Cancer is a multiscale disease, involving processes that act at the molecular, cellular, tissue, and whole-body scales, and that happen at fast (e.g., biomolecular reactions), intermediate (e.g., tissue deformation), and slow (e.g., evolution) time scales. Cancer disrupts tissue morphology and function not only near primary tumors and metastases, but also in organs throughout the body. The body responds to the evolving cancer locally (via the vasculature and stroma) and globally (through the immune system and secreted factors), and these responses alter future tumor growth. All these processes (with their inherent time and space scales) combine to form a complex system, wherein mutated cells respond to and alter their environment to survive and thrive—this complex adaptive system is known as cancer.
While the disease manifests at these tissue and whole-body levels, clinicians generally attack cancer at the cellular and molecular level using therapeutic agents targeting molecular linchpins of the disease, such as protein kinases that drive growth. When we study response, researchers typically focus on a small number of factors at a particular scale, such as tumor shrinkage as a function of drug, or the molecular effect of inhibiting a kinase.
Bridging these scales from the molecular to the organismal (whole-body) remains a unique but critical challenge. At the molecular level, tremendous advances have been made identifying the mutations and epigenetic events that drive individual patients' cancer progression. However, it is unclear how these molecular events map onto cellular, tissue and organismal level disruptions. At the other end of the scale, the histological patterns seen in cancerous tissues are used daily in the clinic to stage the disease and gauge prognosis. However, the precise molecular underpinnings of the aberrant tissue morphology are unclear, and therefore, it is impossible to predict the most effective therapeutic regimen of molecular targets. A multiscale, integrative, multidisciplinary approach is called for—one that brings together scientists from across disciplines to shed light on the complex cancer system.
We are convinced that the diverse perspectives and techniques developed by physical and quantitative scientists can make a significant impact on cancer management and treatment in potentially revolutionary ways. Our work is centered on our strong belief that diverse physical measurements at molecular to organismic scales need to be integrated with sophisticated and diverse modeling approaches, to generate a model of cancer that can be used predict the behavior of cancers during emergence and in response to treatment and other perturbations.
Mouse models of lymphoma
We have developed two mouse models that closely resemble human B-cell lymphoma: Eμ-Myc/Arf-/- (a very aggressive B-cell lymphoma which is slightly resistant to treatment), and Eμ-Myc/p53-/- (a very aggressive B-cell lymphoma that is very resistant to treatment). These murine cells overexpress the oncogene c-myc, the resulting B-cell malignancies closely resemble human non-Hodgkin's lymphomas, tumors arise with relatively short latency and high penetrance, tumor burden is easy to monitor by lymph node palpation or blood smear, therapy can be performed on mice with intact immune systems, and the cells can be cultured and transplanted into syngenic, non-transgenic mice. These characteristics make them ideal for detailed, controlled studies of B-cell lymphoma progression that closely mimic human disease, including the development of resistance to treatments. We have coordinated the dissemination of these cell lines throughout our PSOC member institutions along with rigorous handling protocols to ensure that all scientists are performing measurements on cells that are as identical as possible. See Core 2.
We are working to comprehensively characterize our lymphoma model system with in vitro and in vivo experiments with novel, state-of-the-art measurement platforms, measuring from the molecular to the whole-mouse scale. Among our techniques:
- Multiplex spectral surface plasmon resonance imaging (SPR)
- Multiplex SPR flows proteins across a 2D sensor array of 96 or more protein ligands (binding targets) and detects changes in the refractive index to quantitatively measure protein-ligand binding rates and strengths. (More information here.) We are using this technology in RP1 to measure protein-protein interactions to necessary for building cell signaling networks.
- We use this technique in RP1 with flow cytometry and fluorescence-to measure the protein phosphorylation state (activity) of individual cells, allowing construction of signaling models that better tie to cell phenotype and behavior.
- Liquid chromatography coupled with tandem mass spectrometry (LC-LCMS/MS)
- RP1 uses LC-LCMS/MS to quantitatively probe cell surface and whole-cell protein abundances to help build cellular signaling models.
- RNA microarrays
- RP1 uses RNA microarrays to measure cell gene expression changes to help build cellular signaling models.
- Oligonucleotide-based Comparative Genomic Hybridization (CGH) arrays
- We use this method in RP2 to measure changes in cell epigenetics and genetic structure (copy number, loss of heterozygosity, etc.) as cells evolve in response to therapy. See here for more information on the technique.
- Intravital microscopy (IVM)
- IVM, especially when coupled with surgically-implanted windows, is capable of real-time imaging in live animals at multiple time points, allowing new characterizations of angiogenesis and other dynamic processes at the cell and tissue scales. RP3 is using this method to measure dynamic tumor-associated processes in mouse lymph nodes. See here for some interesting IVM papers and images.
- RP3 uses morphometric, colorimetric, and other measurements on pathology slides to quantify cell activity (e.g., cell cycling by Ki67, apoptosis by cleaved Caspase-3) relative to position within tissues. This helps to calibrate computational tumor growth models.
- Protein microarray
- We make extensive use of protein microarrays in RP4 to quantify host response (in the form of antibodies) to tumor growth. We are using this to help build models of tumor-host interaction.
- MagnetoNano Sensor
- This technology uses an array of magneto-nano sensors coupled with anti-protein probes, allowing accurate, high-throughput measurements of protein-protein interactions. (See here for more information.) This method is used to derive detailed tumor-host interaction data for RP4.
