diaPASEF-Powered Chemoproteomics Enables Deep Kinome Interaction Profiling

Abstract

Kinases control most cellular processes through protein phosphorylation. The 518 human protein kinases, i.e., the kinome, are frequently dysregulated in human disease. Kinase activity, localization, and substrate recognition are controlled by dynamic PPI networks composed of scaffolding and adapter proteins, other signaling enzymes, and phospho-substrates. Mapping kinome PPI networks can, therefore, quantify kinome activation states and kinase-mediated cell signaling, and can be used to prioritize kinases for drug discovery. We introduce our 2^nd generation (gen) kinobead competition and correlation analysis (kiCCA) for kinome PPI mapping. 2^nd gen kiCCA utilizes kinome affinity purification with kinase inhibitor soluble competition, data-independent acquisition with parallel accumulation serial fragmentation (diaPASEF) mass spectrometry (MS), and a redesigned CCA algorithm with improved selection criteria and the ability to predict multiple kinase interaction partners of the same proteins. Using neuroblastoma cell line models of the noradrenergic-mesenchymal transition (NMT), we demonstrate that 2^nd gen kiCCA (1) identified 6-times more kinase PPIs in native cell extracts compared to our 1^st gen approach, (2) determined kinase-mediated signaling pathways that underly the neuroblastoma NMT, and (3) accurately predicted pharmacological targets for altering NMT states. Our 2^nd gen kiCCA approach is broadly useful for cell signaling research and kinase drug discovery.

Keywords: ALK inhibitor; chemoproteomics; kinome; neuroblastoma; noradrenergic-mesenchymal transition; protein−protein interaction.

Figure 1.

Workflow overview and performance of diaPASEF-powered kinobead AP-MS. (A) Detailed overview of our diaPASEF kinobead AP-MS workflow, which enables both kinome-wide KI selectivity profiling and kinase PPI mapping using our kiCCA approach. (B) Comparing the number of quantified kinases and non-kinase proteins found in our diaPASEF kinobead AP-MS analysis of the NB cell lines SK-N-SH and SH-SY5Y to our previous kinobead AP-MS approach. Refers to Fig. 1A. (C) Number of kinases (left panel) and reported non-kinase interaction partners of these kinases (right panel) that were significantly competed off the kinobeads with our 21 KIPs compared to our previous experiments. Results of diaPASEF kinobead AP-MS assays with soluble competition. Statistics: two-sample Student’s t-test p-value ≤ 0.05, log2 FC ≥ 0, N = 2. Refers to Fig. S1B, S1C, and Table S1. (D) Kinases competed at 1 μM competitor concentration of the ALK inhibitors crizotinib, ceritinib, entrectinib, and lorlatinib as determined by diaPASEF kinobead AP-MS with soluble competition in SK-N-SH cell lysates. Statistics: two-sample Student’s t-test, Benjamini-Hochberg (BH)-FDR ≤ 0.05, log2 FC ≥ 0, N = 4. Refers to Table S1. (E) Comparison of significantly competed ALK inhibitor targets from our diaPASEF kinobead AP-MS assay with previously reported targets of these inhibitors. Statistics: see (D). Refers to Table S1. (F) Kinobead AP-MS soluble competition with ALK inhibitor titration at eight different concentrations ranging from 10 μM to 3.33 nM (3-fold dilutions). Curves were plotted and fitted using Origin 2025. Error bars are the S.D. Refers to Fig. S1F and Table S1.

Figure 2.

Re-defining the selection criteria for the CCA. (A) Parameter scanning experiment determining the CCA score (S_CCA) for kinase PPIs predicted by our CCA in the SK-N-SH cell line. Increasingly stringent Student’s t-test log2 FC and p-value cut-off criteria were applied and only kinases or non-kinases fulfilling these criteria were used for CCA. Statistics: two sample Student’s T-test, range of log2 FC ≥ 0 to ≤ 5, and p ≥ 0 to ≤ 0.001. Refers to Fig. S2A. (B) Plotting the S_CCA for varying log 2 FC cut-offs and p-Value cut-offs against one another identified optimal mixed cut-off criteria for kinases input into our CCA. Refers to Fig. S2B. (C) Venn diagrams showing the overlap of reported kinase PPIs that were predicted by our 1^st gen CCA algorithm and PPIs predicted by our 2^nd gen CCA algorithm. Both 1^st and 2^nd gen CCA used the same diaPASEF kinobead AP-MS data from the two NB cell lines SK-N-SH and SH-SY5Y as the input. Refers to Table S1. (D) Heatmap showing the %TDR for PPIs achieved by our 2^nd gen CCA separated into 0.05-unit intervals of the CCA Pearson’s r-value for the 1^st through 10^th predicted kinase interactor of a non-kinase protein. The %TDR reflects the percentage of BioGRID-reported vs. all predicted PPIs. Refers to Fig. S2C. (E) STRING (v12.0) network for interaction partners of the kinases AAK1, BMP2K, and BCR, as predicted by 2^nd gen CCA. Only interactions with a 2^nd gen kiCCA %TDR > 10 were used to construct the networks. (F) Volcano plot showing the results of a co-IP/MS experiment using a specific antibody that binds AP2B1. All proteins shown were predicted to be members of this network by 2^nd gen kiCCA with a %TDR > 10% (see (E)). Statistics: two sample Student’s t-test p < 0.05, Log2 FC > 0, N = 3. Refers to Fig. S2E and S2F. (G) Venn diagrams comparing the total number of predicted kinase PPIs of 1^st predicted kinase vs. 1^st – 10^th predicted kinase interactor of a protein, as was determined by 2^nd gen CCA in the NB cell lines SK-N-SH and SH-SY5Y. All interactions with a %FDR > 10 were used. Refers to Table S1.

