Proteome datasets of lung SqCCs and PPAs
MS-based proteomic analysis was conducted on the FFPE tissue specimens comprising seven SqCCs and eight PPAs. These specimens were selected for their preserved condition, tumor area, and well-clarified pathological diagnosis (Table 1). Presurgical treatment was not performed for any of these lung adenocarcinomas. Statistical t-test on the smoking Brinkmann index (BI) exhibited that SqCC was significantly associated with the extent of smoking (p = 0.018).
A total of 2,108 proteins were identified, among which 1,281 (60.8%) were commonly expressed in the cancerous cells of both SqCC and PPA. One hundred and fifteen (5.5%) and 712 (33.8%) were unique to SqCC and PPA, respectively (Fig. 1A).
We subjected 1,396 and 1,993 proteins expressed in SqCC and PPA, respectively, to gene ontology (GO) analysis using the Protein Analysis Through Evolutionary Relationships (PANTHER) version 16.0 software program (The Thomas Lab, University of Southern California, Los Angeles, CA, USA) [10], and the results were notably similar between the two subtypes (Fig. 1B). Common to both subtypes, proteins were abundantly associated with the cellular process, localization, biological regulation, metabolic process, and response to stimulus in the biological process (GO), and with the cytoskeletal protein, transporter, nucleic acid metabolism protein, protein-binding activity modulator, translational protein, metabolite interconversion enzyme, protein modifying enzyme, membrane traffic protein, chaperone, and hydrolase in the protein class (GO) (Fig. 1B).
Identification of data-driven key protein network modules by WGCNA
We identified 30 protein modules by constructing weighted protein co-expression networks, in which all the identified proteins were clustered (Fig. 2A). The WGCNA analysis was performed with a soft threshold power of 25 selected for approximate scale-free topology, a minimum module size of 15, and a module detection sensitivity deepSplit of 4. The traits used in the WGCNA analysis were the lung cancer subtypes, SqCC and PPA. Correlations were obtained between resultant modules and traits to identify the protein modules significant to respective traits. A heatmap of the eigen protein expressions and samples (Fig. 2B) and pairwise correlations between the modules regarding eigen-protein expressions (Fig. 2C) are presented respectively.
We identified nine modules that showed high and/or moderate correlations (correlation: |r| > 0.5) and statistical significances (multiple testing correction with the Benjamini-Hochberg method: q-value < 0.05) with clinical traits (Fig. S2). We focused on the protein co-expression network modules significantly associated with lung SqCC, which were found to be WM26, WM27, and WM28 (indicated by the red dashed squares) whereas six modules, WM11, WM13, WM15, WM16, WM17, and WM18 (indicated by the blue dashed squares), were significant to PPA.
Functional enrichment analysis of the PPI networks
The biological associations between the proteins in each key protein network significant to SqCC were analyzed by mapping the network proteins in the human protein-protein interaction (PPI) network and by pathway enrichment (Fig. 3).
We used the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database [11] to generate the PPI networks for the three WGCNA network modules associated with SqCC - WM26, WM27, and WM28, and were reconstructed using the Cytoscape version 3.8.2 software program (Institute for Systems Biology, Seattle, WA, USA). Top hub proteins were determined using the cytoHubba plugin by maximal clique centrality (MCC) [12]. The three WGCNA modules, WM26 (r = 0.851, q = 0.0004), WM27 (r = 0.900, q = 7.34 ´ 10-5), and WM28 (r = 0.884, q = 0.0001), were significantly correlated with SqCC, where eigen proteins and/or hub proteins are indicated by blue and red dotted circles, respectively (Fig. 3). For the top three WGCNA modules significant to PPA (WM15, WM16, and WM18), their protein networks and pathways enriched are presented in Fig. S3.
