Identification of miRNAs as diagnostic and prognostic markers in hepatocellular carcinoma

The development of high-throughput technologies has yielded a large amount of data from molecular and epigenetic analysis that could be useful for identifying novel biomarkers of cancers. We analyzed Gene Expression Omnibus (GEO) DataSet micro–ribonucleic acid (miRNA) profiling datasets to identify miRNAs that could have value as diagnostic and prognostic biomarkers in hepatocellular carcinoma (HCC). We adopted several computing methods to identify the functional roles of these miRNAs. Ultimately, via integrated analysis of three GEO DataSets, three differential miRNAs were identified as valuable markers in HCC. Combining the results of receiver operating characteristic (ROC) analyses and Kaplan–Meier Plotter (KM) survival analyses, we identified hsa-let-7e as a novel potential biomarker for HCC diagnosis and prognosis. Then, we found via quantitative reverse-transcription polymerase chain reaction (RT-qPCR) that let-7e was upregulated in HCC tissues and that such upregulation was significantly associated with poor prognosis in HCC. The results of functional analysis indicated that upregulated let-7e promoted tumor cell growth and proliferation. Additionally, via mechanistic analysis, we found that let-7e could regulate mitochondrial apoptosis and autophagy to adjust and control cancer cell proliferation. Therefore, the integrated results of our bioinformatics analyses of both clinical and experimental data showed that let-7e was a novel biomarker for HCC diagnosis and prognosis and might be a new treatment target.


INTRODUCTION
AGING Micro-ribonucleic acids (miRNAs) are short, singlestranded, noncoding RNAs (ncRNAs) [19][20][21][22][23][24][25] nucleotides long that bind to the 3′-untranslated region (UTR) of messenger RNAs (mRNAs), resulting in degradation of the target mRNA molecules or translational inhibition [12]. MiRNAs are multifunctional molecules participating in cell development, differentiation and aging [13][14][15]. However, the miRNA profiles of malignancies are significantly different from those of normal tissues, making them potentially appealing biomarkers in the diagnosis, treatment and prognosis of various cancers [12,16,17]. Recently, the role of miRNAs in tumorigenesis and tumor progression has attracted much attention. A growing amount of evidence has proven the vital role of dysregulated miRNAs in cancer diagnosis and prognosis. Certain miRNAs are significantly correlated with the presence of tumors, even in the early stages, or with worse prognosis [18,19]. In this study, we focused on finding miRNAs that could have value as diagnostic and prognostic biomarkers in HCC.
With the development of high-throughput technologies, a large amount of data has been generated from molecular and epigenetic analyses. The relevant databases could be used to identify novel biomarkers of cancers via computational approaches [20]. Herein, we analyzed Gene Expression Omnibus (GEO) DataSet miRNA profiling datasets to identify miRNAs with potential diagnostic-and prognostic-biomarker value in HCC. Then, we verified the prognostic values of these miRNAs using the Kaplan-Meier (KM) Plotter (http://www.kmplot.com/analysis/). Further analyses of these miRNAs included Gene Ontology (GO) analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) signaling pathway analysis. Finally, we identified the most valuable miRNA, hsa-let-7e. Note, however, that bioinformatics data analyses based on the abovementioned databases often produce conflicting results [21]. Therefore, we performed studies to identify the clinical value, biological function (BF) and molecular mechanism (MM) of let-7e. Differences in let-7e expression in HCC tissues were verified and their clinical relevance analyzed. Finally, via in vitro experiments, we demonstrated the potential BF and MM of let-7e in HCC. Our results suggested that let-7e might be a novel biomarker for diagnosis, treatment and prognosis in HCC.

