Reports on Progress in Physics

Related content
- Invited Article
Biophysical models of tumour growth J S Lowengrub, H B Frieboes, F Jin et al.

- The fundamental role of mechanical

To cite this article: P Tracqui 2009 Rep. Prog. Phys. 72 056701 properties in the progression of cancer
disease and inflammation
Claudia Tanja Mierke

- Dynamic interplay between tumour, stroma

and immune system can drive or prevent
View the article online for updates and enhancements. tumour progression
R J Seager, Cynthia Hajal, Fabian Spill et
al.

Recent citations
- Acid-mediated tumor invasion as a
function of nutrient source location
Silvia A. Menchón

- A cell-based model for analyzing growth

and invasion of tumor spheroids
PengCheng Chen et al

- Aggregation dynamics of active cells on

non-adhesive substrate
Debangana Mukhopadhyay and Rumi De

This content was downloaded from IP address 132.248.227.150 on 08/10/2019 at 19:52

IOP PUBLISHING REPORTS ON PROGRESS IN PHYSICS
Rep. Prog. Phys. 72 (2009) 056701 (30pp) doi:10.1088/0034-4885/72/5/056701

Biophysical models of tumour growth

P Tracqui
CNRS UMR 5525, Laboratoire TIMC-IMAG, DynaCell Group, Institut de l’Ingénierie et de
l’Information de Santé (In3S), Faculté de Médecine de Grenoble, 38706, La Tronche Cedex, France
E-mail: Philippe.Tracqui@imag.fr

Received 9 October 2008, in final form 11 March 2009

Published 29 April 2009
Online at stacks.iop.org/RoPP/72/056701

Abstract
Tumour growth is a multifactorial process, which has stimulated in recent decades the
development of numerous models trying to figure out the mechanisms controlling solid
tumours morphogenesis. While the earliest models were focusing on cell proliferation
kinetics, modulated by the availability of supplied nutrients, new modelling approaches
emphasize the crucial role of several biophysical processes, including local matrix
remodelling, active cell migration and traction, and reshaping of host tissue vasculature. After
a brief presentation of this experimental background, this review will outline a number of
representative models describing, at different scales, the growth of avascular and vascularized
tumours. Special attention will be paid to the formulation of tumour–host tissue interactions
that selectively drive changes in tumour size and morphology, and which are notably mediated
by the mechanical status and elasticity of the tumour microenvironment. Emergence of
invasive behaviour through growth instabilities at the tumour–host interface will be presented
considering both reaction–diffusion and mechano-cellular models. In the latter part of the
review, patient-oriented implications of tumour growth modelling are outlined in the context of
brain tumours. Some conceptual views of the adaptive strategies and selective barriers that
govern tumour evolution are presented in conclusion as potential guidelines for the
development of future models.
(Some figures in this article are in colour only in the electronic version)
This article was invited by Professor G Gillies.

Contents
1. Introduction 2 4. Control of tumour growth by mechanical factors:
2. Biological background underlying tumour growth continuum mechanics based models 13
modelling: a brief overview 4 4.1. Modelling the effect of pressure within the
2.1. Tumour growth as a deregulation of the tumour and associated cell velocity field 13
interplay between intrinsic cellular programs 4.2. Modelling the role that stress responsiveness
and environmental constraints 4 may play in controlling tumour growth 15
2.2. Contribution of active biophysical processes 5. Morphological instabilities at the tumour/stroma
to tumour growth: cell migration and cellular interface and tumour invasion 16
forces 5 6. Tumour–host tissue interactions and tumour
vascularization 18
2.3. Transduction of physical signals 6
7. Multiscale modelling of tumour growth 19
2.4. Tumour-associated angiogenesis 8
8. Tumour growth models incorporating anticancer
2.5. In vitro models of tumour growth: MTSph 9 treatments: model-driven evaluation and
3. Modelling avascular tumour growth 10 optimization of therapeutic designs 20
3.1. From temporal to spatially structured models 10 9. Conclusions and perspectives 22
3.2. Modelling tumour growth sustained by cell Acknowledgments 25
migration in response to environmental cues 12 References 25

0034-4885/09/056701+30$90.00 1 © 2009 IOP Publishing Ltd Printed in the UK

Rep. Prog. Phys. 72 (2009) 056701
Used acronyms
EC endothelial cell
ECM extracellular matrix
EGFR epidermal growth factor receptor
MMP Matrix Metaloproteinase
MTSph multicellular tumour spheroids
TIMP tissue inhibitor of metaloproteinase
VEGF vascular endothelial growth factor
1. Introduction
Tumorigenesis, i.e. the process of initiation and progression
of a tumour, is a multifactorial process, surely emerging
from molecular and genetic abnormalities (Albertson et al
2003, Balmain et al 2003), but also strongly dependent on
cell–cell and cell–extracellular matrix (ECM) macroscopic
regulations and interactions (Bissell et al 2002, Ruiter et al
2002, Vincent and Gatenby 2008). It is known that a
tumour (from the Latin verb tumere, meaning to bulge or
to swell) originates from the transformation of normal cells
into cancer cells (carcinogenesis), which acquire the ability
to become unresponsive to normal physiological controls
(Radisky et al 2001). But this simple statement embraces the
extremely diverse and complex reciprocal dialogues that cells
and molecules comprising a tumour engage through highly
regulated interactions favouring malignancy.
In the past 25 years, a majority of cancer studies have
focused on examining functional consequences of activating
and/or inactivating mutations in critical genes implicated
in cell cycle control. These studies have taught us a
great deal about the functions of tumour promoter genes
(oncogenes) and tumour suppressor genes, as well as on
the signalling pathways regulating cell proliferation and
cell death (apoptosis). However, such studies have often
ignored the fact that cancers are heterogeneous cellular entities
whose growth is dependent upon reciprocal and dynamical
interactions between genetically altered cells and the time-
varying microenvironment in which they live. Especially,
and in view of further modelling approaches, it is worth
underlining that pathological features observed during cancer
progression are still close enough to crucial physiological
processes. Indeed, these series of multistage events involving
cellular and tissular mutual influences also orchestrate organ
morphogenesis during embryonic development and tissue
repair occurring in wound healing. Thus, tumour growth
is often presented as a ‘wound that does not heal’ (Bissell
and Radisky 2001, Chang et al 2004, Dvorak 1986, Kalluri
and Zeisberg 2006), illustrating the paradoxical aspects of
tumorigenesis and the subtle transition existing between
physiological and pathological contexts. Thus, stromas, i.e.
the arrangement of cells and connective tissue at sites of
wound healing and tumour growth, share many characteristics
(Chang et al 2004, Dvorak 1986). Other biological processes,
including cell migration, tissue remodelling and formation
of new vasculature, which are crucial and normal features
utilized during tissues repairing and healing, are also critically
2
P Tracqui
regulated events in growing tumours (Tlsty and Coussens
2006).
Thus, while cancer is still currently viewed as a disease
involving irreversible genomic changes affecting intrinsic
cellular programs, one has to consider that such genomic
alterations act in combination with modifications of extrinsic
programs taking place in environmental conditions defined
by inflammation, immune response, matrix metabolism and
stiffness, gradients of signalling molecules, tissue oxygenation
and vascular status, which all combine and underline tumour
development (Kassis et al 2001, Lunt et al 2008). In the
case of solid tumours, which are considered in this paper,
these interactions include signalling through cell adhesion
molecules, such as cadherins and integrins (Hynes 2002a),
and host tissue remodelling through the action of enzymes,
such as plasmin or metalloproteinases (Egeblad and Werb
2002), as well as differential cell response to growth factors
and signalling molecules (Balkwill 2004, Coussens and Werb
2002, Kulbe et al 2004), all of which act together to
control processes such as proliferation, survival/apoptosis,
migration and invasion. Recent research in cell and tumour
biology notably highlighted that a wide range of cellular
and molecular processes are tightly controlled by mechanical
factors depending on cell cytoskeleton and ECM mechanical
properties (Butcher et al 2009, Tilghman and Parsons 2008,
Zaman et al 2006). It is of particular interest that the expression
of many of the genes involved in biological processes quoted
above can be affected by both the biochemical and biophysical
conditions defining the tumour microenvironment, including
hypoxia and mechanical stresses (Huang and Ingber 2005,
Liotta and Kohn 2001, Lunt et al 2008, Scheurer et al 2004).
In this context, modelling of tumour growth is a rather
old research subject (Araujo and McElwain, 2004, Byrne et al
2006), but which neither stops getting increasing relevance
because of the challenge arising from considering (i) on the
one hand the exponential accumulation of data produced
by the development of new techniques from molecular
biology within all components of the ‘omics’ family, from
genomics, epigenomics and proteomics to metabolomics
(Jones and Baylin 2007, Nielsen and Oliver 2005) (ii) on the
other hand, the necessity of conducting models development
in an integrative or system biology framework in order
to counterbalance the limitation of reductionist approaches
relying only on molecular biology studies, since they are
insufficient to tackle systemic aspects of tumour development
(Anderson and Quaranta 2008, Bissell et al 2005, Bizzarri et al
2008, Ingber 2002).
Early models of tumour growth were based on kinetic
models of cell proliferation, and then mostly investigated from
a mathematical point of view (Araujo and McElwain 2004).
However, an evolution toward a more physical approach of
tumour growth has taken place in the last ten years (Preziosi
2003). Notably, theoretical models have been improved
by increasing consideration of (i) the biomechanics of the
tumour and its microenvironment, (ii) the coupling of tumour
growth with the tumour vascularization process, i.e. tumour
angiogenesis (Chaplain et al 2006, Roose et al 2007). The
evolution of tumour growth models has concomitantly coped
Rep. Prog. Phys. 72 (2009) 056701
with the complexity and non-intuitive end products originating
from multifactorial and mostly nonlinear biological processes,
which require the development and renewal of concepts in this
field (Gatenby and Maini 2002, Nagy 2005, Potter 2001, 2007,
Quaranta et al 2008, Vincent and Gatenby 2008).
It is known that one obstacle that biophysical models of
living systems must overcome is the multiplicity of the time and
length scales, from molecular and intra-cellular processes (e.g.
progression through the cell cycle, signal transduction, etc) to
tissular processes, especially microenvironment remodelling
and vascularization. At a microscopic scale, tumour models
would suitably describe malignant transformation of normal
cells and associated alterations of cell–cell interactions and
tumour heterogeneity. At a macroscopic scale, models must
consider cell response to gradient fields of diverse origins,
such as concentration gradients of diffusible or non-diffusible
molecules as well as strain and stress gradients generated by
the growing tumour mass. Thus, one might at least expect
that biophysical models of tumour growth will be able to
(i) cope with physical phenomena that differ in nature but are
still highly connected, (ii) gather in a manageable way the
most relevant data provided by multimodal tumour imaging
and by biophysical measurements of tumour and tumour-
microenvironment properties, (iii) provide explanations and
predictions helpful for the design and optimization of therapies
targeting tumour development.
Before considering more precisely the set of a priori
relevant data in this modelling perspective, it seems worth
starting with simple outlines that may shed light on the
progressive evolution of the models reviewed in this paper.
One should first observe that the growth of solid tumours
is physically a mechanical process according to which a
heterogeneous tissue expands within a surrounding medium.
Expansion is controlled by some internal driving stresses,
which are counterbalanced by the mechanical resistance
provided by the surrounding tumour environment (figure 1).
As a first implication, the interface of the tumour and its
environment appears as a region of remarkable interest since
complex phenomena could develop there, especially through
instabilities which may break a uniform and compact tumour
front into the wavy and irregular morphologies that can be
observed experimentally, and which dramatically correlate
with increased tumour aggressiveness (Weigelt and Bissell
2008).
A second modelling guideline is the recognition of the
biological specificity of the tumour growth, which is driven
by both physical and active cellular processes. Typically,
internal stresses are mostly generated by cell proliferation
dynamics, influenced by the diffusion of nutrients within the
tumour. Similarly, the physical forces existing at the tumour
front are not just force balance equations involving tumour
surface tension and the resistive pressure of the surrounding
medium. Active cellular traction forces mediate cell adhesion
and migration in this interface region (figure 1) (Gordon
et al 2003, Mierke et al 2008). These mechanical forces
reshape the tumour microenvironment, either directly by
structural changes enforcing medium anisotropy, or indirectly
by activating mechanosensitive cellular processes (Butcher
3
P Tracqui
Traction
Proliferation
activated Adhesion
stroma
Intratumoral
pressure
ECM
Tumour
Deformation
stiffness
Migration
Extracellular
stiffness
Figure 1. Schematic representation of avascular tumour/host tissue
dynamical interactions through coupled cellular and biophysical
processes. Notably, all processes contribute to the modulation of the
mechanical stresses within the tissue, stroma and tumour. Arrows
highlight that both normal and tumour cells dynamically adapt to
such stresses and strains by modifying their behaviour and
remodelling their microenvironment. Thus, the indicated circular
feedback loop might increase ECM compaction and density around
the tumour (encapsulation) or conversly contribute to tumour cell
detachment and disease progression.
et al 2009, Huang and Ingber 2005, Olson and Sahai 2009,
Peyton et al 2007). A third element underlying this schematic
modelling framework of solid tumour growth is the recognition
of the rather large range of physical properties exhibited by the
biological objects themselves, i.e. the diversity and specificity
of the biomechanical properties of individual cells (Suresh
2007), of tumours (Netti et al 2000, Venkatesh et al 2008), and
of the tumour microenvironment and host tissue (Humphrey
2003). Therefore, a rather large spectrum of mechanical and
biological responses to endogenous and exogenous stimuli has
to be expected.
Keeping in mind these introductory remarks, this paper
reviews some of the representative contributions to the
modelling of solid tumour growth. We will focus mainly
on continuous approaches, extended to some recent hybrid
continuous–discrete models coupling physical processes to
cellular automaton or multiagent-based description of cell
behaviour. Owing to the enormous body of journals and
publications devoted to the analysis of solid tumour growth,
no such review could be comprehensive. Nevertheless, we
tried to provide a concise overview, illustrating from selected
representative papers and contributions of research groups
how the field of tumour growth modelling has evolved from
earlier mathematical description to recent models more firmly
grounded on the physical and mechanical aspects sketched
above. Thus, special attention will be given to most recent
studies and their capacity to account for extended sets of
Rep. Prog. Phys. 72 (2009) 056701
in vitro and in vivo data, while underlying the continuity or
breakthrough with regard to earlier models. The main issues,
orientations and perspectives will be highlighted without going
into the details of the models formulation, which can be found
exhaustively in the corresponding quoted references.
The outline of the paper is as follows: in the continuity of
this introductory part, section 2 reviews the main components
of the increasing corpus of biological processes and data that
have supported and inspired the evolution of tumour growth
models, from earlier studies to more recent works. Section 3
presents the leading aspects of these earliest models, largely
devoted to the study of avascular solid tumour growth and
which assign to the diffusion of nutrients a critical role in
determining tumour architecture and growth rate. Recent
extensions of avascular tumour models, taking into account
more explicitly the influence of the mechanical resistance
introduced by the tumour microenvironment, are presented
in section 4. Modelling of the dynamics at the tumour–host
tissue interface is developed in section 5, emphasizing the
prominent role of the morphological instabilities that may
be generated at the tumour front beyond some bifurcation
threshold. Section 6 extends this analysis by considering
architectural modification of the tumour environment due to
tumour angiogenesis. The development of a vascular network
within the tumour vascularization is a multiscale problem that
is presented in section 7, in association with a review of
hybrid models of tumour growth. Tumour growth models are
developed with the hope of being able to predict and optimize
tumour response to treatment. These aspects are reviewed in
section 8, where the evolution of tumours in conjunction with
different therapies is illustrated in the case of brain tumours.
Conclusions and perspectives will be addressed in the final
section, focusing on (i) the ability of biophysical models to deal
with alterations of cell and tissue dynamical architecture that
characterize the first critical stages of tumour development,
(ii) some conceptual evolutions which might stimulate new
modelling strategies and (iii) the models clinical value for
improvement and new design of therapeutic strategies against
malignancy.
2. Biological background underlying tumour growth
modelling: a brief overview
This section reviews some key biological processes and
experimental data, which give consistence to relevant
modelling of tumour growth. Without going into excessive
detail, the challenge is to provide a concrete background
and critical arguments to evaluate the relevance, benefits and
limitations of the models that will be presented in the following
sections.
2.1. Tumour growth as a deregulation of the interplay
between intrinsic cellular programs and environmental
constraints
Let us recall that solid tumours are heterogeneous multicellular
entities containing a diverse population of cells, composed
of genetically mutated cancer cells as well as other cell
4
P Tracqui
types present in the tumour microenvironment, including
fibroblasts, immune cells and cells lining blood vessels.
These cells are embedded in irregular tissue architecture
exhibiting an abnormal microcirculation, with a loss of
vascular hierarchy and large variability in vessel diameter,
length and tortuousness.
Since cancer is still largely considered as a genetic
disease rather than as a system biology disease (Bizzarri et al
2008), much research has focused on genetic and molecular
properties of the tumour cells themselves. Nevertheless, such
genomic alterations act in concert with several biological and
biophysical processes, which modify local tissue homeostasis
by stimulating cell proliferation, motility and differentiation,
while simultaneously impairing cell death (apoptosis), cell
adhesion and microenvironment architecture (Kass et al 2007).
Thus, hypoxic microenvironment of a tumour modifies the
expression of a number of genes contributing to cancer cells
survival and growth, and to metastatic efficiency (Keith and
Simon 2007, Scheurer et al 2004). Tissue vascularization,
immune response and ECM turnover depend on reciprocal
interactions between normal and cancer cells, between cells
and soluble molecules, and between cells and ECM (Tlsty and
Coussens 2006, Weigelt and Bissell 2008). The mechanical
stresses experienced by tumour cells and tumour stroma may
equally affect cell phenotype and the expression of several
genes through mechanotransduction processes (Butcher et al
2009, Huang and Ingber 2005, Liotta and Kohn 2001, Lunt et al
2008, Scheurer et al 2004), as further detailed in section 2.3.
Globally, solid tumour growth cannot be sustained unless
the tumour efficiently co-opts inflammatory cells, vascular
cells and fibroblasts to help reset the balance between ECM
degradation and ECM biosynthesis (Bhowmick et al 2004,
Coussens and Werb 2002, Tlsty 2001).
Fibroblasts, the predominant cells in the ECM (Bosman
and Stamenkovic 2003), were thought to be passive
participants in neoplastic programming of tissues. However,
recent data indicate that they exert an active role: they are able
to develop significant mechanical forces (Gutierrez-Fernandez
et al 2008, Tymchenko et al 2007) and are responsible for
the production of various classes of proteolytic enzymes, their
inhibitors, as well as many soluble paracrine growth factors
that regulate cell proliferation, morphology, survival and death
(Kalluri and Zeisberg 2006, Sugimoto et al 2006, Zeisberg
et al 2007). In response to pathologic signals, fibroblasts
change their phenotype and function and undergo dynamic
changes accompanying tumour progression (Zeisberg et al
2007). While multiple cell types contribute to the maintenance
of appropriate ECM composition at the invasive front, similar
to that observed during wound healing, fibroblasts and vascular
cells are particularly responsible for the new synthesis and
remodelling of ECM components (collagens, fibronectin, etc)
that commonly surround solid tumours.
It is intuitively clear that inappropriate synthesis
or degradation of any ECM molecule can alter the
architecture and the mechanical state of the stroma and
tumour microenvironment, thus modifying host tissue
integrity. Indeed, the turnover of ECM molecules—synthesis
balanced by degradation— is essential for maintaining tissue
Rep. Prog. Phys. 72 (2009) 056701
homeostasis (Wolf et al 2007). When developing neoplasms
undergo malignant conversion and acquire the capacity to
invade normal tissue, increased architectural disorder at the
invasive front of the neoplastic mass is characteristic, with
the balance between ECM synthesis and degradation shifted
in favour of ECM accumulation or fibrosis. Accordingly,
composition and architecture of ECM varies considerably
during tumorigenesis and the newly synthesized peri-tumoral
stroma is mostly characterized by a disorganized orientation
of the collagen fibres (Ruiter et al 2002). Importantly,
such increased collagen synthesis and dense accumulation of
fibrillar collagens result in tight encapsulation of malignant
cells that creates a physical barrier to tumour expansion (Kass
et al 2007).
This deregulation toward increased synthesis of extracel-
lular components is more or less counterbalanced by enzymatic
degradation (proteolysis) of pericellular ECM molecules in the
tumour neighbourhood, predominantly by matrix metallopro-
teinases (MMPs). This large family of proteolytic enzymes,
which normally remodel ECM, are secreted by epithelial
cells, activated fibroblasts and inflammatory cells (Coussens
et al 2002, Egeblad and Werb 2002, Sternlicht and Werb
2001). These enzymes are either soluble or plasma membrane-
associated and play key roles in wound healing and in cancer
progression (Egeblad and Werb 2002). Interestingly, these
enzymes regulate tumour development in several ways: they
were originally described as cleaving ECM substrates, but sev-
eral recent studies indicate they could also have a surprising
and counterintuitive protective role (Gutierrez-Fernandez et al
2008). As in many enzymatic networks, specific inhibitors
control MMPs activity, namely the tissue inhibitors of met-
alloproteinases (TIMPs) (Lafleur et al 2003). Globally, the
expression of many MMPs and TIMPs, and thus ECM remod-
elling, is regulated at the transcription level by a variety of
growth factors, cytokines and chemokines, but also at a post-
transcriptional level in specific cases (Clark et al 2008, Fu
et al 2008). Thus, tumour growth takes place in a highly dy-
namic landscape that develops a multifactorial environmental
response to tumour development.
2.2. Contribution of active biophysical processes to tumour
growth: cell migration and cellular forces
Tumour cells migration is obviously a determinant process
favouring solid tumour expansion. Let us recall that cell
migration involves a series of highly coordinated events,
including the extension of cell membrane protrusions, the
formation of new adhesions, the development of traction forces
and the release of previous adhesion to ensure cell body
translocation (Sheetz et al 1998, 1999). In the context of
cancer cell migration, special attention has been paid to thin
actin-rich protrusions, termed invadopodia, that colocalize
high proteolytic activity and cytoskeletal signalling pathways
to cell–ECM contact points.
All these processes received increasing attention from
the biologists and physicists. Thus, observations of the
nanodynamics of invadopodium by atomic force microscopy
revealed the remarkable directionality of this cell projection
5
P Tracqui
and its sensitivity to cell–cell biochemical signalling (Fillmore
et al 2003, Chasiotis et al 2003). Since the pioneering
work of Harris et al more than two decades ago imaging the
wrinkling of thin elastic substrates by adherent cells (Harris
et al 1981, 1984), analysis of cellular traction forces fosters
the development of a remarkable diversity of experiments.
Indeed, even non-muscle cells, like endothelial cells (ECs)
or fibroblasts, remarkably generate active tension through an
actomyosin filament sliding mechanism similar to the one
used in the contraction of smooth muscles cells. Importantly,
both normal and cancer cells exert traction forces onto their
own environment (du Roure et al 2005). The qualitative
and quantitative features of cell traction forces exerted by
different types of cells, including tumour cells, are now
rather well documented (Delvoye et al 1991, Dembo et al
1996, Shreiber et al 2003), thanks to the large spectrum
of experimental setups developed at different scales. This
spectrum includes measurements of isometric tension or
compaction rates of viscoelastic biogels contracted by cellular
forces (Benkherourou et al 2000, Eastwood et al 1996,
Kolodney and Wysolmerski 1992), as well as the tracking
of microbeads displacements and of the deformations of
substrates with printed micro-lithography patterns (Balaban
et al 2001, Beningo et al 2002, Oliver et al 1999). More
recently, several micro and nanotechnology devices, designed
as discrete deflectable elements such as micro-cantilevers
(Galbraith and Sheetz 1997) or elastic micro-posts (du Roure
et al 2005, Tan et al 2003), allow force quantification from
direct relations between the element deflection and the force
applied on it. The reported magnitude of forces is in the order
of 10 nN, corresponding to traction stresses in the range 2.5–
5.5 nN mm−2 (du Roure et al 2005). However, even if such
efficient cellular machinery may potentially promote tumour
growth through tumour cell migration, we still largely ignore
at this stage how it may become effective in the environmental
context prevailing in the tumour neighbourhood.
In addition to random migration, it is known that
cells can exhibit directional migration, or taxis, in response
to environmental gradients of both soluble chemicals
(chemotaxis) and substrate-attached molecules (haptotaxis)
(Hsu et al 2005, Lamalice et al 2007, Lecaudey and Gilmour
2006, Weiner 2002). In addition, it has been shown more
recently that ECM mechanical properties also influence cell
migration (mechanotaxis) (Kass et al 2007, Zaman et al 2006).
Importantly, cells are able to discriminate between soft and
rigid ECM (Alexander et al 2008), then developing a net cell
motion in the direction of higher ECM rigidity (Zaman et al
2006), a phenomenon known as durotaxis (Lo et al 2000).
Furthermore, architectural properties of ECM are of critical
importance for directional cell migration. Indeed, cellular
shape, orientation, adhesion and migration can be guided by
the topography of the microenvironment (figure 2). Such
topographic guidance of cell migration has been evidenced by
the alignment of cells migrating on micro-machined grooves
in the substrate or on micropillars (Saez et al 2007, Tzvetkova-
Chevolleau et al 2008).
Taken together, these few experimental findings indicate
that many possible regulatory loops may differentially affect
Rep. Prog. Phys. 72 (2009) 056701
(a)
Figure 2. Compilations of 2D cell trajectories exhibited by 3T3 fibroblasts migrating on polydimethylsiloxane (PDMS) flat substrate (a)
and on PDMS micropatterned with lines oriented horizontally and having an equal width and inter-linear space of 750 nm (c). The
departure of each cell trajectory has been shifted to the origin (0,0), with distance given in micrometers. Cell trajectories are random over
the flat substrate (a), while lines promote a cell guidance phenomenon and induce a preferential cell migration in their direction (c). This
directional migration occurs in pace with a highly polarized morphology of the migrating cells (b). Scale bar in (b) is 20 µm (from
Tzvetkova-Chevolleau et al 2008).
the dynamics of normal and cancer cells, and thus the
expansion of a solid tumour. For example, cellular forces may
induce the re-organization of the tumour microenvironment
from a diffuse quasi-isotropic to an anisotropic fibrillar ECM
assembly, thus modifying ECM and stroma architecture, and
thus exerting feedback onto cell migration and associated
tumour growth (Spencer et al 2007). In this case,
mechanosensing processes have a leading role in the
determination of the cellular response.
2.3. Transduction of physical signals
Transduction of physical signals is known to regulate a large
array of physiological processes, from hearing and touch to
bone and vasculature shaping (Jaalouk and Lammerding 2009).
Therefore, it is not surprising that failure of mechanosensitive
processes contributes to cancer if tumour cells, already
stimulated by various chemical signals, additionally sense
and respond to the mechanical cues provided by mechanical
interactions with their extracellular environment (Bissell et al
2002, Sottile 2004). More precisely, it appears clearly that
ECM provides not only the necessary structural support for
cell migration as presented previously, but is also part of the
deformable continuum able to transmit signals through cell–
surface adhesion sites (Chen 2008, Geiger et al 2001, Li et al
2005). Such adhesion regions correspond to different clusters
of proteins, ranging from dot-like adhesion to larger focal
adhesions (Hirata et al 2008). Cell adhesion at these sites
is accompanied by a remodelling of the cytoskeleton network
into densely packed straight fibres made of actin filaments.
Specific transmembrane receptors like integrins, a large family
of heterodimers which differ by the combination of their two
subunits (α, β) (Hynes 2002b), ensure a physical continuity
and dynamic links between extracellular environment on one
side of the plasmic membrane and the complex molecular
bridges that connect the cytoskeleton actin filaments to the
cytoplasmic domains of integrin subunits on the other side
(figure 3) (Berrier and Yamada 2007, Gieni and Hendzel 2008).
These scaffolds define a multi-component mechanosensory
P Tracqui
(b) (c)
switch (figure 4) through which control and adaptation of cell
response to mechanical stresses and strains will operate (Chen
2008, Geiger and Bershadsky 2002, Geiger et al 2001, Wang
et al 2007).
Thus, cellular tension drives cell adhesion and focal
adhesions assembly, growth and maintenance depend on
mechanical forces applied to them (Smith et al 2008).
Such adaptive behaviour was investigated using extracellular
substrates of different mechanical stiffness, with studies
reporting that cells organize their cytoskeleton and adhesive
contacts differently on soft and stiff surfaces (Alexander et al
2008, Collin et al 2006, Geiger et al 2001). In the light of the
proposed mechanosensory switch behaviour, the tension acting
on the adhesion sites of cells attached to a flexible substrate
may be smaller than when cells are spread over a rigid surface.
Consequently, dimensions of focal adhesions formed on soft
elastic substrates are smaller than those formed when cells are
attached on a stiffer ECM (Balaban et al 2001, Riveline et al
2001).
According to these experimental data, the interplay
between mechanical forces and tissue growth and remodelling
appears as a central process in the modelling of tumour
growth, with mechanotransduction, i.e. the conversion of
mechanical force into biochemical information (Huang et al
2004, Li et al 2005, Orr et al 2006), being a key function
for understanding how tumour cells and tumour growth are
affected by mechanical constraints. Since this structural
continuity is ensured up to the cell nucleus and the nuclear
matrix (Wang et al 2009), the tensional state of ECM and
tumour stroma may influence cell dynamics by modifying
cellular phenotypes and expression of mechanosensitive genes
(Gieni and Hendzel 2008, Orr et al 2006, Wang et al 2007).
The ECM and tumour mechanical status depends on several
factors, originating from intracellular contractile machinery,
from external loads due to mechanical activity of neighbouring,
from resistive stresses exerted by the host tissue on the growing
tumour as well as from movement of fluids within tumour
interstitium (Netti et al 2000).
6
Rep. Prog. Phys. 72 (2009) 056701 P Tracqui

