http://www.diva-portal.org

This is the published version of a paper published in Chemical Research in Toxicology.

Citation for the original published paper (version of record):

Zhang, J., Li, Y., Gupta, A A., Nam, K., Andersson, P L. (2016)Identification and Molecular Interaction Studies of Thyroid Hormone Receptor Disruptorsamong Household Dust Contaminants.Chemical Research in Toxicology, 29(8): 1345-1354http://dx.doi.org/10.1021/acs.chemrestox.6b00171

Access to the published version may require subscription.

N.B. When citing this work, cite the original published paper.

Permanent link to this version:http://urn.kb.se/resolve?urn=urn:nbn:se:umu:diva-125629

Identification and Molecular Interaction Studies of Thyroid HormoneReceptor Disruptors among Household Dust ContaminantsJin Zhang,† Yaozong Li,† Arun A. Gupta, Kwangho Nam,* and Patrik L. Andersson*

Department of Chemistry, Umea University, SE-901 87 Umea, Sweden

*S Supporting Information

ABSTRACT: Thyroid hormone disrupting chemicals(THDCs), often found abundantly in the environment,interfere with normal thyroid hormone signaling and inducephysiological malfunctions, possibly by affecting thyroidhormone receptors (THRs). Indoor dust ingestion is asignificant human exposure route of THDCs, raising seriousconcerns for human health. Here, we developed a virtualscreening protocol based on an ensemble of X-ray crystallo-graphic structures of human THRβ1 and the generalized Bornsolvation model to identify potential THDCs targeting thehuman THRβ1 isoform. The protocol was applied to virtuallyscreen an in-house indoor dust contaminant inventory, yielding31 dust contaminants as potential THRβ1 binders. Fivepredicted binders and one negative control were tested usingisothermal titration calorimetry, of which four, i.e., 2,4,5-trichlorophenoxyacetic acid (2,4,5-T), bisphenol A (3-chloro-2-hydroxypropyl) (2,3-dihydroxypropyl) ether (BADGE-HCl-H2O), 2,2′,4,4′-tetrahydroxybenzophenone (BP2), and 2,4-dichlorophenoxyacetic acid (2,4-D), were identified as THRβ1 binders with binding affinities ranging between 60 μM and460 μM. Molecular dynamics (MD) simulations were employed to examine potential binding modes of these binders andprovided a rationale for explaining their specific recognition by THRβ1. The combination of in vitro binding affinitymeasurements and MD simulations allowed identification of four new potential THR-targeting THDCs that have been found inhousehold dust. We suggest using the developed structure-based virtual screening protocol to identify and prioritize testing ofpotential THDCs.

■ INTRODUCTION

Thyroid hormone (TH) homeostasis is important for macro-nutrient metabolism, thermogenesis and energy balance,cardiovascular function, and normal brain development inhumans and wildlife.1 A broad range of environmentalxenobiotics have been reported to disturb normal TH functionsby interfering with TH signaling, potentially causingmalfunctions in blood circulation, brain development, andreproduction.2 These chemicals, referred to collectively asthyroid hormone disrupting chemicals (THDCs), are routinelydetected in our environment, and some, e.g., 2,4,6-tribromo-phenol, 2,3,4,5,6-pentachlorophenol, and polybrominateddiphenylethers (PBDEs), have been found in householddust.3−5 Animal and human studies have indicated thatTHDCs may pose a threat to human health.6−8 Of particularconcern is exposure to these chemicals during critical stages offetal development, such as during neurological development,which may lead to permanent brain function defects, such ascretinism, permanent hearing losses, and structural abnormal-ities in the brain.6,7

Indoor dust is a major source of human exposure toTHDCs.9 Ingestion of dust containing THDCs, such asbromophenols and perfluoroalkyl and polyfluoroalkyl sub-

stances (PFASs), has been associated with alteration of THlevels in humans.5 We have previously identified 485 organiccontaminants with different chemical composition in indoordust.10 These dust contaminants were originated mainly fromconsumer products or outdoor environments.11 Major groupsinclude flame retardants (e.g., PBDEs), pesticides, plasticizers(e.g., phthalates), PFASs, and classical persistent organicpollutants (e.g., polychlorinated biphenyls (PCBs) anddioxins).9 Among them, PBDEs, PCBs, and PFASs have beenreported to cause thyroid-related adverse effects, such asmetabolic diseases, reduced fertility, and thyroid-relatedtumors.1 For some compounds, molecular targets of the THsystem have been identified. For example, PBDE 181 and 190and certain PFASs, such as perfluorooctanoic acid (PFOA) andperfluorooctanesulfonic acid (PFOS), have been shown totarget transthyretin (TTR) and disturb TH transportation.12−14

Thyroid hormone receptors (THRs), which are activated bytriiodothyronine (T3) and thyroxine (T4) under normalphysiological conditions,15 are potential molecular targets forTHDCs.16 For example, bisphenol A and certain hydroxylated

Received: May 16, 2016Published: July 13, 2016

Article

pubs.acs.org/crt

© 2016 American Chemical Society 1345 DOI: 10.1021/acs.chemrestox.6b00171Chem. Res. Toxicol. 2016, 29, 1345−1354

This is an open access article published under an ACS AuthorChoice License, which permitscopying and redistribution of the article or any adaptations for non-commercial purposes.

PBDEs (OH-PBDEs) (e.g., 4-OH-PBDE 69 and 4-OH-PBDE121) have been shown to target THR and affect THR-mediatedgene expressions.16−18 The importance of THRs in THhomeostasis is well-studied in the field of computer-aideddrug discovery;19−22 however, only a few in silico based studiesare published on the interaction with THRs of environmentalpollutants.23−25 Improved knowledge of interactions ofcontaminants with THRs is essential for identification ofTHDCs targeting THRs and human health related riskassessments of dust contaminants.Virtual screening (VS) by molecular docking is a high-

throughput approach for predicting protein-compound bindingposes and identifying and prioritizing bioactive compounds forexperimental testing.26−29 VS has been successfully applied toidentify potential environmental pollutants targeting, e.g., theestrogen receptor alpha and haloalkane dehalogenases, fromlarge libraries of industrial chemicals.30−32 VS has the potentialto reduce costs and animal testing in chemical risk assessmentprocesses by identifying the most hazardous chemicals or themost promising candidates in drug development.33,34

In the present study, we have developed a VS protocol toidentify dust contaminants targeting human THRβ1. In theprotocol, two additional approaches were introduced for robustscreening of compounds with diverse scaffolds and functionalgroups. First, instead of relying on a single X-ray structure, anensemble of protein structures, i.e., the same protein withslightly different ligand binding site (LBS) conformations,35−37

was used to mimic protein flexibility.38 Second, a molecularmechanics based method, i.e., the generalized Born surface area(MM-GBSA) method,39 was used for better estimation ofsolvation effects, thereby improving the performance of the VSprotocol.40 The developed virtual screening model was thenapplied to screen dust contaminants targeting human THRβ1.THRβ1 was selected among the four possible isoforms of THR(i.e., THRα1, THRα2, THRβ1, and THRβ2)41 because of itsimportance to normal liver and kidney function, hearingdevelopment, and the feedback loop of the hypothalamic-pituitary-thyroid axis.42 Disruption of THRβ1 by THDCs hasalso been suggested to be implicated in inadequate braindevelopment and liver malfunctions.17,43 Six representative dustcontaminants (five predicted binders and one predictednonbinder) were selected for isothermal titration calorimetry(ITC) to determine their binding affinities to THRβ1. MDsimulations were then carried out to examine the possiblemechanism of THRβ1 ligand recognition and differentdynamical behavior of THRβ1 upon the binding of diversebinders.

