Comprehensive Kinase Activity Profiling Revealed the Kinase Activity Patterns Associated with the Effects of EGFR Tyrosine Kinase Inhibitor Therapy in Advanced Non-Small-Cell Lung Cancer Patients with Sensitizing EGFR Mutations

EGFR mutations are strong predictive markers for EGFR tyrosine kinase inhibitor (EGFR-TKI) therapy in patients with non-small-cell lung cancer (NSCLC). Although NSCLC patients with sensitizing EGFR mutations have better prognoses, some patients exhibit worse prognoses. We hypothesized that various activities of kinases could be potential predictive biomarkers for EGFR-TKI treatment among NSCLC patients with sensitizing EGFR mutations. In 18 patients with stage IV NSCLC, EGFR mutations were detected and comprehensive kinase activity profiling was performed using the peptide array PamStation12 for 100 tyrosine kinases. Prognoses were observed prospectively after the administration of EGFR-TKIs. Finally, the kinase profiles were analyzed in combination with the prognoses of the patients. Comprehensive kinase activity analysis identified specific kinase features, consisting of 102 peptides and 35 kinases, in NSCLC patients with sensitizing EGFR mutations. Network analysis revealed seven highly phosphorylated kinases: CTNNB1, CRK, EGFR, ERBB2, PIK3R1, PLCG1, and PTPN11. Pathway analysis and Reactome analysis revealed that the PI3K-AKT and RAF/ MAPK pathways were significantly enriched in the poor prognosis group, being consistent with the outcome of the network analysis. Patients with poor prognoses exhibited high activation of EGFR, PIK3R1, and ERBB2. Comprehensive kinase activity profiles may provide predictive biomarker candidates for screening patients with advanced NSCLC harboring sensitizing EGFR mutations.


Introduction
Lung cancer is one of the most common cancers, causing death frequently [1]. Among lung cancers, non-small-cell lung cancer (NSCLC) amounts to approximately 80% of newly diagnosed lung cancers yearly [2]. Notably, most NSCLC cases are diagnosed as lung cancers with advanced stage [2]. For patients with advanced NSCLC, chemotherapy has been considered the front-line remedy. However, patients are provided with limited therapeutic effects and poor clinical outcomes due to its toxicity and adverse events: the outcomes are a median overall survival (OS) of only 8-10 months and a 5-year survival rate of less than 15% [3]. Epidermal growth factor receptor tyrosine kinase inhibitors (EGFR-TKIs) are used to treat patients with advanced or metastatic NSCLC. EGFR-TKIs have demonstrated a significant effect on NSCLC patients harboring EGFR mutations and have improved quality of life [4,5].
The overexpression of EGFR is commonly identified in NSCLC (32-81%) and is known as a reliable target and biomarker for NSCLC treatment [6,7]. EGFR-TKIs including gefitinib and erlotinib demonstrate surpassing clinical effects compared to standard chemotherapy or best supportive care [4,5]. Previous studies have reported that EGFR mutation is a practical predictive marker of increased sensitivity to EGFR-TKIs and concerned with the improvement of progression-free survival with TKIs [8]. Particularly, deletions of exon 19 and point mutations of exon 21 (21-L858R) are commonly observed in 85% of patients with NSCLC harboring EGFR mutations [9][10][11]. These mutations demonstrate a high response rate of 70% against EGFR-TKI [12]. Several studies have confirmed that patients with NSCLC harboring these two mutations show fewer side effects and improved quality of life [13]. However, the heterogeneity of responses to EGFR TKIs has been pointed out to have a wide range from a few months to several years in progression-free survival. Additionally, in around 10-20% of patients with sensitizing EGFR mutations, objective responses to EGFR TKIs are not exhibited. These phenomena have motivated the search for other predictive biomarkers surpassing EGFR mutations that can detect patients with sensitizing EGFR mutations who have a worse prognosis.
Protein phosphorylation is an essential apparatus regulating cellular functions such as apoptosis, cell proliferation and migration, cell cycle, and differentiation [14]. Approximately 500 different kinases orchestrating these pivotal functions are encoded by the human genome [15], and 90% of all proteins are subjected to phosphorylation [16]. Aberrated kinase activity is caused by genetic mutations such as amplification, point mutation, chromosomal translocation, and epigenetic regulation in carcinogenesis and cancer progression. Furthermore, the dysregulation of self-phosphorylation and the kinase-to-kinase regulatory relationship also result in the aberrant activity of kinases. The corrupted kinase activity finally causes various effects including the interruption of important cell functions, the transformation of normal cells into tumor cells, and the determination of malignant features such as invasion, metastasis, and resistance to chemotherapy [17]. Therefore, aberrantly regulated kinases and their substrate proteins are considered biomarkers to affect the process of cancer treatments [18,19]. In addition, protein kinases are crucial therapeutic targets in oncology [20], and a number of approved kinase inhibitors have been used for cancer therapeutics [21]. Therefore, the research of protein phosphorylation will deepen the understanding of fundamental biology and provide novel insights for clinical applications of NSCLC harboring sensitizing mutations.
Although most NSCLC patients with sensitizing EGFR mutations exhibit a better prognosis, some exhibit a worse prognosis. For the identification of predictive biomarkers that are useful to detect these patients with worse prognoses, we performed a comprehensive kinase activity analysis using PamStation 12 (PamGene International, BJ's-Hertogenbosch, The Netherlands) in NSCLC patients at advanced stages with sensitizing EGFR mutations and followed their prognosis prospectively. Finally, kinomic profiles may provide prognostic biomarkers for patients with advanced NSCLC harboring sensitizing EGFR mutations.

