Identification of decisive molecular interactions regulating resistance to apoptosis in tumour cells

Abstract

TRAIL is a member of the TNF family of cytokines which can selectively induce apoptosis in cancerous cells. For this reason recombinant TRAIL is being considered for the treatment of cancer. During their development cancer cells disengage pivotal sections of the machinery required for apoptosis. This is also the case for the TRAIL pathway. Concurrently 50-60% of tumours are known to be resistant to TRAIL. Additionally reports found that TRAIL can initiate NF-kB -pro-survival pathways in resistant cell lines which may promote tumour aggression. Therefore a reliable biomarker that can identify responsive patients is needed. Biomarker discovery is often based on the identification of differentially expressed genes from transcriptome studies. In these analyses the threshold is a minimum arbitrary level (e.g. two fold induction.) in the majority of samples. This gene-by-gene approach cannot assess interaction between genes or proteins. . Minimal forays have been made into identifying genes or proteins which potentially influence each other, i.e. their combined presence or ratio correlates stronger with a given phenotype. These expression ratios, even if they are not robustly overexpressed or repressed, may better identify decisive interactions and predict response. Here we show that by utilizing classification trees we could identify a panel of genes which could predict sensitivity to TRAIL with a high degree of sensitivity and specificity. In addition we identify the caspase-8/Bcl-2 protein expression ratio as a predictive biomarker for sensitivity to TRAIL. Both of these biomarkers perform better than other classifiers of TRAIL sensitivity.