systems biology for drug discovery · key to future medical breakthroughs is systems biology, say...

Dr. Galina Lebedeva

12th March 2010

Systems Biology for

Drug Discovery as applied to NSAID safety problem

Key To Future Medical Breakthroughs Is Systems Biology,

Say Leading European Scientists

•  High Complexity of biochemical networks underlying cell functioning

•  If disturbed can result in a disease (diabetes, cancer..- network diseases)

•  Conventional approaches of biology are not suitable for the analysis of these elaborate webs of interactions, which is why drug design often fails

•  Knocking out one target in a pathway is not productive – disease will by-pass the drug (network robustness)

•  Systems Biology approaches (computational modelling and analysis) should be engaged to develop smarter therapeutic strategies and predict drug safety and efficacy.

ScienceDaily (Jan. 13, 2009)

Key To Future Medical Breakthroughs Is Systems Biology






Reality of molecular complexity

Pathways do not end … they interconnect

Health, disease and therapeutic intervention in terms of network modelling

•  Input – output behaviour of biological networks •  Health: balance of input and output, adequate response to

perturbation •  Disease: balance is disturbed, due to suppression or

activation of certain stages, resulting in inadequate output •  Goal of therapy: to restore the normal balance by targeting

key components/checkpoints of networks by drugs •  Therapeutic resistance – loss of the sensitivity of the

output to the drug

Key To Future Medical Breakthroughs Is Systems Biology,

Say Leading European Scientists






ScienceDaily (Jan. 13, 2009)

Systems medicine

•  aimed at exploitation of systems biology approaches to develop prognostic and predictive models for diagnostics and therapeutic applications

•  there is not yet an established arsenal of computational approaches, methods, and techniques applicable for model-based diagnostics and design of therapeutic strategies

•  specific techniques of model analysis need to be developed to allow for application of general models in the context of therapeutic research.

• Takes into account all properties of the biosystem (protein structure, mechanism, stoichiometry, dynamics and

regulation)

• Clearly describes the key properties of the biosystem in terms of easily understandable and measurable parameters (Vmax,

Km, Kd, Ki, IC50..)

• Reproduces all known responses of the biosystem to external and internal perturbations and therapeutic interventions

Kinetic Modeling (KM) Approach

Kinetic Modeling in Systems Biology, Demin, Goryanin Chapman & Hall/CRC, 2008

Kinetic modelling approach 1.  Pathway reconstruction and static model development:

elucidation of stoichiometry of the network, identifying key cross-talks and regulations

2. Generation of the system of ODEs describing dynamics of the metabolic/signalling network:

dx/dt=N·v(x;e,K) Here, x=[x1,…xm] is vector of compound concentrations and v=[v1,…vn] is vector of rate laws

3.  Modelling metabolic, signalling and transport processes: Detailed description of the catalytic cycles of key proteins, derivation of rate equations

4.  Prameterisation of the model (literature, experiments, fitting) 5.  Validation of submodels on the base of in vitro and cell

extract data 6.  Validation of the whole model using in vivo dynamic data

(genomics, proteomics, metabolomics)

Applica'onofthevalidatedmodeltoprac'calproblems:e.g.inpharmaceu'calindustry‐drugresistance,safety

Practical applications

•  Various analyses of the validated model, e.g.: –  In silico experiments to test various hypothesis on the

mechanisms of disease and drug action •  E.g. test how modifications within the network can affect its input-

output behaviour –  Local and global sensitivity analysis to generate ideas on

•  Drug targets •  Biomarkers •  Combination therapies

Challenges in translating theory into clinical practice:

•  Extrapolation to a multi-layered context, from molecular to cellular, tissue,… and organism level

•  High level of individual variability of the networks, e.g. in cancer – to be addressed by personalised medicine

•  Complex dynamical aspects –  pharmaco-kinetics and pharmaco-dynamics effects, –  drug scheduling, –  circadian rhythms…

•  Incomplete knowledge on the biological networks underlying disease onset and progression

•  limitations imposed by the number of elements which a tractable model can include

Context layers to consider…

•  Intracellular microenvironment –  regulation of protein activity by local concentration of substrates/

products and co-factors… •  Larger network context

–  Subsystem embedded in a larger network is subject to higher level regulation

•  Cell-specific protein expression and gene regulation –  Pathway structure and dynamics vary in different cell types…

•  Organ/ organism level effects – –  Spatial aspects

•  ….

