Our investigation conforms to a three-arm study design and consisted of a genomic, miRNA and kinome profiling arm. We summarize the patient demographics in Fig. 1 and this includes sex, tumour type, TNM stage, tobacco smoking history, anatomical location and drug therapies. The majority are AD cases and the cohort consists of slightly more males (59%) than females (41%). Furthermore, 72.6% patients are TNM stage I-IIIA and based on self-reporting, the tobacco smoking history is surprisingly similar, i.e. 42.1% and 42.1%, respectively of current and never smokers. In regards to tumour location there are 3 major anatomical regions, i.e. inferior and superior right and superior left lobe, and majorly, the patients received adjuvant drug therapy (mainly clinical stages ≥ 2) which consisted of chemo, radio and targeted therapies. About 20% of patients were treated with ICIs either as stand-alone or a combination of chemo and targeted therapy, and 13% of patients received neoadjuvant therapy primarily consisting of chemo- and radiotherapy. See also supplementary Table S1 and S2 for further information.
Genomic profiling of different lung cancer types
To identify disease-regulated genes and miRNAs, we compared the genomes of tumours to NATs. Shown in supplementary Fig. 2 is the principal component analysis (PCA) analysis and the heatmaps clearly segregated tumours from NATs across the different histological LC types. We identified 439, 1240, 383 and 320 significantly upregulated genes for AD, SQ, NET and MT cases and there are 1092, 1477, 609 and 1267 downregulated DEGs (supplementary Table S3). Common to all tumours is the predominant repression of gene expression.
To identify DEGs specific for a histological type of lung cancer, we compared the genomes of AD, SQ, NET and MT cases, and identified 176, 895, 296 and 141 uniquely upregulated and 255, 550, 172 and 433 downregulated genes, respectively (Fig. 2a).
Furthermore, we identified 442, 302, 85 and 88 significantly upregulated miRNAs for AD, SQ, NET and MT cases, and for the same patients there are 17, 8, 10 and 15 downregulated miRNAs. In addition, we identified 150, 47, 1 and 1 uniquely upregulated and 4, 2, 2 and 3 downregulated miRNAs in AD, SQ, NET and MT (Fig. 2b). In summary, miRNAs are primarily upregulated in tumours, whereas transcriptomes are predominantly repressed (Fig. 2b and supplementary Table S4).
We performed gene enrichment analysis for DEGs which are specifically regulated in the various histological types of lung cancer and compared significantly enriched terms between them. As shown in supplementary Fig. S4, and with the exception of the hallmark gene set mTORC1 signalling, which is common to SQ and MT, none of the terms overlapped. Therefore, the gene enrichment analysis is specific for a given histological type of lung cancer. For AD, and among unique terms for upregulated DEGs, we wish to emphasize KRAS and EGFR signalling and protein glycosylation. Significantly enriched terms for SQ are the hallmark gene sets E2F and MYC targets, hypoxia and estrogen response and for NET tumours, neuroepithelial cell differentiation and GABAergic signalling. Finally, enriched terms for MT are the hallmark gene set MTORC1 signalling, activation of GTPase activity, chromosome organization and DNA repair. We also considered enriched terms for downregulated DEGs. Here, TNFα signalling, response to cytokine stimulus, actin filament-based process and immune response, regulation of MAPK cascades and cell adhesion are specific for AD, SQ, NET and MT tumours, respectively (Fig. 2c).
To independently validate the results of the present study (= “Hannover cohort”), we questioned the TCGA database (https://www.cancer.gov/ccg/research/genome-sequencing/tcga) and applied the following criteria: DEGs are defined with an FDR adjusted p-value < 0.05, a |FC|≥2 in ≥ 50% of cases. Note the TCGA repository does only compile LUAD and LUSC cases. Therefore, we compared 524 LUAD cases to 58 NATs and separated the data by clinical stages. We searched for common DEGs between the two cohorts and found > 70% of the up- and downregulated DEGs as commonly regulated. The results are similar when compared to the different histological types of LC (Venn diagram, Fig. 3a, and supplementary Table S5). Therefore, the data agreed reasonably well.
