Open Access

Simulating non-small cell lung cancer with a multiscale agent-based model

  • Zhihui Wang1,
  • Le Zhang1,
  • Jonathan Sagotsky1 and
  • Thomas S Deisboeck1Email author
Theoretical Biology and Medical Modelling20074:50

https://doi.org/10.1186/1742-4682-4-50

Received: 12 June 2007

Accepted: 21 December 2007

Published: 21 December 2007

Abstract

Background

The epidermal growth factor receptor (EGFR) is frequently overexpressed in many cancers, including non-small cell lung cancer (NSCLC). In silico modeling is considered to be an increasingly promising tool to add useful insights into the dynamics of the EGFR signal transduction pathway. However, most of the previous modeling work focused on the molecular or the cellular level only, neglecting the crucial feedback between these scales as well as the interaction with the heterogeneous biochemical microenvironment.

Results

We developed a multiscale model for investigating expansion dynamics of NSCLC within a two-dimensional in silico microenvironment. At the molecular level, a specific EGFR-ERK intracellular signal transduction pathway was implemented. Dynamical alterations of these molecules were used to trigger phenotypic changes at the cellular level. Examining the relationship between extrinsic ligand concentrations, intrinsic molecular profiles and microscopic patterns, the results confirmed that increasing the amount of available growth factor leads to a spatially more aggressive cancer system. Moreover, for the cell closest to nutrient abundance, a phase-transition emerges where a minimal increase in extrinsic ligand abolishes the proliferative phenotype altogether.

Conclusion

Our in silico results indicate that in NSCLC, in the presence of a strong extrinsic chemotactic stimulus (and depending on the cell's location) downstream EGFR-ERK signaling may be processed more efficiently, thereby yielding a migration-dominant cell phenotype and overall, an accelerated spatio-temporal expansion rate.

Background

Non-small cell lung cancer (NSCLC) remains at the top of the list of cancer-related deaths in the United States [1]. The epidermal growth factor receptor (EGFR) is frequently overexpressed in NSCLC [2, 3]. Binding of epidermal growth factor (EGF) or transforming growth factor alpha (TGFα) to the extracellular domain of EGFR produces a number of downstream effects that affect phenotypic cell behavior including proliferation, invasion, metastasis, and inhibition of apoptosis [4]. In particular, increasing the expression of these growth factors leads to EGFR hyperactivity [5, 6], thus increases tumor cell motility and invasiveness, and finally enhances lung metastasis [7, 8]. Since approximately 90% of all cancer deaths originate from the spread of primary tumor cells into the surrounding tissue [9], quantitative measurements of the relationship between the level of the growth factors and the resulting tumor expansion is crucial – all the more so, since EGFR has emerged as an attractive therapeutic target for patients with advanced NSCLC [10].

A number of EGFR-related intracellular signal transduction pathways have been studied [1116], including NSCLC [17], and corresponding computational models at the molecular-level have been developed. These quantitative works mainly focused on signal-response relationships between the binding of EGF to EGFR and the activation of downstream proteins in the signaling cascade. With these in silico approaches, experimentally testable hypotheses can be made on signaling events controlling divergent cellular responses such as cell proliferation, differentiation, or apoptosis [18, 19]. However, most signaling works did not yet consider the cellular level (see [20, 21] for a review), and, conversely, only a few recent EGF/EGFR-mediated cellular-level models have started to incorporate a simple molecular level in studying e.g., cell migration in breast cancer [22], cell proliferation [23], and autocrine receptor-ligand dynamics [24, 25]. We argue that a more detailed understanding of a complex cancer system requires integrating both molecular- and cellular-level works to properly examine multicellular dynamics. To our knowledge, to date, no multiscale model of NSCLC has been developed or published.

Our group has been developing multiscale models to investigate highly malignant brain tumors as complex dynamic and self-organizing biosystems. Since this NSCLC model builds on these works, we will briefly review some milestones. First, an agent-based model for studying the spatio-temporal expansion of virtual glioma cells in a two-dimensional (2D) environment was built and the relationship between rapid growth and extensive tissue infiltration was investigated [26, 27]. This 'micro-macro' framework was then extended 'top-down' by incorporating an EGFR molecular interaction network [28] so that molecular dynamics at the protein level could be related to multi-cellular tumor growth patterns [29]. Most recently, an explicit cell cycle description was implemented to study in more detail brain tumor growth dynamics in a three-dimensional (3D) context [30]. These previous works have provided a computational paradigm in which biological processes have been successfully simulated from the molecular scale up to the cellular level and beyond. This progress led us to test the platform's applicability to and flexibility for other cancer types as well.

In this paper, we have therefore extended these previous modeling works to the case of NSCLC. Necessary modifications include at the molecular level the implementation of a NSCLC-specific EGFR-ERK signal transduction pathway. A novel, data-driven switch that is operated by two key molecules, i.e. phospholipase Cγ (PLCγ) and extracellular signal-regulated kinase (ERK), processes the phenotypic decision at the cellular level. The aim of this in silico work is to provide insights into the externally triggered molecular-level dynamics that govern phenotypic changes and thus impact multicellular patterns in NSCLC. In the following sections, we will first show the detailed design of the model before we present and then discuss the simulation results.

Model

Molecular Signaling Pathway

The kinetic model of the implemented NSCLC-specific molecular signaling pathway, which consists of 20 molecules, is shown in Fig. 1. These proteins, including both receptor (EGFR) and non-receptor kinases (e.g., PLCγ and protein kinase C (PKC) [31, 32], Raf, mitogen-activated protein kinase kinase (MEK), and ERK [3335]), have been experimentally or clinically proven to play an important role in NSCLC tumorigenesis. Although in reality these molecules fulfil their functions by interacting with a multitude of other molecular species from many distinct pathways [36, 37], we choose to start with these proteins not only because of their significance in the case of NSCLC but also since most of their kinetic parameters can be found in the literature. Also, it is reasonable to reduce the number of involved molecules as a starting point for modeling [38]. Amongst these proteins, both PLCγ and ERK are of particular interest for determining the cell's phenotypic changes as we will detail below.
Figure 1

Kinetic model of the NSCLC-specific EGFR signaling pathway. The arrows represent the reactions specified in Tables 1 and 2.

Kinetic equations are written in terms of concentrations and the reaction rates are functions of concentrations. The association and dissociation steps are characterized by first-order and second-order rate constants, respectively. We note that, although in reality chemical reactions of second or higher order are two-step processes, they are usually treated as a one-step process in mathematical modeling [39]. Our model is based on a total of 20 ordinary differential equations (ODEs) and uses exactly the same modeling techniques as other pathway analysis studies (see [11, 12] for detailed definitions). For simplicity, the ODEs for different molecules were calculated by Eq. (1):
d ( X i ) d t = v production v Consumption MathType@MTEF@5@5@+=feaagaart1ev2aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xI8qiVKYPFjYdHaVhbbf9v8qqaqFr0xc9vqFj0dXdbba91qpepeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaqcfa4aaSaaaeaacqWGKbazcqGGOaakcqWGybawdaWgaaqaaiabbMgaPbqabaGaeiykaKcabaGaemizaqMaemiDaqhaaOGaeyypa0ZaaabqaeaacqWG2bGDdaWgaaWcbaGaeeiCaaNaeeOCaiNaee4Ba8MaeeizaqMaeeyDauNaee4yamMaeeiDaqNaeeyAaKMaee4Ba8MaeeOBa4gabeaaaeqabeqdcqGHris5aOGaeyOeI0YaaabqaeaacqWG2bGDdaWgaaWcbaGaee4qamKaee4Ba8MaeeOBa4Maee4CamNaeeyDauNaeeyBa0MaeeiCaaNaeeiDaqNaeeyAaKMaee4Ba8MaeeOBa4gabeaaaeqabeqdcqGHris5aaaa@5B71@
(1)
where X i represents one of these 20 molecular pathway components. In Eq. (1), the change in concentration of molecule X i is the result of the reaction rates producing X i minus the reaction rates consuming it. Each biochemical reaction is then characterized by v i (see Fig. 1) with forward and reverse rate constants. Tables 1 and 2 summarize the kinetic parameters and the ODEs used for the model.
Table 1

Kinetic equations and initial concentrations. See Table 2 for references.