■ MATERIALS AND METHODSThe study was conducted by the following procedure (Figure 1): (I)development of the VS model, (II) initial screening of 485 dustcontaminants, yielding 131 initial hits, (III) refinement of the initialhits, reducing the number of hits to 31 by ligand−protein interactionanalysis, (IV) selection of six representative compounds (five predictedbinders and one predicted nonbinder) based on their structuraldiversity and environmental relevance, and (V) experimental measure-ment of their binding affinities to THRβ1 using ITC followed by MDsimulations. Details of each of these five steps are given below.THRβ1 Structure Selection and Preparation. Among 17 X-ray

crystallographic structures of human THRβ1 in complex with variouscoligands available in the RCSB Protein Data Bank (PDB),44 sevenstructures were selected based on the structural diversity of theircoligands and shape/size of their binding sites of the protein. The PDBID of each structure is given in Table S1. For each structure, any

mutated residue was restored to its corresponding wild-type residue,followed by the addition of hydrogen atoms, optimization of hydrogenbonding networks, and assignment of the protonation state of eachresidue to its corresponding protonation state at pH 7.0. The pronatedstate of each protein residue in or near the LBS was further inspectedvisually and corrected if needed. The structures were then minimizedusing the OPLS2005 force field to remove unfavorable high energycontacts.45 The preparation steps were carried out using the ProteinPreparation Wizard of the Schrodinger Suite.46

As water molecules can mediate interactions between proteinresidues and ligands by forming hydrogen bonds with them, theirpresence can improve virtual screening performance.47−49 However,they can also hinder the identification of ligands by blocking directinteractions with the protein.50 In the present work, both possibilitieswere considered by either including (PDB ID with “-W”) or excludingexplicit waters (PDB ID with “-nW”) in any X-ray structure withidentified crystal waters in its ligand binding site (LBS). This yielded atotal of 12 protein structures (Table S2), which were then used in thedevelopment of the VS protocol.

Ligand Data Sets for Virtual Screening (VS). Two different datasets were used in the study: (1) the THR subset of the Database ofUseful Decoys: Enhanced (DUD-E)51 and (2) an indoor dustcontaminant inventory we compiled previously.10 The THR subsetof the DUD-E comprising 103 THR binders and 7450 in silicogenerated inactive decoys was used for the development andevaluation of the VS protocol. The dust contaminant inventorycontained 485 organic contaminants that humans could potentially beexposed to through household dust. These compounds werepreprocessed using the ChemAxon JChem Standardizer.52 For eachcompound, between 1 and 32 conformers were generated dependingon the total number of chiral centers, and their ionization andtautomeric states were determined at pH 7.0 ± 1.0, using theSchrodinger LigPrep53 and the Epik modules.

Single Structure Docking, Ensemble Docking, and MM-GBSA Rescoring. The LBS of each THRβ1 structure was describedby using a 10 Å cubic grid box centered on its corresponding coligand.The prepared compounds of the DUD-E THR subset were dockedindividually into each structure. For each conformer, a total of 5,000poses were generated and ranked using the standard prevision (SP)scoring function of the Schrodinger Glide module.54 The top-ranked400 poses were selected and subjected to 100 steps of conjugategradient energy minimizations with a dielectric constant of 2.0.

The results of each single structure docking were assessed bydetermining the area under the curve (AUC) value of the receiveroperating characteristic (ROC) curve and the enrichment factors(EFs) (Table S2). For each THRβ1 structure, the AUC value of theROC curve was obtained by plotting the true positive rate (sensitivity)against the false positive rate (1-specficity) (Figure S1). We utilized

Figure 1. Scheme adopted in this study with details of the virtualscreening (VS) model development process, the isothermal titrationcalorimetry (ITC) validation study, and the molecular dynamics (MD)study of ligand−protein interactions. ssDock model, single structuredocking model; esDock model, ensemble structure docking model. Inbrackets are the numbers of hits at different steps.

Chemical Research in Toxicology Article

DOI: 10.1021/acs.chemrestox.6b00171Chem. Res. Toxicol. 2016, 29, 1345−1354

1346

AUC because its value is insensitive to the ratio of actives versusdecoys in the DUD-E data set.55 Enrichment was characterized by twoEFs evaluated at 1% (EF1%) and 10% (EF10%) of the rankeddatabase, respectively.56

Ensemble models were prepared by considering all possiblecombinations of the single structure docking results, from 2 structuresup to all 12 structures. The top 20% of the DUD-E subset ranked bySP docking score was extracted for each single structure docking.Then, for a given combination of structures, the extracted results werecombined to determine the overall performance, evaluated by the totalnumber of active compounds identified by that particular combination.In Table S3, the results of the best performing ensemble model foreach given number of protein structures are presented. The finalensemble model used for screening dust contaminants was selected byconsidering both the number of active compounds identified and thetotal computational cost of docking.To further improve the VS performance of the selected ensemble

model, we applied the MM-GBSA method and rescored the ligand-THRβ1 docking poses. The calculation was performed using theSchrodinger Prime module with the VSGB 2.0 implicit solvent modeland the OPLS2005 force field.57 During the rescoring, the MM-GBSAenergy minimizations were performed by allowing the conformation ofthe ligand to relax while the rest of the complex was held fixed. TheMM-GBSA scores were determined for the resultant optimized poses.The results of the MM-GBSA rescoring were also assessed by thenumber of identified active compounds, AUC values, and enrichmentfactors and compared with other docking results, as shown in TableS4.Compound-THRβ1 Interaction Analysis. The majority of the

amino acid residues constituting the LBS of THRβ1 are hydrophobic.In contrast, only a few residues are hydrophilic, some of which havebeen suggested to play critical roles in the specific recognition ofendogenous ligands and pharmaceuticals targeting THRβ1.22 Sincehydrophilic residues in LBS can mediate electrostatic interactions withligands, we reasoned that these residues, if any, may contribute towardspecific binding of dust contaminants to THRβ1. We analyzed theseinteractions for all seven selected X-ray structures and counted thenumber of H-bonds and salt bridges between each ligand and THRβ1based on the poses proposed by the MM-GBSA rescoring. These datawere used to refine the 131 initial hits identified by the MM-GBSAscoring, yielding 31 final hits (Table S5).Isothermal Titration Calorimetry (ITC). Binding affinities of six

dust contaminants and T3 (positive control) to THRβ1 weremeasured by using ITC. The six tested dust contaminants includedfive predicted binders and one predicted nonbinder as a negativecontrol (Table 1). Their purity, pKa values, and commercial sourcesare provided in Table S6. The human THRβ1 ligand binding domain(LBD) (6xHis plus Gly209-Asp461) was overexpressed in Escherichiacoli BL21 (DE3) host cells by following the procedure outlined by Yeet al.22 The purified THRβ1 LBD was exhaustively dialyzed inphosphate buffer containing 20 mM sodium phosphate, 125 mM

NaCl, and 1 mM TCEP at pH 8.0. We prepared stock solutions of T3(20 mM) and the six dust contaminants (100 mM) in dimethylsulfoxide (DMSO). Before the ITC experiments, all stock solutionswere diluted with phosphate buffer to a final level of 5% DMSO, as is acommonly acceptable DMSO concentration for ITC.58,59 All experi-ments were conducted at 25 °C by injecting aliquots of eachcompound (in syringe) into THRβ1-LBD buffer solution (in cell)using MicroCal Auto-iTC200 (Malvern Instruments Ltd., Worcester,UK). Initial “buffer scouting” was performed for each compound toavoid buffer mismatch caused by DMSO concentration differencesbetween the cell and syringe. Raw data were collected, corrected forcompound heats of dilution, and integrated using MicroCal Origin.The ITC results (with “c” value <5) were fitted to a single-site bindingmodel and fixed stoichiometry parameter (N = 1), allowing estimationof the equilibrium dissociation constant (Kd) and enthalpy change(ΔH).60,61

Molecular Dynamics (MD) Simulations. The systems for MDsimulations were prepared based on the 1Q4X THRβ1 LBD complexstructure,62 which was the top performing single protein structure indocking (Table S2), and the poses from the VS protocol were used forthe three dust contaminants (2,4,5-T, BADGE-HCl-H2O, and BP2;abbreviations are given in Table 1). The resulting complex (2,4,5-Tsystem as an example) was solvated in an 85 Å rhombic dodecahedronwater box containing 12475 TIP3P water molecules.63 The size of thewater box was chosen to allow at least a 10 Å buffer space from eachprotein atom to the closest boundary of the water box. 0.15 M NaClwas added to neutralize each system and mimic biophysical conditions.For example, the resulting system for 2,4,5-T contained 41494 atoms(3981 protein atoms, 18 ligand atoms, 37425 water atoms, 40 Na+, and30 Cl− ions). In addition, one apo THRβ1 LBD system was preparedas a reference by removing its ligand from the X-ray structure. Eachsystem was initially energy minimized for 10,000 steps under a seriesof position restraints and constraints, heated to 300 K, and thenequilibrated under the constant temperature and pressure conditionsfor 1 ns using the CHARMM program (version 38a1).64

