A three miRNAs signature predicts survival in cervical cancer using bioinformatics analysis

Growing evidences showed that a large number of miRNAs were abnormally expressed in cervical cancer tissues and played irreplaceable roles in tumorigenesis, progression and metastasis. The aim of the present study was to identify the differential miRNAs expression between cervical cancer and normal cervical tissues by analyzing the high-throughput miRNA data downloaded from TCGA database. Additionally, we evaluated the prognostic values of the differentially expressed miRNAs and constructed a three-miRNA signature that could effectively predict patient survival. According to the cut-off criteria (P < 0.05 and |log2FC| > 2.0), a total of 78 differentially expressed miRNAs were identified between cervical cancer tissues and matched normal tissues, including 37 up-regulated miRNAs and 41 down-regulated miRNAs. The Kaplan-Meier survival method revealed the prognostic function of the three miRNAs (miRNA-145, miRNA-200c, and miRNA-218-1). Univariate and multivariate Cox regression analysis showed that the three-miRNA signature was an independent prognostic factor in cervical cancer. The functional enrichment analysis suggested that the target genes of three miRNAs may be involved in various pathways related to cancer, including MAPK, AMPK, focal adhesion, cGMP-PKG, wnt, and mTOR signaling pathway. Taken together, the present study suggested that three-miRNA signature could be used as a prognostic marker in cervical cancer.


Identification of three miRNAs associated with OS in cervical cancer.
To identify the miRNAs which would be potentially associated with overall survival of cervical cancer patients, we evaluated the association between miRNAs expression and patients' survival using Kaplan-Meier curve and Log-rank test. The results showed that one miRNA (miR-200c) was negatively correlated with overall survival (OS), and two miRNAs (miR-145 and miR-218-1) were positively related to OS (Fig. 2). The association between three miRNAs and clinical features was evaluated in cervical cancer patients ( Table 2). The results showed that miR-145 was significantly associated with metastasis (P = 0.033) and T stage (P < 0.001); miR-218-1 was associated with stage (P = 0.004), T stage (P = 0.001), and histological type (P < 0.001). No significant difference was found between miR-200c and clinical features (P > 0.05). Prognostic value of three miRNAs signature risk score in cervical cancer. We constructed a prognostic signature by integrating the expression profiles of three miRNAs and corresponding estimated regression coefficient. Then, we calculated a risk score for each patient, and ranked them according to increased score. Thus, a total of 251 patients were classified into a high risk group (n = 125) and a low risk group (n = 126) according to the median risk score. Survival analysis was performed using the Kaplan-Meier method with a Log-rank statistical test. The result showed that patients in high risk group have significantly worse OS than patients in low risk group (P < 0.001, Fig. 3). Taking into account the following clinical features: age, metastasis, lymph node status, stage, T stage, histological type, pregnancy number, and smoking history category, univariate and multivariate Cox regression analysis were used to test the effect of the three-miRNA signature (high risk vs. low risk) on OS. In univariate analysis, age (HR = 0.562, P = 0.037), lymph node status (HR = 2.567, P = 0.010), stage (HR = 2.511, P = 0.001), T stage (HR = 4.640, P < 0.001), and three-miRNA signature (HR = 2.574, P < 0.001) were associated with OS in cervical cancer patients. In multivariate analysis, the three-miRNA signature (HR = 2.183, P = 0.028) was showed to be an independent prognostic factor in cervical cancer patients (Table 3).
Target prediction and function analysis. The target genes of three miRNAs (miR-145, miR-200c, and miR-218-1) were predicted using TargetScan, miRDB, PicTar, and miRanda online analysis tools. A total of 67 overlapping genes of miR-145, 126 overlapping genes of miR-200c, and 5 overlapping genes of miR-218-1 were identified (Fig. 4). Then, enrichment analysis was performed to elucidate the biological function of consensus target genes. The KEGG pathways were significantly enriched in MAPK signaling pathway, AMPK signaling pathway, focal adhesion, cGMP-PKG signaling pathway, wnt signaling pathway, and mTOR signaling pathway. In addition, the GO biological process (BP) terms were mainly enriched in signal transduction, regulation of cell migration, and regulation of transcription.