Identification of significantly dysregulated miRNAs and their diagnostic and prognostic values in HCC tissues
We performed a comprehensive differential analysis of miRNA expression based on three GEO DataSets (GSE6857, GSE22058 and GSE12264) in order to identify dysregulated miRNAs in HCC tissue compared with non-tumor normal controls (NCs). Datasets GSE6857, GSE22058 and GSE12264, respectively, contained 124, 146 and 966 miRNAs. By merging the three datasets, we were able to identify three consistently dysregulated miRNAs (hsa-let-7b, hsa-let-7c and hsa-let-7e) in HCC tissue versus NCs ( Figure  1A). The expression profiles of these distinct miRNAs are shown in the heatmaps in Figure 1B. heatmaps. Considering the widely differing expression levels of hsa-let-7b, hsa-let-7c and/or hsa-let-7e between cancerous and non-tumor tissues, we selected them for further investigation of whether, when dysregulated, they could serve as diagnostic and prognostic markers in HCC.
Next, we used quantitative reverse-transcription polymerase chain reaction (RT-qPCR) to measure differences in these three miRNAs' expression levels between HCC tissue specimens and matched non-tumor tissue specimens. The results showed that let-7b (P=0.038; 45 pairs of specimens) and let-7e (P=0.016; 63 pairs) were upregulated in HCC tissue compared with NCs. However, expression of let-7c (P=0.162; 43 pairs) did not differ between HCC and normal tissues ( Figure 1C).

Functional enrichment analysis
Using a functional-enrichment analysis tool (FunRich; http://www.funrich.org), we analyzed the most relevant neighboring and cross-linked target genes of the three miRNAs. The result was a network (Figure 2A), which AGING was composed of these miRNAs and of 52 genes. Next, we analyzed the relevant transcription factors (TFs) of let-7b, let-7c and let-7e using FunRich. The top-10 relevant enriched TFs were Achaete-Scute family basic helix-loop-helix (BHLH) transcription factor 2 (ASCL2), early growth response 1 (EGR1), specificity protein 1 (SP1), homeobox A7 (HOXA7), transcription factor 3 (TCF3), myelocytomatosis oncogene (MYC), neurofibromin 1C (NF1C), visual system homeobox 2 (VSX2), SP4 and LIM homeobox 3 (LHX3) ( Figure 2B). Then, we measured their expression levels in HCC specimens and matched nontumor specimens using RT-qPCR. Results are shown in Figure 2C.  The target gene modules in the network were often enriched with diverse specific functions that had biological significance. In order to identify the BFs of the three miRNAs' target genes, we performed GO analysis based on information from selected genes. Because miRNAs can promote degradation of miRNAs by binding to mRNA, it is helpful to distinguish up and downregulated mRNA groups for further research and analysis into the mechanism of miRNAs. First, we included the genes negatively correlated with the miRNAs in our analysis. The results showed that these genes were enriched in the regulation of mitogenactivated protein kinase (MAPK) activity and anion transmembrane transport in biological processes (BP) ( Figure 2D). Then, we included the genes positively correlated with the miRNAs in our GO analysis. Our results indicated that three GO terms of BP, two GO terms of Cellular Component (CC) and three GO terms of Molecular Function (MF) were enriched; we identified these as significant ( Figure 2E). In addition, we performed KEGG pathway enrichment analysis to identify potential pathways of the genes positively ( Figure 2F) and negatively ( Figure 2G) correlated with miRNAs.

High expression of let-7e was associated with poor clinical outcomes in HCC
The above results showed that let-7e could be the most significant biomarker for diagnosis and prognosis in HCC; it was upregulated in primary HCC tissues compared with adjacent normal tissues in 63 patients. Next, we explored the relationship between let-7e expression and clinicopathological parameters, using median expression level as the cutoff value between the high-and low-expression groups. Results are summarized in Table 2. High let-7e expression was significantly associated with poor tumor differentiation (P=0.014), larger tumor size (P=0.029) and venous invasion (P=0.031).
To explore whether expression of let-7e was associated with prognosis in HCC patients, we used KM survival analyses to compare the overall survival (OS) and disease-free survival (DFS) rates between the high-and low-expression groups. Results revealed that median OS was 23 months in the high-let-7e expression group versus 38 months in the low-let-7e expression group. The high-let-7e expression group had a significantly worse OS rate than the low-let-7e expression group (P=0.046, log-rank test; Figure 3A). The median DFS rates of patients with high and low let-7e expression were 19 and 28 months, respectively. Compared with the low-let-7e expression group, the high-let-7e expression group had a significantly worse DFS rate (P=0.030, log-rank test; Figure 3B). Therefore, high let-7e expression was significantly associated with poor prognosis in HCC patients.

