KOPI: Kinase inhibitOr Proteome Impact analysis

www.nature.com/scientificreports OPEN KOPI: Kinase inhibitOr Proteome Impact analysis Ginny Xiaohe Li1,4,5, Tianyun Zhao2,3,5, Loo Chien Wang2, Hyungwon Choi1, Yan Ting Lim2,5* & Radoslaw M. Sobota2* Kinase inhibitors often exert on/off-target effects, and efficient data analysis is essential for assessing these effects on the proteome. We developed a workflow for rapidly performing such a proteomic assessment, termed as kinase inhibitor proteome impact analysis (KOPI). We demonstrate KOPI’s utility with staurosporine (STS) on the leukemic K562 cell proteome. We identified systematically staurosporine’s non-kinome interactors, and showed for the first time that it caused paradoxical hyper- and biphasic phosphorylation. Kinases catalyse phosphorylation of serine, threonine or tyrosine residues on target substrate. Their dysregulation underlies many disorders, making them an important protein target class1. Kinase-targeted drug discovery has generated several classes of kinase inhibitors (KIs), of which more than 70 are used c linically2,3. The current strategy seeks to inhibit the aberrant kinase and the hyperactivated pathway, and refine its specificity within the kinome4–8. The assumption that an overactive kinase or hyper-phosphorylation drives oncogenesis has nonetheless been repeatedly challenged. KIs can behave as agonists in vitro9,10. Kinases also show non-catalytic functions such as competition for protein interactions or exert allosteric effects on partner proteins that regulate critical biochemical pathways11,12. KIs, though designed to target specifically within the kinome, can also interact with non-kinase targets13,14. Hence, a systems-wide understanding of the mode of action for KIs gives the first cue to its potential as a viable therapeutic agent. An ensemble of experimental and computational workflows that complements existing kinase-centric compound refinement processes can facilitate the selection of the lead compound and contribute to its success. Proteomics coupled with thermal shift assays can provide an unbiased view of a compound’s effect on the proteome by quantifying its effects on protein expression and turnover, phosphorylation and the i nteractome15–17. A critical bottleneck to the analysis is the lack of an open-source tool for performing user-friendly analytics on such data. Although a few solutions are freely available for analysing thermal stability data on B ioconductor18–20, phosphoproteomics for target tracking is not a part of the process. KOPI fills this gap by providing seamless analysis of drug-induced proteome dynamics based on both thermal stability and phosphoproteomics experiments (Fig. 1A–E). Results To demonstrate KOPI, we generated a training data set with K526 cells treated with staurosporine, a potent KI chosen for its polypharmacology. The incubation duration was 30 min to emphasize transient events such as protein turnover and post-translational modification over protein transcription and translation. We performed thermal proteome profiling using the isothermal dose-response20. The cells were treated with nine different drug doses up to 10 µM and heat pulsed at 37 °C and 52 °C (Fig. 1A), and the soluble proteome was prepared for quantitative proteomic and phosphoproteomic analysis in each experiment (Fig. 1B, Supplementary Information). KOPI is implemented as a web-based Rshiny application, accounting for three levels of proteomic data concurrently; constitutive protein abundance, thermal stability (Fig. 1B,D) and phosphopeptide abundance (Fig. 1B,E). A critical departure from the existing methods is that we did not limit the dose–response analysis to sigmoidal curves. The existing methods select hits with a coefficient of determination above a set threshold, and use the compound’s saturating concentrations for hit identification18,20,21. Instead, we generated a metric, the average slope, to describe the protein/peptide’s dose–response (Fig. 1C). Each dose–response curve was smoothened by a Gaussian kernel with a user-defined bandwidth parameter to summarize the overall trend. Its 1 Department of Medicine, Yong Loo Lin School of Medicine, National University of Singapore, Singapore, Singapore. 2Functional Proteomics Laboratory, Institute of Molecular and Cell Biology, Agency for Science, Technology and Research, Singapore, Singapore. 3School of Biological Sciences, Nanyang Technological University, Singapore, Singapore. 4 Department of Pathology, University of Michigan, Ann Arbor, Michigan, USA. 5These authors contributed equally: Ginny Xiaohe Li, Tianyun Zhao and Yan Ting Lim. *email: limyt@ imcb.a-star.edu.sg; Scientific Reports | (2022) 12:13015 | https://doi.org/10.1038/s41598-022-16557-w 1 Vol.:(0123456789) www.nature.com/scientificreports/ A F Intact cell ITDR with staurosporine 10 37 C o + 52oC + Phosphopeptide enrichment D m/z SRM high 0 MARK2 NUP133 0 BMP2K KRT18 UQCRC1 VIM NUP155 AGL BCR CDK11B MYLK CAMKK2 DVL3 GRWD1 TFAM SPR CCNC DOPEY2 HMGN2 PDPK2P MRPL3 MRPS27 HEBP1 ALDH1B1 SNX5 MAP2K4 SNX6 MRPL2 LUZP1 −1 CTNNB1 CENPC AURKA SNRNP70 HNRNPA1L2 MRPL47 Biphasic 468 (5.5%) 2 Hypo 1027 (12.0%) CAND2 MTX1 ProteomeImpact PRKAR2A destabilizing HNRNPDL MRPL46 TMEM109 HADHA 0.0 SLC25A24 GSK3A PDPK1 TXNDC15 CSNK2A2 RB1 FUS MRPS22 HNRNPA1 NARS2 DPP9 C1QBP CDK2 G 0.4 0.8 1.2 Hyper 78 (0.9%) low high Non−responsive 6972 (81.6%) 1 hyper-phosphorylation hypo-phosphorylation high PAK4 CHEK1 GSK3B STK38 stabilizing biphasic 1.0 AAK1 MARK3 −0.4 non-responsive 0 Average slope 37oC low AURKB Abundance (average slope at 37oC) Phosphorylation patterns 0 MAPK8 MAPK14 1 52oC 2.0 1.0 decreasing high SLK Thermal stability increasing 2.0 FDXR MCCC1 STK26 + E CSK FECH concentration low COL4A3BP PWP1 PHKB soluble fraction low high 0 CAMK2D MAP3K7 rep1 rep2 37oC 1.0 PMPCA SUCLA2 Protein expression 2.0 CDK5 PMPCB STK25 2 Curve fitting and average slope calculation low GAK Thermal stability (average slope at 52oC) C IRAK4 OXSR1 Multiplex quantitative MS Soluble fraction harvest and tryptic digest Intensity B 3 10 µM 0 µM CAMK2G CAMKK1 hyper-phosphorylation −1 biphasic hypo-phosphorylation 0 1 2 PhosphoResponse 3 4 Number of continuous upward intervals 5 6 7 8 Figure 1.  Kinase inhibitor proteome impact analysis (KOPI) workflow and outputs. (A) The cell line is treated with kinase inhibitor/s at different concentrations under normal (37 °C) temperature and a heat challenge (52 °C). (B) The soluble fraction is harvested for tryptic digestion and multiplexed for quantitative mass spectrometry. (C) The fold changes relative to the untreated are fitted with a smoothing curve. (D) The average slope per curve is calculated to indicate the proteins’ response to drug treatment. The same analysis is performed on the enriched phospho-peptides to quantify the phosphoproteome’s resp (...truncated)