Primary human liver cancer tissue and corresponding adjacent tissue (<2 cm away from the tumor area) samples were obtained from 10 patients who underwent primary surgical resection of liver cancer at the Department of Hepatobiliary Surgery, Union Hospital, Tongji Medical College, Huazhong University of Science and Technology. All participants provided written informed consent, and the study protocols were approved by the hospital ethics committee. Medical records of the patients, including age and gender, tumor staging, pathological diagnosis and surgical records were collected. The tissues were immediately snap-frozen in liquid nitrogen and stored at −80°C until total RNA was extracted.

Normal hepatocyte cell line Lo2 and human hepatocellular carcinoma cell lines HepG2 and SMMC7721 were purchased from the Chinese Institute of Biochemistry and Cell Biology and cultured in RPMI-1640 medium (Hyclone, Logan, Utah, USA) containing 10% fetal bovine serum (FBS) (Gibco, Grand Island, NY, USA), 1 U/ml penicillin and 1 µg/ml streptomycin (Gibco). Cells were grown in a humidified atmosphere of 5% carbon dioxide at 37°C. All cell lines were authenticated and characterized by the cell bank. The cells were expanded immediately and multiple aliquots were cryopreserved. Cells were used within 6 months of resuscitation

Total RNA and small RNA were extracted from the cell lines and tissues using TRIzol^® reagent (Invitrogen, Carlsbad, CA, USA). The purity and integrity of the RNA sample were assessed using a NanoDrop 2000 UV-Vis Spectrophotometer (Thermo Scientific, Waltham, MA, USA), and 500 ng RNA was transcribed into cDNA using the PrimeScript™ RT reagent kit (Takara, Dalian, China) according to the manufacturer's instructions. The obtained cDNA was used as a template to perform PCR amplification using the SYBR^® Premix Ex Taq™ II kit (Takara). Human Gal-9, GAPDH, miRNAs and U6 mRNA levels were analyzed using qRT-PCR with the Mx3000P system (Agilent Technologies, Santa Clara, CA, USA). GAPDH was used as an internal control for Gal-9 detection and U6 for miRNA detection. Each 20 µl reaction system consisted of 2 µl of cDNA, 10 µl SYBR^® Premix Ex Taq™ II, 10 µmol/l of both sense and antisense primers. All PCR reactions were carried out as follows: initial denaturation at 95°C for 30 sec, followed by 40 cycles of denaturation at 95°C for 5 sec, annealing at 60°C for 30 sec and extension at 72°C for 30 sec. Each experiment contained at least three replicates, and the results were calculated according to the method 2^−ΔΔCT. The PCR primers used in this study are available upon request and are listed in Table I.

Protein was extracted from the cells using RIPA lysis buffer containing EDTA-free Complete Protease Inhibitors (Roche, Mannheim, Germany). After the protein concentration was measured, it was mixed with SDS loading buffer, separated by 10% SDS-PAGE and transferred onto nitrocellulose membranes. To block nonspecific binding, the membranes were incubated at room temperature for 1 h with 5% skim milk powder, followed by an overnight incubation at 4°C with the primary antibodies anti-Gal-9 (ImmunoWay Biotechnology Company, Newark, DE, USA) or anti-β-actin (Cell Signaling Technology, Boston, MA, USA) and the blots were incubated with horseradish peroxidase-labeled secondary antibodies (Cell Signaling Technology). Signals were visualized by ECL chemiluminescence. Changes in protein expression were quantified using ImageJ software (National Institutes of Health, Bethesda, MD, USA). Equal protein loading was assessed by expression of β-actin.

Bioinformatic analysis was carried out using specific programs: miRanda (http://www.microrna.org) and TargetScan (http://www.targetscan.org/).

