2012 SOT Society of Toxicology annual meeting
L’Oréal Research and Innovation, ImmunoSearch

IRR-IS® - EpiSkinTM based model for Irritation - Evaluation of IRR-IS®, an EpiSkinTM based model for Quantifying Chemical Irritation Potency

Irritation potency classification assays using In vitro 3D reconstructed epidermis models and viability measurement (MTT test) have been developed and validated by ECVAM. The ability of these methods to quantify skin irritation potency in terms of classification according to the EU GSH rules is satisfactory. Validated methods only predict two classes including irritants (Cat. 2) and non Cat. substances. Nevertheless, quantification of the skin irritation potential is of high importance not only for transport labeling and occupational health but also for the toxicological risk evaluation and should address the question of intermediate irritation levels (moderate irritants) in line with new and future regulatory demands.

ImmunoSearch developed IRR-IS®, a new method based on the quantitative analysis of specific biomarkers expressed in 3D reconstructed epidermis (EpiSkinTM). The aim was to provide a tool able to predict irritancy potency (from mild to severe) and to support the risk assessment approach and possible keys to get closer to potency assessment. The selection of tuned set of biomarkers was done by analysis expression profiles in 3D reconstructed epidermis with several reference irritants. Test chemicals were topically applied neat for 30 min then washed and the tissues further incubated for 6 hrs. Tissues were teased, total RNA purified with Trizol and reverse transcription performed. Genes expression was determined by quantitative PCR after. We selected 25 biomarkers and developed a specific algorithm based on the analysis of gene expression magnitude.

A reference set of 45 coded irritant chemicals from the public domain was provided by L’Oréal and tested blind by ImmunoSearch. Sensitivity (irritant) >90%, specificity (non irritant) > 70% and accuracy >80 % were comparable to the validated EpiSkin model performance. These preliminary good results need to be applied to larger sets in order to refine the algorithm.