Verification of let-7e expression in transfected HCC cells
To investigate the BF of let-7e, we transfected let-7e mimic and let-7e inhibitor into HepG2 and Hep3B cell lines, respectively, to construct let-7e-expressing and let-7e-inhibited cell lines. We also transfected control mimic and inhibitor to construct NC cells for these two HCC cell lines. Before conducting experiments to determine the BF of let-7e, we confirmed the expression thereof in the transfected cell lines via RT-qPCR ( Figure 4).
Expression of let-7e was significantly upregulated in HCC cells transfected with let-7e mimic than in those transfected with control mimic ( Figure 4A). Conversely, let-7e expression was significantly downregulated in HCC cells transfected with let-7e inhibitor than in those transfected with control inhibitor ( Figure 4B).
Using these transfected HCC cells, we conducted the following experiments to determine the BF of let-7e.

Expression of let-7e affected the growth and proliferation of HCC cells
To explore whether high let-7e expression affected the growth of HCC cells, we performed cell viability and colony formation assays in HepG2 and Hep3B HCC cells. Cell viability assay results showed that  upregulated let-7e promoted the proliferation of HCC cells ( Figure 5A), while the colony formation assays revealed that upregulated let-7e improved the cells' colony formation efficiency ( Figure 5B). To prove whether downregulated let-7e could inhibit the growth of HCC cells, we transfected HepG2 and Hep3B HCC cells with let-7e inhibitor and NC inhibitor and then performed cell viability and colony formation assays. The results showed that downregulated let-7e inhibited cell proliferation of HCC cells in cell viability assays AGING ( Figure 5C). The colony formation assays showed that downregulated let-7e inhibited the colony formation efficiency of these cells ( Figure 5D).

Expression of let-7e affected apoptosis of HCC cells
To explore whether let-7e's promotion of HCC cell growth and proliferation was relevant to cell apoptosis, we analyzed apoptosis via flow cytometry (FCM). Results showed that the percentage of apoptotic cells was significantly lower in let-7e mimic-transfected cells but higher in let-7e inhibitor-transfected cells, which indicated that let-7e expression affected apoptosis in HCC cells ( Figure 6A, 6B).

Expression of let-7e affected apoptosis and autophagy of HCC cells
To further explore whether the functional mechanism of cell growth and proliferation affected by let-7e was  AGING related to mitochondrial apoptosis, we performed Western blot (WB) analysis to detect the expression of apoptosis-related proteins in the HCC cell lines. Results suggested that the levels of pro-apoptotic protein Bax, Bak, Bad, cleaved Caspase-9 and cleaved Caspase-3 were decreased, while that of the anti-apoptotic protein Bcl-2 was increased, in let-7e mimic-transfected HCC cells ( Figure 6C, 6D).
Autophagy is a conserved cellular process considered to be associated with tumor progression, particularly in cell apoptosis [22,23]. To explore the function of let-7e in autophagy of HCC cells, we detected the expression of autophagy-related proteins using WB analysis. The results suggested that expression levels of LC3, Atg4A, Atg5, Atg 16L and p53 were decreased but that of p62 was increased in let-7e mimic-transfected HCC cells ( Figure 6E, 6F).
Taken together, our results indicated that let-7e might suppress cell apoptosis and autophagy via the p53 pathway, which induced the growth and proliferation of HCC cells.