“Too many pharmacological agents have entered into clinical practice

for which considerable and potential life-threatening outcomes

were recognized only AFTER a large number of patients had been treated”

Wall Street Journal – Thursday, January 26, 2006 quoting an editorial from same week’s New England Journal of Medicine

Drug Safety Problem

Example: Application of Systems Biology Approach to

(NSAIDs) safety problem

•  NSAIDs – popular drugs for pain relief and antipyretic, more recently started to be used in cancer and neurophysiology (depression).

•  Main target – COX1,2 •  COX1 – constitutive, COX2 – induced at inflammation •  Aspirin (targeting both COX1 and COX2)– risk of gastro-intestinal

bleeding at medium/high dose •  Selective COX-2 inhibitors (Coxibs) – developed to overcome GI

side effects: –  efficient in pain relief but with new dangerous side effects (heart

attacks) –  Vioxx withdrawal from the market – cost Merck $billions, with ongoing

legal costs –  FDA suggests Vioxx has contributed to >20 000 heart attacks &

sudden cardiac deaths during its stay on market •  The exact mechanism of NSAID action, and the origin of many

undesirable adverse effects still remain poorly understood.

Non-Steroidal Anti-Inflammatory Drugs • Aspirin

• Ibuprofen

• Naproxen

• Indomethacin

• Celebrex …

(NSAIDs) safety problem

Non-Steroidal Anti-Inflammatory Drugs • Aspirin

• Ibuprofen

• Naproxen

• Indomethacin

• Celebrex …

•  Static pathway reconstruction •  Intracellular level models:

–  Catalytic cycle of Drug Target (COX) –  NSAID action on COX –  Upstream and downstream metabolic and

signalling pathways involved in response to NSAIDs

•  Cell level models –  NSAID action on platelets –  NSAID action on endothelium cells –  Combined platelet-endothelium-plasma

•  Model of Blood circulation system •  Coupling with pharmacokinetic profiles

“Vioxx” project

Biosystems Informatics Institute, Newcastle Moscow States University, ISB SPb

Intracellular

Cell-type – specific

Organ level

Pathway reconstruction for biosynthesis of prostanoids and leukotriens NSAIDs inhibit PGH2 production

Arachidonate PGG2 PGH2

PGI2 PGE2 PGD2 TXA2

6-KetoPGF1α

6-KetoPGE1

PGF2α

PGA2 PGC2

PGB2

PGJ2

Dihydro-PGJ2

TXB2

Dihydro-TXB2

5-HPETE

LTA4

LTC4

LTD4

LTB4

20-OH-LTB4

20-COOH-LTB4

12-Keto-LTB4

Lecithin

15-HPETE

12-HPETE

5-HETE

2O2

NADPH

NADP

AccH2 Acc

FpH2

Fp

NADP

NADPH

O2

O2

1.14.99.1 1.14.99.1

5.3.99.4

1.1.1.189

5.3.99.3 5.3.99.2

5.3.99.5

1.1.1.184

1.1.1.188

1.13.11.34

1.13.11.34

Glutathione

Glutamate

4.4.1.20

2.3.2.2

3.3.2.6

1.11.1.7

3.1.1.4

1.13.11.33 1.13.11.31

1.14.13.30

PGHS-1,2 PGHS-1,2

PTGIS

PTGES2 PGDS

TBXAS1

DHRS4

CBR1 AKR1C3

ALOX5

ALOX12 ALOX15

PLA262D

ALOX5

LTC4S

GGT1

LTA4H

CYP4F3

LPO

NSAID

• PGI2 and PGD2 inhibit platelet aggregation; • TXA2 induces platelet aggregation; • PGE2 potentiates it

The overall response to NSAIDs results from complex interplay of inductions/inhibitions in different branches of prostaglandin synthesis and further signalling

Prostanoid

Prostanoid Receptors (GPCRs)

G-proteins

signaling pathways

PGI2 PGE2 PGD2 TXA2 PGF2α

Gq Gi Gs

Raf

MEK

MAPK

PLC

PKC Ca2+

AC

cAMP

PKA

CREB1

ATP

GSK-3

β-catenin

p38

Ras

IP EP1 EP2 EP3 EP4 FP DP1 TP DP2

IP3

PLA2

Rac PLD

PPARγ PPARγ

COX2 AA

Pathway reconstruction for prostanoid signalling network






“Vioxx” project


Intracellular


Organ level

The Cyclooxygenase Reaction

Arachidonic acid + 2O2 PGG2 + H2O PGG2 PGH2

The enzyme has two activities: Cyclooxygenase and Peroxidase

Cyclooxygenase (COX) is a membrane bound enzyme responsible for the oxidation of arachidonic acid to Prostaglandin G2 (PGG2) and the subsequent reduction of PGG2 to prostaglandin H2 (PGH2).

