Gα12 overexpressed in hepatocellular carcinoma reduces microRNA-122 expression via HNF4α inactivation, which causes c-Met induction.

MicroRNA-122 (miR-122) is implicated as a regulator of physiological and pathophysiological processes in the liver. Overexpression of Gα12 is associated with overall survival in patients with hepatocellular carcinoma (HCC). Array-based miRNA profiling was performed on Huh7 stably transfected with activated Gα12 to find miRNAs regulated by the Gα12 pathway; among them, miR-122 was most greatly repressed. miR-122 directly inhibits c-Met expression, playing a role in HCC progression. Gα12 destabilized HNF4α by accelerating ubiquitination, impeding constitutive expression of miR-122. miR-122 mimic transfection diminished the ability of Gα12 to increase c-Met and to activate ERK, STAT3, and Akt/mTOR, suppressing cell proliferation with augmented apoptosis. Consistently, miR-122 transfection prohibited tumor cell colony formation and endothelial tube formation. In a xenograft model, Gα12 knockdown attenuated c-Met expression by restoring HNF4α levels, and elicited tumor cell apoptosis but diminished Ki67 intensities. In human HCC samples, Gα12 levels correlated to c-Met and were inversely associated with miR-122. Both miR-122 and c-Met expression significantly changed in tumor node metastasis (TNM) stage II/III tumors. Moreover, changes in Gα12 and miR-122 levels discriminated recurrence-free and overall survival rates of HCC patients. Collectively, Gα12 overexpression in HCC inhibits MIR122 transactivation by inactivating HNF4α, which causes c-Met induction, contributing to cancer aggressiveness.


INTRODUCTION
Hepatocellular carcinoma (HCC) accounts for most of primary liver cancer cases and belongs to the leading causes of death by cancer [1]; HCC is often diagnosed at an advanced stage and has a poor prognosis due to its aggressive phenotype [2]. HCC is a malignant tumor that is frequently resistant to conventional cytostatic agent [1]. Although receptor tyrosine kinase inhibitors can be used in patients with advanced HCC, the benefits are modest with prognosis remaining poor [3]. Thus, promising therapy for advanced HCC is unavailable yet. Identification of transducers and/or cell surface receptors responsible for the acquisition of HCC malignant phenotype may be of help for the development of therapeutic strategies.
Modification of the tumor microenvironment and gain of proliferative capacities of tumor cells are the influential factors leading to poor prognosis. Heterotrimeric G proteins transmit extracellular signals from G protein-coupled receptors (GPCRs) to intracellular effector molecules. Much attention has been paid to Gα 12 transforming gep oncogene because the G protein mediates growth, migration, and metastasis [4]. It is expected that Gα 12 overexpression augments pathophysiological functions of the GPCRs interacting with sphingosine-1-phosphate (S1P), lysophosphatidic www.impactjournals.com/oncotarget acid (LPA), thrombin, and angiotensin-II [5][6][7]. Moreover, levels of the ligands are often elevated in HCC and may contribute to proliferation, adhesion, invasion, and metastasis of HCC, representing poor prognosis [8]. However, little information is available on the functional role of Gα 12 in the factors or components that leads to the aggressive phenotype of HCC.
A set of microRNAs (miRNAs) are globally dysregulated in cancer [9]. Mice with conditional deletion of Dicer-1 in hepatocytes provided the evidence that the miRNA in the liver plays a role in inflammation and cell cycle regulation [10,11]. Furthermore, hepatocytespecific Dicer 1 knockout mice developed spontaneous HCC [11]. In particular, miR-122 is a predominant liverenriched miRNA, which may act as a tumor suppressor [12]. Previous studies from our laboratory reported overexpression of Gα 12 in the patients with HCC and the association between Gα 12 dysregulation of p53-responsive miRNAs and epithelial-mesenchymal transition (EMT) of cancer cell [13]. Because miR-122 is the most greatly and significantly suppressed by activated Gα 12 among those down-regulated in the microarray analysis, this study investigated the effect of miR-122 dysregulation on cancer cell malignancy using cell and animal models, and human HCC samples. Here, we report c-Met as a new target of miR-122. Our findings also reveal the role of Gα 12 pathway in the activity of hepatocyte nuclear factor 4α (HNF4α) required for the expression of MIR122. To verify the relationship between decrease of miR-122 by Gα 12 and HCC progression, the levels of Gα 12 , miR-122, or c-Met were measured for human HCC samples and correlated with changes in recurrence-free and overall survival rates of the patients.

