An HNF4α–microRNA-194/192 signaling axis maintains hepatic cell function

Hepatocyte nuclear factor 4α (HNF4α) controls the expression of liver-specific protein-coding genes. However, some microRNAs are also modulated by HNF4α, and it is not known whether they are direct targets of HNF4α and whether they influence hepatic function. In this study, we found that HNF4α regulates microRNAs, indicated by marked down-regulation of miR-194 and miR-192 (miR-194/192) in liver-specific Hnf4a-null (Hnf4aΔH) mice. Transactivation of the shared miR-194/192 promoter was dependent on HNF4α expression, indicating that miR-194/192 is a target gene of HNF4α. Screening of potential mRNAs targeted by miR-194/192 revealed that expression of genes involved in glucose metabolism (glycogenin 1 (Gyg1)), cell adhesion and migration (activated leukocyte cell adhesion molecule (Alcam)), tumorigenesis and tumor progression (Rap2b and epiregulin (Ereg)), protein SUMOylation (Sumo2), epigenetic regulation (Setd5 and Cullin 4B (Cln4b)), and the epithelial-mesenchymal transition (moesin (Msn)) was up-regulated in Hnf4aΔH mice. Moreover, we also found that miR-194/192 binds the 3′-UTR of these mRNAs. siRNA knockdown of HNF4α suppressed miR-194/192 expression in human hepatocellular carcinoma (HCC) cells and resulted in up-regulation of their mRNA targets. Inhibition and overexpression experiments with miR-194/192 revealed that Gyg1, Setd5, Sumo2, Cln4b, and Rap2b are miR-194 targets, whereas Ereg, Alcam, and Msn are miR-192 targets. These findings reveal a novel HNF4α network controlled by miR-194/192 that may play a critical role in maintaining the hepatocyte-differentiated state by inhibiting expression of genes involved in dedifferentiation and tumorigenesis. These insights may contribute to the development of diagnostic markers for early HCC detection, and targeting of the miR-194/192 pathway could be useful for managing HCC.

Liver has diverse functions such as metabolism of carbohydrates, lipids, and proteins; detoxification of ammonia and xenobiotics; and synthesis of bile acids and blood coagulation factors. Various liver-enriched transcription factors such as hepatocyte nuclear factor 1 (HNF1), 2 HNF3, HNF4, and HNF6 are needed to maintain differentiated liver-specific functions (1). Of these, HNF4␣ is a member of the nuclear receptor superfamily and a key regulator involved in the transcription regulatory network in the liver (2,3), and re-expression of HNF4␣ in dedifferentiated hepatoma cells has the potential to revert cells back to the differentiated state (4). As Hnf4a-null mice are embryonic lethal during gastrulation (5), two types of liverspecific Hnf4a-null mice (Hnf4a ⌬H ) were established (6,7). Disruption of HNF4␣ in the fetal livers showed that hepatic HNF4␣ is essential for formation of a hepatic epithelium, liver morphogenesis, and liver homeostasis by positive regulation of a large number of genes during development into the mature liver (6). Disruption of HNF4␣ in the Hnf4a ⌬H mice led to diverse phenotypes such as fatty livers and elevation of serum bile acids and ammonia (7)(8)(9). Hnf4a ⌬H mice also exhibit increased spontaneous hepatocyte proliferation and increased susceptibility to diethylnitrosamine-induced liver cancer, indicating that HNF4␣ has tumor suppressor activity, although the mechanism has not been established (10,11). Because HNF4␣ regulates many liver-enriched genes via HNF4␣-binding sites in the promoter regions of the target genes and HNF4␣ is bound to about 12% of the gene promoters expressed in human hepatocytes (12), the HNF4␣-derived transcriptional network in the liver is expected to be considerably complicated.
The gene expression pattern in the livers of Hnf4a ⌬H mice is markedly changed largely due to loss of expression of direct HNF4␣ target genes that are down-regulated, giving rise to the altered phenotypes of these mice (13). For instance, reduced expression of ornithine transcarbamylase (Otc), encoded by a HNF4␣ target gene, caused elevated serum ammonia due to the inability to degrade ammonia (8). In contrast, many mRNAs are up-regulated in Hnf4a ⌬H mice. Although direct transcriptional repression by HNF4␣ itself could be considered as a mechanism by which these mRNAs are reduced in Hnf4a ⌬H mice, another possibility is the involvement of microRNAs (miRNAs) that could be under direct transcriptional control of HNF4␣. miRNAs are small non-coding RNAs involved in various physiological functions such as development, differentiation, inflammation, and tumorigenesis through their ability to down-regulate protein-coding gene expression by binding to target mRNAs and either promoting mRNA degradation or inhibiting mRNA translation (14,15). miRNAs have important roles in liver diseases such as viral hepatitis, steatohepatitis, liver fibrosis, polycystic liver disease, and hepatocellular carcinoma (HCC) (16,17).
Among the miRNAs, miR-122 is the most abundant in adult liver (18). Expression of miR-122 is reduced in HCC, and overexpression of miR-122 suppresses cell proliferation (19,20). Thus, miR-122 could function as a tumor suppressor gene against HCC. Expression of miR-122 positively correlates with HNF4␣ expression in HCC cell lines, and miR-122 is directly regulated by HNF4␣, indicating that the HNF4␣-directed miR-122 pathway could be essential for inhibition of hepatocyte proliferation and HCC development (21,22). Likewise, HNF4␣ directly regulates the expression of miR-134 and miR-134-suppressed oncogenic KRAS expression, revealing that miR-134 has a suppressive effect on HCC progression (23). Moreover, the following HNF4␣-miRNA inflammatory feedback circuits in human HCC have been proposed. HNF4␣ positively regulates miR-7 and miR-124 followed by suppression of IL6R, STAT3, and RelA-mediated NF-B pathways, eventually resulting in the up-regulation of HNF4␣ through down-regulation of miR-24, miR-629, or miR-21 (24,25). These results also indicate that lack of HNF4␣-controlled miRNAs could result in HCC development by activation of inflammatory signaling. These miRNAs were identified from studies using immortalized hepatocytes or HCC cell lines and thus could be biomarkers for HCC diagnosis and candidates for new target genes for HCC therapy. However, miRNAs regulated by HNF4␣ in normal hepatocytes but not in immortalized or transformed hepatocytes remain to be identified.