Kinetic model of COX-1/2 catalytic cycle (as for purified enzyme)

• Detailed ODE-based model for COX-1,2 catalysis, with all kinetic parameters accurately cross-validated. (>30 ODE, >60 rate equations, >25 parameters)

Heme in the POX site: Fe and PP

Tyr385

Substrate/ inhibitor binding site (COX)

Catalytic mechanism of drug target elucidated and understood Structural knowledge as a base for mechanistic description of COX catalysis

molecular level cellular level organ/organism level

From Model to Simulation

Implement the ODE set in DBSlove or/and MATLAB (via SBML conversion)

Write this as set of ordinary differential equations (ODEs), e.g.:

d[C4] dt

= v4 +v57+v53+v13-v1-v9

d[D5] dt

= v1 - v2 –v14-v29+v18

step 2

step 3

step 1

Transform the scheme into a set of reactions and reaction rates, e.g.:

R2: D5 → E4 ν2 = k2 • D5

R1: C4 → D5 ν1 = k1 • C4 • AA

……..

Time, s 300 250 200 150 100 50 0

[Adr

enoc

hrom

e], m

kM 160

140 120 100 80 60 40 20 0

1.08 mkM

0.81 mkM

0.54 mkM

0.27 mkM

0.16 mkM

Parameters of the COX catalytic cycle identified on the base of experimental data available from literature

120 110 100 90 80 70 60 50 40 30 20 10 0

AA

cons

umpt

ion.

, mkM

2,2 2

1,8 1,6 1,4 1,2

1 0,8 0,6 0,4 0,2

0

Time, s

[AA]=20 mkM

2 mkM

1 mkM

0.5 mkM

[COX-1]=1.61 mkM

Model Validation. Identification of kinetic parameters.






“Vioxx” project


Intracellular


Organ level

Figure A: Extended Cox-1/2 model. Bottom panel – default Cox-1/2 model; Top panel – inhibition of Cox-1/2 by Aspirin and/or second Inhibitor.

Figure B: Cox cartoon of the simplified Cox1/2 inhibition mechanism.

Extended Cox-1/2 Model for Drug Action Modeling

AA

AA

Effects of inhibitors (NSAIDs) introduced to the COX model:

• Aspirin + - 1,2

• Indomethacin + + 1,2

• Naproxen 1-,2+ + 1,2

• Diclofenac + + 1

• Ibuprofen - + 1,2

• Celecoxib 1-,2+ + 2

• Vioxx 1-,2+ + 2

Time dependence

Reversibility of binding

Selectivity to COX1,2

1- COX1; 2 - COX2

0 20 40 60 80

100

0 200 400 600 800 1000 1200 Ibuprofen concentration, M

Ibuprofen

1,2 Preincubation time, sec

Aspirin

Consistent description of experimental data on inhibitory effects of different NSAIDs

0.8 mM

2.35 mM

4.36 mM 0

0,2 0,4 0,6 0,8

1 1,2

0 1000 2000 3000 4000 5000

Rel

ativ

e C

OX

activ

ity

Rel

CO

X-2

activ

ity

Preincubation time, sec Preincubation time, sec

Celecoxib

0.5 M

1 M 2 M

0

0,2

0,4

0,6

0,8

1

0 10 20 30 40 50 60 70

Indomethacin

2.2 M 3.8 M

5.4 M

0

0,2

0,4

0,6

0,8

1

0 500 1000 1500 2000

1.4 M

Rel

ativ

e C

OX

activ

ity

Rel

ativ

e C

OX

activ

ity

NSAID binding parameters used to cross-validate and predict

inhibitor effects on COX

- No curve fitting was employed -

• Selective COX2 inhibitor can block aspirin effect – experimental phenomena observed, not previously explained

50% inhibition plane

- For both single drug and drug combinations:

- Aspirin, Ibuprofen, Naproxen, Celecoxib, Indomethacin, Diclofenac,…

Prediction of NSAID action on target

High Dose Aspirin GI side effects

High Dose Coxib CV side effects

Coxib [M]

ASA [M]

Safer ASA/Coxib Combinations?

