Integrated analysis identifies microRNA-195 as a suppressor of Hippo-YAP pathway in colorectal cancer

Background With persistent inconsistencies in colorectal cancer (CRC) miRNAs expression data, it is crucial to shift toward inclusion of a “pre-laboratory” integrated analysis to expedite effective precision medicine and translational research. Aberrant expression of hsa-miRNA-195 (miR-195) which is distinguished as a clinically noteworthy miRNA has previously been observed in multiple cancers, yet its role in CRC remains unclear. Methods In this study, we performed an integrated analysis of seven CRC miRNAs expression datasets. The expression of miR-195 was validated in The Cancer Genome Atlas (TCGA) datasets, and an independent validation sample cohort. Colon cancer cells were transfected with miR-195 mimic and inhibitor, after which cell proliferation, colony formation, migration, invasion, and dual luciferase reporter were assayed. Xenograft mouse models were used to determine the role of miR-195 in CRC tumorigenicity in vivo. Results Four downregulated miRNAs (hsa-let-7a, hsa-miR-125b, hsa-miR-145, and hsa-miR-195) were demonstrated to be potentially useful diagnostic markers in the clinical setting. CRC patients with a decreased level of miR-195-5p in tumor tissues had significantly shortened survival as revealed by the TCGA colon adenocarcinoma (COAD) dataset and our CRC cohort. Overexpression of miR-195-5p in DLD1 and HCT116 cells repressed cell growth, colony formation, invasion, and migration. Inhibition of miR-195-5p function contributed to aberrant cell proliferation, migration, invasion, and epithelial mesenchymal transition (EMT). We identified miR-195-5p binding sites within the 3’-untranslated region (3′-UTR) of the human yes-associated protein (YAP) mRNA. YAP1 expression was downregulated after miR-195-5p treatment by qRT-PCR analysis and western blot. Conclusions Four downregulated miRNAs were shown to be prime candidates for a panel of biomarkers with sufficient diagnostic accuracy for CRC in a clinical setting. Our integrated microRNA profiling approach identified miR-195-5p independently associated with prognosis in CRC. Our results demonstrated that miR-195-5p was a potent suppressor of YAP1, and miR-195-5p-mediated downregulation of YAP1 significantly reduced tumor development in a mouse CRC xenograft model. In the clinic, miR-195-5p can serve as a prognostic marker to predict the outcome of the CRC patients. Electronic supplementary material The online version of this article (doi:10.1186/s13045-017-0445-8) contains supplementary material, which is available to authorized users.


Background
Colorectal cancer (CRC) has been the fourth leading cause of cancer death worldwide for several decades [1]. The 5year survival rate for localized stage CRC is 90%, whereas as CRC spreads to the regional lymph nodes or distant parts of the body, the 5-year survival rate plunges from 71% to 13% in the United States [2]. These malignancies are ascribed to accumulation of genetic alterations, including dissemination of proto-oncogenes, losing or inactivating of tumor suppressor genes, and EMT, which ultimately lead to tumor evolution and progression [3][4][5][6][7].
Using the miRNA expression profile, dysregulated microRNAs (miRNA/miR) in CRC can be identified. Specific miRNAs have been suggested as biomarkers for CRC pathogenesis, metastasis, recurrence, and as potential therapeutic targets [8]. Previous research, by examining a series of CRC microRNAs expression profile data, recommended several deregulated miRNAs, but most failed to identify efficient miRNA biomarkers. Furthermore, few mechanistic studies address the question of how miRNA drives the alterations of the hallmark CRC signaling pathways [9]. For a collective set of deregulated miRNAs, experimental validation is still lacking, and the potential mRNA targets of these miRNAs and their regulatory mechanisms remain unknown. The multicenter sequencing of 1893 carcinoma/normal paired samples revealed that miR-195-5p was a tumor-suppressing miRNA and was downregulated in tumor tissues [10]. Since predicted targets of miR-195-5p were enriched for the Hippo/YAP signaling pathway, we hypothesize that miR-195-5p may function by targeting the YAP1 mRNA. Recent studies showed that YAP and its transcriptional coactivator with PDZ-binding motif (TAZ) function as a nexus on the crosstalk of multiple signaling pathways, and therefore can be potential therapeutic targets as the chief downstream effectors of the Hippo pathway in a variety of cancers [11][12][13][14][15]. It was documented that YAP and TAZ overexpression resulted in a trend of shorter survival time by multivariate analysis in CRC [16]. However, there have been few studies on miRNAs that regulate the oncogenic alterations in the Hippo signaling pathway.