Transmembrane
receptors

Cytoskeleton
stress fibres
nucleus

ECM Focal adhesion

σcell
Actomyosin
Focal contractility
Integrins
adhesion Transmembrane
Cytoskeleton
receptors
tension

β σext
α
ECM tension

Figure 3. In adhesive microenvironment, here simplified as a 2D substrate (upper figure), normal and cancer cells can generate, transmit and
respond to inside–out and outside–in cues. This signalling process involves the anisotropic distribution of cytoskeleton stress fibres (bundles
of actomyosin) that are anchored into focal adhesions. Fibres contraction promotes the clustering of transmembrane receptors (integrins)
and the accumulation of proteins that make up the focal adhesion complex. Schematic close-up of a focal adhesion (lower figure) illustrates
that the balance of external and internal forces drives the cell response to ECM tension and stiffness cues.

environment through bi-directional (outside–in/inside–out)

Figure 4. The balance of internal and external stresses at a
mechanosensor, like focal adhesion, drives the tensional
homeostasis experienced by the cell and the subsequent up or down
regulation of fundamental cellular functions such as proliferation,
adhesion, migration or differentiation. The stress balance depends
on both active elements (intracellular contraction or traction exerted
by neighbouring cells) and passive components (cystoskeleton and
ECM stiffnesses). Then, softening or hardening, i.e. remodelling of
these viscoelastic components, directly affects the local tensional
homeostasis.
These considerations underline that the development and
relevance of biophysical models of tumours have to be
examined in the light of different concepts. A first one
is the dynamic reciprocity concept proposed by Bissel and
co-workers (Bissell and Aggeler 1987, Nelson and Bissell
2006) to qualify the mutual interactions of cells and their
signalling feedback schemes. As a mechanical extension
of the homeostasis concept, a cornerstone of physiological
regulations, the tensional homeostasis concept, and its
deregulation in cancer, was invoked more recently as a
relevant background for analysing tumour growth (Butcher
et al 2009, Paszek et al 2005). Both concepts are linked,
the latter one qualifying how the tumour microenvironment
specifies a contextual information framework for both normal
and cancer cells to determine the correct (or incorrect)
response to superimposed biochemical and biomechanical
stimuli (figure 4). Dynamic reciprocity usually involves
autocrine and paracrine regulatory loops (Bussolino et al
2006, Lee et al 2007). For example, when carcinomas
develop, epithelial cells can initiate incorrect signalling,
resulting in stromal cell production of growth factors that,
in turn, stimulate inappropriate proliferation of the secreting
cells themselves (autocrine feedback) or of neighbouring
cells (paracrine regulation). Focusing on mechanoreciprocity
(Butcher et al 2009) while keeping for biomechanical
regulations a similar conceptual framework, it may be
useful to introduce autobaric and parabaric pathways through
which cells influence themselves and each other through
mechanical signals (Banes et al 1995, Tracqui 2006). Thus,
autobaric processes define a self-enhancement of the cell
stress state by the cell’s own contraction, as illustrated in
the mechanosensitive development of focal adhesions quoted
previously. Parabaric regulatory loops may be initiated by a
cell increasing its mechanical load over neighbours’ cells by
short-range mechanical deformation transmitted by the ECM.

7
Rep. Prog. Phys. 72 (2009) 056701
Blood vessel
VEGF,...
EC
Endothelial proliferation EC
cells (EC) and sprouting chemotaxis
Figure 5. Tumours induce blood vessel growth, i.e. angiogenesis, by secreting various growth factors like vascular endothelial growth factor
(VEGF), which binds to receptors present on ECs membrane. ECs then release proteases, escape from the original parent vessel walls
(sprouting) and migrate by chemotaxis toward the tumour where they form new vessels. Single tumour cells can break away from the
tumour mass and enter one of these blood vessels, being then carried to a distant site where they can initiate the growth of a secondary
tumour (metastasis).
To what extent biophysical models of tumour growth already
consider these diverse regulatory loops will be presented in
the following sections. One should however stress that it is
often difficult to isolate and analyse separately such regulatory
loops, since they evolve together within a ECM that changes
temporally as an adaptation to varying cell responses (Nelson
and Bissell 2006).
2.4. Tumour-associated angiogenesis
For maintaining tissue homeostasis and influx of oxygen
and nutrients, formation and activation of blood vessels are
essential during all tissue repair and growth processes as
well as during cancer development (Avraamides et al 2008,
Cavallaro and Christofori 2000, Folkman 2002, Nyberg et al
2005). When a primary tumour expands, the development of
a new blood vasculature is required to sustain the metabolic
needs of this rapidly growing cellular mass and to ensure
the elimination of waste products. This is accomplished by
activation of pre-existing vascular networks, a process referred
to as angiogenesis, by opposition to vasculogenesis that defines
the de novo formation of new vessels (Carmeliet 2003).
During angiogenesis, a well-orchestrated series of events
occurs, encompassing EC proliferation and further directional
migration of these cells through remodelled basement
membrane and stroma (figure 5). ECs respond to the
angiogenic stimuli released by inflammatory cells and by the
expanding populations of cancer cells constituting the tumour
(Folkman 2003). Further recruitment of perivascular support
cells enables stabilization of nascent vessels, functional
lumen formation and appropriate blood flow. Thus, tumour
angiogenesis is a multifactorial process, regulated in part by
local changes in the relative balance between soluble and
insoluble either pro- or anti-angiogenic mediators, which
impact on EC proliferation, differentiation, migration and 3D
organization into tubular structures with functional lumen.
Pro-angiogenic polypeptide growth factors comprise most
notably different forms of VEGF (Lee et al 2007), basic
fibroblast growth factor (FGF) and tumour necrosis factor TNF
8
P Tracqui
VEGF, ...
Angiogenic
vessel
Avascular
tumour
Tumour
angiogenesis
(Gharaee-Kermani and Phan 2001, Gruss et al 2003). These
factors are highly expressed in developing tumours, as they are
during physiologic wound healing. Availability of diffusible
angiogenic inhibitors and stimulators is however limited, as
they are either sequestered to ECM molecules or tethered
to cells via membrane anchorage domains. Extracellular
proteases like MMPs regulate the release of these factors,
rendering them available for interaction with receptors on
vascular cells and thus activating the development of tumour-
associated vasculature. Intuitively, MMPs were previously
thought to facilitate tumour expansion by merely degrading
ECM structural components, thereby allowing a cleared
path for migrating tumour or ECs. In fact, recent studies
indicate that proteolytic degradation of ECM by MMPs
release molecules that act as positive or negative regulators of
tumour-associated angiogenesis, depending on host tissues and
development stage of the tumour (Bergers et al 2000, Coussens
et al 2002, Lafleur et al 2003).
Such considerations of biochemical regulatory pathways
do not explicitly account for the morphogenesis of the
tumour-associated angiogenic vasculature, which is distinctly
tortuous and inherently unstable. In fact, while considerable
information is known concerning the diverse assortment
of positively and negatively acting soluble and insoluble
factors regulating angiogenesis, less is known about the
role of mechanical factors in the onset and modulation of
vascular network morphogenesis. Significant insights into
the mechanobiology of angiogenesis have come from the
development and analysis of in vitro assays using cultured
ECs (Vailhe et al 2001). These assays, referred to as
in vitro angiogenesis or tubulogenesis, mimic quite well the
early morphogenesis of pseudo-capillaries networks, during
which ECs cultured on ECM may organize as pseudo-
capillaries around the meshwork of growing areas devoid
of ECM, the lacunae. Such patterning of ECs have in the
last ten years attracted the attention of experimentalist and
theoreticians by the intriguing fact that they appear to follow
self-organizing rules (Gamba et al 2003, Namy et al 2004,
Tosin et al 2006, Tracqui 2006). Several types of intervening
Rep. Prog. Phys. 72 (2009) 056701 P Tracqui

(a)
(b)

Figure 6. Adhesion of transformed EAhY926 ECs onto the surface of Cytodex© microbeads (diameter ∼200 µm) mimics the existence
of outer rim of cells shedding from the tumour surface and interacting with the surrounding ECM. When focusing on one of these
microcarriers (a), significant displacements (arrows (a)) of the viscoelastic fibrin matrix underlying the microcarrier can be reconstructed
from time-lapse videomicroscopy sequences, revealing the field of traction forces generated by the cells. Interestingly, for soft enough fibrin
gels, ECs self-organized into circular cellular chords (arrows (b)). This experimental feature illustrates how vessels morphogenesis could be
sustained and modulated in a tumour microenvironment through stroma stiffness and cell–ECM mechanical interactions.

pathways have been identified, among which are cells–ECM
mechanical interactions. Indeed, ECM may sustain the
early morphogenesis of this capillary-like structures network
by providing positional cues, by mechanically influencing
microvessel cell behaviour in a given range of ECM stiffness
(Ghajar et al 2008, Stephanou et al 2007, Vailhe et al 1997)
and by transmitting mechanical loads imposed to neovessels
during angiogenesis (Krishnan et al 2008).
Interestingly, in vitro assays conducted with microbeads
can illustrate how such self-organizing phenomenon could
be induced from the outer rim of cells shedding from
the tumour surface. Following a previously described
in vitro microcarrier-based angiogenesis assay (Nehls and
Drenckhahn 1995), we cultured transformed ECs (EAhY926
cell line) derived from human umbilical vein endothelial
cells (HUVECs). EAhY926 cells were allowed to migrate
onto microcarrier beads, generating an outer rim of cells.
Then, microcarrier beads are laid down onto a fibrin gel
and the outgrowth of the ECs is recorded by time-lapse
videomicroscopy. Figure 6 illustrates two key features of this
experiment. First, the centripetal displacement of the fibrin gel
toward each microcarrier reveals the existence of significant
cell traction forces generated by cells progressively shedding
from the microbead (figure 6(a)). Second, the migration of
cells from the outer bead surface, combined with cell–cell
and cell–fibrin matrix mechanical interactions, result in a self-
organized morphogenesis of a lacuna within the fibrin gel,
surrounded by a circular chord of cells (figure 6(b)).
This morphogenetic process occurs for a given range of
fibrin gel stiffness, in agreement with data obtained from usual
in vitro angiogenesis assays where ECs are initially uniformly
seeded onto the fibrin gel surface (Stephanou et al 2007, Vailhe
et al 1997, Vernon et al 1992). This feature also illustrated
how the sensitivity of tumour-derived ECs to the elasticity
of the tumour microenvironment may contribute to normal or
abnormal development of the local microvasculature (Ghosh
et al 2008).
A clear limitation of the microcarrier assay is that cell
dynamics and interactions are limited to the microcarrier
surface. This is not the case with multicellular tumour
spheroids (MTSph), the most common in vitro assay developed
to mimic solid tumour growth. Since a large number of the
biophysical models reviewed in this paper have been developed
and discussed with respect to this assay, we briefly present in
the next paragraph the properties and major aspects of this
experimental model.
2.5. In vitro models of tumour growth: MTSph
Many types of mammalian cells can aggregate into 3D
structures named multicellular spheroids (figure 7). Since
monolayer cultures of tumour cells cannot represent the
characteristics of three-dimensional solid tumours, MTSph
are well established in experimental cancer research as 3D
structures of intermediate complexity (Mueller-Klieser 1997,
2000, Sutherland 1988). One major advantage of MTSph is
their reproducible concentric arrangement of the different cell
populations found in tumours microregions. MTSph are thus
excellent in vitro models for studying cell chemosensitivity, as
well as intrinsic and acquired forms of resistance to anticancer
agents. Structurally, the cross-section of MTSph reveals,
starting from the spheroid centre (figure 8): (i) an inner necrotic
core, which is largely variable depending on the cell type
and on the culturing conditions, with a first layer of hypoxic
quiescent cells, (ii) an intermediate well-oxygenated zone
with quiescent cells, (iii) a well-oxygenated outer rim with
actively proliferating cells (Mueller-Klieser 2000). MTSph
are usually classified into three groups according to the
relative emergence of necrotic and hypoxic regions (Mueller-
Klieser 2000). However, other factors may influence MTSph
properties, as illustrated by the fact that not all malignant
cells can form MTSph. Moreover, MTSph can hardly be
vascularized when co-cultured in vitro with ECs, while they
become vascularized following implantation into animals.

