COMPUTATIONAL MODELLING IN MEDICINE Pascal Schulthess 04.03.2016 Thermodynamic modeling explains the regulation of CYP1A1 expression in the liver Thesis defense
C O M P U TAT I O N A LM O D E L L I N G I NM E D I C I N E
Pascal Schulthess
04.03.2016
Thermodynamic modeling explains the regulation of CYP1A1 expression in the liver Thesis defense
C O M P U TAT I O N A LM O D E L L I N G I NM E D I C I N E
Pascal Schulthess
04.03.2016
Thermodynamic modeling explains the regulation of CYP1A1 expression in the liver Thesis defense
Exposure to dioxin through food uptake
welt.de, wikipedia.org, beyondthedish.wordpress.com
CYPs biotransform toxins
Cytochrome P450s (CYPs)
‣ major enzymes in detoxification
‣ harbor broad substrate specificity
‣ increase hydrophilicity of compounds to ease their excretion in bile or urine
wikipedia.org
Dioxin
CYP1A1
Dioxin induces zonated expression of CYP1A1
Braeuning, A. & Schwarz, M. Biol. Chem. 391, 139–148 (2010) Braeuning, A., Köhle, C., Buchmann, A. & Schwarz, M. Toxicol. Sci. 122, 16–25 (2011).
rela
tive
mR
NA
exp
ress
ion [a
. u.]
0
3
6
9
100
150
200
250
300
Dioxin – +
CYP1A1primary mousehepatocytes
n=5
Dioxin induces zonated expression of CYP1A1
CYP1A
WT
P
C
Braeuning, A. & Schwarz, M. Biol. Chem. 391, 139–148 (2010) Braeuning, A., Köhle, C., Buchmann, A. & Schwarz, M. Toxicol. Sci. 122, 16–25 (2011).
rela
tive
mR
NA
exp
ress
ion [a
. u.]
0
3
6
9
100
150
200
250
300
Dioxin – +
CYP1A1primary mousehepatocytes
n=5
Dioxin induces zonated expression of CYP1A1
CYP1A
WT
P
C
β-ca
teni
n k.
o.
P
C
Braeuning, A. & Schwarz, M. Biol. Chem. 391, 139–148 (2010) Braeuning, A., Köhle, C., Buchmann, A. & Schwarz, M. Toxicol. Sci. 122, 16–25 (2011).
rela
tive
mR
NA
exp
ress
ion [a
. u.]
0
3
6
9
100
150
200
250
300
Dioxin – +
CYP1A1primary mousehepatocytes
n=5
Dioxin and β-catenin signaling converge on CYP1A1 promoter
AhR
Dioxin
Arnt
Dioxin and β-catenin signaling converge on CYP1A1 promoter
AhR
Dioxin
Arntβ-catenin
TCF
Wnt
Dioxin and β-catenin signaling converge on CYP1A1 promoter
AhR
Dioxin
Arntβ-catenin
TCF
Wnt
Dioxin and β-catenin signaling converge on CYP1A1 promoter
AhR
Dioxin
Arnt
CYP1A1
F E D T C
β-catenin
TCF
Wnt
Dioxin and β-catenin signaling converge on CYP1A1 promoter
How are the two signals integrated?
AhR
Dioxin
Arnt
CYP1A1
F E D T C
β-catenin
TCF
Wnt
Dioxin and β-catenin signaling converge on CYP1A1 promoter
How are the two signals integrated?
How do binding sites cooperate?
AhR
Dioxin
Arnt
CYP1A1
F E D T C
β-catenin
TCF
Wnt
1. Introduction of the modeling framework
2. Model of synthetic promoter
3. Model of wild-type CYP1A1 promoter
Outline
1. Introduction of the modeling framework
2. Model of synthetic promoter
3. Model of wild-type CYP1A1 promoter
Outline
Modeling approaches for gene expression
Level of detail
(Un)directed graphsBayesian networksBoolean networks
⁝
(Nonlinear) ODEsPDEs⁝
Thermodynamicstate ensembles
Mathematical model combines signaling and promoter logic
+
Thermodynamic model Signaling model
AhR
β-catenin
Dioxin
1
2
Arnt
εBP
εAPεAB
KPKAKB
Thermodynamic state ensemble model of gene expression
KP
Sherman MS, Cohen BA, PLoS Comput Biol 8(3):e1002407 (2012) Bintu L, et al., Curr Opin Genet Dev 15(2):116–124 (2004)
Thermodynamic state ensemble model of gene expression
KP
Sherman MS, Cohen BA, PLoS Comput Biol 8(3):e1002407 (2012) Bintu L, et al., Curr Opin Genet Dev 15(2):116–124 (2004)
Thermodynamic state ensemble model of gene expression
KP
εAP
KA
Sherman MS, Cohen BA, PLoS Comput Biol 8(3):e1002407 (2012) Bintu L, et al., Curr Opin Genet Dev 15(2):116–124 (2004)
Thermodynamic state ensemble model of gene expression
KP
εAP
KA
εBP
εAB
KB
Sherman MS, Cohen BA, PLoS Comput Biol 8(3):e1002407 (2012) Bintu L, et al., Curr Opin Genet Dev 15(2):116–124 (2004)
1. Introduction of the modeling framework
2. Model of synthetic promoter
3. Model of wild-type CYP1A1 promoter
Outline
Model of synthetic promoters1× C 2× C 3× C
4× C 5× C 6× C
1× D 2× D 3× D
CYP1A1
F E D T C
Model of synthetic promoters1× C 2× C 3× C
4× C 5× C 6× C
1× D 2× D 3× D
Rel
ativ
e Lu
cife
rase
Act
ivity
[a.u
.]
