ARTICLE IN PRESS

Progress in Biophysics & Molecular Biology 85 (2004) 353368

Modelling cell population growth with applications to cancer

therapy in human tumour cell lines
Britta Bassea,*, Bruce C. Baguleyb, Elaine S. Marshallb,
Graeme C. Wakea, David J.N. Walla
a

Biomathematics Research Centre, University of Canterbury, Private Bag 4800, Christchurch, New Zealand
Auckland Cancer Society Research Centre, Faculty of Medical and Health Sciences, University of Auckland,
Private Bag 92019, Auckland, New Zealand

Abstract
In this paper we present an overview of the work undertaken to model a population of cells and the
effects of cancer therapy. We began with a theoretical one compartment size structured cell population
model and investigated its asymptotic steady size distributions (SSDs) (On a cell growth model for
plankton, MMB JIMA 21 (2004) 49). However these size distributions are not similar to the DNA (size)
distributions obtained experimentally via the ow cytometric analysis of human tumour cell lines (data
obtained from the Auckland Cancer Society Research Centre, New Zealand). In our one compartment
model, size was a generic term, but in order to obtain realistic steady size distributions we chose size to be
DNA content and devised a multi-compartment mathematical model for the cell division cycle where each
compartment corresponds to a distinct phase of the cell cycle (J. Math. Biol. 47 (2003) 295). We then
incorporated another compartment describing the possible induction of apoptosis (cell death) from mitosis
phase (Modelling cell death in human tumour cell lines exposed to anticancer drug paclitaxel, J. Math. Biol.
2004, in press). This enabled us to compare our model to ow cytometric data of a melanoma cell line where
the anticancer drug, paclitaxel, had been added. The model gives a dynamic picture of the effects of
paclitaxel on the cell cycle. We hope to use the model to describe the effects of other cancer therapies on a
number of different cell lines.
r 2004 Elsevier Ltd. All rights reserved.
Keywords: Mathematical model; Cell cycle; Steady size distribution; DNA prole; DNA histogram; Paclitaxel

*Corresponding author.
E-mail addresses: b.basse@math.canterbury.ac.nz (B. Basse), b.baguley@auckland.ac.nz (B.C. Baguley),
e.marshall@auckland.ac.nz (E.S. Marshall), g.wake@math.canterbury.ac.nz (G.C. Wake),
david.wall@canterbury.ac.nz (D.J.N. Wall).
0079-6107/$ - see front matter r 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.pbiomolbio.2004.01.017

ARTICLE IN PRESS
B. Basse et al. / Progress in Biophysics & Molecular Biology 85 (2004) 353368

354

1. A one-compartment theoretical mathematical model of the cell cycle

According to Arino (1995), the rst age structured population model is due to the 1911 classic
paper of Sharpe and Lotka (1911). However, it was not until the 1967 paper of Bell and Anderson
(1967) that structured population models were related to cells. Since then many researchers have
considered cell size structured dynamical models with particular emphasis on the steady-size
distributions they exhibit. The terminology steady-size distribution describes the phenomenon
whereby if one plots the number density of cells versus size over a period of time, eventually this
size distribution does not change its shape as time evolves (even though the total number of cells
may be changing). The key aim of most of the papers published since the late 1960s has been the
proof of the existence and uniqueness of such steady size distributions (see for example Basse et al.,
2004b; Diekmann, 1983; Diekmann et al., 1984; Grabosch, 1994; Gyllenberg and Webb, 1990;
Hall et al., 1991; Hannsgen and Tyson, 1985; Rossa, 1995). The emphasis of these papers has been
the interesting mathematics that arise in such models, with little or no mention of modern
biological applications. Our initial approach to modelling a cell cohort was also a theoretical one
based on these previous models. We considered a population structured by size, where size is a
generic term encompassing any one of age, volume, DNA content, width, etc. (Basse et al.,
2004b). The number of cells with size between x and x Dx at time t; is denoted nx; tDx; and the
corresponding equation for cell growth is
@
@2
@
nx; t 2 Dx; tnx; t  gx; tnx; t
@t
@x
@x
2
a Bax; tnax; t  Bx; t mx; tnx; t;

where 0oxoN; t > 0; D m2 =s is the diffusion coefcient, g (m/s) is the rate of growth and m (1/s)
is the rate of death. The function B (1/s) is the rate at which cells divide into a equally sized daughter
cells. Here a > 1 is regarded as a constant, and the functions D; g; m and B are all non-negative.
The derivation of Eq. (1), taking D 0 and g constant for simplicity, arises from considering
the limit as Dx and Dt tend to zero of the discrete equation:
Dt
Dt
nx; tDx g
nx  Dx; tDx
nx; t DtDx nx; tDx  g
Dx
Dx
aBax; tDtnax; tDax
 Bx; t mx; tDtnx; tDx:

More detailed derivations of similar models can be found in Diekmann (1983) and Diekmann
et al. (1984) and Tyson and Hannsgen (1985). The derivation of the diffusion term can be found in
Okubo (1980).
The partial differential equation (1) may be supplemented by the boundary conditions;
lim nx; t 0;

x-N

lim

x-N

@
nx; t 0;
@x

@
D0; tn0; t  g0; tn0; t 0:
@x

3
4
5

ARTICLE IN PRESS
B. Basse et al. / Progress in Biophysics & Molecular Biology 85 (2004) 353368

355

Conditions (3) and (4) place decay conditions on n as x-N for any xed time. Eq. (5) is a no
ux condition on the boundary x 0:
Of particular theoretical interest are solutions to the boundary-value problem (1)(5) that
correspond to steady-size distributions (SSDs) for the number density function. SSD solutions to
the boundary-value problem correspond to solutions of the form nx; t Ntyx (i.e., separable
solutions). The rst-order case D 0 with g and B constant, and m 0 was studied by Hall and
Wake (1989). In addition, Hall and Wake (1990) studied the rst-order problem for exponential
growth (i.e. gx; t gx; B bxk ; B bxk Hx  x1 ; where a; b are constants). The second-order
problem Da0 was studied by Wake et al. (2000) and Hall (1998) for constant Da0; g and b;
with m 0: Continuing this work we extended the second order problem to consider the cases
where bx bdx  l and bx bHx  l (Basse et al., 2004b) and we found SSDs in these
cases.

2. Biological application: a multi-compartment model of a human tumour cell line unperturbed by

cancer therapy
At the Auckland Cancer Society Research Centre (New Zealand), 60 cell lines from patients
with a variety of cancer types have been developed for a number of years. Each cell line can be
thought of as an individual cell population. A cell line can be analysed at intervals by ow
cytometry. First, the cellular DNA of each individual cell is labelled with a uorescent dye and
cells are then processed in a ow cytometer, which measures uorescence of single cells. The latter
is proportional to DNA content and thus a DNA distribution (sometimes referred to as a DNA
histogram or DNA prole) may be obtained for a cohort of cells. Once the cell population is
growing exponentially, the DNA prole does not change its shape and we use the terminology
steady DNA distribution (SDD). A typical DNA histogram for a human tumour cell line
unperturbed by cancer therapy is shown in Fig. 3a (solid line).
An unperturbed cell line with a steady DNA distribution is the starting point of an experiment.
The cell line can be perturbed by anticancer drugs or radiation. Typically, several hundred ow
cytometry proles are required to provide an adequate picture of the time courses of drug and
radiation effects for a series of cell lines.
Our rst aim was to model the population dynamics of a cell line unperturbed by cancer
therapy. One common element of the models discussed in Section 1 is that the SSDs are NOT
similar to those obtained experimentally in the human cell lines at the Auckland Cancer Society
Research Centre. With this in mind we determined that our size structure variable, x; should
represent relative DNA content (because DNA content is the experimental measure of cell size for
tumour cells) and that we should require four compartments corresponding to each distinct phase
of the cell cycle, G1 -phase (growth), S-phase (DNA replication), G2 -phase (further growth) and
M-phase (cell division) (Fig. 1). In particular DNA content only increases during S-phase. Our
model equations for a cell line unperturbed by cancer therapy are:
@G1
x; t 4bM2x; t  k1 G1 x; t;
@t

ARTICLE IN PRESS
B. Basse et al. / Progress in Biophysics & Molecular Biology 85 (2004) 353368

356

A Phase

Transition
k1
G1 Phase

S Phase
~10 hours

G2 Phase

Transition
k2
M Phase
~0.5 hours

2 cells

Fig. 1. Cell cycle control.

