Pharmaco‐phosphoproteomic analysis of cancer‐associated KIT mutations D816V and V560G

The CD117 mast/stem cell growth factor receptor tyrosine kinase (KIT) is critical for haematopoiesis, melanogenesis and stem cell maintenance. KIT is commonly activated by mutation in cancers including acute myeloid leukaemia, melanoma and gastrointestinal stromal tumours (GISTs). The kinase and the juxtamembrane domains of KIT are mutation hotspots; with the kinase domain mutation D816V common in leukaemia and the juxtamembrane domain mutation V560G common in GISTs. Given the importance of mutant KIT signalling in cancer, we have conducted a proteomic and phosphoproteomic analysis of myeloid progenitor cells expressing D816V‐ and V560G‐KIT mutants, using an FDCP1 isogenic cell line model. Proteomic analysis revealed increased abundance of proteases and growth signalling proteins in KIT‐mutant cells compared to empty vector (EV) controls. Pathway analysis identified increased oxidative phosphorylation in D816V‐ and V560G‐mutant KIT cells, which was targetable using the inhibitor IACS010759. Dysregulation of RNA metabolism and cytoskeleton/adhesion pathways was identified in both the proteome and phosphoproteome of KIT‐mutant cells. Phosphoproteome analysis further revealed active kinases such as EGFR, ERK and PKC, which were targetable using pharmacological inhibitors. This study provides a pharmaco‐phosphoproteomic profile of D816V‐ and V560G‐mutant KIT cells, which reveals novel therapeutic strategies that may be applicable to a range of cancers.

GISTs most commonly occur in the juxtamembrane domain of the receptor (e.g., V560G mutations) [3,4], while KIT mutations in testicular seminomas, adult mastocytosis and AML are most frequent in the second tyrosine kinase domain (e.g., D816V mutations) [5,8,9].Inhibition of KIT is an attractive approach for treatment of KITmutant disease, however each class of KIT mutations confers differential sensitivity to KIT inhibitors.V560G-mutant KIT cells are sensitive to KIT inhibitors such as imatinib and dasatinib.Imatinib is a standard therapy for GISTs [10] and imatinib has shown benefit for KIT-mutation positive melanoma [11].However, development of imatinib-resistant disease occurs in approximately ∼50% of GIST patients [12].
D816V-mutant KIT cells display increased resistance to imatinib and dasatinib likely due to steric hindrance caused by the mutation [13,14].Multikinase KIT inhibitors midostaurin and nintedanib have shown in vitro efficacy against D816V-mutant KIT cells [15] and midostaurin is approved for the treatment of systemic mastocytosis [16].Kinase inhibitors have not proven successful as single agent therapy in AML [17], however midostaurin in combination with chemotherapy treatment is currently being assessed for KIT-mutation positive AML (NCT01830361, NCT03686345).
Each class of KIT mutations also confers differential activation of downstream signalling pathways, which may affect drug sensitivity.
The D816V kinase domain mutation has been associated with activation of JAK/STAT, ERK and MTOR signalling [13,18].The V560G juxtamembrane mutation has also been associated with activation of JAK/STAT and MTOR signalling [13,18].However, with the rapid improvement in the sensitivity and accuracy of mass spectrometrybased proteomics methods in the last decade [19], a comprehensive unbiased analysis of the signalling pathways activated by different cancer-associated KIT mutations is warranted.This may uncover novel therapeutic approaches for KIT-mutant cancer.
To explore KIT-mutant signalling pathways and identify potential therapeutic targets, our laboratory previously created isogenic cell line models of mutant KIT [13].FDCP1 mouse haematopoietic progenitor cells were transduced with V560G-or D816V-mutant KIT, or an empty vector control (EV).Cells were cultured in DMEM with 10% FBS, as previously described [20].EV cells were supplemented with 0.5ng/ml mouse granulocyte macrophage-colony stimulating factor (GM-CSF, BioLegend), while V560G-and D816V-KIT expressing lines were factor independent.All lines were confirmed to be free of mycoplasma contamination using the MycoAlert mycoplasma detection kit (Lonza), as per manufacturer's instructions.