The three WGCNA protein network modules were significant to SqCC (Fig. 3A). The pathways enriched for WM26 (Fig. 3B) included 1) biological process (GO): cornification, keratinization, and intermediate filament cytoskeleton organization; 2) Reactome pathways: formation of the cornified envelope; and 3) STRING local network cluster: intermediate filament protein and keratin type II head and keratinocyte migration. Subnetwork 1 was significant to translation, in which key proteins and their associated pathways included EEF1B and EEF1D, eukaryotic translation elongation; CCT3 and CCT4, chaperonin-mediated protein folding; EIF4G1, signaling by insulin receptor; RAD23B, nucleotide excision repair; and PSMC1, cell cycle progression, apoptosis, or DNA damage repair. Translation initiation factor eIF4G was overexpressed in SqCC [13]. Subnetwork 2 is significant to the oxidoreduction coenzyme metabolic process and/or glycolysis and gluconeogenesis. The group of CD44 and MIF (macrophage migration inhibitory factor) suggests the negative regulation of DNA damage response and intrinsic apoptotic signaling pathway by the p53 class mediator. The 14-3-3 proteins, SFN (stratifin, 14-3-3 sigma, or epithelial cell marker protein 1), YWHAG (14-3-3 gamma), and YWHAH (14-3-3 eta), are involved in protein insertion into the mitochondrial membrane, which is involved in the apoptotic signaling pathway.
Niemira et al. conducted the RNA-seq based profiling of tissues obtained from 114 patients with NSCLC who received tumor resection surgery, followed by bioinformatics analyses combined with WGCNA [14]. Their GO enrichment analysis of genes differentially expressed in lung SqCC revealed deregulated processes of cornification, epidermis development, keratinization, and epidermal cell differentiation [14], which is consistent with our results. Pak et al. investigated the morphologic characteristics of lung SqCC and concluded that keratinization in lung SqCC was associated with a poor prognosis and mostly with smoking [15]. The hub proteins in this module were the series of intermediate filament proteins including the eigen-protein KRT72 (keratin 72). KRT5 (keratin 5), KRT6A (keratin 6A), KRT6B (keratin 6B), DSG3 (desmoglein 3), and TRIM29 (tripartite motif-containing protein 29) are also highly connected proteins in this module, which were reported as potential biomarkers for distinguishing between SqCC and lung adenocarcinoma [16]. DSG3, a member of seven transmembrane desmosomal cadherins, was upregulated in various SqCC tissues, and its expression level correlated with clinical stages [17]. TRIM29, also known as the ataxia group D complementary gene (ATDC), is a transcriptional regulator involved in cell proliferation, differentiation, infiltration, migration, and invasion [18]. Expressions of TRIM29 were upregulated in numerous cancer types and were suggested to promote lung SqCC cell metastasis by regulating the autophagic degradation of E-cadherin [19].
The enriched pathways of the WM27 module included 1) biological process (GO): translational initiation; 2) Reactome pathways: L13a-mediated translational silencing of ceruloplasmin expression, translation, and nonsense-mediated decay (NMD) independent of the exon junction complex; and 3) STRING local cluster: signal recognition particle-dependent cotranslational protein targeting the membrane, and peptide chain elongation (Fig. 3). Subnetwork 3 is significant for protein refolding, and the Ras-related proteins RAB5A and RAB6A are involved in cytosolic transport. The eigen-protein PAFAH1B2 (platelet-activating factor acetylhydrolase IB subunit alpha2, also known as the PAF-AH 30 kDa subunit) was overexpressed in some types of tumors including lung cancer [20]. The transcription of PAFAH1B2 is directly initiated by HIF1a activated in a hypoxic environment, and its overexpression induces epithelial-mesenchymal transition and subsequent aggressive phenotypes [21]. The hub proteins included the ribosomal proteins -RPLP1, RPL26, RPS25, RPS21, and RPL22, which participate in NMD, and EIF3F. Under cellular stress, translation initiation switches from cap-dependent translation to alternative mechanisms such as internal ribosome entry site (IRES) initiation [22]. RPS25 is the key ribosomal protein mediating c-MYC IRES-dependent translation under endoplasmic reticulum stress [23]. EIF3F (eukaryotic translation initiation factor 3F) was overexpressed in lung cancer cells, which were reported to reprogram cell proliferation and energy metabolism [24].
