Downregulation of miR122 by grainyhead-like 2 restricts the hepatocytic differentiation potential of adult liver progenitor cells

Many developing organs contain multipotent tissue-specific stem/progenitor cells that may be extinguished later in development (Zhou et al., 2007; Pan et al., 2013). Conversely, tissue-specific stem/progenitor cells are thought to contribute to supplying cellular components in mature organs for the maintenance of organ homeostasis and for recovery from injury. In some adult tissues/organs, such as the intestine and skin, it has been shown that residential stem/progenitor cells continuously produce multiple cell types in homeostasis (Fuchs, 2009; Barker et al., 2010). In many epithelial tissues and organs, however, it is not clear whether tissue-specific stem/progenitor cells actually contribute to cellular turnover throughout life.

The liver contains two types of epithelial cells, namely hepatocytes and cholangiocytes (biliary epithelial cells), which originate from fetal liver stem/progenitor cells (LPCs) termed hepatoblasts (Oertel et al., 2003; Tanimizu et al., 2003). Under severe chronic liver injury the ability of hepatocytes and/or cholangiocytes to multiply is exhausted, and then adult LPCs may be activated to supply these cell types. In addition to LPCs in regenerating livers, it has been reported that adult LPCs exist in the normal liver. Adult LPCs have been enriched in cellular fractions and express cholangiocyte markers, including epithelial cell adhesion molecule (EpCAM), CD133 (also known as PROM1), CD13 (also known as ANPEP), SRY-related HMG box transcription factor 9 (SOX9) and osteopontin (OPN; also known as SPP1) (Suzuki et al., 2008; Kamiya et al., 2009; Okabe et al., 2009; Cardinale et al., 2011; Dorrell et al., 2011; Furuyama et al., 2011; Español-Suñer et al., 2012). By genetic lineage tracing it was demonstrated that SOX9⁺ LPCs forming the ductal plate, which constitutes the primitive structure of the bile duct, actually supplied cholangiocytes as well as mature hepatocytes (MHs) (Carpentier et al., 2011). By contrast, recent results showed that adult LPCs barely supplied MHs in normal liver (Malato et al., 2011; Español-Suñer et al., 2012). These results suggest that LPCs may exist in or near bile ducts and that LPCs might alter their differentiation potential during development.

In this study, we used EpCAM as a marker to enrich LPCs, which form bipotential colonies in a clonal condition. We found that neonatal EpCAM⁺ LPCs produce hepatocytes much more efficiently than adult LPCs. We further demonstrated that grainyhead-like 2 (GRHL2), a cholangiocyte-specific transcription factor, downregulates miR122, a microRNA (miRNA) known to be important for hepatocytic differentiation. Thus, our data indicate that one crucial transcription factor is involved in defining the differentiation capability of epithelial tissue-specific stem/progenitor cells.

It has been reported that the number of hepatoblasts (fetal LPCs) decreases from mid- to late fetal stage (Suzuki et al., 2002), that adult LPCs exist in livers beyond 2 weeks after birth, and that their number decreases during postnatal development (Kamiya et al., 2009). In accordance with a previous report (Okabe et al., 2009), we used EpCAM as a marker to enrich adult-type LPCs. We immunohistochemically confirmed that EpCAM⁺ cells are localized next to the portal vein after E17 (Fig. 1A). A single-cell suspension was used to isolate single EpCAM⁺ cells from the CD45⁻ TER119⁻ CD31⁻ non-hematopoietic/non-endothelial cell fraction (Fig. 1B). As shown in Fig. 1C, EpCAM⁺ cells formed bipotential colonies containing albumin (ALB)⁺ hepatocytes and CK19⁺ cholangiocytes. Interestingly, 1-week EpCAM⁺ cells formed bipotential colonies most efficiently (Table 1 and Fig. 1D). By contrast, EpCAM⁺ cells isolated from 6- to 11-week livers mostly formed cholangiocyte colonies containing only CK19⁺ cells (Fig. 1D). These results are consistent with a previous report demonstrating that the number of LPCs possessing the ability to form bipotential colonies decreases during postnatal development (Kamiya et al., 2009). In addition, we noted that the ratio of ALB⁺ to total colony-forming cells was much lower in bipotential colonies derived from 8-week versus 1-week EpCAM⁺ cells (Fig 1Cb,c). This suggests that, in addition to a reduction in LPC numbers, the hepatocytic differentiation potential of LPCs might decrease as postnatal development progresses.