We performed an integrated analysis and identified significantly disturbed miRNAs by using microRNA sequencing data in primary tumors and paired adjacent normal tissue (PANT), both derived from the CRC patients. We further validated aberrant expression patterns of these miRNAs in the TCGA-COAD dataset [17]. We distinguished miR-195-5p as a clinically noteworthy miRNA in CRC as found in lung cancer [18]. We confirmed that the miR-195-5p level was high in patients with favorable survival in TCGA, and may play a critical role in CRC tumorigenesis. Ectopic expression of miR-195-5p could significantly reduce proliferation, migration, invasion, and EMT in two colon cancer cell lines, DLD1 and HCT116. We showed for the first time that the full-length 3′-UTR of the human YAP1 mRNA was a direct target of miR-195-5p. These data offer a plausible mechanism accounting for the tumor-suppressing function of miR-195-5p, further supporting the notion that the miR-195-5p/YAP1/EMT axis can be the therapeutic focus for eliminating CRC tumorigenesis [19].

Methods
Search strategy and data collection, preprocessing, and normalization Public microarray repositories were curated to search through the PubMed, GEO (accession numbers GSE33124, GSE35982, GSE33122, GSE35834, GSE33125, GSE33123, GSE68306, GSE35834, GSE41655, GSE41012, GSE73178, GSE54632, GSE68377, GSE48267, GSE54088, GSE45349, GSE38389, GSE28364, GSE7828, GSE35602, GSE10259 and GSE49246) and Array Express (accession numbers E-MTAB-96, E-GEOD-35834, E-GEOD-28364 and E-MTAB-2479) through September 2015 (the full-detailed search strategy and search terms are shown in Additional file 1: Table S1). The reference sections of the regained studies were analyzed for additional pertinent papers. The checklist and pipeline for properly organizing the integrated analysis were decided as recommended by Ramasamy, via the reporting guidelines of microarray meta-analysis [20]. Only original experimental studies that screened for different miRNAs between CRC and PANT in humans were the first choice for inclusion. Additional criteria for selection dictated that the included datasets should contain at least 30 samples of CRC and PANT correspondingly. Exclusion criteria included: (1) repeated reports by the same institute, hospital; (2) a non-expression gene chip; and (3) a nonwhole-genome chip. All datasets were normalized individually on the base-2 logarithm by Robust Multi-Array Average (RMA) and Linear Models for Microarray (LIMMA) package and annotated by converting different probe IDs to gene IDs. All miRNA names were standardized according to miRBase version 17 via miRBase Tracker and consisted with TCGA for subsequent validation [21]. Any probe that did not map to a gene ID was removed as viral miRNAs or non-miRNA probes.

Integrated analysis of miRNAs expression datasets
To identify differentially expressed (DE) miRNAs between CRC and PANT, we used MetaOmics software (http://www.pitt.edu/~tsengweb/MetaOmicsHome.htm) executed as a MetaDE package [22]. For integrated analysis, the mean and standard deviation (SD) filter thresholds were specified as 10%. Considering the different stringencies of the methods, Fisher's method was performed for statistical analysis of significance: a modified t test and permutation method were used to extrapolate the P values [23]. Onesided tests were applied to classify the upregulated or downregulated DE miRNAs. A P value less than 0.05 was considered statistically significant for the DE miRNAs.

