Chemical Prioritization Methods for Nuclear Receptor Modulators at the U.S. EPA

Keith A. Houck, Ph.D. National Center for Computational Toxicology (NCCT/ORD/EPA) EDTA International Symposium, Dongguk University November 9, 2018 [email protected]

The views expressed in this presentation are those of the author and do not necessarily reflect the views or policies of the U.S. EPA Use of commercial names does not constitute endorsement of those brands U.S. Environmental Protection Agency Regulatory Agencies Make a Broad Range of Decisions on Chemicals…

Number of Chemicals Lack of Data /Combinations 70 60 • Number of chemicals and 50 combinations of chemicals is 40 30 extremely large (>20,000 substances 20 <1% Percent of Chemicals of Percent 10 on active TSCA inventory)

0 Acute Cancer • Due to historical regulatory Gentox Dev Tox requirements, most chemicals lack Repro Tox EDSP Tier 1 Modified from Judson et al., EHP 2010 traditional toxicity testing data Ethics/Relevance Economics • Traditional toxicology testing is Concerns $10,000,000 expensive and time consuming

$1,000,000 • Traditional animal-based testing has $100,000

Cost issues related to ethics and relevance $10,000

$1,000 Toxicology Moving to Embrace 21st Century Methods

3 High-Throughput Assays Used to Screen Chemicals for Potential Toxicity

Hundreds High‐ Thousands Throughput of Chemicals ToxCast/Tox21 Assays

• Understanding of what cellular processes/pathways may be perturbed by a chemical

4 • Understanding of what amount of a chemical causes these perturbations Broad Success Derived from High- Throughput Screening Approaches

Group Chemicals by Provide Mechanistic Prioritization of Chemicals Similar Bioactivity and Support for Hazard ID for Further Testing Predictive Modeling Chemicals

FIFRA SAP, Dec 2014

IARC Monographs Assays/Pathways 5 Focus on Nuclear Receptors and Xenobiotics

• Family of ligand-regulated nuclear transcription factors (48 human) • Conserved, modular domains – DNA-binding domain – Ligand-binding domain • Binds lipophilic, small molecules • Endogenous ligands: steroid hormones, fatty acids • Regulates genes for key physiological processes: endocrine system, growth and differentiation, metabolism • Endogenous ligand physicochemical properties consistent with cell permeable qualities

• Good focus for selective xenobiotic effects http://proteopedia.org/wiki/index.php/Image:3dzy2.png Ligands for Nuclear Hormone Receptors From the EPA’s Endocrine Disruptor Screening Management Plan: “Examine effects of these chemicals on estrogen, androgen and thyroid hormone-related processes”

Sex Steroids ER Vitamin D VDR Estrogens, Progestins, Androgens Unknowns PR OH O OH OH AR

O HO O HO OH Glucocorticoids GR Lipids PPARs OH O O HO OH OH

O O Mineralocorticoids MR Oxysterols LXRs OH HO O O O

O HO Thyroid Hormones TR Bile Acids O FXR I I OH NH O 2 H HO O

I OH

HO OH H Retinoids RARs Xenobiotics PXR

O RXRs Cl CAR OH N N 7 The Estrogen Receptor Model

• Public solicitation for diverse high-throughput assays to cover broad range of bioactivity/toxicity endpoints • Many estrogen receptor assays included – Binding – Nuclear localization – Transactivation – Cell proliferation • No single assay perfect for a variety of reasons • Decided to develop computational model utilizing all data Targeted Pathways

18 In Vitro Assays Measure ER-Related Activity

R3 A1 Receptor (Direct Molecular Interaction) A2 Intermediate Process A3 Assay ER Receptor ER Receptor Binding 3 Binding R2 R1 Noise Process (Antagonist) A4 (Agonist) R4 ER agonist pathway A5 ER antagonist pathway A6 Dimerization N7 N1 Dimerization Interference pathway A7

Cofactor A8 Cofactor N8 N2 Recruitment Recruitment A9 R5 DNA DNA N9 N3 Binding A10 Binding R6

A12 R9 A11 RNA N4 Transcription A13 A17 Antagonist R7 N10 Transcription Protein N5 A14 A18 Suppression Production

A15 ER‐induced R8 N6 Judson et al., Tox Sci. 2015 Proliferation A16 Browne et al., ES&T. 2015 Kleinstreuer et al., EHP 2016 ER Model Performance

In vivo Comparison

Rank Order (ER Agonist AUC)

Judson et al., Tox Sci 2015 Browne et al., Environ. Sci. Technol., 2015 ER Minimal Model

Combinations of four assays provide good balanced accuracy

R.S. Judson et al. / Regulatory Toxicology and Pharmacology 91 (2017) 39e49 Regulatory Applications: EDSP

