Discovering cancer-associated transcripts by RNA sequencing

Abstract

High-throughput sequencing of poly-adenylated RNA (RNA-Seq) in human cancers shows remarkable potential to identify uncharacterized aspects of tumor biology, including gene fusions with therapeutic significance and disease markers such as long non-coding RNA (lncRNA) species. However, the analysis of RNA-Seq data places unprecedented demands upon computational infrastructures and algorithms, requiring novel bioinformatics approaches. To meet these demands, we present two new open-source software packages - ChimeraScan and AssemblyLine - designed to detect gene fusion events and novel lncRNAs, respectively. RNA-Seq studies utilizing ChimeraScan led to discoveries of new families of recurrent gene fusions in breast cancers and solitary fibrous tumors. Further, ChimeraScan was one of the key components of the repertoire of computational tools utilized in data analysis for MI-ONCOSEQ, a clinical sequencing initiative to identify potentially informative and actionable mutations in cancer patients’ tumors. AssemblyLine, by contrast, reassembles RNA sequencing data into full-length transcripts ab initio. In head-to-head analyses AssemblyLine compared favorably to existing ab initio approaches and unveiled abundant novel lncRNAs, including antisense and intronic lncRNAs disregarded by previous studies. Moreover, we used AssemblyLine to define the prostate cancer transcriptome from a large patient cohort and discovered myriad lncRNAs, including 121 prostate cancer-associated transcripts (PCATs) that could potentially serve as novel disease markers. Functional studies of two PCATs - PCAT-1 and SChLAP1 - revealed cancer-promoting roles for these lncRNAs. PCAT1, a lncRNA expressed from chromosome 8q24, promotes cell proliferation and represses the tumor suppressor BRCA2. SChLAP1, located in a chromosome 2q31 ‘gene desert’, independently predicts poor patient outcomes, including metastasis and cancer-specific mortality. Mechanistically, SChLAP1 antagonizes the genome-wide localization and regulatory functions of the SWI/SNF chromatin-modifying complex. Collectively, this work demonstrates the utility of ChimeraScan and AssemblyLine as open-source bioinformatics tools. Our applications of ChimeraScan and AssemblyLine led to the discovery of new classes of recurrent and clinically informative gene fusions, and established a prominent role for lncRNAs in coordinating aggressive prostate cancer, respectively. We expect that the methods and findings described herein will establish a precedent for RNA-Seq-based studies in cancer biology and assist the research community at large in making similar discoveries.