Using Diversity Outbred mice to mimic human population...

Using Diversity Outbred mice to mimic human population

dynamics in susceptibility to environmental chemicals

Alison HarrillBiomolecular Screening Branch

National Toxicology Program / NIEHSNational Institutes of Health

October 6, 2017

• The opinions expressed in this presentation are the author's own and do not necessarily reflect the view of the National Institutes of Health, the Department of Health and Human Services, or the United States government.

Disclaimer

• Many thousands of chemicals in commerce and in the environment

– Flame retardants, pesticides, dyes, manufacturing catalysts, etc..

– Arsenic, lead, diesel exhaust, etc…

• Multiple scales and routes of exposure

– Occupational (long, high dose duration exposure) , user (short high dose exposure), communities (long or short, low or high)

Where we almost never have human outcome data and never will

Human Health Risk Assessment for Chemicals

• Data needed for risk assessment

– Hazards – target tissues and outcomes

– Dose / Potency – what exposure carries an agreeable risk (conservative)

• Acute versus cumulative

– Susceptible subpopulations

• Age

• Life stage

– All of these data come from whole animal guideline and in vitro studies

Human Health Risk Assessment for Chemicals

• Gene x environment differences are important

– 2009: NRC – for risk assessment “attention should be directed to vulnerable individuals and subpopulations that may be particularly susceptible”

– Understanding influence of genes can inform data-driven uncertainty:

• Methyl mercury: individual susceptibility can vary 50,000-fold between individuals

• 5% of population could be 25-fold more susceptible than average

Considering genetic variation in chemical risk assessment

Variation in: Genetic sequenceEpigenome

EnvironmentHealthStatus

• NTP Mouse – B6C3F1 hybrid

• NTP Rat – Sprague Dawley – outbred, but with low allelic variation and many IBD regions

• Human cells – Typically from a single donor

• A “feature”, not a “bug”!

• New tools have made it more desirable to consider genetic diversity in chemical safety assessments

Diversity is rarely incorporated into risk assessment

Rationally interbred population that mimics human genetic diversity, but polymorphisms are highly randomized

Diversity Outbred mice provide genetic diversity

Founder Strains: A/J, C57BL/6J, 129S1/SvImJ, NOD/ShiLtJ, NZO/H1LtJ, CAST/EiJ, PWK/PhJ,

WSB/EiJ

F1 Diallel Cross→F1 Crosses

Establish Inbreeding Funnels

Collaborative Cross (CC)

• >100 RI lines available• Each RI line is genetically

distinct• Within each RI line, all mice

are genetically identical

Diversity Outbred (DO)

• Large population of genetically unique individuals

• High level of genomic heterozygosity

• Each individual animal is genetically unique

Initiation: Interbred 144 segregating CC lines between G2:F4 and G2:F12 generations

• Comprised of dozens of conventional inbred strains

• Each line is genetically distinct• Many strains have a high

degree of genetic relatedness between them, potentially limiting degree of genetic diversity across strains

Mouse Diversity Panel (MDP)

Strain

Selection

• Diversity Outbred mice are highly genetically diverse, with a randomization of polymorphisms that is superior to human populations Harrill and McAllister, Environmental Health Perspectives 2017

DO are a genetic reshuffling of genes from the 8 founder strains

DO Mice: Background of population diversityImages: www.jax.org

• 129S1/SvImJ – Sensitive uterine response to oestrogens, high serum cholesterol

• A/J –resistant to cigarette smoke induced emphysema; late onset muscle disease (homozygous mutation)

• C57BL/6J – refractory to many tumors, high susceptibility to diet-induced obesity/ T2D, atherosclerosis, high incidence of eye abnormalities

• NOD/ShiLtJ – autoimmune T1D, defects in Ag presentation, impaired wound healing

• NZO/HILtJ – obesity on standard diet and T2D

• CAST/EiJ – highly genetically divergent from other strains, improved neuron axonal regeneration, fast and highly active

• PWK/PhJ – Highly genetically divergent from other strains, docile

• WSB/EiJ – Highly active, wild temperament, age-related autosomal dominant deafness

