Transcript Computational Toxicology in EPA’s Office of Research and

Opportunities for Strategic Planning using
Systems Models
or
A Biologist’s View of using Computers in
Risk Assessment
Gary Ankley
Second Annual McKim Conference
September 25-27, 2007
Fate
Exposure
Effects Continuum
Effect/Outcome
Source
Environmental
Concentration
Exposure
Biological
Event
Dose
Ultimate goal should be linked, predictive
models for each aspect of the continuum
Biological
Event
Source
Biological
Event
Environmental
Concentration
Exposure
Toxicant
Chemical
Reactivity
Profiles
Macro-Molecular
Interactions
Receptor/Ligand
Interaction
DNA Binding
Protein Oxidation
QSAR Models
Effect/Outcome
Dose
Cellular
Responses
Organ
Responses
Gene Activation
Altered
Physiology
Protein Production
Altered Signaling
Protein Depletion
Disrupted
Homeostasis
Altered Tissue
Development
or Function
Individual
Responses
Population
Responses
Lethality
Impaired
Development
Impaired
Reproduction
Cancer
Structure
Recruitment
Extinction
QSARs for Regulatory Decision-Making:
Critical Attributes
• Transparent, transferable
• Reflects toxicity mechanism of concern
• Relatable to possible adverse outcome(s)
relevant to risk assessment
• Wide acceptance by all sectors in
regulatory community
Fathead Minnow Narcosis Toxicity/Kow Plot
1
0
Log Molar LC50
-1
-2
-3
-4
-5
-6
-7
-8
-2
-1
0
1
2
Log Kow
3
4
5
6
QSARs for Predicting
Narcosis Toxicity
• Transparent, transferable model based on
well-defined biological response
• Reflects basis of toxicity (membrane
penetration/disruption)
• Relatable to adverse outcome highly
relevant to risk assessment
• Widely used as a basis for regulatory
decision-making/research
Chemical Binding to Estrogen Receptor (ER)
0.1
LogRBA
0.01
0.001
0.0001
0.00001
0.000001
0
1
2
3
4
LogKow
5
6
7
8
QSAR for Predicting ER Binding
• Transparent, transferable model potentially
indicative of chronic response(s)
• Mechanistic reflection of an important point
of control within the vertebrate HPG axis
• ER perturbation could produce adverse
effects relevant to risk assessment
• Translation to regulatory decision-making
remains challenging
Historic Challenges to Implementation of
QSARs for Chronic Effects
• Many attempts not based on mechanistic
understanding of biology/initiating events
(e.g., derived from regression relationships)
• Reflect only limited (usually one) point of
control within biological axis/response of
concern
• Apical outcomes uncertain due to complexity
of the toxicity pathways under consideration
(e.g., multiple biological outcomes, feedback
controls, compensation)
Biological
Event
Toxicant
Chemical
Reactivity
Profiles
Macro-Molecular
Interactions
Receptor/Ligand
Interaction
DNA Binding
Protein Oxidation
QSAR
Models
Cellular
Responses
Organ
Responses
Gene Activation
Altered
Physiology
Protein Production
Altered Signaling
Protein Depletion
Disrupted
Homeostasis
Altered Tissue
Development
or Function
Systems Models
Effect/Outcome
Individual
Responses
Population
Responses
Lethality
Impaired
Development
Impaired
Reproduction
Cancer
Structure
Recruitment
Extinction
1
2
3
4
5
6
7
Brain
8
9
10
11
12
13
14
15
16
18
19
20
Figure Key
state transition
a
Catalysis (including
Liver
activation)
transcriptional activation
b
translational activation
transcription inhibition
c
dissociation
association
d
Genes
mRNA
e
protein
activated
f
protein
g
receptor
Simple
h
molecule
Phenotype
i
j
Pituitary
k
Blood
l
m
n
o
p
Ovary
q
r
Graphical
Systems
Model for
Small Fish
HPG Axis
Systems Model Overview
• Developed for small fish, but due to conserved
nature of vertebrate HPG axis, has broad
applicability
• Reflects interaction of >105 proteins and 40 simple
molecules, regulation of about 25 genes and >300
reactions within six tissues (with multiple cell types)
• Multiple intended uses
 Organizing/understanding genomic data
 Identifying key points of control within the HPG
axis
 Relating molecular initiating events to adverse
outcomes
Chemical Probes
Compartment
GABA
Dopamine
Brain
?
?
?