Figure 3.

Mapping changes in kinome PPIs that are associated with the NB cell NMT using our 2^nd gen kiCCA approach. (A) Schemata: the noradrenergic-neuronal SH-SY5Y cell line and the isogenic mesenchymal-like SK-N-SH cell line serve as in vitro model for the NB cell NMT. (B) Volcano plot showing the differential abundance of kinases between the two opposing NMT states represented by the SH-SY5Y and SK-N-SH cell line. Statistics: one sample Student’s t-test, BH-FDR < 0.05, Log2 FC > 0/ < 0, N = 22. Refers to Table S2. (C) Venn diagrams comparing the differential abundance of kinases and kinase interaction partners between the two NMT states as determined by either 1^st gen or 2^nd gen kiCCA. Statistics: see (B). Refers to Table S2. (D) Results of a GSEA with GOBP terms applied to the results of differential expression analysis of kinase and kinase interaction partners between the two NB cell line SH-SY5Y and SK-N-SH; only pathways that achieved an FDR < 0.05 are shown. Statistics: see (B). Refers to Table S2. (E) Association of kinases with pathways through the sum of %TDRs of their interaction partners that have been associated with specific GOBP pathways terms shown in panel (D). NES is the normalized enrichment score. Refers to Fig. S4 and Table S2.

Figure 4.

Utilizing 2^nd gen kiCCA to prioritize kinase targets for pharmacologically manipulating NB cell NMT states. (A) Physical STRING interaction networks (v12.0) of CK2 PPIs predicted by 2^nd gen kiCCA (%TDR > 10) in the isogenic NB cell lines SK-N-SH (mesenchymal-like) and SH-SY5Y (noradrenergic neuronal-like). Other kinases than CK2 that may be part of the network were omitted. Relates to Fig. S3, Table S1 and S2. (B) Physical STRING interaction networks (v12.0) of proteins that specifically co-precipitated with a AUTS2 antibody compared to an isotype control IgG, and that were part of the 2^nd gen kiCCA network for CK2. Statistics: two-sample Student’s T-test, p < 0.05. Refers to Table S2. (C) DEA of proteins that specifically co-precipitated with a AUTS2 antibody compared to an isotype control IgG, and that were part of the 2^nd gen kiCCA network for CK2, comparing abundance between SH-SY5Y and SK-N-SH cells. Statistics: two-sample Student’s T-test, p > 0.05. Refers to Fig. 4B and Table S2. (D) Results from GSEA of global protein expression data, comparing SH-SY5Y cells treated with either the CK2 inhibitor SGC-CK2–1 or DMSO (vehicle) for 4 days. Only pathway terms that achieved an FDR < 0.05 are shown. Relates to Fig. S3D and Table S3. (E) Results from GSEA of global protein expression data, comparing SK-N-SH cells treated with either the TBK1 inhibitor GSK8612 or DMSO (vehicle) for 4 days. Only pathway terms that showed an FDR < 0.05 are shown. Relates to Table S3. (F) Differences in the expression of proteins that are the transcriptional targets of TGFβ-SMAD3, JNK-AP1, and NF-κB signaling between SK-N-SH cells treated with either the TBK1 inhibitor GSK8612 or DMSO (vehicle) for 4 days. Result of global proteome profiling. All proteins shown significantly differed in expression. Statistics: two sample Student’s t-test, BH-FDR < 0.05, N = 4. Refers to Table S2. (G) Trans well migration assay showing that CK2 inhibition with SGC-CK2–1 promoted migration in both SH-SY5Y cells and SK-N-SH cells, and that TBK1 and tyrosine kinase inhibition using GSK8612 and dasatinib, respectively, significantly inhibited cell migration only in the SK-N-SH cell line. Statistics: two sample Student’s T-test p < 0.05, N = 4; error bars are the S.D.