Uses for DO mice in Chemical Risk Assessment

Mode of ActionHazard ID

Elem

ent

Pot

entia

l App

roac

hes

for P

opul

atio

n-B

ased

R

isk

Ass

essm

ent Predict adverse

effects that only occur in genetically sensitive individuals

Identify hazards that conventional models

may miss

Inform extrapolation of rodent to human via data to replace

standard uncertainty factors

Elucidate shape of dose-response relationship for

variety of endpoints in populations

Quantify threshold doses and BMDL10for adverse events

that occur in sensitive individuals

Dose Response

Estimate population risk with data-driven relationship between exposure and dose

Elucidate interplay between variability in

toxicokinetics with variable

toxicodynamics

‘Omics platform identification of key molecular changes

associated with increased risk

Identify genetic sequence variants

that underlie toxicity sensitivity

Exposure Assessment

Establish exposure biomarkers for biomonitoring

Measure population-wide differences in

toxicokinetics to estimate internal

dose

Harrill and McAllister, Environmental Health Perspectives 2017

Usage Case 1: Identify specific polymorphisms associated with xenobiotic toxicity using the DO

• Idiosyncratic toxicities have been difficult to study because dose-response relationship is unclear and because MOA is debated

– Lack of models: DO may fill a gap

Investigating the cause of rare clinical reports of green tea extract herbal supplement liver injury

DO mice as a model for rare adverse events

Resistant

Responders

Extreme responders “idiosyncratic” DILI

Church et al. Food & Chem Toxicol. 2015

Individual DO Mice

Pharmacogenetic analysis of idiosyncratic DILI

Church et al. Food & Chem Toxicol. 2015

Human translation of GTE PGx risk factors

Gene symbol

SNP (Array)

Gene name Chr P value

Risk/Protective allele

Effect

PER3 exm10762Period circadian clock 3

1 0.004937 T/C Missense (R/W)

MFN2 exm15928 Mitofusin 2 1 0.0067 A/G Missense

(I/V)

VPS13D exm16480

Vacuolar protein sorting 13 homolog D (S. cerevisiae)

1 0.043064 A/T Missense (R/S)

Green tea extract (GTE) supplement hepatotoxicity risk alleles in human clinical cases

PGx: chemotherapy-induced neutropenia

Trp53Abcb1b

N=200-400 DO mice per drug Single dose level

Implications for precision medicine research & gene x environment analysis

Gatti et al. The Pharmacogenomics Journal. 2017

doxorubicin

doxorubicin + CSF3

cyclophosphamide

docetaxel

naive

Usage Case 2: Evaluate biomarker performance toward assessing human clinical adverse outcomes

Using a population model to investigate predictive value of novel biomarkers

• Many groups are working on qualifying novel biomarkers into clinical practice

• But, how well will biomarkers identified in genetically limited rodent strains translate to a larger population?

• Do single strain studies prevent us from making human relevant discoveries?

Harrill et al. Tox Sci 2016

Renal Biomarkers to Investigate in DO Mice

• “Gold standard”

– Creatinine

– Blood urea nitrogen (BUN)

• Novel proteins qualified for preclinical rat toxicology

– KIM-1, IP-10, TIMP-1, OPN, Renin, VEGF, NGAL, EGF, clusterin, cysC, albumin, TFF3

• Other urine markers

– miRNA

Cisplatin Study – Biomarker Performance

Figure 5. Treatment paradigm for cisplatin exposure and sample collection.

Animals: Female DO mice, 8-12 weeks of ageN = 45 cisplatin (5 mg/kg)N = 45 saline

Goals: (1) Evaluate performance of conventional kidney injury biomarkers(2) Evaluate performances of “NEW & IMPROVED” kidney injury biomarkers(3) Determine whether urinary miRNA biomarkers offer advantages over (1) and (2)

Haixia Lin & Julia Tobacyk, UAMS

Population dynamics in conventional kidney biomarkers, single cisplatin dose

Vehicle Cisplatin0

255075

100125150175200225

BU

N (m

g/dL

)

*

*P<0.05

Cisplatin induced histopathologyDifferential susceptibility

*Special thanks to John Seely and NIEHS Pathology

Tubular degeneration or necrosis was not found in vehicle treated animals (excluding 1 animal with minimal severity of tubular degeneration)

No drug-related pathology was found in the liver

Treatment Group Vehicle Cisplatin

Number Examined 45 44

No Degeneration/Necrosis Present 44 13

Degeneration (Tubule) 1* 18

Minimal (Grade 1) 1 15

Mild (Grade 2) - 3

Necrosis (Tubule) - 26

Minimal (Grade 1) - 17

Mild (Grade 2) - 9

Grade 0

Grade 1 Grade 2

A B

C D

Vehicle

Cisplatin

Cisplatin

Vehicle

BUN and urinary proteins only elevate when Grade

2 necrosis is present

Vehicl

e

Grade 0

Grade 1

Grade 2

0

50

100

150

200

250B

UN

(mg/

dL)

