Talk about the evolution of toxicity testing given at the Institute of Environmental Toxicology, Ibaraki, Japan, 26 July 2018

1. THE EVOLUTION OF TOXICOLOGY:
FROM LD50S TO GENOMICS (AND BEYOND)
Institute of Environmental Toxicology
Ibaraki, Japan
26 July 2018
Kevin M. Crofton, PhD
R3Fellows, LLC

2. OUTLINE
2
What’s the Problem?
• Too many chemicals with no data
Who’s your boss?
• Regulatory drivers
Data, Data, Who’s Got What Data?
Nothing it ever perfect
• Turning Criticisms into Challenges

3. PROBLEM: THE CHEMICAL EXPOSURE UNIVERSE
3
TSCA Inventory: ~84,000
P.P. Egeghy et al. Sci Total Environ. 414 (2012) 159–166

4. PROBLEM: THE CHEMICAL HAZARD UNIVERSE
4
Category
Size of
Category
Estimate Mean Percent
In the Select Universe
Pesticides and Inert
Ingredients of Pesticides
Formulations
Cosmetic Ingredients
Drugs and Excipients
Used in Drug Formulations
Food Additives
Chemicals in Commerce:
At Least 1 Million
Pounds/Year
Chemicals in Commerce:
Less than 1 Million
Pounds/Year
Chemicals in Commerce:
Production Unknown or
Inaccessible
Complete
Health
Hazard
Assessment
Possible
Partial
Health
Hazard
Assessment
Possible
Minimal
Toxicity
Information
Available
Some
Toxicity
Information
Available
(But Below Minimal)
No Toxicity
Information
Available
3,350
3,410
1,815
8,627
12,860
13,911
21,752
10 8 82
12 12 76
11 11 78
5 14 1 34 46
18 18 3 36 25
2 14 10 18 56
10 24 2 26 38
US National Research Council, 1984
 MAJOR CHALLENGE IS TOO MANY CHEMICALS AND
NOT ENOUGH DATA
 TOTAL # CHEMICALS = 65,725
 CHEMICALS WITH NO TOXICITY DATA OF ANY KIND =
~46,000
 CURRENTLY BETTER BUT NOT GOOD ENOUGH
Crofton et al Congenit Anom 52;140, 2012

5. Who’s Your Boss? Matching Data Type and
Uncertainties to Decision Context
It is critical to understand the
uncertainties in the data
And match them to the
changing regulatory
decision contexts

6. 66
EPA
Office
Assessment
“Workflows”
Historical
Throughput
Data
Types
OPPTS Premanufacture Notice (PMN)
New chemicals
Significant New Use Rule (SNUR)
Existing chemicals
~1000/yr
90d/chem
~84,000 total
III (II)
Current Chemical Risk (new program) ~10 total I
DFE / Green Chemistry ~2500 I, II, III
OSCP Endocrine Screening Program ~10-20/year
OPP Pesticide registration (PR) ~10 new/yr
~50 old/yr
I
Pesticide re-registration ~1000/yr
24,576 total
I
OW Chemical Contaminant List 6yr
~6,000 total
I,II,III
Regulatory Actions on CCL 6yr
90 total
I
Unregulated Contaminant Monitoring 30/5yr I
Drinking Water Health Advisories (MCLs) ~80 total II, III
ORD
NCEA
IRIS ~3/yr
~540 total
I
PPRTV 400-500 II,III
I. Data rich – Extensive guideline studies
II. Data partial – Some acute in vivo and in vitro data, SAR and exposure modeling
III. Data minimal to none – only chemical structure, SAR and exposure modeling
Slide courtesy of I. Shah
~1000/year
90 days/chemical
Very limited data.
eg structure, LogP
~10 chemicals/year
Lots and lots of data
$ millions/chemical

7. THE EVOLUTION OF MATCHING HAZARD DATA
UNCERTAINTIES TO REGULATORY NEEDS
QSAR and Read
Across
EDSP Prioritization
WOE
Pesticides
Human Perchlorate
Risk Management
Examples
(Crofton, 2018)

8. What’s Necessary to Begin Solving the Problem of Too Many Chemicals With No
Empirical Data on Exposure or Hazard Information
1. Chemical curation
• Everything starts with chemical structure
2. Prediction of hazard (or bioactivity)
• Need fast efficient testing methods
3. Predictions of exposure
• Need new models that predict or measure
exposures
4. Putting it all together
• Models that integrate this into estimates of risk
• Tools that can be used by risk managers
The Beginnings of a Solution

9. THE OLD WAY TO FIND CHEMISTRY STRUCTURES AND
PROPERTIES – LIBRARIES AND BOOKS!

10. The New Way –
Public Access to Chemical Structures and Properties via the Web

11. EVEN IN CHEMISTRY THERE IS UNCERTAINTY
DSSTox_v2
PublicCurated
5. Low
2. Low
6. Untrusted
1. High
3. High
4. Med
7. Incomplete
validated
Public_Untrusted
Public_Low
Public_Medium
Public_High
DSSTox_Low
DSSTox_High 4535
16K
33K
101K
584K
~ 310K pending
~ 150K pending
Chemicals Information Ranked on “Confidence”
Richard et al. Chem Res Toxicol. 29:1225-1251, 2016