Depicted in Fig. 3b are the heatmaps for DEGs of the Hannover cohort and the colour codes represent z-scores which measures the standard deviations from the mean of a gene expression values. We obtained perfect segregation of AD and SQ tumours from NATs.
Subsequently, we performed gene enrichment analysis with the Metascape software 22 and selected the top 100 regulated genes common between the Hannover and TCGA cohorts. The FC of the selected genes ranged from 3 to 14-fold. We queried the Hallmark gene sets and KEGG pathways, and show the results in Fig. 3c. For upregulated genes, the Hallmark gene sets underscored mTOR signalling, G2M checkpoint, E2F targets and glycolysis in lung adenocarcinomas (AD cases). Similar, KEGG emphasized cell cycle, p53 signalling, cellular senescence and cancer pathway.
In regards to downregulated genes (range 5 to 20-fold), the hallmark gene sets underscored apoptosis, inflammatory response and TNFα signalling via NFKB. We obtained similar results with KEGG and prominent examples of enriched terms are TNFα signalling, immune and chemokine signalling and leukocyte migration (Fig. 3c).
In the same way, we analysed SQ tumours (LUSC) and for upregulated DEGs (range 6 to 194-fold), the Hallmark gene sets are similar to AD cases, i.e. mitotic spindle, E2F targets, G2M checkpoints and mTOR signalling while KEGG pathway analysis highlighted p53 signalling, cell cycle and cellular senescence. For repressed DEGs (range 9 to 72-fold) the Hallmark gene sets emphasized inflammatory response, KRAS signalling and DNA repair whereas KEGG underscored phagosome, complement and coagulation cascades and ECM receptor interactions. We summarize the results in supplementary Table S5.
Unlike the cancer transcriptomic data described above, there was little overlap among DEMs between the Hannover and TCGA cohort (Fig. 3d). In fact, for AD cases only 6% of up- and downregulated DEMs are common to both cohorts, whereas for SQ cases, 5% and 25% DEMs are commonly regulated. Given that a single miRNA may regulate multiple genes, whereas one gene can be regulated by multiple miRNAs, it is reasonable, that less miRNAs are commonly regulated between the two cohorts.
Gene networks in different histological types of lung cancer
To gain insight into the molecular wiring across different histological types of lung cancer, we constructed regulatory gene networks, and to ascertain the relationship between DEMs and DEGs, we questioned the miRNet data base. This revealed experimentally proven miRNA gene targets which we compared to DEGs specific for the different types of LC (Fig. 2a). Based on the paradigm that miRNAs repress the expression of gene targets, we identified 253, 417, 121 and 332 DEGs specific for AD, SQ, NET and MT (Fig. 4a). Note all these target genes are repressed in lung tumour samples (Fig. 2a). Conversely, for repressed DEMs, we discovered 48, 291, 29 and 0 upregulated DEGs. Furthermore, there are 85 targets in common, i.e. irrespective of the histological type of lung cancer and the gene enrichment analysis emphasized UV response (= DNA repair), hallmark gene set EMT and TNFα signalling for upregulated DEMs (Fig. 4a). On the other hand, only 1 target gene is common for repressed DEMs, i.e. nucleoside diphosphate kinase 1 which we found upregulated in all lung tumour samples and this kinase supports metastatic spread.
To be able to construct regulatory miRNA-gene networks, we performed Metascape analysis and searched for enriched terms. We selected the top 10 biological processes based on FDR adjusted p-values and requested that the term must cover > 80% of DEGs specific for a given tumour type (Fig. 4a). We likewise selected the top 10 miRNAs based on the number of target genes for each biological process and compile the individual data in supplementary Table S6. In this way, we constructed miRNA-gene networks specific for AD, SQ, NET and MT lung cancers. We compared the terms among the different types of lung cancer and the Venn diagram in Fig. 4b shows that none overlapped amongst the various histological types of lung cancer. We obtained similar results for repressed DEMs for which we considered terms associated with upregulated DEGs. In fact, in the comparison of AD versus SQ, there is only one term in common, i.e., peptidyl amino acid modification. Together, this demonstrates specificity for the miRNA-gene networks.