A 50 ns production MD simulation was performed for each systemusing the NAMD program (version 2.9).65 The temperature of eachsimulated system was maintained at its target value by using aLangevin thermostat with 5 ps−1 friction coefficient. The pressure wascontrolled by the Nose−Hoover Langevin piston method with 200 pspiston period and 100 ps piston decay time. A 2 fs integration timestep was used, and any bond involving hydrogens was constrained totheir corresponding force field bond lengths. Nonbonded van derWaals energies were evaluated by smoothly turning off theirinteraction energies using a switching function at a distance between9 and 11 Å, and electrostatic interactions were evaluated using theparticle mesh Ewald summation (PME) method.66 In all simulations,the CHARMM36 force field with the CMAP correction was used topresent the THRβ1 LBD,64,67 TIP3P for water molecules, and theCGenFF36 force field for the three dust contaminants.68 Details of the

Table 1. Studied Compounds with Chemical Abstract Services (CAS) Registry Numbers, Abbreviations, Industrial Application,Prediction from the Virtual Screening Protocol and Binding Affinities (Kd) Determined Using Isothermal Titration Calorimetry

compounds (abbreviations) CAS no. applications predictionKd

(μM)a

triiodothyronine (T3) 6893-02-3 positive control 1.62,4,5-trichlorophenoxyacetic acid (2,4,5-T) 93-76-5 herbicide active 60bisphenol A (3-chloro-2-hydroxypropyl) (2,3-dihydroxypropyl) ether (BADGE-HCl-H2O)

227947-06-0 plasticizer active 87.5b

2,2′,4,4′-tetrahydroxybenzophenone (BP2) 131-55-5 UV filter active 2002,4-dichlorophenoxyacetic acid (2,4-D) 94-75-7 herbicide active 4631-((2,4-dichlorophenyl)carbamoyl)cyclopropanecarboxylic acid (cyclanilide) 113136-77-9 herbicide active ND5-chloro-2-(2,4-dichlorophenoxy)phenol (triclosan)c 3380-34-5 antimicrobial agent (negative

control)inactive ND

aND refers to nondetected affinity. bThe binding affinity of BADGE-HCl-H2O was only an approximate value, probably due to poor solubility in theITC buffer. cTriclosan was tested as a negative control. It was identified among the 131 initial hits but showed no interaction with the fourhydrophilic residues (Figure 1).

Chemical Research in Toxicology Article

DOI: 10.1021/acs.chemrestox.6b00171Chem. Res. Toxicol. 2016, 29, 1345−1354

1347

partial optimization of the dust contaminant force field parameters areprovided in the Supporting Information.

■ RESULTS AND DISCUSSION

Ensemble Docking Model with MM-GBSA Rescoring.The reliability of the VS using Glide was examined bycomparing the predicted poses of actual coligands found inthe 12 human THRβ1 X-ray structures to their correspondingX-ray structures.54 The results showed that the dockingsimulation reproduced the actual binding poses with root-mean-squared-deviation (RMSD) lower than 2 Å for allcompounds (Table S2).69 The DUD-E THR subset was thensubjected to Glide docking. The VS performance of (singlestructure) docking was compared between the 12 structures byexamining the AUC value, total number of active compoundsidentified, and EF1% and EF10% values (Table S2), and 1Q4X-W showed the best performance with an AUC of 0.863, goodearly enrichment (EF1% = 32.410 and EF10% = 7.180), and atotal of 92 identified binders (out of 103 binders).The results showed that no single structure was able to

identify all 103 binders with high AUC values. Thus, wecombined different structures as an ensemble of structuralmodels in order to identify a broader range of compounds witha high true positive versus false positive discrimination ratio.We examined all possible combinations of the ensemblestructures from 2 structures up to all 12 structures andcompared their performances based on the total number ofidentified binders (see Materials and Methods). Table S3presents the best performing ensemble model for each givennumber of structures. The ensemble model with four structures(i.e., 1N46-nW, 1Q4X-nW, 1R6G-nW, and 1XZX-nW) was

chosen as our “final” ensemble model to identify dustcontaminants targeting THRβ1. This model was chosen notonly because it was one of the top models tested, successfullyidentifying 95 compounds out of the total 103 binders, but alsobecause it showed reasonable computational cost, which isessential for high-throughput VS against a large compoundlibrary. By comparison, the 1Q4X-W model, which is the topsingle structure model, only covered 75 binders whenconsidering the top 20% hits (Table S4). Finally, the MM-GBSA method was applied to refine the selected ensembledocking model and its result, which yielded an AUC value thatincreased to 0.865 from 0.804 and EF10% to 7.478 from 6.109,respectively (Table S4). The ROC curve of the model clearlyshowed the improvement of EF10% and AUC after MM-GBSArescoring (Figure S1).

Virtual Screening of Household Dust Contaminants.Identification of potential binders of THRβ1 was achieved intwo steps (Figure 1). In the first step, the VS protocoldescribed above was applied to a recently compiled inventoryof 485 indoor dust contaminants.10 A total of 131 compounds(top 20% hits) were identified using the developed VSprotocol. For each dust contaminant, all stereoisomers andprotonation states, in the range of pH 7.0 ± 1.0, wereconsidered during the virtual screening. Each stereoisomer andprotonation form was docked into the THRβ1 structures, andthe top scored stereoisomer/protonation form was selected forsubsequent MM-GBSA calculation. For instance, BADGE-HCl-H2O has four stereoisomers. Among them, the RR stereo-isomer has scored the top docking score (Table S7) and wasselected for the MM-GBSA rescoring and MD simulation. Inselecting protonation form, we calculated the energetic penalty

Figure 2. Molecular structures of active compounds and thermograms of their binding to THRβ1 determined by isothermal titration calorimetry.The abbreviations of compounds are defined in Table 1. For BADGE-HCl-H2O, the dissociation constant (Kd) was an approximate value likely dueto its poor solubility in the ITC buffer.

Chemical Research in Toxicology Article

DOI: 10.1021/acs.chemrestox.6b00171Chem. Res. Toxicol. 2016, 29, 1345−1354

1348

for each protonation state based on its population estimatedusing the Epik module available in the Schrodinger program,70

and the penalty was considered in the docking score. Forexample, triclosan has a pKa of 7.9. The compound can exist asboth the protonated and deprotonated forms at pH 7.0 ± 1.0.We selected the protonated form for later MM-GBSArescoring, as the form has a more negative docking score andlower energy penalty (Table S8). In the second step, ligand-THRβ1 interactions were analyzed for all ligand-THRβ1complexes, including the 7 X-ray structures (Table S1), toidentify protein residues that form hydrogen bonds and/or saltbridge interactions with the bound ligands (see Materials andMethod for details). Our rationale for this step was twofold.First, since most endogenous THR binders are recognized viaspecific hydrophilic interactions, this additional step would aidin selecting chemicals that can form such specific interactions.Second, because hydrophobic residues dominate the bindingpocket of THRβ1, it is possible that our VS energy functionunderestimates hydrophilic interaction energies. This wouldthen bias our selection toward certain chemicals that are morehydrophobic. With the aid of the second step, we intended toremedy the bias, thereby improving the selectivity betweenspecific versus nonspecific binders. On the basis of the analysis,we identified four hydrophilic residues, i.e., Arg282, Arg320,Asn331, and His435 in the binding pocket of THRβ1. Werefined the 131 initial hits by analyzing their electrostaticinteractions with these four residues, and 31 dust contaminantsdisplayed at least one electrostatic interaction with any of thehydrophilic residues (Table S5).The 31 potential binders can be classified into four major

groups: (1) PFASs (10), (2) hydroxylated and/or halogenatedaromatic compounds (14), (3) phenoxyacetic acids (3), and(4) miscellaneous (4) (Table S5). Interestingly, the presenttwo-step screening protocol successfully identified known THRbinders, such as bisphenol A (Ki = 200 μM),17 tetrabromobi-sphenol A (TBBPA, EC50 ≈ 100 μM),71 and four PFASs72

including perfluorooctanesulfonate (IC50 = 13 μM),72,73

perfluorononanoic acid (IC50 = 12 μM), perfluoroundecanoicacid (IC50 = 15 μM), and perfluoroheptanoic acid (IC50 = 9μM), which indicated the selectivity of the present screeningprotocol.