Dysregulation of miR-122 by Gα 12
To evaluate whether Gα 12 expression is associated with prognosis of HCC, we first carried out a survival analysis using electronic medical records and 59 primary human HCC samples stratified based on Gα 12 levels measured as in the previous study [13]. The cutoff for strong Gα 12 intensity was set at '>3-fold' difference in HCC/NT avg . The intensity of Gα 12 was strongly detected in 28.8% (17/59) of the cancerous samples and significantly correlated with shorter overall survival in HCC patients ( Figure 1A). Next, we compared miRNA microarray profiles using wild-type (WT) Huh7 and Huh7 cells stably transfected with a constitutively active mutant of Gα 12 (Gα 12 QL-Huh7) ( Figure 1B). In our finding, activated Gα 12 most substantially and significantly repressed miR-122: either stable or transient transfection with Gα 12 QL reduced the levels of mature form of miR-122 in Huh7 or HepG2 cells ( Figure 1C). To verify the effect of siRNA knockdown of Gα 12 on miR-122 expression, we examined the effects of four different siRNAs on the basal Gα 12 expression in HepG2 cells ( Figure 1D). Transfection with each of the four different siRNAs caused sufficient knockdown of Gα 12 . Consistently, miR-122 levels were significantly decreased in the samples except siRNA #2 (siGα 12 #2). Based on the targeting efficacy on miR-122 expression, siGα 12 #1 was selected in the subsequent experiments. In Gα 12 QL-Huh7 or SK-Hep1, siRNA knockdown of Gα 12 promoted increase of miR-122. Our data indicates that Gα 12 overexpressed in liver cancer mostly greatly dysregulates the expression of miR-122.

Inhibition of c-Met by miR-122
Having identified the most evident decrease of miR-122 by the activated form of Gα 12 , we searched for the target of miR-122 as a protein possibly implicated in the aggressiveness of HCC. Bioinformatic analyses using Microcosm program enabled us to select the targets putatively regulated by miR-122. Among the putative, but yet unidentified, targets of miR-122, c-Met was the most enriched interacting molecule of the pathway in cancer (Figure 2A). We found a putative miR-122 binding site within the 3'-untranslated region (3'UTR) of c-Met mRNA using RNA 22 program ( Figure 2B). To clarify the role of miR-122 in regulating c-Met, in vitro functional assays were done after enhancing or silencing the miRNA. Transfection with miR-122 mimic unchanged c-Met mRNA level ( Figure 2C). miR-122 mimic transfection notable decreased c-Met protein levels in three different cell lines, whereas miR-122 inhibitor increased them ( Figure 2D). Consistently, miR-122 mimic diminished luciferase expression from pEZX-c-Met-3'UTR luciferase construct comprising the c-Met 3'UTR region ( Figure 2E). Transfection with miR-122 inhibitor enhanced the 3'UTR reporter activity. These results show that miR-122 directly inhibits c-Met translation by targeting the 3'UTR region.

c-Met overexpression by activated Gα 12 and the effects of LPA and S1P on miR-122 and c-Met expression
Having identified the link between miR-122 and c-Met downstream of Gα 12 , we next confirmed the effect of Gα 12 modulations on c-Met. Either stable or transient transfection of Huh7 (or HepG2) cells with Gα 12 QL increased c-Met levels ( Figure 3A), whereas knockdown of Gα 12 reduced them ( Figure 3B). In addition, transfection with miR-122 mimic diminished the induction of c-Met by Gα 12 QL ( Figure 3C).
Previously, we observed that Gα 12 is expressed to greater levels in mesenchymal cell lines (SK-Hep1 and SNU449) than epithelial cell lines (Huh7 and HepG2) www.impactjournals.com/oncotarget [13]. To further link Gα 12 and miR-122 physiologically, we examined their levels in a panel of human HCC cell lines. The GNA12 transcript levels were both higher in the latter than the former ( Figure 3D, left). Consistently, miR-122 contents were lower in the mesenchymal cell lines ( Figure 3D, right). The c-Met levels were also higher in SK-Hep1 and SNU449 than Huh7 and HepG2, as were GNA12 mRNA levels ( Figure 3E). All of these results indicate that increased levels of Gα 12 causes the induction of c-Met by deregulating miR-122.

