18}. The complexity of this system requires development of in vitro assays for each of the targets to get a comprehensive view of what may be occuring in vivo. To date, several in vitro assays have been developed and used to screen chemicals for interference in processes related to hypothalamic and pituitary signaling, thyroid hormone synthesis, and thyroid hormone signaling. Specifically, the thyroid-relevant assays used to screen the ToxCast chemical libraries include thyroid hormone receptor ^19–22, thyroid stimulating hormone receptor ^{23, 24}, thyrotropin releasing hormone receptor ²⁰, thyroid peroxidase ^{25, 26}, sodium iodide symporter ^{27, 28}, iodothyronine deiodinases ^29–33, and iodotyrosine deiodinase ^{34, 35}. An area not currently covered by these assays are the thyroid hormone distributor proteins.

Transthyretin (TTR), also called prealbumin, is one of the serum binding proteins responsible for delivering thyroid hormones to target tissues and maintaining the balance of free versus bound TH ^36–38. TTR is synthesized in the liver and in the choroid plexus for TH transport in plasma and cerebral spinal fluid ^{36, 39}. TTR is suggested to mediate delivery of TH across the blood-brain barrier and through the placenta ⁴⁰. TTR is a homotetramer with two TH binding sites; however, due to negative cooperativity it is understood to bind only one TH at a time ^41–43. In addition to TH, TTR can bind retinol binding protein, which in turn binds and transports retinol ⁴⁴. Chemical binding to TTR can displace TH from the protein resulting in lack of TH delivered to the target tissue and subsequent disruption of biological pathways dependent on TH activation ⁴⁵. Competitive binding with TTR in cerebrospinal fluid has been shown in vitro to increase free T4 which may negatively regulate production of TH ⁴⁶. Alternatively, TH freed from TTR could be metabolized and excreted resulting in decreased TH levels in serum ^{47, 48}. Assays have been developed to assess chemical binding to TTR. These include a radiolabeled ligand displacement assay ⁴⁹ and the fluorescence-based competitive binding assays using fluorescein isothiocyanate-T4 conjugate (FITC-T4) or 8-anilino-1-naphthalenesulfonic acid ammonium salt (ANSA) ^50–52. The FITC and ANSA assays are currently being validated by the European Union Reference Laboratory for Alternatives to Animal testing ⁵³ tsar.jrc.ec.europa.eu/test-method/tm2019-08).

The objective of this work was to screen the ToxCast ph1_v2, ph2, and e1k libraries ⁵⁴ in the TTR-ANSA assay using a tiered approach to: 1) identify chemicals with the potential to bind TTR, 2) determine an EC50 value for the chemicals that bind to TTR and displace ANSA based on 85% activity or greater in single concentration screening, and 3) evaluate the performance of the assay for use in large-scale screening.

Materials and Methods

Overview

Development of this TTR binding assay for use in our lab was based on methods developed by ⁵¹ to identify chemicals that bind to the protein by measuring a reduction in fluorescence due to displacement of ANSA. The assay was optimized in a 96-well plate with a fluorescence readout that measures displacement of the probe from the protein. The assay used human TTR, 8-anilino-1-naphthalene sulfonic acid (ANSA) as the fluorescent probe, dimethyl sulfoxide (DMSO) as the negative control, and L-thyroxine (T4) as the positive control. TTR, ANSA, DMSO, and T4 were purchased from Sigma-Aldrich (St Louis, MO). The screening format followed ²⁹, ³¹, ³⁴, and ⁵⁵ with initial screening at a single concentration (target of 100 μM) followed by concentration-response testing for a subset of chemicals. Most chemicals screened in concentration-response were tested from 0.015 to 100 μM but overall concentrations ranged from 0.004 to 150 μM due to chemicals being plated in source plates at different concentrations (details in “Chemicals” section).

Chemicals

1813 unique chemicals from the ToxCast libraries ⁵⁴ were obtained from Evotec (Branford, CT) via the ToxCast program. The set included chemicals from the ph1_v2, ph2, and e1k libraries. The chemicals were supplied across twenty-five 96-well chemical source plates with one chemical per well. Quality control of these chemicals relied on requirements of the ToxCast library [e.g., identification, purity, analytical verification ⁵⁴]. Chemical identities were masked until single concentration screening was complete. Chemicals were provided at a target concentration of 20 mM in DMSO; however, for some chemicals the plated concentration differed from the target concentration due to solubility limitations in DMSO (120 chemicals <20 mM, 22 chemicals >20 mM). Further, the concentration for 12 chemicals (oils and mixtures) was provided in mg/mL. Overall, 93.4% of chemicals were provided within 1 mM of the target concentration.