Measuring Binding Affinities of THDCs to THRβ1. Onthe basis of the VS results, six dust contaminants (five predictedbinders and one predicted nonbinder) (Table 1) were selectedfor measurement of binding affinities to THRβ1 by ITC. Theserepresent different chemical structures and applications, and areenvironmentally relevant contaminants. The derived ITCthermograms (Figure 2) show clearly that four out of the fivepredicted THDCs bound to the THRβ1 LBD, with Kd valuesbetween 60 μM and 463 μM (Table 1), which can be comparedwith a typical hit rate around 40% for commonly practiced VScampaigns.74,75 The results also revealed the lack of responsefor triclosan (the predicted nonbinder), suggesting that thedeveloped VS protocol can distinguish binders from non-binders. The derived thermodynamic data revealed that thebinding of the four confirmed hits was largely enthalpy driven(Table S9). The docking poses and ligand-THRβ1 interactionsof the potential THDCs were compared with the cocrystal poseof T3 (Figure S2). We also measured the binding affinity of thepositive control T3 (Kd = 1.6 μM). Our value for T3 iscomparable to the previously reported affinity of T3 to THRα1(Kd = 1.8 μM), measured using an approach similar to ours.59

The similarity between the measured binding affinities of T3may be due to only a single residue difference in their ligand-binding pockets, namely, Ser277 in THRα1 and Asn331 inTHRβ1.76 The present value, however, is lower than the affinity(0.26 nM)22 reported previously based on a radio ligandbinding assay, which may be due to the ITC protocol that usesonly the monomer and the LBD of the THR.59,77 In theSupporting Information, a more thorough discussion on thedifference of the two measurements is provided. In addition, wecompared the ITC measured binding free energy (ΔG) of the

Figure 3. Comparison of docking and MD-averaged poses of potential THR disrupters: (B) 2,4,5-T, (C) BP2, and (D) BADGE-HCl-H2O. In A, theTHRβ1 ligand binding domain is shown in cartoon format, and the location of the ligand binding pocket is indicated with the two helices (H3 andH5), a binder (2,4,5-T, pink ball-and-stick), and two arginine residues (Arg282 and Arg316). In B−D, MD-averaged poses (from the last 10 ns) areshown in stick with different colors, and VS proposed poses in gray stick, respectively. In B, the electrostatic interactions between ligands andresidues are shown by purple dashes.

Chemical Research in Toxicology Article

DOI: 10.1021/acs.chemrestox.6b00171Chem. Res. Toxicol. 2016, 29, 1345−1354

1349

active compounds to their corresponding docking and MM-GBSA scores in Figure S3. Despite displaying reasonablecorrelation between the two sets of data, the scoring resultswere mainly used in the present study to prioritize the dustcontaminants for experimental testing and not for theprediction of binding free energies.One compound was identified as false positive by ITC, i.e., 1-

((2,4-dichlorophenyl)carbamoyl)cyclopropanecarboxylic acid(cyclanilide). The false identification of cyclanilide may havebeen due to the negligence of intramolecular strain energy inthe VS. Cyclanilide is structurally very similar to T3, but itscyclopropanyl group may have high strain energy, whichprevents the formation of the proposed pose in the LBS ofTHRβ1. The poor buffer solubility of bisphenol A (3-chloro-2-hydroxypropyl) (2,3-dihydroxypropyl) ether (BADGE-HCl-H2O) could be due to its incomplete saturation of theTHRβ1 LBD, allowing only an approximate estimation of itsaffinity (Figure 2).Binding Pose Examination by All-Atom MD Simu-

lations. Three compounds (2,4,5-T, BADGE-HCl-H2O, andBP2) were selected for MD simulations based on theirstructural variation and the ITC data (Figure 3). Overall, allthree compounds were stably bound near their correspondingposes predicted by the VS docking procedure during the entireMD simulations (Figure S5). Most notably, the MD-averagedstructure of 2,4,5-T, which was the highest affinity binderamong the six tested chemicals (Table 1), was nearly identicalto its original pose (Figure 3B). Three types of interactionsseemed to contribute to the stable binding of 2,4,5-T. First, thecarboxyl group of 2,4,5-T engaged in salt-bridged interactionswith the two arginine residues (Arg282 and Arg316) in theTHRβ1 binding pocket,78 holding the ligand in placethroughout the entire MD simulations (Figure 3B). Second,chlorine atoms at the 2 and 5 positions of the phenoxy-group(here named Cl2 and Cl5, respectively; Figure S6) formedhydrophobic contacts with Ile275, Ala279, Met310, andAsn331, thereby restricting the motion of the ligand duringthe MD simulations. Third, Cl5 formed a hydrogen-bond-likeinteraction with Ser314 (the detailed interaction is shown inFigure S6). We suggest that this interaction, together with thehydrophobic contact of Cl5 with Met310, contributed to thehigher affinity binding of 2,4,5-T (Kd = 60 μM) than that of2,4-dichlorophenoxyacetic acid (2,4-D) (Kd = 463 μM), whichdiffers by only a single chlorine atom.For the remaining two compounds (BP2 and BADGE-HCl-

H2O), the MD simulations showed ligand orientations thatwere slightly displaced relative to their predicted poses (Figure3C and D). Analysis of the MD trajectories suggested that thedisplacement in both cases was partly caused by rotationsaround the two bonds at the central carbon atom (Figure S7),which may not be captured accurately in the VS docking.Nevertheless, the two ligands still bound in the binding pocket,likely stabilized by hydrophobic interaction between the twobenzyl moieties and the hydrophobic residues in the THRβ1LBD. In the Supporting Information, we provide a detailedanalysis of the MD trajectories, including the distribution of thetwo rotatable bonds and the RMSD of ligands and proteinbackbone atoms.Environmental and Toxicological Implications of the

Identified THDCs. Four new potential THDCs targetingTHRβ1 were identified in this study. To the best of ourknowledge, none of these contaminants has previously beenreported to bind to THR. It can be concluded that they are

weak binders (Kd = 60−460 μM) in the range of previouslyreported THR binders, such as OH-PCBs (Ki = 30−90 μM),79

bisphenol A (Ki = 200 μM),17 and tetrabromobisphenol A(TBBPA, EC50 ≈ 100 μM).71 Exposure to weak THR binders,e.g., bisphenol A during pregnancy, has been associated withreduced TH levels in pregnant women and decreased thyroid-stimulating hormone in male neonates.80 Thus, thesecontaminants warrant further analysis as humans may bechronically exposed at doses which are similar or significantlyhigher than human T3 levels. Typical levels of T3 in humanserum are 0.9−2.8 nM,81 whereas levels of the contaminants inhuman urine of the general population have been reported tobe 167 nM (37 ppb) for 2,4-D,82 220 nM (54.3 ng/mL) forBP2,83 and 8.7 nM (3.4 ng/mL) for BADGE-HCl-H2O.84

Among the four contaminants, 2,4,5-T showed the highestaffinity to THRβ1. This compound has been used as aherbicide for broad-leaf plants and found in dust samplescollected from private homes and daycare centers, such as inNorth Carolina, USA.85 2,4,5-T has been confirmed to bind totransthyretin and disturb TH transportation.10 The compoundis known to be associated with the development of hypo-thyroidism and prostate and skin cancer.86−88 Exposure to2,4,5-T has been shown to be linked to increased occurrence ofgestational diabetes mellitus in women during pregnancy.89

2,4,5-T has also been reported as a weak estrogen receptorantagonist in reporter cell lines.90 The compound was acomponent of the defoliant Agent Orange deployed during theVietnam War, exposure to which has been shown to beassociated with the development of hypothyroidism, auto-immune thyroiditis, and other endocrine disorders; however, asto which compounds in Agent Orange cause these diseases isnot well understood.91 With a similar structure and applicationto 2,4,5-T, 2,4-D is still in use and is among the most frequentlydetected herbicides in the USA.92 Although 2,4-D did not showany interaction with the estrogen, androgen, or aromatase/steroidogenesis pathways as evaluated in several in vitroassays,93 the compound has been reported to increase theprevalence of hypothyroidism in male applicators.88 The levelof 2,4-D in their urine was detected to be 410−2500 ppb.82 2,4-D has been detected in dust samples from homes and daycarecenters in North Carolina, Northern and Central California,and Ohio.94,95 Both 2,4,5-T and 2,4-D are classified as a“possible human carcinogen” by the International Agency forResearch on Cancer (IARC).96

BADGE-HCl-H2O is an additive or starting agent inpersonal care products, food packages, and material coatings.The concentrations of BADGEs in the US urine samples were3−4-fold higher than the corresponding concentrations ofbisphenol A.84 It was found in indoor dust samples from theUSA, China, Korea, and Japan (444 ng/g).97 Nevertheless, wehave not found any reports in the literature on thyroid relatedadverse effects of this compound.BP2 is a high volume production chemical98 and a

representative of the benzophenone-type UV absorbers usedin cosmetics and plastic products.99,100 It was also found inhousehold dust samples from the USA (49.8 ng/g) and EastAsian countries (20.1 ng/g).101 The compound has beenreported to disturb the transport of thyroid hormones bybinding to transthyretin,10 and it inhibits human thyroidperoxidase and causes endometriosis.102,103 In vivo experimentshave shown that the compound induces weak antithyroidalactivity in both rodent studies and an amphibian metamor-phosis assay.104 BP2 has also been reported as an agonist for