Integrated-signature miRNA analysis of TCGA
The results of integrated analysis of miRNAs expression datasets were validated in the TCGA datasets. TCGA-COAD miRNA data and clinical data (level 3) of the corresponding patients (tumor and/or adjacent normal tissue) were downloaded from the TCGA Data portal (up to May 20, 2016). The expression analyses were carried out using BRB-ArrayTools (version 4.5, National Cancer Institute, Bethesda, MD, USA) [24].

Tissue collection
Sixty human CRC tissues and PANT (distance to cancer > 5 cm) were acquired from patients who had been diagnosed with primary CRC by pathological assessment of tissues and undergone surgeries with complete prognostic information at the Zhongnan Hospital of Wuhan University between January 2011 and December 2015. No local or systemic neoadjuvant radiotherapy, or/and chemotherapy and targeted therapy were managed. The study was approved by the Research Ethics Committee of Wuhan University (Wuhan, Hubei, PR China). Informed consents were obtained from all participating patients.
Cell culture and transfection of miR-195 mimic, inhibitor, and siRNA of target gene Human colon cancer cell lines DLD1 and HCT116, and normal intestinal epithelium cell line NCM460 were purchased from the Cell Bank of Wuhan University. Cells were cultured in DMEM medium (Gibco, USA) containing 2 mmol/L glutamine, 10% heat-inactivated (56°C, 30 min) fetal calf serum, streptomycin (100 U/mL) and penicillin (100 U/mL), and maintained in a humidified atmosphere of 5% CO 2 at 37°C. Hsa-miR-195 mimic and mimic negative control (NC), hsa-miR-195 inhibitor, and inhibitor negative control (NC) were purchased from RiboBio Co., Ltd. (Guangzhou, China). Cells were cultured in complete medium at least 24 h before transfection. Cells were washed by phosphate-buffered saline (PBS, pH 7.4) before transient transfection with 50 nM miR-195-5p mimic or miR mimic NC, 100 nM miR-195-5p inhibitor or miR inhibitor NC. Three siRNA duplexes targeting human YAP1 (GenBank accession no. NM_001130145.2) were synthesized by RiboBio Company (Guangzhou, China). Transfections were performed by Lipofectamine 2000 (Invitrogen, USA) according to the manufacturer's protocol with RNA oligonucleotides at a final concentration of 50 nM.
Total RNA extraction and quantitative reverse transcription polymerase chain reaction (qRT-PCR) Total RNA was extracted using a Trizol reagent (Invitrogen). For the qRT-PCR detection of mature miR-195-5p expression, we purchased the Bulge-Loop™ miRNA qRT-PCR Primer Set and Control Primer Set (RiboBio, Guangzhou, China). RNA (2 μg) was converted into cDNA using the RevertAid First Strand cDNA Synthesis Kit (Thermo). QRT-PCR was accomplished using the FastStart Universal SYBR Green Master (Rox) (Roche) in the ABI PRISM ® 7300 real-time PCR system (Applied Biosystems, Foster City, CA, USA). GADPH and U6 were used as endogenous controls. We used melting curves to monitor non-specific amplifications. Relative expression level was computed using 2 -ΔΔCt method. The primer sequences used are listed in Supplemental Experimental Procedures.
Tumor formation in BALB/c nude mice and immunohistochemistry BALB/c athymic nude mice (female, 4 -6 weeks old and 16 -20 g) were purchased from Hubei Research Center of Laboratory Animals (Wuhan, China). All animal experiments were carried out in accordance with the Guide for the Care and Use of Laboratory Animals of Wuhan University. To establish the CRC cancer xenograft model, 5 × 10 6 DLD1 cells were suspended in 150 μL PBS and inoculated subcutaneously into the flanks of the nude mice. After 8 days, the transplanted nude mice were randomly divided into two groups (n = 6 each). MiR-195-5p agomir or miR agomir NC (RiboBio Co., Ltd, Guangzhou, China) was directly injected into the implanted tumor at the dose of 2 nmol/30μL PBS per mouse every 4 days for seven times. Tumor dimension was measured by length (L) and width (W) with a caliper every 4 days, and the volumes were calculated using the formula: (L × W 2 )/2. Mice were sacrificed by cervical dislocation after being anaesthetized with 10% chloral hydrate at day 36, and the tumors were excised and snap-frozen for protein and RNA extraction. Immunohistochemistry of the tumor tissues was performed as described previously [26].

