Tox21: A U.S. Federal Collaboration to Improve the
Human Hazard Characterization of Chemicals
Raymond Tice, Ph.D.
Chief, Biomolecular Screening Branch
(tice@niehs.nih.gov)
BSRC Regulatory Science Symposium
Shinagawa, Japan
February 21, 2014
Topics
• Background
• Phase I – Proof of principle
• Phase II - Expanded compound screening
• Phase III - Improving on biological coverage and
relevance
• Chemical prioritization efforts
2
Our Need for Prioritization
Too Many Chemicals
Too Little Data (%)
9912
Judson et al., EHP 117:685-695 (2009)
3
Toxicity Testing in the 21st Century
This 2007 National Academy
of Science report envisions a
not-so-distant future in
which virtually all routine
toxicity testing would be
conducted in vitro in
human cells or cell lines by
evaluating perturbations of
cellular responses in a
suite of toxicity pathway
assays using high
throughput robotic
assisted methodologies. 4
Activation of a Toxicity Pathway
Exposure
Tissue Dose
Biologic Interaction
Low Dose
Higher Dose
Higher yet
Perturbation
Normal
Biologic
Function
Biologic
Inputs
Early Cellular
Changes
Adaptive Stress
Responses
NAS Report, 2007
Cell
Injury
Morbidity
and
Mortality
5
Formation of the U.S. Tox21 Community
•
•
•
5-year Memorandum of Understanding
(MoU) on “High-Throughput Screening,
Toxicity Pathway Profiling, and
Biological Interpretation of Findings”
released on Feb 14, 2008 signed by
NHGRI (F.S. Collins), NIEHS/NTP
(S.H. Wilson), and EPA (G.M. Gray).
Revised 5-year MoU to add FDA signed on July 19, 2010
http://ntp.niehs.nih.gov/go/28213) by NHGRI (E.D. Green), NIEHS/NTP
(L.S. Birnbaum), EPA (P.T. Anastas), and FDA (J. Woodcock).
Known informally as Tox21 for Toxicology in the 21st Century
6
Tox21 Goals
•
•
•
Identify patterns of compoundinduced biological response in
order to:
−
characterize toxicity/disease
pathways
−
facilitate cross-species extrapolation
−
model low-dose extrapolation
Prioritize compounds for more
extensive toxicological evaluation
Develop predictive models for
biological response in humans
7
Areas of Interest
NIEHS/NTP
NCATS
EPA
FDA
Lab Animal Toxicology



