PG16: Epigenetics of Lung Disease Designing Studies of … · Designing Studies of DNA Methylation....
Transcript of PG16: Epigenetics of Lung Disease Designing Studies of … · Designing Studies of DNA Methylation....
Andrea Baccarelli, MD, PhD, MPHLaboratory of Environmental Epigenetics
Harvard School of Public Health
PG16: Epigenetics of Lung Disease Designing Studies of DNA
Methylation
Presentation Outline
Applying Epigenetics to Human Studies
Intro to Epigenetics
Design a Study
Challenges & Opportunities
Epigenetics
• Programming of gene expression that:– does not depend on the DNA code– (relatively) stable, i.e., replicated through:
• cell mitosis• meiosis, i.e. transgenerational (limited evidence in humans)
• Characteristics of epigenetic programming– Modifiable (can be reprogrammed)– Active or poised to be activated:
• Potentially associated with current health states or predict future events
Epigen
etics &
Music Use th
e Same Markings
pencilmarkings(can be erased)
markings in ink(permanent)
DNA methylationMethyl marks added to certain DNA bases repress gene transcription
Histone modificationsA combination of different molecules can attach to the ‘tails’ of proteins called histones. These alter the activity of the DNA wrapped around them
Epigeneticmarkings
DNA methylatio
nDNA methylation suppresses RNA expression(more accurately: it is usually associated with suppressed RNA)
DNA methylationinactive DNA
demethylationactive or poised to be activated
Histone
mod
ificatio
nsChromosomal structureNucleosome – fundamental unit of chromatin
147 bp DNA wound 1.75 turns around histones
histone octamer:2 x (H2A, H2B, H3, H4)
Histone
mod
ificatio
nsHistone modificationstypes and functions
Ac ‐ acetyl (lysine), Me ‐methyl (lysine), P ‐ phosphoryl (Ser or Thr)
Why Epigenetics in Biomedical Research?
• The epigenome is environmentally sensitive– May provide records of past exposures (including prenatal and transgenerational) and help to reconstruct risk‐factor experience
• The epigenome harbors profiles potentially useful to identify at‐risk individuals– May help to predict future risks of disease
• Epigenomic investigations might bring further mechanistic understanding– Challenges in human studies need to be considered
‘Each living organism has two histories that determine its biology:
an evolutionary history whose duration is in the hundreds of thousands of years, and a developmental history that starts at the time of its conception.’
Ze'ev Hochberg, 2011
Epigenetic history
Disease programming throughout the lifecourse
Fetallife
Childhood & adult life
Beforeconception
Epigenome at birth
In-utero exposures
Epigenome in childhood
Early lifeexposures
Preconceptionalexposures
Epigenome(Parental)
Programming of disease risks
Adult/agingepigenome
Later lifeexposures
Genome(parental)
Genome(offspring)
Exposure of gametes
Figure adapted from Fleisch, Wright & Baccarelli, J Mol Endocrinol, 2013
2,000 CASESRecruit from 14
hospitals
2,000 CONTROLSRecruit from census(letters, phone calls)
Self‐administeredQuestionnaire
Blood orBuccal
Specimens
Tissue Collection Fresh frozen Paraffin blocks Tissue slides
CAPI ClinicalData
Hospitals Homes
CAPIBlood orBuccal
Specimens
Self‐administeredQuestionnaire
Study DocumentationCenter
Scan Data – Verify completeness within 30 days
CentralLaboratory
Data Processing CenterCluster server with mirrored
database
TransportationTeam
within 4hours
Blood Components: Fresh frozen Whole blood PBMC (white cells) RBC (red cells) Serum Plasma Buffy coat → DNA RNA Blood cards
PRINCIPALINVESTIGATORS
documents are shipped
