AP-2β inhibits hepatocellular carcinoma invasion and metastasis through Slug and Snail to suppress epithelial-mesenchymal transition.

Transcription factor AP-2β plays an important role in human cancer, but its clinical significance in hepatocellular carcinogenesis is largely unknown.
Methods: AP-2β expression was detected in human hepatocellular cancer (HCC) tissues and cell lines. The effects of AP-2β on HCC proliferation, migration, invasion, tumor formation and metastasis were evaluated by MTT, colony formation and transwell assays in vitro and mouse experiments in vivo. The association between AP-2β and miR-27a/EMT markers in HCC cell lines and tissues was analyzed.
Results: AP-2β expression was decreased in HCC tissues and cell lines. Reduced expression of AP-2β was significantly associated with more advanced tumor stages and larger tumor sizes. The overexpression of AP-2β reduced HCC proliferation, migration, invasion, tumor formation and metastasis in vitro and in vivo. Additionally, AP-2β overexpression increased the sensitivity of HCC cells to cisplatin. Moreover, AP-2β modulates the levels of EMT markers through Slug and Snail in HCC cell lines and tissues. Furthermore, oncogenic miR-27a inhibits AP-2β expression by binding to the AP-2β 3' untranslated region (UTR) and reverses the tumor suppressive role of AP-2β.
Conclusion: These results suggested that AP-2β is lowly expressed in HCC by inhibiting EMT signaling to regulate HCC cell growth and migration. Therefore, AP-2β in the novel miR-27a/AP-2β/Slug/EMT regulatory axis enhances the chemotherapeutic drug sensitivity of HCC and might represent a potential target for evaluating the treatment and prognosis of human HCC.

Introduction

Hepatocellular carcinoma (HCC) accounts for more than 5% of all cancer types and is the fifth leading cause of cancer mortality worldwide with an extremely poor prognosis 1. Early stage patients are standardly treated with surgical resection, local ablation and liver transplantation, but most patients at an advanced stage preclude these treatments, the 5-year survival rate is 25-39%, and the recurrence rate is as high as 80% 2. Sorafenib, a kinase inhibitor with antiangiogenic and antiproliferative properties, is the only targeted drug approved by the FDA to prolong the overall survival of patients with advanced HCC, but the response rate of sorafenib is as low as 2.2% 3. Currently, none of the new drugs tested have shown positive results in first-line or second-line settings after sorafenib progression 4. Therefore, the molecular events underlying hepatocarcinogenesis should largely be addressed. The identification of the key genes underlying HCC progression and metastasis and the development of more effective strategies for future therapeutic intervention are urgently required.

Transcription factor AP-2β serves as an oncogene or tumor suppressor. In one respect, AP-2β activates the human telomerase reverse transcriptase (hTERT) promoter, enhances telomerase activity, and functions as a novel tumor marker in lung cancer 5. AP-2β also mediates ERK/p38, caspase/cytochrome-c and VEGF/PEDF-dependent signaling pathways and promotes tumor growth with a poor prognosis in human lung adenocarcinoma 6. AP-2β interacts with the glioma amplified sequence 41 (GAS41), which is amplified in 23% of glioblastomas and 80% of grade I astrocytomas, suggesting their potential significance in glioma 7. In the other respect, AP-2β was highly methylated in esophageal squamous cell carcinoma (ESCC) cell lines compared with non-cancerous esophageal mucosae 8. Moreover, the CpG methylation of AP-2β is significantly associated with an unfavorable prognosis as well as adverse patient outcome in primary neuroblastoma 9. AP-2β exerts no effects on HPV16/18 E6/E7 promoters 10, but enhances β-catenin degradation, inhibits the Wnt/ β-catenin pathway, and suppresses the proliferation of cervical cancer cells 11. Some studies have shown that AP-2β was more highly expressed in lobular samples than in ductal breast carcinomas, AP-2β inhibition diminished the proliferation of lobular breast cancer (BC) cell lines and controlled cell growth in this slow-growing BC subtype 12, 13. In contrast, normal and pure ductal carcinoma in situ (DCIS) samples expressed higher levels of AP-2β than invasive tumors, AP-2β might be correlated with less aggressive features and better patient outcomes in invasive breast cancer 14. These data emphasized the complexity and importance of AP-2β in carcinogenesis.