Down-regulation of hepatic miR-194 and miR-192 in Hnf4a ⌬H mice
HNF4␣ positively regulates the expression of many genes through HNF4␣-binding sites in promoter regions (1,3). A marked change of the mRNA expression pattern was observed in the livers of Hnf4a ⌬H mice (13), including many down-regulated mRNAs identified as encoded by direct HNF4␣ target genes (8,9,26). In addition, several up-regulated genes were also detected in Hnf4a ⌬H mice, but their mechanism of increase is not known. Although transcriptional repression of genes such as Snail, Slug, and HMGA2 by HNF4␣ was demonstrated (27), the mechanism of negative regulation of other genes by HNF4␣ has not been established. Another possibility for negatively regulated mRNAs is indirectly through miRNA. Indeed, HNF4␣ is known to regulate the expression of miR-7, miR-122, miR-124, and miR-134 (22)(23)(24)(25). Thus, expression profiling of hepatic miRNAs in Hnf4a-floxed (Hnf4a f/f ) and Hnf4a ⌬H mice was investigated by microarray analysis (supplemental Table  S1). Of 610 miRNAs, expression of 24 and 19 miRNAs was down-regulated or up-regulated more than 2-fold in Hnf4a ⌬H mice, respectively (Table 1). In general, because HNF4␣ positively regulates the target genes, down-regulated miRNAs in Hnf4a ⌬H mice could be direct target genes for HNF4␣. Although other miRNAs, including miR-34a and miR-301a, were increased in Hnf4a ⌬H mice, they were not investigated in the current study. They could be either directly repressed by HNF4␣ or the result of downstream events mediated by loss of expression of other direct HNF4␣ target genes.
The expression of miR-194 was decreased by 90% in Hnf4a ⌬H mice compared with Hnf4a f/f mice. miR-194 was also expressed at considerably higher levels in Hnf4a f/f mice when compared with other miRNAs. Similarly, the expression of miR-192 was also higher in Hnf4a f/f mice and markedly lower in Hnf4a ⌬H mice. Indeed, qRT-PCR quantification showed that hepatic expression of miR-194 and miR-192 was suppressed by about one-tenth in Hnf4a ⌬H mice (Fig. 1A). Expression of other liverenriched miRNAs such as miR-21, -101b, -130a, -193, -193b, -455, and -805 was also decreased in Hnf4a ⌬H mice (Fig. 1A). Others have reported that the expression of miR-194 and miR-192 is regulated under the same promoter, and tissue distribution of miR-194 and miR-192 is almost equivalent to that of HNF4␣ (28,29), suggesting that HNF4␣ could positively regulate the expression of miR-194 and miR-192. Therefore, further studies were performed on miR-194 and miR-192. No significant difference in hepatic expression of miR-7, miR-122, miR-124, and miR-134, which were reported as the target miRNAs for HNF4␣ (22)(23)(24)(25), was observed in Hnf4a ⌬H mice (supplemental Table S1).
In mice, mature miR-194 is located on two chromosomes with miR-194-1 and miR-194-2 clustered with miR-215 and miR-192, respectively. Expression of miR-215 was unchanged in Hnf4a ⌬H mice, and the expression level was very low as compared with miR-194 and miR-192 (Table 1). Thus, miR-194-2, but not miR-194-1, could be highly expressed in normal liver and preferentially suppressed in Hnf4a ⌬H mice. As expected, hepatic expression of pri-miR-194-1 was unchanged, whereas expression of pri-miR-194-2 was markedly suppressed in Hnf4a ⌬H mice (Fig. 1B). These data indicate that HNF4␣ could directly activate the miR-194/192 gene via a single promoter.