at the end of the study
ship weekly in:
•liquid nitrogen
•dry ice
•room
temp
weekly
within 4hours
SIBLINGS
ACD 2x7 ml (real
volume)
spin
PBMC3 vials
5 millions cells/ml in 1.5 ml vials
Plasma6 vials
1 ml each in 1.5 ml
vials
RBC and granulocytes
3 vials1 ml each in 1.5 ml vials
EDTA1x7 ml (real
volume)1x3ml (real vol)
Buffy coat1ml
Plasma3 vials,1 ml
each in 1.5 ml criovials
RBC and granulocytes3vials, 1 ml
each in 1.5 ml criovials
Serum1x7 ml (real
volume)
3 vials 1.5 ml, with 1 ml each
2 Blood cards DNA
extraction 50 g
9 vials 1.5 ml with 1 g each and 2 vials
with the rest(3 in Italy and 6 + 2 in
USA)
7 ml Tube
3 vials 1.5 mlwith 0,2 ml
buffyeach
6 vials 1.5 mlwith 3 g RNA
each vials
Storage: : vapor phase of liquid
nitrogenShipping: vapor phase of liquid
nitrogen Storage: -80°CShipping: dry ice
Storage: vapor phase of liquid nitrogenShipping: vapor phase of liquid
nitrogen
Storage: -80°CShipping: dry
ice
Storage: vapor phase of liquid
nitrogenShipping: vapor phase of liquid
nitrogen
125 l in one
microtube
3 ml Tube
3 vials, whole blood,1 ml each in 1.5 ml criovials
Storage: -80°C
Shipping: dry ice
1 preanalitix
tube
Storage: -80°Shipping: dry
ice
PreAnalytix Tube2x2,5 ml (real
volume)
Storage: -20°C
Shipping: dry ice
Flowchart for blood collection, EAGLE study
Effect of tim
e to storage
on DNA methylatio
nSix samples from each of three placental areas (A, B, C) left at room temperature for
0 to 24 hours before ‐80 C freezing
Villahur N, Epigenomics 2013
Time to storage and DNA methylation
Villahur N, Epigenomics 2013Repeated elements: CV=2.6%LUMA: CV=9.3%
Cross‐sectional correlation between Body Mass Index (BMI) and DNA methylation
Manhattan plot showing the distribution of p‐values of the association of methylation probes with body‐mass index in the discovery cohort
The red dots indicate probes that fall within KLF13 (chromosome 15), CLUH (chromosome 17), and HIF3A (chromosome 19).
Data from the Dick et al., Lancet 2014
The ideal world
Relton and Davey‐Smith Int J Epidemiol 2012
Epigenome
Epigenetic Inheritance Systems
IntermediatePhenotypes / Biomarkers
DiseaseGerm‐line Genetic Variation
Stochastic Events
Environment
The real world
Relton and Davey‐Smith Int J Epidemiol 2012
Epigenome
Epigenetic Inheritance Systems
IntermediatePhenotypes / Biomarkers
DiseaseGerm‐line Genetic Variation
Stochastic Events
Environment
The real world
Relton and Davey‐Smith Int J Epidemiol 2012
Epigenome
Epigenetic Inheritance Systems
IntermediatePhenotypes / Biomarkers
DiseaseGerm‐line Genetic Variation
Stochastic Events
Environment
?
?
?
Reverse Causation
• In reverse causation:– Cause and effect are reversed
• BMI Study:– methyla on → BMI– BMI → methyla on– Either is equally probable
• Study design– Cross‐sectional and case‐control studies are susceptible to reverse causation
– Longitudinal studies should be preferred in epigenetic epidemiology
– Two‐step mendelian randomization can provide an analytical approach to test for causality (Relton and Davey Smith Int J Epidemiol 2012)
Epigenetic markings are Tissue Specific.
Potentially each tissue or cell type has a specific methylation profile.
Tissue specificity
Epigenetics contribute totissue differentiation
Blood Counts and Methylation(combined‐analysis of 5 studies)
Alu LINE‐1
Beta * P‐value * Beta * P‐value *
White blood cells, 103cell/mm3 0.002 0.938 0.078 0.168
Neutrophils, % 0.009 0.226 0.036 0.005
Lymphocytes, % ‐0.009 0.246 ‐0.039 0.004
Monocytes, % ‐0.001 0.981 ‐0.033 0.374
Eosinophils, % ‐0.014 0.643 0.007 0.888
Basophils, % 0.005 0.968 ‐0.202 0.399* Adjusted for age, gender and study.
