Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
Three open access computational tools were used on pesticides with well-characterized toxicological profiles to determine concordance between modeled predictions and measured acute in vivo toxicity (most→least: chlorpyrifos-oxon/chlorpyrifos (CPFO/CPF), carbaryl (CB), endosulfan (ENDO), propyzamide (PZ)). Tools included 1) Toxicity Forecaster (ToxCast) and Tox21 (ToxCast/Tox21) assays (AC50s µM) to identify biological targets associated with modes of action; 2) Toxicological Prioritization Index (ToxPi) incorporating AC50s to rank and score toxicity; 3) Integrated Chemical Environment (ICE) tool for in vitro to in vivo extrapolation one-, three-, and multi-compartment models to predict human equivalent administered dose (EADHuman). ToxPi toxicity ranking was predictive for CPFO, CB and PZ but not for CPF or ENDO. Qualitative graphical visualizations and quantitative fold differences between EADHumans and acute oral in vivo endpoints for each pesticide were most predictive in the three-compartment model. Qualitative modeled toxicity rank among AChE inhibitors (CPFO>CPF>CB) was 100% predictive. Brain AChE inhibition endpoints (EADHuman vs. ADJBMD10) had better predictions (lower fold difference) than those with RBC AChE inhibition endpoints. Overall, the computational tools in this study could be useful in not only prioritizing pesticides for risk assessment but also providing insights into mechanistic data often lacking in traditional testing.
Linde Gas whitepaper 'Drug discovery advances inextricably linked to specialt...
Similar to Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
Marike Kolossa-Gehring: The Chemical Challenge - How HBM4EU Can HelpTHL
Similar to Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study (20)
Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
2. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
There are hundreds of thousand chemicals, including
pesticides in the environment, many with unknown toxicity.
The United States Environmental Protection Agency (US
EPA) and international agencies are committed to
developing rapid and cost-effective computational tools to
support risk assessment efforts for risk management
applications (dos Santos and Nardocci, 2019; TSCA,
2020). In order to accomplish these goals, the
Computational Toxicology Chemistry Dashboard
(CompTox) provides a database repository that includes
the Toxicity Forecaster (ToxCast) and Tox21
(ToxCast/Tox21) in vitro bioassays (US EPA, 2020a).
There are almost 1600 unique assays for use in chemical
screening, prioritization, and toxicity predictions through
identification of potential disruptions in key biological
pathways (Pearce et al., 2017; Williams et al., 2017;
Zurlinden et al., 2020). The CompTox Dashboard allows
easy access to ToxCast/Tox21 assay results for over
882,000 compounds obtained from numerous databases.
These tools may, in some cases, replace the traditional
methods for assessment of toxicity that rely on expensive,
lengthy, and ethically challenging animal studies (Sipes et
al., 2017; Turley et al., 2019; Zurlinden et al., 2020). In light
of this goal, the US EPA has directed the Office of
Chemical Safety and Pollution Prevention and the Office
of Research and Development to “reduce requests for, and
funding of mammal studies by 30% by 2025, and eliminate
all mammal study requests by 2035,” (US EPA, 2020c).
With the motivation in mind of moving to in vitro models,
we conducted a case study examining whether the well-
established acute oral animal bioassay endpoints for five
pesticides coincide with results from three open access
computational toxicology tools. To accomplish this we
selected pesticides from four different classifications:
chlorpyrifos (CPF: organophosphate insecticide) and the
primary CPF toxic metabolite CPF oxon (CPFO),
endosulfan (ENDO: organochlorine), carbaryl (CB: N-
methyl carbamate insecticide) and propyzamide (PZ:
amide herbicide) (CDPR, 2008, 2014, 2018; US EPA,
2007a, 2011, 2015c). Each of these compounds, has been
shown to be highly toxic to agricultural workers, household
residents and bystanders living near application sites
(CDPR, 2008, 2014, 2018; US EPA, 2007b, 2014a, 2015d,
2017a). Data from the open access human health risk
assessments determined the no- and lowest-observed-
effect levels that served as a basis for comparison to in
vitro predictions.
The open access tools for this study included the
ToxCast/Tox21 Dashboard (US EPA, 2020a), the
Toxicological Prioritization Index Graphical User Interface
Program (Marvel et al., 2018; Reif et al., 2013; ToxPi,
2020) and the Integrated Chemical Environment tool (Bell
et al., 2020; NTP, 2020). The tools were used with each
pesticide for their utility in predicting: 1) biological activities
of ToxCast/Tox21 assays at known targets; 2) ToxPi-
generated chemical ranking and Toxicity Score based on
relative activities in ToxCast/Tox21 intended target
families; 3) human equivalent administered dose
(EADHuman) determined from three in vitro to in vivo
extrapolation (IVIVE) and physiologically based
toxicokinetic (PBTK) models in the ICE program (Bell et
al., 2020; ICE, 2020) to compare with the established in
vivo acute oral lowest effect levels. Concordance between
predicted results from computational tools and measured
in vivo bioassay results were examined for use in chemical
screening and prioritization and in support of open access
data associated with the presumptive mechanistic
pathways.
METHODOLOGY
In Vivo Acute Oral No Effect and Lowest Effect Levels
Used in Risk Assessment
In vivo acute oral studies for pesticides performed under
Health Effects Test Guidelines were designed to generate
a dose-response, that included a no observed effect level
(NOEL) and lowest observed effect level (LOEL) (US EPA,
1998, 2014b). These studies have traditionally been
conducted using doses at wide intervals, which may not
accurately estimate a no effect level. However, when data
can be modeled, a Benchmark Dose analysis may better
represent the lowest effect (BMD) or an estimated no effect
level (BMDL) than LOEL/NOELs conventionally used in
risk assessment for non-cancer endpoints (US EPA,
2012). A Benchmark Response (BMR) of 10%
(BMD10/BMDL10) was included in the open access oral
acute data for CPFO, CPF and CB, as an acceptable
normal range of variability for red blood cell (RBC) and
brain acetylcholinesterase inhibition (brain AChE) (CDPR,
2014; Moser et al., 2010; US EPA, 2011, 2017a). The
open access ENDO risk assessment used an acute oral
NOEL as an endpoint for neurotoxicity in pregnant rabbits
(CDPR, 2008). In some cases where a NOEL is not
obtained and/or BMD analysis cannot be modeled, an
estimated no-effect-level (ENEL) can be obtained by use
of a 10-fold uncertainty factor to extrapolate from the LOEL
(LOEL ÷ 10 = ENEL) (US EPA, 2015c).
The open access acute oral no effect and lowest effect
levels for each pesticide were the most sensitive for that
exposure duration (Table 1). This timeframe was also
important because the ToxCast/Tox21 assays are
performed after a single (acute) administration of
chemical. Each in vivo acute low and no effect points of
departure (POD) was reported as a LOEL/NOEL, a
BMD10/BMDL10 or LOEL/ENEL. The toxicity rank of each
pesticide in Table 1 was from most toxic (CPFO) to least
toxic (PZ) based on the listed PODs. These values were
used to compare Toxicity Score/Rank predicted in the
ToxPi program through use of selected ToxCast/Tox21
active hit-calls as data inputs.
For the Integrated Chemical Environment tool (ICE)
models (described below), the acute BMD10/LOELs in
3. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
Table 1 were further adjusted by a default factor of 10
(acute ADJBMD10/ADJLOEL) to account for animal to
human interspecies extrapolation (US EPA, 2002b; WHO,
2017). That is, the acute in vivo BMD10/LOELs were
obtained from animal studies but the ToxCast/Tox21 data
were performed with human tissues/proteins and the ICE
models predict human EADs (Bell et al., 2020; ICE, 2020;
NTP, 2020). Acute ADJBMD10/ADJLOEL are the relevant
data to compare to acute EADHuman generated from IVIVE
models, such as those used in ICE as has been shown
previously (Sipes et al., 2017).
Chlorpyrifos (CPF) and Chlorpyrifos oxon (CPFO): In
the United States and internationally, CPF is widely
registered for use because it is highly effective as an
insecticide, miticide and acaricide for application on food
crops (e.g., nuts, grains, fruit, vegetables, and grains) in
addition to non-food use (e.g., golf course turf, nurseries,
wood products) (CDPR, 2017; PMRA, 2019; US EPA,
2020b). Residues on food, in water and air presented
human health safety concerns and there are increasing
reports indicating that CPF exposure at low doses leads to
irreversible neurodevelopmental effects in humans and
rodents (APVMA, 2017; EFSA, 2014, 2017; Rauh et al.,
2015; Rauh et al., 2012; Silva, 2020). The main acute CPF
MOA is through activation by CYP2B6 (desulfuration) to
the toxic metabolite CPFO to inhibit plasma, or butyryl
cholinesterase (BuChE), red blood cell
acetylcholinesterase (RBC AChE) and brain AChE
resulting in neurotoxicity in humans and animals (Eaton et
al., 2008a; Testai et al., 2010). CPFO is detoxified by
paraoxonase 1 (PON-1/A esterase) to form the main
urinary metabolite 3,5,6-trichlorpyridinol (TCPy) (Eaton et
al., 2008b). A hydrolysis path involving CYP2C19
(pregnane-X-receptor: PXR-associated) and CYP3A4/5
directly detoxifies CPF to TCPy, followed by conjugation
with glucuronide, sulfate, or glutathione-s-transferase.
CYP3A is activated through PXR + retinoid-X-receptor
(RXR) heterodimerization and interaction with the PXR-
Element (PXRE) promotor region in the nuclear DNA
(Wang and Negishi, 2003). Reversal of AChE inhibition by
CPFO in humans occurs over 30 days (US EPA, 2016) and
takes 2-5 weeks in rats (Ellison et al., 2011). Other CYPs
involved with CPF dearylation/dealkylation include
CYP1A2, 2B6, 3A4, 3A5, 3A7, 1A1, 2C19, 2C9, and 2D6
(Buratti et al., 2003; Mutch et al., 2004; Tang et al., 2001).
Data indicated that RBC AChE inhibition is more sensitive
than in brain for CPF and CPFO (Table 1). However,
effects on each endpoint may lead to adverse outcomes
during development as has been proposed (Mattsson et
al., 2000; Moser and Padilla, 1998).
Carbaryl (CB): CB is broadly applied on food (e.g., nuts,
fruits, vegetables, grains, oyster beds) and non-food (e.g.,
turf, golf courses, ornamentals, cut flowers) crops in
residential and agricultural settings in the United States
and Internationally (APVMA, 2012; EFSA, 2006; PMRA,
2016; US EPA, 2017a). In addition to acute neurotoxicity
effects to human health through residues in food, air and
water, CB is considered by the US EPA to be a “likely
human carcinogen,” (US EPA, 2007b). Further, the
benefits from killing mosquitoes carrying malaria, conflicts
with the toxic and deadly effects on honeybees (NPIC,
2003). The acute CB MOA is through inhibition of AChE in
plasma, brain and red blood cells (RBC AChE) by
carbamylation of the serine hydroxyl group at the active
site on the enzyme in humans and animals (Moser et al.,
2010; US EPA, 2017a). While CB itself can inhibit AChE,
CYP-activated metabolic pathways produce carbaryl
methylol (PXR-associated CYP2B6), 5-hydroxcarbaryl
(CYP1A1, CYP1A2-PXR-associated) and 4-
hydroxylcarbaryl (CYP1A1, PXR-associated CYP3A4)
which are also toxic (Nong et al., 2008; Tang et al., 2002;
US EPA, 2007a). Arylhydrocarbon hydroxylase nuclear
receptor (Ahr) regulates CYP1A1 expression (Fallone et
al. 2005) and the CYP3A, CYP1A and CYP2B families are
activated through CAR/RXR and PXR/RXR with PXRE in
the nuclear DNA (Ihunnah et al., 2011; Wang and Negishi,
2003).
Endosulfan (ENDO): ENDO agricultural use as an
insecticide and acaricide was completely phased out in the
United States (USEPA, 2010) from 2009-2016. It was
globally banned for all uses as of 2011 because it was
listed as a persistent organic pollutant by the Stockholm
Convention’s Persistent Organic Pollutants Review
Committee (UNSC, 2010). However, ENDO persistence in
the environment and accumulation in fatty tissues may
pose a risk to human health as well as wildlife (Cone, 2010;
EFSA, 2011). ENDO functions acutely as a neurotoxic
non-competitive inhibitor in the central nervous system
(CNS) by binding to and blocking Cl- channels associated
with the γ -amino-butyric acid receptor (GABAAR) (Casida,
1993; Ffrench-Constant et al., 2000; Lawrence and
Casida, 1984). GABAA receptors are the principal
inhibitory neuroreceptors in the mammalian brain and
antagonism of GABAergic neurons in the CNS causes
generalized brain stimulation (Cole and Casida, 1986;
Gant et al., 1987). Normally when GABA binds to its
receptor, the Cl- ion channels are opened, leading to an
influx of Cl- into neurons through an electrochemical
gradient. The result is hyperpolarization of the cell
membrane and inhibited neuron firing. However, ENDO
prevents Cl- from entering neurons, thus blocking the
effect of GABA binding to its GABAAR, resulting in
uncontrolled excitation (Kamijima and Casida, 2000; Ratra
et al., 2001). This primary effect is observed acutely in both
humans and animals (where clinical signs were recorded).