Regulation of hepatic miR-194/192 by HNF4␣ and HNF1␣
HNF1␣, another liver-enriched transcription factor, directly regulates the miR-194/192 gene through a binding site in the miR-194/192 promoter region (28). Accordingly, qRT-PCR analysis using the livers of Hnf4a ⌬H mice and Hnf1a-null mice was performed to determine which factor, HNF1␣ or HNF4␣, predominates in controlling expression of the miR-194/192 gene. Hepatic expression of both miRNAs in Hnf1a-null mice was reduced to 50%, and the remnant levels were much higher when compared with Hnf4a ⌬H mice (Fig. 1C). Because HNF4␣ positively regulates Hnf1a in the liver (2), decreased expression of miR-194/192 in Hnf4a ⌬H mice could be due to decreased expression of both Hnf1a and Hnf1a. Therefore, the expression of Hnf4a and Hnf1a mRNAs in Hnf1a-null mice was also investigated. The expression of Hnf4a was reduced to 3%, and the expression of Hnf1a mRNA was reduced to 45% in Hnf4a ⌬H mice (Fig. 1D). Conversely, expression of Hnf1a mRNA was not detected, and expression of Hnf4a mRNA was reduced to 42% in Hnf1a-null mice (Fig. 1E), revealing that the remaining expression of Hnf1a in Hnf4a ⌬H mice and Hnf4a in Hnf1a-null mice is similar. Hence, HNF4␣ could be a more essential factor for controlling expression of miR-194/192 as compared with HNF1␣ in the liver.
p53 also has the potential to transcriptionally control the miR-194/192 gene (29,30). However, p53 was up-regulated in Hnf4a ⌬H mice, whereas it was unchanged in Hnf1a-null mice (Fig. 2, A and C). Similarly, the expression of p21 (Cdkn1a) mRNA, encoded by a direct target gene for p53, was also upregulated in Hnf4a ⌬H mice and unchanged in Hnf1a-null mice (Fig. 2, B and D), indicating that up-regulation of p53 in Hnf4a ⌬H mice could not overcome repression of miR-194/192 expression caused by loss of HNF4␣.

The miR-194/192 gene promoter is transactivated by HNF4␣
To investigate the mechanism by which HNF4␣ regulates the miR-194/192 gene, the promoter region of the miR-194/192 gene was analyzed. As described above, HNF1␣ was already found to transactivate the miR-194/192 gene through an HNF1␣-binding site in colorectal adenocarcinoma-derived Caco-2 cells, and two regions located at Ϫ130 to Ϫ118 and Ϫ110 to Ϫ98 upstream of the HNF1␣-binding site were highly conserved among species (28), but the binding proteins were not identified. Because HNF4␣ binding to both regions was predicted by JASPAR, an open access database (Fig. 3A), promoter analysis was performed to determine whether HNF4␣ has the potential to transactivate the miR-194/192 promoter. In HCC-derived and endogenous HNF4␣-expressing HepG2 cells, the Ϫ70/ϩ21 fragment containing an HNF1␣-binding site was transactivated by 50-fold as compared with the promoterless vector pGL4.11 (Fig. 3B). The promoter activity of the Ϫ115/ϩ21 fragment containing the HNF1␣-binding site and a proximal HNF4␣-binding site was much higher, and that of the Ϫ162/ϩ21 fragment was highest in the presence of both HNF4␣-binding sites, indicating that the minimum promoter of the miR-194/192 gene is located between Ϫ162 and ϩ21. Introduction of mutations into the proximal (M1) or distal (M2) HNF4␣-binding site resulted in significant reduction of the activities, and mutations in both binding sites (M1 ϩ M2) did not transactivate the Ϫ162 and Ϫ2032/ϩ21 regions (Fig.  3C). By co-transfection with the HNF4␣ expression vector in HEK293T cells, the promoter activity was slightly induced in the Ϫ115/ϩ21 fragment containing the distal HNF4␣-binding site and strongly activated in the Ϫ162/ϩ21 fragment in the presence of both HNF4␣-binding sites (Fig. 3D). Furthermore, the induced activities in the Ϫ162 and Ϫ2032/ϩ21 promoters were markedly suppressed by introduction of mutations into one or both HNF4␣-binding sites (Fig. 3E). These results indi-cate that both HNF4␣-binding sites cooperatively transactivate the miR-194/192 gene.
To determine whether HNF4␣ can directly bind to both HNF4␣-binding sites in the miR-194/192 promoter, gel mobility shift analysis was performed (Fig. 4A). Nuclear extracts from HepG2 cells bound to the distal and proximal HNF4␣-binding sites (Fig. 4A, lane 2, lower arrows). This complex was diminished by the addition of excess amounts of the unlabeled competitor and OTC competitor that contains an HNF4␣-binding site (Fig. 4A, lanes 3 and 5) but not the competitor that contains mutations in the HNF4␣-binding site (Fig. 4A, lane 4). Moreover, the complex was supershifted by anti-HNF4␣ antibody (Fig. 4A, lane 6, upper arrow) but not the anti-C/EBP␣ antibody (lane 7). These results indicate that HNF4␣ binds to both HNF4␣-binding sites of the miR-194/192 promoter. Next, chromatin immunoprecipitation (ChIP) was used to determine whether HNF4␣ directly binds to the miR-194/192 promoter in vivo. We found that HNF4␣ was bound to the HNF4␣-binding sites in HepG2 cells (Fig. 4B). In addition, ChIP analysis using the livers of Hnf4a f/f and Hnf4a ⌬H mice indicated that hepatic HNF4␣ in Hnf4a f/f mice bound to the promoter region, but the binding was decreased in Hnf4a ⌬H mice (Fig. 4C). Similar results were obtained in the mouse Otc promoter that is dependent on an HNF4␣-binding site (8), suggesting that HNF4␣  physiologically binds to the promoter region of the miR-194/ 192 gene in humans and mice.