Human Toxicology/Exposure
Assessment




Ultra High Throughput Screening
Low to Mid Throughput Assays




Stem Cell Assay Development




Epigenetic Assays


Engineered Tissue Models




‘Omic Based Systems




Lower Organism Models



Genetic Variability in Response


Databases & Informatic Tools




Validation Experience




8
Agency Points of Contact
FDA - Thomas Colatsky, Ph.D.
NCGC/NCATS –Anton Simeonov, Ph.D.
EPA/NCCT –Russell Thomas, Ph.D.
NIEHS/NTP - Raymond Tice, Ph.D.
Assays & Pathways
Working Group
Chemical Selection
Working Group
Informatics
Working Group
Targeted Testing
Working Group
Co-Chairs
Kevin Gaido, Ph.D. (FDA)
Keith Houck, Ph.D. (EPA)
Kristine Witt, M.S. (NTP)
Menghang Xia, Ph.D. (NCGC)
Co-Chairs
William Leister, Ph.D. (NCGC)
Donna Mendrick, Ph.D. (FDA)
Ann Richard, Ph.D. (EPA)
Suramya Waidanatha, Ph.D. (NTP)
Co-Chairs
Ruili Huang, Ph.D. (NCGC)
Richard Judson, Ph.D. (EPA)
Jennifer Fostel, Ph.D. (NIEHS)
Weida Tong, Ph.D. (FDA)
Co-Chairs
Michael DeVito, Ph.D. (NTP)
David Gerhold, Ph.D. (NCGC)
Timothy Shafer, Ph.D. (EPA)
James Weaver, Ph.D. (FDA)
 Identify toxicity
pathways &
corresponding
assays
 Establish compound
 Review nominated
assays and
prioritize for use at
the NCGC
 Establish QC
libraries for qHTS
(10K, mixtures, watersoluble)
procedures for
compound identity,
purity, concentration,
and stability
 Evaluate assay
performance
 Develop prioritization
schemes and
prediction models
 Make all data
publicly accessible
 Evaluate relevance
of prioritization
schemes &
prediction models
 Extrapolate in vitro
concentration to in
vivo dose
9
Tox21 Phase I – Proof of Principle
(2005 – 2010)
•
EPA via ToxCast™ screened 320 compounds (309
unique, primarily pesticide actives and some endocrine
active compounds) in ~550 assays.
–
•
-
Data made public via ACToR (Aggregated Computational
Toxicology Resource; http://epa.gov/actor)
NCGC screened 1408 compounds (1353 unique) from
NTP and 1462 compounds (1384 unique) from EPA in 140
qHTS assays representing 77 predominantly cell-based
reporter gene endpoints.
Data made public via PubChem (http://pubchem.ncbi.nlm.nih.gov/)
and will be available in CEBS (Chemical Effects in Biological Systems;
http://www.niehs.nih.gov/research/resources/databases/cebs/)
10
Quantitative High Throughput Screening (qHTS) at NCATS
•
•
•
•
•
•
•
DMSO soluble compounds
homogeneous assays
1536-well plate format
15-point concentrationresponse curve
5 nM to 92 M typical
~5 L assay volume
~1000-2000 cells/well
11
PhaseII qHTS
NCGC qHTS Assays
Tox21 Phase
•
Phenotypic readouts
−
−
−
−
•
•
•
•
−
Cytotoxicity
Apoptosis: caspase 3/7, 8, 9
Membrane integrity: LDH, protease
release
•
Mitochondrial toxicity (membrane
potential)
Genetox: p53, ATAD5, DNA damage
repair deficient lines (chicken DT40 lines)
Cell Signaling
−
Stress response: ARE, ESRE, HSP,
Hypoxia, AP-1
−
−
Immune response: IL-8, TNF, TTP
Other: AP-1, CRE, ERK, HRE, JNK3,
NFkB
Epigenetics
−
Locus DeRepression (LDR)
Drug metabolism
•
CYP1A2, CYP2C19, CYP2C9, CYP2D6,
CYP3A4
Target specific assays
−
Nuclear receptors: AR, AhR, ER, FXR,
GR, LXR, PPAR, PPARδ, PPARγ, PXR,
RXR, TRβ, VDR, ROR, RORγ
−
−
hERG channel
Isolated molecular targets: 12hLO,
15hLO1, 15hLO2, ALDH1A1, HADH560,
HPGD, HSD17b4, APE1, TDP1, DNA
polymerase III, RECQ1 helicase, RGS4,
BRCA, IMPase, O-Glc NAc Transferase,
Caspase-1/7, CBFβ-RUNX1, PK, Tau,
Cruzain, β-Lactamase, PRX, YjeE , NPS,
Proteasome, SF1, SMN2, beta-globin
splicing, Anthrax Lethal Factor, TSHR
Genetic variability: 87 HapMap CEPH Panel
12
Conclusions from Tox21 Phase I
•
qHTS can be used to screen libraries of environmental compounds
against targets with known biological significance
•
Cell lines are more suitable for qHTS assays than primary cells
•
High quality data are essential for automated data analysis and
interpretation
- Signal to background ratio (S/B)
- CV and Z’ factor
- Reproducibility of duplicate compounds within and between runs
•
Biological relevance is critical: do the results make sense based on
available standard toxicity test data and/or human data
- Importance of reference compounds
•
Significant limitation remains the lack of hepatic metabolic activation
Tice et al. EHP 121:756–765 (2013)
13
Tox21 Phase II – Expanded Compound Screening
(2011 – 2014)
•
•
EPA’s ToxCast™ Phase II: ~700 compounds in ~700 assays,
~1000 compounds in endocrine activity assays
NCGC qHTS Phase II:
–
–
10K compound library screened 3 times at 15 concentrations in each qHTS
assay
qHTS assays focused on:



•
nuclear receptor activation or inhibition
induction of cellular stress response pathways
characterizing human variability in response
Partner-lead projects
–
–
–
–
cardiotoxicity (FDA)
endocrine disruptors (EPA)
genotoxicity (NIEHS/NTP)
mitochondrial toxicity (NCATS)
14
Tox21 10K Compound Library
14000
12000
EPA
(3726)
10000
8000
(1328)
GSIDs
6000
Tox21 IDs
4000
wells
(623)
NTP
(3194)
367
2000
(553)
0
EPA
NTP
NCGC
Total
NCGC (3526)
Total
Unique
Library tested 3x in each assay
Unique
EPA
NTP
NCGC
Total Total Unique
GSIDs
3726
3194
3524
10444
8307
unique substances
Tox21 IDs
3729
3210
3733
10672
10496
unique solution IDs
wells
4224
3726
4224
12174
12174
total number of test cmpd wells
88 single-sourced
cmpds in duplicate
on each plate
2255 replicate substances (GSIDs) across 3 inventories
Compound identity and structures available at http://www.epa.gov/ncct/dsstox/sdf_tox21s.html
15
Tox21 10K Compound Library
NCGC
•
•
•
3524 cmpds
Drugs
Active
pharmaceutical
ingredients
EPA
NTP
•
•
3726 cmpds
•
Antimicrobial
Registration Program
related compounds
•
Endocrine Disruptor
Screening Program
•
OECD Molecular
Screening Working
Group List
reference compounds
from in vivo regulatory
tests
ToxCast I and II
compounds
•
FDA Drug Induced
Liver Injury Project
•
Failed Drugs
•3194 cmpds
•NTP-studied compounds
•NTP nominations and
•NICEATM/ICCVAM
•External collaborators
(e.g., Silent Spring
Institute, U.S. Army Public
Health Command)
•Formulated mixtures
16
Phase II Stress Response qHTS Assays
Oxidative stress
ARE/Nrf2 in HepG2
P53 activation in HCT-116 colon cancer cells
Genotoxic stress
ATAD5 levels in HEK293 cells (ATPase family AAA
domain-containing protein 5 – a DNA damage response
element)
DT40 (DNA-repair mutant isogenic chicken cell clones)
(Rev3 (-/-), rad54/ku70 (-/-), wild type)
pH2AX induction in CHO cells
Heat shock
Hsp70 in HeLa or HepG2 cells
ER stress
ESRE (lipid damage) in HeLa cells
Hypoxia
HRE (HIF-1α) in ME-180 cervical carcinoma cells
Inflammation
NFκB in ME-180 cells
AP-1 activation in ME-180 or HepG2 cells
Multiple stresses, cell death,
specific toxicities
Caspase 3/7 activation
LDH release, ATP levels
mitochondrial membrane potential in HepG2 cells
hERG (ion channel effects) in U2OS cells (cardiotoxicity)
Cell death/viability kinetic studies in 3 cell types
*Bolded text indicates completed assays
17
Phase II Nuclear Receptor and Related qHTS Assays*
hAhR full length receptor in HepG2 cells
hAR full length receptor in MDA kb2 cells; partial receptor in HEK293 cells
hERα full length receptor in BG1 cells; partial receptor in HEK293 cells
hFXR partial receptor in HEK293 cells
hGR full length receptor in HeLa cells
hPPARδ partial receptor in HEK293 cells
NR assays conducted
in agonist and antagonist
modes
hPPARγ partial receptor in HEK293 cells
hPXR full length receptor in HepG2 cells
hRORγ partial receptor in CHO cells
rTRβ full length receptor in GH3 cells; partial human receptor in HEK293 cells
hVDR partial receptor in HEK293 cells
Inhibition of aromatase using MCF-7 cells
*Bolded text indicates completed assays
18
Tox21 10K Screen Performance
Active
match (%)
Inactive
match (%)
AR-BLA agonist
5.4
86.9
7.4
0.3
1.8
89.6
AR-BLA antagonist
AR-MDA agonist
15.5
4.1
75.1
93.7
9.4
2.2
0.1
0.0
1.4
1.4
96.4
99.7
AR-MDA antagonist
13.2
76.4
10.1
0.3
1.5
92.3
ERα-BG1 agonist
16.4
71.2
12.1
0.3
1.5
91.5
ERα-BG1 antagonist
12.0
79.7
8.0
0.3
1.5
95.3
ERα-BLA agonist
7.0
87.1
5.9
0.0
1.4
95.3
ERα-BLA antagonist
9.8
77.9
12.0
0.3
1.5
84.9
FXR-BLA agonist
2.5
93.9
3.7
0.0
1.5
95.1
FXR-BLA antagonist
GH3-TRβ agonist
7.5
2.1
83.0
90.7
9.4
7.2
0.1
0.0
1.7
1.4
88.5
87.8
GH3-TRβ antagonist
17.2
