The induction mechanism of HNF-4α by spherical cell shape in human hepatoma cells, FLC-4, was investigated. To get insight into the induction mechanism of HNF-4α in three-dimensional FLC-4 cells, mRNA microarray analysis was performed. The gene expression related to drug metabolism and nuclear receptors, such as LXRα, was elevated in spherical FLC-4 cells. We found the first time that the expressions of genes related to malignancy of hepatoma cells, such as HIF-1α, c-Myc and VEGFC, were downregulated by spherical cell shape. Network analysis revealed that HNF-4α would elicit both the enhancement of hepatocyte-specific gene expression and suppression of malignancy. Since HNF-4α gene expression was known to be regulated by microRNA, we inferred that spherical cell shape would induce HNF-4α gene expression through microRNA. To investigate the possibility of such a mechanism, mRNA–microRNA interactions were examined using microRNA microarray and bioinformatics analysis. The level of miR-24, a microRNA targeting HNF-4α, was reduced in spherical FLC-4 cells. On the other hand, spherical cell shape-induced miR-194 and miR-320c would directly downregulate SLC7A5 and E2F1 gene expression, respectively, which are both related to malignancy. Our study suggested that spherical cell shape would induce HNF-4α gene expression and consequent enhancement hepatocyte-specific functions. Spherical cell shape itself would suppress malignancy in FLC-4 cells through microRNA, such as miR-194 and miR-320c.
Three-dimensional culture of rat primary hepatocytes could restore cell polarity and upregulated many liver-specific gene expression including hepatocyte nuclear factor (HNF)-4α, albumin and apo A-I (Oda et al., 1995; Oda et al., 2008). Similarly, three-dimensional culture of human primary hepatocytes could improve the expression levels of drug metabolizing enzymes (Pampaloni et al., 2009). However, long term culture of human primary hepatocytes is difficult to achieve since these cells present limited growth activity and life-span, even in three-dimensional culture (Guillouzo and Guguen-Guillouzo, 2008). To overcome this issue, hepatoma cell lines are commonly used to study liver functions on the long term, though hepatoma cell lines generally lack a substantial set of liver-specific gene expression, especially major cytochrome P-450 (CYP) (Wilkening et al., 2003).
Functional liver cell (FLC) cell lines, human hepatoma cell lines, express and secrete liver-specific proteins at a high level (Hasumura et al., 1988; Homma et al., 1990). Spherical cell shape in FLC-4 cells enhanced liver-specific gene expression, including albumin, apo A-I, CYP3A7, at a similar level to that of human liver (Laurent et al., 2012). However, common hepatoma cell lines did not show higher liver-specific functions in three-dimensional culture (Laurent et al., 2012). We concluded that cell shape per se, but not single components of EHS-gel, enhanced this gene expression through HNF-4α, a crucial liver-enriched transcription factor (LETF) regulating a broad range of liver functions (Laurent et al., 2012). Up to now, it is unclear how cell shape induces LETF gene expression.
Recently, microRNA, which are 21–23 nucleotide non-coding RNA, were described as post-transcriptional regulators of gene expression through repressing many specific target mRNA (Ambros, 2004). miR-122, miR-192 and miR-194 are highly expressed in the liver and were shown to regulate liver functions (Girard et al., 2008; Xie et al., 2011). More recently, microRNA–LETF regulatory feedback loops have been found (Takagi et al., 2010; Zeng et al., 2010). In particular, miR-24 was shown to target directly HNF-4α and its downstream target gene expression, such as CYP7A1, in HepG2 cells (Takagi et al., 2010).
In this study, the mechanism of HNF-4α induction by three-dimensional cell shape in FLC-4 cells was explored by performing both mRNA and microRNA microarray analysis and integrating these data by bioinformatics. Gene expression analysis suggested that liver functions were increased and malignant phenotype was repressed in spherical FLC-4 cells. We found that spherical cell shape itself would repress malignancy-related gene expression through microRNA, including miR-194 and miR-320c.
Results and Discussion
Microarray of mRNA revealed that spherical cell shape enhanced drug metabolism and suppressed malignancy in FLC-4 cells
Spherical cell shape of FLC-4 cells on EHS-gel (Fig. 1A) induced HNF-4α gene expression in FLC-4 cells (Fig. 1B). Gene expression of HNF-4α was demonstrated to be regulated by many transcription factors, including LETF (Hatzis and Talianidis, 2001). To understand how spherical cell shape induced HNF-4α gene expression, we performed mRNA microarray and bioinformatics approach. Our results revealed that 83 genes were upregulated, and that 87 genes were downregulated in spherical FLC-4 cells. Various functions were altered by spherical cell shape in FLC-4 cells (Table 1). Among them, gene expression related to drug metabolism was upregulated according to pathway analysis (Table 1). Surprisingly, we found that cancer related gene expression was the most changed by spherical, and was downregulated by spherical cell shape of FLC-4 cells (Table 1).