Shown in Fig. 4c-4f are the regulatory miRNA-gene networks for the different histological types of lung cancer, and we used the Cytoscape software to create the network. Although the terms are more generic in nature, the underlying DEGs are highly specific and encompass potential drug targets. For instance, for the SQ network (Fig. 4d1) the Metascape analysis revealed cell cycle phase transition as a statistically significant enriched term and embedded in this term are highly regulated DEGs, notably CDK1, CHEK1 and GSK3B which are induced by 8-, 4- and 2-fold, respectively. Strikingly, the activities of these kinases are also significantly increased in the same patient derived tumour biopsies (see below). A further example relates to peptidyl-amino acid modification and among genes embedded in this term is the 5-fold upregulated NTRK2. Importantly, the kinase activity of NTRK2 is also significantly increased, and its inhibitor Entrectinib, Larotrecitinib are already approved for the treatment of solid tumours in patients diagnosed with NTRK2 fusion proteins. Together, the miRNA-gene networks are of high relevance for an identification of potential drug targets.
Kinome profiling in LC
We used the PamGene® platform to identify aberrant kinase activities in tumour and NATs of 54 patients. The assay measures the phosphorylation of peptides, and each peptide contains canonical amino acid sequences which are recognized by distinct serine/threonine and tyrosine kinases. We obtained two sets of data. First, direct measurement of the phosphorylation of kinases, and second, an identification of regulated kinase based on the phosphorylation of canonical amino acid sequences. Here, the upstream kinase phosphorylates a specific substrate, and therefore, the phosphopeptide is a read out of the kinase activity. Depicted in Fig. 5a is a Venn diagram of phosphorylated peptides across the different histological types of LC. We measured an increased phosphorylation of 156 peptides with 95 being common between the different types of LC (Fig. 5a). Conversely, there are 28, 1, 2 and 1 phosphorylated peptide uniquely associated with AD, SQ, NET and MT tumours. Of the 156 phosphorylated peptides there are 72 which allow direct measurement of regulated kinases. Note, for some kinases, such as the EGFR, we measured the phosphorylation of several peptides which contain specific tyrosine residues and we show in Fig. 5b significantly regulated kinases between the different types of LC. Together, we identified 53 significantly regulated kinases of which 46 are common (Fig. 5b and supplementary Table S7), and we did not observe repressed kinase activities but noticed primarily upregulation of PTKs, i.e. 67% (Fig. 5c). Furthermore, based on upstream kinase analysis of the 156 peptides, we identified 35 serine/threonine kinases specifically regulated in AD (Fig. 5d) and these kinases phosphorylate 28 peptides (Fig. 5a). Moreover, upstream kinase analysis of the phosphorylated peptides revealed 71 kinases regulated in common (Fig. 5d, supplementary Fig. S5a-5d) of which 67.6% are tyrosine kinases, and we show in supplementary Fig. S4 examples of kinetic plots for tumour and NAT biopsies. Converging the results of Fig. 5b (direct measurement of 53 phosphorylated kinases) and 5d (upstream kinase analysis of phosphorylated peptides) yielded a total of 137 distinct kinases which are significantly upregulated in LC (supplementary Fig. S5e).
The heatmap shown in Fig. 5e informs on tumour regulated kinases, and the data are log2 fold increases in kinase activity across different histological types of LC. In supplementary Table S7 we specify the amino acid sequence of the target protein which becomes phosphorylated. For instance, we observed increased phosphorylation (up to 9-fold) of the peptide sequence GSVQNPVYHNQPL which are the amino acids 1103–1115 of the EGFR receptor. This peptide contains one tyrosine residues, i.e. tyrosine 1110 of the EGFR intracellular kinase domain and its selective phosphorylation allows signalling via GRB2-GAB1 and MAPK to stimulate MYC activity. Likewise, phosphorylation of EGFR tyrosine 1110 stimulates PI3K/AKT signalling and activation of MYC 23. Moreover, we observed increased phosphorylation of scaffolding adapter protein GAB2 which facilitates downstream signalling of EGFR and measured an up to 26-fold increased phosphorylation of its peptide sequence DEKVDYVQVDK, which are the amino acids 638–648 of the GAB2 protein. Meanwhile, the tyrosine residue of amino acid 643 is phosphorylated by JAK2. Importantly, JAK2 activity is upregulated across all histological types of lung cancer (about 5-fold), and we observed increased phosphorylation of its tyrosine residue 570 (see Fig. 5d and supplementary Table S7, amino acid sequence 563–577 of the JAK2 protein VRREVGDYGQLHETE).