Metabolism occurs through activation by CYP2B6 and
CYP3A4/5 to form ENDO-sulfate (Bebe and
Panemangalore, 2003; Casabar et al., 2006). The main
metabolite, ENDO-sulfate, has about the same toxicity as
the parent technical mixture (ATSDR, 2000). Nuclear
receptors (e.g., constitutive androstane receptor (CAR),
peroxisome proliferation receptor alpha (PPARα), PXR)
heterodimerize with RXR and bind at the PXRE (Sugatani
et al., 2004) or at the peroxisome proliferator hormone
response element (PPRE) promotor regions on the gene
4. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
(Viswakarma et al., 2010; Wang and Negishi, 2003) to
initiate the expression of CYP2B6 and CYP3A4/5 and
other proteins involved in ENDO metabolism (Michalik et
al., 2006).
Effects on dopamine and dopamine active transporter
(DA/DAT) resulted from ENDO treatment in rodents during
gestation (in utero), neonatally and postnatally. Weanling
rats had a decrease in DA in the hippocampus after
treatment with ENDO and a lowered ability to learn and
retain a required task (Lakshmanan et al., 2013). ENDO
administered intraperitoneally to neonatal rats resulted in
decreased DA and increased foot-shock fighting behavior
(Seth et al., 1986). Male C57Bl/6j offspring (age 3 months)
treated in utero with ENDO showed a decrease in DAT
expression in the brain. (Wilson et al., 2014). DAT was
further depleted in these pups after treatment with the
neurotoxin MPTP (1-methyl-4-phenyl-1,2,3,6-
tetrahydropyridine) . C57Bl/6 mice treated with ENDO as
juveniles and again in adulthood had decreased DA (Jia
and Misra, 2007). ENDO is also an endocrine disruptor
that affects the androgen and estrogen receptor but
usually at doses equal or greater than those inducing
neurotoxicity (Milesi et al., 2015; Silva and Gammon,
2009).
The acute LOEL/NOEL for ENDO is 1.8/0.7 mg/kg/day
from a developmental toxicity study in rabbits was reported
by CDPR (2008) and, US EPA (2002a). Although a
developmental study in pregnant rabbits involves 22 days
of gavage dosing, neurotoxic effects (hyperactivity) were
observed within the first few days of treatment. This study
exhibited the most sensitive endpoint for ENDO (Table 1).
Propyzamide (PZ): PZ has non-food uses (e.g.,
landscape maintenance, ornamental turf, golf courses,
etc.), as well as food uses (e.g., lettuce, endive, artichokes,
and radicchio) and is registered in the United States (US
EPA, 2015b) and internationally (APVMA, 2020; EFSA,
2016; PMRA, 2009). Although PZ has comparatively low
toxicity, it is widely used on crops and the residues in food
and drinking water are a concern. There are also concerns
for PZ effects on the thyroid from exposure during
development (Marty et al., 2012; US EPA, 2017b). The PZ
mode of action (MOA) involves activation of nuclear
receptors and induction of liver enzymes as seen with
phenobarbital treatment in rodents (Braeuning et al., 2014;
Elcombe et al., 2014; LeBaron et al., 2014). Acute
exposure activates xenobiotic sensors (e.g., CAR, PPARα
and PXR) (Angrish et al., 2016; LeBaron et al., 2014) that
heterodimerize with RXR to bind at PXRE (Sugatani et al.,
2004) or PPRE (Viswakarma et al., 2010; Wang and
Negishi, 2003) to initiate expression of the CYPs and other
proteins involved in lipophilic xenobiotic metabolism
(Michalik et al., 2006). CYP4A10, CYP2B10 (CAR-
associated), CYP1A1 (AhR-associated nuclear receptor)
and CYP3A11 (PXR-associated) are the primary P450s
involved in PZ metabolism (LeBaron et al., 2014). An acute
oral gavage neurotoxicity study in adult rats showed an
increase in landing foot splay (female) and decrease in
motor activity (males and females) at the LOEL of 40
mg/kg/day (US EPA, 2015c). There is not currently an
MOA for acute PZ neurotoxicity and there were no other
studies in the database showing evidence of neurotoxicity
regardless of animal strain or treatment duration.
However, nuclear receptor and CYP induction would have
occurred to metabolize and eliminate PZ, since the effects
were reversed by observation day 2. An ENEL of 4.0
mg/kg/day, achieved by adding an uncertainty factor of 10
to the LOEL (40 mg/kg/day), was determined by the US
EPA to be the acute oral value for PZ (US EPA, 2015c)
(Table 1).
US EPA ToxCast/Tox21 from the CompTox Chemical
Dashboard
ToxCast/Tox21 results on the CompTox Dashboard
characterize chemical interactions with intended target
families and their biological molecular targets (e.g., cell
cycle, CYP [P450s], receptor-ligand binding,
steroidogenesis) were reported as concentration at 50%
maximum activity (AC50) (Judson et al., 2011). These
chemical interactions can provide support for presumptive
MOAs, AOPs and/or downstream toxicity (NAS, 2017).
ToxCast/Tox21 data are from in vitro and zebrafish (in
vivo) HTS assays from numerous vendors and platforms
(US EPA, 2020a). Complete descriptions of assays are
available in the Dashboard download for individual
chemicals (Supplemental Tables 1-5; Summarized in
Supplemental Table 6). Data for each assay are filtered
from level 0 (raw file processing: vendor-specific) through
6 (cautionary flags attached to detect potential false
negative/positive) in the ToxCast Pipeline (Filer et al.,
2017; Sipes et al., 2017; Williams et al., 2017). The
CompTox Dashboard also presents graphical depictions of
each active ToxCast/Tox21 concentration-response as
well as the selection of a winning dose-response model
and AC50 (Browne et al., 2015). Data characterizing the
winning model include notations related to cautionary flags
and the cytotoxicity limit (examples in Supplemental
Figure1) (Judson et al., 2016; Williams et al., 2017). The
cytotoxicity limit occurs in a narrow range of
concentrations where assay results may indicate cell
stress and cytotoxicity rather than compound-specific
activity (“cytotoxicity-associated burst”) (Judson et al.,
2016).
Biomolecular interactions in the ToxCast/Tox21 assays
were measured for several intended target families (e.g.,
DNA binding, nuclear receptor, steroid hormone assay,
etc.) and target enzymes or proteins (e.g., esterase:
acetylcholinesterase inhibition AChE; CYP: cytochrome
P450 CYP1A1; transferase: UDP-glucuronosyltransferase
UGT1A1). All active hit-calls for CPF/CPFO, CB, ENDO
and PZ were downloaded from the CompTox Dashboard
for selection prior to use in the ToxPi ranking analysis.
AC50s close to or greater than the cytotoxicity limit were
less reliable but may be relevant to secondary effects from
treatment (Judson et al., 2016).
5. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
Table 1: Acute In Vivo Endpoints for Chlorpyrifos-oxon, Chlorpyrifos, Endosulfan, Carbaryl, and Propyzamide
Animal Strain Treatment Doses (mg/kg/d) Acute Effect
Acute Endpoints
mg/kg/day
Adjusted
BMD10/LOELa
mg/kg/day
Refb
Chlorpyrifos-oxon
Pup: Sprague-
Dawley Rat
Gavage PND
11
(Tested PND
11)
0 (corn oil), 0.005,
0.01, 0.05, 0.1, 0.5
↓ RBC AChE
BMD10 0.093
BMDL10 0.05
ADJ BMD10
0.0093
1
↓ Brain AChE
BMD10 1.06
BMDL10 0.36
ADJ BMD10
0.106
Chlorpyrifos
Pup: Sprague-
Dawley Rat
Gavage PND
11
(Tested PND
11)
0 (corn oil), 0.005,
0.01, 0.05, 0.1, 0.5
↓ RBC AChE
BMD10 0.5
BMDL10 0.36
ADJ BMD10
0.05
1
↓ Brain AChE
BMD10 1.42
BMDL10 0.91
ADJ BMD10
0.142
Carbaryl
Pup: Male
Long-Evans Rat
Gavage, single
PND 11 (Tested
PND 11)
0 (corn oil), 3, 7.5,
15, 30
↓ RBC AChE
BMD10 1.11
BMDL10 0.78
ADJ BMD10
0.111
2
↓Brain AChE
BMD10 1.46
BMDL10 1.14
ADJ BMD10
0.146
Endosulfan
Pregnant New
Zealand White
Rabbit
Gavage GD 6-
28
0 (corn oil), 0.3, 0.7,
1.8
Neurotoxicity signs from
treatment day one
LOEL 1.8
NOEL 0.7
ADJ LOEL
0.18
3
Propyzamide
Adult:
F344/DuCrl Rat
Gavage, single
(corn oil)
40, 200, 600
↓ Motor activity (♂)
↑ Landing foot-splay (♀)
LOEL 40
ENEL 4.0c ADJ LOEL 4.0 4
Abbreviations: ADJ: adjusted; BMD10/BMDL10: Benchmark Dose Lowest effect level/presumptive no effect level (10% benchmark
response); ENEL: Estimated No Effect Level; GD: Gestation Day; LOEL: Lowest-observed-effect-level; NOEL: no-observed-effect-
level; PND: Postnatal Day
a
- Lowest effect levels (BMD10/LOEL) were divided by a default interspecies factor of 10 to account for animal to human (interspecies)
variability
b-
References: 1. US EPA (2011); 2. Moser et al. (2010), US EPA (2017a); 3. CDPR (2008), US EPA (2002a); 4. US EPA (2015c).
c-
Oral ENEL = 40 (LOEL) ÷ 10 (to extrapolate from a LOEL to an estimated no effect level) = 4.0 mg/kg NOEL was not achieved and the
data were not modeled by BMD analysis (US EPA, 2015c).
Toxicological Prioritization Index Graphical User
Interface (ToxPi GUI)
The ToxPi program (https://toxpi.org/ version 2.3,
accessed 8-2020) (Marvel et al., 2018; Reif et al., 2010;
Reif et al., 2013) was used to rank and prioritize chemicals
for application in risk assessment processes. In this study,
selected AC50s were used in the ToxPi program to
visualize relative toxicity and ranking among the 5
pesticides. Since not all chemicals had active hit-calls for
the same assays, in ToxPi version 2.3, data that are
missing or invalid are disregarded (slice score = 0), and
valid data would have a score between 0 and 1 (after slice
scale modeling) (Marvel et al., 2018). The length of each
ToxPi slice from the center of the circle is proportional to
the normalized potency of the assay value (-log10 (AC50) +
6) for the representative intended target family. This study
used equally weighted slices of the same width
irrespective of the number of measured AC50s. Predicted
toxicity ranking was based on intended target family
category and AC50 potency. ToxPi ranks were compared
to rank of in vivo endpoint values (Table 1) to determine
concordance between predicted and measured toxicity.
Assays included in the ToxPi analysis are described in
Supplemental Tables 2b-6b.
High Throughput Toxicokinetics
Selected AC50s (µM) with active hit-calls below the
cytotoxicity limit and relevant to the MOA were used in the
open access ICE program (Bell et al., 2020) to predict an
in vivo EADHuman (mg/kg/day). Figures 4-8 show the assay
selections for the ICE program input (supplemental Table
7). Human pharmacokinetic parameters (PK) were used in
the 1 (1C) and 3 (3C) compartment and physiologically
based toxicokinetic multicompartment (PBTK) models to
convert the AC50 values to EADHuman (Bell et al., 2020; ICE,
2020). The data inputs (selected AC50s, chemical
CASRN), along with reverse dosimetry (Tan et al., 2007)
generated the estimated EADHuman (mg/kg/day)
(Supplemental Table 7).
Qualitative ranking of the fold difference calculations
(Thomas et al., 2013) between the 1C, 3C or PBTK
EADHuman and in vivo endpoints were represented
graphically for each of the 5 chemicals by plotting in R
package “ggplot2” (Wickham, 2016). They were further
examined for RBC versus brain AChE inhibition adjusted
endpoints for CPFO, CPF and CB in the following manner:
1) Fold differences were determined between EADHuman
and ADJBMD10s based on the RBC AChE inhibition 2) The
same was done for fold differences between EADHuman
6. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
and ADJBMD10s based on the brain AChE inhibition; 3)
Final determinations indicated which endpoint, RBC or
brain AChE inhibition had the most predictive EADHuman
and which model best represented those results. AChE
inhibition is a biomarker for human exposure to
organophosphates and carbamates and is the key event
in the MOA for anticholinesterase compounds (US EPA,
2000). Inhibition initially occurs in blood and US EPA
Guidance Document on use of cholinesterase inhibition in
risk assessment indicated that RBC measurements serve
as a surrogate for brain AChE inhibition (i.e., inhibition of
RBC AChE presumes occurrence of brain AChE inhibition)
(Chen et al., 1999; Coban et al., 2016; US EPA, 2000). For
each anticholinesterase pesticide in this study the brain
BMD10 was higher than the RBC BMD10 for AChE
inhibition, which would be expected since the pesticides
initially enter the blood where RBC AChE is the first line of
defense (Table 1) (Coban et al., 2016; US EPA, 2000).
