Serum-Mediated Responses in Normal and Transformed Oral Keratinocyte Lines

Serum-driven responses, many of which are related to wound healing, are potentially deregulated in cancer development and associated genomic alterations might have prognostic value. The current study assessed fetal bovine serum-induced transcriptomic changes for clinical relevance in head and neck squamous cell carcinoma (HNSCC) using oral keratinocyte models otherwise routinely cultured without serum, including normal keratinocytes (NOK) and the transformed keratinocyte lines SVpgC2a, SqCC/Y1 and LK0412. Bioinformatics-driven analysis of gene expression implicated primarily serum-induced terminal differentiation in NOK including alterations in 99 genes, 13 gene ontology-categories and 6 molecular networks and involvement of 7 key regulator genes. Compared to NOK, the transformed lines expressed around 3-fold lower numbers of differently expressed transcripts, unique sets of gene ontologies, molecular networks and key regulator genes for each line, and consistent absence of terminal differentiation markers. Assessment of the complete in vitro/serum exposure-derived set of differentially expressed genes (totally 180 genes) relative a clinical, information-rich HNSCC data set identified 17 survival-associated genes of which only 12 had previous association to HNSCC. Multi-step validation of the survival-associated genes relative to several independent tumor data sets, including in the Human Gene Expression Map and Human Protein Atlas databases, confirmed novel association to HNSCC for genes COTL1 and INSIG1 and novel poor outcome prediction for the genes CUL4B and PDGFRL. The definition of normal and aberrant serum responses in keratinocyte models therefore coupled new genes to HNSCC including with relevance to prognosis. Analysis of gene expression changes in serum-exposed normal and transformed cells relative the respective un-exposed states. Significantly differentially expressed genes were next assessed by bioinformatics processing using Gene Ontology categories and network analyses. Findings were also validated relative independent HNSCC data sets as well as transcriptomics and proteomics databases.