Identification of miR-194 and miR-192 target mRNAs
It was reported that miR-194 is involved in epithelial-mesenchymal transition (EMT) and HCC metastasis (31). In addition, miR-194 regulates mRNAs encoded by many genes including N-cadherin (Cdh2), Rac1, heparin-binding epidermal growth factor (Hbegf), tyrosine protein phosphatase non-receptor type 2 (Ptpn2), integrin ␣9 (Itga9), suppressor of cytokine signaling 2 (Socs2), DNA methyltransferase 3A (Dnmt3a), Fzd6, and talin 2 (Tln2) that play important roles in tumorigenesis, tumor progression, cell adhesion, cell migration, and EMT (31)(32)(33). miR-192 is also known to negatively regulate mRNAs encoding IGF-1, IGF-1R, and ZEB2 that are also involved in EMT and tumor progression and invasion (30,34). Therefore, qRT-PCR was performed to determine whether hepatic expression of these mRNAs was increased in Hnf4a ⌬H mice. Expression of Fzd6, Hbegf, Ptpn2, Dnma3a, Itga9, and Rac1 mRNAs was significantly increased, whereas expression of Socs2, Cdh2, Tln2, and Zeb2 mRNAs was unchanged in Hnf4a ⌬H mice (Fig. 5, A and C). Next, novel target genes for miR-194 and miR-192 were investigated using TargetScan, DIANA-microT, and miRDB databases, and the resulting high-scoring hits that overlapped such as Hnf1b, Gyg1, hook microtubule-tethering protein 3  6 and 7). The complex between HNF4␣ and the probe and the supershifted complex are indicated by the lower and upper arrows, respectively. B, chromatin immunoprecipitation using HepG2 cells was performed with anti-HNF4␣ antibody (lane 2) and normal goat IgG (lane 3) (left). Input DNA was used as a positive control (lane 1). The regions between Ϫ201 and Ϫ41 containing both HNF4␣-binding sites (ϩHNF4␣) and between ϩ16667 and ϩ16874 without HNF4␣-binding site (ϪHNF4␣) in the human miR-194/192 gene were amplified. The data from qPCR were normalized relative to the input and expressed as -fold enrichment (right). C, chromatin immunoprecipitation using the livers of Hnf4a ⌬H and Hnf4a f/f mice with anti-HNF4␣ antibody and normal goat IgG. The regions between Ϫ164 and Ϫ70 containing both HNF4␣-binding sites in the mouse miR-194/192 promoter (left), between Ϫ264 and Ϫ165 containing two HNF4␣-binding sites in the mouse Otc promoter (right), and between ϩ45820 and ϩ45893 without an HNF4␣-binding site in the mouse Hmgcs2 gene were amplified, respectively. The data from qPCR was normalized relative to the input and expressed as -fold enrichment over data from IgG control. Error bars represent S.D. Data are mean Ϯ S.D. of three independent experiments. *, p Ͻ 0.05. (Hook3), Setd5, selenophosphate synthetase 1 (Sephs1), Sumo2, Cln4b, and Rap2b were selected as the target candidates for miR-194 (supplemental Fig. S1, B-F). Regarding the candidates of miR-192 target gene mRNAs, Ereg, ADP-ribosylation factor guanine nucleotide-exchange factor 1 (Arfgef1), radical S-adenosyl methionine domain-containing 2 (Rsad2), Alcam, and Msn were selected (supplemental Fig. S2, G-I). As expected, the expression of all candidate mRNAs was significantly increased in Hnf4a ⌬H mice (Fig. 5, B and C). The expression of CLN4B protein, a candidate for miR-194 targeting, and ALCAM protein, a candidate for miR-192 targeting, was also increased more than 2-fold in Hnf4a ⌬H mice compared with Hnf4a f/f mice (Fig. 5D). Therefore, studies were done to determine whether miR-194 and miR-192 could directly interact with the 3Ј-UTR of these candidate mRNAs and inhibit the protein expression (Fig. 6). miR-194 and miR-192 markedly repressed the 3Ј-UTR activities in the presence of the complementary sequences of miR-194-and miR-192-binding sites (c-miR-194 and c-miR-192) (supplemental Fig. S2). The 3Ј-UTR activities of Fzd6, Gyg1, Setd5, Sumo2, Cln4B, and Rap2b using a luciferase reporter system were significantly inhibited by miR-194 (Fig. 6A, wild type (WT)). Because two miR-194-binding sites were predicted within the 3Ј-UTR of Fzd6 and Rap2b, mutations were introduced to disrupt the binding of miR-194 in both binding sites. Each mutant (M1 and M2) partially rescued the 3Ј-UTR activity, and both mutants (M1 ϩ M2) completely recovered the activities (Fig. 6A). The same results were also obtained in the mutated 3Ј-UTR regions of the Gyg1, Setd5, Sumo2, and Cln4B (Fig. 6B). Similarly, 3Ј-UTR activities of Ereg, Alcam, and Msn were suppressed in a miR-192-dependent manner, and the activities were recovered by introduction of mutations into the miR-192-binding sites (Fig. 6B). These results indicate that Fzd6, Gyg1, Setd5, Sumo2, Cln4B, and Rap2b are the target mRNAs for miR-194 and that Ereg, Alcam, and Msn mRNAs are targets for miR-192.