“The approach incorporates validated high-throughput assays and a computational model and, based on current research, can serve as an alternative for some of the current assays in the Endocrine Disruptor Screening Program (EDSP) Tier 1 battery.” Androgen Receptor Screening

• Utilize existing ToxCast/Tox21 assays to develop AR model • Cytotoxic chemicals confounded antagonist cell-based assays • Run additional confirmation assay for antagonists – Higher agonist concentration – Competitive antagonists show right-shift in potency ToxCast/Tox21 Assays

Antagonist Mode Evaluation of AR Model

Summary Reference Data • Model has high sensitivity Literature Review • Antagonist mode specificity improved by considering Reference Chemical antagonist assay with high Classifications agonist concentration • Weakly active chemicals most Model Performance difficult to detect Evaluation • Broad screening suggested cytotoxic compounds not all excluded

Validated Model for Chemical Screening

Chem Res Toxicol. 30:946-964, 2017. Agonists versus Antagonist

Selective receptor modulators behave conditionally as agonists and/or antagonists

17-Estradiol

4-Hydroxytamoxifen

Fulvestrant

Brzozowski et al., Nature. 389: 753–8, 1997). Tox21 AR Screening Results ⁓ 8,000 chemicals

Hydroxyflutamide Bis(tributyltin)oxide • Only 102 chemicals positive Dipyrithione using strictest criteria Ziram NTP Mix21 AR2 2‐EQP • Expanding criteria allows for ranking of chemicals based on 17alpha‐Ethinylestradiol strength of evidence Bis(1‐piperidinylthioxomethyl)hexasulfide Triphenyltin acetate • Chemicals that are confounded Tributyltin benzoate by cytotoxicity are not eliminated Nilutamide but evidence is weaker Triethyltin bromide • Potency not currently considered Equilin but is another important factor 17alpha‐Estradiol (Acryloyloxy)(tributyl)stannane Triphenyltin fluoride Ethylestrenol Copper dimethyldithiocarbamate Vinclozolin Challenges with assessing NR antagonism in vitro

• Measuring loss of signal- confounded by cytotoxicity • To address: – Two different assay platforms – Use bootstrapping techniques to determine effect of cytotoxicity – Two concentrations of agonist R1881 – MARCoNI assay for corepressor/activator recruitment Antagonist Reference Chemical Results LUC vs LUC vs Chemical Designation Assay Hitcalls LUC_counterscreen LUC_viability Procymidone Very Weak Antagonist BLA, LUC, LUCcs Yes Yes Fenarimol Very Weak Antagonist BLA, LUC, LUCcs Yes Yes 4-(1,1,3,3- Tetramethylbutyl)phenol Weak Antagonist LUC Yes Yes o,p'-DDT Weak Antagonist BLA, LUC Yes Yes Antagonist p,p'-DDE Weak Antagonist LUC Yes Yes Propiconazole Weak Antagonist BLA, LUC, LUCcs Yes No Screening Zearalenone Weak Antagonist BLA, LUC, LUCcs No No Methoxychlor Weak Antagonist BLA, LUC, LUC2 No No • LUC: R1881 = 0.5 nM Linuron Moderate/Weak Antagonist BLA, LUC Yes No Vinclozolin Moderate/Weak Antagonist BLA, LUC, LUCcs Yes Yes Flutamide Moderate/Weak Antagonist BLA, LUC, LUCcs Yes Yes Bisphenol A Moderate/Weak Antagonist BLA, LUC, LUCcs Yes Yes • LUC_counterscreen: Prochloraz Moderate/Weak Antagonist BLA, LUC, LUCcs Yes Yes Cyproterone acetate Moderate Antagonist BLA, LUC Yes Yes R1881 = 10 nM Nilutamide Moderate Antagonist BLA, LUC, LUCcs Yes Yes Spironolactone Strong/Moderate Antagonist BLA, LUC No Yes Mifepristone Strong/Moderate Antagonist BLA, LUC, LUCcs No Yes Fenitrothion Strong Antagonist BLA, LUC, LUCcs Yes Yes Hydroxyflutamide Strong Antagonist BLA, LUC, LUCcs Yes Yes

Bicalutamide Strong Antagonist BLA, LUC, LUCcs Yes Yes 17-Methyltestosterone Negative Antagonist NA NA NA 4-Androstene-3,17-dione Negative Antagonist NA NA No Atrazine Negative Antagonist NA NA NA Daidzein Negative Antagonist BLA NA NA Deltamethrin Negative Antagonist NA NA NA Methomyl Negative Antagonist LUCcs NA No Simazine Negative Antagonist NA NA NA MARCoNI assay Microarray Assay for Real-time Coregulator-Nuclear receptor Interaction