The full-length Gal-9 3′UTR containing a putative miR-22 binding site was amplified by PCR from genomic DNA and cloned at the EcoRI and HindIII sites into the pCMV vector. The specific primers for the wild-type Gal-9 3′UTR vector were as follows: 5′-ATAGAATTCGCGGCTTCCTGGCCCTG-3′ and 5′-CGCAAGCTTTGAATGTGCCAACAAGCA-3′. The specific primers for the mutant-type Gal-9 3′UTR vector were as follows: 5′-ATAGAATTCGCGGCTTCCTGGCGGAC-3′ and 5′-CGCAAGCTTTGAATGTGCCAACAGCGC-3′. Wild-type and mutant-type inserts were confirmed by sequencing. For the expression of Gal-9, the genomic fragment of Homo sapiens Gal-9 precursor was amplified by PCR using the primer pairs: 5′-GGAGAATTCGAGATGGCCTTCAGCGGTTCCCAG-3 and 5′-CCACTCGAGCGCCTATGTCTGCACATGGGTCAG-3′. The PCR product was cloned into pcDNA3.1 at the EcoRI and XhoI sites and named the Gal-9/pcDNA3.1 vector. miR-22-mimic, miR-491-mimic and miR-control mimic (negative control) were synthesized by Guangzhou RiboBio Co., Ltd. (Guangzhou, China). Cells were transiently transfected by use of Lipofectamine 2000 transfection reagent (Invitrogen, Carlsbad, CA, USA) for 24 h, according to the manufacturer's protocol.

HepG2 cells were plated in a 96-well plate and co-transfected with miRNA mimics or control mimics with pCMV-Gal-9 3′UTR-WT or pCMV-Gal-9 3′UTR-MU and PRL-TK (Promega, Beijing, China), using Lipofectamine 2000. Cells were collected 24 h after transfection and luciferase activity was analyzed using the Dual-Luciferase Reporter Assay System Modulus™ Single Tube Multimode Reader (Turner BioSystems, Madison, WI, USA). The pRL-TK vector that provided the constitutive expression of Renilla luciferase was co-transfected as an internal control to correct for differences in both transfection and harvest efficiencies. Transfections were carried out in duplicates and repeated at least twice in independent experiments.

PBMCs were isolated from the venous blood of healthy blood donors. Aliquots of 3 ml of blood were diluted in 3 ml PBS and layered on 6 ml Lymphocyte Separation Media (MD Pacific, Tianjin, China). After centrifugation at 2,000 rpm for 20 min at room temperature, the PBMCs were recovered from the interphase and washed twice with serum-free RPMI-1640 medium (Hyclone, Logan, Utah, USA) and centrifuged at 1,000 rpm for 10 min. PBMCs were resuspended in complete culture medium and adjusted to 3×10⁶ cells/ml.

For the co-culture experiments, HepG2 liver cancer cells were washed with PBS, diluted in complete medium and transferred to the wells. Transfections were performed using Lipofectamine 2000. After a 24-h transfection, PBMCs were plated in each well for co-culture. After 48 h, the cancer cells and the PBMCs were separated for further analysis.

The WST-1 assay was used to evaluate cell proliferation. After a 48-h co-culture, 10 µl/well cell proliferation reagent WST-1 (Roche, Mannheim, Germany) was added and incubated for 2 h. The cells were shaken thoroughly for 1 min on a shaker and the absorbance was measured using an ELX808 Universal Microplate Reader (Bio Tek, Winooski, VT, USA). The wavelength for measuring the absorption of the formazan product was 450 nm. All samples were performed in triplicate.

A cell apoptosis assay was performed using an Annexin-V-Fluos staining kit (Roche, Mannheim, Germany). After co-culture of the PBMCs and HepG2 cells for 48 h, PBMCs were collected and centrifuged at 1,000 rpm for 5 min and washed with PBS twice. The cell pellet was resuspended in 100 µl of Annexin-V-Fluos labeling solution (Roche) and incubated 10–15 min at 15–25°C. All specimens were assessed on a FACS Calibur flow cytometer (BD, Franklin Lake, NJ, USA), and data were analyzed with FlowJo software.

The statistical significance was determined using the Student's t-test. All data are expressed as the mean ± SD. A P-value <0.05 was considered to indicate a statistically significant result. All experiments were repeated more than three times and each experiment consisted of triple wells. One representative experiment was selected to construct diagrams and for data analysis.

