Nowadays, tumor has become one of the greatest intimidation to human health, which has exceed the cardiovascular disease [25]. Cervical cancer has become the second most common malignancies among females all over the world[26]. Especially, in developing countries, where it is not popular for females to take part in cervical screening, cervical cancer posed a greater threat to woman than developed country [27]. Cervical intraepithelial neoplasias (CIN) was recognized as the precursor lesions for cervical cancer. Persistent infection of human papillomaviruses (HPVs) is one of majority reasons led to CIN [28]. The potential mechanism of CIN is assumed that the infection of virus alterated gene transcription or affected the posttrascription regulation of message RNA. The possible process of posttranscription included two categories. Firstly, microRNA is able to triger degradation of the target message RNA by binding ones 3’ untranslated regions (UTR) [29]. Secondly, RNA binding proteins involve in the process, editing, stable maintenance, transportation and translation of message RNA [9, 30]. Recently, microarray and RNA sequencing technologies have emerged as favourable tools for scientists to investigate the potential variament of gene or gene modification in the development of cancer [31]. In this study, the RNA sequencing data of 306 cervical cancer tissues and 13 normal cervix tissues form GTEx and TCGA databases was intergrated to analyse the expression profile of differently expressed RNA binding proteins (also called DEGs in this article, Differently expressed genes) in cervical cancer. 348 DEGs was retrieved by wilcoxson sum-rank test, of which 177 DEGs were down regulated in tumor samples and 170 DEGs were up regulated in tumor samples. The functional enrichment analysis of GO and KEGG were performed for the downregulated and upregulated DEGs respectively. The PPI network was constructed for sorting the candidate genes of prognosic prediction model by STRING database. Moreover, this DEGs was screened by cox regression with wald X2 test and kaplan-meier analyse with log-rank test. Among these DEGs, WDR43, RBM38, RNASEH2A and HENMT1 with HR < 1 played a protective role in survival. Other six genes (EIF3C, PRPF40B, EEF1D, CTU1, ZC3HAV1L, NUFIP1) were considered as risk factors with HR > 1. The nomograph was drawn to present the prognostic prediction model with FIGO stage and RBPs predictor. It was validated by C-index and calibration curve subsequently. In addition, the enrichment analysis of immune cell and function was performed by ssGSEA package in R software.
We investigated the biological functions of These DEGs by GO analysis. To begin with, the enrichment of cell components was located in the ribosome, cytoplasmic ribonucleoprotein granule and the ribonuclease. They play crucial roles in transmission of genetic information from DNA to protein. Protein was synthesised in ribosome by translating the coding information from RNA. The mutation of ribosomal protein may exert an influence on degradation of p53 protein which involved in the process of many kinds of cancer, such as endometrial cancer, T-cell acute lymphoblastic leukemia, chronic lymphocytic leukemia and colorectal cancer[32]. Many kinds of disease have been reported having something to do with RNA processing or RNA metabolism, which exerted influence on RNA translation[33–35]. The forming of ribonucleoprotein complexes has been recognized as the result of interaction of RNA and RBPs. They sustain the stability of target message RNAs, after which the the efficiency of mRNA translation is promoted. For example, oncogenic RNA binding protein SRSF1 is reported to accelerate the proliferation of lung cancer cells by strengthening the message RNA stability of DNA ligase 1 [36]. What is more, ribonucleoprotein granule was discovered as a crucial region for protein synthesis. The development of cancer is affected by the modification of ribonucleoprotein, because of its significant role in RNA translation [32]. Moreover, the category of molecular function in GO analysis revealed the interactions of RNA and proteins such as RNA methyltransferase activity. RBPs have been discovered to bind with many kinds of RNA such as pre-mRNA, snRNA, tRNA and mRNA. The regulation of various enzyme was also displayed in GO analysis such as endoribonuclease, ribonuclease and nuclease. They are correlated to synthesis or repair of DNA and metabolism of RNA. For example, in the field of correlation between cervical cancer and RNA methylation. Jingxin Pan et al. developed a prognostic prediction model for cervical cancer patients based on m6A RNA methylation regulator [37]. While, most of research concentrate on the methylation of protein or DNA such as the promoters of genes instead of RNA. The underlying mechanism of RNA methylation and CC remains to be revealed. Finally, in term of biological process category of GO analysis, differently expressed RBPs have something to do with the processing of both coding RNA and non-coding RNA such as rRNA and tRNA. Both of RNA splicing and metabolism were adjusted by these differently expressed RBPs. Our result was consistent with the consensus reached before. It is reported that RNA binding protein (RBP) quaking (QKI) was able to interact with the QKI response elements (QREs) in SLC26A4 gene introns, which lies in the 3’UTR (3’ untranslation region) of mRNA after transcription, thereby promoting circSLC26A4 biogenesis. CircSLC26A4 promotes the proliferation of cervical cancer cells in both vivo and vitro [38]. The RBP HuR was discovered to promote the growth of cervical cancer cells by interaction with the 3'UTR of RBP nucleolin (NCL) mRNA, which specifically promoted the translation of NCL without the alteration of NCL mRNA levels[39]. In other kind of cancer, RBP Musashi1 (Msi1) promoted the proliferation of colon cancer cells by target the 3’UTR of p21(cip1) [40]. Then, the items in KEGG pathway analysis suggested that the origination and development of cervical cancer is regulated by RBPs through mRNA surveillance pathway, RNA transport and RNA degradation. The underlying correaltion between RBPs and signal pathways should be under reaserch further.