AP-2β plays an important role in different types of human cancer, while the role of AP-2β in HCC and its associated clinicopathological features are not well investigated. In the present study, we showed the downregulation of AP-2β in human HCC tissues and cell lines. Moreover, the tumor suppressive effects of AP-2β were exerted through the inhibition of EMT signaling pathway via Slug and Snail, which was reversed by oncogenic miR-27a, it binds to the 3′ UTR of AP-2β and inhibits its expression. AP-2β overexpression sensitized HCC cells to cisplatin treatment. Therefore, AP-2β is negatively associated with HCC malignancies, and markedly inhibits HCC growth and metastasis. AP-2β will likely serve as an important therapeutic target in future applications to evaluate HCC patient prognosis and AP-2β agonist or recombinant proteins combined with chemotherapy drug maybe effective tools in confronting HCC.

Methods

Human tissues

70 liver cancer specimens and 10 adjacent non-tumorous tissues were used in this study (Table ). The experiment was approved by Human Ethics Committee of Hunan Normal University and informed consent was obtained from all patients. The immunohistochemical (IHC) analysis was performed on polyformalin-fixed and paraffin-embedded tissues 11. Briefly, sections (5 µm) were deparaffinized by two 10-min washes in xylene, then rehydrated through successive graded ethanol solutions. Endogenous peroxidase was quenched with 3% H2O2 in methanol for 10 min and washed for 10 min in PBS. Antigen retrieval was achieved by microwaving sections in 0.01 M citrate buffer (pH 6.0) for 10 min at 800 W. The tissues were blocked in 10% bovine serum albumin (BSA) in PBS for 1 h before the addition of the mouse monoclonal antibodies against AP-2β (E-8) (1:200, Santa Cruz Biotechnology Inc., Santa Cruz, CA), E-cadherin (4A2), Slug (L40C6) and Snail (L70G2) (1:200, cell signaling technologies, MA, USA) or normal mouse IgG control (1:200, BD Biosciences, CA, USA) at 4 °C for overnight. The sections were incubated with HRP-conjugated goat anti-mouse secondary antibody (diluted 1:1000, Sigma-Aldrich, St. Louis, MO, USA) for 30 min and then with 3,3-diaminobenzidine (DAB)/H2O2 for 5 min. Sections were counterstained with hematoxylin, mounted and photographed using an optical microscope (Olympus CX41, Tokyo, Japan). The protein levels of AP-2β, E-cadherin, Slug and Snail were analyzed according to the following standard. The percentage of tumor cells stained is scored as: 0 (no cell staining), 1 (≤30%), 2 (31% to 60%) and 3 (61% to 100%). Staining between two score values is given 0.5.

HCC cell lines

SMMC7721, MHCC97H, Hep3B, LO2, HepG2, HEK293 and 293T cells were cultured in DMEM medium (Gibco BRL, Grand Island, NY, USA) supplemented with 10% fetal calf serum (FBS, Hyclone, Australia), and 1% penicillin/streptomycin (Gibco). All cells were maintained in a humidified atmosphere containing 5% CO2 at 37 °C.

Plasmid construction

Plasmid pCMV-Myc-AP-2β was described previously 15. The lentiviral vectors pGC-FU-AP-2β-3Flag-IRES-Puromycin and pGC-FU-3Flag-IRES-Puromycin were constructed as described previously 11. The miR-27a stem-loop sequence was inserted into plasmid pEZX-MR03 (GeneCopoeia, Guangzhou, China). The 3′ UTR of the AP-2β gene was obtained from the TargetScan database (http://www.targetscan.org). The 3′ UTR of AP-2β was amplified from gDNA of LO2 cells and inserted into the pmirGLO vector (Promega Corporation, Madison, WI, USA). Site-directed mutagenesis of the AP-2β 3′ UTR was performed by overlapping PCR to mutate potential miR-27a-binding sites as described previously 16. Hsa-miR-27a mimics were synthesized by GenePharma Company (Shanghai, China).

The regulatory regions of Slug and Snail were predicted by Proscan and AP-2 binding sites were predicted by the JASPAR software. Genomic regions upstream of translation start codon (ATG) were amplified from human genomic DNA using primer pairs (Table ) and inserted into pGL3-Basic vector (Promega), denoted pSlug(-973/+225)-Luc and pSnail(-1000/+267)-Luc. All constructs have been sequenced for verification.

