A total of 12 male C57BL/6 mice, aged 6 weeks, weighing 20–25 g, were purchased from the Shanghai Xipuer-Bikai Laboratory Animal Co., Ltd (Shanghai, China). All mice were housed in cages under standard conditions with an ambient temperature of 24°C under a 12 h light/dark cycle. Mice had free access to food and water (as detailed in the subsequent section). All experimental protocols were approved by the Animal Ethics Committee of the Eighth People's Hospital of Shanghai (Permit no. 2016-02-2).

Animals were randomly divided into two groups: Control group (n=6) and NASH group (n=6). All mice received food and water ad libitum. Mice in the NASH group were provided with an MCD diet. Mice in the control group were fed the same diet, but with sufficient quantities of DL-methionine (3 g/kg) and choline bitartrate (2 g/kg). At the end of 8 weeks of administration of the diet, all the mice were anesthetized and sacrificed, and their livers were removed. Serum and liver samples were frozen in liquid nitrogen prior to biochemical analysis. A small piece of the liver sample was fixed in 4% paraformaldehyde (96 h at 4°C) for subsequent histological analysis.

The levels of serum triglycerides (TG) and total cholesterol (TC), and the activities of the liver-associated enzymes, alanine aminotransferase (ALT) and aspartate aminotransferase (AST), were estimated with an automatic biochemical analyzer (Roche/Hitachi P-800 Modular Analytics Diagnostic System; Roche Diagnostics, Indianapolis, IN, USA). Hepatic homogenates were used for determination of the TG and TC content using a kit provided by Nanjing Jianchen Bioengineering Institute (Nanning, China) according to the manufacturer's protocol. Tissue lipids were extracted with methanol/chloroform (1:2).

The liver samples removed from each mouse were fixed in 4% paraformaldehyde for 96 h at 4°C. Following dehydration in graded alcohol (once in 50, 70 and 95%, and twice in 100% for 2–3 h each) and embedding in paraffin wax, the sections were cut to a thickness of 4 µm, and stained with hematoxylin and eosin (H&E) (25). The liver histology was scored for ballooning (0–2), steatosis (0–3), and inflammation (0–3), and the sum of these scores was used to create the NAFLD Activity Score. Picrosirius red staining (0.1% Sirius Red in saturated picric acid stained for 75 min at room temperature) was used to assess collagen fibers in the liver tissues. Each section was assessed under 10×20 light microscopic fields. For lipid staining, 6 µm frozen liver sections were air-dried for 30 min, followed by fixation in 4% formaldehyde. Oil red O staining was performed with staining solution (0.5% Oil Red O in anhydrous isopropanol, sections were stained for 30 min at 37°C) (Sigma-Aldrich; Merck KGaA, Darmstadt, Germany) according to the manufacturer's protocol. Then, six slides from the control and NASH groups were used for quantitative analysis. The stained area was quantified in six randomly selected ×200 microscopic fields per mouse section using ImageJ software (ImageJ2×; National Institutes of Health, Bethesda, MD, USA).

Total RNA was isolated from the different groups using TRIzol^® reagent (Invitrogen; Thermo Fisher Scientific, Inc., Waltham, MA, USA) and purified with an RNeasy Mini kit (Qiagen GmbH, Hilden, Germany) according to the manufacturer's protocol. The purity and concentration of RNA samples were determined with the NanoDrop^® ND-1000 spectrophotometer (Thermo Fisher Scientific, Inc.). The integrity of the RNA was assessed by electrophoresis on a denaturing gel.

Three liver RNA samples from each group were selected for microarray studies. Sample labeling and array hybridization were performed according to the manufacturer's protocol (Arraystar, Inc., Rockville, MD, USA). In brief, the extracted total RNA was treated with RNase R (Epicentre^®; Illumina, Inc., Madison, WI, USA) to remove linear RNAs and enrich the population of circRNAs. The enriched circRNAs were amplified and transcribed into fluorescent complementary RNA (cRNA) using random primers, according to the instructions provided by the Arraystar Super RNA Labeling kit (Arraystar, Inc.). Following purification of the labeled cRNAs with the RNeasy Mini kit (Qiagen GmbH), the concentration and specific activity of the labeled cRNAs were measured using the NanoDrop^® ND-1000 spectrophotometer. The labeled cRNA was hybridized on to Arraystar mouse circRNA array v1.0. Following washing of the slides, the arrays were scanned using the Agilent G2505C scanner (Agilent Technologies, Inc., Santa Clara, CA, USA).