Reactant

Molecular variable

Initial concentration [nM]

ODE

X 1

EGF

to be varied

d(X1)/dt = -v1

X 2

EGFR

80

d(X2)/dt = -v1

X 3

EGF-EGFR

0

d(X3)/dt = v1 - 2v2

X 4

(EGF-EGFR)2

0

d(X4)/dt = v2 + v4 - v3

X 5

EGF-EGFR-P

0

d(X5)/dt = v3 + v7 - v4 - v5

X 6

PLCγ

10

d(X6)/dt = v8 - v5

X 7

EGF-EGFR-PLCγ

0

d(X7)/dt = v5 - v6

X 8

EGF-EGFR-PLCγ-P

0

d(X8)/dt = v6 - v7

X 9

PLCγ-P

0

d(X9)/dt = v7 - v8 - v9 - v10

X 10

PLCγ-P-I

0

d(X10)/dt = v9

X 11

PKC

10

d(X11)/dt = -v10

X 12

PKC*

0

d(X12)/dt = v10 - v11

X 13

Raf

100

d(X13)/dt = -v11

X 14

Raf*

0

d(X14)/dt = v11 - v12 - v14

X 15

MEK

120

d(X15)/dt = v13 - v12

X 16

MEK-P

0

d(X16)/dt = v12 + v15 - v13 - v14

X 17

MEK-PP

0

d(X17)/dt = v14 - v15 - v16 - v18

X 18

ERK

100

d(X18)/dt = v17 - v16

X 19

ERK-P

0

d(X19)/dt = v16 + v19 - v17 - v18

X 20

ERK-PP

0

d(X20)/dt = v18 - v19

- PKC* and Raf * indicate the activated form of PKC and Raf, respectively.

Table 2

Kinetic parameters. Concentrations and the Michaelis-Menten constants (K4, K8, and K11-K19) are given in [nM]. First- and second-order rate constants are given in [s-1] and [nM-1·s-1], respectively. V4, V8, and V11-V19 are expressed in [nM·s-1].

Reaction number

Equation

Kinetic parameter

 

Reference

v1

k1·X1·X2 - k-1·X3

k1 = 0.003

k-1 = 0.06

[11]

v2

k2·X3·X3 - k-2·X4

k2 = 0.01

k-2 = 0.1

[11]

v3

k3·X4 - k-3·X5

k3 = 1

k-3 = 0.01

[11]

v4

V4·X5/(K4 + X5)

V4 = 450

K4 = 50

[11]

v5

k5·X5·X6 - k-5·X7

k5 = 0.06

k-5 = 0.2

[11]

v6

k6·X7 - k-6·X8

k6 = 1

k-6 = 0.05

[11]

v7

k7·X8 - k-7·X5·X9

k7 = 0.3

k-7 = 0.006

[11]

v8

V8·X9/(K8 + X9)

V8 = 1

K8 = 100

[11]

v9

k9·X9 - k-9·X10

k9 = 1

k-9 = 0.03

[11]

v10

k10·X9·X11 - k-10·X12

k10 = 0.214

k-10 = 5.25

Estimate

v11

V11·X12·X13/(K11 + X13)

V11 = 4

K11 = 64

[39]

v12

V12·X14·X15/[K12·(1 + X16/K14) + X15]

V12 = 3.5

K12 = 317

[14]

v13

V13·X16/[K13·(1 + X17/K15) + X16]

V13 = 0.058

K13 = 2200

[12]

v14

V14·X14·X16/[K14·(1 + X15/K12) + X16]

V14 = 2.9

K14 = 317

[12]

v15

V15·X17/[K15·(1 + X16/K13) + X17]

V15 = 0.058

K15 = 60

[12]

v16

V16·X17·X18/[K16·(1 + X19/K18) + X18]

V16 = 9.5

K16 = 1.46 × 105

[12]

v17

V17·X19/[K17·(1 + X20/K19) + X19]

V17 = 0.3

K17 = 160

[12]

v18

V18·X17·X19/[K18·(1 + X18/K16) + X19]

V18 = 16

K18 = 1.46 × 105

[12]

v19

V19·X20/[K19·(1 + X19/K17) + X20]

V19 = 0.27

K19 = 60

[12]

Micro-Environment

The 2D virtual micro-environment is made up of a discrete lattice consisting of a grid with 200 × 200 points (Fig. 2). We use p(i, j) to express each point in the lattice, where i and j indicate the integer location in Euclidean terms. One single, distant nutrient source (simulating a cross-sectional blood vessel) is located at p(150, 150). To start with, a number of M × N cells (in other words, an M-by-N matrix) are initialized in the center of the lattice (and this number can be set to meet different simulation purposes). Each grid point can be occupied by one cell only or remain empty at a time.
Figure 2

Two-dimensional virtual micro-environment. Depicted are the 200 × 200 lattice (left) with the position of the nutrient source, and the seed cells with assignment of the corner cell IDs (0, 6, 42, and 48).