Cells pellets (1 × 10 7 cells) were collected in biological triplicate, from cultures in log phase of growth.This yielded >600 µg protein per sample, of which 200µg were digested for each sample.Samples were digested and phospho-enriched using the EasyPhos method [21], as previously described [20].Label-free proteomes and phosphoproteomes were analysed using an Exploris 480 mass spectrometer.Peptides were desalted using an Acclaim PepMap 100 C18 75 µM × 20 mm trap column (Thermo Fisher) prior to separation on a 75 µM × 25 cm EASY-Spray PepMap C18 column (Thermo Fisher) over 90 min using a 5-35% acetonitrile gradient.Full MS scans of 350 to 1200 m/z were acquired at a resolution of 120,000, incorporating a normalised automatic gain control of 300% and maximum fill time of 50 ms.FAIMSpro compensation voltage was set to −60.Dynamic exclusion was set to 20s, mass tolerance low/high 10ppm.MS/MS scans were acquired using a resolution of 45,000, automatic gain control of 250%, a normalized collision energy of 36 and maximum fill time of 100 ms.
Raw files were analysed using Proteome Discoverer 2.5, as previously described [20].Files were searched against the UniProt Mus Musculus proteome (25,285 sequences, downloaded July 30, 2020), using SEQUEST HT.Search parameters included trypsin digestion, and allowed up to two missed cleavages.Precursor mass tolerance was 10 ppm, and fragment mass tolerance 0.02 Da.Cysteine carbamidomethylation was set as a fixed modification, phosphorylation (S/T/Y), acetylation (N terminus, K), oxidation (M), deamidation (N/Q) and N-terminal methionine loss were allowed as dynamic modifications.For the proteome, methylation (R/K) was also an allowed dynamic modification.Site probability threshold was set to 75/100.The results were filtered to a 1% false discovery rate at the peptide level, using the target-decoy strategy within Percolator [22].Label-free quantification was performed using the nodes "Minora Feature Detector," "Feature Mapper" and "Precursor Ions Quantifier" [23].
Master proteins were filtered for at least 1 unique peptide and no missing values in at least one group (EV, D816V, V560G), identifying 3719 proteins (Table S1).In the phosphoproteome, 5103 phosphopeptides were identified with no missing values in at least one group, with expected ratios of serine, threonine and tyrosine sites (pS:pT:pY 90.5%:9.2%:0.4%;Table S2).To obtain an overview of the differences between cell lines, heatmaps of proteins (proteome) or phosphopeptides (phosphoproteome) with significantly different abundance between groups (ANOVA, p < 0.05) were constructed using Perseus v2.0.7.0 [24] (Figure 1A,B).Row clusters were analysed for pathway enrichment using STRING (Reactome pathways, FDR-adjusted p-value <0.05, background = the full STRING network.Protein accessions were used for phosphoproteome data) [25,26].Clusters with high expression in EV control cells were enriched for Metabolism of RNA, Cell cycle, Splicing, Transcription and Innate immune system pathways (Figure 1A).Clusters with high expression in D816V-mutant KIT cells were enriched for Metabolism, TCA cycle and Metabolism of RNA (Figure 1A).Metabolism, Immune system and Cellular responses to stress were enriched in clusters with high expression in V560G-mutant KIT cells (Figure 1A.The full list of significantly enriched pathways is given in Table S3).
Differentially abundant phosphopeptides mapped to proteins involved in RNA metabolism and processing, cell cycle, DNA repair and transcription pathways (Figure 1B.Full list of significantly enriched pathways is given in Table S4).Clusters with high expression in EV controls were enriched for Metabolism of RNA, Splicing, Chromatin cells (Figure 1B).We have previously identified altered activation of DNA double strand break repair pathways in KIT-mutant AML [20].
We next analysed the proteins and phosphopeptides altered in D816V-and V560G-mutant KIT cells compared to EV controls (Fold change ≥1.5 or ≤0.67 and p-value ≤0.05 (2 tailed t-test) [27]) (Figure S1, Table S2).This revealed a high degree of overlap between D816Vand V560G-mutant cells.One hundred and thirty proteins were increased in D816V cells compared to EV cells, and 125 in V560G cells compared to EV, with 66 of these overlapping (Figure S1A).One hundred and forty-eight proteins were decreased in D816V cells and 113 were decreased in V560G cells compared to EV controls, including 75 common proteins (Figure S1A).
Four hundred and eleven phosphopeptides were increased in D816V cells compared to EV cells, and 248 in V560G cells compared to EV, with 145 of these overlapping (Figure S1B).Three hundred and sixty-five phosphopeptides were decreased in D816V cells and 157 were decreased in V560G cells compared to EV controls, including 86 common phosphopeptides (Figure S1B).