Hepatocyte-specific gene expression was induced and malignant tumor-related gene expression was suppressed by spherical cell shape in FLC-4 cells.
Spherical cell shape enhanced hepatocyte-specific gene expression in FLC-4 cells
Since drug metabolism was upregulated by spherical cell shape in mRNA microarray (Table 1), we quantified this gene expression using qRT-PCR. Gene expression related to phase I drug metabolizing enzymes (CYP2C9, CYP3A7 and CYP4F2) was elevated in spherical FLC-4 cells (Fig. 1C). CYP2C9 and CYP3A are importantly expressed in the liver (Kamiyama et al., 2007; Rana et al., 2010). Spherical cell shape enhanced the gene expression related to phase II (glutathione S-transferase (GST)A3, sulfotransferase (SULT)1B1, UDP-glucuronosyltransferase (UGT)2B15) and drug efflux pump so-called phase III (ATP-binding cassette (ABC)C2, ABCC3) in FLC-4 cells (Fig. 1C). Drug metabolizing enzyme gene expression, such as CYP2C, CYP3A, SULT and ABCC were shown to be induced by HNF-4α (Kamiyama et al., 2007). We hypothesized that spherical cell shape-dependent induction of HNF-4α would enhance the expressions of drug-metabolizing enzymes in FLC-4 cells.
The gene expression of several nuclear receptors was upregulated in microarray experiment (supplementary material Table S1), especially liver X receptor (LXR)α (NR1H3) (nuclear receptor subfamily 1, group H, member 3) which was the most elevated by spherical cell shape in FLC-4 cells (Fig. 1D; supplementary material Table S1). LXRα is crucial in the maintenance of lipid metabolism (Baranowski, 2008). Hence, we focused on this transcription factor. The DNA-binding activity of LXRα was also increased (Fig. 1E). Our data indicated that spherical cell shape would induce lipid metabolism in FLC-4 cells through LXRα.
Malignancy-related gene expression was downregulated by spherical cell shape in FLC-4 cells
Spherical cell shape suppressed gene expression related to cancer malignant phenotype (Table 1; supplementary material Table S2). Among this gene expression, spherical cell shape suppressed hypoxia-inducible factor (HIF)-1α (Fig. 1F). HIF-1α was reported to induce c-Myc and VEGFC, which are involved in malignancy (Yoo et al., 2011; Zampell et al., 2012). Both c-Myc and VEGFC gene expression was reduced in three-dimensional FLC-4 cells (Fig. 1F). The gene expression of E2F1 (Fig. 1F), a downstream target of c-Myc (Li et al., 2003), was inhibited. HIF-1α and c-Myc were responsible for the shift from oxidative phosphorylation to glycolysis in cancer cell metabolism (Yeung et al., 2008). Hence, spherical cell shape would repress HIF-1α pathway, and consequently would reestablish normal metabolism in FLC-4 cells. Spherical cell shape also reduced solute carrier (SLC)7A5 (LAT1) gene expression, an amino acid transporter implicated in malignancy and cell growth (Fuchs and Bode, 2005) (Fig. 1F). Spherical cell shape decreased significantly the gene expression of gap junction (GJ)A1 (Cx43) and increased GJB1 (Cx32) gene expression, as a result of which the ratio of the gene expression of GJA1/GJB1, a marker of malignancy in hepatoma (Zhang et al., 2007), was downregulated (Fig. 1F). We supposed that spherical cell shape would downregulate malignant phenotype in FLC-4 cells.
Spherical cell shape-dependent induction of HNF-4α would elicit the enhancement of hepatocyte-specific gene expression and suppress malignancy-related gene expression in FLC-4 cells
We speculated that the cell-shape dependent induction of HNF-4α would control the gene expression involved in drug metabolism and lipid metabolism (Table 1; Fig. 1). We hypothesized that cell shape-dependent induction of HNF-4α would be involved in the suppression of HIF-1α, VEGFC, and c-Myc gene expression (Table 1; Fig. 1). Actually, network analysis of mRNA microarray by IPA revealed that HNF-4α played a central role in the induction and suppression of gene expression in spherical FLC-4 cells (Fig. 2). HNF-4α was shown to repress the gene expression of malignancy-related genes, such as c-Myc and HIF-1α (Wang et al., 2011; Yin et al., 2008). From these results, it was supposed that cell shape-dependent induction of HNF-4α would mediate the induction of liver-specific gene expression and repression of malignancy-related gene expression (Fig. 2).
