Integrated miRNA profiling and bioinformatics analyses reveal potential causative miRNAs in gastric adenocarcinoma.

Gastric cancer (GC) is one of the leading causes of cancer-related deaths throughout China and worldwide. The discovery of microRNAs (miRNAs) has provided a new opportunity for developing diagnostic biomarkers and effective therapeutic targets in GC. By performing microarray analyses of benign and malignant gastric epithelial cell lines (HFE145, NCI-N87, MKN28, RF1, KATO III and RF48), 16 significantly dysregulated miRNAs were found. 11 of these were validated by real-time qRT-PCR. Based on miRWalk online database scans, 703 potential mRNA targets of the 16 miRNAs were identified. Bioinformatic analyses suggested that these dysregulated miRNAs and their predicted targets were principally involved in tumor pathogenesis, MAPK signaling, and apoptosis. Finally, miRNA-gene network analyses identified miRNA-125b as a crucial miRNA in GC development. Taken together, these results develop a comprehensive expression and functional profile of differentially expressed miRNAs related to gastric oncogenesis. This profile may serve as a potential tool for biomarker and therapeutic target identification in GC patients.


INTRODUCTION
Gastric cancer (GC), one of the most common malignancies worldwide, is also the primary cause of cancerrelated deaths in China and internationally [1]. Due in part to the absence of early diagnostic modalities, GC is usually diagnosed at late stages, with a very poor prognosis. The 5-year survival rate of late-stage GC is only 3.1%, whereas the 5-year survival of early GC is up to 90% [2]. Therefore, it is critical to further elucidate the molecular pathogenesis of GC in order to identify molecular biomarkers for early detection and novel targets for effective therapy.
MicroRNAs (miRNAs) are small, single-stranded RNAs that negatively modulate transcription by sequencespecific interaction with the 3′ untranslated regions (UTRs) of target mRNAs [3]. Their target genes are involved in many cancer-related cellular processes, including cell cycle regulation, differentiation, apoptosis, and invasion/ migration. MiRNAs have been characterized as oncogenes or tumor suppressors [4]. Dysregulated expression of miRNAs is a newly proposed mechanism of cancer pathogenesis [5,6]; specific miRNAs are associated with each type of tumor development and progression [7,8]. Nevertheless, the unique expression profiles of miRNAs and their downstream signaling pathways in GC remain incompletely characterized. Therefore, the purpose of the current study was to delineate the global miRNA expression profile and related signaling networks in GC.
MiRNA microarray analyses were first performed to identify dysregulated miRNAs in human GC cell lines. 16 miRNAs were found to be aberrantly expressed, 11 of which were chosen and validated by real-time qRT-PCR miRNAs-106a, -17, -20a, -92a, -100, -125b, -127-3p, -145, -381, and -455. Results of these qRT-PCRs agreed with miRNA microarray data. 2,532 potential mRNA targets of these miRNAs, related to cell proliferation, cell death, apoptosis, and known vital tumorigenesis signaling pathways, were predicted by further bioinformatics analyses based on the online miRWalk database. In addition, miRNA-gene network analyses suggested miRNA-125b as a key miRNA in GC pathogenesis.

GO analysis
To elucidate functions of dysregulated miRNAs in GC, potential mRNA targets of altered miRNAs were collected using the Validated Target module of miRWalk. 2,532 targets were obtained. Among these targets, deduplication was conducted and 703 (27.7%) targets were finally saved.
The functions of these 703 target genes were then analyzed using GO annotation with the DAVID tool. When setting the cutoff standard at P < 0.01 and FDR < 0.05, 446, 27 and 33 GO terms, respectively, were found from three ontologies: biological processes, cellular components and molecular functions ( Table 1). The 10 most significant GO terms are shown in Table 1. In addition, cluster analysis was also conducted to search for the enrichment of targets represented by the negative logarithm of the P value (-Log (P)) in the ontology of biological processes. The top 10 functional classes are shown in Figure 4. According to cluster analyses for biological functions, most of the predicted targets of altered miRNAs were involved in cell proliferation, transcription, cell death, programmed cell death, RNA metabolic process, and gene expression. These functions are known to be strongly associated with human tumorigenesis and metastasis [9].

KEGG pathway analysis
To gain a better understanding of the functions and regulatory networks of predicted target genes, we searched for target enrichment in KEGG. There were 702 differentially expressed genes identified in 29 pathways with a cutoff standard of P < 0.01 and FDR < 0.05 by KEGG pathway analysis ( Table 2). These results revealed that the predicted targets of dysregulated miRNAs were associated with multiple tumor types, including bladder, pancreatic, lung and colorectal cancer. Predicted targets were correlated with apoptosis, cell cycle, and cell adhesion. Some important signaling pathways known to mediate tumorigenesis and metastasis were also involved, including the TGF-β, p53 and ErbB signaling pathways. Moreover, the MAPK cell cycle pathways were enriched according to-Log (P) values ( Figure 5).