Notably, this includes further peptidases/proteases (MCPT1, TPSB2, CMA1).Tryptase (e.g., TPSB2) levels in serum have been correlated with D816V-KIT allele fractions in systemic mastocytosis patients [28], however to our knowledge, the increase in so many proteases in KIT mutant cells has not previously been reported.
The top proteins that reached significance in D816V-mutant KIT cells but did not reach significance in V560G-mutant KIT cells were mostly regulated in the same direction (Figure S2A,B).Top increased proteins in D816V-mutant KIT cells that did not reach significance in V560G-KIT cells included fatty acid metabolism protein CYP4A12B, gamma tubulin protein TUBGCP3 and albumin (Figure S2A).Top proteins decreased in D816V-KIT cells that did not reach significance in V560G-KIT cells included cell adhesion protein ADGRE1, leucine rich repeat containing LRRC58, and neurogranin (Figure S2A).Top increased proteins in V560G-mutant KIT cells that did not reach sig-nificance in D816V-KIT cells included actin filament-severing protein scinderin, cell adhesion regulator leupaxin, dCTP pyrophosphatase 1, and integrin ITGB7 (Figure S2B).The top proteins decreased in V560G-KIT cells that did not reach significance in D816V-KIT cells included syntaxin binding protein 6, acid sphingomyelinase-like phosphodiesterase 3b and quinone oxidoreductase CRYZ (Figure S2B).There were only 5 proteins that were significantly increased in one mutant KIT line with concurrent significant downregulation in the other mutant KIT line.These were the fatty acyl-CoA reductase FAR2, which was increased in V560G-KIT cells and decreased in D816V-KIT cells, as was actin-binding protein TMSB4X.Proteins increased in D816V-KIT cells and decreased in V560G-KIT cells were transcription factor USF1, annexin ANXA3 and glycogenin GYG1 (Figure S2A,B).
Phosphopeptides increased in D816V-and V560G-mutant KIT cells compared to EV included FYB1 phosphosites (Figure 1D).FYB1 is an adaptor protein involved in cytoskeleton remodelling [29].In line with this, phosphosites on other cytoskeleton and adhesion related proteins RCDS1, CTSG and ITGB7 were also increased (Figure 1D).
While not in the top 20 largest changes, protein expression of ITGA4 was increased and ITGA6 was decreased in both mutants (Table S1), indicating dysregulation of cell adhesion processes.
Top phosphopeptides increased in D816V-mutant KIT cells that did not reach significance in V560G-KIT cells included phosphosites on KIT, transcription factor subunit GTF2F1, and LDL receptor related protein LRP12 (Figure S2C).Top phosphopeptides decreased in D816V-KIT cells that did not reach significance in V560G-KIT cells included phosphosites on splicing factor SCAF1, FAM102B, ribosomal subunit biogenesis protein NOP56, and mRNA splicing cofactor SON (Figure S2C).Top phosphopeptides increased in V560G-KIT cells that did not reach significance in D816V-KIT cells included phosphosites on immunoglobulin epsilon receptor subunit FCER1G, molecular chaperone and chromatin regulator DNAJC2, ubiquitin binding adaptor TAX1BP1 and cyclin-Y (Figure S2D).Top phosphopeptides decreased in V560G-KIT cells that did not reach significance in D816V-KIT cells included phosphosites on proline-rich and coiled-coil-containing protein 2C, ran-specific GTPase-activating protein RANBP1, epigenetic regulator TASOR (Figure S2D).To identify significantly altered pathways in D816V-and V560Gmutant KIT cells compared to EV, the proteomes were analysed using the 2D enrichment package in Perseus [31].KEGG and GO Biological Process annotations were analysed for enrichment using the default parameters (Benjamini-Hochberg FDR <0.02; Table S5) [32,33].For each KEGG and GO annotation term it is tested whether the distribution of the mapped values differs from the global distribution of input values.This identified a high degree of overlap betweenthe pathways increased and decreased in both KIT mutants compared to EV controls, with the only dissimilarity being an increase in ncRNA processing associated pathways in D816V cells but not V560G cells (Figure 2A).Oxidative phosphorylation was significantly increased in both KIT-mutant lines compared to control, with a high degree of overlap with proteins involved in Alzheimer's disease and Parkinson's disease signalling (Figure 2A).Also increased were lipid biosynthesis and protein processing in the endoplasmic reticulum (Figure 2A).Pathways significantly decreased in both KIT-mutant lines were largely associated with RNA metabolism, including transcription, RNA splicing TA B L E 1 Functional signature analysis of phosphoproteome alterations in D816V and V560G-mutant KIT cells compared to EV controls.Kinase and perturbation signatures were analysed using PTM-SEA (Krug et al. [34]) and results were grouped by functional category.A positive Z score indicates upregulation of the signature, a negative Z score indicates downregulation of the respective signature.Note: Kinase and perturbation signatures were analysed using PTM-SEA (Krug et al. [34]) and results were grouped by functional category.A positive Z score indicates upregulation of the signature, a negative Z score indicates downregulation of the respective signature.
