Given the wide variation in prognostic outcomes of NSCLC, it is of great importance to build a robust classifier to segment patients with different risks and prognosis, which is essential to maximize the benefits of accurate assessment, individual therapy and timely long-term follow-up. Numerous data indicate that hypoxia and TME are involved in processes that promote the development of tumor cells and a more malignant phenotype[4, 13]. Consistent with criteria prognostic parameters like nodal status, tumor stage and tumor grade, hypoxia and immune cells infiltrating TME have been proposed as prognostic factors for patient outcome[14, 15]. Although methods such as PET imaging and immunohistochemical marker expression assays have been initially implemented to detect the degree of hypoxia in patients' tumors, the specific identification methods are still fraught with many unknowns. In this study, the integrated mining of transcriptional profiles and microenvironment features was intended to establish a tool to solve this vital clinical issue.
In this study, we profiled the mRNA expression of 205 hypoxia-associated genes in three GEO databases. Two NSCLC subgroups Cluster 1/2 were identified by consistent clustering on the basis of hypoxia-related genes. The Cluster 1/2 subgroup not only affects disease prognosis and key signaling pathways, but it influences immune cells infiltration. Bioinformatic approach uncovered that hypoxia condition mainly associated with the PI3K/Akt/MTOR signaling pathway, the JAK2/STAT3 signaling pathway, the MYC targets V1 signaling pathway, and angiogenesis. Interestingly, a number of studies have found that hypoxia plays a role in the induction of tumor cell growth and metastasis by PI3K/Akt signaling and MYC target V1 signaling [16–18]. There is also much evidence to support the interaction between hypoxia and angiogenesis[19–21]. Moreover, we used Cox and LASSO regression to develop an 11-gene prognostic signature for TCGA cohort. The risk scores were derived by integrating mRNA expression levels and risk coefficients for 11 hypoxia-related genes, and significantly classified prognosis in NSCLC patients into low and high-risk clusters. There are a number of clinically available multigene based risk models that can predict the prognosis of NSCLC. Zhu and colleagues recently reported a 22-autophagy-related signature based on overall survival in patients with lung adenocarcinoma[22]. Recently, Wu et al. reported a seven long non-coding RNAs prognostic model to predict OS in NSCLC patients. So far, there were limited prognostic models based on hypoxia-associated genes for patients with NSCLC. In an effort to exploit hypoxia-induced epithelial-mesenchymal transition gene signatures associated with clinical outcomes in NSCLC, Gao et al. demonstrated a 17 gene prognostic panel for NSCLC[23]. Our model shows good accuracy and stability in clinical outcome prediction either for 1-year (AUC = 0.738), 3-year (AUC = 0.702) and 5-year (AUC = 0.666) OS of NSCLC patients. Our data also revealed that all 11 genes (FBP1, NDST2, ADM, LDHA, DDIT4, EXT1, BCAN, IGFBP1, PDGFB, AKAP12, CDKN3) in our model were associated with poor prognosis in NSCLC patients. These discovered hypoxia-associated genes have been repeatedly reported in cancer[24–27]. Previous studies reported that hypoxia-induced GBE1 expression promotes tumor progression by repressing FBP1 activities in lung adenocarcinoma[28]. In contrast, in the present study, we found a positive coefficient for FBP1, indicating that FBP1 was considered as risk gene in NSCLC. The underlying molecular mechanisms deserve further investigations.
Accumulating evidence indicates that hypoxia is an important feature of TME and that it can drive progression and metastasis by facilitating suppressive cells (TANs, Tregs, and TAMs), and producing immunosuppressive molecules. CIBERSORT revealed that patients in Cluster 2 with high hypoxia risk had significantly more abundant infiltrative TANs and M2 macrophages phenotype. Furthermore, TANs were associated with a poor prognosis, suggesting that our hypoxia model may predict the immune microenvironment. Hypoxia not only tightly regulates the production of specific chemokines, it also controls their action by regulating their receptors[29, 30]. Chemokines play an important role in regulating tumor immunity[31]. Our findings therefore provide insight into the underlying mechanisms of recruitment-related chemokines in TANs infiltration, and found that the production of CXCL6 increased in NSCLC cells under hypoxic condition, and the blockade of CXCL6 almost halted TANs migration, suggesting that the recruitment of TANs is mediated mainly by CXCL6. CXCL6 induced proliferation and metastasis of lung cancer cell lines was confirmed in other studies, and the role of CXCL6 in the TME of NSCLC is unclear[32, 33]. Our study broadened the understanding of CXCL6 in NSCLC progression. As previously mentioned, multiple studies have noted that TANs appear to mostly develop a pro-tumorigenic phenotype[34, 35]. In this study, we revealed TANs enhanced the proliferative and invasive ability of NSCLC cells in vitro. Experimental studies are needed to further clarify the molecular mechanisms underlying the TANs related pro-tumorigenic phenotype in NSCLC.