BUN

a,b

Cisplatin Necrosis ScoreVehicle

Vehicl

e

Grade 0

Grade 1

Grade 2

-20000

0

20000

40000

60000

80000

100000Renin

Ren

in (p

g pe

r 18

hr u

rine

vol.)


a,b

Vehicl

e

Grade 0

Grade 1

Grade 2

-50000

0

50000

100000

150000

200000

250000TIMP-1

TIM

P-1

(pg

per 1

8 hr

urin

e vo

l.)


Vehicl

e

Grade 0

Grade 1

Grade 2

-50000

0

50000

100000

150000B2M

B2M

(pg

per 1

8 hr

urin

e vo

l.)


a,b

Vehicl

e

Grade 0

Grade 1

Grade 2

-20000

0

20000

40000

60000KIM-1

KIM

-1 (p

g pe

r 18

hr u

rine

vol.)


a,b

Vehicl

e

Grade 0

Grade 1

Grade 2

-50000

0

50000

100000

150000IP-10

IP-1

0 (p

g pe

r 18

hr u

rine

vol.)


a,b

A.

B.

C.

F.

G.

H.

Vehicl

e

Grade 0

Grade 1

Grade 2

-5000

0

5000

10000

15000NGAL

KIM

-1 (p

g pe

r 18

hr u

rine

vol.)


a,bE.

Grade 0

Grade 0

Grade 1

Grade 2

15

20

25

30

35

miR-130a

Ct

Vehicle Cisplatin

a,b,c

Grade 0

Grade 0

Grade 1

Grade 2

15

20

25

30

35

miR-221

Ct

Vehicle Cisplatin

a,b,c

Grade 0

Grade 0

Grade 1

Grade 2

15

20

25

30

35

miR-138

Ct

Vehicle Cisplatin

a,b,c

Grade 0

Grade 0

Grade 1

Grade 2

0

10

20

30

40

miR-320

Ct

Vehicle Cisplatin

a,b,c

Grade 0

Grade 0

Grade 1

Grade 2

15

20

25

30

35

miR-151-3p

Ct

Vehicle Cisplatin

a,b,c

Grade 0

Grade 0

Grade 1

Grade 2

0

10

20

30

40

miR-328

Ct

Vehicle Cisplatin

a,b,c

Grade 0

Grade 0

Grade 1

Grade 2

0

10

20

30

40

miR-152

Ct

Vehicle Cisplatin

a,b

Grade 0

Grade 0

Grade 1

Grade 2

0

10

20

30

40

miR-680

Ct

Vehicle Cisplatin

a,b

Grade 0

Grade 0

Grade 1

Grade 2

0

10

20

30

40

miR-218

Ct

Vehicle Cisplatin

a

Grade 0

Grade 0

Grade 1

Grade 2

0

10

20

30

40

miR-685

Ct

Vehicle Cisplatin

a,b,c

The same is true for urinary miRNAs that increase in abundance in response to cisplatin exposure. Elevations in urinary miRNA

abundance are only observed at necrosis grade 2.

Urine miRNA Abundance

Population response dynamics aid biomarker evaluation

Cisplatin miRNA biomarkers in DO - ROC

0 50 1000

25

50

75

100

100% - Specificity%

Sens

itivi

ty%

ROC curve: ROC of miR-130a

0 50 1000

25

50

75

100

ROC curve: ROC of miR-221

100% - Specificity%

Sens

itivi

ty%

0 50 1000

25

50

75

100


100% - Specificity%

Sens

itivi

ty%

0 50 1000

25

50

75

100


100% - Specificity%

Sens

itivi

ty%

0 50 1000

25

50

75

100

ROC curve: ROC of miR-151-3p

100% - Specificity%

Sens

itivi

ty%

0 50 1000

25

50

75

100


100% - Specificity%Se

nsiti

vity

%

0 50 1000

25

50

75

100


100% - Specificity%

Sens

itivi

ty%

0 50 1000

25

50

75

100


100% - Specificity%

Sens

itivi

ty%

0 50 1000

25

50

75

100


100% - Specificity%

Sens

itivi

ty%

0 50 1000

25

50

75

100


100% - Specificity%

Sens

itivi

ty%

Even though performance not great at low end of injury spectrum , ROC

curves are all significant p<0.05 AND

AUROC values 0.71-0.82

Usage Case 3: Tool for population-based estimates of chemical potency for risk assessment

Human BMCL10 7.2 ppm (44 subjects), < 1 ppm FX

DO BMCL10: 0.205 ppm ; B6C3F1 BMCL10: 3.12 ppm

DO: calculate BMDL for inhaled benzene

French et al. EHP 2015.