To further compare the differentiation capability of neonatal and adult LPCs, we examined the ratio of ALB⁺ cells, including both ALB⁺ CK19⁻ and ALB⁺ CK19⁺, among colony-forming cells (Fig. 2A). The results indicated that more than 50% of cells were ALB⁺ in colonies derived from 1-week EpCAM⁺ cells (neonate) as compared with only ∼10% of cells in colonies derived from 6-week EpCAM⁺ cells (adult).

Next, we analyzed the level of albumin expression in each bipotential colony. For this purpose, we prepared cDNA from total RNA extracted from each colony and examined whether each colony expresses albumin (Fig. 2C). We then selected albumin⁺ colonies and quantified the albumin expression level. The level of albumin expression was much higher in neonatal bipotential colonies than in adult colonies (Fig. 2D). Expression of Ck19 (Krt19), Sox9 and Epcam was relatively low in colonies expressing albumin at high level.

To further clarify whether neonatal and adult LPCs can differentiate into hepatocytes, we subcultured colonies derived from EpCAM⁺ cells isolated from 1-week and 6-week livers. We picked up ten colonies of each and successfully established four and five clones from 1-week and 6-week colonies, respectively. Two out of four neonatal clones (clones 2 and 4) retained the original marker EpCAM, whereas it was eventually lost from the other clones (data not shown).

We induced hepatocytic differentiation of LPC clones by oncostatin M (OSM) and Matrigel (MG). After induction, all neonatal clones expressed marker genes related to differentiated functions of hepatocytes, including metabolic enzymes and cytochrome P-450s (Fig. 3A). By contrast, adult clones were induced to express only albumin and carbamoyl-phosphate synthetase 1 (Cps1). We quantified the expression level of Cps1 and tyrosine aminotransferase (Tat) and the data showed that neonatal clones expressed both genes at a level comparable to MHs (Fig. 3B). Neonatal cells showed hepatocytic morphology after induction (Fig. 3Ca,b). Furthermore, the expression of hepatocyte nuclear factor 4α (HNF4α), CCAAT/enhancer binding protein α (C/EBPα), ALB and CPS1 was induced in neonatal clones (Fig. 3Ca-h). Neither morphological changes nor the expression of HNF4α and ALB was clearly induced in adult clones (Fig. 3Ci-m). These results further support the proposal that neonatal LPCs have greater potential than adult LPCs to differentiate into hepatocytes.

In order to examine the differentiation capability of neonatal and adult LPCs in vivo, we transplanted neonatal and adult clones. All the clones were labeled with GFP, and GFP⁺ cells were enriched by FACS (Fig. 4A). Recipient mice were pretreated with retrorsine, which inhibits hepatocyte proliferation, and then 70% partial hepatectomy was performed at the time of transplantation (Fig. 4B). Recipient liver sections were stained with anti-GFP antibody to identify donor cells (Fig. 4C). Engraftments of donor cells were observed in recipients transplanted with either neonatal or adult clones (Table 2). Neonatal clones were engrafted as hepatocytic or ductular cells (Fig. 4Ca,b). By contrast, the engraftment of adult cells was rare and these were mostly found as ductular cells (Fig. 4Cc).

We further examined the recipient livers transplanted with neonatal cells by expression of hepatocyte and cholangiocyte markers (Fig. 4D). Neonatal clones differentiated to HNF4α⁺ hepatocytes or CK19⁺ cholangiocytes (Fig. 4Da-d). Since recipient CD31⁺ sinusoidal endothelial cells entered into a cluster of HNF4α⁺ donor cells, donor-derived hepatocytes were incorporated into recipient liver trabeculae (Fig. 4De-h). We also examined the expression of hepatocyte and cholangiocyte markers in recipient livers transplanted with adult cells and found that most of the donor cells differentiated to CK19⁺ ductular cells (Fig. 4Ea-d). The GFP⁺ ductular structure (dotted circle in supplementary material Fig. S1a-c) was separate from the portal veins (asterisks in supplementary material Fig. S1). The differentiated cells also expressed other cholangiocyte markers (supplementary material Fig. S1f-w). A small proportion of donor cells became CK19⁺ HNF4α⁺ cells, which did not form the ductular structure (Fig. 4Ee). These results indicate that EpCAM⁺ LPCs show bidirectional differentiation capability in vivo, although adult LPCs rarely differentiate into hepatocytes.

These results further support the proposal that the bipotency of LPCs is altered during development. Thus, we considered that a molecular machinery restricting the differentiation potential might exist in adult LPCs.