Patients
Nineteen patients with advanced NSCLC who were treated in the enrolled hospitals between June 2018 and October 2020 were included in the study. All patients had available tumor tissues for biomarker analysis. Tumor tissue from the primary sites was obtained by surgery or biopsy. The samples were rinsed with saline after being taken. The samples were simply put into a −80 • C freezer. The time between biopsy and freezing was conducted within a few minutes in all samples. Serial sections were used for mutational and tyrosine Proteomes 2023, 11, 6 3 of 17 kinase activity analyses. After the diagnosis of NSCLC, EGFR-TKIs were administered as monotherapy (composed of the first therapy in gefitinib, afatinib, or osimertinib) and complete clinicopathological findings were examined in all patients. Treatment with gefitinib (250 mg), afatinib (20-40 mg), or osimertinib (80 mg) alone was maintained unless disease progression, adverse events, or patient refusal occurred. Written informed consent was obtained from all patients. The study protocol was approved by the Institutional Ethics Committee of the National Cancer Center (2018-208) and the Kyoto Prefectural University of Medicine (ERB-C-1106).

Study Design
This prospective observational study was designed to evaluate tyrosine kinase activity for predicting the clinical response to EGFR-TKI treatment in patients with advanced NSCLC harboring activating EGFR mutations. Tumor samples were obtained at the initial diagnosis. Until all clinical data were evaluated, the clinical data were enclosed during laboratory analysis Recorded clinical data consisted of age, sex, smoking history, pathology, stage at diagnosis, treatments, EGFR mutation, and adverse events. After the administration of EGFR-TKIs, the prognosis was followed. Efficacy evaluations included the best response, disease control rate (DCR), objective response rate (ORR), progression-free survival (PFS), and overall survival (OS). The outline of this study is presented in Figure 1. Workflow of this study. A number of NSCLC patients with sensitizing EGFR mutations were enrolled in this study. The prognosis was followed.

Assessments
Tumors were assessed at diagnosis and every 8-12 weeks until the investigators reported disease progression or unacceptable adverse events. According to the Response Evaluation Criteria in Solid Tumors (RECIST), clinical responses to TKIs consisting of complete response (CR), partial response (PR), stable disease (SD), and disease progression (PD) were examined [22]. The definition of PFS was the time from the initiation of TKI treatment to PD or death. The definition of OS was the time from TKI initiation to death. All films were assessed by an independent radiologist who was blinded to the Workflow of this study. A number of NSCLC patients with sensitizing EGFR mutations were enrolled in this study. The prognosis was followed.

Assessments
Tumors were assessed at diagnosis and every 8-12 weeks until the investigators reported disease progression or unacceptable adverse events. According to the Response Evaluation Criteria in Solid Tumors (RECIST), clinical responses to TKIs consisting of Proteomes 2023, 11, 6 4 of 17 complete response (CR), partial response (PR), stable disease (SD), and disease progression (PD) were examined [22]. The definition of PFS was the time from the initiation of TKI treatment to PD or death. The definition of OS was the time from TKI initiation to death. All films were assessed by an independent radiologist who was blinded to the EGFR biomarker status.