AP-2β overexpression lentivirus generation

Lentiviral particles were prepared as described in our previous work 11. Briefly, the lentivirus expression plasmids and packaging plasmids (pHelper 1.0 and pHelper 2.0) were cotransfected into 293T cells, supernatants were harvested 48 h after transfection and filtered through a 0.45-μm pore size filter (Millipore, Billerica, MA, USA) and concentrated by ultracentrifugation. The infectious titer was determined using hole-by-dilution titer assay. SMMC7721 and MHCC97H cells were infected with AP-2β-Flag-lentivirus or negative control (NC)-Flag-lentivirus at the multiplicity of infection (MOI) of 5-10 with 5 μg/mL of polybrene (Sigma) and detected on the 4th day by an invert fluorescence microscope (Zeiss Axioskop 2, LLC, US) followed by the screening of 1.5 μg/mL of puromycin for stable cell lines.

Proliferation assays

For MTT assay, 5,000 HCC cells per well were plated in 48-well plates untreated or treated with 5-30 μM of cisplatin (DDP) or 0.9% NaCl (NS) for 24 h. On day 1 to 5, cells were analyzed with 1 mg/mL 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetra zolium bromide (MTT, Sigma) at 37 °C for 4 h. Then 100 μL dimethyl sulfoxide (DMSO) per well was added to dissolve the formazan crystals. The absorbency at 490 nm was measured with a spectrophotometer (UV-2102C, Changsha, China). Liquid colony formation was performed as described previously 17. Briefly, 1,000 HCC cells were seeded in triplicate in 6-well plates and grown for over 12 days. Colonies were fixed with methanol, stained with Giemsa (BBI International, Cardiff, UK) and photographed with the digital camera (Olympus). Only colonies containing over 30 cells were counted. All experiments were carried out for at least three times.

Migration and invasion assays

For wound-healing assay, HCC cells were cultured in 6-well plates until over 95% confluence. A 100-μL pipette tip was used to generate wounds. After wound creation, the medium was changed to remove cellular debris. Three wounded areas in each well were photographed at 1 and 3 days with the invert microscope (Zeiss Axioskop 2). The transwell cell migration and invasion assays were performed in polyethylene terephthalate (PET)-based migration chambers and BD BioCoat Matrigel Invasion Chambers (BD Biosciences, Bedford, MA) with 8 μm porosity as described previously 18. HCC cells (2×104) in serum-free DMEM/F12 were seeded onto uncoated or Matrigel-coated filters in the upper chambers. DMEM/F12 containing 15% FBS was added to the lower chambers. After 48 h of incubation, cells on the upper surface of the filters were removed with a cotton swab, and the filters were fixed with 100% methanol and stained with Giemsa. The migration and invasive ability of HCC cells was calculated as the mean number of cells in all fields and expressed as the relative ratio compared with control cells. The experiments were carried out three times independently.

In vivo functional assays

The mouse experiments were performed according to the ethical guidelines for laboratory animal use and approved by the Ethics Committee of Hunan Normal University. The in vivo tumorigenic ability was performed in a xeno-graft mouse model. Approximately 1×107 of lentivirus-infected HCC cells in 0.2 mL of sterile PBS were injected subcutaneously into the left and right dorsal regions of 5-week-old female nude mice (n=6 mice/group), respectively. Mice were checked every 2 days; the formed tumors were measured as described previously 19 and analyzed by IHC. After 45 days, mice were sacrificed and tumors were excised, weighed and photographed. In vivo metastatic ability was estimated by tail vein injection, two groups of 6 mice each were given intravenous injections of 2.5×106 HCC cells, respectively. After 6 weeks the mice were sacrificed and the tumor nodules formed on the lung and liver surfaces were counted. Lungs and livers were excised and embedded in paraffin for conventional Hematoxylin and Eosin (HE) staining.

