Cervical cancer is characterized as having a high incidence and mortality rate, particularly in low and middle economic populations of patients. Therefore, it is urgent to determine new prognostic indicators in order to more accurately predict the prognosis of patients with cervical cancer. While there have been many studies directed toward examining the relationship between hypoxia and tumor formation, the relationship between hypoxia and prognosis of patients with cervical cancer remain quite limited. In this study, the hypoxia-related gene signatures was used to predict the prognosis of patients with cervical cancer and demonstrate that this approach achieves good prediction results.
Specifically, we developed a 5-gene signature prognostic risk model and provide a verification of its validity as demonstrated in both train and test data sets. Based on the results of survival analysis, ROC and nomogram, we believe that this constructed risk factor model is quite robust. In addition, in each clinical subtype of TCGA, this risk factor model effectively predicted the risk of cervical cancer. For the first time, we have established a prognostic prediction model based on hypoxia genes. Moreover, the results of qPCR and immunohistochemistry staining assay revealed that the expression of
AK4, HK2, P4HA1, TGFBI and VEGFA is high in cervical cancer. The functional study shown that expression of
AK4, HK2, P4HA1, TGFBI and
VEGFA can regulate the proliferation, migration, and invasion ability of cervical cancer cells. Among them,
AK4, P4HA1 and
TGFBI were first confirmed as oncogene in cervical cancer. In our study, taken together, we believe that this model can be used to evaluate the prognostic risk in cervical cancer patients. And our model is better than other cervical cancer models [
23,
24].
In our analysis, the high expression and high risk of these five hypoxia-related genes,
AK4, HK2, P4HA1, TGFBI and
VEGFA, were identified as risk factors. Adenylate kinase 4 (
AK4) is a member of the adenosine kinase family and has been found to play an important role in malignant tumors and anti-tumor therapy. Results from recent studies have shown that high expressions of
AK4 promote lung cancer metastasis by enhancing HIF-1α stability and EMT under conditions of hypoxia [
25]. High expressions of
AK4 also promote cell proliferation and invasion in ovarian cancer and
HER2 positive breast and esophageal cancers [
26‐
28]. Hexokinase 2 (
HK2) is a rate-limiting enzyme in the glycolysis pathway. In addition to the catalytic activity of this enzyme,
HK2 can also antagonize apoptosis within the mitochondrial pathway, which plays an important role in the invasion and metastasis of malignant tumors. It has been reported that B7-H3 promotes aerobic glycolysis and increases chemoresistance in colorectal cancer cells through the upregulation of
HK2 [
29]. Silencing
HK2 in human hepatocellular carcinoma cells inhibits tumorigenesis and increases cell death, and
HK2 silencing can synergistically inhibit tumor growth with sorafenib [
30]. In colorectal cancer,
PLK3 inhibits glucose metabolism by targeting HSP90/STAT3/HK2 signal transduction. Under conditions of
PLK3 overexpression, expression levels of
HK2 decrease, while silencing
PLK3 increases the expression of
HK2 in tumor cells.
HK2 silencing can inhibit the growth of colorectal cancer cells [
31]. Proline 4-hydroxylase subunit α-1 (
P4HA1) is associated with a variety of malignant tumor development pathways, such as EMT, angiogenesis, invasion, inflammation, tumor metabolism and glycolysis pathways [
32]. Findings from recent studies have shown that
P4HA1 is up-regulated in lung, breast and head/neck cancer tissues, and high expression levels of
P4HA1 are significantly correlated with the clinical characteristics of these cancers. The clinical prognosis of patients with high expressions of
P4HA1 is poor [
33]. In melanoma, depletion of
P4HA1 reduces cell adhesion, invasion and in vitro survival, and in xenotransplantation models, knockdown of
P4HA1 reduces the invasion of melanoma in vivo and the deposition of collagen in interstitial ECM and tumor vascular basement membranes. Such results indicate that
P4HA1 can serve as a potential biomarker for poor prognosis of primary melanoma [
34]. Upregulation of
P4HA1 promotes cell migration and invasion in glioblastoma and head/neck squamous cell carcinoma and, in this way, provides a biomarker for poor prognosis in patients with high expressions of
P4HA1 [
35,
36]. As an extracellular matrix protein,
TGFBI is closely related to the development of different malignant tumors. In glioma, the median survival time for patients with a high expression of
TGFBI was significantly shorter than that of patients showing low expressions of
TGFBI [
37]. An up-regulation of
TGFBI can promote the occurrence and metastasis of breast cancer, increase tumor angiogenesis and increase hypoxia, and
TGFBI overexpression promotes oral squamous cell carcinoma. Knockout of
TGFBI inhibits the proliferation and metastasis of oral squamous cell carcinoma in vivo. Accordingly, low levels of
TGFBI expression can predict a better prognosis [
38], while patients with
TGFBI overexpression have a poor prognosis [
39]. Vascular endothelial growth factor A (
VEGFA) plays an important role in tumor angiogenesis.
VEGFA may be a prognostic gene in clear cell renal cell carcinoma, as significant increases in
VEGFA are associated with a poor prognosis [
40]. Upregulation of
VEGFA also indicates a poor prognosis for lung adenocarcinoma and oral squamous cell carcinoma, suggesting that
VEGFA represents a valuable prognostic biomarker [
41,
42].
The new and important findings of this study are the identification of a prognostic 5-gene signature with a relatively high AUC in the training and test dataset, which can then predict 1, 3 and 5-year survival rates. We also established a series of clinical variables of the nomogram model and verified these predicted genes in a series of experiments, which then substantiated the reliability of the prediction. In the future, genetic diagnosis and treatment will become more effective means. The expression of 5-gene signature in cervical cancer tissues will be detected by qPCR, and the model genes will be converted into risk scores for predicting the prognosis of patients, which is significance of clinical application. The limitation of this study is that some data lacked clinical follow-up information and further genetic and experimental studies along with experimental verification will be required with larger samples. Moreover, direct clinical application tests of this prognosis model will need to be conducted.
In summary, our study developed a 5-gene signature prognostic hierarchical system based on the hypoxic pathway of cervical cancer. This protocol shows an effective AUC in the training and independent test set, and serves as a model which is independent of clinical characteristics. Further, in order to verify the signature, TCGA and GSE44001 datasets was used to test, and finally got a good risk prediction effect in those datasets. We also conducted experimental verifications on these five genes and found that the expression of AK4, HK2, P4HA1, TGFBI and VEGFA were high in cervical cancer tissues. Moreover, silencing these genes inhibited the proliferation, migration and invasion ability of cervical cancer cells as demonstrated in vitro. Therefore, these results suggesting that the signature could potentially be used to evaluate the prognostic risk of cervical cancer patients, and provide potential targets for the treatment of cervical cancer patients.