EGFR-mutation Analysis
EGFR mutations in exons 18-21 were examined using the polymerase chain reaction method for tumor and plasma samples. Sequencing was performed at commercial clinical laboratories (SRL, Inc., Tokyo, Japan).

Comprehensive Tyrosine Kinase Activity Assay
Frozen biopsy tissues were processed at 4 • C for kinomic profiling after grinding. The tissues were lysed in M-PER Mammalian Extraction Buffer (Pierce, Rockford, IL, USA) with Halt protease and phosphatase inhibitors (Pierce cat. 78420, 78415) [23]. Protein quantification was performed using a Bradford reaction assay. The extracted protein (5 mg) was mixed with kinase buffer, ATP, and fluorescently labeled anti-PY20 antibodies. Then, the mixed protein was loaded into the tyrosine (PTK) PamChips. The samples were subjected to the PamStation 12 kinomics workstation (PamGene International, BJ's-Hertogenbosch, The Netherlands) using the PTK PamChip protocol in Evolve12 Software (v. 1.5) (PamGene International), as previously reported [23]. When lysates were pumped through the array, images were captured, analyzed, and quantified using BioNavigator v. 5.1 (PamGene International). The study was performed in duplicate and subjected to activity analysis.

Identification of Upstream Kinases
Comprehensive kinase activity was analyzed using BioNavigator software v. 6.3.67.0 (PamGene International). The analysis of kinase activity was conducted corresponding to the phosphorylated peptide-specific reaction per each "spot" on the PamChip. The intensities of the raw signals of the 144 spots were measured over multiple 50 ms exposures sequentially as lysates were pumped through the array. The measurement was also performed over multiple exposure times (10, 20, 50, 100, and 200 ms) after the lysates were rinsed off. These values were converted to slopes of intensity by exposure time. The slopes were multiplied by 100 and log2-transformed. The unsupervised hierarchical clustering of the kinomic activity data was performed using Euclidean distance. Active kinases were predicted using PhosphoSitePlus (https://www.phosphosite.org, accessed on 25 January 2023) [24] and the UniProt database (https://www.uniprot.org/, accessed on 25 January 2023) [25]. These kinases, which were found in all the databases, were considered positively identified.

STRING Analysis
The Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (https://string-db.org/, accessed on 25 January 2023) was used to examine the proteinprotein interaction (PPI) network. The STRING app in Cytoscape (https://cytoscape.org/ index.html, accessed on 1 February 2023)was used to examine the potential correlation between these signal intensities and each peptide phosphorylation [26]. To perform the PPI network analysis, significant peptides were selected (p-value using a t-test between cluster 1 and cluster 2). These peptides (n = 2) were translated into their corresponding UniProt ID using UniProt (https://www.uniprot.org/, accessed on 25 January 2023). After the removal of duplicated UniProt IDs, UniProt IDs and fold changes were entered into the STRING database. In the setting for organism, "Homo sapiens" was selected.

Pathway Analysis, Network Analysis, and Reactome Analysis
STRING analysis was also used for pathway analysis. The dataset after conversion to UniProt ID was entered as an input into STRING. UniProt IDs were mapped onto pathways based on curated data from the literature. The top 20 pathways were grouped according to the processes. The identified peptides were used for the network analysis. The Uniprot IDs corresponding to the peptides were used in Cytoscape software version 3.9.3 [27] and the STRING database [28]. The network analysis was conducted using a network analyzer. Hub bottlenecks were determined following the degree values and betweenness centrality. Common hub bottlenecks and top nodes were identified as the central nodes according to closeness centrality and stress. Action maps consisting of activation, inhibition, and expression were demonstrated for the central nodes using CluePedia [29]. The Kapa score was considered to be the default value in CluePedia. Reactome analysis was conducted with the ReactomePA package with the p-value cutoff set as 0.05 [30]. Only the top entries with a minimum adjusted p-value were included in the dotplot.