Quantitative fold differences were calculated for each
EADHuman in each model and were reported as 3 separate
categories: 1) Fold differences within one order of
magnitude of the in vivo ADJBMD10/ADJLOELs (<10-fold
difference) were predictive; 2) those with a 1-2 order of
magnitude (11-100-fold) difference were marginally
predictive and 3) those with greater than 2 orders of
magnitude (>101-fold) difference were not predictive.
One Compartment IVIVE Model: A single compartment
in a population-based PK model assumed 100% chemical
absorption (route not applicable; Bell et al. (2020)) and
treatment at 1 mg/kg/day (Wetmore et al., 2012).
Measured plasma concentration at steady state (Css µM)
is simulated at the 50th (median) and upper 95th percentile
by Monte Carlo distribution to incorporate interindividual
variability. Hepatic metabolism and passive glomerular
filtration of chemicals is the assumed form of elimination
(ICE, 2020). The predicted oral 1C EADHuman is related in
a linear fashion to the in vitro AC50 (µM) and inversely
related to the Css (µM) (Wetmore et al., 2012), hence the
upper 95th percentile Css was used as a conversion factor,
along with reverse dosimetry (Tan et al., 2007) to provide
a conservative 1C EADHuman (mg/kg/day; 50th and 95th
percentile Css data in Supplemental Table 9). The
EADHuman calculation for the steady state 1C model used
in this study corresponded to the total chemical
concentration administered (Casey et al., 2018; Wetmore
et al., 2012) according to Equation (1):
Steady-state EAD Human mg/kg/day =
AC50(µM) x (1 mg/kg/day ÷ 95th
percentile Css (µM))
Equation (1)
Three Compartment IVIVE Model: The 3C model has
perfusion rate-limited compartments (i.e., equilibrium is
achieved rapidly for tissue, RBC and plasma compared to
flow of blood) comprised of gut, liver, and rest-of-body
(e.g., fat, brain, bones). The “Solve_3comp” model is from
the open access high-throughput TK-R package that
calculates plasma concentration over time (httk v 1.9.2)
(Pearce et al., 2017). An oral dose of 1 mg/kg/day in an
acute interval (24-hr) was used to calculate maximum
human plasma concentration (Cmax µM) at the 50th
percentile using the average values for the PKTK
parameters. Selected inputs (AC50s and chemical
CASRN), Cmax and reverse dosimetry produced acute oral
EADs (mg/kg/day). Hepatic metabolism and passive
glomerular filtration of chemicals is the assumed form of
elimination (Bell et al., 2020; ICE, 2020).
PBTK Model: The “Solve_pbtk” multicompartment (gut,
artery, vein, lung, liver, kidney, rest-of-body) function
calculates Cmax at the 50th percentile using the average
values for the PKTK parameters over time (httk v 1.9.2;
Bell et al. (2020); Pearce et al. (2017)). Each compartment
is perfusion rate-limited and has mass balance differential
equations describing rate of change for quantity of
chemical. The model used median PBTK parameters to
calculate Cmax (µM) at an acute oral dose of 1 mg/kg/day
for 24 hours. Selected inputs (AC50s and chemical
CASRN), Cmax and reverse dosimetry produced acute
PBTK EADs (mg/kg/day).
Both the “Solve_3comp” and Solve_pbtk” models, used
the Cmax in the reverse dosimetry calculations as follows in
Equation (2) (ICE, 2020):
Acute Oral EAD Human (mg/kg/day) =
AC50 (µM ) x (1 mg/kg/day ÷ Cmax ((µM))(24hr exposure)
Equation (2)
Visual depictions of and calculations for each model are in
Pearce et al. (2017).
RESULTS
ToxCast/Tox21 Active Hit-Calls
The ToxCast/Tox21 data downloaded from the CompTox
Chemicals Dashboard (US EPA, 2020a) are summarized
in Table 2 (Data download: Supplemental Tables 2a-6a).
The pesticides in this study were tested in 854-980 assays
from 14 different vendors (Supplemental Table 1), but only
11-33% of the total assays had active hit-calls. Further,
most of the active hit-calls had cautionary flags (Filer et al.,
2017; Sipes et al., 2017; Williams et al., 2017). These
potential false positives were reported as noisy data (low
signal-to-noise ratio), low efficacy (i.e., capacity of a drug
to activate or inactivate a receptor), borderline active (only
at highest concentration above baseline), or confounding
by overfitting the data (Browne et al., 2015; US EPA,
2020a). Of the total active hit-calls 32 to 68% were not
selected for ToxPi because their activities were based on
background measurements, cell viability/toxicity (Judson
et al., 2016), were false positives or were otherwise
unassociated with MOAs or other toxicity pathways.
7. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
Table 2: Summary of ToxCast/Tox21 Assay Performance for Each Pesticidea
Parameter CPFO CPF ENDO CB PZ
Assays Performed 978 854 953 978 980
Number Active Hit-Calls (of Total Assays %) with
reported AC50s
217 (22%) 150 (17%) 311 (33%) 103 (11%) 101 (11%)
Unflagged Active Hit-Calls (of Total Actives %)b
46 (21%) 36 (24%) 89 (9.3%) 32 (31%) 43 (42%)
Cytotoxicity Limitc
11.72 µM 15.74 µM 12.31 µM 11.72 µM 11.05 µM
Actives Below the Cytotoxicity Limit (of Total Actives %) 53 (24%) 21 (14%) 45 (14%) 32 (31%) 38 (38%)
Active Hit-Calls Selected for ToxPid
49 41 49 40 34
Actives with Cautionary Flags Excluded from ToxPi 2 2 3 1 1
Number with AC50s below the Cytotoxicity Limit 23 4 24 16 18
Parameters Not Specific to the Primary Mode of Action/Associated Toxic Endpoints and/or Not Used in ToxPi
Group A: Cell viabilitye, f
64 40 125 24 15
Group B: Cytokine and cytokine receptorf
18 13 40 5 3
Group C: Cell Adhesionf
7 10 16 4 3
Group D: Miscellaneousf, g
19 10 30 21 11
Groups A-D: Total (of active hit-calls %) 108 73 211 54 32
Total Active Hit-Calls Excluded from ToxPih
(% Total Actives)
109 (50%) 74 (49%) 214 (68%) 55 (53%) 33 (32%)
a-
Data from the CompTox Dashboard (US EPA, 2020a)
b
-Assays with cautionary flags were possibly false positives.
c-
The Cytotoxicity Limit delineates a narrow concentration range where assay results may indicate cell stress and cytotoxicity
rather than compound-specific activity.
d-
Assays with cautionary flags indicating a false positive were excluded (Supplemental Tables 2b-6b)
e-
These assays include non-specific cell viability measurements “background measurement”, “cell cycle” and “cell
morphology”, “channel 1 and channel 2.
f-
These assays include measurements for endpoints that are not necessarily related to the respective chemical MOAs
g
Miscellaneous assays include: “apolipoprotein,” “catalase,” “filaments,” “growth factor,” “histones,” “ion channel,” “kinase,”
“lysase,” “malformation,” “membrane protein,” “metabolite,” “microna,” “miscellaneous proteins,” “mutagenicity response,”
“oxidase,” “protease,” “protease inhibitor,” “phosphatase”
h
-All active hit-calls from parameters unrelated to the MOA and/or associated toxic endpoints and assays with cautionary
flags with unacceptable dose-response curves.
Abbreviations: CB: carbaryl; CPF: chlorpyrifos; CPFO: chlorpyrifos-oxon; ENDO: endosulfan; PZ: propyzamide
Vendors with the greatest number of active hit-calls below
the cytotoxicity limit were from Novascreen (NVS: cell-
free) and CellzDirect (CLD: primary human liver cells) and
were related to liver metabolic processes for all 5
chemicals (Supplemental Tables 2b-6b). Other intended
target families with active hit-calls below the cytotoxicity
limit, included estrogenic (CPFO, ENDO), androgenic
(ENDO, PZ), neurotoxicity (CPFO, ENDO, CB) and steroid
hormone (CPF, PZ). Active hit-calls with cautionary flags
were selected for ranking in ToxPi given the following
considerations: 1) Appearance of the dose-response
curve; 2) relevance of the assay to the primary MOA as
well as secondary effects; 3) if the flagged AC50 is below
the cytotoxicity limit and there are more than one active hit-
call for a similar target, then the data may be selected; 4)
Selected assays had less than 3 flags, and AC50 value
greater than the lowest concentration screened (Friedman
et al., 2019). Assays with cautionary flags indicating the
results were likely false positives were excluded from the
ToxPi and ICE model analyses.
Chlorpyrifos-oxon: The intended target families
associated with the CPFO metabolic pathway included
nuclear receptors (PXR, PPAR, FXR, RXR and CAR)
involved with CYP induction (Chang et al., 2003; Holick,
2005; Li et al., 2017; Michalik et al., 2006; Wang and
Negishi, 2003), that may be specific to CPFO metabolism
(Supplemental Figure 1; Supplemental Table 2a). Some
were below the cytotoxicity limit of 11.72 µM, including
many CYPs related to the MOA: CYP2C19, CYP2B6
(Foxenberg et al., 2007), CYP1A2 (Foxenberg et al., 2011)
and AChE inhibition (Testai et al., 2010)). The cell-free
Novascreen assay for BuChE (NVS_ENZ_hES) had a
cautionary flag (AC50 below the lowest concentration
tested) and was eliminated for use in ToxPi and ICE model
analyses because the dose response curve indicated
results were likely false positive.
There were active hit-calls for nuclear receptor estrogenic
(8), androgenic (4), thyroid (2) and neurotoxicity pathways
(4), some with AC50s below the cytotoxicity limit (3
estrogenic, 1 neurotoxicity). One estrogen receptor assay
(ATG_Era_TRANS_up) appeared to be a false positive
based on the cautionary flag (hit-call potentially
confounded by overfitting: Only highest conc above
baseline, active). It also had an unacceptable curve
appearance on the CompTox Dashboard; hence, it was
not used in the ToxPi or ICE model analyses. The
8. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
neurotoxicity endpoints related to dopamine target (DRD1)
and opiate receptors have been associated with CPF
treatment (Aldridge et al., 2005; Slotkin and Seidler, 2010).
Active hit-calls DRD1 and GHB (Tspan1 gene) are
associated with autism spectrum disorder (Li et al., 2017)
which is also linked to CPF exposure (De Felice et al.,
2015; Lan et al., 2019). There were two mitochondria
assays that, although the AC50s were above the
cytotoxicity limit and had only 50% efficacy, were selected
for use in ToxPi because CFPO is a mitochondrial toxicant
(Cole et al., 2011).
Chlorpyrifos: Of the active hit-calls for liver metabolism:
nuclear receptors (i.e., PXRE, FXR, Ahr, RXR, CAR),
CYPs (CYP1A1, 1A2D, 2B6, 3A4, 2C19) and transfer
factor (UGT1A1), a few had AC50s below the CPF
cytotoxicity limit of 15.74 µM (Table 2, Supplemental
Figure 1, Supplemental Table 3a) (Michalik et al., 2006;
Wang and Negishi, 2003). The cell-free Novascreen assay
for BuChE inhibition had an active hit-call, although it is
primarily inhibited by CPFO (Testai et al., 2010). RBC
AChE inhibition, as a sensitive endpoint, is preferable to
BuChE inhibition because plasma has a mix of AChE and
BuChE, where RBCs have only AChE (US EPA, 2000). As
stated previously, RBC AChE inhibition also serves as a
surrogate for brain AChE inhibition (US EPA, 2000).
There were active hit-calls with the estrogen (7),
progesterone (1), androgen (5), thyroid (2) and steroid
hormone (8) targets but none was below the cytotoxicity
limit. CPF was tested in the Endocrine Disruptor Screening
Program (EDSP: Tier 1) for potential effects on the
estrogen, androgen and thyroid pathways, but based on
the weight-of-evidence, it was not considered to be an
endocrine disruptor (https://www.epa.gov/endocrine-
disruption/endocrine-disruptor-screening-program-tier-1-
screening-determinations-and). An estrogen nuclear
receptor assay by Odyssey Thera (OT:
OT_ER_ERaERa_1440) had 3 cautionary flags (<50%
efficacy: hit-call potentially confounded by overfitting:
Borderline active) and graphically appeared to be a false
positive, therefore it was not included in the ToxPi or ICE
model analyses (Supplemental Table 3b). While endocrine
ToxCast/Tox21 results may be unrelated to the primary
CPF toxicity, depending on exposure, disruptions of these
targets may have effects on endocrine parameters during
development.
A single mitochondria assay had an active hit-call with CPF
and although it was above the cytotoxicity limit, with 2
cautionary flags, it was retained for ToxPi because CPF is
a mitochondrial toxicant (Park et al., 2017; Singh et al.,
2018; Yamada et al., 2017).
