Towards an integrated systems-based modelling framework for drug transport and its effect on tumour cells
- Cong Liu^{1}Email author,
- Cong Krishnan^{1, 2} and
- Xiao Yun Xu^{1}
https://doi.org/10.1186/1754-1611-8-3
© Liu et al.; licensee BioMed Central Ltd. 2014
Received: 7 August 2013
Accepted: 26 December 2013
Published: 13 January 2014
Abstract
Background
A systematic understanding of chemotherapeutic influence on solid tumours is highly challenging and complex as it encompasses the interplay of phenomena occurring at multiple scales. It is desirable to have a multiscale systems framework capable of disentangling the individual roles of multiple contributing factors, such as transport and extracellular factors, and purely intracellular factors, as well as the interactions among these factors. Based on a recently developed systems-based modelling framework, we have developed a coupled system in order to further elucidate the role of drug transport, and its interplay with cellular signalling by incorporating intra- and extra-vascular drug transport in tumour, dynamic descriptions of intracellular signalling and tumour cell density dynamics.
Results
Different aspects of the interaction between transport and cell signalling and the effects of transport parameters have been investigated in silico. Limited drug penetration is found to be a major constraint in inducing drug effect; many aspects of the interaction of transport with cell signalling are independent of the details of cell signalling. A sensitivity analysis indicates that the effect of drug diffusivity depends on the balance between interstitial drug transport and the specific requirement for triggering apoptosis (governed by highly nonlinear signalling networks), suggesting that the effect of drug diffusivity in such cases must be considered in conjunction with descriptions of cellular dynamics.
Conclusions
The modelling framework developed in this study provides qualitative and mechanistic insights into the effect of drug on tumour cells. It provides an in silico experimental platform to investigate the interplay between extracellular factors (e.g. transport) and intracellular factors. Such a platform is essential to understanding the individual and combined effects of transport and cellular factors in solid tumour.
Keywords
Drug transport Drug effect Intracellular signalling in silico experimental platform Systems-based modelling frameworkIntroduction
The efficacy of chemotherapy is strongly dependent on the transport of anticancer drugs to tumour cells and their responses to the administrated drug [1]; both can be compromised significantly by complexities associated with cancer, which is a disease of collective dysregulations across multiple scales.
To exert therapeutic effects, anticancer drugs must reach tumour cells with a sufficiently high concentration [2]. Limited penetration is one of the major causes of failure of chemotherapy treatment of solid tumours. The first obstacle to most blood-borne chemotherapeutic agents is posed by the abnormal and chaotic tumour vasculature, which limits tumour blood flow and consequently the supply of drugs and nutrients [3]. After crossing the capillary wall, anticancer drugs must penetrate through the tumour interstitium, where drug distribution is determined by the effectiveness of drug transport by diffusion and convection as well as drug consumption [2]. Elevated interstitial fluid pressure in solid tumours hinders convective transport, rendering diffusion the dominant mechanism for interstitial drug transport [4]. Drug diffusivity depends strongly on the physicochemical properties of the specific drug, such as molecular weight, shape, charge and solubility. Drug consumption involves drug binding, sequestration and metabolism, which can be altered by microenvironmental conditions, such as extracellular matrix composition and structure, cell packing density and the presence of tumour acidity [3, 5, 6].
For most anticancer drugs, it is necessary for drug molecules to transport across cell membranes to reach the target molecules and interact with them, as a consequence, triggering cellular signal transduction. Cellular signalling is one of the important characteristics of every living cell in that it governs the basic cellular activities by perceiving and correctly responding to external/internal stimuli. Anticancer drugs as stress stimuli can regulate/trigger cell signalling to kill cells or affect their cellular responses (cell apoptosis, cell proliferation, differentiation and migration), which might be directly associated with fatal consequences of cancer [7–9]. The way by which cellular signalling functions is extremely complex as cellular pathways are not isolated from each other but are interconnected through a complex network [10]. In particular, highly non-linear input-output relationships are usually displayed in cell signalling networks, with a number of emergent properties, such as adaptive responses and robust switching by positive feedback [11, 12]. It is worth emphasizing that the dynamic interactions and signal transmission in these chemical networks control cellular decision making and cellular responses such as movement, apoptosis etc.
Considerable effort has been devoted to developing mathematical models to predict drug concentration and drug effect on solid tumours (for a review, see [13]). Compartmental models have been widely adopted for prediction of temporal profiles of drug concentration in designed compartments, particularly drug concentration in blood in pharmacokinetic studies [14–16]. To obtain spatio-temporal drug distributions in a given tumour geometry, it is necessary to explicitly account for drug transport (diffusion or/and convection) [17–19]. After drug concentrations are obtained, suitable pharmacodynamic models can be used to predict the effect of drug as a function of drug concentration and/or as a function of time following drug administration in a phenomenological and empirical manner, with elaboration of observed data thus neglecting detailed underlying mechanisms [20–22]. On the other hand, deterministic models can be used to describe the tumour response by assuming a drug concentration-dependent tumour growth characteristic or tumour death kinetics [23–25]. Unfortunately, mathematical models addressing the above mentioned areas have been developed separately; furthermore, they often bypass a key component that is the dynamic process of cellular signal transduction. While all these models are capable of providing certain levels of insights, none of them offers a transparent and integrated description of drug transport and drug effect accounting for the associated cellular signalling.