9
Rep. Prog. Phys. 72 (2009) 056701 P Tracqui

(a)

Tumour volume (mm3)

(c)
(b)

Figure 7. Growth of multicellular spheroids of liver epithelial cell line proliferating in soft agar gel and visualized at different days Ji after
cells seeding (a). (b) provides a close-up view of one spheroid at day 11. When inoculated subcutaneously into nude mice, these cells grew
as a tumour. The tumour growth curve exhibits a typical shape (c), with a plateau corresponding to the inhibition of tumour expansion (see
section 3.1). (Courtesy of A Bessard and G Baffet, INSERM U522, Rennes.)

R(t) by embedding MTSph in extracellular matrices, like agar

Large (figure 7) or collagen, with tunable composition and stiffness.
proliferating Low oxygen
cells level Thus, Helmlinger et al (1997) investigate the influence
of external mechanical loading on MTSph by conducting
experiments where carcinoma and sarcoma cells were growing
in an agarose matrix. The matrix stiffness was tuned by
varying the volume fraction of the solid component during
nutrients the preparation. The spheroid growth was then constrained by
Necrotic resistive stresses fields with different magnitude. Experimental
centre results showed that the size reached by the growing spheroid
decreases as the stiffness of the gel increases. Concomitantly
High oxygen to the reduction of spheroid size, decreasing apoptosis rate and
level increasing cellular packing were observed.
Biophysical models of tumour growth took the benefit
Small
proliferating cells of such direct experimental analysis of structure to function
waste products
properties, where the proliferative and metabolic properties of
cells in spheroids are related to their radial position. Thanks
Figure 8. Schematic cross-section of a multicellular spheroid. to their well-defined spherical geometry, theoretical analysis
Within the spherical shape, a concentric arrangement of large of MTSph growth may use simple, radially symmetric or
proliferating tumour cells is found in the periphery, while smaller
non-proliferating cells concentrate in deeper regions. Diffusional asymmetric models that will be repeatedly encountered in the
constraints to oxygen and nutrients supply increase with spheroid different following sections, considering notably mechanical
size and a central necrotic core emerges as a result of cell death for a properties of MTSph in section 4 and irregular MTSph
certain spheroid radius R. This value, which largely depends on the morphologies explained by shape instability scenarios in
cell type and on cell culture conditions, is one of the main variables section 5.
considered in MTSph growth models.

3. Modelling avascular tumour growth

This indicates that MTSph properties are determined both
by intrinsic properties of cells and by cell–cell and cell–
microenvironment interactions.
These considerations stimulate the design of experiments
for mimicking tumour stroma influence. Evolution of the
above standard MTSph outgrowth assay has been proposed
3.1. From temporal to spatially structured models
The modelling of avascular tumour growth is a first step
toward an integrated description of tumour development and
metastasis. The earliest and simplest models described the
changes over time of the number of cancer cells within a

10
Rep. Prog. Phys. 72 (2009) 056701
solid tumour, with cell proliferation following an unlimited
exponential growth law or more realistically Verhulst’s
(logistic), von Bertalanffy or Gompertz’s growth laws (Afenya
and Calderon 2000, Kozusko et al 2007a, Marusic et al 1994a,
Retsky et al 1990). The latter include slowdown and inhibition
of tumour’s evolution, due to contact inhibition between
packed cells and deprivation of environmental resources. In
addition to these standard models, alternative descriptions have
been recently proposed (Bru et al 2003, Delsanto et al 2004,
Kozusko and Bourdeau 2007b), looking for some universality
in the characterization of the growth process. Restricted to
the temporal aspects of tumour growth, these models reduce
to a single ordinary differential equation whose solution can
be successfully compared with experimental growth curves
exhibited by in vitro growing MTSph (figure 7(c)) or in vivo
growing tumours (Marusic et al 1994b, Retsky et al 1990).
However, as pointed out in Araujo and McElwain (2004),
such models are essentially mathematical descriptions, with
model parameters hardly related to physical processes. This
limits their explicative power, even if they can be extended
to deal with tumour heterogeneity and the existence of sub-
populations of cancer cells (polyclonality) emerging from
genetic mutations and characterized by different proliferation
rates and sensitivity to environmental or chemical cues.
In order to understand more precisely the influence
environmental conditions have on tumour growth, notably
nutrients delivery and acidity levels, spatially structured
reaction–diffusion models have been developed as moving
boundary problems, in which the position of the outer tumour
boundary is unknown (Ferreira et al 2002, Smallbone et al
2005). The generic physical driving processes that will
be commonly encountered in models are the diffusion of
nutrients and oxygen, provided by the surrounding host tissue
and consumed by tumour cells. Since oxygen and nutrient
availability limits cell proliferation, the expected outcome of
such models is the definition of conditions leading to the
appearance of necrotic and hypoxic regions. When the tumour
is small, nutrient availability stimulates the cell proliferation.
As the tumour expands, cells located in the central part of
the tumour are progressively starved of nutrient. Cell death
increases, generating a necrotic core in regions when nutrient
falls below some critical values (figure 8).
As a simplification, tumour cell density is no longer
explicitly considered as a variable of these spatial models. The
nutrient concentration n(r, t) and the tumour radius R(t) are
the generic variables, with tumour cells proliferation assumed
proportional to the local nutrient concentration. From a formal
point of view, nutrient concentration is given by a reaction–
diffusion equation. Coupled to the diffusion of nutrient
(diffusion coefficient Dn ), the rate of nutrient consumption
depends on the tumour radius R(t) to account for tumour
heterogeneity (figure 8), leading to the simple mass balance
equation
∂n
= Dn n − f (c, R). (1)
∂t
Tumour growth is given by a differential equation, describing
the rate of change of the tumour volume from the balance
between the nutrient-dependent rates of cell proliferation
11
P Tracqui
and death. The main assumption (and limitation) of such
models is to consider for simplicity the growth of a spherical
tumour. This symmetry condition allows the determination
of the tumour boundary expansion from the one-dimensional
movement of the outer tumour radius R(t) in the radial
direction (0, r) as the solution of the integro-differential
equation (Byrne and Chaplain 1996a):