@S
@2 S
@S
% t; TS ;
x; t D 2  g x; t k1 G1 x; t  Sx;
@t
@x
@x

@G2
% t; TS  k2 G2 x; t;
x; t Sx;
@t

@M
9
x; t k2 G2 x; t  bMx; t;
@t
where 0oxoX and t > 0: Initial distributions at time t 0 in G1 ; S; G2 and M-phase are the
arbitrary functions: G1 x; 0 G10 ; Sx; 0 S0 ; G2 x; 0 G20 and Mx; 0 M0 ; respectively
0oxoX : The zero ux boundary conditions in S-phase are:
@S
10
D 0; t  gS0; t 0; t > 0;
@x
@S
X ; t  gSX ; t 0; t > 0;
11
@x
where X is the maximum DNA content. Model parameters are summarised in Table 1 and we
now give a description of model equations.
D

2.1. G1 -phase
In Eq. (6), describing G1 -phase, the rst term on the right-hand side is the source term provided
by the inux of newly divided daughter cells from M-phase. A cell in M-phase divides into two
identical daughter cells at a rate of b divisions per unit time. The factor of 4 in this term arises
from the fact that, at division, all cells with DNA content in the interval 2x; 2x 2Dx are
doubled in number and mapped to an interval with half the DNA content, namely x; x Dx:
Thorough derivations of similar non-local terms can be found in both Diekmann (1983) and

ARTICLE IN PRESS
B. Basse et al. / Progress in Biophysics & Molecular Biology 85 (2004) 353368

357

Table 1
Model parameters and variables. Mesh size: Dt 0:5; Dx 0:01; xmax 2:5
Parameter (Dim.)

Description

x
t (h)
G1 x; t
Sx; t
G2 x; t
Mx; t
Ax; t


k1 t1

2
D x
t


g T1S x
t

Relative DNA content

Time
Number density of cells
Number density of cells
Number density of cells
Number density of cells
Number density of cells

TS 1g (h)


k2 t1


b t1


mM t1


gA x
t

Division rate

TM (h)
Tc (h)

Value

LB

UB

Eventual death rate in M-phase

Average disintegration rate of DNA in A-phase

2=Dt

Time in M-phase before A-phase onset

Total cell cycle time

0
12

40
250

in
in
in
in
in

G1 -phase
S-phase
G2 -phase
M-phase
A-phase

Transition probability of cells from G1 to S-phase

Dispersion coefcient

0.0004

Average growth rate of DNA in S-phase

0.1

Time in S-phase

10

Transition probability of cells from G2 to M-phase

2

LB and UB are the lower and upper bounds, respectively, as applied during the optimisation routine of tting model
outputs to those obtained experimentally.

Tyson and Hannsgen (1985). The second term in Eq. (6) describes the loss of cells from G1 -phase
due to transition to the S-phase k1 G1 x; t: Therefore k1 provides the stochastic transition from
the G1 -phase to S-phase and represents the probability per unit time per unit cell that the cell will
enter the S-phase.
2.2. DNA synthesis (S-phase)
In Eq. (7), describing S-phase, g is the average rate at which the relative DNA content of a cell
increases with time. For a homogeneous tumour cell population, cells in G1 -phase and G2 and Mphase, will each have constant DNA contents. However, in ow cytometry the cells are not all
illuminated evenly as they pass through the ow cytometer and this causes an apparent variation
of DNA content. The dispersion term, D@2 S=@x2 ; can be used to account for this variability. The
inclusion of this parameter in the model allows us to compare model output DNA proles directly

ARTICLE IN PRESS
B. Basse et al. / Progress in Biophysics & Molecular Biology 85 (2004) 353368