Quantitative real-time PCR

Total RNA from HCC cell lines was extracted using TRIzol reagent (Invitrogen, Carlsbad, CA), and reverse transcription was performed using a first strand cDNA synthesis kit (Roche Diagnostics, IN) according to the manufacturer's instructions. Real-time PCR analysis was performed using SYBR Green kit (Takara Bio Inc., Shiga, Japan) on 7900HT Fast Real-time PCR system (Applied biosystems, Weiterstadt, Germany). The reactions were incubated in a 384-well plate at 95 °C for 10 min followed by 40 cycles of 95 °C for 15 s and 60 °C for 30 s. The designed primers were listed in Table . The Ct value was measured during the exponential amplification phase. The relative expression levels of target genes were given by 2-∆∆Ct and log2 values were indicated as the relative difference compared to the controls.

Luciferase reporter assays

HEK293 cells were cultured in 12-well plates and transfected with reporter plasmids and pCMV-Myc-AP-2β plasmid or miR-27a mimics using Lipofectamine 2000 as described previously 20, 21.

Immunofluorescence analysis

HCC cells were cultured on glass coverslips in a 12-well plate and grown to 70% confluence. After 24 h, cells were treated as described previously 15. The primary antibodies used were rabbit polyclonal anti-Slug and Snail antibodies (AbClonal Technology, USA) while the secondary antibodies were Alexa 488 goat anti-Rabbit antibodies (Molecular Probes, USA). The nucleus was stained with Hoechst 33258 (Sigma). The fluorescence signals were collected using an upright fluorescence microscope (Zeiss Axioskop 2).

Western blotting

HCC cells were lysed in RIPA buffer as described previously 11, 15. Western blotting was performed according to the standard protocol. Information of antibodies was followed, rabbit polyclonal antibodies against MMP9 and AP-2β, mouse monoclonal antibodies against cyclin D1 (CCND1), Vimentin, β-catenin, α-catenin, Flag, β-actin and GAPDH were from Santa Cruz Biotechnology. Mouse monoclonal antibodies against AKT and phosphorylated AKT473, ERK, p-ERK, p-GSK-3β were from Cell Signaling Technology. Rabbit polyclonal antibodies against Ki67, Slug and Snail were from AbClonal Technology. Rabbit polyclonal antibodies against E-cadherin and N-cadherin were from Sangon Biotech (Shanghai, China). HRP-conjugated goat anti-rabbit and goat anti-mouse secondary antibodies were from Sigma.

Hsa-miR-27a overexpression lentivirus generation

The sequence of miR-27a pre-miRNA was cloned to pLenti-GFP lentiviral vector pEZX-MR03, recombinant and control vectors were then transfected into HEK293T cells with the Lenti-Pac HIV Packaging Mix (GeneCopoeia), and viral supernatants were collected 48 h after transfection. The lentivirus particles were purified and stored at -80 °C. After 96 h infection, stable miR-27a-infected HCC cells were screened at 1 μg/mL of puromycin.

Statistical analysis

Statistical analyses were performed using the SPSS 16.0 (SPSS Inc., Chicago, IL, USA), GraphPad software (SanDiego, California, USA) and SigmaPlot 12.5 (Systat Software Inc., San Jose, CA, USA). The Pearson's χ2 test was used to analyze the association of AP-2β expression with clinicopathologic characteristics. The expression levels of AP-2β or target genes in HCC tumor tissues and adjacent normal tissues were compared using a paired Student's t-test. Differences between gene expression were assessed by Fisher׳s exact test. Survival analyses were assessed by Kaplan-Meier plotter. Data are shown as mean ± SD from at least 3 independent experiments. Results were considered statistically significant when P < 0.05.

Results

Downregulation of AP-2β in HCC tissues and cell lines

To determine the clinical significance of AP-2β in HCC, the expression levels of AP-2β in 10 normal and 70 HCC liver tissue samples were examined by IHC staining. AP-2β was highly expressed in 70% of normal liver tissues. Strong AP-2β staining (2.5/3+) was detected in 7 (10%) of the 70 HCC tissue samples, while moderate staining (1.5/2+) was detected in 26 (37%) HCC tissue samples, and weak or negative AP-2β staining (0~1+) was detected in 37 (53%) HCC samples (Figure ). Low AP-2β expression was observed in low-grade hepatocellular cancers (I and II) and especially in high grade HCCs (III and IV) compared with that in normal liver tissues (P<0.001, paired Student's t-test; Figure ). Clinicopathological association analyses of the 70 HCCs revealed that AP-2β expression was significantly associated with advanced clinical stage and HCC tumor size (Pearson's χ2 test, P=0.004; Table ). Moreover, AP-2β expression was negatively correlated with tumor grade (Pearson's correlation coefficient, -0.415, P=0.01; Figure ). Therefore, AP-2β expression was significantly lower in high-grade HCC tissues than in normal liver tissues.