To identify kinases and signalling processes altered by either KIT mutation compared to EV controls, the phosphoproteomes were analysed using PTM-SEA [34].Currently, annotations for mus musculus are not as comprehensive as for homo sapiens, therefore phosphosite sequence tags were searched against both mus musculus and Table 1).P38 MAPKs were also significantly activated in V560Gmutant cells, with a trend towards activation in D816V-mutant cells.
Consistent with this, signatures of ERK/MAPK inhibitors were concurrently downregulated (Table 1).Together, these results indicate activation of ERK/MAPK signalling in both D816V-and V560G-mutant cells, suggesting these cells may be sensitive to ERK/MAPK inhibitors.
In accordance with the PTM-SEA results, we have previously shown that both KIT mutant cells are also more sensitive to ERK/MAPK inhibitors, selumetinib and ASTX029 [20].
EGFR signalling was also activated in both D816V-and V560Gmutant KIT cells, with activation of EGF treatment and EGFR1 pathway signatures (Table 1).EGFR pathway signalling kinase Akt1 also displayed activation (Table 1).The signature for HER2/EGFR inhibitor lapatinib was significantly downregulated in V560G-mutant cells, consistent with activation of EGFR signalling (Table 1).Conversely, the signature for EGFR inhibitor erlotinib was significantly activated in V560G-mutant cells (Table 1).Possible reasons for this disparity could include the limitations of searching mouse data against a homo sapiens database, or perhaps detection of negative-feedback loops within the erlotinib signature.Taken together, the consensus results suggest activation of EGFR signalling in KIT-mutant cells (Table 1).In agreement with this, D816V-and V560G-mutant KIT cells were more sensitive to the EGFR inhibitors erlotinib and lapatinib, and the AKT inhibitor MK2206, in comparison to EV controls (Figure 2B,C).
Phorbol ester signalling was significantly increased in V560Gmutant cells (Table 1), suggestive of activated PKC signalling.PKA activity was also significantly increased in V560G-mutant cells, with a trend toward activation in D816V-mutant cells (Table 1).The signature for staurosporine, a multikinase PKC and PKA inhibitor, was significantly downregulated in D816V-mutant cells (Table 1).Together, this suggests activation of PKA and PKC activity in KIT-mutant cells.
To test the functional relevance of this, sensitivity to midostaurin (PKC and KIT multikinase inhibitor) and enzastaurin (PKC-beta inhibitor) was tested (Figure 2B).Both D816V-and V560G-mutant KIT lines were more sensitive to midostaurin compared to EV controls (Figure 2B), however this is likely due to the KIT inhibitory activity of midostaurin.However, both D816V-and V560G-mutant KIT lines were also more sensitive than EV control cells to the specific PKC inhibitor enzastaurin, with V560G cells displaying the lowest IC50 (Figure 2B,C).
Kinase enrichment analysis identified increased CDK1 activity in both KIT mutant lines (Table 1), therefore we tested sensitivity to R0-3306 (CDK1 inhibitor), riviciclib (CDK1/4/9 inhibitor) and CDKI73 (CDK1/2/4/9 inhibitor).For each CDK inhibitor, IC50s were very similar across the 3 cell lines (Figure 2C), with a small increase in sensitivity of D816V cells to CDKI73 (Figure 2B,C).These results may be confounded by a lack of target specificity of these inhibitors, as has been recently reported [35].Other possible explanations include data incompleteness and the presence of negative feedback loops.
Mutant KIT has been previously associated with activation of JAK/STAT signalling [18].Paradoxically, signatures for JAK inhibitors tofacitinib and TG101348 displayed increased activity in V560Gmutant KIT cells (Table 1).In support of this, V560G-mutant cells were significantly more resistant to the JAK inhibitor ruxolitinib, with the EV control cells displaying the highest sensitivity of the 3 lines (Figure 2B,C).It should be noted however that the FDCP1-EV cells are reliant on GM-CSF, which activates JAK signalling [36].
