Human HCC Hep3B and PLC/PRF/5 cells and HEK 293T cells were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM, Invitrogen) with 10% Fetal Bovine Serum (FBS), 1% penicillin/streptomycin, and 1% glutamine. All cells were cultured in a 37°C temperate box in a 5% (v/v) CO2 condition.

1x10⁵ HCC cells were seeded into an 8 μm transwell chamber (Corning Life Science) with 100 μl diluted Matrigel (BD Corning) coated first in a serum free medium, then the bottom chamber was filled with 700 μL medium (10% FBS), and after being cultured for 24 hours, we collected the transwell chamber for fixation and dyeing of the invaded cells, then observed and counted the cells under the microscope. Each sample was designed in three duplicate chambers and several pictures under different horizons were taken for analysis.

The HCC cells were seeded in a 6 well plate, and when the cells were more than 90% full, a sterile tip was used to make a scratch and it was set as 0h. Pictures were taken with a microscope and record. After being incubated for 24 hours in the incubator, pictures were taken again and the distance of cell migration was measured. Each sample was designed in three duplicate wells and several pictures under different horizons were taken for analysis.

Firstly, the culture medium containing 10% fetal calf serum was used to prepare a single cell suspension and inoculate 5000 cells per well, with a volume of 200ul per well in a 96-well plate. Secondly, the cells were cultured for 1-6 days respectively in the general cell culture conditions. Thirdly, 20ul of MTT solution (5mg/ml in PBS, pH=7.4) was added to each well and continued incubating for 4 hours. Fourthly, the culture supernatant was carefully aspirated and discarded in the well and 150ul DMSO was added to each hole and shaken for 10 minutes to fully melt the crystals. Fifthly, the light absorption value of each hole was measured on the enzyme-linked immunosorbent monitor at 490nm, the result recorded, and the cell growth curve drawn with the time as the abscissa and the absorbance as the ordinate.

Firstly, the cells were digested in the logarithmic growth phase with 0.25% trypsin and a pipette into single cells, and suspended in DMEM culture medium of 10% fetal bovine serum for later use. Secondly, 300 cells were added to a six-well with 3ml of culture medium, gently rotated to mix the cells, and then incubated in a cell culture incubator for 2 weeks. Thirdly, the supernatant was discarded, and carefully washed twice with PBS. 4% paraformaldehyde was added to fix for 15 minutes, the fixative removed, and GISMA was added for dyeing for 10-30 minutes, then the dyeing solution was slowly washed off with running water and dried in the air. Fourthly, the clone formation rate was counted and calculated. Clone formation rate = (number of clones/number of inoculated cells)x100%.

30 ug proteins were loaded into 10% SDS/PAGE gel and then transferred onto a PVDF membrane. After being incubated with the primary antibody in a 4°C condition for overnight and secondary antibody at room temperature for 1-2 hours, the membrane was visualized with an ECL system. The antibodies information are listed below: RABL6 (PA5-31048, Thermo Fisher Scientific), CPEB1 (ab127739, Abcam), RPL4 (PA5-98127, Thermo Fisher Scientific), CLIC5 (ab191102, Abcam), NOL4L (ab237758, Abcam), GAPDH (ab9485, Abcam).

Total RNA was extracted from tissues and cells using TRIzol reagent (Invitrogen) and subjected to reverse transcription with a TakaRa PrimeScript™ RT kit (Takara, Dalian, China). qRT-PCR assay was performed with the SYBR^® Premix Ex Taq™ II (Takara, Dalian, China) in a CFX96 Touch™ real-time PCR detection system (Bio-Rad Laboratories, Hercules, CA, USA) following the instructions. The expression of miRNAs was measured by TaqMan miRNA assays (Applied Biosystems, Foster City, CA, USA). All the primers were ordered from IDT Company.

The pLKO.1-shAR, pWPI-oeAR, pLV-miR-122-5p (oemiR-122-5p), pLKO.1-shRABL6, pWPI-RABL6 plasmids, the psPAX2 packaging plasmid, and pMD2.G envelope plasmid were transfected into HEK-293T cells using the standard calcium chloride transfection method for 48 h to get the lentivirus soup. The lentivirus soups were collected and concentrated by density gradient centrifugation and used immediately or frozen at -80°C for later use. Lentivirus soups were added into the medium until the cells were 60% confluent, polybrene (final concentration was 10 ug/ml) was added to improve transfection efficiency. 24 hours later, the medium was refreshed and transfected with lentivirus soups again. The transfected cells were used for further experiments after the transfection efficiency had been checked.