In this study, an integrated systems-based mathematical modelling framework is employed and extended, which captures the information flow from drug delivery to the outcome, thus including biological transport processes of drugs and cellular response and accounting for dynamics of the relevant signal transduction. This allows us to begin to probe and elucidate different aspects of the roles and the interaction of transport and intracellular signalling dynamics. In a spatially distributed system, intracellular signalling is triggered in response to heterogeneous drug stimuli delivered through transport pathways. It must be emphasized here that drug stimuli are dynamic (being delivered through complex vascular networks), while drug transport can be affected by many tissue level features and intracellular dynamics is highly nonlinear. Finally the dynamic coupling of these factors is not necessarily unidirectional. For instance the tissue scale properties and features could be a potential factor in affecting drug transport; as an example, it is found that apoptosis-inducing pretreatment enhances drug delivery [26, 27]. Furthermore, with considerable progress in unravelling the intracellular and intercellular signal transduction in systems biology, mathematical modelling approaches can begin to go beyond empirical and transport centric models to integrate dynamic descriptions of transport and cellular signalling for a systematic examination of their interactions.
This paper examines some basic aspects of the interaction of transport and cellular dynamics. As a first step, the model is formulated on a simplified geometry of tumour vasculature, in which explicit coupling of blood flow between vascular and interstitial space is incorporated, along with drug transport. The effects of anticancer drugs are addressed by integrating the above with dynamics of intracellular apoptosis signalling. The integrated model is used to evaluate treatment strategies and to analyse other factors that may influence the response of tumour cells, in order to provide insights into the complex interplay between the different processes involved.
Methods
For mathematical modelling of drug transport, a commonly adopted approach is to avoid an explicit representation of the tumour vasculature which, instead, is treated as a distributed source term in the governing equations [19, 28–31]. In doing so, descriptions of transport processes are incomplete without accounting for vascular transport and the spatial relationship between blood vessels and tumour interstitium. However, incorporating realistic tumour vasculature geometry is highly challenging, given the fact that the tumour vasculature is abnormal, irregular and heterogeneous. Further complexities in evaluating drug effects are added when dynamic intracellular signalling processes are incorporated, which are triggered in response to spatio-temporal drug stimuli and exhibit highly non-linear dynamics. To obtain clear-cut and transparent insights into transport mechanisms, cellular signalling and their interaction, we employ the modelling framework as an in silico experimental platform which describes a well-defined tumour-drug system with minimal essential elements, definite information flow and a controlled source of variability and heterogeneity. The in silico experimental platform depicts an idealized tumour with no heterogeneity, in a simplified geometry. This setup is designed as an initial effort to contain the minimal components necessary for understanding the effects of drugs on tumours and elucidating the effects of transport and cellular factors in a transparent manner without consideration of other factors.
Computational geometry
Mathematical models
Parameters and values used in the mathematical models
Parameter | Symbol | Value | Reference |
---|---|---|---|
Tumour blood flow | |||
Tumour capillary radius | R_{C} | 100 (μm) | [33] |
Tumour interstitium radius | R_{T} | 1200 (μm) | Estimated |
Tumour length | L | 1 (cm) | [33] |
Blood density | ρ | 1000 (kg/m^{3}) | [32] |
Blood viscosity | μ | 0.004 (Pa∙s) | [35] |
Vascular hydraulic conductivity | L _{ p } | 0.36-2.8 × 10^{-6} (cm/mmHg∙s) | [36] |
Tissue hydraulic conductivity | K | 0.4-2.0 × 10^{-7} (cm^{2}/mmHg∙s) | [36] |
Vascular inlet pressure | P _{ A } | 15 (mmHg) | [37] |
Vascular outlet pressure | P _{ V } | 0 | [36] |
Ambient pressure | P _{ 0 } | 0 | [36] |
Drug transport (Doxorubicin) | D _{ E } | 1.5778 × 10^{-6} (cm^{2}/s) | [17] |
DOX diffusion coefficient | |||
Rate of transmembrane transport | V_{ 1 }, V_{ 2 } | 0.28 (ng/(10^{5}cells)/min) | [14] |
(V_{1} = V_{2}) | |||
Diffusive permeability for DOX | P _{ E } | 2.778 × 10^{-4} (cm/s) | [17] |
Michaelis constant for transmembrane transport | k _{ 1 } | 0.219 (μg/ml) | [14] |
Michaelis constant for transmembrane transport | k _{ 2 } | 1.37 (ng/(10^{5}cells)) | [14] |
Tumour cell density | |||
Initial tumour cell density | c _{ t,0 } | 10^{6} (cells/mm^{3}) | [17] |
Tumour cell growth rate | a _{ 1 } | 0.5 (day^{-1}) | Estimated |
Saturation constant in logistic equation | b | 0.02592 (mm^{3}/(10^{5}cells)/day) | Estimated |
Tumour cell natural decay rate | a _{ 2 } | 0.24 (day^{-1}) | Estimated |
Intracellular signalling: | |||
Bistable switch | |||
Michaelis Menten constants | V _{ f } | 27 (hr^{-1}) | [38] |
Michaelis Menten constants | V _{ r } | 0.459 (hr^{-1}) | [38] |
Michaelis Menten constants | K _{ m1 } | 100 | [38] |
Michaelis Menten constants | K _{ m2 } | 0.01 | [38] |
Kinetic parameter mediating feedback strength | k _{ fb } | 2.927 (hr^{-1}) | [38] |
Basal level in the bistable switch | p | 0.7 | Estimated |
Parameter mediating input regulation in the bistable switch | q | 0.3(ng/(10^{5}cells))^{-1} | Estimated |
Monostable switch | |||
Kinetic parameter reflecting the time scale of the response | k | 0.432 (hr^{-1}) | Estimated |
Associated constant | k _{ h } | 1 (ng/(10^{5}cells)) | Estimated |
Hill coefficient | n | 10 | Estimated |
R_{1} protein activation rate | k _{ f } | 3.6 (hr^{-1}) | Estimated |
R_{1} protein degradation rate | k _{ r } | 0.144 (hr^{-1}) | Estimated |
R_{1} Threshold for apoptosis switch | R _{ 1 } , _{ th } | 0.9 | Estimated |
Variables and their definitions in the mathematical models
Symbol | Definition | Symbol | Definition |
---|---|---|---|
u _{ v } | Blood velocity vector in vascular domain | c _{ v } | DOX concentration in vascular domain |
u _{ i } | Blood velocity vector in interstitial domain | c _{ E } | Extracellular DOX concentration |
P _{ v } | Vascular blood pressure | c _{ I } | Intracellular DOX concentration |
P _{ i } | Interstitial fluid pressure | R | Hypothetical downstream intermediate protein in apoptosis signalling cascade |
J _{ F } | Transmural fluid velocity, determined by Starling’s law | R _{1} | Hypothetical protein responsible for triggering apoptosis |
J _{ s } | Transmural drug flux, determined by Kedem-Katchalsky equation | c _{ t } | Tumour cell density |
Tumour blood flow
The model accounts for the coupling between vascular, transmural and interstitial fluid flow since tumour blood vessels are highly permeable. Blood flow is assumed to be steady, which is acceptable here as it is capable of serving as a fundamental platform to investigate the dynamic behaviour of drug transport and intracellular events without imposing further complexities related to pulsatile blood flow.
