EAAP 30 Aug- 4th Sept 2015- Warsaw
Epigenetic databases in cattle and the
prediction of phenotypes from genotypes
Jean-Paul Renard
Hélène Jammes - Hélène Kiefer
Académie d’Agriculture de France - Paris
INRA UMR 1198- Jouy en Josas
66th EAAP Annual Meeting
30 August – 4th september 2015
Warsaw, Poland
Eve Devinoy INRA GABI- 1113- Jouy en Josas
van Driel, R. et al., 2003 Journal of Cell Science
Between genotype and phenotype…
…3 levels of genes regulation…
Sequence
Chromatin
…Controlled by epigenetic information
Histone modifications
DNA methylation
Non coding RNAs
Nuclear compartments Nuclear organization
ex: interpreting non coding genetic variation
in complex human traits at a single locus
a- Genetic association with an organismal trait (GWAs) b- Genetic association with a molecular trait (eQTL)
c- Genetic association allelic activity (ASE)
d-molecular biomarker for an physiological trait (EpiWAs)
Ward and Kellis, 2012, Nat.Botech.
Involves the use of reference
epigenome maps of primary
and cultured cells
Epigenetic information in the prediction of traits from GWAS..
in human
EAAP 30 Aug- 4th Sept 2015- Warsaw
first cell lineages establishment
of body plan
waves of epigenetic reprogramming shape transcriptomic patterns
during both embryonic, foetal and post natal periods
organogenesis,
Post-natal
development
BIR
TH
posterior pole
anterior pole
2 weeks 2 months perinatal
adult Blastocyst placenta foetus
gastrulation
EAAP 30 Aug- 4th Sept 2015- Warsaw
First step: simplify the approach using
- genetically identical animals: through cloning
- DNA methylation
Second step: develop a wide genome approach using key
tissues: blood, liver, muscle placenta, mammary gland
Third step: combine phenotypic and epigenetic data
Outcome: - extension to individuals from non related genotypes
- integration into a bovine data platform
Measuring the contribution of epigenetics
to the variability of phenotype in the bovine species
*
16
8
12
10
14
6
4
0
2 Su
ccss
ra
tes
of
recl
on
ing
Wakayama et al. 2013 Cell Stem cell
Cloning techniques do not select for somatic cells
more amenable to full reprogramming
Lif
esp
ans
in y
ears
0.5
1.0
1.5
2.0
2.5
3.0
3.5
recloning number recloning number
mouse cloning
(%)
-optimisation of in vitro conditions (empirical)
**
** **
no impact on viability no drop in efficiency
-recloning from adult cloned mouse
EAAP 30 Aug- 4th Sept 2015- Warsaw
fibroblastes
- Ca++
In vitro culture
donor cells
enucleated recipient oocytes
in
viv
o
In
vit
ro
in
vit
ro
in
vit
ro
in
viv
o
transfer into
recipient
One-cell stage embryo blastocyste
different nuclear microenvironment from the one-cell stage
same nuclear genome
in vivo in vitro (culture) in vivo (oocyte ) - in vitro (culture)
relatively efficiency at birth 40%
9 months- long gestation
1998 2002 2006 2000 2004 2008
MOET
IVP
Cal
vin
g r
ate
(%)
40
30
20
10
0
- 40%
Watanabe and Nagai2011 and EFSA 2012
SCNT
from 482 calving
Pregancy rate
Transfer (D7)
D21 D70 D90 calving
IVF 0
30
50
70
100
D35 Clone
Bovine cloning: a way to profile epivariants in a context of high
environmental constraints applied on the genome.