Gα 12 inactivation of HNF4α necessary for the basal expression of MIR122
To precisely define the underlying basis of miR-122 repression by Gα 12 signaling, the levels of miR-122 primary transcript and of its precursor form were measured in Gα 12 QL-Huh7 cells. Activated Gα 12 decreased both the primary and the precursor forms of miR-122 transcript levels ( Figure 4A, left and right), suggesting that activated Gα 12 inhibits MIR122 gene transcription. HNF4α, a transcription factor belonging to the HNF family members, may regulate the MIR122 gene [16]. As a continuing effort to find the basis of miR-122 dysregulation by Gα 12 , we assessed the enhancing or silencing effect of Gα 12 on HNF4α; transfection with Gα 12 QL diminished HNF4α level, whereas siRNA knockdown of Gα 12 accumulated it ( Figure 4B). In quantitative real-time polymerase chain reaction (qRT-PCR) analysis, stable transfection of Huh7 cells with Gα 12 QL nullified HNF4α mRNA levels ( Figure  4C, upper). However, transient transfection of Gα 12 QL did not change them, suggesting that the decrease in HNF4α mRNA in Gα 12 QL-Huh7 cells may have resulted from the adaptive change. Next, we assessed whether Gα 12 facilitates post-translational modification of HNF4α for destabilization. In Huh7 cells, activated Gα 12 increased HNF4α ubiquitination for degradation ( Figure 4C, lower). In addition, Gα 12 QL overexpression promoted c-Met level, which was reversed by overexpression of HNF4α (Figure Immunoblotting for c-Met. HepG2 cells were transfected with control siRNA or Gα 12 siRNA for 48 h and were continuously treated with either LPA or S1P as described in panel F. www.impactjournals.com/oncotarget 4D). These results provide evidence that Gα 12 decreases miR-122 levels by inhibiting HNF4α activity, which may contribute to c-Met up-regulation.

Inhibition of cancer cell aggressiveness by miR-122
c-Met activation triggers a variety of cellular responses, including survival, proliferation, and angiogenesis [17]. Multiple signaling pathways such as the mitogen-activated protein kinase/extracellular signalregulated kinases (ERK), signal transducer and activator of transcription 3 (STAT3), and Akt/mammalian target of rapamycin (mTOR) pathways are involved in tumor biology [18]. We investigated the effect of activated Gα 12 and miR-122 mimic transfection on the signaling pathways of c-Met using a cell model. Gα 12 QL transfection promoted phosphorylation of ERK, STAT3, Akt, and mTOR in Huh7 cells, whereas transfection with miR-122 mimic abrogated this effect ( Figure 5A). Evasion of apoptosis is a crucial event during malignant transformation [19].
To further understand the mechanism of Gα 12 oncogenic activity, we assessed the role of Gα 12 -miR-122 pathway in tumor cell death. Transfection of the cells with miR-122 mimic diminished the ability of Gα 12 QL to increase procaspase 3 and B-cell lymphoma 2 (Bcl-2) levels, but enhanced poly[ADP-ribose]polymerase 1 (PARP1) cleavage ( Figure 5B), supporting the induction of cell death. Adaptation to nutrient deprivation is presumed to be one of the prerequisites for cancer cells to survive in the tumor microenvironment [20]. At 3 days after serum starvation, up to 21% of WT-Huh7 cells underwent late apoptosis while stable transfection of Gα 12 QL significantly decreased the population of Annexin V + /PI + , indicative of rescue of the cells from the loss of membrane integrity and death ( Figure 5C). As compared with control miRNA mimic, miR-122 mimic transfection significantly (~3-fold) facilitated the apoptosis of Gα 12 QL-Huh7 cells caused by serum starvation for 3 days, strengthening the concept that miR-122 plays a critical role in sensitizing cancer cell apoptosis to stimulus.
In Gα 12 QL-Huh7 cells, DNA synthesis rate was much augmented as compared to control, which was repressed by transfection with miR-122 mimic ( Figure  5D). Consistently, stable transfection with Gα 12 QL promoted Huh7 cell proliferation under anchorageindependent condition, and this effect was attenuated by  miR-122 mimic transfection ( Figure 5E). To explore the biological significance of Gα 12 signaling in angiogenesis, we evaluated capillary tube formation of bovine aortic endothelial cells (BAECs) using the conditioned media collected from WT-Huh7 or Gα 12 QL-Huh7 cells transfected with control mimic or miR-122 mimic. A clear difference was found in capillary tube formation after miR-122 mimic transfection ( Figure 5F).