Where ρ is blood density, μ is blood viscosity, u is blood velocity vector with subscript v denoting vascular space, and P_{ v } is vascular blood pressure.
where K represents interstitial hydraulic conductivity, and u_{i} and P_{ i } are blood velocity vector and fluid pressure in the interstitium.
Drug transport
Three variables are considered in the drug transport processes: the intravascular drug concentration (c_{ v }), the interstitial extracellular free drug concentration (c_{ E }), and intracellular drug concentration (c_{ I }). Within the blood vessel, the drug is transported mainly by convection with negligible reactions (e.g. metabolism, binding/unbinding, uptake). Once extravasated into the tumour interstitium, drug particles penetrate through the intersititum via diffusion and convection and at the same time they may be taken up by tumour cells. The drug is treated as a blood-borne solute and its transport is governed by a diffusion-convection-reaction equation.
BC (8a) prescribes a pulse injection at the vessel inlet with a constant intensity S and infusion time T, in which Heaviside term H() indicates infusion occurs during the period of t = 0 to t = T. BC (8b) defines a convective flux at the outlet; BC (8c) sets an outward solute flux across the leaky wall boundary.
Solute dynamics in the interstitium (Ω_{ i }) is governed by the extracellular and intracellular drug transport.
Where P is drug diffusive permeability across the vessel wall, σ_{ f } is osmotic reflection coefficient, Δc_{ lm } is the log-mean concentration across the vessel wall, and J_{ F } is the fluid flux across the vessel wall, which is determined by Starling’s law.
Intracellular drug concentration
It is noted that drug binding to plasma proteins (mainly albumin) is neglected in the current study. In many cases, the binding/unbinding process is described simply by first-order kinetics; therefore its effect is relatively simple in that it reduces the amount of free extracellular drug available to tumour cells. Incorporation of drug binding would not alter the system output qualitatively, but would dramatically increase computational burden. However, it is essential to take this into account when quantitative modelling is required or sophisticated binding mechanisms are involved.
Intracellular apoptosis signalling
The modelling of intracellular signalling processes, a substantial core of systems biology, plays a very important role in the entire modelling framework as decisions on cell fate are processed and determined by molecular signalling networks. It is thus very important to have a dynamic representation of this in the modelling framework. Given the complexity of signalling which contains many missing biological details and unknown parameters, it is essential to choose an appropriate level of description in the model to start with, so that the most important known aspects of the signalling and cellular decision making are included. Thus, the strategy adopted here is to start with coarse grained descriptions of the cellular signalling dynamics, which are capable of representing correctly the nature of the information flow, and ensuring that the qualitatively important features of detailed models are accounted for. We believe that this is more appropriate than starting with detailed models containing many unknown factors and other details whose correctness may be difficult to establish. At the same time, this provides a platform that allows more detailed mechanistic models to be incorporated in the future.
When modelling the intracellular processes in response to chemotherapy, the main process of interest is apoptosis (programmed cell death) following the administration of anticancer agents. Based on systems biology investigations and existing models [39–45], we recognize that two key features must be reflected by any model regardless of its complexity. Firstly, some threshold effect must be present; secondly, the “switch” to apoptosis must be realized in an irreversible way.
The apoptosis models adopted here are based on the two types of switches commonly observed in cellular signalling: bistable and monostable apoptosis switches. As bistable switches can exhibit irreversibility intrinsically, they have been used in modelling irreversible cell fate decision-making in apoptosis [40, 43, 45]. Generally, positive feedback (biologically existing in the intracellular caspase network) and cooperativity (in apoptosome formation) are regarded as sources of bistability. However, as far as apoptosis is concerned, it is not obvious whether representing the irreversible fate (cell death) as a steady state is necessary or even reasonable (at the cell fate decision level). It is possible that an irreversible decision is made when critical cellular events are triggered, from which there is no turning back. Thus the irreversibility could result from a simple irreversible reaction, which is kept under tight control and triggered only under very special circumstances [39, 41, 42]. Noting this, and the fact that both mechanisms have been discussed in the context of apoptosis, the two models employed in this study are: (1) a bistable switch with self-contained threshold behaviour and irreversibility; (2) a sequential interconnection of a monostable switch and a downstream irreversible reaction effect [34].
Bistable switch
Where K_{ m1 } and K_{ m2 } are the Michaelis-Menten parameters. k_{ fb } is a kinetic parameter which parametrizes the feedback strength. The constants p and q serve to set the basal level and dynamic range of the module.
Irreversible monostable switch
Where n denotes the Hill coefficient, k_{ h } an associated constant in the Hill term, and k is a parameter representing the time scale of the response.
