Posters

      Flavor Ingredients in E-Vapor Products: A Structure-Based Grouping Approach to Predict Their Biological Activity

      Sciuscio, D.; Langston, T. B.; Ashutosh, K.; Smith, D. C.; Marescotti, D.; Martin, F.; Lee, K. M.; Hoeng, J.; Vanscheeuwijck, P.

      Conference date
      Mar 15, 2021
      Conference name
      Society of Toxicology (SOT) Annual Meeting 2021
      Topic
      Summary

      Electronic Nicotine Delivery Systems (ENDS) contain a wide variety of flavor ingredients. While most flavor ingredients used in today’s ENDS are ‘generally recognized as safe’ (GRAS) for oral consumption, there are insufficient data on their safety via inhalation. Considering the range of different flavors used in ENDS and the resulting flavor mixtures, it is highly impractical to test all possible flavor combinations for inhalation toxicity. In this study, we developed a structure-based approach where more than 200 flavors commonly used in ENDS were clustered in groups of structurally related compounds and a total of 38 Flavor Group Representatives (FGRs) were selected. We propose that a representative FGR mixture could then be tested both in vitro and in vivo and the data generated could support the toxicological assessment of all structurally-related individual flavors, based on the “readcross” concept that structurally-related compounds would have comparable metabolic and biological activities (as outlined in the European Commission Regulation (EC) No 1565/2000). Prior to FGR selection, toxicological predictions were made for key endpoints using the TOPKAT (e.g., pIrritancy, pCarcinogenicity, pChronic LOAEL pDevToxicity) and QSAR toolbox (e.g., pCramerClass) to fill gaps, if experimental data were not available in literature (e.g., LD50, ExpCarcinogenicity). In addition, we characterized experimentally the cytotoxic potential of more than 200 flavors by real-time cellular analysis (XCelligence) and a set of most cytotoxic substances was evaluated using High Content Screening (HCS). HCS data for the remaining flavors was obtained using a predictive model based on pCramer, pIrritancy, pChronicLOAEL, ExpCarcinogenicity and XCelligence (pToxPiHCS). Finally, flavors within each group were ranked based on: LD50, pDevToxicity, PToxPiHCS, pChronicLOAEL and pIrritancy scores, in order to select the predicted most toxic FGR for each structural group. This approach allowed the selection of FGRs that could be tested alone or in combination as mixtures to generate in vitro and in vivo data and to determine acceptable levels for their use in ENDS.