RT-qPCR was used to confirm the circRNA expression profiles obtained from the microarray data. Total RNA was isolated from liver tissues using TRIzol^® reagent (Invitrogen; Thermo Fisher Scientific, Inc.) and reverse transcribed into cDNA using SuperScript™ III Reverse Transcriptase (Invitrogen; Thermo Fisher Scientific, Inc.) according to the manufacturer's protocol. The relative gene expression was determined using the Agilent Mx3000P qPCR system. Amplifications were performed using a SYBR Green PCR kit (Takara Biotechnology Co., Ltd., Dalian, China). The qPCR program comprised an initial denaturation step of 3 min at 95°C, followed by 40 cycles of 15 sec at 95°C, 30 sec at 60°C, and 30 sec at 72°C. All samples were normalized to the signal generated from GAPDH. Results were obtained from three independent wells and shown as the fold change, according to the 2^−ΔΔCq method (26). Primers were designed to amplify the circRNA-specific back splice junctions (Table I).

Scanned images were imported into the Agilent Feature Extraction software (version 11.0.1.1) for raw data extraction. Quantile normalization of raw data and subsequent data processing were performed using the R software package (R version 3.1.2). The circRNAs that exhibited fold changes (FC) ≥1.5 (P≤0.05) between the two groups were categorized as significantly differentially expressed. The selected differentially expressed circRNAs were identified through volcano plotting and FC filtering. Hierarchical clustering analysis was performed to show the distinguishable circRNA expression pattern among samples.

The circRNA-miRNA interaction analysis was performed using Arraystar's miRNA target prediction software, which includes TargetScan (www.targetscan.org/) and miRanda (www.microrna.org/). Potential miRNA response elements (MREs) for circRNA were predicted using the miRNA support vector regression (mirSVR) which obtained from miRanda as previously described (27). Gene Ontology (GO) analysis was performed to explore the functional roles of host linear transcripts in the terms biological processes (BPs), cellular components (CCs) and molecular functions (MFs). Kyoto Encyclopedia of Genes and Genomes (KEGG) was used to determine the target genes in different biological pathways.

Data are presented as the mean ± standard deviation for triplicate measurements. Statistically significant differences between groups were estimated using the Student's t-test with SPSS software, version 13.0 (SPSS, Inc., Chicago, IL, USA). P<0.05 was considered to indicate a statistically significant difference.

At the end of 8 weeks of feeding the mice on the prescribed MCD diet, the body weight of the mice in the model group was significantly decreased compared with that in the control group (P<0.01; Table II). The H&E-stained sections exhibited typical histopathological features of NASH. As presented in Fig. 1, sections from the NASH livers exhibited disordered hepatic lobules, full fat vacuoles in lobule cells, infiltration of inflammatory cells, and cell swelling (Fig. 1A). The livers of MCD diet-fed mice exhibited marked lipid accumulation and liver fibrosis (Fig. 1B and C). Serum AST and ALT levels indirectly reflect the failure of liver function. As indicated in Table II, the serum ALT (406.7±132.1 U/l) and AST (573.3±169.5 U/l) activities were significantly increased following the administration of the MCD diet, as compared with the normal group (25.0±7.5 and 175.3±29.5, respectively; P<0.05). The TG and TC levels were lower in the MCD diet-fed mice compared with that in control mice (P<0.05 and P<0.01, respectively). The hepatic TG content (705±69.3 µmol/g protein) was higher in the MCD diet-fed mice compared with the control mice (179±17.6 µmol/g protein), whereas the hepatic TC content did not exhibit any difference between the two groups (Table II). These changes in serum biomarkers and liver tissue indicated the successful establishment of the NASH model in mice.