Carbaryl: The intended target families in the CB metabolic
pathway included DNA Binding (Ahr) and nuclear receptor
(PXRE, PPRE, PPAR and CAR) assays; all of which are
potentially involved with the CYPs related to CB
metabolism (Chang et al., 2003; Faucette et al., 2007;
Ihunnah et al., 2011; Lin et al., 2009; Michalik et al., 2006)
(Table 2, Supplemental Figure 1 and Supplemental Table
4a). The CYP LTEA_HepaRG_CYP2C19_up assay had 4
cautionary flags (only highest conc above baseline, active:
Borderline active: <50% efficacy: Hit-call potentially
confounded by overfitting) and appeared graphically on
the CompTox Dashboard as a false positive. It was not
included in the ToxPi or ICE model analyses. The Ahr
assays (induces CYP1A1; Ko et al. (1996)) related to liver
metabolism, had AC50s below the cytotoxicity limit (11.72
µM). Nuclear receptor assays for PPRE, PXRE, PPARɣ
and CAR were above the cytotoxicity limit. Of the active
hit-calls for the CYP intended target families, most had
AC50s below the cytotoxicity limit, including CYP1A2 (Tang
et al., 2002), CYP1A1 (Tang et al., 2002) and Ahr (through
induction of CYP1A1 metabolic activation; Ko et al.
(1996)). These targets are associated with the CB MOA.
Esterase intended target family had active hit-calls for
AChE inhibition which is focal to CB metabolism and
toxicity (Blacker et al., 2010; CDPR, 2014). While AChE
inhibition is rapidly reversed with CB, there remains
concerns for non-cholinergic long-term effects on the
developing brain, even after a single administration (Lee et
al., 2015). For example, an assay for the serotonin-binding
transporter (SERT) in the brain had an AC50 below
cytotoxicity limit. CB effects on this transporter is
associated with increased stress sensitivity and stress-
related disorders in rats (Brivio et al., 2019) as well as
effects on motor activity and cognition (Piper et al., 2005)
(Table 6). CB was also active with oxidoreductase
monooxygenases (MAO) receptors that affect pathways in
the liver and the brain (Beliaev et al., 2009; Gaweska and
Fitzpatrick, 2011).
CB was tested in EDSP because it was shown to have
potential for effects on reproduction in females, and
thyroid, testes, pituitary and adrenal glands (reviewed in
CDPR (2014)) but the weight of evidence did not support
CB as having endocrine disrupting activity (US EPA,
2015a). The nuclear receptor and oxidoreductase active
hit-calls associated the estrogen (7), androgen (3), and
steroid hormone (5) pathways exceeded the cytotoxicity
limit and their activity was likely incidental.
Endosulfan: ENDO metabolic pathways in liver had active
hit-calls for targets that were directly relevant to the MOA
(e.g., CYP2B6, CYP3A4, FXR, PXRE, CAR, PBREM
[phenobarbital response enhancer module], UGT1A).
Many had AC50s at or below the cytotoxicity limit (12.31
µM) (Table 2; Supplemental Figure 1 and Supplemental
Table 5a-b) (Bebe and Panemangalore, 2003; Casabar et
al., 2006; Ihunnah et al., 2011; Pavek and Dvorak, 2008).
Two nuclear receptor assays and one CYP assay
appeared to be false positives with unacceptable dose-
response curves on the CompTox Dashboard
(LTEA_HepaRG_CYP2C19_up,
NVS_NR_hCAR_Antagonist, ATG_PBREM_CIS_up) and
they were excluded from the ToxPi or ICE model analysis.
9. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
ENDO had active hit-calls for nuclear receptor target
families related to estrogen (11), progesterone (1), and
androgen metabolism (9), many of which had AC50s below
the cytotoxicity limit (Supplemental Table 6b). It is well-
documented that in vivo ENDO treatment affects
estrogenic and androgenic pathways (Silva and Gammon,
2009; Silva et al., 2015).
The thyroid target had active hit-calls, although ENDO is
generally not known to interact with the thyroid receptor as
a major pathway (CDPR, 2008; US EPA, 2002a). An acute
high dose of ENDO in pubertal male rats (6.12 mg/kg/day)
resulted in down regulation of thyroid stimulating hormone
(TSH) along with decreased plasma concentrations
(Caride et al., 2010). Caride et al. (2010) suggested that
these effects were associated with pituitary toxicity, since
prolactin, luteinizing hormone, growth hormone and TSH
expressions were also modified by ENDO. ENDO had an
active hit-call with the sulfotransferase (SULT) target
which is associated with androgen and estrogen
metabolism (Ihunnah et al., 2011; Khor et al., 2010) as well
as xenobiotic detoxification (Duffel, 2010). The DAT
assays in human and rat were the only active hit-calls
related to neurotoxicity but effects on this target are
associated with ENDO treatment (Jacob et al., 2008; Jia
and Misra, 2007; Seth et al., 1986; Wilson et al., 2014).
Propyzamide: An intended target family in the PZ
metabolic pathway included nuclear receptors (PXRE,
PPRE, PPAR, PXR), with one AC50 below the cytotoxicity
limit (11.05 µM). These endpoints are potentially involved
with induction of CYPs related to PZ metabolism in the liver
(Chang et al., 2003; Ihunnah et al., 2011; LeBaron et al.,
2014; Lin et al., 2009; Michalik et al., 2006) (Table 2;
Supplemental Figure 1 and Supplemental Table 6a-b).
The CYP target family (CYP1A1, CYP1A2, CYP2B6 [PXR
associated], CYP2A1, CYP3A4) had active hit-calls for
CYPs that are known or likely to be involved with PZ
metabolism (LeBaron et al., 2014). CYP1/2 families can be
activated in liver by lipophilic xenobiotics and repeated
dose studies with PZ in rodents have shown increased
hepatocellular hypertrophy associated with CYP induction
(LeBaron et al., 2014; Papineni et al., 2015; Rasoulpour et
al., 2015).
The transferase intended target families had active hit-
calls for UGT1A1, associated with elimination of thyroxine
(T4) in rodents, and disruption of the hypothalamus-
pituitary-thyroid (HPT) axis (Hood et al., 2003; Papineni et
al., 2015). This MOA involves PZ induction of UGT1A1,
which then conjugates to T4 leading to increased
elimination (Papineni et al., 2015). Sustained thyroid
stimulation by TSH will lead to thyroid follicular cell
hypertrophy, hyperplasia, and tumorigenesis. UGT1A1
assays had AC50s below the cytotoxicity limit. PZ also had
an active hit-call below the cytotoxicity limit for
sulfotransferase (SULT) targets. Sulfotransferases are
associated with androgen and estrogen metabolism
(Ihunnah et al., 2011; Khor et al., 2010).
PZ treatment leads to disruption of the hypothalamus-
pituitary-gonad (HPG) axis leading to androgenic
imbalance (Rasoulpour et al., 2015). Further, estrogen (1),
progesterone (1), androgen (3), and steroid hormone (8)
assays had active hit-calls below the cytotoxicity limit. One
of the three androgen assays (ACEA_AR_agonist_80hr)
had 4 cautionary flags (AC50 < lowest concentration tested:
< 50% efficacy: Hit-call potentially confounded by
overfitting: Borderline active: Only one conc above
baseline, active). The dose-response curve on the
CompTox Dashboard indicated a false positive and it was
not used in the ToxPi or ICE model analyses. The other
endocrine assays were likely non-specific interactions due
to the lipophilicity of PZ.
The ToxPi Program
ToxPi program inputs were AC50s for the selected assays
that could be associated with the known MOAs or other
toxic endpoints (Supplemental Tables 2b-6b). The ToxPi
model ranked the five chemicals from lowest to highest
toxicity (rank 1 to 5) for the intended target family
categories designated by the CompTox Dashboard
(Figure 1 and Table 4). The dimensionless Toxicity Score
is the sum of all the individual intended target family scores
divided by the 11 categories (i.e., CYP, DNA Binding,
Esterase, GCPR, Growth Factor Inhibitor, Mitochondria,
Nuclear Receptor, Steroid Hormone, Transferase,
Transporter, Oxidoreductase). For example, using CPFO
data from Figure 1 in Equation (3) (Marvel et al., 2018):
CPFO Toxicity Score = (ΣIntended Target Family
scores [5.6316]) ÷ (No. of categories [11]) = 0.5120)
Equation (3)
The scores then determined the ranks, with the highest
score having the highest rank and greatest toxicity (Reif et
al., 2010). Output data with the rank from low (least toxic)
to high (most toxic) and Toxicity Score (Figure 1: shown
as color-coded slices) were as follows: PZ (Rank 1:
0.2633) < CPF (Rank 2: 0.3271) < CB (Rank 3: 0.3800) <
ENDO (Rank 4: 0.4548) < CPFO (Rank 5: 0.5120).
The acute in vivo ranking as determined by the
LOEL/BMD10s was as follows: PZ (Rank 1: ENEL 4.0
mg/kg/day) < ENDO (Rank 2: LOEL 1.8 mg/kg/day) <. CB
(Rank 3: BMD10 RBC AChE/Brain AChE: 1.11/1.46
mg/kg/day) < CPF (Rank 4: BMD10 RBC AChE/Brain
AChE: 0.5/1.42 mg/kg/day) < CPFO (Rank 5: BMD10 RBC
AChE/Brain AChE: 0.093/1.06 mg/kg/day). Table 3 shows
concordance between ToxPi ranking and in vivo endpoints
for CPFO, CB and PZ, however ENDO and CPF were not.
Since the ToxPi model was based on ToxCast/Tox21
AC50s, as well as selection of assays with active hit-calls a
certain degree of variability was introduced. Further, three
of the in vivo lower effect levels had endpoints that were
almost equivocal (i.e., ENDO: 1.8 mg/kg/day; CB:
1.11/1.46 mg/kg/day; CPF: 0.5/1.42 mg/kg/day). Variability
in the in vivo endpoints was dependent on expert
10. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
decisions by regulatory agencies (i.e., US EPA, CDPR),
that must rely on protocols (e.g., treatment levels, dose
spacing, duration, animal strain, chemical route of
administration and other factors) determined by pesticide
registrants or study authors. ToxPi, concordant in 3 of the
5 pesticide rankings, was predictive considering the
numerous variables in both in vivo and in vitro models
(Table 3).
Table 3: Concordance Between ToxPi Score/Ranking and Acute In Vivo Low Effect Level Rank
Chemical
Toxicity
Score
ToxPi
Toxicity Rank
In Vivo
(mg/kg/day)
In Vivo
Toxicity Rank
Concordance ToxPi Predictions and
Acute In Vivo POD Values
PZ 0.2633 1 40 1 +
CPF 0.3271 2 0.5/1.42 a 4 --
CB 0.3800 3 1.11/1.46a 3 +
ENDO 0.4548 4 1.8 2 --
CPFO 0.5120 5 0.093/1.06a 5 +
Green Boxes indicate concordance and red boxes indicate no concordance
Toxicity ranking: 1 (least toxic); 5 (most toxic)
a- BMD10 for RBC/brain AChE inhibition, respectively
Figure 1: ToxCast/Tox21 AC50 values for all ENDO, CPFO, CPF, CB and PZ active hit-calls are categorized and
scored by intended target family. The top panel has the color codes for ToxPi Intended Target Families. The lower
panel shows the calculated scores for each intended target family, the overall Toxicity Scores, and the rank, which goes
from high (5=most toxic) to low (1=least toxic). Selected active hit-calls included assays associated with primary MOA or
toxic endpoint determinations. Toxicity Score is calculated as:
Toxicity Score = (ΣIntended Target Family scores) ÷ (number of Intended Target Family categories)
11. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
High Throughput Toxicokinetics
Assays selected for the ICE 1C, 3C and PBTK modeling
are shown in Table 4 (Supplemental Table 8). AC50 values
were all below the cytotoxicity limit and there were 4, 14,
7, 10 and 6 AC50s for CPF, CPFO, CB, ENDO, and PZ,
respectively as model inputs. Selection also involved
elimination of identical assays performed at different
durations for each chemical. For example, the AC50 for
CYP2B6 at 24 hours (CLD_CYP2B6_24hr) with CPFO
was 0.404 µM but at 48 hours it was 11.2 µM. The
selection was made for the lower value, meaning that at
24 hours, this assay is a more sensitive endpoint.
Qualitative Comparisons Among Model Predictions
Qualitative data analysis demonstrated visually the
performance among the 1C, 3C and PBTK models.
Figures 2 and 3 show the "spread" of data as a means to
compare the performance of the different models. The
“ggplot2” calculated on R-studio provided an ideal
graphical depiction of the overall fold differences per
chemical as quartiles in a box-and-whiskers format with an
overlay of individual assay points (Pearce et al., 2017;
Wickham, 2016).
One Compartment Model: The 1C model had numerous
outliers beyond the median value (Figures 2-3), indicating
that it was not the best selection for modeling our data.
Qualitatively, CPFO EAD Human fold differences had the
most variability and the most outliers (least predictive),
where PZ EAD Human fold differences had no outliers (highly
predictive) (Figure 2).
