Three RNA-seq datasets (GSE28619, GSE143318, and GSE59492) were retrieved from Gene Expression Omnibus (GEO). GSE28619 included mRNA expression profiles of 15 ALD samples and seven normal liver samples, whereas GSE143318 included five and five, respectively. MicroRNAs expression data for 13 ALD tissues and six normal liver tissues were collected from GSE59492. The expression matrix was constructed with the R package affyPLM, and probe IDs were converted to gene symbols according to the corresponding platform annotation files. Gene expression values were averaged if matched with multiple IDs.

The original expression data were pre-processed with the robust multiarray average algorithm in R studio software, and the classical Bayesian algorithm was used to screen differentially expressed genes (DEGs) and differentially expressed miRNAs (DEMs) with the “limma” package. FDR (adj. P-value) < 0.05 and |log2 fold change (FC)| ≥1.5 were set as the cut-off criteria. Notably, the differential expression analysis between ALD and normal samples has been completed in GSE143318, which was viewed as a validation group of DEGs.

Only common DEGs between GSE28619 and the validation group (GSE143318 and our previous work) were selected for further analysis. The protein–protein interaction (PPI) relationship of DEGs was mapped onto the Search Tool for the Retrieval of Interacting Genes (STRING, version 10.0, http://string.embl.de/). According to the default settings, the PPI network was constructed with a minimum required interaction score of 0.4, and disconnected genes were hidden. The results were visualized with Cytoscape 3.8.2 software. The R package clusterProfiler was used to perform the functional annotation analyses, comprising gene ontology (GO) enrichment and Kyoto encyclopedia of genes and genomes (KEGG) pathway analyses, which were depicted graphically with the “ggplot2” package.

The prediction of DEM-target genes was achieved with two independent online databases: TargetScanHuman 7.2 (http://www.targetscan.org/vert_72/) and miRDB (http://mirdb.org/index.html). The fundamental criteria were based on the experiment, reference, and support type (functional miRNA–target interactions) to explore the most relevant miRNA–mRNA interactions. Afterwards, FunRich analysis between DEMs target genes and DEGs was performed to screen the overlapping genes, thereby establishing the miRNA–mRNA network and identifying hub genes and miRNAs of ALD.

For the ALD model, C57BL/6 male mice, 6–8 week-old, each weighing 18–20 g, were purchased from SIPEIFU Biotechnology Ltd. (Beijing, China) and were housed in an animal facility with constant temperatures. The animal experimental procedures were approved by the Ethics Committees of Southwest Medical University (No.201809121). The mice were then randomly divided into ALD and control groups after acclimating to the laboratory conditions. Alcoholic liver disease mice were fed Lieber-DeCarli diets (Trophic Animal Feed High-tech Co., Ltd., China), and the model was constructed as follows: (1) adaptive feeding with an ordinary liquid diet for 3-days; (2) gradual addition of an alcohol liquid diet to the feed (9, 14, 19, 24, and 28% for 2 days, respectively); and (3) feeding with a 28% alcohol liquid diet for 26 days. The control mice were fed an ordinary liquid diet. Subsequently, the livers were quickly excised, and a portion of the tissues was fixed in 10% neutral-buffered formalin for pathological analysis; the remaining tissues and blood were stored at −80°C for further analyses.

Normal liver cell L02 line was obtained from the Laboratory Medicine of Chongqing Medical University (Chongqing, China) and cultured in RPMI 1640 (Gibco, United States) supplemented with 10% fetal bovine serum (Gibco, United States) and 100 U/ml Penicillin-Streptomycin Solution (Thermo Fisher Scientific, China). After reaching 50–60% confluence, L02 cells were stimulated with 100 mM alcohol medium for 48 h to construct the ALD cell model (9, 13). All cells were grown at 37°C in a humidified incubator containing 5% CO₂.

H&E and Oil red O staining kits were purchased from Sangon Biotech Co., Ltd. (Shanghai, China). First, the mouse liver tissues were embedded in paraffin blocks, and L02 cells were fixed with 4% paraformaldehyde for 24 h and 60 min, respectively. Second, H&E and Oil red O staining was performed according to the manufacturers' protocols. The serum biochemical characteristics of the mice, such as alanine aminotransferase (ALT), aspartate aminotransferase (AST), triglycerides (TG), and total cholesterol (TC), were detected with a BS2000M automatic biochemical analyzer (Mindray, Shenzhen, China); The TG level of L02 cells was tested with a Tissue and cell TG assay kit (Applygen, Beijing, China). The absorbance was measured at 550 nm. Image pro plus software was applied to calculate the integrated optical density (IOD) of the oil-red O staining, thereby performing the quantitative analysis.

To explore the importance of miRNA and the regulated relationship of miRNA–mRNA in ALD, we transfected L02 cells with 50 nM miRNA mimics, inhibitor, negative control (NC), and FOXO1 siRNA with Lipofectamine 2000 (Invitrogen, United States) and Opti-MEM (Gibco, United States), on the basis of the manufacturer's instructions. After 48 h of transfection, cells were collected for further analyses. The RNA oligo sequences were as follows: miRNA mimics: 5′-UUUGGCAAUGGUAGAACUCACACU-3′; miRNA inhibitor: 5′-AGUGUGAGUUCUACCAUUGCCAAA-3′; FOXO1 siRNA: 5′-UGACUUGGAUGGCAUGUUC-3′; NC: 5′-UUGUACUACACAAAAGUACUG-3′.

