Abstract A1-19: A comprehensive transcriptional portrait of human cancer cell lines

Abstract

Abstract Tumor-derived cell lines have served as vital models to advance our understanding of oncogene function and therapeutic responses. They represent a wide range of tumor types and show great diversity in their response to perturbations, thereby providing an important model system to capture the disease heterogeneity that defines the cancer patient population. Although substantial effort has been directed to defining the genomic constitution of cancer cell line panels, the transcriptome remains understudied. Most previous efforts have used probe-based hybridization technologies, which fail to reveal certain RNA features, such as oncogenic fusion transcripts, non-coding transcripts, non-human RNAs and expressed single nucleotide variants (SNVs). To shed light on these important features of human cancer cell lines we generated comprehensive sequence analysis of the poly-A+ transcriptome of 675 commonly used cancer cell lines with matched SNP array data. Analysis of the RNA-seq data revealed a comprehensive portrait of expression in these cell lines. Many cell lines, despite their unique names, share a common origin. We identified 109 cell lines in our study with > 85% concordance based on SNP array measurement. All subsequent analyses were performed using a reduced set of cell lines without genomically highly similar lines. Examining the expression of genes over the entire set of cell lines showed a dominant epithelial/mesenchymal gene expression signature dividing solid tumor cell lines. Interestingly, we identified several long intergenic non-coding (linc)RNAs associated with either epithelial or mesenchymal cell state. Global correlation analysis led to the identification of new gene regulation patterns for the known oncogenes MET and EGFR. Detection of non-human sequences allows for the detection of pathogenic sequences present in cancer cell lines. We identified presence of sequence belonging to known oncogenic viruses – such as human papilloma virus and Epstein-Barr virus – as well as viruses that may be remnants of viral transfection experiments or indicative of persistent contamination. We used a splicing-aware mapping approach to 2,200 predicted in-frame gene fusion pairs. Of the 2,200 gene fusions catalogued, 1,435 consist of genes not previously found in fusions, providing many leads for further investigation. We combined the fusion results from the cancer cell lines with fusions detected in 6,730 primary tumors from The Cancer Genome Atlas (TCGA) and a database of previously described fusion genes found in literature. In total, we identified 232 cell lines that can serve as a model system to investigate fusion genes detected in clinical samples. Finally, we combine multiple genome and transcriptome features in a pathway-based approach to enhance prediction of response to targeted therapeutics. We combined multiple types of genomic aberration (mutation, copy number, fusion and overexpression) of given pathway genes to score every cell line as pathway aberrant or normal. Prediction of drug sensitivity for PIK3CA inhibitors, MEK inhibitors and an FGFR inhibitor was improved by an order of magnitude for the pathway-based approach compared to single gene predictors. It is becoming increasingly apparent that cell lines exhibit significant genomic and transcriptomic diversity comparable to the diversity observed among solid tumors. Comprehensive characterization of existing cell lines will allow selection of appropriate cell line models for biological studies and drug discovery. Our study greatly expands the current understanding of human cancer cell lines and provides a foundation for many additional discoveries that will enable further cancer studies and the successful development of novel therapeutics. Citation Format: Christiaan Klijn, Jeff Settleman, Somasekar Seshagiri, Zemin Zhang. A comprehensive transcriptional portrait of human cancer cell lines. [abstract]. In: Proceedings of the AACR Special Conference on Translation of the Cancer Genome; Feb 7-9, 2015; San Francisco, CA. Philadelphia (PA): AACR; Cancer Res 2015;75(22 Suppl 1):Abstract nr A1-19.