R
R
d 3
(R ) = G(n, . . .)r 2 dr − A(n, . . .)r 2 dr, (2)
dt Rnc 0
where the functions G(n, . . .) and A(n, . . .) model indirectly
cell proliferation and cell death, respectively, with Rnc being
the radius of the necrotic core. As growth proceeds, a
multilayer architecture emerged within the tumour from the
imbalance between supply and demand of nutrient. Thus,
such models are simple and successfully reproduced the
multilayered architecture of MTSph cultured in vitro, with an
outer rim of nutrient-rich, proliferating cells, a central core of
nutrient-starved, necrotic debris and an intermediate region of
nutrient-poor, quiescent cells, as described in section 2.5. In
addition, analytical and quantitative predictions are possible
regarding (i) the time at which the necrotic core of the
idealized tumour occurs, (ii) the size of the tumour at the
onset of necrotic core formation, which depends on the model
parameters values, notably the diffusion coefficient of the
nutrient.
Generalization of this diffusion-consumption scenario to
other diffusing molecules (hydrogen ions, glucose, lactate,
etc) and growth factors within reaction–diffusion models
allows a more refined description of the control exerted
by the host microenvironment onto the growing tumour
(Casciari et al 1992, Kelly et al 2008, Smallbone et al
2005). Thus, Gatenby et al (2006) explored an acid-mediated
tumour invasion hypothesis, according to which acidification
of the tumour microenvironment is toxic to normal cells, but
advantageous to cancer cells, thus triggering the transition
from benign to malignant tumour growth. Their model
consists of three coupled reaction–diffusion equations for
cancer cells, normal cells and pH, and simulates the death
of normal cells induced by an increasing acidity generated
by neighbouring cancer cells. The latest extension of this
model gives a still more accurate description of the role
of acidosis in tumour growth by considering the respective
contribution of active and quiescent cells (Smallbone et al
2008). Additionally, Swietach et al (2008) used diffusion–
reaction modelling to explore the potential of cell membrane
enzyme for controlling the gradients of intracellular pH within
a tumour, a more uniformly alkaline tumour favouring further
growth.
These models, incorporating a limited number of
biological processes and simple geometries, do not account
for the observed irregular shape of multicellular spheroids or
tumours. This consideration stimulated the development of
models trying to include cell motility processes and cellular
response to environmental cues.
Rep. Prog. Phys. 72 (2009) 056701
3.2. Modelling tumour growth sustained by cell migration in
response to environmental cues
Modelling tumour growth in connection with the invasion
of tumour cells into the host tissue is a major challenge,
since the switching on or off of this migratory stage has a
dramatic influence on the induction of the metastasis process.
We know from section 2.2 that direction and speed of
cells migration depend on environmental stimuli, especially
gradients of soluble chemicals (chemotaxis), gradients of
substrate-attached molecules (haptotaxis) or ECM mechanical
properties (mechanotaxis), including gradient of matrix
rigidity (durotaxis) and ‘substrate guidance’ monitored by the
topography and anisotropy of the microenvironment (figure 2).
However, spatio-temporal evolution of such directional
migration processes may become hardly predictable when the
tumour microenvironment continuously varies in composition
as a result of growth factors and ECM turnover, or in
organization because of ECM deformations induced by cellular
traction forces. A crucial property of biophysical models is
then their ability to determine whether the coupling between
tumour microenvironment evolution, cellular forces and active
cell migration will sustain tumour growth or, conversely, might
prevent its expansion (figure 1).
Theoretical models have incorporated different degrees
of complexity to account for cell motility, either passively in
a diffusive manner or actively in response to external cues.
Mathematical differential expressions of some associated
cellular fluxes are briefly introduced below. A rather
exhaustive presentation of such models can be found in the
seminal books of J D Murray (2002, 2003). Random cell
motility is classically described by cellular diffusion, in which
the diffusive cell flux Jd is proportional to the local cell
gradient ∇m(r, t) at location r, i.e. Jd = −D∇m, where
the cellular diffusion coefficient D can be identified from the
analysis of cell tracking data (figures 2(a) and (c)) based on
random walk or persistent random walk models (Rosello et al
2004, Selmeczi et al 2005). Instead of considering a constant
diffusion coefficient, some models of tumour growth assumed
that this parameter is a function D(m) of the cell density (self-
reinforcement of cell motility) or depends locally on the ECM
properties. We especially reported in section 2 that cell traction
forces induce architectural reorganization and anisotropy of
the fibrous tumour microenvironment at the invasive front.
In agreement with experimental observations (Korff and
Augustin 1999), such influence of dynamic reshaping and
distortion of the microenvironment onto cellular diffusion can
be modelled by a strain-dependent diffusion tensor D(ε) here
ε is the Cauchy’s infinitesimal strain tensor (Cook 1995). The
cellular diffusive flux Jd then reads Jd = −∇ · [D(ε)m].
However, the most common directional cell motion invoked
when modelling tumour growth remains cell chemotaxis,
with cells moving up (more rarely down) gradients of a
diffusible molecule (growth factor, cytokine, etc) like VEGF
(Lee et al 2007, Selmeczi et al 2005). Cell chemotaxis
has notably a pivotal role in modelling the reaction of the
host immune system to tumour cells (Adam and Bellomo
1997). In the simplest models, the chemotactic flux of
cells Jg is assumed proportional to the cell density m(r, t)
12
P Tracqui
and to the chemical gradient signal ∇g(r, t), with Jg =
αm∇g. The coefficient α defines the cell’s sensitivity to
the diffusible chemoattractant. In a similar way, active cell
migration by haptotaxis, i.e. up to ligand-density gradients (or
adhesivity gradient) is modelled by considering an haptotactic
cell flux Jh . In a first approximation, this flux may be
considered proportional to both the local cell density and
the ECM density gradient ∇e(r, t). Assuming that the
sensitivity of the cell response to this gradient is monitored
by a coefficient h (haptotactic cell sensitivity), the associated
cellular flux is described by the phenomenological expression
Jh = hm∇e. Figure 9 illustrates that such modelling of
haptotactic cell migration, in combination with cellular forces,
can generate a self-enhanced accumulation of cells in specific
regions of a deformable medium. More generally, the above
expressions of cellular fluxes can be modified by considering
that chemotactic or haptotactic coefficients are no longer
constant parameters, but rather functions that depend on cell
density or on microenvironment composition and structure, as
illustrated above for cellular diffusion fluxes.
The above theoretical expression of cellular fluxes has
been considered in a large class of continuous differential
tumour models coupling reaction–diffusion equations with
chemotactic and possibly haptotactic cell migration processes.
Analysis of travelling-wave solutions exhibited by reaction–
migration models provides a theoretical basis for studying
in 1D domains tumour cells infiltration in the surrounding
tissue in different contexts (Marchant et al 2000, Sherratt and
Chaplain 2001). In 2D domains, consideration of cell motility
in addition to competition for nutrients among normal and
cancer cells was proven to give rise to a wide range of tumour
morphologies. Thus, Ferreira et al (2002) derived from a
nutrient-limited growth model a phase-like diagram exhibiting
the transition from compact and circular morphologies to
papillary and finger-like tumour patterns. Considering a
3D-geometry, Stein et al (2007) modelled the invasion of
collagen gels by glioblastoma tumour spheroids. The model
is designed to describe the dynamics in two distinct regions
of the dispersive spheroid: the central core contains cells
that proliferate rapidly but move slowly; the invasive rim
contains cells that proliferate slowly but are highly motile, with
radially directed motility. The model simulates the growth and
morphologies of both the core radius and the invasive radius,
which compare favourably to experiments on the patterns of
growth and dispersion conducted with wild type and epidermal
growth factor-transfected tumour cell lines. Interestingly,
the authors reported that in two-dimensional radial migration
assays, transfected cells exhibit a higher dispersion rate than
wild-type cells, in agreement with observations made in animal
models and with the reported implication of epidermal growth
factor activity in the increased aggressiveness of many cancers.
But surprisingly, the wild-type cells are more invasive than
the transfected ones in the 3D spheroid/collagen assay, which
points out the different influences of 2D and 3D extracellular
environments on cell phenotype and behaviour (Martins and
Kolega 2006). In this case, wild-type cells would have a better
ability to reshape the collagen matrix, forming preferential
migration tracks which could increase directed cell motility.
Rep. Prog. Phys. 72 (2009) 056701
Figure 9. Finite-element simulation of the 2D haptotactic migration of cells on an ECM exhibiting a bumpy variation of density.
(Left) Initial spatial profile of ECM density, slightly patterned by the superimposition of initial cell density (quasi-homogeneous distribution
with slight random fluctuations around a normalized unit density value). (Right) Haptotatic migration of cells over the ECM gradient,
combined with ECM deformation generated by cell traction forces, lead to the inhomogeneous steady-state shown on the right, with
accumulation of cells in the central part of the ECM.
Such differences plead for a more accurate modelling of
directional cell migration. Interestingly, a recent work of
Gerisch and Chaplain (2008) introduced a refined analysis of
cell–matrix adhesion by considering a sensing region for cells,
with cell velocity being enhanced in the direction where cells
can form the most bonds within the sensing region.
4. Control of tumour growth by mechanical factors:
continuum mechanics based models
Including mechanical factors as control parameters for tumour
growth is a conceptual change with respect to the above
modelling frameworks focusing on cell population dynamics
modulated by diffusible molecules. As outlined previously,
tumour growth is fundamentally a biomechanical process
with multiple facets, ranging from inherent balance between
proliferative pressure and resistive force of surrounding host
tissue to the far more complex processes of mechanical control
of cells dynamics and tumour vascularization. Then, tumours
should not be considered as passive mechanical structures, but
rather as dynamic active media where mechanical signals can
be transduced in different ways to determine the differential
response of normal and cancer cells, respectively (figure 4).
Biophysical models that are grounded within a mechanical
framework aim at answering questions about the influence
of strains and stresses on the evolution of the tumour
structural and mechanical properties itself, on the mechanical
interactions between the different types of cells inside and
outside the tumour, and between these cells and the tumour
microenvironment. This section highlights representative
modelling approaches proposed in this context, from the
earliest models introducing pressure and surface tension of the
tumour to more recent approaches dealing with multiphase
models and introducing diverse strain energy functions to
13
P Tracqui
describe constitutive stress/strain relationships within the
tumour.
4.1. Modelling the effect of pressure within the tumour and
associated cell velocity field
Following the work of Greenspan (1976), the role of cell–
cell adhesion as a physical constraint regulating tumour
compactness was addressed by several groups (Byrne and
Chaplain 1996b, Cristini et al 2003). In these models, cell–
cell adhesion was introduced by considering surface tension
at the tumour outer rim as a control parameter for tumour
reshaping during growth. Tumour growth itself is driven by
the intratumoral pressure gradient ∇p generated by localized
cell proliferation. Such gradients induce motion of tumour
cells, away from the proliferation regions (high pressure) and
towards regions where cell death predominates. However, cell
density does not appear explicitly as a model variable and a key
variable remains as usual the nutrient concentration delivered
by the microenvironment. Its evolution still obeys a reaction–
diffusion equation, assuming additionally that a steady-state
distribution has been reached, since the time scale of the
growing tumour is much slower than the diffusion time scale
of the nutrient. Considering the tumour as an incompressible
fluid, it is assumed that cell proliferation and death produce
a velocity field v linked to the nutrient concentration through
the continuity equation ∇ · v = (n), where (n) represents
the net balance between the nutrient-dependent proliferating
rate and the cell death rate (Cristini et al 2003, Graziano and
Preziosi 2007). A constitutive equation is needed to relate the
velocity field to the internal pressure p. The usual assumption
is to view the tumour as a porous medium through which
the tumour cells flow. Then Darcy’s law is used to relate
the velocity v to the pressure gradient, i.e. v = −γ ∇p,
where γ is a positive constant characterizing the hydrostatic,
Rep. Prog. Phys. 72 (2009) 056701
viscous-like properties of the tumour cells (Byrne and Preziosi
2003, Macklin and Lowengrub 2007). On the free boundary
of the tumour, which moves with velocity v , different types
of conditions expressing tumour compactness are provided
to close the model formulation. Thus, mean curvature and
surface tension accounting for cell–cell adhesion can be used
to prescribe boundary conditions on the pressure or on the
nutrient concentration (Byrne and Chaplain 1996b).
A main departure of such models from those assuming
radially symmetric tumour morphologies is their ability to deal
with the influence of tumour shape and local curvature, and
thus to cope more realistically with the diversity of irregular
tumour morphologies observed experimentally. Furthermore,
it becomes theoretically possible to characterize parameter
regions for which the radially symmetric solution becomes
unstable to asymmetric perturbations. This represents a
first basis for understanding the emergence of tumours with
branching morphologies, as detailed in the next section. On the
other hand, several limitations of these modelling approaches
can be pointed out. From a physical point of view, the validity
of Darcy’s law to describe cell motion within the tumour
may be questionable. Modelling tumours as freely expanding
bodies, keeping a relative compactness thanks to cell–cell
adhesion processes, is also a rather simplified view of the
real experimental situations, only justified in the case of free
floating growing MTSph. In other cases, a growing mass of
tumour cells contends with increasing mechanical resistance
from the surrounding host tissue, as experienced in vitro by
MTSph growing in gels (section 2.5). In vivo, this constrained
growth has fundamental implications regarding the formation
of a dense capsule of extracellular material around the tumour.
A crucial issue is then to know under which conditions such
physical barrier will be sufficient to contain tumour growth and
to prevent dissemination of cancer cells and metastasis.
In the light of preceding sections, analysis of tumour
encapsulation may invoke two kinds of processes: (i) a passive
mechanical response of the surrounding tissue, which is
simply compressed into a capsule by expansion of the tumour
(expansive growth hypothesis), (ii) active response of the host
tissue, with induction of ECM biosynthesis by normal cells.
These two processes have been analysed by Jackson and Byrne
(2002), with spatio-temporal evaluation of cells and ECM
(the model constituents or phase variables) simulated when
both processes are considered to act either independently or
synergistically. Stresses inside the tumour include for each
constituent intraphase stresses, depending on the constituent
volume fraction, and interphase stresses, including frictional
drag due to the relative motion between two constituents.
Cell proliferation and matrix production are inhibited at
high cell and ECM densities. The model is furthermore
considered as a closed system, with the sum of all constituents
being constant. One-dimensional model simulations indicate
that without ECM production, tumour growth generates thin
capsule, without growth suppression. With ECM production,
ECM accumulates within the tumour. With both processes
acting, tumour is growing and surrounded by a thin and dense
capsule. Closer to the biological observations, and in order
to analyse how the local host stroma is affected by degrading
14
P Tracqui
enzymes produced by the tumour itself, the model extension
proposed by Jackson and Byrne (2002) considered that tumour
cells produce protease at a constant rate when the pressure they
experience exceeds a threshold value. In addition, proteases
enhance motility of the cell by weakening cell–cell interactions
and adhesion. As expected, protease activity induces capsule
breakdown and favours tumour expansion.
In section 1, we highlighted that the active response of the
host tissue to tumour expansion is similar to a wound-healing
response, which, in addition to collagen and extracellular
proteins synthesis, can induce differentiation of fibroblasts
into migrating myofibroblasts able to contract the tissue. In
an attempt to analyse the role of tumour microenvironment
contraction on potential tumour encapsulation in the absence
of ECM biosynthesis by normal cells, Lubkin and Jackson
(2002) proposed a fluid mechanics approach to discriminate
between expansive growth and active contraction scenarios.
They consider both the host tissue and the tumour to be
composed of two interpenetrating material phases, a first
aqueous phase for the interstitial fluid and a second phase
consisting of normal and tumour cells embedded in fibrous
extracellular components. They assume that, on the time
scale of tumour growth, the second phase behaves as a Stokes
fluid. Still commonly assuming that tumour cell proliferation
is a function of available nutrients concentration, the model
introduces two kinds of stresses, the so-called solvation and
contractile stresses, respectively. The solvation stress is a
measure of the affinity of one phase for another, with positive
values if the cell–ECM phase is strongly hydrophilic, and
negative values if the cell–ECM phase is hydrophobic. The
contractile stress, modelling the potential contraction of the
host tissue, is proposed as an interaction term proportional to
the normal and tumour cells densities. Simulations of this
simple three-parameter model indicates that contractility of
the surrounding host tissue is not needed for the formation of
a capsule, but that this process may be enhanced if host tissue
contractility additionally takes place.
More recently, Guiot et al (2006) developed a rather
different approach by considering a mechanistic description of
tumour invasion in the framework of solid inclusion growing
in an elastic fracturable medium. The transition from a non-
invasive growth to invasive patterns takes place when a failure
stress, corresponding to a threshold for fracture propagation,
is reached. The so-generated cracks provide a mechanistic
analogy for the generation of infiltration paths for tumour
cells within the host tissue. The considered tumour growth
model implements the isotropic growth of a spherical and
linear elastic inclusion inside a linear elastic medium, the
growth rate being inhibited by the interfacial pressure. In the
standard case of a linear elastic microenvironment, the model
predicts very satisfactorily the time evolution of the volume
of non-invasive spheroids of human adenocarcinoma cells
growing in agarose gels with increasing stiffness (Helmlinger
et al 1997). When a fracturable elastic microenvironment
is considered, Guiot et al (2007a) also reported a very
good agreement between the model simulations and observed
rates of volume growth obtained experimentally with invasive
glioma spheroids (Gordon et al 2003).
Rep. Prog. Phys. 72 (2009) 056701
4.2. Modelling the role that stress responsiveness may play in
controlling tumour growth
As underlined previously, solid tumour expansion occurs
within a deformable host medium that mechanically constrains
the growth process. Quite naturally, biophysical modelling of
tumour growth is thus part of a general framework dealing
with development and influence of stress field on growing
biological tissues (Cowin 2004, Humphrey 2003, Lin and
Taber 1995, Volokh 2006), occurring during both physiological
and pathological morphogenetic processes.
In this context, a rather large diversity of biophysical
models have been considered, with tumour being considered
as one component of multiphasic medium exhibiting viscous
fluid-like or elastic solid-like properties. Experimental data
indicate that, according to their cellular origin (carcinoma,
sarcoma, glioma, etc), tumours develop with different
mechanical properties. Thus, Netti et al (2000) reported
significant differences in the stress relaxation behaviour of
tumours obtained from various human tumour cell lines
and grafted in mice. When submitted to a sequence of
compressive strain within a confining chamber, the studied
carcinomas exhibit a viscous behaviour, with almost complete
stress relaxation in between the imposed strains, which
is typical of macromolecular solutions. Conversely, the
tested glioblastoma and sarcoma display a viscoelastic solid
type of behaviour, with sequential increase of stress with
strain. Additionally, a significant nonlinear dependence of the
relaxation phase with strain was observed with glioblastoma
(Netti et al 2000), which closely resembled the one observed
with extracellular fibrin matrix submitted to similar loading
conditions (Benkherourou et al 2000). Such tumours thus
behave as structured soft tissues. Thus, these experiments
provide quite relevant quantitative data for modelling tumour
growth in a continuum mechanics framework.
In the continuum mechanics approach of Ambrosi and
Mollica (2004), the tumour spheroid was considered as
compressible elastic solid whose increase in mass with time is
modulated by nutrients levels. A known modelling difficulty
is then to describe simultaneously the change in mass and
in stresses occurring during growth. The usual approach
(Rodriguez et al 1994) consists of describing the evolution
of the mechanical properties of the growing tumour as a
composition of two separate transformations. The first one
considers the change in mass due to growth from the initial
configuration up to an intermediate and virtual, so-called
natural configuration, and is described by a growth tensor G
(figure 10). The second transformation describes the stress
response of the tumour up to the current configuration. For
simplicity, the newly synthesized material is introduced with
the same stress state as the existing material. A constitutive
relationship has to be provided to model the stress response
from the natural configuration.
By analogy with polymeric foams, Ambrosi and Mollica
(2002) considered the tumour as an isotropic nonlinear
elastic and compressible material characterized by a strain
energy function, from which the Cauchy stress tensor T in
the tumour can be derived. This constitutive equation is
15
P Tracqui
F=Fe G
T=0
T=Tr
Initial
configuration
G Current
Fe configuration
T=0
Virtual
configuration
Figure 10. Three states considered in finite growth of an elastic
tumour: the growth tensor G maps the original stress-free (Cauchy
stress T = 0) initial configuration into an intermediate virtual
configuration (unconstrained growth). Since this stress-free
configuration may differ from the observed current configuration, an
elastic transformation Fe describes the stress response of the tumour
from the natural to the current configuration, tumour mass being
preserved. The overall deformation gradient F indicates how the
growing tumour is deforming locally when going from the initial to
the current configuration. The transformation Fe gives rise to a
residual stress Tr in the current configuration.
supplemented by the usual mass balance equations for the
tumour mass and the nutrient, and by the balance of linear
momentum, which sums up to divT = 0 when inertial terms
and body forces are neglected. Key points in the model
formulation are the expression of the growth rate appearing
in the tumour mass balance equation and the associated
relationship between the growth rate and the growth tensor
G. Isotropic growth of a sphere can be modelled with a
growth tensor of the form G = g(t)I , where the identity
tensor is multiplied by an explicit function of time g(t). More
interestingly, and with reference to Helmlinger et al (1997)
experiments reporting inhibition of tumour spheroid growth
with increasing stiffness of the surrounding agarose matrix
(section 2.5), Ambrosi and Mollica (2004) considered that the
spherical growth tensor G depends on nutrient concentration
and on intratumoral stress. Conservation of mass leads
to a growth model where the function g(t) is given by a
differential equation gt = h(trace P , n)g, where n is the
nutrient concentration and P the first Piola–Kirchoff stress
tensor, introduced to solve the mechanical problem in the
reference configuration. The dependence on stress is modelled
through a decreasing exponential function, such that stress
always inhibits growth. Additional constitutive equations
are prescribed for the mechanical behaviour of the agarose
gel. The mechanical response of this matrix, considered as a
poroelastic material, is derived from a strain energy function
depending both on the third invariant and, exponentially, on the
first invariant of the right Cauchy–Green deformation tensor.
Corresponding simulations of tumour spheroid growth
showed that in the absence of external loads (floating spheroid),
the change in growth rate was essentially due to the reduced
availability of nutrient that occurs when the diameter of
the spheroid increases. A significant nutrient concentration
Rep. Prog. Phys. 72 (2009) 056701
only subsists in a thin layer below the spheroid surface,
which may be interpreted as the outer rim of proliferative
cells observed experimentally with MTSph assays. When
the spheroid is growing within the surrounding poroelastic
medium, its size increases linearly from the very beginning,
until it tends to plateau. In this case, growth is essentially
inhibited by extracellular stresses. Let us remark that an
additional complexity in the description of MTSph growth
is the intrinsic generation of residual stresses, a rather
common feature observed with different physio-pathological
processes (Alford et al 2008, Ohayon et al 2007), but which
may significantly influence the tensional homeostasis of the
growing and remodelled tissue (Alford et al 2008, Ambrosi
et al 2008, Lin and Taber 1995).
While the tumour spheroid model of Ambrosi and Mollica
(2004) is one–component, the growth of MTSph embedded
in a compressible elastic matrix was alternatively modelled
by considering a growing structure composed of solid and
fluid phases in interaction. Following Landman and Please
(2001), Chen et al (2001) considered the tumour spheroid as a
mixture of cells and extracellular fluid characterized by their
volume fraction and velocity, related through force balance
equations involving hydrodynamic drag force, hydrostatic
pressure and intercellular forces. Growth takes place through
cell proliferation and mass exchange between phases, with
dead cells assumed to transform into fluid phase. The
spherically symmetric spheroid then consists of a necrotic
region, resulting as usual from diffusion-limited nutrient
supply, and of a compact but active outer proliferative rim.
The spheroid’s expansion is restrained by the surrounding
matrix, still modelled as an infinite hyperelastic isotropic
and compressible medium characterized by a strain energy
function, which is a slight modification of the strain energy
function proposed by Helmlinger et al (1997): it depends on
