steven r mcdougall institute of petroleum engineering heriot-watt university edinburgh scotland...

102
Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced Angiogenesis Tumour Growth and Angiogenesis, 11-19 January 2006, Hsinchu, Taiwan

Post on 21-Dec-2015

216 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

Steven R McDougallInstitute of Petroleum Engineering

Heriot-Watt University

Edinburgh

Scotland

Mathematical Modelling of Dynamic Adaptive

Tumour-Induced Angiogenesis

Tumour Growth and Angiogenesis, 11-19 January 2006, Hsinchu, Taiwan

Page 2: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 2

Co-Investigators

• M A J Chaplain Dept. of MathematicsUniversity of Dundee

• A R A Anderson Dept of Mathematics, University of Dundee

• A Stéphanou Institute of Petroleum Engineering,

Heriot-Watt University

Page 3: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 3

Outline• Background

• Modelling Tumour-Induced Vasculature

– Continuum scale formulation of growth

– Discrete (capillary) scale growth

– Improved continuous-discrete coupling through MMP

• Incorporating Flow

– Flow in porous media

– Blood flow and drug delivery

• 3D Dynamic Adaptive Tumour-Induced Angiogenesis

– Vascular remodelling and evolving bed architecture

• Conclusions and future work

Page 4: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 4

Petroleum Engineering and Medicine??

• Multiphase flow at the micro-scale has been investigated over the past 15 years at Heriot-Watt University

• Software resulting from this research is widely used by the oil industry to interpret laboratory experiments and reservoir behaviour

• However, the underlying modelling framework can also be adapted to examine a wide range of problems in the clinical arena

• We will focus here upon a particular application in clinical oncology – tumour-induced angiogenesis

Page 5: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 5

Modelling Rationale

• Use mathematical modelling techniques to give some insight into the underlying physical and biological processes governing tumour-induced angiogenesis

• Reproduce experimental observations and suggest additional laboratory studies

• Use insights gained from modelling to suggest new treatment strategies and scheduling protocols

Page 6: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

Modelling Tumour-Induced Vasculature

Page 7: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 7

3 Phases of Solid Tumour Growth

• Early avascular phase– tumour spheroid 106 cells– max diameter ~ 2mm– necrotic core– thin proliferating rim

• Angiogenesis and vascularisation– capillary network formation– blood supply to tumour– additional growth

• Invasion and metastasis– attachment to BM– collagenase production– addressins at distant sites

Page 8: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 8

• Hypoxic tumour cells secrete tumour angiogenic factors (TAFs)

• Endothelial cells degrade their basement membrane

• These migrate towards the tumour

• Capillary branching increases towards the tumour

• Tumour penetrated and nutrient supply begins

Tumour-Induced Angiogenesis

Page 9: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 9

Vascular Growth SchematicParent Vessel

Tumour Surface

TAF

FN

TISSUE

Page 10: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 10

Vascular Growth - Continuum Approach

f(x,f(x, y,y, t)t)tt

==

n(x,n(x, y,y,t)t)

tt == DD n n 22 (n(nf )f )

n fn f

c(x,c(x, y,y,t)t)

tt ==

ncnc

(n(nc )c )

nn

c (x,y,t) ~ c (x,y,t) ~ TAF ConcentrationTAF Concentration

f (x,y,t) ~ f (x,y,t) ~ Matrix Macromolecule (fibronectin)Matrix Macromolecule (fibronectin)

n (x,y,t) ~ n (x,y,t) ~ Endothelial Cell DensityEndothelial Cell Density

Page 11: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 11

EC Density Profile

Page 12: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 12

• Use discretised form of continuum equations to migrate capillary sprouts and grow capillaries