• Cell-free assay measuring co-regulator recruitment to AR-LBD – 154 co-regulators – 3 concentrations (1, 10, 100 uM) – log fold-change of binding compared to DMSO

• Tested 318 suspected AR antagonists

• Reduced variables (co-regulators) to 28 most affected

• Goal: to see if patterns of coregulatory recruitment can distinguish between true antagonists and false antagonists (cytotoxicity/artifacts)

Image: pamgene.com High confidence: Cluster 1-2 Lower confidence: Cluster 3 2 1 34 No confidence: Clusters 4-7 5 67

7

Cyproterone 6

5 Bicalutamide

DHT 4

Hydroxyflutamide Unique 3 chemicals 2

1

Rank Spearman Dissimilarity/Wards Co-regulator Recruitment Patterns

• Mean value of cluster 1 2 3 4 5 6 7 plotted per coregulator

• Loss of binding seen 1 in cluster 2&3 versus 1 (red oval) 2 • These represent SRC 3 coactivators that have 4 6 histone acetyl 7 5 transferase activity • Selective receptor modulators; likely would influence biological response Thyroid Hormone Receptor Modulators: Tox21 qHTS Campaign

Primary Screen Hit Characterization

ASSAYS ASSAYS Rat pituitary GH3 cell line* GH3 expressing endogenous TRα GAL4-TR (human) and TRβ, with TRE regulating GAL4-RXR (human) luciferase expression TR/TR coactivator recruitment Cell viability TR nuclear translocation

* Developed by Albertinka Murk, Wageningen University, the Netherlands TR Modulator Hit Characterization: TR Coactivator Assay (Invitrogen):

GST-tagged TR-ligand binding domain

• Direct-acting T3 (TR agonist) modulators should TR- TR- GST GST regulate coactivator LBD LBD recruitment • Test in both agonist (recruitment) and Tb dye-labeled anti-GST Fluorescein-labeled antagonist antibody coactivator SRC2-2 peptide (disassociation) format

FRET = Iem, Fl/Iem, Tb Low FRET High FRET Optimization of TR Coactivator Assay

.

2.5 nM TR-LBD S/B CV% EC50 (M) TR-LBD S/B CV% EC50 (M) 10 min 4.4 3.9 5.911E-10 10 min 5.0 3.6 2.786E-10 20 min 5.3 4.7 4.786E-10 20 min 5.9 3.3 2.605E-10 30 min 6.1 4.6 4.661E-10 30 min 6.7 4.2 2.629E-10 60 min 6.8 4.8 4.380E-10 60 min 7.0 4.6 2.571E-10 120 min 7.4 5.7 4.507E-10

U.S. Environmental Protection Agency Example Agonists

Direct TR Agonist Indirect TR Agonist/RXR Agonist

13-cis retinoic acid Betamipron TR_GH3 TR_SRC2 TR_GH3 TR_SRC2

TR_HEK293 RXR_ HEK293 TR_HEK293 RXR_HEK293 TR Agonists

Direct

Indirect TR-FRET TR Coactivator Assay, Antagonist Mode

Three known TR antagonists tested with the MLS000389544 tested on various coactivator peptides: SRC2-2 peptide:

No understanding of why this assay failed.

U.S. Environmental Protection Agency Reference for MLS000389544: J Biomol Screen. 2011 Jul;16(6):618-27. Development of a TR Nuclear Localization Assay

Stavreva et al., Toxicology 368–369: 69-79, 2016. TR Nuclear Translocation Assay

Nuclear/Cytoplasm GFP Nuclear/Cytoplasm Screened 300 chemicals Antagonist Characterization Examples

TR Antagonist

Risarestat

Cytotoxic Other?

Dichlofluanid Carfilzomib TR Antagonist Candidates Modes of TR Modulation

Ligand-Independent Repression Agonism Agonism by Permissive Antagonism Heterodimer

CoR

apo receptor

Physiological/Toxicological Relevance? Thyroid Axis Targets Some Existing Limitations in High- Throughput and In Vitro Test Systems

Biological Coverage Chemical Coverage and Specific (Gene Basis) Chemical Types (e.g., VOCs)