Total RNA was extracted from mouse liver tissues and L02 cells with the Trizol method. According to the manufacturer's instructions, the cDNA for miRNAs was first synthesized with miRNA first strand cDNA synthesis (Sangon Biotech, Shanghai, China). For mRNA, a PrimeScript™ RT reagent Kit (TaKaRa, Shiga, Japan) was used. Then quantitative real-time PCR (RT-qPCR) was performed with 2 × SG Fast qPCR Master Mix (Low Rox) (Sangon Biotech, Shanghai, China) on Applied Biosystems 7500 Real-Time PCR Detection System (Applied Biosystems, United States) according to the manufacturer's instructions. SnRNA U6 and GAPDH were used as the internal controls for miRNA and mRNA, respectively. Primer sequences are shown in Table 1. The relative levels were calculated with the 2-^ΔΔCt method, and all experiments were performed in triplicate and were repeated at least three times.

The binding sites of the mRNA 3′UTR regions targeted by miRNAs were predicted with the TargetScan database, and the sequences, including wild type and mutant with spacer mutations, were then cloned and inserted into the luciferase reporter pmirGLO vector (Promega, WI, USA). Subsequently, 50 nM miRNA mimic, normal control, and 1 μg pmirGLO vector were co-transfected into HEK 293T cells. Cells were plated on 24-well plates, and harvested after 48 h of transfection. The relative activity of firefly luciferase was detected with the Dual-Glo luciferase assay kit (Beyotime, Shanghai, China).

Protein was extracted with Western Cell Lysis Buffer with 1% phenylmethyl sulfonyl fluoride (Sangon Biotech, Shanghai, China). Next, a modified BCA Protein Assay Kit (Sangon Biotech, Shanghai, China) was used to determine the protein concentration according to the absorbance at 562 nm. The 30 μg total protein was loaded onto an SDS polyacrylamide gel (Solarbio, Beijing, China), including 5% stacking gel and 8% resolving gel. After electrophoresis at 75 V for 30 min and 100 V for 60 min, protein was transferred to a 0.45 μm PVDF membrane (Millipore, Billerica, MA, United States) at 100 V for 2–4 h, blocked in 5% non-fat milk for 60 min, and incubated with primary antibodies for 12 h at 4°C (Abcam, UK) (FOXO1 69KD, 1:1,000, ab179450; sirtuin 1 (SIRT1) 110KD, 1:1,000, ab189494; adipose triglyceride lipase (ATGL) 55KD, 1:1,000, ab109251; sterol regulatory element-binding protein-1c (SREBP-1c) 127KD, 5 μg/ml, ab3259; fatty acid synthase (FASN) 273KD, 1:1,000, ab128856; GAPDH 37 KD, 1:10,000, ab181602); CYP2E1 (57KD, 1:2,000, 19937-1-AP) (Proteintech, UK), and the HRP conjugated secondary antibody (Beyotime, Shanghai, China) for 60 min. Subsequently, a BeyoECL Star kit (Beyotime, Shanghai, China) was used to conduct the western blotting detection, and the relative quantitative analyses were performed in Image J software.

All statistical analyses were performed in GraphPad Prism 8.0 and R studio 3.6.0. Quantitative data are represented with means ± standard deviation. Analysis of variance and Student's t-test were used to explore the differences among groups. P-values < 0.05 were considered statistically significant.

The flowchart of the current study is shown in Figure 1. On the basis of the defined criteria, a total of 416 DEGs and 21 DEMs were identified in GSE28619 and GSE59492 between ALD samples and normal liver samples, respectively. Volcano plots (Figures 2A,B) and heatmaps (Figures 2C,D) were constructed to reveal the details. In total, 123 DEGs were confirmed to be significantly differentially expressed in GSE143318 gene profiling (Figure 2E), including 80 up-regulated and 44 down-regulated DEGs. Gene ontology enrichment analysis (Figure 2F) further illustrated that common DEGs were highly involved in signal transduction, oxidation–reduction process, and extracellular matrix organization; Additionally, KEGG pathway analysis showed that the PI3K–Akt signaling pathway, focal adhesions, and ECM–receptor interaction were significantly associated with ALD development (Figure 2G).

With the STRING online tool, a total of 74 nodes and 207 edges were identified to be highly connected in the PPI network (Figure 3A). Target genes of DEMs were predicted with two independent databases, and only common genes were retained for subsequent analyses (Figure 3B). Afterwards, the miRNA–mRNA network was established to reveal the potential molecular mechanisms of ALD, including 16 common genes and eight miRNAs (Figure 3C). Of note, FOXO1 has been demonstrated to participate in lipid metabolism in our previous study and other publications (14–17). However, the expression of miR-182-5p between ALD tissues and normal liver tissues has remained controversial, and its molecular mechanisms were unclear (11, 12). Therefore, we reasoned that the hypothesis that miR-182-5p plays an essential role in ALD development by targeting FOXO1 was worthy of study in depth.