• Capillary branching, anastomosis and cell mitosis all included in the model

• 2 tumour geometries considered

– linear TAF source

– circular TAF source

Discrete Model

Page 13: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 13

Discrete Model

l,l, mmqq++11

l,l, mmqq

00 ll++ 1,1, mmqq

11 ll-1,-1, mmqq

22 l,l, mm + + 11qq

33 l,l, mm -1-1qq

44

l,l, mmqq++11

l,l, mmqq

nn == nn PP nn PP nn PP nn PP nn PP

cc == cc

ff == ff l,l, mmqq++11

l,l, mmqq kk nn

l,l, mmqq

1 1 kk l,l, mmqqnn

1 1

withwith x=lhx=lh, , y=mhy=mh andand t=qkt=qk

00 PP

33 PP

22 PP

44 PP

11 PP

kk nnl,l, mmqq

Page 14: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 14

Movement Weighting

11 22 22

22 22 22

33 22 22

44 22 22

PP ==k Dk Dhh

kk

44hh

PP == k Dk Dhh

++kk

44hh

PP == k Dk Dhh

kk

44hh

PP == k Dk Dhh

++ kk

44hh

) ) (( ll++1,1,mmqq

ll-1,-1,mmqqcc cc

l,l,mm++11qqff ff l,l,mm-1-1

qq(( ))

) ) (( ll++1,1,mmqq

ll-1,-1,mmqqff ff++

) ) (( ll++1,1,mmqq

ll-1,-1,mmqqcc cc ) ) (( ll++1,1,mm

qqll-1,-1,mmqqff ff++

l,l,mm++11qqcc cc l,l,mm-1-1

qq(( )) ++

l,l,mm++11qqff ff l,l,mm-1-1

qq(( )) l,l,mm++11qqcc cc l,l,mm-1-1

qq(( )) ++

00PP == 1 1 44 k Dk Dhh

22

ll++1,1,mmqq

ll-1,-1,mmqq

l,l,mmqq

l,l,mm++11qq

l,l,mm-1-1qq kk

hh(( cc ++ cc -- 44cc ++ cc ++ cc )) 22

ll++1,1,mmqq

ll-1,-1,mmqq

l,l,mmqq

l,l,mm++11qq

l,l,mm-1-1qq kk

hh(( ff ++ ff -- 44ff ++ ff ++ ff ))22

Page 15: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 15

Numerical Simulation

• At each time step

– Solve fibronectin and TAF equations

– Generate 5 directional coefficients (7 in 3D) P0 to P4

– Compute probability ranges R0 to R4

– Generate random number in (0,1)

– Determine appropriate growth direction

Page 16: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 16

Branching and Anastamosis

Sprout SproutSprout

Loop formed by anastomosis

Branching at sprout tipto form two new sprouts

Parent Vessel

Sprout tip

Page 17: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 17

Sample Results

Linear Source Circular Source

Page 18: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 18

• Matrix metalloproteinases explicitly included

• MMP produced locally by individual ECs

• MMP also diffuses and decays

• n equation now only used to extract directional coefficients

,)(2 haptotaxischemotaxisrandom

fncncnDt

n

.