Statistical Analysis
Progression-free survival (PFS) was calculated from the date of EGFR-TKI administration to the date of disease progression or death from any cause, and overall survival (OS) was calculated from the date of EGFR-TKI administration to the date of death from any cause. Fisher's exact test or χ2 test for categorical variables and the t-test for continuous variables were used to analyze the clinicopathological features of the two groups divided by the cut-off value. Survival curves were plotted using the Kaplan-Meier method and compared using the log-rank test. All statistical analyses were performed using the GraphPad Prism software (v.9.0; GraphPad Software, San Diego, CA, USA).

Patient Characteristics
Patient characteristics are summarized in Table 1. The median age was 73.0 years (range, 46-88 years). There were six men and thirteen women, of whom six were nonsmokers and thirteen were smokers. All patients had stage IV adenocarcinoma. A total of 11 cases with exon 19-del mutation and 8 cases with exon 21 L858R mutation were identified. As EGFR-TKI therapy, two patients received first-line therapy as gefitinib, three patients as afatinib, and fourteen patients as osimertinib. One case was not evaluated using a comprehensive kinase activity assay because the amount of extracted protein did not meet the required amount for the experiment. Thus, 18 cases were analyzed. Based on RECIST 1.1, fifteen patients were categorized as PR, one patient was categorized as SD and PD, and two patients were not evaluated as NE. Based on the results of the RECIST, the objective response rate (ORR) was 88.24% and the disease control rate (DCR) was 94.12%.

Comprehensive Kinase Activity Analysis in NSCLC Patients with Sensitizing EGFR Mutations
The tyrosine kinase activity profiles of advanced NSCLC patients with sensitizing EGFR mutations were examined in nineteen NSCLC patients with pathologically confirmed NSCLC who underwent panel sequencing. In one patient, the amount of protein was too low to measure the protein kinase activity. Therefore, this sample was excluded from the analysis. Except for this case, all NSCLC tissue samples demonstrated protein kinase activity profiles. As a quality control, peptides that showed no increase in signal intensity over time were eliminated. After the quality control, 102 of 144 PTK peptides were evaluated. The values of mean signal intensity per peptide were calculated and log2-transformed (Supplementary Table S1). In all samples, the kinase activity assay was performed in duplicate.

Identification of Peptides Showing Significant Activation of Phosphorylation in Advanced NSCLC with Sensitizing EGFR Mutations
The results were imaged as a heat map ( Figure 2a); the rows of the heatmap represent each peptide, and the columns of the heatmap represent each sample. The peptides were grouped according to the signal intensity of NSCLC patients with sensitizing EGFR muta-  Table S2). The samples were classified into two groups: cluster 1, the lower phosphorylated group, and cluster 2, the higher phosphorylated group (Figure 2a). Cluster 1 included 19 samples and cluster 2 comprised 17 samples.

Putative Upstream Kinases
Of these 102 peptides, 29 peptides were identified to be able to phosphorylate the tyrosines (upstream kinases). The putative tyrosine kinases for each peptide are listed in Table 2, Supplementary Table S3, and Figure 2b. Six receptor tyrosine kinases (RTK), ERRB2, FGFR2, MET, PDGFRB, RET, and VEGFR2, were identified, as well as members of the four non-receptor tyrosine kinases (NRTK), including Abl1, Lck, SRC, and Syk in cluster A, the intermediately phosphorylated group. In the highly phosphorylated group, cluster B, the two RTKs, PDGFRB and VEGFR2, the three NRTKs, FER, FES, and Syk, and the serine/threonine kinase, WEE1, were identified. The six RTKs, including EGFR, EPHA4, ERBB4, FGFR1, INSR, and VEGFR2, the two NRTKs, including ABL1 and SRC, and the five serine/threonine kinases (STKs), including MAP2K1, MAP2K2, MAP2K3, MAP2K4, and MAP2K6, were poorly phosphorylated in cluster C (Figure 2b). Based on the Venn diagram, one kinase, VEGFR2, was phosphorylated in all groups. Except for VEGFR2, two kinases were commonly identified in the two clusters: PDGFRB and SYK were detected both in clusters A and B, and SRC and ABL1 were detected both in clusters A and C.