PACAP
Pituitary
GnRH
Neuronal
System
GnRH
NPY
GABAA
R
D2 R
GABAB
R
Y2 R
GnRH
R
PAC1 R
2
Muscimol (+)
3
Apomorphine (+)
4
Haloperidol (-)
Y1 R
Gonadotroph
Activin R
Fipronil (-)
Y2 R
Follistatin
Activin
1
D1 R
D2 R
GPa
FSHb
Blood
Circulating LDL, HDL
LHb
Circulating LH, FSH
LDL R
LH R
5
Trilostane (-)
6
Ketoconazole (-)
FSH R
HDL R
Cholesterol
Outer mitochondrial
membrane
StAR
Inner mitochondrial
membrane
Gonad
Activin
(Generalized, gonadal,
steroidogenic cell)
Inhibin
P450scc
pregnenolone
3bHSD
17α-hydroxyprogesterone
Fadrozole (-)
8
Prochloraz (-,-)
progesterone
P450c17
20βHSD
7
androstenedione
17βHSD
17α,20β-P (MIS)
testosterone
P450arom
9
P45011β.
Vinclozolin (-)
11βHSD
11-ketotestosterone
Blood
Androgen / Estrogen
Responsive Tissues
(e.g. liver, fatpad, gonads)
Circulating Sex Steroids / Steroid
Hormone Binding Globlulin
ER
AR
10
Flutamide (-)
11
β-Trenbolone (+)
12
Ethynyl estradiol (+)
estradiol
Types of Genomic Data
Proteomics
Transcriptomics
Fathead Minnow Microarray
Intens. [a.u.]
Peptide Mass Fingerprinting
x10 4
1091.620
1.5
1799.879
1347.669
1.0
2143.156
1615.722
890.612
Representative protein expression
profile in testes of control zebrafish
1214.658
0.5
1504.667
1978.039
2460.281
2801.340
0.0
Data from EPA/ EcoArray© CRADA
1000
1500
2000
2500
3000
m /z
Data from EPA-Cincinnati
Metabolomics
Fathead Minnow Liver NMR Scan
Fathead Minnow (male)
Data from EPA-Athens
Some General Observations to Date
• Despite the large number of genomic endpoints
examined in fathead minnow and zebrafish studies
with probe chemicals to date, only a relative
handful related to HPG axis function are affected
(although many changes are observed in other
“non-HPG” related parameters)
• Chemical probes with different MOA within the HPG
axis often affect the same genes, suggesting
common nodes of perturbation and/or control (e.g.,
20bHSD, FSHb, CYP19A)
Compartment
GABA
Common
Responsive Genes
in Fish HPG Axis
Dopamine
Brain
?
?
?
PACAP
Pituitary
GnRH
Neuronal
System
GnR
H
D1 R
Y2 R
NPY
GAB
AA R
D2 R
GAB
AB R
Y2 R
Follistatin
GnRH
R
PAC1
R
Y1 R
D2 R
Activi
n
Gonadotroph
Activin
R
GPa
FSHb
Circulating LDL,
HDL
Blood
LHb
Circulating LH, FSH
LDL R
LH R
FSH R
HDL R
Gonad
Outer mitochondrial
membrane
(Generalized, gonad,
steroidogenic cells and
oocytes)
Cholestero
l
Activin
StAR
Inner mitochondrial
membrane
Inhibin
P450scc
pregnenolone
(oocytes)
3bHSD
17α-hydroxyprogesterone
progesterone
P450c17
20βHSD
androstenedion
e
17βHSD
17α,20β-P
(MIS)
P450arom
testosterone
AR
P45011β
(steroidogenic cells)
11βHSD
11-ketotestosterone
Blood
Circulating Sex Steroids / Steroid
Hormone Binding Globlulin
Androgen / Estrogen
Responsive Tissues
(e.g. liver, fatpad, gonads)
estradiol
Estradiol
Vtg
+
ER
General Observations cont’.
• The further “up” the axis in terms of perturbation,
the less profound the apical effects (e.g.,
agonists/antagonists of the GABA and dopamine
receptors seem to produce less pronounced effects
than inhibitors of terminal steroidogenic enzymes
and ER, AR agonists/antagonists)
 Differences in innate chemical potency?
 Differences in specificity of interaction with HPG
vs. non-HPG function?
 Opportunity for biological
adaptation/compensation within the HPG axis?