We next analyzed the expression of AP-2β proteins in human HCC lines. Low expression or the loss of AP-2β proteins was detected in human hepatoma cell lines MHCC97H, SMMC7721, Hep3B and HepG2, while high AP-2β expression was found in the normal human hepatic cell line LO2 (Figure ). Thus, low metastatic HCC cell line SMMC7721 and high metastatic HCC cell line MHCC97H were further selected to overexpress AP-2β proteins, not non-metastatic cell lines Hep3B and HepG2.

AP-2β overexpression inhibits HCC cell proliferation and tumorigenicity

To explore the role of AP-2β in human HCC, AP-2β was cloned into the lentiviral vector pGC-FU-3Flag-IRES-Puromycin, and stable HCC cell lines with AP-2β overexpressed or not (pFLAG-AP-2β and pFLAG-NC) were screened and established. The overexpression of AP-2β decreased the viability, colony number and size of HCC cells (Figure ), revealing the strong anti-tumorigenic ability of AP-2β. To further examine the effect of AP-2β on the in vivo tumorigenicity of HCC cells, AP-2β LV-infected and control NC-infected HCC cells were subcutaneously injected into the left and right dorsal flanks of nude mice, respectively. Within 45 days, the body weight of mice showed no obvious change (data not shown), but the average tumor weight and tumor volume of the AP-2β overexpressed group were markedly reduced when compared with the controls (Figure ). H&E staining confirmed that cells are more loosely arranged in the AP-2β group (Figure ). Ki67 IHC staining further confirmed that AP-2β overexpression inhibits proliferation marker Ki67 expression in HCC (Figure ). These results indicated that AP-2β overexpression markedly inhibits the tumorigenic ability of HCC cells.

AP-2β suppresses HCC cell migration and metastasis

To further investigate the effects of AP-2β on HCC cell invasion and metastasis, the in vitro invasion and in vivo metastasis assays were performed. Wound-healing assays showed that AP-2β overexpression inhibits cell migration approximately 50% at the edge of exposed regions in HCC cells (Figure ). Transwell migration assays indicated that overexpression of AP-2β leads to a marked decrease in cell motility (Figure ). Moreover, Matrigel invasion assays revealed that AP-2β overexpressed cells exhibit a significantly slower rate of cell invasion than control cells (Figure ). These results suggested that AP-2β decreases cell invasion, which was further demonstrated in vivo. The pFLAG-AP-2β cells and control cells were injected into 6-week-old nude mice via tail-vein injection to mimic cell metastasis through circulation. Six weeks after injection, the metastatic nodules that formed on the surface of the lungs were counted (Figure ). The metastatic lesions on the surfaces of the mouse lungs were demonstrated by H&E staining (Figure ). The mice in the control group had more and larger lung metastatic nodules, whereas fewer and smaller metastatic nodules were observed in mice injected with pFLAG-AP-2β cells (P<0.05, Student's t-test; Figure ). The liver cells had no obvious morphological changes (data not shown). Taken together, these findings demonstrated that AP-2β inhibits HCC cell invasion and metastasis in vitro and in vivo.

AP-2β increases the sensitivity of HCCs to cisplatin

Because of a tendency of metastasis and recurrence of HCC, post-operation chemotherapy is necessary; but, drug resistance is easily induced in HCC 22. We investigated the influence of AP-2β on the chemosensitivity of HCC cells. A 35% reduction of MTT absorbance was detected in AP-2β overexpressed cells compared with a 23% inhibition detected in control cells treated with DDP for 24 h (Figure ). These results showed that AP-2β could sensitize HCC cells to the cytotoxicity of cisplatin, consistent with the findings in endometrial cancer cells 23.