A further example relates to the adapter protein/proto-oncogene CRK which binds to tyrosine-phosphorylated proteins. We measured the phosphorylation of the amino acid sequence 214–226, i.e. GPPEPGPYAQPSV of the CRK protein which contains one tyrosine residue at amino acid 221 and its tyrosine phosphorylation was up to 25-fold induced in LC. The CRK protein is phosphorylated by the tyrosine kinase ABL1.
Among the common regulated kinases, we observed an up to 10-fold induced phosphorylation activity for the platelet-derived growth factor receptor-β (PDGFRB). The receptor undergoes autophosphorylation 24 and we assayed the phosphorylation of 6 peptides of PDGFRB (supplementary Table S7). We observed differences in the activity for the different peptides and show in Fig. 5e highly induced autophosphorylation of the amino acid sequence 771–783 (YMAPYDNYVPSAP) in SQ. Conversely, the phosphorylation of amino acid sequence 709–721 RPPSAELYSNALP increased by about 3- and 4-fold, respectively in AD and SQ cases (supplementary Table S7).
In stark contrast, we measured increased phosphorylation of the peptide sequence ELLCLRRSSLKAY, i.e. amino acid 338–350 of the β2-adrenergic receptor (ADRB2) in AD cases only. This peptide contains two serine residues and can be phosphorylated by protein kinase A (PKA). Note PKA is a cAMP-dependent kinase and upregulation of ADRB2 in LC has been reported 25,26. In fact, targeting ADRB2 sensitizes tumour cells to VEGFR-TK inhibitors 27. Among the AD specific STKs we observed induced phosphorylation of the CREB1, i.e. sequence EILSRRPSYRKIL at serine 129 and 133 and this transcription factor is likewise phosphorylated by PKA. Similar, we observed induced phosphorylation of the serine/threonine kinase RAF1 by Src, and this kinase phosphorylates the tyrosine residues 340 and 341 of the RAF1 protein sequence PRGQRDSSYYWEI, i.e. amino acid residues 332–344.
Another interesting example relates to the histone H3-like centromeric protein a (CENPA). We assayed the peptide MGPRRRSRKPEAPR at amino acid residues 1–14 and noted its > 2-fold induced phosphorylation in AD cases. Importantly, CENPA can be phosphorylated by the serine/threonine aurora kinase A and B 28 which we found induced at the gene expression level by > 3 and 5-fold, respectively in the Hannover AD and TCGA cohort.
Collectively, there are many possibilities to block signalling events in different histological types of LC, and for three-quarters of patients (supplementary Fig. S5f) we identified upregulation of > 20 kinases. We provide information on kinase inhibitors approved by the FDA (Fig. 5e, blue coloured quadrants). Together, 83 kinase inhibitors target 24 kinases all of which are approved by the FDA, and 20 inhibitors targeting 8 kinases (ALK, BRAF, EGFR, ERBB2, MEK, MET, RET, NTRK2) are already cleared for the treatment of lung cancer. In supplementary Table S8 we compile all kinase inhibitors and their targets.