In addition to test chemicals, DMSO, T4, and TTR were screened on each plate. DMSO, the solvent control, served as a negative control due to its inability to displace ANSA from TTR. T4 is a known ligand of TTR and served as a positive control. The final amount of DMSO in the DMSO and T4 controls was 0.5% to match the amount added in test chemical reactions. A TTR protein-only control was included to correct fluorescence values. Source plates were received with column 1 and several random wells left empty for addition of reference controls. A T4 standard curve was added in column 1 using a 10 mM stock solution prepared in 0.4 N NaOH and diluted to final concentrations of 1.8, 0.81, 0.3645, 0.164, 0.0738, 0.0332. 0.0149, and 0.0067 μM. The maximum T4 concentration (1.8 μM) served as the positive control correlating with complete displacement of ANSA from TTR. T4 concentrations were selected based on assay optimization, and the low concentration of T4 (0.0067 μM ) was selected such that fluorescence was comparable to fluorescence observed in the DMSO negative control reactions. Each replicate plate contained the T4 curve, additional high T4 and low T4 wells, DMSO, and TTR. Plate layouts for single concentration screening and concentration-response testing are in Supplemental Figure 1. A schematic diagram of assay setup is shown in Figure 1A.

Figure 1.

Diagram of assay setup (A) and a depiction of how the assay works (B). When bound to TTR, ANSA fluoresces at an emission wavelength of 475 nm. When T4 is added to the reaction, ANSA is displaced from the protein resulting in reduced relative fluorescence units (RFUs). RFUs from sample chemical wells are compared to the T4 curve to determine binding activity of chemicals to TTR.

Autofluorescence

Reduction in fluorescence is used to determine displacement of ANSA from TTR (Figure 1B); thus, autofluorescence of test chemicals is a potential interference of this assay. Specifically, one concern would be that autofluorescence of the chemical would mask any reduction in ANSA fluorescence resulting in a false negative response. To address interference, 1 μL of each chemical was added to 199 μL of phosphate buffer, incubated for 2 hours at 4°C, and fluorescence was measured using a BioTek Synergy Neo2 plate reader (Agilent, Santa Clara, CA) with an excitation wavelength of 380 nm and an emission wavelength of 475 nm.

Binding assay optimization

Several iterations of the binding assay were performed while optimizing this assay for use in our lab. Initial testing followed the parameters outlined by ⁵¹ with reactions containing 0.5 μM TTR, 0.6 μM ANSA, and a T4 curve ranging from 100–1000 nM. Additionally, we assessed the difference in using clear vs black 96-well plates and flat vs round bottom plates in an effort to minimize background fluorescence. Black round-bottom plates (Corning 3356) were selected for use as assay plates. Further optimization included testing different concentrations and ratios of TTR and ANSA with the goal of reducing the amount of TTR in each reaction to ultimately lower the cost of the assay.

Assay conditions

Each 200 μL reaction contained 0.125 μM TTR, 1.2 μM ANSA, and 0.5% DMSO in 0.1 M phosphate buffer, pH 7.5. Four individual master mixes were prepared: a main master mix for test chemicals which contained all components except DMSO, the T4 and DMSO master mixes which contained all components, and a TTR master mix which did not contain ANSA (Supplemental Table 1). Master mixes were prepared in low light to prevent ANSA degredation. Each master mix was loaded into the appropriate wells of a 2 mL 96-well reservoir plate. Using a Biomek i5 liquid handler (Beckman Coulter, Indianapolis, IN), six 96-well black assay plates were loaded with master mix. Subsequently, 2 μL T4 and 1 μL test chemical were added to the assay plates using a 20 μL Rainin Liquidator 96-channel benchtop pipettor (Mettler Toledo, Columbus, OH). Reactions were mixed by pipetting using a 200 μL Rainin Liquidator, sealed with a polyolefin plate seal, and shaken on a Jitterbug microplate shaker (Boekel Scientific, Feasterville, PA) for 2 minutes at 1000 rpm. Assay plates were covered with foil and incubated at 4°C for 2 hours. Following incubation, fluorescence was measured using a BioTek Synergy Neo2 plate reader (Agilent, Santa Clara, CA) using an excitation wavelength of 380 nm and an emission wavelength of 475 nm (Figure 1).

Chemical screening

Chemical screening was performed using a tiered approach. The first tier was comprised of screening each chemical at a single concentration. The target concentration for single concentration screening was 100 μM with some variation based on solubility. Chemical source plates were thawed before use and were used to load three individual assay plates (n=3 for each chemical, one well on each assay plate; see Supplemental figure 1A for plate layout). Additionally, a “T4 source plate” was prepared. The T4 source plate contained an adequate volume of each concentration of the T4 standard curve and the additional high and low T4 concentration wells. The T4 source plate was used to load individual assay plates using the Liquidator as described previously. Data were normalized to percent of control with the low concentration of T4 representing maximum fluorescence (no displacement of ANSA) and the highest concentration of T4 representing no fluorescence (100% ANSA displaced). In single concentration screening, chemicals were considered inactive if they displayed less than 20% activity compared to the high T4 concentration control.