the first and third strain invariants and is similar to the one
used in (Ambrosi and Mollica 2004). As intuitively expected,
the ECM inhibits the tumour spheroid’s growth, but the 1D
simulations of the model allow a quantitative tracking of the
sigmoı̈dal pattern exhibited by the evolution with time of the
spheroid matrix interface, together with the associated rapidly
varying displacement of the necrotic core.
The biomechanical models presented above can be
certainly extended by taking into account additional properties
characterizing tumours mechanical response to extracellular
strains and stresses. One may notably consider more precisely
how the mechanical confinement surrounding the tumour
mass feeds back onto tumour remodelling through variation
of mechanical moduli characterizing the tumour properties.
Such analyses can be conducted in a multiphase modelling
framework (Preziosi and Tosin 2009) or transposed from
studies developed in other fields, e.g. when analysing stress
adaptation of bone or heart tissue: here tissue stiffness is a
function of microstructures which evolve in time depending
on the external loading conditions (Cowin 2004, Taber and
Chabert 2002, Tozern and Skalak 1988). Aside from pure
tumour mass expansion, the mechanical effects of cell traction
forces onto the alignment and anisotropy of the fibrous tumour
microenvironment also appear as critical active controls of
tumour expansion, as developed in the following section.
16
P Tracqui
5. Morphological instabilities at the tumour/stroma
interface and tumour invasion
Tumour invasion involves a variety of processes that
ultimately lead to cell detachment from the primary tumour
and infiltration into adjacent tissue and before possible
dissemination to secondary metastatic sites with the aid of
the bloodstream system (figure 5). We already approached
in the preceding section a challenge in modelling tumour
growth, namely understanding how the propensity of limiting
tumour expansion by compression and fibrosis of the host
tissue may be redirected to the capacity of triggering invasive
progression and further metastasis capabilities of tumour cells
(figure 1). Considering the invasive branching morphology
commonly observed in vitro with growing MTSph, a promising
view consists of approaching this pattern formation process
as the result of the amplification of growth instabilities at
the tumour/host tissue interface. Thus, the landmark of
tumour aggressiveness would not be solely a localized tumour
cell proliferation, modifying the tumour’s surface-to-volume
ratio, but a global switch or bifurcation between ‘smooth
margin’ and ‘fingering protrusions’ surface patterns. In
this framework, biophysical models are expected to shed
light on the selection of invasive patterns by the tumour
microenvironment characteristics, i.e. nutrient availability,
growth factors activity and ECM properties.
We outlined in the preceding section that 1D models
focusing on surface tension (Greenspan 1976) or cell–cell
adhesion (Byrne and Chaplain 1996b) as the biophysical
processes maintaining the tumour’s compactness exhibit
parameter regions for which the radially symmetric solution
becomes unstable to asymmetric perturbations. Both
processes yielded similar qualitative predictions: regular
tumour growth occurs if surface tension or cell–cell adhesion is
strong enough, but irregular, infiltrative structures resembling
irregular outer boundaries of tumours emerge if these
parameters are smaller.
Alternatively, the emergence of non-equilibrium steady
states as dissipative structures is well documented in reaction–
diffusion models or in reaction–diffusion–migration models
of pattern formation (Murray 2002, Nicolis and Prigogine
1977). In the original model proposed by Turing in 1952
(Turing 1990), the so-called Turing’s instability arises from the
coupling between nonlinear reaction kinetics and the relative
diffusion of two competing (inhibitor and activator) chemical
species. In reaction–diffusion–migration models, simplest
scenarios involve cells that produce their own chemoattractant.
Then a small perturbation in an initially homogeneous
distribution of cells gives rise to a gradient of chemoattractant
that enhances the heterogeneity of cell density. On the
other hand, cellular diffusion, as well as chemoattractant
diffusion, tends to eliminate such heterogeneities. Clearly,
spatial patterns will be generated if the aggregative process
is greater than the dispersive effects. A similar self-enhance
aggregative effect can be driven by adhesivity gradients
(haptotaxis), exhibited by inhomogeneous ECM densities
(figure 9) (Murray 2002, Tracqui 2006). This aggregative
effect is still counterbalanced by cellular diffusion, but since
Rep. Prog. Phys. 72 (2009) 056701
Figure 11. Illustrative mixture of spots and stripes obtained at the
surface of a sphere as an asymptotic spatial pattern emerging beyond
a Turing’s instability. This conceptual background for tumour
morphological instabilities and branching is illustrated here from the
finite-element simulation of a simple two-variable
reaction–diffusion model including a nonlinear autocatalytic
reaction term. Then, colour scale would correspond to concentration
variations of a potential growth-promoting chemical diffusing on the
surface of a spherical tumour.
ligands are bounded to the ECM and do not diffuse, it is only
the ECM viscoelastic properties which can act as a stabilizing
factor in this mechanical instability scenario (figure 9).
In view of these many possibly different sources of
instabilities, several theoretical and computational approaches
have been developed to sustain the view that nonlinear and
self-organizing processes may drive tumour morphological
instabilities, with clusters of cells organized and distributed
non-homogeneously at the tumour edge as a response to a
biochemical pre-patterning (figure 11). Thus, in the spirit
of Turing’s pre-patterning theory, Chaplain et al (1995,
2001) studied the evolution of the concentrations of growth-
inhibiting and growth-promoting diffusive chemicals, which
are produced by the tumour cells on the surface of a growing
spherical tumour. This surface is defined as a thin layer of
proliferative cells, within which the production of the chemical
factors is restricted. By assuming that the two chemical
species u and v interact nonlinearly through cross activator–
inhibitor mechanisms, it is known theoretically that linearly
stable steady-state values of u and v in the absence of diffusion
may bifurcate toward unstable solutions when diffusion comes
into play. The growing tumour is modelled as a sphere of radius
R(t), and the reaction–diffusion model is defined as a moving-
boundary problem that reads
1 1
ut = u+f (u, v), vt = dv+g(u, v),
[R(t)]2 [R(t)]2
(3)
where f (u, v) and g(u, v) are polynomial functions of u and v
incorporating an order-2 autocatalytic formation of chemical
17
P Tracqui
u as the driving nonlinearity of the reaction terms. Given
a set of parameter values, which satisfy the conditions for
Turing’s instability, different spatially heterogeneous steady-
state distributions of the two chemicals u and v can be obtained
over the tumour surface. As in rectangular geometries, higher
order unstable modes become excited as tumour grows and
tumour surface becomes larger. This leads to a series of
spatial pre-patterns of growth-inhibitor and growth-activator
concentrations (figure 11) that may be associated with the
progression and asymmetries of tumour morphologies.
Another conceptual approach, based on the principle
of ‘least resistance, most permission and highest attraction’
has been proposed by Deisboeck and coworkers (Deisboeck
et al 2001, Habib et al 2003). In this scenario, cells
follow each other because of reduced mechanical resistance
of the ECM, enhanced haptotactic gradient, and increased
chemical attraction, all as part of a self-organizing adaptive
system. Following these guidelines, Sander and Deisboeck
(2002) analysed and modelled the emergence of branching
pattern around the central core of muticellular glioma
spheroids, generated from a mutant cell line in which the gene
coding for the epidermal growth factor receptor (EGFR) was
amplified. The spheroid was cultured between two layers of
ECM. In addition to random cell motion, the model implements
two types of directional migration for tumour cells, called
heterotype and homotype chemotaxis, respectively. The first
one refers to the usual chemoattraction of cells in response
to gradient of a diffusible nutrient supplied by the tumour
environment. The second one refers to paracrine stimulation,
according to which tumour cells produce soluble growth
factors during their migration. Neighbouring cells are attracted
by the so-created growth factor gradient, with motile cells
following each other because of this chemotactic process. The
2D variation of the concentration of motile tumour cells is thus
given by an extended diffusion–chemotaxis transport equation,
which reads
ct = ∇ · (Dc ∇c) − ∇ · (c∇χ (n)n) − ∇ · (cη∇h). (4)
Here c(r, t) is the concentration of cells, and n(r, t) is the
nutrient concentration. The coefficient χ (n) is a nonlinear
function of n, which defines the sensitivity of cells to
heterotype chemotaxis. The last term is the homotype
chemotaxis, mediated by the concentration h(r, t) of the
paracrine mediator produced by motile cells. Proliferation
of cells is assumed negligible in the invasive region, but cell
generation counterbalancing cell shedding is implemented as
a boundary condition on the spheroid surface. Diffusion–
reaction equations are additionally provided for nutrient
and homotype factor concentration, assuming a steady-state
distribution for the nutrient due to its much faster diffusion.
A linear stability analysis of a schematic 2D portion of the
region surrounding the spheroid shows that the steady-state
concentration of the growth factor is unstable. Then, a branch
of invasive tumour cells may extend from the spheroid surface
as a result of this intrinsic instability due to the combination
of heterotype and homotype chemotaxis.
One should notice that formally, the last term in
equation (4) looks like a haptotactic flux, except that molecules
Rep. Prog. Phys. 72 (2009) 056701
of h can diffuse. Tumour cell haptotaxis at the tumour
surface was proposed in Tracqui (1995) as a possible driving
mechanism for surface pre-patterning of invasive tumours,
in conjunction with deformation of the tumour–host tissue
interface by cellular forces. Considering the few layers of
active cells at the tumour surface, the model assumption is
that haptotactic cell movement initiates local cell aggregation,
with an associated local increase in cellular forces amplitude
and tumour stroma deformations which positively feedback
onto cell motion. If haptotaxis is sufficiently large compared
to cellular diffusion, this autocatalytic process can be strong
enough to overcome the mechanical resistance of the tumour
stroma, thus initiating and stabilizing the formation of pre-
invasive spurs of cancer cells at the tumour surface.
Quite recently, Macklin and Lowengrub (2007) proposed
a more refined analysis of the crucial role played by the
tumour microenvironment. In particular, they show that, in
a nutrient-poor microenvironment, the only viable invasion
strategy for the tumour is to break into small fragments, which
penetrate the surrounding tissue. Conversely, under nutrient-
rich conditions, the tumour behaviour depends mostly on the
biomechanical property of the medium: a soft, mechanically
responsive medium enhances ‘smooth margin’ invasion while,
in the opposite situation, long invasive fingers are produced,
whose characteristics largely depend on cell-to-cell adhesion.
Alternatively, the instability scenario explored by Khain and
Sander (2006) is based on a dynamical switch of the tumour
cells phenotype. They compared the invasive patterns obtained
with mutant MTSph, where mutant cells were also obtained by
amplification of the EGFR gene, to the spherically symmetric
pattern observed with MTSph composed of wild-type, i.e.
non mutated, cells. The model assumptions are that cells
have high motility but slow proliferation in the spheroid
neighbourhood, whereas rapid proliferation occurs in the inner
spheroid region. Thus, Khain and Sander (2006) introduced an
autocatalytic-type of density-dependent proliferation, leading
to a two-variable nonlinear reaction–diffusion model similar
to the one used in Chaplain et al (2001). As known from
2D-analysis of such nonlinear systems, plane wave solutions
can become transversally unstable if the ratio of the nutrient
and cell diffusion coefficients exceeds a threshold value.
Bifurcation of spherical spheroid shape toward developing
branches of tumour cells is then associated with this instability
phenomenon. When considering additionally that ECM
reorganization due to cell/ECM interaction enhances tumour
cell migration in the radial direction, the simulated developing
branches appear much thinner than in the isotropic diffusion
case. Thus, this simple reaction–diffusion model captures
different in vitro morphologies of mutant and wild-type
spheroids. It deserves nevertheless further experimental
validation of the associated model predictions, i.e. that
wild-type cells have a larger diffusion coefficient but a
smaller proliferation rate than mutant cells in the spheroid
microenvironment.
A diffusional instability scenario that takes place during
tumour growth was also considered by Cristini and co-
workers (Cristini et al 2003, Frieboes et al 2006). In their
model, the local specific mass growth rate is given as the
18
P Tracqui
divergence of the tumour cell velocity field v , related to the
pressure in the tissue by Darcy’s law. Tumour cells and
ECM are treated as comprising porous media with hydraulic
conductivity, while cell adhesion forces are modelled using
an equivalent surface tension at the tumour/tissue interface.
Diffusing nutrient and oxygen are supplied by neovasculature
and captured by tumour cells. The onset of morphologic
instability and tumour fragmentation is based on the 3D linear
stability analysis of the reaction–diffusion model, where mass
growth and shape evolution of the tumour depend on four
microphysical variables: cell mitosis rate, diffusion length
L, ratio A of cell death rate over proliferation rate and
ratio Gm of tumour cell mitosis rate over relaxation rate,
where the relaxation rate is associated with cell–cell adhesive
forces and cell motility coefficient introduced in Darcy’s law.
The bifurcation diagram associated with the model defines
transitions between tumour morphologies as a function of A
and Gm. For small Gm, the spherical shape is stable and
the balance between cell proliferation and death regulates the
tumour size. When too weak cell adhesion forces lead to larger
values of Gm, shape instability may develop. These theoretical
results are in agreement with observations of MTSph invasive
morphologies and support a diffusional instability scenario
where emergence of irregular tumour morphologies is driven
by the differential cell proliferation induced by gradients of
nutrients and growth factors in the tumour microenvironment
(Frieboes et al 2006). However, extrapolation to in vivo
situations is not straightforward, since tumours in vivo possess
varying degrees of vascularization and exhibit a complex
interface with normal cells and the ECM. This provides a clear
incitation to develop more complex models including explicitly
tumour vascularization.
6. Tumour–host tissue interactions and tumour
vascularization
The switch from the slow and relatively harmless avascular
tumour growth phase modelled above to a rapid vascular
growth phase occurs because of tumour vascularization by
angiogenesis (figure 5). Therefore, biophysical modelling
of tumour growth should be ideally able to couple tumour
mass growth to the well differentiated stages of vascularization
presented in section 2 and which build up tumour vasculature
from the derivation of existing host tissue vessels (figure 5).
Thus, several approaches, based on continuous (Holmes
and Sleeman 2000, Levine et al 2001) or discrete formalisms
(Anderson and Chaplain 1998, Bauer et al 2007, Chaplain et al
2006, Merks et al 2006), tried to provide a global description of
successive events leading to tumour vascularization, from loss
of interconnection between the ECs and release of proteolytic
enzymes in response to diffusible angiogenic factors secreted
by tumour cells, to capillary sprouts, tips formation and
interconnection into closed loops (anastomosis), until new
capillaries penetrate the tumour (figure 5). In models of
capillary network morphogenesis and adaptation to blood
circulation, tumour is usually only viewed as a source of
angiogenic factors. We will concentrate here on models
explicitly coupling the influence of tumour vasculature on
Rep. Prog. Phys. 72 (2009) 056701
tumour growth, which have been significantly improved in
recent years.
Cristini et al (2005) related tumour shape and
microvascular density to provide evidence that spatially
homogeneous oxygen and nutrient act by suppressing
morphological instability. Their model incorporates an
angiogenesis component as reinforced random walk of new
blood vessels drawn into the tumour, and blood vessel
will collapse when the blood-to-tumour pressure difference
approaches zero. The two-dimensional simulations of the
model were interpreted and summarized into a two-parameter
phase diagram summarizing possible tumour morphologies:
one parameter is associated with cell adhesion forces, while
the other one quantifies microvascular density. Model
computations show that even after angiogenesis, the tumour
maintains a compact morphology because of high cell
adhesion. However, with a smaller value of the cell adhesion
parameter, a morphologic instability of the tumour mass arises
as the tumour develops. Such computational results can be
discussed in the light of the tortuousness and loss of vascular
hierarchy observed in solid tumours (Fukumura and Jain 2007).
Indeed, ideal angiogenesis will lead to a compact and spherical
tumour mass confined in a limited volume of tissue. In this
case, further tumour growth requires additional angiogenesis
to increase nutrient access. In contrast, disordered vascularity
leads to instability and tumour branching since the same
tumour mass does not critically require angiogenesis for further
growth and thus might extend across a larger region within the
host tissue.
Recently, the same group (Frieboes et al 2007)
proposed an extended 3D model of tumour growth
still focusing on the influence on tumour angiogenesis
and associated nutrients diffusion gradients onto tumour
invasion. Their hybrid continuum–discrete model of tumour
angiogenesis is based on convection–reaction–diffusion
equations governing the dynamics of tumour cells, vital
nutrients and cytokines. Mechanical interactions of the
ECM with evolving neovasculature are modelled by adhesion
energy and momentum equations based on Darcy’s law and
equilibrium among pressure and elastic forces. The model
predicts gross morphological features of growing tumours,
with regions of viable cells, necrotic areas and fractal-like
neovasculature. Interestingly, the model accounts for the
tumour engulfing and compressing the vessels, thus increasing
necrosis. This result highlights the parabaric feedback
loop that affects the angiogenic pathway described above:
growing around capillaries that are usually immature, the
tumour compresses them and might cause their collapse. But
in turn, this increases hypoxia of the surrounding tissue,
which stimulates secretion of angiogenic factors and thus the
formation of new capillaries.
Additionally, adaptation of the vessel network to
circulating hydrodynamic blood flow, with local vessel
dilation, is a regulation process increasingly considered both
during host tissue vascularization and during tumour growth
(Chaplain et al 2006, Stephanou et al 2005). Thus, Lee
et al (2006) considered the tumour–vessel system as a
dynamically evolving graph interacting with tumour growth.
19
P Tracqui
Interactions between tumour and vasculature take place via
two concentration fields: the oxygen supplied by the vessel
network and growth factors secreted by tumour cells. Each
edge of the generated graph represents a tubular vessel carrying
a hydrodynamic blood flow that exerts a shear force upon the
vessel walls. Vessel dilatation takes place in regions of large
growth factor concentrations. Conversely, weakly perfused
vessels can collapse when the hydrodynamic shear force is too
low. As a result, geometry of the tumour vasculature emerges
in the model as a combination of dense and void regions that
might possess fractal properties.
By coupling tumour growth to tumour angiogenesis,
such models have obviously strong medical implications for
understanding potential failure of anticancer chemotherapy
due to insufficient drug delivery to the tumour. They also
reinforced the multiscale properties of tumour growth and
the benefit that could be expected by developing models
addressing the specificity of biological and biophysical
processes expressed at each level. The next section
briefly reviews some representative models developed in this
perspective.
7. Multiscale modelling of tumour growth
A major issue in cancer research, as in other domains
of physiopathology, is to handle the enormous amount
of data produced at the molecular, cellular and tissular
levels, both in normal and pathological contexts. Thus,
multiscale biophysical models of tumour growth are especially
important as formal and explicative tools for relating molecular
processes and gene expression to global tumour morphologies
(figure 12).
In the Jiang et al (2005) model, avascular tumour
growth spans three distinct scales. At the cellular level,
a lattice Monte Carlo model describes cellular dynamics
(proliferation, adhesion and viability). At the subcellular
level, a Boolean network regulates the expression of
proteins that control the cell cycle. At the extracellular
level, reaction–diffusion equations describe the chemical
dynamics (nutrient, waste, growth promoter and inhibitor
concentrations). This model produces an avascular tumour
that quantitatively mimics experimental measurements in
multicellular spheroids. Based on the simulations, the model
predicts (i) the microenvironmental conditions required for
tumour cell survival and (ii) the pattern of spheroid growth
curves under different external nutrient supply conditions.
Later, Anderson et al (2006) analysed the evolution of
tumour morphology driven by selective pressure from the
microenvironment, thanks to a multiscale model of cancer
invasion where cell mutations and interaction with different
types of microenvironment, referred to as homogeneous,
‘bumpy’ and ‘grainy’, are considered. Sharing some analogies
with the granular solid considered by Guiot et al (2007a),
bumpy and grainy media represent density variations of the
microenvironment taken at a cellular scale or at a macroscale,
respectively. The approach of Anderson et al (2006) is
based on a hybrid discrete–continuum model in which cells
are treated as discrete entities, while microenvironmental
Rep. Prog. Phys. 72 (2009) 056701
Figure 12. Illustration of a multiscale approach of tumour growth
based on a mixed continuous–discrete (hybrid) model.
Spatio-temporal variations of continuous extracellular gradient are
combined to the probabilities P1–P4 of a single tumour cell located
at the lattice coordinates (i, j ) to move in one of four directions.
Each probability may depend on the steady-state value of a protein
or gene network describing the controlled expression of proteins
regulating tumour cells adhesion and migration.
factors are modelled as concentrations. Tumour morphologies
emerge as a consequence of the cell collective behaviour,
controlled by microenvironmental factors that influence cell
mutations. In the model, these mutations occur at the time
of cell division, either linearly through four increasingly
aggressive phenotypes, or randomly through a mutation
jumping process in which cells are allowed to acquire one
of one hundred predefined phenotypes. A striking result
of the random mutation scheme is that, even though each
phenotype is equally likely to be chosen, only a few, aggressive
phenotypes, are actually selected. Thus, the simulated
growing tumour evolves toward increased aggressiveness,
characterized by tumour cells exhibiting the shortest lag
time to proliferation, the highest production of ECM-
degrading proteases and the highest haptotactic migration
rate. Interestingly, these phenotypic differences correlate
with emergent tumour morphologies that are affected by the
microenvironment structural properties: under homogeneous
conditions, tumour morphology is non-invasive with smooth
margins, as opposed to the invasive morphology with fingering
margins simulated when tumour develops in the heterogeneous
matrices. Thus, heterogeneous tumour microenvironment
seems to drive cancer cells mutations toward phenotypes
characterized by higher motility and higher haptotactic
sensitivity.
8. Tumour growth models incorporating anticancer
treatments: model-driven evaluation and
optimization of therapeutic designs
A major application of tumour growth model development is
the expected ability to derive from the dynamics and behaviour
20
P Tracqui
of virtual tumours some practical guidelines based on a
model-driven optimisation of the tumour response to different
types of anticancer treatments, including chemotherapy,
radiotherapy, surgery, etc, while diminishing the unavoidable
side effects of such treatments. Successes and limitations
of theoretical models explicitly including and comparing
anticancer therapies are exemplified here, with special
consideration of brain tumours.
Briefly, the development of brain tumours, after
diagnosis, is routinely recorded by different and accurate
medical imaging techniques (computerized tomography (CT),
magnetic resonance imaging (MRI), positron emission
tomography (PET) (Spence et al 2008), etc). Typically,
serial scans can lead to a quantification of the expensive or
treatment-induced recessive dynamics of the brain tumour
for each patient. Different models have been developed to
analyse the different stages of a brain tumour growth (Murray
2003 chapter 11, Harpold et al 2007). Their biological
and physical bases do not differ significantly from those
considered in the models reviewed previously. All models
consider cell proliferation and cell infiltration as fundamental
processes, possibly together with the compression of the
brain tissue surrounding the tumour. In addition, geometrical
constraints to brain tumour growth can be accurately taken into
account in the modelling approaches, thanks to the different
modality of image acquisition than can be used. These
constraints encompass anatomical boundaries, either fixed as
the skull, or displaced and deformed by tumour expansion,
like the brain ventricles. Such geometric confinements
significantly affect the size and shape of the tumour, notably
if the tumour initiates close to these boundaries. These
properties were highlighted with continuous tumour growth
models, in which tumour growth-permitting regions versus
tumour growth-prohibiting regions are defined through space-
dependent diffusion coefficients (Tracqui et al 1995, Tracqui
and Mendjeli 1999). Alternatively, similar conclusions can
be drawn using the cellular automaton approach, as recently
reported by Gevertz et al (2008), with automaton cells defined
from the Voronoi tessellation of the considered tissue. A
benefit of this latter approach is to consider intercellular
mechanical stress between cells as a control parameter for the
tumour expansion, thus incorporating the mechanical feedback
exerted by the environmental confinement on the tumour
growth.
Brain tissue heterogeneity also provides anatomical
tracks favouring tumour cell dissemination, leading to model
formulations that explicitly consider the larger cell motility
in white matter compared with grey matter (Swanson et al
2000), as well as differences in the mechanical resistance
between both tissues. In the 3D model developed by Frieboes
et al (2007), brain structural properties play a critical role in
the formation of rapidly developing tumour buds. Tumour
buds invading the white matter encounter less mechanical
resistance, leading to the formation of relatively slender
invasive fingers, compared with the blunt bulbs invading the
grey matter. The tumour buds rapidly elongate and grow
along the folds of the cortex where the tumour cells are
highly mobile. Interestingly, the overall tumour shape is also
Rep. Prog. Phys. 72 (2009) 056701 P Tracqui