Discussion
In this study, HNF4␣ was found to directly activate the miR-194/192 gene promoter. Down-regulation of HNF4␣ target genes partially could explain the onset of phenotypes in Hnf4a ⌬H mice (7,8,26). From microarray analysis using livers of Hnf4a ⌬H mice (13), most mRNAs were markedly down-regulated, probably due to the loss of direct positive regulation of the respective genes by HNF4␣. However, up-regulation of many mRNAs was also observed in Hnf4a ⌬H mice. Thus, HNF4␣ might negatively regulate these genes via direct positive regulation of miRNAs that in turn could bind to destabilize mRNAs. HNF4␣ was shown to bind to the promoter region of miR-122, the most abundant miRNA in the liver, and transactivate the miR-122 in an HNF4␣-dependent manner (22). However, hepatic expression of miR-122 was unchanged in Hnf4a ⌬H mice. HNF4␣ knockdown also inhibited miR-122 expression in mouse liver (22), suggesting that transactivation of miR-122 by HNF4␣ would be transient in Hnf4a ⌬H mice as observed, probably due to secondary effects. Likewise, HNF4␣ positively regulates miR-7, miR-124, and miR-134 in human HCC and tumor-derived cell lines (23)(24)(25). However, unexpectedly, the hepatic expression of these miRNAs was also unchanged in Hnf4a ⌬H mice. Because the expression levels of miR-7, miR-124, and miR-134 were very low or undetectable in normal livers of Hnf4a f/f mice, HNF4␣ may not directly regulate these miRNAs as found in normal liver. Thus, HNF4␣ might regulate these miRNAs only during the later stages of HCC progression but not the early stages of HCC development because Hnf4a ⌬H mice do not exhibit early spontaneous HCC development.
Unlike miR-7, miR-122, and miR-134, miR-194 and miR-192 were not reported to be down-regulated in human HCC specimens used in these studies, indicating that down-regulation of miR-194 and miR-192 in the liver could be an early event in the onset of liver inflammation and pretumorigenesis. miR-194 is involved in the differentiation of intestinal epithelial cells, and HNF1␣ was also reported to transactivate miR-194/192 via an HNF1␣-binding site (28). HNF1␣ is a direct target for HNF4␣ (2), and hepatic expression of HNF1␣ was actually down-regulated in Hnf4a ⌬H mice as expected. Thus, both HNF4␣ and HNF1␣ might cooperatively control the miR-194/192 gene promoter. However, down-regulation of miR-194 and miR-192 in Hnf1a-null mice was not as pronounced when compared with Hnf4a ⌬H mice because sufficient HNF4␣ still remains in Hnf1a-null mice, indicating that HNF4␣ might be a predominant factor in regulation of the miR-194/192 gene as compared with HNF1␣.
p53 was also reported to control expression of the miR-194/ 192 gene via a p53-binding site (29,30). The p53-binding site is located upstream of the distal HNF4␣-binding site, and the promoter of Ϫ162/ϩ21 fragment including the p53-and HNF4␣-binding sites was activated in an HNF4␣-dependent manner, suggesting that p53 and HNF4␣ might synergistically transactivate the miR-194/192 gene. Expression of miR-194 and miR-192 was markedly down-regulated in a p53-dependent manner in B cell neoplasm multiple myeloma, which rarely expresses endogenous HNF4␣ (30). Conversely, the expression of miR-194 and miR-192 was slightly down-regulated in human HCC including mutated or deleted p53 without statistical significance as compared with wild-type p53 (34), indicating that HNF4␣ and HNF1␣ mainly transactivate the miR-194/192 gene despite p53 status.
EMT is an important process for cells to acquire tumor metastatic potential. Expression of miR-194 was also down-regulated in liver mesenchymal-like cancer cell lines as compared with liver epithelial cancer cell lines, and overexpression of miR-194 suppressed invasion and migration of hepatoma cells (31). Thus, miR-194 and miR-192 expression in liver might retain the differentiated status of hepatocytes and suppress proliferation, migration, and malignant transformation. Re-expression of HNF4␣ in dedifferentiated hepatoma cells resulted in reversion to differentiated phenotypes (4). Moreover, reduced expression of HNF4␣ correlated with human HCC progression (25), and forced expression of HNF4␣ reduced the proliferation and induced epithelial formation (25,35,36). These results strongly suggest that HNF4␣ is one of the most important factors to induce hepatocyte differentiation by direct transactivation of liver-specific genes and to inhibit dedifferentiation of hepatocytes by indirect negative regulation of oncogenes and EMT-related genes via direct up-regulation of miR-194 and miR-192. However, hepatic expression of EMTactivating genes including Cdh2 as an miR-194 target and Zeb2 as an miR-192 target mRNA was not increased in mice, showing that down-regulation of miR-194 and miR-192 in inflammatory and non-/pretransformed hepatocytes in Hnf4a ⌬H mice could not yet activate EMT-related genes because Hnf4a ⌬H mice die at about 8 weeks of age before showing tumorigenesis (7).
