Identification of the miRNAs associated with clinical outcome of gastric cancer patients.
The association between miRNAs expression patterns and clinical outcomes was analyzed using LinkedOmics database, which are web-based tools to deliver fast and customizable functionalities based on The Cancer Genome Atlas (TCGA) data[12]. The significances (-log10 P value) of the correlation between the expression of all microRNAs and the overall survival, T stage (or pathological stage), and M stage of GC were shown in the four Volcano Plots. The results showed that GC patients with higher expression of miR-100 (miR-653, miR-125a, miR-218, etc.) or lower expression of miR-182 (miR-7, miR-96, miR-15b, miR-194, etc.) usually have a shorter overall survival time (p < 0.05, Fig 1 A). The expression of miR-217 (miR-181a, miR-132, miR-100, etc.) was positively correlated with the T stage of GC, while the expression of miR-320a (miR-7, miR-182, miR-183, miR-194, etc.) was negatively correlated with the T stage of GC (p < 0.05, Fig 1B). The expression of miR-130a (miR-217, miR-132, miR-100, etc.) was positively related to the pathological stage of GC, while the expression of miR-7 (miR-320a, miR-182, miR-183, miR-194, etc.) was negatively related to the pathological stage of GC (p < 0.05, Fig 1 C). In addition, we also note that GC tissues in M1 stage usually possessed higher expression of miR-152 (miR-181a, miR-125b, miR-100, etc.) and lower expression of miR-194 and miR-147b (p < 0.05, Fig 1 D) compared to the GC tissues in M0 stage.
To determine the most appropriate diagnostic and prognostic biomarkers for GC, we plotted Venn diagrams based on the number of miRNAs that significantly positively associated with overall survival rate, T stage, pathological stage and M stage. Table S1 lists miRNAs that are significantly related to at least two clinical characteristics (such as survival rate, T stage, pathological stage, and M stage). As shown in Figure 1E, F and Table S1, miR-100, miR-125b, miR-199a and miR-194 were the 4 most promising biomarkers that can accurately predict cancer stage and reflect an individual’s cancer risk in GC. Additionally, our data also suggested that miR-100, miR-125b, miR-199a were unfavorable prognostic biomarkers in GC, while miR-194 was favorable prognostic biomarker in GC.
Lower expression of miR-194 predicts poorer prognosis in TCGA gastric cancer cohort.
In order to better understand the correlation between miR-194 expression and clinical pathology of gastric cancer, we further analyzed the RNA-Seq data of miR-194 in the stomach cancer tissues from TCGA database (n > 375). The results showed that miR-194 expression level in the diffuse type GC was significantly lower than that in the intestinal type GC (Fig 2A, p < 0.0001). Moreover, miR-194 expression tends to be higher in moderately or highly differentiated gastric cancer tissues than that in poorly differentiated gastric cancer tissues (Fig 2B, p < 0.01).
Furthermore, there was a significant positive correlation between miR-194 expression and the extent of GC progression (Fig 2C). Similarly, miR-194 was more highly expressed in the gastric mucosa (T1) than tumors extending beyond the gastric mucosa layer (T2+T3+T4) (Fig 2D). Patients with lower expression of miR-194 were more likely to have distant tumor metastasis (Fig 2E). More importantly, patients with lower miR-194 expression tended to have a shorter overall survival time than those patients with higher miR-194 expression (Fig 2F). Collectively, these results together suggested that miR-194 may be tumor suppressor in GC and miR-194 could serve as an independent diagnostic and prognostic biomarker for GC.
The preferences of miR-194-5p in regulating target genes
To figure out the preference of miR-194 in regulating target genes, RNA sequencing studies was performed in the two gastric cancer cell lines (SGC7901 and BGC823) that transfected with miR-194-5p mimics and corresponding negative control siRNAs. The RNA sequencing results were uploaded on the NCBI website. And the gene expression omnibus accession number was GSE134308. The heat map showed that there were hundreds of coding genes (non-coding genes were not included in this study) have altered their expression levels after treatment with miR-194-5p mimics (Fig 3A, |FC| > 1.5). Of these, approximately 138 coding genes were relatively strongly downregulated by miR-194 mimics, while almost 70 coding genes that relatively severely upregulated by miR-194 mimics in both SGC7901 and BGC823 cell lines (Fig 3B). And the top 23 coding genes that most strongly downregulated by miR-194 were listed in the Table 1 (log2 FC < -0.9).