(a) (b)

Figure 13. Example of a case study where clinical data were obtained from serial MRI scans taken during the terminal year of a patient with
a glioblastoma and who received radiotherapy treatment. (a) Patient’s MRI showing the brain tumour location, close to the neighbouring
ventricles, (b) simulated 3D constrained growth of the treated tumour (b insert) considering density dependent diffusion model with logistic
(dotted line) or Gompertzian (solid line) nonlinear proliferation rates (Tracqui and Mendjeli 1999). Data points correspond to the tumour
volume measured from patient scans at the time of diagnosis and in course of therapy (black rectangles along the time axis). In each case,
the left part of the curves, i.e. from the time of diagnosis back to the appearance of the first cancer cell, has been obtained by reversing time
in the simulations, thus giving access to a predictive description of the silent (non-observable) phase of tumour growth.

determined by the distribution of vessels within the tumour,
with the considered multiscale model providing simulations
of growing glioma and neovascular morphologies that reflect
histopathology data.
The severity and poor prognosis of brain cancers is
associated with a still limited understanding of the large
diversity of the kinetics and growth patterns exhibited by
the different types of brain tumours. Thus gliomas, highly
invasive primary brain tumours accounting for almost 50%
of all brain tumours, are known to grow at a wide range of
possible proliferation and diffusion rates (Alvord 1995). This
stimulates modelling approaches trying to integrate in a single
description tumour growth before and after diagnosis and
therapy. Indeed, once detected, tumours are not left untreated
in humans, and the imaged tumour growth results from the
combined effect of initial endogenous tumour development
and its response to therapy. The ability to distinguish between
these two components of tumour growth is clearly a modelling
challenge, especially when considering that therapy is not
only translated into an additional cell death rate, but also
modifies intrinsic tumour growth, for example by inducing
mutations and emergence of new, usually therapy resistant,
mutated subpopulations of cancer cells (Tracqui et al 1995).
Since the time of onset of the brain tumour is not known
and thus the duration of the pre-diagnosis tumour growth
is not clinically measurable, model simulation starts from a
more or less well-defined initial tumour morphology, which
determines a subsequent growth pattern against which the
clinical data extracted from the analysis of serial scans may be
compared. However, if assuming as a first simplification that
the endogenous tumour dynamical features remain unchanged
after therapy, modelling tumour growth pattern after diagnosis
interestingly provides a basis for reconstructing pre-diagnosis
tumour morphology and associated tumour growth history.
Indeed, by reversing time in model simulation starting from
diagnosis, it becomes possible to track back the temporal
pattern of the glioblastoma development up to the appearance
of the first cancer cell (figure 13). Such estimations may be
used as a quantitative basis to define a range of patient ages
where early detection of primary tumour can be planed.
An additional benefit of brain tumour growth models is
their ability to evaluate virtually, but on quantitative bases,
the comparative effects of different and independent or more
often combined therapies, in a patient-dedicated way. Special
attention has been paid to post-surgical behaviour of gliomas
after resection through in silico experiments conducted with
virtual gliomas (Swanson et al 2003, 2008a, Woodward
et al 1996). Even if the resection seems optimal from a
surgical point of view, i.e. presumably removes almost all
the cancer cells, it is indeed well known that the small
number of cells left behind could be sufficient for inducing
a complete re-appearance of the tumour some months later.
The recurrence of the tumour can start from the tumour cells
left behind, and this minority of surviving tumour cells is
aggressive enough to drastically reduce patient life expectancy
(figure 14) (Swanson et al 2008a, Tracqui and Mendjeli 1999).
Similarly, Alvord and colleagues investigated the response
to radiation therapy, focusing on the effect of radiation dose
fractionation on gliomas growth (Rockne et al 2009, Swanson
et al 2008b). Thus, simple biophysical models, specified by a
very limited number of parameters, allow a coherent qualitative
and quantitative description of the wide spectrum of gliomas
growth and invasiveness patterns. Moreover, their ability to
predict quite successfully many aspects of real gliomas growth