683bp of RABL6 3’UTR was chosen and the potential binding site of miR-122-5p was obtained by using the miRDB and Genome databases. Based on which, we constructed the wild type and mutant (the binding site was replaced by GGATCCG) RABL6 3’UTR psiCHECK-2 plasmids. Then, the HCC cells were transduced with the wild type/mutant plasmids and pLV/oemiR-122-5p or control/miR-122-5p inhibitor (Thermo Fisher Scientific). After being cultured for 48 hours, the cells were collected and the luciferase activity was measured by a dual luciferase reporter assay according to the protocol.

Six to eight week old nude mice were purchased from NCI and divided into three groups for injection of 1 × 10⁶ PLC/PRF/5 cells transduced with Luciferase (luc) and with/without shAR and oemiR-122-5p for 3 groups (1: pLKO+pLV; 2: shAR+pLV; 3: shAR+oemiR-122-5p). The various PLC/PRF/5 cells at 1 × 10⁶ were mixed with Matrigel (1:1) and injected into the mice under the left lobes of the liver capsules. Tumor development and metastasis were monitored by non-invasive In Vivo Fluorescent Imager (IVIS Spectrum, Caliper Life Sciences) once a week. Mice were sacrificed after 8 weeks, and tumors and any metastases were removed for studies. The study was carried out under the approval of the ethics committee of Xiangya Hospital Central South University and followed the Interdisciplinary Principles and Guidelines for the Use of Animals in Research, Testing, and Education by the New York Academy of Sciences, Ad Hoc Animal Research Committee.

All statistical analysis was conducted by SPSS 22.0 software, and the data were shown as mean ± SD. Differences in mean values between two groups were analyzed by two tailed Student’s t test and the mean values of more than two groups were compared with one-way ANOVA. We thought it as significant difference when p ≤ 0.05.

To check AR’s role in the regulation of HCC progression, we first manipulated AR (over-express or knock down) in HCC Hep3B and PLC/PFR/5 cells, then we conducted transwell invasion assay and wound healing migration assay to check the cells’ invasion and migration capacities. The results showed that oeAR (over-expressing AR) could decrease HCC Hep3B cells’ invasion and migration capacities ( Figures 1A–C ), and shAR (knocking down AR) led to the increase of the invasion and migration capacities of HCC PLC/PRF/5 cells ( Figures 1D–F ), which was also consistent with our previous data (16, 17). At the same time, we constructed the 2# shAR plasmid to knock down AR in HCC PLC/PRF/5 cells, and the results showed which could also increase the cells invasion and migration capacities, but the promotion effect was not as strong as the 1# shAR plasmid( Supplementary Figures 1A, B ), so we chose the 1# shAR plasmid for further study. To study the function of ARs in the same HCC cells, we also studied shAR in HCC Hep3B cells and oeAR in HCC PLC/PRF/5 cells, and the results of the invasion and migration capacities were also consistent with our previous data in Figure 1 and Supplementary Figures 1C–E .

Together, the data from Figure 1 and Supplementary Figure 1 indicated that AR could suppress HCC cells invasion and migration capacities.

To study the detailed mechanism between AR and HCC, and find a new pathway to provide new strategies to better suppress HCC progression, we focused on miRNAs, for the reason that many papers have reported that miRNAs might play important roles in the regulation of progression of many cancers (8–11). We first searched the published papers related to RNA sequencing in HCC, and selected 9 miRNAs that were reported to be downregulated in HCC compared to healthy group (18) ( Figure 2A ). Then we checked the miRNAs expression after knocking down/overexpression AR in HCC cells, and the results showed that 3 of 9 miRNAs (hsa-miR-122-5p, miR-192-5p, miR-199a-5p) were decreased after shAR in HCC PLC/PRF/5 cells, and 2 of the 3 miRNAs (hsa-miR-122-5p and miR-192-5p) were increased after oeAR in HCC Hep3B cells ( Figures 2B, C ). Next, we checked the miR-122-5p and miR-192-5p functions to HCC cells after adding miRNAs inhibitor or overexpressing miRNAs. The results showed that adding miR-122-5p inhibitor could partly reverse AR’s role in the regulation of the invasion capacity of HCC Hep3B cells, but not miR-192-5p inhibitor ( Figure 2D ), and wound healing migration assay also confirmed miR-122-5p’s role in HCC progression ( Figure 2E ). At the same time, we checked the proliferation capacities of the HCC cells by using MTT assay and colony formation assay, and the results also showed that oeAR could decrease the proliferation capacities of HCC Hep3B cells, and adding miR-122-5p inhibitor could partly reverse AR’s role in the regulation of the proliferation capacity of HCC Hep3B cells ( Figures 2F, G ). We also examined oemiR-122-5p in HCC PLC/PRF/5 cells to check its function, and the data showed that adding miR-122-5p could block shAR’s role in promoting HCC cells invasion, migration, and proliferation capacities ( Figure 2H and Supplementary Figures 2A–C ).