The enriched pathways of WM28 (Fig. 3) included 1) biological process (GO): translational initiation, nuclear-transcribed mRNA catabolic process, and NMD; 2) Reactome pathways: L13a-mediated translational silencing of ceruloplasmin expression, GTP hydrolysis, and joining of the 60S ribosomal subunit, peptide chain elongation, and NMD independent of the exon junction complex; and 3) STRING local network cluster: GTP hydrolysis and joining of the 60S ribosomal subunit and protein export, and peptide chain elongation. Thus, WM27 and WM28 shared almost the same enriched pathways. The hub proteins in this module were the ribosomal proteins including the eigen-protein RPS3A. The expression of RPS3A was upregulated, and its frequent enhancement was reported in patients with SqCC [25]. RACK1 (receptor for activated C kinase 1) is a member of the 40S ribosomal subunit involved in translational repression and the initiation of ribosome quality control via the regulatory ubiquitylation of 40S ribosomal proteins [26]. SERBP1 (SERPINE1/PAI1 mRNA-binding protein 1, also known as PAI-RBP1), a member of the serine protease inhibitor, was identified as a partner of RACK1 [27]. SERBP1 was overexpressed in various cancers including breast cancer, ovarian carcinoma, glioblastoma, and also lung SqCC, and this might be associated with tumorigenicity and resistance to anticancer drugs [28]. RPS6 (40S ribosomal protein S6), a substrate for p70S6 kinase (p70S6K), is known to play important roles in tumorigenesis and development. The highly expressed RPS6 and its phosphorylated form have been observed in various cancers including NSCLC, where upstream of Akt2/mTOR/p70S6K signaling pathway was aberrantly regulated [29]. The dephosphorylation of RPS6 can inhibit the mTOR pathway, resulting in the inhibition of tumor growth and metastasis [30]. Eukaryotic translation initiation factors, EIF4A1 and EIF4A2, are the RNA helicases belonging to the EIF4F initiation complex that unwind mRNA during translation [31]. EIF4A1was upregulated in various cancers while the downregulation of EIF4A2 in NSCLC was associated with a poor prognosis [32].
Multivariate correlation analysis of semiquantitative key protein expressions
Representative proteins expressed throughout all the 30 modules were subjected to multivariate correlation analysis (MVA). As a result, the spectral count-based semiquantitative expression of 89 key proteins including eigen proteins and/or hub proteins was clustered into several groups (a, b, c, d, and e; Fig. 4). The clusters c, d, and e were characteristic of the SqCC trait whereas the clusters a and b were characteristic of the PPA trait. Of these, cluster e included hub proteins of the WM26 module and eigen proteins of the SqCC characteristic modules. Cluster d included most of the hub proteins in WM28.
Upstream regulators, canonical pathways, and downstream regulator effects predicted by IPA
Causal network and upstream regulator analysis together with downstream annotation was performed for the WGCNA modules significant to SqCC, using the Ingenuity Pathway Analysis (IPA) (http://www.ingenuity.com) software [33]. Table 2 briefly summarizes the top master regulators, diseases or functions, and canonical pathways predicted for the three SqCC-characteristic WGCNA modules. Causal and upstream regulators were predicted to be activated or inhibited (|z-value|> 2.0) and upregulated (1.5 < z-value < 2.0) with the significance of network bias-corrected p-value < 0.005. The top master regulators with high values in activation or inhibition score (z-score) in causal networks, which were significantly associated with SqCC (WM26, WM27, and WM28) and PPA (WM15, WM16, and WM18), are presented in Tables S1 and S2, together with their participating regulators and target molecules in the datasets.
