To what end? - Computational tools to uncover regulators of pre-mRNA polyadenylation site selection

In eukaryotic cells, remarkably orchestrated regulatory steps ensure the availability of proteins and non-coding RNAs at the right spot, at the right time. While many of these steps, such as splicing, have been well studied since decades, the choice of the mRNA 3' end, which leads to expression of one of the many possible primary transcripts from a single locus has been recognized as key mechanism of post-transcriptional gene regulation only in recent years. Transitions between cell states have been found to be associated with specific patterns of change in poly(A) site usage, leading to coordinated changes in the length of 3' untranslated regions (3' UTRs). As 3' UTRs carry a plethora of cis-regulatory elements, their systematic shortening or lengthening has global effects on the responsivity of the transcriptome to regulation, which in turn affects essentially every aspect of RNA metabolism, including stability, transport and translation. However, the mechanisms underlying alternative polyadenylation (APA) under physiological or pathological conditions remain largely unknown. Likely, changes in poly(A) site choice are caused by changes in the availability of regulators that bind in the vicinity of poly(A) sites and impact their processing efficiency. The projects summarized in this thesis were devoted to a better understanding of the regulation of APA. Integrative analysis of a large number of data sets allowed us to establish a comprehensive annotation of poly(A) sites in the human genome. Tools developed for the projects described here could then exploit this resource to quantify and model the changes of poly(A) site usage in different contexts. In particular, the application of PAQR to quantify 3' end processing from RNA-seq data and of KAPAC to relate the abundance of individual sequence motifs to changes in poly(A) site usage led to intriguing insights into the regulation of APA in cancer. For glioblastoma, a CU-dinucleotide repeat motif was most significantly associated with the observed 3' UTR shortening, an effect that is likely to be explained by the binding of PTBP1, a factor previously known for its role in splicing regulation. Together with HNRNPC, another splicing factor that was implicated in the regulation of poly(A) site choice through analyses presented here, these results suggest an extensive coupling between splicing and 3' end processing. In particular, it appears that many regulators of both mRNA maturation steps exist and remain to be uncovered. Previous results from glioma cell lines indicated that PTBP1 levels directly affect proliferation and migration. Considering its role in splicing and 3' end processing, PTPB1 may emerge as an important regulator of gene expression with direct implications for tumor progression in glioblastoma. Potentially, PTBP1 can serve as therapeutic target or diagnostic marker in brain tumor. In summary, the work of this thesis illustrates how the deployment of computational tools can condense the information contained in large-scale data sets into biologically relevant results, shedding light on novel aspects of mRNA 3' end processing in physiological and pathological conditions. The uncovered regulators may be amenable to targeting by small molecules, thereby restoring the RNA processing patterns specific to the healthy states