Three external chemical cues are employed in the model: EGF, glucose and oxygen tension. As we have done in previous studies [29, 30], the nutrient source carries the highest value of these three diffusive cues, which implicates that it is the most attractive location for the chemotactically acting tumor cells. Then, by means of normal distribution, each grid point of the lattice is assigned a concentration profile of these three cues. The levels of these distributions are weighted by the distance, d ij , of a given grid point from the nutrient source. The distributions of these three cues are described by the following equations:
E G F i j = T m exp ( 2 d i j 2 / σ t 2 ) MathType@MTEF@5@5@+=feaagaart1ev2aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xI8qiVKYPFjYdHaVhbbf9v8qqaqFr0xc9vqFj0dXdbba91qpepeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaGaemyrauKaem4raCKaemOray0aaWbaaSqabeaacqWGPbqAcqWGQbGAaaGccqGH9aqpcqWGubavdaWgaaWcbaGaemyBa0gabeaakiabgwSixlGbcwgaLjabcIha4jabcchaWjabcIcaOiabgkHiTiabikdaYiabdsgaKnaaDaaaleaacqWGPbqAcqWGQbGAaeaacqaIYaGmaaGccqGGVaWliiGacqWFdpWCdaqhaaWcbaGaemiDaqhabaGaeGOmaidaaOGaeiykaKcaaa@4AAA@
(2)
G l u c o s e i j = G a + ( G m G a ) exp ( 2 d i j 2 / σ g 2 ) MathType@MTEF@5@5@+=feaagaart1ev2aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xI8qiVKYPFjYdHaVhbbf9v8qqaqFr0xc9vqFj0dXdbba91qpepeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaGaem4raCKaemiBaWMaemyDauNaem4yamMaem4Ba8Maem4CamNaemyzau2aaWbaaSqabeaacqWGPbqAcqWGQbGAaaGccqGH9aqpcqWGhbWrdaWgaaWcbaGaemyyaegabeaakiabgUcaRiabcIcaOiabdEeahnaaBaaaleaacqWGTbqBaeqaaOGaeyOeI0Iaem4raC0aaSbaaSqaaiabdggaHbqabaGccqGGPaqkcqGHflY1cyGGLbqzcqGG4baEcqGGWbaCcqGGOaakcqGHsislcqaIYaGmcqWGKbazdaqhaaWcbaGaemyAaKMaemOAaOgabaGaeGOmaidaaOGaei4la8ccciGae83Wdm3aa0baaSqaaiabdEgaNbqaaiabikdaYaaakiabcMcaPaaa@594B@
(3)
O x y g e n i j = O a + ( O m O a ) exp ( 2 d i j 2 / σ o 2 ) MathType@MTEF@5@5@+=feaagaart1ev2aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xI8qiVKYPFjYdHaVhbbf9v8qqaqFr0xc9vqFj0dXdbba91qpepeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaGaem4ta8KaemiEaGNaemyEaKNaem4zaCMaemyzauMaemOBa42aaWbaaSqabeaacqWGPbqAcqWGQbGAaaGccqGH9aqpcqWGpbWtdaWgaaWcbaGaemyyaegabeaakiabgUcaRiabcIcaOiabd+eapnaaBaaaleaacqWGTbqBaeqaaOGaeyOeI0Iaem4ta80aaSbaaSqaaiabdggaHbqabaGccqGGPaqkcqGHflY1cyGGLbqzcqGG4baEcqGGWbaCcqGGOaakcqGHsislcqaIYaGmcqWGKbazdaqhaaWcbaGaemyAaKMaemOAaOgabaGaeGOmaidaaOGaei4la8ccciGae83Wdm3aa0baaSqaaiabd+gaVbqaaiabikdaYaaakiabcMcaPaaa@5852@
(4)
Moreover, the three chemotactic cues continue to diffuse over the lattice throughout the entire process of a simulation with a fixed rate, using the following equation:
M i j t = D M 2 M i j , t = 1 , 2 , 3 , ... . MathType@MTEF@5@5@+=feaagaart1ev2aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xI8qiVKYPFjYdHaVhbbf9v8qqaqFr0xc9vqFj0dXdbba91qpepeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaqbaeqabeGaaaqaaKqbaoaalaaabaGaeyOaIyRaemyta00aaWbaaeqabaGaemyAaKMaemOAaOgaaaqaaiabgkGi2kabdsha0baakiabg2da9iabdseaenaaBaaaleaacqWGnbqtaeqaaOGaeyyXICTaey4bIe9aaWbaaSqabeaacqaIYaGmaaGccqWGnbqtdaahaaWcbeqaaiabdMgaPjabdQgaQbaakiabcYcaSaqaaiabbsha0jabg2da9iabigdaXiabcYcaSiabikdaYiabcYcaSiabiodaZiabcYcaSiabc6caUiabc6caUiabc6caUiabbccaGiabc6caUaaaaaa@4EAD@
(5)
where M represents one of the three external cues, and t represents a time step. The coefficients in Eqs. (2–5) are listed in Table 3 (see also [30] for more details). It is evident then that the closer a given location is to the nutrient source, the higher the levels of the three cues will be at this grid point. Glucose will be continuously taken up by cells to support their metabolism. Only the nutrient source, p(150, 150), is replenished at each time step while all other grid points are not. In addition, cells take up both their own EGF and that secreted by adjoining cells in our model, because cancer cells act in both autocrine and paracrine manner in consuming EGF [40, 41].
Table 3

Coefficients of distribution and diffusion of EGF, glucose and oxygen tension. Values are taken from the literature [72, 73].

Coefficient

Value

Units

Description

T m

2.56

Nm

Maximum concentration of EGF

G a

17.0

mM

Normal concentration of glucose

G m

57.0

mM

Maximum concentration of glucose

O a

0.0017

DC

Normal concentration of oxygen

O m

0.0025

DC

Maximum concentration of oxygen

D EGF

6.7 × 10-11

m2·s-1

Diffusion coefficient of EGF

D Glucose

5.18 × 10-11

m2·s-1

Diffusion coefficient of glucose

D Oxygen

8.0 × 10-9

m2·s-1

Diffusion coefficient of oxygen

Each cell encompasses a self-maintained molecular interaction network (shown in Fig. 1) and the simulation system records the molecular composite profile at every time step to determine the cell's phenotype for the next step. In between time steps, the chemical environment is being updated, including EGF and glucose concentration as well as oxygen tension (according to Eq. (5)). When the first cell reaches the nutrient source the simulation run is terminated.

Cellular Phenotype Decision

Four tumor cell phenotypes are considered in the model: proliferation, migration, quiescence and death. Cell death is triggered when the on site glucose concentration drops below 8 mM [42]. A cell turns quiescent when the on site glucose concentration is between 8 mM and 16 mM, when it does not meet conditions for migration or proliferation (see below), or when it cannot find an empty location to migrate to or proliferate into.

The most important two phenotypic traits for spatio-temporal expansion, i.e. migration and proliferation, are decided by evaluating the dynamics of the following critical intracellular molecules. (1) PLCγ is known to be involved in directing cell movement in response to EGF [4345]; PLCγ dynamics are accelerated during migration in cancer cells [46]. Therefore, in our model, the rate of change of PLCγ (ROCPLC) decides if a cell proceeds to migration or not. That is, if ROCPLC exceeds a certain set threshold, TPLC, the cell has the potential to migrate. (2) Similarly, the rate of change of ERK (ROCERK) decides if a cell proceeds with proliferation. ERK has been found experimentally to have a strong influence on cell proliferation [33, 47, 48], and transient activation of ERK with EGF leads to cell replication [49, 50]. If a cell decides to migrate or proliferate, it will search for an appropriate location to move to or for its offspring to reside in. Candidate locations are those grid points surrounding the cell. Implementing a cell surface receptor-mediated chemotactic evaluation, the most appropriate location is detected by using a 'search-precision' mechanism [27] according to:T ij = ψ·L ij + (1 - ψε ij

where T ij represents the perceived attractiveness of location p(i, j), L ij represents the result of an evaluation function for location p(i, j) (see [27] for the definition of L ij ), and ε~ N(μ, σ2) is an error term following a normal distribution with mean μ and variance σ2. ψ [0,1] denotes the search-precision parameter that for a given run is held constant for all cells. Briefly, for a given cell at a certain location, when ψ = 0 the cell performs a pure random walk, whereas when ψ = 1 the cell always selects the location with the highest glucose concentration. Based on previous results [27], we set ψ = 0.7 because this value tends to lead to the highest average velocity of the tumor's spatial expansion.

It is worth noting that even if ROCPLC or ROCERK exceed their corresponding thresholds, it does not necessarily have to lead to cell migration or proliferation. Rather, if nowhere else to go, the cell remains quiescent and continues to search for an empty location at the next time step.

Any cell in the process of changing its phenotype will fall into one of these four categories: (i) ROCPLC < TPLC and ROCERK < TERK; (ii) ROCPLC > TPLC and ROCERK < TERK; (iii) ROCPLC < TPLC and ROCERK > TERK; and (iv) ROCPLC > TPLC and ROCERK > TERK. Figure 3 lists these conditions and their phenotypic consequences, respectively. Following the first three cell decisions is straightforward; first, if a cell experiences condition (i) no phenotypic change results as both ROCPLC and ROCERK remain below their corresponding thresholds; however, if a cell faces condition (ii) the cell migrates because of ROCPLC exceeding its threshold while in the presence of (iii) the cell proliferates due to ROCERK exceeding its threshold. However, for (iv), and in the absence of any specific experimental data, i.e. for the case that both ROCPLC and ROCERK exceed their corresponding thresholds, we explored two hypotheses: 'rule A' yielding migration advantage (i.e., the cell decides to migrate) whereas 'rule B' resulting in a proliferation advantage (i.e., the cell decides to proliferate). For simplicity, decision rules for the first three conditions are referred to 'general rules', while rules A and B are referred to 'special rules' hereafter. In the following section, we will describe the corresponding simulation results.
Figure 3

Cell phenotypic decision algorithm. See text for more details.

