Genome-wide microRNA and messenger RNA profiling in rodent liver development implicates mir302b and mir20a in repressing transforming growth factor-beta signaling


  • Potential conflict of interest: Nothing to report.


MicroRNAs (miRNAs) are recently discovered small RNA molecules that regulate developmental processes, such as proliferation, differentiation, and apoptosis; however, the identity of miRNAs and their functions during liver development are largely unknown. Here we investigated the miRNA and gene expression profiles for embryonic day (E)8.5 endoderm, E14.5 Dlk1+ liver cells (hepatoblasts), and adult liver by employing Illumina sequencing. We found that miRNAs were abundantly expressed at all three stages. Using K-means clustering analysis, 13 miRNA clusters with distinct temporal expression patterns were identified. mir302b, an endoderm-enriched miRNA, was identified as an miRNA whose predicted targets are expressed highly in E14.5 hepatoblasts but low in the endoderm. We validated the expression of mir302b in the endoderm by whole-mount in situ hybridization. Interestingly, mir20a, the most highly expressed miRNA in the endoderm library, was also predicted to regulate some of the same targets as mir302b. We found that through targeting Tgfbr2, mir302b and mir20a are able to regulate transforming growth factor beta (TGFβ) signal transduction. Moreover, mir302b can repress liver markers in an embryonic stem cell differentiation model. Collectively, we uncovered dynamic patterns of individual miRNAs during liver development, as well as miRNA networks that could be essential for the specification and differentiation of liver progenitors. (HEPATOLOGY 2013)

Generation of hepatocyte-like cells from differentiated pluripotent stem cells or reprogrammed cells provides a potential cell source for liver transplantation and drug testing. However, hepatocyte-like cells generated through in vitro culture cannot fully recapitulate the characteristics of their in vivo counterparts.1 Improving methods for hepatocyte derivation in vitro may benefit from enhancing our understanding of molecular networks regulating liver development in vivo. During mouse embryonic development, liver progenitor cells are specified from definitive endoderm at the 7-8 somite stage (embryonic day [E]8.5).1 At E9.5, cells in the liver domain, known as hepatoblasts, undergo epithelial-to-mesenchymal transition (EMT), invade the surrounding septum transversum, form the liver bud, and rapidly proliferate.2 Around E14.5-15.5, hepatoblasts start to differentiate into hepatocytes or cholangiocytes, which mature into the major functional cells of the liver.3

MicroRNAs (miRNAs) are small, noncoding RNAs that regulate gene expression by binding to complementary sequences within messenger RNAs (mRNAs). They are transcribed as primary miRNAs and processed by DROSHA to generate pre-miRNAs, which are further cleaved by DICER to produce mature miRNAs.4 miRNAs have been shown to contribute to organogenesis4; however, little is known about their roles in liver development. Disruption of Dicer, which leads to the loss of mature miRNAs,4 causes lethality by E7.5 due to defects in proliferation and maintenance of pluripotency in extraembryonic tissues.5, 6 Deletion of Dicer in the embryo proper causes death around E9.5, with developmental retardation and massive apoptosis.6 Thus, Dicer has distinct roles in embryonic and extraembryonic tissues. Deletion of Dicer specifically in liver by mating Dicerfl/fl mice with Alb-Cre mice shows defects in liver zonation and promotes hepatocarcinogenesis.7-9 However, this model did not address miRNA functions in embryonic liver development because Dicer activity is not depleted until birth.

By using microarrays, Hand et al.10 found a set of miRNAs that are differentially expressed in E15.5, E18.5, and postnatal day 2 livers in the mouse. By in situ hybridization in zebrafish, two ductal plate and bile duct specific miRNAs, mir30a and mir30c, were identified. Knock-down of mir30a resulted in defects in biliary development. By combining miRNA expression patterns from E16.5, E17.5, postnatal day 1, and adult mouse liver with mRNA expression data from in vitro liver differentiation, the mir-23/24/27 cluster was found to regulate liver stem cell differentiation through modulating TGFβ signaling.11 These studies demonstrate the importance of miRNAs during the later differentiation stages of liver development; however, how miRNAs contribute to critical phases of early liver specification and progenitor cell proliferation remains unexplored.