Putative Upstream Kinases
Of these 102 peptides, 29 peptides were identified to be able to phosphorylate the tyrosines (upstream kinases). The putative tyrosine kinases for each peptide are listed in Table 2, Supplementary Table S3, and Figure 2b. Six receptor tyrosine kinases (RTK), ERRB2, FGFR2, MET, PDGFRB, RET, and VEGFR2, were identified, as well as members of the four non-receptor tyrosine kinases (NRTK), including Abl1, Lck, SRC, and Syk in cluster A, the intermediately phosphorylated group. In the highly phosphorylated group, cluster B, the two RTKs, PDGFRB and VEGFR2, the three NRTKs, FER, FES, and Syk, and the serine/threonine kinase, WEE1, were identified. The six RTKs, including EGFR, EPHA4, ERBB4, FGFR1, INSR, and VEGFR2, the two NRTKs, including ABL1 and SRC, and the five serine/threonine kinases (STKs), including MAP2K1, MAP2K2, MAP2K3, MAP2K4, and MAP2K6, were poorly phosphorylated in cluster C (Figure 2b). Based on the Venn diagram, one kinase, VEGFR2, was phosphorylated in all groups. Except for VEGFR2, two kinases were commonly identified in the two clusters: PDGFRB and SYK were detected both in clusters A and B, and SRC and ABL1 were detected both in clusters A and C.   Bold characters indicate predicted kinases using two database analysis including UniProt and Phosphosite Plus.

Pathway Analysis and Network Analysis
Pathway analysis (STRING) using the UniProt IDs of the 102 peptides and the signal intensity values for each peptide yielded many pathways ( Table 3). The identified pathways are highly significant with the p-values ranging from 4.27 × 10-23 to 2.21 × 10-12. The significantly enriched pathways were the PI3K-Akt, Ras, Rap1, and MAPK signaling pathways. Network analysis revealed that a network including 102 UniProt IDs was constructed (Figure 3a). The network contained seven isolated nodes and a main connected component (78 nodes and 619 edges). The top 10% of nodes based on degree values, including CTNNB1, EGFR, PIK3R1, ERBB2, PTPN11, PLCG1, CRK, and CBL, were selected as hubs. The top 10 nodes regarding betweenness centrality, CTNNB1, EGFR, PIK3R1, PLCG1, ERBB2, MAPK1, PTPN11, and CRK, were determined as bottlenecks. Common hubs and bottlenecks, including CTNNB1, EGFR, PIK3R1, ERBB2, PTPN11, PLCG1, and CRK, were identified (Table 4 and Supplementary Table S3). The action map of the identified seven kinases is shown in Figure 3b.

Kinase Profile Different between Highly Phosphorylated and Lower Phosphorylated Group
In the previous heat map, cases were grouped according to the signal intensity of NSCLC patients with sensitizing EGFR mutations (Figure 2a). The cases were classified into two groups: cluster 1, the lower phosphorylated group, and cluster 2, the highly phosphorylated group. The 19 samples from the 10 cases were classified into cluster 1, and the other 19 samples from the other 10 cases were classified into cluster 2 (Supplementary Table S3).
Peptides showing significant differences in phosphorylation between clusters 1 and 2 were identified. Thirty-five peptides were identified as differentially phosphorylated (p < 0.05, FDR = 0.090) (Supplementary Table S4). For these 35 peptides, kinases reported to be able to phosphorylate tyrosine as upstream kinases were identified. The putative tyrosine kinases upstream of each peptide are listed in Table 5. Several receptor tyrosine kinases such as EGFR and JAK2 have been identified.

Pathway Analysis, Reactome Analysis, and Network Analysis between Highly Phosphorylated Group and Low Phosphorylated Group
Pathway analysis using the UniProt IDs of the 35 peptides and the value of signal intensities for each peptide yielded many pathways (Table 6). These pathways were highly significant, as the p-values range from 2.74 × 10 -23 to 1.10 × 10 -12 . The 20 most significant pathways were enriched in the PI3K-Akt, Ras, Rap1, and MAPK signaling pathways, which were similar to those of NSCLC patients with sensitizing EGFR mutations.
Reactome analysis using the UniProt IDs of the 29 peptides with p-values less than 0.05 in the t-test yielded many pathways (Figure 4). The pathways, including the PI3K-AKT and RAF/MAP pathways, were involved in the highly phosphorylated group.
Network analysis revealed that a network consisting of 35 UniProt IDs was constructed ( Figure 5). The network contained three isolated nodes and a main connected component (26 nodes and 93 edges). The top 10% of the nodes based on degree values, including EGFR, PIK3R1, and ERBB2, were selected as hubs. The top ten nodes regarding betweenness centrality contained EGFR, PIK3R1, and ERBB2 as bottlenecks. Common hubs and bottlenecks, including EGFR, PIK3R1, and ERBB2, were identified. All hub bottlenecks were included in the top nodes based on closeness centrality and stress (Table 7  and Supplementary Table S5).