Discussion
With the introduction of vaccination and screening programs, the incidence of mortality associated with cervical cancer in developed areas have dramatically declined in recent decades, but the mortality in developing countries remains high, up to 87% of cervical cancer deaths [12][13][14][15] . The cervical cancer patient prognosis would be improved considerably if tumor behavior could be predicted reliably at the time of initial diagnosis. Therefore, understanding the molecular mechanisms of cervical cancer development and identification of novel biomarkers are needed. In the present study, a total of 78 differentially expressed miRNAs were identified, and three of them were associated with overall survival in cervical cancer patients. The three-miRNA (miR-145, miR-200c, and miR-218-1) signature was established and was identified to be an independent prognostic factor for cervical cancer patients. Moreover, we screened the target genes of these three miRNAs, and predicted the enrichment pathways and biological functions of target genes using bioinformatics methods.
In the last decade, MiRNAs, as the master modulators of multiple biological and pathological processes, are a hot research topic in the field of cancer development. Mounting evidence has demonstrated that miR-NAs established a complex combinatorial system of gene expression and pathway regulation, as well as prognostic indicators and therapeutic targets in different cancers, including cervical cancer 16,17 . Previous studies have demonstrated that many miRNAs are crucial for the initiation, progression and metastasis of cervical cancer by regulating various processes, including cancer cell proliferation, differentiation, apoptosis, adhesion, cell cycle arrest, migration and invasion 18 . To date, several studies had identified a number of miRNAs with prognostic values, such as miR-155 19 , miR-425-5p 20 , miR-196a 21 , miR-503 22 , miR-26b 23 , miR-335 24 , miR-3 25 , miR-215 26 , miR-224 27 , and so on. However, previous studies were based on small sample size, sample types, different detection platforms, various assay methods, and relatively limited numbers of miRNAs. In the present study, we     In the present study, we found that miR-145, miR-200c, and miR-218-1 were differentially expressed, and significantly associated with overall survival in cervical cancer patients. While efficacy of a single marker was limited, multi-markers based model may provide more powerful information for the prognosis prediction of patients. We constructed three-miRNA signature, and the results suggested that the three-miRNA signature (high risk and low risk) predicted survival well, and was an independent prognostic factor in cervical cancer.
To gain a deep insight into the molecular functions of three miRNAs, we predicted the target genes and analyzed the related pathways and GO annotations. Abnormal signaling pathways play crucial roles in the pathogenesis and progression of cervical cancer. We found that three miRNAs could regulate several key signaling pathways, including MAPK signaling pathway, AMPK signaling pathway, focal adhesion, cGMP-PKG signaling pathway, wnt signaling pathway, and mTOR signaling pathway. Accumulating evidence has demonstrated that activation of MAPK signaling pathway is important in cervical cancer progression, invasion, and metastasis 36 . Yung M. M., et al. reported that activation of AMP-activated protein kinase (AMPK), a metabolic sensor, hampers cervical cancer cell growth through blocking the Wnt/β-catenin signaling activity 37 . The transformation of HPV expressing human keratinocytes requires activation of the Wnt pathway and that this activation may serve as a screening tool in HPV-positive populations to detect malignant progression 38 . Moreover, it has been well established that the PI3K/Akt/mTOR signaling pathway plays a crucial role in cervical cancer development 39 , and inhibition of mTOR kinase activity suppress tumor growth 40 . Therefore, further molecular investigations are needed to confirm these predictions, and it can provide new therapeutic interventions in cervical cancer.
Taken together, we identified three-miRNA signature as a potential prognostic predictor for cervical cancer patients. Further studies are needed to validate our findings in large sample size, and further function investigation are also required to explore the molecular mechanism of these miRNAs in cervical cancer progression.

Materials and Methods
Data processing. The raw sequencing data and clinical information were downloaded from TCGA database (https://cancergenome.nih.gov/). The inclusion criteria were set as follows: (1) the sample with both miRNA sequencing data and clinical information; (2) the sample with prognosis information. Finally, a total of 254 samples were enrolled in this study, including 251 cervical cancer tissues and 3 matched normal tissues. The detailed  Table 3. Univariate and multivariate Cox regression analysis in CC patients.
clinical characteristics and differentially expressed miRNAs were list in Table S2. The miRNA sequencing data were processed using R language package. The differentially expressed miRNAs between cervical cancer and normal tissues were analyzed by limma package in R. The fold changes (FCs) in the expression of individual miRNA were calculated and differentially expressed miRNAs with log 2 |FC| > 2.0 and P < 0.05 were considered to be significant.
Association of differentially expressed miRNAs and patient prognosis. The differentially expressed miRNA profiles were normalized by log2 transformed. The prognostic value of each differentially expressed miRNA was evaluated using Kaplan-Meier curve and Log-rank method. The miRNAs that were significantly associated with overall survival were identified as prognostic miRNAs, and then subjected to a binary logistic regression analysis. Subsequently, a prognostic miRNA signature was constructed, and the miRNA signature could calculate a risk score for each cervical cancer patient. With the miRNA signature, cervical cancer patients were classified into high risk and low risk groups using the median risk score. Then, the differences in patients' survival between the high risk group and low risk group were evaluated by Kaplan-Meier method.
The target gene prediction of prognostic miRNA signature. The target genes of prognostic miR-NAs were predicted using TargetScan (http://www.targetscan.org/), miRDB (http://www.mirdb.org/miRDB/), PicTar (http://pictar.mdc-berlin.de/), and miRanda (http://www.microrna.org/) online analysis tools. To further enhance the bioinformatics analysis reliability, the overlapping target genes were identified using Venn diagram. Then, the overlapping genes were analyzed by The Database for Annotation, Visualization and Integrated Discovery (DAVID) bioinformatics tool (https://david.ncifcrf.gov/). DAVID is a web-based online bioinformatics resource that aims to provide a comprehensive set of functional annotation tools for the investigators to understand the biological mechanisms associated with large lists of genes/proteins 41 . Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were then performed for the target genes. The P-value < 0.05 and gene count ≥ 3 were set as the cut-off criteria.
Statistical analysis. The data were expressed as mean ± standard deviation (SD). The expression levels of miRNAs in cervical cancer and matched normal tissues were analyzed by unpaired t test. The chi-square and t tests were performed to assess the relationship between miRNA expression and clinical features. Kaplan-Meier survival analysis and univariate/multivariate Cox proportional hazard regression analysis were carried out to compare each miRNA (low vs. high level) and prognostic miRNA signature (low vs. high risk). P value less than 0.05 was considered as statistical significant. The statistical analysis was performed using IBM SPSS Statistics software program version 22.0 (IBM Corp., NY, USA).