Others have reported that HNF1␣-regulated miR-194 repressed the 3Ј-UTR activity of FZD6 (32). In this study, two functional miR-194-binding sites were identified in the 3Ј-UTR of Fzd6 mRNA. FZD6 is a membrane receptor for Wnt ligand signaling and controls cell differentiation (37). Because overexpression of FZD6 was confirmed in human HCC (38) and hepatic expression of FZD6 was not significantly increased in HNF1␣-null mice (32), repression of FDZ6 expression through negative regulation by miR-194 via HNF4␣ could be important for suppression of tumorigenesis.
There are no reports that GYG1 in involved in HCC and metastasis. Glycogen synthesis starts from self-glucosylation of GYG1 followed by elongation of glucose chains by glycogen synthase (39). Hepatic expression of glycogen synthase 2, a limited enzyme for glycogen synthesis in the liver and a direct target for HNF4␣, was markedly decreased in Hnf4a ⌬H mice (40). Eventually, expression of GYG1 might be increased to compensate for reduced production of de novo glycogen synthesis in Hnf4a ⌬H mice.
CUL4B is the main component of Cullin4B-Ring ubiquitin ligase complex and promotes tumorigenesis (41) and SETD5 is a histone methyltransferase on the histone H4 tail at lysines 5, 8, and 12 in yeast (42). Loss of function mutations in the SETD5 gene were found in patients with intellectual disability (43); however, the physiological significance of H4 methylation by SETD5 remains unsolved. Histone H4 lysines 5, 8, and 12 are also acetylated by histone acetyltransferases (44), and methylation of these sites by SETD5 might change chromatin structure and influence transcription. HNF4␣-dependent suppression of Cul4B and Setd5 could partially control the epigenetic program and be important to retain the differentiated status in normal hepatocytes.
SUMO2 has an important role to regulate a variety of biological functions such as transcription, DNA repair, cell cycle, and subcellular transport by SUMOylation to the target proteins (45). Similar to an ES cell-based model of hepatocyte differentiation in which SUMO2-modified proteins were decreased during hepatocyte differentiation (46), increased activity of SUMOylation by SUMO2 could induce hepatocyte dedifferentiation in Hnf4a ⌬H mice.
RAP2B functions as a p53-mediated prosurvival factor. RAP2B was also found to be up-regulated in many types of tumors, but RAP2B had very weak transformation activity, indicating that RAP2B itself is not an oncogene but maintains the tumorigenic status (47). Because hepatic expression of p53 and Cdkn1a was increased in Hnf4a ⌬H mice, induced p53 might be responsible for additional induction of Rap2b mRNA.
In addition, Ereg, Alcam, and Msn mRNAs are targets for miR-192 in the liver. EREG is one of the ligands for epidermal growth factor receptor and is involved in inflammation and tumorigenesis (48). Knockdown of EREG suppressed cell growth in human hepatoma cell lines (49), and Ereg-null mice exhibited a significant reduction of tumor number and size in HCC model experiments (50), indicating that EREG may contribute to HCC progression. Unlike other target genes for miR-194 and miR-192 examined in this study, EREG is more highly expressed in hepatic stellate cells than in hepatocytes (50), suggesting that miR-192 synthesized in hepatocytes might be secreted into hepatic stellate cells followed by repression of EREG in normal liver. ALCAM, a cell adhesion molecule, is a marker of colorectal cancer stem cells (51). Although ALCAM expression was inhibited by miR-192 in a gastric cell line, the 3Ј-UTR of Alcam mRNA failed to respond to miR-192 (52). By contrast, the 3Ј-UTR activity of Alcam mRNA was dependent on a miR-192-binding site. ALCAM was overexpressed in human hepatoma cell lines, and silencing of ALCAM suppressed cell growth and promoted apoptosis (53), suggesting that ALCAM could accelerate tumor cell survival by an antiapoptotic role in HCC.
MSN functions as a cross-linker between the plasma membrane and actin filament. High expression of moesin was observed in breast cancers and implicated in EMT (54,55). However, expression of Msn mRNA was induced only 1.6-fold in Hnf4a ⌬H mice. Because other EMT-related mRNAs such as Cdh2 and Zeb2 were not induced in Hnf4a ⌬H mice, these EMT-related mRNAs might not be induced sufficiently in inflammatory and non-transformed hepatocytes of Hnf4a ⌬H mice.
In conclusion, HNF4␣ was found to repress the expression of a variety of mRNAs through positive regulation of miR-194 and miR-192 in non-transformed liver. Most of these mRNAs encode proteins involved in tumorigenesis, prosurvival, epigenetic change, and EMT. These observations revealed an HNF4␣ cascade that inhibits hepatocyte dedifferentiation. Further studies may contribute to the development of novel diagnostic markers for the early detection of HCC using the HNF4␣-miR-194/192 axis, and inhibition of the miR-194/192 targets by induced expression and/or activation of miR-194/ 192 could be useful for HCC therapy.