Highly activated master regulators predicted for the WM26 module were MNK1/2, ROCK2, EFNA4, EFNA3, EFNA5, EFNA2, DSP, EFNA1, and SFN, while KMT2D and MXD1 were highly inhibited. MNK1/2 encodes mitogen-activated protein kinases interacting protein kinases 1 and 2 (Mnk1 and Mnk2) which are known to play important roles in controlling signals involved in mRNA translation. ROCK2 encodes the Rho-associated coiled-coil-containing protein kinase 2 (also known as Rho-kinase 2 or ROCK-II), belonging to mammalian serine/threonine kinases and downstream effectors of the small GTPase RhoA, which are key regulators of keratinocyte adhesion and terminal differentiation [34]. ROCK2 is an oncoprotein that acts as a prognostic marker in various solid tumors [35]. EFNA4, EFNA3, EFNA5, EFNA2, and EFNA1 are members of the Eph (erythropoietin-producing hepatoma or Ephrin) receptors of the A-type, which are the most important family of receptor tyrosine kinases involved in signaling pathways of embryogenesis and tissue patterning. Eph signaling regulates cell morphology and migration by modifying cell adhesion and organizing actin cytoskeleton and then affects cell proliferation and differentiation [36]. Eph receptors are expressed in cancer cells and the tumor microenvironment involved in tumorigenesis and metastasis [37]. However, Eph receptors can act both as tumor promoters and suppressors, depending on the cancer type [37, 38]. Regarding lung cancer, the upregulation or overexpression of EFNA1, EFNA2, EFNA4, EFNA5, and EFNA7 are indicative of tumor-promoting roles in lung cancer [39]. KMT2D encodes histone-lysine N-methyltransferase 2D, formerly named MLL2 (myeloid/lymphoid or mixed-lineage leukemia 2), which methylates “Lys-4” of histone H3 (H3K4me), inducing epigenetic transcriptional activation. This epigenetic regulator KMT2D is the most frequently mutated in all cancers, and its mutations are notably associated with keratinocyte cancers [40]. Lin-Shiao et al recently reported that KMT2D interacts with the transcription factor TP63 on chromatin and regulates TP63 target enhancers to coordinate epithelial homeostasis and GO analysis for genes upregulated in shKMT2D-treated keratinocytes showed epithelial cornification, keratinization, differentiation, and development as the most enriched category [41]. Tumor protein 63 (TP63), the master regulatory transcription factor of epithelial tissues, was predicted to be upregulated (z-value = 1.897). TP63 regulates most of the same target genes involved in vitamin D and retinoid signaling that are regulated by KMT2D [41]. Vitamin D receptor regulates the c-MYC/MXD1 network, in which the transcriptional repressor MXD1 is the antagonist, to suppress c-MYC function, preventing epidermal tumor formation [42]. Interestingly, the MXD1 causal network inhibited in this study suggested the activation of c-MYC. SFN was activated, encoding SFN, which is a cell cycle checkpoint protein and is present mainly in tissues enriched in the stratified squamous keratinizing epithelium. SFN binds to translation and initiation factors, and especially stimulates tumor initiation and the progression of early-stage lung adenocarcinoma [43]. DSP (desmoplakin) encodes the major high-molecular-weight protein of desmosomes. Desmosomal genes are expressed differently between lung adenocarcinoma and lung SqCC although the mechanism regulating their expression remains unknown. The protein networks of the WM26 module demonstrated the involvement of upregulated desmosomal proteins including DSP, DSC2 (desmocollin 2), DSC3 (desmocollin 2), JUP (junction plakoglobin), PKP1 (plakophilin 1), and PKP3 (plakophilin 3). Martin-Padron et al. reported that PKP1 was overexpressed and increased cell proliferation and cell survival in lung SqCC, and found that PKP1 enhances MYC translation together with the translation initiation complex by binding to the MYC mRNA [44]. Kudo et al. demonstrated via immunohistochemical staining that the expression of DSC3, SFN, DSP,and JUP among cell adhesion and growth inhibitor genes was highly increased in TP53-mutated tumors and that TP53-mutated tumors exhibited high nuclear staining of the TP53 protein only in tumor cells at the tumor margins adjacent to the stroma but not in the tumor interior; thus, exhibiting tumor cell heterogeneity in the expression of mutated TP53 protein between the tumor interior and margins [45]. Keratinization of the epidermis and cell proliferation of SqCC cell lines were significantly annotated to the WM26 network module (Table 2), with key regulators being HIF1A (hypoxia-inducible factor 1 subunit alpha) and IGF1 (insulin-like growth factor 1).