Zhu et al., Int J Epidemiology 2012
Need to account for signals from cell type differences Adjust for cell type in multivariate analysis Normalize methylation data for cell types before data analysis
DNA methylation arrays as surrogate measures of cell mixture distribution
Houseman et al, BMC Bioinformatics 2012
Predicting DNA methylation level across human tissues
• Two datasets:– 450K Illumina data (480,000 CpGs, n=14) on PBLs, atrium, and internal
mammary artery (IMA).– HumanMethylation27 data (27,000 CpGs, n=39) on peripheral blood
leukocytes (PBLs) and Epstein‐Barr Virus (EBV) transformed lymphoblastoid cell lines (LCLs)
• Between tissue patterns:– Relatively high ‘background’ correlations between tissues – Differences between tissues highly consistent and reliably
reproducible across multiple individuals• Linear regression and Support Vector Machine (SVM) models for
each CpG site to predict methylation in ‘target’ tissues based on ‘surrogates’.– PBLs→Atrium, Raw R2=0.83; calibrated R2=0.99– PBLs→IMA, Raw R2=0.81; calibrated R2=0.94– LCLs→PBLs, Raw R2=0.92; calibrated R2=0.99
Ma B et al. Nucleic Acid Research 2014 (Epub ahead of print)
Baccarelli et al, Epigenomics, 2012
Nasal brush in 36 Children with asthma(studied twice, n=72)
DNA methylation and asthma‐related inflammation
p=0.001
30 40 50 60 70 80
2
3
4
5
6
iNOS promoter methylation (%5mC)
Exha
led
Nitr
ic O
xide
(log
scal
e)
Baccarelli et al. Epigenomics, 2012
DNA Methylation in Nasal Epithelial Cells vs. Exhaled Nitric Oxide and FEV1
p=0.003
20 40 60 80 1001
2
3
4
IL-6 promoter methylation (%5mC)
FEV-
1(lo
g sc
ale)
MT-
TF &
MT-
RN
R1
Met
hyla
tion
(%) P=0.002
Controls(n=20)
High-exposed steel workers
(n=20)
Mitochondrial mtDNA methylationin steel workers exposed to metal‐rich air particles (PM1)
Byun et al, Particle Fib Tox, 2013
Other cytosine modifications
Relation between 5‐hydroxy‐methyl‐cytosines and 5‐methylcytosine in human blood DNA
Hou et al, in preparation
N=237R=0.21 p=0.001
Different mechanisms, same design issues?
• Interest in other epigenetic modifications is growing– mtDNA methylation– 5‐hydroxy‐methylcytosine, 5‐formylcytosine and 5‐carboxylcytosine
– Non CpG DNA methylation– Others (including histone modifications, miRNAs, etc)
• Most of the same design considerations will apply– Sample collection– Time to storage– Reverse causation– Tissue specificity
Challenges in epigenetics
• How many epigenomes in one body?– Tissue specificity
• Most studies in humans are on blood DNA• Need to investigate tissues relevant for the exposure‐disease of interest (challenging in epidemiology)
• How many epigenomes in one lifetime?– The epigenome changes over time
• Reverse causation is always a potential issue• Need for longitudinal studies
• How many epigenomic markings in one epigenome?– DNA methylation, histone modifications, others
• Which is most informative?
Opportunities in epigenetics
• How many epigenomes in one body?– Opportunities for screenings of multiple epigenomes:
• Multiple tissues • Multiple cell types (e.g., blood subpopulations)
• How many epigenomes in one lifetime?– Opportunities for lifecourse epigenetics:
• The epigenome might record recent or past experiences• The epigenome might predict future risk of disece
• How many epigenomic markings in one epigenome?– Integrate multiple epigenomic markings
• Coordinated and complementary mechanisms