In vitro human ion channel assays predictive of drug-induced seizure

Seizure is among the most severe FDA black box warnings of neurotoxicity reported on drug labels. Gaining a better mechanistic understanding of off-targets causative of seizure will improve the identification of potential seizure risks preclinically. In the present study, we evaluated an in vitro panel of 9 investigational (Cav2.1, Cav3.2, GlyRA1, AMPA, HCN1, Kv1.1, Kv7.2/7.3, NaV1.1, Nav1.2) and 2 standard (GABA-A, NMDA) ion channel targets with strong correlative links to seizure, using automated electrophysiology. Each target was assessed with a library of 34 preclinical compounds and 10 approved drugs with known effects of convulsion in vivo and/or in patients. Cav2.1 had the highest frequency of positive hits, 20 compounds with an EC30 or IC30 ≤ 30 µM, and the highest importance score relative to the 11 targets. An additional 35 approved drugs, with categorized low to frequent seizure risk in patients, were evaluated in the Cav2.1 assay. The Cav2.1 assay predicted preclinical compounds to cause convulsion in nonclinical species with a sensitivity of 52% and specificity of 78%, and approved drugs to cause seizure in nonclinical species or in patients with a sensitivity of 48% or 54% and specificity of 71% or 78%, respectively. The integrated panel of 11 ion channel targets predicted preclinical compounds to cause convulsion in nonclinical species with a sensitivity of 68%, specificity of 56%, and accuracy of 65%. This study highlights the utility of expanding the in vitro panel of targets evaluated for seizurogenic activity, in order to reduce compound attrition early on in drug discovery.