358

with those obtained experimentally. At time t; the k1 Gx; t  TS cells which entered S-phase TS
% t; TS in Eq. (7)
hours previously, are due to exit S-phase and enter G2 -phase. The loss term Sx;
provides for this (see Basse et al. (2003, 2004a)), and is given by
(R
X
0 k1 G1 y; t  TS gTS ; x; y dy; tXTS ;
% t; TS
12
Sx;
0
toTS ;
where
2
1
gt; x; yE p exgty =4Dt :
13
2 pDt
% t; t is the solution of the differential equation
The term Sx;
@S%
@S%
@2 S%
@S%
x; t; t
x; t; t D 2  g x; t; t
14
@t
@t
@x
@x
% t; t 0 k1 G1 x; t and boundary conditions similar to those for
with initial condition Sx;
Eq. (7). The variable t is an age like variable and is the time that a cell spends in S-phase. From
this we see that
Z TS
% t; t dt:
Sx; t
Sx;
15

2.3. G2 -phase and mitosis (M-phase)

% t; TS ; becomes the corresponding source term in Eq. (8) representing G2 The loss term, Sx;
phase. The second term in equation (8) accounts for the loss due to exit to M-phase. Thus k2 is the
transition rate of cells from G2 -phase to M-phase. In Eq. (9), we have cells dividing at a rate of b
divisions per unit time.
2.4. Separable solutions and steady DNA distributions
We set (with the assumption that the dependent variables are separable):
G1 x; t y1 xNt;

16

Sx; t yS xNt;

17

G2 x; t y2 xNt;

18

Mx; t yM xNt;

19

and emphasise that y1 x; yS x; y2 x and yM x represent the SDD solution mode in G1 - , S- , G2 and M-phase respectively. Substituting equations (16)(19) into (6)(9) and letting l be our
separation constant, we obtain the following delay equation (20) when D 0 and Fredholm
integral equation (21) for the case where Da0:

x
; x > 0; D 0;
20
y1 x  1 Ly1
2

ARTICLE IN PRESS
B. Basse et al. / Progress in Biophysics & Molecular Biology 85 (2004) 353368

359

g2x; z; TS y1 z dz Ly1 x;

x > 0; Da0;

21

where
L F l

l k1 l k2 l belTS
4bk1 k2

22

is determined as an eigenvalue of Eq. (20) and (21) and y1 x is the corresponding eigenfunction.
To gain the steady DNA distribution of the other phases we solve the auxillary separated
equations where 0oxoX :
4byM 2x  l k1 y1 x 0;

23

k2 y2 x  l byM x 0;

24

Dy00S x  gy0S x  lyS x k1 y1 x  l k2 y2 x 0;

25

Dy0s 0  gyS 0 Dy0s X  gyS X 0;

Z X
0o
yS x dxoN:

26
27

The D 0 case, Eq. (20), can be solved using Fourier transforms in the space of Schwarzs
distributions (Stakgold, 1979). The Fredholm integral equation, (21), can be solved numerically
using methods described in Basse et al. (2004a). We see from Fig. 2 that the D 0 case results in
point distributions in G1 ; G2 and M-phases, while when D becomes non-zero its effect is to smooth
the point distributions into classical distributions.
For the total cell cohort we incorporate all phases and the SDD is given by
yT x y1 x yS x y2 x yM x;
where
Z

28

yT x dx 1;

29

and
nT x; t N0 yT xelt
N0 y1 x yS x y2 x yM xelt

30
31

is the asymptotic (assuming nx; t is an attracting solution) form of the total cell cohort. Note that
a positive or negative value of l corresponds to population growth or decay, respectively, and may
be useful in the future as a measure of the effectiveness of cancer treatment.
Experimentally, cells are cultured until they have a steady DNA distribution, they are then
perturbed by cancer treatments such as radiation or chemotherapy. Thus the steady DNA
distributions gained in this section represent the starting point for experiments. In the next section
we describe one such experiment, the addition of the anticancer drug, paclitaxel, to an
unperturbed cell line.