Once a threshold of R_{ 1 } is crossed, cell death is triggered. Normalised concentrations of molecules R and R_{ 1 } are adopted. Two additional points are worth emphasizing here. Firstly, the actual choice of monostable/bistable model has a very minor effect, as analysis with other model variants has led to very similar results. Secondly, detailed characterization and comparison of monostable and bistable models demonstrate their similarities and differences, which provide a basis for their use in contexts such as this [46].
Tumour cell density dynamics
Where a_{ 1 } is tumour growth rate, a_{ 2 } is tumour natural decay rate and b is a saturation constant in the logistic tumour growth equation.
Eqn. 15 naturally describes the growth and death of cells with saturating growth effect leading to a finite steady state (models with some variations have very similar net results: see [46]). These equations provide an explicit representation of the key features of interest. Although the population balance formalism provides a more comprehensive description of birth and death of cells, it is computationally highly demanding and can be difficult to handle if additional cellular complexities are included.
When tumour cells are perturbed by anticancer drugs, the intracellular apoptosis signalling is initiated, resulting in cell death at the population level. Since cell density is described in a continuous, rather than discrete form, triggering of the intracellular threshold is represented by a sharp fall in growth rate at the population level. Clearly, the dynamics of the logistic model imply that if the growth rate becomes sufficiently low, the zero steady state (c_{ t } = 0) would be the only biologically relevant state, indicating that all cells will eventually die. It should be noted that computational implementation of this threshold effect is achieved in a reversible way in the bistable model, but in a unidirectional (irreversible) way in the monostable model.
Initial conditions
Except for tumour cell density, all other variables are set to be zero initially. A uniform tumour cell density is assumed prior to drug injection.
Numerical methods
Computational procedures
All numerical simulations presented in this paper are implemented in the finite element based software Comsol Multiphysics. The simulation of blood flow is decoupled from that of drug transport and tumour cell density by assuming the velocity field is independent of the drug concentration field and tumour cell density distribution. Steady-state simulation of blood flow is carried out first. Upon obtaining the pressure and velocity fields, drug transport is resolved by solving the diffusion-convection-reaction equation. The boundary conditions at the vessel wall are implemented in accordance with the physical settings in Comsol Multiphysics. With regard to fluid flow, transmural velocity (J_{ F }) is positive in both the vessel and interstitial domain as it points away from the surface in the vessel domain, and at the same time, it represents the inflow to the interstitial domain. J_{ F } is set as a variable in accordance with Starling’s law, which enables the coupling of the vascular fluid pressure (P_{ v }) to the interstitial fluid pressure (P_{ i }). For drug transport, an inflow flux is set by default, which means a negative transmural flux (J_{ S }) in the vessel domain, and a positive flux in the interstitial domain. Intracellular signal transduction is triggered by the local intracellular drug concentration, the response of which is manifested through tumour cell density owing to decreased tumour growth rate or increased tumour death rate. The tumour cell density, in turn, affects drug transport.
The equations are discretised and solved on a pre-generated computational mesh. Mesh sensitivity study is carried out first to provide mesh independent solutions. The final mesh consists of 75,000 and 325,000 mapped meshes for drug transport and tumour cell density, respectively.
Model parameters
Values of all parameters as well as variables used in the integrated model are defined in Tables 1 and 2, respectively. These are extracted from a variety of sources as they span multiple scales of description and are not available in a single tumour-drug system. Values for blood flow related parameters are mainly extracted from similar mathematical models found in the literature. Doxorubicin (DOX) is one of the anticancer drugs commonly used in clinics and a large amount of experimental data is available on its physical and pharmacokinetics properties; therefore it is chosen as a representative anticancer drug for parameterization purpose. Values for tumour growth parameters are chosen to reflect both the range of steady state, as well as the appropriate time scale. In the intracellular dynamics, parameters are not generally available and their values are determined based on appropriate reflection of the time scales involved in apoptosis signalling. Further, the downstream threshold (R_{ 1,th }) is specially chosen to ensure complete R_{ 1 } activation while maintaining the upstream signal for a sufficient time period. This in turn directly reflects the cellular behaviour to exposure of drugs for a sufficient time. Thus our intracellular descriptions are parameterized so that they are capable of triggering apoptosis decisions for comparable signals as are seen experimentally. It should be emphasized that most of the essential conclusions drawn from this study are not strongly dependent on the particular numerical choice of parameters.
Results
In this section, numerical results of blood flow are presented first, followed by results of drug transport and distribution. The effect of drug examined in terms of tumour cell density distribution by considering both bistable and irreversible monostable intracellular apoptosis models under various pulse drug injections. In addition, a sensitivity analysis is performed on parameters involved in drug transport. Results for drug concentration and tumour cell density are presented in the dimensionless form, which are normalised by their corresponding reference values: 0.001 μg/mm^{3} for vascular and extracellular drug concentrations, 1 ng/10^{5}cells for intracellular drug concentration and 10^{6} cells/mm^{3} for tumour cell density.
Blood flow
Drug transport and distribution
Drug transport
where L is the characteristic length, u is a representative velocity, and D is the diffusion coefficient for a given anticancer drug. Since D is 1.578 × 10^{-10} m^{2}/s for doxorubicin (Table 1), the corresponding Pe in the blood vessel is on the order of 10^{6}, suggesting that intravascular drug transport is dominated by convection.
Drugs extravasate into the tumour interstitium by diffusion and convection, determined by diffusive permeability P and blood filtration velocity J_{ F }. The filtration velocity shown in Figure 3(b) is of the order of 10^{-7} m/s, which is an order of magnitude lower than diffusive permeability (2.78 × 10^{-6} m/s) across the vessel wall for doxorubicin, suggesting that the total transmural flux is dominated by the diffusive flux in this case.