A subset of chemicals with the greatest activity moved forward to concentration-response testing. Specifically, the 282 chemicals with 85% or greater median displacement of ANSA were tested in 12-point concentration-response curves in an effort to establish EC50s for these chemicals binding to TTR thereby displacing ANSA. The 85% activity level for moving chemicals forward for concentration-response testing differed from previous thyroid screening assays (e.g., deiodinases used a 50% activity threshold). In this assay, a large proportion of chemicals showed activity >50%; thus, selecting 85% activity as the threshold for moving chemicals onto concentration-response testing provided better discrimination from a prioritization standpoint. For concentration-response testing, chemicals were taken from the original chemical source plate and aliquoted in a new 96-well polypropylene plate (“CR source plate”) to prepare dilutions in DMSO. Chemicals were tested in 12-point curves with target concentrations of 0.015, 0.034, 0.076, 0.168, 0.374, 0.83, 1.845, 4.101, 9.113, 20.25, 45, and 100 μM, one chemical per row with different concentrations in each column tested in three separate assay plates (n=3 for each chemical at each concentration, see Supplemental Figure 1B for plate layout).

Data Processing and Analysis

Data were processed and analyzed in R [version 4.2.1, R Core Team ⁵⁶] using an automated pipeline that normalized data, calculated plate diagnostics, and assigned assay-specific flags. Data were processed by: 1) adjusting relative fluorescence units (RFUs) by subtracting fluorescence in the TTR-only wells, and 2) normalizing test chemical responses to the concentration-response curve of the model chemical T4 fit to the Hill model with high concentration T4 (1.8 μM) representing 100% of ANSA displacement (0% of control) and low T4 (0.0067 μM) representing 0% ANSA displacement (100% of control). Results are reported as percent activity which was calculated as 100 minus the percent of control. Results for single concentration screening are reported as percent activity calculated from the median of the three replicates with range used to display variability.

Concentration-response data were analyzed using the ToxCast Analysis Pipeline (tcpl) package version 3.0.1 ⁵⁷ using the tcplLite option. In this analysis, percent activity was used as the response value and 20% activity as the threshold cutoff. This 20% cutoff was selected to have a consistent threshold across assays. The three replicates for each concentration of a chemical were included to fit dose-response curves based on the constant, constrained Hill, and constrained gain-loss model. The best model fit was identified based on the lowest Akaike Information Criterion (AIC) value. For chemicals that fit the Hill or gain-loss model, the EC20, EC50, and Hill slope were calculated from the model fit parameters. Results from concentration-response screening are displayed as displacement from the maximum response with chemical concentration (log10 scale) on the x-axis and percent activity on the y-axis.

Assay quality

Quality control measures were calculated following methods previously described ^{31, 34}. Briefly, variability and separation between the positive control (1.8 μM T4) and solvent/negative control (DMSO) were evaluated with plate-wise Z’ factors. The Z’ factor is an indicator of assay quality, and values above 0.5 indicate good separation between positive and negative controls ⁵⁸. A Z’ factor greater than or equal to 0.5 was used as a guideline for acceptable plate runs. Typically assay plates were rerun to replace data from plate runs where quality criteria were not met (e.g., low Z’ factor, poor control data). In one instance, a Z’ factor less than 0.5 was not re-run because the two other replicates for that plate met the quality criteria.

Data that fell outside of acceptable parameters for variability and fluorescence were flagged for manual review. Flags included high variability when the absolute difference between the mean and median of three replicate runs was greater than 10% and potential assay interference indicated by fluorescence outside of normal ranges. Individual wells with observed problems (e.g., outliers, low volume in a well in one replicate) were excluded from analyses with only two replicates used.

Results

Assay Optimization, Performance, Quality Control

Optimization of the TTR binding assay in our lab resulted in a fluorescence-based assay suitable for screening large chemical libraries for binding to the human TTR protein. To evaluate data quality, standard curves, Z’ factor, and variability of the controls were evaluated for each assay plate and summarized across all plates. Quality control metrics were consistent over the duration of data collection in triplicate testing of 25 single concentration plates and 52 concentration-response plates (Supplemental Table S2). Standard curves for the positive control (T4) were consistent with mean EC50 of 0.075 μM ± 0.02 SD and mean Hill slope of −1.113 ± 0.29 SD. Standard curves for all replicates had a Z’ factor >0.5 except for one plate which failed acceptance criteria with a Z’ factor of 0.302. The low Z’ factor in this plate was attributable to unusually low fluorescence in a DMSO control well; however, the two other replicate plates had Z’ factors >0.5, and this set of plates was a repeat run with data similar to the initial assay run. Additionally, the mean Z’ factor across all plates was 0.883± 0.06 SD and ranged from 0.542 to 0.98 when the one replicate plate was excluded.