To identify miRNAs expressed during liver development in vivo, we used Illumina sequencing to generate miRNA libraries from E8.5 foregut, E14.5 hepatoblasts, and adult liver. Notably, the majority of miRNAs were expressed in all three libraries but exhibited temporal changes in expression levels. By integrating mRNA and miRNA expression patterns, we identified two endoderm-enriched miRNA, mir302b and mir20a, that can regulate Tgfbr2 and Kat2b expression, both of which function in TGFβ signaling. We verified that both mir302b and mir20a can modulate TGFβ signal transduction. Suppressing TGFβ signaling by SB505124 or mir302b reduced liver marker expression during mouse embryonic stem cell (ESC) differentiation, suggesting mir302b and mir20a may regulate liver development through TGFβ signaling. Together, our microRNAome profiling provides detailed information on miRNA expression during liver development in vivo.

Materials and Methods

Animals, RNA Preparation, and Sequencing.

C57BL/6J mice were maintained according to protocols approved by the Animal Care Committee of the University of British Columbia following guidelines established by the Canadian Council on Animal Care. Endoderm isolation was previously described.12 Hepatoblast isolation is described in the Supporting Methods. Adult liver was obtained from a 6-month-old female mouse. Small RNA preparation and library construction was previously described.13 Libraries were sequenced on Illumina GAIIx.

Small RNA Analysis.

Reads were aligned to NCBI37/mm9 reference genome and miRNA annotation was based on miRBase V15. Read processing, quantification, and annotation was previously described.13 Expression of miRNAs was normalized to total reads aligned to genome and expressed as reads per million (RPM). Two replicates from foregut, two from hepatoblasts, and one from adult liver were generated. The expression correlation between replicates was r2 = 0.9282 and r2 = 0.9718 for foregut and hepatoblasts, respectively (Supporting Fig. S1A). We used one replicate for further analysis. miRNAs expressed at greater than 10 RPM were used in K-means clustering analysis.14 Novel miRNA prediction was previously described.13

Real-Time Quantitative Reverse Transcription Polymerase Chain Reaction (qRT-PCR).

RNA was extracted using mirVana miRNA isolation kit (Ambion). Real-time quantification was performed using TaqMan MicroRNA Assays (Applied Biosystems) or SYBR Green (Roche) according to the manufacturer's instructions. All expression results were normalized to U6 or Actin for miRNA and gene expression, respectively.

Vectors and Cell Culture Assays.

The mir302b overexpression vector, pCMV-mir302b-IRES-GFP (302b_OE), and its control vector, pCMV-mir-IRES-GFP (Ctrl_OE), were purchased from Origene. A lentiviral-mediated gene expression system, including mir302b expression vector, pCDH302b, and its control vector, pCDH, was purchased from SBI. The mir20a knockdown vector, pCAG-d2eGFP-20a (20a_KD), and its control vector, pCAG-d2eGFP-Cxcr4 (Ctrl_KD), were subcloned from pCMV-d2eGFP-20a and pCMV-d2eGFP-Cxcr4,15 respectively (Addgene). Wildtype or mutant 3′ untranslated region (UTR) miRNA targets were cloned into pmirGLO (Promega). Tgfbr2 expression vectors, pBOS-Tgfbr2 and pBOS-Tgfbr2(Dominant negative, DN), were subcloned from pCMV5-Tgfbr2 and pCMV5-Tgfbr2(DN),16 respectively. 3TP-lux was described previously.16 TK-Renilla controlled for transfection efficiency. Western results were quantified by ImageJ. Luciferase assay is described in the Supporting Methods. ESC differentiation was previously described.17

Whole-Mount In Situ Hybridization (WISH).

Probes for WISH were obtained from Exiqon and experiments were performed according to Sweetman's protocol18 at a temperature of 20° below the melting temperature of probes. No probe and mir29a served as negative controls (Fig. S2).


All data presented are representative of at least three independent experiments unless indicated otherwise. Statistical analysis was performed using Student's t test or one-way analysis of variance (ANOVA), followed by the SNK test if necessary. Statistical significance was inferred at *P < 0.05 and **P < 0.01.


MicroRNAome of Foregut, Hepatoblast, and Adult Liver.

