Transcriptome analysis uncovers the diagnostic value of miR-192-5p/HNF1A-AS1/VIL1 panel in cervical adenocarcinoma

Despite the fact that the incidence of cervical squamous cell carcinoma has decreased, there is an increase in the incidence of cervical adenocarcinoma. However, our knowledge on cervical adenocarcinoma is largely unclear. Transcriptome sequencing was conducted to compare 4 cervical adenocarcinoma tissue samples with 4 normal cervical tissue samples. mRNA, lncRNA, and miRNA signatures were identified to discriminate cervical adenocarcinoma from normal cervix. The expression of VIL1, HNF1A-AS1, MIR194-2HG, SSTR5-AS1, miR-192-5p, and miR-194-5p in adenocarcinoma were statistically significantly higher than that in normal control samples. The Receiver Operating Characteristic (ROC) curve analysis indicated that combination of miR-192-5p, HNF1A-AS1, and VIL1 yielded a better performance (AUC = 0.911) than any single molecule -and could serve as potential biomarkers for cervical adenocarcinoma. Of note, the combination model also gave better performance than TCT test for cervical adenocarcinoma diagnosis. However, there was no correlation between miR-192-5p or HNF1A-AS1 and HPV16/18 E6 or E7. VIL1 was weakly correlated with HPV18 E7 expression. In summary, our study has identified miR-192-5p/HNF1A-AS1/VIL1 panel that accurately discriminates adenocarcinoma from normal cervix. Detection of this panel may provide considerable clinical value in the diagnosis of cervical adenocarcinoma.

www.nature.com/scientificreports/ chemotherapy and radiotherapy than that of SCC 14,15 . In view of this clinical challenge, there is an urgent need to elucidate molecular mechanisms and explore new reliable prevention strategy for cervical adenocarcinoma. Various types of biomarkers, especially RNA biomarkers, have been applied to clinical diagnosis for diseases for decades. Recently, high-throughput RNA sequencing technology facilitates the quantification measurements of RNA expression at the entire transcriptome level. These large-scale expression profiles of RNAs enable the detection of protein-coding RNAs (i.e., mRNAs) and non-coding RNAs (i.e., ncRNAs), which could be of profound value in terms of disease characterization. As the best-characterized type of RNA biomarkers, the mRNA biomarkers have successfully programmed into multi-gene panels for cancer diagnosis [16][17][18][19] . Lots of functionally important ncRNAs can also be used as biomarkers for increasing evidence of abnormal expression of ncRNAs closely associated with various cancers. For instance, a blind study of 22 different tumor types showed that microRNA (miRNA) expression pattern could accurately classify tumors according to tissue of origin 20 . It was reported that the expression of miR-27a is decreased in cervical cancer cell lines and it functions as a tumor suppressor in cervical adenocarcinoma by inhibiting TGF-βRI signaling pathway 21 . MiR-362-3p downregulation in cervical adenocarcinoma is associated with advanced clinical stage and poor prognosis via targeting minichromosome maintenance protein 5 (MCM5) 22 . In addition, long non-coding RNAs (lncRNAs) could be considered as biomarkers. For instance, lncRNA HOTAIR plays important roles in the development of cancer [23][24][25][26] . In cervical cancer, it was reported that high levels of circulating HOTAIR are correlated with tumor recurrence and short overall survival 27,28 . To date, the lncRNA profile in cervical adenocarcinoma remains largely unknown.
In this study, we comprehensively examined the mRNA and ncRNA expression profiles (including miRNAs and lncRNAs) in cervical adenocarcinoma and normal cervix by using transcriptome sequencing. We aim to identify and validate a mRNA/ncRNA-combined signature that could improve the screening or early detection of cervical adenocarcinoma. We revealed that a panel with miR-192-5p, HNF1A-AS1 and VIL1 could accurately discriminate adenocarcinoma from normal cervix.

