We calculated a risk score for each patient based on this model and evaluated the performance of the prognostic to Kaplan-Meier survival analysis and time-dependent ROC curve analysis, which demonstrated significant prognostic performance in training, and validation as well as on the entire dataset. Recently, various proteins-based prognostic models in malignancies have been successfully validated protein prognostic model in malignancies, including breast cancer[14], ovarian cancer [15], stomach adenocarcinoma [16], bladder urothelial carcinoma[17], head and neck squamous cell carcinoma[18]. The four-protein prognostic signature developed in our study comprised three proteins (Histone-H3, HSP27_pS82, and CHK2) that were protective and one protein (PAXILLIN) that was related to bleak prognosis, which are not captured by genomics or transcriptomics. We conducted a multivariate Cox proportional hazards regression analysis to determine whether the 4 protein prognostic signature was independent of other biomarkers like age, AJCC Stage. A previous study reported a 6-protein (CHK1_pS345, CHK2, IRS1, PAXILLIN, BRCA2, and BRAF_pS445) signature that can predictor of survival for patients with LUSC (AUC = 0.699) [19]. CHK2 and PAXILLIN are identical to our model. In this study, we found that the protective protein was significantly upregulated in low-risk group compared to the high-risk group and deleterious proteins were upregulated in high-risk group compared to the low-risk group. The risky proteins might promote various malignant properties like proliferation, migration and invasion whereas the protective protein might act as inhibitory switch to various oncogenic pathways. As part of our ongoing research efforts toward improving biomarker development in cancer patients, we employed Alphafold2 for 3D structure prediction [20].
Cancer initiation and progression are often accompanied by signaling pathway changes as well as alterations in multiple protein PTMs [21]. HSP27 (Small heat shock protein, HSPB1) is a 205 amino acid-containing member of the small heat shock protein family and have an impact on the Protein Quality Control (PQC) machinery, phosphorylated in response to diverse stress stimuli, including reactive oxygen species, inflammatory cytokines and osmotic stress [22]. HSP27 is expressed at high levels in some cancer tissues including LUSC [23], liver cancer [24], and ovarian cancer [25]. The phosphorylation level of HSP27_pS82 are typically expressed at low levels under normal physiological conditions but are dramatically upregulated in response to cellular stress and cancer. Highly phosphorylated, HSP27_pS82 increase its protective effect to rescuing capacity of improperly folded/partially denatured proteins and prevent mitochondrial release of cytochrome c and prevent formation of the apoptosome.
Checkpoint kinase 2 (CHK2) is a serine/threonine kinase tumor-suppressor that plays critical roles in aspects of DNA damage response, DNA repair, and apoptosis [26]. In response to DNA damage, CHK2 is initially phosphorylated at Thr-68, which leads to its full activation. Our analysis found that CHK2_pT68 has a higher positive correlation with CHK2 only in the low-risk group, not in the high-risk group. CPTAC data revealed that CHK2 was overexpressed in tumor tissues at mRNA and protein level. The IHC images in the Human Protein Atlas also demonstrate that CHK2 was overexpressed in tumor tissues.
Histone-H3 (H3C1) is a core component of nucleosome encoded by the tumor suppressor gene H3C1, which plays a central role in transcription regulation, DNA replication, DNA repair, and chromosomal stability [27]. Featuring a main globular domain and a long N-terminal tail which play a pivotal role for nucleosome integrity. Since up-regulation of Histone-H3 re-activates tumor suppressor genes through chromatin acetylation and modification of transcription factors may evoke several intracellular mechanisms relevant for cellular differentiation, alterations of cell signaling and cell cycle, and apoptosis induction [28]. We presume that maintaining a consistently high level of histone H3 acetylation would be benefit to patients.
PAXILLIN is a main dynamic protein complexes of Focal adhesion (FAs) [29] and a unique focal adhesion protein with its ability to interact with certain oncogenes [30]. PAXILLIN overexpression may promote tumor malignancy, including breast carcinoma [31], colorectal cancer[32, 33], and non–small cell lung cancer [34]. Lung cancer cell lines indicated that the absence or low level of PAXILLIN protein expressed on certain lung cancers, such as LUSC (but LUAD) [35]. However, our research has shown that PAXILLIN mRNA and protein level is downregulation in LUSCs as well as poorer patient survival rates. Because paxillin is involved in cell motility. It was postulated that low levels of paxillin possibly involved in cell motility and promoted more invasive tumor behavior.
We identified enrichment of oncogenic pathways like focal adhesion and ECM-receptor interaction pathways played an important role in tumor proliferation, adhesion, tumor invasion, and metastasis in multiple cancers [36, 37]. Originally, PCA analysis show PAXILLIN has the major contribution for high-risk group. Based on this observation, it seems that the PAXILLIN is a major factor in determining focal adhesion and ECM-receptor interaction pathways within tumor cell. PI3K-Akt signaling pathway was identified to be significantly activated in non-small-cell lung cancer [38, 39]. GSEA results show low-risk group enriched cytochrome P450 metabolism in cancer cell that means active this way could have effective effects.
As a side note, there remain some limitations in our current study. Firstly, the prognostic predictive value of 4 proteins markers is only based on a single cohort with a relatively small sample size and there was no external validation, which could bring in some bias. Since this study utilized the RPPA data, it is limited to hundreds of pre-selected proteins that does not provide a system-wide view of the proteome compared to other high-throughput technologies.
Based on online datasets, we identified a new and reliable prognostic proteomics signature and a variety of features that were significantly correlated with our model protein's expression. Nevertheless, further studies are required to properly explore the relevance of underlying LUSC development, as well as how these mechanisms may be targeted for therapeutic intervention.