ARTICLE IN PRESS
360

B. Basse et al. / Progress in Biophysics & Molecular Biology 85 (2004) 353368

G1
S
G2+M

Cell Count

D=0

0.5

1.5

2.5

X Relative DNA content

Fig. 2. The SDD solutions of the model for unperturbed cell lines (Eqs. (6)(9)) have been plotted for the cases D 0
and Da0; illustrating the smoothing effect that D has on the SDD. Arbitrary parameter values are k1 0:05; k2
0:2; b 2; D 0:0004; g 0:1; xmin 0; xmax 4; Dx 0:1 D 0; Dx 0:0133 Da0; F l 0:5 ) l
0:0218:

3. Cancer therapy
Paclitaxel, a chemotherapeutic agent used in the treatment of ovarian and other types of cancer,
causes the arrest of cells in M-phase and the subsequent degradation of cellular DNA as a result
of apoptosis (programmed cell death). Flow cytometry suggests that cells build up in G2 =M-phase
and then after a time the cellular DNA begins to degrade. To model the addition of paclitaxel we
start with a steady DNA prole, set the division rate parameter b to zero and set the transition
rate parameter from M-phase, mM to a non-zero value when tXTM : We have incorporated
another compartment into the model Ax; t to describe the cellular degradation. We assume that
cells stay in M-phase for TM hours before transferring to A-phase. This means that we must keep
track of how long a cell spends in M-phase. We introduce a new parameter tM ; the time a cell has
%
been in M-phase and we let Mx;
t; tM be the number density of cells that have been in M-phase
for tM hours. We assume that cells degrade their cellular DNA at a constant rate gA (DNA units
per time). The multi-compartment model for a cell line perturbed by the anticancer drug paclitaxel
is:
@G1
x; t k1 G1 x; t;
@t

32

@S
@2 S
@S
% t; TS ;
x; t D 2  g
x; t k1 G1 x; t  Sx;
@t
@x
@x

33

ARTICLE IN PRESS
B. Basse et al. / Progress in Biophysics & Molecular Biology 85 (2004) 353368

@G2
% t; TS  k2 G2 x; t;
x; t Sx;
@t
%
%
@M
@M
%
x; t; tM
x; t; tM mM Mx;
t; tM ;
@t
@tM
Z N
%
Mx;
t; tM dtM ;
Mx; t

361

34
35
36

@A
@gA A
x; t
x; t
@t
@x

%
mM Mx;
t; tM dtM ;

37

TM

where 0oxoX and t; tM > 0; and G10 ; S0 ; G20 and M0 ; the initial distributions in G1 - , S- , G2 - and
M-phase, respectively, at time t 0; represent the starting point for an experiment and are the
SDD solutions obtained from the model of an unperturbed cell line. Boundary conditions in S
and side conditions (i.e. boundary and initial distribution) in A-phase are:
@S
@S
0; t  gS0; t D
X ; t  gSX ; t 0; t > 0;
38
D
@x
@x
A0; t 0;

t > 0;

39

Ax; 0
0;

0oxoX :

40

During A-phase, cells degrade their DNA content. Thus the equation representing A-phase is
similar to the S-phase equation (without dispersion). As DNA content degrades to x 0; it is no
longer registered by ow cytometry and the boundary condition in this phase accounts for this.

4. Optimisation and model results

We have used the Matlab function fmincon to nd the least squares t between model and
experimental DNA proles provided by the Auckland Cancer Society Research Centre (see Basse
et al., 2004a). The human cancer cell line, denoted NZM13 (New Zealand melanoma cell line
number 13) was exposed to the anticancer drug paclitaxel and subsequently analysed by ow
cytometry 0, 18, 48, 72 and 96 h later. The experimentally obtained DNA proles are shown in
Figs. 35 as solid lines.
We chose to nd the minimum distance (in the least squares sense) between model and data,
over the parameter set b k1 ; k2 ; TM ; mM ; gA ; where k1 and k2 are the transition rates from G1 phase to S-phase and G2 -phase to M-phase, respectively. TM is the time in M-phase before the
onset of A-phase, mM is the eventual transition rate of cells from M- to A-phase and gA is the
degradation rate during A-phase. These parameters were chosen because they are the most
difcult to estimate experimentally. The bounds for these parameters and values and descriptions
of other model parameters are given in Table 1. Remaining model parameters were xed at the
values: D 4  104 ; g 0:1 and b 2: The values of g and b were chosen because the average
time that a cell spends in S- and M-phase is TS 1=g 10 h and 1=b 30 min; respectively. The

ARTICLE IN PRESS
B. Basse et al. / Progress in Biophysics & Molecular Biology 85 (2004) 353368