The WM27 and WM28 modules were found to share the same master regulators that representatively included highly activated MLXIPL and MYCN,and highly suppressed LARP1 and RICTOR (rapamycin-insensitive companion of mTOR), which were remarkable in the WM28 module: LARP1, overlap p-value = 3.59 ´ 10-32 and z-value = -3.873; MLXIPL, overlap p-value = 8.54 ´ 10-30 and z-value = 3.873; MYCN, overlap p-value = 3.27 ´ 10-22 and z-value = 3.742; RICTOR, overlap p-value = 4.10 ´ 10-12 and z-value = -3.0. Highly activated ZEB and highly inhibited Mir200 were characteristic of the WM27 module. ZEB encodes the zinc finger E-box-binding homeobox proteins ZEB1 and ZEB2 which regulate the epithelial-mesenchymal transition pathway as both transcriptional activators and repressors. The miRNA-200 (miR-200) family can repress ZEB proteins to regulate epithelial differentiation [46]. Activated ZEB and inhibited Mir200 suggested progressive tumorigenesis acquiring a mesenchymal phenotype in lung SqCC. The high expression of ZEB1 is associated with tumor grade in NSCLC or distant metastasis in lung SqCC [47].
LARP1 (La-related protein 1) is an RNA binding protein and mTORC1 effector involved with terminal oligopyrimidine (TOP) mRNA translation. Surprisingly, LARP1 was found to be highly suppressed in this study. Significant upregulated LARP1 was frequently reported to correlate with adverse prognosis in several cancers including NSCLC [48]. In contrast, clear cell renal cell carcinoma progression was promoted by decreased LARP1 derived from the downregulation of the long noncoding RNA ASB16-AS1 by inhibiting miR-185-5p and miR-214-3p [49]. Our result should be understood along contexts underpinning disease mechanisms and because the function of LARP1 is highly controversial [50]. The highly activated MYC, MLXIPL, and MYCN and highly suppressed LARP1 and RICTOR were predicted for the WM28 module. MLXIPL encodes a carbohydrate-responsive element-binding protein, which is a basic helix-loop-helix leucine zipper (bHLH-LZ) transcription factor of the MYC/MAX/MAD superfamily, promotes aerobic glycolysis through inhibition of TP53, resulting in tumor cell proliferation [51]. MYCN, a member of the MYC family of oncogenes, also encodes a bHLH-LZ protein MYCN, and its deregulation was reported in various cancer types and was often associated with a poor prognosis. MYCN-amplified cancer cells exhibit the enhanced expression of genes and proteins involved in aerobic glycolysis (referred to as the Warburg effect), oxidative phosphorylation, and the detoxification of reactive oxygen species (ROS) [52]. Numerous chemical drug interventions including sirolimus (rapamycin) were inhibited (Table 2), suggesting an involvement of potential therapeutic targets in those data-driven networks. The translation initiation of protein, nonsense-mediated mRNA decay, and negatively regulated cell death of tumor and cancer cells were annotated to the WM28 module (Table 2). Interestingly, the EIF2 signaling pathway exhibited the highest significance in overlapping p-values among the top canonical pathways predicted for both the WM27 and WM28 network modules, and it was remarkably activated especially for WM28 (overlap p-value = 2.77 ´ 10-35 and z-value = 2.828). The upregulation of eIF2a and eIF2b, which are members of translation initiation factors (eIFs), has been reported in several cancer types including lung cancer, and is often associated with poor prognosis in patients [53]. Bilguun et al. found that STXBP4 (syntaxin binding protein 4) which targets TP63 was crucially associated with lesion growth in lung SqCC patients, in which the eIF2 signaling pathway was the most significantly activated [54].