21
Rep. Prog. Phys. 72 (2009) 056701
Figure 14. Predictive power of models simulating virtual tumour
growth under different therapies. The grey line corresponds to the
tumour evolution already reported in figure 13 when a radiation
therapy was used. The hatched line corresponds to the re-growth of
the tumour after a simulated surgical resection, which had left some
cancer cells behind. With such aggressive brain tumours,
proliferation and diffusion of these cells from the border of the
resected brain region (insert) lead rather rapidly to a lethal tumour
volume (horizontal dashed line), in agreement with clinical data.
Estimation of the expected life expectancy of the patient can be read
directly from the intersection of the simulated growth curve with the
lethal tumour volume.
favours their systematic use as tumour grading tools enabling
a dynamic identification of biologically distinct groups of
gliomas.
Biophysical models of brain tumours can also support
investigation of rather different therapeutic strategies. Cristini
et al (2005) analysed targeting of brain tumours by
anti-angiogenic therapy. Whereas it seems clear that
a direct action against tumour growth was to reduce
angiogenesis, the modelling approach they developed suggests
that such intuitive consideration has to be modulated by
more subtle considerations. Indeed, a properly working
tumour microvasculature can help maintain compact non-
infiltrating tumour morphologies by minimizing oxygen and
nutrient gradients. In contrast, anti-angiogenic therapy, by
increasing microenvironmental heterogeneity, may promote
morphological instability, leading to invasive patterns even
under conditions in which the overall tumour mass shrinks.
The model of Cristini et al (2005) suggests that anti-
angiogenesis solely with an end to eradicating a tumour
may not be the most prudent course. These conclusions
modify the conventional view on angiogenesis leading to
invasion and metastasis: a well-vascularized tumour may be
morphologically stable due to homogeneous distribution of
oxygen and nutrient supply, while local hypoxia or lack of
cell nutrients due to inefficient neovasculature leads to regions
of heterogeneous cell proliferation and migration, which, in
turn, promotes an invasive tumour morphology (Macklin and
Lowengrub 2007). The model proposed by Welter et al (2008),
implementing a dynamically evolving vessels network within a
22
P Tracqui
growing tumour, also pointed out unexpected results: collapse
and vessel regression did not decrease drug delivery to the
tumours in their computations, a result implying that this
mechanism would not be the major reason for diminishing
drug delivery in real tumours.
Taken together, these different results highlight the ability
of models in providing predictive guidelines for therapeutic
strategies that, solely or in combination, and in the light of
accurate estimations of expectable survival times, are geared
toward optimization of patient-dedicated therapy.
9. Conclusions and perspectives
It is now becoming evident that modelling tumour growth
is much more complex than it was thought not so long
ago. Earliest approaches tended to consider that models,
developed with regard to the multicellular spheroid paradigm,
already address some key issues on cell proliferation and
death controlled by availability of diffusible nutrients, with
progressive development of necrosis in solid tumours. It turns
out that it is not that simple, as stressed by Mueller-Klieser
(2000): the onset of necrosis is not caused by simply one factor,
such as hypoxia, as it had been widely accepted for many
years, and the development of necrosis in tumour spheroids
is a complex phenomenon involving other factors, such as
cell–cell and cell–matrix interactions, expression of growth
factor receptors and adhesion molecules, etc. Thus, even if
more detailed models of MTSph are still actively developed
invoking diffusion-limited phenomena, other recent models
explore alternative views, allowing necrotic regions to develop
under conditions of adverse mechanical stress rather than in
regions of low nutrient concentrations (Landman and Please
2001). In parallel, tumour spheroids models can be revisited
and improved for mimicking tumour stroma and vasculature
(Alajati et al 2008, Lin and Chang 2008, Mueller-Klieser
2000).
We highlighted in this review how a transition toward
a more mature approach of solid tumour modelling has
been taken this last decade, especially with the increasing
consideration of biophysical factors in the control exerted
by the tumour microenvironment onto the growth process.
Indeed, the microenvironment is no longer modelled only as
a provider of diffusible nutrients and growth factors, or as a
medium in which inflammatory and immune processes take
place, but as a deformable and remodelled tissue, with time-
varying structural and mechanical properties which participate
to the ‘selective pressure’ imposed to cell dynamics during
tumour development.
This transition is correlated with the increasing
development, in this post-genome sequencing era, of
two research domains at the interface between physics,
mathematics and cell biology. The first one is the recognized
limitation of the reductionist approach, which, despite
outstanding progresses made in normal and cancer cells
molecular biology, is inadequate when used alone to capture
the complexity of tumour progression and the onset of
metastasis (Bizzarri et al 2008). In addition, the overwhelming
amount of data collected by molecular biology techniques
Rep. Prog. Phys. 72 (2009) 056701
has highlighted the need for a systemic approach of cellular
and tissular biology. In this system biology framework, the
tumour microenvironment is now a crucial element of the
cancer puzzle. Thus, models of tumour growth are no longer
limited to an inside transformation in cellular organization,
resulting from mutations and controlled by external supply,
but more globally as an altered signalling context in which
dramatic modifications of the signalling-regulatory pathways
originating from the tumour stroma and surrounding host tissue
may markedly affect the structure, functioning and growth of
the tumour (Bissell and Radisky 2001, Bissell et al 2005).
These considerations introduce a second facet of the
system biology of tumour growth, namely the increasingly
recognized and documented importance of the biomechanical
contexts as determinants of the cell signalling pathways.
Notably, mechanotransduction processes profoundly influence
all aspects of normal and cancer cells behaviour, including
proliferation apoptosis, adhesion, migration, secretion, etc
(Huang and Ingber 2005, Ingber 2002). These elements
highlight the need for tumour models development considering
the superimposition of both biochemical and biomechanical
regulatory pathways.
In this context, relevant representations of cell, tumours
and host tissues as mechanical entities are necessary for the
evaluation of stress and strains experienced and transmitted
by the tumour cells. We illustrated in section 4 the diversity
of models used to describe tumour mechanical properties,
with tumours visualized as amorphous solids, viscous fluids or
granular solid (Guiot et al 2007a). Central to an understanding
of tissue stress evolution is also the necessity of identifying the
precise constitutive nature of growing tumour tissues, taking
into account tumour heterogeneity. Different modelling choice
can be made, either trying to define a homogenized continuum,
or taking into account specifically tumour heterogeneity,
anisotropy and multiphasic nature.
As reviewed here, such biomechanical models already
help to increase our understanding of necrosis formation in
tumour and MTSph development. But improved mechanical
models of tumours and cancerous tissue also give qualitative
and quantitative insights into the regulation of cell adhesion and
migration. This includes both the subcellular levels, in terms of
stress transmitted by the cells and controlling cell adhesion, as
well as the cellular level, where we underlined how haptotaxis
can be a key determinant of tumour cell migration, controlled
by both the gradients of adhesion molecules and gradients of
rigidity created within the surrounding matrix by proteases
activity and ECM biosynthesis. We additionally highlighted
that accurate modelling of tumours mechanical properties
has a profound impact on our understanding of tumour
vascularization and on the collapse of tumour blood vessels,
which is of fundamental importance for the administration of
anticancer agents.
In a broader modelling perspective, considering the
superimposition, at different scales, of both biochemical and
biomechanical regulatory pathways enforces the consideration
of a tumour as a complex biological system governed
intrinsically by nonlinear processes. The potential existence of
multiple asymptotic solutions exhibited by nonlinear systems
23
P Tracqui
indicate that the development path of a tumour is not unique for
a given set of genomic parameters, but depends on the initial
condition and environmental perturbations experienced by the
growing tumour.
We reviewed in this paper another systemic property of
several tumour models, namely the emergence of specific
instability modes in the spatial distribution of cancer cells
at the tumour surface and its interpretation as a bifurcation
process sustained by intrinsic instabilities. Such a bifurcation
scenario is clearly very helpful to predict whether a particular
tumour type in a given host tissue would exhibit at some
developmental stage a transition from a compact shape with
smooth margins toward a stellate morphology with fingering
or filopodial extensions, the latter being the signature of an
invasive phenotype. The problem is that a rather large spectrum
of instability mechanisms can trigger the onset of finger-
like invasive patterns. Existing models mostly considered
diffusive instabilities, either as a simple Turing-like pre-
patterning of the tumour surface alone (Chaplain et al 2001),
or alternatively nonlinear interactions and development of
cellular forces at the tumour–host tissue interface. In these
models, the irregular tumour morphology results from a global
and synchronized patterning of the tumour surface, but one
may ask whether the invasive branching rather results from
a desynchronized activity that takes place at different sites
across the tumour surface. This pleads for a refined modelling
of microenvironment influence on transition toward invasive
types of tumour morphologies. It is known that hypoxia
or nutrient deprivation leads to invasive tumour behaviour,
but the work of Anderson et al (2006) additionally showed
that alterations of the microenvironment, from homogeneous
to inhomogeneous structure, could produce a change from
smooth, non-invasive margins, to morphologically invasive,
fingering margins.
Such requirements for refined biophysical models of
tumour growth feedbacks onto the integrative modelling
strategy quoted in the introductory part of this section, which
tends to consider cancer as a systemic disease and more
specifically to provide an all-comprehensive model of tumour
growth (Anderson and Quaranta 2008). A clear implication
of this perspective is to address the fundamental problem of
coupling different phenomena at different scales, attempting to
integrate detailed subcellular information to make predictions
at the tumour scale. Filling the gap between microscopic
and macroscopic descriptions is a challenge of fundamental
importance for tumour growth modelling, which leaves an
open space for the development of models that has just
begun to be explored (Ingber 2008b). One may think
to a juxtaposition of models or to intermediate, so-called
mesoscopic models (Anderson and Quaranta 2008) in order
to present a unified and integrative model of the entire process
of tumour development. In this system biology framework,
most promising models are likely to be based on hybrid
continuous–discrete descriptions and formalisms. Indeed,
continuum descriptions of tumour growth are supported by
the existing bodies of phenomenological and constitutive
laws elaborated in agreement with fundamental physical
principles, and thus take benefit from the knowledge gained
Rep. Prog. Phys. 72 (2009) 056701
in other fields of physics. This aspect has been especially
emphasized in the above sections with the various approaches
proposed to describe a solid tumour in the framework of
continuum mechanics, with tumour description based on
analogies with fluid dynamics or well-known mechanisms of
fracture mechanics (Guiot et al 2006, 2007b). Additionally,
the spatio-temporal evolution of the associated vector fields is
also partly amenable to theoretical analysis, with emerging
simulated behaviours of the models controlled by a rather
limited number of parameters or parameter groupings, as
illustrated with diffusion-instabilities predictions. Thus, Guiot
et al (2007a) defined an invasion parameter (IP) through a
formula that identifies the most relevant physical parameters,
e.g. the tumour’s surface tension, its radius and the confining
host tissue pressure, whose control should be the target
of dedicated therapies. Eventually, a cancer type, organ
site and patient-specific IP may be of significant value for
diagnostic purposes, as most of this multicellular behaviour
occurs well below the current non-invasive imaging resolution
limits. In the same perspective, Chaplain and Sleeman (1993)
originally proposed earlier a classification and grading of solid
tumours based on strain energy functions. However, we are
still looking for appropriate constitutive laws for modelling
tumours mechanical properties, knowing that these laws may
change with tumour types and grading.
Despite these advantages, the averaging over space
realized in continuum formalisms can hardly account for
the diversity of cellular and subcellular dynamical features
and genetic or epigenetic regulatory mechanisms exhibited at
the individual cell level. Alternatively, discrete models are
perfectly adapted for modelling internal signalling networks
within each cell. Thus, hybrid models, in which individual
cells are treated discretely but interact with other chemical
and mechanical continuum fields, are likely to combine the
benefits of both descriptions. Under appropriated formalisms,
such models will be able to couple the different spatial
scales impacted by the growth process with biochemical and
biomechanical information passed between multiple scales.
Especially, these multiscale biophysical models should help
to understand how the emerging shift of architectural and
tensional homeostasis of the host tissue is shifted toward
malignancy by bifurcation processes taking place within
interactive complex signalling networks. In addition to the
diffusive or mechanochemical instabilities quoted here, these
modelling approaches will put into a broader perspective
and larger basis the consideration of tumour branching and
invasiveness as instability phenomena.
Nevertheless, this is clearly an enormous challenge
when considering the increasing number of components
to consider within the ‘omics family’ which characterize
complex functions of many cellular components. In
addition to genomics, transcriptomics and proteomics (Doucet
and Overall 2008), metabolomics has been added more
recently, as the discipline that describes ‘the complete set
of metabolites/low-molecular-weight intermediates, i.e. the
metabolome, that vary according to the developmental or
pathological state of the cell, tissue, organ or organism’
(Nielsen and Oliver 2005). Similarly, let us add by analogy
24
P Tracqui
mechanomics as the discipline analysing the variation of
the complete set of mechanical interactions at the cell–
ECM interface, i.e. the mechanome, in a context dependent
way. The point is that, to increase their relevance,
biophysical models of tumours should tell us something
about how genome, transcriptome, proteome, metabolome
and mechanome networks and charts evolve with variation
of physical fields defining macroscopically the environmental
context. However, such multiscale models are likely to
become more and more complicated, with large increase in
the model parameters. Then, a risk is that they may succeed in
supporting partly intuition through computational results, but
without giving real understanding and insight into underlying
biological principles, and without ensuring to get relevant
predictions and exploration of all possible model behaviours.
A guard against building over complicated models might
be to keep a functional view of tumorigenesis and malignancy,
as a drift from a morphofunctional field, defined as the
physiological spatio-temporal coherence and regulation of
cells and tissues organization. This view is closely connected
to the earlier concept of epigenetic landscape introduced by
C H Waddington to deal with cell fate in embryogenesis and
morphogenesis (Waddington 1968). Thus, the appearance
of a tumour indicates a breakdown of this morphofunctional
field, the organizing properties of the tissue homeostasis
being gradually lost as the tumour develops. Within this
framework, tumorigenesis is no longer studied as a merely
deterministic consequence of altered gene function, but rather
as an emerging response of network-structured interactions.
For nonlinear systems, these interactions would define the
number and topologies of each attraction bassin of the phase
portrait associated to the considered model (Huang and Ingber
2006, Huang et al 2005). The dynamic of the model variables
in response to the different environmental and endogenous
stimuli experienced by the tumour cells may correspond to
transitions between attraction basins (domains), i.e. to the
crossing of the boundaries (or barriers) that delineate each
of them. In this spirit, Gatenby and Gillies (2008) recently
developed the concept of adaptive landscape to qualify the
microenvironmental forces and selection mechanisms acting
at work on tumour cells population during carcinogenesis.
Considering that the tumour stroma is modified at critical
steps during cancer development, they analyse the sequence
of multiple and self-interacting proliferation barriers that
are overcome in carcinogenesis, emphasizing that multiple
strategies may be used to overcome a tissue barrier to
proliferation, while adaptation to one barrier may decrease or
increase the height of subsequent barriers. In this perspective,
relevance and explanatory power of biophysical models of
tumorigenesis might be evaluated through their capacity to
include one adaptation to every barrier, while exploring
different strategies serving as adaptations to the considered
barrier. At the same time, the breakdown of the normal
morphofunctional field becomes predominantly explainable
from the analysis of the processes and functions ensuring
the adaptation to the barrier (e.g. adhesion or proteolysis),
rather than from the detailed knowledge of the biological actors
coming into play (e.g. the precise integrin or metaloprotease
at work).
Rep. Prog. Phys. 72 (2009) 056701
In conclusion, the complexity of the system biology of
cancer, even restricted to the development of solid tumours,
still overwhelms us, leaving a huge gap between the enormous
amount of experimental information that has been collected,
and the explicative and predictive power of theoretical models.
But even if open problems in the study of tumour development
are still numerous, biophysics models have still a lot to offer to
improve our understanding of tumour morphodynamics. We
outlined the different ways recent models provide mechanistic
insights on tumour growth and vascularization through
quantitative simulations that determine how combinations of
microenvironmental and cellular parameters trigger invasive
tumour morphologies. Consideration of the many possible
feedback loops, either autobaric or parabaric, which regulate
biomechanical interactions between normal and cancer cells
and the tumour extracellular environment is still largely
lacking, but analysing how mechanical stresses acting onto cell
secretory activity come into play through mechanotransduction
processes is a promising orientation for new integrative models
(Butcher et al 2009, Macklin and Lowengrub 2007). A long-
term goal of the modelling effort is the ability to predict which
combination of the key functional switches (proliferation,
death, degradation, biosynthesis, migration, traction, etc)
might reverse the invasive behaviour of individual and specific
tumours, based on actual measurements of molecular, cellular
and microenvironmental parameters (Ingber 2008a). A more
accessible, short-term goal, of the modelling effort, enhanced
by the percolation of biophysical models in biology and
medicine, is to actively participate to the design of new
experimental models. Progress in tissue engineering is
encouraging for the larger development of assays targeting
tumour–stroma interactions, interactions between tumour and
immune cells and tumour vascularization (Ghajar et al 2007,
Ingber 2008a). Another short-term goal for biophysical
models development, with strong implications in health
care, is their use as predictive tools in drug discovery
and in the management of cancer patients. In addition to
molecular or biochemical studies focusing on cell sensitivity
and chemoresistance, they will contribute to explain from a
more systemic and virtual approach why some therapies fail
while others prove to be effective in locally controlling tumour
expansion. Models already appeared as efficient tools for
gathering the information on tumour morphology, biology and
metabolism in conjunction with multimodal high-resolution
imaging techniques. As an adjunctive aid to customize
therapy in individual patients, such practical successes will be
highly encouraging, since even if an increasing effort is being
made for the understanding complexity of tumorigenesis,
elaborating new modelling paradigms that integrate all
available knowledge remains a challenge facing us.
Acknowledgments
We thank Dr. Bruno Vailhé for assistance with the
microcarrier-based angiogenesis assay.
25
P Tracqui
References
Adam J A and Bellomo N 1997 A Survey of Models for
Tumor-immune System Dynamics (Boston: Birkhaüser)
Afenya E K and Calderon C P 2000 Diverse ideas on the growth
kinetics of disseminated cancer cells Bull. Math. Biol.
62 527–42
Alajati A et al 2008 Spheroid-based engineering of a human
vasculature in mice Nat. Methods 5 439–45
Albertson D G, Collins C, McCormick F and Gray J W 2003
Chromosome aberrations in solid tumors Nat. Genet.
34 369–76
Alexander N R, Branch K M, Parekh A, Clark E S, Iwueke I C,
Guelcher S A and Weaver A M 2008 Extracellular matrix
rigidity promotes invadopodia activity Curr. Biol. 18 1295–9
Alford P W, Humphrey J D and Taber L A 2008 Growth and
remodeling in a thick-walled artery model: effects of spatial
variations in wall constituents Biomech. Modelling
Mechanobiol. 7 245–62
Alvord E C, Jr 1995 Patterns of growth of gliomas AJNR Am. J.
Neuroradiol. 16 1013–7
Ambrosi D and Mollica F 2002 On the mechanics of a growing
tumor Int. J. Eng. Sci. 40 1297–316
Ambrosi D and Mollica F 2004 The role of stress in the growth of a
multicell spheroid J. Math. Biol. 48 477–99
Ambrosi D, Guillou A and Di Martino E S 2008 Stress-modulated
remodeling of a non-homogeneous body Biomech. Modelling
Mechanobiol. 7 63–76
Anderson A R and Chaplain M A 1998 Continuous and discrete
mathematical models of tumor-induced angiogenesis Bull.
Math. Biol. 60 857–99
Anderson A R and Quaranta V 2008 Integrative mathematical
oncology Nat. Rev. Cancer 8 227–34
Anderson A R, Weaver A M, Cummings P T and Quaranta V 2006
Tumor morphology and phenotypic evolution driven by
selective pressure from the microenvironment Cell
127 905–15
Araujo R P and McElwain D L 2004 A history of the study of solid
tumour growth: the contribution of mathematical modelling
Bull. Math. Biol. 66 1039–91
Avraamides C J, Garmy-Susini B and Varner J A 2008 Integrins in
angiogenesis and lymphangiogenesis Nat. Rev. Cancer
8 604–17
Balaban N Q et al 2001 Force and focal adhesion assembly: a close
relationship studied using elastic micropatterned substrates
Nat. Cell. Biol. 3 466–72
Balkwill F 2004 Cancer and the chemokine network Nat. Rev.
Cancer 4 540–50
Balmain A, Gray J and Ponder B 2003 The genetics and genomics
of cancer Nat. Genet. 33 238–44
Banes A J, Tsuzaki M, Yamamoto J, Fischer T, Brigman B, Brown T
and Miller L 1995 Mechanoreception at the cellular level: the
detection, interpretation, and diversity of responses to
mechanical signals Biochem. Cell Biol. 73 349–65
Bauer A L, Jackson T L and Jiang Y 2007 A cell-based model
exhibiting branching and anastomosis during tumor-induced
angiogenesis Biophys. J. 92 3105–21
Beningo K A, Lo C M and Wang Y L 2002 Flexible polyacrylamide
substrata for the analysis of mechanical interactions at
cell-substratum adhesions Methods Cell Biol. 69 325–39
Benkherourou M, Gumery P Y, Tranqui L and Tracqui P 2000
Quantification and macroscopic modeling of the nonlinear
viscoelastic behavior of strained gels with varying fibrin
concentrations IEEE Trans. Biomed. Eng. 47 1465–75
Bergers G et al 2000 Matrix metalloproteinase-9 triggers the
angiogenic switch during carcinogenesis Nat. Cell Biol.
2 737–44
Berrier A L and Yamada K M 2007 Cell-matrix adhesion J. Cell.
Physiol. 213 565–73
Rep. Prog. Phys. 72 (2009) 056701
Bhowmick N A, Neilson E G and Moses H L 2004 Stromal
fibroblasts in cancer initiation and progression Nature
432 332–7
Bissell M J and Aggeler J 1987 Dynamic reciprocity: how do
extracellular matrix and hormones direct gene expression?
Prog. Clin. Biol. Res. 249 251–62
Bissell M J and Radisky D 2001 Putting tumours in context Nat.
Rev. Cancer 1 46–54
Bissell M J, Kenny P A and Radisky D C 2005 Microenvironmental
regulators of tissue structure and function also regulate tumor
induction and progression: the role of extracellular matrix and
its degrading enzymes Cold Spring Harb. Symp. Quant. Biol.
70 343–56
Bissell M J, Radisky D C, Rizki A, Weaver V M and Petersen O W
2002 The organizing principle: microenvironmental influences
in the normal and malignant breast Differentiation 70 537–46
Bizzarri M, Cucina A, Conti F and D’Anselmi F 2008 Beyond the
oncogene paradigm: understanding complexity in
Cancerogenesis Acta Biotheor. 56 173–96
Bosman F T and Stamenkovic I 2003 Functional structure and
composition of the extracellular matrix J. Pathol. 200 423–8
Bru A, Albertos S, Luis Subiza J, Garcia-Asenjo J L and Bru I 2003
The universal dynamics of tumor growth Biophys. J.
85 2948–61
Bussolino F, Valdembri D, Caccavari F and Serini G 2006
Semaphoring vascular morphogenesis Endothelium 13 81–91
Butcher D T, Alliston T and Weaver V M 2009 A tense situation:
forcing tumour progression Nat. Rev. Cancer 9 108–22
Byrne H and Preziosi L 2003 Modelling solid tumour growth using
the theory of mixtures Math. Med. Biol. 20 341–66
Byrne H M and Chaplain M A J 1996a Growth of necrotic tumors in
the presence and absence of inhibitors Math. Biosci.
135 187–216
Byrne H M and Chaplain M A J 1996b Modeling the role of
cell–cell adhesion in the growth and development of
carcinomas Math. Comput. Modelling 24 1–17
Byrne H M, Alarcon T, Owen M R, Webb S D and Maini P K 2006
Modelling aspects of cancer dynamics: a review Phil. Trans. A
Math. Phys. Eng. Sci. 364 1563–78
Carmeliet P 2003 Angiogenesis in health and disease Nat. Med.
9 653–60
Casciari J J, Sotirchos S V and Sutherland R M 1992 Variations in
tumor cell growth rates and metabolism with oxygen
concentration, glucose concentration, and extracellular pH
J. Cell. Physiol. 151 386–94
Cavallaro U and Christofori G 2000 Molecular mechanisms of tumor
angiogenesis and tumor progression J. Neurooncol. 50 63–70
Chang H Y, Sneddon J B, Alizadeh A A, Sood R, West R B,
Montgomery K, Chi J T, van de Rijn M, Botstein D and
Brown P O 2004 Gene expression signature of fibroblast serum
response predicts human cancer progression: similarities
between tumors and wounds PLoS Biol. 2 E7
Chaplain M A and Sleeman B D 1993 Modelling the growth of solid
tumours and incorporating a method for their classification
using nonlinear elasticity theory J. Math. Biol. 31 431–73
Chaplain M A, Ganesh M and Graham I G 2001 Spatio-temporal
pattern formation on spherical surfaces: numerical simulation
and application to solid tumour growth J. Math. Biol.
42 387–423
Chaplain M A, Giles S M, Sleeman B D and Jarvis R J 1995 A
mathematical analysis of a model for tumour angiogenesis
J. Math. Biol. 33 744–70
Chaplain M A, McDougall S R and Anderson A R 2006
Mathematical modeling of tumor-induced angiogenesis Annu.
Rev. Biomed. Eng. 8 233–57