Results

Our algorithm was implemented in C/C++. A total of 49 seed cells were initially set up in the center of the lattice, and these cells were arranged in a 7 × 7 square shape (i.e., M = 7 and N = 7, see Fig. 2 for the configuration of the seed cells). We defined cell IDs from 0 to 48 (left to right, bottom to top). To investigate cell expansion dynamics, we monitored all cells and recorded their molecular profiles at every time step. We are particularly interested in the following four boundary cells: Cell No 0 (bottom-left corner, farthest from the source), Cell No 6 (top-left corner), Cell No 42 (bottom-right corner), and Cell No 48 (top-right corner, closest to the source). Through the distinct micro-environmental conditions they face, these corner cells exemplify the impact of location on single cell behavior, while they however still grasp the nature of the entire system. As described before, both rules A and B were tested for each different simulation condition.

Multi-Cellular Dynamics

Figure 4 shows two simulation results for rules A and B, respectively. The simulations were conducted with a standard EGF concentration of 2.56 nM. Note that this concentration is derived from the literature [51, 52] and has been rescaled to fit our model as a benchmark starting point for further simulations. In the upper panel of Fig. 4(a) for rule A, tumor cells first display on site proliferation prior to exhibiting extensive migratory behavior towards the nutrient source. However, for rule B (lower panel), cells remain stationary proliferative throughout, thereby increasing the tumor radius yet without substantial mobility-driven spatial expansion. The run time for the latter case (rule B) was considerably longer than for rule A. Based on the criterion chosen for terminating the run, i.e. the first cell reaching the nutrient source, this result is somewhat expected since rule A favors migration whereas rule B promotes proliferation. This is further supported by analysis of the evolution of the various phenotypes and the change of [total] cell numbers (Fig. 4(b)). While both rules generate all three cell phenotypes (proliferation (dark blue), migration (red), and quiescence (green)), rule A (left panel) indeed appears to result in a cancer cell population that exhibits a larger migratory fraction than the one emerging through rule B (right panel) which, however, yields a larger portion of proliferative cells (light blue). It is thus not surprising that for rule B, the [total] cell population of the tumor system exceeds the one achieved through rule A by a factor of 10.
Figure 4

Multicellular tumor expansion dynamics. (a) Shows the multicellular patterns that emerge through rule A (upper panel) and rule B (lower panel), respectively. (b) Describes the numeric evolution (y-axis) of each cell phenotype as well as of the [total] cell population (light blue) over time (x-axis) for rule A (left panel) and rule B (right panel), respectively. Note: proliferative tumor cells are labeled in dark blue, migratory cells in red, quiescent cells in green and dead cells in grey.

Influence of Decision Rules on Phenotypic Changes

To better understand the significance of each rule for the tumor system, we have investigated its influence on generating the intended phenotype. Figure 5 shows the weight of rule A on migration (a), and that of rule B on proliferation (b). (The results are taken from the two simulation runs reported in Fig. 4). In Fig. 5(a), migrations derive from two sources: (1) general rule, i.e. [ROCPLC > TPLC and ROCERK < TERK] and (2) rule A; proliferations stem from one source only, i.e. if [ROCPLC < TPLC and ROCERK > TERK]. Rule A plays a more dominant role in triggering migrations than the general rule does, yet does not contribute to increasing proliferations. Likewise, rule B has influence on proliferation only (Fig. 5(b)) and it contributes more to inducing proliferations than the corresponding general rule does.
Figure 5

Weight of decision rules on changing cell phenotypes. Influence on changing cell migration (left panel) and proliferation (right panel) when following the corresponding rule (see Fig. 3). The dashed red line indicates rule A-mediated migrations in (a), while the dashed blue line denotes rule B-mediated proliferations in (b). Fitting curves in solid black are calculated using a standard linear least squares method. Slopes of the fitting curves are 1.40 cells/step in (a) and 0.03 cells/step in (b), respectively. Note: The drop of the dashed red line in the left panel of (a) is caused by the termination of the simulation when a cell reached the source (in this case, no further computation on remaining cells will be performed).

However, as documented in the linear least square fittings, the rate at which rule A causes an increase in migration exceeds by far the one by which rule B induces an increase in proliferation. This indicates that the influence of rule A on increasing migrations is more substantial than that of rule B on increasing proliferations. Being particularly interested in gaining insights into spatially aggressive tumors, we continue in the following with investigating the implications of rule A on microscopic and molecular level dynamics of the cancer system.

Phase-Transition at Molecular Level

To further investigate (for rule A) the relationship between EGF concentration and phenotypic changes we varied the extrinsic EGF concentration from the standard value of 2.65 × 1.0 nM to 2.65 × 50.0 nM by an incremental increase of 0.1 nM in each simulation. As a result of the model's underlying chemotactic search paradigm, expectedly a simulation under the condition of a higher extrinsic EGF concentration finished faster than that with a lower one. However, cells turn out not to exhibit completely homogeneous behavior.

Specifically, we focus on Cell No 48, the cell closest to the nutrient source, and report its corresponding molecular changes in Fig. 6. One can see that as the standard EGF concentration increases, the number of proliferations (blue) decreases gradually up to a phase transition between 2.65 × 31.1 and 2.65 × 31.2 nM. That is, if the standard EGF concentration is less than 2.65 × 31.1 nM, proliferation still occurs in this particular cell, but if the ligand concentration starts to exceed 2.65 × 31.2 nM, its proliferative trait entirely disappears. In the presence of nutrient abundance, a very minor increase in extrinsic EGF can apparently abolish the expression of a phenotype. Even more intriguing, although the subcellular concentration change appears to be rather similar with regards to its patterns, on a closer look, the peak maxima of the rate changes for PLCγ and the turning point of the rate changes for ERK occur at an earlier time point for increasing EGF concentrations. This finding suggests that in the presence of excess ligand, the here implemented intracellular network switches to a more efficient signal processing mode. We note that for cell IDs 0, 6, and 42, no such phase transition emerged (data not shown) hence further supporting that this behavior is concentration dependent, and that geography, i.e. a cell's position relative to nutrient abundance, matters. Confirming the robustness of our finding for Cell No 48 we note that this cell continued to experience a phase transition when the coordinates of the center of the initial 49 cells was set randomly within a square region where p(100,100) is the lower left corner and p(110,110) is the upper right corner (5 runs, data not shown).
Figure 6

Changes at the molecular level for Cell No 48 with an increasing extrinsic EGF concentration (rule A). Four simulation runs are depicted where (from left to right) the EGF concentration increases from 2.65 × 1.0 to 2.65 × 31.1, 2.65 × 31.2, and finally, to 2.65 × 50.0 nM. (From top to bottom) plotted are the absolute change of PLCγ, rate of change of PLCγ, and rate of change of ERK. Note that the number of proliferations is decreasing gradually and finally disappears at a phase transition between the EGF concentrations of 2.65 × 31.1 and 2.65 × 31.2 nM. (For phenotype labeling see Fig. 4).

Discussion & Future Works

While using mathematical models to investigate the behavior of signaling networks is hardly new, understanding a complex biosystem, such as a tumor, by focusing on the analysis of its molecular or cellular level separately or exclusively is insufficient, particularly if it excludes the interaction with the surrounding tissue. Recent analyses of signaling pathways in mammalian systems have revealed that highly connected sub-cellular networks generate signals in a context dependent manner [53]. That is, biological processes take place in heterogeneous and highly structured environments [54] and such extrinsic conditions alone can induce the transformation of cells independent of genetic mutations as has been shown for the case of melanoma [55]. Taken together, modeling of cancer systems requires the analysis and use of signaling pathways in a simulated cancer environment (context) across different spatial-temporal scales.