Statistical analysis
The significance of the differences between the groups was determined with an ANOVA test or Student's t test, and P < 0.05 was considered statistically significant. The results were expressed as the mean ± SD from at least three independent experiments. Kaplan-Meier method and Cox's proportional hazards regression model were used to calculate overall survivals, and the differences were analyzed by a log-rank test. Receiver operating characteristic (ROC) curve analysis was performed using the code of Mihaly (Additional file 2: Supplemental R script) by R Software (version 3.2.3; http:// www.r-project.org) [27].
Supplemental experimental procedures include the following information: dual luciferase reporter assays, BrdU immunofluorescence assay, colony formation assay, CCK8 assay, wound healing assay in vitro, transwell migration/invasion assay, transfection reagents, primers, western blot analysis, and antibody (Additional file 3: Supplemental Experimental Procedures).

Integrated analysis of seven CRC miRNAs expression datasets identified 21 significantly deregulated miRNAs in CRC
We located and manually curated 23 published and publicly available CRC datasets (see Methods; Additional file 4: Table S2). We identified seven datasets that met the criteria listed in Methods (387 CRC samples and 386 PANT samples; Additional file 5: Table S3). The study selection flow chart for this integrated analysis is shown in Fig. 1a. After merging the datasets, 32 differential miRNAs were identified. Twenty-five miRNAs were retained after filtering by the mean and SD (10%) to filter small expression intensities and small variation genes. We identified 21 miRNAs showing consistent DE patterns using a moderated t test by adding a fudging parameter, and Fisher's method by summarizing -log(P value) across studies and running 300 permutations to eliminate significant influence of the large number of samples [28]. One-sided tests revealed 10 upregulated (highly expressed) miRNAs in CRC compared with PANT (Additional file 6: Table S4) and 11 downregulated (lowly expressed) ones in CRC compared with PANT (Additional file 7: Table S5). After combining the effect size, we identified a total of 21 DE miRNAs that showed similar profiles of DE genes to those using Fisher's method of combining P value (P < 1E-19; Additional file 8: Table S6). As expected, hierarchical clustering of the seven datasets using the 11 downregulated miRNAs could distinguish the CRC from the PANT samples (Fig. 1b).
To further investigate whether the deregulated miR-NAs correlate with the survivals of the CRC patients, we performed Kaplan-Meier and Cox's proportional hazards regression model analysis and found that low miR-195 level was significantly correlated with poor overall survivals of CRC patients (Fig. 2c, Table 2), suggesting the prognostic value of miR-195-5p in clinical CRC diagnosis. The downregulation of miR-195 has been consistently observed in the CRC patients with poor overall survivals, indicating that miR-195 may be functionally important in CRC pathogenesis. In contrast, low levels of let-7, miR-125b, or miR-145 in TCGA CRC tissues were correlated with improved overall survivals (Additional file 10: Figure S1b).

Functional enrichment and target prediction of miR-195
Significantly, in spite of the presence of biological and technical confounding factors, such as the differences in sample cohorts, treatments, and microarray technologies, in all eight independent CRC expression microarrays, our results indicated the importance of miR-195 expression in human CRC (Fig. 2d). We conducted target prediction for validated miR-195 with high-stringency. Target genes were obtained from both prediction algorithms and experimentally supported databases. The counts of predicted targets, experimentally validated targets, prediction-based consensus targets, and 71 consensus target genes were summarized (Fig. 2e, Additional file 11: Table S8). In addition, we performed KEGG pathway analysis to elucidate the potential biological functions of miR-195 integratedsignature. Interestingly, the top KEGG pathways enriched for the miR-195 targets were mainly associated with cancer-specific pathways (Additional file 12: Table S9), including the Hippo signaling pathway, proteoglycans in cancer, viral carcinogenesis, pathways in cancer, and prostate cancer (Fig. 2f ). In previous analysis, the Hippo signaling pathway was significantly enriched for the predicted targets of miR-195-5p (P = 6.47E-05) (Additional file 13: Figure S2). Taken together, these results indicate that deregulated miR-195 expression plays a critical role in human CRC.
