ibm research icahn school of medicine at mt sinai … · j.l. borges: el hacedor, 1960. all models...
TRANSCRIPT
Introduction to Systems Biology of Cancer
Lecture 4
Gustavo Stolovitzky
IBM Research
Icahn School of Medicine at Mt Sinai
DREAM Challenges
Classification of cancer
From MIT Course: Statstical
Learning Theory and Applications
Modeling in biological processes of
relevance in cancer
Modeling
"… In that Empire the Art of Cartography was so Perfect that the map of a single province occupied a whole city, and the map of the Empire, a whole Province. With the passing of time, these Huge Maps wouldn't be enough and the Colleges of the Cartographers erected a Map of the Empire that equaled in width the Empire itself…”
J.L. Borges: El Hacedor, 1960.
All models are wrong … but some are useful.
Attributed to George Box
Roles of p53
• Transcription factor
• Central role in defending genomic stability
• Decides on cell cycle arrest and apoptosis
• Implicated in over 50% of cancers
Cell cycle
arrest
p53 is highly regulated
ATM
COP-1 PIRH-2
WIP-1
P38
MAPK
MDM-2
p53
MDM-2
ARF E2F-1
Rb
p21
Cyclin E—cdk-2
SIAH-1
b-catenin
PTEN
AKT
Cyclin G
Cyclin G
PP2A
PIP-3
Bar-Or et al., PNAS, 2000
Reflects population but not single cells
Protein level after irradiation (IR)
p53 – MDM2 auto-regulation
Oscillations are not damped at single cell level
Coordinated oscillations of
P53 and MDM2
Lahav et al., Nature Genetics, 2004
P53 protein
Digital Clock: individual cells
Lahav et al., Nature Genetics, 2004
Digital behavior at single cell level: mean number of pulses
but not the amplitude or frequency depends on input signal.
Fraction of cells with zero, one,
two or more pulses as a function
of -IR dose:
Pulse width and height as a function
of -IR dose:
Digital Clock: individual cells
First pulse Second pulse
Pathways to oscillations
From Ciliberto, Novak, Tyson model, cell-cycle, 2005
p53
MDM-2
PTEN
AKT
PIP-3
Pathways to oscillations
p53
MDM-2
mdm2 rna – transport out of nucleus
MDM2 translation – transport into nucleus
x1
a x2
x3
Xn-1
xn
… a
a
w2
a
l
l l
l
l
a
x
1
x
n w2
l l
w1 (time delay t)
nn /)2(
2
1
w
aaw
?
=
( )2/1
2
wlw
t
n
n
n/1
2
a
waw
l/w
tw
w
l
wltw arcsin
)/(1
22/12
Basic structure of the model
DNA
damage
initiation
& repair
ATM:
DNA
damage
detection
P53-
MDM2
oscillator IR
Cell Cycle
Arrest ,
Apoptosis
and DNA
Repair
?
Modeling digital behavior
• Distribution of initial DSBs ~ Poisson Distribution
• Mean of number of DSBs proportional to IR dose (30-40 Gy-1 cell -1)
Rothkamm & Löbrich, PNAS, 2003 DSB
Repair of double strand breaks (DSBs)
3 min
24 h
• Biphasic repair process: rapid repair of simple lesions + slower repair
of complex lesions
• Two repair mechanisms: NHEJ (Non-Homologous End-Joining) & HR
(Homologous Recombination)
Löbrich et al., PNAS, 1995 Rothkamm et al., MCB, 2003
Two-Lesion-Kinetics (TLK)
• Limiting pool of repair proteins
• DSB-enzyme complexes necessary for DNA damage repair
RP: repair protein (Mre11/Rad50/Nbs1 cmplx)
D: intact DSB
C: DSB-enzyme complex
F: fixed DSB
Model: stochastic TLK of DSB repair
A B C
0 500 1000 15000
5
10
15
20
DS
Bc
Time (min)
a b
0 500 1000 15000
0.2
0.4
0.6
0.8
1
DS
B+
DS
Bc
Time (min)
a b
500 1000 15000
20
40
60
80
His
togra
m o
f C
CP
Time (min)
Implemented using Monte-Carlo method:
CCT: critical crossing times
Remaining DSBs DSB complexes Distribution of CCT
Simulation: DNA repair process
DNA
damage
initiation
& repair
ATM:
DNA
damage
detection
P53-
MDM2
oscillatorIrradiation
Cell Cycle
Arrest and
DNA Repair
?