Dioxin [nM]
5101520
5101520
0 0.5 5 50 250
5101520
0 0.5 5 50 250 0 0.5 5 50 250
CYP1A1
F E D T C
Model of synthetic promoters1× C 2× C 3× C
4× C 5× C 6× C
1× D 2× D 3× D
Rel
ativ
e Lu
cife
rase
Act
ivity
[a.u
.]
Dioxin [nM]
5101520
5101520
0 0.5 5 50 250
5101520
0 0.5 5 50 250 0 0.5 5 50 250
Exp.
CYP1A1
F E D T C
Model of synthetic promoters1× C 2× C 3× C
4× C 5× C 6× C
1× D 2× D 3× D
model parameterssignaling 3
C constructs 13 (28)D constructs 1 (9)
Rel
ativ
e Lu
cife
rase
Act
ivity
[a.u
.]
Dioxin [nM]
5101520
5101520
0 0.5 5 50 250
5101520
0 0.5 5 50 250 0 0.5 5 50 250
Exp.
CYP1A1
F E D T C
Model of synthetic promoters1× C 2× C 3× C
4× C 5× C 6× C
1× D 2× D 3× D
model parameterssignaling 3
C constructs 13 (28)D constructs 1 (9)
Rel
ativ
e Lu
cife
rase
Act
ivity
[a.u
.]
Dioxin [nM]
5101520
5101520
0 0.5 5 50 250
5101520
0 0.5 5 50 250 0 0.5 5 50 250
Exp.Fit
CYP1A1
F E D T C
Model parameters unravel TF interactions
Binding Energy [kBT]
-20 -4 -6 -8
Model parameters unravel TF interactions
Only first TF interacts with polymerase
Binding Energy [kBT]
-20 -4 -6 -8
Model parameters unravel TF interactions
Only first TF interacts with polymerase
Only nearby TFs cooperate
Binding Energy [kBT]
-20 -4 -6 -8
Model parameters unravel TF interactions
Only first TF interacts with polymerase
Only nearby TFs cooperate
Binding Energy [kBT]
-20 -4 -6 -8
19bp 49bp
156bp 292bp
Model parameters unravel TF interactions
Only first TF interacts with polymerase
Only nearby TFs cooperate
Binding Energy [kBT]
-20 -4 -6 -8
19bp 49bp
156bp 292bp
Rel
ativ
e Lu
cife
rase
Act
ivity
[a.u
.]
Dioxin [nM]
2
6
10
14
0 0.5 5 50 250
2
6
10
14
0 0.5 5 50 250
Reduced induction for 4+ binding sites
0.1
0.2
0
0.5
1
1.5
0
mean number of occupied binding sites
probability that first binding site is occupied
5
10
15
20
maximal induction
Reduced induction for 4+ binding sites
0.1
0.2
0
0.5
1
1.5
0
mean number of occupied binding sites
probability that first binding site is occupied
5
10
15
20
maximal induction
Reduced induction for 4+ binding sites
0.1
0.2
0
0.5
1
1.5
0
mean number of occupied binding sites
probability that first binding site is occupied
5
10
15
20
maximal induction
Reduced induction for 4+ binding sites
0.1
0.2
0
0.5
1
1.5
0
mean number of occupied binding sites
probability that first binding site is occupied
5
10
15
20
maximal induction
Reduced induction for 4+ binding sites
0.1
0.2
0
0.5
1
1.5
0
mean number of occupied binding sites
probability that first binding site is occupied
5
10
15
20
maximal induction
Reduced induction for 4+ binding sites
0.1
0.2
0
0.5
1
1.5
0
mean number of occupied binding sites
probability that first binding site is occupied
5
10
15
20
maximal induction
Reduced induction for 4+ binding sites
0.1
0.2
0
0.5
1
1.5
0
mean number of occupied binding sites
probability that first binding site is occupied
5
10
15
20
maximal induction
Reduced induction for 4+ binding sites
Sequestration responsible for reduced induction
0.1
0.2
0
0.5
1
1.5
0
mean number of occupied binding sites
probability that first binding site is occupied
5
10
15
20
maximal induction
Prediction of dioxin and β-catenin signaling integration
Dioxin [nM]
β-ca
t Act
ivity
[%]
ExperimentPrediction3x C
100
28
93
6753
00 0.5 5 50 250
2015105Relative Luciferase Activity [a.u.]
