Edison Liu

1
DNA Technologies and Strategies in Cancer
Medicine Pre-IMPAKT Brussels May 6, 2009
Edison Liu liue_at_gis.a-star.edu.sg 65-6478-8007
2
We are in the midst of a medical revolution

• Based on
• Availability of the complete human genome
  sequence
• Genomic technologies
• Computational advances
• All delivered to clinical doorsteps

3
Genomics and Genetics

• Genetics study of genes and their function
• Genomics study of all genes and how they
  function together
• Discover all possible genes involved in a
  cellular process or human disease

4
Advances in Breast Cancer Genetics

• Linkage analysis and Genome-wide Association
  studies
• Germline
• Hereditable
• Whole genome sequencing
• Somatic
• Cellular evolution

5
Mapping of cancer susceptibility alleles
A
a
A
a
X
Linkage
B
B
b
B
c
C
C
C
Cancer
a
A
A
a
a
A
A
A
B
B
b
B
B
B
b
B
c
C
C
C
C
c
c
C
Cancer
Cancer
6
Population based discoverycase-control
association studies
General Population (controls) Alzheimers
Patients (cases)
APOE 2/ 3 85 APOE 2/ 3 60 APOE
4 15 APOE 4 40
7
Infinium I II Assays
BeadChip
8
High-penetrance breast cancer susceptibility
genes BRCA1, BRCA2, TP53
Frequency Rare 0.1 Risk of breast cancer
10-20X RR
Moderate-penetrance breast cancer susceptibility
genes ATM, BRIP1, CHEK2, PALB2
Frequency Rare 0.6 Risk of breast cancer 2-4X
RR
Low-penetrance breast cancer susceptibility
alleles rs2981582 (FGFR2, 10q), rs3803662
(TNRC9 (renamed TOX3)), rs889312 (MAP3K1, 5q),
Rs3817198 (LSP1, 11p), rs13281615 (8q),
rs13387042 (2q), rs1045485 (CASP8_D302H)
Frequency Common 550 Risk of breast cancer
1.1-1.65X relative risk
After Stratton Rahman Nat Gen 4017 (2008)
9
26,258 cases 26,894 controls
Gene Position Allele OR
P-trend frequency FGFR2 10q
0.38 1.26 (1.6) 2 X 10-76 TNRC9
16q 0.46 1.11 (1.2) 2 X
10-19 MAP3K1 5q 0.28 1.13
(1.3) 7 X 10-20 LSP1 11p
0.30 1.07 (1.2) 3 X
10-9 Rs13281615 8p 0.40 1.08
(1.2) 5 X 10-12 Potentially 449 candidate
susceptibility alleles
Easton, et al. Nature 4471087 (2007)
10
Few Genes Big EffectMany Genes Small
EffectMany interactions Small effect and
combinatorial complexity
Wright 1968
BRCA1
BRCA2
If there are gt400 common susceptibility
alleles, the combinatorial possibilities present
in the population are immense
Intensity of Effect
Frequency in the population
11
ImplicationsSame phenotype, Different
genotypesin different ethnic groupsrequiring
different treatments
BRCA
? Caucasians
BRCA
BRCA
? Asians
12
Genes common to gt1 diseaseAutoimmune
disordersChallenge todayPhenotype-genotype
correlations on a population scale
Chris Cotsapas Mark Daly
13
The precision is to the nucleotide DNA
sequencing
14
Common features massively parallel
sequencing on two dimensional arrays. Short reads
(25-500 bp), High computational requirements
454 Roche Emulsion PCR, pyrosequencing Illum
ina Bridging PCR, reversible dye
terminator ABI SOLiD oligonucleotide
hybridization
15
7 orders of magnitude increase in throughput
MR Stratton et al. Nature 458, 719-724 (2009)
16
Ph1 Chromosome CML
Digital Karyogenomics
17
Sequencing cancer genomes Historically 350
genes (1.6 of 22K coding genes) with recurrent
mutations in cancer Mouse models suggest that
2000 genes may contribute to a cancer
phenotype Estimates 10 of genes have
non-silent mutations in primary human cancers
(2000) Range of point mutations found gt100,000
to lt1,000 Estimated average number of specific
cancer gene mutations in any breast cancer 14
18
Genes commonly mutated in cancer TP53 K-ras,
N-ras, H-ras BRAF APC Pathways
Ras-MAPK ,EGFR, ERBB2, FGFR1, FGFR2, FGFR3,
RET, PDGFRA and PDGFRB AKT-PI3Kinase Wnt
Notch
19
TP53
Mutations occur in clusters around genes
and Pathways. Most mutations occur in lt5 of
tumors
From Wood L, et al. Science 3181108 (2007)
20
Driver Cancer Mutations Mutations in nodal
genes critical to cancer biology Passenger
Mutations Mutations that arose in the
evolutionary development of a cancer and does
not participate in cancer phenotype Estimated
ratio 1150 and likely to be much lower.
21
Power Law Relationship A few highly
connected nodes in a network has the greatest
impact on a system
Effect
Frequency (of interaction) Frequency (of event)
22
Perhaps the passenger mutations represent
permission mutations. Perhaps there are many
genes (now involved in passenger mutations)
that contribute in the composite to the cancer
phenotype.
23
Genomic scale information provides unique but
complex biological insights
24