Next, we addressed the questions whether the genes coding for the kinases are also regulated at the gene expression level and show the results in Fig. 5f. There are 4 (PTK6, CDK1, AURKA and EPHB2) and 10 kinases (PAK1, PKMYT1, AURKB, CHEK1, CDK1, JAK3, EPHB2, PRKCG, PTK6 and AURKA), respectively of the Hannover-AD and TCGA cohort whose gene expression and kinase activity are regulated in the same way. In other words, the gene coding for the kinase and the activity of the coded protein is increased. Similar, for SQ there are 8 (CDK1, CHEK1, EPHA4, GSK3B, MAP2K6, NTRK2, PAK1 and PTK2) and 5 kinases (CDK1, CHEK1, EPHB2, PAK1 and PKMYT1) in common between the two cohorts (Fig. 5f). For neuroendocrine tumours, RET is the only kinase where the gene expression (11-fold) and kinase activity (5-fold) is induced. Together, the data implies that only a few kinases are simultaneously regulated at the gene expression and kinase activity level. Therefore, kinase profiling is of critical importance.
Additionally, we determined the prognostic value and computed univariate Cox regression proportional hazard models for genes that code for 110 significantly regulated kinase based on the upstream analysis (Fig. 5d and supplementary Table S9). We identified 37 kinase coding genes in AD with a statistically significant HR. However, only for 18 kinases the HR was > 1 (range 1.1–1.79) where as for the remaining kinase coding genes the HR was < 1 (range 0.34–0.87). Similar for the SQ cases there are 3 and 3 kinase coding genes, respectively with either a statistically significant increased (range 1.2–1.37) or reduced HR (range 0.65–0.9, see supplementary Table S9). As described above only a few kinases are simultaneously regulated at the gene expression and kinase activity level, and therefore, the prognostic value of the univariate Cox regression proportional hazard models based on the expression of the kinase coding genes is less obvious.
Notwithstanding, we were able to compute meaningful Kaplan-Meier survival curves for kinases where the change in gene expression and kinase activity agreed, and depicted in Fig. 5g are the plots for the kinases AURKA, AURKB, CDK1, CHEK1, EPHB2 and PKMYT1. Thus, the prognostic value for these kinases could be established and in the case of CDK1, clinical phase II trials in NSCLC and SCLC are already ongoing (NCT05651269 and NCT02161419).
Furthermore, for the different histological types of LC, we constructed networks that describe the relationship between miRNAs, their gene targets and the activity of the kinase protein. (Fig. 5h). As an example, the expression of the genes coding for PDGFRA, KDR and NTRK2 kinase are down regulated whereas the miRNAs targeting these kinases are upregulated. Although this fits the paradigm, i.e. miRNAs repress gene translation, we assayed induced kinase activity and this implies independent mechanisms for their regulation. We also identified upregulated miRNAs linked to induced expression of the kinase coding genes, as denoted for CDK1, ARUKA and PTK6, and the kinase activity was likewise induced. Our results reinforce the notion, that expression of the kinase coding gene and the activity of the coded protein cannot easily be correlated. Despite this limitation, the networks help to define therapeutic targets as described below (see master regulatory networks).
Commonly regulated kinases in lung cancer
Depicted in Fig. 6a are the complex signalling networks in LC, and we specify the signalling networks in tumour and immune cells. Based on the paradigm that ≥ 70% of patients show the same change, i.e. induced kinase activity, we identified 45 and 47 significantly regulated kinases in adeno- and squamous lung cancer cases of which 34 are regulated in common (Fig. 6b, Venn diagram). Note we did not observe repressed kinase activities. In the same way we analysed the neuroendocrine and metastatic tumours and identified 74 and 27 kinases as significantly upregulated. Additionally, an irrespective of the histologic type of LC we identified 17 tyrosine and 4 serine-threonine kinases as commonly upregulated between 54 LC patients (Fig. 6c).