Gα 12 knockdown effects in a tumor-xenograft model
In a previous study, shRNA inhibition of Gα 12 resulted in a profound anti-tumor effect in a tumor xenograft animal model using SK-Hep1, a mesenchymal type of tumor cell [13]. These samples were used in the present study to further evaluate Gα 12 impact on tumor aggressiveness. Immunoblottings and immunohistochemistry showed that shRNA knockdown of Gα 12 suppressed c-Met expression, but increased HNF4α levels in the xenograft tumor tissue ( Figure 6A and B). Consistently, the phosphorylation of ERK, STAT3, Akt and mTOR, which are downstream molecules from c-Met, was all diminished in the tumors depleted of Gα 12 ( Figure  6C). Similarly, procaspase 3, PARP1, Bcl-2, and vascular endothelial growth factor (VEGF) levels were attenuated ( Figure 6C). Moreover, knockdown of Gα 12 facilitated tumor cell death, as indicated by an increase in terminal transferase-mediated dUTP nick-end labeling (TUNEL) staining intensity ( Figure 6D). Similarly, Ki67 staining intensities were also reduced in the xenograft tumor samples ( Figure 6D). Our results provide strong evidence that Gα 12 inhibition impedes the survival and growth advantage of mesenchymal liver tumor cells, which may be associated with c-Met suppression.

Association of Gα 12 /miR-122/c-Met changes with HCC patient survival
To further explore the relationship between Gα 12 and miR-122 (or c-Met), we examined the expression of miR-122 and c-Met in tissues from 59 human primary HCC and matched non-tumorous (NT) tissues. The chi-square test showed a significant association between Gα 12 and miR-122 expression ( Figure 7A), whereas the Pearson or Spearman analysis failed to do so (Supplementary Figure  1). Immunoblottings confirmed overexpression of Gα 12 and c-Met in the HCC compared to adjacent NT ( Figure 7B, upper). In addition, a positive and significant correlation existed between Gα 12 and c-Met in the samples ( Figure  7B, lower). In a subgroup analysis, we found that miR-122 repression and c-Met induction were distinct in TNM stage II and III tumors (n=25) compared to TNM stage I tumors (n=34) ( Figure 7C), consolidating the clinical relevance of miR-122 and c-Met changes with tumor stage progression (i.e., aggressive feature). Moreover, HCC patients with high Gα 12 and low miR-122 had the poorest prognosis (i.e., the lowest overall survival and highest probability of tumor recurrence), whereas those with low Gα 12 and high miR-122 had the best outcomes ( Figure 7D, upper). Also, we verified that the patients with high c-Met in HCC had shorter overall survival and higher possibilities of tumor recurrence as compared with the patients with low c-Met in HCC ( Figure 7D, lower). These results support the conclusion that Gα 12 overexpression causes miR-122 dysregulation, promoting c-Met induction, which may deteriorate the prognosis, recurrence-free and overall survival rates of HCC patients.