Downregulation of miR-195-5p is concomitant with upregulation of YAP1 in primary human CRC As miR-195 showed a consistent pattern of downexpression in human CRC, we wondered about the expression pattern of miR-195-5p in the colon cancer cell lines. Similarly, we found a lower expression of miR-195-5p in DLD1 and HCT116 cell lines, compared with the normal intestinal epithelium cell line NCM460 (Fig. 3a). Next, we measured the mature miR-195-5p level in an independent validation sample cohort (n = 60) by qRT-PCR (normalized with U6). The results showed that miR-195-5p expression in the tumors was significantly (P < 0.05) reduced in the tumors relative to PANT (Fig. 3b). Additionally, Kaplan-Meier and Cox's proportional hazards regression model survival analysis revealed that patients with low expression levels of miR-195-5p had shorter overall survival (Fig. 3c, Table 2). Thus, given the results of the integrated analysis and the in vitro experiments, we hypothesize that the decreased expression of miR-195-5p may promote CRC progression and development.
Since the Hippo pathway emerged in our previous KEGG analysis for the miR-195 targets, it is conceivable that miR-195 may directly regulate this pathway during CRC tumorigenesis. Meanwhile, the protein-protein interaction network (PPI) suggested that YAP1 is the hub gene of the miR-195 targets (Additional file 14: Figure S3). YAP1 is the pivot in the Hippo signaling pathway leading to potent oncogenicity, and can promote cell growth, migration, invasion, and EMT in many tumors including CRC. Therefore, we also examined YAP1 expression in this sample cohort comprising the 60 human CRC and PANT. Both qRT-PCR and western blot results showed that YAP1 was significantly upregulated in CRC compared with PANT (Fig. 3d). Significantly, the miR-195-5p level was inversely correlated with the YAP1 level in the tumors as calculated by Pearson's correlation (R 2 = 0.531, P = 4.02E − 11) (Fig. 3e), suggesting that miR-195-5p may downregulate YAP1 in CRC (Fig. 3e). Of note, Kaplan-Meier and Cox's proportional hazards regression model also revealed that patients with high levels of YAP1 had shorter overall survivals (Fig. 3f, Table 2).
(See figure on previous page.) Fig. 1 Eleven downregulated genes were identified by integrated analysis of CRC miRNA expression datasets. a The flowchart of the integrated analysis and functional validation. In silico bioinformatics data analysis pipeline consists of curation of seven publically available datasets, data preprocessing, metaDE and integrated analysis, TCGA dataset validation, validation for diagnostic/prognostic values, KEGG pathway and target gene validation, and in vitro and in vivo functional validation. b Clustering of the 11 genes in CRC vs. PANT across each independent dataset. Each column represents a sample and each row represents the expression level of a given miRNA. The color scale represents the raw Z score ranging from blue (low expression) to red (high expression). Dendrograms by each heatmap correspond to the hierarchical clustering by expression of the 11 miRNAs. c Expression of the downregulated miRNAs is plotted for tumor and normal tissues (TCGA dataset). Expression values of miRNAs are log 2 -transformed MiR-195-5p inhibits proliferation and colony formation of colon cancer cells We next examined the pathophysiological significance of miR-195-5p downregulation in CRC and its underlying regulatory mechanisms. We transfected DLD1 and HCT116 cells with miR-195-5p mimic or miR mimic NC, and miR-195-5p inhibitor or miR-195-5p inhibitor NC, separately. We first examined the effect of miR-195-5p on proliferation of DLD1 and HCT116 cells. BrdU incorporation assay revealed that miR-195-5p inhibited DNA synthesis in DLD1 and HCT116 cells (Fig. 4a-b). In contrast, miR-195-5p inhibitor treatment could lift this inhibition ( Fig. 4a-b). In addition, clonogenic assay showed that miR-195-5p mimic treatment decreased the clonogenic survivals of DLD1 and HCT116 cells compared with blank controls, while miR-195-5p inhibitor-treated DLD1 and HCT116 cells showed a reversed phenotype (Fig. 4c-d), suggesting miR-195-5p negatively regulates cancer cell proliferation. Supporting this notion, a colorimetric based cellular proliferation assay (i.e., CCK8 assay) showed consistent phenotypes when treating with miR-195-5p mimic or its inhibitor (Fig. 4e-f).