68.4
14.2
0.3
1.4
87.8
GR-BLA agonist
6.7
87.5
5.8
0.1
1.4
94.9
GR-BLA antagonist
9.7
75.1
13.1
2.0
1.8
77.4
Assay
Inconclusive Mismatch
AC50 fold Score*
(%)
(%)
*Score = 2x%active match + %inactive match - %inconclusive – 2x%mismatch
19
Tox21 10K Screen Performance (Continued)
Assay
PPARδ-BLA agonist
PPARδ-BLA antagonist
PPARγ-agonist
PPARγ-antagonist
VDR-BLA agonist
VDR-BLA antagonist
Aromatase
AhR
ARE
ATAD5
DT40 (100)
DT40 (653)
DT40 (657)
HSE-BLA
Mitochondria Memb. Pot.
P53-BLA
Active
match (%)
Inactive Inconclusive Mismatch
match (%)
(%)
(%)
AC50
fold
Score*
3.5
90.8
5.7
0.0
1.7
91.9
5.7
8.8
8.6
2.3
5.4
15.9
8.9
15.3
5.3
23.9
21.1
22.4
7.6
17.6
11.1
86.7
83.9
79.9
92.5
86.4
73.1
78.8
68.3
91.8
59.5
58.9
58.1
86.4
67.5
84.9
7.5
7.0
11.1
5.2
8.1
10.7
12.2
15.7
2.8
16.5
18.7
19.3
6.1
14.3
4.0
0.1
0.3
0.4
0.0
0.1
0.3
0.1
0.7
0.1
0.1
1.4
0.3
0.0
0.6
0.0
1.7
1.6
1.9
1.6
1.5
1.4
1.8
1.8
1.3
1.3
1.5
1.4
1.5
1.5
1.3
90.4
93.8
85.2
91.8
89.0
93.7
84.1
81.9
99.5
90.6
79.6
83.0
95.5
87.2
103.0
*Score = 2x%active match + %inactive match - %inconclusive – 2x%mismatch
20
NIEHS-NCATS-UNC TOXICOGENETICS PROJECT:
qHTS for Cytotoxicity in a Population-Based in vitro Model
Population-wide study design:
Goal: use crowdsourcing to better predict the toxicity of chemicals
1086
1. Use the biological
data (SNPs, basal gene expression) to develop a model that
cell
lines
accurately
predicts individual responses to compound exposure
2. Use the intrinsic chemical properties to develop a model that accurately predicts
how a particular population will respond to certain types of chemicals
3. for information, go to https://www.synapse.org/#!Synapse:syn1761567
• To understand how genetic variation affects individual response to common
environmental and pharmaceutical chemicals
• The largest ever population-based ex-vivo cytotoxicity study
–
–
–
–
–
1086 cell lines
179 common, pharmaceutical, or important environmental chemicals (9 duplicates)
8 concentrations (0.33 nM – 92 M)
1-3 plate replicates
~2,400,000 data points + 2-5x106 SNPs
21
Tox21 Phase II Advantages and Limitations
• Extent of pathway coverage
• Focus on the use of reporter gene assays using immortal
cell lines
• Focus on single compounds
• Limited capability for xenobiotic metabolism
• Focus on simple biological systems
• Limited to acute exposure scenarios
• Limited availability of “big” data analysis tools
• Limited availability of high quality human toxicological data
22
Tox21 Phase III – Improving on Biological
Coverage and Relevance (2013 - ?)
•
•
•
•
•
•
Goal to develop more physiologically-relevant
HepaRG Cells
/predictive in vitro and lower organism
approaches (models/assays/data interpretation
strategies) to assess chemical toxicity potential
Incorporation of xenobiotic metabolism &
longer-term cumulative exposures into current
and new approaches
Increased use of in silico models and
quantitative extrapolation models
Integrated, data-rich assay approaches
capturing various molecular pathways &
cellular phenotypes
Mechanistic studies integrating to Adverse
Outcome Pathways (AOPs)
Expand collaborations and interactions
Spheroids/Organoids
Near-Term Targeted Assays
• High Content screening
-
Hoechst: Cell loss & nuclear size
DHE: Oxidative stress/ROS
p53: DNA damage
pH2A.X: Genotoxicity
JC-10: Mitochondrial damage (MMP)
Caspase 3: Apoptosis
Lipitox: Steatosis & Phospholipidosis
Reactive metabolites/ROS: GSH depletion
• Gene expression assays
-
~1000 genes, multiple species
23
In Silico Analysis of the 10k Library for Xenobiotic Metabolism
Identify Practicable Approaches for 10k chemicals
•
•
Metabolite Structure Predictions (human enzymes)