,

,

2 mmnt

m

mfnt

f

cnt

c

i

i

i

Improved Continuum-Discrete Coupling

Page 19: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 19

Simulation Results – Large Tumour

Tumour Surface Tumour Surface

Parent Vessel Parent Vessel

Capillary Network Enzyme Concentrations

Page 20: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 20

Tumour Surface Tumour Surface

Parent Vessel Parent Vessel

Capillary Network Enzyme Concentrations

Simulation Results – Small Tumour

Page 21: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 21

Comparison with Experiment

Experiment Simulation

Page 22: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 22

Experiment Simulation

Comparison with Experiment

Page 23: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

Incorporating Flow

Page 24: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 24

a

N S

ShaleAmalgamation surfaceOutcrop termination

100 m

Flow Through Porous Media

Page 25: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 25

Flow Through Porous Media

(q)i . t = (VbS)i

Qo

Qw

Qo

Qw

Vb, ,So, Sw

Page 26: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 26

Network Model

Page 27: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 27

Observing the PhysicsPressure depletion in a reservoir

Page 28: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 28

Simulation vs Experiment

Oil displacing water from a water-wet micromodel

Page 29: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 29

We Have the Fluid Distributions - What About

Flow?Pin Pout

qi=0

qi

Large set of pressure equations

Page 30: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 30

Viscous-Dominated Gasflood

Isolate each phase and flow

its networkGAS OILOIL

Page 31: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 31

Apply to Vasculatures

Vasculature 1 - Linear TAF source Vasculature 2 - Circular TAF source

Page 32: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 32

Chemotherapy Modelling

• Single-phase tracer algorithm developed

• “Chemotherapy drug” at concentration Cmax is injected into the upstream end of the parent vessel

• At each timestep

– total mass of drug flowing into each node calculated

– perfect mixing assumed at nodes

– drug uptake instantaneous within 40m of tumour surface

Page 33: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 33

Initial Assumptions

• Rigid, impermeable cylinders used for capillaries

• Blood treated as Newtonian fluid (constant viscosity)

– no correlation with haematocrit

• No reaction kinetics included in uptake function

• Vascular network is 2D

Page 34: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 34

Input Data• Four suites of simulations performed

– 2 vasculatures x continuous infusion

– 2 vasculatures x 30s bolus injection

• 5 sets of input data for each suite (0<t<2500s)

Base Run Run 1 Run 2 Run 3 Run 4

blood (Pa.s) 4 x 10-3 1 x 10-3 8 x 10-3 4 x 10-3 4 x 10-3

P (Pa) 800 800 800 800 800Rcap (m) (4, 4.01) (4, 4.01) (4, 4.01) (2, 2.01) (3, 3.01)Rart (m) 10 10 10 10 10

Page 35: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 35

Continuous Infusion into Vasculature 1

Vasculature 1

Page 36: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 36

Effect of Blood ViscosityMass in Vasculature

0

2

4

6

8

10

12

0 5 10 15 20 25 30

t*

M*

M* 1cP <r>=4

M* 4cP <r>=4

M* 8cP <r>=4

Uptake in Tumour

0

2

4

6

8

10

12

14

16

18

0 5 10 15 20 25 30

t*

Upt

ake

in T

umou

r

MT* 1cP <r>=4

MT* 4cP <r>=4

MT* 8cP <r>=4

Page 37: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 37

Effect of Mean Capillary Radius

Mass in Vasculature

0

5

10

15

20

25

30

0 5 10 15 20 25 30

t*

M*

M* 4cP <r>=2

M* 4cP <r>=4

M* 4cP <r>=6

Uptake in Tumour

0

2

4

6

8

10

12

14

16

18

20

0 5 10 15 20 25 30

t*

Upt

ake

in T

umou

r

MT* 4cP <r>=2

MT* 4cP <r>=4

MT* 4cP <r>=6

Page 38: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 38

Continuous Infusion into Vasculature 2

Vasculature 2

Page 39: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 39

Drug Delivery Comparison

Tot Mass in Vasc Circ

0

2

4

6

8

10

12

0 5 10 15 20 25 30

t*

M* M* 4cP <r>=4

Mcirc* 4cP <r>=4

Mass in Tum Circ

0

0.5

1

1.5

2

2.5

3

3.5

0 5 10 15 20 25 30

t*

MT

* MTcirc* 4cP <r>=4

MT* 4cP <r>=4

Page 40: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 40

Bolus Injection into Vasculature 2

Vasculature 2

Page 41: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 41

Drug Delivery Comparison

M* Linear vs Circular Bolus

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0 5 10 15 20 25 30

t*

M* M* 4cP <r>=4 bolus

M* 4cP <r>=4 circ bolus

Uptake in Tum. Lnear vs Circular Bolus

0

0.01

0.02

0.03

0.04

0.05

0.06

0 5 10 15 20 25 30

t*

MT

* MT* 4cP <r>=4 bolus

MT* 4cP <r>=4 circ bolus

Page 42: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 42

Early Findings

• Bolus injection into a capillary network generated from a circular TAF source

– drug bypasses the tumour completely

• Important implications for chemotherapy strategies

– structure of associated vasculature should be considered when planning chemotherapy treatments