3 Compartment and PBTK Models: Qualitative visual
EADHuman predictions for both 3C and PBTK models
showed the most fold difference variability around the
median with CPFO and the least with both ENDO and PZ.
It was also evident that the magnitude of qualitative
differences was dependent on the number of assay inputs
(Figure 2). However, ENDO and PZ do not need to be
activated to a toxic metabolite; as such, their AC50s might
be more representative of target activity. From a qualitative
assessment, it is evident that the 3C and PBPK models
provided equivocal fit-for-purpose in determining EADHuman
predictions.
Figure 2: Quanlitative fold difference comparison among
models and pesticides. Box and whiskers plots show fold
difference median, minimum and maximum values in
quartiles. Colored circles indicate fold differences for
individual assays in the three models.
Qualitative Comparisons Among Models for RBC
versus Brain AChE Inhibitors
Qualitative visual depictions of fold difference between
EADHuman for RBC vs. brain AChE inhibitors and in vivo
RBC vs. brain ADJBMD10, respectively, were compared
(Wickham, 2016). The 3C and PBTK were better
predictors of EADHuman for RBC and brain AChE inhibition
based on lower fold differences around the median (Figure
3). Brain AChE inhibitor EADHuman were more predictive in
all ICE models than RBC AChE. The predicted order of
toxicity compared to in vivo BMD10 for both RBC and brain
AChE inhibition was CPFO>CPF>CB. This qualitative
order was 100% predictive of the measured in vivo order
(Table 1).
Figure 3: Qualitative fold differences among the 3 models
for EADHuman predictions and in vivo RBC or brain AChE
ADJBMD10s are shown along with direct RBC vs. brain
AChE inhibition fold difference comparisons. Box and
whiskers plots show fold difference median, minimum and
maximum values in quartiles. Colored circles indicate
relative values for individual assays in the three models.
12. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
Quantitative Comparisons Among Model Predictions
Overall quantitative fold differences for each model and
pesticide are shown in Figures 4-8. Quantitative fold
differences are categorized as <10, 11 to 100 and greater
than 101. Data are described below as number of assays
in a category per total assays.
One Compartment Model: Quantitatively the 1C model
was predictive of EADHuman (< 10-fold difference) in 6/6
assays for PZ and 4/14 assays for CPFO for the brain
AChE ADJBMD10 endpoint (Figures 4-8; Supplemental
Table 9). 1C EADHuman with fold differences between 11
and 100 were 4/14 CPFO (RBC) and CPFO 10/14 (brain),
2/4 CPF (brain), 1/7 CB (brain) and 6/10 ENDO. 1C with
EADHuman with fold differences greater than 101 were 10/14
CPFO (RBC), 4/4 CPF (RBC) and 2/4 CPFO (brain), 7/7
CB (RBC) and 6/7 CB (brain) and 4/10 ENDO.
Three Compartment Model: The 3C model was
quantitively predictive of EADHuman in 3/14 CPFO (RBC)
and 14/14 CPFO (brain), 1/4 CPF and (RBC), 2/4 CPF
(brain), 1/7 CB (brain), 10/10 ENDO and 3/6 PZ (Figures
4-8; Supplemental Table 9). 3C EADs with fold differences
between 11-100 were 11/14 CPFO (RBC) and 14/14
CPFO (brain), 3/4 CPF (RBC) and 2/4 CPF (brain), 7/7 CB
(RBC) and 6/7 CB (brain) and 3/6 for PZ.
PBTK Model: Quantitatively the PBTK model was
predictive of EADHuman in 3/14 CPFO (RBC) and 12/14
CPFO (brain), 2/4 CPF (brain), 10/10 ENDO and 5/6 PZ
(Figures 4-8; Supplemental Table 9). PBTK EADs with fold
differences between 11 and 100 were 9/14 CPFO (RBC)
and CPFO 2/14 (brain), 4/4 CPF (RBC) and 2/4 CPF
(brain), 1/6 PZ. PBTK with EADs with fold differences
greater than 101 were 2/14 CPFO (RBC).
13. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
Figure 4: Data identify fold difference between predicted EADHuman for RBC and Brain AChE and their ADJBMD10 in vivo endpoints using ToxCast/Tox21 AC50 and
ICE models.
Figure 5: Data identify fold difference between predicted EADHuman for RBC and Brain AChE and their ADJBMD10 in vivo endpoints using ToxCast/Tox21 AC50 and
ICE models.
752
66
97
8.5
132
12
337
30
43
3.8
59
5.2
362
32
47
4.1
63
5.6
468
41
61
5.3
82
7.2
303
27
39
3.4
53
4.7
36
3.1
4.62
2.5
6.3
1.8
96
8.4
12
1.1
17
1.5
27
2.4
3.5
3.2
4.8
2.4
405
36
52
4.6
71
6.2
29
2.5
3.7
0.32
5
2.3
201
18
26
2.3
35
3.1
375
33
48
4.3
66
5.8
360
32
46
4.1
63
5.5
737
65
95
8.4
129
11
0
100
200
300
400
500
600
700
800
RBC AChE 1C Brain AChE 1C RBC AChE 3C Brain AChE 3C RBC AChE PBTK Brain AChE PBTK
CHLORPYRIFOS OXON
NVS_ADME_hCYP1A1 NVS_ADME_hCYP1A2 NVS_ADME_hCYP2C19 CLD_CYP1A1_6hr
CLD_CYP1A2_24hr CLD_CYP2B6_24hr TOX21_PPARg_BLA_antagonist_ratio OT_FXR_FXRSRC1_1440
CLD_UGT1A1_48hr NVS_ENZ_hAChE TOX21_ERa_BLA_Antagonist_ratio TOX21_ERb_BLA_Antagonist_ratio
TOX21_PR_BLA_Antagonist_ratio NVS_GPCR_hDRD1
322
113
2.3
0.81
35
12
220
78
16
5.7
24
8.5
711
250
52
18
78
27
119
42
51
18
13
4.6
0
100
200
300
400
500
600
700
800
RBC AChE 1C Brain AChE 1C RBC AChE 3C Brain AChE 3C RBC AChE PBTK Brain AChE PBTK
CHLORPYRIFOS
ATG_PXRE_CIS_up ATG_PXR_TRANS_up LTEA_HepaRG_CYP2B6_up ATG_Ahr_CIS_up
14. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
Figure 6: Data identify fold difference between predicted EADHuman for RBC and Brain AChE and their ADJBMD10 in vivo endpoints using ToxCast/Tox21 AC50 and
ICE models.
Figure 7: Data identify fold difference between predicted EADHuman for neurotoxicity and the ADJLOEL in vivo endpoint using ToxCast/Tox21 AC50 and ICE models.
The tables were divided based on the fold difference scale for 1C (24 to 309) compared to the 3C and PBTK models (0.03 to 11).
197
260
23
18
63
48
183
241
21
16
59
45
89
117
10
8.0
29
22
263
346
31
23
85
64
136
179
16
12
44
33
139
183
16
12
45
34
122
161
14
11
39
30
0
50
100
150
200
250
300
350
400
RBC AChE 1C Brain AChE 1C RBC AChE 3C Brain AChE 3C RBC AChE PBTK Brain AChE PBTK
CARBARYL
ATG_Ahr_CIS_up NVS_ADME_hCYP1A2 CLD_CYP1A2_24hr NVS_ENZ_hAChE CLD_CYP1A1_48hr CLD_CYP2B6_48hr TOX21_AhR_LUC_Agonist
15. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
Figure 8: Data identify fold difference between predicted
EADHuman for effects on motor activity and the ADJLOEL in
vivo endpoint using ToxCast/Tox21 AC50 and ICE model
DISCUSSION
In pesticide risk assessment, the traditional in vivo Health
Effects Test Guideline studies are designed to observe a
low or no effect dose level and, if possible, identify a
targeted system in a dose-related manner. This type of
design can have large gaps between doses and events
leading to a “tipping point,” where an organism may no
longer be able to recover from chemical insult, can be
masked (Frank et al., 2018; Saili et al., 2020; Shah et al.,
2016). However, currently there are numerous open
access computational tools available that can help reveal
and prioritize chemical activities at their tipping points prior
to irreversibly overwhelming metabolic systems. More
sophisticated tools are in development that will replace
animal studies for pesticide risk assessment by 2035
(Nyffeler et al., 2020; US EPA, 2020a)}. The user-friendly
open access tools, such as ToxCast/Tox21 and ToxPi are
used to identify MOAs, AOPs, and previously unknown
molecular interactions that may impact pesticidal effects
on human health (Browne et al., 2015; Kleinstreuer et al.,
2017; Kleinstreuer et al., 2013; Kleinstreuer and Knudsen,
2011; Pham et al., 2016). Even if a molecular event is
minimal, in combination with other disruptions it may prove
critical to the understanding of chemical action that can
support and facilitate the risk assessment process.
The acute in vivo endpoints used in this study were
obtained from open access documents, including Health
Effects Test Guideline studies as reported in risk
assessments by the California Department of Pesticide
Regulation (CDPR, 2008, 2014, 2018) or the US EPA (US
EPA, 2011, 2015c, 2017a). BMD10 analyses for CPF,
CPFO and CB were performed by the US EPA and had
the advantage of a modeled endpoint, where ENDO and
PZ data relied on registrant-determined dose levels for
endpoint determination (i.e., NOEL/ENEL/LOEL). An
ENEL could be considered an artifact of dose selection
and the uncertainty factor used to derive it. On the other
hand, comparative data across several laboratories
indicate that motor activity is quite variable, such that an
effect of more than 30% difference from control must occur
to distinguish signal from noise (Crofton et al., 1991;
Wolansky et al., 2006).
Open access ToxCast/Tox21 assays produced active hit-
calls that supported, for the most part, the molecular
mechanistic functions related to each pesticide. For
example, many active hit-calls were for targets associated
with reported liver metabolic pathways, as well as
pathways for neurotoxicity and endocrine disruption for
ENDO, CPFO, CB and PZ. Several, albeit a small
percentage of overall active hit-calls, were below the
cytotoxicity limit and supported the use of the
ToxCast/Tox21 assays to accurately predict potential
pathways where disruptions could lead to adverse effects.
CPF had the fewest active hit-calls and none was specific
to the MOA, which strengthened previous findings
requiring activation of CPF to CPFO for adverse effects to
occur (i.e., AChE inhibition). PXRE/PXR are xenosensors
that are activated as adaptive responses leading to
chemical activation or detoxification, but they are not
specific to the CPF MOA (Kliewer, 2003; Knudsen et al.,
2013; Knudsen et al., 2011). CPF results also indicated
that human primary cells or human-derived cell lines and
cell-free biochemical assays may not adequately allow for
chemical interaction at the target at a concentration that
would be observed in vivo. CPFO and CB inhibited AChE
and esterase activities for both compounds were below the
cytotoxicity limit, indicating a likely specificity. The two cell-
free Novascreen assays are the only AChE assays
available in the ToxCast/Tox21 battery. The results for
CPF show that this compound is not likely to inhibit AChE
without metabolic activation.
The primary metabolic pathways for ENDO in liver (e.g.,
CYP2B6 and CYP3A4/5) had active hit-calls but the
GABAAR, critical to the MOA, did not (Kamijima and
Casida, 2000; Ratra et al., 2001). The GABAARβ3 subunit
is the major site of binding for this pesticide (Jacob et al.,
2008; Kamijima and Casida, 2000; Ratra et al., 2001) but
there was not an assay for this endpoint on the CompTox
Dashboard. Only one GABAAR assay
(NVS_LGIC_bGABARa5; inactive) was performed with
ENDO but it was not active. Although neurotoxicity assays
were for DA and DAT, both of which are associated with
ENDO toxicity in vivo and in vitro (Jacob et al., 2008; Jia
and Misra, 2007; Lakshmanan et al.; Seth et al., 1986;
Wilson et al., 2014).
The primary MOA for PZ is initiated in the liver and while
there were numerous active hit-calls involving liver
metabolism, ToxCast/Tox21 assay results also indicated
effects on endocrine endpoints that also occur in vivo
(LeBaron et al., 2014; Papineni et al., 2015; Rasoulpour et
al., 2015). However, the lack of activity with thyroid and
16. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
testosterone assays supported the in vivo data indicating
that effects on these hormones occur through disruption of
the HPG/HPT axes and are secondary to effects in the liver
(Papineni et al., 2015; Rasoulpour et al., 2015).
The open access ToxPi model used in this study was
dependent on individual ToxCast/Tox21 AC50s for selected
assays within relevant intended target families. Therefore,
the ranking was only as predictive as the potency (AC50
value) of the active hit-calls for each chemical/target. While
the selected assays in the intended target families may be
associated with the MOAs, their relative AC50 values may
or may not represent what would occur in vivo. For
example: 1) cell-free assays involve chemical to target
protein activity that is not physiological; 2) many of the
AC50s were above the cytotoxicity limit, thereby rendering
the relevance of those assay results as undetermined; 3)
while selected active hit-calls identified many of the
enzymes and proteins in the metabolic pathways
supporting the mechanistic weight-of-evidence for each
pesticide, these targets would also be active for a variety
of lipophilic compounds; 4) active hit-calls may be false
positives based on cautionary flags due to automated
processes that are prone to error (Filer, 2019; Filer et al.,
2017; Ryan, 2017). However, the use of ToxCast/Tox21
AC50s is supported by the knowledge that ToxPi is a model
for generalized ranking (prioritizing) chemical toxicity.