Chemical Research in Toxicology Article

DOI: 10.1021/acs.chemrestox.6b00171Chem. Res. Toxicol. 2016, 29, 1345−1354

1350

both human and fish (rainbow trout) estrogen receptors (ERαand ERβ).105

■ CONCLUSIONSHere, we developed a robust VS protocol to identify andprioritize potential THDCs targeting THRβ1 for further in vitroor in vivo toxicological evaluation of their thyroid-relatedadverse effects. The VS protocol was applied to screen anindoor dust contaminants inventory, and the results suggested31 dust contaminants including musk compounds, PFASs, andbisphenol A derivatives (Table S5) as potential binders toTHRβ1. Five representative potential binders were selected andtested in vitro for binding to THRβ1 by ITC. The experimentsconfirmed that four compounds, i.e., 2,4,5-T, BADGE-HCl-H2O, BP2, and 2,4-D, are weak binders to THRβ1. Theirpotential binding modes were also validated and refined usingMD simulations. The developed VS protocol was shown to becapable of identifying potential THDCs, and we suggest furtherexperimental testing and analysis of the top ranked dustcontaminants including the four novel THR binders.

■ ASSOCIATED CONTENT*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acs.chemres-tox.6b00171.

Details of the crystal structures used in the study, the VSperformance and enrichment of the models using singlecrystal structure, the ensemble models and “best”ensemble models rescored with MM-GBSA; analysis ofinteractions between compounds and the hydrophilicresidues, the compounds tested using ITC; force fieldparameters of the THRβ1 binders, comparison betweendocking poses of THR disrupters and cocrystal pose ofT3, superposition of existing crystal structures, andfigures describing MD simulation results (PDF)3D-coordinates of ligand-THRβ1 complexes used for theMD simulations (ZIP)

■ AUTHOR INFORMATIONCorresponding Authors*(K.N.) Tel: +46-90-786-6570. E-mail: kwangho.nam@umu.se.*(P.L.A.) Tel: +46-90-786-5266. E-mail: patrik.andersson@umu.se.Author Contributions†J.Z. and Y.L. contributed equally to this work.FundingThis study was financed by the MiSSE project through grantsfrom the Swedish Research Council for the Environment,Agricultural Sciences and Spatial Planning (Formas) (210-2012-131), by Umea University, by the Kempe foundation, andby the Swedish Research Council (VR) (521-2011-6427 and2015-04114).NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThe MM-GBSA calculations and MD simulations wereconducted using the resources provided by the SwedishNational Infrastructure for Computing (SNIC) at HighPerformance Computing Center North (HPC2N) and Na-

tional Supercomputing Center (NSC). The THRβ1-LBD wasexpressed by the Protein Expertise Platform, Umea University.

■ ABBREVIATIONS2,4,5-T, 2,4,5-trichlorophenoxyacetic acid; 2,4-D, 2,4-dichlor-ophenoxyacetic acid; AUC, area under the curve; BADGE-HCl-H2O, bisphenol A (3-chloro-2-hydroxypropyl) (2,3-dihydrox-ypropyl) ether; BP2, 2,2′,4,4′-tetrahydroxybenzophenone;c yc l an i l i d e , 1 - ( (2 , 4 -d i ch l o ropheny l ) c a rb amoy l ) -cyclopropanecarboxylic acid; DMSO, dimethyl sulfoxide;DUD-E, Database of Useful Decoys: Enhanced; EC50, halfmaximal effective concentration; EF, enrichment factor; ER,estrogen receptor; IARC, international agency for research oncancer; IC50, half maximal inhibitory concentration; ITC,isothermal titration calorimetry; LBD, ligand binding domain;LBS, ligand binding site; MD, molecular dynamics; MM-GBSA,molecular mechanics-generalized Born surface area; OH-PBDE,hydroxylated polybrominated diphenylether; PBDE, polybro-minated diphenylether; PCB, polychlorinated biphenyl; PDB,Protein Data Bank; PFAS, perfluoroalkyl and polyfluoroalkylsubstance; PFOA, perfluorooctanoic acid; PFOS, perfluorooc-tanesulfonic acid; PME, particle mesh Ewald; RMSD, rootmean squared deviation; ROC, receiver operating character-istic; SP, standard precision; T3, triiodothyronine; T4,thyroxine; TBBPA, tetrabromobisphenol A; TCEP, tris(2-carboxyethyl)phosphine; TH, thyroid hormone; THDC,thyroid hormone disrupting chemical; THR, thyroid hormonereceptor; TTR, transthyretin; VS, virtual screening; ΔG,binding free energy; ΔH, enthalpy change

■ REFERENCES(1) Murk, A. J., Rijntjes, E., Blaauboer, B. J., Clewell, R., Crofton, K.M., Dingemans, M. M., Furlow, J. D., Kavlock, R., Kohrle, J., Opitz, R.,Traas, T., Visser, T. J., Xia, M., and Gutleb, A. C. (2013) Mechanism-based testing strategy using in vitro approaches for identification ofthyroid hormone disrupting chemicals. Toxicol. In Vitro 27, 1320−1346.(2) Crofton, K. M. (2008) Thyroid disrupting chemicals:mechanisms and mixtures. Int. J. Androl. 31, 209−223.(3) Suzuki, G., Takigami, H., Watanabe, M., Takahashi, S., Nose, K.,Asari, M., and Sakai, S. (2008) Identification of brominated andchlorinated phenols as potential thyroid-disrupting compounds inindoor dusts. Environ. Sci. Technol. 42, 1794−1800.(4) Meeker, J. D., Johnson, P. I., Camann, D., and Hauser, R. (2009)Polybrominated diphenyl ether (PBDE) concentrations in house dustare related to hormone levels in men. Sci. Total Environ. 407, 3425−3429.(5) Johnson, P. I., Stapleton, H. M., Mukherjee, B., Hauser, R., andMeeker, J. D. (2013) Associations between brominated flameretardants in house dust and hormone levels in men. Sci. TotalEnviron. 445−446, 177−184.(6) Zoeller, T. R., Dowling, A. L., Herzig, C. T., Iannacone, E. A.,Gauger, K. J., and Bansal, R. (2002) Thyroid hormone, braindevelopment, and the environment. Environ. Health Perspect. 110(Suppl.3), 355−361.(7) Miller, M. D., Crofton, K. M., Rice, D. C., and Zoeller, R. T.(2009) Thyroid-disrupting chemicals: interpreting upstream bio-markers of adverse outcomes. Environ. Health Perspect. 117, 1033−1041.(8) Noyes, P. D., Lema, S. C., Macaulay, L. J., Douglas, N. K., andStapleton, H. M. (2013) Low level exposure to the flame retardantBDE-209 reduces thyroid hormone levels and disrupts thyroidsignaling in fathead minnows. Environ. Sci. Technol. 47, 10012−10021.(9) Mercier, F., Glorennec, P., Thomas, O., and Le Bot, B. (2011)Organic contamination of settled house dust, a review for exposureassessment purposes. Environ. Sci. Technol. 45, 6716−6727.