To identify miRNAs expressed during liver development, we performed small RNA sequencing from E8.5 foregut, containing liver progenitor cells, E14.5 hepatoblasts, and adult female liver (∼70% hepatocytes). Sequence-based approaches are quantitative and expression levels can be compared between different miRNAs. Foregut endoderm was isolated by mechanical dissociation from embryos at somite stage 8-12.12 Since nearly 70% cells of E14.5 liver are hematopoietic cells,3 hepatoblasts were isolated to >90% purity using fluorescence-activated cell sorting (FACS) for the surface marker, Dlk1 (Dlk1+) (Figs. S3, S4). Dlk1+ cells comprise a bipotential population that can differentiate into cholangiocytes and hepatocytes.19, 20 Dlk1+ cells overlap hepatocyte nuclear factor 4 alpha (HNF4α) expression at E14.5, but not myeloid, endothelial, or mesenchymal markers (Fig. S3). Reads from foregut (4,286,769), hepatoblasts (5,160,511), and adult liver (4,176,620) that uniquely mapped to the genome were compared to annotated miRNAs. We also compared our adult female liver library to an adult male liver library which employed a similar method.21 Of the 592 miRNA/miRNA* identified, more than 60% were expressed in both libraries, and the expression level correlation was 0.5807 (Fig. S1B). Thus, while gender differences likely exist, there is substantial overlap in miRNA expression in male and female livers.

In the foregut and hepatoblast libraries, 59% and 64% of reads aligned to 430 and 384 known miRNA genes, respectively, while 91% aligned to 315 miRNA genes in adult liver (Fig. 1A). The remaining reads mapped to known transcripts, rRNA and tRNA, or resulted from degradation products and genomic repeats. Surprisingly, among the sequences that mapped at unannotated regions, only three had features resembling novel miRNAs, all of which were present in embryonic tissues but were low in adult (Supporting Table S1). Most of the annotated miRNAs had an expression level ranging from 10 RPM to 1,000 RPM; only a few were expressed at more than 104 RPM (Fig. 1B, Table S2). Of note, more than 60% of miRNAs were present in all three libraries (Fig. 1C).

Figure 1.

Summary of small RNA sequencing. (A) Distribution of annotated small RNA sequences in the foregut, hepatoblast and adult liver libraries. (B) The frequency of miRNAs that are expressed at the defined levels. The majority of miRNA/miRNA* are expressed between 10-1,000 RPM (reads per million). (C) Venn diagram showing the overlap of expressed miRNA in the three libraries.

To compare expression patterns of individual miRNAs in the three libraries, we used K-means clustering analysis. Thirteen temporally related groups (designated Clusters A-M) of miRNAs were identified, including three clusters in which the miRNAs were highly and specifically enriched in one library (Fig. S1D). Cluster A contained miRNAs with high expression in foregut, including miR302b and the mir17-92 group. Cluster H, containing miRNAs expressed highly in hepatoblasts, was enriched for mir379. Cluster L, containing miRNAs expressed highly in adult liver, was enriched for let7 family members (Fig. 2A; Table S3). Thus, our data identified groups of miRNAs dynamically expressed during liver development.

Figure 2.

The dynamic expression of miRNA genes during liver development. (A) Based on the expression pattern of each miRNA, miRNA clusters consisting of expression enriched in foregut (Cluster A), hepatoblast (Cluster H), and adult liver (Cluster L) were identified. (B) Validation of mir302b and mir24 expression by qRT-PCR. U6 was used as an endogenous control. All data presented are representative of at least three independent experiments.

Integration of Gene and miRNA Expression Patterns.

miRNAs can regulate gene expression by binding to the 3′UTR of specific mRNA transcripts, resulting in their degradation. Thus, the expression level of target mRNAs could be inversely correlated with that of the miRNA. To identify reciprocal mRNA-miRNA patterns, RNA-Seq libraries were generated from E8.5 endoderm and E14.5 Dlk1+ liver cells. Of the total 355,195,544 reads sequenced from endoderm RNA, 57.42% were assigned to ENSEMBL transcripts, while 71% of 193,450,752 reads from hepatoblast RNA mapped to known transcripts (further RNA-Seq details are given at To identify differentially expressed transcripts in endoderm and hepatoblasts, we employed Alexa-Seq.22 Briefly, the cumulative base coverage of a feature is normalized to feature length and library size, generating “Normalized Average Coverage” (NAC) values. Transcripts were considered differentially expressed if NAC values differed by a factor of two and their corrected P-value was less than 0.05 (by Fisher's exact test). A total of 5,227 transcripts were enriched in endoderm compared to hepatoblasts (Endoderm-enriched), while 1,599 genes were found to be more highly expressed in the fetal liver (Hepatoblasts-enriched).

