Abstract LB-197: An alternative splicing signature that identifies breast cancer stem cells

Abstract

Abstract Tumors frequently contain a sub-population of cells, referred to as cancer stem cells (CSCs) or tumor-initiating (TI) cells, with the abilities to self-renew and to regenerate all cell types within the tumor. Recent studies suggest that the process of epithelial-mesenchymal transition (EMT) may contribute to generation of CSCs within the tumor. Although alternative splicing has been shown to play a role in the biology of EMT, the occurrence and role of alternative splicing in CSC biology is largely unexplored. Given the role of CSCs in the recurrence and spread of cancer, there is an urgent need to develop new agents that target CSCs. Development of CSC-targeted drugs will be greatly facilitated by biomarkers that can identify CSCs to aid in patient selection and determination of drug response. Analysis of alternative splicing in CSCs may provide valuable new CSC-specific markers. In this study, a TI gene expression signature (Creighton et al, 2009), an EMT gene expression signature (Gupta et al., 2009, Taube et al, 2009) and a Basal B/Luminal breast cancer subtype classifier were used in Support Vector Machine (SVM) analysis of 41 human breast cancer cell lines to identify changes in alternative splicing. We discovered 209 cassette exon splicing events from the union of these 3 classifiers, of which 68 splicing events were concordant. Interestingly, GO and KEGG pathway analysis using these 68 alternatively spliced events demonstrated enrichment of genes encoding key drivers of CSC phenotype, including cell migration, motility, and cell adhesion pathways, as well as extracellular matrix-receptor interactions. SVM analysis of an independent NCI-60 cancer cell line dataset determined that the top 60 exons from the breast cancer cell line training group identified 96% of the CSC-high cell lines and 90% of CSC-low cell lines with high accuracy. To extend the analysis to human tumor samples, we assessed the whole transcriptome data from human breast cancers (81 patients, Lin et al., 2009) using a CSC centroid gene signature model that clustered tumor samples into CSC-high and CSC-low subgroups. Interestingly, a centroid model based on the 68 alternative splicing events similarly identified the CSC-high and CSC-low breast cancers with high accuracy. The CSC-high subgroup contained mostly triple negative breast cancers, known to have increased frequencies of CSC and EMT phenotypes, suggestive of the therapeutic importance of this alternative splicing signature for identification of CSCs in patients. Q-PCR analysis showed that several of the alternative splicing events observed in the CSC-high subgroup were further enriched in tumorspheres grown from human breast cancer cell lines. The CSC-associated alternative splicing signature identified here will be further refined to develop new CSC-specific diagnostic markers to stratify breast cancer patients and monitor response to novel CSC-targeted therapies. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr LB-197. doi:1538-7445.AM2012-LB-197