PF223 CHARACTERIZATION OF KINASE ACTIVITIES INFERRED BY PHOSPHOPROTEOMICS IN MYELOID CELL LINES TREATED WITH TARGETING COMPOUNDS FOR THE IDENTIFICATION OF DRIVING AND BYPASSING ONCOGENIC SIGNALING PATHWAYS

Background: Chronic Myeloid Neoplasms (MN), such as Myelodysplastic Syndromes and Myeloproliferative Neoplasms, are heterogeneous disorders caused by sequential accumulation of genetic lesions in haematopoietic stem cells (HSC). They are characterized by cytopenia or cytosis with a propensity to evolve towards secondary acute myeloid leukemia (sAML). The response to targeted treatment as well as clonal evolution from chronic MN to sAML is currently not faithfully predictable based on genetic approaches alone. Therefore, identification of more reliable biomarkers than mutational status is required to improve stratification of patients in potential responders or non‐responders to targeting compounds. This goal remains a relevant and unmet need in the post‐genomic era of “precision medicine”. Aims: The aim of this project is to build a bioinformatics pipeline that uses phosphoproteomic (PP) data to i) identify differentially phosphorylated sites and ii) infer targetable kinases of overactive oncogenic pathways. Here, we present the validation phase focusing on PP characterization and the inference of kinase‐activity from two myeloid cell lines perturbed with two kinase inhibitors. Methods: To validate our analysis pipeline, two myeloid cell lines K562 and MOLM13, driven by BCR‐ABL1 and FLT3 kinases, respectively, were used. They were inhibited with kinase inhibitors Nilotinib (NILO) and Midostaurin (MIDO), respectively. PPs were enriched with titanium‐dioxide and analyzed by mass spectrometry (nanoLC‐MS 2 ). A Kinase Activity Enrichment Analysis (KAEA) pipeline was developed in R to infer kinase activity. This pipeline is based on the enrichment algorithm in SetRank package integrating substrate‐kinase datasets from five experimentally validated databases complemented by NetworKIN in‐silico predictions. The pipeline is supported by a shiny web‐app interface to visualize the analysis and allow interactive interrogation of the data on three levels: differential phosphosites, kinase activity and involved oncogenic signaling pathways. Results: K562 treated with NILO showed expected inhibition of ABL1, KIT and down‐stream effectors of MAPKs as well as additional vulnerable kinases such as RPS6K, MET and RET (Figure 1A). MOLM13 treated with MIDO showed expected inhibition of PRKC and downstream kinases of FLT3 (AKT1, JAK and MAPK) as well as additional vulnerable kinases such as RPS6K and SGK1 (Figure 1B). At the same time, overactive kinases emerged in both cell lines such as CDK1/2, CSNK1D and PRKs in K562 and TLK2, CLK1 and Casein‐kinases in MOLM13. Summary/Conclusion: Our perturbed cell line data validated the utility of our analysis pipeline and showed the ability to detect expected inhibition of kinase activities as well as additional vulnerable and bypassing kinases from differential PP MS data. Overactive kinases may represent alternative pathways for cancer cells to escape killing by targeting compounds and are potential candidates for combinatorial treatment. Our pipeline has, therefore, the potential to provide an unsupervised characterization of dynamic changes of kinase‐activities, characterize mechanisms of response/resistance and identify targets for combinatorial treatment in primary patient samples with myeloid neoplasms. We are currently validating our analysis pipeline in primary AML patient samples.