To identify miRNAs that may contribute to the regulation of differentially expressed genes, we used DIANA mirExTra (, which identifies miRNAs whose predicted target genes are overrepresented in a subset of genes.23 For Endoderm-enriched genes, mirExTra predicted over 300 miRNAs that were potential regulators of these genes. Of these miRNAs, 44 (14.67%) were more highly expressed in hepatoblasts compared to endoderm (Table S4). Surprisingly, of the miRNAs predicted to regulate Hepatoblasts-enriched genes, only mir302b showed a reciprocal pattern of expression. Analysis showed that mir302b potentially targets 575 out of the total 1,599 (35%) Hepatoblasts-enriched genes (compared to 22% for a random set of genes). Thus, mir302b could play an important role in regulating hepatoblast genes in the endoderm.

mir302b Is Expressed in Endoderm at the Onset of Organogenesis.

To explore mir302b expression in early development, we tested the expression of mir302b by qRT-PCR. Confirming the library sequencing, mir302b was found to be high in endoderm and rapidly down-regulated by E10.5 in Dlk1+ hepatoblasts (Fig. 2B; Fig. S6A). mir24, a miRNA in Cluster J (Fig. S1D), which is upregulated during liver development,11 was validated as a positive control (Fig. 2B). We next analyzed the expression patterns of mir302b in early embryos by WISH. During early gastrulation (E6.5), mir302b was expressed in epiblast (Fig. 3A,C, white arrowhead; Fig. S5A,B) and weakly in the mesoderm (Fig. 3C, black arrow; Fig. S5A,B). As gastrulation progresses (E7.5), mir302b expression was observed mainly in the embryonic ectoderm and mesoderm (Fig. 3B,D, black arrows), but not in newly formed definitive endoderm (Fig. 3D, black arrowhead; Fig. S5C-E). The expression of mir302b in definitive endoderm was first detected at approximately the 3-somite stage (E8.25) (Fig. 3E,F; Fig. S5G-L). mir302b expression was evident throughout the foregut (Fig.3F, black arrowhead), encompassing the region that contains liver progenitors, and in the hindgut region but was excluded from the midgut (Fig. S5J-L). As the embryos developed (E8.75), mir302b was ubiquitously expressed but was absent from the heart (Fig. 4). Sections of these embryos showed that mir302b expression expanded to the entire gut (Fig. 4A-D, black arrowhead). Thus, induction of mir302b in definitive endoderm initiates at the 3-somite stage, corresponding to the stage when liver and pancreatic progenitors are first specified.1

Figure 3.

mir302b is expressed in the endoderm at the time of organogenesis. (A,C) mir302b expression in early streak embryo (E6.5); lateral view (A) and transverse section (C). (B,D) mir302b expression in E7.5 embryo; lateral view (B) and transverse section (D). (E,F) Lateral view of mir302b expression in 3-somite stage embryo (E8.5) (E) and transverse sections (F). a, anterior; p, posterior; Epi, epiblast; m, mesoderm; ve, visceral endoderm; en, endoderm; PS, primitive streak; n ect, neural ectoderm; ect, ectoderm; Fg, foregut. All data presented are representative of at least three independent experiments.

Figure 4.

mir302b is expressed in the whole endoderm at E8.75. (A-D) mir302b expression in E8.75 embryo; lateral view (A) and transverse sections (B-D). nt, neural tube; ht, heart; a, allantois. All data presented are representative of at least three independent experiments.

Two Endoderm-Enriched miRNAs, mir302b and mir20a, Repress TGFβ Signal Transduction.

To identify functional miRNA-mRNA targeting pairs, we utilized a more stringent miRNA target prediction algorithm, mirWalk.24 mirWalk integrates five additional prediction algorithms, each using different approaches including TargetScan, DIANA-microT, miRDB, miRanda, and PITA. Sixteen of 575 Hepatoblast-enriched genes were predicted to be mir302b targets by all six algorithms (Table S5). Of note, six of these have been implicated in TGFβ signaling, including Tgfbr2, Nuclear Factor 1A and 1B (NF1A/B), Bcl6, Kat2b (also known as P/CAF), and Camk2n1. Since one gene can be regulated by multiple miRNAs, we investigated whether other miRNAs in Cluster A were predicted to target these genes. Although mir20a targets were not overrepresented in the 1,599 Hepatoblast-enriched genes, we noted that mir20a, the most highly expressed miRNA in the foregut library, was also predicted to target Tgfbr2, Kat2b, and Camk2n1, using the above six algorithms. Tgfbr2 is a type II receptor required for TGFβ ligand signaling. Kat2b and Camk2n1 can modulate TGFβ signaling.25, 26 By qRT-PCR, mir20a was abundantly expressed in endoderm and dynamically expressed during early liver development (Fig. S6B,C). Collectively, our findings suggest that endoderm enriched miRNAs, mir302b and mir20a, target Tgfbr2, Kat2b, and Camk2n1.