After being fed a Lieber-DeCarli diet for 28 days, ALD mice markedly differed from normal mice in their weight and biochemical characteristics. The statistical differences in weight variation between EtOH-fed and normal mice are shown in Figure 4A. Moreover, ALT and TG levels in EtOH-fed mice were markedly higher than those in normal mice. However, no significant difference was observed in TC and AST levels (Figure 4B). H&E staining and Oil Red O staining (Figure 4C) demonstrated that EtOH consumption significantly stimulated hepatic steatosis and accelerated fat accumulation in mice. Therefore, in this work, the ALD mouse model was successfully constructed. RT-qPCR was used to evaluate the relative expression levels of miR-182-5p and FOXO1 between ALD and normal liver tissues. The expression of miR-182-5p was dramatically up-regulated in ALD mice, whereas that of FOXO1 was lower, as compared with the levels in normal mice (Figures 4D,E).

Next, L02 cells were exposed to 100 mM alcohol medium for 48 h, as described above, to establish the ALD cell model. As shown in Figure 5A, the cellular Oil Red O staining revealed that the lipid accumulation in ALD cells was much higher than that in normal L02 cells. Quantitative analysis indicated that the TG content and IOD of L02 cells dramatically increased as the EtOH stimulus was presented (Figure 5B). The differential expression of miR-182-5p and FOXO1 between ALD cells and L02 cells was consistent with the results in the mice model, thus further validating the RNA-seq expression chip results (Figures 5C,D). CYP2E1 expression was remarkably increased in ALD cell compared to normal liver cells (Figure 5E), suggesting that the alcohol metabolism ability of L02 cell was enhanced under EtOH stimulation.

According to TargetScan and miRDB databases, miR-182-5p binding to the 3'UTR of FOXO1 was identified (Figure 6A). Dual-luciferase reporter assays were performed to verify the targeting interaction. As shown in Figure 6B, the luciferase activity of the wild-type group that contained FOXO1 specific binding sites was clearly inhibited by co-transfection of miR-182-5p mimics. Importantly, the expression of FOXO1 mRNA was suppressed by over-expression of miR-182-5p (Figures 6C,D). Moreover, western blotting assays indicated that FOXO1 protein expression was markedly decreased in L02 cells transfected with miR-182-5p mimics (Figure 6E). In contrast, FOXO1 expression was positively up-regulated after miR-182-5p knockdown. Therefore, the above results demonstrated that miR-182-5p directly targets FOXO1 and inhibits its mRNA and protein expression.

FOXO1, a hub transcription factor has been reported to have a critical function in fatty liver by regulating lipid metabolism-related gene expression. Hence, several lipid metabolism-associated downstream genes of FOXO1 previously confirmed by experimental studies (SIRT1, ATGL, SREBP-1c, and FASN) were detected to reveal the mechanism underlying the effects of FOXO1 in ALD. RT-PCR and western blotting assays indicated that the expression of FOXO1 was significantly decreased after transfection with FOXO1 siRNA (Figure 7). Further experiments showed that inhibiting FOXO1 in L02 cells promotes SREBP-1c and FASN, but represses SIRT1 expression. In comparison, ATGL and CYP2E1 expression levels were found no significant difference between siRNA and NC groups (P > 0.05).

To further investigate the potential mechanism of miR-182-5p in ALD, we performed miR-182-5p overexpression and knockdown experiments. Quantitative real-time PCR showed that miR-182-5p expression in the mimic group was significantly higher than that in the ALD group, whereas an opposite trend in the knockdown group was observed, thus suggesting successful transfection in this work (Figure 8A). Interestingly, the TG level in the ALD model significantly decreased with miR-182-5p inhibitor treatment, whereas transfection of miR-182-5p mimics accelerated TG accumulation (Figure 8B). Moreover, Oil Red O staining and IOD analysis demonstrated that miR-182-5p knockdown markedly attenuated EtOH-induced hepatic steatosis (Figure 8C). Together, the above results revealed that miR-182-5p plays an essential role in TG deposition in ALD. As shown in Figures 8D,E, the expression of FOXO1 in ALD cells was clearly suppressed by EtOH exposure and upregulated miR-182-5p, but this effect was alleviated by transfection with miR-182-5p inhibitor. Meanwhile, SIRT1 expression was inhibited, whereas the expression of SREBP-1c, and FASN was greater in ALD than in the controls, thus verifying that an EtOH stimulus led to imbalanced lipid metabolism. Additionally, enhancement of the miR-182-5p/FOXO1 signaling axis dramatically stimulated SREBP-1c and FASN overexpression while repressing the expression of SIRT1. In contrast, its inhibition had opposite effects on ALD. These results demonstrated that the miR-182-5p/FOXO1 axis participates in lipid metabolism in alcoholic liver by regulating the expression of key lipid biosynthesis and decomposition-related genes (Figure 9).