Effect of Tumor Microenvironment on Tumor VEGF During Anti-VEGF Treatment: Systems Biology Predictions

DOI:10.1093/jnci/djt093 © The Author 2013. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: . Article Effect of Tumor Microenvironment on Tumor VEGF During Anti-VEGF Treatment: Systems Biology Predictions Stacey D. Finley, Aleksander S. Popel Manuscript received June 28, 2012; revised March 8, 2013; accepted March 22, 2013. Correspondence to: Stacey D. Finley, PhD, Department of Biomedical Engineering, School of Medicine, Johns Hopkins University, 720 Rutland Ave, 613 Traylor Research Bldg, Baltimore, MD 21205 (e-mail ). Vascular endothelial growth factor (VEGF) is known to be a potent promoter of angiogenesis under both physiological and pathological conditions. Given its role in regulating tumor vascularization, VEGF has been targeted in various cancer treatments, and anti-VEGF therapy has been used clinically for treatment of several types of cancer. Systems biology approaches, particularly computational models, provide insight into the complexity of tumor angiogenesis. These models complement experimental studies and aid in the development of effective therapies targeting angiogenesis. Methods We developed an experiment-based, molecular-detailed compartment model of VEGF kinetics and transport to investigate the distribution of two major VEGF isoforms (VEGF121 and VEGF165) in the body. The model is applied to predict the dynamics of tumor VEGF and, importantly, to gain insight into how tumor VEGF responds to an intravenous injection of an anti-VEGF agent. Results The model predicts that free VEGF in the tumor interstitium is seven to 13 times higher than plasma VEGF and is predominantly in the form of VEGF121 (>70%), predictions that are validated by experimental data. The model also predicts that tumor VEGF can increase or decrease with anti-VEGF treatment depending on tumor microenvironment, pointing to the importance of personalized medicine. Conclusions This computational study suggests that the rate of VEGF secretion by tumor cells may serve as a biomarker to predict the patient population that is likely to respond to anti-VEGF treatment. Thus, the model predictions have important clinical relevance and may aid clinicians and clinical researchers seeking interpretation of pharmacokinetic and pharmacodynamic observations and optimization of anti-VEGF therapies. J Natl Cancer Inst;2013;105:802–811 Vascular endothelial growth factor (VEGF) promotes various processes involved in angiogenesis, including endothelial cell proliferation, adhesion, migration, and chemotaxis (1). Angiogenesis is a hallmark of cancer (2) and has been targeted by various cancer therapies, with a focused effort on drugs that inhibit VEGF. Several antiangiogenic agents have been approved by the US Food and Drug Administration (FDA) to treat various cancers and other diseases. Bevacizumab (Genentech, South San Francisco, CA), a recombinant humanized monoclonal antibody to VEGF, is approved for the treatment of metastatic colorectal and kidney cancer, glioblastoma, and non–small cell lung cancer. Ziv-aflibercept (Regeneron, Tarrytown, NY), a soluble decoy receptor for VEGF, is an FDA-approved agent for the treatment of metastatic colorectal cancer and is currently in clinical trials for the treatment of several other cancer types. Other FDA-approved antiangiogenic cancer therapeutics include axitinib, pazopanib, regorafenib, sorafenib, and sunitinib. These agents are small molecule kinase inhibitors with various targets such as VEGF receptors, platelet-derived growth factor receptors, fibroblast growth factor receptors, and Raf kinase. 802 Articles | JNCI Systems biology approaches are useful in gaining a broader understanding of the complexity of angiogenesis. Computational models can be applied to generate and test biological hypotheses and can aid in the development of effective therapies that target angiogenesis (3). Additionally, models can provide a framework to predict promising drug targets and identify patient populations that will respond to a particular therapy. We have developed a molecular-detailed compartment model that is useful in understanding VEGF dynamics in the body. The model is based on detailed biochemical kinetics and molecular transport and has been validated against available experimental data. It is a predictive tool that can provide insight into the distribution of VEGF in the body and the effects of systemic administration of anti-VEGF therapeutics, such as bevacizumab and aflibercept. We have applied the model to understand and explain clinical observations of anti-VEGF agents (4) and predict the effect of the drugs (5,6). Here, we present three important model predictions regarding the pretreatment levels of VEGF121 and VEGF165 and the dynamic response of plasma and tumor Vol. 105, Issue 11 | June 5, 2013 Background VEGF to anti-VEGF treatment. We compare our results with available experimental data and propose clinical applications of the model predictions. Methods The whole-body model includes normal tissue (“normal compartment,” represented by skeletal muscle), the vasculature (“blood compartment”), and diseased tissue (“tumor compartment”) and has been described in our previous articles (5,6). The normal and tumor compartments consist of parenchymal and endothelial cells and interstitial space (Figure 1A). We include molecular interactions between two major VEGF isoforms (VEGF121 and VEGF165), VEGF receptors (VEGFR1 and VEGFR2), and coreceptor neuropilins (NRP1 and NRP2) (Figure 1B). In this study, we also include VEGF interactions with two soluble factors: soluble VEGFR1 (sVEGFR1) and α-2-macroglobulin (α2M), introduce VEGF secretion by endothelial cells, and modify the permeability between the blood and tumor. The tumor is parameterized as a breast tumor with a volume of 33 cm3; however, the model is broadly applicable to any solid tumor. Model elements reflect quantitative experimental characterization of the VEGF system. The model is described in detail in the Supplementary Methods (available online). The model predicts the concentration of 154 species using 154 ordinary differential equations. We are able to predict VEGF level in the multiple tissues in the body as well as the distribution of VEGF in the form of unbound ligand or matrix- and receptorbound complexes. The predicted levels of free VEGF and sVEGFR1 in muscle interstitium (7–13) and plasma (14–19) are within the range of experimental data (Table 1). Figure 1. Molecular-detailed compartmental model of vascular endothelial growth factor (VEGF) kinetics and transport in the body. A) The model includes three compartments: normal tissue, blood, and tumor tissue. VEGF is secreted by muscle fibers and tumor cells in the normal tissue and tumor, respectively (qv). VEGF receptors are localized on the luminal and abluminal endothelial surfaces and tumor cells. Only neuropilin 1 (NRP1) is present on muscle fibers. Free and ligand-b (...truncated)