DNA
damage
initiation
& repair
ATM:
DNA
damage
detection
P53-
MDM2
oscillatorIrradiation
Cell Cycle
Arrest and
DNA Repair
?
Ataxia telangiectasia mutated (ATM): mutated in disease AT, a human
genetic disorder characterized by neural degeneration, immunodeficiency,
sterility, cancer predisposition, etc.
Basic structure of model
• Dimer in normal cells
• Intermolecular autophosphorylation
• Direct activation by DSBs
• Nucleation formed by DSB and ATM*
Bakkenist & Kastan, Nature 2003
ATM activation
ATMD
kundim
kdim
ATM ATM*
kaf
kar
DSB complex
ATMD
kundim
kdim
ATM ATM*
kaf
kar
DSB complex
Where
ATMD: ATM dimer
ATM : inactive ATM monomer
ATM* : active ATM monomer
2ATMD+ATM+ATM*=ATMT
and C is DSB complex
Model: ATM activation
10 -2
10 -1
10 0
10 1
0
0.2
0.4
0.6
0.8
1
IR dose ( Gy )
Concentr
ation o
f A
TM
*
0 500 1000 1500 0
0.2
0.4
0.6
0.8
1
Concentr
ation o
f A
TM
*
Time (min)
Bakkenist & Kastan, Nature 2003
Time response:
ON-to-OFF signal
Correlation b/w ATM* & IR
1 2 3 4 5 6 7 8 9 10 111
1.5
2
0
0.1
0
.2
0.3
0
.4
0.5
0
.6
0.7
0
.8
0.9
Gy
Normalized by ATMT
n2.2
sim theo
Simulation: Switch like behavior of ATM*
DNA
damage
initiation
& repair
ATM:
DNA
damage
detection
P53-
MDM2
oscillatorIrradiation
Cell Cycle
Arrest and
DNA Repair
?
DNA
damage
initiation
& repair
ATM:
DNA
damage
detection
P53-
MDM2
oscillatorIrradiation
Cell Cycle
Arrest and
DNA Repair
?
Basic structure of model
2])([22
53
53253
53
5353
53
5353
53
5325353
53
2)](53[
)](53[2
5353
*
*
2222
**
*
53
***
**
535353
2*
*
22
5353
*
MDMKATM
ATMmdmr
dt
dMDM
KTP
TPMDMTPk
KTP
TPATMk
dt
dTP
KTP
TPATMkTPk
KTP
TPMDMTPpr
dt
dTP
mdmKtTP
tTPks
dt
dmdm
psdt
dp
a
MDMMDMMDMMDM
dTPrp
p
fp
p
fprp
d
TPTPTP
mdmnn
n
mdmmdm
pp
t
t
mRNA: p53, mdm2
Protein: TP53 (inactive), TP53* (active / phosphorylated), MDM2
n=4
p53 – Mdm2 oscillator: equations
Lahav et al., Nature
Genetics 2004
0 500 1000 15000
1
2
3
4
5
TP53
mdm2
MDM2
Time (min)
Mo
lecula
r co
nce
ntr
atio
n (
nM
)
0
20
40
60
80
100
0 1 2
Number of pulses
Pe
rce
nta
ge
of ce
lls 10 Gy
5 Gy
2.5 Gy
0.3 Gy
0 Gy
0 1 2+0
20
40
60
80
100
Number of pulses
Pe
rce
nta
ge
of
ce
lls (
%)
0 500 1000 15000
1
2
3
4
5
Time (min)Mo
lecula
r co
nce
ntr
atio
n (
nM
)
TP53
mdm2
MDM2
0 500 1000 15000
1
2
3
4
5
Time (min)Mo
lecula
r co
nce
ntr
atio
n (
nM
)
TP53
mdm2
MDM2
A B
C D
Mole
cula
r in
ten
sity (
fold
)
Predict drop of MDM2 at the
beginning of time course
0
1
2
3
4
5
0 100 200 300 400 500 600
MD
M2
Time (min)
MDM2
0 15’30’1h 2h 3h 4h 5h 6h 7h 8h 10h
Experiment
Time (min)
Response to 5Gy
Complete Model Results
0 500 1000 15000
1
2
3
4
5
Time (min)
Mole
cula
r in
ten
sity (
fold
)
TP53
mdm2
MDM2
0 500 1000 15000
1
2
3
4
5
Time (min)
Mole
cula
r in
ten
sity (
fold
)
TP53
mdm2
MDM2
0 500 1000 15000
1
2
3
4
5
Time (min)
Mole
cula
r in
ten
sity (
fold
)
TP53
mdm2
MDM2
Stochasticity in oscillation: IR of 5 Gy
induces one, two or three oscillations
Note: molecular intensity is
normalized by respective
basal concentration.