Extent of Metabolism Predictions (metabolic clearance)
Assess Predictivity of Approaches
•
•
•
Assess accuracy via known substrates
Do we predict true metabolites
Assess extent of metabolism predictions
Analyze 8,307 Chemical Structures (10k Library)
•
•
P450 & UGT Substrate Predictions
CLINT Predictions (combined via liver expression levels)
Predict Metabolite Structures
•
Assess accuracy of metabolite structure predictions
Assess Predicted Structures via QSAR models
•
•
•
Toxicity Predictions
ADME/Permeabilities
Physicochemical Properties
Rank/Prioritize Chemicals for Assays in
Metabolically Competent Model Systems
•
•
•
Predicted Toxicities via chemical structure alerts
Extent of metabolism predictions
Reactive Metabolites?
QSAR Predictions
Physicochemical
•LogP, Lipinski Ruleof5, pH, solubility(water,
salt,GI),blood/plasma ratio
Human Metabolism
•Substrate, inhibitor, CLINT, SOM, Metabolite
Structures,
ADME
•Gut absorption & permeability, skin
permeability, cornea permeability, blood-brain
barrier permeability, uptake transport, efflux
transport,
Toxicity
•22 QSAR model toxicity predictions
•Human, Rat, Mouse, Aquatic
•LD50, Mutagenicity, ALT, AST, Alk. Phos.,
LDH, phospholipidosis, reproductive,
ER
binding, AR binding
24
10k Library Xenobiotic Metabolism Predictions
Substrate
Calls
Extent of Metabolism Predictions
• CLINT predicted for 5 individual P450s
• Combined CLINT from 5 Enzymes in
ToxPi
− Weighted each enzyme CLINT contribution
by relative P450 expression levels
#4
• 168,805 unique
metabolites predicted
• Evaluating ability to predict
known metabolites in library
• Assessing 10k library &
169k metabolites with
various toxicity prediction
models
25
High Throughput Transcriptomics Workshop
On Gene Prioritization Criteria
September 16-17, 2013
National Institute of Environmental Health Sciences
29 July 2013 Federal Register Request for Information
•
•
The nomination and prioritization of ~1000 environmentally
responsive genes per species for use in screening large numbers of
substances in cells or tissues from human, rat, mouse, zebrafish,
and C. elegans, using high throughput toxicogenomic technologies.
Recommendations on criteria to use for prioritizing the genes that
potentially would be the most useful in a screening paradigm, with a
focus on effects that reflect general cellular responses, independent
of cell type, and gene expression changes that are specific by cell
type.
26
The NCATS BioPlanet: the universe of biological pathways
for assay selection and prioritization
• Hosts the universe of human
pathways (~1600 unique)
• All pathway annotations from
manually curated, public sources
(e.g., KEGG, WikiPathways,
Reactome, Science Signaling)
• Integrates pathways from >10
different data sources
• Annotates pathways by source, biological function/process,
disease/toxicity relevance, assay availability
• Easy visualization, browsing, analysis of pathways
• Facilitates pathway assay selection/prioritization for Tox21 production
phase
• Web version in process for public release
From Ruili Huang, NCATS
27
Alternative Organisms – C. elegans and Zebrafish
C. elegans (NIEHS/NTP)
• Screened ToxCast Phase II compounds in growth
assay
• Screening subsets of compounds in assays that
measure
-
feeding
-
larval lethality
-
reproduction
Zebrafish – R. Tanguay (Sinnhuber Aquatic Research
Laboratory, Oregon State University, Corvallis, OR)
•Screened ToxCast Phase II compounds
•Screening 3455 NTP compounds at ~ 64 μM
•Assays include
-
1 day photo induced behavior
-
1 day assessment of mortality/developmental progression
-
5 day photo motor response
-
5 day assessment of 20 morphological endpoints
