Gene expression profiling and modeling of cervical cancer

Cervical cancer is the number one cancer killer in women worldwide, but we still do not know how well our cell line models compare to the actual disease. Microarrays were used to create gene expression profiles of normal cervix and cervical cancer and were compared to nine different cervical cell lines typically used in research. Normal cervical expression was compared to cervical cancer expression and 140 genes were identified as differentially expressed. These genes were subjected to strict statistical testing and were further validated by a literature search. This validation increased our confidence in the analysis methods and enabled us to compare the data to cell lines. The Pearson correlation was used to quantitatively assess the expression similarity between cervical tissue and cell lines. The Primary normal cell line was the best global model of cervical tissue, with C4-I and C4-II the next best models. Cell lines were cultured in different types of media as well as in 3-dimensions (raft) that mimics the in vivo environment. These different culture conditions had varying changes in correlation to tissue; for instance simply using a different type of media could increase the correlation. Cell lines cultured as rafts had the largest change in correlation and were the best models of cervical tissue. The correlation of specific pathways were calculated to move beyond simple global comparisons. There were many cases where the correlation of a cell line in a specific pathway increased when cultured in a different environment. The cell adhesion pathway had a much higher correlation when cells were cultured as a raft instead of monolayer. In conclusion, new biomarkers of cervical cancer were discovered and current ones reconfirmed. A quantitative analysis of how well cell lines model cervical cancer was performed and revealed that small changes to the culture environment can improve their correlation to cervical tissue. This research improved our understanding of our ability to model cancer and will help translate in vitro results into in vivo care.