Signatures for several epigenetic-targeted therapies were significantly activated in both KIT-mutant lines (Table 1).While epigenetic modifiers azactidine and decitabine have been shown to be growth inhibitory in KIT-mutant mast cell leukaemia cell lines [37], D816Vmutant KIT has been associated with progression on midostaurin and azacitidine combination treatment in a patient with systemic mastocytosis [38].Disease progression on venetoclax and azacitidine therapy was also reported for two AML patients with tyrosine kinase domain KIT mutations [39].Thus, mutant KIT may be associated with poor response to these hypomethylating therapies.
Based on the functional enrichment results obtained from the proteome identifying upregulation of proteins involved in oxidative phosphorylation in KIT-mutant cells (Figure 2A), we lastly tested the OXPHOS inhibitor IACS-010759.Both D816V-and V560G-mutant KIT lines were more sensitive to IACS-010759 compared to EV controls (Figure 2B,C), with EV cells not reaching IC50 in the dose range tested (Figure 2C).
In summary, the work herein provides a comprehensive pharmacophosphoproteomic characterisation of common cancer-associated KIT mutations, D816V (tyrosine kinase domain) and V560G (juxtamembrane domain).Functional enrichment analyses using the proteome and phosphoproteome has identified activation of a number of pathways for which therapeutic inhibitors are available.In vitro drug screening provided proof-of-principle follow up demonstrating the power of proteomic and phosphoproteomic characterisation for identifying rational targets for therapy.
modifying enzymes and Transcription pathways (Figure 1B).Clusters with high expression in D816V-mutant KIT cells were enriched for Metabolism of RNA, Splicing, Transcription and Cell cycle pathways (Figure 1B).Cell cycle and DNA double strand break repair pathways were enriched in clusters with high expression in V560G-mutant KIT F I G U R E 1 Proteomic and phosphoproteomic analysis of FDCP1 cells transduced with D816V-or V560G-mutant KIT or EV control.(A) Heatmap of proteins and (B) phosphopeptides with significantly altered abundance between any two groups (EV, V560G or D816V.ANOVA, p ≤ 0.05).Grey shading denotes missing values.Row clusters were analysed for enrichment using STRING (Reactome database).No significant enrichment was detected for the dark blue and gold row clusters in B).Full lists of enriched pathways are provided in Tables S3 and S4.Heatmaps created using Perseus.(C) Bubble plot of the top 20 proteins increased and top 20 decreased in D816V-and V560G-mutant KIT cells compared to EV controls (p < 0.05).(D) Bubble plot of the top 20 phosphosites increased and top 20 decreased in D816V-and V560G-mutant KIT cells compared to EV controls (p < 0.05).
The top phosphopeptides that reached significance in one mutant KIT line but not the other were mostly regulated in the same direction (FigureS2C,D).Notable exceptions included transcription factor SOX4 (phosphorylation decreased in D816V-KIT cells), and cytoskeletal protein PALLD (phosphorylation decreased in V560G-KIT cells).There were no phosphopeptides that were significantly regulated with opposite directions in the two mutant KIT lines (p < 0.05, fold change > ±1.5).

F I G U R E 2
Functional analysis of proteome and phosphoproteome alterations in D816V-and V560G-mutant KIT cells compared to EV controls.(A) 2D enrichment analysis identifies pathways enriched in D816V-and V560G-mutant KIT cells compared to EV controls.A positive enrichment score indicates upregulation in KIT-mutant cells, whereas a negative score indicates downregulation.Pink shading: upregulated in both D816V and V560G cells.Blue shading: downregulated in both D816V and V560G cells.(B) Sensitivity to the indicated drugs was measured by resazurin assay, after 72 h of treatment.(C) IC50s generated from (B) using individual biological replicates and Fit Lowess/Spline in GraphPad Prism.All IC50s are in µM.T-test: significance compared to EV controls.
homo sapiens PTMSigDB libraries.Input parameters included "rank" normalisation, weight = 0.75, 1000 permutations, and a minimum signature overlap of 10.PTM-SEA tests whether the phosphosites mapping to each signature are systematically larger or smaller than the distribution of the full input list.Activation of ERK/MAPK signalling was identified in both D816V-and V560G-mutant KIT cells, with significant activation of ERK2/MAPK1 kinase activity (Z score >0 = activation, FDR <0.25;