Population-Based Analysis of Chemical Sensitivity in vitro – Developmental

Neurotoxicity

More refined recombination structure in the CC/DO result in smaller more refined co-expression networks and correlations among behavioral phenotypes at sample sizes comparable to conventional behavioral QTL mapping studies

Chesler. Mamm Genome. 2014; 25(1-2): 3–11.

• Predictive Biology has archived hundreds of ES cells from DO mice

• Over 100 male and 100 female DO ES cell lines (representing 200 individual and genetically unique mice) have been differentiated into neural progenitor cells

• Chromosomal aberrations checked by RNA-Seq and Giga MUGA (mouse universal genotyping array)

DO Neural Progenitor Cells

ESCells

Neural Stem cells

Neural Progenitor

Cells

Neuron-Restricted

Progenitors

Motor Neuron

Progenitors

Glial-Restricted

Progenitors

DNT population responses – differ by chemical and by DO cell line

Estradiol

Methyl Mercuric (II) Chloride

Dieldrin

2,2',4,4',5‐Pentabromodiphenyl ether

Rotenone

Phenol, isopropylated, phosphate (3:1)

Wide variability in response across DO cell lines / individuals for a single chemical

Wide variability in response across DO cell lines / individuals for a single chemical

00.5

11.5

22.5

33.5

44.5

5

0 5 10 15 20

IC50

(uM

)

Cell Lines

Dieldrin

Replicate 1

Replicate 2

Wide variability in response across chemicals for a single DO cell line

Wide variability in response across chemicals for a single DO cell line

Compound P valuesDieldrin 0.000239Ether 0.00071

Estradiol 0.007Methyl mercury 0.02

PBDE 0.0218Rotenone 0.038

Lead 0.07BPA 0.11

Caffeine 0.12Triphenylphosphate 0.24

BPAF 0.25Phthalate 0.3Dopamine 0.43

Hexachlorophene 0.8Phenol 0.939

Correlation between runs

• Standard uncertainty factor for extrapolating rodent to human NOAEL is 100 – is that protective enough?

• Use DO dose-response data to inform a point of departure for regulatory agencies to set a reference dose (RfD)

Potential Study Outcomes

Wide variability in dose responsivenes

Chemical XBMD – 13.4 uM

NTP, EPA, & FDA/NCTR

Phase 2: Tox21 Cross-Partner Project

Chemical Exposures

Expose DO NPCs and fix cells

High Content Imaging

Label targets with fluorescent probes

Scan/Image cells

Image analysis / quantify fluorescent intensity and

morphometry

Predictive Biology NCTR/FDA

Data Analysis and BMD Assessment

Calculate BMD/BMDL for each chemical/sex and for each endpoint

Determine relative sensitivity of each HCA probe and DO cell line

Determine sex differences in susceptibility for each compound

NIEHS/NTP

Dose Concentration Selection

Input into study design, chemical selection, and target doses

Cross-Partner Data Contextualization (NTP/FDA)

Model Evaluation: Compare BMDs calculated using DO NPCs to in vivo data collected by NTP and other stakeholders; determine concordance with human epidemiological data where available, compare variability with other DNT screens

under development (human iPSC neurite outgrowth, zebrafish, others)

AcknowledgementsNIH/NIEHS/NTPRick PaulesMamta BehlKristine Witt

UAMSHaixia Lin (postdoc)Julia Tobacyk (former student)Sam Luo

UNC-Chapel Hill &The Hamner InstitutesRachel Church Merrie MosedaleScott EaddyPaul Watkins

The Jackson LaboratoryDaniel GattiGary Churchill

FDA/NCTRBill SlikkerBill MattesDan AcostaDonna MendrickMark Avigan

Predictive BiologyTed ChoiLesley Page

Funding*UAMS*FDA*The Burroughs Wellcome Fund

*Prior to my joining the NIH

EPLJohn Seely

Using Diversity Outbred mice to mimic human population...

Documents

Transcript of Using Diversity Outbred mice to mimic human population...