AP-2β inhibits HCC epithelial-mesenchymal transition by binding to the 5′ regulatory regions of Slug and Snail

Since EMT is a key process in tumor invasion and metastasis, we next analyzed the effect of AP-2β on EMT by detecting the expression levels of EMT markers. The results of qRT-PCR showed the increased expression of epithelial markers (E-cadherin and α-catenin) with decreased expression of mesenchymal markers (Vimentin and N-cadherin) and MMP9 in AP-2β overexpressed SMMC7721 cells (Figure ). Western blotting further demonstrated that AP-2β increases the protein levels of the epithelial markers and decreases the protein levels of mesenchymal markers in HCC cell lines (Figure ) and mouse tumor tissues (Figure ), respectively. The correlation between AP-2β and E-cadherin was further analyzed by IHC staining in HCC samples. As reported, E-cadherin expression was significantly decreased in hepatocellular cancer tissues when comparing to that in normal liver tissues (paired Student's t-test; Figure and Table ) 24. The expression of E-cadherin was positively correlated with AP-2β expression in HCC samples (Pearson correlation coefficient, 0.667, P=0.01; Figure ), and low levels of E-cadherin were primarily observed in HCCs with low AP-2β expression (Fisher's exact test, P<0.05; Table ).

To investigate the regulatory relationship between AP-2β and EMT genes, three EMT-related transcription factors (Snail, Slug and ZEB1) were detected by qRT-PCR, and the results showed that AP-2β decreases the mRNA levels of these genes (Figure ). We next used the JASPAR database to search for potential AP-2 binding sites in the regulatory regions of these genes. Two AP-2 consensus binding sites (Slug, -783~-775 bp GCCCCTGGC and Snail -501~-493 GCCCGAGGC) were identified within the upstream regions of Slug and Snail. We first examined the effect of AP-2β on Slug and Snail transcription activities. The luciferase activities of pGL3-Slug (-943/+253) and pGL3-Snail (-1073/ +260) reporter plasmids were significantly decreased in AP-2β-overexpressed cells compared with control cells (Figure ). Furthermore, the expression of Slug and Snail was downregulated by AP-2β overexpression in HCC cell lines. Intriguingly, upstream phosphorylated AKT and GSK-3β as well as p-ERK and CCND1 were downregulated in AP-2β-overexpressed cells (Figure ). The fluorescence intensity of Slug and Snail was significantly decreased by AP-2β in MHCC97H cells (Figure ). Consistent with these findings, the correlation between AP-2β and Slug/Snail was further analyzed in HCC samples by IHC staining (paired Student's t-test; Figure , Table , and Table ). High levels of Slug were primarily observed in HCCs with low AP-2β expression, but this trend did not reach statistical significance (Fisher's exact test, P>0.05; Table ). A significant inverse correlation between the expression of Snail and AP-2β was detected in HCC samples (Fisher's exact test, P<0.05; Table ). Importantly, the negative correlation between AP-2β and Slug/Snail showed statistical significance (Fisher's exact test, P<0.05, Table ). Therefore, AP-2β efficiently inhibits Slug and Snail gene expression.

MiR-27a targets the 3′ UTR of AP-2β, inhibits AP-2β expression, subsequently decreases the inhibitory effects of AP-2β in HCC

To identify miRNAs that regulate AP-2β, we used four computational programs (TargetScan, RNA22, miRWalk and miRanda) to search potential miRNAs binding to the 3′ UTR of AP-2β gene and identified the conserved miR-27a (Figure ). The wild-type or mutated 3′ UTR of AP-2β was co-transfected with miR-27a mimics into HEK293 cells. As shown in Figure , the overexpression of miR-27a suppressed the firefly luciferase activities of the 3′ UTR of AP-2β. Furthermore, the miR-27a repressed AP-2β protein expression (Figure ). However, the luciferase activity was restored by using MRE-mutated MT1 of AP-2β, but not MT2 (Figure ), suggesting that the predicted MT1 mediates the binding of the miR-27a to AP-2β. These data indicated that miR-27a binds to the 3′ UTR of AP-2β and negatively regulates AP-2β expression.

To verify the crucial role of miR-27a in AP-2β-regulated HCC cell proliferation and invasion, AP-2β-overexpressed HCC cells were infected with miR-27a LV. The results showed that miR-27a could dramatically rescue the inhibited cell proliferation (Figure ) and migration (Figure ) of AP-2β-overexpressed HCCs. Additionally, miR-27a is negatively associated with the overall survival of patients from the Cancer Genome Atlas (TCGA) data (Figure ) 25. Moreover, miR-27a also reversed the EMT transition by downregulating epithelial marker expression and upregulating mesenchymal marker expression (Figure ). These results showed that oncogenic miR-27a reverses the inhibitory effects of AP-2β in HCC.

