Large-scale analysis of genome and transcriptome alterations in multiple tumors unveils novel cancer-relevant splicing networks

Alternative splicing is regulated by multiple RNA-binding proteins and influences the expression of most eukaryotic genes. However, the role of this process in human disease, and particularly in cancer, is only starting to be unveiled. We systematically analyzed mutation, copy number, and gene expression patterns of 1348 RNA-binding protein (RBP) genes in 11 solid tumor types, together with alternative splicing changes in these tumors and the enrichment of binding motifs in the alternatively spliced sequences. Our comprehensive study reveals widespread alterations in the expression of RBP genes, as well as novel mutations and copy number variations in association with multiple alternative splicing changes in cancer drivers and oncogenic pathways. Remarkably, the altered splicing patterns in several tumor types recapitulate those of undifferentiated cells. These patterns are predicted to be mainly controlled by MBNL1 and involve multiple cancer drivers, including the mitotic gene NUMA1. We show that NUMA1 alternative splicing induces enhanced cell proliferation and centrosome amplification in nontumorigenic mammary epithelial cells. Our study uncovers novel splicing networks that potentially contribute to cancer development and progression.Alternative splicing alterations are emerging as important signatures to further understand tumor formation and to develop new therapeutic strategies (Grosso and Carmo-Fonseca 2014). Specific alternative splicing changes that confer tumor cells with a selective advantage may be caused by mutations in splicing regulatory sequences (Dorman et al. 2014) and/or regulatory factors (Brooks et al. 2014). Various splicing factors have been described to be mutated in cancer, including SF3B1, SRSF2, ZRSR2, and U2AF1 in myelodysplastic syndromes and lymphoid leukemias (Yoshida et al. 2011), RBM10 and U2AF1 in lung tumors (Imielinski et al. 2012; Brooks et al. 2014), and SF3B1 in breast tumors (Ellis et al. 2012; Maguire et al. 2015). These mutations generally impair the recognition of regulatory sites, thereby affecting the splicing of multiple genes, including oncogenes and tumor suppressors (Kim et al. 2015). On the other hand, increasing evidence shows that changes in the relative concentration of splicing factors can also trigger oncogenic processes. For instance, splicing factors from the SR and hnRNP families are overexpressed in multiple tumor types and induce splicing changes that contribute to cell proliferation (Karni et al. 2007; Golan-Gerstl et al. 2011). Similarly, down-regulation of splicing factors that act as tumor suppressors has also been observed (Wang et al. 2014; Zong et al. 2014).Importantly, specific alternative splicing events can substantially recapitulate cancer-associated phenotypes linked to mutations or expression alterations of splicing factors. This is the case of NUMB, for which the reversal of the splicing change induced by RBM10 mutations in lung cancer cells can revert the proliferative phenotype (Bechara et al. 2013). Events that contribute to cancer are often controlled by multiple factors, like the exon skipping event of MST1R involved in cell invasion, which is controlled by SRSF1 (Ghigna et al. 2005), HNRNPA2B1 (Golan-Gerstl et al. 2011), HNRNPH1, and SRSF2 (Moon et al. 2014). Furthermore, some events may be affected by both mutations and expression changes in splicing factors. For instance, mutations in RBM10 or down-regulation of QKI lead to the same splicing change in NUMB that promotes cell proliferation (Bechara et al. 2013; Zong et al. 2014). Alternative splicing changes that characterize and contribute to the pathophysiology of cancer (Sebestyén et al. 2015) are thus potentially triggered by alterations in a complex network of RNA binding proteins, which remain to be comprehensively described. To elucidate the complete set of alterations in these factors and how they globally affect alternative splicing that may contribute to cancer, we analyzed RNA and DNA sequencing data from The Cancer Genome Atlas (TCGA) project for 11 solid tumor types.