MiR-195-5p inhibits migration and invasion of colon cancer cells
Invasion and migration through the basement membrane are characteristics of metastatic cancer cells. We assessed the role of miR-195-5p on the migration and invasion of DLD1 and HCT116 cells. In the scratch wound healing assay, cell motility of cells was monitored at different time points after generation of the wound. The miR-195-5p-expressing cells migrated toward the wound at a much slower rate than the control or the cells treated by miR-195 inhibitors (Fig. 5a-b). In the transwell invasion and migration assay, we found that invasion and migration of the miR-195-5p-expressing cells was reduced, and this effect could be reversed by the miR-195-5p inhibitor (Fig. 5c-f). These results, taken together, clearly demonstrate that miR-195-5p negatively regulates invasion and migration of colon cancer cells.
To confirm the direct regulation of miR-195-5p on YAP1 expression, we transfected DLD1 and HCT116 cells with miR-195-5p mimic, miR mimic NC, miR-195-5p inhibitor, and miR-195-5p inhibitor NC. Both qRT-PCR and western blotting revealed that the YAP1 level was reduced in miR-195-5p-expressing cells, while its level was restored in miR-195-5p inhibitor-treated cells (Fig. 6d-f). The conversion from epithelial cells to mesenchymal cells, characterized by spindle-type cell morphology, was observed in HCT116 cells treated with miR-195-5p inhibitor (Additional file 15: Figure S4). So far, the results suggest that miR-195-5p negatively regulates YAP1 levels in CRC cells. Additionally, expressions of TAZ, Vimentin, ZEB2, and SMAD3 protein were also negatively regulated by miR-195-5p, while E-cadherin was positively regulated by miR-195-5p, suggesting the negative regulations of miR-195-5p on the Hippo pathway and EMT in CRC cells. Multivariate analysis used stepwise addition and removal of clinical covariates found to be associated with survival in univariate models (P < 0.05) and final models include only those covariates that were significantly associated with survival (Wald statistic, P < 0.05). Bold italics indicate statistically significant values (P < 0.05) Silence of YAP1 expression inhibits colon cancer cell growth, clone formation, invasion, migration, and rescue assay We next evaluated the potential tumorigenicity of YAP1 in CRC. Silence of YAP1 expression by siRNA significantly inhibited the expression of YAP (Additional file 16: Figure S5). Moreover, loss of YAP1 expression also contributed to inhibition of colon cancer cell (DLD1 and HCT116 cells) growth (Additional file 16: Figure S5a), clone formation (Additional file 16: Figure S5b-c), invasion, and migration (Additional file 16: Figure S5d-e). These results further verified the powerful tumorigenicity of YAP1 in CRC. These results indicated that the anti-cancer and reverse EMT efficacy of miR-195-5p is partly attributed to its inhibitory role on YAP, which was confirmed by qRT-PCR and western blot of YAP1 (Additional file 16: Figure S5f-g) in DLD1 cells.