Chemical Research in Toxicology Article

DOI: 10.1021/acs.chemrestox.6b00171Chem. Res. Toxicol. 2016, 29, 1345−1354

1351

(10) Zhang, J., Kamstra, J. H., Ghorbanzadeh, M., Weiss, J. M.,Hamers, T., and Andersson, P. L. (2015) In Silico Approach ToIdentify Potential Thyroid Hormone Disruptors among CurrentlyKnown Dust Contaminants and Their Metabolites. Environ. Sci.Technol. 49, 10099−10107.(11) Lioy, P. J., Freeman, N. C., and Millette, J. R. (2002) Dust: ametric for use in residential and building exposure assessment andsource characterization. Environ. Health Perspect. 110, 969−983.(12) Weiss, J. M., Andersson, P. L., Zhang, J., Simon, E., Leonards, P.E., Hamers, T., and Lamoree, M. H. (2015) Tracing thyroid hormone-disrupting compounds: database compilation and structure-activityevaluation for an effect-directed analysis of sediment. Anal. Bioanal.Chem. 407, 5625−5634.(13) Hamers, T., Kamstra, J. H., Sonneveld, E., Murk, A. J., Kester,M. H., Andersson, P. L., Legler, J., and Brouwer, A. (2006) In vitroprofiling of the endocrine-disrupting potency of brominated flameretardants. Toxicol. Sci. 92, 157−173.(14) Weiss, J. M., Andersson, P. L., Lamoree, M. H., Leonards, P. E.,van Leeuwen, S. P., and Hamers, T. (2009) Competitive binding ofpoly- and perfluorinated compounds to the thyroid hormone transportprotein transthyretin. Toxicol. Sci. 109, 206−216.(15) Evans, R. M. (1988) The steroid and thyroid hormone receptorsuperfamily. Science 240, 889−895.(16) Zoeller, R. T. (2007) Environmental chemicals impacting thethyroid: targets and consequences. Thyroid 17, 811−817.(17) Moriyama, K., Tagami, T., Akamizu, T., Usui, T., Saijo, M.,Kanamoto, N., Hataya, Y., Shimatsu, A., Kuzuya, H., and Nakao, K.(2002) Thyroid hormone action is disrupted by bisphenol A as anantagonist. J. Clin. Endocrinol. Metab. 87, 5185−5190.(18) Freitas, J., Cano, P., Craig-Veit, C., Goodson, M. L., Furlow, J.D., and Murk, A. J. (2011) Detection of thyroid hormone receptordisruptors by a novel stable in vitro reporter gene assay. Toxicol. InVitro 25, 257−266.(19) Zloh, M., Perez-Diaz, N., Tang, L., Patel, P., and Mackenzie, L.S. (2016) Evidence that diclofenac and celecoxib are thyroid hormonereceptor beta antagonists. Life Sci. 146, 66−72.(20) Schapira, M., Raaka, B. M., Das, S., Fan, L., Totrov, M., Zhou,Z., Wilson, S. R., Abagyan, R., and Samuels, H. H. (2003) Discovery ofdiverse thyroid hormone receptor antagonists by high-throughputdocking. Proc. Natl. Acad. Sci. U. S. A. 100, 7354−7359.(21) Wang, F. F., Yang, W., Shi, Y. H., and Le, G. W. (2015)Molecular determinants of thyroid hormone receptor selectivity in aseries of phosphonic acid derivatives: 3D-QSAR analysis andmolecular docking. Chem.-Biol. Interact. 240, 324−335.(22) Ye, L., Li, Y. L., Mellstrom, K., Mellin, C., Bladh, L. G., Koehler,K., Garg, N., Garcia Collazo, A. M., Litten, C., Husman, B., Persson, K.,Ljunggren, J., Grover, G., Sleph, P. G., George, R., and Malm, J. (2003)Thyroid receptor ligands. 1. Agonist ligands selective for the thyroidreceptor beta1. J. Med. Chem. 46, 1580−1588.(23) Vedani, A., Zumstein, M., Lill, M. A., and Ernst, B. (2007)Simulating alpha/beta selectivity at the human thyroid hormonereceptor: consensus scoring using multidimensional QSAR. Chem-MedChem 2, 78−87.(24) Politi, R., Rusyn, I., and Tropsha, A. (2014) Prediction ofbinding affinity and efficacy of thyroid hormone receptor ligands usingQSAR and structure-based modeling methods. Toxicol. Appl.Pharmacol. 280, 177−189.(25) Li, F., Xie, Q., Li, X., Li, N., Chi, P., Chen, J., Wang, Z., and Hao,C. (2010) Hormone activity of hydroxylated polybrominated diphenylethers on human thyroid receptor-beta: in vitro and in silicoinvestigations. Environ. Health Perspect. 118, 602−606.(26) Taylor, R. D., Jewsbury, P. J., and Essex, J. W. (2002) A reviewof protein-small molecule docking methods. J. Comput.-Aided Mol. Des.16, 151−166.(27) Kitchen, D. B., Decornez, H., Furr, J. R., and Bajorath, J. (2004)Docking and scoring in virtual screening for drug discovery: methodsand applications. Nat. Rev. Drug Discovery 3, 935−949.

(28) Shoichet, B. K., McGovern, S. L., Wei, B., and Irwin, J. J. (2002)Lead discovery using molecular docking. Curr. Opin. Chem. Biol. 6,439−446.(29) Jorgensen, W. L. (2004) The many roles of computation in drugdiscovery. Science 303, 1813−1818.(30) McRobb, F. M., Kufareva, I., and Abagyan, R. (2014) In silicoidentification and pharmacological evaluation of novel endocrinedisrupting chemicals that act via the ligand-binding domain of theestrogen receptor alpha. Toxicol. Sci. 141, 188−197.(31) Rabinowitz, J. R., Little, S. B., Laws, S. C., and Goldsmith, M. R.(2009) Molecular modeling for screening environmental chemicals forestrogenicity: use of the toxicant-target approach. Chem. Res. Toxicol.22, 1594−1602.(32) Daniel, L., Buryska, T., Prokop, Z., Damborsky, J., andBrezovsky, J. (2015) Mechanism-based discovery of novel substratesof haloalkane dehalogenases using in silico screening. J. Chem. Inf.Model. 55, 54−62.(33) Rabinowitz, J. R., Goldsmith, M. R., Little, S. B., and Pasquinelli,M. A. (2008) Computational molecular modeling for evaluating thetoxicity of environmental chemicals: prioritizing bioassay require-ments. Environ. Health Perspect. 116, 573−577.(34) Muralidharan, A. R., Selvaraj, C., Singh, S. K., Sheu, J. R.,Thomas, P. A., and Geraldine, P. (2015) Structure-Based VirtualScreening and Biological Evaluation of a Calpain Inhibitor forPrevention of Selenite-Induced Cataractogenesis in an in VitroSystem. J. Chem. Inf. Model. 55, 1686−1697.(35) Fukunishi, Y. (2010) Structural ensemble in computational drugscreening. Expert Opin. Drug Metab. Toxicol. 6, 835−849.(36) Ellingson, S. R., Miao, Y., Baudry, J., and Smith, J. C. (2015)Multi-conformer ensemble docking to difficult protein targets. J. Phys.Chem. B 119, 1026−1034.(37) Sun, X. Q., Chen, L., Li, Y. Z., Li, W. H., Liu, G. X., Tu, Y. Q.,and Tang, Y. (2014) Structure-based ensemble-QSAR model: a novelapproach to the study of the EGFR tyrosine kinase and its inhibitors.Acta Pharmacol. Sin. 35, 301−310.(38) Amaro, R. E., Baron, R., and McCammon, J. A. (2008) Animproved relaxed complex scheme for receptor flexibility in computer-aided drug design. J. Comput.-Aided Mol. Des. 22, 693−705.(39) Bashford, D., and Case, D. A. (2000) Generalized born modelsof macromolecular solvation effects. Annu. Rev. Phys. Chem. 51, 129−152.(40) Du, J., Sun, H., Xi, L., Li, J., Yang, Y., Liu, H., and Yao, X. (2011)Molecular modeling study of checkpoint kinase 1 inhibitors bymultiple docking strategies and prime/MM-GBSA calculation. J.Comput. Chem. 32, 2800−2809.(41) Harvey, C. B., and Williams, G. R. (2002) Mechanism of thyroidhormone action. Thyroid 12, 441−446.(42) Flamant, F., Baxter, J. D., Forrest, D., Refetoff, S., Samuels, H.,Scanlan, T. S., Vennstrom, B., and Samarut, J. (2006) InternationalUnion of Pharmacology. LIX. The pharmacology and classification ofthe nuclear receptor superfamily: thyroid hormone receptors.Pharmacol. Rev. 58, 705−711.(43) Ibhazehiebo, K., Iwasaki, T., Kimura-Kuroda, J., Miyazaki, W.,Shimokawa, N., and Koibuchi, N. (2011) Disruption of thyroidhormone receptor-mediated transcription and thyroid hormone-induced Purkinje cell dendrite arborization by polybrominateddiphenyl ethers. Environ. Health Perspect. 119, 168−175.(44) RCSB-PDB. (2015) RCSB Protein Data Bank (PDB). http://www.rcsb.org/pdb/.(45) Jorgensen, W. L., Maxwell, D. S., and TiradoRives, J. (1996)Development and testing of the OPLS all-atom force field onconformational energetics and properties of organic liquids. J. Am.Chem. Soc. 118, 11225−11236.(46) Schrodinger. (2015) Protein Preparation Wizard 2015-2,Schrodinger, LLC, New York, NY.(47) Verdonk, M. L., Chessari, G., Cole, J. C., Hartshorn, M. J.,Murray, C. W., Nissink, J. W., Taylor, R. D., and Taylor, R. (2005)Modeling water molecules in protein-ligand docking using GOLD. J.Med. Chem. 48, 6504−6515.