Ketoconazole
O
H3C
N
N
N
O N Cl
O
H
• Model conazole fungicide
Cl
• Reversible, competive inhibitor of cytochrome
P450 (CYP) activities
• Reduces testosterone production in mammals
Mean Cumulat iv e Number of Eggs Spawned/ Female
Effect of Ketoconazole on Fathead
Minnow Reproduction
350
Ket oc onaz ole ( µg/ L)
300
Cont r ol
6
250
25
100
200
400
150
100
*
50
*
0
0
2
4
6
8
10
12
Exposur e ( d)
14
16
18
20
Effect of Ketoconazole on Ex vivo Steroid
Production in Fathead Minnows
12
A
♀a
AB
10
T (ng/ml)/g
8
6
B
4
B
B
2
0
600
♂b
A
T (ng/ml)/g
500
400
AB
300
AB
AB
200
B
100
0
0
6
25
Ketoconazole (µg/L)
100
400
Effects of Ketoconazole on Fathead Minnow
In vivo Steroid Levels
Male
Female
0.60
0.50
Estradiol (ng/ml)
8
6
4
2
0.40
0.30
0.20
0.10
0
0.00
10
15
Testosterone (ng/ml)
Testosterone (ng/ml)
Estradiol (ng/ml)
10
8
6
4
2
0
12
9
6
3
0
0
6
25
100
Ketoconazole (ug/L)
400
0
6
25
100
Ketoconazole (ug/L)
400
Ketoconazole Effects on Male Gonad
3
b
b
GSI
2
a
a
6
25
Ketoconazole (ug/L)
a
1
0
0
100
400
Proliferation of Interstitial Cells Involved in Steroid Synthesis
A, B = Controls; C= 6 g/L; D= 400 g/L
Male Gonad Module from Systems Model
Steroidogenic Compensation to Ketoconazole
Cholesterol
CYP11A
CYP17
(hydroxylase)
Pregnenolone
CYP17
(lyase)
17a-OH-Pregnenolone
3b-HSD
Progesterone
DHEA
3b-HSD
3b-HSD
CYP17
(hydroxylase)
CYP21
17a-OH-Progesterone
CYP21
CYP17
(lyase)
20b-HSD
CYP19
Androstenedione
17b-HSD
estrone
17b-HSD
CYP19
11-deoxycorticosterone
CYP11B1
Testosterone
11-deoxycortisol
corticosterone
CYP11B2
CYP11B2
17α20β-dihydroxy-4pregnen-3-one
CYP11B1
11β-OHTestosterone
11βHSD
aldosterone
cortisol
11-Ketotestosterone
17b-estradiol
Relating Molecular Alterations to Adverse
Outcomes
• Critical both to use of genomic data and mechanistic
(QSAR) predictions
• Toxicity pathway concept essential to establishing
linkage across biological levels of organization, but this
can only be successful if pathway is considered as
network/web rather than linear chain of events
 Feedback/homeostatic processes can modulate
biological responses
 Single initiating event can elicit multiple responses
 Multiple initiating events (mechanisms) may trigger
toxicity via same mode of action
• Systems models facilitate consideration of pathway
complexity
Key Nodes in Toxicity Pathways:
Initiation versus Response
• Molecular initiating event (e.g., receptor
activation, enzyme inhibition) is logical focus of
mechanistic QSAR models
• But, this is not necessarily the key “choke point”
modulating adverse apical responses
• Need understanding/depiction of toxicity
pathway to discern between the two different
types of nodes and relate them to one another
Key Nodes in Toxicity Pathways:
Illustration from the HPG Axis in Fish
• Vitellogenein (vtg), egg yolk protein, is produced
normally by oviparous female vertebrates in
response to stimulation of the ER by 17β-estradiol
• Commonly used exposure biomarker in males for
exposure to exogenous estrogens
• Effective production of vtg in females critical to
successful egg production
• Vtg production in females can hypothetically be
decreased via several discreet mechanisms
within the HPG axis
Chemical Inhibitors of VTG Synthesis
in the Fathead Minnow
• Fenarimol: conazole fungicide with multiple
hypothesized MOA, including ER antagonism
• Prochloraz: conazole fungicide which inhibits
several CYPs involved in steroid production
(CYP17, CYP19)
• Fadrozole: specific pharmaceutical inhibitor of
CYP19
• 17β-trenbolone: anabolic androgen that causes
feedback inhibition of steroid production
• 17α-trenbolone: anabolic androgen metabolite that
causes feedback inhibition of steroid production
Compartment
GABA
Molecular
Mechanisms of
Inhibition of VTG
Production
Dopamine
Brain
?
?
?