To screen for circRNAs that were differentially expressed between the NASH and control mice livers, circRNA microarrays were performed in three samples from each group. A box plot showed that the distributions of circRNA for the tested samples were essentially identical following normalization (Fig. 2A). Hierarchical clustering, scatter plot visualization and volcano plot analyses revealed the differences in circRNA expression levels in NASH and control mice liver (Fig. 2B-D; FC >2.0; P<0.05). Consequently, 450 circRNAs were identified as dysregulated in the livers of NASH mice, with FC >1.5 and P<0.05. Among these, 298 circRNAs were upregulated, and 152 circRNAs were downregulated, in the NASH group compared with those in the control group (Fig. 3A). Additionally, 21 circRNAs were upregulated, and 12 circRNAs were downregulated, in the livers of NASH mice with FC >3.0 and P<0.05 (Fig. 3A). CircRNAs on mouse chromosomes are illustrated in the pie chart format in Fig. 3B. Types and counts of differentially regulated circRNAs are presented in Fig. 3C. The top 10 upregulated and downregulated circRNAs sorted by their FC values are summarized in Table III, along with their potential miRNA targets. Taken together, these data indicated that the circRNA expression patterns in NASH livers were different from those in the control group.

To further validate the results of the present study, Arraystar's miRNA target prediction software, based on TargetScan and miRanda, was used to predict the circRNA-miRNA interactions. A total of 20 differentially expressed exonic circRNAs with the highest FC values were selected (FC >3; P<0.05) to predict their miRNA response elements (MREs), including 10 upregulated circRNAs and 10 downregulated circRNAs. Four MREs that were identified to have ‘good’ mirSVR scores for each circRNA are presented in Table III.

To validate the microarray results, seven differentially expressed exonic circRNAs were randomly selected (FC >3.5; P<0.05), including five upregulated circRNAs and two downregulated circRNAs. Primers for circRNAs were designed (shown in Table I), and the appearance of a single peak in the melting curve confirmed the specificity of the amplified products for each circRNA (Fig. 4). The results demonstrated that five circRNAs (circRNA-28028, circRNA-29981, circRNA-004372, circRNA-0043021 and circRNA-36096) were upregulated, and these findings were broadly in accord with the data obtained from the microarray analysis. However, only circRNA-29981 reached the required level of statistical significance to become a candidate for further study (Fig. 5). By contrast, circRNA-008234 was observed to have the opposite expression trend in microarray and RT-qPCR analysis, although the difference was revealed not to be statistically significant. Additionally, the expression levels of circRNA-22646 were not significantly different between the NASH and control groups (Fig. 5). Furthermore, the detailed annotations for circRNA-29981 interaction with various miRNAs (miR-7661-3p, miR-181b-5p, miR-6952-3p, miR-3077-3p, and miR-1912-3p) are presented in Fig. 6. The effect of the circRNA-29981 regulation pathways in NASH is worth further exploration.

CircRNAs are generally formed by alternative splicing of pre-mRNA species, which are usually spliced to generate a linear mRNA (11). A more recent study has demonstrated that a class of circRNAs, termed exon-intron circRNAs, enhances the expression of their pre-mRNAs (28). CircRNAs may also be potentially in competition with their linear transcripts (24). To provide a comprehensive landscape of the origination of circRNAs, GO and KEGG analyses were performed to annotate the biological functions of host linear transcripts. The top 10 generally altered GO terms between NASH and control groups were classified according to the BP, CC and MF (Fig. 7). For increased mRNAs, the most enriched and meaningful BP terms were ‘cellular process’, ‘single-organism process’ and ‘single-organism cellular process’. The most enriched CC terms were ‘cell’, ‘cell part’, ‘intracellular’, and ‘intracellular part’. As for MF, the most enriched terms were ‘binding’, ‘protein binding’, and ‘organic cyclic compound binding’. In addition, the data from the GO analysis on the downregulated transcripts revealed that the top two GO processes were identical with those in the increased transcripts (Fig. 7). Furthermore, KEGG analysis revealed that certain pathways, including regulation of the actin cytoskeleton, 2-oxocarboxylic acid metabolism, and the endocytosis signaling pathway, were associated with the upregulated transcripts, whereas pathways such as the pyruvate metabolism, fatty acid biosynthesis and fatty acid metabolism signaling pathways were associated with the downregulated transcripts (Fig. 8).