Chasiotis I, Fillmore H L, and Gillies G T 2003 Atomic force
microscopy measurement of cytostructural elements involved
in the nanodynamics of tumour cell invasion Nanotechnology
14 557–61
26
P Tracqui
Chen C S 2008 Mechanotransduction—a field pulling together?
J. Cell Sci. 121 3285–92
Chen C Y, Byrne H M and King J R 2001 The influence of
growth-induced stress from the surrounding medium on the
development of multicell spheroids J. Math. Biol. 43 191–220
Clark I M, Swingler T E, Sampieri C L and Edwards D R 2008 The
regulation of matrix metalloproteinases and their inhibitors Int.
J. Biochem. Cell Biol. 40 1362–78
Collin O, Tracqui P, Stephanou A, Usson Y, Clement-Lacroix J and
Planus E 2006 Spatiotemporal dynamics of actin-rich adhesion
microdomains: influence of substrate flexibility J. Cell Sci.
119 1914–25
Cook J 1995 Mathematical models for dermal wound healing:
wound contraction and scar formation PhD Thesis Department
of Applied Mathematics, University of Washington, Seattle
Coussens L M and Werb Z 2002 Inflammation and cancer Nature
420 860–7
Coussens L M, Fingleton B and Matrisian L M 2002 Matrix
metalloproteinase inhibitors and cancer: trials and tribulations
Science 295 2387–92
Cowin S C 2004 Tissue growth and remodeling Annu. Rev. Biomed.
Eng. 6 77–107
Cristini V, Lowengrub J and Nie Q 2003 Nonlinear simulation of
tumor growth J. Math. Biol. 46 191–224
Cristini V, Frieboes H B, Gatenby R, Caserta S, Ferrari M and
Sinek J 2005 Morphologic instability and cancer invasion Clin.
Cancer Res. 11 6772–9
Deisboeck T S, Berens M E, Kansal A R, Torquato S,
Stemmer-Rachamimov A O and Chiocca E A 2001 Pattern
of self-organization in tumour systems: complex growth
dynamics in a novel brain tumour spheroid model Cell Prolif..
34 115–34
Delsanto P P, Guiot C, Degiorgis P G, Condat C A, Mansury Y and
Deisboeck T S 2004 Growth model for multicellular tumor
spheroids Appl. Phys. Lett. 85 4225–27
Delvoye P, Wiliquet P, Leveque J L, Nusgens B V and Lapiere C M
1991 Measurement of mechanical forces generated by skin
fibroblasts embedded in a three-dimensional collagen gel
J. Invest Dermatol. 97 898–902
Dembo M, Oliver T, Ishihara A and Jacobson K 1996 Imaging the
traction stresses exerted by locomoting cells with the elastic
substratum method Biophys. J. 70 2008–22
Doucet A and Overall C M 2008 Protease proteomics: revealing
protease in vivo functions using systems biology approaches
Mol. Aspects Med. 29 339–58
du Roure O, Saez A, Buguin A, Austin R H, Chavrier P, Siberzan P
and Ladoux B 2005 Force mapping in epithelial cell migration
Proc. Natl Acad. Sci. USA 102 2390–5
Dvorak H F 1986 Tumors: wounds that do not heal. Similarities
between tumor stroma generation and wound healing N. Engl.
J. Med. 315 1650–9
Eastwood M, Porter R, Khan U, McGrouther G and Brown R 1996
Quantitative analysis of collagen gel contractile forces
generated by dermal fibroblasts and the relationship to cell
morphology J. Cell. Physiol. 166 33–42
Egeblad M and Werb Z 2002 New functions for the matrix
metalloproteinases in cancer progression Nat. Rev. Cancer
2 161–74
Ferreira S C Jr, Martins M L and Vilela M J 2002
Reaction–diffusion model for the growth of avascular tumor
Phys. Rev. E Stat. Nonlinear Soft Matter Phys. 65 021907
Fillmore H L, Chasiotis I, Cho S W and Gillies G T 2003 Atomic
force microscopy observations of tumour cell invadopodia:
novel cellular nanomorphologies on collagen substrates
Nanotechnology 14 73–6
Folkman J 2002 Role of angiogenesis in tumor growth and
metastasis Semin. Oncol. 29 15–8
Folkman J 2003 Fundamental concepts of the angiogenic process
Curr. Mol. Med. 3 643–51
Rep. Prog. Phys. 72 (2009) 056701
Frieboes H B, Zheng X, Sun C H, Tromberg B, Gatenby R and
Cristini V 2006 An integrated computational/experimental
model of tumor invasion Cancer Res. 66 1597–604
Frieboes H B, Lowengrub J S, Wise S, Zheng X, Macklin P,
Bearer E L and Cristini V 2007 Computer simulation of glioma
growth and morphology Neuroimage 37 S59–70
Fu X, Parks W C and Heinecke J W 2008 Activation and silencing
of matrix metalloproteinases Semin. Cell Dev. Biol. 19 2–13
Fukumura D and Jain R K 2007 Tumor microvasculature and
microenvironment: targets for anti-angiogenesis and
normalization Microvasc. Res. 74 72–84
Galbraith C G and Sheetz M P 1997 A micromachined device
provides a new bend on fibroblast traction forces Proc. Natl
Acad. Sci. USA 94 9114–8
Gamba A, Ambrosi D, Coniglio A, de Candia A, Di Talia S,
Giraudo E, Serini G, Preziosi L and Bussolino F 2003
Percolation, morphogenesis, and Burgers dynamics in blood
vessels formation Phys. Rev. Lett. 90 118101–4
Gatenby R A and Gillies R J 2008 A microenvironmental model of
carcinogenesis Nat. Rev. Cancer 8 56–61
Gatenby R A and Maini P 2002 Modelling a new angle on
understanding cancer Nature 420 462
Gatenby R A, Gawlinski E T, Gmitro A F, Kaylor B and Gillies R J
2006 Acid-mediated tumor invasion: a multidisciplinary study
Cancer Res. 66 5216–23
Geiger B and Bershadsky A 2002 Exploring the neighborhood:
adhesion-coupled cell mechanosensors Cell 110 139–42
Geiger B, Bershadsky A, Pankov R and Yamada K M 2001
Transmembrane crosstalk between the extracellular
matrix–cytoskeleton crosstalk Nat. Rev. Mol. Cell Biol.
2 793–805
Gerisch A and Chaplain M A 2008 Mathematical modelling of
cancer cell invasion of tissue: local and non-local models and
the effect of adhesion J. Theor. Biol. 250 684–704
Gevertz J L, Gillies GT and Torquato S 2008 Simulating tumor
growth in confined heterogeneous environments Phys. Biol.
5 36010–19
Ghajar C M, Suresh V, Peyton S R, Raub C B, Meyskens F L Jr,
George S C and Putnam A J 2007 A novel three-dimensional
model to quantify metastatic melanoma invasion Mol. Cancer
Ther. 6 552–61
Ghajar C M, Chen X, Harris J W, Suresh V, Hughes C C, Jeon N L,
Putnam A J and George S C 2008 The effect of matrix density
on the regulation of 3-D capillary morphogenesis Biophys. J.
94 1930–41
Gharaee-Kermani M and Phan S H 2001 Role of cytokines and
cytokine therapy in wound healing and fibrotic diseases Curr.
Pharm. Des. 7 1083–103
Ghosh K, Thodeti C K, Dudley A C, Mammoto A, Klagsbrun M and
Ingber D E 2008 Tumor-derived endothelial cells exhibit
aberrant Rho-mediated mechanosensing and abnormal
angiogenesis in vitro Proc. Natl Acad. Sci. USA 105 11305–10
Gieni R S and Hendzel M J 2008 Mechanotransduction from the
ECM to the genome: are the pieces now in place? J. Cell
Biochem. 104 1964–87
Gordon V D, Valentine M T, Gardel M L, Andor-Ardo D,
Dennison S, Bogdanov A A, Weitz D A and Deisboeck T S
2003 Measuring the mechanical stress induced by an
expanding multicellular tumor system: a case study Exp. Cell
Res. 289 58–66
Graziano L and Preziosi L 2007 Mechanics in tumor growth
Modelling of Biological Materials ed F Mollica et al (Basel:
Birkhäuser) pp 267–328
Greenspan H P 1976 On the growth and stability of cell cultures and
solid tumors J. Theor. Biol. 56 229–42
Gruss C J et al 2003 Stroma formation and angiogenesis by
overexpression of growth factors, cytokines, and proteolytic
enzymes in human skin grafted to SCID mice J. Invest.
Dermatol. 120 683–92
27
P Tracqui
Guiot C, Pugno N and Delsanto P P 2006 Elastomechanical model
of tumor invasion Appl. Phys.Lett. 89 233901–3
Guiot C, Pugno N, Delsanto P P and Deisboeck T S 2007a Physical
aspects of cancer invasion Phys. Biol. 4 P1–6
Guiot C, Delsanto P P and Deisboeck T S 2007b Morphological
instability and cancer invasion: a ‘splashing water drop’
analogy Theor. Biol. Med. Modelling 4 4
Gutierrez-Fernandez A et al 2008 Matrix metalloproteinase–8
functions as a metastasis suppressor through modulation of
tumor cell adhesion and invasion Cancer Res. 68 2755–63
Habib S, Molina-Paris C and Deisboeck T S 2003 Complex
dynamics of tumors: modeling an emerging brain tumor
system with coupled reaction–diffusion equations Physica A
327 501–04
Harpold H L, Alvord E C Jr and Swanson K R 2007 The evolution
of mathematical modeling of glioma proliferation and invasion
J. Neuropathol. Exp. Neurol. 66 1–9
Harris A K, Stopak D and Wild P 1981 Fibroblast traction as a
mechanism for collagen morphogenesis Nature 290 249–51
Harris A K, Stopak D and Warner P 1984 Generation of spatially
periodic patterns by a mechanical instability: a mechanical
alternative to the Turing model J. Embryol. Exp. Morphol.
80 1–20
Helmlinger G, Netti P A, Lichtenbeld H C, Melder R J and Jain R K
1997 Solid stress inhibits the growth of multicellular tumor
spheroids Nat. Biotechnol. 15 778–83
Hirata H, Tatsumi H and Sokabe M 2008 Mechanical forces
facilitate actin polymerization at focal adhesions in a
zyxin-dependent manner J. Cell Sci. 121 2795–804
Holmes M J and Sleeman B D 2000 A mathematical model of
tumour angiogenesis incorporating cellular traction and
viscoelastic effects J. Theor. Biol. 202 95–112
Hsu S, Thakar R, Liepmann D and Li S 2005 Effects of shear stress
on endothelial cell haptotaxis on micropatterned surfaces
Biochem. Biophys. Res. Commun. 337 401–9
Huang H, Kamm R D and Lee R T 2004 Cell mechanics and
mechanotransduction: pathways, probes, and physiology Am.
J. Physiol. Cell Physiol. 287 C1–11
Huang S and Ingber D E 2005 Cell tension, matrix mechanics, and
cancer development Cancer Cell 8 175–6
Huang S and Ingber D E 2006 A non-genetic basis for cancer
progression and metastasis: self-organizing attractors in cell
regulatory networks Breast Dis. 26 27–54
Huang S, Eichler G, Bar-Yam Y and Ingber D E 2005 Cell fates as
high-dimensional attractor states of a complex gene regulatory
network Phys. Rev. Lett. 94 128701
Humphrey J D 2003 Continuum biomechanics of soft biological
tissues Proc. R. Soc. Lond. A 459 3–46
Hynes R O 2002a Integrins: bidirectional, allosteric signaling
machines Cell 110 673–87
Hynes R O 2002b A reevaluation of integrins as regulators of
angiogenesis Nat. Med. 8 918–21
Ingber D E 2002 Cancer as a disease of epithelial-mesenchymal
interactions and extracellular matrix regulation Differentiation
70 547–60
Ingber D E 2008a Can cancer be reversed by engineering the tumor
microenvironment? Semin. Cancer Biol. 18 356–64
Ingber D E 2008b Tensegrity-based mechanosensing from macro to
micro Prog. Biophys. Mol. Biol. 97 163–79
Jaalouk D E and Lammerding J 2009 Mechanotransduction gone
awry Nat. Rev. Mol. Cell Biol. 10 63–73
Jackson T L and Byrne H M 2002 A mechanical model of tumor
encapsulation and transcapsular spread Math. Biosci.
180 307–28
Jiang Y, Pjesivac-Grbovic J, Cantrell C and Freyer J P 2005 A
multiscale model for avascular tumor growth Biophys. J.
89 3884–94
Jones P A and Baylin S B 2007 The epigenomics of cancer Cell
128 683–92
Rep. Prog. Phys. 72 (2009) 056701
Kalluri R and Zeisberg M 2006 Fibroblasts in cancer Nat. Rev.
Cancer 6 392–401
Kass L, Erler J T, Dembo M and Weaver V M 2007 Mammary
epithelial cell: influence of extracellular matrix composition
and organization during development and tumorigenesis Int. J.
Biochem. Cell Biol. 39 1987–94
Kassis J, Lauffenburger D A, Turner T and Wells A 2001 Tumor
invasion as dysregulated cell motility Semin. Cancer Biol.
11 105–17
Keith B and Simon M C 2007 Hypoxia-inducible factors, stem cells,
and cancer Cell 129 465–72
Kelly C, Smallbone K and Brady M 2008 Tumour glycolysis: the
many faces of HIF J. Theor. Biol. 254 508–13
Khain E and Sander L M 2006 Dynamics and pattern formation in
invasive tumour growth Phys. Rev. Lett. 96 1881031–4
Kolodney M S and Wysolmerski R B 1992 Isometric contraction by
fibroblasts and endothelial cells in tissue culture: a quantitative
study J. Cell Biol. 117 73–82
Korff T and Augustin H G 1999 Tensional forces in fibrillar
extracellular matrices control directional capillary sprouting
J. Cell Sci. 112 3249–58
Kozusko F and Bourdeau M 2007b A unified model of sigmoid
tumour growth based on cell proliferation and quiescence Cell
Prolif. 40 824–34
Kozusko F, Bourdeau M, Bajzer Z and Dingli D 2007a A
microenvironment based model of antimitotic therapy of
Gompertzian tumor growth Bull. Math. Biol. 69 1691–708
Krishnan L, Underwood C J, Maas S, Ellis B J, Kode T C,
Hoying J B and Weiss J A 2008 Effect of mechanical boundary
conditions on orientation of angiogenic microvessels
Cardiovasc. Res. 78 324–32
Kulbe H, Levinson N R, Balkwill F and Wilson J L 2004 The
chemokine network in cancer—much more than directing cell
movement Int. J. Dev. Biol. 48 489–96
Lafleur M A, Handsley M M and Edwards D R 2003
Metalloproteinases and their inhibitors in angiogenesis Expert
Rev. Mol. Med. 5 1–39
Lamalice L, Le Boeuf F and Huot J 2007 Endothelial cell migration
during angiogenesis Circ. Res. 100 782–94
Landman K A and Please C P 2001 Tumour dynamics and necrosis:
surface tension and stability IMA J. Math. Appl. Med. Biol.
18 131–58
Lecaudey V and Gilmour D 2006 Organizing moving groups during
morphogenesis Curr. Opin. Cell Biol. 18 102–7
Lee D S, Rieger H and Bartha K 2006 Flow correlated percolation
during vascular remodeling in growing tumors Phys. Rev. Lett.
96 058104
Lee S, Chen T T, Barber C L, Jordan M C, Murdock J, Desai S,
Ferrara N, Nagy A, Roos K P and Iruela-Arispe M L 2007
Autocrine VEGF signaling is required for vascular homeostasis
Cell 130 691–703
Levine H A, Pamuk S, Sleeman B D and Nilsen-Hamilton M 2001
Mathematical modeling of capillary formation and
development in tumor angiogenesis: penetration into the
stroma Bull. Math. Biol. 63 801–63
Li S, Huang N F and Hsu S 2005 Mechanotransduction in
endothelial cell migration J. Cell Biochem. 96 1110–26
Lin I E and Taber L A 1995 A model for stress-induced growth in
the developing heart J. Biomech. Eng. 117 343–9
Lin R Z and Chang H Y 2008 Recent advances in three-dimensional
multicellular spheroid culture for biomedical research
Biotechnol. J. 3 1172–84
Liotta L A and Kohn E C 2001 The microenvironment of the
tumour-host interface Nature 411 375–9
Lo C M, Wang H B, Dembo M and Wang Y L 2000 Cell
movement is guided by the rigidity of the substrate Biophys. J.
79 144–52
Lubkin S R and Jackson T 2002 Multiphase mechanics of capsule
formation in tumors J. Biomech. Eng. 124 237–43
28
P Tracqui
Lunt S J, Chaudary N and Hill R P 2008 The tumor
microenvironment and metastatic disease Clin. Exp. Metastasis
26 19–34
Macklin P and Lowengrub J 2007 Nonlinear simulation of the effect
of microenvironment on tumor growth J. Theor. Biol.
245 677–704
Marchant B P, Norbury J and Perumpanani A J 2000 Traveling
shock waves arising in a model of malignant invasion SIAM J.
Appl. Math. 60 463–76
Martins G G and Kolega J 2006 Endothelial cell protrusion and
migration in three-dimensional collagen matrices Cell Motil.
Cytoskeleton 63 101–15
Marusic M, Bajzer Z, Freyer J P and Vuk-Pavlovic S 1994a
Analysis of growth of multicellular tumour spheroids by
mathematical models Cell Prolif. 27 73–94
Marusic M, Bajzer Z, Vuk-Pavlovic S and Freyer J P 1994b Tumor
growth in vivo and as multicellular spheroids compared by
mathematical models Bull. Math. Biol. 56 617–31
Merks R M, Brodsky S V, Goligorksy M S, Newman S A and
Glazier J A 2006 Cell elongation is key to in silico replication
of in vitro vasculogenesis and subsequent remodeling Dev.
Biol. 289 44–54
Mierke C T, Rosel D, Fabry B and Brabek J 2008 Contractile forces
in tumor cell migration Eur. J. Cell Biol. 87 669–76
Mueller-Klieser W 1997 Three-dimensional cell cultures: from
molecular mechanisms to clinical applications Am. J. Physiol.
273 C1109–23
Mueller-Klieser W 2000 Tumor biology and experimental
therapeutics Crit. Rev. Oncol. Hematol. 36 123–39
Murray J D 2002 Mathematical Biology 3rd edn vol I (Berlin:
Springer)
Murray J D 2003 Mathematical Biology 3rd edn vol II (Berlin:
Springer)
Nagy J D 2005 The ecology and evolutionary biology of cancer: a
review of mathematical models of necrosis and tumor cell
diversity Math. Biosci. Eng. 2 381–418
Namy P, Ohayon J and Tracqui P 2004 Critical conditions for
pattern formation and in vitro tubulogenesis driven by cellular
traction fields J. Theor. Biol. 227 103–20
Nehls V and Drenckhahn D 1995 A novel, microcarrier-based in
vitro assay for rapid and reliable quantification of
three-dimensional cell migration and angiogenesis Microvasc.
Res. 50 311–22
Nelson C M and Bissell M J 2006 Of extracellular matrix, scaffolds,
and signaling: tissue architecture regulates development,
homeostasis, and cancer Annu. Rev. Cell Dev. Biol.
22 287–309
Netti P A, Berk D A, Swartz M A, Grodzinsky A J and Jain R K
2000 Role of extracellular matrix assembly in interstitial
transport in solid tumors Cancer Res. 60 2497–503
Nicolis G and Prigogine I 1977 Self-organization in
Non-equilibrium Systems (New York: Wiley)
Nielsen J and Oliver S 2005 The next wave in metabolome analysis
Trends Biotechnol. 23 544–6
Nyberg P, Xie L and Kalluri R 2005 Endogenous inhibitors of
angiogenesis Cancer Res. 65 3967–79
Ohayon J, Dubreuil O, Tracqui P, Le Floc’h S, Rioufol G,
Chalabreysse L, Thivolet F, Pettigrew R I and Finet G 2007
Influence of residual stress/strain on the biomechanical
stability of vulnerable coronary plaques: potential impact for
evaluating the risk of plaque rupture Am. J. Physiol. Heart
Circ. Physiol. 293 H1987–96
Oliver T, Dembo M and Jacobson K 1999 Separation of propulsive
and adhesive traction stresses in locomoting keratocytes J. Cell
Biol. 145 589–604
Olson M F and Sahai E 2009 The actin cytoskeleton in cancer cell
motility Clin. Exp. Metastasis 26 273–87
Orr A W, Helmke B P, Blackman B R and Schwartz M A 2006
Mechanisms of mechanotransduction Dev. Cell 10 11–20
Rep. Prog. Phys. 72 (2009) 056701
Paszek M J et al 2005 Tensional homeostasis and the malignant
phenotype Cancer Cell 8 241–54
Peyton S R, Ghajar C M, Khatiwala C B and Putnam A J 2007 The
emergence of ECM mechanics and cytoskeletal tension as
important regulators of cell function Cell Biochem. Biophys.
47 300–20
Potter J D 2001 Morphostats: a missing concept in cancer biology
Cancer Epidemiol. Biomarkers Prev. 10 161–70
Potter J D 2007 Morphogens, morphostats, microarchitecture and
malignancy Nat. Rev. Cancer 7 464–74
Preziosi L 2003 Cancer Modelling and Simulation (London:
Chapman and Hall, CRC Press)
Preziosi L and Tosin A 2009 Multiphase modelling of tumour
growth and extracellular matrix interaction: mathematical tools
and applications J. Math. Biol. 58 625–56
Quaranta V, Rejniak K A, Gerlee P and Anderson A R 2008
Invasion emerges from cancer cell adaptation to competitive
microenvironments: quantitative predictions from multiscale
mathematical models Semin. Cancer Biol. 18 338–48
Radisky D, Hagios C and Bissell M J 2001 Tumors are unique
organs defined by abnormal signaling and context Semin.
Cancer Biol. 11 87–95
Retsky M W, Swartzendruber D E, Wardwell R H and Bame P D
1990 Is Gompertzian or exponential kinetics a valid description
of individual human cancer growth? Med. Hypotheses
33 95–106
Riveline D, Zamir E, Balaban N Q, Schwarz U S, Ishizaki T,
Narumiya S, Kam Z, Geiger B and Bershadsky A D 2001 Focal
contacts as mechanosensors: externally applied local
mechanical force induces growth of focal contacts by an
mDia1-dependent and ROCK-independent mechanism J. Cell
Biol. 153 1175–86
Rockne R, Alvord E C Jr, Rockhill J K and Swanson K R 2009 A
mathematical model for brain tumor response to radiation
therapy J. Math. Biol. 58 561–8
Rodriguez E K, Hoger A and McCulloch A D 1994 Stress-dependent
finite growth in soft elastic tissues J. Biomech. 27 455–67
Roose T, Chapman S J and Maini P K 2007 Mathematical models of
avascular tumor growth SIAM Rev. 49 179–208
Rosello C, Ballet P, Planus E and Tracqui P 2004 Model driven
quantification of individual and collective cell migration Acta
Biotheor. 52 343–63
Ruiter D, Bogenrieder T, Elder D and Herlyn M 2002
Melanoma-stroma interactions: structural and functional
aspects Lancet Oncol. 3 35–43
Saez A, Ghibaudo M, Buguin A, Silberzan P and Ladoux B 2007
Rigidity-driven growth and migration of epithelial cells on
microstructured anisotropic substrates Proc. Natl Acad. Sci.
USA 104 8281–6
Sander L M and Deisboeck T S 2002 Growth patterns of
microscopic brain tumors Phys. Rev. E Stat. Nonlinear Soft
Matter Phys. 66 051901
Scheurer S B, Rybak J N, Rosli C, Neri D and Elia G 2004
Modulation of gene expression by hypoxia in human umbilical
cord vein endothelial cells: a transcriptomic and proteomic
study Proteomics 4 1737–60
Selmeczi D, Mosler S, Hagedorn P H, Larsen N B and Flyvbjerg H
2005 Cell motility as persistent random motion: theories from
experiments Biophys. J. 89 912–31
Sheetz M P, Felsenfeld D P and Galbraith C G 1998 Cell migration:
regulation of force on extracellular-matrix–integrin complexes
Trends Cell Biol. 8 51–4
Sheetz M P, Felsenfeld D, Galbraith C G and Choquet D 1999 Cell
migration as a five-step cycle Biochem. Soc. Symp. 65 233–43
Sherratt J A and Chaplain M A 2001 A new mathematical model for
avascular tumour growth J. Math. Biol. 43 291–312
Shreiber D I, Barocas V H and Tranquillo R T 2003 Temporal
variations in cell migration and traction during
fibroblast-mediated gel compaction Biophys. J. 84 4102–14
29
P Tracqui
Smallbone K, Gatenby R A and Maini P K 2008 Mathematical
modelling of tumour acidity J. Theor. Biol. 255 106–12
Smallbone K, Gavaghan D J, Gatenby R A and Maini P K 2005 The
role of acidity in solid tumour growth and invasion J. Theor.
Biol. 235 476–84
Smith A S, Sengupta K, Goennenwein S, Seifert U and Sackmann E
2008 Force-induced growth of adhesion domains is controlled
by receptor mobility Proc. Natl Acad. Sci. USA 105 6906–11
Sottile J 2004 Regulation of angiogenesis by extracellular matrix
Biochim. Biophys. Acta 1654 13–22
Spence A M et al 2008 Regional hypoxia in glioblastoma
multiforme quantified with [18F]fluoromisonidazole positron
emission tomography before radiotherapy: correlation
with time to progression and survival Clin. Cancer Res.
14 2623–30
Spencer V A, Xu R and Bissell M J 2007 Extracellular matrix,
nuclear and chromatin structure, and gene expression in normal
tissues and malignant tumors: a work in progress Adv. Cancer
Res. 97 275–94
Stein A M, Demuth T, Mobley D, Berens M and Sander L M 2007
A mathematical model of glioblastoma tumor spheroid
invasion in a three-dimensional in vitro experiment Biophys. J.
92 356–65
Stéphanou A, McDougall S R, Anderson A R A and Chaplain M A J
2005 Mathematical modelling of flow in 2D and 3D vascular
networks: applications to anti-angiogenic and
chemotherapeutic drug strategies Math. Comput. Modelling
41 1137–56
Stephanou A, Meskaoui G, Vailhe B and Tracqui P 2007 The
rigidity in fibrin gels as a contributing factor to the dynamics of
in vitro vascular cord formation Microvasc. Res. 73 182–90
Sternlicht M D and Werb Z 2001 How matrix metalloproteinases
regulate cell behavior Annu. Rev. Cell Dev. Biol. 17 463–516
Sugimoto H, Mundel T M, Kieran M W and Kalluri R 2006
Identification of fibroblast heterogeneity in the tumor
microenvironment Cancer Biol. Ther. 5 1640–6
Suresh S 2007 Biomechanics and biophysics of cancer cells Acta
Biomater. 3 413–38
Sutherland R M 1988 Cell and environment interactions in tumor
microregions: the multicell spheroid model Science
240 177–84
Swanson K R, Alvord E C Jr and Murray J D 2000 A quantitative
model for differential motility of gliomas in grey and white
matter Cell Prolif. 33 317–29
Swanson K R, Bridge C, Murray J D and Alvord E C Jr 2003 Virtual
and real brain tumors: using mathematical modeling to
quantify glioma growth and invasion J. Neurol. Sci. 216 1–10
Swanson K R, Rostomily R C and Alvord E C Jr 2008a A
mathematical modelling tool for predicting survival of
individual patients following resection of glioblastoma: a proof
of principle Br. J. Cancer 98 113–9
Swanson K R, Harpold H L, Peacock D L, Rockne R, Pennington C,
Kilbride L, Grant R, Wardlaw J M and Alvord E C Jr 2008b
Velocity of radial expansion of contrast-enhancing gliomas and
the effectiveness of radiotherapy in individual patients: a proof
of principle Clin. Oncol. (R. Coll. Radiol.) 20 301–8
Swietach P, Wigfield S, Cobden P, Supuran C T, Harris A L and
Vaughan-Jones R D 2008 Tumor-associated carbonic anhydrase
9 spatially coordinates intracellular pH in three-dimensional
multicellular growths J. Biol. Chem. 283 20473–83
Taber L A and Chabert S 2002 Theoretical and experimental study
of growth and remodeling in the developing heart Biomech.
Modelling Mechanobiol. 1 29–43
Tan J L, Tien J, Pirone D M, Gray D S, Bhadriraju K and Chen C S
2003 Cells lying on a bed of microneedles: an approach to
isolate mechanical force Proc. Natl Acad. Sci. USA 100 1484–9
Tilghman R W and Parsons J T 2008 Focal adhesion kinase as a
regulator of cell tension in the progression of cancer Semin.
Cancer Biol. 18 45–52
Rep. Prog. Phys. 72 (2009) 056701
Tlsty T D 2001 Stromal cells can contribute oncogenic signals
Semin. Cancer Biol. 11 97–104
Tlsty T D and Coussens L M 2006 Tumor stroma and regulation of
cancer development Annu. Rev. Pathol. 1 119–50
Tosin A, Ambrosi D and Preziosi L 2006 Mechanics and chemotaxis
in the morphogenesis of vascular networks Bull. Math. Biol.
68 1819–36
Tozern A and Skalak R 1988 Interaction of stress and growth in a
fibrous tissue J. Theor. Biol. 130 337–50
Tracqui P 2006 Mechanical instabilities as a central issue for in
silico analysis of cell dynamics Proc. IEEE 94 710–24
Tracqui P 1995 From passive diffusion to active cellular migration
in mathematical models of tumour invasion Acta Biotheor.
43 443–64
Tracqui P, Cruywagen G C, Woodward D E, Bartoo G T, Murray J D
and Alvord E C Jr 1995 A mathematical model of glioma
growth: the effect of chemotherapy on spatio-temporal growth.
Cell Prolif. 28 17–31
Tracqui P and Mendjeli M 1999 Modelling three-dimensional
growth of brain tumours from time series of scans Math.
Models Methods Appl. Sci. 9 581–98
Turing A M 1990 The chemical basis of morphogenesis 1953 Bull.
Math. Biol. 52 153–97 (discussion 19–52)
Tymchenko N, Wallentin J, Petronis S, Bjursten L M, Kasemo B and
Gold J 2007 A novel cell force sensor for quantification of
traction during cell spreading and contact guidance Biophys. J.
93 335–45
Tzvetkova-Chevolleau T, Stephanou A, Fuard D, Ohayon J,
Schiavone P and Tracqui P 2008 The motility of normal and
cancer cells in response to the combined influence of the
substrate rigidity and anisotropic microstructure Biomaterials
29 1541–51
Vailhe B, Vittet D and Feige J J 2001 In vitro models of
vasculogenesis and angiogenesis Lab. Invest. 81 439–52
Vailhe B, Ronot X, Tracqui P, Usson Y and Tranqui L 1997 In vitro
angiogenesis is modulated by the mechanical properties of
fibrin gels and is related to alpha(v)beta3 integrin localization
In Vitro Cell Dev. Biol. Anim. 33 763–73
Venkatesh S K, Yin M, Glockner J F, Takahashi N, Araoz P A,
Talwalkar J A and Ehman R L 2008 MR elastography of liver
tumors: preliminary results AJR Am. J. Roentgenol.
190 1534–40
30
P Tracqui
Vernon R B, Angello J C, Iruela-Arispe M L, Lane T F and
Sage E H 1992 Reorganization of basement membrane
matrices by cellular traction promotes the formation of cellular
networks in vitro Lab. Invest. 66 536–47
Vincent T L and Gatenby R A 2008 An evolutionary model for
initiation, promotion, and progression in carcinogenesis Int. J.
Oncol. 32 729–37
Volokh K Y 2006 Stresses in growing soft tissues Acta Biomater.
2 493–504
Waddington C H 1968 The basic ideas of biology Towards a
Theoretical Biology ed C H Waddington (Edinburgh:
Edinburgh University Press) pp 1–41
Wang J H, Thampatty B P, Lin J S and Im H J 2007
Mechanoregulation of gene expression in fibroblasts Gene
391 1–15
Wang N, Tytell J D and Ingber D E 2009 Mechanotransduction at a
distance: mechanically coupling the extracellular matrix with
the nucleus Nat. Rev. Mol. Cell Biol. 10 75–82
Weigelt B and Bissell M J 2008 Unraveling the microenvironmental
influences on the normal mammary gland and breast cancer
Semin. Cancer Biol. 18 311–21
Weiner O D 2002 Regulation of cell polarity during eukaryotic
chemotaxis: the chemotactic compass Curr. Opin. Cell Biol.
14 196–202
Welter M, Bartha K and Rieger H 2008 Emergent vascular network
inhomogeneities and resulting blood flow patterns in a growing
tumor J. Theor. Biol. 250 257–80
Wolf K, Wu Y I, Liu Y, Geiger J, Tam E, Overall C, Stack M S and
Friedl P 2007 Multi-step pericellular proteolysis controls the
transition from individual to collective cancer cell invasion Nat.
Cell. Biol. 9 893–904
Woodward D E, Cook J, Tracqui P, Cruywagen G C, Murray J D and
Alvord E C Jr 1996 A mathematical model of glioma growth:
the effect of extent of surgical resection Cell Prolif. 29 269–88
Zaman M H, Trapani L M, Sieminski A L, Mackellar D, Gong H,
Kamm R D, Wells A, Lauffenburger D A and Matsudaira P
2006 Migration of tumor cells in 3D matrices is governed by
matrix stiffness along with cell–matrix adhesion and
proteolysis Proc. Natl Acad. Sci. USA 103 10889–94
Zeisberg E M, Potenta S, Xie L, Zeisberg M and Kalluri R 2007
Discovery of endothelial to mesenchymal transition as a source
for carcinoma-associated fibroblasts Cancer Res. 67 10123–8