362

value of D can be estimated by comparing model and data proles prior to the optimisation
procedure.
In order to t our mathematical model to the ow cytometric prole data, rst a model
parameter set is chosen at values which are biologically realistic. Using this parameter set, the
model SDD is found via the numerical solution of equation (21). Once the SDD is found, we have
the starting point for the model of the perturbed cell line described in Section 3. The perturbed
model does not have separable solutions and so we use the numerical method of nite differences
to solve Eq. (32)(37). From this procedure we may obtain the total DNA distribution nT x; t
(Eq. (31)) as predicted by the model and we can compare model DNA proles with those obtained
experimentally (in the least squares sense) at discrete time 0, 18, 48, 72 and 96 h following the
addition of paclitaxel.
Figs. 35 show the model results and an excellent correspondence between model and data. The
model also enables us to calculate the percentages and numbers in each phase as a function of time
(Figs. 6 and 7), the rate of entry of cells into A-phase (Fig. 8), the degradation rate in A-phase (via
the parameter gA ), and the rate of eventual cell loss (Fig. 9). The parameters TM ; the time in Mphase prior to the transition to A-phase, mM the eventual rate of entry into A-phase and gA ; the
degradation rate during A-phase are all parameters that have not been able to be measured
experimentally and the model provides a means of doing this. This is biologically important
because the aim of treatment with anticancer drugs or radiation is to increase the rate of cancer
cell death, and the results of therapy depend both on the proportion of cells killed and the rate of
cancer cell repopulation.

NZM13 DNA profile 0 hours after the addition of Taxol

NZM13 DNA profile 18 hours after the addition of Taxol

800

600

Data
Model

Data
Model

700
500
600

Cell Count

Cell Count

400
500

400

300

300
200
200
100
100

(a)

0.5

1.5

2.5

x (Relative DNA Content)

3.5

(b)

0.5

1.5

2.5

3.5

x (Relative DNA Content)

Fig. 3. Model and experimental DNA distributions for the human cancer cell line NZM13 at times 0 and 18 h (graphs
(a) and (b), respectively) after the addition of paclitaxel. This anticancer drug causes mitotic arrest and the subsequent
build up of cells in the G2 =M-phase.

ARTICLE IN PRESS
B. Basse et al. / Progress in Biophysics & Molecular Biology 85 (2004) 353368

NZM13 DNA profile 72 hours after the addition of Taxol

NZM13 DNA profile 48 hours after the addition of Taxol

350

363

250

Data
Model

Data
Model

300
200

Cell Count

Cell Count

250

200

150

150

100

100
50
50

(a)

0.5

1.5

2.5

3.5

x (Relative DNA Content)

0.5

1.5

2.5

3.5

x (Relative DNA Content)

(b)

Fig. 4. Model and experimental DNA distributions for the human cancer cell line NZM13 at times 48 and 72 h (graphs
(a) and (b), respectively) after the addition of paclitaxel. This anticancer drug causes mitotic arrest and the subsequent
build up of cells in the G2 =M-phase.

NZM13 DNA profile 96 hours after the addition of Taxol

180

Data
Model

160
140

Cell Count

120
100
80
60
40
20

0.5

1.5

2.5

3.5

x (Relative DNA Content)

Fig. 5. Model and experimental DNA distributions for the human cancer cell line NZM13 at 96 h after the addition of
paclitaxel. This anticancer drug causes mitotic arrest and the subsequent build up of cells in the G2 =M-phase.

ARTICLE IN PRESS
364

B. Basse et al. / Progress in Biophysics & Molecular Biology 85 (2004) 353368

G1
S
G2
M
A

70

Percentage of Cells

60

50

40

30

20

10

20

40

60

80

t (hours)

Fig. 6. Percentage of cells in each phase as a function of time as predicted by the model for human cell line NZM13.
Division stopped at t 0 h; prior to this the model had a steady DNA distribution as seen by constant percentages in
each phase. Model parameters that were tted were: k1 ; k2 ; TM and mM and gA : For parameter descriptions and units see
Table 1.