ToxPi predicted Toxicity Scores and ranks were least to
most potent: PZ < CPF < CB < ENDO < CPFO, while the
in vivo toxicity lowest effect levels were as follows: PZ <
ENDO < CB < CPF < CPFO. The comparison showed
concordance for 3 of 5 compounds, which is fairly
predictive considering the following potential confounders:
1) number of active hit-calls relevant to the MOA or
associated toxic effects, which is dependent on metabolic
capability of the cells, assay design and other limitations of
cell or cell-free systems (NAS, 2017); 2) selection of
relevant, MOA-related ToxCast/Tox21 assays for ToxPi
analysis; 3) Variability of the in vivo endpoints are
dependent on study protocols determined by Health
Effects Test Guidelines (US EPA, 1998), pesticide
registrants or study authors. Due to the degree of
variability among in vitro and in vivo parameters, ToxPi
ranking may not be exact.
The open access ICE program provided the 3 IVIVE/PBTK
models for this study (Bell et al., 2020; Bell et al., 2018).
The 1C linear model was comparable to a constant
infusion exposure (Css 95th percentile) and was only
predictive for PZ at less than 10-fold difference from in vivo
endpoints (Figures 2-8). The oral acute PBTK model for
exposure at Cmax was less predictive than the 3C model.
One explanation could be that the PBTK model, with
multiple compartments, incorporating a wide range of
mechanistic information and complexity may also
introduce increased variability with each added PBTK
parameter. The 3C model may have a more appropriate
fit-for-purpose, considering the data available in this study
(Sipes et al., 2017). That is, the PBTK model, with higher
data requirements may require more precision than we
needed.
Qualitatively fold differences between EADHuman
predictions and acute oral in vivo ADJBMD10/ADJLOEL
values for each compound and model were graphed in
quartiles and indicated that rank among models and
chemicals was: 1C < PBTK < 3C. Quantitative fold
differences for each pesticide provided a more in-depth
evaluation of the EADHuman predictions for each model.
Divided into three categories of fold difference (<10;
>11<100; >101), three of four of the direct acting
pesticides (CPFO [brain], ENDO, PZ) had the most
predictive EADHuman, primarily in the 3C model, but also in
the 1C model where PZ EADHuman were 100% predictive.
The exception was CB which had almost 100% non-
predictive EADHuman in the 1C and only one predictive EAD
(brain AChE inhibition endpoint) in the 3C model. EADHuman
predictions for AChE inhibitors (CPFO, CPF, CB),
especially CB was not as well captured in the
ToxCast/Tox21 assays since liver metabolism would likely
be the initial pathway but not the toxic pathway for these
compounds. The most striking overall observation was the
difference among the three ICE models indicating that
model refinement/complexity did not necessarily provide
more accurate results.
In pesticide risk assessment, where open access in vivo
Health Effects Test Guideline results were available,
ToxCast/Tox21 data, ToxPi ranking and ICE modeled
EADHuman predictions provided support for the presumptive
MOAs and mechanistic pathways in most cases. These
tools, as applied in this case study, were easily accessible,
user friendly and provided a variety of methods for
assessing chemical toxicity. They will prove to be even
more useful as technology progresses in not only
prioritizing pesticides for risk assessment but also
providing insights into mechanistic data often lacking in
traditional testing.
FUNDING: This research did not receive any specific grant
from funding agencies in the public, commercial, or not-
for-profit sectors.
ACKNOWLEDGEMENT: We would like to thank Emeritus
Professor Warren K. Musker (retired from the Chemistry
Department, University of California, Davis, CA, USA) for
his help with the chemistry and selection of relevant
pesticides for our case study.
DECLARATION of INTEREST: I have no competing
interests (financial/personal) to declare.
17. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
REFERENCES
Aldridge, JE, Meyer, A, Seidler, FJ, Slotkin, TA. (2005a).
Alterations in central nervous system serotonergic and
dopaminergic synaptic activity in adulthood after
prenatal or neonatal chlorpyrifos exposure. Environ
Health Perspect, 113(8), 1027-1031. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/16079074.
Accessed June 30, 2020).
Angrish, MM, Kaiser, JP, McQueen, CA, Chorley, BN.
(2016). Tipping the Balance: Hepatotoxicity and the 4
Apical Key Events of Hepatic Steatosis. Tox Sci,
150(2), 261-268.
APVMA. (2012). CARBARYL. Australian Pesticides and
Veterinary Medicines Authority (APVMA), Part 2: Uses
of carbaryl in agricultural situations. www.apvma.gov.a.
Accessed Sept 27, 2020.
APVMA. (2017). Reconsideration of chlorpyrifos:
supplementary toxicology assessment report. Director
Public Affairs and Communication, Australian
Pesticides and Veterinary Medicines Authority. PO Box
6182, KINGSTON ACT 2604 Australia: Licensed from
the Australian Pesticides and Veterinary Medicines
Authority (APVMA) under a Creative Commons
Attribution 3.0 Australia Licence Retrieved from
https://apvma.gov.au/sites/default/files/publication/268
31-
chlorpyrifos_supplementary_toxicology_assessment_r
eport_1.pdf. Accessed Sept 27, 2020
APVMA. (2020). Standard for propyzamide active
constituent. Australian Pesticides and Veterinary
Medicines Authority, March 25, 2019;
https://apvma.gov.au/node/2751. Accessed Sept 27,
2020.
ATSDR. (2000). Toxicological Profile for Endosulfan.
Agency for Toxic Substances and Disease Registry,
Division of Toxicology/Toxicology Information Branch.
https://www.atsdr.cdc.gov/toxprofiles/tp41.pdf#:~:text=
v%20.%20This%20toxicological%20profile%20is%20p
repared%20in,profile%20will%20be%20revised%20an
d%20republished%20as%20necessary. Accessed July
10, 2020.
Bebe, FN, Panemangalore, M. (2003). Exposure to Low
Doses of Endosulfan and Chlorpyrifos Modifies
Endogenous Antioxidants in Tissues of Rats. J Environ
Sci Health, Part B, 38(3), 349-363. doi:10.1081/PFC-
120019901.
Beliaev, A, Ferreira, H, Learmonth, DA, Soares-da-Silva,
P. (2009). Dopamine β-Monooxygenase: Mechanism,
Substrates and Inhibitors. Curr Enz Inhib, 5(1), 27-43.
Bell, S. M., Abedini, J., Ceger, P., Chang, X., Cook, B.,
Karmaus, A. L., . . . Kleinstreuer, N. (2020). An
integrated chemical environment with tools for chemical
safety testing. Tox in Vitro, 67, 104916.
Bell, SM, Chang, X, Wambaugh, JF, Allen, DG, Bartels, M,
Brouwer, KLR, . . . Kleinstreuer, NC. (2018). In vitro to
in vivo extrapolation for high throughput prioritization
and decision making. Toxicol in Vitro, 47, 213-227.
Blacker, A.M, Lunchick, C, Lasserre-Bigot, D,
Payraudeau, V, Krolski, ME. (2010). Toxicological
Profile of Carbaryl. In R. Krieger (Ed.), Hayes'
Handbook of Pesticide Toxicology (3rd ed., Vol. II, pp.
1607-1617). Academic Press, Elsevier.
Braeuning, A, Gavrilov, A, Brown, S, Wolf, CR,
Henderson, CJ, Schwarz, M. (2014). Phenobarbital-
Mediated Tumor Promotion in Transgenic Mice with
Humanized CAR and PXR. Toxicol Sci, 140(2), 259-
270.
Brivio, P, Homberg, J.R, Riva, MA, Calabrese, F. (2019).
Alterations of Glutamatergic Markers in the Prefrontal
Cortex of Serotonin Transporter Knockout Rats: A
Developmental Timeline. Cell Mol Neurobiol, 39(5),
715-720.
Browne, P, Judson, R, Casey, WM, Kleinstreuer, NC,
Thomas, RS. (2015). Screening Chemicals for
Estrogen Receptor Bioactivity Using a Computational
Model. Environ Sci Technol, 49, 8804-8814.
Buratti, FM, Volpe, MT, Meneguz, A, Vittozzi, L, Testai, E.
(2003). CYP-specific bioactivation of four
organophosphorothioate pesticides by human liver
microsomes. Toxicol Appl Pharmacol, 186(3), 143-154.
Caride, A, Lafuente, A, Cabaleiro, T. (2010). Endosulfan
effects on pituitary hormone and both nitrosative and
oxidative stress in pubertal male rats. Toxicol Lett,
197(2), 106-112.
Casabar, RCT, Wallace, AD, Hodgson, E, Rose, RL.
(2006). Metabolism of Endosulfan-α by Human Liver
Microsomes and Its Utility as a Simultaneous in Vitro
Probe for CYP2B6 and CYP3A4. Drug Metab Dispos,
34(10), 1779-1785.
Casey, WM, Chang, X, Allen, DG., Ceger, PC, Choksi, NY,
Hsieh, J-H, . . . Kleinstreuer, NC. (2018). Evaluation
and Optimization of Pharmacokinetic Models for In Vitro
to In Vivo Extrapolation of Estrogenic Activity for
Environmental Chemicals. Environ Health Perspec,
126(9), 097001.
Casida, JE. (1993). Insecticide action at the GABA-gated
chloride channel: Recognition, progress, and
prospects. Arch Insect Biochem Physiol, 22(1‐2), 13-
23.
CDPR. (2008). Endosulfan Risk Characterization
Document.
https://www.cdpr.ca.gov/docs/whs/active_ingredient/in
dex.htm, Medical Toxicology and Worker Health and
Safety Branches Department Of Pesticide Regulation
California Environmental Protection Agency. Accessed
June 6, 2020.
CDPR. (2014). Carbaryl (1-naphthyl methylcarbamate):
Occupational and Bystander Risk Characterization
Document. Medical Toxicology Branch, Department of
Pesticide Regulation, California Environmental
Protection Agency, Sacramento, CA, June 23, 2014.
https://www.cdpr.ca.gov/docs/whs/active_ingredient/c
arbaryl.htm. Accessed Sept 29, 2020.
CDPR. (2017). Draft Evaluation of Chlorpyrifos as a Toxic
Air Contaminant: Risk Characterization of Spray Drift,
Dietary, and Aggregate Exposures to Residential
Bystanders. Human Health Assessment Branch,
Department of Pesticide Regulation, California
Environmental Protection Agency, Sacramento, CA,
http://www.cdpr.ca.gov/docs/whs/active_ingredient/chl
orpyrifos.htm. Accessed July 20, 2020.
18. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
CDPR. (2018). Evaluation of Chlorpyrifos as a Toxic Air
Contaminant: Addendum to the Risk Characterization
of Spray Drift, Dietary, and Aggregate Exposures to
Residential Bystanders Human. Human Health
Assessment Branch, Department of Pesticide
Regulation, California Environmental Protection
Agency, Sacramento, CA, http://www.cdpr.ca.gov/
docs/whs/active_ingredient/chlorpyrifos.htm. Accessed
July 20, 2020.
CFR. (2020). Code of Federal Regulations Title 40
(40CFR). https://www.epa.gov/pesticide-registration/
about-pesticide-registration#laws. accessed Sept 6,
2020.
Chang, TKH, Bandiera, SM, Chen, J. (2003). Constitutive
Androstane Receptor and Pregnane X Receptor Gene
Expression in Human Liver: Interindividual Variability
and Correlation with CYP2B6 mRNA Levels. Drug
Metab Dispos, 31(1), 7.
Chen, WL, Sheets, JJ, Nolan, RJ, Mattsson, JL. (1999).
Human Red Blood Cell Acetylcholinesterase Inhibition
as the Appropriate and Conservative Surrogate
Endpoint for Establishing Chlorpyrifos Reference Dose.
RegToxicol Pharm, 29(1), 15-22.
Coban, A, Carr, RL, Chambers, HW, Willeford, KO,
Chambers, JE. (2016). Comparison of inhibition
kinetics of several organophosphates, including some
nerve agent surrogates, using human erythrocyte and
rat and mouse brain acetylcholinesterase. Toxicol Lett,
248, 39-45.
Cole, LM, Casida, JE. (1986). Polychlorocycloalkane
insecticide-induced convulsions in mice in relation to
disruption of the GABA-regulated chloride ionophore.
Life Sci, 39(20), 1855-1862.
Cole, TB, Beyer, RP, Bammler, TK, Park, SS, Farin, FM,
Costa, LG, Furlong, CE. (2011). Repeated
developmental exposure of mice to chlorpyrifos oxon is
associated with paraoxonase 1 (PON1)-modulated
effects on cerebellar gene expression. Toxicol Sci,
123(1), 155-169.
Cone, M. (2010). Endosulfan to Be Banned, Pesticide
Poses "Unacceptable Risks," EPA Says. Scientific
American, June 10, 2010; https://www.
scientificamerican.com/article/endosulfan-banned-
epa/. Accessed Sept 27, 2020.
