A systems biology approach reveals common metastatic pathways in osteosarcoma

Abstract

Background: Osteosarcoma (OS) is the most common malignant bone tumor in children and adolescents. The survival rate of patients with metastatic disease remains very dismal. Nevertheless, metastasis is a complex process and a single-level analysis is not likely to identify its key biological determinants. In this study, we used a systems biology approach to identify common metastatic pathways that are jointly supported by both mRNA and protein expression data in two distinct human metastatic OS models.

Results: mRNA expression microarray and N-linked glycoproteomic analyses were performed on two commonly used isogenic pairs of human metastatic OS cell lines, namely HOS/143B and SaOS-2/LM7. Pathway analysis of the differentially regulated genes and glycoproteins separately revealed pathways associated to metastasis including cell cycle regulation, immune response, and epithelial-to-mesenchymal-transition. However, no common significant pathway was found at both genomic and proteomic levels between the two metastatic models, suggesting a very different biological nature of the cell lines. To address this issue, we used a topological significance analysis based on a "shortest-path" algorithm to identify topological nodes, which uncovered additional biological information with respect to the genomic and glycoproteomic profiles but remained hidden from the direct analyses. Pathway analysis of the significant topological nodes revealed a striking concordance between the models and identified significant common pathways, including "Cytoskeleton remodeling/TGF/WNT", "Cytoskeleton remodeling/Cytoskeleton remodeling", and "Cell adhesion/Chemokines and adhesion". Of these, the "Cytoskeleton remodeling/TGF/WNT" was the top ranked common pathway from the topological analysis of the genomic and proteomic profiles in the two metastatic models. The up-regulation of proteins in the "Cytoskeleton remodeling/TGF/WNT" pathway in the SaOS-2/LM7 and HOS/143B models was further validated using an orthogonal Reverse Phase Protein Array platform.

Conclusions: In this study, we used a systems biology approach by integrating genomic and proteomic data to identify key and common metastatic mechanisms in OS. The use of the topological analysis revealed hidden biological pathways that are known to play critical roles in metastasis. Wnt signaling has been previously implicated in OS and other tumors, and inhibitors of Wnt signaling pathways are available for clinical testing. Further characterization of this common pathway and other topological pathways identified from this study may lead to a novel therapeutic strategy for the treatment of metastatic OS.

Figure 1

Pathway analysis of up-regulated genes from HOS/143B and SaOS-2/LM7 models. Top 10 significant pathways identified by MetaCore using all the significantly up-regulated genes from (A) HOS/143B and (b) SaOS-2/LM7 models. Only one common pathway, the “Development_WNT signaling pathway. Part 2”, was identified in the top 10 significant pathways between the two models. Bars quantify the –log (p-value) indicating the enrichment of the individual pathways identified. False discovery rate (FDR) is calculated for all identified pathways and a cutoff of FDR < 0.05 is set to define pathways as significant. Asterisks denote significant pathways shown.

Figure 2

Pathway analysis of up-regulated glycoproteins from HOS/143B and SaOS-2/LM7 models. Top 10 significant pathways identified by MetaCore using all the significantly up-regulated glycoproteins from (A) HOS/143B and (B) SaOS-2/LM7 models. No common significant pathways were identified in the top 10 pathways between the two models. All pathways shown are significant. Refer to Figure 1 legend for graph details.

Figure 3

Western blot validation of atase and vimentin. The glycoproteomic analysis identified atase and vimentin to be significantly up-regulated in metastatic 143B and LM7 cell lines, respectively. ß-Tubulin was used as a loading control (second row)

Figure 4

Topological nodes of up-regulated genes from HOS/143B and SaOS-2/LM7 models. Top 10 significant pathways identified by MetaCore using all the topological significant nodes from the up-regulated genes from (A) HOS/143B and (B) SaOS-2/LM7 models. Two common significant pathways were identified in the top 10 pathways of the two models, namely the “Cytoskeleton remodeling/TGF/WNT” and the “Cytoskeleton remodeling/Cytoskeleton remodeling”. All pathways shown are significant. Refer to Figure 1 legend for graph details.

Figure 5

Topological nodes of down-regulated genes from HOS/143B and SaOS-2/LM7 models. Top significant pathways identified by MetaCore using all the topological significant nodes from the down-regulated genes from (A) HOS/143B and (B) SaOS-2/LM7 models. Four common significant pathways were identified in the top 10 pathways of the two models, namely the “Cytoskeleton remodeling/TGF/WNT”, “Cytoskeleton remodeling/Cytoskeleton remodeling”, “Cell adhesion_Chemokines and adhesion”, and “Cell adhesion_ECM remodeling”. All pathways shown are significant. Refer to Figure 1 legend for graph details.

Figure 6

Map of Cytoskeleton remodeling/TGF/WNT pathway showing the molecules identified by the topological significance analysis using significantly up-regulated genes and up-regulated proteins from the two metastatic models. The highlighted molecules in the pathway represent the molecules identified by the topological analysis of the genomic and/or proteomic datasets.