It has been demonstrated that certain transcription factors are crucial in defining cellular lineages. We previously found that cholangiocyte transcription factors, including GRHL2 and SOX9, are expressed at a higher level in adult than in neonatal EpCAM⁺ cells (Tanimizu et al., 2013). However, we had not examined whether neonatal EpCAM⁺ cells equally expressed these markers at low level or if they contain cells positive and negative for the markers. We prepared smear samples of FACS-purified EpCAM⁺ cells and examined the expression of HNF1β, SOX9, GRHL2 and HNF4α, as well as CK19 and OPN in each neonatal and adult EpCAM⁺ cell (Table 3). Consistent with the data in Table 3, HNF1β, SOX9 and OPN, but not HNF4α, are expressed in neonatal and adult EpCAM⁺ cells in liver tissues (supplementary material Figs S2 and S3). Interestingly, ∼25% of neonatal EpCAM⁺ cells were GRHL2⁻ (Fig. 5Aa-c), whereas the adult EpCAM⁺ fraction barely contained any GRHL2⁻ cells (Fig. 5Ad-f).

We further confirmed that neonatal EpCAM⁺ cells were GRHL2⁺ (Fig. 5Bb-d) or GRHL2⁻ (Fig. 5Be-g) by examining its expression in neonatal liver sections. Neonatal EpCAM⁺ GRHL2⁻ cells formed smaller ductules in the periphery. By contrast, bile ducts in the adult liver consist of EpCAM⁺ GRHL2⁺ cells (Fig. 5Bh-n). Additionally, the expression pattern of CK19 was similar to that of GRHL2 (supplementary material Fig. S3). Previously, we demonstrated that GRHL2 not only promotes the maturation of cholangiocytes but also inhibits hepatocytic differentiation (Tanimizu et al., 2013). Therefore, these results suggest that substantial differences in differentiation potential might exist among neonatal EpCAM⁺ cells, as well as between neonatal and adult cells. We further examined the possibility that GRHL2 expression might regulate the hepatocytic differentiation potential of LPCs.

To correlate the expression profile of GRHL2 with the bipotency of LPCs, we examined its expression in colonies derived from neonatal and adult EpCAM⁺ cells. We found that intense and uniform expression of GRHL2 was maintained in colonies derived from adult EpCAM⁺ cells (Fig. 6Aa-e). By contrast, neonatal EpCAM⁺ cells formed two types of colonies: one consisted of GRHL2⁻ cells (Fig. 6Af-j) and the other of GRHL2⁺ and GRHL2⁻ cells (Fig. 6Ak-o). Although the neonatal GRHL2⁻ and GRHL2⁺ colonies were similar in size, the GRHL2⁻ colonies contained more ALB⁺ cells (Fig. 6B). In addition, in GRHL2⁺ colonies ALB was restricted to cells in which GRHL2 was spontaneously downregulated (Fig. 6C). Based on these results, we considered the possibility that targets of GRHL2 might be involved in reducing the hepatocytic differentiation potential of LPCs.

To identify molecules downstream of GRHL2, we performed mRNA and miRNA microarray analyses using a hepatic progenitor cell line HPPL with or without GRHL2. Since HPPL cells do not express GRHL2, we expected that its targets would be clearly upregulated or downregulated in HPPL cells with GRHL2. The results of the mRNA microarray suggested that some metabolic genes, such as arginase and Cps1 were downregulated by overexpression of GRHL2. The results of the miRNA microarray, by contrast, suggested that miR122 was remarkably downregulated by GRHL2 (Table 4; supplementary material Table S3). We confirmed this result by quantitative PCR analysis, showing that overexpression of GRHL2 substantially downregulated miR122 in HPPL cells (Fig. 7A).