Figure 5.
Network analysis of kinase activity in poor prognosis. Altered kinases from each patient were mapped onto a network using the Cytoscape STRING app. Red nodes represent higher fold change in kinase activity, and white nodes represent lower fold change in kinase activity. The size of a node represents the number of connections between the nodes.

Discussion
The present study investigated comprehensive kinomic profiles to identify commonly upregulated kinases and develop prognostic markers for patients with NSCLC harboring sensitizing EGFR mutations. This is the first study to demonstrate the activity of multiple tyrosine kinases in NSCLC patients harboring sensitizing EGFR mutations. Sam-

Discussion
The present study investigated comprehensive kinomic profiles to identify commonly upregulated kinases and develop prognostic markers for patients with NSCLC harboring sensitizing EGFR mutations. This is the first study to demonstrate the activity of multiple tyrosine kinases in NSCLC patients harboring sensitizing EGFR mutations. Samples from 18 NSCLC patients with EGFR mutations were used to perform kinase activity analysis using PamStation12. This was a prospective study that integrated kinase information and clinical outcomes. Notably, prominent tyrosine kinases of the ErbB receptor family, EGFR and ERBB2, and kinases belonging to downstream signaling pathways, including CRK, CTNNB1, PIK3R1, PLCG1, and PTPN11, were highly activated in all patients. In particular, tyrosine receptor kinases, including EGFR, ERBB2, and PIK3R1, represent key components that affect patient prognosis. Pathway analysis and Reactome analysis revealed that PI3K-AKT and RAF/MAPK signaling pathways were enriched in NSCLC with sensitizing EGFR mutations.

Common Activated Kinases in NSCLC with Sensitizing EGFR Mutations
Our study revealed that seven kinases, CTNNB1, CRK, EGFR, ERBB2, PIK3R1, PLCG1, and PTPN11, were highly activated in all NSCLC patients harboring sensitizing EGFR mutations. Among these seven kinases, CTNNB1, CRK, PLCG1, and PTPN11 were newly identified as highly activated kinases in our study. B-catenin, encoded by the CTNNB1 gene, plays an important role in a signaling pathway affecting cell proliferation and differentiation [31]. Despite the low frequency of mutations in CTNNB1 in NSCLC, CTNNB1 contributes to the development of NSCLC through EGFR mutations in vitro and in vivo from Nakamura et al. [32]. The CT10 regulator of kinase (CRK) is a family of widely expressed adaptor proteins involved in signal transduction from various oncoproteins, including Bcr-Abl, EGFR, PDGF, and VEGFR [33,34] and plays essential roles in cytoskeletal changes, cell proliferation, adhesion, and migration [35]. The overexpression of CRK has been reported in NSCLC [36] and the high phosphorylation of CRK has been verified in NSCLC cell lines [37]. Phospholipase C gamma 1 (PLCG1) is a subtype of phospholipase C gamma (PLCg), a lipase activated by receptors in the cellular membrane including RTKs and adhesion receptors [38]. Wenqiang et al. reported that using in vitro and in vivo models of NSCLC and the phosphorylation of PLCG1 promote tumor growth in NSCLC, which is consistent with our results [38]. PTPN11, protein tyrosine phosphatase non-receptor type 11, is a member of the protein tyrosine phosphatase (PTP) family [39] and regulates several molecules involved in Ras signaling [40]. Mutations in PTPN11 have been linked to the pathogenesis of leukemia and breast cancer. However, a low prevalence of somatic PTPN11 mutations has been detected in lung cancer [41]. In addition to the prevalence of PTPN11 mutations, our results suggest that the phosphorylation of PTPN11 is related to NSCLC harboring sensitizing EGFR mutations. Altogether, these newly identified kinases with high activation are important for deepening the pathogenesis of NSCLC with sensitizing EGFR mutations.

