In locally advanced rectal cancer, neoadjuvant chemoradiotherapy is performed prior to surgery to downstage the tumour. Thirty to 40 % of patients do not respond. Defects in apoptotic machinery lead to therapy resistance; however, to date, no study quantitatively assessed whether B cell lymphoma 2 (BCL2)-dependent regulation of mitochondrial apoptosis, effector caspase activation downstream of mitochondria or a combination of both predicts patient responses. In a cohort of 20 rectal cancer patients, we performed protein profiling of tumour tissue and employed validated ordinary differential equation-based systems models of apoptosis signalling to calculate the ability of cancer cells to undergo apoptosis. Model outputs were compared to clinical responses. Systems modelling of BCL2-signalling predicted patients in the poor response group (p = 0.0049). Systems modelling also demonstrated that rectal cancers depended on BCL2 rather than B cell lymphoma-extra large (BCL(X)L) or myeloid cell leukemia 1 (MCL1) for survival, suggesting that poor responders may benefit from therapy with selective BCL2 antagonists. Dynamic modelling of effector caspase activation could not stratify patients with poor response and did not further improve predictive power. We deliver a powerful patient stratification tool identifying patients who will likely not benefit from neoadjuvant chemoradiotherapy and should be prioritised for surgical resection or treatment with BCL2 antagonists.
Key messages: Modelling BCL2-family proteins identifies patients unresponsive to therapy. Caspase activation downstream of mitochondria cannot identify these patients. Rectal tumours of poor responders are BCL2- but not BCL-XL-dependent. DR_MOMP allows clinicians to identify patients who would not benefit from therapy. DR_MOMP is also a useful patient stratification tool for BCL2 antagonists.