Complete Model Results
Lahav et al., Nature
Genetics 2004
Number of pulses increases as IR dose increases
Less stochasticity than experiment
Complete Model Results
0
20
40
60
80
100
0 1 2
Number of pulses
Pe
rce
nta
ge
of
ce
lls
Number of pulses
Perc
enta
ge o
f cells
(%
)
0 1 2+0
20
40
60
80
1000 Gy
0.3 Gy
2.5 Gy
5 Gy
10 Gy
*
0 2 4 6 8 10 120
100
200
300
400
500
600
Irradiation dose (Gy)
Period (
min
)
0 2 4 6 8 10 120
100
200
300
400
500
600
Irradiation dose (Gy)
Period (
min
)
0 2 4 6 8 10 120
0.5
1
1.5
Irradiation dose (Gy)
Puls
e h
eig
ht
0 2 4 6 8 10 120
0.5
1
1.5
Irradiaiton dose (Gy)
Puls
e h
eig
ht
* *
A B
C D
Period as function of IR dose:
Pulse height as function of IR dose:
Digital behavior
Pulse width and height as a function of
-IR dose:
P53
GAPDH
0 15’30’1h 2h 3h 4h 5h 6h 7h 8h 10h
Simulation
Simulating a cell population
P53
Mdm2
Fo
ld c
ha
ng
e
0
3
6
9
12
15
0 100 200 300 400 500 600
Time (min) 0 500 1000 1500 2000
0
1
2
3
4
5
Time (min)Mole
cula
r concentr
ation (
nM
)
p53 with PS
Mdm2 with PS
p53 no PS
Mdm2 no PSF
old
ch
an
ge
Lahav et al., Nature Genetics 2004
0 500 1000 15000
1
2
3
4
5
TP53
mdm2
MDM2
Time (min)
Mole
cula
r concentr
ation (
nM
)
0
20
40
60
80
100
0 1 2
Number of pulses
Pe
rce
nta
ge
of
ce
lls 10 Gy
5 Gy
2.5 Gy
0.3 Gy
0 Gy
0 1 2+0
20
40
60
80
100
Number of pulses
Perc
enta
ge o
f cells
(%
)
0 500 1000 15000
1
2
3
4
5
Time (min)Mole
cula
r concentr
ation (
nM
)
TP53
mdm2
MDM2
0 500 1000 15000
1
2
3
4
5
Time (min)Mole
cula
r concentr
ation (
nM
)
TP53
mdm2
MDM2
A B
C D
Mole
cula
r in
ten
sity (
fold
)
Time (min)
Response to 5Gy
DNA
damage
Digital response of tumor suppressor p53 to
IR
Ma, Wagner, Rice, Hu, Levine and Stolovitzky, A plausible model for
the digital response of p53 to DNA damage, Proc. Natl. Acad. Sci. U S A.
102, 14266 (2005).
Wagner; Ma; Rice; Hu; Levine; Stolovitzky, p53-Mdm2 loop
controlled by a balance of its feedback strength and effective dampening
using ATM and delayed feedback, IEE PROCEEDINGS SYSTEMS
BIOLOGY, 152, 3, 109-118 (2005).
Predictions
Basal
Protein
(Fast)
(Slow) ATM*
mRNA
p 53
P53-induced
transcription
53 p
DNA
Mdm2
DNA
Mdm2
Protein
Protein
p53
p53*
Delayt1
Basal
mRNA
Mdm2
Delayt2
Figure 4 (From Ma, Wagner et.al.) - Diagram of the
p53-Mdm2 oscillator. p53 is translated from p53
mRNA and inactive for induction of its targets.