DISCUSSION
HCC is still one of the most malignant tumors, with increasing incidence and a high mortality rate [2,24]. Though a variety of therapeutic methods have been developed, the recurrence and metastasis rates of HCC remain stubbornly high, leading to unsatisfactory and poor prognoses [25]. Because of the disease's complicated pathogenesis and molecular mechanism [26,27], finding an effective treatment is still a worldwide problem. Therefore, it is imperative to identify underlying molecular mechanisms and valuable biomarkers of HCC.
With the development of high-throughput technologies and bioinformatics analysis, omics sciences have been promoted over the last decade, and petabytes of molecular data on various human diseases have been collected [28]. This ever-growing amount of bioinformatics data contributes to our understanding of the physiopathological aspects of diverse diseases, including cancers [29]. In the present study, we aimed to identify novel miRNAs that could serve as biomarkers for diagnosis and prognosis in HCC, as well as to understand the MMs of HCC by analyzing and integrating several miRNA profiling datasets using several computational approaches. However, the huge quantity of data from the relevant databases is often not completely accurate, which generates conflicting results [21]. Therefore, after selecting the miRNAs most likely to be valuable from the results of our bioinformatics analyses, we explored their probable clinical values, BFs and MMs in HCC via analysis of clinical data and our own experiments.
MiRNAs, which are short, single-stranded ncRNAs 19-25 nucleotides long, are pivotal in the cellular processes of development, differentiation and aging [12][13][14][15]. In the current study, we first analyzed and integrated the GEO DataSets GSE6857, GSE22058 and GSE12264 to identify miRNA expression profiling datasets. By analyzing the differing expression levels of miRNA profiles, as well as performing ROC analyses of the GEO DataSets and survival analyses using the KM Plotter datasets, we identified three dysregulated miRNAs for further analyses: let-7b, let-7c and let-7e.
The let-7 family contains 13 members encoding nine mature miRNAs, including let-7a, let-7b, let-7c, let-7d, let-7e, let-7f, let-7g, let-7i and miR-98 [30]. The multiple functions of this family are extensively pleiotropic including oncogenic behavior; repression of oncogenes; and regulation of signaling pathways, cell cycle, apoptosis, epithelial-mesenchymal transition (EMT) and chemosensitivity in cells . In previous studies, let-7b expression has been associated with prognosis in hepatoblastoma, HCC, melanoma and prostate cancer [32,34,58,59]. Let-7c expression was related to the development of acute promyelocytic leukemia; HCC; and prostate, lung and endometrial cancers [34,43,[60][61][62][63]. Other studies have found let-7e expression to be associated with melanoma; endometrial, prostate and ovarian cancers; and esophageal carcinoma [58,59,[63][64][65]. The above referenced studies revealed that let-7b, let-7c and let-7e were multifunctional miRNAs involved in the development of various cancers. In the present study, we analyzed their most relevant neighboring and crosslinked genes using a functional-enrichment analysis tool. We also analyzed the top 10 relevant enriched TFs, which were ASCL2, EGR1, SP1, HOXA7, TCF3, MYC, NF1C, VSX2, SP4 and LHX3. These TFs participate in the occurrence and development of diverse cancers [66][67][68][69][70][71][72][73][74][75][76][77]. The selected miRNAs might play roles in HCC progression by influencing levels of these TFs. Cancer is regarded as a disease of communication between and within cells [78]. To explore how the selected miRNAs might affect progression of HCC via their BFs and via signaling pathways, we analyzed the target genes of the three miRNAs using GO and KEGG analyses. BF and pathway enrichment analysis results showed that these miRNAs regulated the phosphatidylinositol-4,5bisphosphate 