Our group has been focusing on the development of such multiscale models for studying highly malignant brain tumors [27, 29, 30, 56]. Here, on the basis of these previous works, we presented a 2D multiscale agent-based model to simulate NSCLC. Specifically, we monitored how, dependent on microenvironmental stimuli, molecular profiles dynamically change, and how they affect a single NSCLC cell's phenotype and, eventually, the resulting multicellular patterns.

Proceeding top-down in our analysis, we first evaluated the multicellular readout of molecular 'decision' rules A and B (versus general rules; Fig. 3). The patterns of a more stationary, concentrically growing cancer system (following rule B) are quite different from the rapid, chemotactically-guided, spatial expansion that can be seen in the tumor regulated by rule A (Fig. 4(a)). Not surprisingly, the latter also operates with many more migratory albeit overall less [total] cells (Fig. 4(b)). Furthermore, examining in more detail the influence of the two distinct rules on their respective phenotypic yield, we found that the impact of rule A on increasing cell migration is more substantial than rule B's influence on furthering proliferation (Fig. 5). This finding suggests that the migratory rule A can operate the cancer system through incrementally smaller changes (while the simulation system is more robust for rule B). Such sensitivity to migratory cues corresponds well with experimental data on the response of human breast cancer cells, which showed that a spatially successful expansive system reacts rather quickly to even miniscule changes in chemotactic directionality [57, 58].

Continuing therefore with rule A, our effort was then geared to gain insights into tumor expansion dynamics not only with regards to extrinsic stimuli but also to cell geography, i.e. a cell's location relative to the replenished nutrient source. Most interestingly, we found a phase transition in the cancer cell closest to the nutrient source (i.e. Cell No 48, while none of the other three corner cells showed similar behavior). Specifically, for a tumor cell at this location, i.e., facing nutrient abundance, proliferation is completely abolished once the extrinsic EGF concentration exceeds a certain level. While this at first may seems rather unexpected, this finding however only confirms the experimentally sound notion that EGF stimulates the spatial expansion of a cancer system [58]. Moreover, with increasing EGF concentrations, the maxima of ROCPLC (Fig. 6) gradually occur earlier which seems to indicate that, under these conditions, the downstream signal is processed faster. Interestingly, such a 'no proliferation, just migration' behavior in the presence of chemo-attractant has indeed already been reported in several in vitro studies using a variety of cancer cell lines [59, 60] as well as in non-cancerous human cells [61]. (While admittedly, for the reasons stated, rule B did not receive similar attention in our analysis), we nonetheless argue that, on the basis of our results and the experimental reports they seem to correspond with, rule A and thus a migratory decision prompted by a [ROCPLC > TPLC and ROCERK > TERK] condition is a reasonable outcome for the signaling process taking place in NSCLC also in vitro and in vivo.

However, moving the model closer to reality will require a multitude of adjustments, one of which is its ability to account for up- or down-regulation in key molecules as a result of tumorigenesis. As a first step, and since experimental data on over-expression of EGFR in a variety of cancer types, including NSCLC, are ample [6265] we have begun to simulate the impact of an increasing number of receptors on the cancer system (Fig. 7; simulations conducted with an EGFR concentration of 800 nM (per system)). Comparing this preliminary data with those reported in Fig. 6 (simulations conducted with an EGFR concentration of 80 nM (per system)), we find that an EGFR-overexpressing NSCLC tumor seems to operate with even more migration and does so earlier on. The result is a spatially even more aggressive cancer system, which seems to correspond well with the aforementioned experimental studies. And, intriguingly, while the phase transition itself is preserved, it however occurs already at a smaller EGF concentration, hence indicating that the increase in receptor density leads to an amplification of the downstream signal, which again corresponds well with experimental results in examining signaling activities generated by different EGFR family members [66]. Taken together, while preliminary, this finding demonstrates applicability and confirms flexibility of this multiscale platform, hence warrants its further expansion.
Figure 7

Changes at the molecular level for Cell No 48 with an increasing extrinsic EGF concentration (rule A), at an EGFR concentration of 800 nM. Three simulation runs are depicted where (from left to right) the EGF concentration increases from 2.65 × 1.0 to 2.65 × 5.9 and 2.65 × 6.0 nM. (From top to bottom) plotted are the absolute change of PLCγ, rate of change of PLCγ, and rate of change of ERK. Note that a phase transition emerges again between the EGF concentrations of 2.65 × 5.9 and 2.65 × 6.0 nM, hence at a lower concentration compared to the one depicted in Fig. 6 (EGFR concentration of 80 nM). In the two simulations around the phase transition, the maximum rates of change for both PLCγ and ERK (i.e., ROCPLC and ROCERK at 2.65 × 5.9 and at 2.65 × 6.0 nM) are lower compared with those in Fig. 6 (i.e., ROCPLC and ROCERK at 2.65 × 31.1 and at 2.65 × 31.2 nM). (For phenotype labeling see Fig. 4).

There are a number of research tracks that can and should be pursued in future works. First, it will be intriguing to see if, in the presence of a non-replenished nutrient source, the proliferative phenotype eventually can be recovered once extrinsic ligand concentrations fall beyond the phase-transition threshold. More generally, while most of the pathway's parameters, including rate constants and initial component concentrations were obtained from the experimental literature, this data naturally originated from a variety of often stationary experimental settings and different cell types. It therefore represents a less desirable and reliable input than time series data that come from one experimental setting only. Also, some parameters had to be estimated, much like in other well-established pathway models [11, 12]. Taken together, future works will have to include not only proper experimental verification of the estimated parameters and evaluation of the simulation results but also, on the in silico side, techniques such as sensitivity analysis to help determine the effects of parameter uncertainties on model outcome [67] and to identify control points for experiment design [68]. While a pathway model cannot be a biological representation in every detail [38] we plan on adding, in incremental steps, other pathways of relevance for NSCLC such as e.g. PI3K/PTEN/AKT [69]. Moreover, simulating a more heterogeneous biochemical environment and implementing both cell-cell and cell-matrix interactions [70] are planned steps at the cellular level that should help representing the cancer system of interest in more detail.

Regardless, we believe that the current model already provides useful insights into NSCLC from a systematic view in terms of quantitatively understanding the relationship between extrinsic chemotactic stimuli, the underlying properties of signaling networks, and the cellular biological responses they trigger. Our results yield several experimentally testable hypotheses and thus further support the use of multiscale models in interdisciplinary cancer research. To our knowledge, this presents the first multiscale computational model of Non-Small Cell Lung Cancer and is thus potentially a significant first step towards realizing a fully validated in silico model for this devastating disease.

Abbreviations

EGF: 

Epidermal growth factor

EGFR: 

EGF receptor

ERK: 

Extracellular signal-regulated kinase

MAPK: 

Mitogen-activated protein kinase

MEK: 

MAPK kinase

NSCLC: 

Non-small cell lung cancer

PLC: 

Phospholipase Cγ

PKC: 

Protein kinase C

TGF: 

Transforming growth factor

Declarations

Acknowledgements

This work has been supported in part by NIH grant CA 113004 (The Center for the Development of a Virtual Tumor, CViT [71]) and by the Harvard-MIT (HST) Athinoula A. Martinos Center for Biomedical Imaging and the Department of Radiology at Massachusetts General Hospital. We would like to acknowledge helpful discussions with Drs. Raju Kucherlapati, Victoria Joshi and David Sarracino from the Harvard-Partners Center for Genetics and Genomics (HPCGG).

