Mode of action-based classification and prediction of activity of uncouplers for the screening of chemical inventories

Abstract

A new approach for classification of uncouplers of oxidative and photophosphorylation, also suitable for screening of large chemical inventories, is introduced. Earlier fragment-based approaches for this mode of toxic action are limited to phenols but weak acids of extremely diverse chemical classes can act as uncouplers. The proposed approach overcomes the limitation to phenolic uncouplers by combining structural fragments with the global information of physico-chemical descriptors. In a top-down approach to reduce the number of candidate chemicals, firstly substructure definitions for the detection of weak acids were applied. Subsequently, conservative physico-chemical thresholds for the two most important properties for the uncoupling activity were defined: an acid dissociation constant (pK _a) between 3 and 9, and a sufficiently low energy barrier for the internal permeability of anions (17 kcal/mol). The later was derived from a novel approach to calculate the distribution of compounds across membranes. The combination of structural and physico-chemical criteria allowed a good separation of active from inactive chemicals with high sensitivity (95%) and slightly lower (more than 75%) specificity. Applying this approach to several thousand high and low production volume chemicals retrieved a surprisingly small number of 10 compounds with a predicted excess toxicity above 10. Nevertheless, uncoupling can be an important mode of action as highlighted with several examples ranging from pesticide metabolites to persistent organic compounds.

Keywords:

• classification
• mode of action
• uncouplers of oxidative phosphorylation
• database screening
• mechanistic QSAR

Acknowledgments

The authors thank Nadine Bramaz for experimental assistance, Philippe Perisset for programming the data evaluation and other features of the Kinspec system, Johann Gasteiger and Dimitar Hristozov of the Computer-Chemie-Centrum of the University of Erlangen and Markus Sitzmann from the National Institute of Health for support with software and in programming questions. Simon Spycher thanks the 3R-Research Foundation for funding of this work (3R-Project 95-05).