To identify the expression levels of the Gal-9 mRNA and protein in normal liver and liver cancer cells, we performed qRT-PCR and western blot analysis in Lo2 cells as well as in HepG2 and SMMC7721 cells. The Gal-9 mRNA level in HepG2 cells was ~5.3-fold higher and the level in the SMMC7721 was 2.4-fold higher than that in the Lo2 cells. Furthermore, the Gal-9 protein level in the HepG2 cells was ~2-fold higher than that in the Lo2 cells. The results revealed that Gal-9 was significantly upregulated in the liver cancer cells at the mRNA and protein levels compared to these levels in the normal liver cells (p<0.05, Fig. 1A and B), particularly in the HepG2 cells.

Target prediction programs miRanda (http://www.microrna.org) and TargetScan (http://www.targetscan.org/) were used to predict and identify miRNAs that possibly target Gal-9 in liver cancer. The overlapping prediction analysis suggested that 4 miRNAs (miR-22, 296-3p, 455-5p and 491-5p) were potential regulators of Gal-9 (Fig. 2A).

We determined the expression of these 4 miRNAs in normal liver (Lo2) and liver cancer (HepG2) cells. The results showed that miR-22 and miR-491 were significantly down-regulated in liver cancer cells than these miRNAs in normal liver cells, particularly miR-22 (Fig. 2B). miR-22 was focused on and its expression level was determined in liver cancer and corresponding adjacent tissues. miR-22 was downregulated in liver cancer tissues compared to that in the corresponding adjacent tissues (Fig. 2C).

To investigate whether the expression of Gal-9 is regulated by miR-22, the miR-control or miR-22 mimic were transfected into HepG2 cells and the expression levels of Gal-9 mRNA and protein were detected using qRT-PCR and western blot analysis, respectively. We found that the expression of Gal-9 mRNA was slightly decreased, but did not reach a significant level (Fig. 3A). However, the expression of Gal-9 protein was significantly downregulated after transfection of the miR-22 mimics (Fig. 3B) compared with the negative control and miR-control.

To further investigate the mechanisms underlying the regulation of Gal-9 by miR-22, the 3′UTR of Gal-9 was cloned into a reporter vector linking the luciferase open reading frame downstream to generate pCMV-Gal-9-3′UTR wild-type (Gal-9 3′UTR-WT) and pCMV-Gal-9-3′UTR mutant-type (Gal-9 3′UTR-MU). The two plasmids were transfected into HepG2 cells, respectively, together with the miR-control or miR-22 mimic, respectively. The luciferase activity was measured after a 24-h transfection. The luciferase activity in the Gal-9 3′UTR-WT group, but not in the Gal-9 3′UTR-MU, was significantly reduced in the cells co-transfected with the miR-22 mimic when compared with the miR-control (P=0.0268, Fig. 3C).

To determine whether the miR-22/Gal-9 axis affects lymphocyte and tumor cell apoptosis, HepG2 cells were divided into four groups: negative control, cells transfected with Gal-9 overexpression vector only, cells transfected with the miR-22 mimics only and cells co-transfected with the Gal-9 overexpression vector and miR-22 mimics. After a 24-h transfection, the liver cancer cells were co-cultured with PBMCs for 48 h and the apoptosis of PBMCs was detected. Our results showed that the apoptosis rate of the PBMCs co-cultured with the HepG2 cells transfected with the Gal-9 overexpression vector was higher than that of the negative control. The apoptosis rate of the PBMCs co-cultured with the HepG2 cells co-transfected with miR-22 mimics and the Gal-9 overexpression vector was lower than that of the cells transfected with Gal-9 only (Fig. 4A).

In order to further examine the role of the miR-22/Gal-9 axis in tumor cell proliferation, a WST-1 assay was used. The results showed that HepG2 cells transfected with the Gal-9 overexpression vector exhibited higher proliferation compared with the negative control. However, HepG2 cells transfected with the miR-22 mimics only exhibited slower proliferation when compared with the negative control. While HepG2 cells co-transfected with the miR-22 mimics and the Gal-9 overexpression vector showed no significantly different cell proliferation compared to the negative control (Fig. 4B).