Discussion

Hepatocarcinogenesis is a multi-factor and multi-step complex process, associated with genetic alterations that involve dominant gain-of-function mutations, the amplification and/or overexpression of oncogenes as well as the recessive loss-of-function, deletion and/or epigenetic silencing of tumor suppressor genes 26-28. AP-2β, first identified in embryogenesis 29, has been associated with tumor malignancy in lung cancer 5, and higher differentiation and a better clinical outcome in breast cancer, and neuroblastoma 9, 14. In the present study, we observed that the expression level of AP-2β was decreased in HCC tissues, low AP-2β expression was detected in 53% of HCC samples, and the AP-2β expression was negatively associated with tumor size and tumor grade (p=0.004). The correlation between AP-2β and E-cadherin, Slug/Snail was further confirmed in HCC tissues and cell lines. Moreover, 55.8% of patients with decreased AP-2β expression displayed decreased E-cadherin levels, while 53.5% of patients with decreased AP-2β expression revealed enhanced Slug and Snail levels, and only 32.6% of patients with decreased AP-2β/E-cadherin expression showed enhanced Slug and Snail levels (p<0.05, Table ). These results indicated that the combination of AP-2β with target gene expression signatures might provide novel molecular approaches for the diagnosis and clinical therapy of HCC.

HCC invasion is a key step that leads to metastasis and poor prognosis 30. Functional assays confirmed that AP-2β suppresses the cell proliferation, migration and invasion of hepatocellular cancer in vitro. These effects were efficiently reversed by oncogenic miR-27a molecules 31, which target the 3′ UTR of AP-2β. The role of AP-2β in inhibiting HCC cell migration and invasion was further identified in a mouse model of pulmonary metastasis. As expected, AP-2β inhibited cell motility in vivo. Moreover, AP-2β mediated the suppression of angiogenesis during corneal development 32. As reported, HCC is one of the most chemoresistant cancers to systemic chemotherapy, with a response rate varying from 0% to 20% 33. However, AP-2β overexpression significantly enhanced the cisplatin-induced death of HCC cells, suggesting that AP-2β has strong tumor inhibition ability and functions as an auxiliary enhancer of chemotherapeutic drugs.

EMT is well considered as a central mechanism for tumor invasion and metastasis including HCC 34-36. The downregulation of E-cadherin is the key step of EMT. Several transcription factors that can repress E-cadherin expression, including Snail, Slug, ZEB-1/2, Twist and E47 proteins, have been identified 37. Cancer-associated cascades, which include phosphoinositide 3-kinase (PI3K)/Akt-, Wnt-, Notch-, Hedgehog- or NF-kB-dependent pathways, have emerged as important regulatory signaling for EMT 38. Based on the function of AP-2β in carcinogenesis, we proposed that AP-2β participates in HCC epithelial-to-mesenchymal transition and cancer metastasis. Here, we found that AP-2β overexpression exerts a significant effect on EMT by increasing the mRNA and protein levels of epithelial markers and decreasing the expression of mesenchymal markers. These results were further supported by Western blot analysis of cultured HCC cells and immunohistochemical staining of HCC tissue samples. AP-2β inhibited the metastatic potential of HCC cells and led to fewer and smaller tumor nodules in the lungs of nude mice. Moreover, AP-2β overexpression decreased the levels of phosphorylated Akt, which correlated with decreased levels of phosphorylated GSK-3β and the EMT regulators Snail and Slug as reported 39, 40. However, the effects of AP-2β could be reversed by oncogenic miR-27a (Figure ). AP-2β was previously reported to decrease the phosphorylation of AKT and insulin receptor substrate 1 (IRS-1) in preadipocytes 41, but it remains unknown how AP-2β inhibits the versatile protein AKT in HCC. Additionally, AP-2β is involved in the cross-talk of other signaling pathways, which could decrease the levels of phosphorylated ERK and β-catenin as the same in cervical cancer 11, and CCND1 as downstream targets of ERK and Wnt signaling. Taken together, these data showed that AP-2β regulates HCC cell proliferation and migration, at least in part, by affecting EMT-related cell environment.