Chemical Research in Toxicology Article

DOI: 10.1021/acs.chemrestox.6b00171Chem. Res. Toxicol. 2016, 29, 1345−1354

1352

(48) Osterberg, F., Morris, G. M., Sanner, M. F., Olson, A. J., andGoodsell, D. S. (2002) Automated docking to multiple targetstructures: incorporation of protein mobility and structural waterheterogeneity in AutoDock. Proteins: Struct., Funct., Genet. 46, 34−40.(49) Li, Y., Shen, J., Sun, X., Li, W., Liu, G., and Tang, Y. (2010)Accuracy assessment of protein-based docking programs against RNAtargets. J. Chem. Inf. Model. 50, 1134−1146.(50) Bortolato, A., Tehan, B. G., Bodnarchuk, M. S., Essex, J. W., andMason, J. S. (2013) Water network perturbation in ligand binding:adenosine A(2A) antagonists as a case study. J. Chem. Inf. Model. 53,1700−1713.(51) Mysinger, M. M., Carchia, M., Irwin, J. J., and Shoichet, B. K.(2012) Directory of useful decoys, enhanced (DUD-E): better ligandsand decoys for better benchmarking. J. Med. Chem. 55, 6582−6594.(52) ChemAxon (2014) JChem 6.2.0, ChemAxon, Budapest,Hungary, http://www.chemaxon.com.(53) Schrodinger (2014) Maestro, Schrodinger, LLC, New York, NY.(54) Schrodinger (2015) Glide, version 6.7, Schrodinger, LLC, NewYork, NY.(55) Jain, A. N., and Nicholls, A. (2008) Recommendations forevaluation of computational methods. J. Comput.-Aided Mol. Des. 22,133−139.(56) Wei, B. Q., Baase, W. A., Weaver, L. H., Matthews, B. W., andShoichet, B. K. (2002) A model binding site for testing scoringfunctions in molecular docking. J. Mol. Biol. 322, 339−355.(57) Li, J., Abel, R., Zhu, K., Cao, Y., Zhao, S., and Friesner, R. A.(2011) The VSGB 2.0 model: a next generation energy model for highresolution protein structure modeling. Proteins: Struct., Funct., Genet.79, 2794−2812.(58) Iakovleva, I., Brannstrom, K., Nilsson, L., Gharibyan, A. L.,Begum, A., Anan, I., Walfridsson, M., Sauer-Eriksson, A. E., andOlofsson, A. (2015) Enthalpic Forces Correlate with the Selectivity ofTransthyretin-Stabilizing Ligands in Human Plasma. J. Med. Chem. 58,6507−6515.(59) Putcha, B. D., and Fernandez, E. J. (2009) Direct interdomaininteractions can mediate allosterism in the thyroid receptor. J. Biol.Chem. 284, 22517−22524.(60) Tellinghuisen, J. (2008) Isothermal titration calorimetry at verylow c. Anal. Biochem. 373, 395−397.(61) Turnbull, W. B., and Daranas, A. H. (2003) On the value of c:can low affinity systems be studied by isothermal titration calorimetry?J. Am. Chem. Soc. 125, 14859−14866.(62) Borngraeber, S., Budny, M. J., Chiellini, G., Cunha-Lima, S. T.,Togashi, M., Webb, P., Baxter, J. D., Scanlan, T. S., and Fletterick, R. J.(2003) Ligand selectivity by seeking hydrophobicity in thyroidhormone receptor. Proc. Natl. Acad. Sci. U. S. A. 100, 15358−15363.(63) Jorgensen, W. L., Chandrasekhar, J., Madura, J. D., Impey, R.W., and Klein, M. L. (1983) Comparison of Simple PotentialFunctions for Simulating Liquid Water. J. Chem. Phys. 79, 926−935.(64) Brooks, B. R., Brooks, C. L., Mackerell, A. D., Nilsson, L.,Petrella, R. J., Roux, B., Won, Y., Archontis, G., Bartels, C., Boresch, S.,Caflisch, A., Caves, L., Cui, Q., Dinner, A. R., Feig, M., Fischer, S., Gao,J., Hodoscek, M., Im, W., Kuczera, K., Lazaridis, T., Ma, J.,Ovchinnikov, V., Paci, E., Pastor, R. W., Post, C. B., Pu, J. Z.,Schaefer, M., Tidor, B., Venable, R. M., Woodcock, H. L., Wu, X.,Yang, W., York, D. M., and Karplus, M. (2009) CHARMM: TheBiomolecular Simulation Program. J. Comput. Chem. 30, 1545−1614.(65) Phillips, J. C., Braun, R., Wang, W., Gumbart, J., Tajkhorshid, E.,Villa, E., Chipot, C., Skeel, R. D., Kale, L., and Schulten, K. (2005)Scalable molecular dynamics with NAMD. J. Comput. Chem. 26, 1781−1802.(66) Petersen, H. G. (1995) Accuracy and Efficiency of the ParticleMesh Ewald Method. J. Chem. Phys. 103, 3668−3679.(67) Best, R. B., Zhu, X., Shim, J., Lopes, P. E., Mittal, J., Feig, M.,and Mackerell, A. D., Jr. (2012) Optimization of the additiveCHARMM all-atom protein force field targeting improved samplingof the backbone phi, psi and side-chain chi(1) and chi(2) dihedralangles. J. Chem. Theory Comput. 8, 3257−3273.

(68) Vanommeslaeghe, K., and MacKerell, A. D., Jr. (2012)Automation of the CHARMM General Force Field (CGenFF) I:bond perception and atom typing. J. Chem. Inf. Model. 52, 3144−3154.(69) Warren, G. L., Andrews, C. W., Capelli, A. M., Clarke, B.,LaLonde, J., Lambert, M. H., Lindvall, M., Nevins, N., Semus, S. F.,Senger, S., Tedesco, G., Wall, I. D., Woolven, J. M., Peishoff, C. E., andHead, M. S. (2006) A critical assessment of docking programs andscoring functions. J. Med. Chem. 49, 5912−5931.(70) Shelley, J. C., Cholleti, A., Frye, L. L., Greenwood, J. R., Timlin,M. R., and Uchimaya, M. (2007) Epik: a software program for pKaprediction and protonation state generation for drug-like molecules. J.Comput.-Aided Mol. Des. 21, 681−691.(71) Kitamura, S., Jinno, N., Ohta, S., Kuroki, H., and Fujimoto, N.(2002) Thyroid hormonal activity of the flame retardantstetrabromobisphenol A and tetrachlorobisphenol A. Biochem. Biophys.Res. Commun. 293, 554−559.(72) Ren, X. M., Zhang, Y. F., Guo, L. H., Qin, Z. F., Lv, Q. Y., andZhang, L. Y. (2015) Structure-activity relations in binding ofperfluoroalkyl compounds to human thyroid hormone T3 receptor.Arch. Toxicol. 89, 233−242.(73) Du, G., Hu, J., Huang, H., Qin, Y., Han, X., Wu, D., Song, L.,Xia, Y., and Wang, X. (2013) Perfluorooctane sulfonate (PFOS) affectshormone receptor activity, steroidogenesis, and expression ofendocrine-related genes in vitro and in vivo. Environ. Toxicol. Chem.32, 353−360.(74) Zhu, T., Cao, S., Su, P. C., Patel, R., Shah, D., Chokshi, H. B.,Szukala, R., Johnson, M. E., and Hevener, K. E. (2013) Hitidentification and optimization in virtual screening: practicalrecommendations based on a critical literature analysis. J. Med.Chem. 56, 6560−6572.(75) Schneider, G. (2010) Virtual screening: an endless staircase?Nat. Rev. Drug Discovery 9, 273−276.(76) Wagner, R. L., Huber, B. R., Shiau, A. K., Kelly, A., Cunha Lima,S. T., Scanlan, T. S., Apriletti, J. W., Baxter, J. D., West, B. L., andFletterick, R. J. (2001) Hormone selectivity in thyroid hormonereceptors. Mol. Endocrinol. 15, 398−410.(77) Putcha, B. D., Wright, E., Brunzelle, J. S., and Fernandez, E. J.(2012) Structural basis for negative cooperativity within agonist-boundTR:RXR heterodimers. Proc. Natl. Acad. Sci. U. S. A. 109, 6084−6087.(78) Martinez, L., Nascimento, A. S., Nunes, F. M., Phillips, K.,Aparicio, R., Dias, S. M., Figueira, A. C., Lin, J. H., Nguyen, P.,Apriletti, J. W., Neves, F. A., Baxter, J. D., Webb, P., Skaf, M. S., andPolikarpov, I. (2009) Gaining ligand selectivity in thyroid hormonereceptors via entropy. Proc. Natl. Acad. Sci. U. S. A. 106, 20717−20722.(79) Cheek, A. O., Kow, K., Chen, J., and McLachlan, J. A. (1999)Potential mechanisms of thyroid disruption in humans: interaction oforganochlorine compounds with thyroid receptor, transthyretin, andthyroid-binding globulin. Environ. Health Perspect. 107, 273−278.(80) Chevrier, J., Gunier, R. B., Bradman, A., Holland, N. T., Calafat,A. M., Eskenazi, B., and Harley, K. G. (2013) Maternal urinarybisphenol a during pregnancy and maternal and neonatal thyroidfunction in the CHAMACOS study. Environ. Health Perspect. 121,138−144.(81) Bunevicius, R., Kazanavicius, G., Zalinkevicius, R., and Prange,A. J., Jr. (1999) Effects of thyroxine as compared with thyroxine plustriiodothyronine in patients with hypothyroidism. N. Engl. J. Med. 340,424−429.(82) Burns, C. J., and Swaen, G. M. (2012) Review of 2,4-dichlorophenoxyacetic acid (2,4-D) biomonitoring and epidemiology.Crit. Rev. Toxicol. 42, 768−786.(83) Asimakopoulos, A. G., Thomaidis, N. S., and Kannan, K. (2014)Widespread occurrence of bisphenol A diglycidyl ethers, p-hydroxybenzoic acid esters (parabens), benzophenone type-UV filters,triclosan, and triclocarban in human urine from Athens, Greece. Sci.Total Environ. 470−471, 1243−1249.(84) Wang, L., Wu, Y., Zhang, W., and Kannan, K. (2012)Widespread occurrence and distribution of bisphenol A diglycidylether (BADGE) and its derivatives in human urine from the UnitedStates and China. Environ. Sci. Technol. 46, 12968−12976.