12. GENERATING HAZARD DATA FOR CHEMICALS
• In the beginning there
was Paracelsus
• Then there was the LD50,
followed by acute tox
testing
• In the 80s, and to this
day, it is guideline
testing…
• BUT it is too costly and
too time consuming to do
this for all chemicals……
*Slide courtesy of Rusty Thomas

13. ToxCast and Tox21 Programs
• ToxCast – EPA program
–Multi-year research program started in 2007
–Use automated in vitro chemical screening technologies to expose living cells or
isolated proteins to chemicals where changes in biological activity may suggest
potential toxic effects
–Chemical library
• ~3400 environmentally relevant chemicals
http://www.epa.gov/ncct/toxcast/
• Tox21 – Collaborative project
• US EPA, NIH/NCATS, NIH/NIEHS/NTP and FDA
–aimed at developing better toxicity assessment methods using HTS.
–Chemical library
• ~8,500 chemicals, including environmental chemicals, food additives and
pharmaceuticals http://www.ncats.nih.gov/research/reengineering/tox21/tox21.html13
Generating Bioactivity Data for Lots of ChemicalsGenerating Bioactivity Data for Lots of Chemicals

14. 14
INCREASED THROUGHPUT REQUIRED SHIFT TO HTP IN
VITRO MOLECULAR TARGET APPROACHES
INCREASED THROUGHPUT REQUIRED SHIFT TO HTP IN
VITRO MOLECULAR TARGET APPROACHES
ToxCast
~600 Cell &
biochemical
assays
~1,000
Chemicals
Tox21
~30 Cell &
biochemical
assays
~8,000
Chemicals
Courtesy of Rusty Thomas

15. 15
Tox21 & ToxCast
Chemicals, Data and Release Timelines
Set Chemicals Assays Endpoints Completion Available
ToxCast Phase I 293 ~600 ~700 2011 Now
ToxCast Phase II 767 ~600 ~700 03/2013 Now
ToxCast E1K 800 ~50 ~120 03/2013 Now
Tox21 ~8300 ~80 ~150 Ongoing Ongoing
ToxCast Phase III + ~900 ~300 ~300 In Progress Fall 2018?
~72 million data points
~3 million conc response curves

16. EXPOSURE SCIENCE EVOLUTION
FROM SINGLE CHEMICAL ANALYTICS TO
HTP EXPOSURE PREDICTIONS
16
• For years exposure science has lagged behind
• Most models require extensive information on production, use, fate and transport
and rely on empirical data (no measurement = no exposure?)
• If you tell an analytical chemist you are interested in Chemical X – they will design the
methods needed to analyze for Chemical X…..
• Recent Breakthroughs
• HTP Exposure Predictions for thousands of chemicals
• Exposure predictions based on:
• pChem, production values, fate and transport, and product use categories
(e.g., industrial, pesticide use, consumer personal care)
• Industrial vs consumer use
• Yields predictions of exposure estimates and Baysian confidence

17. 17
EXPOSURE PREDICTIONS FOR 7968 CHEMICALS &
COMPARISON TO NHANES
EXPOSURE PREDICTIONS FOR 7968 CHEMICALS &
COMPARISON TO NHANES
• NHANES – US NATIONAL STUDY – MEASURES EXPOSURES IN HUMAN SERUM AND URINE
• CHEMICALS CURRENTLY MONITORED BY NHANES ARE DISTRIBUTED THROUGHPUT THE PREDICTIONS
• SHOWS ACCURACY OF THE PREDICTION MODEL
Wambaugh et al Environ. Sci. Technol., 2014, 48 (21), pp 12760–12767
NHANES
LoD
Median
Upper
95%tile

18. 18
THE NEAR FUTURE
NON-TARGETED CHEMICAL SCREENING
THE NEAR FUTURE
NON-TARGETED CHEMICAL SCREENING
“One group of tools, known as "non-targeted analysis" (NTA) methods, allows the
rapid characterization of thousands of never-before-studied compounds in a wide
variety of environmental, residential, and biological media” Sobus et al (2017)*
• Humans are exposed to an Exposome of potentially thousands of chemicals
• Methods used allow for detection of hundreds of suspect chemicals (e.g. include
hybrid quadraple-time-of-flight mass spectrometry)
• New computational analyses allow for ranking of tentative candidates based on
retention and spectral matching
NTA is the 96-well plate for analytical chemists
*J Expo Sci Environ Epidemiol. 2017 Dec 29. doi: 10.1038/s41370-017-0012-y. [Epub]

19. 19
Risk is the product of hazard
and exposure
Use rTK to convert bioactive
concentrations to daily dose
Combine with exposure
prediction
Judson et al., (2011) Chemical Research in Toxicology
Potential Exposure from
ExpoCast
mg/kg BW/day
Potential Hazard from
ToxCast with Reverse
Toxicokinetics
Low
Risk
Med
Risk
High
Risk
Putting It All Together - HT Risk Prioritization

