Systems biology approach to identification of biomarkers for metastatic progression in cancer

Abstract

Background: Metastases are responsible for the majority of cancer fatalities. The molecular mechanisms governing metastasis are poorly understood, hindering early diagnosis and treatment. Previous studies of gene expression patterns in metastasis have concentrated on selection of a small number of "signature" biomarkers.

Results: We propose an alternative approach that puts into focus gene interaction networks and molecular pathways rather than separate genes. We have reanalyzed expression data from a large set of primary solid and metastatic tumors originating from different tissues using the latest available tools for normalization, identification of differentially expressed genes and pathway analysis. Our studies indicate that regardless of the tissue of origin, all metastatic tumors share a number of common features related to changes in basic energy metabolism, cell adhesion/cytoskeleton remodeling, antigen presentation and cell cycle regulation. Analysis of multiple independent datasets indicates significantly reduced oxidative phosphorylation in metastases compared to primary solid tumors.

Conclusion: Our methods allow identification of robust, although not necessarily highly expressed biomarkers. A systems approach relying on groups of interacting genes rather than single markers is also essential for understanding the cellular processes leading to metastatic progression. We have identified metabolic pathways associated with metastasis that may serve as novel targets for therapeutic intervention.

Figure 1

Biological pathways significantly overrepresented in the shortlist of genes differentially expressed between primary solid and metastatic tumors (Ramaswamy et al. data set).

Figure 2

Genes differentially expressed between primary and metastatic cancers in the oxidative phosphorylation pathway. Relative change and direction of change in transcript abundance of differentially expressed are marked with color flags. Red color designates higher and blue color designates lower transcript abundance compared to average between primary tumor (1) and metastatic samples (2). The legend for GeneGo pathway maps is given in Supplemental Figure 6 (Additional File 1).

Figure 3

Glycolysis pathway. In spite of the fragmentary nature of the composed meta-set, the Warburg effect is still reflected in the pathway map through increased abundance of lactate dehydrogenase (LDHB). Relative change and direction of change in transcript abundance of differentially expressed are marked with color flags. Red color designates higher and blue color designates lower transcript abundance compared to average between primary tumor (1) and metastatic samples (2). The legend for GeneGo pathway maps is given in Supplemental Figure 6 (Additional File 1).

Figure 4

Alterations in extraceullular matrix and secreted proteins associated with metastatic cancer. Relative change and direction of change in transcript abundance of differentially expressed are marked with color flags. Red color designates higher and blue color designates lower transcript abundance compared to average between primary tumor (1) and metastatic samples (2). The legend for GeneGo pathway maps is given in Supplemental Figure 6 (Additional File 1).

Figure 5

Alterations in adhesion-mediated signaling and cytoskeleton remodeling in metastatic cancer. Relative change and direction of change in transcript abundance of differentially expressed are marked with color flags. Red color designates higher and blue color designates lower transcript abundance compared to average between primary tumor (1) and metastatic samples (2). The legend for GeneGo pathway maps is given in Supplemental Figure 6 (Additional File 1).

Figure 6

Alterations in the antigen presentation pathway observed in metastatic tumors. Relative change and direction of change in transcript abundance of differentially expressed are marked with color flags. Red color designates higher and blue color designates lower transcript abundance compared to average between primary tumor (1) and metastatic samples (2). The legend for GeneGo pathway maps is given in Supplemental Figure 6 (Additional File 1).