It has been reported that miR122 is important for hepatocytic differentiation (Laudadio et al., 2012). Consistently, overexpression of GRHL2 in HPPL cells, in which miR122 was downregulated (Fig. 7A), inhibited hepatocytic differentiation (supplementary material Fig. S4). We also found that knockdown of GRHL2 increased miR122 in adult EpCAM⁺ cells (Fig. 7B). We further examined the effect of miR122 on hepatocytic differentiation by introducing a miR122 mimic, and found that it slightly upregulated albumin and downregulated Ck19 in colonies derived from adult EpCAM⁺ cells (Fig. 7C). We also found that the miR122 mimic upregulated albumin, Cps1 and Tat but not Hnf4a in primary EpCAM⁺ cells (supplementary material Fig. S5). Consistent with the notion that the Notch signaling pathway promotes cholangiocyte differentiation (Tanimizu and Miyajima, 2004; Zong et al., 2009), DAPT, which is a γ-secretase inhibitor and thereby a potent inhibitor of the Notch signaling pathway, further promoted upregulation of Cps1 and Tat by miR122 (supplementary material Fig. S5). Consistent with their reduced capacity to differentiate into hepatocytes, miR122 expression in adult EpCAM⁺ cells was significantly lower than in neonatal cells, whereas Grhl2 expression in adult cells was higher than in neonatal cells (Fig. 7D). Furthermore, in neonatal and adult LPC colonies, Grhl2 and Mir122 expression levels are negatively and positively correlated with albumin expression, respectively (supplementary material Fig. S6). These results further support the hypothesis that downregulation of miR122 by GRHL2 is one of the key molecular pathways restricting the hepatocytic differentiation potential of LPCs.

It has been reported that Mir122 expression is controlled by the proximal promoter and enhancers (Xu et al., 2010; Laudadio et al., 2012). In order to determine whether GRHL2 transcriptionally regulates Mir122, we examined its effect on reporter constructs containing the Mir122 proximal promoter/enhancer (Fig. 7E). The results indicate that GRHL2 negatively regulates Mir122 promoter/enhancer activity.

Each tissue/organ may contain tissue-specific stem/progenitor cells that contribute to organogenesis, homeostasis and the recovery from injury. However, it remains unclear whether tissue stem/progenitor cells retain the same differentiation potential throughout life. In this study, by examining characteristics of LPCs continuously between late fetal and adult periods, we demonstrate that the number of LPCs and their differentiation capability are restricted during postnatal liver development. We further identified that downregulation of miR122 by GRHL2 is one of the molecular pathways regulating the differentiation capability of LPCs.

We have shown (Fig. 7) that miR122 alone can increase the expression of albumin in adult LPCs. However, given that overexpression of GRHL2 remarkably inhibited hepatocytic differentiation (supplementary material Fig. S4), the effect of miR122 on hepatocytic differentiation is relatively weak. We previously found that overexpression of GRHL2 inhibited the expression of HNF4α in neonatal EpCAM⁺ cells, which key to suppressing hepatocytic differentiation. However, introduction of a miR122 mimic did not significantly increase Hnf4a expression in adult EpCAM⁺ cells (supplementary material Fig. S5). Taken together, we consider that, in addition to miR122, other downstream targets of GRHL2 also participate in modulating the differentiation potential of LPCs.

LPCs have been thoroughly characterized beyond 2 weeks after birth (Kamiya et al., 2009). In addition, by examining the differentiation potential of late fetal, neonatal and adult LPCs, we found that neonatal LPCs showed a strong ability to differentiate into functional hepatocytes in vitro and in vivo, suggesting that EpCAM⁺ LPCs are a candidate for supplying hepatocytes during early postnatal development as embryonic ductal plate cells (Carpentier et al., 2011). However, since hepatocytes show the ability to proliferate even in a clonal condition by 4 weeks (data not shown), they certainly play an important role in supplying new hepatocytes during early postnatal development. Therefore, further experiments are necessary to prove whether neonatal EpCAM⁺ LPCs also supply new hepatocytes during development.

Adult LPCs have been enriched in cellular fractions that are positive for cholangiocyte markers and they have potential to differentiate to hepatocytes in vitro and in vivo (Suzuki et al., 2008; Okabe et al., 2009; Cardinale et al., 2011; Dorrell et al., 2011; Furuyama et al., 2011; Español-Suñer et al., 2012). Consistent with previous work, we demonstrated that adult LPCs are enriched in the EpCAM⁺ fraction. However, compared with 1-week LPCs, their capability to differentiate into hepatocytes was very low. Given that recent lineage-tracing studies failed to detect MHs derived from LPCs in vivo without injury (Español-Suñer et al., 2012), the differentiation capability of LPCs appears extremely limited at the late postnatal and adult period. Specific stimulation, including Wnt plus R-spondin (Huch et al., 2013), and/or inhibition of the Notch signal (Boulter et al., 2012) might be necessary to release adult LPCs from the molecular machineries that serve to block hepatocytic differentiation.

