Concepts of predictive toxicology.

‬

Currently, much effort is invested into the development of non-animal testing strategies to identify the potential of compounds to induce systemic toxicity (Hammad, 2013[9]; Stewart and Marchan, 2012[19]). Organotypical in vitro systems are particularly popular in the fields of kidney (Limonciel et al., 2012[15]; Jennings et al., 2012[11]; Valente et al., 2012[20]), heart (Maayah et al., 2014[16]; Bonifacio et al., 2014[1]), liver (Grinberg et al., 2014[8]; Godoy et al., 2013[7]; Schug et al., 2013[18]) and developmental toxicity (Weng et al., 2014[22]; Waldmann et al., 2014[21]; Krug et al., 2013[14]). However, it is also clear that in vitro systems represent valuable tools to study certain mechanisms and endpoints but do not reach the complexity of organs or organisms (Ghallab, 2013[6]).

Recently, Daston et al. (2014[3]) published a concept, how future research on non-animal methodology should be designed to overcome current limitations. The authors recommend two complementary and interconnecting concepts. A first work stream should focus on toxicity characterization. Here, critical biological targets and mechanisms leading to toxic effects should be elucidated based on in vitro systems. For this purpose methods such as high-throughput and high content screening and computational modelling will be applied (Daston et al., 2014[3]). A second work stream should focus on translation into regulation. Specific aims are for example methods for grouping, read-across strategies and in vitro methods to derive no-effect levels. In recent years much has been written about grouping strategies and general concepts to improve chemical risk evaluation (Geenen et al., 2012[5]; Kalkhof et al., 2012[12]; Keller et al., 2009[13]; Renwick, 2004[17]; Zbinden, 1993[23]; Gebel et al., 2014[4]; Calabrese, 2013[2]). The present concept paper of Daston and colleagues (2014[3]) belongs certainly to the most fundamental papers in this field and is a must-read for anyone interested in predictive toxicology and alternative methods. However, the authors neglect one major limitation of their strategy: The concept may lead to reasonable predictions for chemicals with unspecific mechanisms of action, meaning that many mechanisms are simultaneously active that lead to the breakdown of cellular functions. However, the concept may fail for highly specific mechanisms of action. The reason for this limitation is that Daston et al. (2014[3]) in agreement with the SEURAT concept recommend focusing on 'critical biological targets' in in vitro systems only (Jennings et al., 2014[10]). This bears the risk of establishing an illusory in vitro world which lacks critical components of real organs or organisms. Let us assume a compound specifically inhibits reabsorption of bile salts in cholangiocytes in bile ducts. How should this mechanism be recognized in an in vitro system that contains hepatocytes only? Moreover, it cannot be excluded that a compound may alter kidney cells in a way that triggers the attack of immue cells. Can we be sure that this specific mechanism would be identified in an in vitro system containing renal proximal tubular epithelial cells only? A research program to in vitro systems only has a high probability to fail. Therefore, a third work stream is painfully missing in the concept of Daston et al. (2014[3]); namely research that systematically compares mechanisms of toxicity in vitro and in vivo. Do the currently available in vitro systems really recapitulate the mechanisms that finally lead to adverse effects in vivo? Finally, it should not be ignored that many mechanisms leading to toxicity in vivo are far from being fully understood. Further research is needed to identify key mechanisms of toxicity in vivo to be able to establish in vitro systems recapitulating these mechanisms. Although it may seem paradox: the successful development of non-animal methodology requires animal experiments.