• Model based analysis allows for prediction of safer drug combinations

COX in the context of intracellular micro- conditions

COX degradation

Description includes:

•  Detailed catalytic cycle of COX

•  COX self-inactivation

•  COX synthesis and degradation

•  In-fluxes and out-fluxes of substrates and products

Allows for properties of COX and NSAID inhibition to be translated into in vivo conditions

COX catalysis and NSAID effects in real time and within real physiological substrate / product concentration range

Intracellular microenvironment controls COX activity and dictates sensitivity to NSAIDs model allowed to explain/predict many experimental phenomena:

•  Discrepancies between in vitro / in vivo estimates of IC50 for Aspirin

•  Origin of variability of in vivo experimental values of Aspirin IC50 – intracellular micro- environmental concentrations of substrates

•  Variability in COX-1/COX-2 selectivity – results from intracellular conditions

•  Attenuation of ASA effect by Celecoxib

COX activity is controlled by intracellular micro-environmental conditions: Arachidonic Acid and Reducing Co-substrate concentration

AA, M RC, M

Velo

city

of P

GH

2 pr

oduc

tion

Functionally active

Non-functional

In vivo COX-1 Whole blood assay 1.3 [1], Platelet 1.3 [2] Endothelial cells 1.5 [3] Fibroblasts 2.6 [4]

In vitro COX-1 Purified enzyme: 30-200 [5]

“in vivo” model

IC50=2 M

IC50=300 M

in vitro model

ASA, uM

First valid explanation of discrepancies between

in vitro / in vivo estimates of IC50 for Aspirin

Experimental estimates of IC50 (M) for Aspirin:

PGH

2 pr

oduc

tion,

rel.u

nits

1. Warner T., Giuliano F., et al PNAS 96, 1999, 7563 2. Quellet M., Riendeau D., Percival M. PNAS 98, 2001, 14583 3. Mitchell J. , Akarasereenont P., et al PNAS 90, 1994, 11693 4. Chulada P., Langenbach R. J. Pharm. Exp. Ther. 280, 1997,606 5. Kargman S., Wong E. et al Bioch. Pharm. 52, 1996, 1113

Accumulation of acetylated COX gives rise to additional COX inhibition due to retained peroxidase activity

Modelling drug action on target under various regimes of drug administration

Low dose Aspirin, once daily

ASA ASA ASA ASA

A higher dose Aspirin reduced with time






“Vioxx” project


Intracellular


Organ level

Detailed scheme

Inactive AC Active AC

ATP

Inactive PKA Active PKA

CREB1

CREB1-phosphate

Protein Phosphatase

Gi Gs

cAMP PPi

AMP

Cyclic-nucleotide Phosphodiesterase

ATP

ADP

Pi

TRANSCRIPTION

Abbrevations: Gs = stimulating Gα-protein Gi = inhibiting Gα-protein AC = Adenylyl Cyclase PKA = cAMP-dependent Protein Kinase CREB1 = cyclic AMP response element

binding protein-1 RED – constant metabolites (parameters) BLUE – variable metabolites

Example: Adenylate Cyclase and Protein Kinase A Signalling

Input Signal

Output Signal

1. AC activation

2. cAMP synthesis

3. PKA activation

4. Protein phosphorylation

AC AC-Gi

AC-Gs AC-Gi-Gs

Gs Gs

Gi

Gi

K_AC_Gi

K_AC_Gs_Gi

K_AC_Gi_Gs

K_AC_Gs

ATP cAMP PPi

ATP cAMP PPi

Identification of constants for AC activation model

* Chen-Goodspeed M, Lukan AN, Dessauer CW; Modeling of Galpha(s) and Galpha(i) regulation of human type V and VI adenylyl cyclase; J Biol Chem. 2005; 280(3):1808-16

Gi concentration (µM) 3 2 1 0 In

itial

rate

of c

AM

P sy

nthe

sis

(pm

ol/m

in/m

g)

2 1.8 1.6 1.4 1.2

1 0.8 0.6 0.4 0.2

0

Gs = 3.0 µM

Gs = 1.0 µM

Gs = 0.25 µM

Gs = 0.1 µM Gs = 0.05 µM Gs = 0.02 µM Gs = 0.01 µM






“Vioxx” project


Intracellular


Organ level

PGD2 (ext)

TXA2 (ext)

AA

PLA2

COX-1

PGH2

TBXAS

HHT TXA2

TXB2

inactivation

Ph.Lip.-AA

PGH2 (ext)

TXB2 (ext)

AA (ext)

cAMP

ATP

PKA

IP3R

Ca2+

AC

Ca2+ ER

PLC

IP3

PIP2

AMP

Ca-ATPase

degradation PGE2

(ext)

R1

Gq

R2

Gs

thrombin,

ADP,

TXA2 (ext)

PGI2 (ext) ,

iloprost (IP);

PGD2(ext) (DP)

PKC

degradation

DAG

Platelet model •  TXA2 biosynthesis •  transmembrane transport •  signalling pathways activated by prostanoids •  Ca2+ fluxes involved in platelet activation •  NSAIDs action on cyclooxygenase

