PTM-POSE: A computational tool for exploring the functional consequences of splicing-control of post translational modifications (PTMs) in different biological contexts

Introduction: Both post translational modifications (PTMs) and alternative splicing are important for modulating protein function and interactions, but they are rarely studied in tandem. Isoform-specific PTM information tends to be limited in proteomic databases due to an inability of traditional proteomic experiments to effectively capture unique protein isoforms. By merging information from genomic and proteomic databases, we previously illustrated that almost a third of post-translational modifications are regulated by alternative splicing, potentially influencing the functional role of these proteins within different tissue- and disease-contexts. Many diseases, including cancer, can be characterized by the dysregulation of both splicing regulation and PTM networks, often driven by mutations or copy number alterations of either kinases, which mediate phosphorylation, or splicing factors, which guide expression of specific isoforms. Further, splicing factors are themselves commonly regulated by PTMs like phosphorylation, suggesting the potential for crosstalk between different cell signaling and splicing regulatory networks. However, the extent of this relationship is poorly understood, with only a few available tools that allow for a limited exploration of the PTMs that are affected by splicing events in specific biological contexts. Here, we have developed a computational tool, called PTM Projection Onto Splice Events (PTM-POSE, https://github.com/NaegleLab/PTM-POSE), to annotate the outputs of most available splicing quantification tools with PTM sites and analyze functional consequences of these changes. We then applied PTM-POSE to RNA-binding protein (RBP) related datasets from TCGA, ENCORE, and other publicly available datasets to identify coordinated regulation of specific signaling pathways and kinases by individual splicing factors.

Materials and

Methods: To project PTMs onto splice events, we first utilize our in-house computational pipeline (ExonPTMapper: https://github.com/NaegleLab/ExonPTMapper) to obtain the genomic coordinates of all known PTMs for the most recent Ensembl build (GRCh38.p14), using all experimentally identified PTMs recorded in ProteomeScout and PhosphoSitePlus. With PTM-POSE, we can then take this input and annotate any quantified splicing events by projecting PTMs onto impacted genomic regions using only four types of information outputted by most splicing quantification tools: chromosome, DNA strand, and the start and end of the impacted genomic region. We then further annotate the splicing-controlled PTMs with functional annotations pulled from various gene-level (Reactome, Gene Ontology), exon-level (Exon Ontology, DIGGER, NEASE), and PTM-level (PhosphoSitePlus, RegPhos, PTMcode, KSTAR) databases to identify significant individual PTM sites and broadly impacted PTM-regulated biological processes. To test the utility of this approach, we first applied PTM-POSE to patient-specific measurements of exon inclusion from the The Cancer Genome Atlas (TCGA) prostate cancer cohort downloaded from the TCGASpliceSeq database, focusing on the role of ESRP1 expression on survival outcomes, a splicing factor important in regulating epithelial specific splicing events that has been implicated in worsened prognosis in several cancers. Next, given that many splicing factors are commonly regulated by phosphorylation, we downloaded RBP-knockdown data from the Encyclopedia of RNA Elements (ENCORE), applied PTM-POSE to the splicing events identified by MATS upon RBP-knockdown, and extracted the phosphorylation sites whose inclusion across isoforms was regulated by each RBP.

Results, Conclusions, and Discussions: We identified a total of 317 PTM-containing exons (1238 total PTMs) whose inclusion across isoforms is related to ESRP1 expression across the TCGA prostate cancer cohort. While many different types of PTMs were identified, the majority were phosphorylation sites (877 of the 1238 PTMs). Given this, we were curious whether certain kinase targets were more likely to be impacted by ESRP1-mediated splicing than others, perhaps indicating changes to the regulation of specific signaling pathways. Using a kinase enrichment analysis adapted from our kinase activity inference algorithm, KSTAR, we found that substrates of SGKs, PDK1, and GRK2 were enriched in patients with high ESRP1 expression, suggesting that these kinases may play a more important role in cancer progression in these tumors. Of particular note, both SGK and PDK1 are involved in PI3K/PTEN/mTOR signaling, and we found that patients exhibiting both low PTEN and high ESRP1 expression had significantly worse progression-free survival than patients with either condition alone, suggesting that ESRP1 may increase cancer progression by elevating mTOR signaling. As many other splicing factors and kinases have been associated with cancer progression, we became interested in identifying the extent of crosstalk between splicing regulation and cell signaling through specific splicing factors. From application of PTM-POSE to a suite of RBP-knockdown experiments from ENCORE, we found specific RBPs exhibited distinct regulation on phosphorylation sites and kinases in K562 cells, with the less than 1% median overlap between phosphorylation sites regulated by each splicing factor. Substrates of certain kinases like PRKD1, ROCK1, and AURKA were more commonly enriched among the splicing-controlled phosphorylation sites. From this work, we have begun to construct the first RBP-kinase crosstalk network, which will help to uncover mechanisms by which changes to expression or activity of various splicing factors and kinases can lead to splicing-induced changes to cell signaling and phenotype, such as ESRP1’s connection to SGK1/mTOR signaling. Using this information, we will be able to better identify the consequences of alternative splicing in disease, potentially identifying new therapeutic targets for patients with overactive kinases and/or splicing factors.

Acknowledgements (Optional): The results discussed here are in part based upon data generated by the TCGA Research Network: https://www.cancer.gov/tcga. Research reported in this publication was supported by the National Institute Of General Medical Sciences of the National Institutes of Health under Award Number R35GM138127 and by the National Science Foundation Graduate Research Fellowship under Grant No. 1842490.