The relationship between obesity-induced fatty liver and HCC development has long been determined 42, 43, and some studies have showed that obesity increases cancer risk from elevated insulin and insulin-like growth factor 1 (IGF-1), sex steroids and cytokines in adipose tissues 44. Obesity-promoted HCC development was dependent on the enhanced production of the tumor-promoting cytokines IL-6 and TNF 45. Several studies have revealed that AP-2β enhances lipid accumulation and insulin resistance 41 by directly inhibiting the secretion of insulin 46 and the expression of adiponectin and leptin in adipocytes 47, enhancing the expression of inflammatory adipokines IL-6 and MCP-1 48, and decreasing the tyrosine phosphorylation of IR and IRS-1 49, which suggests the enhancing role of AP-2β in obesity and inflammation. AP-2β might enhance obesity and fat liver inflammation by IL-6 in obese patients. Thus, AKT inactivation is unlikely to enhance HCC development in obese persons, and there may be a negative feedback mechanism or epigenetic change for AP-2β to activate AKT and enhance HCC development. Similarly, rapamycin is a universal inhibitor of mTORC1-dependent S6K1 phosphorylation, but the strong negative feedback loop from S6K1 to AKT signaling could promote cell survival and chemoresistance 50. Thus, it would be interesting to elucidate the interplay of AP-2β with inflammatory cytokines, immune cells and other critical proteins in high-fat diet (HFD)-induced obesity, fat liver and hepatocellular carcinogenesis, which will expand the current knowledge of AP-2β and reveal the molecular importance of AP-2β from obesity to HCC for the development of pharmaceutical agents.

Conclusions

AP-2β functions as a novel tumor suppressor to inhibit the proliferation, migration, and invasion of HCCs by decreasing the expression of EMT regulators and markers. AP-2β could increase the cytotoxicity of chemotherapeutic drugs in HCCs. Additionally, the downregulated AP-2β expression in high-grade HCC suggests it might serve as a marker for diagnosis and combined therapy. Therefore, understanding the underlying mechanisms of AP-2β in regulating HCC has new implications for the future intervention and inhibition of HCC metastasis.

Table 1

AP-2β expression and clinical characteristics.

Features	Total	Overexpression	Low expression	P value
Sex
Male	62	24	38
Female	8	3	5	0.376
Age (median, 48 years)
<48 years	31	8	23
≥48 years	39	19	20	0.084
Tumor size (cm)
≤5	47	24	23
>5	23	3	20	0.004
Cell differentiation
Well/ Moderately	53	25	28
Poor	17	2	15	0.009
Tumor stage
I/II	47	24	23
III/IV	23	3	20	0.004

Table 2

Correlation between AP-2β expression and Slug/Snail expression in HCC samples by IHC analysis.

AP-2β expression	Cases	Snail/Slug expression		p value*
		Low, No (%)	High, No (%)
Low (≤30%)	43	12(27.9%)	23(53.5%)	0.008
High (>30%)	27	9(33.3%)	7(25.9%)
Total	70	21(30%)	30(42.9%)

All 70 samples were divided according to the proportion score to define AP-2β or Slug/Snail expression with low or high staining in HCCs.

*Fisher′s exact test.

Table 3

Correlations among AP-2β, E-cadherin, Slug/Snail expression in HCC samples by IHC analysis.

AP-2β expression	Cases	E-cadherin expression				p value*
		Low (≤30%)		High (>30%)
Low (≤30%)	43	24(55.8%)	Snail/Sluglow7(16.3%)	19(44.2%)	Snail/Sluglow5(11.6%)	0.027
			Snail/Slughigh14(32.6%)		Snail/Slughigh9(20.9%)
High (>30%)	27	4(14.8%)	Snail/Sluglow0(0%)	23(85.2%)	Snail/Sluglow9(33.3%)
			Snail/Slughigh2(7.4%)		Snail/Slughigh5(18.5%)
Total	70	28(40%)		42(60%)

All 70 samples were divided according to the proportion score to define AP-2β, E-cadherin or Slug/Snail expression with low or high staining in HCCs.

*ANOVA.