Potential causative miRNAs in GC
To further elucidate correlations between miRNAs and potential target genes, miRNA-gene network analyses were generated by Cytoscape ( Figure 6). Relationships between miRNAs and target genes were analyzed by integrating the top GO analysis results with the KEGG pathway data. In this Figure, the size of each node is proportional to the purported functional connectivity of each miRNA. According to its size in this representation, miRNA-125b is the hub node with the most interactions, suggesting that it is a key causative point within this network.

DISCUSSION
During the past decade, miRNAs have emerged as key regulators in numerous oncogenic processes. Their discovery has initiated revolutionary progress in cancer research [10][11][12]. Thus, it is relevant to delineate the expression profiles, functions, and potential regulatory mechanisms of miRNAs in cancers. The current study investigated the expression signature of miRNAs in GC and analyzed their downstream targets, functions, and signaling pathways in GC.
Although our microarray data was not validated in primary GC tissue samples, our findings are consistent with previous literature in GC as well as other cancer types. For example, miRNA-106a exhibits oncogenic functions in pancreatic cancer [13]. It was also found to be overexpressed in GC and to regulate invasion and metastasis [14][15][16]. MiRNAs-17, -20a and -92a, members of the miRNA-17-92 cluster, acting as oncogenes, are also overexpressed in GC tissues [17]. The expression level of miRNA-20b was also found to be increased in GC tissues [18]. Meanwhile, miRNAs-125b, -127, and -145 were downregulated in GC tissues and cell lines. MiRNA-145 was shown to suppress invasion, metastasis, and angiogenesis in GC cells [19][20][21]. The remaining dysregulated miRNAs that we identified are reported for the first time in GC. Thus, our results provide a rationale for further study of these novel dysregulated miRNAs in gastric tumorigenesis.
Since most miRNAs are believed to be involved in carcinogenesis by inhibiting target mRNAs, we predicted targets of these 16 altered miRNAs. In total, 2,532 targets were identified using the Validated Target module of miRWalk database. Further GO analyses revealed that the majority of predicted target genes functioned in modulating transcriptional regulation, cell proliferation, cell death, and programmed cell death. These functions Signaling pathways regulated by validated targets of the 16 dysregulated miRNAs were assessed via KEGG pathway analysis. In the current study, the most significant pathways were closely associated with tumor pathogenesis, including pathways in cancer, pancreatic cancer, prostate cancer, chronic myeloid leukemia, and bladder cancer. In addition to pathways in cancer, the MAPK signaling pathway was also shown to be important in GC development. MAPKs (mitogen-activated protein kinases) principally comprise extracellular signalingrelated kinases (ERKs), c-Jun NH2-terminal kinases (JNKs), and p38 MAPKs. These pathways play an important role in cell proliferation and differentiation, survival, apoptosis, and anti-apoptosis [22,23] and are implicated in various tumor types. Furthermore, Liang et al. [24] showed that MAPKs, particularly the ERK subclass, are overexpressed and correlate with tumorigenesis and metastatic potential in GC. In our results, 53 potential mRNA targets were implicated in the MAPK pathway. Thus, elucidating oncogenic mechanisms by which miRNAs regulate the MAPK pathway represents a promising strategy for identifying new therapeutic targets in GC.
Finally, by combining GO terms and KEGG pathway analysis with network analyses, we identified a potential crucial miRNA in gastric carcinogenesis, miRNA-125b. MiRNA-125b can act as either a tumor suppressor or an oncogene, depending on tumor type, with either increased or decreased expression. For example, miRNA-125b expression is decreased in hepatocellular carcinoma (HCC) tissues and HCC cell lines vs. normal tissues and cells, suggesting a tumor-suppressive role in this organ [25]. Conversely, in pancreatic cancer, prostate cancer and acute myeloid leukemia, miRNA-125b levels are elevated, implying that it functions as an oncogene in these tissues [26][27][28]. MiRNA-125b expression levels are heterogeneous in GC cell lines and tissues, according to discoveries by several research groups. Matteo et al [19] found that miRNA-125b levels were significantly downregulated in GC tissues. However, another group showed that miRNA-125b was upregulated in GC (HGC-27, MKN-28, BGC-823, SGC-7901, AGS, and MKN-45 GC cell lines) vs. the normal gastric cell line GES-1 [29]. In our study, miRNA-125b was downregulated in all GC cell lines studied (NCI-N87, MKN28, RF1, KATO III and RF48) vs. HFE145. Inconsistent findings in previous studies may have been due to histologic heterogeneity in GC. Moreover, previous studies have indicated that expression of miRNA-125b is altered according to pathophysiological conditions [30][31][32]. For example, in papillary thyroid cancer, miRNA-125b is upregulated, while in anaplastic thyroid carcinoma, expression of miRNA-125b is reduced relative to normal thyroid tissue [32]. Notwithstanding the heterogeneous and controversial expression of miRNA-125b in various cancers, miRNA-125b plays a key role in regulating signaling  pathways that mediate cell proliferation, differentiation, and apoptosis [33]. Thus, miRNA-125b may represent a potential biomarker for the diagnosis and independent prognostication of GC. Admittedly, no biological functional experiments have been conducted to elucidate the causative role of miRNA-125b. Nevertheless, further investigations are indicated to overcome this gap in knowledge.
In summary, 16 significantly dysregulated miRNAs were discovered in GC, including 6 up-and 10 downregulated miRNAs. In addition, 2,532 potential target transcripts were predicted for these 16 miRNAs. Integrated GO and KEGG pathway analyses, combined with miRNAgene network analyses, uncovered a potential key gastric oncogenic miRNA, miRNA-125b. These findings imply that dysregulated expression of miRNAs and their potential target mRNAs are prominently involved in the pathogenesis of GC, suggesting that dysregulated miRNAs may serve as potential biomarkers and therapeutic targets in GC patients.