Autofluorescence

The assessment of chemical autofluorescence was done after screening all chemicals in the TTR assay to better understand an allowable threshold for chemical autofluorescence. For this analysis, the threshold for allowable chemical autofluorescence was set at 10% of maximum reaction fluorescence (i.e., 10% of the average DMSO negative control fluorescence across all single concentration and concentration-response plates). The average maximum fluorescence in DMSO wells across all plates was 10263, so the threshold for chemical autofluorescence was 1026.3. Based on this threshold, 262 chemicals were determined to be autofluorescent (Figure 2). The 262 chemicals interfering with the assay are listed in Supplemental Table S3. Additionally, 39 chemicals (Supplemental Table S4) were flagged during data processing due to fluorescence being more than 20% greater than DMSO negative controls (i.e., RFUs > DMSO+20%). Note, the flag for RFUs > DMSO=20% was intended to identify wells with extreme fluorescence values, but in the case of these 39 chemicals all three replicates were flagged resulting in their exclusion from further analysis. One explanation for this could be that these chemicals fluoresce when bound to TTR, similar to ANSA, which also explains why they were not identified in the screening of chemicals for autofluorescence. Due to these exclusions, results from 1512 of the 1813 chemicals are reported herein.

Figure 2.

Fluorescence of chemicals from the ToxCast phase 1_v2, phase 2, and e1k libraries at a target concentration of 100 uM. Chemicals are ordered from low to high according to the average relative fluorescence units (RFUs) of three replicate plates. The dashed line denotes 10% of the maximum fluorescence value from DMSO wells across all plates, which was the maximum chemical autofluorescence set for this study. Chemicals above the dashed line (i.e., those with RFUs >1026) were excluded from further analysis. There were 68 additional chemicals not depicted on the plot due to RFUs ranging from 10,200 to “overflow” (≥100,000).

Single Concentration Screening

In single concentration screening, 624 chemicals had <20% activity compared to the activity of high concentration T4 controls and were considered inactive (Table 1, Figure 3). Of the 888 chemicals with activity of 20% or greater, 282 had 85% or greater activity and were moved onto concentration-response testing (Supplemental Table S5).

Table 1.

Summary of single point screening at an initial target concentration of 100 μM: percent and number of chemicals in each library that produced less than 20% or greater than/equal to 20%, 50%, and 85% activity.^a

Chemical library	# chemicals testedb	% with < 20% activity (#)	% with ≥ 20% activity (#)	% with ≥ 50% activity (#)	% with ≥ 85% activity (#)
ToxCast ph1_v2	262	20.6 (54)	79.4 (208)	47.3(124)	17.6 (46)
ToxCast ph2	578	46.5 (269)	53.5 (309)	34.4 (199)	20.2 (117)
ToxCast e1k	672	44.8 (301)	55.2 (371)	32.3 (217)	17.7 (119)
Total	1512	41.3 (624)	58.7 (888)	35.7 (540)	18.7 (282)

^aChemical names, CAS registry numbers, maximum tested concentrations, and median % activity produced can be found in Supplementary Table S3.

^bChemical plates received from ToxCast included 1813 unique chemicals; however, 301 compounds had evidence of interfering with the TTR assay and were excluded from all summaries and analyses.

Figure 3.

Displacement of ANSA from transthyretin (TTR) by chemicals in single-concentration screening of the ToxCast phase 1_v2, phase 2, and e1k libraries at a target concentration of 100 uM. Results are displayed as percent of control with median (black square) and range (line) of three replicates. (A) All chemicals, (B) Expanded plot of chemicals with ≥85% displacement (≤15% of control; displacement is 100-percent of control). Chemicals are plotted by rank order based on median % of control from low to high. Chemicals further tested in concentration-response screening are those below the orange dashed line which denotes 85% displacement.

Concentration-response Testing

Results from concentration-response testing confirmed single concentration screening results for 96% of the 282 chemicals tested. Results were considered consistent when the percent difference between median percent activity in single concentration screening and median percent activity at the maximum concentration tested in concentration-response was less than 25%. Two of the chemicals, prednisone and tripropylene glycol monomethyl ether, were active in single concentration screening and had less than 20% activity at the highest concentration tested in concentration-response. Different sets of source plates were used for single concentration and concentration-response testing; thus, it is possible that there was an error in dosing in either of the assays. The remaining nine chemicals that were inconsistent ranged from 86–102% activity in single concentration screening and 32–73% activity at the maximum concentration tested in concentration-response.

