systems pharmacology model to control cardiac fibrosis...see poster. zeigler+ j mol cell cardiol...
TRANSCRIPT
Systems pharmacology model to control cardiac fibrosis
Jeff SaucermanIMAG MSM March 2019
Signaling complexity hinders targeting of fibrosis after myocardial infarction
Objective: Develop computational models that predict and mechanistically explain therapeutic strategies to control myofibroblast activation and fibrosis
Travers et al Circ Res 2016Stuart et al JMCC 2016
Every year, ~1M Americans have a myocardial infarction
Reconstruction of the fibroblast signaling network
89 nodes, 129 reactions reconstructed from 177 papers2+ citations per reaction, emphasis on cardiac fibroblasts
Cytokines TGFβBeta-adrenergic
Integrin
Zeigler+ J Mol Cell Cardiol 2016
From network structure to dynamics: logic-based differential equations
• Proteins/genes in units of normalized activity, not concentration• Reactions use normalized-Hill functions and AND/OR logic gates.• Default parameter values: EC50,(0.5), Hill coeff (1.4), rxn weight (1) time
constant (signaling: 10 s, transcription: 1 h, translation: 10 h)Kraeutler+ BMC Sys Bio 2010
𝑓𝑓𝑎𝑎𝑎𝑎𝑎𝑎 𝐴𝐴 =𝐴𝐴𝑛𝑛
𝐸𝐸𝐶𝐶50𝑛𝑛 + 𝐴𝐴𝑛𝑛𝑑𝑑𝐶𝐶𝑑𝑑𝑑𝑑
=1𝜏𝜏𝐶𝐶
(𝑊𝑊𝐴𝐴𝐶𝐶 𝑓𝑓𝑎𝑎𝑎𝑎𝑎𝑎 𝐴𝐴 𝑌𝑌𝑀𝑀𝐴𝐴𝑀𝑀,𝐶𝐶 − 𝐶𝐶)
𝐴𝐴𝐴𝐴𝐴𝐴 𝑋𝑋,𝑌𝑌 = 𝑓𝑓 𝑋𝑋 𝑓𝑓 𝑌𝑌𝑂𝑂𝑂𝑂 𝑋𝑋,𝑌𝑌 = 𝑓𝑓 𝑋𝑋 + 𝑓𝑓 𝑌𝑌 − 𝑓𝑓 𝑋𝑋 𝑓𝑓 𝑌𝑌
Z
ANDX Y
Z
ORX Y
Predicted network response to TGFβTGFβ
Logic-based differential equation model
ET1IL6
Angela Zeigler
latent TGFβ
Methods: Kraeutler et al, BMC Sys Bio 2010
Model accurately predicts 66 of 82 input-output relationships
other outputs
Cardiac fibroblast validation data from 34 papers not used for model buildingValidation accuracy = 80%; see poster
Zeigler+ J Mol Cell Cardiol 2016
TGFβRi: SD208
Predicted and validated role for TGFβautocrine loop in mechano-induced αSMA
3D culture of cardiac fibroblasts in either mechanically restrained or floating collagen gelsGreen: αSMA; purple: DAPI
Will Richardson
Zeigler+ J Mol Cell Cardiol 2016
TGFβ stretch
αSMA
Applying the network model to predict matrix remodeling dynamics after
myocardial infarction9 model inputs driven by experimentally measured cytokine/hormone dynamics after myocardial infarction
Post-MI experimental data from rator human hearts, from infarct zone where available
0 1 2 3 4 5 6 7 8Time (Weeks)
0
0.1
0.2
0.3
0.4
0.5
0.6
Norm
alize
dIn
put L
evel
1
1.5
2
FoldC hange
inv iv o
Angiotensin II
Idealized CurveScheiffer 1994White 2001
0 1 2 3 4 5 6 7 8Time (Weeks)
0
0.1
0.2
0.3
0.4
0.5
0.6
Norm
alize
dIn
put L
evel
0
5
10
15
20
FoldC hange
inv iv o
TGFBIdealized CurveHerskowitz 1995Hao 1999
0 1 2 3 4 5 6 7 8Time (Weeks)
0
0.1
0.2
0.3
0.4
0.5
0.6
Norm
alize
dIn
put L
evel
0
20
40
60
FoldC hange
inv iv o
IL6Idealized CurveHerskowitz 1995Deten 2002
0 1 2 3 4 5 6 7 8Time (Weeks)
0
0.1
0.2
0.3
0.4
0.5
0.6
Norm
alize
dIn
put L
evel
0
5
10
15
20
FoldC hange
inv iv o
IL1Idealized CurveHerskowitz 1995Deten 2002
0 1 2 3 4 5 6 7 8Time (Weeks)
0
0.1
0.2
0.3
0.4
0.5
0.6
Norm
alize
dIn
put L
evel
0
2
4
6
8
10 FoldC hange
inv iv o
TNFaIdealized CurveHerskowitz 1995Heba 2001