5. Future work
There are many possibilities for extension and application of the model and we list a few of the
immediate ones in this section.
Firstly, the lack of exact t of the DNA prole in the region near x 0 at 96 hours after the
addition of paclitaxel (Fig. 5) suggests a non-constant degradation rate of DNA content and
needs further investigation. There is some biological evidence that the DNA degradation does in
fact slow down as the DNA content becomes small.
Secondly, there are a number of different cell lines available for study from the Auckland
Cancer Society Research Centre. Fifteen of these have been exposed to paclitaxel and analysed at
later times using ow cytometry as described in the previous section. We would like to apply our
model to each of these fteen cell lines in the hope that the model can explain why some cell lines
respond differently to the drug.
Thirdly, we aim to develop a mathematical model of the cellular response to the anticancer drug
carboplatin, an anticancer drug that has broad clinical use in cancer treatment but with an action
differing substantially from that of paclitaxel. This development involves considerable
modication of the current mathematical model, incorporating additional terms into the
equations to represent additional cell death mechanisms (including senescence) and cell cycle
arrest. We already have ow cytometry data for 15 cell lines exposed to carboplatin for use in
model tting.

ARTICLE IN PRESS
B. Basse et al. / Progress in Biophysics & Molecular Biology 85 (2004) 353368

365

Number of Cells

15000

G1
S
G2
M
A
Total

10000

5000

20

40

60

80

t (hours)

Fig. 7. Number of cells in each phase as a function of time as predicted by the model for human cell line NZM13.
Division stopped a t 0 h; prior to this cells can be seen to be growing exponentially. Model parameters that were tted
were: k1 ; k2 ; TM and mM and gA : For parameter descriptions and units see Table 1.

Fourthly, we would like to use the mathematical model to examine the effects on cancer cells of
a combination of paclitaxel and carboplatin. In some cases, the individual induced cell death
mechanisms of each drug will act independently of the other, while in others, cell cycle arrest
caused by carboplatin will reduce the rate of cell death induced by paclitaxel. The behaviour
predicted by the model will be compared to that of cell lines exposed to this combination.
It may be possible to develop a generalised model which can describe a number of cancer
therapies in vitro, including combination therapies and radiation. The model equations discussed
in this paper currently involve only the relative DNA content of a cell and time. These equations
could be extended by a spatial coordinate. This gives the possibility of modelling not only cell
populations but actual (in vivo or in vitro) three-dimensional tumour growth. Studies using
molecular biological techniques have provided an excellent basis for the mechanisms involved in
the processes of cell proliferation and cell death, but have also indicated that an astonishingly high
number of molecular interactions are involved. This information cannot at present be used to
describe the dynamic interaction responsible for growth and regression of human cancers.
An extensive search of other cell population models has revealed only a handful of researchers
who directly compare ow cytometric proles with cell population model outcomes (see
Montalenti et al., 1999; Sena et al., 1999; Ubezio, 1990). Whether ours is a better approach

ARTICLE IN PRESS
366

B. Basse et al. / Progress in Biophysics & Molecular Biology 85 (2004) 353368

Proportion of M-phase cells becoming apoptotic

0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0
-20

20

40

60

80

100

Fig. 8. Model estimation of the rate of entry of cells into A-phase as a function of time for human cell line NZM13.
This rate quickly converges to the optimal value of mM 0:0394: For parameter descriptions and units see Table 1.

Total Number of Cells Lost = 3487.2207

120

Rate of cell loss from A-phase

100

80

60

40

20

10

20

30

40

50

60

70

80

90

100

Fig. 9. Model estimation of the rate of eventual cell loss from the apoptotic phase as a function of time for human cell
line NZM13. Division stopped at t 0 h: Model parameters that were tted were: k1 ; k2 ; TM and mM and gA : For
parameter descriptions and units see Table 1.

ARTICLE IN PRESS
B. Basse et al. / Progress in Biophysics & Molecular Biology 85 (2004) 353368

367

remains to be seen, however our ultimate aim is to eventually be able to model the behaviour of
cells directly from patients tumours. These cells will be a mix of both normal and cancer cells and
a mathematical model structured by DNA content may easily describe this situation and may give
insight as to why some patients fail to respond to cancer treatments.

Acknowledgements
The diagrams, Figs. 19 and Table 1, have been reproduced from our paper, Basse et al. (2004a)
(see Figs. 1, 310 and Table 1) with the kind permission of r Springer-Verlag Berlin Heidelberg,
which is gratefully acknowledged.