Authors’ Affiliations

(1)
Complex Biosystems Modeling Laboratory, Harvard-MIT (HST) Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital

References

  1. Sharma SV, Bell DW, Settleman J, Haber DA: Epidermal growth factor receptor mutations in lung cancer. Nat Rev Cancer. 2007, 7: 169-181. 10.1038/nrc2088.View ArticlePubMedGoogle Scholar
  2. Hackel PO, Zwick E, Prenzel N, Ullrich A: Epidermal growth factor receptors: critical mediators of multiple receptor pathways. Curr Opin Cell Biol. 1999, 11: 184-189. 10.1016/S0955-0674(99)80024-6.View ArticlePubMedGoogle Scholar
  3. Hirsch FR, Varella-Garcia M, Bunn PA, Di Maria MV, Veve R, Bremmes RM, Baron AE, Zeng C, Franklin WA: Epidermal growth factor receptor in non-small-cell lung carcinomas: correlation between gene copy number and protein expression and impact on prognosis. J Clin Oncol. 2003, 21: 3798-3807. 10.1200/JCO.2003.11.069.View ArticlePubMedGoogle Scholar
  4. Ciardiello F, De Vita F, Orditura M, Tortora G: The role of EGFR inhibitors in nonsmall cell lung cancer. Curr Opin Oncol. 2004, 16: 130-135. 10.1097/00001622-200403000-00008.View ArticlePubMedGoogle Scholar
  5. Putnam EA, Yen N, Gallick GE, Steck PA, Fang K, Akpakip B, Gazdar AF, Roth JA: Autocrine growth stimulation by transforming growth factor-alpha in human non-small cell lung cancer. Surg Oncol. 1992, 1: 49-60. 10.1016/0960-7404(92)90056-Q.View ArticlePubMedGoogle Scholar
  6. Rusch V, Baselga J, Cordon-Cardo C, Orazem J, Zaman M, Hoda S, McIntosh J, Kurie J, Dmitrovsky E: Differential expression of the epidermal growth factor receptor and its ligands in primary non-small cell lung cancers and adjacent benign lung. Cancer Res. 1993, 53: 2379-2385.PubMedGoogle Scholar
  7. Price JT, Wilson HM, Haites NE: Epidermal growth factor (EGF) increases the in vitro invasion, motility and adhesion interactions of the primary renal carcinoma cell line, A704. Eur J Cancer. 1996, 32A: 1977-1982. 10.1016/0959-8049(96)00207-9.View ArticlePubMedGoogle Scholar
  8. Xue C, Wyckoff J, Liang F, Sidani M, Violini S, Tsai KL, Zhang ZY, Sahai E, Condeelis J, Segall JE: Epidermal growth factor receptor overexpression results in increased tumor cell motility in vivo coordinately with enhanced intravasation and metastasis. Cancer Res. 2006, 66: 192-197. 10.1158/0008-5472.CAN-05-1242.View ArticlePubMedGoogle Scholar
  9. Christofori G: New signals from the invasive front. Nature. 2006, 441: 444-450. 10.1038/nature04872.View ArticlePubMedGoogle Scholar
  10. Silvestri GA, Rivera MP: Targeted therapy for the treatment of advanced non-small cell lung cancer: a review of the epidermal growth factor receptor antagonists. Chest. 2005, 128: 3975-3984. 10.1378/chest.128.6.3975.View ArticlePubMedGoogle Scholar
  11. Kholodenko BN, Demin OV, Moehren G, Hoek JB: Quantification of short term signaling by the epidermal growth factor receptor. J Biol Chem. 1999, 274: 30169-30181. 10.1074/jbc.274.42.30169.View ArticlePubMedGoogle Scholar
  12. Schoeberl B, Eichler-Jonsson C, Gilles ED, Muller G: Computational modeling of the dynamics of the MAP kinase cascade activated by surface and internalized EGF receptors. Nat Biotechnol. 2002, 20: 370-375. 10.1038/nbt0402-370.View ArticlePubMedGoogle Scholar
  13. Brightman FA, Fell DA: Differential feedback regulation of the MAPK cascade underlies the quantitative differences in EGF and NGF signalling in PC12 cells. FEBS Lett. 2000, 482: 169-174. 10.1016/S0014-5793(00)02037-8.View ArticlePubMedGoogle Scholar
  14. Hatakeyama M, Kimura S, Naka T, Kawasaki T, Yumoto N, Ichikawa M, Kim JH, Saito K, Saeki M, Shirouzu M, Yokoyama S, Konagaya A: A computational model on the modulation of mitogen-activated protein kinase (MAPK) and Akt pathways in heregulin-induced ErbB signalling. Biochem J. 2003, 373: 451-463. 10.1042/BJ20021824.PubMed CentralView ArticlePubMedGoogle Scholar
  15. Resat H, Ewald JA, Dixon DA, Wiley HS: An integrated model of epidermal growth factor receptor trafficking and signal transduction. Biophys J. 2003, 85: 730-743.PubMed CentralView ArticlePubMedGoogle Scholar
  16. Sasagawa S, Ozaki Y, Fujita K, Kuroda S: Prediction and validation of the distinct dynamics of transient and sustained ERK activation. Nat Cell Biol. 2005, 7: 365-373. 10.1038/ncb1233.View ArticlePubMedGoogle Scholar
  17. Hendriks BS, Griffiths GJ, Benson R, Kenyon D, Lazzara M, Swinton J, Beck S, Hickinson M, Beusmans JM, Lauffenburger D, de Graaf D: Decreased internalisation of erbB1 mutants in lung cancer is linked with a mechanism conferring sensitivity to gefitinib. Syst Biol (Stevenage). 2006, 153: 457-466.View ArticleGoogle Scholar
  18. Khalil IG, Hill C: Systems biology for cancer. Curr Opin Oncol. 2005, 17: 44-48. 10.1097/01.cco.0000150951.38222.16.View ArticlePubMedGoogle Scholar
  19. Spencer SL, Gerety RA, Pienta KJ, Forrest S: Modeling somatic evolution in tumorigenesis. PLoS Comput Biol. 2006, 2: e108-10.1371/journal.pcbi.0020108.PubMed CentralView ArticlePubMedGoogle Scholar
  20. Ma'ayan A, Blitzer RD, Iyengar R: Toward predictive models of mammalian cells. Annu Rev Biophys Biomol Struct. 2005, 34: 319-349. 10.1146/annurev.biophys.34.040204.144415.PubMed CentralView ArticlePubMedGoogle Scholar
  21. Chaplain MA, McDougall SR, Anderson AR: Mathematical modeling of tumor-induced angiogenesis. Annu Rev Biomed Eng. 2006, 8: 233-257. 10.1146/annurev.bioeng.8.061505.095807.View ArticlePubMedGoogle Scholar
  22. Wang W, Goswami S, Sahai E, Wyckoff JB, Segall JE, Condeelis JS: Tumor cells caught in the act of invading: their strategy for enhanced cell motility. Trends Cell Biol. 2005, 15: 138-145. 10.1016/j.tcb.2005.01.003.View ArticlePubMedGoogle Scholar
  23. Walker D, Wood S, Southgate J, Holcombe M, Smallwood R: An integrated agent-mathematical model of the effect of intercellular signalling via the epidermal growth factor receptor on cell proliferation. J Theor Biol. 2006, 242: 774-789. 10.1016/j.jtbi.2006.04.020.View ArticlePubMedGoogle Scholar
  24. Monine MI, Berezhkovskii AM, Joslin EJ, Wiley HS, Lauffenburger DA, Shvartsman SY: Ligand accumulation in autocrine cell cultures. Biophys J. 2005, 88: 2384-2390. 10.1529/biophysj.104.051425.PubMed CentralView ArticlePubMedGoogle Scholar