By examining the expression of marker genes at the single-cell level, we found that neonatal EpCAM⁺ cells can be divided into GRHL2⁻ and GRHL2⁺. Interestingly, EpCAM⁺ GRHL2⁻ cells formed small ductules in the periphery of the neonatal liver. According to the recent model of bile duct development, in which development/morphogenesis proceed from the liver hilum to the periphery (Antoniou et al., 2009), these EpCAM⁺ GRHL2⁻ cells might be recently committed from hepatoblasts and thereby retain bidirectional differentiation capability. In addition, in contrast to adult cells, neonatal EpCAM⁺ GRHL2⁺ cells spontaneously lose expression of GRHL2 and then express ALB during clonal expansion (Fig. 6C). Therefore, it can be assumed that GRHL2 expression is not necessarily fixed even in EpCAM⁺ GRHL2⁺ cells at neonatal stage, and, thereby, that neonatal EpCAM⁺ GRHL2⁺ cells retain the potential to differentiate into hepatocytes.

In this study, we demonstrated that neonatal and adult LPCs are distinct in terms of their hepatocytic differentiation potential. It has been demonstrated that the combination of GATA4, FOXA3 and HNF1α (Huang et al., 2011) or of HNF4α and one of the FOXA factors (Sekiya and Suzuki, 2011) converts fibroblasts into hepatocytes. By contrast, overexpression of HNF4α in the presence of endogenous FOXA1 did not induce hepatocytic characteristics in EpCAM⁺ cells (Tanimizu et al., 2013). Given that recent reports suggested that, even in chronically injured livers, LPCs do not efficiently supply new hepatocytes (Tarlow et al., 2014), adult LPCs might be equipped with molecular machineries blocking hepatocytic differentiation. The present finding that miR122 is downregulated by GRHL2 might provide an explanation for such mechanisms. Recently, it was demonstrated that key transcription factors defining neural lineage cells, including PAX6 and ETV6, inhibit dedifferentiation induced by OCT3/4 (POU5F1), SOX2, KLF4 and cMYC during reprograming (Hikichi et al., 2013). As we have previously demonstrated (Senga et al., 2012; Tanimizu et al., 2013), GRHL2 is a key transcription factor in inducing the cellular properties of mature cholangiocytes. Therefore, transcription factors conferring specific functions and/or structures to certain cell types might generally define the differentiation potential of tissue-specific stem/progenitor cells. In order to use highly proliferative LPCs in vitro and in vivo for the purposes of producing functional hepatocytes, it might be important to release them from inhibitory machineries in order to make them responsive to inductive signals for hepatocytic differentiation.

Growth factor-reduced Matrigel (MG) and purified laminin 111 were from BD Biosciences. Epidermal growth factor (EGF), hepatocyte growth factor (HGF) and oncostatin M (OSM) were purchased from R&D Systems. C57BL6 and nude mice were purchased from Sankyo Labo Service Corporation (Tokyo, Japan). All animal experiments were approved by the Sapporo Medical University Institutional Animal Care and Use Committee and were carried out under the institutional guidelines for ethical animal use.

E17 mouse livers were cut into small pieces and then digested in PBS containing Mg²⁺, Ca²⁺ and Librase-TM (Roche Applied Sciences). For neonatal and adult mouse livers, a two-step collagenase perfusion method was used to isolate cells as previously reported (Tanimizu et al., 2013). Liberated cells were used to isolate EpCAM⁺ cells by FACS. Non-specific binding of antibodies was blocked by antibody against Fcγ receptor (anti-CD16/CD32 antibody). Cells were labeled with FITC-conjugated anti-EpCAM, APC-conjugated anti-CD45 (PTPRC), APC-conjugated anti-CD31 (PECAM1) and APC-Cy7-conjugated anti-TER119 (LY76) antibodies, and with propidium iodide (PI; Sigma-Aldrich). EpCAM⁺ cells were isolated from PI⁻ live cells by FACS using a FACSAria II (BD Biosciences). Antibodies used are listed in supplementary material Table S1.

Single EpCAM⁺ cells were isolated by FACSAria II as described above. Five to eight thousand cells were plated in a 35-mm dish coated with laminin 111 (BD Biosciences). The cells were cultured in DMEM/Fischer F-12 medium supplemented with 10% FBS, 10 ng/ml EGF, 10 ng/ml HGF, 0.1 μM dexamethasone (Dex; Sigma-Aldrich) and 1× insulin-transferrin-selenium (ITS; Gibco). After 9 days of culture, cells were fixed in 4% paraformaldehyde (PFA) and stained with anti-cytokeratin 19 (CK19) and anti-albumin antibodies (supplementary material Table S1). Signals were visualized with Alexa Fluor-conjugated secondary antibodies (Molecular Probes). Nuclei were counterstained with Hoechst 33258 (Dojindo). Images were acquired on a Nikon X-81 fluorescence microscope. Colonies containing more than 50 cells were categorized into three groups: bipotential colonies, which consist of ALB⁺ hepatocytes and CK19⁺ cholangiocytes; hepatocyte colonies, which consist of only ALB⁺ cells; and cholangiocyte colonies, which consist of only CK19⁺ cells. We define a cell that forms bipotential colonies as an LPC in the current study.