We examined whether Tgfbr2, Kat2b, and Camk2n1 are true targets of mir302b and mir20a by using a reporter assay where wildtype (WT) or mutated versions of the putative 3′UTR miRNA targeting sites of Tgfbr2, Kat2b, or Camk2n1 were inserted into a luciferase reporter vector (Fig. 5A; Fig. S7A). Constructs were transfected into HEK293T cells, which do not express endogenous mir302b but do express mir20a (Fig. S8). Addition of exogenous mir302b (302b_OE) inhibited luciferase activity of the vector containing WT Tgfbr2 3′UTR compared with empty vector and the mutated version. In contrast, knockdown of mir20a (20a_KD) increased luciferase activity in cells containing WT Tgfbr2 3′UTR report vector but not the mutated version (Fig. 5B). Moreover, expression of mir302b in HEK293T cells reduced Tgfbr2 protein expression, while knockdown of mir20a increased Tgfbr2 expression (Fig. 5C). For Kat2b, luciferase activity was reduced by mir302b in cells containing the WT but not the mutated 3′UTR (Fig. 5B); however, mir20a showed no targeting effect on the Kat2b 3′UTR (Fig. 5B). Neither mir302b overexpression nor mir20a knockdown significantly affected the luciferase activity of Camk2n1 3′UTR reporter vector (Fig. S7B). Together, these data demonstrate that both mir302b and mir20a are able to regulate Tgfbr2 expression, while only mir302b can target Kat2b.

Figure 5.

mir302b and mir20a target genes contributing to Tgfβ signaling. (A) 3′UTR sequences from Tgfbr2 and Kat2b showing the predicted miRNA binding regions were cloned into luciferase vectors. The numbers indicate the distance (base) to stop codon of each gene. Mutated versions are shown below. (B) Relative luciferase activity of Tgfbr2 or Kat2b wildtype (WT), mutant (Mut) and backbone (Vector) reporters transfected into HEK293T cells containing mir302b overexpression vector (302b_OE, ↑) or its control vector (Ctrl_OE, −), or mir20a knockdown vector (20a_KD, ↓) or its control vector (Ctrl_KD, +). (C) Western blot showing levels of Tgfbr2 in HEK293T cells transfected with 302b_OE, Ctrl_OE, 20a_KD, or Ctrl_KD. Numbers under each band show relative quantification. All data presented are representative of at least three independent experiments.

Since both Tgfbr2 and Kat2b are associated with TGFβ signaling, we tested whether mir302b and mir20a can affect TGFβ signal transduction. We employed a reporter assay, 3TP-lux, in which a TGFβ-responsive promoter drives the expression of luciferase.16 Irrespective of addition of TGFβ, the promoter activity was reduced in cells expressing mir302b but increased in cells with mir20a knockdown (Fig. 6). Notably, mir302b cannot repress TGFβ signaling when Tgfbr2, lacking the 3′UTR, is overexpressed, and knockdown of mir20a does not increase the signal with dominant-negative Tgfbr2 (Tgfbr2(DN)) (Fig. 6), demonstrating that both mir302b and mir20a are able to suppress TGFβ signal transduction by targeting Tgfbr2.

Figure 6.

mir302b and mir20a repress TGFβ signal transduction. Luciferase activity of the TGFβ responsive reporter, 3TP-lux, 48h after transfection of HEK293T cells with mir302b_OE (↑), Ctrl_OE (−), mir20a_KD (↓), Ctrl_KD (+), pBOS (−), pBOS-Tgfbr2 (+), or pBOS-Tgfbr2(DN) (+) in the presence (+) or absence (−) of TGFβ. Mean values in row with same letter are not significantly different. Values in row with different letters are significantly different.