ODE system: 42 equations, 62 rate laws, 152 parameters

Time, min

uM IP3

cAMP

Ca2+

COX-1

thrombin iloprost

Part of “active” IP3R

High dose of iloprost effectively inhibits platelets activation by thrombin

Stimulation of platelets by iloprost leads to cAMP-dependent PKA activation that phosphorilates IP3-dependent calcium channels (IP3R) on endoplasmic reticulum (ER). This results in in the inhibition of IP3R sensitivity to IP3, and a decrease in Ca2+ outflux from ER

Data from JBC(2002)277:29321

Platelet model validated on perturbation experimental data with receptor agonists






“Vioxx” project


Intracellular


Organ level

AA

PLA2

COX-1,-2

PGH2

TXAS

HHT TXA2

TXB2

inactivation…

Ph.Lip.-AA

PGH2 (ext)

TXB2 (ext)

AA (ext)

cAMP

ATP

PKA

IP3R

Ca2+

AC

Ca2+ ER

PLC

IP3

PIP2

AMP

Ca-ATPase

degradation

PGE2 (ext)

R1

Gq

R2

Gs

thrombin, TXA2 (ext)

TXA2 (ext)

PKC

degradation

DAG

PGE2

PGES

PGI2 PGI2 (ext)

PGIM (ext)

PGIS

PGI2 (ext) ,

iloprost (IP);

PGD2(ext) (DP)

Endothelium cell model •  prostanoid biosynthesis (PGI2, PGE2, TxA2) •  transmembrane transport •  signalling pathways activated by prostanoids •  Ca2+ fluxes involved in EC activation •  NSAIDs action on cyclooxygenase

ODE system: 47 equations, 69 rate laws, 184 parameters






“Vioxx” project


Intracellular


Organ level

Endothelial cell

Arachidonic acid

(exogenous)

PGH2

,PGD2, PGE2, PGF2a

PGI2 (prostacyclin)

BLOOD PLASMA Platelet

Arachidonic acid

Phospholipids

AA transporter

PGH2

PGI2 (prostacyclin)

PGH2 transporter

AA transporter

Phospholipase A2

PGD2, PGE2, PGF2a

Transporter

COX-1 COX-2

Prostacyclin synthase

Nonenzymatic

Arachidonic acid

Phospholipids

AA transporter

PGH2

TXA2 Thromboxane A2

PGH2 transporter

TXA2 transporter

Phospholipase A2

PGD2, PGE2, PGF2a

Transporter

COX-1

Thromboxane synthase

Nonenzymatic TXA2

Thromboxane A2 Nonenzymatic

Platelet-Endothelium-Plasma model allows to assess potential cardiovascular risk after NSAID administration


Inflammatory EC

Non-inflammatory EC

Risk of clot formation increases with increase in COX2 selective inhibitor concentration under inflammatory conditions

Dose-response of the platelet-endotelium plasma system. Prediction of clotting risks for NSAIDS and combinations


Heart muscle

Head (Brain)

AORTA

LUNG Upper body (arms)

Kidneys

GIT

Liver VENA CAVA

Lower body (Legs)

Kinetic model of endothelium cell

Endothelium cells

Platelets Kinetic model of platelet

•  Several phases (immobile, mobile,…) •  Several cell types •  Interaction between cells via secreted

metabolites •  Models of intracellular processes in each cell

type •  Anatomical/geometrical features of the system •  Changes in properties of the cells located at

different parts of the system

8 organs have been taken into account in our models

Scaling up to organism level: Blood circulation

capi

llarie

s &

ven

ules

veins

lung

vena

cav

a

aort

a

vena

cav

a

lower body (legs)

capi

llarie

s

arte

ries

&

arte

riole

s

Activation (TXA2)

veins

capi

llarie

s &

ven

ules

upper body (arms)


Monitoring state of platelets in different parts of blood circulation

1)   Clotting risks monitored (proportional to intracellular Ca2+ concentration in platelets) 2)   Maximal risk of clot formation is observed in leg veins 3)   Capillaries of lungs and arms decrease risk of clot formation substantially

•  In silico methods help to understand the efficacy, mode of action and potential dangerous side effects of drugs •  Model-based analysis and prediction of drug safety and efficacy •  Informed advice on drug dosing, scheduling and combinations

Systems Approach applied to drug safety

Acknowledgement

Dr. Oleg Demin

Dr. Yury Kosinsky and team

Dr. Jörg Lippert

Prof. Tim Warner

Dr. Alexey Goltsov

Dr. Maciej Swat

Prof. Igor Goryanin University of Edinburgh

systems biology for drug discovery · key to future medical breakthroughs is systems biology, say...

Documents