Crofton, KM, Howard, JL, Moser, VC, Gill, MW, Reiter, LW,
Tilson, HA, MacPhail, RC. (1991). Interlaboratory
comparison of motor activity experiments: Implications
for neurotoxicological assessments. Neurotoxicol
Teratol, 13(6), 599-609.
De Felice, A, Scattoni, ML, Ricceri, L, Calamandrei, G.
(2015). Prenatal exposure to a common
organophosphate insecticide delays motor
development in a mouse model of idiopathic autism.
PLoS One, 10(3), e0121663.
dos Santos, CEM, Nardocci, AC. (2019). Prioritization of
pharmaceuticals in drinking water exposure based on
toxicity and environmental fate assessment by in silico
tools: An integrated and transparent ranking. Comput
Toxicol, 9, 12-21.
Duffel, MW. (2010). 4.18 - Sulfotransferases. In McQueen,
CA. (Ed.), Comprehensive Toxicology (Second Edition)
(pp. 367-384). Oxford: Elsevier.
Eaton, DL, Daroff, RB, Autrup, H, Bridges, J, Buffler, P,
Costa, LG, . . . Spencer, PS. (2008a). Review of the
Toxicology of Chlorpyrifos with an Emphasis on Human
Exposure and Neurodevelopment. Crit Rev Toxicol, 38
Suppl 2, 1-125.
Eaton, DL, Daroff, RB, Autrup, H, Bridges, J, Buffler, P,
Costa, LG, . . . Spencer, PS. (2008b). Review of the
Toxicology of Chlorpyrifos with an Emphasis on Human
Exposure and Neurodevelopment. Crit Rev Toxicol, 38
Suppl 2, 1-125.
EFSA. (2006). Conclusion Regarding the Peer Review of
the Pesticide Risk Assessment of the Active Substance
Carbaryl. EFSA Journal.
EFSA. (2011). Statement on Oral Toxicity of Endosulfan In
Fish: EFSA Panel on Contaminants in the Food Chain
(CONTAM). EFSA Journal, 9(4), 2131-2157.
EFSA. (2014). Conclusion on the Peer Review of the
Pesticide Risk Assessment of the Active Substance
Chlorpyrifos. European Food Safety Authority Journal,
12, 3640-3674.
EFSA. (2016). European Food Safety Authority
Conclusion on the Peer Review of the Pesticide Risk
Assessment of the Active Substance Propyzamide
EFSA Journal, 14 (8), 25.
EFSA. (2017). Investigation into Experimental
Toxicological Properties of Plant Protection Products
Having a Potential Link to Parkinson's Disease and
Childhood Leukaemia. EFSA Journal, 15(3), 1-325.
Elcombe, CR, Peffer, RC, Wolf, DC, Bailey, J, Bars, R,
Bell, D, . . . Lake, BG. (2014). Mode of action and
human relevance analysis for nuclear receptor-
mediated liver toxicity: A Case Studywith Phenobarbital
as a Model Constitutive Androstane Receptor (CAR)
Activator. . Crit Rev Toxicol, 44, 64-82.
Ellison, CA, Smith, JN, Lein, PJ, Olson, JR. (2011).
Pharmacokinetics and Pharmacodynamics of
Chlorpyrifos in Adult Male Long-Evans Rats Following
Repeated Subcutaneous Exposure to Chlorpyrifos.
Toxicology, 287(1), 137-144.
Faucette, SR, Zhang, T-C, Moore, R, Sueyoshi, T,
Omiecinski, CJ, LeCluyse, EL, . . . Wang, H. (2007).
Relative Activation of Human Pregnane X Receptor
versus Constitutive Androstane Receptor Defines
Distinct Classes of CYP2B6 and CYP3A4 Inducers. J
Pharmacol Exp Ther, 320(1), 72.
Ffrench-Constant, RH, Anthony, N, Aronstein, K,
Rocheleau, T, Stilwell, G. (2000). Cyclodiene
Insecticide Resistance: From Molecular to Population
Genetics. Annu Rev Entomol, 45, 449-466.
Filer, D. (2019). The New ToxCast Analysis. Available
through filer.dayne@epa.gov
Filer, DL, Kothiya, P, Setzer, RW, Judson, R, Martin, MT.
(2017). tcpl: the ToxCast Pipeline for High-Throughput
Screening Data. Bioinformatics, 33(4), 618-620.
doi:0.1093/bioinformatics/btw680
19. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
Foxenberg, RJ, Ellison, CA, Knaak, JB, Ma, C, Olson, JR.
(2011). Cytochrome P450-Specific Human PBPK/PD
Models for the Organophosphorus Pesticides:
Chlorpyrifos and Parathion. Toxicology, 285(1-2), 57-
66.
Foxenberg, RJ, McGarrigle, BP, Knaak, JB, Kostyniak, PJ,
Olson, JR. (2007). Human Hepatic Cytochrome P450-
Specific Metabolism of Parathion and Chlorpyrifos.
Drug Metabol Dispos 35(2), 189.
Frank, CL, Brown, JP, Wallace, KB, Wambaugh, JF, Shah,
I, Shafer, TJ. (2018). Defining toxicological tipping
points in neuronal network development. Toxicol Appl
Pharmacol, 354, 81-93.
Friedman, KP, Gagne, M, Loo, LH, Karamertzanis, P,
Netzeva, T, Sobanski, T, . . . Thomas, RS. (2019). Utility
of In Vitro Bioactivity as a Lower Bound Estimate of In
Vivo Adverse Effect Levels and in Risk-Based
Prioritization. Toxicol Sci.
https://academic.oup.com/toxsci/advance-
article/doi/10.1093/toxsci/kfz201/5571376 Accessed
Sept 18, 2019.
Gant, DB, Eldefrawi, ME, Eldefrawi, AT. (1987).
Cyclodiene Insecticides Inhibit GABAA Receptor-
Regulated Chloride Transport. Toxicol Appl Pharmacol,
88(3), 313-321.
Gaweska, H, Fitzpatrick, PF. (2011). Structures and
Mechanism of the Monoamine Oxidase Family.
Biomolecular concepts, 2(5), 365-377.
doi:10.1515/BMC.2011.030
Holick, MF. (2005). Stay Tuned to PXR: an Orphan Actor
that may not be D-structive only to Bone. J Clinical
Investigations, 115(1), 32-34.
Hood, A, Allen, ML, Liu, YP, Liu, J, Klaassen, CD. (2003).
Induction of T4 UDP-GT Activity, Serum Thyroid
Stimulating Hormone, and Thyroid Follicular Cell
Proliferation in Mice Treated with Microsomal Enzyme
Inducers. Toxicol Appl Pharm, 188, 6-13.
ICE. (2020). ICE IVIVE Tool.
https://ice.ntp.niehs.nih.gov/downloads/UserGuides/IV
IVE_userguide.pdf. Accessed Sept 27, 2020.
Ihunnah, CA, Jiang, M, Xie, W. (2011). Nuclear Receptor
PXR, Transcriptional Circuits and Metabolic
Relevance. Biochim et Biophys Acta (BBA) - Mol Basis
Dis, 1812(8), 956-963.
Jacob, TC, Moss, SJ, Jurd, R. (2008). GABAA Receptor
Trafficking and Its Role in The Dynamic Modulation of
Neuronal Inhibition. Nature Rev Neurosci, 9(5), 331-
343.
Jia, Z, Misra, HP. (2007). Developmental Exposure to
Pesticides Zineb and/or Endosulfan Renders the
Nigrostriatal Dopamine System More Susceptible to
These Environmental Chemicals Later in Life.
Neurotoxicol, 28(4), 727-735.
Judson, R, Houck, K, Martin, M, Richard, AM, Knudsen,
TB, Shah, I, . . . Thomas, RS. (2016). Analysis of the
Effects of Cell Stress and Cytotoxicity on In Vitro Assay
Activity Across a Diverse Chemical and Assay Space.
Toxicol Sci, 152(2), 323-339.
Judson, R, Kavlock, RJ, Setzer, RW, Hubal, EAC, Martin,
MT, Knudsen, TB, Dix, DJ. (2011). Estimating Toxicity-
Related Biological Pathway Altering Doses for High-
Throughput Chemical Risk Assessment. Chem Res
Toxicol, 24, 451–462.
Kamijima, M, Casida, JE. (2000). Regional Modification of
[3H]Ethynylbicycloorthobenzoate Binding in Mouse
Brain GABAA Receptor by Endosulfan, Fipronil, and
Avermectin B1a. Toxicol Appl Pharmacol, 163(2), 188-
194.
Khor, VK, Dhir, RD, Yin, X, Ahima, RS, Song, W-C. (2010).
Estrogen Sulfotransferase Regulates Body Fat and
Glucose Homeostasis in Female Mice. Amer J Physiol-
Endocrinol Metabol, 299(4), E657-E664.
Kleinstreuer, NC, Ceger, P, Watt, ED, Martin, M, Houck,
K, Browne, P, . . . Judson, R. (2017). Development and
Validation of a Computational Model for Androgen
Receptor Activity. Chem Res Toxicol, 30(4), 946-964.
Kleinstreuer, NC, Dix, DJ, Houck, KA, Kavlock, RJ,
Knudsen, TB, Martin, MT, . . . Judson, R. (2013). In
Vitro Perturbations of Targets in Cancer Hallmark
Processes Predict Rodent Chemical Carcinogenesis.
Toxicol Sci, 131(1), 40-55.
Kleinstreuer, NC, Knudsen, TB. (2011). Predictive
modeling and computational toxicology. Hood, RD.
(Ed.) 3rd ed. Vol. Chapter 23. Boca Raton, FL: CRC
Press Taylor and Francis Group.
Kliewer, SA. (2003). The Nuclear Pregnane X Receptor
Regulates Xenobiotic Detoxification. J Nutrition, 133(7),
2444S-2447S.
Knudsen, T, Martin, M, Chandler, K, Kleinstreuer, NC,
Judson, R., Sipes, N. (2013). Predictive Models and
Computational Toxicology. In Barrow, PC. (Ed.),
Teratogenicity Testing: Methods and Protocols (pp.
343-374). Totowa, NJ: Humana Press.
Knudsen, TB, Houck, KA, Sipes, NS, Singh, AV, Judson,
RS, Martin, MT, Kavlock, RJ. (2011). Activity Profiles of
309 ToxCast™ Chemicals Evaluated Across 292
Biochemical Targets. Toxicology, 282(1), 1-15.
doi:https://doi.org/10.1016/j.tox.2010.12.010
Ko, HP, Okino, ST, Ma, Q, Whitlock, JP, Jr. (1996). Dioxin-
induced CYP1A1 Transcription In Vivo: The Aromatic
Hydrocarbon Receptor Mediates Transactivation,
Enhancer-Promoter Communication, and Changes in
Chromatin Structure. Mol Cell Biol, 16(1), 430-436.
doi:10.1128/mcb.16.1.430
Lakshmanan, PS, Eeckhaut, T, Van Huylenbroeck, J, Van
Bockstaele, E. (2013). Micronucleation by Mitosis
Inhibitors in Developing Microspores of Spathiphyllum
wallisii Regel. Plant Cell Rep, 32(3), 369-377.
Lan, A, Stein, D, Portillo, M, Toiber, D, Kofman, O. (2019).
Impaired Innate and Conditioned Social Behavior in
Adult C57Bl6/J Mice Prenatally Exposed to
Chlorpyrifos. Behav Brain Funct, 15(1), 2.
Lawrence, LJ, Casida, JE. (1984). Interactions of Lindane,
Toxaphene and Cyclodienes with Brain-Specific T-
Butylbicyclophosphorothionate Receptor. Life Sci,
35(2), 171-178.
20. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
LeBaron, MJ, Rasoulpour, RJ, Gollapudi, BB, Sura, R,
Kan, HL, Schisler, MR, . . . Eisenbrandt, DL. (2014).
Characterization of Nuclear Receptor-Mediated Murine
Hepatocarcinogenesis of the Herbicide Pronamide and
Its Human Relevance. Toxicol Sci, 142(1), 74-92.
Lee, I, Eriksson, P, Fredriksson, A, Buratovic, S, Viberg,
H. (2015). Developmental Neurotoxic Effects of Two
Pesticides: Behavior and Biomolecular Studies on
Chlorpyrifos and Carbaryl. Toxicol Appl Pharmacol,
288(3), 429-438.
Li, J, Wang, L, Guo, H, Shi, L, Zhang, K, Tang, M, . . . Xia,
K. (2017). Targeted Sequencing and Functional
Analysis Reveal Brain-Size-Related Genes and Their
Networks in Autism Spectrum Disorders. Mol
Psychiatry, 22(9), 1282-1290.
Lin, YS, Yasuda, K, Assem, M, Cline, C, Barber, J, Li, C-
W, . . . Schuetz, EG. (2009). The Major Human
Pregnane X Receptor (PXR) Splice Variant, PXR.2,
Exhibits Significantly Diminished Ligand-Activated
Transcriptional Regulation. Drug Metabol Dispos,
37(6), 1295.