The Hill model was identified in the ToxCast pipeline as the best fit model for 230 of 282 chemicals, and the gain-loss model was the best fit model for 50 chemicals. EC50 values were obtained for 276 of the chemicals screened in concentration-response (Supplemental Table 6, Supplemental Figure 2). For the 6 chemicals with no EC50 reported, two were inactive at all concentrations tested (prednisone and tripropylene glycol monomethyl ether) and the constant model was the winning fit. The other four chemicals (citronellol, 2-butyloctan-1-ol, 1,3,5,7−tetramethyl−1,3,5,7−tetravinylcyclotetrasiloxane, and difenzoquat) did not reach 50% activity at the highest concentration tested in concentration-response.

The 25 most potent chemicals, ranked by EC50, are listed in Table 2. Twenty-four of the chemicals had EC50 values less than or equal to the mean EC50 of T4 indicating that these chemicals are equally or more potent as T4 in binding to TTR. Example curves for selected chemicals are displayed in Figure 4. Complete concentration-response curves were not achieved for the nine chemicals presented from the top twenty-five ranked inhibitors. These chemicals were more (Figures 4B-I) or equally (C.I. disperse Black 6 dihydrochloride, Figure 4J) potent compared to T4 based on EC50. These chemicals could have been tested at even lower concentrations to achieve 100% of control (zero activity) and establish a full concentration-response curve.

Table 2.

The top 25 ranked chemicals for binding to transthyretin (TTR) based on EC50.

Rank	Chemical	CASRN	Max Tested Conc., μM	% Activity at Max Conc.a	EC20 (μM)	EC50 (μM)
1	Tetrac	67–30-1	100	107.7	0.011	0.017
2	o-Aminoazotoluene	97–56-3	100	109.4	0.007	0.018
3	4,6-Dinitro-o-cresol	534–52-1	100	109.5	0.004	0.020
4	Tiratricol	51–24-1	100	104.7	0.008	0.021
5	Bromoxynil	1689–84-5	80	103.5	0.007	0.022
6	3,3',5,5'-Tetrabromobisphenol A	79–94-7	100	107.2	0.015	0.024
7	Diethylstilbestrol	56–53-1	100	104.3	0.011	0.025
8	2,4,6-Tribromophenol	118–79-6	100	106.3	0.014	0.028
9	Acid Red 337	67786–14-5	100	107.0	0.013	0.033
10	2,2',6,6'-Tetrachlorobisphenol A	79–95-8	100	106.9	0.012	0.033
11	2-Chloro-4-phenylphenol	92–04-6	100	106.1	0.015	0.034
12	4-Bromo-2-(4-chlorophenyl)-1-(ethoxymethyl)-5-(trifluoromethyl)-1H-pyrrole-3-carbonitrile	122453–73-0	100	103.1	0.017	0.035
13	Apigenin	520–36-5	100	104.4	0.012	0.035
14	Dinocap	39300–45-3	100	112.0	0.012	0.036
15	2,4,6-Trichlorophenol	88–06-2	100	103.9	0.011	0.042
16	CP-100829	135080–03-4	50	107.6	0.013	0.042
17	2,4-Dihydroxybenzophenone	131–56-6	100	105.5	0.010	0.045
18	Phenolphthalein	77–09-8	100	104.5	0.007	0.049
19	Chlorthal-dimethyl	1861–32-1	100	103.2	0.012	0.055
20	Chrysin	480–40-0	100	105.3	0.013	0.055
21	Fluazinam	79622–59-6	100	109.8	0.016	0.062
22	C.I. Acid Orange 8, monosodium salt	5850–86-2	100	109.1	0.018	0.072
	L-thyroxine (T4, positive control)	51–48-9	1.8			0.075 ±0.02
23	Ammonium perfluorooctanoate	3825–26-1	100	103.8	0.011	0.075
24	C.I. Disperse Black 6 dihydrochloride	20325–40-0	100	111.4	0.018	0.080
25	Dinoseb	88–85-7	50	110.6	0.19	0.081

^aMedian of three replicates at maximum concentration of chemical in concentration-response testing.

Figure 4.

Concentration-response curves for binding to transthyretin by T4 at eight concentrations of the model ligand T4 (A) and selected chemicals at twelve concentrations with three replicates at each concentration (black circles). Examples include nine of the top twenty-five ranked inhibitors (B-J) selected from Table 2 and two chemicals with ^~90% activity in single concentration screening (K-L) to represent a variety across types of compounds and inhibition curves.