20. 20
Bioactivity
Converted to
mg/kg/day
with HTTK
(Wetmore et
al., 2012)
Exposure
Predictions
(Wambaugh et
al., 2014)
ToxCast Chemicals
COMBINING BIOACTIVITY-BASE DOSE AND EXPOSURE
EXAMPLE - ESTROGEN ACTIVE CHEMICALS
COMBINING BIOACTIVITY-BASE DOSE AND EXPOSURE
EXAMPLE - ESTROGEN ACTIVE CHEMICALS
Higher Priority for
Further Testing
Prioritization = test the chemicals that might be the worst, first! Slide courtesy of John Wambaugh

21. 21
TURNING CRITICISMS INTO CHALLENGESTURNING CRITICISMS INTO CHALLENGES
21
• There is a lack of biological coverage of in vitro HTP assays
• There is no metabolism in your in vitro assays
• In vitro assays are not normal biology
• You can’t test my favorite chemicals because of limitations in your methods (e.g.,
solvents, high LogP)
• You can’t possibly predict kinetics with simple rTK models!
• You can’t do HT Exposure predictions based on simple use models!
• If not this, then what? What other methods could we use for prioritization of
thousands of chemicals?
Updated version of slide from Robert Kavlock

22. 22
CHALLENGES: INCREASING BIOLOGICAL COVERAGECHALLENGES: INCREASING BIOLOGICAL COVERAGE
Whole Transcriptome
Technologies
Diverse Cell Types
& Culture Systems
Gene coverage in
ToxCast/Tox21 is low = ~300
Ongoing pilot projects using cost-
efficient whole gene sequencing
platforms
Goal: use multiple cell types to
cover entire genome
Modified from Slide from Rusty Thomas

23. 23
EFFORTS TO EXPAND BIOLOGICAL COVERAGE
GLOBAL TRANSCRIPTOMICS
EFFORTS TO EXPAND BIOLOGICAL COVERAGE
GLOBAL TRANSCRIPTOMICS
• What if we could measure the changes in every gene in a cell, or organ or individual?
• 2006 - Broad Institute’s L1000 project
• Early work to predict drug impacts on all human genes using ~1000 genes to map all pathways
in MCF7 cells based on Connectivity Map (Lamb et al. Science 2006. 313:1929–35)
• 2017 - National Toxicology Project S1500+
• Currently exploring the use of TempOSeq for about 2000 genes for use in short term rodent
studies (Mav et al. PLoS One. 2018 Feb 20;13(2):e0191105)
• Ongoing - US EPA Global Transcriptomics Project
• Using BioSpyder’s TempOSeq for the entire human genome
• Starting with one cell type MCF7
• Challenge is to determine how many cell types are needed to adequately cover biological
space.

24. 24
CHALLENGES: RETROFITTING ASSAYS WITH METABOLIC ACTIVITYCHALLENGES: RETROFITTING ASSAYS WITH METABOLIC ACTIVITY
DeGroot and Simmons, Unpublished
Cyp activity over time
of encapsulated S9
fraction
MSTI Assay - An increase in electrophilicity was detected as a decrease in the
fluorescent signal.
S9 encapsulated in
a alginate bead

25. 25
• Public release policy of from Tox21 and ToxCast Programs
(raw & processed data, all publications, all processing and
modeling code)
• Transparent ToxCast data analysis pipeline
• Data quality flags to indicate concerns with chemical purity and
identity, noisy data, and systematic assay errors
• Publicly available as an R package
• Tox21 and ToxCast chemical libraries have undergone
analytical QC and results publicly available
• Public posting of all procedures
• Chemical Procurement and QC
• Data Analysis
• Assay Characteristics and Performance
• Recently completed external audit on ToxCast data and data
analysis pipeline
Challenges: Regulatory Applications Require Transparency
(example: EPA)

26. HOW DO WE SPEED UP ACCEPTANCE AND USE OF NEW METHODS
AND MODELS FOR REGULATORY USE
• Tech changes are coming rapidly and science does not wait around for regulators
• Time for acceptance of new methods (i.e., tipping point) can be long
Cheap global
transcriptomics
Non-Targeted
ID of ‘true’
exposures
Organs on
Chips
Tox21
ToxCast
Goal
Rapid & cost efficient
in silco predictions that
improve public and
environmental health
Time
~30 years*
* From launch of 1984 NRC Report to the start of Tox21 Project

27. 27
NEXT PHASE… EVOLUTION TOWARDS A TRULY
PREDICTIVE SCIENCE
NEXT PHASE… EVOLUTION TOWARDS A TRULY
PREDICTIVE SCIENCE
c. 1970s - 2004 2005 - 2007 2007 - 2010 2011 – 2018+c.1500 – 1960s 2018 - ?
New ToolsThe NAS
Challenges
Tox and
Exposure
Birth of Early
Computational
Toxicology
Efforts
Classic
Toxicology
The Dose
Makes the
Poison
Computationally
Predicted
Risk
Courtesy of Rusty Thomas

28. 28
THANK YOU FOR LISTENINGTHANK YOU FOR LISTENING