References
Arino, O., 1995. A survey of structured cell population dynamics. Acta Biotheoret. 43, 325.
Basse, B., Baguley, B.C., Marshall, E., Joseph, W.R., van Brunt, B., Wake, G.C., Wall, D.J.N., 2003. A mathematical
model for analysis of the cell cycle in cell lines derived from human tumours. J. Math. Biol. 47, 295312.
Basse, B., Baguley, B.C., Marshall, E., Joseph, W.R., van Brunt, B., Wake, G.C., Wall, D.J.N., 2004a. Modelling cell
death in human tumour cell lines exposed to the anticancer drug paclitaxel. J. Math. Biol., in press, http://
www.springerlink.com/link.asp?id=gbdctlgym73bxp73
Basse, B., Wake, G.C., Wall, D.J.N., van Brunt, B., 2004b. On a cell growth model for plankton. MMB JIMA 21 (1),
4961.
Bell, G., Anderson, E., 1967. Cell growth and division. Mathematical model with applications to cell volume
distributions in mammalian suspension cultures. Biophys. J. 7, 329351.
Diekmann, O., 1983. Growth, ssion and the stable size distribution. J. Math. Biol. 18, 135148.
Diekmann, O., Heijmans, H.J.A., Thieme, H.R., 1984. On the stability of the cell size distribution. J. Math. Biol. 19,
227248.
Grabosch, A., 1994. Compactness properties and asymptotics of strongly coupled systems. J. Math. Anal. Appl. 187,
411437.
Gyllenberg, M., Webb, G.F., 1990. A nonlinear structured population model of tumour growth with quiescence. J.
Math. Biol. 28, 671694.
Hall, A.J., 1998. Advanced second order functional differential equations. Ph.D. Thesis, Massey University,
Palmerston North, New Zealand.
Hall, A.J., Wake, G.C., 1989. A functional differential equation arising in the modelling of cell-growth. J. Aust. Math.
Soc. Ser. B 30, 424435.
Hall, A.J., Wake, G.C., 1990. A functional differential equation determining steady size distributions for populations of
cells growing exponentially. J. Aust. Math. Soc. Ser. B 31, 344353.
Hall, A.J., Wake, G.C., Gandar, P.W., 1991. Steady size distributions for cells in one-dimensional plant tissues. J.
Math. Biol. 30 (2), 101123.
Hannsgen, K.B., Tyson, J.J., 1985. Stability of the steady-state size distribution in a model of cell growth and division.
J. Math. Biol. 22, 293301.
Montalenti, F., Sena, G., Cappella, P., Ubezio, P., 1999. Simulating cancer-cell kinetics after drug treatment:
application to cisplatin on ovarian carcinoma. Phys. Rev. E 57 (5), 58775887.
Okubo, A., 1980. Diffusion and Ecological Problems: Mathematical Models. Springer, Berlin.
Rossa, B., 1995. Asynchronous exponential growth in a size structured cell population with quiescent compartment.
Carcinogenesis and Cell & Tumour Growth, Vol. 2. Wuerz Publishing Ltd, Winnipeg, Canada, pp. 183200
(Chapter 14).

ARTICLE IN PRESS
368

B. Basse et al. / Progress in Biophysics & Molecular Biology 85 (2004) 353368

Sena, G., Onado, C., Cappella, P., Montalenti, F., Ubezio, P., 1999. Measuring the complexity of cell cycle arrest and
killing of drugs: kinetics of phase-specic effects induced by taxol. Cytometry 37, 113124.
Sharpe, F.R., Lotka, A., 1911. A problem in age-distribution. Philos. Mag. 21, 435438.
Stakgold, I., 1979. Greens Functions and Boundary Value Problems. Wiley-InterScience, New York.
Tyson, J.J., Hannsgen, K.B., 1985. Global asymptotic stability of the cell size distribution in probabilistic models of the
cell cycle. J. Math. Biol. 6168.
Ubezio, P., 1990. Cell cycle simulation for ow cytometry. Comput. Methods Programs Biomed. Sec. II. Systems and
Programs 31 (3697), 255266.
Wake, G.C., Cooper, S., Kim, H.K., van Brunt, B., 2000. Functional differential equations for cell-growth models with
dispersion. Commun. Appl. Anal. 4, 561574.