Mice heterozygous for both smad2 and smad3 die at midgestation with liver hypoplasia and anemia.27 To investigate whether inhibition of TGFβ signaling affects hepatoblast development, we used a stepwise hepatoblast differentiation protocol in ESCs (Fig. S9A). ESC-derived endoderm, expressing Foxa2, Gsc, and Sox17 (Fig. S9B), was generated with medium containing Activin A and exposed to liver specification factors of bone morphogenetic protein 4 (BMP4), beta fibroblast growth factor (bFGF), Activin A, and vascular endothelial growth factor (VEGF). Cells were further cultured in hepatoblast expansion medium with growth factor cocktails. The hepatic markers Alb, AFP, Hnf4α, transthyretin (Ttr), hemopexin (Hpx), and Serpina1a were induced during differentiation (Fig. S9C). Of note, both mir20a and mir302b showed dynamic expression (Fig. 7) with mir302b expression highest at the endoderm stage. Forced expression of mir302b through lentiviral vector during hepatoblast expansion resulted in decreased expression of Tgfbr2 and liver markers, compared to control cells (Figs. 7, 8). A similar reduction of liver markers, but not Tgfbr2, was observed with the TGFβ inhibitor, SB505124 (Fig. S9D). These results demonstrate that mir302b represses liver development during ESC differentiation and suggests that de-repressing TGFβ signaling by down-regulation of mir302b provides a favorable environment for hepatoblast development.

Figure 7.

Ectopic expression of mir302b in ESC differentiation. A lentiviral vector-based system was employed to ectopically express mir302b. Lentivirus generated from the control vector (pCDH) or mir302b overexpression vector (pCDH302b) was added at day 6 (d6) of differentiation. The expression of mir302b, mir20a, and Tgfbr2 during ESC differentiation with pCDH or pCDHmir302b virus was analyzed by qRT-PCR. All data presented are representative of at least three independent experiments.

Figure 8.

Forced expression of mir302b represses liver development in ESC differentiation. The expression of liver markers, Alb, Afp, Hnf4α, Ttr, Hpx, and Serpina1a, were detected at the end of the differentiation. All data presented are representative of at least three independent experiments.


miRNAs Are Dynamically Expressed During Liver Development.

Little is known about miRNA expression during early liver development due to the difficulty in isolating specific embryonic tissues. Here, we describe the first miRNA libraries from dissected E8.5 foregut and E14.5 Dlk1+ hepatoblasts. Our data illustrate the dynamic patterns of miRNA expression that occur during liver development. By way of cluster analysis, we identified three groups of highly expressed miRNAs that are enriched in either foregut, hepatoblasts, or adult liver (Fig. 2A; Fig. S1D). In the adult liver, which contains ∼70% hepatocytes, let7 members comprise 15% of the total miRNA reads, with several members (let7f, c, and b) showing particularly high expression in adult compared to foregut and hepatoblasts (Fig. S1C, Table S3). Let7 functions by repressing self-renewal and promoting differentiation.28 Of note, the most specifically enriched miRNAs in the adult liver are mir29a and mir29b, being virtually absent in foregut and hepatoblasts. Previous studies suggest that mir29b is more highly expressed in mature cells than progenitors during neuronal maturation,29 similar to our observation in adult liver (Fig. S1C, Table S3).

In hepatoblasts, mir379, mir434-3p, and mir127 were highly enriched compared to adult liver and foregut (Fig. S1C, Table S6). These miRNAs are located in the imprinted Dlk1-Dio3 locus, which includes Dlk1. The Dlk1-Dio3 locus contains over 60 miRNAs, comprising 41% of the total miRNA reads obtained in the hepatoblast library. Since Dlk1 was used to isolate hepatoblasts, miRNA expression in the Dlk1-Dio3 region may be coordinated with Dlk1 expression. Many miRNAs from the Dlk1-Dio3 locus were expressed in the foregut at lower levels than the hepatoblasts. In particular, mir-541 and mir-379 were expressed 1.7− and 4.6-fold lower, respectively, in the foregut than hepatoblasts. Recent findings suggested a positive correlation between the activation of the imprinted Dlk1-Dio3 region and pluripotency in induced pluripotent stem (iPS) cells.30 In addition, mir379, mir541, mir434-3p, and mir127, are highly expressed in germline-competent iPS cells.30 Whether expression of microRNAs from the Dlk1-Dio3 locus associate with tissue pluripotency in liver development remains to be elucidated.

Dlk1 expression may only mark a subset of hepatoblasts at E14.5. We observed that at E14.5 nearly all cells expressing HNF4α also express Dlk1+. However, not all Dlk1+ cells express HNF4α (Fig. S3). Thus, different subpopulations may exist. Recent studies suggest that at earlier stages of liver development, Dlk1+ cells can be subdivided based on epithelial cell adhesion molecule (EpCAM) expression.31 It would be of great interest to investigate the miRNA profile of these cells.