EAAP 30 Aug- 4th Sept 2015- Warsaw
10-3
CpG
Sequence
Environment
Stochastic
Active DNA methylation Dnmt1: maintenance Inheritable (mitosis) Dnmt3: de novo in response to stimuli
DNA methylation is highly dependant
on genomic sequence
DNA methylation loss Passive Active
EAAP 30 Aug- 4th Sept 2015- Warsaw
Sonication
Denaturation
Anti-5meC
Methylated DNA analysis
Analyse by Chip hybridization or sequencing
2. Whole genome analysis
MeDIP/Chip
MeDIP/sequencing
RRBS (Reduced ReprensentativeBisulfite Sequencing)
1. Global analysis
Cytosine methylation electrophoretic assay
LUMA (Luminometric Methylation Assay)
Immunocytochemistry
3. Sequence specific approach
Bisulfite conversion
/pyrosequencing
Tools used to characterize, identify and measure DNA
methylation variations
Bovine clones are genetically identical but epigenomically variants
A
Clones Holstein
Clones Simmental
Jumelles Simmental
Vache donneuse
% d
e C
yto
sin
es m
éth
ylé
es
B
Dev
iati
on
in
div
idu
elle
du
% d
e cy
tosi
nes
mét
hylé
es
Holstein,
4 genotypes
Simmental,
5 genotypes
Twins pairs, n=13
Simmental
Variability of global
methylation rate
in lymphocytes of
adult bovine clones
De Montera et al., 2010 Cytosine methylation electrophoretic assay
BIR
TH
posterior pole
anterior pole
2 weeks 2 months perinatal adult
Blastocyst placenta foetus
gastrulation
From : http://www.epigenome-noe.net/researchtools/protocol.php?protid=33
bovine microarray targeting
21 416 genes
Methylated Unmethylated
Epigenetic signatures at perinatal stage in liver
MeDIP/Chip- Kiefer et al. submitted
Post-natal
development
BIR
TH
posterior pole
anterior pole
2 weeks 2 months perinatal
adult Blastocyst placenta foetus
gastrulation
-2000 pb 0 +1360 pb
(~34 probes per promoter region ) + 12 larger regions
Data analysis framework
Enriched in MeDIP « methylated » Not enriched in MeDIP
Identification of 83 Differentially
Methylated Regions (DMRs)
between clones and AI SpatStat R package
3
Identification of highly
methylated regions containing clusters of enriched probes in at least one condition
2
4 Validation by
pyrosequencin
80% regiond vamidaterd
1 Identification of
methylated probes ChIPmix R package 720 000
Data analysis framework
Identification of
methylated probes
among 720 000 probes ChIPmix R package
Identification of 83 Differentially
Methylated Regions (DMRs)
between clones and AI SpatStat R package
1
3
Identification of 15885
highly methylated
regions containing clusters of enriched probes in at least one condition
2
4 Validation by
pyrosequencing
Data analysis framework
Identification of
methylated probes ChIPmix R package
Identification of 83 Differentially
Methylated Regions (DMRs)
between clones and AI SpatStat R package
1
3
Identification of highly
methylated regions containing clusters of enriched probes in at least one condition
2
4 Validation by
pyrosequencing
Perinatal AI
Perinatal clones
Data analysis framework
Identification of
methylated probes ChIPmix R package
Identification of 83 Differentially
Methylated Regions (DMRs)
between clones and AI SpatStat R package
1
3
Identification of highly
methylated regions containing clusters of enriched probes in at least one condition
2
4 Validation by
pyrosequencing
C U T 39%
mC C C 61%
Native DNA Treated DNA PCR products Pyrogram
Bisulfite conversion
PCR amplification
Pyro-sequencing
35 40 45 50
180
220
260
Cell a
rea (
µm
2)
35 40 45 50
2
3
4
5
6
7
Nu
CV
Sh
ap
e F
acto
r (%
)
35 40 45 50
0.2
0.4
0.6
0.8
NL
DH
A/A
A
35 40 45 50
14
18
22
26
Methylation (%)
Nu
CV
are
a (
%)
35 40 45 50
0.2
0.4
0.6
0.8
1.0
Methylation (%) P
L C
14.0
(%
)
Fetus AI
Calf AI
Fetus clone
Calf clone
Perinatal
AI
Perinatal
clone
% m
eth
yla
tio
n
p=0.018
35
40
45
50
979
76
Perinatal AI
Perinatal clones
Microarray data
Pyrosequencing validation
Correlation methylation-phenotypic markers
r=-0.86
r=-0.69
r=-0.89
r=0.61
r=-0.55
Permutation test
Spearman’s rank correlation test
Metabolic parameters Cellular parameters
One example (among hundred others) of DMR
methylation rate correlated to phenotypic markers of
liver physiology
35 40 45 50
180
220
260
Cell a
rea (
µm
2) Fetus AI
Calf AI
Fetus clone
Calf clone
Correlation methylation-phenotypic markers
r=0.61
Clone with no glycogen storage (average size of hepatocytes is smaller)
Clone with normal glycogen storage (average size of hepatocytes is similar to AI)
An example of DMR methylation rate
correlated to phenotypic markers of the liver’s physiology
In human, the associated gene
shows differential methylation in
the blood of newly diagnosed,
drug naive patients with type 2
diabetes Canivell et al. 2014,
PLoS one)
Ongoing validation from the blood of dairy cows at early lactation.