Together, the data from Figure 2 and Supplementary Figure 2 indicated that AR may decrease HCC invasion, migration, and proliferation capacities via altering miR-122-5p expression.

To check the underlying mechanism of how AR/miR-122-5p signaling regulates HCC cells progression, we searched the miRDB database (http://mirdb.org/index.html) and predicted 5 potential genes (RABL6, CPEB1, RPL4, CLIC5, NOL4L) that targeted by miR-122-5p. Then we checked these genes’ expression after shAR/oeAR in HCC Hep3B or PLC/PRF/5 cells. The results from Figures 3A, B showed that RABL6 may be the potential gene that is regulated by AR. Next we conducted the reverse assay to check if shRABL6 could reverse shAR’s function on HCC cells progression. The result showed that shRABL6 could reverse shAR’s function on HCC PLC/PRF/5 cells invasion and proliferation capacities ( Figure 3C and Supplementary Figures 3A, B ), and at the same time, oeRABL6 could reverse oeAR’s function on HCC Hep3B cells invasion, migration, and proliferation capacities ( Figures 3D, E and Supplementary Figures 3C, D ).

Together, the data from Figure 3 and Supplementary Figures 3A–D indicated that AR/miR-122-5p axis may influence HCC cells invasion and migration capacities via altering the RABL6 expression.

As the Argonaute 2 IP assay indicated miRNAs participated in the AR and RABL6 pathway ( Supplementary Figures 3E, F ), and our data proved miR-122-5p was the candidate miRNA, so, to check how miR-122-5p regulate RABL6 expression, we first searched the miRDB database (http://mirdb.org/index.html) to find out the binding sites of miR-122-5p on the 3’UTR of RABL6 mRNA and constructed wild type and mutant sequences ( Figure 4A ), and the results from the luciferase assay indicated that the luciferase activity was increased when adding the miR-122-5p inhibitor in HCC Hep3B RABL6 wild type cells compared to the control cells, but the activity in the mutant groups did not increase significantly when compared to the control group ( Figure 4B ).What is more, the luciferase activity decreased significantly when oemiR-39-5p was found in HCC PLC/PRF/5 cells in the wild type group but not mutant groups ( Figure 4C ).

Together, the data of Figures 4A–C proved that miR-122-5p could directly target 3’UTR of RABL6 mRNA to regulate the expression of RABL6.

To check the clinical significance of AR/miR-122-5p/RABL6 axis in liver cancer, we collected the data from UALCAN (http://ualcan.path.uab.edu/index.html), Kaplan-Meier Plotter (http://www.kmplot.com/analysis/index.php?p=service), GEPIA (http://gepia.cancer-pku.cn/index.html), and ENCORI (http://starbase.sysu.edu.cn/panCancer.php) databases. The results showed that AR expression was decreased in primary tumors ( Figure 5A ), and as the stages increase, the expression of AR gradually decreases ( Figure 5B ). What is more, the data from GEPIA database (Quartile) showed that low expression of AR in LIHC patients often meant worse survival rate ( Figure 5C ). At the same time, the data from ENCORI and Kaplan-Meier Plotter databases showed that miR-122-5p expression was decreased in LIHC patients and low expression of miR-122-5p often meant worse survival rate for LICH patients ( Figures 5D, E ). What’s more, the data from ENCORI and GEPIA databases concluded that RABL6 expression was increased in LIHC patients and high expression of RABL6 was connected to low survival rate in LIHC patients ( Figures 5F, G ). The co-expression analysis through ENCORI database proved that AR and RABL6 expression were negatively correlated ( Figure 5H ), which was consist with our in vitro data.

Together, the data from Figures 5A–H indicated that AR/miR-122-5p/RABL6 axis played important roles in the progression of liver cancer.

To examine the validity of the above in vitro cell line data in the in vivo mouse model, we applied the orthotopic HCC xenograft mouse model. We generated HCC PLC/PRF/5 cells with luciferase (luc) expression with/without shAR and oemiR-122-5p for 3 groups (1: pLKO+pLV; 2: shAR+pLV; 3: shAR+oemiR-122-5p). The HCC cells were inoculated into the left lobes of liver capsule of nude mice and tumor size and metastases were monitored weekly via the non-invasive In Vivo Imaging Systems (IVIS) for 8 weeks. The results revealed that shAR could promote the tumor progression in the mouse model, and oemiR-122-5p could partly reverse AR’s function in the mouse tumor model ( Figure 6A ). Importantly, results from IHC staining demonstrated that shAR led to increase the RABL6 expression, which could be partly reversed after oemiR-122-5p ( Figure 6B ).

Together, results from our preclinical study using an in vivo mouse model in Figure 6 prove that AR may play a protective role to suppress the HCC progression via altering the AR/miR-122-5P/RABL6 axis.