Discussion

Investigating chemical binding to TTR is important due to TTR’s role in transporting thyroid hormone to target tissues as well as maintaining thyroid hormone homeostasis. It is believed that displacement of T4 from TTR in some mammals results in enhanced clearance and reduced serum TH concentrations ^59–61. While some research has been done to investigate chemical binding to serum transport proteins, this study presents the largest library of chemicals screened to date. Screening 1512 chemicals from the ToxCast chemical library through the TTR-ANSA binding assay presented herein resulted in identification of 888 active chemicals (20% activity or greater). Of these, 282 produced 85% activity or greater including 247 chemicals that, to our knowledge, have not previously been shown to bind TTR.

Autofluorescence was addressed as a potential interference of this assay due to the possibility of false-negative results (i.e., chemical autofluorescence may mask the reduction in fluorescence that occurs when ANSA is displaced). An alternate interference of the ANSA assay would be fluorescence quenching by the test chemical resulting in false positives (i.e., the reduction in fluorescence would suggest displacement of ANSA from TTR). No further investigation was done into quenching since a false positive would be moved onto further testing rather than being identified as inactive.

The data set was evaluated to determine whether any functional groups are enriched in the highly active (>85% activity) chemicals compared to the less active chemicals screened in this assay ^{62, 63}. SMILES were downloaded from the CompTox Chemicals Dashboard for the ph1_v2, ph2, and e1k library lists. Following the approach of Vorberg and Tetko ⁶³, SMILES were uploaded into the Online Chemical Modeling Environment (OCHEM, https://ochem.eu/home/show.do ). SMILES were recognized for 1504 individual chemicals. The chemicals were discretized by setting 85% activity as the threshold. There were 280 chemicals recognized as having “high” activity (>85%), and the remaining chemicals were categorized as “low” activity. The SetCompare Utility was then used to compare functional groups enriched among the highly active chemicals (Table 3). This analysis revealed that azo compounds and phenols are overrepresented in the highly active compounds with enrichment factors of 68 and 6.6, respectively. Interestingly, hydroxy compounds containing alcohols or phenols were grouped together and found to be enriched among the highly active compounds; however, this is largely due to enrichment of phenols (36.2% of high-actives vs 5.5% of low actives) since alcohols are enriched in the low-active group (enrichment factor 4.8). The enrichment of aromatic ethers binding to TTR is in agreement with a previous analysis of functional groups enriched among TTR-binding chemicals from a radiolabeled T4 assay ⁶⁴. The top 25 most active chemicals (Table 2) were assessed for the presence of enriched functional groups. Most of the 25 most potent chemicals screened in this TTR assay are arenes, which were enriched by a factor of 1.7 in highly active chemicals, and many of them are phenols. Further, thirteen of the chemicals are halogen derivatives, which had an enrichment factor of 1.6. Three azo compounds are present among the list of 25 most potent TTR binders.

Table 3.

Analysis of functional groups enriched in high and low TTR binding activity groups using the SetCompare Utility of OCHEM.

Descriptor	High activity		Low activity		Enrichment factor	p value
	#	%	#	%
Phenols	101	36.2	66	5.5	6.6	3.86E-38
Arenes	217	77.8	555	46	1.7	8.27E-23
Aromatic compounds	225	80.6	624	51.7	1.6	6.05E-20
Azo	19	6.8	1	0.1	68	1.57E-13
Hydroxy compounds: alcohols or phenols	110	39.4	258	21.4	1.8	1.02E-09
Alcohols	11	3.9	196	16.2	4.2	−2.19E-09
Sulfonic acid derivatives	34	12.2	50	4.1	3	1.54E-06
Sulfonic acids	20	7.2	18	1.5	4.8	2.04E-06
Aryl halides	71	25.4	164	13.6	1.9	2.29E-06
Halogen derivatives (alkyl, alkenyl, aryl)	99	35.5	268	22.2	1.6	4.68E-06