Animals
Hnf4a ⌬H mice carrying a mixed SvJ129, FVB, and C57BL/6 background were generated as described previously (7). All experiments were performed with 45-day-old male Hnf4a f/f and Hnf4a ⌬H mice. Hnf1a-null mice were also used (56). The mice were housed in a pathogen-free animal facility under a standard 12-h light/12-h dark cycle with ad libitum water and chow. All experiments with mice were carried out under Association for Assessment and Accreditation of Laboratory Animal Care guidelines with approval of the National Cancer Institute Animal Care and Use Committee and Gunma University Animal Care and Experimentation Committee.

miRNA microarrays
Total RNA from the livers of Hnf4a f/f and Hnf4a ⌬H mice was purified using the miRNeasy mini kit (Qiagen, Tokyo, Japan). Extracted RNA was labeled with Cy5 using the Label IT miRNA labeling kit (Takara Bio, Otsu, Japan). Labeled RNAs were hybridized onto 3D-Gene Mouse miRNA Oligo chips (v.11.1; Toray Industries, Tokyo, Japan). The annotation and oligonucleotide sequences of the probes conformed to the miRBase miRNA database (http://www.mirbase.org/). 3 After stringent washes, fluorescence signals were scanned with the ScanArray Express Scanner (PerkinElmer Life Sciences) and analyzed using GenePix Pro version 5.0 (Molecular Devices, Sunnyvale, CA). The raw data of each spot were normalized by subtraction of the mean intensity of the background signal determined by all blank spots' signal intensities of 95% confidence intervals. Measurements of both duplicate spots with signal intensities greater than two standard deviations (S.D.) of the background signal intensity were considered to be valid. A relative expression level of a given miRNA was calculated by comparing the signal intensities of the averaged valid spots with their mean value throughout the microarray experiments after normalization by their median values adjusted equivalently. miRNA microarray data are described in supplemental Table S1. The data are also available at the Gene Expression Omnibus under accession number GSE70516.

Reverse transcription and real-time PCR
Total RNA was extracted using TriPure Isolation Reagent (Roche Applied Science). Reverse transcription for mature miRNA was performed using a Mir-X miRNA First-Strand Synthesis kit (Clontech). Real-time PCR was performed on a LightCycler 480 system II (Roche Applied Science). cDNAs encoding mature miRNA and U6 RNA were amplified using a Mir-X miRNA qRT-PCR SYBR kit (Clontech). The levels of mature miRNA were expressed relative to U6 RNA as an internal control using the ⌬⌬Ct method. For detection of two transcripts for pri-miR-194 and mRNAs, cDNA from the livers of Hnf4a f/f and Hnf4a ⌬H mice was transcribed using ReverTraAce qPCR RT Master with gDNA Remover (Toyobo, Osaka, Japan), and real-time PCR was performed using FastStart SYBR Green Master (Roche Applied Science). The levels of expression were normalized relative to Gapdh mRNA as an internal control using the ⌬⌬Ct method. Sequences for the primers are shown in supplemental Table S2.

Cloning of the human miR-194/192 gene promoter region
The Ϫ2032, Ϫ1003, Ϫ162, Ϫ115, Ϫ70, and Ϫ51/ϩ21 fragments of the human miR-194/192 promoter containing KpnI and XhoI sites were amplified with genomic DNA from HepG2 cells by PCR and cloned into the luciferase reporter vector pGL4.11 (Promega, Madison, WI). Mutations were introduced into the HNF4␣-binding sites in the miR-194/192 reporter vector by PCR-based site-directed mutagenesis. Sequences for the primers are shown in supplemental Table S3.

Transient transfection and luciferase assay
HepG2 and HEK293T cells were cultured at 37°C in Dulbecco's modified Eagle's medium (Wako, Osaka, Japan) containing 10% fetal bovine serum (HyClone, Logan, UT) and 1,000 units of penicillin and 0.1 mg of streptomycin/ml (Sigma-Aldrich). For suspension transfection, miR-194 promoters cloned into pGL4.11 and pGL4.74 encoding Renilla luciferase regulated under herpes simplex virus thymidine kinase promoter (Promega) were transfected into HepG2 cells with polyethylenimine Max (Polyscience, Warrington, PA) as a transfection reagent. For co-transfection using HEK293T cells, HNF4␣ expression plasmid was used. After 48 h of transfection, the cells were washed with phosphate-buffered saline (PBS), and promoter activities were measured using the Dual-Luciferase Reporter Assay System (Promega).

Western blotting
Liver samples from Hnf4a f/f and Hnf4a ⌬H mice and HepG2 cells treated with siRNA against Hnf4a mRNA were homogenized in lysis buffer (7 M urea, 2 M thiourea, and 1% Triton X-100) and allowed to sit on ice for 30 min. The homogenate was centrifuged at 12,000 ϫ g for 30 min at 4°C, and the supernatants were used as whole-cell lysates. Total protein (40 g) with Laemmli sample buffer was incubated at 65°C for 15 min and fractionated by 10% SDS-polyacrylamide gel electrophoresis. The gels were stained with Coomassie Brilliant Blue R-250 or blotted onto a PVDF membrane (GE Healthcare). The membrane was incubated for 1 h with mouse monoclonal antibodies against human HNF4␣ (Perceus Proteomics, Tokyo, Japan) and rabbit polyclonal antibodies against human CLN4B (GeneTex, Irvine, CA) and human ALCAM (Sigma-Aldrich). After washing, the membrane was incubated for 1 h with horseradish peroxidase-conjugated anti-mouse IgG (Cell Signaling Technology, Tokyo, Japan), and the reaction product was visualized using SuperSignal West Pico Chemiluminescent Substrate (Pierce) on an ImageQuant LAS4000 (GE Healthcare). Expression of CLN4B and ALCAM proteins was quantified by densitometric analysis using ImageJ software, and the expression normalized by total protein in Hnf4a ⌬H was presented as expression differences relative to the Hnf4a f/f mice.