Hypothetic regulation of gene expression by HNF-4α by spherical cell shape in FLC-4 cells.
Changes in microRNA expression would mediate cell shape-dependent induction of HNF-4α and the suppression of malignant phenotype in spherical FLC-4 cells
The purpose of this study was to get insight into the induction mechanism of HNF-4α gene expression by spherical cell shape in FLC-4 cells. To clarify this mechanism, we performed microRNA microarray analysis. The results indicated that 23 microRNA were upregulated, and 6 microRNA were downregulated significantly in spherical FLC-4 cells (Table 2). We then analyzed integrated data of mRNA and microRNA microarray by bioinformatics to get insight into mRNA and microRNA interactions. Although 29 microRNAs were changed by spherical FLC-4 cells on EHS-gel, many of them were not related to changes in mRNAs. And only four microRNAs were selected to be associated with changes in mRNAs (Table 3). Therefore, we focused on these four microRNAs. None of them were not related to HNF-4α.
We also found that miR-24 level was slightly but significantly downregulated (Table 2), which directly suppressed HNF-4α gene expression (Takagi et al., 2010). Other microRNA predicted to target HNF-4α in Targetscan database were not affected by spherical cell shape. Although the changes in miR-24 might partly explain the induction of HNF-4α by spherical cell shape, we supposed that the change in miR-24 was too small to explain the induction of HNF-4α.
Microarray of mRNA indicated that hepatocyte-specific gene expression, including drug metabolism and insulin signaling, was induced by spherical cell shape in FLC-4 cells (Table 1; supplementary material Table S1). In contrast, malignancy-related gene expression was repressed by spherical cell shape (Table 1; supplementary material Table S2). To understand how microRNA would regulate the gene expression changed by spherical cell shape, we integrated our microRNA and mRNA expression data using Targetscan database to predict mRNA–microRNA interactions. miR-29b was downregulated by spherical cell shape in FLC-4 cells, and was predicted to induce phosphatidylinositol 3-kinase regulatory subunit (PIK3R)1 and PIK3R3 gene expression (Table 3). Since these genes are involved in insulin signaling and malignancy (Anderson, 2010), spherical cell shape-dependent reduction of miR-29b level would regulate these functions in FLC-4 cells. In contrast, miR-194 was upregulated by spherical cell shape, and was predicted to suppress SLC7A5 gene expression, implicated in malignant phenotype (Table 3). These results suggested that spherical cell shape in FLC-4 cells would repress malignancy through miR-194, which was shown to suppress this gene expression in the liver (Meng et al., 2010). Three-dimensional cell shape induced miR-320c, which was predicted to inhibit E2F1 gene expression in FLC-4 cells (Table 3). miR-320 family and E2F1 were shown to inhibit and promote malignant progression respectively (Ladu et al., 2008; Wentz-Hunter and Potashkin, 2011). This indicated that miR-320c would also mediate the suppression of malignancy-related gene expression by spherical cell shape in FLC-4 cells. Based on these results, the enhancement of miR-194 and miR-320c level by spherical cell shape would suppress malignancy in FLC-4 cells.
In conclusion, the present study showed that spherical cell shape in FLC-4 cells greatly induced gene expression related to drug metabolism and lipid metabolism. The suppression of malignancy-related gene expression by spherical cell shape was put into evidence in FLC-4 cells. The induction of HNF-4α gene expression would have a central role in the enhancement of hepatocyte-specific gene expression and the suppression of the gene expression related to malignancy by spherical cell shape in FLC-4 cells. The integration of microRNA and mRNA microarray data indicated that spherical cell shape would elicit the repression of malignancy through enhancing miR-194 and miR-320c levels.
Materials and Methods
FLC-4 cells were maintained in ASF104 serum-free medium (Ajinomoto Co. Ltd., Tokyo, Japan). Cells were inoculated into 100-mm culture dishes (Corning Inc., Corning, NY, USA) at 40% confluency (number of cells: 2.7×104/cm2) and cultured with medium changes every 3 days. Cell cultures were carried out in 100-mm dishes (Cat. no. 1029; Falcon, Becton Dickinson, Lincoln Park, NJ, USA) under a constant temperature of 37°C with highly humidified 95% air and 5% CO2. Type I collagen (TIC)-coated dishes were prepared by adding 100 µg/mL TIC (Nitta Gelation, Japan) to plastic dishes (Cat. no. 1029; Falcon, Becton Dickinson, Lincoln Park, NJ, USA). EHS-gel-coated dishes were prepared as described previously (Oda et al., 2008).