The top canonical pathways predicted for the three SqCC characteristic modules are also listed in Table 2. The WM26 module was most significantly annotated to glucocorticoid receptor signaling and inhibited HIPPO signaling, and the WM27 and WM28 were significant to eIF2 signaling, mTOR signaling, and the regulation of eIF4 and p70S6K signaling.
Genomic alteration landscape of early-stage SqCCs based on the TCGA database
Kim et al. performed a comparative genomic analysis between Korean and North American lung SqCC samples and demonstrated a spectrum of genomic alterations similar to the two ethnically different cohorts, which contrasts with the differences noted in lung adenocarcinoma [55]. The TCGA lung SqCC sub-datasets (T1A-T2A: n = 184) were selected to match our early tumor stage patients’ group, and their genomic alteration profiles were visualized using the cBioPortal Pan-Lung cancer (TCGA, Cell 2018) (https://www.cbioportal.org/) (Fig. S4). Frequently altered driver mutation candidates were as follows: TP53 (80%) and CDKN2A (40%) in cell cycle; PIK3CA (43%), PTEN (19%), and FGFR1 (17%) in mitogenesis and RAS signaling; SOX2 (39%) and TP63 (31%) in squamous cell differentiation; KMT2D (23%) and FAT1 (20%) in transcription and gene expression; and SYNE1 (29%) and NFE2L2 (21%) in cell survival. Characteristics regarding mRNA-level expressions types of genomic alterations represent PIK3CA, frequent missense (driver) and high amplification; PTEN, frequent truncating (driver) and missense (driver); SOX2 and TP63, high amplification; KMT2D, highly frequent truncating (driver) and splice (driver); and NEF2L2, frequent missense (driver) (Fig. S5). Shang et al. reported a mutational landscape of Chinese lung SqCC patients, in which mutation frequencies of PIK3CA, NFE2L2, and KMT2D were most significant [56], which was quite similar to our TCGA-based study, although most of their cohorts were at advanced tumor stages. In our TCGA-based genomic alterations, PIK3CA, SOX2, and TP63 co-occurred with high significances of q-value < 0.001, and NOTCH2 and DDR2 co-occurred with q = 0.002. No driver mutation candidates were mutually exclusive. TP63 was co-expressed with PKP1, KRT6A, DSG3, KRT6C, NFE2L2, KRT6B, and SOX2 at the mRNA level (Spearman’s correlation > 0.6 and q-values < 5.0 ´ 10-18), which was found to be consistent with co-expression networks of the WM26 module (Fig. 2A). Our IPA analysis of the WM26 module annotated the proliferation of SqCC, in which the key regulator HIF1A was important in the adaptive response to hypoxia and angiogenesis. The ROS-responsive transcription factor NRF2 encoded by NEF2L2 can bind and transactivate an antioxidant response element upstream of HIF1A, through which the expression of HIF1α encoded by HIF1A is regulated directly by NRF2 [57]. Moreover, the accumulation of HIF1α directly upregulates SLC2A1 (known as GLUT1, glucose transporter-1), as we indeed observed in the co-expression networks of WM26, most likely suggesting hypoxia-induced metabolic changes to aerobic glycolysis (the Warburg effect).