to predict early lactation-associated
hepatosteatosis in dairy cows
BIR
TH
posterior pole
anterior pole
Intercaruncule
Caruncule
2 weeks 2 months perinatal
Extension to muscle epigenomic analysis:
% m
eth
yla
tio
n
DMR1 DMR2 DMR3 DMR4
AI controls, 18 months n=9
Healthy clones, 18 months n=8
Muscle biopsy at 18th months ~ 500 µg genomic DNA
*p<0.05
Permutation test
A first step towards the tracability of cloned products ? Kiefer et al. submitted
screening AI and cloned adults
EAAP 30 Aug- 4th Sept 2015- Warsaw
Genomic data
phenotypical data
environmental data
Genome and
functions
• Genomics (QTL, SNPs…
• Epigenomics
• Metabolomics
• Ruminomics
Traits
size, growth,
Productive traits
Reproductive traits
puberty
ovulation rates
fertilisation
embryo death
fetal losses…..
Phenomes
• more data
• higher resolution
environment
Animal management
Animal health/treatment
What happened during the reproductive cycle?
•What was the outcome?
through EOL
capture more Data on an
individual basis
through ATOL
• more data
• higher resolution
and from now …
The individual animal within its environement as the reference
through FANG
Leading Data Science Platform in the Industry
Software Engineering
How can it be built?
Domain
Science
What is important?
Statistics
How can
predictions be
made?
DATA INSIGHTS
Models:
Predict the Future
In agriculture
EAAP 30 Aug- 4th Sept 2015- Warsaw
Storing Data is Getting Cheaper In Agriculture, Environmental and Operating Data
Are Taking Shape
Low Cost Computing is Proliferating
Average Price Per Gigabyte, 1980-2013 $192,308
$200,000
$180,000
$160,000
$140,000 $119,950
$120,000
$100,000
$80,000
$60,000
$40,000
$20,000
$0
$44,950
$176 $0.70 $6.21
$0.07 $0.04
1980 1984 1989 1996 2000 2004 2009 2013
Source: Wayback Machine (Statistics Brain)
Distributed Computing Allows Multiple Computers to Use More Data and Solve a Problem Together
97% drop in price
per gigabyte over
past 10 Years
Computing Applications Make Data Relevant to Animal production.
They are already economically operational in Agriculture
ex: Hadoop platform for Big data management
EAAP 30 Aug- 4th Sept 2015- Warsaw
Domain
Science
What is important?
Statistics
How can
predictions be
made?
DATA
Software Engineering
How can it be built?
Domain
Science
What is important?
Statistics
How can
predictions be
made?
DATA INSIGHTS
Unified
platform
Continue
capability
development
Establish
coordinated
industry
framework
to address
data privacy
Establish broad,
engaged
foundation base
experience with
product & service
offerings
Establish
anchor
producers
distribution
partnerships
Models: Predict
the Future
EAAP 30 Aug- 4th Sept 2015- Warsaw
• Oral session
• n°37wed.17h
Poster
n°15-21249
Jean Philippe Perrier
Maxime Gasselin
Hala Al Adhami
François Piumi
Luc Jouneau
Audrey Prézelin
Evelyne Camplion
Hélène Kiefer
Hélène Jammes
Coll. i Biostatistical Analysis
Marie-Laure Magniette
AgroParisTech
Coll. Methylation Analysis
(RRBS) CNRS, Strasbourg
EpigRAni partners
• Eve Devinoy, GABI, INRA, Jouy en Josas
• Dominique Rocha, GABI, INRA, Jouy en Josas
• Marion Boutinaud, Pegase, INRA, Rennes
• Jorg Tost, CEA/CNG, Evry
Poster
n°14-21257