As shown in Fig. 6a, RAS is not regulated in the present study (grey coloured) while the rectangular marked proteins are based on upstream kinase analysis. Proteins marked by the symbol ℗ are direct measurements of phosphorylated kinases. Additionally, we identified regulated kinases by upstream analysis by assaying the phosphorylation of canonical peptides. An important finding of our study is an upregulation of a wide range of receptor tyrosine kinases in tumours such as the vascular endothelial growth factor receptor 2 (VEGF2/KDR), epidermal growth factor receptors (EGFR and ERBB2), ephrin type-A&B receptor (EPHA & EPHB2), platelet derived growth factor receptor (PDGFR), the RET receptor (RET) and the MER, MET and RON proto-oncogenes. Additionally, we assayed peptides typically phosphorylated by AXL (Tyro3-Axl-Mer (TAM) receptor) tyrosine kinase, the insulin (InSR) and insulin-like growth factor 1 receptor (IGF1R), the anaplastic lymphoma receptor (ALK), neurotrophic TK, the kit proto-oncogene and the fibroblast growth factor and obtained clear evidence for their increased activity in tumours of LC patients (range in AD 2 to 4-fold, in SQ 2 to 24-fold). Moreover, for some of the receptor tyrosine kinases, we determined the phosphorylation sites and show the downstream signalling of kinases which are regulated as well. For instance, we assayed the phosphorylation of 3 peptides of the EGFR, notably the amino acid sequence (AS) 1103–1115 (GSVQNPYHNQPL), the AS 1165–1177 (ISLDNPDYQQDFF) and AS 1190–1202 (STAENAEYLRVAP) and observed differences between the various histological types of LC. The differences in the tyrosine phosphorylation are due to variant SRC and ABL1 kinase activity with implications for downstream signalling events. Similar, we assayed the tyrosine phosphorylation of the MAPK1 peptide HTGFLTEYVATRW (AS 180–192) which we found significantly increased in SQ cases only (range 1.2-3.7-fold). This peptide can be phosphorylated by the kinases RAF1, RET, JAK2, ERK1 and LYN and all of them are upregulated in the Hannover cohort of LC patients. Likewise, we observed increased phosphorylation of the tyrosine residue 185 of the p38gamma/MAPK12 peptide ADSEMTGYVVTRW, and one investigation reported the precisely ordered phosphorylation reactions of the p38 Mitogen-activated protein (MAP) kinase which involves the kinases MKK3/6, MEKK6, SEK1 and ASK 29.
Tumor cell infiltrating lymphocytes
A major finding of our study is the unique regulation of kinases in tumour infiltrating lymphocytes. Importantly these kinases were regulated in all tumours irrespective of the histological type, and the data are based on 54 individual cases. In regards to the T-cell receptor (TCR), the data are complex. In general, the TCR αß and γδ heterodimers engage with CD3 molecules 30 and the tyrosine residues of the CD247/CD3ζ chain are of critical importance in activating TCR. We observed 10-fold induced phosphorylation of the immunoreceptor tyrosine-based activation motif (ITAM) KDKMAEAYSEIGM of CD247/CD3ζ. The phosphorylation of this ITAM is catalysed by the ZAP70 kinase, and we found its activity likewise increased by 3 to 4-fold in LC. Moreover, the lymphocyte cell-specific protein tyrosine kinase LCK phosphorylates the ZAP70 protein (= ζ chain of TCR associated protein kinase 70), and we measured a 4-fold increased LCK activity in tumour biopsies of LC patients. Additionally, the CD3ε polypeptide PVPNPDYEPIRKG is phosphorylated by LCK and phosphorylation of this peptide is also increased by 4-fold. Together, we observed increased phosphorylation of the CD3ε and CD3ζ chains and measured increased activity of kinases catalysing the phosphorylation of these peptide chains. Obviously, this will support assembly of the TCR-CD3 complex. Importantly, the multiple signalling roles of CD3ε were recently discovered and through a series of mechanistic studies, it was shown that the kinase CSK, and the protein tyrosine phosphatases SHP1 and SHP2 inhibit TCR signalling 31. We assayed the CD3ε specific peptides PVPNPDYEPIRKG at AS 182–194 and KGQRDLYSGLNQR at AS 193–205; however, determined only for the first ITAM (AS 182–194) a 4-fold increased phosphorylation. Furthermore, the phosphatases SHP1 and SHP2 are activated by tyrosine phosphorylation 32. We assayed the phosphorylation of the SHP1 at AS 558–570 KHKEDVYENLHTK and of Shp2 at AS 580–590 SARVYENVGLM and for both peptides the tyrosine phosphorylation increased up to 4-fold. Moreover, the SHP1 peptide is phosphorylated by LCK and its activity is similarly upregulated by 4-fold (supplementary Table S7). Furthermore, SHP2 regulates SRC family kinase activity and RAS/ERK activation by controlling CSK recruitment 33.