DISCUSSION
miR-122 levels in the liver amount to 135,000 copies per normal human hepatocyte [21], representing 72% of all miRNAs. miR-122 is necessary for the control of lipid and glucose metabolism, and other physiological activities in the liver [22,23]. Mir122a-/-mice spontaneously develop liver tumors [24]. In addition, miR-122 is frequently under-expressed in human HCC [25]. Moreover, the loss of miR-122 alters hepatic phenotype, assisting gain of metastatic properties, which strengthens the concept that miR-122 may be an intrinsic tumor suppressor gene in the liver [24][25][26]. Nevertheless, the upstream regulator of miR-122 and the basis underlying miR-122 dysregulation in HCC had been elusive. Our results shown here demonstrate for the first time that Gα 12 overexpressed in the tumor decreases miR-122, accounting for cancer aggressiveness and poor prognosis of the patients with HCC.
Gα 12 may also play a role in the progression of cancer malignancy through other pathways. EMT has been implicated in tumor invasion and metastasis [27]. Previously, we found that Gα 12 was overexpressed in HCC, which caused ZEB1 induction through p53responsive miRNAs deregulation, promoting EMT of liver tumor [13]. It is noteworthy that hepatocyte growth factor (HGF) elicits mitogenic, and morphogenic properties [28,29], and reduces the expression of E-cadherin with increase of N-cadherin [30]. Promotion of cancer EMT program and cell migration by HGF depends on c-Met [31]. A novel finding of this study is the identification of c-Met as a new target of miR-122, as evidenced by the outcomes of in vitro functional assays using a construct comprising c-Met 3'UTR and bioinformatic analysis. Several miRNAs including miR-103 and -203 also affect c-Met expression [32]. Of the miRNAs normally enriched in hepatocytes, activated Gα 12 most greatly and significantly decreased miR-122 levels. Since miR-122 is the most abundant in the liver, the repression of miR-122 by Gα 12 would greatly alter cell biology in association with c-Met-mediated aggressiveness (e.g., anti-apoptosis, proliferation, and angiogenesis). Because miRNAs have overlapping targets, other targets of miR-122 including cyclin G1 and ADAM17 may additionally be involved in HCC pathogenesis [25,33]. Hence, c-Met and others would work together for tumor malignancy [33]. miRNAs are transcribed by RNA polymerase II as pri-miRNAs which undergo nuclear export and cytoplasmic cleavage to generate mature forms [34]. Pri-miR-122 had been identified as a non-coding RNA, hcr [21]. The expression of miR-122 relies on liver-enriched transcription factors in the developing liver or cell lines [16]. In particular, HNF4α is abundantly expressed in the liver, and directly binds to the promoter region of the MIR122 gene [35]. In the present study, activated Gα 12 decreased the levels of primary and precursor forms of miR-122, supporting the role of Gα 12 in inhibiting MIR122 transcription. An important finding of our study is the ability of Gα 12 to negatively control HNF4α, a transcription factor required for the constitutive MIR122  Figure 1A. The weak and strong intensities were defined by Gα 12 expression level in HCC/NT avg ≤3-fold; and Gα 12 expression level in HCC/NT avg >3-fold, respectively. The data were analyzed by Chi-square test. B. Correlation between Gα 12 and c-Met levels in HCC. Immunoblottings for Gα 12 and c-Met were carried out on the homogenates of 59 pairs of HCC samples and were normalized to those of β-actin. Shown above are the representative blots for HCC (T) and non-tumorous (N) samples. Two variables were correlated by Pearson (r) or Spearman (ρ) correlation coefficients (lower). C. The relative levels of miR-122 or c-Met in the patients with HCC in different TNM stages. The line indicates the mean value. Statistical analysis was done using Student's t-test. D. Kaplan-Meier survival curves for 9 HCC patients with respect to Gα 12 and miR-122 expression (upper). Tumor samples were divided into four groups according to mean fold changes (T/N) of Gα 12 or miR-122. Recurrence free survival or overall survival of patients with low (≤2-fold) or high (>2-fold) c-Met transcript levels (lower). P value was generated by log-rank test and Breslow test. E. A schematic diagram illustrating the proposed mechanism by which Gα 12 dysregulation of miR-122 contributes to poor prognosis in patients with HCC. gene expression [35]. Gα 12 transduces c-Jun N-terminal kinases (JNK)-dependent signaling for Nrf2 and IκBα ubiquitination [5,37]. It also decreased p53 and FOXO1 levels through the induction of MDM2, an E3 ubiquitin ligase [13,38]. Since Gα 12 activates JNK1 [36], miR-122 repression by Gα 12 may be associated with JNK1dependent inhibition of HNF4α. Our data showing an increase in HNF4α ubiquitination by Gα 12 supports the idea that JNK1 activated by Gα 12 may decrease miR-122 through the inhibitory phosphorylation and ubiquitination of HNF4α [23].
There has been an increasing interest which focuses on heterogeneous inter-receptor networks. A majority of GPCRs has growth-promoting activity by trans-activating receptor tyrosine kinases [40]. Our result shown here identifies c-Met as a novel target of the Gα 12 pathway. Ligand activation of c-Met causes phosphorylation of two tyrosine residues, which activates Ras/MAPK and PI3K/ AKT pathways through recruitment of adaptor proteins, promoting tumor growth and metastasis [17]. Since certain GPCR ligands transactivate c-Met [41], this may act as a point of convergence for different cell-surface receptors. Gα 12 increases the activities of Rho/Rac-dependent AP-1 and others (e.g., STAT3) [42,43], conferring on cells the ability to recruit multiple receptors, non-receptors, and Ser/Thr kinases for neoplastic transformation and progression. Hence, Gα 12 overexpression in HCC makes a positive feed-forward loop in activating signaling such as ERK1/2, STAT3, Akt, and mTOR through up-regulation of c-Met as a consequence of decrease of miR-122 in the tumor tissue.
The balance between proliferation and apoptosis is frequently disrupted in tumor tissues, and the acquisition of abnormal growth rates and anchorage-independent growth advances tumor malignancy [44]. In our findings, Gα 12 QL transfection prevented apoptosis, but promoted soft agar colony growth of tumor cells. The results that Gα 12 QL transfection initiated capillary tube formation (i.e., a late stage of angiogenesis) and this effect was antagonized by miR-122 mimic transfection support the notion that miR-122 acts as a suppressor of liver tumor progression in severity. This concept is re-enforced by the finding that Gα 12 knockdown not only attenuated c-Met, the downstream signals, and Ki67 intensities in a xenograft model, but increased tumor cell death.
In patients with HCC, Gα 12 levels correlated with either decrease of miR-122 or c-Met induction in the HCC samples. Moreover, these changes correlated with TNM stages, suggestive of the role of miR-122 and c-Met dysregulation in the tumor stage progression. Similarly, miR-122 was depressed in a subset of HCC harboring c-Met signature [26]. Our findings support the reciprocal link between miR-122 and c-Met expression downstream from increase of Gα 12 , extending basic scientific information to clinical arena. The result also supports the role of Gα 12 as an independent prognostic factor for tumor recurrence particularly in combination with low miR-122. Overall, our findings provide an insight into (1) the inhibitory role of Gα 12 in miR-122 targeting c-Met, and (2) the crosstalk between GPCR and c-Met in HCC, implying that intervention of the Gα 12 pathway may be of help to improve c-Met-targeted therapy.