3-kinase (PI3K-)-protein kinase B (Akt signaling pathway), p53 and MAPK signaling pathways, all of which are considered classic signaling pathways involved in the development of cancers [78][79][80]. P53 is downstream protein affected by Akt, a AGING central protein of the PI3K-Akt pathway; it mediates many signaling pathways in cancers and generates multifarious biological responses [81]. In addition, p53 can functionally interact with components of the MAPK signaling pathway, including p38 MAPK, extracellularsignal-related kinase (ERK) and stress-activated protein kinase (SAPK)-c-Jun N-terminal protein kinase (JNK) [82]. Therefore, we supposed that the selected miRNAs AGING might play roles in HCC progression by affecting these interactional signaling pathways.
From the comprehensive results of above analyses, we supposed that let-7e was likely to play the most significant role in the diagnoses and prognoses of HCC patients. However, due to discrepancies in the relevant databases, bioinformatics data analyses often produce conflicting results [21]. Therefore, to identify the clinical value, BF and MM of let-7e, we based this study on our own data. First, we explored the association between let-7e expression and clinicopathological parameters. Our results suggested that high let-7e expression was significantly associated with poor tumor differentiation, larger tumor size and venous invasion. The results of KM survival analyses indicated that high let-7e expression was correlated with worse OS and DFS rates in HCC. Next, we performed cell viability and colony formation assays to identify the BF of let-7e in HCC cells. We found that upregulated let-7e promoted cell growth and colony formation efficiency in HCC cells, which indicated that this miRNA might act as an oncomiR.
MiRNAs promote cellular proliferation through mechanisms such as suppression of apoptosis and autophagy [83,84]. In our study, we found that upregulated let-7e promoted cellular proliferation by suppressing apoptosis and autophagy in HCC cells. These are both highly conserved processes participating in cellular proliferation, death and homeostasis, and dysfunctions in either process can result in various human diseases [85,86].
Autophagy usually occurs at the contact sites between the endoplasmic reticulum and mitochondria with the formation of the isolation membrane/phagophore. This membrane/phagophore contains macroproteins or even whole organelles which are sequestered into lysosomes for degradation [90]. Autophagy is a highly conserved process relying on the function of a core set of ATGs [91]. A series of signaling pathways initiate or regulate autophagy cascades, including the adenosine monophosphate-activated protein kinase (AMPK-)mammalian target of rapamycin complex 1 (mTORC1), class I PI3K, Akt-mTOR, Ras-rapidly accelerated fibrosarcoma (Raf-1-)-mitogen-activated protein kinase 1/2 (MEK1/2-)-extracellular signal-regulated kinase 1/2 (ERK1/2,) and p53 signaling pathways [92]. Recent studies have found autophagy to be an upstream initiator of apoptosis and to regulate cell apoptosis by modulating Caspase and Bcl-2 family proteins [93]. Autophagy might be a guardian or executioner of apoptosis, depending on the surrounding microenvironment, therapeutic intervention and stage of carcinoma [93]. Therefore, in this study, we also detected the expression of autophagy-related proteins to explore the effect of let-7e on autophagy in HCC cells.
We found that upregulated let-7e decreased the expression levels of LC3, Atg 4A, Atg5, Atg 16L and p53, but increased that of p62. From these results, we supposed that let-7e might suppress cell autophagy and apoptosis via the p53 signaling pathway, inducing the growth and proliferation of HCC cells.
In conclusion, let-7e was associated with poor prognosis in HCC patients and acted as an oncogene by suppressing autophagy and apoptosis in HCC cells, suggesting that it could be a novel biomarker for prognosis and target of treatment in HCC.

