Analyses of transcriptomic responses generated by hepatocarcinogens in a battery of liver-based in vitro models

For assessing the cancer-causing potential for humans of a chemical compound, the conventional approach is the use of the 2-year rodent carcinogenicity bioassay, thus alternatives such as in vitro toxicogenomics are highly desired. In the present study, the transcriptomics responses following exposure to genotoxic (GTX) and non-genotoxic (NGTX) hepatocarcinogens and non-carcinogens (NC) in five liver-based in vitro models, namely conventional and epigenetically-stabilized cultures of primary rat hepatocytes, the human hepatoma-derived HepaRG and HepG2 cell lines and the human embryonic stem cell-derived hepatocyte-like cells hES-Heps are examined and compared. The global gene expression profiles of five commonly used liver-based in vitro systems are investigated, namely conventional cultures of primary rat hepatocytes (HepsC), Trichostatin A (TSA)-stabilized cultures of primary rat hepatocytes (HepsT), human embryonic stem cell-derived hepatocyte-like cells (hES-Hep), HepG2 and HepaRG cells. These models are exposed to 15 prototypical compounds which have been carefully chosen and belong to 3 toxic classes i.e. (i) genotoxic (GTX) (aflatoxin B1, AFB1; 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone, NNK; 2-nitrofluorene, 2NF; benzo(a)pyrene, BaP; cyclophosphamide, CYCLO), (ii) non-genotoxic (NGTX) (methapyrilene hydrochloride, MPH; piperonylbutoxide , PIPB; Wy-14643, WYE; phenobarbital sodium, SPB; 12-O-tetradecanoylphorbol-13-acetate, TPA) carcinogens and (iii) non-carcinogens (NC) (nifedipine, NIF; clonidine, CND; D-mannitol, MAN; tolbutamide, TOL; diclofenac sodium, SDF). For the gene- and pathway- based analysis, the raw microarray data was re-annotated to Ensembl version 61 genome and Gene Chip Robust Multi-array Average (GC-RMA) normalized (Dai et al. 2005). This resulted in 11,187 and 18,919 probe sets for the rat and the human models, respectively. For the classification analysis, the raw microarray data was annotated using the chip description files from Affymetrix and then GC-RMA normalized. In order to eliminate batch effects, data from the different studies were half-z normalized. Half-z-normalization adjusts the logarithmic expression values of transcripts within a group such that each transcript has zero mean.