• Strategy
• P53 is a transcription factor whose expression
  profile affects clinical outcome
• Map all binding sites for p53 in an unbiased
  manner (technology development)
• Chialin Wei Huck Hui Ng
  Yijun Ruan

25
Gene Identification Signature (GIS) analysis
For the identification of fusion oncogenes
Full-length cDNA Cloning Terminal tag extraction
Ditag formation
GIS library
GIS single ditag library
Ng, et al. Nature Methods 9 Jan 2005
26
A proof-of-concept experiment, ChIP-GIS
Genome-wide localization of p53 binding sites
Human cell HCT116 Treated with 5FU to induce p53
p53 ChIP
qPCR
400X, p21 site
ChIP enriched DNA (0.5-4 kb)
GIS Pair-End diTag library

• Mapping to hg17
• 17-17 bp minimal
• within 4 kb

512,876 PET sequences
382,741 (75) mapped to single location
Wei CL, et al. Cell (2006) 124(1)207-19
27
Predicting the number of specific p53 binding
sites by extrapolation
3741
28
chr636,740,278-36,765,927, 21,200 bp
CDKN1A, p21
CDKN1A, p21
qPCR validation
29
Bona fide p53 binding sites (relative to
idealized gene) are predominantly gt5 kb from TSS
Upregulated genes

• 100 new direct p53 target
• genes identified by ChIP-PET analysis

Downregulated genes
30
These p53 binding sites act as enhancers and
reside next to genes whose expression reflects
p53 status in tumors
P53 /
P53 -/-
Expression of putative p53 regulated genes
are dependent on p53 for transcriptional induction
31
Genes identified as direct downstream targets of
p53 can distinguish p53 status in primary breast
cancers and give prognostic significance
P 7 X10-5
32
Do sequence polymorphisms at p53 binding sites
have functional consequences?Discovery of
genetic susceptibility loci to cancer
Consequences of the observations to populations

• Kartiki Desai, Lee Yew Kok and JJ Liu

33
Identification of polymorphisms within p53 motifs
237 binding Sites with p53 motif
Blast against dbSNP database
91
215 binding Sites with 967 SNPs
6
14 SNPs are within the motif
12 SNPs screened in CEPH Lymphoblast Cell Lines
5 SNPs are polymorphic in LCLs
2
34
Preliminary Analysis of Two Motif SNPs
rs1860746 GAACATGTTCAGACATT/GCTC
rs13249982 AA/GACTTGCCTGGGCCTGTCC
T 20 G 80
A 88 G 12
Location FLJ32440
Location PRKAG2
GM07057 A GM12767 G
GM07057 T GM12767 G
35
Experimental DesignDo the motif SNPs affect
binding and gene regulation?
Two cell lines with different motif Sequences
No drug treatment
6 or 10 hrs 5Fu Treatment
DNA for CHIP Analysis
RNA for Expression Analysis
36
Preliminary Study both binding site
polymorphisms alter p53 Binding
rs1860746_PRKAG2
Rs13249982_FLJ32440
GM12767 (G)
GM07057 (A)
GM12767 (G)
GM07057 (T)
Quantitative Chromatin Immunoprecipitation
37
P53 binding Motif SNPs at the PRKAG2 locus
rs1860746 GAACATGTTCAGACATT/GCTC
PRKAG2 AMPKg
T 20 G 80
Location PRKAG2
AMPKg
AMPKb
AMPKa
Mutations in AMPKg are causative in Wolf
Parkinson White syndrome and disorders
of glycogen storage in cardiac muscle
Catalytic domain
GM07057 T GM12767 G
38
Motoshima et al. J Physiol. 2006 July 1 574(Pt
1) 6371.
Metabolic Stress Exercise Ischemia Glucose
deprivation
AMPK is also involved in cell growth
and proliferation Role in Cancer?
39
Reporter assays show p53 dependent allele
specific transcription
40
Genotypes of the p53 binding motif SNP influence
PRKAG2 (AMPK) associated gene repression
rs1860746_PRKAG2

