Effective Combination Therapies for B-cell Lymphoma Predicted by a Virtual Disease Model

Abstract

The complexity of cancer signaling networks limits the efficacy of most single agent treatments and brings about challenges in identifying effective combinatorial therapies. In this study, we used chronic active B cell receptor (BCR) signaling in diffuse large B cell lymphoma (DLBCL) as a model system to establish a computational framework to optimize combinatorial therapy in silico. We constructed a detailed kinetic model of the BCR signaling network, which captured the known complex crosstalk between the NFκB, ERK and AKT pathways and multiple feedback loops. Combining this signaling model with a data-derived tumor growth model, we predicted viability responses of many single drug and drug combinations in agreement with experimental data. Under this framework, we exhaustively predicted and ranked the efficacy and synergism of all possible combinatorial inhibitions of eleven currently targetable kinases in the BCR signaling network. Ultimately, our work establishes a detailed kinetic model of the core BCR signaling network and provides the means to explore the large space of possible drug combinations. Major Findings Using chronic active B cell receptor (BCR) signaling in diffuse large B cell lymphoma (DLBCL) as a model system, we developed a kinetic-modeling based computational framework for predicting effective combination therapy in silico. By integrative modeling of signal transduction, drug kinetics and tumor growth, we were able to directly predict drug-induced cell viability responses at various dosages, which were in agreement with published cell line experimental data. We implemented computational screening methods that identified potent and synergistic combinations in silico and validated our independent predictions in vitro. Quick Guide to Equations and Assumptions A kinetic model was constructed for the BCR signaling network according to the following rules. For protein-protein binding interactions, we assumed that this type of reaction were under equilibrium and solved the steady-state level of protein complex analytically in the model, false[ AB false] = 1 ( 1 + K ( [ T A ] + [ T B ] ) ) - false( 1 + K false( false[ T A false] + false[ T B false] false) false) 2 - 4 K 2 [ T A ] [ T B ] 2 K where [TA] and [TB] stand for the total concentration of protein species A and B respectively, [AB] represents the concentration of the protein complex, K is the inverse of the dissociation constant Kd. For kinase catalyzed reactions, we adopted the classic Michaelis-Menten kinetics, j 1 d P d t = 1 k cat [ E ] [ S ] K m + [ S ] where 1dPdt is the rate of catalytic product formation, kcat is the turnover rate, Km is the Michaelis-Menten constant, [E] and [S] are concentration of enzyme and substrate respectively. Each interaction in the BCR signaling network was depicted by corresponding terms according to above rules. The full kinetic model consists of 28 state...

Patient-Specific Modeling of Diffuse Large B-Cell Lymphoma

“…Roy et al (2019) [ 50 ] focused on B-cell differentiation processes downstream of NF-κB signaling in a population-based model, with a prediction for the role of RELA and cREL in DLBCL. Du et al (2017) [ 69 ] modeled the upstream regulation of NF-κB by BCR signaling in DLBCL using ODEs. The model was fitted to DLBCL cell line data and used to predict drug sensitivity and synergies for this cell line.…”

Cell signaling heterogeneity is modulated by both cell-intrinsic and -extrinsic mechanisms: An integrated approach to understanding targeted therapy

“…In particular, several recent studies developed integrated approaches of mathematical modeling with systematic perturbation experiments applying various kinase inhibitors to cancer cells. Some of these studies proposed novel combination therapies [ 43 , 51 , 54 ], just as we have. However, in addition to predicting average cell viability, we also consider post-treatment heterogeneity and, critically, the impact of the microenvironment.…”

“…The model assumes that the rate of change of a variable is determined by the linear combination of neighboring nodes with corresponding weights. This additive linear function has successfully described protein reaction networks [ 54 , 55 , 65 ] although other functions such as Michaelis-Menten kinetics are viable options [ 51 ]. In the experiments we carried out, the microenvironmental conditions are growth factor and HGF.…”

Executable cancer models: successes and challenges

“…It is challenging to accurately measure or fit parameters, especially in non-linear cases. They can be used to model the population of cells in a tumour 94,112 , or chemical reactions, such as in a regulatory pathway 42,113,114 . An important branch of these models is pharmacokinetic/dynamic (PKPD) models for modelling the effects of drugs 115 .…”