Phosphorylated by ATM*, p53 becomes active (p53*),
and able to transcribe (after a time delay) Mdm2 which
also has a basal transcription rate. Mdm2 protein
promotes a fast degradation of p53 and a slow
degradation of p53*. In addition to a basal self-
degradation, Mdm2 is degraded by a mechanism
stimulated by ATM*.
Figures 4 and 8 from Ma, Wagner et al.,
A B
Figure 8 (From Ma, Wagner, et. al.) - One-
dimensional bifurcation diagrams of steady-state
p53 versus single parameter variation of Mdm2
basal transcription rate (A) or p53 basal
transcription rate (B). The stable equilibrium is
represented by solid line. The lower and upper
bounds of stable oscillation are represented by
paired dotted lines.
0 2 4 6 80
5
10
15
20
25
30
p5
3 c
on
ce
ntr
atio
n (
fold
)
equilibrium
oscillation
p53 basal transcription (fold)0 1 2 3
0
2
4
6
8
mdm2 basal transcription (fold)
p53 c
oncentr
atio
n (
fold
)
equilibrium
oscillation
Specific Predictions About Bifurcations
0 1 2 30
2
4
6
8
mdm2 basal transcription (fold)
p53 c
oncentr
atio
n (
fold
)
equilibrium
oscillationA
0 2 4 6 80
5
10
15
20
25
30
p5
3 c
on
ce
ntr
atio
n (
fold
)
equilibrium
oscillation
p53 basal transcription (fold)
B
Fig. 8 of Ma et al., PNAS, 2005
P53 b
asal tr
anscription
Mdm2 basal transcription
Low High
Lo
w
Hig
h
Oscillatory region
Non-oscillatory region
WP
Mdm2 SNP309: T → G SNP Increases Mdm2
•Quantitative data at cellular level
explained by a point mutation.
•T → G increases affinity for Sp1
•MCF-7 cell line is T/G
•SNP status correlates with cancer risk
Bond et al, Cell, 2004.
Frequency
T/T 48%
G/T 40%
G/G 12%
T/T T/G G/G
Mdm2
mRNA
1X 8X
Mdm2
Protein
1X 2X 4X
Predictions T/T & T/G Oscillate, G/G Does Not P
53 b
asal tr
anscription
Mdm2 basal transcription
Low High
Lo
w
Hig
h
Oscillatory region
Non-oscillatory region
0 1 2 30
2
4
6
8
mdm2 basal transcription (fold)
p53
conc
entra
tion
(fold
)
equilibrium
oscillation
T/T
1X
T/G
2X
G/G
4X
T/T
1X
T/G
2X G/G
4X
T/T & T/G Cell Lines H460 and MCF-7 Oscillate
T/T
T/G
Cells wild type (T/T, H460) or heterozygous (T/G, MCF-7) for SNP309
oscillate in response to 5Gy IR.
NB: First p53 peak at 2 hours, second peak at 7 to 8 hours.
G/G Cell Lines A875 and Manca Do Not Oscillate
Two cell lines homozygous (G/G) for SNP309 do NOT oscillate in response to
5Gy IR.
Manca A875
NB: p53 peaks at 10 to 12 hours.
Model Predicts Hopf Bifurcations WRT p53 Production P
53 b
asal tr
anscription
Mdm2 basal transcription
Low High
Lo
w
Hig
h
Oscillatory region
Non-oscillatory region
0 2 4 6 80
5
10
15
20
25
30
p5
3 c
on
ce
ntr
atio
n (
fold
)
equilibrium
oscillation
p53 basal transcription (fold)
Oscillations Require Intermediate p53 Production Rates
P53-null H1299-SW24 cells expressing p53 under tetracycline-
dependent promoter.
0 2 4 6 8 10 120
0.5
1
1.5
2
2.5
3
Re
lative
in
du
ction
fo
ld o
f M
dm
2
0 2 4 6 8 10 120
1
2
3
4
5
Re
lative
in
du
ction
fo
ld o
f p
53
h
h h
h
Low Tetracycline
High P53
Medium Tetracycline
Medium P53
High Tetracycline
Low P53
p53 Mdm2
Experi
ment
Model