Development of an integrated system biology model for predicting mixtures of chemicals: the case of the azole fungicides

Evaluating the toxicity of mixtures is one of the major objectives of today toxicology. Mixture experiments can provide an indication of the joint effects of substances, but it is practically infeasible to test all mixture combinations. Therefore, a mechanistic approach and the development of integrated models are crucial to assessing mixtures' risks. Data previously obtained by our research group described specific teratogenic effects (reduction and fusion of branchial arches mediated by altered hindbrain segmentation and neural crest migration) in postimplantation rat whole embryos cultured in vitro in presence of a number of azole fungicides, including triadimefon (FON) and flusilazole (FLUSI) alone or in binary mixtures. Results showed a clear concentration-response effect for single fungicides while co-exposure resulted in an additive effect. The observed additive effect of the binary mixture supports the hypothesis that the tested individual substances share the same mode of action (inhibition of CYP26 enzymes involved in retinoic acid, RA, catabolism with subsequent local increase in endogenous RA levels) and that azole fungicides constitute a common mechanism group. The aim of the present work is to obtain an in silico tool useful to predict the effects of any possible embryo co-exposure to azole fungicides. Using an integrated approach combining mathematical modeling, molecular docking and in vitro experiments, we have developed an integrated system biology model able to simulate both the formation of a RA gradient in the rat embryo hindbrain and its perturbation after exposure to FON, to FLUSI and their binary mixtures. We simulated the RA gradient formation in the hindbrain as in Goldbeter et al. [1], imposing a linear gradient for RA synthesis. Using GNU MCSim version 5.5.0, we fitted an empirical RA concentration-response model on the basis of previous experimental data (percentage of malformations at the branchial apparatus) obtained in rat embryos cultured in vitro in presence of increasing concentrations of RA. The parameters for FON and FLUSI were adjusted according to the fit to the previous experimental data and their affinities for CYP26 were computed following a computational approach based on molecular docking. Predictive simulations for the mixtures were then performed. The model appears to be reasonably predictive for the mixtures' effects (experimental data and model predictions agree at 90% confidence level) and the RA levels in rat hindbrain can be reasonably estimated after azole exposures. An experimental procedure for verifying other azole mixtures is now in progress. This project was funded by the Horizon 2020 Framework programme of the European Union (EuroMix project). [1] Goldbeter, A., Gonze, D. and Pourquié, O., Sharp developmental thresholds defined through bistability by antagonistic gradients of retinoic acid and FGF signaling. Dev. Dyn. 236 (2007) 1495–1508.