• 13 cell lines with different
• genotypes at rs1860746
• 10 hrs drug treatment
• vs. no drug treatment
• Triplicate real-time PCR
• analysis for each cell line

Fold Down-Regulation
41
5FU
MUT
WT
TT
GG
Western blot analysis of AMPK after 5FU shows
allele associated changes
Time
42
Genotoxic stress
Transactivation
p21
p53
cell cycle arrest
Apoptosis
Is this regulatory SNP for AMPKg involved in
human cancer?
43
CAHRES Cohort A Population-based Swedish Study
of Breast Cancer Per Hall, Keith Humphreys.
Karolinska Drawn from all Postmenopausal women,
Swedish born, residing in Sweden between 1993 to
1995 1534 breast cancers cases 1504 controls
Helsinki University Familial and Sporadic Breast
Cancer Cases. Heikkinen, Aaltonen, Muranen,
Aittomäki, von Smitten, Nevanlinna, Helsinki
University Drawn from women with familial and
sporadic breast cancer presenting to the Helsinki
University Central Hospital residing in Finland
between 1997 to 2004 2244 breast cancers cases
1256 controls
44
The rs1860746 locus where p53 exerts its
regulatory control on AMPK confers cancer risk
in human populations
JJ Liu, Per Hall, Heli Nevanlinna
45
Hypothetical Consequences

• Genotoxic or nutritional insult
• P53 induction
• AMPK downregulated (WT)
• AMPK sustained (MUT/TT 5 population)
• TT allele homozygotes more susceptible to cancer
• Better survival of cancer cells under hypoxic
  conditions?
• TT allele homozygotes more responsive under
  metabolic stress?
• Extent of infarction after MI. Modifier of
  outcome
• Arrhythmias after MI (WPW syndrome)?

46
DNA Technologies and Strategies in Cancer
Medicine IMPAKT Brussels May 6, 2009
Edison Liu liue_at_gis.a-star.edu.sg 65-6478-8007
47
Understand how the cell decides which ERE to use

• Do good binding sites interact with other
• binding sites?
• Chromatin interaction analysis-pair end diTag
• ChIA-PET for an unbiased, comprehensive
• map of binding site interactions

Edwin Cheung
Yijun Ruan
48
The ChIA-PET methodology

• Cross-link cells
• Sonicate chromatin
• ChIP

B
GGCCGCGATATCGGATCCAAC CGCTATAGCCTAGGTTG
Add linker
MmeI

• Circularize DNA
• self-ligation
• Inter-ligation

Inter-ligation
Self-ligation

• Extract PETs
• MmeI digestion

• Purify PETs
• Amplify PETs
• Sequence PETs
• Map PETs

ChIA-PET
49
gt2/3 of all ERa binding sites are involved in
distant Interactions (looping)
Whole genome ER bindings and long range (LR)
Interactions In human breast cancer cells
50
GREB1
RNAPII ChIP-PET
1
2
3
4
5
6
ER ChIA-PET inter-ligation PETs
Complex interactions are valid
51
ERa-mediated interactions are functional
and identify more ER responsive genes
Interactions are associated with up-regulated
genes
Guillaume Bourque
52
Recruitment of Pol II at ERa-binding sites
E2F6
EST
GREB1
ChIP-chip data
ER binding sites
ERE motif
ER interactions
RNAPII ChIP-PET
Pan You Fu
53
DNA looping is a common phenomenon involved in TF
function
54
Complex interactions with stronger binding sites
at distal regions than target gene
promoters115 kb
55
Anchor site genes should be coordinately
expressed whereas loop genes may not.
56
A special class of anchor gene Enclosed Anchor
Gene
5 kb
57
Enclosed anchor genes are enriched for
up-regulated genes
58
How can we explain ER facts of life?

• ER binds only to a fraction of response element
  motifs in the genome
• Bona fide binding sites tend to be gene centric,
  but bind at great distances from TSS both 5, 3,
  and intragenic
• ER binding on genomic DNA oscillates with
  discernable periodicity
• Poor evolutionary conservation of bona fide ER
  binding sites
• ER binding sites loop at great distances
• Most ER binding sites also recruit RNAPolII
• ER binding site topology associated with gene
  induction and not gene repression

59
Hypothesis Equilibrium distribution of ER
improves local concentration of receptor and
provides boundaries for an oscillator
E R
60
Mechanistic parsimony in topologyCycling
transcriptional machinery by loops