Further investigation was done for the 25 most potent chemicals in this assay (lowest EC50s) to determine whether they have demonstrated thyroid activity in other in vitro or in vivo assays. Specifically, in vitro activity was assessed in thyroid assays with data available through the US EPA’s CompTox Chemicals Dashboard ⁶⁵. The majority of the top 25 potent chemicals are active in multiple thyroid relevant assays. All of the chemicals except for 2,4,6-trichlorophenol were active in at least one other in vitro assay. Three chemicals (bromoxynil, chlorthal-dimethyl, and ammonium perfluorooctanoate) were active in only one other in vitro assay, which was the Tox21_TR_Luc_GH3_antagonist assay. Twenty-three of the twenty-five chemicals were active in the Tox21_TR_Luc_GH3_antagonist assay; however, eighteen of them were also active in the associated viability assay indicating potential cytotoxicity. It has been reported that the Tox21_TR_Luc_GH3 antagonist assay is sensitive and identifies chemicals as active even if they are not direct thyroid receptor antagonists ⁶⁶. Chemical activity in the TTR binding assay and in other assays in which TH is the substrate (iodothyronine deiodinases) or ligand (thyroid receptor) makes sense due to TH being endogenous ligands for TTR. Similarly, activity of these chemicals in the iodotyrosine deiodinase assay (recycles iodide from mono- and diiodotyrosine) makes sense due to the structural similarity of iodothyronine and iodotyrosine (i.e., a phenolic ring flanked by iodine). However, several chemicals are also active in the thyroid peroxidase (stimulated by thyroid-stimulating hormone) assay, which is not regulated by TH, suggesting that these chemicals have non-specific interactions with multiple targets in the thyroid axis. To further understand thyroid activity of these chemicals, a basic review of the literature was performed to identify in vivo thyroid activity. The review was performed by searching "(thyroid OR T4) AND (chemical name)" in Web of Science (search conducted July 24, 2023). As anticipated, there are many reports of in vivo thyroid activity for the thyroid hormone analogs Tetrac (ranked 1) and Triac (ranked 4). No in vivo studies were found for nineteen of the chemicals. Thyroid activity was demonstrated in vivo for the remaining four chemicals: ammonium perfluorooctanoate ⁶⁷ and apigenin ⁶⁸ in rodents, 2,4,6-tribromophenol ^{69, 70} in rodents and zebrafish, and diethylstilbestrol ⁷¹ in zebrafish; however, it is unknown whether TTR is the mode of action for these chemicals.

The EC50 values generated in this study were compared with those from previous screening efforts using radiolabeled ligand (¹²⁵I-T4) as well as fluorescein isothiocyanate (FITC-T4) and 8-anilino-1-napthalene sulfonic acid (ANSA) assays (Table 4). There was a total of thirty-five chemicals and the T4 control overlapping between previous TTR screening assays and the ToxCast libraries. There are some chemicals (e.g., amiodarone, benzoic acid, methimazole) with reported EC50s from previous ANSA screening assays greater than 24 μM which was corroborated by the ANSA assay reported herein by low percent activity in single concentration screening which precluded the chemicals from being screened in concentration-response. Similarly, several chemicals screened in the ¹²⁵I-T4 assay had indeterminable EC50 values (2-perfluorohexyl ethanol, bisphenol A diglycidyl ether, hexanoic acid, and octanoic acid) which were not screened in concentration-response in this study due to low activity in single concentration screening. Interestingly, bisphenol A had 96% activity in single concentration screening but a relatively high EC50 value of 8.63 μM. The EC50 for bisphenol A was not determined in a previous ¹²⁵I-T4 assay suggesting this chemical does not have a particularly high affinity for TTR which agrees with the higher EC50 value determined in this study. In general, the ANSA-TTR assay conditions in this study appear to be more sensitive than other TTR binding assays with amiodarone in the FITC assay as the only exception. It has been reported that TTR is sensitive to changes in assay conditions ^{72, 73}; thus, the variety of methods and assay protocols used to investigate chemical binding to TTR hinder quantitative comparisons ⁷³. However, the T4 EC50 values reported in this study are in the range of those reported in literature (Table 4).

Table 4.