According to the binding sequence differences in the miR-194-5p seed region, the miR-194-5p binding sites in mRNAs can be divided into three types, including 7-mer-A1, 7-mer-m8 and 8-mer (Fig 3C). Then, we annotated the information of miR-194 binding sites in the coding genes that downregulated by miR-194 using miRcode web-based tools. The results showed that nearly 60% of the coding genes contained miR-194-5p binding sites (Fig 3D). And the distribution proportion of the three binding modes is very close, suggested that any type of binding manner between miR-194 and mRNAs would be effective (Fig 3E). Next, we further analyzed the binding location preference of miR-194 in mRNA sequence. The results showed that almost 86% of miR-194 binding sites were located at 3' untranslated region of target genes (Fig 3F). In addition, no significant correlation was observed between expression alteration (log2FC) and the number of miR-194 binding sites in both SGC7901 and BGC823 cell lines (Fig 3G and H).
MiR-194 negatively regulated CCND1 expression by binding on the 3' untranslated region in GC.
Next, we conducted KEGG pathway analysis to determine the most affected pathways by miR-194 in GC. As shown in Supplementary Figure, the cell cycle pathway was one of the most affected pathways by miR-194. RNA-seq data showed that CCND1 was strongly downregulated in both SGC7901 and BGC823 cell lines (log2FC value was -0.84 in SGC7901, -0.71 in BGC823). To confirm the RNA-seq results, we investigated the CCND1 expression in two GC cell lines after treatment with miR-194 mimics and inhibitors by qPCR analysis and the western blot assay. The results showed that miR-194 overexpression significantly decreased CCND1 expression in GC, while miR-194 inhibitor significantly increased CCND1 expression in GC (Fig 4A-D). These results strongly indicated that CCND1 was negatively regulated by miR-194.
Based on the above results, we further analyzed the information of miR-194 binding site in the CCND1 transcript using the miRcode web-based tools and predicts the folding energy between miR-194 and CCND1 transcript using RNA22 web-based tools [15, 16]. As shown in Figure 5A, in the 3' untranslated region of CCND1, there was only one miR-194 binding site (7-mer-A1). The folding energy between miR-194 and CCND1 transcript was -9 Kcal/mol, suggested that the negative regulation of miR-194 on CCND1 expression might be through regulating mRNA stability by the direct binding on CCND1 transcripts. To further validate this hypothesis, luciferase reporter assay was performed. Since the length of 3' untranslated region of CCND1 was too large to be amplified, we selected to clone the 735-length (from 1900 nt to 2642 nt) into the luciferase reporter vector with original sequence (wildtype) or with the 4 nt changed (mutant) in the miR-194 binding site (Fig 5B). As expected, compared to the BGC823 cells that transfected with negative control siRNA and wildtype luciferase vector, a significant decrease and increase was observed in the BGC823 cells that transfected with miR-194 mimics and inhibitors, respectively. However, no obvious changes were detected in the BGC823 cells that transfected with mutant CCND1 3’UTR luciferase vector (Figure 5C). These results strongly suggested that miR-194 negatively regulate CCND1 expression at post-transcriptional level.
MiR-194 suppressed GC cell proliferation through induction of GC cell cycle arrest.
The cyclin protein CCND1 plays an essential role in the cell cycle process. Increasing studies have demonstrated that downregulation of CCND1 induces G1 phase arrest and then impairs cell growth[17]. Therefore, we performed the cell proliferation assay in SGC7901 and BGC823 cells that transfected with miR-194 mimics and inhibitors. The results showed that miR-194 mimics markedly suppressed the cell proliferation in GC, while miR-194 inhibitors slightly promoted the cell proliferation in GC (Fig 6 A and B). On the other hand, we conducted the flow cytometry assays in the two GC cell lines after treatment with miR-194 mimics and corresponding negative control siRNAs. The flow cytometry analysis showed that compared to NC group, miR-194 significantly decreased the number of the sub G0 phase and S phase cells, but increased the number of G1 phase cells in both SGC7901 and BGC823(Fig 6 C-F). These results together suggested that miR-194 inhibits GC cell growth via induction of cell cycle arrest.