Human liver samples
A total of 59 paired samples of HCC and NT tissue were obtained from the Bio-Resource Center at the Asan Medical Center, Seoul, Korea [13]. Informed consent was provided in accordance with the ethical guidelines of the 1975 Declaration of Helsinki. Written informed consent was obtained from all patients. The study protocol was approved by institutional review board of Asan Medical Center (#2012-0133)

Microarray
The purified labeled miRNA probes were hybridized to 8×15 K human miRNA microarrays from Agilent Technologies previously [13]. Our dataset is available from NCBI's Gene Expression Omnibus (accession number GSE44079).

Real-time PCR assays
Total RNA was extracted using Trizol (Invitrogen, Carlsbad, CA). qRT-PCR assays for miRNAs were performed using miScript SYBR Green PCR kit (Qiagen, Valencia, CA), whereas those for mRNAs were done using LightCycler ® DNA master SYBR Green-I kit (Roche, Mannheim, Germany) according to the manufacturer's instruction.

Immunoblot analysis
Sodium dodecyl sulfate-polyacrylamide gel electrophoresis was carried out using whole cell lysates or liver homogenates according to the previous method [23]. The proteins of interest were visualized using an ECL chemiluminescence detection kit (Amersham Biosciences, Amersham, UK). At least three independent experiments were performed. Scanning densitometry of the immunoblots was performed with the Image Scan and Analysis System (Alpha Innotech Corp, San Leandro, CA). The band intensity was measured using Adobe Photoshop (Adobe Systems, Inc., San Jose, CA).

3'UTR luciferase assay
The miRNA 3'UTR target clone (Luc-MET-3'UTR) was purchased from GeneCopoeia (Rockville, MD), which contains renilla luciferase as internal control fused downstream to a firefly luciferase. The cells were cotransfected with control or c-Met 3'UTR luciferase vector and miR-122 mimic (or inhibitor) or its relative control using FuGENE ® HD Reagent (Roche, Indianapolis, IN). After 48 h of transfection, firefly and renilla luciferase activities were measured using Luc-Pair miR Luciferase Assay (GeneCopoeia) according to the manufacturer's protocols.