Gel mobility shift analysis
Nuclear extracts from HepG2 cells were prepared using NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Fisher Scientific), and gel shift analysis carried out using the LightShift Chemiluminescent EMSA kit (Thermo Fisher Scientific). The following double-stranded probes were used: wildtype and mutants of distal and proximal HNF4␣-binding sites for the human miR-194/192 promoter and the HNF4␣-binding site for the mouse Otc promoter as a positive control. Nuclear extracts and the 5Ј-biotin-labeled probe of the HNF4␣-binding sites for the miR-194/192 promoter (wild type) were added, and the reaction mixture was incubated on ice for 10 min. For competition experiments, a 50-fold excess of unlabeled probe was added to the reaction mixture, and the mixture was incubated on ice for 10 min prior to the addition of the 5Ј-biotin-labeled probe. For supershift analysis, anti-HNF4␣ or anti-C/EBP␣ antibodies (Santa Cruz Biotechnology, Santa Cruz, CA) were added to the mixture, and the mixture was incubated on ice for 10 min after the addition of the 5Ј-biotin-labeled probe. DNA-protein complexes were fractionated by 7% PAGE and blotted onto a Biodyne B nylon membrane (Pall, Tokyo, Japan). After washing, complexes were visualized using the detection module in the kit on an ImageQuant LAS4000. Sequences for the primers are shown in supplemental Table S4.

Chromatin immunoprecipitation
HepG2 cells were fixed in 0.5% formaldehyde and quenched by 125 mM glycine at room temperature. After washing, cells were resuspended in 3 ml of lysis buffer 1 (50 mM HEPES-KOH (pH 7.5), 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% Nonidet P-40, 0.25% Triton X-100, and protease inhibitor (Roche Applied Science)) on ice for 10 min and then centrifuged at 1,400 ϫ g for 5 min. The cell pellet was resuspended in 3 ml of lysis buffer 2 (10 mM Tris-HCl (pH 8.0), 200 mM NaCl, 1 mM EDTA, and 5 mM EGTA) for 10 min at room temperature and then centrifuged at 1,400 ϫ g for 5 min. The cell pellet was resuspended in 1 ml of lysis buffer 3 (10 mM Tris-HCl (pH 8.0), 300 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, and 0.1% sodium deoxycholate). Liver samples were ground into pieces with a pestle and mortar under liquid nitrogen and fixed in PBS containing 20 mM sodium butyrate, 1% formaldehyde, and protease inhibitor mixture for 10 min at room temperature. After centrifugation, the pellet was resuspended in lysis buffer (50 mM Tris-HCl (pH 8.0), 10 mM EDTA, 1% sodium dodecylsulfate, 20 mM sodium butyrate, and protease inhibitor). The cell lysate from HepG2 cells and liver samples was disrupted by sonication (UR-20P, Tomy, Tokyo, Japan) for 5 min on ice and then diluted 10-fold with RIPA ChIP buffer (10 mM Tris-HCl (pH 8.0), 140 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 1% Triton X-100, 0.1% sodium deoxycholate, 20 mM sodium butyrate, and protease inhibitor) followed by centrifugation at 20,000 ϫ g for 10 min. A small volume of the supernatant was stored at 4°C as the input sample. The remaining supernatant was precleared by adding of 15 l of a 50% slurry of protein G-Sepharose 4 Fast Flow (GE Healthcare) with sonicated salmon sperm DNA and rotated for 4 h at 4°C followed by centrifugation at 1,900 ϫ g for 5 min. The supernatant was divided into two aliquots and incubated with 4 g of anti-HNF4␣ antibody or control normal goat IgG (Santa Cruz Biotechnology) for 16 h at 4°C and incubated with 40 l of a 50% slurry of protein G-Sepharose 4 Fast Flow. The reaction mixture was centrifuged at 1,900 ϫ g for 5 min at 4°C, and the pellet was collected. The pellet was washed for 5 min at 4°C with 1 ml of RIPA-1 buffer (50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% SDS, and 0.1% sodium deoxycholate), RIPA-1 buffer containing 300 mM NaCl, LiCl wash solution (10 mM Tris-HCl (pH 8.0), 0.25 M LiCl, 1 mM EDTA, 0.5% Nonidet P-40, and 0.5% sodium deoxycholate), and Tris-EDTA. Then ChIP direct elution buffer (10 mM Tris-HCl (pH 8.0), 300 mM NaCl, 5 mM EDTA, and 0.5% SDS) was added to the pellet and incubated for 16 h at 65°C for decrosslinking. After treatment with RNase A for 30 min at 37°C and proteinase K for 2 h at 55°C, DNA was purified and amplified by PCR and real-time PCR using the ⌬⌬Ct method. Enrichment of HNF4␣ binding was normalized to the input samples and expressed as -fold enrichment as compared with the control normal IgG antibody. Sequences for the primers are shown in supplemental Table S4.

3-UTR assays
3Ј-UTRs of the mouse candidate genes for the miR-194 and miR-192 were amplified by PCR and cloned downstream of firefly luciferase gene in pGL3-control (Promega). As positive controls, complementary sequences for miR-194 and miR-192 (c-miR-194 and c-miR-192) were self-annealed and then cloned into the XbaI site of the pGL3-control. For 3Ј-UTR assay, pGL3-control/3Ј-UTR plasmids and pGL4.74 with 20 nM miR-194 or miR-192 mimic were co-transfected into HEK293T cells using HilyMax (Dojindo, Kumamoto, Japan) as a transfection reagent and assayed by the Dual-Glo Luciferase Assay System (Promega). Sequences for the primers of the 3Ј-UTRs are shown in supplemental Table S5.

Statistical analysis
All values are expressed as the mean Ϯ S.D. All data were analyzed by unpaired Student's t test for significant differences between the mean values of each group.