Methods
Human tissue samples. Three groups of cervical adenocarcinoma and normal cervical tissue samples were all collected from the Women's Hospital, Zhejiang University School of Medicine. Cervical samples, including 4 normal and 4 adenocarcinoma tissue samples for RNA sequencing were obtained from January 2018 to February 2018. The second panel of cervical tissue samples containing 20 normal and 20 adenocarcinoma tissue samples for the first step validation by RT-qPCR analysis, were collected from March 2018 to March 2019. The third panel of cervical tissue samples including 57 normal and 141 adenocarcinoma tissue samples for the second step validation by RT-qPCR analysis, were obtained from January 2009 to December 2017. None of these patients received radiotherapy or chemotherapy before operation. Diagnosis of these samples was confirmed independently by two senior pathologists before further analysis. After removal from the patients, tissue samples were put immediately in liquid nitrogen and stored at -80 °C until use. This study was approved by the Ethics Committee of Women's Hospital of Zhejiang University (IRB-2019062-R). All steps of the study were performed with the relevant guidelines and regulations (according to the Declaration of Helsinki Principles). Written informed consent was obtained from each patient prior to the study. The three different phases of our study design are shown in Fig. 1A and the patients' clinical-pathological information is summarized in supplementary Table S1. transcriptome sequencing. Four cervical adenocarcinoma tissues and 4 normal cervical tissues were used for transcriptome sequencing. Transcriptome Sequencing was performed by NovelBio Corp. Laboratory, Shanghai, China. Our transcriptome sequencing data have been deposited in NCBI Gene Expression Omnibus (GEO) datasets with the accession number GSE145372 (https ://www.ncbi.nlm.nih.gov/geo). RNA was utilized to construct both rRNA depletion library and miRNA library. The products were sequenced on Hiseq Sequencer and HiSeq Xten platform (Illumina, San Diego, CA), respectively. The reads were then mapped to Human genome (GRCh38 NCBI) utilizing HISAT2 (-5 5 -3 5 -p 8 -min-intronlen 20 -max-intronlen 500,000 -k 3) 29 and HTSeq was used to calculate the gene count of mRNA and lncRNA annotated by NCBI genome gff3 file 30 . To achieve miRNA expression, the reads were filtered (length ranged from 15 to 33 bp; Q20 > 20%; Trim end Q15) and mapped to Human miRNA database (miRBase v22.0) and Human genome (GRCh38 NCBI), together with mapping to the RFAM database (https ://rfam.xfam.org/) for miRNA quality control by BWA (bwa aln -n 0.04 -e 3 -l 32 -k 2 -t 12) 31 . We applied DESeq2 package 32 to discover the differentially expressed mRNA, lncRNA, and miRNA under the following criteria: Absolute Fold Change > 2 and FDR < 0.05.
For heatmap generation, the sample FPKM (Fragments per kilobases million reads) values of Different expression genes were clustered using Software Cluster 3.0 (https ://bonsa i.hgc.jp/~mdeho on/softw are/clust er/softw are. htm). The rows and columns expression matrix was centered according to the mean value and finally normalized to value ranged from − 1 to 1. Average linkage was applied to cluster genes and samples with the uncentered correlation method. The heatmap was finally visualize by TreeView (https ://bonsa i.hgc.jp/~mdeho on/softw are/ clust er/softw are.htm).
Combined with the validation results, we finally chose VIL1, HNF1A-AS1, MIR194-2HG, SSTR5-AS1, miR-192-5p, and miR-194-5p as the most promising candidates for further investigation because of their large fold changes in expression and relatively little variation in each group. Similarly, all these 5 candidates were abnormally expressed in cervical adenocarcinoma (Fig. 3).
We  (Table S3). This combined miR-192-5p/HNF1A-AS1/VIL1 panel achieved greater AUC value than any single candidate detection, indicating a promising biomarker pattern for discriminating cervical adenocarcinoma from normal cervix. The corresponding ROC curve is shown in Fig. 4B.