Activated Kinases in NSCLC Patients with Poor Prognosis
This study revealed that three kinases, EGFR, ERBB2, and PIK3R1, were prognostic biomarkers because they were detected as highly activated kinases in the patient group with poor prognosis. The overexpression of EGFR is commonly observed in NSCLC patients (40-80%) [42,43] and is associated with poor prognosis [44][45][46]. HER2 is a member of the erbB transmembrane receptor family. Increased HER2 expression has an association with inferior survival in patients with NSCLC, and high EGFR and HER2 co-expression has an additive impact on unfavorable prognosis [47]. Additionally, Rikova et al. reported that NSCLC tumors express highly phosphorylated EGFR and ERBB2 at above-average concentrations [48], which was concordant with our results. Increased PI3K/Akt activity has also been observed in NSCLC. However, the finding that p-Akt has no association with EGFR-TKI efficacy is conflicting [49]. PIK3R1 encodes the regulatory subunit (p85a) of PIK3CA.
Activating mutations in PIK3R1 have been reported in several cancers, including colon cancer and glioblastoma, and lead to the activation of the PI3K-AKT pathway [49]. Our study identified PIK3R1 phosphorylation as a prognostic biomarker in patients with NSCLC with sensitizing EGFR mutations. In addition, we also elucidated that the co-occurrence of EGFR, HER2, and PIK3R1 phosphorylation was associated with poor prognosis in patients with NSCLC. Our gene ontology analysis and Reactome analysis demonstrated that the PI3K-AKT and RAF/MAPK signaling pathways were enriched in the group with poor prognosis, which supports newly identified kinases as predictive biomarkers. These observations strongly suggested the utility of the kinase profiling approach to the prognostic biomarker development, and warrant further validation studies using the additional samples.

Limitations
Our study had several limitations. First, the number of patients with a worse prognosis was too small compared with those with a favorable prognosis. Ideally, an equal number of patients with both prognoses should be included in the data-mining analysis. Secondly, since most of the enrolled patients did not have any adverse events, we could not determine whether the peptides and kinases were significantly associated with the adverse events. Third, three samples in cluster 2 seemed to be more similar to those in cluster 1. We examined the clinical features of those samples, but we could not identify the factor that would make an appropriate explanation for the obvious discordance. Fourth, many confounding factors such as the different efficacies of osimertinib vs other EGFR TKIs, the site of metastases, tumor burden, and performance status have not been taken into consideration. These limitations stemmed from the limited number of cases, and we could solve them by including more patients. Toward clinical applications, further investigation should be required to identify the predictive biomarkers with statistical significance. Fifth, the EGFR undergoes other posttranslational modifications which are not only Y phosphorylation but also S phosphorylation as proteoforms. In the current study, a single modification has been evaluated and the addition of other modification data will deepen our understanding of the proteoforms in advanced NSCLC patients harboring sensitive EGFR mutations. The current molecular targeted drugs were largely targeting Y phosphorylation, and we employed the PamChip which allows the investigation of only Y phosphorylation. The PamChip for S phosphorylation will further our understanding of the kinome backgrounds of cancer progression. Sixth, the STRING and Reactome depend on the literature review, and the kinase functions that were not reported yet were not considered for the interpretation. Thus, multiple interpretations other than those in this discussion will be possible in the near future. Overall, our analysis strongly suggested the utility of kinase profiling for the development of predictive biomarkers and warrants further investigation.

Conclusions
Comprehensive kinase activity analysis using 18 samples derived from patients with NSCLC harboring sensitizing EGFR mutations identified common kinomic profiles and kinases that are specific to patients with poor prognoses. We need more samples from different patients and more experiments using different methods to obtain conclusive experimental results. Continuous and collaborative efforts will thus be required to identify prognostic markers using comprehensive kinomic profiles.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/proteomes11010006/s1, Table S1: Raw data of kinase activity in 18 samples; Table S2: List of clustering analysis in 18 NSCLC patients; Table S3: Network analysis of kinases in NSCLC patients with common EGFR mutations; Table S4: List of peptides associated with prognosis; Table S5: Network analysis of kinases in the NSCLC patients with poor prognosis.
Funding: This study was supported in part by research funding from Nippon Boehringer Ingelheim Co., Ltd.
Institutional Review Board Statement: This study was conducted in accordance with the Declaration of Helsinki, and approved by the ethics committee of the National Cancer Center.
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.
Data Availability Statement: Not applicable.