Metabolic and protein interaction sub-networks controlling the proliferation rate of cancer cells and their impact on patient survival

Abstract

Cancer cells can have a broad scope of proliferation rates. Here we aim to identify the molecular mechanisms that allow some cancer cell lines to grow up to 4 times faster than other cell lines. The correlation of gene expression profiles with the growth rate in 60 different cell lines has been analyzed using several genome-scale biological networks and new algorithms. New possible regulatory feedback loops have been suggested and the known roles of several cell cycle related transcription factors have been confirmed. Over 100 growth-correlated metabolic sub-networks have been identified, suggesting a key role of simultaneous lipid synthesis and degradation in the energy supply of the cancer cells growth. Many metabolic sub-networks involved in cell line proliferation appeared also to correlate negatively with the survival expectancy of colon cancer patients.

Figure 1. Panel a shows the Spearman correlation coefficients and associated p-values of some of the metabolic genes that show the most significant correlations.

The panels b and c show the distributions of Spearman correlation coefficients and p-values. The higher frequency of low p-values shows that there are more growth correlated genes than those that could be expected as an artifact of multiple testing (which would correspond to a flat distribution of p-values).

Figure 2. Enrichment p-values for several transcription factor binding motifs in the JASPAR and TRANSFAC databases.

The p-values have been computed for 3 different sets of top correlated genes (defined by false discovery rates of 0.01, 0.03 and 0.05).

Figure 3. Protein interaction graph corresponding to the growth correlated genes.

The numbers in each node correspond to the identifiers in supplementary file S1. The yellow nodes correspond to proteins involved in regulatory feedbacks such as SKP2 (node 62) and p130 (node 73). We can see that p130 interacts with 17 other growth correlated proteins, which correspond to cytoplasmic ribosomal proteins and heterogeneous ribonucleoprotein particles (hnRNPs). The whole list of genes in the network can be found in the supplementary file S1. The orange nodes correspond to the most connected hubs of the main cluster. They are ribosomal proteins such as RPS9, RPS6, RPL4 or RPL13A. The green nodes correspond to proteins that also have a metabolic activity and the purple nodes are the most connected hubs in the second cluster, which correspond to mitochondrial ribosomal proteins.

Figure 4. Examples of 4 growth correlated sub-networks.

The reactions marked in red show positive transcriptional correlation with the growth rate. The three sub-networks involved in lipid biosynthesis and degradation re-appear among the metabolic sub-networks found in the patient mortality analysis.

Figure 5. Correlation of cell growth rate with lactate production and gene expression of citrate synthase.

The correlation with lactate production is actually negative, which contradicts the hypothesis of glycolysis being the main source of ATP for cancer cells.

Figure 6. Schematic representation of the hypothesized cycle combining simultaneous fatty acid synthesis and degradation.

Figure 7. Panel a represents how the number of metabolites and the robustness of each metabolic sub-network are distributed (for both the growth correlated sub-networks and the mortality correlated sub-networks).

Panel b shows the overlap between metabolites in the growth correlated sub-networks and the mortality correlated sub-networks.