Development of a high throughput in vitro toxicity screen predictive of high acute in vivo toxic potential

At an early stage of drug discovery high throughput screens are an invaluab le tool to de-select compounds with undesirable properties. A high througho ut in vitro toxicity screen has been developed and validated to identify co mpounds that have a high potential to be acutely toxic in vivo. This screen is based on treating Chinese hamster ovary (CHO) cells with test compounds for 24 h and then determining the degree of cytotoxicity by the reduction of Resazurin. Twenty-six structurally unrelated compounds were chosen that spanned a range of acute LD50 values and mechanisms of toxicity. The acute LD50 values (intraperitoneal and intravenous routes) from rat and mouse wer e taken from the RTECS database. Experimentally derived in vitro IC35 resul ts were compared to the 'most toxic' (lowest) LD50 values for each compound . The resulting correlation was statistically significant (r=0.8475). Howev er, due to the scatter of the data points, it was considered not appropriat e to rank compounds according to their degree of in vivo toxicity on the ba sis of the in vitro result. However, by defining cut-off concentrations for both the in vivo (LD50) and the in vitro (IC35) values it was possible, us ing the in vitro result (IC35<10<mu>M), to identify compounds that had a hi gh potential to be acutely toxic in vivo ('most toxic' LD50 <25 <mu>mol/kg) . Further development led to a high throughput screen capable of giving a ' Yes', 'No' or 'Borderline' classification as to whether a compound has a hi gh acute in vivo toxic potential. This screen is highly specific (no false positive classifications) and has a sensitivity of approximately 80%. This is deemed acceptable for a first tier toxicity screen at an early stage in the drug discovery process. Transfer of this screen from GlaxoSmithKline UK to sites in Italy, Spain and the USA resulted in very similar findings ind icating the inter-laboratory robustness of this screen and therefore the ab ility to compare results across the GlaxoSmithKline sites. (C) 2001 Publish ed by Elsevier Science Ltd.