  25. Maly IV, Wiley HS, Lauffenburger DA: Self-organization of polarized cell signaling via autocrine circuits: computational model analysis. Biophys J. 2004, 86: 10-22.PubMed CentralView ArticlePubMedGoogle Scholar
  26. Mansury Y, Kimura M, Lobo J, Deisboeck TS: Emerging patterns in tumor systems: simulating the dynamics of multicellular clusters with an agent-based spatial agglomeration model. J Theor Biol. 2002, 219: 343-370. 10.1006/jtbi.2002.3131.View ArticlePubMedGoogle Scholar
  27. Mansury Y, Deisboeck TS: The impact of "search precision" in an agent-based tumor model. J Theor Biol. 2003, 224: 325-337. 10.1016/S0022-5193(03)00169-3.View ArticlePubMedGoogle Scholar
  28. Athale C, Mansury Y, Deisboeck TS: Simulating the impact of a molecular 'decision-process' on cellular phenotype and multicellular patterns in brain tumors. J Theor Biol. 2005, 233: 469-481. 10.1016/j.jtbi.2004.10.019.View ArticlePubMedGoogle Scholar
  29. Athale CA, Deisboeck TS: The effects of EGF-receptor density on multiscale tumor growth patterns. J Theor Biol. 2006, 238: 771-779. 10.1016/j.jtbi.2005.06.029.View ArticlePubMedGoogle Scholar
  30. Zhang L, Athale CA, Deisboeck TS: Development of a three-dimensional multiscale agent-based tumor model: simulating gene-protein interaction profiles, cell phenotypes and multicellular patterns in brain cancer. J Theor Biol. 2007, 244: 96-107. 10.1016/j.jtbi.2006.06.034.View ArticlePubMedGoogle Scholar
  31. Mai H, May WS, Gao F, Jin Z, Deng X: A functional role for nicotine in Bcl2 phosphorylation and suppression of apoptosis. J Biol Chem. 2003, 278: 1886-1891. 10.1074/jbc.M209044200.View ArticlePubMedGoogle Scholar
  32. Lahn M, Su C, Li S, Chedid M, Hanna KR, Graff JR, Sandusky GE, Ma D, Niyikiza C, Sundell KL, John WJ, Giordano TJ, Beer DG, Paterson BM, Su EW, Bumol TF: Expression levels of protein kinase C-alpha in non-small-cell lung cancer. Clin Lung Cancer. 2004, 6: 184-189.View ArticlePubMedGoogle Scholar
  33. Brognard J, Dennis PA: Variable apoptotic response of NSCLC cells to inhibition of the MEK/ERK pathway by small molecules or dominant negative mutants. Cell Death Differ. 2002, 9: 893-904. 10.1038/sj.cdd.4401054.View ArticlePubMedGoogle Scholar
  34. Raben D, Helfrich B, Bunn PA: Targeted therapies for non-small-cell lung cancer: biology, rationale, and preclinical results from a radiation oncology perspective. Int J Radiat Oncol Biol Phys. 2004, 59: 27-38. 10.1016/j.ijrobp.2004.01.054.View ArticlePubMedGoogle Scholar
  35. Hemstrom TH, Sandstrom M, Zhivotovsky B: Inhibitors of the PI3-kinase/Akt pathway induce mitotic catastrophe in non-small cell lung cancer cells. Int J Cancer. 2006, 119: 1028-1038. 10.1002/ijc.21927.View ArticlePubMedGoogle Scholar
  36. Adjei AA, Hidalgo M: Intracellular signal transduction pathway proteins as targets for cancer therapy. J Clin Oncol. 2005, 23: 5386-5403. 10.1200/JCO.2005.23.648.View ArticlePubMedGoogle Scholar
  37. Zochbauer-Muller S, Gazdar AF, Minna JD: Molecular pathogenesis of lung cancer. Annu Rev Physiol. 2002, 64: 681-708. 10.1146/annurev.physiol.64.081501.155828.View ArticlePubMedGoogle Scholar
  38. Aldridge BB, Burke JM, Lauffenburger DA, Sorger PK: Physicochemical modelling of cell signalling pathways. Nat Cell Biol. 2006, 8: 1195-1203. 10.1038/ncb1497.View ArticlePubMedGoogle Scholar
  39. Bhalla US, Iyengar R: Emergent properties of networks of biological signaling pathways. Science. 1999, 283: 381-387. 10.1126/science.283.5400.381.View ArticlePubMedGoogle Scholar
  40. Rusch V, Klimstra D, Venkatraman E, Pisters PW, Langenfeld J, Dmitrovsky E: Overexpression of the epidermal growth factor receptor and its ligand transforming growth factor alpha is frequent in resectable non-small cell lung cancer but does not predict tumor progression. Clin Cancer Res. 1997, 3: 515-522.PubMedGoogle Scholar
  41. Shvartsman SY, Wiley HS, Deen WM, Lauffenburger DA: Spatial range of autocrine signaling: modeling and computational analysis. Biophys J. 2001, 81: 1854-1867.PubMed CentralView ArticlePubMedGoogle Scholar
  42. Freyer JP, Sutherland RM: Regulation of growth saturation and development of necrosis in EMT6/Ro multicellular spheroids by the glucose and oxygen supply. Cancer Res. 1986, 46: 3504-3512.PubMedGoogle Scholar
  43. Mouneimne G, Soon L, DesMarais V, Sidani M, Song X, Yip SC, Ghosh M, Eddy R, Backer JM, Condeelis J: Phospholipase C and cofilin are required for carcinoma cell directionality in response to EGF stimulation. J Cell Biol. 2004, 166: 697-708. 10.1083/jcb.200405156.PubMed CentralView ArticlePubMedGoogle Scholar
  44. Piccolo E, Innominato PF, Mariggio MA, Maffucci T, Iacobelli S, Falasca M: The mechanism involved in the regulation of phospholipase Cgamma1 activity in cell migration. Oncogene. 2002, 21: 6520-6529. 10.1038/sj.onc.1205821.View ArticlePubMedGoogle Scholar
  45. Friedl P, Wolf K: Tumour-cell invasion and migration: diversity and escape mechanisms. Nat Rev Cancer. 2003, 3: 362-374. 10.1038/nrc1075.View ArticlePubMedGoogle Scholar
  46. Dittmar T, Husemann A, Schewe Y, Nofer JR, Niggemann B, Zanker KS, Brandt BH: Induction of cancer cell migration by epidermal growth factor is initiated by specific phosphorylation of tyrosine 1248 of c-erbB-2 receptor via EGFR. Faseb J. 2002, 16: 1823-1825.PubMedGoogle Scholar
  47. Li M, Lee TW, Yim AP, Mok TS, Chen GG: Apoptosis induced by troglitazone is both peroxisome proliferator-activated receptor-gamma- and ERK-dependent in human non-small lung cancer cells. J Cell Physiol. 2006, 209: 428-438. 10.1002/jcp.20738.View ArticlePubMedGoogle Scholar
  48. Vicent S, Garayoa M, Lopez-Picazo JM, Lozano MD, Toledo G, Thunnissen FB, Manzano RG, Montuenga LM: Mitogen-activated protein kinase phosphatase-1 is overexpressed in non-small cell lung cancer and is an independent predictor of outcome in patients. Clin Cancer Res. 2004, 10: 3639-3649. 10.1158/1078-0432.CCR-03-0771.View ArticlePubMedGoogle Scholar
  49. Marshall CJ: Specificity of receptor tyrosine kinase signaling: transient versus sustained extracellular signal-regulated kinase activation. Cell. 1995, 80: 179-185. 10.1016/0092-8674(95)90401-8.View ArticlePubMedGoogle Scholar