PACAP
Pituitary
GnRH
Neuronal
System
GnR
H
D1 R
Y2 R
NPY
GAB
AA R
D2 R
GAB
AB R
Y2 R
Follistatin
GnRH
R
PAC1
R
Y1 R
D2 R
Activi
n
Gonadotroph
Activin
R
GPa
FSHb
Circulating LDL,
HDL
Blood
LHb
Circulating LH, FSH
LDL R
LH R
FSH R
HDL R
Gonad
Outer mitochondrial
membrane
(Generalized, gonad,
steroidogenic cells and
oocytes)
Cholestero
l
Activin
StAR
Inner mitochondrial
membrane
Inhibin
P450scc
pregnenolone
(oocytes)
3bHSD
17α-hydroxyprogesterone
progesterone
Fadrozole
P450c17
20βHSD
androstenedion
e
17βHSD
17α,20β-P
(MIS)
Prochloraz
P450arom
testosterone
P45011β
(steroidogenic cells)
Blood
estradiol
Circulating Sex Steroids / Steroid
Hormone Binding Globlulin
Androgen / Estrogen
Responsive Tissues
(e.g. liver, fatpad, gonads)
α trenbolone
11βHSD
11-ketotestosterone
AR
β trenbolone
Estradiol
Vtg
+
ER
Fenarimol
Effects of Aromatase Inhibition on
Reproduction in the Fathead Minnow
150
N
10
N
a
Aromatase Activity
(fmol/mg-1 hr-1)
2
10
6
Male
Female
Control
50
*
*
*
4
Fadrozole
75
c
2
c
CN
0
-20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0
2 4
6
8 10 12 14 16 18 20
0
Exposure (d)
0
6
E2 (ng/ml)
50
Fadrozole (µg / L)
8
4
*
2
*
0
Vtg (mg/ml)
Cumulative Number of Eggs
(Thousands)
Fadrozole (ug/L)
8
b
20
10
*
*
0
Control
2
10
Fadrozole (µg/l)
*
50
Linking Molecular Responses to Apical
Effects: VTG and Fecundity
Chemical
Fathead Minnow Fecundity vs Vtg
Exposure Concentrations
1
0.005µg/l, 0.05µg/l, 0.5µg/l, 5µg/l, and 50µg/l
0.003µg/l, 0.01µg/l, 0.03µg/l, and 0.1µg/l
0.03mg/l, 0.1mg/l, and 0.3mg/l
0.1mg/l and 1mg/l
2µg/l, 10µg/l, and 50µg/l
Fecundity = -0.042 + 0.95 * Vtg (R2 = 0.88)
0.9
0.8
0.7
Relative Fecundity
17β-trenbolone
17α-trenbolone
Prochloraz
Fenarimol
Fadrozole
0.6
0.5
0.4
0.3
0.2
0.1
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Relative Vitellogenin
0.8
0.9
1
Biological
Event
Toxicant
Chemical
Reactivity
Profiles
Macro-Molecular
Interactions
Receptor/Ligand
Interaction
DNA Binding
Protein Oxidation
QSAR
Models
Cellular
Responses
Organ
Responses
Gene Activation
Altered
Physiology
Protein Production
Altered Signaling
Protein Depletion
Disrupted
Homeostasis
Altered Tissue
Development
or Function
Systems Models
Effect/Outcome
Individual
Responses
Population
Responses
Lethality
Impaired
Development
Impaired
Reproduction
Structure
Recruitment
Extinction
Cancer
Population
Models
Population Forecasts Based on Molecular Responses
Measurement of vtg concentrations and
fecundity for female fathead minnows
Fecundity
17β-trenbolone
Projection of density dependent
logistic population trajectories for
the fathead minnow population
based upon change in vtg
17α-trenbolone
prochloraz
fenarimol
fadrozole
Vtg
Life table with age specific vital
rates of survival and fecundity for
the fathead minnow population
Carrying capacity for the fathead minnow
population
Population projection for populations
at carrying exposed to stressors that
depress vitellogenin production
Average
Population
Size Size
Average
Population
(Proportion
of
Carrying
Capacity)
(Proportion of Carrying Capacity)
Forecast Population Trajectories
1
1
A A
0%
0.8
0.8
0.6
0.6
0.4
0.4
B B
25%
0.2
0.2
E D
D
0 >95%E 75%
0
C C
50%
0
0
5
5
10
10
Time (Years)
Time (Years)
15
15
20
20
Summary: Conceptual Systems Models in
Research and Regulatory Ecotoxicology
• Provide a framework whereby data from multiple
biological levels of organization (including “omics”) can
be integrated and understood in the context of toxicity
pathways
• Guide hypothesis-driven testing of chemicals/pathway
components
• Help identify key molecular initiating events within an
axis/pathway that subsequently can be represented by
in vitro assay systems and/or QSAR models
• Serve as a basis for defining linkages between
molecular/biochemical changes and adverse outcomes,
in part, through identification of key response nodes
Future Steps
• Proof-of-concept studies focused on welldefined axes such as HPG/HPT
• Cataloging other pathways and building
first-generation conceptual systems
models
• Linkage of systems frameworks with other
models (e.g., QSAR, PB-PK, population)
as a basis for making predictions/guiding
testing