0 1 2 3 4 5 6 7 8Time (Weeks)
0
0.1
0.2
0.3
0.4
0.5
0.6
Norm
alize
dIn
put L
evel
1
1.5
2
FoldC hange
inv iv o
NEIdealized CurveWhite 2001
0 1 2 3 4 5 6 7 8Time (Weeks)
0
0.1
0.2
0.3
0.4
0.5
0.6
Norm
alize
dIn
put L
evel
1
1.5
2
FoldC hange
inv iv o
ET1Idealized CurveWhite 2001
0 1 2 3 4 5 6 7 8Time (Weeks)
0
0.1
0.2
0.3
0.4
0.5
0.6
Norm
alize
dIn
put L
evel
0
2
4
6
8
10 FoldC hange
inv iv o
BNP
Idealized CurveWhite 2001
0 1 2 3 4 5 6 7 8Time (Weeks)
0
0.1
0.2
0.3
0.4
0.5
0.6
Norm
alize
dIn
put L
evel
1
1.5
2
2.5
3 FoldC hange
inv iv o
PDGFIdealized CurveZhao 2011
Experimental validation of predicted collagen mRNA and protein outputs after
myocardial infarction
Experimental data from rat infarcts:Deten 2001Zimmerman 2001Fomovsky 2010
Network model coupled to single ODE model with experimentally measured time-dependent MMP activity and fibroblast proliferation
0 1 2 3 4 5 6 7 8 9Time (weeks)
0
0.05
0.1
0.15
0.2
0.25
Mod
el
0
0.5
1
1.5
Experimental
(FoldChange)
Collagen I mRNAModel PredictionRat Infarct Deten 2001Rat Infarct Zimmerman 2001
0 1 2 3 4 5 6 7 8 9Time (weeks)
0
0.05
0.1
0.15
0.2
0.25
Mod
el
0
0.5
1
1.5
Experimental
(FoldChange)
Collagen III mRNAModel PredictionRat Infarct Deten 2001Rat Infarct Zimmerman 2001
Simulating FDA approved drugs targeting fibroblast signaling network
25 FDA approved drugs that uniquely target 24 network nodes
Competitive inhibitors/activators
Non-competitive inhibitors/activators
shift [X] adjust reaction weights Dose: 80% of saturation
Virtual screen for drugs that modulate fibroblasts after myocardial infarction
Time post-MI:
Fraccarollo et al Hypertension 2015
Chu et al Card Res 2012
Leuder+ Circ HF ‘15
Network mechanisms contributing to fibrosis induced by arsenic trioxide
Ani Chandrabhatla
Chu et al Card Res 2012
Network mechanisms contributing to fibrosis attenuated by Entresto
Ani Chandrabhatla
Entresto
Rats 4 weeks post-MI Trivedi+ JAHA 2018
0 1 2 3 4 5 6 7 8 9Time (Weeks)
0
5
10
15
20
25
30
Col
lage
nAr
ea F
ract
ion
(%) Untreated
Entresto
Model prediction after MI
Model prediction 1 week after MI
Integrated agent-based and signaling network model of fibrosis
4
Michaela Rikardw. Peirce-Cottler, Holmes
Summary• Developed and validated a large-scale model of cardiac
fibroblast signaling• Validated model prediction that stretch-mediated
myofibroblast differentiation involves a TGFβ autocrine loop• Framework for predicting in vivo signaling dynamics and
effects of drugs after myocardial infarction• Predicted pro- and anti-fibrotic FDA approved drugs (e.g.
arsenic trioxide, nitrates, Entresto)
AcknowledgementsSaucerman LabAni ChandrabhatlaBryan ChunSteve ChristiansenDeborah FrankMonika GrabowskaAli KhalilimeybodiCody NarcisoAnders NelsonBethany WissmannLaura WooAngela Zeigler
CollaboratorsJeff Holmes (UVA)Shayn Peirce-Cottler (UVA)Michaela Rikard (UVA)Will Richardson (Clemson)
Funding:
Posters on automated validation, fibrosis gene therapies
Grad student and post-doc positions availablehttp://bme.virginia.edu/saucerman
@sauce_lab