Concerning the B cell receptor, the only peptide available on the PamGene platform is the ITAM of the Igα chain with a AS 181–193 EYEDENLYEGLNL. Its tyrosine specific phosphorylation is catalysed by SYK and this kinase is activated by LYN. We observed a significant 4-and 5-fold increased activity of the LYN and SYK kinases and determined an up to 4-fold increased ITAM phosphorylation of the Igα chain in neuroendocrine LC cases. The data implies activation of the pro to the pre-B cell stage but is confounded by the limited information available. Furthermore, the BCR induces phosphorylation of PLC-γ2 on tyrosine Y1217 which contributes to an activation of this phospholipase 34.
A further example relates to the TEC tyrosine kinase which is known to have multiple functions on the immune system and T cell signalling 35. We assayed the peptide RYFLDDQYTSSSG at AS 512–524 of the TEC kinase and found its activity significantly increased by 4-fold.
Collectively, while some essential components of the TCR complex are activated, inhibitors of TCR signalling are likewise upregulated. The results are suggestive for a dysfunctional TCR. Interestingly, none of the gene coding for the TCR complex were significantly regulated in LC patients (supplementary Table S3), thus emphasizing the need to perform phosphopeptide analysis.
Apart from the complex TCR signalling events in the tumour microenvironment, we wish to emphasize activation of the PI3K/AKT which is mediated through a range of kinases in response to insulin receptor and insulin receptor substrate signalling (Fig. 6a). In LC, both kinases were upregulated 4 to 5-fold, and its aberrant regulation is frequently observed in cancers 36. Within this pathway GSK3B is also 3-fold upregulated. A further example relates to an activation of the MAPK pathway, and in LC tumour biopsies, we measured an average 5-fold increased phosphorylation of the peptide PRGQRDSSYYWEI at AS 332–344 of the RAF1 protein. Similar, the activity of the ERK1 and ERK5 kinases increased by 5-fold as evidenced by the phosphorylation of the ERK1 peptide GFLTEYVATR at AS 199–208 and the ERK5 peptide AEHQYFMTEYVAT at AS 212–224. Moreover, we observed an extraordinary increased activities of the CRK and EFS kinases, i.e. 25- and 7-fold, respectively. Upon MAPK activation RAF phosphorylates MEK and MEK phosphorylates ERK. Notwithstanding, an alternative route of ERK phosphorylation involves the CRK phosphorylation of the EFS kinase which in turn phosphorylates Src and eventually ERK 37. Noteworthy, CRK selectively regulates T cell migration 38.
Different kinomes in lung adeno- and squamous cell carcinoma
Unlike lung adenocarcinomas, targeted therapy in SQ is still in its infancy 39. We compared the kinase activities between different histological types of LC and report the results for AD and SQ. Shown in Fig. 6d is the fold change difference of significantly regulated kinases between the two histological types of LC. Specifically, we measured the activity of kinases in tumour tissue relative to NATs and compared the results between AD and SQ patients. Remarkably, of the 46 commonly regulated kinases (Fig. 5b), the activities of 17 kinases are even higher in SQ, and for some patients, we measured unprecedented high activities as denoted for the FGFR1, PGFRB; ANXA2, CD3ζ, GAB2, LYN and MAPK12. It was not unexpected to see higher FGFR1 activities given its common alterations in SQ. However, we also observed up to 10-fold induced activities of the phosphoinositide-3-kinase regulatory subunit 1 and therefore demonstrate high activity of the phosphatidylinositol 3-kinase subunit. Together, the data underscores the great opportunities for targeted therapies in LC, and we obtained strong evidence for a broad range of kinases that are also potential targets for the treatment of SQ.