Page 43: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 43

Targeting Strategies• 3 a posteriori pruning

algorithms considered

– random

– low-flow

– bottlenecks

• Motivation comes from possible targeting of different areas of vasculature with different cytotoxic compounds

Vessel maturation

Page 44: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 44

Targeting “High Shear” VesselsDrug targeting high-shear vessels

Page 45: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 45

However …

Page 46: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 46

Limitations of Early Model

• Two-dimensional

– is bypassing as significant in 3D?

• Blood treated as a Newtonian fluid

– blood is non-Newtonian and biphasic

– blood viscosity is not constant

• Capillaries treated as rigid and impermeable

– in reality, vasodilation/vasoconstriction occurs

– capillary architecture (i.e. bed topology) is not static

Page 47: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 47

Additional Limitations• Flow only incorporated after growth has

ended

– in fact, perfusion occurs during growth

• Growth and flow towards a tumour

– what about growth and flow within a tumour

– tumour cords around blood vessels

• Other limitations?

– coupled low-pressure venous system absent

– …….??

Page 48: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 48

Now begin to relax some of the earlier

assumptions…

Page 49: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 49

Relaxing Earlier Assumptions

3D Extension

Page 50: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 50

)1(

d

dPZ th

• Bypassing in 3D not necessarily less than that seen in 2D

• Delivery initially slower in 3D

• Anastomosis density, dimensionality, and duration of infusion all affect uptake

• 97-99% of drug injected bypasses the tumour

• Percolation theory can help here

• Really need Z(y) distribution to understand impact of vascular architecture

Relaxing Earlier Assumptions

Page 51: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 51

• Blood is non-Newtonian

• Capillaries are dynamic entities

• Dynamic feedback must be addressed

– Blood rheology

– Capillary radius = f1 (shear stress)

– Branching probability = f2 (shear stress)

Relaxing Earlier Assumptions

Page 52: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 52

(i) Blood is non-Newtonian

Red Blood Cells

Endothelial Cells

22

45.0 1.12

2

1.12

2)()1(1

R

R

R

RHf Drel Blood viscosity depends upon

haematocrit and capillary radius…

1211121115.0

)2(06.017.045.0

)2(101

1

)2(101

11)8.0(

1)45.01(

1)1()(

,44.22.360645.0

RReC

HHf

ee

R

C

CD

D

RR

…parameterised as follows

Relaxing Earlier Assumptions

Page 53: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 53

),( Dapp HRLHR

PRQ

Dapp ),(8

4

Relaxing Earlier Assumptions

(i) Blood is non-Newtonian

Page 54: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 54

sD

refmerefwttt

tt

kQH

QkPtRRR

RRR

log)(loglog1

1

Wall shear stress

Transmural pressure

Metabolic stimulus

Shrinking tendency

(ii) Shear stress affects vessel radii

Relaxing Earlier Assumptions

Page 55: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 55

Haematocrit distribution Radius distribution

R(m)% red cells

HD, R, app are interrelated(Haematocrit/Radius -> Viscosity -> Shear stress -> Radius -> Flow -> Haematocrit. . .)

Relaxing Earlier Assumptions

(ii) Shear stress affects vessel radii

Page 56: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 56

Sprout Branching

- TAF Concentration

- age of the vessel > thr

Vessel Branching

- magnitude of the WSS

- TAF concentration

- min < age of the vessel < max

(iii) Shear stress affects branching

Relaxing Earlier Assumptions

Page 57: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 57

Age of the vessels in days

(iii) Shear stress affects branching

Relaxing Earlier Assumptions

Page 58: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 58

WSS / WSS in Parent Vessel

connection connection

(iii) Shear stress affects branching

Relaxing Earlier Assumptions

Page 59: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 59

Coupling Growth and Flow

Angiogenesis

(cell migration)

Flow Modelling and

Blood Rheology

Dynamic Adaptive Tumour-Induced

Angiogenesis

Page 60: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 60

Dynamic Adaptive Tumour-Induced

Angiogenesis(DATIA)

Page 61: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 61

DATIA• A mathematical model which

simultaneously couples vessel growth with blood flow through the vessels – dynamic adaptive tumour-induced angiogenesis

• Radial adaptations and network remodelling occurs as immediate consequences of primary anastomoses

• Capillary network architectures from the dynamically adaptive model differ radically from those obtained using earlier models.