Chemical Research in Toxicology Article

DOI: 10.1021/acs.chemrestox.6b00171Chem. Res. Toxicol. 2016, 29, 1345−1354

1353

(85) Morgan, M. K., Wilson, N. K., and Chuang, J. C. (2014)Exposures of 129 preschool children to organochlorines, organo-phosphates, pyrethroids, and acid herbicides at their homes anddaycares in North Carolina. Int. J. Environ. Res. Public Health 11,3743−3764.(86) Ansbaugh, N., Shannon, J., Mori, M., Farris, P. E., and Garzotto,M. (2013) Agent Orange as a risk factor for high-grade prostatecancer. Cancer 119, 2399−2404.(87) Clemens, M. W., Kochuba, A. L., Carter, M. E., Han, K., Liu, J.,and Evans, K. (2014) Association between Agent Orange exposure andnonmelanotic invasive skin cancer: a pilot study. Plast. Reconstr. Surg.133, 432−437.(88) Goldner, W. S., Sandler, D. P., Yu, F., Shostrom, V., Hoppin, J.A., Kamel, F., and LeVan, T. D. (2013) Hypothyroidism and pesticideuse among male private pesticide applicators in the agricultural healthstudy. J. Occup. Environ. Med. 55, 1171−1178.(89) Saldana, T. M., Basso, O., Hoppin, J. A., Baird, D. D., Knott, C.,Blair, A., Alavanja, M. C., and Sandler, D. P. (2007) Pesticide exposureand self-reported gestational diabetes mellitus in the AgriculturalHealth Study. Diabetes Care 30, 529−534.(90) Lemaire, G., Mnif, W., Mauvais, P., Balaguer, P., and Rahmani,R. (2006) Activation of alpha- and beta-estrogen receptors bypersistent pesticides in reporter cell lines. Life Sci. 79, 1160−1169.(91) Yi, S. W., Hong, J. S., Ohrr, H., and Yi, J. J. (2014) AgentOrange exposure and disease prevalence in Korean Vietnam veterans:the Korean veterans health study. Environ. Res. 133, 56−65.(92) Ensminger, M. P., Budd, R., Kelley, K. C., and Goh, K. S. (2013)Pesticide occurrence and aquatic benchmark exceedances in urbansurface waters and sediments in three urban areas of California, USA,2008−2011. Environ. Monit. Assess. 185, 3697−3710.(93) Coady, K. K., Kan, H. L., Schisler, M. R., Gollapudi, B. B., Neal,B., Williams, A., and LeBaron, M. J. (2014) Evaluation of potentialendocrine activity of 2,4-dichlorophenoxyacetic acid using in vitroassays. Toxicol. In Vitro 28, 1018−1025.(94) Morgan, M. K., Sheldon, L. S., Thomas, K. W., Egeghy, P. P.,Croghan, C. W., Jones, P. A., Chuang, J. C., and Wilson, N. K. (2008)Adult and children’s exposure to 2,4-D from multiple sources andpathways. J. Exposure Sci. Environ. Epidemiol. 18, 486−494.(95) Deziel, N. C., Colt, J. S., Kent, E. E., Gunier, R. B., Reynolds, P.,Booth, B., Metayer, C., and Ward, M. H. (2015) Associations betweenself-reported pest treatments and pesticide concentrations in carpetdust. Environ. Health 14, 27−38.(96) Landrigan, P. J., and Benbrook, C. (2015) GMOs, Herbicides,and Public Health. N. Engl. J. Med. 373, 693−695.(97) Wang, L., Liao, C., Liu, F., Wu, Q., Guo, Y., Moon, H. B.,Nakata, H., and Kannan, K. (2012) Occurrence and human exposureof p-hydroxybenzoic acid esters (parabens), bisphenol A diglycidylether (BADGE), and their hydrolysis products in indoor dust from theUnited States and three East Asian countries. Environ. Sci. Technol. 46,11584−11593.(98) Rannar, S., and Andersson, P. L. (2010) A novel approach usinghierarchical clustering to select industrial chemicals for environmentalimpact assessment. J. Chem. Inf. Model. 50, 30−36.(99) Liao, C., and Kannan, K. (2014) Widespread occurrence ofbenzophenone-type UV light filters in personal care products fromChina and the United States: an assessment of human exposure.Environ. Sci. Technol. 48, 4103−4109.(100) Nakajima, D., Asada, S., Kageyama, S., Yamamoto, T.,Kuramochi, H., Tanaka, N., Takeda, K., and Goto, S. (2006) Activityrelated to the carcinogenicity of plastic additives in the benzophenonegroup. J. UOEH 28, 143−156.(101) Wang, L., Asimakopoulos, A. G., Moon, H. B., Nakata, H., andKannan, K. (2013) Benzotriazole, benzothiazole, and benzophenonecompounds in indoor dust from the United States and East Asiancountries. Environ. Sci. Technol. 47, 4752−4759.(102) Kunisue, T., Chen, Z., Buck Louis, G. M., Sundaram, R.,Hediger, M. L., Sun, L., and Kannan, K. (2012) Urinary concentrationsof benzophenone-type UV filters in U.S. women and their associationwith endometriosis. Environ. Sci. Technol. 46, 4624−4632.

(103) Schmutzler, C., Bacinski, A., Gotthardt, I., Huhne, K.,Ambrugger, P., Klammer, H., Schlecht, C., Hoang-Vu, C., Gruters,A., Wuttke, W., Jarry, H., and Kohrle, J. (2007) The ultraviolet filterbenzophenone 2 interferes with the thyroid hormone axis in rats and isa potent in vitro inhibitor of human recombinant thyroid peroxidase.Endocrinology 148, 2835−2844.(104) OECD (2008) Report of the Validation of the AmphibianMetamorphosis Assay (Phase 3). https://www.oecd.org/chemicalsafety/testing/41620749.pdf (accessed May 6, 2016).(105) Molina-Molina, J. M., Escande, A., Pillon, A., Gomez, E.,Pakdel, F., Cavailles, V., Olea, N., Ait-Aissa, S., and Balaguer, P. (2008)Profiling of benzophenone derivatives using fish and human estrogenreceptor-specific in vitro bioassays. Toxicol. Appl. Pharmacol. 232,384−395.

Chemical Research in Toxicology Article

DOI: 10.1021/acs.chemrestox.6b00171Chem. Res. Toxicol. 2016, 29, 1345−1354

1354