Discussion
Cervical adenocarcinoma is one of the main causes for cervical cancer-related deaths, and is a heterogeneous disease with multiple molecular dysregulation.   www.nature.com/scientificreports/ is not excluded that combination of different RNA types is more valuable as effective diagnosis biomarkers for cervical adenocarcinoma. In this study, we identified distinct mRNAs, lncRNAs, and miRNAs to gain insights into the molecular events associated with cervical adenocarcinoma. Accordingly, a panel of mRNA/ncRNA model has been built to discriminate cervical adenocarcinoma from normal tissues. First, we identified mRNA, lncRNA and miRNA signatures in cervical adenocarcinoma tissues by transcriptome sequencing that enabled us to efficiently find potential novel RNA targets for diagnosis. Second, we assessed the transcriptome sequencing data by choosing 46 significantly altered candidates, including 21mRNAs, 11lncRNAs and 13 miRNAs, to derive reliable RNA markers. Third, we examined the expression signatures of the 6 putative mRNAs/ncRNAs in another independent cohort of cervical adenocarcinoma tissues and identified a panel containing miR-192-5p, HNF1A-AS1 and VIL1 as the effective biomarker set for cervical adenocarcinoma. Finally, we revealed that the combination of miR-192-5p/HNF1A-AS1/VIL1 panel and TCT test achieves better AUC than any single detection, providing important clues for clinical implementation.
HR-HPV infection plays important roles in the etiology of cervical adenocarcinoma. The continuous expression of viral oncogenes E6 and E7 is critical in the malignant transformation. Our identification of miR-192-5p/ HNF1A-AS1/VIL1 panel offers an exciting novel insight into the development of cervical adenocarcinoma. We further compared these panel markers with HPV16 or HPV18 E6 or E7. We found VIL1 was positively correlated with HPV18 E7. However, no significant correlation was observed between miR-192-5p or HNF1A-AS2 with HPV16 or HPV18 E6 or E7 in cervical adenocarcinoma tissues. Our results indicate that the alteration of non-HPV-driven host molecules may also play important roles in the pathogenesis of cervical adenocarcinoma. www.nature.com/scientificreports/ Aberrant expression of miR-192-5p has been reported in several tumors, such as hepatocellular carcinoma (HCC), non-small cell lung cancer (NSCLC), pancreatic cancer, and cervical cancer [42][43][44][45] . It was reported that miR-192-5p expression was significantly decreased in HCC, especially in cancer stem cell (CSC) + HCC. miR-192-5p could functionally suppressed CSC features in HCC cells through targeting PABPC4 44 . Differently, the levels of serum miR-192-5p was significantly upregulated in nasopharyngeal carcinoma and cervical cancer 43,46 . This controversy indicate that miR-192-5p acts as either tumor-suppressive or oncogenic miRNA in different tumor types. Our results have shown a significant increase of miR-192-5p expression in cervical adenocarcinoma tissues compared with the normal group, suggesting miR-192-5p as an important oncogene in cervical adenocarcinoma. Strikingly, miR-192-5p alone yielded an excellent AUC in discriminating adenocarcinoma from the healthy controls, which further points out the potential applicable value of miR-192-5p as a diagnostic marker for cervical adenocarcinoma. The potential predicted targets of miR-192-5p include CADM1, SBSPON and FAM229B, which were down-regulated in adenocarcinoma from our transcriptome sequencing data. To test whether miR-192-5p expression is regulated by HPV that contributes to the cervical tumorigenesis, we analyzed the correlation between miR-192-5p and HPV16 or 18 E6 or E7. However, the altered expression of miR-192-5p was not correlated with viral E6 or E7 expression and further studies are needed to address the roles of miR-192-5p in cervical adenocarcinoma. The association between miR-192-5p and HPV16E6, HPV E7, HPV18 E6, and HPV18 E7 was analyzed by Spearman rank correlation , respectively. (B) The relationship between HNF1A-AS1 and HPV16E6, HPV E7, HPV18 E6, and HPV18 E7 was shown. (C) The