Page 62: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 62

Modelling Rationale

• Examine the effects of changing various physical and biological model parameters on the developing vascular architecture

• Simulate chemotherapeutic treatments under different parameter regimes

– identify a number of new therapeutic targets for tumour management.

Page 63: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 63

Simulation Procedure• Capillaries migrate via the discrete form of the

extended PDE formulation (MMP)

– only tip branching possible initially

– no remodelling without flow

• Vessel branching and remodelling considered only after first anastomoses form

• However

– timescale of EC migration ~ days

– timescale of network perfusion ~ minutes

– so we cannot simply remodel as the migration simulation proceeds

Page 64: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 64

• Idealised procedure

– model the growth of the capillary network using the endothelial cell migration model;

– pause the migration model whenever a new anastomosis (loop) forms;

– switch timescales (t = MIN(Vcap/Qcap))

– flow/remodel the entire capillary bed using flow model until a new steady-state has been reached (~100s of perfusion);

– resume network growth using the cell migration model on the longer timescale.

• Practical procedure

– flow the network to steady-state at regular intervals during the growth process (determine optimum interval).

Simulation Procedure

Page 65: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 65

A Posteriori Remodelling

t=0.8 t=2.4 t=3.0

t=4.5 t=72.0 t=300.0

Page 66: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 66

Full DATIA Simulation

4.0 8.0 8.0+1 12.0

12.0+2 12.0+2+3 30.0

Page 67: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 67

Increased Sensitivity to max

7.5 8.0 8.5

11.5 12.0

Page 68: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 68

Additional Sensitivities

P(vb)/2.5 P(vb)/5.0 =0.18 pl*=4.0 HD=0.675

HD=0.225 Pin Pout Pin Pout kp

Page 69: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 69

Transport Through Adapted Networks

• Use tracer algorithm to quantify the efficiency of different networks in carrying blood-borne material e.g. nutrients, chemotherapy drugs, to the tumour

• Almost all of the drug flows through the dilated backbone

• Poor treatment efficiency

• The architecture of the backbone determines delivery to tumour

t=1.0 t=5.0 t=10.0

t=20.0 t=50.0 t=300.0

Page 70: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 70

• All masses normalised to the total mass injected into the parent vessel over the course of the simulation.

• Only around 1.5% of the infused tracer-drug even enters the capillary network

• Total mass in the network reaches a plateau after approximately 50s

• It takes another 200-250s before uptake commences.

Transport Through Adapted Networks

0

0.005

0.01

0.015

0 100 200 300 400 500

Time (s)

Mas

s in

Bed

(n

orm

alis

ed)

0.00E+00

2.00E-07

4.00E-07

6.00E-07

0 100 200 300 400 500

Time (s)

Up

take

(n

orm

alis

ed)

Page 71: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 71

Comparison with Homogeneous Bed

• Uptake values are approximately three orders of magnitude higher that those obtained from the remodelled vasculature

• Highlights the need for incorporating vessel adaptations (dilation/constriction) into any angiogenesis model involving transport issues

0.00E+00

2.00E-04

4.00E-046.00E-04

8.00E-04

1.00E-03

0 100 200 300 400 500

Time (s)

Up

take

(n

orm

alis

ed)

0.00E+00

2.00E-07

4.00E-07

6.00E-07

0 100 200 300 400 500

Time (s)

Up

take

(n

orm

alis

ed)