Comparison of EC50 values from literature and from this study.^a

	Previous TTR screeningb			This study
Chemical	125I-T4 EC50 (μM)	FITC-T4 EC50 (μM)	ANSA EC50 (μM)	ANSA EC50 (μM)	% Activity in Single conc.
T4	0.04972 0.05575 0.09176 0.08849 0.08777 0.0846 0.06178	0.179 0.05780 0.2674 0.2652 0.03181	0.03 50 0.2155 0.4280 0.2651 0.68582 0.68583	0.075
Tetraiodothyroacetic acid (Tetrac)		0.02980	0.2880	0.016	105
Triiodothyroacetic acid (Triac, Tiratricol)		0.02080	0.2880	0.021	105
2,4,6-tribromophenol	0.004875 0.067277	0.574	0.3783	0.028	105
2,4,6-trichlorophenol			1.4183	0.04	100
2-perfluorohexyl ethanol	ND78			NA	56.8
4-nonylphenol		1676		3	89.2
Amiodarone		1.980	>2480	NA	48.4
Benzoic acid			>100082	NA	−17.1
Bisphenol A	ND40			8.63	96
Bisphenol A diglycidyl ether	ND40			NA	39.5
Decanoic acid	ND78			7.4	93.7
Dodecanoic acid			1151	1.7	98.9
Fluorotelomer alcohol 8:2		ND81		NA	45.8
Hexanoic acid	ND78			NA	11.2
Linoleic acid			6.551	NA	74.5
Mono(2-ethylhexyl)phthalate (MEHP)		13.0874	31.5574	1.34	99
Methimazole		>1080	>2480	NA	60.3
N-ethyl perfluorooctane sulfonamide	ND78			1.0	94.7
Octanoic acid	ND78			NA	80.5
Oleic acid			3.351	3.0	98.2
Perfluorobutane sulfonate	19.4678c	13.33182c		1.76	99.4
Perfluorodecanoic acid	8.954785	3.177 1.62381	0.7155	0.65	102.9
Perfluoroheptanoic acid	1.56578	1.277 1.12881	0.455	0.21	102.2
Perfluorohexane sulfonate	0.71778	0.59481 c		0.09	102.4
Perfluorohexanoic acid	8.2278	5.677 3.18981	1.455	1.05	96.4
Perfluorononanoic acid	2.73778	1.177 1.97781		0.2	104.7
Perfluorooctane sulfonamide	6.12478	0.7377		0.61	103.1
Perfluorooctane sulfonate	0.9478	0.1381 c 2.184 c		0.13	104.2
Perfluorooctane sulfonic acid		0.2577 1.2174		0.09	104.9
Perfluorooctanoic acid	0.94978	1.177 2.0274 0.37881		0.15	103.5
Perfluoroundecanoic acid	21.5678	3.177 5.33981	1.9955	0.69	101.4
Propylparaben		19.077		5.0	96.8
Tetrabromobisphenol A	0.03175 0.007777	0.03477 0.2274		0.02	105.8
Tetrachlorobisphenol A	0.106877			0.03	105.2
Triclosan		1.177 0.9374		0.16	106

^aAssays include the TTR-radiolabeled ligand binding assay (¹²⁵I-T4) as well as the fluorescein-5-isothiocyante (FITC-T4) and 8-anilino-1-naphthalene sulfonic acid (ANSA) fluorescent binding assays. Percent activity from single point screening in this study is also reported to explain why some chemicals do not have a reported EC50 (e.g., <85% activity in SC was not run in concentration response).

^bChemicals that bind TTR in the literature are often referred to as “TTR inhibitors” because they potentially inhibit TH from binding TTR. As such, results from TTR assays in found in the literature are often reported as IC50s compared to EC50s reported in this study. For clarity, all the values will be referred to as EC50s in this discussion.

^cMay be different formulation from what was screened in this study; no CAS was provided in the reference literature for comparison. These chemicals in the ToxCast libraries were provided as potassium salt.

There are costs and benefits to each of the three screening methods previously used. The radio-labeled method uses the endogenous ligand of TTR to assess competition of chemicals for binding to the protein; however, the radiolabeled substrate can be expensive and hazardous, and the assay is not easy to adapt to a medium or high throughput format. ANSA is relatively inexpensive and non-toxic, but it is known to bind to many proteins by nonselective hydrophobic forces and cannot be used to specifically explore ligand binding sites ⁵². FITC may be more sensitive than ANSA due to the fluorescent probe being conjugated to the endogenous ligand; however, the FITC assay may be less sensitive than the radiolabeled ligand binding assay due to steric hindrance of the FITC moiety reducing the binding affinity of the probe ⁷⁴.

Based on assay performance metrics and the high Z’ factors over the duration of screening, this ANSA fluorescence-based assay is suitable for high throughput screening of chemicals for binding to TTR; however, the large percentage (58.7%) of active chemicals may indicate that the assay may not be sufficiently specific for use in risk assessment. It is also possible that the TTR homotetrameric structure is not maintained in vitro and the reduction in fluorescence for the presumably active chemicals may be a result of protein degradation as opposed to chemical binding; however, this was previously tested in our lab with several per- and polyfluoroakyl substances and degradation of the protein structure was not observed. This supports the need for development of orthogonal high-throughput screening assays to confirm these chemicals are truly active. Overall, results from this assay, especially in conjunction with additional thyroid relevant assays, may be used to identify chemicals to prioritize for higher tier testing.

Supplementary Material

Supporting Information

• Supplemental tables (xlsx) including reaction components, reaction plate quality control, autofluorescent and excluded chemicals, chemicals tested in single concentration and concentration-response

• Supplemental figures (pdf) including reaction plate layouts and concentration-response curves

Abbreviations:

TTR

transthyretin

NAMs

new approach methodologies

EDCs

endocrine disrupting compounds

TH

thyroid hormone