Developing an Adaptable In Silico Model of the Arachidonic Acid Cascade

Abstract

A major contributor to the regulation of inflammation and immunity, is a class of potent bioactive lipid mediators known as the eicosanoids. New links between these lipid mediators and physiological and pathological conditions are constantly being discovered. The production of eicosanoids is mediated by several enzymes such as cyclooxygenases (COX-1/-2) and lipoxygenases (5-/12-/15-LOX). Eicosanoids are derivatives of 20 carbon polyunsaturated fatty acids, including arachidonic acid. The arachidonic acid cascade generates numerous eicosanoid species, including prostaglandins, leukotrienes, thromboxanes and prostacyclin. As a result of the complexity of the arachidonic acid cascade, it is difficult to intuitively predict the outcome of perturbations. By combining targeted lipidomics with mechanistic mathematical modeling a predictive, adaptable computational model of the network of eicosanoids was created. This multi-disciplinary approach was employed in order to develop a tool for experimental and computational research on the arachidonic acid cascade. In this work, several improvements were made to existing computational models of the arachidonic acid cascade, including employment of ensemble modeling, extensive documentation of parameter values, addition of enzymatic reactions, explicit consideration of eicosanoid export, and detailed validation using mediator lipidomics. Monte Carlo ensemble modeling was used to capture the uncertainty associated with each parameter value whilst taking into account thermodynamic consistency of correlated parameters, allowing for quantitative assessment of confidence levels of model predictions. To validate and refine the model, lipidomic profiles of eicosanoids produced by two cell lines (human HaCaT keratinocytes and 46BR.1N fibroblasts) were analysed by UPLC-ESI-MS/MS, following treatment with pro-inflammatory stimuli in time-course studies. Each modeling condition was replicated in vitro, so that the biological plausibility of the predictions could be assessed and refinements made when required. As a result, a family of improved models were generated which capture some of the key features of the arachidonic acid cascade. The models are capable of predicting the concentration of eicosanoids produced after cell stimulation by three different inflammatory stimuli (calcium ionophore A23187, UV radiation and ATP). However, the calcium ionophore A23187 models were more accurate than others. Furthermore, the prediction of eicosanoid concentrations 6 h post stimulation was typically found to be more accurate than the prediction of eicosanoid concentrations 0.5 h post stimulation. As a result of these discrepancies, further areas of refinement were identified and interesting relationships in the network were highlighted, such as the influence of 12-LOX, ABC, COX-1, LTC4S, PGDS and PGT reactions on the prediction accuracy. This work could help generate novel hypotheses, design future experiments and prioritise experiments which generate interesting results. As a result, this work has the potential to save time, resources and money for researchers across the biosciences.

Date of Award	1 Aug 2020
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Rainer Breitling (Supervisor) & Anna Nicolaou (Supervisor)