PA is one of the most common intracranial tumors with an incidence of 10–15% . Generally, PAs are considered benign, but some of them are invasive, which invade the adjacent structures such as sphenoid sinus, cavernous sinus, and diaphragma sellae [2,3,4]. Therefore, IPAs are not only more difficult to complete surgical resection, but also more likely to recurrence after surgery.
MiRNAs are single-strand non-coding RNAs of approximately 19–23 nt, which regulate gene expression at the post-transcriptional level [5, 6], and they can also act as tumor suppressor genes or oncogenes in various tumors . For instance, miR-193b exerts tumor suppressive effects in human acute myeloid leukemia by inducing tumor cell apoptosis and G1/S arrest , while miR-210-3p plays an oncogene role in prostate cancer by promoting cancer cell epithelial–mesenchymal transition and bone metastasis via NF-κB signaling pathway . Altered expression of many miRNAs has been described in PAs, and specific miRNA signatures are related to clinical and therapeutic characteristics of the tumors . However, comprehensive and specific researches of relationships between miRNAs and invasiveness of PAs are still rare.
In order to better understand the mechanism of invasiveness in PAs, it is necessary to clarify the miRNA regulatory network in IPAs. In this study, we detected DEMs in IPAs and NPAs by RNA sequencing, and established the co-expression network contain miRNAs and predicted target genes by Cytoscape. In addition, the expression of the most upregulated miR-665 and the most downregulated miR-149-3p in IPAs was screen out. Moreover, we explained the potential functions of the two key miRNAs in invasive behavior of PAs by GO analysis and KEGG pathway analysis.
Patients and samples
Seven tumor samples were obtained from patients with PAs who underwent operation at the Department of Neurosurgery, 1st Affiliated Hospital of Kunming Medical University for identification of miRNAs by high-throughput sequencing. None of these patients has been received radiotherapy or chemotherapy before surgery. Tumor samples were divided into 2 groups according to invasive behavior proved by surgical findings and pathology: IPA and NPA. All patients were informed according to inform consent approved by the Ethics Boardof the 1st Affiliated Hospital of Kunming Medical University. Immediately following separation, the fresh tumor samples were placed in sterile, RNase-free 2.0-mL cryotubes. Then, samples were soaked in Trizol and stored at − 80 °C for following analysis.
RNA isolation, library preparation, and sequencing analysis
Total RNA was extracted from tissue samples by Trizol regent. The integrity of total RNA was detected by agarose electrophoresis and which was quantified by NanoDrop spectrophotometer. Then, the sequencing sample library was constructed by the following steps: ribosomal RNA removal, fragmentation, first-strand complementary DNA (cDNA) synthesis, second-strand cDNA synthesis, terminal repair 3′ terminal addition, ligation, and enrichment. The libraries were sequence on an Illumina Hiseq 2500/2000 platform.
MiRNA expression analysis
MiRNA expression levels were estimated by the TPM (transcript per million) through the following criteria: Normalized expression = mapped readcount/Total reads × 106 . All data were analyzed using the DESeq2 R package (1.8.3). log2FC > 1 and p < 0.05 were considered as the cutoff values for DEMs screening .
MiRNA-mRNA network construction
Based upon results of DEM analysis and target gene prediction, the miRNA-mRNA pairs were extracted to construct the miRNA-mRNA regulatory network. Then, the regulatory network was visualized using Cytoscape_v3.5.1.
Target gene prediction, gene ontology, and pathway enrichment analysis
Target genes of the DEMs were predicted using major online tools, including miRanda (http://miranda.org.uk/), PITA (http://genie.weizmann.ac.il/pubs/mir07/mir07_data.html), and RNAhybrid (https://bibiserv.cebitec.uni-bielefeld.de/rnahybrid/) . In order to analyze the main functions of the predicted target genes for the DEMs, we performed GO analysis . Moreover, KEGG  pathway enrichment analysis was used to find out the significant pathway of predicted target genes for the DEMs. A Go term or KEGG pathway with FDR < 0.05 was considered statistically significant. Top 10 enriched GO terms and pathways of DEMs were ranked by enrichment score (− log10(p value)).