Drug distribution
Displayed in Figure 5(b,d,f) are snapshots of vascular, extracellular and intracellular drug concentrations at t = 2 h, half an hour after drug injection. In response to the sudden termination of drug input, reversal of concentration gradient is observed in the near wall region inside the blood vessel and in the interstitium. In this context, the interstitium acts as a reservoir, from which drugs are transported back to the blood vessel and eventually leave the blood vessel by convection. The reverse transport of drugs is confined to a thin layer close to the vessel wall, while drugs outside this layer are transported outward in the radial direction by diffusion and convection. Therefore, the extracellular drug concentration profile experiences a rise and reaches a peak before falling off; the same is observed for the intracellular drug concentration profile.
Drug effect--Tumour cell density distribution
Baseline case
Effects of pulse infusion time and intensity
The predicted tumour cell density shown here has a weaker non-linear response to increases in infusion time or pulse intensity compared to our previous study [34]. Although the in silico experiments in both studies use the same basic modelling elements, they differ in the geometrical scale of interest and the level of integration between different elements; the former may significantly affect the role of drug transport and intracellular signalling in cell killing. In our previous study [34], the time scale for transport (diffusion) in a microscopic tumour cord geometry is of the order of seconds, and the results suggest that transport over such a small time scale plays a less important role, which results in a more obvious nonlinear response. However, in the present in silico simulations, a relatively large tumour interstitium is adopted to mimic drug penetration through the entire tumour tissue. Within this context, the time scales for drug transport (dominated by diffusion) and intracellular signalling are comparable; therefore, the transport limitation poses a major constraint in inducing the effect of anticancer drugs since reactions involved in cellular signalling are triggered by intracellular drug concentration. It is concluded that limitations in transport can act as a buffer to reduce the sensitivity of cell-killing region to changes in the characteristics of stimuli. This is further demonstrated through the sensitivity analysis on the size of tumour domain (presented in sensitivity analysis section).
Effects of pulse fractionations
Sensitivity analysis
Parameters used in the mathematical models are either related to drug transport or involved in drug effect. With regard to drug transport, parameters can be further divided into three groups: (i) diffusion related parameters, namely drug diffusivity (D), diffusive permeability (P), (ii) convection related parameters, such as hydraulic permeability (L_{ p }), hydraulic conductivity in the interstitium (K), (iii) tumour geometric parameters, for example, the radius of blood vessel (R_{C}) and tumour interstitium (R_{T}).
Based on the analysis presented above (in drug transport section), it is deduced that convection plays a minor role in the transmural and interstitial drug transport. Since convection is dependent on hydraulic conductivity L_{ p } and tissue hydraulic conductivity K, their effects on blood flow are examined, and both are found to have a marginal effect on enhancing the transmural velocity (J_{ F }) (results not shown here).
Parameters involved in the intracellular signalling models are estimated to reflect a time scale and threshold value in a reasonable range, which agree qualitatively with those obtained from a cascade of signal transduction. Overall, transparent effects of these parameters are observed as expected from the coarse-grained intracellular signalling models. A slower kinetic rate or an elevated threshold would make it more difficult to trigger apoptosis while in the opposite scenario, relieving the constraints of apoptosis may exert a further effect on improving interstitial drug transport. As the study is oriented towards an integrative understanding of drug effect with account for mechanistic drug action, the intracellular signalling models together with the estimated parameter values adopted in the present study are sufficient to serve the purpose.
Therefore, the sensitivity study presented here is focused on how interstitial drug transport may be perturbed by altering diffusion related parameters, namely drug diffusivity and diffusive permeability, and geometric parameter (the size of tumour interstitium).
Effect of drug diffusivity
Effect of diffusive permeability
Effect of the size of tumour interstitium
Discussion
In this study, the first steps have been taken towards developing an in silico experimental platform with integration of blood flow, drug transport (both vascular and interstitial) and cellular signalling dynamics to provide an overall framework to systematically evaluate the effect of anticancer drugs on tumour cells. This platform, described by our model, contains the minimal essential elements to understand the basic aspects of the interplay between drug transport and cellular effects. The model set-up represents an idealized tumour with a well controlled environment, which is essential for understanding the interactions of many complicating factors. The tumour cell density (the output) is determined by the combined action of drug input (determined by tumour blood flow and drug transport) and intracellular signalling (involving non-linear dynamics), when subject to a particular form of drug stimulus (the input), which is itself determined by the mode of delivery (not modelled in detail). Basic descriptions of the processes involved are incorporated with an aim to create a model system which captures the information flow from drug delivery to causal effect, and is capable of serving as a platform to understand the interplay between transport and extracellular factors on one hand and cellular features on the other hand.
Drug transport
Rather than treating the tumour blood as a homogeneous source term in the transport equation, the models describe tumour blood flow explicitly by coupling vascular blood flow with the interstitial fluid flow through an elevated tumour vascular hydraulic conductivity. Drug delivery is subsequently examined following the same transport routes. From the simulation results, few differences can be observed between the drug concentration profiles at different axial locations, indicating that intravascular transport (convection dominant) is not a rate limiting step in the drug transport processes. Of course it must be borne in mind that this finding is based on the highly simplified geometry adopted in this study – a single vessel similar to the Krogh model. It is acknowledged that a single vessel model is not always the best model; it fails to capture the heterogeneous distribution of functional blood vessels in tumours that result in large avascular regions. However, the present study focuses on providing essential and clear-cut insights into drug distribution and drug effect by conceptualizing the information flow in an integrated manner. Potentially important factors, such as the complex tumour vascular geometry (or even non-Newtonian blood rheological properties) may significantly increase resistance to tumour blood flow, thus leading to insufficient drug delivery through the tumour vasculature and even more heterogeneous drug distribution in the tumour interstitium. A description of more realistic tumour vascular network, based on advanced imaging techniques, could potentially be incorporated into the modelling framework for more realistic prediction of drug concentration distribution in specific tumours. We also note that to bring new insights into drug delivery through a complex vascular network in the tumour, it is necessary to first understand what the effect is in a simpler vasculature.