Immunoprecipitation assay
To assess HNF4α ubiquitination, cells were transfected with a plasmid encoding His-tagged ubiquitin (His-Ubi) for 6 h. Transfected cells were then maintained in Eagle's minimum essential medium containing 1% FBS for 18 h. Cell lysates were incubated with anti-HNF4α antibody overnight at 4°C. After immunoprecipitation, the antigen-antibody complex was precipitated following incubation for 2 h at 4°C with protein G-agarose. The immune complex was solubilized in 2×Laemmli buffer and boiled for 5 min. The samples were immunoblotted with anti-ubiquitin antibody. www.impactjournals.com/oncotarget

Flow cytometric analysis of apoptosis
Apoptosis was analyzed by the FITC-Annexin V plus PI staining method. The transfected cells were harvested by trypsinization. After washing with phosphate buffered saline (PBS) containing 1% FBS, the cells were stained with 5 µl FITC-Annexin V and 2 µg/ml PI. The fluorescence intensity in the cells was assessed using BD FACSCalibur II flow cytometer and the CellQuest software (BD Biosciences, San Jose, CA). In each analysis, 20,000 gated events were recorded.

Thymidine incorporation
The rate of DNA synthesis was measured using [methyl-3 H]-thymidine incorporation assay. Post-confluent cells in 12-well plates were incubated with 10% FBS for 24 h after transfection. The cells were pulse-labeled with 1 µCi/ml [methyl-3 H]-thymidine for 8 h, washed with PBS twice, fixed with 5% trichloroacetic acid for 30 min, and finally dissolved in 0.5 N NaOH containing 0.1% sodium dodecyl sulfate. The radioactivity was measured using a liquid scintillation counter (PerkinElmer, Waltham, MA).

Agarose colony-forming assay
A total of 5×10 3 cells were suspended in 1.5 ml of DMEM medium containing 10% FBS and 0.3% agarose, were plated in 35-mm 0.6% base agar dishes, and were incubated in 37°C and 5% CO 2 incubator for 3 weeks. Then, the cells were fixed in 3.7% paraformaldehyde and stained with 0.005% crystal violet. Colonies (diameter of more than 20 µm) were counted under a microscope. The colony formation assay was performed in triplicate.

Capillary tube formation assay
BAECs were starved for 5 h before seeding 1.5×10 4 cells onto growth factor-reduced Matrigel-coated 96well plates (Nalge Nunc Int. Corp., Rochester, NY), and were incubated in a conditioned medium harvested from transfected Huh7 cells at 37°C in 5% CO 2 for 6 h. The cells were photographed under a light microscope (magnification, ×100). Tube formation was quantified by counting number of branches per microscopic field in 3 randomly selected fields using Image J software (NIH).

Xenograft mouse model
Animal studies were conducted in accordance with the institutional guidelines for care and use of laboratory animals. A subcutaneous xenograft tumor model was previously established in BALB/c nu/nu mice using shCon-or shGα 12 -SK-Hep1 cells (n = 8-9) [13].

Immunohistochemistry
The tumor tissue sections were subjected to immunohistochemistry. Tumor xenografts were fixed in 10% formalin, and then embedded in paraffin. The 4-μmthick tissue sections were immunostained with antibodies of interest.

TUNEL assay
TUNEL assay was carried out using the DeadEnd Colorimetric TUNEL System (Promega, Madison), according to the manufacturer's instruction.

Statistical analysis
Data were shown as the mean±S.E. from at least three independent experiments. Statistical significance was assessed using SPSS 20.0 by one-way analysis of variance procedures and Student's t-test. The Kaplan-Meier method was used for survival analysis. The log-rank test and/or Breslow test were used to compare survival between groups. Chi-square tests were used to compare the categorical variables. Coefficients of correlation were determined by the Pearson or Spearman analysis. P values of <0.05 were considered statistically significant. www.impactjournals.com/oncotarget PCR, quantitative real-time polymerase chain reaction; S1P, sphingosine-1-phosphate; STAT3, signal transducer and activator of transcription 3; TUNEL, terminal deoxynucleotidyl transferase dUTP nick end labeling; UTR, untranslated region; VEGF, vascular endothelial growth factor; WT, wild-type