Human miRNAs sources download

The hairpin and mature sequences of miRNA were identified in the miRBase database. In addition, all sequence and annotation data are also available for download. We downloaded the human miRNA data through miRBase (http://www.mirbase.org/, version 18.0).

The CpG islands of miRNAs analysis

We identified CpG islands of these human miRNAs using several genomics browser analyses, such as the UCSC Genome Browser (http://genome.ucsc.edu/), CpG Island Searcher (http://www.cpgislands.com/), and the European Bioinformatics Institute (EBI, http://www.ebi.ac.uk/Tools/emboss/cpgplot/). In this study, the threshold criteria for CpG islands of miRNAs were as follows: (1) UCSC Genome Browser: GC% ≥ 50%, length > 200 bp, Obs/Exp CpG > 0.6. (2) CpG Island Searcher: GC% = 60%, ObsCpG/ExpCpG = 0.7, length = 300 bp, gap between adjacent islands = 100 bp. (3) EBI: Obs/Exp = 0.7, MinPC = 60, Length = 300. We excluded miRNAs located on the X chromosome, and for two miRNAs with similar positions, we chose miRNAs with a large ObsCpG/ExpCpG value. To improve the accuracy of the CpG islands of prediction, we chose an intersection among the three databases.

Cell culture

Het-1A, Eca109, TE-1, CaES-17 and EC9706 cells were obtained from the Shanghai Institute of Biochemistry and Cell Biology and were cultured in Dulbecco's modified Eagle's medium (Sigma, USA) supplemented with 10% foetal bovine serum (FBS) (HyClone, USA). The cells were cultured at 37 °C in a humidified 5% CO2 incubator (Thermo, USA).

Cell treatment

To estimate the role of epigenetic mechanisms in miRNAs, 5-aza-2'-deoxycytidine (DAC) was used as follows. Two groups were created: ESCC cell lines vs. the Het-1A cell line and ESCC cell lines with 5-aza-2ʹ-deoxycytidine vs. ESCC cell lines without 5-aza-2 ʹ -deoxycytidine. The cells were treated with 1 μM 5-aza-dC for 72 h. After incubation, the cell miRNAs were isolated using the Total TRIzol Reagent Kit (Life Technologies, America). The relative quantifications of the miRNAs between the ESCC and Het-1A cell lines and between the ESCC cell lines with DAC and without DAC were detected by real-time RT-PCR (qPCR), and miProfileTM miRNA qPCR microarray technology was used. U6, U44, U48 and U47 snRNAs were used as internal controls, and the premier sequences of the miRNAs and controls are presented in Table S1. Additionally, we used a reverse transcription control, a positive PCR control, and a no template control. miRNAs with fold changes (ESCC cell lines vs. Het-1A cell line) < 0.5 and fold changes (ESCC cell lines with DAC vs. ESCC cell lines without DAC) > 2 were screened.

Clinical samples and follow-up

We conducted a study of patients newly diagnosed with ESCC at the Zhangzhou Affiliated Hospital of Fujian Medical University in Fujian, China, from December 1, 2012, to August 31, 2014. Participants who underwent surgery at Zhangzhou Affiliated Hospital of Fujian Medical University were recruited. The diagnosis was confirmed by two pathologists. Patients with other cancers were excluded. Before the patients underwent surgical excision, they did not receive chemotherapy or radiotherapy. A total of 145 ESCC tissues were collected. This study was approved by the Institutional Review Boards of Fujian Medical University (NO. 2,011,052), and all participants signed informed consent forms.

MiRNAs extraction

MiRNAs were extracted from cells using the Total TRIzol Reagent Kit (Life Technologies, America). A NanoDrop ND-1000 system (NanoDrop, Thermo, America) was used to detect the concentration of all RNA samples.

Quantitative real-time PCR

To validate the miRNA expression levels, qPCR was performed using an SYBR kit (Takara, Japan) on a 7500 PCR System (Applied Biosystems, Thermo, America). The premier for the candidate miRNAs and internal controls were purchased from Guangzhou Funeng (Guangzhou, China). Table S1 shows the premier sequences. The 2−ΔΔCT method was applied to calculate the relative expression values for the target miRNA, which were normalized to the internal control.

Methylation-specific high-resolution melting (MS-HRM) analyses

Genomic DNA was extracted from ESCC tissue specimens using an adsorption column. Tissue DNA was treated with bisulfite using an EpiTect PLUS DNA Bisulfite kit (QiaGen, Germany) according to the manufacturer’s protocol. Methylation-sensitive high-resolution melting (MS-HRM) assays were used to detect the methylation levels. Briefly, bisulfite-modified unmethylated and methylated standard DNA (Qiagen GmbH) was mixed, giving the samples 0, 25, 50, 75, and 100% methylation degrees for calibration. A standard curve with known methylation degrees was included in each run. The hsa-miR-3665 premier sequence is listed in Table S2. The hsa-miR-3665 amplification conditions were 39 cycles at 95 °C for 30 s, 56.7 °C for 30 s and 72 °C for 15 s. High-resolution melting analysis with LightScanner TM was used to analyse the products after PCR. The HRM data were calculated using High-Resolution Melting Software version 2.0.1 (Applied Biosystems). We divided 145 ESCC patients into two groups according to their 50% methylation levels.

Prediction of target genes of hsa-miR-3665

The target genes of hsa-miR-3665 were predicted using the following miRPathDB (https://mpd.bioinf.uni-sb.de/overview.html), miRWalk (http://mirwalk.umm.uni-heidelberg.de/) and TargetScan (http://www.targetscan.org/index.html). The intersection of these three software results was used as the final target gene of hsa-miR-3665.

GO and KEGG pathway enrichment analysis of target genes

To analyze the functions of the predicted target genes of hsa-miR-3665, Gene Ontology (GO) (http://www.geneontology.org) was used for gene functional enrichment of predicted target genes. Kyoto Encyclopedia of Genes and Genomes (KEGG) data (http://www.kegg.jp/kegg/) was used for analyzing the roles of target genes. The human genome was used as a genetic background. P value < 0.05 was considered statistically significant in GO terms and KEGG pathways.

Statistical analysis

The paired sample t-test was applied to evaluate the miRNA expression level between tumour tissue and the corresponding adjacent tissues. The association between the methylation ratio of miRNAs and the clinical factors was assessed using Chi-square test and multivariate logistic regression. Cox proportional hazard regression was used to investigate the prognostic factors for ESCC. All of the participants were randomly divided into either the training or validation set (the split ratio was 2:1), with the training and validation sets being used to establish the predictive model and to construct the nomogram [22]. The C-indices and calibration curves were created to determine whether the predicted survival and actual survival were in concordance. All statistical analyses and visualizations were performed using R software (version 4.1.2). Differences were identified as considered statistically significant for two-sided P values < 0.05.