The higher vascular permeability normally observed in tumour tissues facilitates transmural transport of drugs into the tumour interstitium. However, it is noted that drug penetration is restricted to the region close to the blood vessel, with poor drug distribution in the interstitium. With regard to drug transport into tumour cells, in the case of the specific anticancer drug selected (doxorubicin), it is found that the drug is preferentially sequestered in tumour cells. Therefore, limited drug penetration in the interstitium is a major difficulty to overcome in order to improve drug efficacy. Compared to drug diffusion in the interstitium, drug consumption by tumour cells determined by the tumour cell density plays a dominant role in impairing interstitial penetration (discussed later). On the other hand, the heterogeneous drug distribution found in the spatially distributed system also implies that homogeneous compartment models may not be sufficient for accurate predictions of drug efficacy. It is worth pointing out that the modelling framework developed is not only applicable to chemotherapeutic agents, but (with minor modifications) also to oxygen and other endocrine signals, which share the same transport pathways. Further, interstitial drug transport may be implicitly affected by cellular responses, such as the presence of low pH due to the adaptive metabolism towards oxygen starvation, and detoxifying anticancer drugs due to acquirement of drug resistances. Understanding the interplay of such factors is beyond the scope of this study, but can be built on the existing framework presented here.
Drug effect
Drug response is evaluated broadly by empirical or mechanistic approaches. This study is motivated by the desire to provide a mechanistic understanding of drug effect by taking into account the interactions between drugs and targets, the downstream signalling reactions and ultimately cell fate decision making. In the current model, apoptosis is the primary cellular response triggered by anticancer drugs and dynamic analysis of drug effect is performed at both cellular and tissue level.
It is important to choose an appropriate level of cellular effects to be included in the model given the fact that many of the biochemical details are still unknown (or may be questionable or may vary between cell types). Coarse grained descriptions of intracellular process are adopted here, with a view to qualitatively capture the nature of signal transduction in the cell and to retain the relevant input-output signalling characteristics. Future work to include more biochemical details of the signalling network (including the detailed dynamics of the caspase network and its regulation) is needed in order to obtain a more detailed depiction of the cellular signalling and understand the roles of multiple intracellular regulatory mechanisms.
In the current work, two types of apoptosis modules with qualitatively distinct dynamics of signal transduction are included and examined separately: a bistable switch and an irreversible monostable switch. The two models are examined to address whether such different dynamic characteristics at the cellular level would result in different drug effects at the population level. It has been found that in most respects the two models predict broadly similar effects. A related point to be made is that when cellular signalling is included, one should be very cautious about claims of validating models from scanty data. The extent to which signalling models can be distinguished in such a context is discussed elsewhere [46]. It is worth emphasizing that our intracellular models are essentially minimal models but have the features which would be expected of all apoptosis models.
Drug transport vs. drug effect
Our analysis reveals that in pulse injections, the drug killing effect is typically confined to the proximal region due to limited drug penetration for both types of apoptosis models (Figures 5,6,7). Our analysis reveals in multiple ways, the need for incorporating the dynamics of intracellular signalling in mechanistic predictions of drug effect and investigating drug transport and drug effect in an integrated manner. Within the present modelling framework, it is found that interstitial drug transport and drug effect are strongly coupled: drug-induced pharmacological effect (apoptosis) can enhance drug penetration in the interstitium, which constrains the exertion of drug action. This indicates that an optimal staged treatment schedule might help to result in a rapid penetration and the subsequent cell killing in regions further away from blood vessels.
It is natural to expect that greater diffusivity enhances drug penetration in the interstitium, resulting in a more homogeneous distribution of drug concentration. However, as our analysis shows, it is not necessarily the case that greater diffusivity would improve drug effect on tumour cells for a given stimulus. This is due to the specific requirement for apoptosis being triggered and the interplay between intracellular dynamics and interstitial drug transport. Drug diffusivity in tumour tissues is not only related to the physiological properties of anticancer agents, i.e. their charge, size or shape, but also to the structures of tumour tissues [50]. Therefore, the analysis presented here may help to refine drug infusion strategies by noting the different effects of drug diffusivity in a drug- and tumour- specific context.
Conclusions
In this paper, an in silico experimental platform is employed, which describes the flow of information from drug delivery to drug effect combining tumour blood flow, anticancer drug transport and cellular dynamics. Within the simplified model setting, a series of investigations on different drug stimuli and parameters is presented, providing explicit insights into the effect of drug and the interplay between multiple transport processes and intracellular signalling dynamics. Although the modelling platform in the current study is coarse-grained and qualitative in nature, it is capable of accommodating other mathematical models and allowing for fine graining and augmented descriptions of individual subprocesses systematically. Quantitative understanding of the contributing factors could be achieved for specific tumour types and specific anticancer drugs, in conjunction with the relevant experimental data at multiple scales. The modelling platform can serve as a computational tool to perform a thorough sensitivity analysis for the control and optimization of chemotherapeutic processes.