Cell lines
Human immortalized normal gastric epithelial cells (NC) HFE145 and GC cells (NCI-N87, MKN28, RF1, KATO III and RF48) were obtained from Howard University and the American Type Culture Collection, respectively. These cells were grown in DMEM medium containing 10% fetal bovine serum (FBS). Cell lines were cultured in a humidified air supplemented with 5% CO 2 at 37°C.

Preparation of Total RNA
According to the manufacturer's instructions, total RNA, including the miRNAs, was extracted from cells by using the miRNeasy mini kit (Qiagen, Hilden, Germany).The RNA concentration was measured using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, USA). The integrity and the quality of RNA were determined by Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, USA). RNAs with a 2100RIN (RNA integrity number) 6.0 and 28S/18S 0.7 were used for the miRNA array analysis and reverse transcription (RT).

MicroRNA microarrays
MicroRNA microarray analyses were performed on two groups of cell lines: normal gastric epithelial cells (NC) HFE145 and GC cells (NCI-N87, MKN28, RF1, KATO III and RF48). MicroRNA labeling and microarrays were performed with MicroRNA Labeling Reagent and Hybridization Kits (Agilent, Santa Clara, CA, USA) and Human microRNA Microarray Kits (V2, Agilent), respectively. All procedures were performed as described previously [34]. Every miRNA was represented by 2 probes and 15 replicates on each microarray, which contained 723 human miRNAs and 76 human viral miRNAs. 100 ng of total RNA from each cell line was treated with phosphatase and then labeled with cyanine 3-pCp. Hybridization was carried out according to manufacturers' instructions. Hybridization signals were detected using an Agilent  Microarray Scanner (Agilent) with Agilent's Scan Control version A.7.0.1 software. Quantile normalization was conducted. The value of two probes were averaged and used as the final expression level. Differential analysis was carried out using the Student t-test; P < 0.05 was set as the cut-off. MiRNAs that were up-regulated or down-regulated by at least two-fold between the two cell line groups were regarded as significantly dysregulated miRNAs.

Real-time quantitative reverse transcription PCR
Procedures for real-time quantitative reversetranscription PCR (qRT-PCR) analyses have been described previously [35]. TaqMan MicroRNA Assays, Human (Applied Biosystems, Foster City, CA, USA) and U6 small nuclear RNA TaqMan RT-PCR amplicon (RNU6B TaqMan microRNA Assay kit, Applied Biosystems) were used to confirm miRNA and U6 expression, respectively. U6 was used as an internal control. Fold-change of expression was calculated by using Ct values in comparison with controls. Expression levels were determined by the comparative ∆Ct method.

Target prediction and bioinformatic analyses
In silico prediction of miRNA target genes was carried out by miRWalk (http://www.umm.uni-heidelberg .de/apps/zmf/mirwalk/) [36]. The miRWalk database focuses on predicted as well as validated miRNA binding sites in all known genes in human, mouse and rat. In current study, the potential target genes were processed by the Validated Target module of the miRWalk database. For functional analyses of miRNA potential targets, GO term analysis was applied to organize genes into categories on the basis of biological processes, cellular components and molecular functions [35]. Biological pathways defined by Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis were identified by DAVID (Database for Annotation, Visualization and Integrated Discovery) software (http://david.abcc.ncifcrf.gov/tools.jsp). DAVID online provides a set of functional annotations of a large number of genes. P-values of each pathway were adjusted by the method of Benjamini-Hochberg to control the false discovery rate (FDR). In current study, GO terms and signaling pathways were selected with the threshold of significance being defined as P < 0.01 and FDR < 0.05.

ACKNOWLEGDEMENTS AND FUNDING
We are grateful to Drs. Hassan Ashktorab and Duane Smoot for the provision of HFE-145 cells.

CONFLICTS OF INTEREST
The authors disclose no potential conflicts of interest.