Marty, MS, Andrus, A, Thomas, J. (2012). Pronamide:
Pubertal Development and Thyroid Function In Intact
Juvenile Peripubertal Male Crl:CD(SD) Rats.
https://www.epa.gov/endocrine-disruption/endocrine-
disruptor-screening-program-tier-1-screening-
determinations-and. Accessed March 15, 2020.
Marvel, SW, To, KT, Grimm, FA, Wright, A, Rusyn, I, Reif,
DM. (2018). ToxPi Graphical User Interface 2.0:
Dynamic Exploration, Visualization, and Sharing of
Integrated Data Models. BMC Bioinformatics, 19(1), 80.
Mattsson, JL, Maurissen, JP, Nolan, RJ, Brzak, KA.
(2000). Lack of Differential Sensitivity to Cholinesterase
Inhibition in Fetuses and Neonates Compared to Dams
Treated Perinatally with Chlorpyrifos. Toxicol Sci, 53(2),
438-446.
Michalik, L, Auwerx, J, Berger, JP, Chatterjee, VK, Glass,
CK, Gonzalez, FJ, . . . Wahli, W. (2006). International
Union of Pharmacology. LXI. Peroxisome Proliferator-
Activated Receptors. Pharmacol Rev, 58(4), 726-741.
Milesi, MM, Alarcón, R, Ramos, JG, Muñoz-de-Toro, M,
Luque, EH, Varayoud, J. (2015). Neonatal Exposure to
Low Doses of Endosulfan Induces Implantation Failure
and Disrupts Uterine Functional Differentiation at the
Pre-Implantation Period in Rats. Mol Cellular
Endocrinol, 401, 248-259.
Moser, VC, McDaniel, KL, Phillips, PM, Lowit, AB. (2010).
Time-Course, Dose-Response, and Age Comparative
Sensitivity of N-Methyl Carbamates in Rats. Toxicol Sci,
114(1), 113-123.
Moser, VC, Padilla, S. (1998). Age- and Gender-Related
Differences in the Time Course of Behavioral and
Biochemical Effects Produced by Oral Chlorpyrifos in
Rats. Toxicol Appl Pharmacol, 149(1), 107-119.
Mutch, E, Williams, FM. (2006). Diazinon, Chlorpyrifos and
Parathion are Metabolised by Multiple Cytochromes
P450 In Human Liver. Toxicology, 224(1), 22-32.
NAS. (2017). Using 21st Century Science to Improve Risk-
Related Evaluations. National Academies Press (US);
Copyright 2017 by the National Academy of Sciences.
All rights reserved.
Nong, A, Tan, Y-M., Krolski, ME, Wang, J, Lunchick, C,
Conolly, RB, Clewell III, HJ. (2008). Bayesian
Calibration of a Physiologically Based
Pharmacokinetic/Pharmacodynamic Model of Carbaryl
Cholinesterase Inhibition. J Toxicol Environ Health, 71,
1363-1381.
NPIC. (2003). Carbaryl. National Pesticide Information
Center. http://npic.orst.edu/ingred/aifact.html.
Accessed Sept 27, 2020.
NTP. (2020). National Toxicology Program Integrated
Chemical Environment. https://ice.ntp.niehs.nih.gov/
downloads/UserGuides/IVIVE_userguide.pdf.
Accessed Sept 27, 2020.
Nyffeler, J, Willis, C, Lougee, R, Richard, A, Paul-
Friedman, K, Harrill, JA. (2020). Bioactivity Screening
of Environmental Chemicals Using Imaging-Based
High-Throughput Phenotypic Profiling. Toxicol Appl
Pharmacol, 389, 114876.
Papineni, S, Marty, MS, Rasoulpour, RJ, LeBaron, MJ,
Pottenger, LH, Eisenbrandt, DL. (2015). Mode of Action
and Human Relevance of Pronamide-Induced Rat
Thyroid Tumors. Regul Toxicol Pharmacol, 71(3), 541-
551.
Park, JH, Ko, J, Park, YS, Park, J, Hwang, J, Koh, HC.
(2017). Clearance of Damaged Mitochondria Through
PINK1 Stabilization by JNK and ERK MAPK Signaling
in Chlorpyrifos-Treated Neuroblastoma Cells. Mol
Neurobiol, 54(3), 1844-1857.
Pavek, P, Dvorak, Z. (2008). Xenobiotic-Induced
Transcriptional Regulation of Xenobiotic Metabolizing
Enzymes of the Cytochrome P450 Superfamily in
Human Extrahepatic Tissues. Curr Drug Metabol, 9,
129-143.
Pearce, RG, Setzer, RW, Strope, CL, Sipes, NS,
Wambaugh, JF. (2017). httk: R Package for High-
Throughput Toxicokinetics. J Statistical Software,
79(4), 1-26.
Pham, N, Iyer, S, Hackett, E, Lock, BH, Sandy, M, Zeise,
L, . . . Marty, MS. (2016). Using ToxCast to Explore
Chemical Activities and Hazard Traits: A Case Study
With Ortho-Phthalates. Toxicol Sci, 151(2), 286-301.
Piper, BJ, Fraiman, JB, Meyer, JS. (2005). Repeated
MDMA (“Ecstasy”) Exposure in Adolescent Male Rats
Alters Temperature Regulation, Spontaneous Motor
Activity, Attention, and Serotonin Transporter Binding.
Develop Psychobiol, 47(2), 145-157.
PMRA. (2009). Propyzamide. Health Canada Pest
Management Regulatory Agency, 23, 2009;
http://publications.gc.ca/collections/collection_2010/arl
a-pmra/H113-28-2009-12-1-eng.pdf. Accessed Sept
27, 2020.
PMRA. (2016). Re-evaluation Decision RVD2016-02,
Carbaryl. Pest Management Regulatory Agency,
Health Canada, 31 March 2016; https://www.canada.
ca/en/health-canada/services/consumer-product-
safety/reports-publications/pesticides-pest-
management/decisions-updates/reevaluation-
21. Open Access ToxCast/Tox21, Toxicological Priority Index (ToxPi) and Integrated Chemical Environment (ICE) Models Rank and Predict Acute Pesticide Toxicity: A Case Study
decision/2016/document-carbaryl-rvd2016-
02.html#a4. Accessed Sept 27, 2020.
PMRA. (2019). Proposed Re-evaluation Decision
PRVD2019-05, Chlorpyrifos and Its Associated End-
use Products: Updated Environmental Risk
Assessment. Pest Management Regulatory Agency,
Health Canada, 31 May 2019;
https://www.canada.ca/en/health-
canada/services/consumer-product-safety/pesticides-
pest-management/public/consultations/proposed-re-
evaluation-
decisions/2019/chlorpyrifos/document.html#a1.
Accessed Sept 27, 2020.
Rasoulpour, RJ, Andrus, AK, Marty, MS, Zhang, F,
Thomas, J, LeBaron, MJ., . . . Eisenbrandt, DL. (2015).
Pronamide: Human Relevance of Liver-Mediated Rat
Leydig Cell Tumors. Reg Toxicol Pharmacol, 72, 394-
404.
Ratra, GS, Kamita, SG, Casida, JE. (2001). Role of Human
GABAA Receptor β3 Subunit in Insecticide Toxicity.
Toxicol Appl Pharmacol, 172(3), 233-240.
Rauh, VA, Garcia, WE, Whyatt, RM, Horton, MK, Barr,
DB,Louis, ED. (2015). Prenatal Exposure to the
Organophosphate Pesticide Chlorpyrifos and
Childhood Tremor. Neurotoxicol, 51:80-86.
Rauh, VA, Perera, FP, Horton, MK, Whyatt, RM, Bansal,
R, Hao, X, . . . Peterson, BS. (2012). Brain Anomalies
in Children Exposed Prenatally to a Common
Organophosphate Pesticide. Proc Natl Acad Sci U S A,
109(20), 7871-7876.
Reif, DM, Martin, MT, Tan, SW, Houck, KA, Judson, R,
Richard, AM. (2010). Endocrine Profiling and
Prioritization of Environmental Chemicals Using
ToxCast Data. Environ Health Perspect, 118, 1714-
1720.
Reif, DM, Sypa, M, Lock, EF, Wright, FA, Wilson, A,
Cathey, T, . . . Rusyn, I. (2013). ToxPi GUI: an
Interactive Visualization Tool for Transparent
Integration of Data From Diverse Sources of Evidence.
Bioinformatics, 29(3), 402-403.
Ryan, N. (2017). A User’s Guide for Accessing and
Interpreting ToxCast™ Data.
https://lri.americanchemistry.com/Users-Guide-for-
Accessing-and-Interpreting-ToxCast-Data.pdf.
Accessed April, 2020
Saili, KS, Antonijevic, T, Zurlinden, TJ, Shah, I,
Deisenroth, C, Knudsen, TB. (2020). Molecular
Characterization of a Toxicological Tipping Point During
Human Stem Cell Differentiation. Repro Toxicol, 91, 1-
13.
Sams, C, Cocker, J., Lennard, MS. (2004).
Biotransformation of Chlorpyrifos and Diazinon by
Human Liver Microsomes and Recombinant Human
Cytochrome P450s (CYP). Xenobiotica, 34(10), 861-
873.
Seth, PK, Saidi, NF, Agrawal, AK, Anand, M. (1986).
Neurotoxicity of Endosulfan in Young and Adult Rats.
Neurotoxicol, 7(2), 623-635.
Shah, I., Setzer, R. W., Jack, J., Houck, K. A., Judson, R.
S., Knudsen, T. B., . . . Kavlock, R. J. (2016). Using
ToxCast™ Data to Reconstruct Dynamic Cell State
Trajectories and Estimate Toxicological Points of
Departure. Environ Health Perspect, 124(7), 910-919.
Silva, MH. (2020). Effects of low-dose chlorpyrifos on
neurobehavior and potential mechanisms: A review of
studies in rodents, zebrafish, and Caenorhabditis
elegans. Birth Defects Res, 112(6), 445-479.
Silva, MH., Gammon, D. (2009). An Assessment of the
Developmental, Reproductive, and Neurotoxicity of
Endosulfan. Birth Defects Res (Part B), 86(1), 1-28.
Silva, MH, Pham, N, Lewis, C, Iyer, S, Kwok, E, Solomon,
G, Zeise, L. (2015). A Comparison of ToxCast Test
Results with In Vivo and Other In Vitro Endpoints for
Neuro, Endocrine, and Developmental Toxicities: A
Case Study Using Endosulfan and Methidathion. Birth
Defects Res (Part B), 104, 71-89.
Singh, NS, Lawana, V, Luo, J, Phong, P, Abdalla, A,
Palanisamy, B, . . . Kanthasamy, A. (2018).
Organophosphate Pesticide Chlorpyrifos Impairs
STAT1 Signaling to Induce Dopaminergic
Neurotoxicity: Implications for Mitochondria Mediated
Oxidative Stress Signaling Events. Neurobio Disease,
117, 82-113.
Sipes, NS, Wambaugh, JF, Pearce, RG, Auerbach, SS,
Wetmore, BA, Hsieh, J-H, . . . Ferguson, SS. (2017). An
Intuitive Approach for Predicting Potential Human
Health Risk with the Tox21 10k Library. Environ. Sci.
Technol., 51(18), 10786-10796.
Slotkin, TA, Seidler, FJ. (2010). Diverse Neurotoxicants
Converge on Gene Expression for Neuropeptides and
Their Receptors in an In Vitro Model Of
Neurodifferentiation: Effects Of Chlorpyrifos, Diazinon,
Dieldrin and Divalent Nickel in PC12 Cells. Brain Res,
1353, 36-52.
Sugatani, J, Nishitani, S, Yamakawa, K, Yoshinari, K,
Sueyoshi, T, Negishi, M, Miwa, M. (2004).
Transcriptional Regulation of Human UGT1A1 Gene
Expression: Activated Glucocorticoid Receptor
Enhances Constitutive Androstane Receptor/
Pregnane X Receptor-Mediated UDP-
Glucuronosyltransferase 1A1 Regulation with
Glucocorticoid Receptor-Interacting Protein 1.
Molecular Pharmacol, 67(3), 845-855.
Tan, YM, Liao, KH, Clewell, HJ, III. (2007). Reverse
dosimetry: Interpreting Trihalomethanes Biomonitoring
Data Using Physiologically Based Pharmacokinetic
Modeling. J Expo Sci Environ Epidemiol, 17, 591–603.
Tang, J, Cao, Y, Rose, RL, Brimfield, AA, Dai, D,
Goldstein, JA, Hodgson, E. (2001). Metabolism of
Chlorpyrifos by Human Cytochrome P450 Isoforms and
Human, Mouse, and Rat Liver Microsomes. Drug
Metab Dispo, 29(9), 1201-1204.
Tang, J, Cao, Y, Rose, RL, Hodgson, E. (2002). In vitro
Metabolism of Carbaryl by Human Cytochrome P450
and Its Inhibition by Chlorpyrifos. Chemico-Biological
Interactions, 141(3), 229-241.