For establishing clones, each colony raised in the low density culture was treated with trypsin after being surrounded by a cloning ring (Asahi Glass Co., Tokyo, Japan). Cells were then plated in a well of a 24-well plate coated with laminin 111 and kept in the medium that was used for the colony assay. Expanded cells were replated every 4 days.

To induce neonatal and adult LPC clones to differentiate to hepatocytes, 50,000 cells were plated on gelatin-coated wells of 24-well plates. After cells became confluent, they were incubated with 10 ng/ml OSM, 1% DMSO, 0.1 μM Dex and 1× ITS for 4 days and then overlaid with 5% MG for an additional 4 days. Stealth RNA against Grhl2 (Invitrogen) and miR122 mimic (Ambion) were introduced into cells using RNAiMAX (Invitrogen).

Total RNA was isolated from purified EpCAM⁺ cells using an RNeasy Mini Kit (Qiagen). cDNA was synthesized using the Omniscript Reverse Transcription Kit (Qiagen). Primers used for PCR are listed in supplementary material Table S2. TaqMan probes for Hprt, Cps1, Grhl2, miR122, Sox9, Tat and U6 snRNA were purchased from Applied Biosystems and used for quantitative PCR.

Nude mice were used for cell transplantation as described previously (Kamiya et al., 2009). LPC clone cells were labeled by GFP-encoding lentivirus vector and GFP⁺ cells were isolated by FACSAria II. Recipient 4-week nude mice were intraperitoneally injected with 60 mg/kg retrorsine (Sigma-Aldrich) four times each week and then 70% of the liver was surgically removed at transplantation. GFP⁺ cells (5×10⁵-1×10⁶) were injected into spleen. Six weeks after the transplantation, the liver was collected to examine the engraftment of donor cells. For detecting GFP⁺ cells in recipient livers, frozen sections were incubated with anti-GFP antibody and then GFP signals were visualized using 3,3′-diaminobenzidine (DAB) (Sigma-Aldrich) as a chromogenic substrate for horseradish peroxidase, which was conjugated to anti-rabbit IgG antibody. Areas of sections were measured using Olympus CellSense Software.

Liver tissues were fixed in PBS containing 4% PFA or in Zamboni solution at 4°C overnight. After washing in PBS and in PBS containing 20% sucrose, they were embedded in OCT compound (Sakura Finetek). Thin sections were prepared with a cryostat (CM1950, Leica Microsystems). After permeabilization with 0.2% Triton X-100 and blocking with Blockace (DS Pharma), sections were incubated with primary antibodies (supplementary material Table S1). Images were obtained with Zeiss 510 and 780 confocal microscopes.

Grhl2 cDNA was amplified and inserted into a pcDNA3-myc vector, which was generated by inserting a myc-tag into pcDNA3 (Life Technologies). Mir122 proximal promoter and enhancer sequences were amplified using mouse genomic DNA as a template and inserted into pUC118 vector (Takara Bio). The promoter and enhancer were inserted into the KpnI/XhoI sites of a pGL4.10 vector (Promega), which contained minimal TK promoter. HEK293T cells were co-transfected using pcDNA3 with or without Grhl2, pGL4.10 with or without Mir122 promoter/enhancer, and a pRL-TK vector containing Renilla luciferase. At 48 h after transfection, luciferase activity was measured with the Dual-Luciferase Reporter Assay System (Promega) on an Infiniti M1000-Pro microplate reader (Tecan).

Statistical analyses to determine s.e.m. for colony assay, qPCR and luciferase assay and two-tailed Student's t-tests were performed using Excel (Microsoft).

Total RNA including microRNAs was extracted from HPPL cells infected with pMXs-Neo or with pMXs-Neo-Grhl2 using miRNeasy (Qiagen). Hybridization and data analysis were performed by Miltenyi Biotec using miRXplore. Microarray data are available at GEO under accession number GSE62862.

We thank Ms Minako Kuwano and Ms Yumiko Tsukamoto for technical assistance.