Additionally, we compared kinase activities of patients diagnosed with driver mutations in EGFR, KRAS, p53, BRAF and MYC (supplementary Table S2) to patients without driver mutations and compile the results in supplementary Table S10. Essentially, we did not observe significant differences in kinase activities between carries of KRAS mutations and KRAS wild type. For instance, we compared the kinase profile of > 3-fold significantly regulated kinases of 5 AD patients with the KRAS mutations G13D, G12A, G12C, G12V and KRAS amplifications to 9 AD patients without driver mutations and did not observe significant differences between the two cohorts (Fig. 6e). Therefore, apart from an identification of mutated kinases other kinases are regulated as well and therefore are bona fide therapeutic targets.
Lastly, we identified kinases which are specifically regulated in AD and show in supplementary Fig. S6 the signalling networks for 28 peptides. The peptides are phosphorylated by 35 different serine/threonine kinases (supplementary Table S7) and with the exception of PI3K, GSK3B, RAF and ERK the kinases are defined by upstream analysis, i.e. kinases known to phosphorylate these peptides (Fig. 5d). Our findings underscore the therapeutic opportunities in blocking tumour associated signalling networks.
Identification of molecular hub proteins to block tumour specific signalling networks
We define hub proteins as master regulatory molecules which are at the apex of a signalling cascades, and their inhibition results in disintegration, fragmentation and abrogation of signalling events. By combining the kinome and genomic data and based on computational analysis of DEGs, we searched for master regulators (MRs) in LC associated signalling networks. By querying the TRANSPATH database, we obtained information on signalling molecules in different pathways and downstream reactions. We identified several kinases as MRs and show in Fig. 7 the results for the different histological types of LC. Although the number of MRs differed between AD and SQ, i.e. 7 vs 3 we did not identify specific MRs for these tumour entities. Conversely, for NET and MT there are 5 and 3 MRs, respectively which are specific to these histological types of LC (supplementary Fig. S7). Importantly, the networks shown in Fig. 7 depict significantly regulated MRs, which function as kinases. Additionally, there are significantly downstream regulated kinases, in addition to significant DEGs coding for signalling molecules in the networks. Lastly, miRNAs control the expression of DEGs. For a given network, we visualize the hierarchy of signalling molecules (supplementary Fig. S8-S11) and the MRs function primarily as receptor tyrosine kinases, MAPKs and cyclin dependent kinases to influence cell proliferation, cytoskeletal dynamics and TCR/immune cell responses.
Validation of tumour regulated kinases in human lung cancer cell lines by siRNA and cell viability assays.
In an earlier study, Campbell and colleagues performed large-scale profiling of kinase dependencies in cancer cell lines 40. By considering their published siRNA mediated gene knock-down and lung cancer cell viability data, we were able to independently validate nearly half of the kinases found to be significantly upregulated in > 70% of patients. Furthermore, we queried the DepMap database (https://depmap.org/portal/) to retrieve kinase dependencies data from lung cancer cell lines, and the results of gene knock down of kinases on cell viability in a panel of human lung cancer cell lines are given in supplementary Table S11.
Among the 21 commonly upregulated kinases (Fig. 6c), we confirmed loss of cell viability following gene knock-down of CDK4, FER, FES, JAK1&2, LCK and MET in 12 human LC cell lines (supplementary Table S11). Moreover, the detrimental effects of CRISPR-mediated gene knockdown of JAK1 on cell viability was confirmed. In regards to kinases specifically regulated in AD (Fig. 6b), siRNA mediated gene-silencing of CDK1, EPHA1, JAK3, RET and ZAP70 caused loss of cell viability in a panel of 7 human adenocarcinoma cell lines (supplementary Table S11). Moreover, siRNA of the kinase ephrin receptor EPHA1 caused cell death in the large cell lung carcinoma cell line BEN. Concerning squamous cell carcinoma, siRNA of FGFR2 and PDPK1 led to significant reductions in cell viability of the human lung cancer cell lines H460, H23 and A427. Finally, we confirmed significantly reduced cell viability in the carcinoid cell line H727 following siRNA of TYK2, i.e. tyrosine kinase 2 which is specifically upregulated in carcinoid neuroendocrine tumours (supplementary Table S7).