MiR-195-5p suppresses tumor growth in vivo
To confirm the tumor suppressor role of miR-195-5p in vivo, we established a BALB/c nude mouse xenograft model using DLD1 cells. Starting day 8 post-implantation, we injected miR-195-5p agomir or miR agomir NC intratumorally every 4 days for 7 treatments. We evaluated in vivo tumor growth by measuring the tumor volumes and weights. Both the volumes and weights of the tumors treated with miR-195-5p agomir were significantly reduced relative to those treated with miR agomir NC (Fig. 7a-d). Therefore, miR-195-5p significantly inhibits the tumorigenicity of DLD1 cells in vivo. Additionally, consistent with the proposed regulations of miR-195-5p on the Hippo pathway and EMT, expression patterns of YAP, Ki67, Vimentin, ZEB2, and E-cadherin in miR-195-5p-agomir-treated tumors were similar to the in vitro results, substantiating the tumor suppressor function of miR-195-5p in CRC tumorigenesis (Fig. 7e-g).

Discussion
Integrated analysis, a conscientious approach to unify cross-platform standardization of gene profiling and overlook the unavailability of raw data, has been used to identify differentially expressed genes at the mRNA and miRNA level in CRC [29,30]. Chen et al. overcame the drawback of lack of conformity among miRNA expression profiling studies in CRC by means of directly combining the ranks of prioritized miRNA lists but not re-analyzing the raw miR-NAs profiles [31]. However, they failed to consider the effects of small sample size and heterogeneity of the individual studies because they included 17 studies whose sample size was under 30 pairs in total 24 studies. Conversely, our integrated analysis of the raw data of the seven studies whose sample size is over 30 pairs reliably reveals the differential miRNAs across various platforms.
Our current study supports the validity and effectiveness of this integrated strategy. Firstly, reduced expression of miR-195-5p in primary CRC was revealed independently by both integrated analysis and TCGA-COAD validation analysis, suggesting miR-195-5p can be an effective diagnostic and prognostic marker in the clinical setting. Secondly, ectopic expression of miR-195-5p in colon cancer cell lines considerably decreased cell growth, migration, and invasion. Thirdly, miR-195-5p can directly regulate expression of YAP1 by targeting its 3′-UTR. It can inhibit expression of YAP1, TAZ, Vimentin, ZEB2, and SMAD3, The goals of the current study were to identify the candidate driver miRNAs in CRC through meta-analysis of multiple datasets of miRNA expression arrays and to validate their biological functions and significance for clinical diagnosis and prognostic prediction. The integrated analysis method described here was designed to exclude the low precision studies, account for biases inherent in single studies, and nominate prevalently dysregulated miRNAs. Taking into account these considerations, we have included seven miRNAs expression data from seven different platforms in CRC using the MetaDE method, which provided options for gene merging, matching across studies, and gene filtering [22]. We screened 10 upregulated and 11 downregulated miRNAs in CRC samples using moderated t test and Fisher's method by summarizing -log(P value). In addition, we validated and obtained 4 downregulated miRNAs (i.e., hsa-let-7a, hsa-miR-125b, hsa-miR-145, and hsa-miR-195) from over 250 tumors from the TCGA-COAD database which granted us the opportunity to systematically analyze the potential molecular mechanisms associated with the pathophysiology of CRC. These four miRNAs seem to have a contextspecific value as a diagnostic marker of CRC. For hsa-let-7a, hsa-miR-125b, and hsa-miR-145, previous reports have shown their downregulation was correlated with the antitumor effect in CRC [31,32]. In this study, we found that hsa-let-7a, hsa-miR-125b, and hsa-miR-145 were downregulated in CRC and correlated with improved survivals of CRC patients. Only the low level of miR-195 in tumors was substantially correlated with reduced overall survivals of CRC patients, suggesting miR-195 may negatively regulate CRC progression.