expression of VIL1 was positively correlated with HPV18 E7, but not HPV16 E6, HPV16 E7, or HPV18 E6. Data of the expression levels for correlation analysis were shown as ΔCT. GAPDH was used for normalization. www.nature.com/scientificreports/ Among the cancer-related lncRNAs, lncRNA HNF1A antisense RNA 1 (HNF1A-AS1) has been reported to be highly expressed and act as an oncogene in various human malignancies [47][48][49] . Studies have demonstrated that HNF1A-AS1 participates in various cellular processes, including proliferation, apoptosis, autophagy, migration, and invasion, and thus promotes tumorigenesis and cancer progression. For instance, HNF1A-AS1 promotes hepatocellular carcinoma cell proliferation by repressing the NKD1 and p21 expression via interacting with EZH2, or by sponging miR-30b-5p to promote autophagy 50,51 . HNF1A-AS1 also promotes tumor cell proliferation and metastasis in a Wnt/β-catenin-dependent manner in osteosarcoma and colorectal cancer 52,53 . However, its clinical significance has not been elucidated in cervical adenocarcinoma. In this study, we have shown that the expression levels of HNF1A-AS1 are markedly upregulated in cervical adenocarcinoma tissue samples compared to normal cervical tissue samples. ROC curve analysis showed that HNFA1-AS1 could discriminate adenocarcinoma from normal cervix with an AUC of 0.774. ceRNA analysis showed that HNF1A-AS1 may act as a sponge of miR-3141 and miR-6743-5p and regulate the downstream targets of these two miRNAs.
Villin 1 (VIL1) is a calcium-regulated, actin-binding protein that is associated with the microfilament bundles of brush border microvilli. Under normal physiological conditions, VIL1 is expressed in epithelial cells of gastrointestinal and urogenital tracts. Recently, VIL1 has been reported to be overexpressed in several tumors including gastrointestinal neuroendocrine tumor 54 , lung cancer 55 , hepatocellular carcinoma 56 , colon cancer 57 , and endometrial adenocarcinoma 58 . Nagashio et al. showed that VIL1 was specifically expressed in adenocarcinoma and large cell neuroendocrine carcinoma in sera from pulmonary carcinoma patients, suggesting VIL1 as a useful marker to distinguish adenocarcinoma/large cell neuroendocrine carcinoma from squamous cell carcinoma/ small cell carcinoma in lung cancer 55 . Nakamura et al. also reported VIL1 as a potential diagnostic marker for cervical adenocarcinoma with poor radioresponse by immunohistochemical analysis. Consistent with other reports, we observed significant upregulation of VIL1 mRNAs in human cervical adenocarcinoma tissues. Wang et al. reported that VIL1 functions as an anti-apoptotic protein via maintaining mitochondrial integrity 59 . The anti-apoptotic function can also be induced by either inhibition of the caspase-9 and caspase-3 or activation of the pro-survival proteins, phosphatidylinositol 3-kinase and phosphorylated Akt. The results of co-expression analysis shows that VIL1 is co-expressed with many genes up-regulated in cervical adenocarcinoma. Moreover, we found the expression of VIL1 was positively correlated with HPV18 E7. Thus, we propose that VIL1 might be a candidate of HPV18-related oncogene in cervical adenocarcinoma. Further functional studies are required to verify this possibility.
Although above findings of this study show important clinical implications for diagnosis of cervical adenocarcinoma, our study has some limitations. Firstly, our results were derived from a clinic-based patients instead of a general population. Also, the TCT tests were not detected in all of the cervical adenocarcinoma patients and it may possess biases. Thus, future studies are needed to confirm the role of the multimarker panel in a general population.

conclusion
In summary, we have determined specific mRNA, lncRNA, and miRNA signatures between cervical adenocarcinoma and normal cervix and identified a novel panel consisting of miR-192-5p, HNF1A-AS1 and VIL1 for discriminating cervical adenocarcinoma form normal tissues. This panel has promising clinical value in the early diagnosis of cervical adenocarcinoma.