(534a) Identifying Perturbed Pathways In Glioblastoma Through Network Reconstruction and Analysis | AIChE

(534a) Identifying Perturbed Pathways In Glioblastoma Through Network Reconstruction and Analysis

Authors 

Eddy, J. A. - Presenter, University of Illinois at Urbana Champaign
Price, N. D. - Presenter, University of Illinois at Urbana-Champaign

Introduction: Perturbations to biomolecular networks in glioblastoma multiforme (GBM)—the most common and devastating form of brain cancer—eventually result in the symptoms of disease observed by the patient, including seizure, nausea, vomiting, headache, and a progressive deterioration of memory and personality. These phenotypes arise not from any single mutation, but from the sum effect of complex interactions among multiple aberrant genes and other molecular agents. The combinatorial nature of GBM tumor development therefore mandates a systems-level approach to elucidate underlying mechanisms of the cancer. Reconstructing detailed in silico models of biochemical reaction networks (e.g., metabolic, signaling, regulatory) at the genome scale establishes a platform on which genetic perturbations can be related to emergent malignant functions and phenotypes. We have reconstructed the first genome-scale network of glioblastoma metabolism, as altered function in several metabolic pathways has been shown to be critical in the development of GBM as well as other cancers [1]. Constraint-based modeling with the network allows us to simulate cell growth under physiologically relevant conditions and investigate mechanisms for tumor development. Curated pathways in the reconstructed network also provide functional context for statistical approaches to study gene expression in disease states. These combined approaches enable the identification and detailed characterization of key perturbed pathways in GBM.

Materials and Methods: Utilizing the recent generic human metabolic reconstruction—a well-curated repository of reaction information—we have computationally integrated gene and protein expression data from multiple sources to define a network specific to GBM. Specifically, we first used the Model Building Algorithm (MBA) [2] to systematically prune reactions from Human Recon 1 [3]. Physico-chemical and environmental constraints were then applied to interrogate properties of the reconstructed network in silico, and to simulate cell growth under physiologically relevant conditions. Experimental measurements from the U87 GBM cell line—including metabolomics, siRNA knockdown phenotypes, and growth rates under various media conditions—were used, in conjunction with simulations, to refine, expand, and validate the model. To study pathway-level changes in gene expression relevant to GBM, we employed a method previously developed in our lab. Differential Rank Conservation (DIRAC) [4] was used to assess how pathway rankings (i.e., the relative expression ordering among genes within model pathways) differ, either among GBM samples, or between tumors and healthy astrocytes.

Results and Discussion: The model represents the subset of human metabolic capability predicted to be active in GBM, based on tissue-specific gene and protein expression, as well as experimental and literature data. Using the model, we were able to accurately predict experimental values we observed for U87 cell lines, including growth rates and lethality of gene deletions. Furthermore, the model is able to capture known metabolic phenotypes in cancer cells, including the Warburg effect. We examined GBM metabolic pathways with DIRAC and identified networks that are (i) tightly regulated within tumors and astrocytes, as defined by high conservation of transcript ordering; or (ii) variably expressed (i.e., shuffled) between phenotypes, resulting in specific classifiers that predict cancer versus healthy with high accuracy.

Conclusions: Cancer is an intrinsically complex and heterogeneous disease, making it particularly amenable to systems biology approaches. Using the mechanistically-detailed model we have built for GBM, we can begin to interrogate the link between mutations in the cancer and key metabolic processes that contribute to tumorigenesis. The model also serves as a powerful tool for studying pathway-level gene expression changes in GBM; perturbed pathways identified by DIRAC represent robust differences between healthy and diseased states, and in turn, serve as focal points for ongoing model simulation and development. In the future, systems-level analysis of GBM with the model should continue to provide important biological insights into underlying mechanisms and potential treatment routes for the disease.

References:

  1. Cairns, RA, et al., Nat Rev Cancer, 2011, 11:85-95.
  2. Jerby, L, et al., Mol Syst Biol, 2010,  6:401.
  3. Duarte, NC, et al., P Nalt Acad Sci USA, 2007, 104:1777-82
  4. Eddy, JA, et al., PLoS Comput Biol, 2010,  6:e1000792.