Declarations
Authors’ Affiliations
References
- Venkatasubramanian R, Henson MA, Forbes NS: Integrating cell-cycle progression, drug penetration and energy metabolism to identify improved cancer therapeutic strategies. J Theor Biol 2008, 253: 98-117. 10.1016/j.jtbi.2008.02.016MathSciNetView ArticleGoogle Scholar
- Minchinton AI, Tannock IF: Drug penetration in solid tumours. Nat Rev Cancer 2006, 6: 583-592. 10.1038/nrc1893View ArticleGoogle Scholar
- Trédan O, Galmarini CM, Patel K, Tannock IF: Drug resistance and the solid tumor microenvironment. J Natl Cancer Inst 2007, 99: 1441-1454. 10.1093/jnci/djm135View ArticleGoogle Scholar
- Boucher Y, Baxter LT, Jain RK: Interstitial pressure gradients in tissue-isolated and subcutaneous tumors: implications for therapy. Cancer Res 1990, 50: 4478-4484.Google Scholar
- Grantab R, Sivananthan S, Tannock IF: The penetration of anticancer drugs through tumor tissue as a function of cellular adhesion and packing density of tumor cells. Cancer Res 2006, 66: 1033-1039. 10.1158/0008-5472.CAN-05-3077View ArticleGoogle Scholar
- Netti PA, Berk DA, Swartz MA, Grodzinsky AJ, Jain RK: Role of extracellular matrix assembly in interstitial transport in solid tumors. Cancer Res 2000, 60: 2497-2503.Google Scholar
- Kreeger PK, Lauffenburger DA: Cancer systems biology: a network modeling perspective. Carcinogenesis 2010, 31: 2-8. 10.1093/carcin/bgp261View ArticleGoogle Scholar
- Kaufmann SH, Earnshaw WC: Induction of apoptosis by cancer chemotherapy. Exp Cell Res 2000, 256: 42-49. 10.1006/excr.2000.4838View ArticleGoogle Scholar
- Kohn KW, Jackman J, Oconnor PM: Cell-cycle control and cancer-chemotherapy. J Cell Biochem 1994, 54: 440-452. 10.1002/jcb.240540411View ArticleGoogle Scholar
- Martin GS: Cell signaling and cancer. Cancer Cell 2003, 4: 167-174. 10.1016/S1535-6108(03)00216-2View ArticleGoogle Scholar
- Cloutier M, Wang E: Dynamic modeling and analysis of cancer cellular network motifs. Integr Biol 2011, 3: 724-732. 10.1039/c0ib00145gView ArticleGoogle Scholar
- Tyson JJ, Baumann WT, Chen C, Verdugo A, Tavassoly I, Wang Y, Weiner LM, Clarke R: Dynamic modelling of oestrogen signalling and cell fate in breast cancer cells. Nat Rev Cancer 2011, 11: 523-532. 10.1038/nrc3081View ArticleGoogle Scholar
- Liu C, Krishnan J, Stebbing J, Xu XY: Use of mathematical models to understand anticancer drug delivery and its effect on solid tumors. Pharmacogenomics 2011, 12: 1337-1348. 10.2217/pgs.11.71View ArticleGoogle Scholar
- El-Kareh AW, Secomb TW: A mathematical model for comparison of bolus injection, continuous infusion, and liposomal delivery of doxorubicin to tumor cells. Neoplasia 2000, 2: 325-338. 10.1038/sj.neo.7900096View ArticleGoogle Scholar
- Kuh HJ, Jang SH, Wientjes MG, Au JL: Computational model of intracellular pharmacokinetics of paclitaxel. J Pharmacol Exp Ther 2000, 293: 761-770.Google Scholar
- Chappell MJ, Evans ND, Errington RJ, Khan IA, Campbell L, Ali R, Godfrey KR, Smith PJ: A coupled drug kinetics-cell cycle model to analyse the response of human cells to intervention by topotecan. Comput Methods Programs Biomed 2008, 89: 169-178. 10.1016/j.cmpb.2007.11.002View ArticleGoogle Scholar
- Eikenberry S: A tumor cord model for doxorubicin delivery and dose optimization in solid tumors. Theor Biol Med Model 2009, 6: 16. 10.1186/1742-4682-6-16MathSciNetView ArticleGoogle Scholar
- Thurber GM, Weissleder R: A systems approach for tumor pharmacokinetics. PLoS One 2011, 6: e24696. 10.1371/journal.pone.0024696View ArticleGoogle Scholar
- Goh Y-MF, Kong HL, Wang C-H: Simulation of the delivery of doxorubicin to hepatoma. Pharm Res 2001, 18: 761-770. 10.1023/A:1011076110317View ArticleGoogle Scholar
- El-Kareh AW, Secomb TW: A mathematical model for cisplatin cellular pharmacodynamics. Neoplasia 2003, 5: 161-169.View ArticleGoogle Scholar
- El-Kareh AW, Secomb TW: Two-mechanism peak concentration model for cellular pharmacodynamics of Doxorubicin. Neoplasia 2005, 7: 705-713. 10.1593/neo.05118View ArticleGoogle Scholar
- El-Kareh AW, Labes RE, Secomb TW: Cell cycle checkpoint models for cellular pharmacology of paclitaxel and platinum drugs. AAPS J 2008, 10: 15-34. 10.1208/s12248-007-9003-6View ArticleGoogle Scholar
- Simeoni M, Magni P, Cammia C, De Nicolao G, Croci V, Pesenti E, Germani M, Poggesi I, Rocchetti M: Predictive pharmacokinetic-pharmacodynamic modeling of tumor growth kinetics in xenograft models after administration of anticancer agents. Cancer Res 2004, 64: 1094-1101. 10.1158/0008-5472.CAN-03-2524View ArticleGoogle Scholar
- Kozusko F, Chen P, Grant SG, Day BW, Panetta JC: A mathematical model of in vitro cancer cell growth and treatment with the antimitotic agent curacin A. Math Biosci 2001, 170: 1-16. 10.1016/S0025-5564(00)00065-1MathSciNetView ArticleMATHGoogle Scholar
- Jackson TL: Intracellular accumulation and mechanism of action of doxorubicin in a spatio-temporal tumor model. J Theor Biol 2003, 220: 201-213. 10.1006/jtbi.2003.3156View ArticleMathSciNetGoogle Scholar