Prediction of dioxin and β-catenin signaling integration
Dioxin [nM]
β-ca
t Act
ivity
[%]
ExperimentPrediction3x C
100
28
93
6753
00 0.5 5 50 250
2015105Relative Luciferase Activity [a.u.]
Prediction of dioxin and β-catenin signaling integration
10093675328
0 0.5 5 50 250
Dioxin [nM]
β-ca
t Act
ivity
[%]
ExperimentPrediction3x C
100
28
93
6753
00 0.5 5 50 250
2015105Relative Luciferase Activity [a.u.]
Prediction of dioxin and β-catenin signaling integration
Integration of the two signals follows AND gate logic
10093675328
0 0.5 5 50 250
Dioxin [nM]
β-ca
t Act
ivity
[%]
ExperimentPrediction3x C
100
28
93
6753
00 0.5 5 50 250
2015105Relative Luciferase Activity [a.u.]
Conclusion for synthetic promoter
How do binding sites cooperate?‣ only first binding site interacts with polymerase ‣ only nearby binding sites cooperate
How are the two signaling pathways integrated?‣ integration follows an AND gate logic
Conclusion for synthetic promoter
How do binding sites cooperate?‣ only first binding site interacts with polymerase ‣ only nearby binding sites cooperate
How are the two signaling pathways integrated?‣ integration follows an AND gate logic
Conclusion for synthetic promoter
How do binding sites cooperate?‣ only first binding site interacts with polymerase ‣ only nearby binding sites cooperate
How are the two signaling pathways integrated?‣ integration follows an AND gate logic
Does the same hold true for the
wild-type CYP1A1 promoter?
1. Introduction of the modeling framework
2. Model of synthetic promoter
3. Model of wild-type CYP1A1 promoter
Outline
Model of wild-type CYP1A1 promoterT CT
WT
TD TE
TF CTD CTE CTF TDE
TDF TEF TD CE
TDF C TF CE
TDFEDF CE
Model of wild-type CYP1A1 promoterT CT
WT
TD TE
TF CTD CTE CTF TDE
TDF TEF TD CE
TDF C TF CE
TDFEDF CE
Dioxin [nM]
Rel
ativ
e Lu
cife
rase
Act
ivity
[a.u
.]1
3
5
7
0 0.5 5 50 250
1
3
5
7
1
3
5
7
1
3
5
7
0 0.5 5 50 2501
3
5
7
0 0.5 5 50 250
0 0.5 5 50 250 0 0.5 5 50 250
0 0.5 5 50 250
1
3
5
7
Model of wild-type CYP1A1 promoterT CT
WT
TD TE
TF CTD CTE CTF TDE
TDF TEF TD CE
TDF C TF CE
TDFEDF CE
Experiment
Dioxin [nM]
Rel
ativ
e Lu
cife
rase
Act
ivity
[a.u
.]1
3
5
7
0 0.5 5 50 250
1
3
5
7
1
3
5
7
1
3
5
7
0 0.5 5 50 2501
3
5
7
0 0.5 5 50 250
0 0.5 5 50 250 0 0.5 5 50 250
0 0.5 5 50 250
1
3
5
7
Model of wild-type CYP1A1 promoterT CT
WT
TD TE
TF CTD CTE CTF TDE
TDF TEF TD CE
TDF C TF CE
TDFEDF CE
Experiment
model parameterssignaling 3
thermodynamic 16 (21)
Dioxin [nM]
Rel
ativ
e Lu
cife
rase
Act
ivity
[a.u
.]1
3
5
7
0 0.5 5 50 250
1
3
5
7
1
3
5
7
1
3
5
7
0 0.5 5 50 2501
3
5
7
0 0.5 5 50 250
0 0.5 5 50 250 0 0.5 5 50 250
0 0.5 5 50 250
1
3
5
7
Model of wild-type CYP1A1 promoterT CT
WT
TD TE
TF CTD CTE CTF TDE
TDF TEF TD CE
TDF C TF CE
TDFEDF CE
Experiment Fit
model parameterssignaling 3
thermodynamic 16 (21)
Dioxin [nM]
Rel
ativ
e Lu
cife
rase
Act
ivity
[a.u
.]1
3
5
7
0 0.5 5 50 250
1
3
5
7
1
3
5
7
1
3
5
7
0 0.5 5 50 2501
3
5
7
0 0.5 5 50 250
0 0.5 5 50 250 0 0.5 5 50 250
0 0.5 5 50 250
1
3
5
7
Model parameters unravel TF interactions
Binding Energy [kBT]0 -4 -8 -12 -16 -20-18-14-10-6-2
synthetic promoter wild-type CYP1A1 promoter
Model parameters unravel TF interactions
Binding Energy [kBT]0 -4 -8 -12 -16 -20-18-14-10-6-2
synthetic promoter wild-type CYP1A1 promoter
Model parameters unravel TF interactions
More and stronger TF-RNAP interactions
Binding Energy [kBT]0 -4 -8 -12 -16 -20-18-14-10-6-2
synthetic promoter wild-type CYP1A1 promoter
Model parameters unravel TF interactions
β-catenin is important interaction partner
More and stronger TF-RNAP interactions
Binding Energy [kBT]0 -4 -8 -12 -16 -20-18-14-10-6-2
synthetic promoter wild-type CYP1A1 promoter
Prediction of dioxin and β-catenin signaling integration
Dioxin [nM]
β-ca
teni
n Ac
tivity
[%]
ExperimentPrediction
WT
0 0.5 5 50 250
10093675328
0 0.5 5 50 250
10093
6753
28
0
RLA
[a.u
.]