Total RNA and small RNA isolation and analysis
Total RNA was extracted according to the method of Chomczynski and Sacchi (Chomczynski and Sacchi, 1987). Total RNA containing small RNAs was extracted from spread FLC-4 cells or spherical FLC-4 cells using the mirVanaTM microRNA isolation kit according to the manufacturer's protocol (Cat. #AM1560, Ambion, Austin, TX, USA). Quality of RNA was determined using Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA) and subsequently quantified using NanoDrop 1000 spectrophotometer (Nanodrop, Wilmington, DE, USA). cDNA was prepared using a High Capacity cDNA Reverse Transcription kit (Cat. #4368814, Applied Biosystems, Foster City, USA) following the manufacturer's instructions. Real-time RT-PCR amplifications were performed in ABI StepOne (Applied Biosystems, Foster City, USA) using 2× Power SYBR Master Mix (Cat. #4309155, Applied Biosystems) as described previously (Laurent et al., 2012). Primer sequences are indicated in supplementary material Table S3. Quantitative RT-PCR melting curve data were collected to check the specificity of PCR. Expression was calculated using the standard curve method. The expression of the chosen genes was normalized to that of 18S rRNA as a reference.
Nuclear protein extraction and electrophoretic mobility-shift assay (EMSA)
Nuclear proteins were extracted from FLC-4 cells and electrophoretic mobility-shift assay (EMSA) was performed as described previously (Laurent et al., 2012). EMSA was performed using the following double-stranded oligonucleotide: DR-4 (5′-agctTCAGGTCACTTCAGGTCAC-3′, 5′-tcgaGTGACCTGAAGTGACCTGA-3′).
Microarray of mRNA and data analysis
Total RNA from spread and spherical FLC-4 cells were used for mRNA microarray analysis. We utilized a DNA oligonucleotide microarray containing duplicate cDNA spots of 1262 well annotated genes of various functional classes, including cytokines/growth factors and their receptors, oncogenes, drug metabolizing enzyme, transcription factors and housekeeping genes (Hitachi Life Science, Saitama, Japan). Five micrograms of total RNA isolated was in vitro amplified, and the antisense RNA (aRNA) from spread FLC-4 cells was labeled with a fluorescent dye Cy5, while aRNA from spherical FLC-4 cells was labeled with Cy3. The arrays were hybridized at 62°C for 10 h in the hybridization buffer containing equal amounts of Cy3- or Cy5-labeled cDNA, and they were then scanned by the ScanArray 5000 scanner (GSI Lumonics, Boston, MA, USA). The data were analyzed by using the QuantArray software (GSI Lumonics, Boston, MA, USA). The average of fluorescence intensities of duplicate spots was obtained after global normalization between Cy3 and Cy5 signals. Genes were considered as differentially expressed when fold change was above 1.8 or below −1.8 and P < 0.05. List of differentially expressed genes was imported in IPA (Ingenuity Systems, Redwood City, CA, USA), and no filter criterion was used for this analysis. Network analysis and biofunction analysis were performed using the IPA software.
MicroRNA microarray and data analysis
200 ng of total RNA from FLC-4 cells cultured on EHS-gel or plastic dishes for 48 hours (n = 3) were labeled using the miRNA Complete Labeling and Hybridization Kit (Cat. #5190-0456, Agilent, Santa Clara, CA, USA). Labeled RNA was hybridized in Agilent Human microRNA Microarray V3 for 20 h at 20 rpm, 55°C (Cat. #G4470C, Agilent, Santa Clara, CA, USA). Slides were washed and scanned according to the manufacturer's instructions. Images were quantified using Feature Extraction (Agilent, Santa Clara, CA, USA). The miRNA array contained the complete content sourced from Sanger database 12.0, i.e. 851 probes for human and 88 probes for viral miRNA transcripts (Agilent, Santa Clara, CA, USA). The raw dataset was normalized and analyzed using the “AgiMicroRna” package of the Bioconductor (http://www.bioconductor.org) suite of software for the R statistical programming language (http://www.r-project.org). Quantile normalization was then used to standardize the data across arrays, and a linear model was fitted to each microRNA using “AgiMicroRna” package. The resultant P values were obtained using a moderated t-test statistics, adjusted for multiple testing by using the Benjamini–Hochberg correction of the false-discovery rate. MicroRNA were selected according to the following criteria: False Discovery Rate (FDR) < 0.05. The results are expressed in log 2 ratio (EHS-gel versus plastic). MicroRNA were selected according to the following criteria: log 2 ratio ≤ −1 or log 2 ratio ≥ 1. The miRNA target prediction was performed using TargetScan database (http://www.targetscan.org).
The significance of differences among values was analyzed by ANOVA and Student's t-test. When P value was below 0.05, differences were considered significant. Values in the text are expressed as means ± s.e.m.
We are grateful to Yasuko Matsuyama for technical assistance.
The authors have no competing interests to declare.