Overview, limitations, and conclusion
We identified protein co-expression networks significantly associated with lung SqCC by WGCNA following MS-based proteomic analysis. Multivariate correlation analysis for semiquantitative expressions of key proteins exhibited protein clusters characteristic of the SqCC trait, which were well differentiated from those characteristics of the PPA trait. Strikingly, pathways enriched for the WM26 module predominantly involved keratinization. The predicted causal networks were also annotated to cell morphology and keratinization of the epidermis, in which key master regulators were highly activated ROCK2 and Ephrs (EFNA1-5), while an epigenetic regulatorand lung tumor suppressor KMT2D were highly inhibited. Downstream regulator effects were annotated to the cell proliferation of SqCC, and a key regulator, HIF1A, was involved in hypoxia-induced metabolic changes to aerobic glycolysis. Correspondingly, upregulated TP63 was predicted for the WM26 module (Table 2 and Fig. 5), and genomic alterations in the early-stage TCGA SqCC database exhibited highly amplified and frequently truncated driver mutationsof KMT2D (Fig. S5).
Pathways enriched and master and upstream regulators of causal networks predicted for both the WM27 and WM28 modules were annotated to translation initiation and nonsense-mediated mRNA decay. The co-expression networks of the WM27 module were characteristic of mesenchymal transformation by activated ZEB, and of upregulated hub ribosomal proteins including the key ribosomal protein RPS25 of IRES-dependent translation. Hong et al. revealed that non-phosphorylated LARP1 interacts with ribosomal protein mRNAs and inhibits their translation while LARP1 phosphorylated by mTORC1 and Akt/S6K1 allows ribosomal protein mRNA translation [58]. However, this switching mechanism of translation of ribosomal protein mRNAs depending on phosphorylation/non-phosphorylation of LARP1 does not seem to interpret our results of both the highly suppressed LARP1 and upregulated expressions of the hub ribosomal proteins including RACK2, RPS6, and RPS25, which were predicted to be indirectly targeted by LARP1. Suppression of the cell death of tumor and cancer cells was centrally annotated as downstream regulator effects to the WM28 module, in which key master regulators were highly suppressed LARP1 and highly activated MYC and MLXIPL (Table 2 and Fig.6). Even more, interestingly, the eIF2 signaling was annotated commonly to both two modules with the highest significance. Collectively, all the results we obtained in this study might allow a possible scenario as follows. Physiological stressors such as hypoxia and/or ROS downregulate mTOR activity which represses cap-dependent translation and reduces overall protein synthesis, in which the eIF2 ternary complex mostly plays an important role [59]. The inhibition of protein synthesis leads to the activation of alternative cap-independent translation of mRNA subsets which utilizes IRESs located in the 5’ untranslated region of mRNA [60]. These mRNAs encode oncogenic proteins such as HIF1a, MYC, c-MYC, VGFA, and BCL-2, which promote the progression of tumorigenesis, angiogenesis, and cancer cell survival. Thus, the eIF2 complex assembly importantly functions in switching from cap-dependent to cap-independent translation [60, 61]. This might explain why LARP1, the master regulator in the cap-dependent TOP mRNA translation, was highly suppressed and why this loss of LARP1 caused the reduction of mTOR activity and downstream mTORC1 signaling including RICTOR.
The limitation of this study was the number of patients examined. We plan to validate the results obtained from this study using a larger sample size of the external cohort in the future.
In conclusion, we successfully applied WGCNA to clinical proteomic datasets. Our results could identify data-driven network profiles and their upstream regulators characterizing cancerous cells microdissected from FFPE tissues of lung SqCCs, exhibiting the progression of SqCC concomitantly with aberrant keratinization, epithelial-mesenchymal transformation, aerobic glycolysis, and negatively regulated cell death. Collectively, the results obtained in this study suggest an underlying disease mechanism of lung SqCC progression caused largely by switching to the cap-independent, IRES-dependent translation of mRNA subsets encoding oncogenic proteins. We plan to conduct a larger-sample cohort study, including genomic alteration analysis, to investigate data-driven proteogenomic networks, which will further provide clinically important information on the proteogenomic landscape of lung SqCCs.