  50. Santos SD, Verveer PJ, Bastiaens PI: Growth factor-induced MAPK network topology shapes Erk response determining PC-12 cell fate. Nat Cell Biol. 2007, 9: 324-330. 10.1038/ncb1543.View ArticlePubMedGoogle Scholar
  51. Fang K: An enhanced and sensitive autocrine stimulation by transforming growth factor-alpha is acquired in the brain metastatic variant of a human non-small-cell lung cancer cell line. Br J Cancer. 1996, 74: 1776-1782.PubMed CentralView ArticlePubMedGoogle Scholar
  52. Kurachi H, Morishige K, Amemiya K, Adachi H, Hirota K, Miyake A, Tanizawa O: Importance of transforming growth factor alpha/epidermal growth factor receptor autocrine growth mechanism in an ovarian cancer cell line in vivo. Cancer Res. 1991, 51: 5956-5959.PubMedGoogle Scholar
  53. Jordan JD, Landau EM, Iyengar R: Signaling networks: the origins of cellular multitasking. Cell. 2000, 103: 193-200. 10.1016/S0092-8674(00)00112-4.PubMed CentralView ArticlePubMedGoogle Scholar
  54. Di Ventura B, Lemerle C, Michalodimitrakis K, Serrano L: From in vivo to in silico biology and back. Nature. 2006, 443: 527-533. 10.1038/nature05127.View ArticlePubMedGoogle Scholar
  55. Postovit LM, Seftor EA, Seftor RE, Hendrix MJ: Influence of the microenvironment on melanoma cell fate determination and phenotype. Cancer Res. 2006, 66: 7833-7836. 10.1158/0008-5472.CAN-06-0731.View ArticlePubMedGoogle Scholar
  56. Deisboeck TS, Berens ME, Kansal AR, Torquato S, Stemmer-Rachamimov AO, Chiocca EA: Pattern of self-organization in tumour systems: complex growth dynamics in a novel brain tumour spheroid model. Cell Prolif. 2001, 34: 115-134. 10.1046/j.1365-2184.2001.00202.x.View ArticlePubMedGoogle Scholar
  57. Segall JE, Tyerech S, Boselli L, Masseling S, Helft J, Chan A, Jones J, Condeelis J: EGF stimulates lamellipod extension in metastatic mammary adenocarcinoma cells by an actin-dependent mechanism. Clin Exp Metastasis. 1996, 14: 61-72. 10.1007/BF00157687.View ArticlePubMedGoogle Scholar
  58. Wang SJ, Saadi W, Lin F, Minh-Canh Nguyen C, Li Jeon N: Differential effects of EGF gradient profiles on MDA-MB-231 breast cancer cell chemotaxis. Exp Cell Res. 2004, 300: 180-189. 10.1016/j.yexcr.2004.06.030.View ArticlePubMedGoogle Scholar
  59. Kim S, Li Q, Dang CV, Lee LA: Induction of ribosomal genes and hepatocyte hypertrophy by adenovirus-mediated expression of c-Myc in vivo. Proc Natl Acad Sci USA. 2000, 97: 11198-11202. 10.1073/pnas.200372597.PubMed CentralView ArticlePubMedGoogle Scholar
  60. Wroblewski LE, Noble PJ, Pagliocca A, Pritchard DM, Hart CA, Campbell F, Dodson AR, Dockray GJ, Varro A: Stimulation of MMP-7 (matrilysin) by Helicobacter pylori in human gastric epithelial cells: role in epithelial cell migration. J Cell Sci. 2003, 116: 3017-3026. 10.1242/jcs.00518.View ArticlePubMedGoogle Scholar
  61. Wang L, Zeng H, Wang P, Soker S, Mukhopadhyay D: Neuropilin-1-mediated vascular permeability factor/vascular endothelial growth factor-dependent endothelial cell migration. J Biol Chem. 2003, 278: 48848-48860. 10.1074/jbc.M310047200.View ArticlePubMedGoogle Scholar
  62. Creighton CJ, Hilger AM, Murthy S, Rae JM, Chinnaiyan AM, El-Ashry D: Activation of mitogen-activated protein kinase in estrogen receptor alpha-positive breast cancer cells in vitro induces an in vivo molecular phenotype of estrogen receptor alpha-negative human breast tumors. Cancer Res. 2006, 66: 3903-3911. 10.1158/0008-5472.CAN-05-4363.View ArticlePubMedGoogle Scholar
  63. El Sheikh SS, Domin J, Abel P, Stamp G, Lalani el N: Phosphorylation of both EGFR and ErbB2 is a reliable predictor of prostate cancer cell proliferation in response to EGF. Neoplasia. 2004, 6: 846-853. 10.1593/neo.04379.View ArticlePubMedGoogle Scholar
  64. Onn A, Choe DH, Herbst RS, Correa AM, Munden RF, Truong MT, Vaporciyan AA, Isobe T, Gilcrease MZ, Marom EM: Tumor cavitation in stage I non-small cell lung cancer: epidermal growth factor receptor expression and prediction of poor outcome. Radiology. 2005, 237: 342-347. 10.1148/radiol.2371041650.View ArticlePubMedGoogle Scholar
  65. Raspollini MR, Castiglione F, Garbini F, Villanucci A, Amunni G, Baroni G, Boddi V, Taddei GL: Correlation of epidermal growth factor receptor expression with tumor microdensity vessels and with vascular endothelial growth factor expression in ovarian carcinoma. Int J Surg Pathol. 2005, 13: 135-142. 10.1177/106689690501300202.View ArticlePubMedGoogle Scholar
  66. Hendriks BS, Orr G, Wells A, Wiley HS, Lauffenburger DA: Parsing ERK activation reveals quantitatively equivalent contributions from epidermal growth factor receptor and HER2 in human mammary epithelial cells. J Biol Chem. 2005, 280: 6157-6169. 10.1074/jbc.M410491200.View ArticlePubMedGoogle Scholar
  67. El-Samad H, Kurata H, Doyle JC, Gross CA, Khammash M: Surviving heat shock: control strategies for robustness and performance. Proc Natl Acad Sci USA. 2005, 102: 2736-2741. 10.1073/pnas.0403510102.PubMed CentralView ArticlePubMedGoogle Scholar
  68. Bentele M, Lavrik I, Ulrich M, Stosser S, Heermann DW, Kalthoff H, Krammer PH, Eils R: Mathematical modeling reveals threshold mechanism in CD95-induced apoptosis. J Cell Biol. 2004, 166: 839-851. 10.1083/jcb.200404158.PubMed CentralView ArticlePubMedGoogle Scholar
  69. Vivanco I, Sawyers CL: The phosphatidylinositol 3-Kinase AKT pathway in human cancer. Nat Rev Cancer. 2002, 2: 489-501. 10.1038/nrc839.View ArticlePubMedGoogle Scholar
  70. Griffith LG, Swartz MA: Capturing complex 3D tissue physiology in vitro. Nat Rev Mol Cell Biol. 2006, 7: 211-224. 10.1038/nrm1858.View ArticlePubMedGoogle Scholar
  71. The Center for the Development of a Virtual Tumor, CViT.http://www.cvit.org
  72. Moskal TL, Huang S, Ellis LM, Fritsche HA, Chakrabarty S: Serum levels of transforming growth factor alpha in gastrointestinal cancer patients. Cancer Epidemiol Biomarkers Prev. 1995, 4: 127-131.PubMedGoogle Scholar
  73. Thorne RG, Hrabetova S, Nicholson C: Diffusion of epidermal growth factor in rat brain extracellular space measured by integrative optical imaging. J Neurophysiol. 2004, 92: 3471-3481. 10.1152/jn.00352.2004.View ArticlePubMedGoogle Scholar

Copyright

© Wang et al; licensee BioMed Central Ltd. 2007

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Advertisement