Page 72: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 72

Effect of Reducing Haptotaxis• Total mass of tracer-drug

entering the supplying vasculature is almost identical to that observed in the base case simulation (=0.28)

• But drug uptake by the tumour is fifty times greater when lateral migration and vessel branching are reduced

• This suggests that tumours supplied by this type of vasculature would be well-supplied with nutrients and could be expected to grow rapidly

• Paradoxically, such tumours would also be highly susceptible to infused treatments

0

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0 100 200 300 400 500

Time (s)

Mas

s in

Bed

(n

orm

alis

ed)

Rho=0.28

Rho=0.16

0.00E+00

5.00E-06

1.00E-05

1.50E-05

2.00E-05

2.50E-05

3.00E-05

0 100 200 300 400 500

Time (s)

Up

take

(n

orm

alis

ed)

Rho=0.28

Rho=0.16

Page 73: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 73

Effect of Reducing Haematocrit

• A depressed haematocrit was found to lead to the formation of highly dilated arcades close to the parent vessel

• Could be a possible mechanism for generating vasculatures that are detrimental to tumour growth

• The therapeutic implications of this

– more drug enters the capillary network than entered in the base-case simulation

– but drug delivery to the tumour is reduced by more than three orders of magnitude.

0

2E-11

4E-11

6E-11

8E-11

1E-10

0 100 200 300 400 500

Time (s)

Up

take

(n

orm

alis

ed)

0.00E+00

2.00E-07

4.00E-07

6.00E-07

0 100 200 300 400 500

Time (s)

Up

take

(n

orm

alis

ed)

Page 74: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 74

Other Pieces of the Puzzle

Page 75: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 75

Flow Within the Tumour• Tumour

vasculature poorly connected

• Tumour capillary radii larger on average

• Capillaries “leaky”

• Examine flow to and within tumour

• Treatment scheduling

Page 76: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 76

Flow Within the Tumour

Optimise scheduling to build-up high local drug concentrations

Page 77: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 77

Examine heterogeneous distributions of nutrient

supply and drug delivery

Transmural Diffusion

Feed oxygen tensions into model => evolving TAF sources

Page 78: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 78

Conclusions and Future Work

Page 79: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 79

Conclusions• An extensive theoretical investigation of the process of

tumour-induced angiogenesis has been presented incorporating:

– blood rheological properties

– metabolic constraints

– vessel branching = f(wall shear stress )

• Results from computational simulations have highlighted a number of possible new targets for therapeutic intervention

– manipulating sensitivity to wall shear stress

– haptotactic response of the endothelial cells

– haematocrit

– intravascular pressure

Page 80: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 80

Conclusions• Explicit coupling of growth and flow leads to

network architectures that differ radically from those found in all previous models.

• Dilated loops (anastomoses) form at an earlier stage close to the parent vessel

– positively reinforcing,

– proximally-dilated capillaries undergo further vessel branching.

– subsequent migration of these additional branches result in high capillary densities in regions distal to the parent vessel

– the number of high-conductivity pathways is consequently greatly reduced close to the tumour surface

Page 81: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 81

Conclusions• It is apparent from the transport simulations that

highly-dilated loops proximal to the parent vessel remove any possibility of effective treatment via intravenous or intra-arterial infusion.

• However, if a tumour-induced capillary network could be forced to develop in just such a way, by means of some clinical intervention perhaps, then nutrient supply to the tumour could be effectively curtailed.

• The DATIA model provides a useful biomechanical framework within which to examine the possibility of managing high conductance pathways as a means of effectively treating solid tumours

Page 82: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 82

Future Work• Couple flow model to continuum and CA

tumour models

• Use digitised images of real 2D and 3D vasculatures

• Migration of tumour fragments -> metastases

• Heterogeneous tissues

• Other applications (lymphangiogenesis, therapeutic angiogenesis, wound healing, retinopathy…)

Page 83: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 83

Page 84: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 84

Discussion Topics

• Tumour-vessel coupling

– increased tumour pressure

– leaky vessels

• Venous system

– low pressure

• Lymphangiogenesis

– how does this differ from angiogenesis?