54321
3x C
100
28
93
6753
0
20
15
10
5 RLA
[a.u
.]
Prediction of dioxin and β-catenin signaling integration
Dioxin [nM]
β-ca
teni
n Ac
tivity
[%]
ExperimentPrediction
WT
0 0.5 5 50 250
10093675328
0 0.5 5 50 250
10093
6753
28
0
RLA
[a.u
.]
54321
3x C
100
28
93
6753
0
20
15
10
5 RLA
[a.u
.]
Prediction of dioxin and β-catenin signaling integration
Dioxin [nM]
β-ca
teni
n Ac
tivity
[%]
ExperimentPrediction
WT
0 0.5 5 50 250
10093675328
0 0.5 5 50 250
10093
6753
28
0
RLA
[a.u
.]
54321
3x C
100
28
93
6753
0
20
15
10
5 RLA
[a.u
.]
Prediction of dioxin and β-catenin signaling integration
Gradual AND gate enables finer regulation
Dioxin [nM]
β-ca
teni
n Ac
tivity
[%]
ExperimentPrediction
WT
0 0.5 5 50 250
10093675328
0 0.5 5 50 250
10093
6753
28
0
RLA
[a.u
.]
54321
3x C
100
28
93
6753
0
20
15
10
5 RLA
[a.u
.]
Conclusion for wild-type CYP1A1 promoter
How do binding sites cooperate?‣ more complex interaction patterns ‣ increased importance of TF-RNAP interactions ‣ β-catenin is an important interaction partner
How are the two signaling pathways integrated?‣ integration follows an AND gate logic ‣ gradual signal integration enables a finer regulation of
promoter response
Conclusion for wild-type CYP1A1 promoter
How do binding sites cooperate?‣ more complex interaction patterns ‣ increased importance of TF-RNAP interactions ‣ β-catenin is an important interaction partner
How are the two signaling pathways integrated?‣ integration follows an AND gate logic ‣ gradual signal integration enables a finer regulation of
promoter response
Conclusion for wild-type CYP1A1 promoter
How do binding sites cooperate?‣ more complex interaction patterns ‣ increased importance of TF-RNAP interactions ‣ β-catenin is an important interaction partner
How are the two signaling pathways integrated?‣ integration follows an AND gate logic ‣ gradual signal integration enables a finer regulation of
promoter response
Model qualitatively predicts hepatic zonationEx
perim
ent
Pred
ictio
n
Dioxin [nM]
PC
PC
PC
P
CP
C
P
CP
C
Model qualitatively predicts hepatic zonationEx
perim
ent
Pred
ictio
n
Dioxin [nM]
PC
PC
PC
P
CP
C
P
CP
C
Conclusion
The interactions at the signaling level, together with the TF cooperativity in the CYP1A1 promoter enable the
spatial expression pattern observed in vivo.
Schulthess P, et al. (2015) Signal integration by the CYP1A1 promoter - a quantitative study. Nucleic Acids Research 43(11):5318–5330.
β-cateninDioxin
Acknowledgements
C O M P U TAT I O N A LM O D E L L I N G I NM E D I C I N E
Nils BlüthgenManuela Benary Torsten Gross Bertram Klinger Johannes Meisig Mattias Rydenfelt
Jörn Schmiedel Anja Sieber Besray Ünal Florian Uhlitz Franziska Witzel
Albert BraeuningLuise Kreft Alexandra Löffler Michael Schwarz Silvia Vetter