In the foregut, 28% of the miRNA reads belong to the mir17-92 locus of miRNAs, which includes mir20a and mir17, both of which are expressed over 19-fold higher in the foregut than hepatoblasts or liver (Fig. S1C, Table S3). A second locus, mir183-182, comprises 13% of the total miRNA reads in the foregut, and is also highly enriched compared to hepatoblasts (>35-fold) and liver (>250-fold). Of note, mir302b was the most highly enriched miRNA in the foregut, with essentially no expression in hepatoblasts or adult liver.

Studies of mir302b orthologs in Xenopus (mir427) and zebrafish (mir430) propose it is critical for mesendodermal fate specification by balancing Nodal and Lefty activity.32 The mir302 family is highly expressed in human ESCs (hESC)33 and knockdown during hESC differentiation leads to decreased expression of mesoderm and endoderm genes but increased ectoderm genes, while overexpression of mir302 family results in increased mesoendoderm.32 By using in situ hybridization in mouse embryos, we observed that mir302b was expressed throughout the ectoderm and newly formed mesoderm at E7.5 (Fig. 3B-D; Fig. S5C-F), similar to results for mir302a,32 supporting a role in pluripotency. We also found that mir302b expression was low or absent in newly formed endoderm at E7.5 (Fig. 3D). However, by the 3-somite stage (∼E8.25), expression was observed throughout the foregut. Later expression was also observed in the hindgut (Fig. 4). Together, our data show that mir302b is highly expressed at the time of endoderm patterning.

Endoderm Enriched miRNAs, mir302b and mir20a, Repress TGFβ Signaling.

Our data show that mir302b reduces expression of murine Tgfbr2 and Kat2b. Tgfbr2 is an essential component of the TGFβ signaling pathway and is specific for TGFβ ligands. Recently, mir302b was found to promote reprogramming of human fibroblasts into iPS cells in part by targeting human TGFBR2 and thus promoting a mesenchymal to epithelial transition.33 Kat2b is a histone acetyltransferase that can interact directly with the intracellular TGFβ signaling component, Smad3, to induce TGFβ-dependent transcriptional responses.25 Thus, mir302b appears to modulate TGFβ signaling at multiple levels, including extracellularly though Lefty,32 directly in the signaling pathway through Tgfbr2, and during transcriptional regulation through Kat2b.

In addition to mir302b, we show that mir20a can target Tgfbr2 expression and repress TGFβ signaling. Expression of mir20a promotes neuroblastoma development by regulating TGFβ signaling.34 Complete depletion of the mir17 family, including mir20a, causes embryonic lethality, with embryos dying around E14.5, exhibiting increased apoptosis in the liver.35 In the endoderm, mir302b may function to compensate for loss of mir20a.

The TGFβ family members, NODAL and BMP4, are required for endoderm formation and patterning.2 However, the role for TGFβ ligands themselves in endoderm organ formation is less well established. Mice lacking Tgfbr2 do not survive beyond E10.5,36 and its role in liver development has not been characterized. It has been proposed that TGFβ signaling must be inhibited during early organogenesis. By culturing 2-somite stage half embryos, Wandzioch and Zaret1 found that TGFβ signaling inhibits the expression of Alb1, suggesting that TGFβ signaling restrains cell specification in foregut. Studies in hESC also showed that the TGFBR1 inhibitor, SB431542, enhances hepatic lineage specification.37 However, mice heterozygous for both smad2 and smad3, which partially disrupts TGFβ signaling, die at midgestation with liver hypoplasia and anemia, demonstrating the importance of TGFβ signaling in hepatoblast proliferation.27 We observed that repressing TGFβ signaling by SB505124 or mir302b reduced the expression of liver markers during mouse ESC differentiation, partially resembling the phenomena of liver hypoplasia in smad2 and smad3 heterozygous mouse. Therefore, down-regulation of mir302b and mir20a during early liver development may relieve the suppression of TGFβ signaling to promote hepatoblast proliferation.


We thank Wenbo Xu and David Ho for technical assistance, and Dr. Nagarajan Kannan, Dr. Jeremy Parker, and Jeff Lam for helpful discussion. P.A.H., M.A.M., and S.J.M.J. are Senior Scholars of the Michael Smith Foundation Health Research.