Multiscale modeling approach
We are approaching computational modeling with a variety of techniques, each suited to the scale of measurements being obtained. Statistical and ordinary differential equation models being developed in RP1 use detailed in vitro measurements to understand the relationship between cell microenvironment (oxygenation, drug exposure, etc.) to cell phenotype. RP2 is building sophisticated models of the evolution of tumor epigenetics in response to treatment. RP3 is integrating sophisticated model of tumor and tissue mechanics, substrate and drug transport, and angiogenesis to model tumor development in the lymph node, with calibration to novel in vivo mouse data. RP4 is developing statistical models of the mouse immune response based upon cutting-edge microarray data. And RP5 is working to tie these data and modeling approaches together through a state-of-the-art compartmental modeling framework, combined with control theory and other engineering approaches.
Principal Investigator: W. Daniel Hillis, Ph.D.
Senior Co-Investigator: David B. Agus, M.D.
RP1: Dynamic State Modeling of Cancer Cell Response to Therapy
Project Lead: Garry Nolan, Ph.D.
Project Co-Lead: Parag Mallick, Ph.D.
RP1 aims to develop a computational model that operates at and below the cellular scale and across multiple time scales to describe how the genetic background and chemical/environmental context of a cell lead to alterations in cellular physiology that ultimately impact the tumor and host. We will focus our efforts on the response of Burkitt's lymphoma to the chemotherapeutic cyclophosphamide and to micro-environment directed therapy. We will focus on interventions that include treatment with cytotoxic chemotherapy (cytarabine, doxorubicin, cyclophosphamide, etc.), environmentally targeted agents (anti-VEGF antibodies, mTOR inhibitors, etc.), and genetic manipulations including targeted gene suppression (with RNAi) or addition (with retroviral mediated gene transfer). To help train our model we will collect coordinated measurements at many molecular and cellular levels with a level of inter-coordination not seen in any publicly available data set and will integrate measurements of:
- protein interaction dynamics from a novel multiplex SPR instrument;
- high-throughput measurements of single-cell protein phosphorylation;
- traditional measurements of mRNA expression; and
- measurement of the proteome.
RP2: A Cancer Evolution Space-Time Machine
Project Lead: Matteo Pellegrini, Ph.D.
This project hypothesizes that tumor genomes become polymorphic after transformation, and these variations record ancestry, or how and how fast it took for a single transformed cell to become the present day tumor population. We will apply this hypothesis to develop a detailed model of cancer evolution using data from single cell genetic analysis of tumor specimens generated by RP3. The detailed molecular phylogenetic model will describe how tumors evolve during standard growth and when under stress from cytotoxic chemotherapy and micro-environment targeted therapy.
RP3: Multi-scale Cancer Modeling: From Cell Phenotypes to Growth and Therapy Response
Project Lead: Sam Gambhir, M.D. Ph.D.
Project Co-Lead: Vittorio Cristini, Ph.D.
We systematically investigate the physical principles governing tumor growth and therapeutic response across broad spatiotemporal scales (seconds to months and molecular to whole-organ) by developing an integrative, multi-scale in silico/in vivo cancer model. To quantify the complex relationships between cancer phenomena at different scales, we dynamically couple discrete (cell-scale) and continuum (tissue-scale) models developed by Cristini and co-workers in a hybrid, multi-scale framework. We integrate this framework with state-of-the-art intravital microscopy (IVM) time-course measurements of tumor growth and chemotherapy response by Gambhir and co-workers.
RP4: Integrated Multi-Scale Analysis of Tumor and Host Response to Therapy
Project Lead: Josh LaBaer, M.D. Ph.D.
Project Co-Lead: Shan Wang, Ph.D.
Many factors contribute to treatment failure in cancer including lack of tumor response to the drugs, toxicity to the host, tumor growth in sanctuary sites, and the emergence of resistance to drugs. Typically, these factors are studied individually, but it is their action in concert that ultimately overwhelms the patient. RP4 is dedicated to multi-scale measurements of the host response to cancer and its therapy and integrating this information with the tumor responses measured by the other projects into a comprehensive, predictive functional interaction tumor-host (TH) model. This model will be grounded on a novel data set of two key mechanisms mediating TH interactions: 1) host immune response and 2) cytokines that mediate intercellular communication.
Key Personnel: Dan Ruderman, Ph.D. Dean Felsher, M.D. Ph.D. Paul Macklin, Ph.D.
This project aims to integrate the models and data being created at the intracellular and cellular scales (RP1-RP2), the tissue and organ scales (RP3), and the host response scale (RP4) into a comprehensive, quantitative model that predicts disease progression and therapy response in individual mice. Methods include a combination of statistical modeling, upscaling, control theory, and mechanistic compartmental models. It is our plan and hope that the integration methods we develop will help facilitate faster analysis and interpretation of multi-platform, multi-scale data sets being generated in the cancer community.
Core 1: Coordination and Dissemination of Biomodels and Samples
Core Lead: Mitchell Gross, M.D., Ph.D.
The primary aim of this core is to provide the basic biologic specimens that will inform the modeling approaches across multiple parts of this project. The Aims include: Standardizing sample processing and storage within this PS-OC and providing high-quality and uniform samples for coordinated analysis.
Core 2: Data and Computational Models Dissemination
Core Lead: Carl Kesselman, Ph.D.
The primary aim of this core is to provide the information technology infrastructure needed to enable sharing of the heterogeneous data produced by the various research projects, allowing these data to be integrated into multi-scale models. This core additionally coordinates model archiving and sharing.