MiR-195 as a tumor suppressor has been reported in various types of cancer. It is a member of paralogous, evolutionarily conserved miRNAs termed the miR-15/ 107 family that has been suggested to have considerable potential in prognostic prediction [33][34][35][36][37][38]. Soon   f Total number of cells in five fields was counted manually. Mean ± SD are shown. Statistical analysis was conducted using one-way ANOVA. *P < 0.05. **P < 0.01. ***P < 0.001 poorer prognosis in adrenocortical carcinoma [39]. Li et al. found that extracellular vesicles delivering miR-195 to intrahepatic cholangiocarcinoma could decrease the tumor size and improve survivals in a rat model [40]. Luo et al. found miR-195 might tender a novel tactic for the diagnosis and treatment of breast cancer patients [41]. In our study, we confirmed that miR-195 was downregulated in CRC tissues, which was significantly correlated with poor survivals of CRC patients. We investigated the mechanism of miR-195-5p-mediated regulations on colon cancer cells and found that overexpression of miR-195-5p markedly inhibited cellular proliferation and invasion of CRC cells. Supported by miRNA target analysis and luciferase reporter assays, we propose that miR-195-5p targets the 3′-UTR of YAP1 mRNA and inhibits YAP expression, leading to proliferative inhibition in CRC cells. Furthermore, the present study also shed light on the potential role of miR-195-5p in CRC metastasis, where ectopic expression of miR-195-5p could reduce cell migration and invasion and The 3′-UTR of YAP1 mRNA harbors two miR-195-5p cognate sites. c Relative luciferase activity of reporter plasmids carrying wild-type or mutant YAP1 3′-UTR in DLD1 and HCT116 cells co-transfected with negative control (NC) or miR-195-5p mimic or miR-195-5p inhibitor. Expression of YAP1 mRNA in DLD1 (d) and HCT116 cells (e) transfected with the indicated miRNAs. f Western blots of YAP, TAZ, Vimentin, ZEB2, SMAD3, and Ecadherin in DLD1 and HCT116 cells transfected with different miRNAs. Assays were performed in triplicates. Mean ± SD is shown. Statistical analysis was conducted using Student's t test. *P < 0.05. **P < 0.01 promote expression of the EMT marker, E-cadherin. To our knowledge, this is the first meta-analysis that reveals the detailed mechanism where loss of miR-195-5p leads to malignant progression of CRC via unleashed expression of YAP1. It is well known that Hippo signaling controls cell growth and differentiation [42]. In several tumors, aberrant activation of the Hippo/YAP signaling pathway represents one of the most important mechanisms accounting for oncogenic progression and invasiveness, and has been recognized as a prognostic factor of poor outcome [11,[43][44][45][46][47][48][49][50][51][52]. Major signal transducers of the Hippo pathway were proposed as prognostic factors in colorectal cancer [53,54]. We also identified the Hippo signaling pathway as being enriched in the prediction analysis of the miR-195-5p targets. Experimentally, we verified that the Hippo pathway core effector YAP1 mRNA 3′-UTR had two conserved miR-195-5p cognate sites. Interestingly, we also found that loss of miR-195-5p accelerated YAP expression and promoted nuclear accumulation of YAP and EMT in vitro [19,55,56]. Hence, major players in this pathway may be tightly regulated by miR-195-5p in the colorectum, and miR-195-5p restoration could target YAP/TAZ/ EMT pathway to suppress CRC progression. Furthermore, it is likely that other molecules or signaling pathways will be discovered that are also targeted by miR-195-5p in CRC. Future work will focus on revealing additional functions of miR-195-5p in CRC carcinogenesis and progression. Fig. 7 Ectopic expression of miR-195-5p suppresses tumor growth in vivo. The volumes (a) and weights (b) of the tumors grown in the xenograft mouse model. Data are presented as mean ± SD (n = 6); *P < 0.05. Photographs of the mice (c) and dissected tumors (d) that were treated with or without Agomir-195-5p. e Expression of YAP1 mRNA in DLD1 tumors. f Western blots of YAP, Ki67, E-cadherin, vimentin, and ZEB2 in the tumors. β-actin was used as a loading control. g Immunohistochemistry of tumor tissues treated with or without miR-195-5p. Assays were performed in triplicates. Mean ± SD is shown. Statistical analysis was conducted using Student's t test.