5 days
28
Human Stem Cell Related Projects
•
•
•
•
•
Collaboration with Cellular Dynamics and Molecular Devices to screen 80 compounds
(focus on neurotoxicants, cardiotoxicants, mitochondrial toxicants) in:
–
Neurite outgrowth assay/mitochondrial membrane potential
–
Beating cardiomyocyte assay/mitochondrial membrane potential
Collaboration with QPS, PhoenixSongs Biologicals, & the Hamner Institutes to evaluate
biological activity of the 80-compound library in various human and rat neuronal cell
culture systems (e.g., primary, embryonic stem cell–derived, induced pluripotent stem cellderived, transformed neural cell lines)
Collaboration with XCell to characterize response of iPSC-derived neural populations
(e.g., dopaminergic) from Parkinson’s disease (familial & sporadic) using the 80compound library
Collaboration with the Buck Institution to screen the 80-compound library in hTERT
astrocytic cell lines to identify senescence-inducing agents and verify results from the
screen in primary human astrocytic cultures.
Collaboration with the Univ. Konstanz to screen the 80-compound library in assays that
evaluate migration of neural crest cells/neurite outgrowth in a human cell line.
29
NTP-CDI-Molecular Devices Stem Cell Project
30
In Vitro Neurite Outgrowth Assay
Cell Body
Processes
(4)
connected to
cell body
Branch
31
Tox21 Phase III
• Greatly increased pathway coverage
• Focus on high content screening using normal
human cells
• Focus is still on single compounds but increased
interest in mixtures
• Expanded capability for xenobiotic metabolism
• Focus on more complex biological systems
• Potential for more extended exposure scenarios
• Expanded availability of “big” data analysis tools
32
BioActivity Network for
Prioritization
Data
+
=
+
33
NICEATM Uterotropic Network Paint
Uterotropic Actives
Uterotropic Inactives
34
Filter for 1st Neighbors of Uterotropic+
Utero Actives
Utero Inactives
1st Neighbors
35
Tox21/ToxCast Data Release Activities
• Tox21 10K library data - Phase II data released in PubChem
(86 entries to date at
https://www.ncbi.nlm.nih.gov/pcassay?term=tox21)
• On Dec 17, EPA publicly released ToxCast Phase II highthroughput screening data at
http://www.epa.gov/ncct/toxcast/data.html
- Data on 1,800 chemicals, screened in 700+ assays
- Accessible via Chemical Safety for Sustainability Dashboards
(http://actor.epa.gov/dashboard/) and raw data files
- Release will include announcement of data challenges through
TopCoder and InnoCentive to figure out how to use new data to
predict chemical no effect levels. Winners will receive monetary prizes
(http://epa.gov/ncct/challenges.html
- EPA is hosting stakeholder workshops (information at:
http://epa.gov/ncct/workshops.html
36
Success depends on
• Robust scientific collaborations
• Well-characterized chemical libraries
• Well-characterized assays in terms of reliability and relevance,
with broad biological coverage
• Incorporating xenobiotic metabolism into in vitro assays
• Informatic pipelines/tools that integrate and mine diverse data
streams
• Understanding the relationships between pathways and disease
in humans and animal models
• Making all data public
• Outreach to the scientific community on the usefulness and
limitations of Tox21 data
37
What will success bring?
•
Test methods for toxicity testing that are
scientifically sound and more economically efficient
•
An increased ability to evaluate the large numbers
of chemicals that currently lack adequate
toxicological evaluation
•
Models for risk assessment that are more
mechanistically based
•
Reduction and/or replacement of animals in
regulatory testing
38