The relationships between many RBPs and cancer has been reported by former studies which were consist with our study. We discovered that CTU1 is a risk factor for CC patients. It has been reported that CTU1/2, which is partner enzymes in U34 mcm5s2-tRNA modification, sustains metastasis and invasion of breast cancer[41]. Francesca Rapino et al. reported that the inhibition of CTU1 and proteins synergizesing with it could kill melanoma cells[42]. Ming Zhang et al. identified the copy number amplifications of CTU1 in 25% of myxopapillary ependymomas by means of whole exome sequencing [43]. CTU1 has been identified as one of prognostic predictors for prostate cancer and bladder cancer[44, 45]. We also identified NUFIP1 as a risk factor of CC patients. The forming of ETV6-NUFIP1 fusion gene has been reported as a potential cause of acute lymphoblastic leukemia in Mexico [46]. Gaurang P Deshpande et al. reported that NUFIP1 had something to do with genome stability maintenance [47] which may help cancer cells survive the pressure from environment. Mutated genes NUFIP1 had a higher level of expression in metastasis tumor than primary tumor in neuroblastoma indicating its oncogenic driver role [48]. However, the potential role of NUFIP1 in the process of cervical cancer development remains to be revealed.
The risk score calculated by expression level of DEGs was demonstrated to be a risk indicator. Patients in high risk group shows a significant lower survival than low risk group. ROC curve for risk factors suggested that risk score predicted the prognosis better than other factors which may be valuable in clinical application. It suggested that more precise theraputic strategy should by applied to CC patients with higher risk score. At last the expression of each DEGs were analysed with patients’clinical features. CTU1 and ZC3HAV1L were significantly associated with clinic FIGO stage and T stage. Their oncogenic role was exposed gradually with the progress of clinical stage. What is more, the expression level of CTU1 increased with the N stage, which showed that it might promote the lymph node metastasis of cervical cancer. The expression level of EEF1D increased with M stage, which showed that It might had something to do with organ metastasis. Both of WDR43, CTU1 and RBM38 were correlated with pathology class. The expression level of EEF1D, RBM38 and WDR43 ascended with the progression of cancer pathology grade. This information may be a clue for further research about correlation between RNA binding proteins and clinical feature in cervical cancer.
Thanks to public database such as TCGA and GTEx, the correlation between prognosis of CC patients and RBPs was analysed. A robust statistical support could be used to help the RBPs researchers in future. More clinical tharaputic shemes should be developed concentrating on RBPs genes in CC patients. There are some limits in this study. To begin with, the clinical stage, pathology grade and the treatment shemes downloaded from TCGA were incomplete. The HPV infection status of each patient was unknown. These deficiencies affected the acuracy of prediction model we constructed at last. Moreover, the potential mechanisms of how RBPs regulate the development of CC and their interaction relationship were remained to be explained. Finally, the nomograph has to be validated in a larger cohort, that will be helpful for further epidemical research. These deficiencies could be solved with a larger scale of clinical data appeared in future.