Page 85: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 85

Tumour/Vessel Coupling• Potts model for

tumour integrity?

• Appropriate for breast cancer modelling

– used to examine interface region between disrupted tumour surface and healthy perimeter tissue

Page 86: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 86

Wound Healing

• Similar techniques are being used to study healing rates in dermal wounds.

• Future work will focus upon the design of dressings that could accelerate the healing process and reduce scarring

Page 87: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 87

Therapeutic Applications

Page 88: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 88

Random Targeting

What if we end treatment here?

Broad-based indiscriminate drug

Page 89: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 89

Targeting Poorly Perfused Vessels

Anti-angiogenic drug targeting immature, poorly perfused vessels

Page 90: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 90

Outcomes• Random targeting

– Increased flow in distal regions

– Decrease in Z4 and increase in Z3 and Z2

– 130% increase in delivery

– Delivery optimised if we can decrease Z4 proximal to parent vessel but maintain good connectivity close to tumour

• Low-flow targeting

– Flow distribution essentially unchanged

– Treatment accelerated in modified network

Page 91: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 91

• High shear targeting

– Bottleneck (percolation radius) capillaries removed

– Network shut down effectively after 5% vessels removed

– No delivery of drug but also no delivery of nutrients

• Delivery highly sensitive to network architecture

• Main flowing backbone plays a dual role

– Helps carry treatment to tumour

– Increases bypassing

• Treatments should be aimed at managing this backbone

Outcomes

Page 92: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

S R McDougall1 and J A Sherratt2

1 Dept of Petroleum Engineering, Heriot-Watt University

2 Dept of Mathematics, Heriot-Watt University

Discrete Modelling of Collagen Deposition and Alignment During Dermal

Wound Repair

Page 93: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 93

Dermal Wound

Page 94: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 94

Collagen Alignment

• Post wounding, fibroblasts respond chemotactically and migrate into wound area

• Collagen deposited

• Fibrin degraded

• Scarring due to collagen alignment

• How can we reduce this?

Page 95: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 95

Matrix Orientation Model

(After Dallon et al, 1999)

)

)s )

(t

(t(t

dt

t dii

v

vf ) ( f

)

)

(t

(tt t ti

ii

f

fv

) ), ( ( ) 1( ) (f c

) sin() , (

fdt

t dx

)

)

(t

(tt x w ti

i N

ii

f

f

1

) , ( ) , (x f

C

FB

f(x,y)

Page 96: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 96

Results

Speed=5m/hr Speed=15m/hr

Page 97: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 97

Tissue Regeneration Model

)

)s( )

(t

(t(t

dt

t dii

v

vf) ,

) (b c

f

)

)

(t

(t

t t

t tti

i

i

i

f

fv

) ), ( (

) ), ( () 1( ) (

f u

f u

) sin() , (

fdt

t dx

) , ( ) , ( ) 1( ) , (t t tx x xb c u

C

b

FB

f(x,y)c

b

)

)

(t

(tt x w ti

i N

ii

f

f

1

) , ( ) , (x f

Page 98: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 98

Tissue Regeneration Model

),()),((),(

1

twtdpdt

td N

iicc xxc

xc

),(),(),(

1

twtddt

td N

iib xxb

xb

C

b

FB

Page 99: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 99

Results

Page 100: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 100

Leukocyte Signalling

212

2

kaka

aD

t

a

aka

aD

t

a12

2

HEALTHY PERIMETER

212

2

kaka

aD

t

a

aka

aD

t

a12

2

Page 101: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 101

Results

Page 102: Steven R McDougall Institute of Petroleum Engineering Heriot-Watt University Edinburgh Scotland Mathematical Modelling of Dynamic Adaptive Tumour-Induced

www.pet.hw.ac.uk 102

Future Work

• Clinical applications (TGF isoform issue)

• Wound contraction

• Blood flow and angiogenesis