Human Focus

Organ and Tissue Metabolic Responses Competence

34 Assessing Cross-Species

Differences in Response NR family NR Class Species Sequence ID ER1 Danio rerio BC162466 ER2a Fish Danio rerio BC044349 Multispecies Attagene Trans Reporter Assay ER2b Danio rerio BC086848 ER1 Xenopus laevis NM_001089617 Amphibian Estrogen ER2 Xenopus laevis NM_001130954 ER1 Reptilian Chrysemys picta NM_001282246 ER1 Avian Gallus gallus NM_205183 ERa Homo Sapiens NM_000125 Mammalian ERb Homo Sapiens NM_001437 AR Fish Danio rerio NM_001083123 AR Amphibian Xenopus laevis NM_001090884 Androgen AR Reptilian Chrysemys picta XM_005279527 AR Avian Gallus gallus NM_001040090 AR Mammalian Homo Sapiens NM_000044 TRa Danio rerio BC096778 Fish TRb Danio rerio BC163114 TRa Amphibian Xenopus laevis NM_001088126 Thyroid TRa Reptilian Chrysemys picta XM_005294120 THRa Homo Sapiens NM_199334 Mammalian Highly multiplexed THRb Homo Sapiens NM_000461 PPARg Fish Danio rerio NM_131467 PPAR PPARg Mus musculus NM_001127330 reporter gene Mammalian assay PPARg Homo Sapiens BC006811 • Host cell: human HepG2 • Agonist mode for all receptors • Antagonist for ER and AR

Houck et al., unpublished Cross-Species Differences in Nuclear Receptor Responses

• 180 Chemicals tested in concentration- response • Chemicals selected for NR activity Assays Retrofit for Xenobiotic Metabolism: Extracellular

37 Assays Retrofit for Xenobiotic Metabolism: Intracellular

38 Environmental Monitoring Application: Nationwide Streams Surveillance

• 38 total sites (4 reference sites) across US and PR • Water samples collected 2012-2014 • Locations varied by watershed drainage area, ag/urban use, population density

39 Brett Blackwell/ORD/EPA Bioassay Analysis Workflow

Ambient Filtered DMSO “Unknown” Water Chemical Sample Extracted 1000x 200mg Mixture conc. HLB

trans-FACTORIAL cis-FACTORIAL 8 15 AhRE PXR PXRE Extract Analysis ERa 6 ERE PPARg • 6-point curve; 3-fold dilution 10 NRF2_ARE GR • 24h exposure 4 VDRE HIF1A 5 • Area Under Curve (AUC)

Fold Induction Fold 2 • Response relative to extract blank

0 0 0.01 0.1 1 10 100 0.01 0.1 1 10 100 Relative Enrichment Factor (REF) Relative Enrichment Factor (REF) 40 Bioassay Results

TGFb PPARd1 HNF6 NURR1VDR TCF/b-catE-Box HNF4a 16 PPRE RXRb NFI RORg PXR • 26/70 endpoints AUC >1.25-fold PPARa GRE CAR 8 AP-1 (borderline active) LXRb ISRE THRa1 MRE PXR 4 STAT3 ERRa TAL LXRa NF-kB • 11/70 endpoints AUC >1.5-fold (active) 2 ERa FoxA2 RORb CMV

ERRg 1 Xbp1 Active Endpoints PPARg CRE RARa 0.5 Ahr • PXRE, PXR, AhRE – 30-36 sites RARb EGR GR NRF2/ARE • ERE – 17 sites RXRa TA RARg ERE • ERα, PPARγ – 10 sites AR Oct-MLP FXR DR4/LXR • GR, VDRE, NRF2 – 6-8 sites Myc HSE Sp1 SREBP Sox p53 • RORE, RXRβ – 2 sites FoxO BRE DR5 Pax6 AP-2 HIF1a IR1 VDRE PBREM RORE 41 Myb Ets C/EBP E2F NRF1GLI GATA Concluding Remarks

• In vitro/alternative and computational approaches are valuable for chemical prioritization, especially where we understand the targets and toxicity pathways • Nuclear receptors are an inherently important target of environmental chemicals

• Endocrine disruption is one important mode of action mediated by NR’s but there are many more receptors with varied, important physiology • Using high-throughput approaches will require systematically addressing key technical and data analysis challenges • assay interferences • selective receptor modulators • cytotoxicity Thank You for Your Attention!

• ICCVAM • NCATS/NIH – Nichole Kleinstreuer – Menghang Xia – Patricia Ceger – Ruili Wang – Warren Casey – Chia-Wen Hsu – David Allen • OECD • EPA/ORD/NHEERL – Patience Browne – Mike Hornung • NCI/NIH – Susan Laws – Diana Stareva – Tammy Stoker – Vikas Soni – Jun Wang – Lyuba Varticovski – Daniel Hallinger – Razi Raziuddin – Ashley Murr – Gordon Hager – Angela Buckalew • NCCT/ORD/EPA – Joseph Korte – Katie Paul-Friedman – Jennifer Olker – Richard Judson – Jeff Denny – Kevin Crofton – Carsten Knutsen – Rusty Thomas – Phillip Hartig – Audrey Bone – Mary Cardon – Ann Richard – Sigmund Degitz – Matt Martin – Dan Villeneuve – Tom Knudsen – Anthony Schroeder EPA’s National Center for Computational – Nisha Sipes – Gerald Ankley Toxicology – Eric Watt – Brett R Blackwell – David Dix – Woody Setzer