Analysis of GEO databases
To identify miRNAs potentially involved in the development and progression of HCC, we reviewed the National Center for Biotechnology Information (NCBI; Bethesda, MD, USA) GEO databases. To find miRNA profiles significantly dysregulated in HCC, we analyzed three GEO DataSets: GSE6857, GSE22058 and GSE12264. Differences in miRNA expression between the tumor group (HCC) and normal group of the three datasets were compared using Student's t test. P < 0.05 and fold change ≥ 1.5 were considered statistically significant. Then, we conducted Venn selections of differentially expressed miRNAs among the three lists using the online tool Venny version 2.1.0 (http://bioinfogp.cnb.csic.es/tools/ venny/).

AGING
Heatmaps of differential miRNA expression between both groups in all three datasets were generated using the pheatmap package in R software (Ihaka and Gentleman, 1999). P < 0.05 was considered statistically significant.

Diagnostic prediction of selected miRNAs
We evaluated the predictive power of the three selected miRNAs in HCC diagnosis using ROC curves. Clinical data were derived from GEO database GSE12264.

Kaplan-Meier plotter
To evaluate the prognostic value of extracted-miRNA expression in liver cancers, we performed analyses using the KM Plotter. Cancer patients were divided into two groups, high and low miRNA expression, according to median values of miRNA expression; and KM survival curves were drawn in the plotter. P < 0.05 was considered statistically significant.

Prediction of target genes of miRNAs
We analyzed the predictive power of target genes of the selected miRNAs and discovered the top 10 enriched TFs using a functional-enrichment analysis tool (FunRich).
We also used FunRich to analyze the interaction network between the selected miRNAs and their target genes.

Enrichment analysis of target genes
Using ClusterProfiler software version 3.11 (https://bio conductor.org/packages/release/bioc/html/clusterProfile r.html) [94,95], we conducted functional-enrichment analysis of the predicted target genes of the selected miRNAs. GO analysis was conducted to describe functions of predicted genes, including BP, CC and MF. KEGG was used for pathway enrichment analysis based on significance at P < 0.05.

Patients and tissue samples
We

Cell transfection
We purchased let-7e mimic and let-7e inhibitor from Ribobio. After being cultured in a six-well culture dish for 24 h until the degree of cell fusion reached about ~70%, the HCC cell lines HepG2 and Hep3B were transfected with let-7e mimic or inhibitor and normal control (NC) mimic or inhibitor via transfection reagents (riboFECT CP Transfection Kit; Ribobio). The transfected concentration of mimic was 50 nmol, and that of inhibitor was 100 nmol.

Cell viability assay
We used a Cell Counting Kit-8 (CCK-8; Dojindo Molecular Technologies, Inc., Kumamoto, Japan) to evaluate transfected cells' viability. Cells were dispersed into and cultured in 96-well plates (2500 cells/well). To each well, we added 10μL CCK-8 reagent at the indicated time point. After incubating the transfected cells for 1 h at 37° C, we measured the absorbance of each well at 450 nm.

Colony formation assay
Transfected HCC cell lines HepG2 and Hep3B were seeded into 6-well plates (1000 cells/well) for colony formation assays. Fourteen days later, we used 70% ethanol to immobilize the cell colonies for 15 min, after which we stained them for 10 min using 0.1% crystal violet. Finally, we counted the colonies and used this number to evaluate the transfected HCC cells' colony formation capacity. Colony count was based on three different experiments and colonies in each well were manually counted three times.

Cell apoptosis analysis
Cell apoptosis was measured using an Annexin V-FITC/PI apoptosis assay kit (Sangon Biotech, Shanghai, China). We analyzed the percentage of apoptosis on a CytoFlex flow cytometer (Beckman Coulter Life Sciences, Brea, CA, USA) per manufacturer's instructions.

Western blot
RIPA buffer (Pierce, Rockford, IL, USA) and BCA Protein Assay Kit (Pierce) were respectively used to extract proteins from cells and determine protein amounts. We separated the same amount of protein (20μg) via 6-12% SDS-PAGE. Then, the protein was transferred to PDVF membranes (MilliporeSigma, Burlington, MA, USA), which we blocked in 5% nonfat milk for 1 h. Next, we incubated the membranes overnight at 4° C with the following indicated primary antibodies, all of which were purchased from Cell Signaling Technology

Statistical analysis
The relationship between HCC patients' clinicopathological characteristics and let-7e expression was analyzed via χ 2 test. We used the KM method with a log-rank test for analysis of OS and DFS rates. All experimental data were analyzed using Student's t test (two-sided). P < 0.05 was considered statistically significant. We used SPSS software version 22.0 (IBM, Armonk, NY, USA) for all statistical analyses.