- Jang SH, Wientjes MG, Au JL: Enhancement of paclitaxel delivery to solid tumors by apoptosis-inducing pretreatment: effect of treatment schedule. J Pharmacol Exp Ther 2001, 296: 1035-1042.Google Scholar
- Zheng JH, Chen CT, Au JL, Wientjes MG: Time- and concentration-dependent penetration of doxorubicin in prostate tumors. AAPS PharmSci 2001, 3: E15. 10.1208/ps030318View ArticleGoogle Scholar
- Baxter LT, Jain RK: Transport of fluid and macromolecules in tumors. I. Role of interstitial pressure and convection. Microvasc Res 1989, 37: 77-104. 10.1016/0026-2862(89)90074-5View ArticleGoogle Scholar
- Baxter LT, Jain RK: Transport of fluid and macromolecules in tumors. II. Role of heterogeneous perfusion and lymphatics. Microvasc Res 1990, 40: 246-263. 10.1016/0026-2862(90)90023-KView ArticleGoogle Scholar
- Baxter LT, Jain RK: Transport of fluid and macromolecules in tumors. III. Role of binding and metabolism. Microvasc Res 1991, 41: 5-23. 10.1016/0026-2862(91)90003-TView ArticleGoogle Scholar
- Teo CS, Hor Keong Tan W, Lee T, Wang C-H: Transient interstitial fluid flow in brain tumors: effect on drug delivery. Chem Eng Sci 2005, 60: 4803-4821. 10.1016/j.ces.2005.04.008View ArticleGoogle Scholar
- Netti PA, Roberge S, Boucher Y, Baxter LT, Jain RK: Effect of transvascular fluid exchange on pressure-flow relationship in tumors: a proposed mechanism for tumor blood flow heterogeneity. Microvasc Res 1996, 52: 27-46. 10.1006/mvre.1996.0041View ArticleGoogle Scholar
- Pozrikidis C, Farrow DA: A model of fluid flow in solid tumors. Ann Biomed Eng 2003, 31: 181-194.View ArticleGoogle Scholar
- Liu C, Krishnan J, Xu XY: A systems-based mathematical modelling framework for investigating the effect of drugs on solid tumours. Theor Biol Med Model 2011, 8: 45. 10.1186/1742-4682-8-45View ArticleGoogle Scholar
- Jain RK: Determinants of tumor blood flow: a review. Cancer Res 1988, 48: 2641-2658.Google Scholar
- Baish JW, Gazit Y, Berk DA, Nozue M, Baxter LT, Jain RK: Role of tumor vascular architecture in nutrient and drug delivery: an invasion percolation-based network model. Microvasc Res 1996, 51: 327-346. 10.1006/mvre.1996.0031View ArticleGoogle Scholar
- Peters W, Teixeira M, Intaglietta M, Gross JF: Microcirculatory studies in rat mammary carcinoma. I. Transparent chamber method, development of microvasculature, and pressures in tumor vessels. J Natl Cancer Inst 1980, 65: 631-642.Google Scholar
- Seaton D, Krishnan J: Modular systems approach to understanding the interaction of adaptive and monostable and bistable threshold processes. IET Syst Biol 2011, 5: 81-94. 10.1049/iet-syb.2009.0061View ArticleGoogle Scholar
- Albeck JG, Burke JM, Spencer SL, Lauffenburger DA, Sorger PK: Modeling a snap-action, variable-delay switch controlling extrinsic cell death. PLoS Biol 2008, 6: 2831-2852.View ArticleGoogle Scholar
- Legewie S, Bluthgen N, Herzel H: Mathematical modeling identifies inhibitors of apoptosis as mediators of positive feedback and bistability. Plos Comput Biol 2006, 2: 1061-1073.View ArticleGoogle Scholar
- Gu C, Zhang J, Chen Y, Lei J: A trigger model of apoptosis induced by tumor necrosis factor signaling. BMC Syst Biol 2011,5(Suppl 1):S13. 10.1186/1752-0509-5-S1-S13View ArticleGoogle Scholar
- Zhang T, Brazhnik P, Tyson J: Computational analysis of dynamical responses to the intrinsic pathway of programmed cell death. Biophys J 2009, 97: 415-434. 10.1016/j.bpj.2009.04.053View ArticleGoogle Scholar
- Eissing T, Conzelmann H, Gilles ED, Allgower F, Bullinger E, Scheurich P: Bistability analyses of a caspase activation model for receptor-induced apoptosis. J Biol Chem 2004, 279: 36892-36897. 10.1074/jbc.M404893200View ArticleGoogle Scholar
- 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.200404158View ArticleGoogle Scholar
- Bagci EZ, Vodovotz Y, Billiar TR, Ermentrout GB, Bahar I: Bistability in apoptosis: roles of bax, bcl-2, and mitochondrial permeability transition pores. Biophys J 2006, 90: 1546-1559. 10.1529/biophysj.105.068122View ArticleGoogle Scholar
- Krishnan J, Liu C: An investigation of signal transduction and irreversible decision making through monostable and bistable switches. In Systems and Synthetic Biology: A Systematic Approach. Edited by: Kulkarni V, Stan G-B, Raman K. Springer Verlag; 2012. In pressGoogle Scholar
- Edelstein-Keshet L: Mathematical Models in Biology. SIAM, 3600 Market Street, Floor 6, Philadelphia, PA 19104: Society for Industrial and Applied Mathematics; 1988.MATHGoogle Scholar
- Eksborg S, Andersson M, Domellöf L, Lönroth U: A pharmacokinetic study of adriamycin and 4′epi-adriamycin after simultaneous intra-arterial liver administration. Med Oncol Tumor Pharmacother 1986, 3: 105-110.Google Scholar
- Robert J: Pharmacokinetics of adriamycin in patients with breast cancer: correlation between pharmacokinetic parameters and clinical short-term response. Eur J Canc Clin Oncol 1982, 18: 739-745. 10.1016/0277-5379(82)90072-4View ArticleGoogle Scholar
- Jain RK: Transport of molecules in the tumor interstitium: a review. Cancer Res 1987, 47: 3039-3051.Google Scholar
Copyright
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.