Liver cancer is an extraordinarily heterogeneous malignant tumor, which is one of main obstacles to the implementation of precision medicine [26]. Although the histopathological classification of liver cancer has been modified and refined, it still cannot solve the problem that predict patient prognosis or response to therapy accurately [27]. With the development of high throughput sequencing technology, more and more genes with oncogenic functions or tumor suppressive functions were identified. The expression patterns of these genes are recurrently altered in HCC, which give tumor cells corresponding abilities to influence malignant behaviors, such as recurrence, metastasis and drug resistance, leading to different clinical outcomes [28]. One of most important objectives of precision medicine is to provide each cancer patient with most accurate and effective treatment based on the genomic characteristics of the cancer [29]. Without doubt, accurate prognostic biomarkers need to be identified for executing precision medicine in a personalized manner. The heterogeneity of tumor exists not only in the genotypes but also in the tumor microenvironment. The accumulation of genomic alterations may result in changes of microenvironment of liver cancer [27]. Immune cells and tissue hypoxia, the elements of tumor microenvironment, have been verified to play an important role in the development of HCC and treatment response [30]. In this study, therefore, we focus on the effect of tumor immunity and hypoxia on the prognosis of HCC, which is convincingly beneficial to understanding the mechanism of tumorigenesis and tumor development, and the improvement of personalized therapeutic approaches. Using the RNA-seq transcriptome data and corresponding clinical data of HCC obtained from the TCGA dataset, we constructed the 8-hypoxia-related immune gene prognostic signature. The 8-gene signature exhibited a good performance to predict OS of HCC in the training set and expressed robust when validated in another cohort.
Our prognostic signature contained eight genes (S100A10, MAPT, CACYBP, BIRC5, KITLG, SPP1, STC2, GHR). These genes were correlated with OS of HCC separately and showed a better performance when combined, for one reason that gene signature analysis can reflect the complex interaction of various genes affecting hypoxia-related tumor immunity in cancer. S100A10, a member of S100 family, mediates the process of transforming plasminogen to an active protease and has been reported to link to the HCC tumorigenesis and progression [31]. As a binding partner of S100 family proteins, CACYBP was highly expressed in HCC, which was significantly associated with elevated AFP level, increased dead and recurrence events, and reduced OS [32]. BIRC5, also known as survivin, is the most potent member of the inhibitor of apoptosis protein (IAP) family and a vital promoter of development and progression of HCC. The overexpression of survivin affects the prognosis of patients with HCC by promoting cell proliferation, inhibiting cell apoptosis and inducing tumor stromal angiogenesis [33]. The osteopontin protein is encoded by the SPP1 gene. Osteopontin overexpression increases proliferation, stem-like properties, glycolysis, and resistance to chemotherapy of HCC cells, which correlates with poor prognosis of HCC patients [34]. Importantly, osteopontin can associate with HIF-1α to promote tumorigenesis [34]. In addition, it plays a critical role in the formation of immunosuppressive microenvironment of HCC [35]. Likewise, STC2 is a proliferation-facilitating gene and correlates with occurrence, development, and prognosis of HCC [36]. Summarily, these genes above have been verified to be associated with the poor prognosis of HCC, which was consistent with the result of our study. Moreover, in this model, GHR was positively correlated with OS of the patients with HCC. One study demonstrated that HCC patients with a significant down-regulation of GHR expression showed higher incidence of recurrence and poor survival rates [37], which also supported our results. Among these genes, MAPT had the highest coefficient and was considered to be the main contributing gene. Many researches have demonstrated that MAPT was correlated with the prognosis of many cancers. The controversial roles of MAPT in both oncogenesis and tumor suppression were reported in these cancers, depending on the cell type and context. In addition, MAPT is also involved in resistance to paclitaxel, platinum, and bicalutamide, which contributed to the poor prognosis [38, 39]. Several lines of evidence have also demonstrated the effect of KITLG in tumorigenesis [40]. However, the biological functions of MAPT and KITLG in HCC have not been elucidated and need further research. Thus it can be seen that the gene signature was closely related to the malignant behaviors of HCC and could be an accurate predictor.
As previously mentioned, multiple pathways were significantly enriched in the high-risk group, including ECM-receptor interaction pathway. The signature contained two ECM-related genes, S100A10 and SPP1 [35, 41]. As one of the critical components in the tumor microenvironment, dysregulated ECM has been involved in the development and progression of HCC [42]. Hypoxia is one of the key inducers of the dysregulated ECM, by affecting ECM deposition, remodeling and degradation [43]. As shown in our analyses, HCC samples in the high-risk group had higher proportions of macrophages. Interestingly, Hypoxia can recruitment macrophages and induce the alteration of macrophages toward protumor phenotype in the tumor microenvironment, which contribute to ECM remodeling in favor of tumorigenesis and cancer progression [43, 44]. At the same time, ECM stiffness through induction of hypoxia is also a barrier for delivery of chemotherapeutic drugs to the tumor site, which weakens the efficacy of antitumor drugs [44]. The high resistance to antitumor drugs is one of the main contributors to the poor outcome of patients with advanced HCC who have no opportunity for surgical resection. The eight genes included in the signature are all linked to the drug resistance. Additionally, in the study, the patients in the high-risk group expressed the lower sensitivity to vinblastine. Apart from the mechanism mentioned above, hypoxia also contributes to the drug resistance by regulating the glycolytic metabolism. As a master regulator of glycolytic metabolism, HIF-1α activation can enhance the glycolytic metabolism at the transcriptional level, which influences the drug sensitivity in HCC [45]. The glycometabolism-related pathways were enriched in the high-risk group in the study. And glycolytic metabolism also takes roles for the regulation of proliferation, immune evasion, invasion, metastasis, angiogenesis in the liver cancer [46]. More importantly, metabolites released from cancer cells have an impact on the function of immune system cells, such as tumor-associated macrophages, NK cells, dendritic cells (DCs), regulatory T cells (Tregs) and cytotoxic T lymphocytes, leading to inhibition of the immune response [47, 48]. Notably, in addition to predicting the OS, the present results suggest that the eight-gene signature has a good performance in predicting the response of HCC patients to antitumor drugs. Therefore, the identified signature is a practical tool to help making clinical decision.
In addition to the above-mentioned macrophages, high-risk patients had significantly higher proportions of Tregs. Tregs belong in one category of tumour-infiltrating lymphocytes and play an important role in immune suppression [49]. Hypoxia has been identified in promoting the recruitment of Tregs to the tumour microenvironment by inducing the expression of CCL28, leading to promote tumour tolerance and angiogenesis [50]. In addition, these cells can suppress the functions of T cells, NK cells and DC [49]. As shown in the results, the score of type_II_IFN_Reponse was significantly lower in the high-risk group. T cell stimulation under hypoxia conditions inhibits the IFN-γ production [51]. In general, Hypoxia participates in accentuating the immunosuppressive characteristics of immune cells in the tumor environment of HCC, which may have important clinical significance for cancer immunotherapy and predicting the prognosis.
To our knowledge, this was the first study constructing a prognostic signature based on the hypoxia-related immune genes in HCC. Compared with the models based on the genes with single property, the signature contained higher information context. Additionally, the robustness and high predictive value strictly guaranteed the applicability of the signature. However, there were still some limitations in this study. We did the research by utilizing published retrospective data sets. Therefore, our findings should be validated in prospective studies. Furthermore, some important immune genes with prognostic potential were excludes in this research only based on hypoxia-related immune genes.